Professional Documents
Culture Documents
1
Fundamentals of Semiconductor
Electrochemistry and
Photoelectrochemistry
Krishnan Rajeshwar
The University of Texas at Arlington, Arlington, Texas
reader is referred to the many authorita- electrons (e.g. Al with [Ne]3s2 3p1 ) will have
tive accounts that exist before this time a partially occupied frontier band in which
frame for a thorough coverage of details the electrons are delocalized. On the other
on semiconductor–electrolyte interfaces hand, a solid with an even number of va-
in general. Entry to this early literature lence electrons (e.g. Si having an electron
may be found in the references cited configuration of [Ne]3s2 2p2 ) will have a
earlier. In some instances, however, the fully occupied frontier band (termed a va-
discussion that follows necessarily delves lence band, (VB)). The situation for Si is
into research dating back to the 1970s and schematized in Fig. 1.
1980s. As Fig. 2 illustrates, the distinction
To facilitate a self-contained descrip- between semiconductors and insulators
tion, we will start with well-established is rather arbitrary and resides with the
aspects related to the semiconductor en- magnitude of the energy band gap (Eg )
ergy band model and the electrostat- between the filled and vacant bands.
ics at semiconductor–electrolyte inter- Semiconductors typically have Eg in the
faces in the ‘‘dark’’. We shall then 1 eV–4 eV range (Table 1). The vacant
examine the processes of light ab- frontier band is termed a conduction band,
sorption, electron-hole generation, and (CB) (Fig. 2). We shall see later that Eg
charge separation at these interfaces. has an important bearing on the optical
The steady state and dynamic aspects response of a semiconductor.
of charge transfer are then briefly con- For high density electron ensembles
sidered. Nanocrystalline semiconductor such as valence electrons in metals, Fermi
films and size quantization are then statistics is applicable. In a thermodynamic
discussed as are issues related to elec- sense, the Fermi level, EF (defined at 0 K
tron transfer across chemically modi-
fied semiconductor–electrolyte interfaces.
Finally, we shall introduce the vari- Tab. 1 Some elemental and compound
ous types of photoelectrochemical devices semiconductors for photoelectrochemical
applications
ranging from regenerative and photo-
electrolysis cells to dye-sensitized solar
Semi- Conductivity Optical
cells. conductor type(s) band gap
energy [eV]a
1.2
Si n, p 1.11
Electron Energy Levels in Semiconductors
GaAs n, p 1.42
and Energy Band Model GaP n, p 2.26
InP n, p 1.35
Unlike in molecular systems, semiconduc- CdS n 2.42
tor energy levels are so dense that they CdSe n 1.70
form, instead of discrete molecular or- CdTe n, p 1.50
bital energy levels, broad energy bands. TiO2 n 3.00(rutile)
3.2(anatase)
Consider a solid composed of N atoms. ZnO n 3.35
Its frontier band will have 2N energy
eigenstates, each with an occupancy of a The values quoted are for the bulk
two electrons of paired spin. Thus, a solid semiconductor. The gap energies increase in the
having atoms with odd number of valence size quantization regime (see Sect. 7).
1.2 Electron Energy Levels in Semiconductors and Energy Band Model 5
4 N states
0 electrons
6 N electrons
2 N states
3p
Electron energy Eg
4 N states
4 N electrons 3s
2 N states
2 N electrons
ro
Distance
Fig. 1Schematic illustration of how energy bands in
semiconductors evolve from discrete atomic states for the specific
example of silicon.
CB
Eg
Eg
VB VB
(a) (b)
as the energy at which the probability and those above EF are likely to be
of finding an electron is 1/2) can be empty.
regarded as the electrochemical potential Electrons in semiconductors may be
of the electron in a particular phase (in regarded as low-density particle ensembles
this case, a solid). Thus, all electronic such that their occupancy in the valence
energy levels below EF are occupied and CBs may be approximated by the
6 1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry
e−
(+) (−)
e− h+
Si Si Si
h+
Si Si Si Si Fig. 3 A localized picture of
electron-hole pair generation
(see also Fig. 2a) in silicon.
1.2 Electron Energy Levels in Semiconductors and Energy Band Model 7
Energy
tor is termed n-type. By analogy, p-type
semiconductors have p # n. The terms
minority and majority carriers now be-
come appropriate in these cases. For a
p-doped semiconductor case, the Fermi (b)
level now lies close to the VB edge (Fig. 4c).
The movement of EF with dopant concen-
tration can also be rationalized via the
Nernst formalism [6].
Doping can be accomplished by adding
EF
altervalent impurities to the intrinsic
semiconductor. For example, P (a Group
15 or VA element) will act as a donor (c)
in Si (a Group 14 or IVA element). This
can be rationalized on chemical terms complex manner, a discussion of which is
by noting that P needs only four valence beyond the scope of this section. Excellent
electrons for tetrahedral bonding (as in the treatises are available on this topic and on
Si lattice) – the fifth electron is available the solid-state chemistry of semiconduc-
for donation by each P atom. The donor tors in general [16–22].
density, ND nominally is approximately The foregoing discussion strictly refers
1017 cm−3 . Thus, assuming that n $ ND to semiconductors in single-crystal form.
(complete ionization at 300 K), p will Amorphous and polycrystalline counter-
be only approximately 103 cm−3 [recall parts present other complications caused
that the product ni pi is ∼1020 cm−6 (see by the presence of defects, trap states,
preceding section)], bearing out the earlier grain boundaries, and the like. For this rea-
qualitative assertion that n # p. son we orient the subsequent discussion
Impurity addition, however, is not the mainly toward single crystals, although
only doping mechanism. Nonstoichiome- comparisons with less ideal cases are
try in compound semiconductors such as made where appropriate. The distinction
CdTe (Table 1) also gives rise to n- or p- between metal and semiconductor elec-
type behavior, depending on whether Cd or trodes is important when we consider
Te is in slight excess, respectively. The de- the electrostatics across the corresponding
fect chemistry in these solid chalcogenides solid–liquid interfaces; this distinction is
controls their conductivity and doping in a made in the following section.
8 1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry
Vacuum reference
χ ∅
CB ox
Energy
EF λ
Eg o
E redox
VB red
V SC E
E CB
EF E CB
E F, equil
E F, redox
E VB E VB
(b)
E CB E CB
E F, redox
E F, equil
EF
E VB V SC EVB
(c)
Fig. 5 (a) Energy levels in a semiconductor (left-hand side) and a
redox electrolyte (right-hand side) shown on a common vacuum
reference scale. χ and φ are the semiconductor electron affinity and
work function, respectively. (b) The semiconductor-electrolyte
interface before (LHS) and after (RHS) equilibration (i.e. contact of
the two phases) shown for a n-type semiconductor. (c) As in (b) but
for a p-type semiconductor.
within it. Thus, when a metal electrode rather variant behavior of a metal and a
comes into contact with an electrolyte, semiconductor at equilibrium with a redox
almost all the potential drop at the inter- electrolyte.
face occurs within the Helmholtz region
in the electrolyte phase. On the other 1.3.2
hand, the interfacial potential drop across The Depletion Layer
a semiconductor-electrolyte junction (see
following) is partitioned both as VSC and There is a characteristic region within the
as VH leading to a simple equivalent cir- semiconductor within which the charge
cuit model comprising two capacitors (CSC would have been removed by the equi-
and CH ) in series (Fig. 6). Further refine- libration process. Beyond this bound-
ments of the equivalent circuit description ary, the ionized donors (for a n-type
are given later but the point to note is the semiconductor), have their compensating
10 1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry
Distance coordinate
VG Gouy
region
VH
Semiconductor
V SC Electrolyte
Potential
Helmholtz
layer
C SC CH
Fig. 6 Electrostatics at a semiconductor–electrolyte interface.
A highly simplified equivalent circuit for the interface at
equilibrium is shown at the bottom. The Gouy layer is
neglected in the latter case (see text).
# $
charge (electrons), and the semiconductor eo ND
VSC = − W2 (8)
as a whole is electrically neutral. This 2εs
layer is the space charge region or the
depletion layer, so termed because the where W is the depletion layer width.
layer is depleted of the majority carriers. Further reflection shows how the mag-
The potential distribution in this interfa- nitude of W should depend on the semi-
cial region can be quantified by relating conductor parameter ND . Consider two
the charge density and the electric field cases of a semiconductor, one that is lightly
strength as embodied by the Poisson doped (say ND ∼ 1016 cm−3 ) and another
equation [14, 26]. Under restrictive con- that is heavily doped (ND ∼ 1018 cm−3 ).
ditions, more fully discussed in Refs. 6 Obviously in the former case, the charge
and 14, we obtain a particularly simple needed for Fermi level equilibration has
expression to come from deeper into the solid and
# $ so the magnitude of W will be larger.
eo ND This suggests a strategy for chemical
V (x) = − x2 (7)
2εs control of the electrostatics at the semicon-
× (o ≤ x ≤ W ) ductor–electrolyte interface [6]. Nominal
dimensions of W are in the 10–1000 nm
In Eq. (7), eo is the electronic charge and range. This may be compared with the
εs is the static dielectric constant of the corresponding Helmholtz layer width, typ-
semiconductor. The potential distribution ically 0.4–0.6 nm. With the capacitor-in-
is mapped in Fig. 6. We are now in a series model (see earlier section), we can
position to quantify the parameter VSC : see that the semiconductor space charge
1.3 The Semiconductor–Electrolyte Interface at Equilibrium 11
layer is usually the determinant factor in In Eq. (10), Vt is the potential as measured
the total capacity of the interface. Once between an ohmic contact on the rear
again, the contrast with the correspond- surface of the semiconductor electrode
ing metal–electrolyte interface is striking. and the reference electrode (Fig. 6). The
Only when the semiconductor is degener- problematic factors in placing the semi-
ately doped (leading to rather large space conductor and solution energy levels on
charge layer charge, QSC and ‘‘thin’’ deple- a common basis involve VH and VG . In
tion layer widths) or when its surface is in other words, theoretical predictions of the
accumulation does the situation become magnitude of VSC (and how it changes as
akin to the metal–electrolyte interface (see the redox couple is varied) are hampered
following). by the lack of knowledge on the magnitude
of VH and VG . A degree of simplification
1.3.3 is afforded by employing relatively con-
Mapping of the Semiconductor Band-edge centrated electrolytes such that VG can be
Positions Relative to Solution Redox Levels ignored.
As with metals, the Helmholtz layer
Considerations of interfacial electron is developed by adsorption of ions or
transfer require knowledge of the relative molecules on the semiconductor surface,
positions of the participating energy levels by oriented dipoles, or especially in the
in the two (semiconductor and solution) case of oxides, by the formation of sur-
phases. Models for redox energy levels face bonds between the solid surface and
in solution have been exhaustively treated species in solution. Recourse to band
elsewhere [27, 28]. Besides the Fermi level edge placement can be sought through
of the redox system (Eq. 6), the thermal differential capacitance measurements on
fluctuation model [27, 28] leads to a Gaus- the semiconductor–redox electrolyte inter-
sian distribution of the energy levels for the face [29].
occupied (reduced species) and the empty In the simplest case as more fully dis-
(oxidized species) states, respectively, as cussed elsewhere [14, 15, 29], one obtains
illustrated in Fig. 5(a). The distribution the Mott-Schottky relation (for the spe-
functions for the states are given by cific instance of a n-type semiconductor) of
% & the semiconductor depletion layer capaci-
E − EF,redox − λ2 tance (CSC ), again by invoking the Poisson
Dox = exp − (9a)
4 kT λ equation
% & ! "
E − EF,redox + λ2 2 2 kT
Dred = exp − (9b) 1/CSC = (V − Vfb ) − (11)
4 kT λ ND eo εs eo
In Eqs. (9a) and (9b), λ is the solvent In Eq. (11), Vfb is the so-called flat band
reorganization energy. potential, that is the applied potential
Now consider the relative disposition of (V ) at which the semiconductor energy
these solution energy levels with respect to bands are ‘‘flat’’, leading up to the solution
the semiconductor band edge positions at junction. Several points with respect to the
the interface. The total potential difference applicability of Eq. (11) must be noted.
across this interface (Fig. 6) is given by The Mott-Schottky regime spans about
1 V in applied bias potential for most
Vt = VSC + VH + VG (10) semiconductor–electrolyte interfaces (i.e.
12 1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry
in the region of depletion layer formation process itself [31]. Figure 7 contains repre-
of the semiconductor space charge layer, sentative Mott-Schottky plots for both n-
see preceding section) [15]. The simple and p-type GaAs electrodes in an ambient
case considered here involves no mediator temperature molten salt electrolyte [32].
trap states or surface states at the inter- Once Vfb is known (from measure-
face such that the equivalent circuit of the ments), the Fermi level of the semicon-
interface essentially collapses to its most ductor at the surface is defined. It is
rudimentary form of CSC in series with then a simple matter to place the ener-
the bulk resistance of the semiconductor. gies corresponding to the conduction and
Further, in all the earlier discussions, it is VBs at the surface (ECB and EVB , respec-
reiterated that the redox electrolyte is suffi- tively) if the relevant doping levels are
ciently concentrated that the potential drop known. The difference between ECB and
across the Gouy (diffuse) layer in the solu- EVB should approximately correspond to
tion can be neglected. Specific adsorption the semiconductor band gap energy, Eg
and other processes at the semiconduc- (see Figs. 4 and 7). Alternatively, if Vfb is
tor–electrolyte interface will influence Vfb ; measured for one given state of doping
these are discussed elsewhere [29, 30] as of the semiconductor (n- or p-doped), the
are anomalies related to the measurement other band edge position can be fixed from
× 1013 × 1015
8.0 4.0
6.4 3.2
4.8 2.4
[farad−2]
[farad−2]
I C−2
I C−2
3.2 1.6
1.6 0.8
0 0
−0.5 −0.1 0.3 0.7 1.1 1.5
Potential
0/3 +
[V vs Al ]
Fig. 7 Mott-Schottky plots for n- and p-type GaAs electrodes in an AlCl3 /n-butylpyridinium chloride
molten-salt electrolyte. (Reproduced with permission from Ref. 32.)
1.3 The Semiconductor–Electrolyte Interface at Equilibrium 13
Vacuum scale
[eV] SHE
0 [V]
CdS
−3.0 −1.5
GaAs GaP
CdSe Si CdTe
−4.0 −0.5 1.4 eV TiO2 InP
0 2.4 eV
−4.5
2.3 eV 1.1 eV ∼1.3 eV 1.5 eV
−5.0 1.7 eV
Potential
Energy
+1.0
∼ 3.0 eV
−6.0
+2.0
−7.0
+3.0
−8.0
CB E redox CB CB
EF EF EF
Energy
E SS o
E redox
o
E redox
VB VB VB
(a) (b) (c)
Fig. 9 Three situations for a n-type surface states (b), and with the semiconductor
semiconductor–electrolyte interface at VB (c). A discrete energy level is assumed for the
equilibrium showing overlap of the redox energy surface states as a first approximation.
levels with the semiconductor CB (a), with
1.2
I
0.8
Vph
[V]
slope = 1.0
0.4
Fig. 10 Plot of the open-circuit
photovoltage for amorphous
Si-methanol interfaces
0.0 containing a series of
−0.4 0.0 +0.4
one-electron redox couples.
Vredox (Reproduced with permission
[V vs. SCE] from Ref. 33.)
1.3 The Semiconductor–Electrolyte Interface at Equilibrium 15
high [37], they act as a ‘‘buffer’’, in that versus Eredox (see Fig. 10). On the other
in the extreme case, carriers in the semi- hand, complete pinning results in a slope
conductor energy bands are completely of zero. Intermediate cases of Fermi level
excluded from the equilibration process. exhibit slopes between 0 and 1 [41]. As
stated earlier, there is a direct analogy
1.3.4 here with the metal/semiconductor junc-
Surface States and Other Complications tion counterparts [42, 43]:
1.0
13
12
0.8
11
∆V H /∆VSC + ∆VH
10
0.6
9
0.4 8
7
6
0.2 5
3 4
2
1
0
0 0.2 0.4 0.6 0.8 1.0
∆V SC + ∆VH /V
Fig. 11 The ratio of the potential drop in the Helmholtz to the
total potential change computed as a function of the total
potential change. A static dielectric constant of 173 (typical of
TiO2 ) and a Helmholtz capacitance of 10 µF/cm2 were assumed
and the doping density was allowed to vary from 1016 cm−3
(curve 1) to 1020 cm−3 (curve 13). (Reproduced with permission
from Ref. 50.)
potential drop across the interface [50]. reverse (partial) currents exactly balance
Clearly VH can become a sizable fraction of each other and there is no net current
the total potential drop (approaching the flow across the interface. In fact, the
situation for metals) under certain con- situation here is similar to that occurring
ditions. It has been shown [51] that the for a metal–redox electrolyte interface at
Mott-Schottky plots will still be linear but the rest potential. We can write down
the intercept on the potential axis is shifted expressions for the net current using a
from the Vfb value. kinetics methodology as in Ref. 6 with
some minor changes in notation:
1.4 ic = −eo Aket cox (ns − nso ) (14)
Charge Transfer Processes in the Dark
1.4.1 In Eq. (14), ket is the rate constant for
Current-potential Behavior electron transfer, cox is the concentra-
tion of empty (acceptor) state in the redox
Let us return to the equilibrium situation electrolyte, ns and nso are the surface con-
of an n-type semiconductor in contact centrations of electrons, the subscript ‘‘o’’
with a redox electrolyte and reconsider in the latter case denoting the equilibrium
the situation in Fig. 9(a). This is shown situation. Thus, as long as the semiconduc-
again in Fig. 12(a) to underscore the fact tor–electrolyte interface is not perturbed
that the interface is in a state of dynamic by an external (bias) potential, ns ≡ nso
equilibrium. That is, the forward and and the net current is zero. The voltage
1.4 Charge Transfer Processes in the Dark 17
E E E
E CB E CB Vbias E CB
E F, redox Vbias E F, redox E F, redox
E VB E VB E VB
dependence of the current is embodied in similar to the metal case. The major dif-
the ratio, ns /nso , which can be regarded ference resides in the vastly different state
as a measure of the extent to which the densities in the solid and the existence of
interface is driven away from equilibrium. an energy gap region. The two nonequilib-
It must be noted that nso is not the bulk rium situations are shown in Figs. 12(b)
concentration of majority carriers (n) in and 12(c), respectively. Away from equilib-
the semiconductor because of the potential rium, we have the analogous Boltzmann
drop across the space charge layer [6, 35]. expression counterpart to Eq. (15)
# $ ! "
eo VSC eo (VSC + V )
nso = n exp − (15) ns = n exp − (16)
kT kT
A few words about the units of the terms leading, in turn, to
in Eq. (14) are in order at this juncture.
The term i/eo A may be regarded as a ic = −eo Aket cox nso
flux (J ) in units of number of carriers ! # $ "
eo V
crossing per unit area per second [1, 3, × exp − −1 (17)
kT
8]. The concentration terms are in cm−3 ;
thus ket has the dimensions of cm4 s−1 The assumption is inherent in the pre-
because of the second-order kinetics na- ceding discussion that all of the applied
ture stemming from the two multiplied bias (V ) drops across the space charge
concentration terms in Eq. (14) [1, 3, 8]. layer such that we are modulating only
Consider now the application of a bias the majority carrier population at the sur-
potential to the interface. Intuitively when face (and not the potential drop across the
it is such that ns > nso , a reduction cur- Helmholtz layer). In other words, the band
rent (cathodic current) should flow across edge positions are pinned or there is no
the interface such that the oxidized redox Fermi level pinning (see Sect. 1.3.4).
species are converted to reduced species Analogous expressions may be devel-
(Ox → Red). On the other hand, when oped for majority carrier flow for a p-type
nso > ns , the current flow direction is re- semiconductor in contact with a redox
versed and an anodic current should flow. electrolyte, with the important caveat that
Once again the situation here is somewhat the VB is involved in this process instead.
18 1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry
Equation (17) suggests that the cathodic readily rationalized. Finally, io will increase
current is exponentially dependent on po- with doping because of the ‘‘thinness’’ of
tential for V < 0. This is the so-called the resultant barrier at the surface.
forward-bias regime. On the other hand, When EF,redox is moved ‘‘down,’’ that is
when V > 0 (reverse-bias regime) the cur- more positive, the band bending increases,
rent is essentially independent of potential VSC increases and thus nso decreases. A
and, importantly, is of opposite sign. Sim- similar alteration in cox affects EF,redox
ply put, in this case, the electron flow through the Nernst expression. In both
direction (i.e. anodic) is from the occu- instances, we are influencing the Fermi
pied redox states into the semiconductor level at the interface at equilibrium. Thus,
CB (Fig. 12c). It should not, thus, be sur- in a sense, io is a quantitative measure
prising that this process is independent of of the extent of rectification of a given
potential. Both bias regimes are contained interface; that is, a smaller io value
in curve 1 in Fig. 13. translates to better rectification.
Of particular interest to this discussion The reverse bias current remains at a
is the ‘‘preexponential’’ term in Eq. (17): very low value because of the lack of
minority carriers (i.e. holes for n-type
io = eo Aket cox nso (18)
semiconductor) in the dark. Alternatively,
Analogous to the metal case, we can call injection of electrons from occupied redox
this term the exchange current; it is the levels (also an anodic current) has to ther-
current that flows at equilibrium when the mally surmount the surface barrier [5, 34,
partial cathodic and anodic components 35]. Under extreme reverse bias, however,
exactly balance one another. Of particular this barrier becomes ‘‘thin’’ and electrons
interest is the dependence of io on nso . can tunnel through it, leading to an abrupt
Also, variations in cox will affect the increase in the anodic current. This pro-
magnitude of io . Both these trends can be cess was studied even in the early days of
Light, I 2 (> I 1)
Light, I 1
Anodic
Current
0 Dark
Cathodic
semiconductor electrochemistry [52] and a such that the band bending is modest and
detailed discussion is found in a book chap- the holes recombine with electrons (either
ter [14]. Ultimately the junction ‘‘breaks via the surface states or in the space charge
down’’ (at the so-called Zener limit). This region itself), a cathodic current flows that
dark current flow is not shown in Fig. 13 is carried by the majority carriers in the
(curve 1). bulk. This recombination current pathway
Returning to the forward-bias (cathodic) is schematized in Fig. 14 and is further
current flow, Eq. (17) bears some analogy discussed in the next section. Hole injec-
to the famous Tafel expressions in electro- tion has been extensively studied especially
chemical kinetics. Thus, ignoring the unity for III–V (Group 13–15) semiconductors
term within the square brackets, Eq. (17) such as GaAs and GaP because of the
predicts a Tafel slope of 60 mV per decade relevance of this process to electroless etch-
at 298 K. In many instances [53, 54], such ing and device fabrication technology. This
a slope indeed is observed. In many cases, topic has been reviewed [61–64].
however, the slopes are higher than the The invokement of either the CB or the
‘‘ideal’’ value [14, 55–59]. VB of the semiconductor in charge ex-
The causes for this anomalous behavior change in the dark with solution redox
are still not fully understood. It appears species is not always straightforward. This
likely that many factors are involved: is particularly true for multielectron redox
surface film formation, varying potential processes to be discussed later. Movement
drop across the Helmholtz region caused, of the semiconductor band edge positions
for example, by surface state charging, and (i.e. band edge unpinning) relative to the
so on. Even crystallographic orientations redox energy levels also presents a com-
appear to be important [59]. These aspects plicating situation (see following). Some
have been discussed by other authors [14, cases (e.g. Eu2+/3+ in contact with GaAs
55, 60]. electrodes) are interesting in that the same
We have so far considered only (majority
carrier) processes involving the CB (again
assuming for illustrative purposes a n-type
E
semiconductor). Consider the interfacial
situation depicted in Fig. 9(c). The energy
Diffusion
states of the redox system now overlap with e− e− ECB
the VB of the semiconductor such that hole
injection in the ‘‘dark’’ is possible. When e−
e−
the band bending is large, the injected
holes remain at the surface and attack the
Et
semiconductor itself, causing the latter to E SS
undergo corrosion. If the bias potential is Electrolyte
e−
Fig. 14 Hole injection into the VB of a
n-type semiconductor from an oxidant E VB
(e.g. Fe3+ ) and the injection or
recombination pathway. Both surface e−
state–mediated and depletion layer e−
trap–mediated routes are shown for the
h+ h+
recombination.
20 1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry
couple can interact with both bands [55, of n-type and p-type samples of a given
65]. Thus, the oxidation of Eu2+ is a VB semiconductor. It may be noted that for
process (occurring at p-GaAs but not at n- a redox process occurring via one of the
GaAs in the dark), whereas the reduction bands (e.g. VB), the cathodic currents (elec-
of Eu3+ (a facile process that reportedly tron transfer from VB to Ox) are expected
occurs at rates close to the thermionic to be equal for n- and p-type materials.
emission limit, Ref. 55) is mediated by CB This idea has been pursued using the so-
electrons [65]. The [Cr(CN)6 ]3−/4− redox called quasi-Fermi level concept [55, 72,
system behaves in a similar manner with 73]. This model has been demonstrated
respect to GaAs [66]. quantitatively by studying the anodic de-
Electroluminescence, (EL), is a versatile composition of GaAs and the oxidation of
probe for studying such carrier injection redox species such as Cu+ and Fe2+ at n-
processes. Thus, hole injection into the VB and p-type GaAs electrodes [72, 73].
of a n-type semiconductor leads to cathodic
EL, whereas electron injection into the CB 1.4.2
of a p-type semiconductor leads to anodic Dark Processes Mediated by Surface States
EL [67]. Examples of studies of cathodic or by Space Charge Layer Recombination
EL are commonplace [68–70]; however,
anodic EL is not very common because Surface states were considered earlier
the energy requirement for the redox (Sect. 1.3.4) from an electrostatic perspec-
couple has a very negative redox potential. tive. Here we examine their dynamic
Nonetheless, anodic EL has been reported consequences. There are two principal
for the p-InP-[Cr(CN)6 ]4− interface [71]. charge transfer routes involving surface
Radical intermediates can also cause EL states. Consider again an n-type semicon-
as discussed later for multielectron redox ductor; the forward-bias current can either
processes. EL is treated in more depth in involve direct exchange of electrons be-
another chapter. tween the semiconductor CB and Ox states
This finally brings us to the compa- in solution (Fig. 12b) or can be mediated
rability of the current-potential behavior by surface states (Fig. 15). The second
e−
ECB
EF
ESS
Electrolyte
Fig. 15 Surface
state–mediated electron
EVB injection from the CB of a n-type
semiconductor into the
electrolyte.
1.4 Charge Transfer Processes in the Dark 21
route involves hole injection into the semiconductor CB also came from AC
semiconductor VB again from Ox states in impedance spectroscopy measurements
solution (Figs. 9c and 14). The recombina- on n-ZnO, n-CdS, and n-CdSe in contact
tion current is mediated either by surface with [Fe(CN)6 ]3− species [76, 77].
states or via space charge layer recombina- The electrochemical impedance for
tion. We will consider first the CB process. surface state–mediated charge transfer
Initial evidence for the intermediacy of has been computed recently [78]. The
surface states came from dark current key results are summarized in Fig. 16.
measurements on n-TiO2 and n-SrTiO3 Figure 16(a) contains the proposed equiv-
in the presence of oxidizing agents such as alent circuit for the process and features a
[Fe(CN]6 3− , Fe3+ , and [IrCl6 ]2− [74, 75]. parallel connection of the impedance for
Similar early evidence that the charge the Faradaic process [ZF (ω)] (ω = angular
transfer process was more complex than frequency, 2πf ) and the capacitance of the
direct transfer of electrons from the semiconductor depletion layer, CSC . The
C1
C2
R1
(a) R2 ∆
−1500
−1000
Xs / Ω
−500 1 Hz
100 kHz 1 kHz 100 mHz
100 Hz
10 Hz
0 10 kHz 10 mHz
1 MHz
Csc
the surface, the potential is redistributed the interface itself. However, values for ket
between the semiconductor spacecharge approximately three orders of magnitude
layer and the Helmholtz region. Now higher have also been reported for simi-
iodine is reduced by hole injection as lar interfaces, namely, n-GaAs-acetonitrile-
gauged by EL and AC impedance mea- containing cobaltocenium [Co(Cp)2 + ] ac-
surements [95]. ceptors [99, 101]. Similarly, high values
were reported for p-GaAs-acetonitrile in-
1.4.3 terfaces with ferricenium and cobaltoce-
Rate-limiting Steps in Charge Transfer nium redox species [102]. In these lat-
Processes in the Dark ter cases, alternative mechanisms (e.g.
thermionic emission, see following) must
The assumption is implicit in the discus-
be invoked in a rate-limiting role. Quartz
sion in Sect. 1.4.1 (leading to Eq. 18) that
crystal microbalance measurements have
charge transfer kinetics at the semicon- yielded evidence for adsorption of redox
ductor–electrolyte interface is the rate- species (and consequently high local sub-
limiting step. Fundamentally, we have strate concentration) in some of these
to differentiate majority carrier capture ‘‘anomalous’’ instances [101].
and minority carrier injection processes In the majority carrier capture process, if
in the dark. In the former case, tran- the interfacial charge transfer kinetics are
sit through the semiconductor itself or facile, the transport of majority carriers
charge exchange with the surface states through the space charge region can
can be potentially rate-limiting. In the lat- play a rate-limiting role. The thermionic
ter case, there are three steps involved: emission theory [34] assumes that every
hole injection into the semiconductor VB, electron that reaches the semiconductor
charge exchange between the recombina- surface, and has the appropriate energy to
tion center and the semiconductor CB, and overcome the potential barrier there, will
diffusion of majority carriers (electrons) cross the interface with a tunnel probability
from the neutral region. Finally, mass of unity. However, if the interfacial kinetics
transport processes in the electrolyte phase are sluggish, some of the electrons will be
itself can be a limiting factor in the over- reflected at the interface. In this case, the
all current flow. We shall examine carrier exchange current io is no longer described
capture and injection processes in turn. by Eq. (18) but by Eq. (19) [34]
The vast majority of outer-sphere, non- # ∗$ ' (
∗ m ns
adsorbing redox systems to date have io = AA T2 (19)
yielded values for ket in the 10−17 − me n
10−16 cm4 s−1 range [3, 8]. These include In Eq. (19), A is the electrode area, A∗ is the
n-Si-CH3 OH [96, 97], n-InP-CH3 OH [98], Richardson constant (120 A K−2 cm−2 ),
GaInP2 -coated n-GaAs-acetonitrile [99], m∗ /me is the relative effective electron
and p-GaAs-HCl [100] interfaces. The re- mass in the CB, and T is the absolute
dox couples in these studies have mostly temperature.
involved metallocenes that show low pro- In many of the reported instances [53,
clivity to adsorb on the semiconductor sur- 55, 103], the current calculated from
face. In these cases, the rate-determining Eq. (19) is much higher than that mea-
step in the overall current flow undoubt- sured experimentally, signaling that inter-
edly lies in the electron transfer event at facial charge-transfer kinetics are limiting
24 1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry
the overall rate. On the other hand, in the n- process was found to describe the charge
GaAs-acetonitrile-Co(Cp)2 + case [101], AC transfer at n-GaAs in 6 M HCl containing
impedance spectroscopy data appear to Cu+ as the hole injecting species [55, 73].
support the assumption that thermionic Whatever the limiting mechanism, ul-
emission is the current-limiting transport timately the current becomes limited by
mechanism. concentration polarization, that is, by the
Another factor that enters into this transport of redox species from the bulk
discussion is the mobility of the majority electrolyte to the semiconductor surface.
carriers. It has been argued [14] that in the The situation in this regard is no different
case of low mobility materials (e.g. µn ∼ from that at metal electrode–electrolyte
1 cm2 V−1 s−1 ), carrier transport from the interfaces. As in the latter case, hydro-
semiconductor bulk to the interface itself dynamic voltammetry is best suited to
can become limiting. Clearly a multitude study mass transport. AC impedance spec-
of factors are important in majority carrier troscopy can be another useful tool in this
capture: ket , acceptor concentration in the regard [105].
electrolyte and carrier mobility. In the former case, the data can be
What about the minority carrier injec- processed via Levich plots of current vs.
tion process depicted in Fig. 14? Here,
ω1/2 (ω = angular frequency). If processes
contrasting with the process considered
other than solution mass transport become
earlier, the hole injection step is usually
rate-limiting, then the plot will show a
very fast (see following). Then the current
curvature and the current will even become
is limited by diffusion or recombination
independent of the electrode rotation rate.
described by the Shockley equation [104]
In this case, inverse Levich (or Koutecky-
eo ADp n2i Levich) plots of 1/i vs. ω−1/2 can be
io = (20a) used for further analyses. Such analyses
nLp
have been done, for example, for n-GaAs-
for bulk recombination, and acetonitrile-Co(Cp)2 + interfaces [101] and
n- and p-GaAs electrodes in contact
io = 0.5 eo AW σ νth Nt ni (20b) with the Cu+/2+ redox couple in HCl
for recombination within the semiconduc- electrolyte [55, 73].
tor space charge region. In Eqs. (20a) and The diffusion impedance at semicon-
(20b), Dp is the diffusion coefficient for ductor electrodes has been considered
holes, Lp is the hole diffusion length, ni is recently [105]. This author described the
the intrinsic carrier concentration, σ is an applicability of AC impedance spec-
average capture cross-section for electrons troscopy for the study of electron capture
and holes, νth is the thermal velocity of and hole injection processes at n-GaAs-
charge carriers, and Nt is the areal den- H2 O/C2 H5 OH-methyl viologen, p-InP-
sity of recombination (trap) centers in the aq. KOH-Fe(CN)6 3 , n-GaAs-H2 SO4 -Ce4+ ,
middle of the energy gap (Fig. 14). and n-InP-aq. KOH-Fe(CN)6 3− interfaces.
The diffusion or recombination mecha- In the case of electron capture processes, a
nism results in considerable overpotential Randles-like equivalent circuit was found
for (cathodic) current flow in the dark to be applicable [105]. On the other hand,
(again assuming a n-type semiconduc- no Warburg component was present in
tor for illustration). Such a rate-limiting the hole injection case when the reverse
1.5 Light Absorption by the Semiconductor Electrode and Carrier Collection 25
reaction was negligible (Fig. 17). For a non- value (shorter wavelengths). Both types of
ideal semiconductor–electrolyte contact transitions can also be seen in the same
(see Sect. 1.3.4), a Warburg impedance ap- material, for example GaP [107].
peared in the electrochemical impedance Within the present context, the impor-
of an injection reaction as well, as exem- tant point to note is that the absorption
plified by the n-InP-Fe(CN)6 3− case [105]. depths (given by 1/α) are vastly different
for direct and indirect transitions. While in
the former case absorption depths span the
1.5
100–1000 nm range, they can be as large
Light Absorption by the Semiconductor
as 104 nm for an indirect transition [9].
Electrode and Carrier Collection
Optical transitions in semiconductors
1.5.1 can also involve localized states in the band
Light Absorption and Carrier Generation gap. These become particularly impor-
tant for semiconductors in nanocrystalline
The optical band gap of the semiconduc- form (see following). Sub–band gap transi-
tor (Sect. 2) is an important parameter tions can be probed with photons of energy
in defining its light absorption behavior. below the threshold defined by Eg .
In this quantized process, an electron-
hole pair is generated in the semicon- 1.5.2
ductor when a photon of energy hν Carrier Collection
(ν = frequency and hν > Eg ) is absorbed.
Optical excitation thus results in a delo- The number of carriers collected (in an
calized electron in the CB, leaving behind external circuit, for example) versus those
a delocalized hole in the VB; this is the optically generated defines the quantum
band-to-band transition. Such transitions yield (+) – a parameter of considerable in-
are of two types: direct and indirect. In the terest to photochemists. The difficulty here
former, momentum is conserved and the is to quantify the amount of light actually
top of VB and the bottom of CB are both absorbed by the semiconductor as the cell
located at k = 0 (k is the electron wave walls, the electrolyte, and other compo-
vector). The absorption coefficient (α) for nents of the assembly are all capable of
such transitions is given by [106] either absorbing or scattering some of the
incident light. Unfortunately, this problem
α = A+ (hν − Eg )1/2 (21)
has not been comprehensively tackled, un-
In Eq. (21), A+ is a proportionality con- like in the situation with photocatalytic
stant. Indirect transitions involve phonon reactors involving semiconductor partic-
modes; in this case Eq. (21) takes the form ulate suspensions, where such analyses
are available [108–111]. Pending these, an
α = A+ (hν − Eg )2 (22)
effective quantum yield can still be defined.
A given material can exhibit a direct or in- Returning to the carrier collection
direct band-to-band transition depending problem, consider Fig. 18 for an n-
on its crystal structure. For example, Si semiconductor–electrolyte interface. As
single crystals have an indirect transition can be seen, the electron-hole pairs are
located at 1.1 eV (Table 1). On the other optically generated, both in the field-free
hand, amorphous Si is characterized by a and in the space charge regions within the
direct optical transition with a larger Eg semiconductor. Recombination of these
26 1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry
1/α
LD + W
W
CB
hν
VB
Redox
Semiconductor electrolyte
Fig. 18 Photogeneration of electron-hole pairs in the field-free
region and depletion layer for a n-type semiconductor–
electrolyte interface. The characteristic regions defined by the
depletion layer (W), Debye length (LD ), and the light
penetration depth (1/α) are also compared.
carriers must be considered in the bulk, in depletion layer is dictated by the diffusion
the space charge layer, and on the semicon- process. A diffusion length, L, can be
ductor surface (the latter in contact with defined
the redox electrolyte). We are assuming
here that light is incident from the elec- ) )
Lp = Dp τp = kT µp τp
trolyte side. Rear illumination geometry
can be profitably employed and is consid- (for n-type semiconductor) (23)
ered later for the nanocrystalline film case.
The direction of the electric field at the
interface (Fig. 6, Sect. 1.3.2) is such that The subscripts in Eq. (23) remind us
the minority carriers (holes in this case) that we are dealing with minority carrier
are swept to the surface and the electrons collection; µp is the hole mobility and τp
are driven to the rear ohmic contact. How is the hole life-time. The characteristic
fast the holes are drained away (by Faradaic length Lp defines the region within
reactions involving the redox electrolyte) which electron-hole pair generation is fully
will dictate how the Fermi levels compare effective. Pairs generated at depths longer
with the equilibrium situation discussed than the Debye length, LD (LD = W + L)
earlier. The departure from equilibrium will simply recombine. Thus, the effective
has been quantified in terms of the quasi- quantum yield for a given interface will
Fermi level concept discussed later. depend on the relative magnitudes of
The extent of collection of minority LD and the light penetration depth, 1/α
carriers from the region beyond the (Fig. 18) [112, 113].
1.5 Light Absorption by the Semiconductor Electrode and Carrier Collection 27
0.001
.. 520
Fig. 19 Gartner plots (see Eq. 25) for
the p-GaP −0.5 M H2 SO4 interface. The
numbers on the plots refer to the 530
excitation wavelength; Efb is the flat
band potential and E is the bias
potential. (Note that this notation is
different from that employed in the 0 0.5 1.0 1.5
text.) (Reproduced with permission
(Efb − E )1/2/V 1/2
from Ref. 117.)
28 1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry
ECB
e−
ESS
hν
hν
E VB
h+
Fig. 21 Surface state mediation of both minority carrier (i.e. hole)
transfer and recombination for a n-type semiconductor electrolyte
interface.
current-potential profiles at the semicon- the minority carrier current (iph ) is directly
ductor–electrolyte interface to this recom- proportional to the photon flux (Eq. 24).
bination mode [137]. Thus, the net current is given by:
Other models taking the above nonideal- ! # $ "
eo V
ities to varying extent have been proposed; i = iph − io exp − −1 (27)
kT
a detailed discussion of these lies be-
yond the scope of this section [141–147]. The minus sign in Eq. (27) underscores the
However, it is worth noting here that fact that the majority carrier component
in some instances involving high-quality flows opposite (or ‘‘bucks’’) the minority
semiconductor–electrolyte interfaces the carrier current flow. This photocurrent
rate-determining recombination step does component is shown as curves 2 and 3
indeed appear to lie in the bulk semi- in Fig. 13.
conductor [1, 148]. Silicon photoelectrodes Equation (27) shows that the diode
in methanolic media containing fast, one- equation is offset by the iph term; this is
electron, outer-sphere redox couples were exactly what is seen in Fig. 13. The plateau
studied in these cases. photocurrent is proportional to the photon
flux, Io , as illustrated for two different
1.5.3 values of the incident light intensity in
Photocurrent-potential Behavior Fig. 13. At the open-circuit condition of the
interface, i = 0 (and neglecting the unity
The current-voltage characteristics of an term within the square brackets relative to
illuminated semiconductor electrode in the exponential quantity), Eq. (27) leads to
contact with a redox electrolyte can be
kT iph
obtained by simply adding together the Voc $ ln (28)
majority and minority current compo- eo io
nents. The majority carrier component is Equation (28) underlines two important
given by the diode equation (Eq. 17) while trends: First, Voc increases logarithmically
30 1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry
with the photon flux (with a slope of curves, as illustrated in Fig. 22. Thus,
∼60 mV at 298 K). Second, Voc decreases the superposition principle [149–154] is
with an increase of io (again logarithmi- not obeyed in this instance. The dashed
cally). This underlines the importance of line in Fig. 22 is produced by translat-
ensuring that the majority carriers do not ing the photocurrent-voltage data by jSC ,
‘‘leak’’ through the interface. Because of the short-circuit current density. If the su-
the diode nature of the interface, from perposition principle is held, this dashed
a device perspective, the semiconductor curve would have overlaid the dark current-
surface must be designed to have fast mi- voltage curve. Thus, this ‘‘excess’’ (forward
nority carrier transfer kinetics (and thus bias) current embodies the APE, and the
high iph ), but must be blocking to the failure of superposition is quantified as the
flow of majority carriers (from the CB voltage difference (%V ) between the dark
for a n-type semiconductor) into the re- j − V data and the dashed line.
dox electrolyte. This challenge is similar to What are the ramifications of the cross-
what the solid-state device physicists face, over or the APE? First, mathematical
but relative to metals (with a high density modeling of carrier transport in a junction-
of acceptor states), chemical control of re- based solar photovoltaic system, accord-
dox electrolytes offers a powerful route to ing to
performance optimization of liquid-based j = jSC − jbk (V ) (27a)
interfaces as also pointed out by previous
authors [1, 6, 8, 149–154]. is not valid in the presence of this effect.
Referring back to Fig. 13, the current- (In Eq. (27a), jbk is the ‘‘bucking’’ cur-
potential curves under illumination of rent density in the forward-bias regime,
the semiconductor simply appear shifted see previous section.) That is, a fully lin-
‘‘up’’ relative to the dark i − V counter- earized system of differential equations
part. This, however, is the ideal scenario. and boundary conditions cannot be used
Anomalous photoeffects (APEs) are of- to model the interface carrier transport.
ten observed that manifest as a cross- Second, computation and modeling of the
over of light and dark current-voltage open-circuit voltage for such devices by
Light
Current density
∆V
Dark
Fig. 22 Anomalous photoeffect (APE)
showing cross-over of the dark and light
current-voltage curves again for a n-type
semiconductor-based interface. The
dashed line is obtained as described in
Potential the text.
1.5 Light Absorption by the Semiconductor Electrode and Carrier Collection 31
5 × 10−2
[mA cm−2]
Current
5 × 10−3
5 × 10−4
5 × 10−5
−50 −150 −350 −550
Voltage vs. (Al/Al+3)
[mV]
More recent studies on n-CdS [164, 165] Figure 23 contains an example of the
and n-TiO2 photoanodes [166] implicate APE for the n-GaAs-AlCl3 /n-butylpyridi-
the formation of photoconductive layers nium chloride molten salt electrolyte inter-
in the APEs. Thus, the foregoing review face [167]. The bottom curve in Fig. 23 is
suggests the following the measured dark current-voltage profile.
The top curve was obtained from the
1. APE is a very general phenomenon that photocurrent-voltage data (under irradi-
has been observed for solid-state junc- ation of the semiconductor). Clearly, if
tions for n- and p-type semiconductors the superposition principle held, the two
alike, and for a wide variety of semicon- curves would have coincided with one
ductor materials. another.
2. The reported results and trends are APEs have also been observed for
often contradictory. It is quite possible nanocrystalline and chemically modified
that the experimental conditions in films, as discussed in a subsequent section.
these studies were quite variant, thus Light-induced changes in the electro-
precluding direct comparison of the statics at the semiconductor–electrolyte
results. interface are conveniently probed by
3. The mechanistic reasons given for capacitance-voltage measurements in the
the APE are possibly many, and gen- dark and under illumination of the semi-
eralizations may not be warranted. conductor electrode. If charge trapping
Clearly, more research is needed on at the interface plays a decisive role
this topic. (whatever be the mechanism), the voltage
1.5 Light Absorption by the Semiconductor Electrode and Carrier Collection 33
drop across the illuminated interface is What are the processes important in a
altered, and consequently the semicon- dynamic interrogation of the semiconduc-
ductor band edge positions are shifted. tor–electrolyte interface?
This, of course, is the Fermi-level pin-
ning situation that was encountered earlier 1. Carrier generation within the semicon-
(Sect. 1.3.4). Examples of studies address- ductor,
ing this aspect may be found, for ex- 2. diffusion of minority carriers from the
ample on p – GaAs [124] and CdTe [168, field-free region to the space charge
169]–based aqueous electrolyte interfaces. layer edge,
3. transit through this layer,
1.5.4 4. charge transfer across the interface, and
Dynamics of Photoinduced Charge Transfer 5. carrier recombination via surface states
or via traps in the space charge layer.
So far the discussion has centered on
the steady state aspects of carrier gen- Other phenomena such as thermaliza-
eration and collection at semiconduc- tion also are important as discussed later in
tor–electrolyte interfaces. As with their the context of hot carrier effects. The time
metal electrode counterparts, a wealth constant (τcell ) of the cell and the measure-
of information can be gleaned from ment circuitry has complicated matters
perturbation-response type of measure- further and have caused some confusion
ments. An important difference, however, in the interpretation of transient data. If
lies in the vastly different timescale win- a potentiostat is not used, then this time
dows that are accessible in the two cases. constant is given by Eq. (29).
The critical RC time-constant of the cell in One can envision three types of per-
a transient experiment is given by turbation: an infinitesimally narrow light
pulse (a Dirac or δ-functional), a rectan-
τcell = C(Rm + Rel ) (29) gular pulse (characteristic of chopped or
In Eq. (29), Rm is the measurement interrupted irradiation), or periodic (usu-
resistor (across which the current or ally sinusoidal) excitation. All three types
photocurrent is measured) and Rel is the of excitation and the corresponding re-
electrolyte resistance. The term C is the sponses have been treated on a common
capacitance, which, in the metal case, is the platform using the Laplace transform ap-
Helmholtz layer capacitance, CH . (Once proach and transfer functions [170]. These
again, the Gouy region is ignored here.) perturbations refer to the temporal be-
For semiconductor–electrolyte interfaces, havior adopted for the excitation light.
we have seen that two layers are involved However, classical AC impedance spec-
in a series circuit configuration with troscopy methods employing periodic po-
corresponding capacitances of CSC and tential excitation can be combined with
CH (Fig. 6). Because CH # CSC , C $ steady state irradiation (the so-called PEIS
CSC . This assumption is usually justified experiment). In the extreme case, both
because CH $ 10−5 F cm−2 and CSC = the light intensity and potential can be
10−8 − 10−0 F cm−2 . If the composite modulated (at different frequencies) and
resistance (Rm + Rel ) is 100 ohm, then the (nonlinear) response can be mea-
τcell for metal electrodes is ∼10−3 s sured at sum and difference frequencies.
and that for the semiconductor case is The response parameters measured in
10−6 − 10−7 s. all these cases are many but include
34 1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry
comprising these layers are comparable to detected, for example, by electron param-
(or even smaller than) nominal depletion agnetic resonance spectroscopy [215, 216].
layer widths. In either case, the resultant negative
The rather complete interpenetration of charge generated by electron accumula-
the electrode and electrolyte phases must tion at the internal surfaces has to be
mean that the Helmholtz double layer ex- balanced by cations (from the electrolyte
tends throughout the interior surface of the phase) for charge compensation. Such
nanoporous network, much like a superca- ion insertion reactions have been stud-
pacitor [9, 210] situation. Finally, unlike in ied using techniques such as voltammetry,
the cases treated earlier, the semiconduc- reflectance or absorption spectroscopy,
tor (especially if it is a metal oxide) is not chronoamperometry, and electrochemi-
heavily doped such that free majority carri- cal quartz-crystal microgravimetry [213,
ers are not present in appreciable amounts. 217–22]. Both aqueous and aprotic elec-
This is indicated, for example, by the sensi- trolytes have been deployed for these
tivity of the conductivity of nanocrystalline studies.
TiO2 layers to UV light – the conductivity Unlike in the single-crystal cases treated
is strongly enhanced on UV exposure, sim- earlier, placement of the semiconductor
ilar to a photoconductive effect. This effect energy band positions at the interface
has been interpreted, in terms of trap fill- via Mott-Schottky analyses is not straight-
ing with recombination times considerably forward for nanocrystalline films. Abrupt
slower than the trapping processes under changes in slope and other nonideali-
reverse bias [211, 212]. The light sensitivity ties [215, 227, 223] have been observed,
also is diagnostic of the fact that the low for example, in the Mott-Schottky plots for
electronic conductivity in the dark is not TiO2 films and attributed to the influence
due to high interparticle resistances (i.e. in of the conductive glass that is normally
the ‘‘neck’’ regions), but rather is indicative employed to support these films. This be-
of the low electron concentrations. havior is especially prevalent under reverse
The electron concentration can be in- bias. The onset of majority carrier op-
creased by forward-biasing the nanocrys- tical absorption (in the visible and near
talline electrode–electrolyte interface po- IR range) under forward-bias instead has
tentiostatically. The interface is driven thus been profitably employed to place the CB
into the accumulation regime for the ma- positions of TiO2 in aqueous media [208].
jority carriers, and if a transparent rear Impedance spectroscopy and electro-
contact (e.g. F-doped, SnO2 or Sn-doped chemical dye desorption experiments have
indium oxide) is used, the resultant blue been employed [224] to study the electri-
(or bluish-black) coloration of the film cal characteristics of TiO2 nanocrystalline
can be spectroscopically monitored [208, films in the dark. This study as well as the
209, 213]. Whether the optical response others cited earlier demonstrate how the
arises from CB electrons or from electrons conductivity changes (as a result of elec-
trapped in surface states is not entirely tron injection from the support electrode)
clear. It has been claimed [214] that the can cause the porous or nanocrystalline
absorption spectrum of the latter differs layer to manifest itself electrically, such
significantly from CB electrons. Electrons that the active region moves away (i.e.
in surface states can be chemically iden- outward) from the support as the forward-
tified with Ti3+ defect sites that can be bias voltage is increased. The potential
38 1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry
distribution has also been analyzed de- by the electrolyte redox (specifically Red)
pending on whether the depletion layer species, collection of the photogenerated
width exceeds or is smaller than the typi- electrons at the rear contact becomes
cal dimension of the structural units in the the determinant factor in the quantum
nanocrystalline network [223]. yield. Thus, the quasi-Fermi level for holes
remains close to EF,redox and that for
1.7.3
electrons, EF,e moves ‘‘up’’, as depicted
Photoexcitation and Carrier Collection:
in Fig. 24b. Illumination thus induces an
Steady State Behavior
electron flux, Jn (x) through the nanocrys-
Figure 24 contains a schematic representa- talline phase. Under steady state condi-
tion of the nanocrystalline semiconductor tions, the photocurrent density (jph ) is
film–electrolyte interface at equilibrium equal to eo Jn (x = d). The driving force for
(Fig. 24a) and the corresponding situation electron diffusion through the network of
under band gap irradiation of the semicon- nanocrystallites has been calculated from
ductor (Fig. 24b) [9]. Because the diffusion first principles [225]. It has been found that
length of the photogenerated carriers is the driving force is approximately kT /eo
usually larger than the physical dimen- divided by the thickness of the network.
sions of the structural units, holes and Importantly, this free-energy gradient is
electrons can reach the impregnated elec- found to be independent of the incident
trolyte phase before they are lost via bulk photon flux.
recombination. This contrasts the situa- It is important to reiterate that the
tion with the single-crystal cases discussed charge separation in a nanocrystalline
earlier. semiconductor–electrolyte interface does
If, as is the case with TiO2 nanocrystal- not depend on a built-in electric field
line films, the holes are rapidly scavenged at the junction as in the single-crystal
E
EC
E F,redox
EV
Substrate
d 0
(a) x
E
− − −
je(d ) E F,e
E F,redox Fig. 24 Schematic representation of a
nanocrystalline semiconductor–
+
electrolyte interface in the dark (a) and
under illumination from the electrolyte
Substrate Red side (b). Ec and Ev correspond to ECB
Ox
and EVB in our notation. (Reproduced
d 0
with permission from the authors of
(b) x Ref. 9.)
1.7 Nanocrystalline Semiconductor Films and Size Quantization 39
case. Instead, the differential kinetics for sensitization experiments) although other
the reactions of photogenerated electrons redox electrolytes [e.g. SCN− /(SCN)2 − ;
and holes with electrolyte redox species 228] have been employed as well. For the
account for the charge separation (and the chalcogenide films, sodium selenosulfite
generated photovoltage). The molecular was employed [227]. It must be noted that,
factors underlying the sluggish scavenging aside from losses caused by the surface
of electrons at the nanocrystalline film- recombination and back-reactions, an ad-
electrolyte boundary (by the redox species) ditional loss component from the increase
are as yet unclear. Clearly, the competition in film resistance must also be recog-
between surface recombination of these nized, especially as the film thickness is
electrons (with the photogenerated holes) increased. The resistance loss manifests
and collection at the rear contact dictates as a deterioration in the photovoltage and
the magnitude of the quantum yield that is fill factor.
experimentally measured for a particular In the discussion to this point, we have
junction. not considered trapping or release of the
Photocurrent losses have been recorded photogenerated electrons as they undergo
for electrolytes dosed with electron accep- transit to the rear contact. However, elec-
tors such as O2 and iodine [226]. Nanocrys- trons trapped in localized interfacial states
talline TiO2 electrodes with thicknesses induce a countercharge in the Helmholtz
ranging from 2 µm to 38 µm were included double-layer, as discussed in the preced-
in this study. In the presence of these ing discussion. The resultant voltage drop
electron-capture agents, electron collec- can introduce a nonnegligible field com-
tion (i.e. photocurrent) at the rear contact ponent into the diffusional process. The
was seriously compromised. On the other time-dependence of the electron density,
hand, as high as 10% of the photons were n(x, t) is given by [9]
converted to current for a 38 µm thick film
in a N2 -purged solution [226]. ∂n(x, t) ∂ 2 n(x, t)
= ηαIo e−αx + Dn
The result was obtained with front-side ∂t ∂x 2
illumination geometry. As one would in- n(x, t) − no
tuitively expect, carrier collection is most − (32)
τ
efficient close to the rear contact. In-
deed, marked differences have been ob- In Eq. (32), η is the electron injection
served for photoaction spectra with the efficiency, Dn is the diffusion coefficient
two irradiation (i.e. through the elec- of electrons, and τ is the pseudo first-
trolyte side vs. through the transparent order lifetime of electrons determined by
rear contact) geometries for TiO2 , CdS, back-reaction with Ox.
and CdSe nanocrystalline films [227, 228]. Even if the migration component is
Obviously, the relative magnitudes of negligible (but see following), solution of
the excitation wavelength and the film Eq. (32) presents difficulties because of the
thickness critically enter into this variant possible dependence of Dn on n and x.
behavior. Similarly, τ may depend on these two vari-
In the vast majority of cases, the ables also. Nonetheless, the steady state
iodide/triiodide redox couple has been em- solution of Eq. (32) has been obtained [229]
ployed (presumably because of its success by assuming that D and τ are constant
in shuttling the photooxidized dye in the and that η = 1. Under these conditions,
40 1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry
the photocurrent is predicted to be in- insulating and (2) the electrolyte perme-
dependent of voltage – a rather physically ates into the network and possibly contacts
implausible situation. In the forward-bias the rear support electrode. The transition
regime, η is expected to decrease and the in the profiles from spiked at potentials
back-reaction of the photogenerated elec- near Von (photocurrent onset) to more
trons (with Ox) can no longer be neglected. rectangular at positive bias potentials (not
This brings us to the rear support-film shown in Fig. 25) must mean that the volt-
interface. What sort of barrier exists at this age does exert an effect on carrier transit
junction? Are the electron exchange ki- through the network. No satisfactory expla-
netics voltage-dependent at this interface? nation appears to exist at present to resolve
The effect of changing the work function of this apparent anomaly.
the substrate on the current-voltage curves
(in the dark and under illumination) has 1.7.4
been investigated for TiO2 nanocrystalline Photoexcitation and Carrier Collection:
films [230]. The onset potential for the Dynamic Behavior
photocurrent is found to be the same re-
gardless of whether SnO2 , Au, or Pt is In this section, we briefly consider the
used to support the film. A Fermi level response of nanocrystalline semiconduc-
pinned rear interface was used to explain tor–electrolyte interfaces to either pulsed
the results. or periodic photoexcitation. Several points
In general, the voltammograms for are noteworthy in this respect: (1) The
nanocrystalline electrodes are similar to photocurrent rise-time in response to an il-
what is observed for their single-crystal lumination step is nonlinear. Further, the
counterparts. An example of a pho- response is faster when the light intensity
tovoltammogram for CdS is contained is higher. (2) The decay profiles exhibit fea-
in Fig. 25. The fact that a S-type pro- tures on rather slow timescales extending
file is observed culminating in a pho- up to several seconds. (3) The photocur-
ton flux-limited plateau regime is rather rent decay transients exhibit a peaking
surprising given that (1) the film is rather behavior. The time at which this peak
2 µA
Fig. 25 Photovoltammogram
−0.8 −0.6 −0.4 −0.2 0 0.2 under interrupted illumination
of a nanocrystalline CdS-sodium
Potential sulfite electrolyte interface in the
[V vs. Ag/AgCl] reverse-bias regime.
1.8 Chemically Modified Semiconductor–Electrolyte Interfaces 41
occurs varies with the square of the film for semiconductors is that the energy band
thickness, d. (4) A similar pattern is also gap (Eg ) increases, or equivalently, the ab-
seen in IMPS data where the transit time, sorption threshold exhibits a blue shift.
τ is seen to be proportional to d 2 . The critical dimension for size quantiza-
These observations have been inter- tion effects to appear in semiconductors
preted within the framework of two dis- depends on the effective mass (m∗ ) of the
tinct models, one involving trapping or electronic charge carriers. For m∗ ∼ 0.05,
detrapping of the photogenerated elec- the critical dimension is about 300 Å; it
trons [231, 232] and the other based on decreases approximately linearly with in-
electron diffusion (or field-assisted dif- creasing m∗ [7].
fusion) not attenuated by electron local- Size quantization effects and quantum
ization [233, 234]. The millisecond transit dot photoelectrochemistry are discussed in
times also mean that the transit times more detail elsewhere in this volume.
are very long compared to equilibration
of majority carriers in a bulk semiconduc-
tor or electron-hole pair separation within 1.8
the depletion layer of a flat electrode. The Chemically Modified
slow transport is rationalized by a weak Semiconductor–Electrolyte Interfaces
driving force and by invoking percolation 1.8.1
effects [223]. Single Crystals
It is interesting that the response pat-
terns differ for different nanocrystalline Much of the research in the early
electrodes [223]. For example, while trap- 1980s on chemically modified semicon-
ping or detrapping effects appear to be ductor–electrolyte interfaces was directed
relatively unimportant for GaP, the re- toward protecting them from photocor-
sponse for TiO2 , especially at low photon rosion; this body of work has been re-
fluxes, is governed by electron trapping viewed [226]. Parallel efforts also went
or detrapping kinetics. This accounts for into improving minority carrier transfer
the faster response at higher photon fluxes at the interface by chemisorbing metal
(see preceding section). ions such as Ru3+ on the semiconductor
surface. Chemical agents such as sulfide
1.7.5
Size Quantization ions are known to passivate the semicon-
ductor against surface recombination [6].
When electronic particles such as electrons A study [22a] on electron transfer dynam-
and holes are constrained by potential ics at p-GaAs-acetonitrile interfaces where
barriers to regions of space that are com- the semiconductor surface was sulfide-
parable to or smaller than their de Broglie passivated exemplifies this fact. In general,
wavelength, the corresponding allowed en- the mechanistic issue of whether these
ergy states become discrete (i.e. quantized) chemical agents improve minority carrier
rather than continuous. This manifests charge transfer by minimizing surface re-
in the absorption (or emission) spectra combination or by a true catalytic action
as discrete lines that are reminiscent of has not been completely resolved [1].
atomic (line) transitions; these sharper Yet another tactic involves perturbing
features often appear superimposed on a the electrostatics at the semiconduc-
broader envelope. Another manifestation tor–electrolyte interface by adsorbing
42 1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry
(or even chemically attaching) electron course, such ‘‘fixed-charge’’ effects have
donors or acceptors on the semiconductor long been known to the solid-state device
surface [237]. In favorable cases, this physics and gas phase catalysis communi-
increases the band bending at the interface ties. Other agents that have been deployed
by thus introducing a fixed counter- for chemical tuning of the interfacial en-
charge of opposite polarity (negative for ergetics at the semiconductor–electrolyte
a n-type semiconductor) at the junc- interface are listed in Table 2.
tion. Chloride ion adsorption on the Native semiconductor surfaces are fairly
n-GaAs surface from ambient temper- inactive from a catalysis perspective.
ature AlCl3 /n-butylpyridinium chloride Thus, noble metal or metal oxide islands
melts [30, 238] is a case in point; this have been implanted on photoelectrode
process serves to improve the junction surfaces as electron storage centers to
and the photovoltage that it delivers. Of drive multielectron redox processes such
1.8 Chemically Modified Semiconductor–Electrolyte Interfaces 43
elsewhere in this series of volumes, we of the composite because of the high local
briefly discuss in what follows, some concentration of the substrate resulting
simple multicomponent assemblies based from the matrix adsorption process. In
on semiconductors. principle, high surface area supports of the
Early examples in the 1980s were aimed sort that are normally used in the gas-phase
at the design of composite systems for catalysis community can also be used in
photoelectrolytic generation of H2 . Thus, conjunction with TiO2 [286, 287]. These
Nafion and SiO2 were used as supports would include materials such as Al2 O3 ,
for coprecipitated ZnS and CdS for pho- SiO2 , or diatomaceous earth. The resul-
toassisted HER from aqueous sulfide me- tant composite films, however, cannot be
dia [270]. Subsequent work has addressed used as electrodes because of their poor
the mechanistic role of the support in the electronic conductivity.
photoassisted HER [271]. Vectorial elec- The second important feature of a
tron transfer was demonstrated in bipolar metal-semiconductor composite approach
TiO2 /Pt or CdSe/CoS photoelectrode pan- is that the metal can function as a
els arranged in series arrays for the template for chemical or electrochemical
photodecomposition of water to H2 and derivatization to afford a film compris-
O2 [272, 273]. ing molecular redox-semiconductor (or
More recently, matrix-semiconductor even semiconductor-semiconductor) con-
composites, that is, films comprising tacts. Figure 26 generically illustrates the
of semiconductor particles that are dis- occlusion electrosynthesis approach for
persed in a nonphotoactive continuous preparing M/TiO2 composite films and
matrix have been developed. Examples of a subsequent derivatization with ferri/
matrix candidates are metals and poly- ferrocyanide to afford the correspond-
mers [274–279]. Occlusion electrosynthe- ing metal hexacyanoferrate (MHCF)/TiO2
sis is a versatile method for preparing counterparts [288]. These chemically mod-
such composite films as exemplified by the ified films exhibit interesting ‘‘bipolar’’
Ni/TiO2 and Ni/CdS family [280–282]. photoelectrochemical behavior [289] and
Matrix-semiconductor composite films
photoelectrochromic properties [290].
have two virtues from a photoelectro-
chemical perspective. First, their com-
ponents can be separately chosen and 1.9
optimized for a specific function. Thus, Types of Photoelectrochemical Devices
the matrix component can be chosen to
have good adsorption tendency toward As Fig. 27 illustrates, there are basi-
a targeted substrate. The semiconductor cally three types of photoelectrochemi-
component then functions in the role cal devices for solar energy conversion.
of photogenerating charge carriers either The first type is regenerative in nature
for reducing or oxidizing this sequestered and the species that are photooxidized
substrate. This photocatalytic strategy has at the n-type semiconductor electrode
been recently demonstrated both for or- are simply re-reduced at the counter-
ganic substrates (methanol and formate electrode (Fig. 27a). Instead of an elec-
ion) [283, 284] and an inorganic substrate trocatalytic electrode [291, 292] where the
(sulfite) [285]. The net result in either case counterelectrode reaction occurs in the
is an enhanced photocatalytic performance dark (this is the situation schematized
1.9 Types of Photoelectrochemical Devices 45
Au
TiO2
Mn+
TiO2 occlusion
Au
M-TiO2
composite
Derivatization
Au
MHCF-TiO2
composite
transfer dynamics across both these types 9. L. M. Peter, D. Vanmaekelbergh, Adv. Elec-
of interfaces have been described in the trochem. Sci. Eng. 1999, 6, 77.
10. L. M. Peter in Comprehensive Chemical Ki-
dark and under photoexcitation of the netics, (Eds.: R. G. Compton, G. Hancock),
semiconductor. Finally, the various types Elsevier, Amsterdam, 1999, pp. 223–280.
of photoelectrochemical devices for solar 11. R. J. D. Miller, G. McLendon, A. J. Nozik
energy conversion are introduced. Sub- et al., in Surface Electron-transfer Processes,
VCH Publishers, New York, 1995.
sequent chapters in this volume provide
12. K. Rajeshwar, L. M. Peter, A. Fujishima
further elaboration of some of these topics et al., Eds., in Photoelectrochemistry Proc.,
considered herein. The Electrochemical Society, Pennington,
NJ, 1997, Vol. 20.
Acknowledgments 13. N. Sato in Electrochemistry at Metal and
Semiconductor Electrodes, Elsevier, Amster-
dam, 1998.
Research in the author’s laboratory on 14. A. Hamnett in Comprehensive Chemical
semiconductor electrochemistry and pho- Kinetics (Eds.: R. G. Compton), Elsevier,
toelectrochemistry since 1995 is funded, Amsterdam, 1987, p. 61, Vol. 27.
in part, by the Office of Basic Energy Sci- 15. J. S. Newman in Electrochemical Systems,
2nd edition, Prentice-Hall, Englewood
ences, the US Department of Energy. A
Cliffs, NJ, 1991, p. 496.
number of talented and dedicated cowork- 16. R. A. Smith in Semiconductors, Cambridge
ers and colleagues have been involved in University Press, Cambridge, 1964.
collaborative research with the author over 17. F. A. Kröger, H. J. Vink in Solid State Physics
the past twenty years; their names appear (Eds.: F. Seitz, D. Turnbull), Academic
Press, New York, 1956, Vol. 3.
in the publications cited from this labora- 18. F. A. Kröger in The Chemistry of Imperfect
tory. I also thank the University of Texas Crystals, North Holland, Amsterdam, 1964.
at Arlington for providing the facilities 19. N. B. Hannay, Ed., in Semiconductors, Rein-
and infrastructure. I am grateful to Prof. hold, New York, 1959.
20. A. R. West in Solid State Chemistry and Its
S. Licht for comments on an earlier ver-
Applications, John Wiley, New York, 1984.
sion of the manuscript. Last but not least, 21. P. A. Cox in The Electronic Structure and
I thank Ms. Gloria Madden and Ms. Rita Chemistry of Solids, Oxford University Press,
Anderson for assistance in the preparation Oxford, England, 1987.
of this chapter. 22. R. Hoffmann in Solids and Surfaces, VCH,
New York and Weinheim, 1988.
23. K. Uosaki, H. Kita in Modern Aspects of
References Electrochemistry (Eds.: R. E. White, J. O’M.
Bockris, B. E. Conway), Plenum, New York,
1. N. S. Lewis, Acc. Chem. Res. 1990, 23, 176. 1986, pp. 1–60, Vol. 18.
2. L. M. Peter, Chem. Rev. 1990, 90, 753. 24. S. Trasatti in The Absolute Electrode Potential:
3. N. S. Lewis, Annu. Rev. Phys. Chem. 1991, An Explanatory Note, IUPAC Commission
42, 543. I.3 (Electrochemistry), 1984.
4. C. A. Koval, J. N. Howard, Chem. Rev. 1992, 25. H. Reiss, J. Electrochem. Soc. 1978, 125, 937.
92, 411. 26. Yu. V. Pleskov, Yu. Ya. Gurevich in Semi-
5. A. Kumar, W. C. Á. Wilisch, N. S. Lewis, conductor Photoelectrochemistry, Consultants
Crit. Rev. Solid State Mater. Sci. 1993, 18, Bureau, New York, 1986.
327. 27. H. Gerischer, Adv. Electrochem. Eng. 1961,
6. M. X. Tan, P. E. Laibinis, S. T. Nguyen et al., 1, 139.
Prog. Inorg. Chem. 1994, 41, 21. 28. H. Gerischer in Physical Chemistry: An Ad-
7. A. J. Nozik, R. Memming, J. Phys. Chem. vanced Treatise (Eds.: H. Eyring, D. Hender-
1996, 100, 13061. son, W. Jost), Academic Press, New York,
8. N. S. Lewis, J. Phys. Chem. 1998, 102, 4843. 1970, Vol. 9A.
48 1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry
130. M. Nishida, J. Appl. Phys. 1980, 51, 1669. 158. H. Morisaki, M. Hariya, K. Yazawa, Appl.
131. F. E. Guibaly, K. Colbow, Can. J. Phys. 1981, Phys. Lett. 1977, 30, 7.
59, 1682. 159. H. Minoura, M. Tsuiki, Chem. Lett. 1978,
132. D. L. Ullman, J. Electrochem. Soc. 1981, 128, 205.
1269. 160. W. Gissler, J. Electrochem. Soc. 1980, 127,
133. S. Ramakrishna, S. K. Rangarajan, 1713.
J. Electroanal. Chem. 1991, 308, 49. 161. B. Vainas, G. Hodes, J. DuBow, J. Elec-
134. W. Shockley, W. T. Read, Phys. Rev. 1952, troanal. Chem. 1981, 130, 391.
87, 835. 162. N. Müller, G. Hodes, B. Vainas, J. Elec-
135. R. N. Hall, Phys. Rev. 1982, 87, 387. troanal. Chem. 1984, 172, 155.
136. S. J. Fonash in Solar Cell Device Physics, Aca- 163. R. Tenne, W. Giriat, J. Electroanal. Chem.
demic Press, San Diego, 1981, Chapter 6, 1985, 186, 127.
pp. 262–326. 164. D. P. Amalnerkar, S. Radhakrishnan,
137. C. T. Sah, R. N. Noyce, W. Shockley, Proc. H. Minoura et al., Sol. Energy Mater. 1988,
IRE 1957, 45, 1228. 18, 37.
138. J. Reichman, Appl. Phys. Lett. 1980, 36, 574. 165. D. P. Amalnerkar, S. Radhakrishnan,
139. W. I. Albery, P. N. Bartlett, A. Hamnett H. Minoura et al., J. Electroanal. Chem.
et al., J. Electrochem. Soc. 1981, 128, 1492. 1989, 260, 433.
140. W. J. Albery, P. N. Bartlett, J. Electrochem. 166. S.-E. Lindquist, H. Vidarsson, J. Mol. Catal.
Soc. 1983, 130, 1699. 1986, 38, 131.
141. H. Reiss, J. Electrochem. Soc. 1978, 125, 937. 167. P. Singh, R. Singh, K. Rajeshwar et al.,
142. J. Thomchick, A. M. Buoncristiani, J. Appl. J. Electrochem. Soc. 1981, 128, 1145.
Phys. 1980, 51, 6265. 168. D. Lincot, J. Vedel, J. Electroanal. Chem.
1987, 220, 179.
143. W. Lorenz, M. Handschuh, J. Electroanal.
169. D. Lincot, J. Vedel, J. Phys. Chem. 1988, 92,
Chem. 1980, 111, 181.
4103.
144. W. Lorenz, C. Aegerter, M. Handschuh,
170. L. M. Peter in Photocatalysis and the Envi-
J. Electroanal. Chem. 1987, 221, 33.
ronment (Ed.: M. Schiavello), Kluwer Aca-
145. M. Handschuh, W. Lorenz, J. Electroanal.
demic Publishers, Dordrecht, 1988, p. 243,
Chem. 1988, 251, 1.
Vol. 237.
146. P. Lemasson, A. Etcheberry, J. Gautron,
171. B. Ba, H. Cachet, B. Fotouhi et al., Elec-
Electrochim. Acta 1982, 27, 607. trochim. Acta 1992, 37, 309.
147. S. Ramakrishna, S. K. Rangarajan, 172. Y-E. Sung, F. Galliard, A. J. Bard, J. Phys.
J. Electroanal. Chem. 1994, 369, 289. Chem. B 1998, 102, 9797.
148. M. L. Rosenbluth, C. M. Lieber, N. S. Lewis, 173. Y-E. Sung, A. J. Bard, J. Phys. Chem. B 1998,
Appl. Phys. Lett. 1984, 45, 423. 102, 9806.
149. S. Licht, Nature 1987, 330, 148. 174. M. P. Dare-Edwards, A. Hamnett, J. B.
150. S. Licht, D. Peramunage, Nature 1990, 345, Goodenough, J. Electroanal. Chem. 1981,
330. 119, 109.
151. S. Licht, N. Myung, J. Electrochem. Soc. 175. J. Schefold, J. Phys. Chem. 1992, 96, 8692.
1995, 142, 840. 176. K. Uosaki, H. Kita, J. Am. Chem. Soc. 1986,
152. S. Licht, N. Myung, J. Electrochem. Soc. 108, 4294.
1995, 142, 845. 177. S. Kaneko, K. Uosaki, H. Kita, J. Phys. Chem.
153. S. Licht, F. Forouzan, J. Electrochem. Soc. 1986, 90, 6654.
1995, 192, 1539. 178. K. Uosaki, Y. Shigematsu, S. Kaneko et al.,
154. S. Licht, D. Peramunage, J. Phys. Chem. B J. Phys. Chem. 1989, 93, 6521.
1996, 100, 9082. 179. M. Matsumura, S. R. Morrison, J. Electro-
155. H. J. Hovel in Semiconductors and Semimet- anal. Chem. 1983, 144, 113.
als, Solar Cells, Academic Press, New York, 180. M. Matsumura, S. R. Morrison, J. Elec-
1975, Vol. 11. troanal. Chem. 1983, 147, 157.
156. R. R. Potter, J. R. Sites, IEEE Trans. Electron 181. H. J. Lewerenz, J. Stumper, L. M. Peter,
Devices 1984, ED-31, 571. Phys. Rev. Lett. 1988, 61, 1989.
157. R. R. Potter, J. R. Sites, Appl. Phys. Lett. 182. L. M. Peter, A. N. Borazio, H. J. Lewerenz
1983, 43, 843. et al., J. Electroanal. Chem. 1990, 290, 229.
1.10 Conclusion 51
183. S. R. Morrison, T. Freund, J. Chem. Phys. 206. S. Y. Huang, L. Kavan, I. Exnar et al., J. Elec-
1967, 47, 1543. trochem. Soc. 1995, 142, L142.
184. W. P. Gomes, T. Freund, S. R. Morrison, 207. J. G. Highfield, M. Grätzel, J. Phys. Chem.
Surf. Sci. 1968, 13, 201. 1988, 92, 464.
185. W. P. Gomes, T. Freund, S. R. Morrison, 208. B. O’Regan, M. Grätzel, D. Fitzmaurice,
J. Electrochem. Soc. 1968, 115, 818. Chem. Phys. Lett. 1991, 183, 89.
186. K. Micka, H. Gerischer, J. Electroanal. 209. B. O’Regan, M. Grätzel, D. Fitzmaurice,
Chem. 1972, 38, 397. J. Phys. Chem. 1991, 95, 10525.
187. J.-S. Lee, T. Kato, A. Fujishima et al., Bull. 210. B. E. Conway, J. Electrochem. Soc. 1991, 138,
Chem. Soc. Jpn. 1984, 57, 1179. 1539.
188. G. H. Schoenmakers, D. Vanmaekelbergh, 211. R. Könenkamp, R. Henninger, P. Hoyer,
J. J. Kelly, J. Phys. Chem. 1996, 100, 3215. J. Phys. Chem. 1993, 97, 7328.
189. G. H. Schoenmakers, D. Vanmaekelbergh, 212. R. Könenkamp, R. Henninger, Appl. Phys.
J. J. Kelly, J. Chem. Soc., Faraday Trans. A 1994, 58, 87.
1997, 93, 1127. 213. I. Bedja, S. Hotchandani, P. V. Kamat,
190. M. Miyake, H. Yoneyama, H. Tamura, J. Phys. Chem. 1996, 100, 19489.
Chem. Lett. 1976, 635. 214. G. Boschloo, D. Fitzmaurice, J. Phys. Chem.
191. N. Hykaway, W. M. Sears, H. Morisaki B 1999, 103, 2228.
et al., J. Phys. Chem. 1986, 90, 6663. 215. F. Cao, G. Oskam, P. C. Searson et al.,
192. D. J. Fermin, E. A. Ponomarev, L. M. Peter, J. Phys. Chem. 1995, 99, 11974.
in Photoelectrochemistry, Proc., (Eds.: K. 216. R. F. Howe, M. Grätzel, J. Phys. Chem.
Rajeshwar, L. M. Peter, A. Fujishima et al.), 1985, 89, 4495.
The Electrochemical Society, Pennington, 217. L. A. Lyon, J. T. Hupp, J. Phys. Chem. 1995,
99, 15718.
NJ, 1997, Vol. 97–20.
218. H. Lindström, S. Södergren, A. Solbrand
193. G. Nogami, J. H. Kennedy, J. Electrochem.
et al., J. Phys. Chem. B 1997, 101, 7710.
Soc. 1989, 136, 2583.
219. H. Lindström, S. Södergren, A. Solbrand
194. W. J. Albery, N. L. Dias, C. P. Wilde, J. Elec-
et al., J. Phys. Chem. B 1997, 101, 7717.
trochem. Soc. 1987, 134, 601.
220. S. Lunell, A. Stashans, L. Ojamäe et al.,
195. P. Herrasti, L. M. Peter, J. Electroanal.
J. Am. Chem. Soc. 1997, 119, 7374.
Chem. 1991, 305, 241.
221. L. A. Lyon, J. T. Hupp, J. Phys. Chem. B
196. J. Li, L. M. Peter, J. Electroanal. Chem. 1985, 1999, 103, 4623.
182, 399. 222. R. van de Krol, A. Goossens, J. Schoonman,
197. B. H. Erné, D. Vanmaekelbergh, I. E. Ver- J. Electrochem. Soc. 1997, 144, 1723.
meir, Electrochim. Acta 1993, 38, 2559. 223. J. J. Kelly, D. Vanmaekelbergh, Electrochim.
198. J. Li, R. Peat, L. M. Peter, J. Electroanal. Acta 1998, 43, 2773.
Chem. 1986, 200, 333. 224. A. Zaban, A. Meier, B. A. Gregg, J. Phys.
199. R. Peat, L. M. Peter, Electrochim. Acta 1986, Chem. B 1997, 101, 7985.
31, 731. 225. D. Vanmaekelbergh, P. E. de Jongh, J. Phys.
200. R. Peat, L. M. Peter, J. Electroanal. Chem. Chem. B 1999, 103, 747.
1986, 209, 307. 226. H. Rensmo, H. Lindström, S. Södergren
201. D. Vanmaekelbergh, J. J. Kelly, J. Phys. et al., J. Electrochem. Soc. 1996, 143, 3173.
Chem. 1990, 94, 5406. 227. G. Hodes, I. D. J. Howell, L. M. Peter,
202. W. P. Gomes, S. Lingier, D. Vanmaekel- J. Electrochem. Soc. 1992, 139, 3136.
bergh, J. Electroanal. Chem. 1989, 269, 237. 228. A. Hagfeldt, U. Björkstén, S.-E. Lindquist,
203. R. Tenne in Semiconductor Micromachining: Sol. Energy Mater. Sol. Cells 1992, 27, 293.
Fundamentals Electrochemistry and Physics 229. S. Södergren, A. Hagfeldt, J. Olsson et al.,
(Eds.: S. A. Campbell, H.-J. Lewerenz), John J. Phys. Chem. 1994, 98, 5552.
Wiley, New York, 1994, pp. 139–175, Vol. 1. 230. A. Shiga, A. Tsujiko, T. Ide et al., J. Phys.
204. J. van de Ven, H. J. P. Nabben, J. Elec- Chem. B 1998, 102, 6049.
trochem. Soc. 1990, 137, 1603. 231. K. Schwarzburg, F. Willig, Appl. Phys. Lett.
205. G. A. Somorjai in Chemistry in Two Dimen- 1991, 58, 2520.
sions: Surfaces, Cornell University Press, 232. P. E. de Jongh, D. Vanmaekelbergh, Phys.
Ithaca, NY, 1981. Rev. Lett. 1996, 77, 3427.
52 1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry
287. N. Takeda, M. Ohtani, T. Torimoto et al., 293. A. Fujishima, K. Honda, Nature 1972, 238,
J. Phys. Chem. B 1997, 101, 2644. 37.
288. N. R. de Tacconi, K. Rajeshwar, R. O. Lezna, 294. B. O’Regan, M. Grätzel, Nature 1991, 353,
Electrochim. Acta 2000, 45, 3403. 737.
289. N. R. de Tacconi, J. Carmona, K. Rajeshwar, 295. J. Manassen, G. Hodes, D. Cahen,
J. Phys. Chem. B 1998, 101, 10151. J. Electrochem. Soc. 1977, 124, 532.
290. N. R. de Tacconi, J. Carmona, W. L. Balsam 296. S. Licht, G. Hodes, R. Tenne et al., Nature
et al., Chem. Mater. 1998, 10, 25. 1987, 326, 863.
291. G. Hodes, J. Manassen, D. Cahen, J. Appl. 297. S. Licht, B. Wang, T. Soga et al., Appl. Phys.
Electrochem. 1977, 7, 182. Lett. 1999, 74, 4055.
292. G. Hodes, J. Manassen, D. Cahen, J. Elec- 298. S. Licht, J. Phys. Chem. B 2001, 105,
trochem. Soc. 1977, 127, 544. 6281.
59
e φSC EF,n
Energy
EF,n ERed / Ox ERed / Ox
Distance
(a) (b)
Fig. 1 A scheme of the energetics at an n-type semiconductor
electrode in contact with a redox system in an electrolyte solution.
(a) The situation under conditions of electronic equilibrium. The
electrochemical potential of the electrons is the same in both
phases, i.e. the electron Fermi-level in the semiconductor EF,n has
the same value as the Fermi-level of the electrons in the redox
system ERed/Ox . (b) Case in which the energy bands in the
semiconductor are flat; this situation, corresponding to maximum
photovoltage, is reached under strong illumination at open circuit.
WSC is the width of the depletion layer and eφSC is the
band-bending.
band gap of the solid. Electrons have been oxidizing the reduced species
transferred from the conduction band (CB)
of the semiconductor to solution creating Red + hVB + −−−→ Ox+ (1)
a space charge layer (a depletion layer
of width WSC ) within the semiconductor The electrons are detected as photocurrent
with a band bending eφSC . In equilibrium, in the external circuit. This ‘‘short-circuit
the Fermi level is constant throughout the photocurrent’’ depends on the competition
system. There are various ways in which between hole transfer across the interface
equilibrium can be disturbed, for example, (the Faraday reaction) and electron-hole
by illumination, by an externally applied recombination in the bulk semiconductor
potential, or by charge carrier injection and at the surface of the electrode. Under
from solution. open-circuit conditions no photocurrent
A photovoltaic (two-electrode) cell oper- can flow in the external circuit. The
ates without an external voltage source [22, Fermi level in the semiconductor is
23]. The working electrode is illuminated. raised with respect to that in solution
Two limiting cases can be distinguished. and a photovoltage is established. In the
If the electrodes are short-circuited, then limiting case at high light intensity, the
the electrons and holes, generated by energy bands become flat (Fig. 1b). The
supra–band gap light, are separated by mi- maximum photovoltage is determined by
gration within the depletion layer and by the difference in Fermi level under flat
diffusion. The holes react at the interface, band and equilibrium conditions. Clearly,
62 2 Experimental Techniques
such a system can supply electrical power into a p-type semiconductor can give rise
(Sect. 2.1.2.2). to light emission [25].
Unlike photovoltaic cells, most photo- Macroporous semiconductor electrodes
electrochemical systems do not operate resemble in many respects the bulk elec-
without an external source. In a more trodes described earlier. However, there
general approach, the potential of the are clear differences between porous and
working electrode can be varied with re- nonporous systems that become even
spect to the equilibrium potential Veq (or more pronounced in nanoporous elec-
to the potential of a RE) by means of an trodes. Porous systems are considered in
external voltage source (e.g. a potentiostat) Sect. 2.1.5.
connected between WE and CE [21, 24]. Recently, insulating nanocrystals have
The current density (indicated here by j ) been attached to conducting substrates
is measured both in the dark and under (metal or indium tin oxide) by van der
illumination as a function of the applied Waals interactions or covalent bonding.
potential V . This is described in more When such a system is used as a working
detail in Sect. 2.1.2.2. electrode in a photoelectrochemical cell,
A strong oxidizing agent (OxS + ), that a photocurrent (in the µA range) can be
is one with electron acceptor levels cor- measured [19, 20, 26]. Clearly, a monolayer
responding to the valence band of the of nanocrystals can absorb a sufficient
semiconductor, can extract electrons from fraction of the incident light to give to a
the band thus creating holes measurable photocurrent. The absorption
of a photon in a nanocrystal leads to a
OxS + −−−→ RedS + hVB + (2)
photoexcited state; an electron is promoted
In an n-type semiconductor, these holes from a lower energy level (HOMO or
can recombine with majority carriers (elec- valence band) to a higher energy level
trons) from the CB [25]. Electrolumines- (LUMO or CB) (Fig. 2). A photocurrent
cence (EL) is expected from a semiconduc- will be observed if the electron and
tor with a direct band gap for conditions in hole are separated effectively before the
which the surface electron concentration system relaxes to the ground state. In
is significant. Similarly, electron injection Fig. 2, the electron is transferred to the
metal, the hole to an oxidizable species in mechanisms two complex systems are
solution. This leads to an upward shift considered. Finally, charge carrier trans-
of the electrochemical potential of the port in porous semiconductor systems is a
electrons µe and a downward shift of topic deserving special attention.
the electrochemical potential of the holes
µh (and thus to a photovoltage Vphoto = 2.1.2
(µe − µh /e). Such a photoelectrochemical Photoelectrochemical Characterization
Methods
system shows a strong analogy with
molecular (e.g. dye) systems adsorbed on
For scientific research on photoelectro-
metal or semiconductor electrodes. chemical systems, the photoelectrode is
The photoelectrochemical solar cell of generally the working electrode in a three-
O’Regan and Grätzel [16] combines the electrode electrochemical cell. Using this
special features of semiconductor poros- set up, the system can be investigated by
ity and sensitization described earlier. perturbing it and recording the system’s
This system consists of nanometer-sized response. In general, such perturbation-
TiO2 crystallites interconnected to form a response methods may be classified ac-
three-dimensional porous assembly. Dye cording to the time dependence of the
molecules acting as the light absorber are perturbation (steady state versus time-
chemisorbed on the internal surface. The resolved) or to the physical nature of
efficiency of light absorbance is close to the perturbation and the response. In
unity as a result of the fact that photons this section, the principles of steady state
encounter an adsorbed dye molecule many and time-resolved methods in general are
times. The electrolyte solution, containing first discussed. Later on, two different
an I − /I3 − redox system, permeates the perturbation-response techniques used to
pores of the system. The TiO2 assembly study semiconductor-based photoelectro-
acts as an electron conductor and the elec- chemical systems are discussed. Finally,
trolyte as the ‘‘hole conductor’’. Because methods in which illumination or minority
of the interpenetration of the electron and carrier injection gives rise to light emission
hole conducting phases on a nanometer from the electrode are considered.
scale, back transfer of the electron to I3 −
is the main source of recombination. 2.1.2.1 Steady State and Time-resolved
The three systems described earlier have Methods
a number of features in common. To ob- Steady state methods are basically simple:
serve photocurrent, light obviously has one applies a time-independent pertur-
to be absorbed by the system. The spa- bation x to a system and records – if
tial separation of photogenerated electrons it exists – the time-independent response
and holes must be more effective than y(x) of the system. This may be repeated
their recombination. This requires effi- for different levels of the perturbation. In
cient kinetics for electron transfer across this way, a functional relation between
the solid/solution interface. These are perturbation and response is determined,
topics that are addressed in the remain- which should provide information on the
der of the chapter. Electrode reactions system studied. Consider, for example, an
in photoelectrochemical systems may be electrical resistor through which a current
quite complicated. To illustrate the gen- is passed (perturbation). Instantaneously,
eral approach to the elucidation of reaction the system responds by maintaining a
64 2 Experimental Techniques
times with success, for example, to identify by illumination are effectively separated
(light-intensity dependent) photocurrent by migration and diffusion (Fig. 4). PL,
multiplication processes (see Sect. 2.1.3.3). on the other hand, is expected when the
The optoelectrical transfer function relates photogenerated carriers recombine radia-
an incident sinusoidally modulated light tively, that is, at potentials approaching the
flux and the resulting modulated pho- flat band potential Vfb (see Sect. 2.1.3.1).
tocurrent density. This quantity is defined Emission may result either from direct
as [30] band-band recombination or from indirect
recombination via a band gap state [25]. In
#j˜PH (ω) the simplest case in which surface recom-
YOE (ω) = (8)
!
e#%(ω) bination can be disregarded, the potential
dependence of the emission intensity IPL
Again, it should be stressed that the opto- is described by the Gärtner equation:
electrical transfer function is defined using " #
small-signal perturbations, superimposed exp(−αWSC )
IPL = κ% (9)
on time-independent bias signals. Mea- 1 + αLmin
surement of the optoelectrical impedance
where % is the incident photon flux and κ
is often referred to as intensity modulated
is the ratio of the rate of radiative recombi-
photocurrent spectroscopy (IMPS).
nation to the total recombination rate. In
principle, it is possible to obtain values for
2.1.2.4 Luminescence-based Techniques Lmin , WSC , and α from the potential depen-
Photoluminescence (PL), like photocur- dence of IPL . The Gärtner model has been
rent, is a technique in which a light flux extended by Gerischer and coworkers [31,
incident on the working electrode acts as a 32] to account for surface recombination.
perturbation. In this case, the response of In this case, surface recombination rates
the system is followed by measuring the in- can be obtained from photoluminescence
tensity of the emitted light. As in photocur- measurements.
rent measurements, the electrode potential While photoluminescence has been
can be used as an additional variable. mainly studied under steady state con-
As explained in Sect. 2.1.2.2, photocur- ditions or with transient techniques,
rent is obtained when, under depletion harmonic-modulation measurements are
conditions, electrons and holes created possible. Beckmann and Memming [33]
PL PC
Intensity /Current
EL
Potential
10
Vfb − kT/e
Fig. 6 Mott-Schottky curve determined
5 at an n-InP electrode in a 1.2 M HCl
aqueous solution. The interfacial
capacitance is determined from the
0 electrochemical impedance measured at
8.2 kHz using a parallel connection of a
−500 0 500 resistor and a capacitor in series with
V / mV vs. SCE the cell resistor.
2.1 Photoelectrochemical Systems Characterization 71
understood if majority charge carriers are Obviously, such so-called frequency dis-
depleted from the semiconductor surface. persion hampers a proper determination
In that case, the capacitance CSC is given of Vfb although reliable values are generally
by the Mott-Schottky equation, which for obtained at high measuring frequencies
an n-type semiconductor reads [36, 37, 39]: (>10 kHz). The origin of this nonideal
$ % $ % behavior is not well understood [37, 41].
eND εε0 1/2 kB T −1/2 However, to check the reliability of Mott-
CSC = φSC −
2 e Schottky measurements, the capacitance
(10) should be measured in a broad frequency
where T is the absolute temperature range [42–44].
and the constants e and kB indicate
the elementary electrical charge and the
2.1.3.2Electron-hole Recombination
Boltzmann constant, respectively. The Dynamics at the s/e Interface
capacitance CSC can be identified directly
with the interfacial capacitance if CSC % 2.1.3.2.1 Introduction The dynamics of
CEL . Although exact values of CEL are electron-hole recombination at the semi-
scarce in the literature, this inequality conductor surface has been extensively
is generally assumed to be fulfilled for studied both at illuminated and at dark
moderately doped semiconductors under s/e interfaces [45–53]. For recombination,
depletion conditions [36, 37, 39, 40]. As minority charge carriers should be present
a consequence, a change of the potential at the interface. For n-type semiconduc-
drop φ across the interface results mainly tors, holes may be supplied to the surface
in a change of φSC (i.e. φEL is approximately by illumination under depletion condi-
constant). Hence, φSC may be written tions using supra-band gap light (see
as the difference between the applied Sect. 2.1.2.2). Alternatively, holes may be
electrode potential V and the so-called flat injected into the valence band by a strong
band potential Vfb [37]: oxidizing agent in the electrolyte solution.
If a depletion layer exists, the injected holes
φSC = V − Vfb (11)
accumulate at the semiconductor surface.
At the flat band potential, there is no As it gives a nice and relatively simple
potential drop across the semiconductor illustration of the use of various charac-
and, hence, the semiconductor energy terization methods, we will discuss the
bands are flat from the bulk up to subject of electron-hole recombination dy-
the semiconductor surface (see Fig. 1b). namics in the next sections, taking the
Moreover, because φEL is approximately n-GaAs photoanode as the main example.
constant, the energy bands are fixed at More complicated topics – related to inter-
the semiconductor surface. Hence, the facial transfer of photogenerated charge
position of the band edges at the surface carriers – are discussed in Section 2.1.3.3.
may be calculated once the flat band
potential is known. 2.1.3.2.2 Current Density Versus Potential
A problem often encountered when Measurements Figure 7 shows a current
calculating the flat band potential from density versus potential curve for an il-
Mott-Schottky data is the frequency de- luminated n-GaAs electrode (λ = 480 nm)
pendence of both the slope and the extrap- in a 0.1 M H2 SO4 solution. As reported
olation point of the C −2 versus V curve. for various photoanodes in the literature,
72 2 Experimental Techniques
(see text).
Vfb
0
−1000 −750 −500 −250 0
V / mV vs. SCE
the j versus V plot shows – at potentials For various illumination intensities, the
more positive than the flat band poten- diameter of the semicircle fitting the data
tial – three different regions, indicated as at high frequencies equals approximately
A, B, and C in Fig. 7 [45, 47, 49, 50, 54]. In kT /e|jPH | [45–47, 49]. In addition, it was
region A, that is closest to Vfb , no current shown that upon illumination, a capacitive
passes through the s/e interface, despite peak appears in the C −2 versus V plot of
the minority charge carrier flux towards the n-GaAs|0.1 M H2 SO4 interface [45, 46,
the surface. On the other hand, a pho- 51]. The peak value proved to be a func-
tocurrent plateau appears in region C – at tion of the frequency and the photocurrent
the most positive potentials. The interme- density as measured in region C [51]. This
diate region B shows a transition between behavior is markedly different from the
these two extreme situations. Clearly, the purely capacitive impedance (vertical line
occurrence of a photocurrent plateau is in the Nyquist plane and straight Mott-
in accordance with the Gärtner-equation: Schottky plot) expected for a blocking s/e
in this potential region, all photogenerated interface (see Sect. 2.1.3.1).
holes reach the semiconductor surface and The appearance of both the semicircle
participate in a charge-transfer reaction and the capacitive peak were accounted
(the oxidation of the semiconductor). On for by Vanmaekelbergh and coworkers,
the other hand, since α −1 % WSC in the by considering recombination of photo-
case of n-GaAs, the absence of photocur- generated holes with CB electrons at the
rent in region A is not accounted for by semiconductor surface [51, 55–57]. The
the Gärtner-equation. This indicates a loss recombination mechanism assumed by
of photogenerated holes, which can be at- these authors consists of the successive
tributed to electron-hole recombination at capture of an electron in an empty surface
the semiconductor surface. state and of a hole in an occupied sur-
face state. Taking the rates of the electron
2.1.3.2.3 The Electrochemical Impedance (hole) capture steps to be first order in
of Surface Recombination Figure 8 shows the CB electron density nS (valence band
the impedance spectrum of the illumi- hole density pS ) and the density of empty
nated n-GaAs|0.1 M H2 SO4 interface, as (filled) surface states, an electrochemical
recorded in the potential region A (no impedance corresponding to the equiva-
steady state photocurrent). One can see lent circuit shown in Fig. 9 was calculated
a small semicircle at high frequencies. for this recombination mechanism. The
2.1 Photoelectrochemical Systems Characterization 73
−100
R1
[Ω cm2]
Im(Z )
−50 512 Hz
32 kHz
4 kHz
0
0 50 100 150
Re(Z )
[Ω cm2]
Fig. 8 High-frequency electrochemical impedance spectrum,
obtained at an illuminated n-GaAs electrode in a 0.1 M H2 SO4
aqueous solution. Bias potential: – 600 mV versus SCE (i.e.
potential region A); limiting photocurrent density: 380 µA cm−2 .
The resistance R1 equals 71 + cm2 , that is, 1.06 × kT/ejPH .
impedance of the three different circuit space charge layer of the semiconductor,
elements read [51]: the overall impedance of the s/e interface
consists of the parallel connection of the
kT
R1 = (12) surface-recombination impedance and the
e|jREC | space charge layer capacitor. At high fre-
kT βn nS 1 quencies, the recombination impedance
R2 = (13)
e|jREC | βp pS 1 − |jREC /jPH | reduces to the resistor R1 . Hence, the
e|jREC | 1 high-frequency impedance of the inter-
C1 = (14) face corresponds to the parallel connection
kT βn nS
of R1 and CSC . This parallel combination
where jREC equals the current density accounts for the experimentally observed
associated with recombination and βn (βp ) high-frequency semicircle (note the cor-
denote the rate constant for electron (hole) respondence between the experimental
capture. diameter and R1 ). In addition, the features
Because the photocurrent density adds of the capacitive peak in the Mott-Schottky
to the current density that (dis)charges the plot could be qualitatively explained by
74 2 Experimental Techniques
considering the parallel connection of the addition, it was shown that the typical
recombination impedance and the space semicircle resulting from the parallel con-
charge layer capacitor [51]. nection of the space charge layer capacitor
Clearly, electron-hole recombination is and a resistor with resistance kT /e|j | is not
not limited to illuminated semiconductors. uniquely related to surface-recombination
Minority charge carriers also may be in- processes. Any reaction step, the rate of
jected by an oxidizing or reducing agent in which is proportional to the density of
the electrolyte solution (see Sects. 2.1.1.2 majority charge carriers at the interface,
and 2.1.2.4). Also in this case, the parallel may contribute this feature to the over-
connection of the surface recombination all impedance [59–61]. Hence, for direct
impedance and the space charge capac- transfer of majority charge carriers or
itor provides an accurate description of for surface-state mediated transfer, this
the experimental impedance [45]. For in- semicircle also may appear in the over-
stance, the impedance spectrum of the n- all impedance spectrum [59]. These latter
GaAs|Ce4+ system – Ce4+ is a well-known examples demonstrate the need for a
hole-injecting agent for n-GaAs – shows reliable dynamic model of the charge-
a capacitive semicircle with a diameter transfer reaction for the interpretation of
equal to kT /e|j | at high frequencies [45, impedance data.
58]. It was also demonstrated that a capac-
itive peak, exhibiting the same functional 2.1.3.2.4 The Optoelectrical Transfer Func-
dependence on frequency and current tion of Surface Recombination If the n-
density as obtained with the n-GaAs pho- GaAs|0.1 M H2 SO4 system, polarized in
toanode, is present in a Mott-Schottky potential region A, is suddenly exposed
plot measured with the n-GaAs|Ce4+ sys- to a constant illumination, a photocur-
tem [45]. rent transient decaying from the value as
Both examples attribute convincingly the measured in potential region C to zero is
loss of photogenerated holes at the n-GaAs recorded [30]. This photocurrent decay is
photoanode polarized in potential region related to the increase from zero recom-
A to electron-hole recombination at the bination at the moment of the exposure
electrode surface. Because similar results (maximum photocurrent) to complete re-
have been obtained at the n-CdTe photoan- combination (no photocurrent). The same
ode [50], one could think of the surface- picture arises from the optoelectrical trans-
recombination impedance as a general fer function (Fig. 10), which corresponds
fingerprint of surface-recombination steps to a semicircle ranging from YOE = 1
in an overall reaction mechanism. This at high frequencies to YOE = 0 at low
is, however, not the case. For the n-InP frequencies [62]. As for the transient at
photoanode, the features of the surface short times, recombination is ruled out
recombination are absent although re- at high modulation frequencies leading
combination was shown to occur at the to a maximum value of the modulated
semiconductor surface [54]. In this partic- photocurrent, whereas at low frequencies
ular case, this discrepancy was resolved recombination is fully operative, causing
by assuming that recombination does not the disappearance of the photocurrent.
occur at fixed recombination centers but This impedance could be analyzed using
rather at intermediates of the anodic de- the same dynamic model that describes the
composition of the semiconductor [53]. In surface-recombination impedance. If the
2.1 Photoelectrochemical Systems Characterization 75
0.5
104 Hz
i˜R
i˜h
)
ω max = 2π (4000) s−1
Im(1 −
103 Hz
0.0
0.0 0.5 1.0
i˜
Re(1 − R )
i˜h
Fig. 10 Plot of the optoelectrical transfer function measured at an illuminated
n-GaAs electrode in a 1 M H2 SO4 aqueous solution. Bias potential: – 600 mV
versus SCE (i.e. potential region A); limiting photocurrent density:
800 µA cm−2 (from Ref. 62).
1.0
3
−In(IPL/Imax)
0.8 2
1
0.6
[a.u.]
IPL
0
0 10 20 30 40 50 60 70
0.4
WSC
[nm]
0.2
0.0
−1.0 −0.5 0.0 0.5 1.0 1.5 2.0
V vs SCE
[V ]
Fig. 11 The intensity of the PL measured at 2.2 eV (the ‘‘yellow’’ emission) from an n-type
GaN electrode as a function of potential. The excitation energy was 4 eV. In the inset the
results are plotted according to Eq. (16) (from Ref. 25).
approach and to determine the absorption cathodic reduction. These processes intro-
coefficient [63] (see Fig. 11). Alternatively, duce surface or near-surface states that
if α is known accurately then Eq. (16) provide pathways for nonradiative recom-
gives information about the potential bination. Gerischer and coworkers [31, 32]
distribution within the semiconductor, as have studied such systems using an ex-
shown by Ellis and coworkers for GaAs [63, tended Gärtner model.
64]. If α values are either available or There are numerous reports on the po-
determined for different wavelengths, then tential dependence of the electrolumines-
the minority carrier diffusion length [Lmin cence for n-type electrodes under (quasi)
in Eq. 9] can be obtained. steady state conditions and a limited num-
In many cases, the photoluminescence ber for p-type electrodes [25]. Such studies
intensity of n-type electrodes at nega- were mainly used for diagnostic purposes,
tive potential does not attain a constant that is, to investigate the role of minority
value as would be expected from Eq. (16) carriers in electrode processes.
[25]. Hysteresis is also frequently encoun- While there have been reports of time-
tered in cyclic scanning experiments [25]. resolved photoluminescence [65–68] and
Such effects are due to potential-dependent electroluminescence experiments [69–71]
changes in the surface chemistry. Under for the study of recombination, trapping
accumulation conditions in aqueous so- and detrapping of photogenerated charge
lution, hydrogen is evolved and may be carriers, potential modulation has scarcely
incorporated into the electrode. In ad- been used. An illustration of the possi-
dition, the semiconductor may undergo bilities of this approach is provided by
2.1 Photoelectrochemical Systems Characterization 77
work of Decker and coworkers [72] who that is capable of injecting an electron
used a frequency-dependent train of po- into the CB. This electron also contributes
tential pulses to study the effect of surface to the photocurrent. As a result, for each
chemistry on the radiative recombination photon absorbed in the system two charge
of holes injected into n-type GaAs from carriers are registered as current. This cor-
Fe(CN)6 3− in solution. An unexpected responds to a ‘‘quantum efficiency’’ of two.
potential-dependence of the emission was For formate oxidation at n-type ZnO, Mor-
attributed to a change in the surface termi- rison and coworkers [74, 75] suggested a
nation from hydroxide to hydride coverage mechanism of the type
on going from positive to negative po-
HCOO− + hVB + −−−→ H+ + COO −
•
tentials. A time constant of 0.1 ms was
(17)
estimated for this transition. This trans-
COO − −−−→ CO2 + eCB − (18)
•
formation was subsequently observed by
Erné and coworkers using in situ in-
frared spectroscopy [73]. It is clear that Results obtained by Honda and cowork-
potential-modulated luminescence tech- ers [76] showed that the reaction is more
niques deserve wider attention. complicated. The ZnO electrode is, in fact,
dissolved during the current-doubling re-
action. The stoichiometry found in their
2.1.3.3 Complex Charge-transfer
work is given by
Processes
In this section we use two types of com- ZnO + 2HCOO− + 2hVB + −−−→
plex charge-transfer reaction to illustrate
the general approach to the elucidation of Zn2+ + H2 O + CO2 + 2eCB − (19)
reaction mechanisms at single crystal elec- This result calls into question mecha-
trodes. These reactions are photocurrent nisms in which the reducing agent reacts
doubling at n-type and p-type semicon- directly with valence-band holes. Honda
ductors and the (photo)anodic oxidation of and coworkers proposed that formate is
the semiconductor itself. oxidized by atomic oxygen formed in the
photoanodic oxidation of ZnO.
2.1.3.3.1 PC-doubling Reactions PC doub-
ZnO + 2hVB + −−−→ Zn2+ + Oads (20)
ling refers to a type of charge-transfer
reaction in which both bands of the semi- A subsequent study of the dependence of
conductor are involved, thus emphasizing the kinetics of the photocurrent-doubling
the distinctive features of semiconduc- reaction on the light intensity supports
tor electrochemistry. The first examples this idea [77]. The following mechanism
of such reactions relate to the photoan- was suggested
odic oxidation of species such as formate
Oads + HCOO− −−−→ OH + COO −
• •
and tartrate at wide band gap n-type elec-
(21)
trodes [74, 75]. A photon generates an
OH + HCOO− −−−→ H2 O + COO −
• •
electron-hole pair in the semiconductor.
The electron and hole are separated by the (22)
electric field of the depletion layer. The The stoichiometry of this scheme agrees
electron is detected as photocurrent in the with that reported by Honda and cowork-
external circuit. The hole oxidizes a species ers. This result could be checked by study-
from solution producing an intermediate ing the kinetics of two types of competing
78 2 Experimental Techniques
yield information about the reaction kinet- the p-type semiconductor dissolves anod-
ics. Peter and coworkers [86, 87] were the ically in the dark showing an exponential
first to show that rate constants for major- increase of current with increasing poten-
ity carrier injection could be determined tial, whereas the n-type semiconductor can
for such systems by IMPS. They studied only be oxidized if minority carriers are
current doubling for oxygen reduction at generated by light under depletions condi-
p-type GaAs and GaP. From a complex- tions (see Fig. 7). The anodic oxidation of
plane representation of the optoelectrical most semiconductors is a complicated pro-
impedance, rate constants in the range cess. The reaction of GaAs, for example,
104 –105 s−1 were calculated. This result which can be represented schematically by
suggests that hole injection is a thermally
•
activated process. HO2 , postulated as the
GaAs + 6hVB + −−−→ Ga(III) + As(III)
injecting species, gives rise to a surface
(25)
energy level; in the case of GaAs, this is
requires six valence-band holes to form
located about 0.4 eV above the valence-
trivalent gallium and arsenic species.
band edge.
Chemical reactions are obviously also
involved. The final products depend on
2.1.3.3.2 Dissolution of Semiconductors the nature of the electrolyte solution. The
Another class of complex reactions that information that can be obtained from
has been widely studied is the oxidative cyclic voltammetry is clearly rather limited.
dissolution of elemental and compound Kinetic studies in which oxidation of a re-
semiconductors. There are a number of ducing agent present in solution competes
reasons for the interest in these systems.
with oxidation of the semiconductor have
The possible application of semiconduc-
yielded a wealth of information, allowing
tors in regenerative photoelectrochemical
quite detailed dissolution mechanisms to
solar cells required the complete suppres-
be proposed [88]. Generally, in these stud-
sion of corrosion of the photoelectrode [5,
ies an RRDE is used; the products formed
6, 22, 23]. On the other hand, with the de-
at the semiconductor disk are detected
velopment of the optoelectronics industry
electrochemically at a noble metal ring.
based on III–V materials there was a need
for a more fundamental understanding of Semiconductors can be etched under
etching processes [15]. The revival of inter- open-circuit conditions with an oxidiz-
est in porous semiconductors, triggered by ing agent that is capable of being re-
the discovery of the unusual optical prop- duced by extracting electrons from the
erties of porous Si, led to a general revival valence band, that is, creating holes in
of interest in the mechanisms of porous the band (see Sect. 2.1.3.2.3) [15]. Infor-
etching [18]. mation on hole injection can be obtained
Because the bonding states correspond from electroluminescence studies on n-
to the valence band of a semiconductor, type electrodes. Electrochemical studies of
one expects holes to be important for such ‘‘electroless’’ etching systems have
the oxidation reaction [15, 24]. This is the shown that, during dissolution, interme-
case with most etching systems. The im- diates are formed with energy levels in
portance of holes is immediately clear the band gap [89]. From such states, elec-
from simple current-potential or cyclic trons can be thermally excited into the CB.
voltammetric measurements. Generally, If this occurs in an n-type semiconductor
80 2 Experimental Techniques
under depletion conditions then the in- InP systems, providing information about
jected electron can be detected as current the sequence of the reaction steps and
in the external circuit. By measuring this the magnitude of the rate constants for
anodic current as a function of injection majority carrier injection [54, 92].
rate and temperature, one gets extensive Because electron injection is detected
information about both chemical and elec- during dissolution of n-type semiconduc-
trochemical processes [88, 89]. Gomes and tors under illumination, it seems likely that
coworkers have shown that electron injec- it should also occur during dissolution of
tion and RRDE competition studies are p-type semiconductors in the dark. There
complementary [90]. The contribution of are two ways in which this can be checked.
electron injection to the total anodic oxi- If the injected electrons recombine radia-
dation rate is, for most etching systems, tively with majority carriers (the holes),
small compared to that of holes. light is emitted. EL corresponding to band-
Electron injection from reaction inter- band recombination has been observed
mediates of the oxidation of n-type semi- during anodic dissolution of p-type InP in
conductors can be observed as quantum ef- HCl solution (Fig. 12) [93]. The quantum
ficiency larger than unity in photocurrent- efficiency for light emission, that is, the
potential measurements. There are two number of photons emitted per electron
striking examples in the literature: the passed through the external circuit, was
photoanodic dissolution of n-type silicon very low (approximately 10−6 ). This proba-
in HF solution [91, 92] and of n-type InP bly is due to a low rate of electron injection
in HCl solution [54]. In these cases the and a high rate of nonradiative recom-
quantum efficiency at low light intensity bination at the etching surface. Another
is exceptionally high, four for silicon and approach that allows one to distinguish
two for InP. In the case of Si, this means quantitatively between electron and hole
that only one photon (and thus one hole) contributions to the current is the p-n junc-
is required to dissolve each silicon atom; tion configuration [94, 95]. This technique
three electrons are injected into the con- has been used for the study of the anodic
duction band dissolution of silicon in alkaline solution.
This is an unusual system showing a novel
Si + hVB + −−−→ Si(IV) + 3eCB − (26)
coupling of chemical and electrochemical
At higher light intensity, the quantum effi- steps [95]. Both p-type and n-type silicon
ciency drops indicating that reaction steps can be oxidized electrochemically in the
involving valence-band holes take over dark. In both cases, passivation occurs if
from electron injection steps. A quantum a limiting current is exceeded. The anodic
efficiency of two for InP means that three peak current observed as n-type silicon
of the six oxidation steps require minority passivates is comparable to that of p-type
carriers and thus photons. As for silicon, silicon. Anodic current from an n-type
the quantum efficiency decreases (from 2 electrode in the dark can, under normal
to 1) as the light intensity is increased. As circumstances, only result from electron
described in the previous section, IMPS injection into the CB. Because there is no
can be used very effectively to study the electron donor present in solution, elec-
mechanisms of ‘‘photocurrent multiplica- tron injection must occur from a surface
tion’’ reactions. The method has proved species. This species very likely arises as
particularly successful for the silicon and a result of the chemical-etching reaction
2.1 Photoelectrochemical Systems Characterization 81
1.0
1.0
0.8
0.8
0.6
0.4
0.6 0.2
[a.u.]
IEL
0.0
800 900 1000
0.4 Wavelength
[nm]
0.2
0.0
3.0
[mA/cm2]
2.0
j
1.0
0.0
0.0 0.2 0.4 0.6
V vs SCE
[V]
Fig. 12 The potential dependence of the EL intensity IEL and of
the current density j for a p-type InP electrode, dissolving
anodically in a 1 M HCl solution. The inset shows the EL
spectrum (from Ref. 93).
with photogenerated or injected holes; this also can be used to get information about
process competes with the oxidation of the electron injection processes [97].
semiconductor by holes. Electrochemical measurements of the
The assumption of pinned band edges is type described earlier give indirect evi-
very often not valid. Lincot and Vedel [50] dence about dissolution processes. More
in an early study of the photoanodic dis- direct chemical information can be ob-
solution of n-type CdTe used EIS to show tained from in-situ spectroscopies, in
that the Fermi level becomes pinned in particular from IR and Raman methods.
a wide potential range positive with re- Chazalviel and coworkers have showed the
spect to Vfb . This means that in this range power of this approach in studies on sili-
the band-bending within the semiconduc- con and GaAs [73, 98, 99]. Electrochemical
tor remains constant while the potential and spectroscopic techniques are macro-
changes across the Helmholtz layer. In scopic methods giving a view of the whole
their analysis, Lincot and Vedel consider electrode surface. To study semiconduc-
the rate constants for change transfer to be tor dissolution at the microscopic (atomic)
exponentially dependent on the Helmholtz level, one needs techniques such as scan-
potential. ning tunneling microscopy (STM) and
In many early studies of anodic oxida- atomic force microscopy (AFM). The an-
tion of p-type semiconductors it was tacitly odic and chemical dissolution of silicon
assumed that the reaction was controlled has been studied in very elegant work by
solely by the surface hole concentration, Allongue and coworkers [100–102].
that is, by the potential drop across the
space charge layer of the solid. The role 2.1.4
of the Helmholtz potential was neglected. Sensitizer-based Photoelectrochemical
For p-type electrodes, anodic oxidation oc- Systems
curs at potentials close to the flat band
value and under accumulation conditions. Dye molecules adsorbed on an electrode
In addition, there will be a high den- form, in principle, the simplest class
sity of surface states as a result of the of photoelectrochemical system. The ba-
breaking of surface bonds. In this case, sic scheme presented at the end of
a change in the applied potential is likely Sect. 2.1.1.2 is directly applicable for un-
to be distributed partly or even completely derstanding the mechanism of photocur-
over the Helmholtz layer. The potential rent generation. The dye molecules act
dependence of the anodic current then is as the light-absorbing species. If injec-
due to (in part) the changes in the rate tion of a photogenerated charge carrier
constants for hole capture resulting from into the electrolyte or electrode com-
changes in the Helmholtz potential. Elec- petes effectively with relaxation of the
trical impedance spectroscopy is necessary excited state, a photocurrent might be
to decide whether etching is under space observed in the external circuit. Singlet-
charge layer or Helmholtz-layer control. to-triplet crossing often leads to a relatively
The latter was shown to be the case for long-lived excited state, which allows
the dissolution of p-GaAs in acidic elec- electron-hole separation to occur by an
trolytes [96]. Vanmaekelbergh and Searson electron transfer process. The photochem-
in a study of the dissolution of p-type istry and photoelectrochemistry of dye-
silicon in HF solution showed that EIS sensitized electrodes has been studied
2.1 Photoelectrochemical Systems Characterization 83
extensively, and the research is still contin- and bottom of the CB and to an increase
uing. Dye-sensitization of photochemical of the band gap energy with respect to that
reactions is a key topic in photographic of a macroscopic crystal. The energy-level
research [103, 104]. At present, there is spectrum shifts from that of a classical
a considerable effort being devoted to insulator to that of a molecule with a re-
dye-sensitized porous photoelectrochem- duction of the dimensions in the 20–1-nm
ical solar cells [16, 105]. The fundamental range. Second, a considerable fraction of
research is focused on the chemical bond- the atoms of a nanocrystal lies at the sur-
ing between the light-absorbing molecules face. This leads to surface-electron states.
and the electrode surface, the mechanism States of energy in the band gap can have
of light absorption, and the dynamics of a strong influence on the optical proper-
charge injection and recombination in the ties of nanocrystals. It is clear that the
dyes. In this research, scanning probe surface chemistry is extremely important.
methods [106, 107] and time-resolved op- Organic and inorganic molecules can pas-
tical spectroscopy [108, 109] are used to- sivate surface states, thus removing them
gether with photoelectrochemical charac- from the optical gap. Functionalized cap-
terization [24]. The reader is referred to a ping molecules play an important role in
more specialized review for details [24]. providing stability against coagulation and
Electrodes to which insulating light-
in allowing the attachment of nanocrys-
absorbing nanocrystals, instead of dyes,
tals to solid (electrode) surfaces (see the
are attached form a relatively new class of
following text).
system. With colloidal solution chemistry,
Early reports have shown photoelec-
a large variety of insulating nanocrystals
trochemical activity when, for example,
can be prepared [110–113]. Well-known
CdS and PbS nanocrystals are attached
examples are II–VI compounds (CdS,
to a metal electrode in a sub-monolayer
CdSe, and ZnO), III–V compounds (InP,
array [19, 20, 116–120]. Clearly, photoex-
InAs, and GaN), and transition-metal
oxides. In addition, there are various citation of the nanocrystals can lead to a
methods in solid-state chemistry and long-lived state, which allows one of the
electrochemistry for preparing nanocrys- charge carriers to be transferred from the
tals directly on surfaces. For instance, nanocrystal before recombination occurs
CdS, CdSe, and PbS nanocrystals can be (see Sect. 2.1.1.2).
electrochemically deposited on gold elec- A macroscopic PbS crystal has a band
trodes [114]. Electrochemical oxidation of gap of 0.41 eV. Because of the small ef-
Si leads to porous silicon that may contain fective mass of the electrons and holes
a large number of Si nanocrystals [115]. (me,eff = mh,eff = 0.09 × me ), strong size-
Nanocrystalline colloidal systems have quantization occurs. The absorption spec-
been extensively characterized with opti- trum of an aqueous suspension (of
cal spectroscopy [110–113]. There are two polyvinyl alcohol-capped) PbS nanocrys-
effects that are essential for understand- tals, 6.5 nm in diameter (see TEM picture)
ing the electronic and optical properties is shown in Fig. 13. The HOMO-LUMO
of insulating nanocrystals. First, quan- optical transition occurs at 2.1 eV, and two
tum confinement of the electron waves other absorption peaks are seen at 3.2 and
in the nanocrystal leads to discrete elec- 4.3 eV. When a gold electrode is immersed
tron states at the top of the valence band in this colloidal solution, PbS nanocrystals
84 2 Experimental Techniques
0.2 10 0
[nm]
0
1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
Energy
[eV]
are attached and form a monolayer array, state contains an electron located at 0.4 eV
see Fig. 14 [121, 122]. versus SCE, thus considerably below the
The photoelectrochemical activity of LUMO of Q-PbS. The much faster kinetics
a Au/Q-PbS electrode, illuminated with of the photoexcited hole transfer indicates
chopped green light, is shown in Fig. 14 that the hole in the long-lived excited state
for three different aqueous electrolyte is still delocalized, that is, it occupies the
solutions. With a 1 M KCl solution (case a), HOMO. It was thus concluded that the
photocurrent transients are observed when excited state in Q-PbS leading to pho-
the electrode potential is more positive toinduced electron transfer consists of a
then −0.4 V (versus SCE). Illumination trapped electron and a delocalized hole in
gives a cathodic current that decreases to the HOMO (denoted as Q(etrap , hHOMO )).
zero with increasing time; turning off the The following scheme describes the decay
light induces an anodic current. Clearly, il- dynamics of the long-lived state in Q-PbS:
lumination leads first to the transfer of
an electron from gold to the photoex- hν relaxation
Q −−−→ Q∗ −−−−−−→
cited hole in the PbS nanocrystals (a
‘‘cathodic’’ transfer), followed by trans- relaxation
fer of the photoexcited electron to empty Q(etrap , hHOMO ) −−−−−→ Q + heat
states in the gold (an ‘‘anodic’’ transfer). (27)
Energy relaxation in the gold due to elec- transfer from Au
tron transfer competes with electron-hole Q(etrap , hHOMO ) −−−−−−−−−→ Q− (etrap )
recombination in Q-PbS. For this to oc- (28)
cur, the electrochemical potential in the transfer to Au
gold electrode must be located between Q− (etrap ) −−−−−−−−→ Q (29)
the energy levels corresponding to the
photoexcited electron and hole. Recombi- Interestingly, when tartrate is added as an
nation transients are suppressed when the electron donor to the solution, an anodic
Fermi level is above 0.4 eV (versus SCE), photocurrent is observed in the poten-
which indicates that the long-lived excited tial range positive with respect to −0.4 V
2.1 Photoelectrochemical Systems Characterization 85
1.0
0.0
−1.0 Light on
−2.0
−3.0 100
Light off
(a)
Light off 50
1.0
[µA /cm2]
0
J
0.0
Light on
−1.0
Light off
(c)
−0.4 −0.3 −0.2 −0.1 0.0
V/V (SCE)
Fig. 14 Current-potential curves, under chopped light (514 nm), measured with a Q-PbS
covered gold electrode in a photoelectrochemical cell with three aqueous electrolytes. The
insert gives an STM picture of a part of the Q-PbS covered electrode showing a disordered
monolayer coverage. (a) 1 M KCl; (b) 1 M KCl + 0.2 M tartrate (electron donor);
(c) 1 M KCl + 0.01 M K3 Fe(CN)6 (electron acceptor) (Ref. 122).
(Fig. 14, case b). This means that electron above 0.4 eV. This is logical, because
donation from tartrate can compete effec- the cathodic photocurrent corresponds to
tively with electron donation from the gold electron transfer to the oxidized species,
electrode [i.e. Eq. (28)]: not to the electrode.
Photoelectrodes, consisting of CdS
transfer from tartrate
Q(etrap , hHOMO )+Red −−−−−−−−−−−−→ nanocrystals that are directly attached
to an electrode (gold or conducting ox-
Q− (etrap ) + Ox+ (30) ide) or indirectly via a dithiol molecular
linker have been studied extensively [19,
In contrast, a cathodic photocurrent is 20, 116–120]. Here, we will focus on
observed when Fe(CN)6 3− is added to the work that demonstrates the power of
solution (Fig. 14, case c). Thus, electron small-amplitude methods in photoelectro-
donation to the oxidizing species in chemical characterization [26, 123, 124].
solution competes effectively with electron By measuring the photoelectrochemical
donation to gold; it can be clearly seen activity of gold/Q-CdS electrodes, as the
in Fig. 14, case c that the electrochemical potential of the gold electrode was varied,
activity in the latter solution does not stop it could be concluded that photoinduced
when the electrode Fermi level is increased electron transfer between the gold surface
86 2 Experimental Techniques
and the quantum dots is due to a long- photoinduced electron transfer from the
lived state in Q-CdS. This state, denoted as LUMO in Q-CdS to empty states in the
[Q(eLUMO , htrap )], consists of an electron gold electrode:
in the LUMO and a hole trapped in a level
ωb(
about 0.6 eV below the LUMO:
Q(eLUMO , htrap ) −−−→ Q+ (htrap ) (32)
hν relaxation
Q −−−→ Q∗ −−−−−−−→ ωc
Q+ (htrap ) + Red −−−→ Q + Ox+ (33)
relaxation
Q(eLUMO , htrap ) −−−−−→ Q + heat With a 1 M KCl solution, the low-fre-
(31) quency limit ∂jPH /e∂% of the optoelectri-
cal transfer function is zero; this agrees
Time-resolved photobleaching experi- with the fact that there is no steady state
ments on Q-CdS colloidal solutions photocurrent. The high-frequency limit of
showed the existence of a long-lived the optoelectrical transfer function is also
state (lifetime 50 ms) that, very probably, zero. Starting from 105 s−1 , an anodic vec-
corresponds to Q(eLUMO , htrap ) [125]. tor develops with decreasing modulation
Figure 15 shows the optoelectrical trans- frequency. This corresponds to electron
fer function (see Sect. 2.1.2.4), measured transfer from the Q-CdS LUMO to gold.
with near UV-light from an argon laser The rate ωb of this process cannot be
for a 1 M KCl solution (a, b) and a KOH- clearly distinguished from the discharge
tartrate solution (c, d). In a and c, the frequency of the photoelectrochemical cell.
transfer function is plotted in the complex Electron transfer from the LUMO to the
plane with the frequency as a parameter, gold is probably faster than 3 × 104 s−1 .
in b and d, the modulus of the transfer At lower frequencies, the modulus of the
function is plotted versus the modulation optoelectrical transfer function shrinks to
frequency ω. In the KOH-tartrate solu- zero. This is because of the electron back-
tion (c, d), an anodic vector is observed transfer from gold to the empty level in
at sufficiently low frequency. This means Q-CdS following the faster LUMO-to-gold
that ∂jPH /e∂% > 0 (see Sect. 2.1.2.2). This electron transfer:
agrees with the observation of a steady ωa
state anodic photocurrent, which increases Q+ (htrap ) −−−→ Q (34)
with increasing light intensity. The anodic
vector shrinks with increasing modula- The rate of this transfer (ωa ) is 6 × 103 s−1
tion frequency ω and, eventually, becomes and is independent of temperature. Thus,
zero for ω > 5 × 104 s−1 . Two semicircles at sufficiently low modulation frequency,
with characteristic frequencies ωc = 6 s−1 both the processes represented by Eqs. (32)
and ωb = 700 s−1 are observed (measure- and (34) are in phase with the modulated
ments at 25 ◦ C). Measurements in the light intensity, and the resulting photocur-
temperature range 0–60 ◦ C showed that rent is zero. When no redox system is
ωb is temperature-independent, while ωc present, relaxation of the long-lived excited
increases with increasing temperature. state in Q-CdS, Q(eLUMO , htrap ), occurs via
Hence, ωc corresponds to the rate of elec- consecutive steps involving electron trans-
tron donation from tartrate to the hole fer from the LUMO to the gold, and from
trapped in Q-CdS, while ωb is the rate of the gold to the trapped hole. This relaxation
2.1 Photoelectrochemical Systems Characterization 87
103 × 2 π rad/s
1.0 ωa
Imaginary [ j(ω)/e Φ(ω)]
0.5
0 1.5
[10−3]
10 × 2 π rad/s
−0.5 1.0
0.5
−1.0 104 × 2 π
rad/s ωb
10−1 × 2π rad/s 4
Modulus [ j(ω)/e Φ(ω)]
0
3
[10−2]
[10−2]
ωc 2
−1
10 × 2 π
rad/s 1
103 × 2 π rad/s 102 × 2 π rad/s
ω b′
−2
0 1 2 3 4 100 101 102 103 104
(c) Real [ j(ω) /e Φ(ω)] (d) Frequency
[10−2] [s−1]
Fig. 15 The optoelectrical transfer function discussed in the text. In (b), the modulus of the
measured for a Q-CdS covered gold electrode in transfer function is plotted vs. the modulation
an electrochemical cell. The attached quantum frequency. (c,d) Results obtained in an aqueous
dots have an average diameter of 5 nm. (a,b). solution (pH = 14) with 0.5 M tartrate as an
Results obtained in an aqueous solution electron donor. (c) Plot of the transfer function
(pH = 7) with 1 M KCl (297 K). In (a), a plot of in the complex plane with the modulation
the transfer function is shown in the complex frequency as a parameter. (d) Plot of the
plane with the modulation frequency as a modulus of the transfer function versus the
parameter. The meaning of the transfer function modulation frequency (from Ref. 123).
and of the characteristic frequencies (ωa , ωb ) is
linked covalently at the other end of the heterogeneous catalysis, storage batter-
molecule via the S-termination. The rates ies, fuel cells, and super-capacitors [41].
of photoinduced electron transfer from Porous insulating or semiconducting
the LUMO of Q-CdSe to gold (kL,Au ), phases have been considered as photo-
and from gold to the trapped hole in chemical devices for the light-stimulated
Q-CdSe (kAu,T ) were obtained from the oxidation of organic waste components
optoelectrical transfer function. An expo- [heterogeneous photocatalysis] [110–113,
nential decrease of the electron transfer 129]. Porous photoelectrochemical sys-
rate with increasing length of the spacer tems have been studied extensively only in
molecule is found, with a decay parameter the last ten years [16, 105, 130–137]. Their
β = 0.5 Å−1 . This low value (the value in fabrication is more demanding than that
vacuum is typically four times larger) in- of photochemical systems, because electri-
dicates a strong through-bond electronic cal work is delivered in an external circuit
coupling in the cyclohexylidene spacer upon illumination (see Sect. 2.1.2.2). In a
molecules. This agrees with the result of porous photoelectrochemical system, the
quantum chemical calculations and with solid phase is insulating or semiconduct-
ing, whereas the permeated phase is an
the strong S–S coupling shown by photo-
electrolyte solution. The current in the
electron spectroscopy [126, 127].
solid phase is due to the motion of elec-
The measurement of changes in the op-
trons or holes; the current in the solution
tical properties of nanocrystals attached
is due to the motion of an oxidized or
to an electrode caused by variations in the
reduced species.
electrode potential forms a challenging but
A porous photoelectrochemical system
interesting characterization method. Re-
can be prepared by the deposition of
cently, a bleaching of the HOMO-LUMO colloidal particles on a conducting sub-
transition of CdS nanocrystals has been ob- strate. It is essential that the particles are
served when the electrode Fermi level was electrically connected and that there is elec-
in resonance with the LUMO [128]. This trical contact between the particles and
led the authors to conclude that the pho- the conducting substrate. Furthermore,
toexcited state in CdS contains an electron the pore system should form a single-
in the LUMO (and a trapped hole). This permeated phase. A well-known example
is in agreement with the results obtained of such a system is the particulate TiO2
by photoelectrochemical characterization photoelectrode, which forms the basis of
of gold/Q-CdS electrodes (see earlier). the photoelectrochemical solar cells pro-
posed by O’Regan and Grätzel and other
2.1.5 groups [16, 130–137]. A SEM picture of
Porous Photoelectrochemical Systems a TiO2 network, consisting of intercon-
nected spheres with a diameter of 30 nm,
2.1.5.1 Introduction is shown in Fig. 16(a); electrical contact
Here, we define a porous solid as a phase between the particles and between the con-
that contains empty spaces that are in- ducting substrate and the particles was
terconnected. Thus, there is a single-solid achieved by slight sintering at 450 ◦ C.
phase that can be permeated with a second An alternative route exists for the prepa-
phase. Porous metals have been exten- ration of porous semiconductors. Many
sively studied, and are widely applied in n-type single crystals, such as GaP, GaAs,
2.1 Photoelectrochemical Systems Characterization 89
Acc.V. Magn.
10.0 kV 108639 200 nm
(a)
(b) 00018 3 µm
Fig. 16 (a) A particulate network consisting of 25 nm TiO2 particles prepared by
deposition from a colloidal solution followed by slight sintering; (b) A porous GaP
crystalline network prepared by anodic etching of an n-type GaP crystal at positive
potential. The structural units have dimensions in the 100–300 nm range.
Si, SiC, and TiO2 can be transformed into would explain why anodic dissolution is
a porous network by anodic etching un- so strongly nonuniform over the surface
der conditions of severe band-bending [18, and eventually leads to the formation of
115, 138–144]. Under such conditions, a porous semiconductor. However, many
surface electrons located at the top of the questions remain regarding porous pho-
valence band or in band gap states can toelectrochemical etching. A well-studied
tunnel through the gap into the CB; the example is macroporous GaP [139, 143].
surface localized holes generated in such When n-type GaP is subjected to a poten-
a way are consumed in anodic dissolu- tial of 5 V or more (versus SCE) a highly
tion of the material. The rate of inter- porous network is formed (Fig. 16b) with
band tunneling (Zener breakdown [145]) dimensions of the structural units and
is strongly dependent on the electric field the pores in the 100-nm range. The GaP
and the presence of surface defects. This network retains its crystallinity [146, 147].
90 2 Experimental Techniques
A ‘‘single-crystal GaP sponge’’ is different much smaller but also larger than the di-
from a particulate network in that there mensions of the solid structures and the
are no grain boundaries and the connec- pores. Therefore, porous semiconductors
tion between the network and the bulk can show remarkable charge-storage prop-
GaP matrix is ideal. erties. Other relevant length scales are the
wavelength of visible light (400–700 nm)
2.1.5.2 Special Properties of Porous and the diffusion length of charge carri-
Photoelectrochemical Systems ers before recombination. These length
Porous photoelectrochemical systems con- scales also can be in the same range
sist of an insulating or semiconducting as the dimensions in the porous struc-
solid network permeated with a conduct- ture; this leads to striking optical and
ing electrolyte solution; the dimensions electrodynamic properties. Extensive re-
of the solid structures and pores are in search is being performed in these fields,
the 1–500-nm range. A typical semicon- and a comprehensive review is beyond
ductor/electrolyte interface has a width the scope of this section. Instead, the
of between 0.5 nm (the Helmholtz layer electrostatic, optical, and electrodynamic
in a concentrated electrolyte solution) properties of porous semiconductor (elec-
and 100 nm (typical depletion layer in a trodes) are briefly discussed and the
semiconductor). Thus, the width of the reader is referred to more detailed pub-
solid/electrolyte interfacial layer can be lications.
EF,n EF,n
Depletion
n-type
Semiconductor Depletion
Solution Solution
200 nm 200 nm
(a) (b)
EF,n
Depletion
Solution
20 nm
(c)
Fig. 17 Schemes of the semiconductor/electrolyte interface for a macroporous and a
nanoporous electrode. (a) An n-type macroporous electrode under moderate depletion:
structural units contain a depleted region and a bulk region (free electrons in the
nondepleted region). (b) A macroporous n-type electrode at a strongly positive potential:
the entire porous electrode is depleted of free electrons. (c) A nanoporous electrode in
which depletion occurs without band bending.
2.1 Photoelectrochemical Systems Characterization 91
density depends on the nature of the coefficient of the light can be estimated
electrolyte solution; this clearly shows from effective medium theory.
the importance of the charging energy.
Size-quantization and single-dot charging 2.1.5.2.5 Electron-hole Photoexcitation by
energy play an important role in electron Sub-band Gap Light In a dye-sensitized
transport in metal–nanodot–metal double porous photoelectrode, an electron from
barrier tunnel junctions [157–161]. Study the dye is photoexcited into the CB by a
of electron transport in two-dimensional photon of energy considerably below the
or three-dimensional assemblies consist- band gap of the semiconductor. The dye
ing of nanometer-sized particles is still in molecules are anchored on the internal
its infancy; one may expect that single-dot surface of the porous semiconductor. Light
charging (leading to Coulomb-blockade) absorption is very effective because of
and size-quantization will result in inter- multiple interactions of a single photon
esting and novel transport phenomena. with the dye molecules. Similarly, a porous
semiconductor without dye molecules may
2.1.5.2.4 Light Scattering in Macroporous absorb sub-band gap light, and this may
Semiconductors The dimensions of the lead to photogeneration of free electrons
structural units and pores in macro- and holes. The mechanisms of free carrier
porous semiconducting and insulating generation with sub-band gap light in
networks are often in the 100-nm range. macroporous GaP photoelectrodes have
This is the same range as the wave- been investigated in detail [166, 167].
length of visible and UV light. Because Surface-localized electrons involved in
of the structural variation of the refrac- two-photon transitions and in a coupled
tive index on the wavelength scale, visible optical-thermal transition were found to
light can be strongly scattered in macro- give rise to significant sub-band gap
porous networks [162–165]. In macrop- photocurrent in this system.
orous GaP, for instance, the propagation
of red (sub–band gap) light is strongly 2.1.5.2.6 Effective Electron-hole Separation
attenuated [146, 147]. The importance for The ability of porous photoelectrochemi-
photoelectrochemical systems lies in the cal systems to separate effectively electrons
fact that the effective absorption length of and holes is widely known since the pre-
supra–band gap light in macroporous sys- sentation of the dye-sensitized particulate
tems is reduced considerably with respect TiO2 solar cell [16, 105, 130–137]. In this
to that in single crystals. For example, the system, the photocurrent quantum yield
penetration depth 1/α of green light in a (the number of electrons counted in the
bulk GaP single crystal is about 10 µm, external circuit as photocurrent divided by
whereas macroporous GaP networks with the number of absorbed photons) is close
a thickness of only 2 µm completely absorb to unity. This means that electron-hole pair
green light [138, 139]. Visible and near- recombination is essentially absent. Effi-
UV light is not scattered in nanoporous cient separation of photogenerated elec-
systems because the structural variation trons and holes was demonstrated with
in the refractive index occurs on a scale several other photoelectrochemical sys-
much smaller than the wavelength of light. tems [105, 130–137]. Photovoltaic devices
In such a case, the effective absorption based on permeated hole-conducting and
94 2 Experimental Techniques
1.0
(b)
0.8 Fig. 18 Comparison of the
photocurrent quantum yield
0.6 [Q = jPH /e(1 − R)%] versus the
(a) wavelength of the incident light,
Q
photoluminescence in the visible spectral with respect to the flat band potential
range [171]. Of the various explanations (Fig. 19) [172, 173]. Because of the absence
given for this phenomenon, the most of an electric field in the porous layer and
widely accepted is that of size quanti- the strong confinement of the carriers, the
zation. As a result of the confinement electron and hole are not separated in the
of charge carriers within nanometer-sized potential range corresponding to depletion
structures in the porous matrix, the ef- in a bulk electrode. As at a single-crystal
fective band gap is widened while the electrode, hole injection from a strongly
oxidizing species such as SO4 − (gener-
•
ratio of radiative to nonradiative recom-
bination is considerably enhanced [171]. ated electrochemically by the reduction of
In situ luminescence measurements pro- S2 O8 2− at the electrode) gives rise to vis-
vide information about the physical and ible electroluminescence in porous n-type
chemical properties of porous silicon and silicon [172, 173]. The emission increases
about charge-transfer reactions at the sili- in the range in which the photolumines-
con/solution interface [25]. cence decreases [172, 173] (see Fig. 19). An
In contrast to the photoluminescence interesting aspect of the electrolumines-
from a single-crystal electrode (see Fig. 11), cence is the voltage tunability of the colour.
the emission from a porous n-type silicon The emission maximum shifts to shorter
electrode is constant at positive potentials wavelength as the potential is scanned to
and decreases only in the range negative negative values, until finally the emission
1.0
EL
PL
0.8
Intensity [a.u.]
0.6
0.4 700 nm
750 nm
800 nm
0.2
0.0
EC e e e
Fig. 20 Energy scheme for
photogeneration and diffusion of
electrons in a porous photoelectrode
Energy
solid network and diffuse, because of a time that electrons need to travel through
gradient of the electrochemical potential, the system before collection, that is, the
through the network over a considerable transit time τtran (d) is in the millisec-
distance toward the metal contact, where ond to second range [179–183, 187, 188].
they are collected (Fig. 20). The thickness, As a result, photogenerated electrons can
d, of a porous electrode and, hence, the be lost before collection, by transfer to
length of the electron pathway is between the oxidized species in the solution, a
1 and 100 µm. process characterized by a time con-
The transport of electrons in porous stant τrec . Electron back-transfer forms
semiconductors is of interest to a wide an important recombination process in
audience, not only to electrochemists. On dye-sensitized photoelectrochemical sys-
one hand, porous semiconducting and in- tems [16, 179–183].
sulating networks can be considered as
disordered systems, which show a strong 2.1.5.3.2 Characterization of Electron Dif-
resemblance to amorphous semiconduc- fusion and Back-transfer by Light Intensity
tors [177, 178]. In porous semiconductors, Modulated Techniques In Sect. 2.1.2, it
diffusing electrons can be scattered not was shown that time-resolved methods
only by the lattice but also by the surface of are required to obtain information on
the matrix and by grain boundaries. Scat- the kinetics and dynamics of a system.
tering is a friction phenomenon reducing The measurement of the photocurrent re-
the free-electron mobility µ. In addition, sponse upon a small modulation of the
porous networks have a large interfacial light intensity is a very effective method to
area. Therefore, a huge volume density of study electron dynamics in a porous sys-
interfacial electronic states, distributed in tem [184–186, 187, 188]. Neglecting trap-
the band gap, can be expected. The volume ping/detrapping of diffusing electrons and
density of band gap states can be much assuming the electrochemical potential
larger than that of a macroscopic crystal gradient in the porous system to be in-
and comparable to that of an amorphous dependent of the spatial coordinate x (see
semiconductor [177, 178]. A diffusing elec- Fig. 20), Vanmaekelbergh and coworkers
tron can be trapped in a state in the gap and demonstrated that the optoelectrical trans-
hence, become temporarily localized. The fer function may be written as a function
electron is promoted back into the CB by of τtran en τrec [189, 190]:
thermal excitation (trapping/detrapping).
It is clear that electron scattering and #j˜PH (ω)
multiple trapping/detrapping are different e(1 − R)#%̃(ω)
physical processes; scattering reduces the 1 − e−iωτtran (d) e−τtran (d)/τrec
electron flux in the system by reducing = (37)
iωτtran (d) + [τtran (d)/τrec ]
the mobility of the diffusing free electrons,
whereas trapping decreases the electron Plots of the optoelectrical transfer func-
flux by reducing the density of free elec- tion in the complex plane are presented in
trons. Fig. 21. Attention is drawn to two limiting
Many recent experimental results show cases. If the transit time of photogener-
that electron transport through a porous ated electrons τtran (d) through the porous
semiconducting network is a slow pro- network is much smaller than the electron
cess [179–183]. For instance, the average lifetime τrec , photogenerated electrons will
98 2 Experimental Techniques
0.0
τrec / τd = 0.1
1/ τ rec
−0.2
Im [ j(ω)/e Φ(ω )]
−0.4 τrec / τd = 1
−0.6 1/ τ d
τrec / τd = 100
reach the collecting contact without being recombination is dominant (τtran (d) +
lost; the collection efficiency is nearly one, τrec ), the optoelectrical transfer function
and thus also the low-frequency limit of the reduces to
transfer function. In this case (Fig. 21), the #j˜(ω) τrec /τtran (d)
optoelectrical transfer function describes = (39)
e(1 − R)#%̃(ω) iωτrec + 1
electron propagation. The shape of the
function is determined by electron diffu- This transfer function corresponds to a
sion only and enters, surprisingly, the third semicircle in the complex plane (Fig. 21).
quadrant of the complex plane at high fre- The recombination lifetime follows from
quencies. The average transit time through the characteristic frequency of the transfer
the porous network follows from the angu- function: 1/ωm = τrec ; the transit time can
lar frequency ωm at which the imaginary then be extracted from the low-frequency
part of the optoelectrical transfer function limit given by
shows its first minimum (characteristic ∂jphoto τrec
frequency, starting from zero frequency): = (40)
(1 − R)e ∂φ τtran (d)
3
τtran (d) = (38) Including trapping/detrapping in the
ωm model that leads to the transfer func-
It is easily inferred from Eq. (38) that the tion expressed by Eq. (37) does not alter
factor e−τtran (d)/τrec damps the oscillating the optoelectrical transfer function itself.
function e−iωτtran (d) . As a result, the However, the transit time that is ob-
optoelectrical transfer function becomes tained from ωm is increased by a factor
increasingly semicircular if τtran (d)/τrec [1 + 0(EF,n )]. The value of the trapping
increases toward and above unity. When parameter 0(EF,n ) (with respect to 1)
2.1 Photoelectrochemical Systems Characterization 99
0.0
−0.1
−0.2
−0.3
ωm
−0.4
−0.5
−0.6
−0.7
−0.2 0.0 0.2 0.4 0.6 0.8 1.0
0.0
−0.1
−0.2
−0.3
ωm
−0.4
−0.5
−0.6
−0.7
10−2 10−1 100 101 102 103
Modulation frequency
(b) [Hz]
Electrolyte solution is 0.5M H2SO4 Fig. 22 (a) Complex-plane
200 Porosity is 25% representation of the measured
Trapping level density: 2kBTΣS(EF)
29. W. Gärtner, Phys. Rev. 1959, 116, 84. 54. B. H. Erné, D. Vanmaekelbergh, I. E. Ver-
30. L. M. Peter, Chem. Rev. 1990, 90, 753. meir, Electrochim. Acta 1993, 38, 2559.
31. B. Smandek, G. Chmiel, H. Gerischer, Ber. 55. D. Vanmaekelbergh, F. Cardon, J. Phys. D:
Bunsen-Ges. Phys. Chem. 1989, 93, 1094. Appl. Phys. 1986, 19, 643.
32. G. Chmiel, H. Gerischer, J. Phys. Chem. 56. D. Vanmaekelbergh, F. Cardon, Semicond.
1990, 94, 1612. Sci. Technol. 1988, 3, 124.
33. K. H. Beckmann, R. Memming, J. Elec- 57. D. Vanmaekelbergh, F. Cardon, Electrochim.
trochem. Soc. 1969, 116, 368. Acta 1992, 37, 837.
34. H. Gerischer, Z. Phys. Chem. 1960, 26, 325. 58. Z. Hens, W. P. Gomes, J. Phys. Chem. B
35. H. Gerischer, Z. Phys. Chem. 1961, 27, 48. 1997, 101, 5814.
36. H. Gerischer, Semiconductor Electrochem- 59. Z. Hens, W. P. Gomes, J. Electroanal. Chem.
istry, in Physical Chemistry, an Advanced 1997, 437, 77.
Treatise, Vol. IXA/Electrochemistry, Aca- 60. Z. Hens, J. Phys. Chem. B 1999, 103, 122.
demic Press, New York, 1970. 61. Z. Hens, W. P. Gomes, J. Phys. Chem. B
37. W. P. Gomes, F. Cardon, in Progress in Sur- 1999, 103, 130.
face Science (Eds.: S. G. Davison), Pergamon 62. D. Vanmaekelbergh, A. R. de Wit, F. Car-
Press, 1982, p. 155, Vol. 12. don, J. Appl. Phys. 1993, 73, 5049.
38. J. O’ M. Bockris, A. K. N. Reddy in Modern 63. A. A. Burk, P. B. Johnson, W. S. Hobson
Electrochemistry, Plenum Press, New York, et al., J. Appl. Phys. 1986, 59, 1621.
1977, Vol. 2. 64. A. B. Ellis, Luminescent Properties of Semi-
39. Y. V. Pleskov, Y. Y. Gurevich in Semicon- conductor Electrodes, in Chemistry and
ductor Photoelectrochemistry, Consultants Structure of Interfaces, New Laser and Op-
Bureau, New York, 1986. tical Techniques, Chapter 6 (Eds.: R. B. Hall,
40. R. Degryse, W. P. Gomes, F. Cardon et al., A. B. Ellis), VCH, Deerfield Beach (FL),
J. Electrochem. Soc. 1975, 122, 711. 1986.
41. T. Pajkossy, J. Electroanal. Chem. 1994, 364, 65. B. A. Balko, G. L. Richmond, J. Phys. Chem.
111. 1993, 97, 9002.
42. J. F. McCann, S. P. S. Baldwall, J. Elec- 66. O. Kruger, Ch. Jung, Ber. Bunsen-Ges. Phys.
trochem. Soc. 1982, 129, 551. Chem. 1994, 98, 1022.
43. Z. Hens, W. P. Gomes, Phys. Chem. Chem. 67. J. F. Kauffman, C. S. Liu, M. W. Karl,
Phys. 1999, 1, 3607. J. Phys. Chem. 1998, 102, 6766.
44. Z. Hens, W. P. Gomes, Phys. Chem. Chem. 68. G. N. Ryba, C. N. Kenyon, N. S. Lewis,
Phys. 1999, 1, 3617. J. Phys. Chem. 1993, 97, 13814.
45. K. Schröder, R. Memming, Ber. Bunsen-Ges. 69. A. Manivannan, K. Hashimoto, A. Fuji-
Phys. Chem. 1985, 89, 385. shima, J. Phys. Chem. 1992, 96, 3766.
46. D. Vanmaekelbergh, W. P. Gomes, Ber. 70. A. Manivannan, K. Itoh, K. Hashimoto
Bunsen-Ges. Phys. Chem. 1985, 89, 994. et al., J. Electrochem. Soc. 1990, 137, 3121.
47. J. E. A. M. van den Meerakker, J. J. Kelly, 71. J. Ouyang, F.-R. Fan, A. J. Bard, J. Elec-
P. H. L. Notten, J. Electrochem. Soc. 1985, trochem. Soc. 1989, 136, 1033.
132, 638. 72. F. Decker, F. Prince, P. Motisuke, J. Appl.
48. J. E. A. M. van den Meerakker, Electrochim. Phys. 1985, 57, 2900.
Acta 1985, 30, 435. 73. B. H. Erné, F. Ozanam, J.-N. Chazalviel,
49. D. Vanmaekelbergh, W. P. Gomes, F. Phys. Rev. Lett. 1998, 80, 4337.
Cardon, Ber. Bunsen-Ges. Phys. Chem. 1986, 74. S. R. Morrison, T. Freund, J. Chem. Phys.
90, 431. 1967, 47, 1543.
50. D. Lincot, J. Vedel, J. Electroanal. Chem. 75. W. P. Gomes, T. Freund, S. R. Morrison,
1987, 220, 179. J. Electrochem. Soc. 1968, 115, 818.
51. D. Vanmaekelbergh, W. P. Gomes, F. 76. A. Fujishima, T. Kato, E. Maekawa et al.,
Cardon, J. Electrochem. Soc. 1987, 134, 891. Bull. Chem. Soc. Jpn. 1981, 54, 1671.
52. D. Vanmaekelbergh, R. P. ter Heide, W. 77. G. H. Schoenmakers, D. Vanmaekelbergh,
Kruijt, Ber. Bunsen-Ges. Phys. Chem. 1989, J. J. Kelly, J. Chem. Soc., Faraday Trans.
93, 1103. 1997, 93, 1127.
53. Z. Hens, W. P. Gomes, J. Phys. Chem. B 78. J. R. Harbor, M. L. Hair, J. Phys. Chem.
2000, 104, 7725. 1979, 83, 652.
2.1 Photoelectrochemical Systems Characterization 103
128. S. G. Hickey, J. D. Riley, E. Tull, J. Phys. 153. D. Cummins, G. Boschloo, M. Ryan et al.,
Chem. B 2000, 104, 7623. J. Phys. Chem. B 2000, 104, 11449.
129. M. R. Hoffmann, S. T. Martin, W. Choi 154. P. Hoyer, R. Eichberger, H. Weller, Ber.
et al., Chem. Rev. 1995, 95, 69. Bunsen-Ges. Phys. Chem. 1993, 97, 630.
130. G. Hodes, J. Electrochem. Soc. 1992, 139, 155. P. Hoyer, H. Weller, J. Phys. Chem. 1995,
3136. 99, 14096.
131. R. Grünwald, H. Tributsch, J. Phys. Chem. 156. E. A. Meulenkamp, J. Phys. Chem. B 1999,
B 1997, 101, 2564. 103, 7831.
132. S. Y. Huang, J. Phys. Chem. B 1997, 101, 157. D. L. Klein, P. L. McEuen, J. E. B. Katari
2576. et al., Appl. Phys. Lett. 1996, 68, 2574.
133. D. Liu, R. W. Fessenden, G. L. Hug et al., 158. D. L. Klein, R. Roth, A. K. L. Lim et al., Na-
J. Phys. Chem. B 1997, 101, 2583. ture 1997, 389, 699.
134. R. Argazzi, C. A. Bignozzi, T. A. Heimer 159. U. Banin, Y. Cao, D. Katz et al., Nature
et al., J. Phys. Chem. B 1997, 101, 2591. 1999, 400, 542.
135. H. Rensmo, K. Keis, H. Lindstroem et al., 160. O. Millo, D. Katz, Y. Cao et al., Phys. Rev. B
J. Phys. Chem. B 1997, 101, 2598. 2000, 61, 16773.
136. A. Wahl, J. Augustynski, J. Phys. Chem. B 161. E. P. A. M. Bakkers, D. Vanmaekelbergh,
1998, 102, 7820. Phys. Rev. B 2000, 62, R7743.
137. W. Kubo, K. Murakoshi, T. Kitamura et al., 162. J. D. Joannopoulos, P. R. Villeneuve, S. Fan,
Chem. Lett. 1998, 12, 1241. Nature 1997, 386, 143.
138. B. H. Erné, D. Vanmaekelbergh, J. J. Kelly, 163. U. Grüning, V. Lehmann, C. M. Engelhardt,
Adv. Mater. 1995, 7, 739. Appl. Phys. Lett. 1995, 66, 3254.
139. B. H. Erné, D. Vanmaekelbergh, J. J. Kelly, 164. J. E. G. J. Wijnhoven, W. L. Vos, Science
1998, 281, 802.
J. Electrochem. Soc. 1996, 143, 305.
165. Y. A. Vlasov, N. Yao, D. J. Norris, Adv.
140. G. Oskam, A. Natarajan, P. C. Searson,
Mater. 1999, 11, 165.
F. M. Ross, Appl. Surf. Sci. 1997, 119, 160.
166. D. Vanmaekelbergh, L. van Pieterson, Phys.
141. J. van de Lagemaat, M. Plakman, D.
Rev. Lett. 1998, 80, 82.
Vanmaekelbergh et al., Appl. Phys. Lett.
167. D. Vanmaekelbergh, M. A. Hamstra, L. van
1996, 69, 2801.
Pieterson, J. Phys. Chem. B 1998, 102, 7997.
142. T. Sugira, T. Yoshida, H. Minoura, Elec-
168. J. J. M. Halls, C. A. Walsh, N. C. Greenham
trochem. Solid-State Lett. 1998, 1, 175. et al., Nature 1995, 376, 498.
143. A. O. Konstantinov, C. I. Harris, E. Janzén, 169. G. Yu, J. Gao, J. C. Hummelen et al., Sci-
Appl. Phys. Lett. 1994, 65, 2699. ence 1995, 270, 1789.
144. A. O. Konstantinov, A. Henry, C. I. Harris 170. U. Bach, D. Lupo, P. Comte et al., Nature
et al., Appl. Phys. Lett. 1995, 66, 2250. 1998, 395, 583.
145. C. Zener, Proc. R. Soc. London 1934, 145, 171. A. G. Cullis, L. T. Canham, P. D. J. Calcott,
523. J. Appl. Phys. 1997, 82, 909.
146. F. J. P. Schuurmans, D. Vanmaekelbergh, 172. E. A. Meulenkamp, L. M. Peter, D. J. Riley
J. Van de Lagemaat et al., Science 1999, 284, et al., J. Electroanal. Chem. 1995, 392, 97.
141. 173. L. M. Peter, D. J. Riley, R. I. Wielgosz et al.,
147. F. J. P. Schuurmans, M. Megens, D. Thin Solid Films 1996, 276, 123.
Vanmaekelbergh et al., Phys. Rev. Lett. 1999, 174. A. Bsiesy, F. Gaspard, R. Herino et al.,
83, 2183. J. Electrochem. Soc. 1991, 138, 3450.
148. L. M. Peter, D. J. Riley, R. I. Wielgosz, Appl. 175. M.Ligeon, F. Muller, R. Herino et al.,
Phys. Lett. 1995, 66, 2355. J. Appl. Phys. 1993, 74, 1265.
149. L. M. Peter, R. I. Wielgosz, Appl. Phys. Lett. 176. E. S. Kooij, A. R. Rama, J. J. Kelly, Surf. Sci.
1996, 69, 806. 1997, 370, 125.
150. W. H. Lubberhuizen, D. Vanmaekelbergh, 177. R. A. Street, Cambridge Solid State Science
E. Van Faassen, Journal of Porous Materials Series, in Hydrogenated Amorphous Silicon,
2000, 7, 147. (Eds.: R. W. Cahn, E. A. Davis, I. M. Ward),
151. B. Enright, B. G. Redmond, D. Fitzmaurice, 1991.
J. Phys. Chem. 1994, 98, 6195. 178. H. Scher, M. F. Shlesinger, J. T. Bendler,
152. M. Gratzel, Nature 2001, 409, 575. Phys. Today 1991, 44, 26.
2.1 Photoelectrochemical Systems Characterization 105
Among these four types of dislocations, diethylzinc (DEZ), AsH3 and H2 Se, re-
we do not need to take into account spectively. The Si substrate orientation was
item (1) because the dislocation density (001) 2–4◦ tilted toward [110] direction. Si
of Si is generally very low. In order to substrate was rinsed in organic solvents,
reduce the dislocation density according followed by the repetition of the oxidation
to items (2), (3), and (4), many efforts by H2 SO4 : H2 O = 4 : 1 and the removal
have been made. The efforts to change of the oxides by 25% HF solution. After
the initial growth mode from three- loading the substrate into the reactor, the
dimensional (3D) to two-dimensional (2D) substrate was heated at 1000 ◦ C for 10 min
have been made to avoid the dislocations at hydrogen atmosphere. The V/III ratio
generated by the coalescence of the islands was varied from 100 to 6400 with chang-
at the initial stage of the growth [10]. ing the PH3 flow rate, keeping the TMG
Although many methods using strained flow rate constant for the growth of GaP
layer superlattice (SLS) buffer layer [3, on Si substrate. The gas pressure was
11, 12], rapid thermal annealing [13], varied from 76 to 380 torr. The epitaxial
thermal cycle annealing (TCA) [14, 15], layer thickness was varied from 20 nm to
and so on have been adopted to reduce 3.7 µm. The growth temperature was kept
the dislocation density of item (3), the constant at 900 ◦ C. 3D growth was not ob-
dislocation density is still on the order of served under these growth conditions for
106 cm−2 . The low temperature growth the case of GaP growth on GaP substrate.
has been investigated to decrease the The samples were examined using No-
generation of the dislocation by the marski optical microscopy, cross-sectional
thermal stress [16, 17]. It is expected that transmission electron microscopy (TEM),
the number of dislocations generated by and so on. GaAs was grown on Si sub-
the thermal stress is reduced when the strate by the two-step growth method with
growth temperature is low. Although the 10-nm-thick GaAs buffer layer grown at
dislocation density of GaAs-on-Si on the 400 ◦ C.
order of 104 cm−2 has been obtained at
the growth temperature of 350 ◦ C, thermal 2.2.3
stability is the problem. Deposition of Gallium Phosphide on Silicon
Substrate
2.2.2
MOCVD 2.2.3.1 Nucleation of GaP on Si Substrate
In general, the growth mode is divided
The epitaxial growth was performed us- into three categories, namely, 2D type,
ing low-pressure or atmospheric pressure Volmer-Weber (3D) type, and Stranski-
MOCVD. The former consists of lamp- Krastanov (2D + 3D) type. In the case of
heated and the latter consists of rf-heated the 2D mode, the dislocations are gener-
horizontal reactor with load lock chamber. ated when the layer thickness exceeds the
The substrate was put on the SiC-coated critical thickness [18]. On the other hand,
carbon susceptor and the temperature was the dislocations and stacking faults are
controlled by the thermocouple inserted generated at the coalescence of the islands
into the susceptor. Source gases for Al, formed at the initial stage of the growth
Ga, Zn, As and Se are trimethylallu- in the case of the 3D mode [19]. There-
minum (TMA), trimethylgallium (TMG), fore, the defect density will be significantly
108 2 Experimental Techniques
Ga
P
GaP Si
Si
(a)
GaP
Si
(b)
Fig. 2Schematic atomic arrangement of 3D GaP on Si (a) and 2D
GaP on Si (b).
necessary for lower growth pressure to velocity v is proportional to the total flow
produce a layer-type growth as shown in rate (F ) and inverse of the growth pressure
Fig. 2(b). Pg , that is,
In the growth of GaP, a simple reaction F
v=b (4)
involving PH3 is considered to be the Pg
decomposition of PH3 and the production
and Eq. (3) is changed into
of phosphorus molecules, that is, # $
Pg
PH3 = P + 3H (1) pPH3 = pPH30 exp −c (5)
F
Using simple reaction kinetics and where c is a modified constant including
solving the differential equations, that is, a, b, and x. On the other hand, using these
d[PH3 ] notations, V/III ratio is given by
= a[PH3 ] (2)
dt pPH30
the partial pressures of PH3 in these cases [V/III] = (6)
pTMGa
are given by
! x" The ratio of the phosphorus concen-
pPH3 = pPH30 exp −a (3) tration([P]) to the gallium concentration
v
([Ga]), which is available to the growth on
where pPH3 is the partial pressure of PH3
the substrate surface is expressed as,
at the distance x from the entrance of
the heated zone in the flow direction with [P/Ga] = [P]/[Ga] (7)
the flow velocity v, pPH30 is the initial
partial pressure of PH3 or the supplied where [P] and [Ga] are proportional to
partial pressure into the reactor, and a the concentration of the decomposed PH3
is a constant including the reaction rate and TMGa, respectively. [Ga] is supposed
constant, the cross-sectional area of the to be constant on the growing surface
reactor and other parameters. The flow for various growth pressures because the
2.2 Deposition of (Multiple Junction) Semiconductor Surfaces 111
TMGa is almost decomposed at the growth In the TEM picture with the thickness
temperature. Thus, substituting the partial of 3.7 µm, two kinds of dislocations with
pressures and the relation for the V/III an extra-half plane in the Si substrate
ratio into Eq. (6), the above ratio is given by (A-type) and the GaP layer (B-type) are
observed [28]. The schematic models of
(pPH30 − pPH3 )
[P/Ga] = a = a[V/III] atomic arrangements for these structures
pTMGa are shown in Fig. 3. In general, the
% # $&
Pg dislocation generation should take place
× 1 − exp −c (8)
F at random in the isotropic crystal [29], that
is, four kinds of A-type dislocations with
where a is the proportionality constant. Burgers’ vector of 1/2[1 0 1], 1/2[0−11],
Therefore, the initial growth mode of GaP, 1/2[0 1 1] and 1/2[−101] can be generated.
which is governed by [P]/[Ga] ratio is However, one salient feature came to be
expressed using the V/III ratio, Pg and seen; instead of these four directional
F . From these results, it can be said that A-type dislocation, only two directional A-
at lower growth pressure, a higher V/III type dislocations with Burgers vector of
ratio is required for the layer growth. 1/2[1 0 1] or 1/2[0−11] are observed. A
possible interpretation for this generation
2.2.3.2 Generation of Dislocation of A-type dislocation with only two-kinds
In the heteroepitaxial growth, the gener- of Burgers’ vectors instead of the four
ation of misfit dislocation and the stress previously reported types of dislocations
relaxation are related to each other. The is the difference of the situation of the site,
misfit dislocation generation and the stress that is, dislocations are created at the step
relaxation of GaP layer on Si substrate
grown under high V/III ratio (layer-type
growth mode) are described. Because the
defects associated with the coalescence of GaP
the islands are not generated, the observed
dislocations are generated after the layer
thickness exceeds the critical thickness.
From the cross-sectional TEM micro- Si
graph, it can be seen that the GaP surface Si
is very flat, and defects such as disloca- P
A-type
tions or structural defects are not observed Ga
at all when the thickness is thinner than
90 nm [27]. This means that GaP grows on
Si coherently with compressive stress at
the initial stage. The TEM measurement GaP
for GaP on Si with various layer thickness
shows that the dislocations at the interface
are observed when the GaP layer thickness
exceeds 90 nm. Si
edge of the misoriented Si substrate rather lattice strain of GaP is relaxed completely
than randomly isotropic generation. The at the growth temperature by introducing
direction of Burgers’ vector for the B-type A-type misfit dislocations, the tensile stress
dislocations is the reverse of that of the is produced in the GaP layer during the
A-type dislocation. cooling process because of the difference of
Assuming that the strain is completely the thermal expansion coefficients of GaP
accommodated at the growth temperature and Si. The thermal expansion coefficient
by 60◦ dislocations, the spacing between of GaP is about 2.5 times larger than that
dislocations corresponds to 110 nm. How- of Si. In order to relax the tensile stress
ever, the spacing between A-type disloca- in the GaP layer, the dislocations with the
tions is much smaller than the calculated extra-half plane in the GaP layer should
value. This difference might be due to the be introduced. Accordingly, it is proved
uniformity of the Si substrate steps. that the dislocations with the extra-half
If the dislocations are generated by plane in the Si substrate are formed by
the lattice mismatch, an extra-half plane the lattice mismatch and that those in the
should occur in the Si substrate, which GaP layer are formed during the cooling
has a small lattice constant [29]. Because process to relax the thermal stress. This
both the room temperature lattice constant experiment supports the report that proved
and the thermal expansion coefficient of the generation of threading dislocations
GaP are larger than those of Si, GaP in GaP on Si during the cooling down
should have the larger lattice constant process [30].
than Si at the growth temperature. Hence, The stress applied to the GaP layer as
the lattice mismatch relaxation at the a function of the thickness measured by
growth temperature is responsible for the X-ray diffraction is shown in Fig. 4 [20].
generation of the A-type dislocation. The The dotted line shows the stress value for
B-type dislocations cannot be explained the thermal stress calculated by the bimetal
merely on the grounds of lattice mismatch. model. A thin GaP layer has a compres-
Because no dislocation is generated at sive stress, and the stress changes to tensile
the initial stage of the growth, the B-type with increasing thickness. Considering the
dislocations should be generated during thermal stress between the growth temper-
the growth or cooling down process. If the ature and room temperature, the GaP layer
Compressive
1
[× 109 dyn cm−2]
0
Stress
1
2
Tensile
0 1 2 3 4
Fig. 4 Stress of GaP layer on Si
Thickness substrate as a function of the
[µm] thickness.
2.2 Deposition of (Multiple Junction) Semiconductor Surfaces 113
GaAs
3D, and type-II dislocations are dominant Figure 10 shows the dark spot defect
when the initial growth mode is 2D. (DSD) density as a function of the number
of TCA1 for various number of TCA2. The
2.2.4.2 Effect of TCA DSD density of GaAs on Si without SLS
In0.1 Ga0.9 As/GaAs SLS was inserted in the is also shown. The DSD density decreases
GaAs layer to reduce the dislocation den- with increasing the number of TCA1 and
sity. The individual layer thickness and the TCA2. The lowest DSD density obtained in
total layer numbers of SLS are 20 nm and this study is 3.8 × 106 cm−2 . It is indicated
10, respectively. The TCA was performed that the TCA both before and after the
just before the SLS growth (TCA1) and SLS growth is more effective in reducing
after the SLS growth (TCA2). The upper the DSD density. Even if the dislocation
and lower temperatures of TCA are 900 density is reduced by TCA1, dislocations
and 300 ◦ C, respectively. Numbers of TCA are generated in the SLS by the lattice
were changed in this study. The intentional mismatch of SLS and GaAs. Although
doping was not performed in GaAs on Si. the individual layer thickness of SLS is
The sample structure is shown in Fig. 9. thinner than the critical thickness, the total
TCA 2
InGaAs (20 nm)/GaAs (20 nm) SLS
TCA 1
GaAs (1.0 µm)
Si
108
Without SLS
DSD Density
107
SLS thickness is thicker than the critical is due to the enhancement of dislocation
thickness [12]. It is suggested from the movement, large compressive stress, and
experimental results that the dislocations the generation of point defects at higher
generated at SLS are bended by TCA2, temperature, which in turn reduce the dis-
resulting in the low dislocation density. location density effectively.
Until now, the low etch pit density on the Figure 12 shows the minority carrier
order of 106 cm−2 has been obtained using lifetime of GaAs, Al0.15 Ga0.85 As and
SLS and TCA for the total epitaxial layer Al0.22 Ga0.78 As grown on Si for various
thickness of more than 3.5 µm [40–43]. TCA temperatures. The minority carrier
Few papers have been reported on the lifetime is also improved with increasing
growth of GaAs on Si, with the dislocation the TCA temperature, which is supported
density of 106 cm−2 at the epitaxial layer by the decrease of DSD at high temper-
thickness of less than 3 µm. ature. The minority carrier lifetime of
The crystal quality of GaAs with 900 ◦ C GaAs grown on Si with 1000 ◦ C TCA is
TCA and SLS is inferior to that on a GaAs 3.36 ns. Those for GaAs and AlGaAs for
substrate because a high density of dislo- various Al compositions, are shown in
cations is generated in the epitaxial layer, Fig. 13. The lifetime of GaAs grown on
which degrades the minority carrier life- GaAs substrate is also plotted for compar-
time. The TCA temperature was optimized ison [44–48]. Although it is impossible to
to improve the minority carrier lifetime. compare the lifetime because the carrier
In this experiment, SLS buffer later was concentration is not the same for all the
not used. samples, it is estimated that the lifetime of
Figure 11 shows the DSD density of GaAs and AlGaAs on Si is approximately
GaAs grown on Si substrate revealed by one order shorter than those grown on
electron beam–induced current (EBIC) GaAs substrate.
measurement for various TCA temper-
ature. The DSD density decreases with 2.2.4.3 Effect of Hydrogenation
increasing the TCA temperature gradu- Hydrogenation was carried out in a
ally and is on the order of 106 cm−2 at quartz tube, where a hydrogen plasma
1000 ◦ C. The crystal quality improvement was excited by rf power via a copper
by the relatively high TCA temperature coil encircling the quartz tube. The
108
Dark spot density
[cm−2]
107
106
900 950 1000
Fig. 11 DSD density of GaAs
TCA Temperature on Si as a function of TCA
[°C] temperature.
2.2 Deposition of (Multiple Junction) Semiconductor Surfaces 119
x = 0.15
1
x = 0.22
0
850 900 950 1000
TCA Temperature
[°C]
100
Ahrenkiel et al. (1988)
'tHooft et al. (1981)
Timmons et al. (1990) on GaAs
Zarem et al. (1989)
Minority carrier lifetime τ
10
[ns]
plasma power, the treatment time, and the at various temperatures ranging from
substrate temperature during the plasma 350 ◦ C to 450 ◦ C for 10 minutes. The TCA
treatment were 90 W, 2 hours, and 250 ◦ C, temperature was 900 ◦ C and the SLS buffer
respectively. In order to recover shallow layer was not used.
level passivation and the damage induced Carrier concentration profiles of unin-
by the plasma treatment, post annealing tentionally doped GaAs grown on Si sub-
was performed in an AsH3 + H2 ambient strates, before and after hydrogenation, are
120 2 Experimental Techniques
1018
As-grown
Carrier concentration
1017
[cm−3 ]
Hydrogenated
1016
1015
0 1 2 3 4
Depth
[µm]
Fig. 14 Carrier concentration profile of GaAs on Si with and without
hydrogenation.
shown in Fig. 14. Undoped GaAs-on-Si is Figure 15 shows the 4.2 K PL spectra
n-type (1 × 1017 cm−3 ) because of Si auto- of GaAs on Si for as-grown sample and
doping during the growth [49]. For the hydrogenated sample, hydrogenated sam-
hydrogenated sample, the carrier concen- ple annealed at 450 ◦ C. The major peaks
tration is reduced to about 3 × 1016 cm−3 are peak B corresponding to the heavy
at depth exceeding 1 µm. This is due to the hole-associated free exciton and peak C
electrical passivation of the shallow levels. corresponding to the carbon impurity-
Because the major donor in GaAs grown bound exciton. After the sample is treated
on Si substrates is Si via auto-doping from by the hydrogen plasma, the full width
the substrate, passivation will occur by the at a half maximum (FWHM) of peak B
formation of SiH0 complexes via the reac- narrows from 4.49 meV to 3.83 meV. This
tion narrowing is due to the passivation of local-
ized states. With 450 ◦ C annealing where
Si+ + H0 + e− −−−→ SiH0
the shallow level is completely recovered,
The SiH0 complex will then be dissociated the FWHM is a little narrower than that of
by heat or applied electric fields. There is a the as-grown sample.
kink in the concentration profile curve at a Figure 16 shows the minority carrier life-
depth of nearly 0.8 µm, which corresponds time derived from the time resolved pho-
to the plasma-induced damage [50], as the toluminescence decay curve. The minority
knee goes deeper with increased plasma carrier lifetime increases from 1.66 ns
treatment time. After a 10-minute anneal- (as-grown) to 4.66 ns after the plasma
ing at 450 ◦ C in AsH3 + H2 ambient, the treatment, and gradually decreases with
donor electrical activities were completely increasing annealing temperature. It is dif-
restored to their initial levels. ficult to judge the crystal quality only by the
2.2 Deposition of (Multiple Junction) Semiconductor Surfaces 121
4.2 K
A B C D E
4.19 meV
Hydrogenated +
PL Intensity
3.83 meV
Hydrogenated
4.49 meV
As-grown
longer minority carrier lifetime of the hy- passivated by hydrogenation. The passi-
drogenated sample (before annealing) is vation effect remains even after 450 ◦ C
due to defect passivation and shallow level annealing, where the shallow level is com-
passivation. Species in the plasma include pletely restored.
free radicals, ions, and electrons. Among From these experiments it can be
these species, free radicals can effectively concluded that (1) the shallow level that
passivate the defects. Furthermore, it is has been passivated by hydrogenation
well known that ions can damage the semi- is completely recovered by annealing at
conductor surface. Therefore, during the 450 ◦ C in AsH3 + H2 ambient, (2) the
hydrogenation process, defect passivation deep levels are still passivated by hydrogen
and damage formation take place at the after annealing at 450 ◦ C in AsH3 + H2
same time. The minority carrier lifetime ambient, and (3) hydrogenation followed
of the hydrogenated sample annealed at by 450 ◦ C annealing produces a longer
450 ◦ C is longer than that of the as-grown minority carrier lifetime at the same
sample because the defects generated dur- shallow carrier concentration.
ing the plasma treatment are passivated.
The DLTS spectra show that the peak of 2.2.4.4 Application to Photovoltaic Device
the Si-defect related level becomes smaller Figure 17 shows the schematic cross-
after hydrogen plasma treatment. It sug- sectional view of a GaAs/Si tandem solar
gests that the Si-related defect level is cell. It consists of n+ -GaAs buffer layer,
Au/AuZn
ZnS/MgF2
p + -GaAs
p +-Al0.8Ga0.2As 0.05 µm
p +-Alx Ga1−x As 0.2 µm (x : 0 → 0.29)
n -GaAs 0.6 µm
n +-Si 1.1 µm
p +-Si 0.5 µm
Au
Fig. 17Schematic cross-sectional view of three-terminal GaAs/Si
tandem solar cell.
2.2 Deposition of (Multiple Junction) Semiconductor Surfaces 123
1.1
Open-circuit voltage
1.0 on GaAs
[V]
on Si
0.9
0.8
900 950 1000
Fig. 18 Open-circuit voltage of
GaAs solar cell on Si as a TCA Temperature
function of TCA temperature. [°C]
124 2 Experimental Techniques
Au-Zn/Au
p+-GaAs MgF2 /ZnS
p+-Al0.8Ga0.2As 1 × 1018 cm−3 0.05 µm
p+-Alx Ga1−x As 1× 1018 cm−3 0.3 µm
(x : 0.15 → 0.30)
n+-GaAs 20 nm
p+-Si 1.0 µm
n+-Si 0.8 µm
Au-Sb/Au
30 Fig. 20 Current-voltage
characteristics of two-terminal
25 Al0.15 Ga0.85 As/Si tandem solar
cell.
Current density
20
[mA cm−2 ]
15
Jsc (mA cm−2) Voc (V) FF (%) η (%)
10
23.6 1.57 77.2 21.2
5
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
Voltage
[V]
characteristics are shown in Figs. 19 and the two-terminal configuration has been
20, respectively. It is observed that the obtained [52].
short-circuit current of the top cell is per- It has been pointed out that a 2-terminal
fectly matched to that of the bottom cell. tandem solar cell with an efficiency higher
A conversion efficiency of 21.2% utilizing than 30% can be obtained by using a
2.2 Deposition of (Multiple Junction) Semiconductor Surfaces 125
top cell material with a band gap energy pressure. When the V/III ratio or the gas
of 1.7–1.8 eV over an Si bottom cell. pressure is increased, the growth mode
This structure resulted in photocurrent changes from island-type to layer-type for
matching between the top cell and the thin GaP layer thickness. The two-type of
bottom cell. However, in our study, the misfit dislocations, which are generated
current matching was obtained by using by the lattice mismatch and the thermal
the Al0.15 Ga0.85 As top cell, of which the expansion mismatch, are observed. A high
band gap energy is 1.61 eV. This is because density of APDs that propagate to the
the short-circuit current of the top cell is surface and are annihilated during growth
inferior to the ideal one. The main reason has been observed. The nucleation of GaAs
for the degradation is the short minority on Si and the effects of SLS buffer layer,
carrier lifetime caused by a high density of TCA, and hydrogenation are described.
dislocation in the AlGaAs layer on Si. If it The crystal quality has been improved by
becomes possible to grow an AlGaAs layer using SLS buffer layer, increasing the TCA
on an Si substrate with a long minority temperature and using hydrogen plasma
carrier lifetime, comparable to that grown treatment. The GaAs layer grown on Si
on GaAs substrate, a higher efficiency substrate has been applied to photovoltaic
tandem solar cells can be obtained by devices.
increasing the Al composition so that the
photocurrent matching between the top References
cell and the bottom cell is retained.
The improvement of the Si bottom cell is 1. J. P. Benner, J. M. Olson, T. J. Coutts, in
also important for the increase of the total Advances in Solar Energy, (Ed.: K. W. Boer),
American Solar Energy Society, Inc., 1992,
conversion efficiency. The main problem is
p. 125, Vol. 7, Chapter 4.
that the conversion efficiency is degraded 2. M. A. Green, K. Emery, K. Bucher et al.,
after the crystal growth process [53]. The Progr. Photovolt. 1999, 11, 31.
junction depth becomes deeper and the As 3. T. Soga, S. Hattori, S. Sakai et al., Electron.
atoms diffuse into the Si substrate during Lett. 1984, 20, 916–918.
4. M. Akiyama, Y. Kawarada, K. Kaminishi,
the growth. A low temperature growth
Jpn. J. Appl. Phys. 1984, 23, L843–L845.
process is necessary for the improvement 5. W. I. Wang, Appl. Phys. Lett. 1984, 44,
of the bottom cell. Therefore, in the future, 1149–1150.
the technology to grow an AlGaAs layer 6. B.-Y. Tsaur, G. M. Metze, Appl. Phys. Lett.
with long minority carrier lifetime at 1984, 45, 535–537.
7. W. T. Masselink, T. Henderson, J. Klem
low temperature should be investigated to
et al., Appl. Phys. Lett. 1984, 45, 1309–1311.
increase the efficiency of the tandem cell. 8. T. Egawa, H. Tada, Y. Kobayashi et al., Appl.
Phys. Lett. 1990, 57, 1179–1181.
2.2.5 9. T. Soga, T. Kato, M. Yang et al., J. Appl. Phys.
Summary 1995, 78, 4196–4199.
10. T. Soga, T. George, T. Suzuki et al., Appl.
The nucleation, dislocation generation, Phys. Lett. 1991, 58, 2108–2110.
stress relaxation, and the annihilation of 11. R. Fischer, D. Neuman, H. Zabel et al., Appl.
APD in the GaP/Si heteroepitaxial growth Phys. Lett. 1986, 48, 1223–1225.
12. Y. Watanabe, Y. Kadota, H. Okamoto et al.,
have been reviewed. The growth mode J. Cryst. Growth 1998, 93, 459–463.
appears to be island-type for the low 13. N. Chand, R. People, F. A. Baicocchi et al.,
values of V/III ratio and the low gas Appl. Phys. Lett. 1986, 49, 815–817.
126 2 Experimental Techniques
Vacuum level
Electron affinity
Φ
Work χ
function CB
Empty Band bending
surface Vs
states
Occupied EF
surface
states Eg
Surface
resonances
VB
Energy
Distance
Density Space charge
of states region
(SCR)
Distance
Distance
Distance
Fig. 1 (a) One-electron energy level diagram of an n-type doped semiconductor. VB
represents the valence band, comprising largely filled, closely spaced energy levels. CB
is the conduction band, which consists of largely empty, closely spaced energy levels.
EF is the semiconductor Fermi level and Eg is the band gap energy. The surface states
are occupied until sufficient space charge is created corresponding to an electrostatic
field, which prevents further electrons from going to the surface. It is this field that
causes the shift in the band energies, i.e. Vs . (b) Side view of the crystal. The surface
localizes electrons at the surface states, and thus is negatively charged. Figure 1 also
shows the presence of surface energy levels that lie outside the band gap. These states
are referred to as surface resonances and unlike surface states they are degenerate
with the bulk states and can mix with them [2, 3]. A surface resonance has a varying
degree of localization in the surface region, although it should be noted that there is
no absolute definition for how strong a surface localization should be in order for it to
be defined as a surface resonance.
2.3 Grafting Molecular Properties onto Semiconductor Surfaces 129
cross-sections of electrons and holes are determines the surface electronic prop-
higher on the surface than in the bulk, sur- erties. Therefore, to control the surface
face states also serve as fast recombination electrical properties and to make the semi-
centers of electrons and holes. This phe- conductor useful for electronic devices,
nomenon can critically affect the electron- we look for treatments that will interact
transfer efficiency in electronic devices and with the surface and modify it chemically.
thus their performance. The recombina- The requirements of molecular surface
tion rate of electrons and holes at the treatments, which are also named passi-
surface is expressed in units of velocity vation treatments, are the following:
(cm s−1 ) and is called surface recombina-
tion velocity (SRV or s). It reflects the rate 1. Remove surface and interface states or
at which minority carriers (holes in n-type at least eliminate them in the energy
materials) are consumed at the surface. In interval of the band gap. Because the
many cases, SRV is determined by the sur- presence of surface states is conducive
face states close to the middle of the band for charge-trapping and recombination
gap where the chances for electron-hole effects, moving them out of the band
recombination are the highest [5]. gap is necessary for the construction
Other surface electrical properties in- of high-speed semiconductor devices
dicated in Fig. 1 are the work function, like the ones that are based on GaAs
!, and the electron affinity, χ. In semi- and InP.
conductors, electron affinity is defined as 2. Tuning the surface electron affinity.
the difference in energy between the lo- Control over the electron affinity is
cal vacuum level [6] and the bottom of the essential for band edge engineering
conduction band (CB) at the surface. The of interfaces such as those in photo-
(local) work function is the minimum en- voltaic solar cells and for controlling
ergy required for an electron to escape the barrier height for electron loss to
into vacuum (just outside the range of the the surroundings. The second effect is
crystal potential) from the Fermi level. Al- demonstrated for the GaAs/(Al,Ga)As
though this definition is straightforward system where a band offset of 0.4 eV [7]
for metals, for semiconductors it should did not block electron loss from the
be borne in mind that there are mostly no GaAs into the passivating (Al,Ga)As
real energy levels at the Fermi level. The layer.
work function is determined by the elec- 3. Strong bonding that can withstand
tron affinity, the band-bending, and the device-processing. Ideally, the molecu-
energy difference between the Fermi level les will chemisorb on the surface.
and the conduction/valence band in the It should be noted that the binding
bulk. groups themselves, while interacting
with the surface, could modify the
2.3.2 surface chemistry and thus the surface
Requirements of Molecular Surface energetics on binding.
Treatments 4. Chemical protection from adsorbates,
especially atmospheric adsorbates. In
Because the surface charge and dipole are those cases where surface oxidation is
determined by the surface chemistry, it is detrimental, there is the daunting task
the surface chemistry that to a large extent of protection from O2 .
130 2 Experimental Techniques
CB Surface CB
treatment EF
EF
Occupied Vs′
surface Vs
states VB
VB
SCR SCR
Fig. 2 Left to right: Reduction of the density of surface states of n-type semiconductor,
as a result of surface treatment, which leads to decrease in width of the space charge
region (SCR) and a decrease in the surface band-bending from Vs to Vs# .
2.3 Grafting Molecular Properties onto Semiconductor Surfaces 131
surface charge changed. Change in the from a change in the chemical environ-
distribution of energies of the surface ment caused by adsorbed extrinsic atoms.
states can also affect the SRV. Because
the SRV depends on the density of surface 2.3.3.2 Controlling the Electron Affinity, χ
states with energies around the midgap, By definition, χ depends on the energy
shifting of the surface-state energies away difference between the vacuum level and
from the midgap can lead to reduction in the bottom of the CB at the surface.
the probability of charge recombination. Therefore, any treatment that influences
The strategy to modify surface states, the surface potential will modify χ.
and thus the Vs and SRV, is based on in- Modification of the surface potential can
teraction of chemically grafted molecules be achieved by utilizing polar molecules
with these states. The key is to find that will bind to the surface and change
molecules that will modify the semicon- its potential. Figure 3 shows schematically
ductor surface chemistry in a way that in- the manner in which a polar molecule can
volves the surface states. In this respect, the modify the surface χ. The surface poten-
origin of the surface states should be con- tial will be reduced if the molecular dipole
sidered. Intrinsic surface states originate is pointing toward the surface and will
from the termination of the crystal bulk increase if the dipole is directed in the
and the breaking of chemical bonds at the opposite direction. This approach of us-
surface, whereas extrinsic surface states ing dipolar molecules was applied to tune
originate from crystal imperfections, such the χ of metals [8–11] and semiconduc-
as missing surface atoms, line defects, or tors [8, 12–14]. It should be mentioned
(a) A
+ q
A d
m
Decrease of c
Binding −
d
group
(b) B
q
B +d
Increase of c
−d
that because short-range atomic forces or band-bending, we can tune the work
determine the energy positions of the function
surface states inside the band gap, they
need not be affected by the presence of $! = $χ ± $Vs (2)
the macropotential. Therefore, looking at
with + for n-type and − for p-type.
Fig. 1, a polar molecule modifies the en-
ergy of the bands with respect to the
2.3.4
vacuum level, but not necessarily those
Strategy for Molecule Selection
of the surface states with respect to the
band edges or the surface-state (and band)
The chemical strategy to control the sur-
energy level densities.
face energetics using organic molecules is
The change in the electron affinity
illustrated schematically in Fig. 4. The idea
caused by the molecule’s dipole moment
can be described in terms of a parallel is to incorporate several molecular proper-
plate capacitor, using the well-known ties simultaneously in one molecule and to
Helmholtz equation allow systematic modification of one spe-
cific property, independent of others. This
cos θ approach provides
$V = N × µ × (1)
εεo
1. a simple tool to investigate the re-
where $V is the potential drop caused
lation of the macroscopic proper-
by the dipole layer, µ is the dipole
ties of the semiconductors and the
moment, N the dipole density per unit
molecular properties of the adsorbed
area, θ the angle between the dipole
molecules;
and the surface normal, ε is the relative
2. a simple tool to enable the development
dielectric permittivity of the film, and εo
of models for surface engineering.
the permittivity of free space [15]. If values
for N , cos θ, and ε are known, $V (and A different approach, which is more suit-
thus $χ) can be calculated. The equation able for nonmolecular, extended bonded,
also demonstrates that controlling χ can electrically conducting inorganic materi-
be achieved by tuning the dipole direction als, is to use different molecular layers,
(toward/from the surface), its magnitude each of which has a different function.
and its orientation with respect to the For example, the first layer can electri-
surface normal. It has to be noted that cally passivate the surface and a second
the concept of a dielectric constant for a layer can give long-term chemical protec-
monolayer is problematic. It can be related tion [7].
to the molecular polarizability using the
Clausius-Mosotti relation (cf. discussion 2.3.5
in [14]). Organic and Inorganic Molecular Surface
Treatments
2.3.3.3 Controlling the Work Function, "
As shown in Fig. 1, the work function is In general, we can classify the chemi-
the energy difference between the Fermi cal surface treatments into two classes,
and vacuum levels and depends on the organic and inorganic. Table 1 summa-
electron affinity and band-bending. Thus, rizes the advantages and disadvantages of
by controlling either the electron affinity each group. Although, as noted in the
2.3 Grafting Molecular Properties onto Semiconductor Surfaces 133
Synthesis
A A A
Adsorption
A A A
Macroscopic
properties
Φ, χ, Vs
Tab. 2 Examples of substrates for which inorganic chemical surface modifications have been
developed
Silicon Oxygen 2, 18
Hydrogen 18, 28
Bromine 29
As 30
GaAs Sulfide and Selenide 20, 24, 27, 31, 32, 33–37
Ruthenium 38
Phosphine (PH3 ) 39
P2 O5 /NH4 OH 40, 41
Chlorine 42
Cesium 43
Sb 44, 24
H2 S 26
Iodine 45
InP CdS 46
Ruthenium 47, 48
CdTe (polycrystalline) Hydrogen plasma 49
Ruthenium 50, 51
Reactive metal interlayer 52
Oxygen anneal 53
(Hg, Cd)Te Sulfide 54
Hydrogen 55, 56
CuInSe2 Oxygen anneal 53, 57
(polycrystalline)
2.3 Grafting Molecular Properties onto Semiconductor Surfaces 135
GaAs(100) and (110) [2] surfaces in the support this hypothesis and reveal a
ambient, high surface state densities pin strong reduction in the band-bending of
the Fermi level and generate a high SRV. GaAs on treatment with Se-containing
Covering GaAs(100) with (Al,Ga)As was reagents. The stable phase was found to
found to be effective in terms of reducing be Ga2 Se3 [32], which has a close lattice
surface electron-hole recombination, but match to GaAs, and therefore creates an
could not block loss of electrons to almost strain-free layer.
the passivating (Al,Ga)As layer. Another Another class of surface treatments is
approach to modify the surface chemistry based on halogens. Halogens are mostly
of GaAs is the use of inorganic sulfides. used for surface etching and were found
These treatments were found to remove to dissociate upon adsorption [19]. The ef-
part of the surface oxide and form ficacy of halogen treatment for modifying
sulfides of Ga and As [19]. Studies of the surface electronic properties was found
metal-insulator semiconductor (MIS) and to depend on the morphology and compo-
Schottky diode structures of GaAs, treated sition of the surface. In the case of GaAs,
with (NH4 )2 S and Na2 S [20, 21], revealed halogens showed high reactivity toward the
that aqueous sulfide treatments induce Ga-rich surface [61, 62].
only minor changes in the net surface Parkinson and coworkers demonstrated
state density. Their main effect is reduction improvement in the open-circuit voltage
of the density of the trap states, which (Voc ) and fill factor of n-GaAs/K2 Se-K2 Se2 -
are farther in energy from the band KOH/C photoelectrochemical solar cells
edges than the shallow, doping levels. on treatment with Ru (III) [38]. The im-
This means that these treatments modify provement in the cell performance was
the positions of the surface-state energy ascribed to the shift of the surface-state
levels [19, 22]. This observation agrees with energies, which was thought to lead
results from other studies where sulfide to a reduction in the main power-loss
treatment was claimed to repin the surface mechanism, electron-tunneling through
Fermi level at a different energy [22–25]. the surface states. Treatment with Ru
A different approach was introduced by was also found to improve the perfor-
Shen and coworkers [26] who used plasma mance, specially the stability, of InP-based
H2 S treatment. In contrast to the aqueous photoelectrochemical solar cells [47]. This
sulfide treatments, this treatment leads to improvement was related to an increased
significant reduction of the surface-state barrier height [48].
density (up to three orders of magnitude)
on sulfidization. The main drawback 2.3.5.2 Organic Surface Treatments
of the inorganic sulfide treatments is Table 3 summarizes examples of organic
the instability of the sulfur-GaAs bond surface treatments used for changing the
in the ambient (vs oxygen attack, in surface electronic properties. As a first
particular) [19, 27]. example we consider self-assembly of
According to Lunt [58], the efficacy of organo-silanes on oxidized silicon. These
a selenide treatment applied to GaAs self-assembled monolayers are known to
is expected to be stronger than that of be stable and well ordered [63, 64]. Mod-
a sulfide one because of the electron- ification of the monolayer properties can
deficient nature of the GaAs binding be achieved in a predetermined fashion
site. Indeed, several studies [59, 60, 31] by changing the chemical structure of the
136 2 Experimental Techniques
self-assembling molecules. On the basis of data that were obtained after adsorption
that, a systematic modification of the sur- of several series of organic molecules (in-
face electron affinity and band-bending of cluding aniline and carbonyl compounds)
n-type silicon was demonstrated by using were interpreted in terms of changes in the
self-assembly of substituted quinoline band-bending [96–99, 102], whereas CPD
chromophores [65]. The chemical scheme changes induced by adsorption of benzoic
included two steps: acids and aniline derivatives were rational-
ized in terms of changes in electron affin-
• self-assembly of organo-silanes that add ity [13]. The different conclusions from lu-
organic functionality to the surface, and minescence intensity measurements and
• grafting of a series of substituted CPD measurements were ascribed to the
quinoline chromophores that add the fact that the PL measurements were done
variable (in this case polar) functionality. in solution, whereas the CPD measure-
ments were performed in air. This differ-
On porous silicon, which has a high ence also reflects the effects of interaction
surface-to-volume ratio and thus a large with the surrounding medium.
number of accessible surface states, a To provide information on molecule-
substantial change in PL was recorded surface interaction and on the feasibility of
on exposure to amines and different sol- chemical treatments for use in electronic
vents [66, 67]. On GaAs, organic sulfides devices, like sensors, attempts have been
and thiols were reported to induce PL made to establish a correlation between
changes that were attributed to changes molecular properties and changes in sur-
in both the SRV and the Vs [58, 68]. face properties. Table 4 summarizes works
Adsorbing a series of benzoic acids in which such a correlation was found.
with systematically varying dipole mo-
ment led to a different response. Changes 2.3.5.3.1 Correlation of Molecular Param-
in contact potential differences (CPD) eter with Changes in the Surface Electron
were correlated to changes in the elec- Affinity As mentioned earlier, control of
tron affinity without observable effect on the electron affinity or surface potential
the surface band-bending [14]. Substituted can be achieved by modifying the molec-
dicarboxylic acids, which were found to ular dipole moments of the adsorbed
bind by a two-site mechanism, rather molecules. Figure 5 shows the change
than a one-site one, such as the ben- in the electron affinity of CdTe, CdSe,
zoic acids [112], modified both the band- and GaAs, which was deduced from CPD
bending and the electron affinity of the measured by a Kelvin probe, upon graft-
GaAs [72, 74]. This difference in the elec- ing a series of substituted benzoic acids
trical effect on treatment with carboxylic onto the semiconductor surface. The lin-
and dicarboxylic acids can be explained ear relation between the change in the
by the different binding group and the electron affinity and the dipole moment
change in the molecular energy levels [75]. of the substituted benzoic acid is clearly
On CdTe, CPD measurements showed observed. The dipole moments of the sub-
that organic benzoic and dicarboxylic acids stituted phenyl groups reflect the electron
could modify the electron affinity and/or withdrawing or donating power of the
the surface band-bending [12, 75]. On the substituents. Bruening et al. used elec-
wider band gap semiconductor, CdSe, PL trochemical and CPD measurements in
138 2 Experimental Techniques
combination with ellipsometry and FTIR can be understood from the increase in
data to study the effects of binding a series effective dielectric constant of the molecu-
of cyclic disulfides with systematically vary- lar layer [130].
ing dipole moment and different degrees
of hydrophobicity to Au and CuInSe2 [8]. 2.3.5.3.2 Correlation of Molecular Parame-
They found that the magnitude and di- ters with Changes in the Band-Bending In
rection of the change in electron affinity general, the change in Vs can be viewed by
depend on the surface coverage, the ori- two mechanisms,
entation of the molecular dipole relative
• molecule-induced surface oxidation/
to the surface normal, and the mode of
reduction (generalized acid–base reac-
binding. Knowing the molecular surface
tion) where the molecules either ac-
coverage and tilt angle and thus the resid-
cept or donate electrons [117] accord-
ual molecular dipole moment, as well as
ing to the difference in the oxida-
the atomic electronegativity difference at
tion/reduction potentials, or alterna-
the molecule-surface interface allowed the-
tively by
oretical estimates of the experimentally
• frontier orbital interaction mechanism,
observed change in the electron affinity,
which involves the energy levels of the
that is, it gave predictive power to Eq. (1).
molecular frontier orbitals and of the
In further work Wu et al. tested Eqn. (1)
semiconductor surface states [118, 119].
and used a series of molecules similar to
that shown in Fig. 7, but completely con- In the second mechanism, we con-
jugated. The main effect was found to be sider the interaction between the high-
a smaller influence of the dipoles which est occupied molecular orbital (HOMO)
2.3 Grafting Molecular Properties onto Semiconductor Surfaces 139
600
R = CH3O, H, F, Br, NO2
R
400
200
COOH
[mV]
Nitro
∆χ
Bromo
Hydro Fluoro
0
Methoxy CdSe
−200 GaAs
CdTe
−400
−2 0 2 4
Dipole moment
[Debye]
Fig. 5 Linear correlation of the change in the electron affinity of n CdTe,
CdSe, and GaAs, as function of the dipole moment of benzoic acids [116],
adsorbed on the semiconductors [12–14].
level of the surface or the molecule and in Fig. 6(b), the molecule’s LUMO level
the lowest unoccupied molecular orbital interacts with occupied surface states,
(LUMO) level of the other. It is an ex- which are below the Fermi level and there-
tension of the well-known frontier orbital fore assume the role of the HOMO level.
interaction between the energy levels of On interaction, the surface-state energies
two molecules forming a complex, as increase (with respect to vacuum), that is,
shown in Fig. 6(a). On interaction, the they are ‘‘pushed’’ down in energy toward
HOMO level is stabilized to lower energy, the VB, whereas the molecular LUMO level
whereas the LUMO level is destabilized is ‘‘pushed’’ up in energy toward the CB.
and pushed up in energy. As a result, electrons that were formerly
Table 4 summarizes studies in which localized on the surface states may now oc-
the molecular ionization potential, which cupy surface resonances with only partial
is related to the HOMO level, and the localization at the surface. Therefore, they
molecular electron affinity, related to the do not anymore (or little) contribute to the
molecular LUMO level, are correlated to net surface charge. Reduction of Vs is thus
the changes in Vs . This correlation fits the expected. A different result is expected if
orbital interaction mechanism. the surface states are close to the semicon-
Figure 6 shows several scenarios for ductor midgap, as shown in the second
the interaction of a given molecular case (Fig. 6c). Because in this case the
LUMO level with surface states at dif- surface states are well removed from the
ferent energies. In the first case, shown band edge, only part of them will turn into
140 2 Experimental Techniques
(a) LUMO
HOMO
CB
(b) LUMO EF
e- VB
CB
(c) LUMO
EF
VB
(d) h+ CB
EF
LUMO VB
surface resonances and the net effect on on the density of the midgap (deep) sur-
the total surface state density, and thus face states [5].
on Vs , is expected to be moderate, com- In the third scenario, shown in Fig. 6(d),
pared to the first case (Fig. 6b). The SRV, we consider the interaction of a molecular
on the other hand, is expected to change LUMO level with empty surface states of a
significantly because it depends critically p-type semiconductor. Because the LUMO
2.3 Grafting Molecular Properties onto Semiconductor Surfaces 141
level of the molecule is well below the • a constant change of 170 mV because of
Fermi level, electron transfer from the sur- the binding group and
face to the molecule is expected and the • a second contribution that could be
molecule plays the role of the HOMO level correlated with the molecules’ LUMO
in the interaction. As in the other cases, energy level and that increased with
the surface states that were pushed in en- decreasing energy separation between
ergies to more than the CB minimum will the molecule’s LUMO state- and the
turn into surface resonances that will lead surface state-energy level [75].
to a reduction in Vs .
The energy levels of the molecule after This second contribution is denoted
interaction should also be considered. schematically in Fig. 7 and can be at-
These levels are modified (i.e. changed tributed to extended coupling of the
in energy, relative to the band edges) on molecules’ energy level and the semicon-
interaction with the surface states, as was ductor surface states, as the molecules’
demonstrated [120, 133] for the interaction LUMO energies become closer in energy
of thiophenol derivatives with Cu(111). to those of the surface states. The predic-
If, on interaction, the molecular LUMO tive power of the frontier orbital interaction
level is at an energy that corresponds to scheme, noted earlier, was demonstrated
somewhere within the band gap and below for n-CdSe, n-GaAs, n-InP, and p-GaAs
EF , the state will act as an occupied surface crystals when changes in the Vs and in the
state and will increase (decrease) the net SRV of the crystals on interaction with a
surface charge of n-(p-)type surfaces. On given dicarboxylic acid molecule (DCDC
the other hand, if the molecular LUMO and DHDC) could be explained [112]. On
level is pushed below the VB maximum, the basis of the studies mentioned earlier,
then the state acts as a surface resonance it was found that the ability of the chemical
and only a moderate effect on the band- treatment to modify the electronic proper-
bending is expected. It should be noted ties of semiconductor surfaces depends on
that although our discussion in this the following parameters:
article is limited to the molecular LUMO
level, analogous interactions of molecular 1. The molecule’s frontier orbital energy
HOMO levels with surface states can also level and the difference in energy
be energetically favorable [119] and lead to between that level and those of the
modification of surface states and/or SRV. interacting surface states. The smaller
Figure 7 shows an example of the the energy distance, the stronger is
change in Vs of an etched n-CdTe sin- the molecule-surface coupling and the
gle crystal as a function of the LUMO larger can be the induced change in the
energy level of a series of dicarboxylic surface electronic properties.
acid derivatives [75]. The systematic mod- 2. The surface-states’ energy levels and
ification of the molecular LUMO level densities. In surfaces where the dom-
was achieved by changing the molecule’s inant surface states are close to the
substituents from electron-donating to band edges, the main effect on interac-
electron-accepting groups. The change tion is a change in band-bending. On
in Vs was suggested to stem from two surfaces with dominant states close to
contributions: midgap, the main effect on interaction
142 2 Experimental Techniques
(a)
X X X
( )B
600 ∞ ∞ ∞∞
o o O O
O O O O
500 HO HO OH VI
OH
I. X = OMe; DMDC II. X = OMe; PMDC
III. X = H; DHDC IV. X = CN; PCDC V
400 V. X = CF3; DFDC
[mV]
∆V S
200
I II III
100
−3.5 −4 −4.5 −5 −5.5
LUMO energy
[eV]
(b) I
CdTe electron energy levels
−4 II −4
CdTe
LUMO energy
III
[eV]
[eV]
−5 −5
IV EF
VI
−6 −6
−7 −7
Fig. 7 (a) Change in Vs of n-CdTe upon adsorption of dicarboxylic
acid derivatives as a function of the benzoyl substituents’ LUMO
energy [75]. $Vs is the change in Vs relative to the etched surface.
(b) Energy diagram of bare CdTe and of LUMO of isolated molecule
before adsorption. The CdTe energy levels were experimentally
measured by CPD, and the Fermi level was calculated from the doping
density.
2.3 Grafting Molecular Properties onto Semiconductor Surfaces 143
Window layer
CdS
microcrystals
Functional
molecules
CuIn(Ga)Se2
Back contact
deposition of the conjugated polymer. n- and p-type surface. Gal et al. [111] used
The molecular treatment was suggested an experimental scheme similar to Fig. 8
to promote phase segregation, that is, to show the effect of organic acids on the
structural order at the interface. The electrical characteristics of polycrystalline
earlier-mentioned studies reveal the po- CuIn(Ga)Se2 (CIGSe)/CdS solar cells (see
tential of molecular treatments for tuning Fig. 9). The changes in the I –V char-
the charge transfer and charge transport acteristics could be correlated with the
across (semi)conductor/(semi)conductor molecules’ dipole moments that modify
interfaces. the band line-up at the interface, rather
A different potential application of a than as a direct effect on the surface Vs .
molecular treatment is their use as a Specially in view of the later work [77, 107,
blocking layer against charge leakage. For 128, 130], these results indicate that it is not
example, Langmuir-Blodgett (LB) films of essential to form an ordered layer in order
22-tricosanoic, CH2 =CH(CH2 )20 −COOH to modify device characteristics. However,
[67] and hexadecanol [68] were shown to what is required is a molecule that will
have charge-blocking properties on GaAs- modify the energetics at the interface so
based devices that are superior to those that any charge carrier passing from one
shown with them on (Al,Ga)As ones. side of the junction to the other will be
Using capacitance measurements of the influenced by it.
metal-(molecular insulator)-semiconduc-
tor (MIS) structure indicated reduction of 2.3.7
the interface trap density [83] and of the Summary
barrier characteristics [84].
Solar cells are another field of applica- Fine-tuning of semiconductor surface
tions for functional molecules. This idea is electrical properties can be achieved by
schematically presented in Fig. 8 for a het- grafting multifunctional organic mole-
erojunction solar cell, where the molecules cules onto the surface. In such molecules,
are located at the interface between the one function takes care of the binding
146 2 Experimental Techniques
4 a
c
2
Current
[mA]
−2
to the semiconductor surface, whereas and help design molecular tools for
another one conveys a desired property controlling semiconductor surfaces. Re-
to that surface and thus to the semicon- cent work shows that this can be carried
ductor. The way in which this can happen over to interfaces of semiconductors in
is conveniently studied by using a series of devices where molecules will directly or in-
molecules in which the desired property is directly influence active device characteris-
varied in a systematic fashion. tics. In this way, even small modifications
Implementation of molecular tools pro- in the molecular structure can induce sig-
vides new opportunities for hybrid or- nificant and predictable changes in device
ganic–inorganic systems. The increasing behavior.
interest in, and importance of, nanoscale
technologies and the use of nanocrys-
tals, where surface-related effects often Acknowledgments
dominate bulk properties [106, 125–127],
adds to the importance of controlling For our own contributions mentioned
and engineering surface properties. Cor- in this chapter we thank Leeor Kro-
relating the molecular properties of the nik, Ellen Moons, Merlin Bruening,
grafted molecules to the changes in Stephane Bastide, Dori Gal, Ayelet Vi-
the macroscopic semiconductor proper- lan, Tamar Moav and Rahel Lazar, as
ties can extend our understanding of well Ron Naaman and his group, for
molecule-surface interaction mechanisms fruitful collaborations. We acknowledge
2.3 Grafting Molecular Properties onto Semiconductor Surfaces 147
the pivotal contributions of the late 10. S. D. Evans, A. Ulman, Surface Potential
Dr. Jacqueline Libman to the early parts of Studies of Alkyl-Thiol Monolayers Ad-
sorbed on Gold, Chem. Phys. Lett. 1990,
our work and Art Ellis (Univ. of Wisconsin,
170(5, 6), 462–466.
Madison) for collaboration and inspira- 11. I. H. Campbell et al., Controlling Schottky
tion for the molecule-surface interaction Energy Barriers in Organic Electronic
mechanism. Partial support from the US- Devices using Self-assembled Monolayers,
Israel Binational- and the Israel Science Phys. Rev. B 1996, 54, R1432–R14 324.
Foundations, the Minerva Foundation 12. M. Bruening et al., Polar Ligand Adsorption
Controls Semiconductor Surface Potentials,
(Munich), the Fussfeld Fund and the Ger-
J. Am. Chem. Soc. 1994, 116, 2972–2977.
man BMBF, through its bilateral Energy 13. M. Bruening et al., Controlling The Work
Research Program with the Israel Ministry Function of CdSe by Chemisorption of Ben-
of Science is gratefully acknowledged. zoic Acid Derivatives and Chemical Etching,
J. Phys. Chem. 1995, 99, 8368–8373.
14. S. Bastide et al., Controlling the Work
References Function of GaAs by Chemisorption of
Benzoic Acid Derivatives, J. Phys. Chem.,
1. E. H. Rhoderick, R. H. Williams, Metal- 1997, 101, 2678–2684.
Semiconductor Contacts, 2nd edition, Ox- 15. D. M. Taylor, O. N. Oliveira, H. Morgan,
ford, Clarendon, 1988. Models for Interpreting Surface Potential
2. H. Lüth, Surfaces and Interfaces of Solid Measurements and their Application to
Materials, 3rd edition, Springer, Germany, Phospholipid Monolayers, J. Coll. Int. Sci.
1998. 1990, 139(2), 508–518.
3. G. V. Hansson, R. I. G. Uhrberg, Photo- 16. J. F. Dorsten, J. E. Maslar, P. W. Bohn, Near-
electron Spectroscopy of Surface States
Surface Electronic Structure in GaAs (100)
on Semiconductor Surfaces, Sur. Sci. Rep.
Modified with Self-assembled Monolayers
1988, 9, 197–292.
of Octadecylthiol, Appl. Phys. Lett. 1995, 66,
4. A. Many, Y. Goldstein, N. B. Grover, Semi-
1755–1757.
conductor Surfaces, Amsterdam, North-
17. S. M. Sze, Physics of Semiconductor Devices,
Holland, 1965.
7th edition, 2nd Chapter, John Wiley &
5. D. V. Lang, C. H. Henry, Nonradiative Re-
Sons, USA, 1981.
combination at Deep Levels in GaAs
18. W. Mönch, Semiconductor Surfaces and In-
and GaP by Lattice-Relaxation Multi-
phonon Emission, Phys. Rev. Lett. 1975, 35, terfaces, 2nd edition, in Surface Science
1525–1528. (Ed.: E Gerhard), Springer, Berlin, 1995,
6. H. Ishii et al., Energy Level Alignment pp. 274–284.
and Interfacial Electronic Structures at 19. F. Seker et al., Surface Chemistry of Proto-
Organic/Metal and Organic/Organic Inter- typical Bulk II–VI and III–V Semiconduc-
faces, Adv. Mater. 1999, 11(8), 605–625. tors and Implications for Chemical Sensing,
7. A. M. Green, W. E. Spicer, Do We Need a Chem. Rev. 2000, 100, 2505–2536.
New Methodology for GaAS Passivation? 20. J.-F. Fan, H. Oigawa, Y. Nannichi, The
J.Vac. Sci. Technol. A 1993, 11, 1061–1069. Effect of (NH4 )2 S Treatment on the
8. M. Bruening et al., Simultaneous Con- Interface Characteristics of GaAs MIS
trol of Surface Potential and Wetting of Structures, Jap. J. Appl. Phys. 1988, 27,
Solids with Chemisorbed Multifunctional L1331–L1333.
Ligands, J. Am. Chem. Soc. 1997, 119, 21. J.-F. Fan, H. Oigawa, Y. Nannichi, Metal-
5720–5728. Dependent Schottky Height with the
9. S. D. Evans et al., Self-Assembled Mono- (NH4 )2 Sx Treated GaAs. Jap. J. Appl. Phys.
layers of Alkanethiols Containing a Polar 1988, 27(11), L2125–L2127.
Aromatic Group: Effect of the Dipole Posi- 22. D. Liu et al., Deep level transient spec-
tion on Molecular Packing, Orientation, and troscopy study of GaAs surface states treated
Surface Wetting Properties, J. Am. Chem. with inorganic sulfides, Appl. Phys. Lett.
Soc. 1991, 113, 4121–4131. 1988, 53, 1059–1061.
148 2 Experimental Techniques
23. H. Hasegawa et al., Control of Fermi Level 36. C. J. Sandroff et al., Electronic Passivation
Pinning and Recombination Processes at of GaAs Surfaces Through the Formation
GaAs Surfaces by Chemical and Photo- of Arsenic – Sulfur Bonds, Appl. Phys. Lett.
chemical Treatments, J. Vac. Sci. Technol. 1989, 54(4), 362–364.
B 1988, 6, 1184–1192. 37. J. S. Herman, F. L. Terry, Hydrogen Sulfide
24. S. Hildebrandt et al., Optical Study of Band- Plasma Passivation of Gallium Arsenide,
Bending and Interface Recombination Appl. Phys. Lett. 1992, 60, 716–717.
at Sb, S, and Se Covered Gallium 38. B. A. Parkinson, A. Heller, B. Miller, Ef-
Arsenide Surfaces, Appl. Surf. Sci. 1993, 63, fects of Cations on the Performance of
153–157. the Photoanode in the N-GaAs/K2 Se-K2 Se2 -
25. R. S. Besser, C. R. Helms, Effect of Sodium KOH/C Semiconductor Liquid Junction
Sulfide Treatment on Band Bending in Solar Cell, J. Electrochem. Soc. 1979, 126(6),
GaAs, Appl. Phys. Lett. 1988, 52, 1707–1709. 954–960.
26. H. Shen, W. Zhou, J. Pamulapati, Photore- 39. P. Viktorovitch et al., Improved Electronic
flectance Study of H2 S Plasma-Passivated Properties of GaAs Surfaces Stabilized with
GaAs Surfaces, Appl. Phys. Lett. 1999, 74, Phosphorous, Appl. Phys. Lett. 1991, 58(21),
1430–1432. 2387–2389.
27. V. L. Berkovits et al., Chemical and Photo- 40. H. H. Lee, R. J. Racicot, S. H. Lee, Surface
chemical Processes in Sulfide Passivation of Passivation of GaAs, Appl. Phys. Lett. 1989,
GaAs(100), In Situ Optical Study and Pho- 54(8), 724–726.
toemission Analysis, J. Vac. Sci. Technol. A 41. L. Ferrari et al., Optical and Spectroscopic
1998, 16, 2528–2538. Characterization of GaAs Passivated Sur-
28. M. Stutzmann, C. P. Herrero, Hydrogen faces, Surf. Sci. 1995, 331–333, 447–452.
Passivation of Shallow Acceptors in Silicon, 42. D. Troost et al., Pinning of the Fermi level
Phys. Scripta. 1989, T25, 276–282. Close to the Valence-Band Top by Chlorine
29. M. J. Bedzyk, W. M. Gibson, J. A. Golov- Adsorbed on Cleaved GaAs (110) Surfaces,
chenko, X-ray Standing Wave Analysis for J. Vac. Sci. Technol. 1987, B5(4), 1119–1124.
Bromine Chemisorbed on Silicon, J. Vac. 43. H. J. Clemens, J. Von Wienskowski,
Sci. Technol. 1982, 20, 634–637. W. Mönch, On the Interaction of Cesium
30. R. I. G. Uhrberg et al., Electronic Structure, with GaAs(110) and Ge(111) Surfaces: Work
Atomic Structure, and the Passivated Na- Function Measurements and Adsorption
ture of the Arsenic-Terminated Si (111) Site Model, Surf. Sci. 1978, 78, 648–666.
Surface, Phys. Rev. B 1987, 35, 3945–3951. 44. P. Skeath et al., Bonding of Antimony on
31. B. A. Kuruvilla et al., Passivation of GaAs GaAs (110): A Prototypical System for
(100) Using Selenium Sulfide, J. Appl. Phys. Adsorption of Column-V Elements on
1993, 73, 4384–4387. III–V Compounds, Phys. Rev. B 1983, 27,
32. J. Sun et al., Chemical Bonding and 6246.
Electronic Properties of SeS2 -Treated 45. K. Jacobi, G. Steinert, W. Ranke, Iodine
GaAs(100), J. Appl. Phys. 1999, 85, Etching of the GaAs(-1-1-1) Surface Studied
969–977. by LEED, AES, and Mass Spectroscopy, Surf.
33. E. Yablonovitch et al., Band-Bending, Fermi Sci. 1976, 57, 571–579.
Level Pinning and Surface Fixed Charge on 46. K. Vaccaro et al., Indium Phosphide Passi-
Chemically Prepared GaAs Surfaces, Appl. vation Using Thin Layers of Cadmium Sul-
Phys. Lett. 1989, 54, 555–557. fide, Appl. Phys. Lett. 1995, 67(4), 527–529.
34. E. Yablonovitch et al., Nearly Ideal Elec- 47. D. N. Bose, Y. Ramprakash, S. Basu, Char-
tronic Properties of Sulfide-Coated GaAs acterization of n-InP Surfaces Before and
Surfaces, Appl. Phys. Lett. 1987, 51(6), After Surface Modification, Mater. Lett.
439–441. 1989, 88(9), 364–368.
35. C. J. Sandroff et al., Dramatic Enhance- 48. D. N. Bose, J. N. Roy, S. Basu, Improved
ment in the Gain of a GaAs/AlGaAs Schottky Barrier on n-InP by Surface Modi-
Heterostructure Bipolar Transistor by Sur- fication, Mater. Lett. 1984, 2(5B), 455–457.
face Chemical Passivation, Appl. Phys. Lett. 49. A. J. Nelson, S. P. Frigo, R. A. Rosenberg,
1987, 51(1), 33–35. Chemistry and Electronic Structure of the
2.3 Grafting Molecular Properties onto Semiconductor Surfaces 149
H2 Plasma Passivated Surface of CdTe, 63. R. Maoz, J. Sagiv, On the Formation and
J. Appl. Phys. 1994, 73(3), 1632–1637. Structure of Self-Assembling Monolayers,
50. D. N. Bose, S. Basu, K. C. Mandal, Char- J. Coll. Int. Sci. 1984, 100(2), 465–496.
acterization of Chemically Modified CdTe 64. A. Ulman, An Introduction to Ultrathin
Surfaces, Thin Solid Films 1988, 164, 13–19. Organic Films, Academic Press, New York,
51. D. N. Bose et al., XPS Investigation of 1991.
CdTe Surfaces: Effect of Ru Modification, 65. R. Cohen et al., Molecular Electronic Tun-
Semicon. Technol. 1989, 4, 866–870. ing of Si Surfaces, Chem. Phys. Lett. 1997,
52. J. L. Shaw et al., Chemically Controlled 279, 270–274.
Deep Level Formation and Band-Bending 66. B. Sweryda et al., A Comparison of Porous
at Metal-CdTe Interfaces, Appl. Phys. Lett. Silicon and Silicon Nanocrystallite Pho-
toluminescence Quenching with Amines,
1988, 53(18), 1723–1725.
J. Phys. Chem. 1996, 100(32), 13 776–13 780,
53. D. Cahen, R. Noufi, Surface Passivation of
and reference therein.
Polycrystalline, Chalcogenide Based Pho-
67. J. M. Lauerhaas, M. J. Sailor, Chemical
tovoltaic Cells, Solar Cells 1991, 30(1–4),
Modification of the Photoluminescence
53–59. Quenching of Porous Silicon, Science 1993,
54. Y. Nemirovsky, D. Rosenfeld, Surface Pas- 261, 1567–1568.
sivation and 1/f Noise Phenomena in 68. S. R. Lunt, P. G. Santangelo, N. S. Lewis,
HgCdTe Photodiodes, J. Vac. Sci. Technol. Passivation of GaAs Surface Recombination
1990, A8(2), 1159–1167. with Organic Thiols, J. Vac. Sci. Technol. B
55. Y. F. Chen et al., Hydrogen Passivation in 1991, 9(4), 2333–2336.
Cd1−x Znx Te Studied by Photolumines- 69. J. K. M. Chun et al., Proton Gated Emission
cence, Appl. Phys. Lett. 1991, 58(5), 493–495. from Porous Silicon, J. Am. Chem. Soc.
56. Y. F. Chen, W. S. Chen, Influence of Hy- 1993, 115, 3024–3025.
drogen Passivation on the Infrared Spectra 70. R. R. Chandler et al., Steric Considerations
of Hg0.8 Cd0.2 Te, Appl. Phys. Lett. 1991, in the Amine-Induced Quenching of Lumi-
59(6), 703–705. nescent Porous Silicon, J. Phys. Chem. 1995,
57. D. Cahen, R. Noufi, Defect Chemical Ex- 99(21), 8851–8855.
planation for the Effect of Air Anneal on 71. J. M. Lauerhaas et al., Reversible Lumines-
CdS/CuInSe2 Solar Cell Performance, Appl. cence Quenching of Porous Si by Sol-
Phys. Lett. 1989, 54(6), 558–560. vents, J. Amer. Chem. Soc. 1992, 114(5),
58. S. R. Lunt et al., Chemical Studies of the 1911–1912.
Passivation of GaAs Surface Recombination 72. A. Vilan, Chemical Modification of the
Using Sulfides and Thiols, J. Appl. Phys. electronic Properties of GaAs Surface, M.Sc
1991, 70(12), 7449–7467. thesis, Weizmann Institute of Science,
59. C. J. Sandorff et al., Enhanced Electronic Rehovot, 1996.
Properties of GaAs Surfaces Chemically 73. D. G. Wu et al., Novel NO biosensor based
on surface derivatization of GaAs by
Passivated by Selenium Reactions, J. Appl.
‘‘hinged’’ Fe porphyrin, Angew. Chem. Int.
Phys. 1990, 67(1), 586–588.
Ed. 2000, 39, 4496–4500.
60. S. A. Chambers, V. S. Sundaram, Struc-
74. R. Cohen et al., Frontier Orbital Model of
ture, Chemistry, and Band-Bending at Se-
Semiconductor Surface Passivation: Dicar-
Passivated GaAs(001) Surfaces, Appl. Phys. boxylic Acids on n- and p-GaAs, Adv. Mater.
Lett. 1990, 57, 2342–2344. 2000, 12, 33–37.
61. W. C. Simpson, J. A. Yarmoff, Fundamen- 75. R. Cohen et al., Controlling Electronic Prop-
tal studies of halogen reactions with III–V erties of CdTe by Adsorption of Dicarboxylic
semiconductor surfaces, Ann. Rev. Phys. Acid Derivatives: Relating Molecular Pa-
Chem. 1996, 47, 527–554. rameters to Band-Bending and Electron
62. Y. Liu, A. J. Komrowski, A. C. Kummel, Affinity Changes, Adv. Mater. 1997, 9,
Side–selective reaction of Br2 with second 746–749.
layer Ga atoms on the As-rich GaAs(001) 76. R. Cohen et al., Molecular Control over
2 × 4 surface, Phys. Rev. Lett. 1998, 81, Semiconductor Surface Electronic Proper-
413–416. ties: Dicarboxylic Acids on CdSe, CdTe,
150 2 Experimental Techniques
GaAs and InP, J. Am. Chem. Soc. 1999, 121, 90. M. Sturzenegger, N. S. Lewis, An X-Ray
10 545–10 553. Photoelectron Spectroscopic and Chemical
77. A. Vilan, A. Shanzer, D. Cahen, Molecular Reactivity Study of Routes to Functionaliza-
Control over Au/GaAs Diodes, Nature 2000, tion of Etched InP Surfaces, J. Am. Chem.
404, 166–168. Soc. 1996, 118, 3045–3046.
78. C. D. Bain, A New Class of Self-Assembled 91. J. W. Thackeray et al., Interactions of Di-
Monolayers: Organic Thiols on Gallium ethyldithiocarbamate with n-type Cadmium
Arsenide, Adv. Mater. 1992, 4(9), 591–594. Sulfide and Cadmium Selenide: Effi-
79. W. C. Sheen et al., A New Class of Orga- cient Photoelectrochemical Oxidation to the
nized Self-Assembled Monolayers: Alkane Disulfide and Flat-Band Potential of the
Thiols on GaAs (100), J. Am. Chem. Soc. Semiconductor as a Function of Adsorbate
1992, 114, 1514–1515. Concentration, J. Am. Chem. Soc. 1986, 108,
80. O. S. Nakagawa et al., GaAs Interfaces with 3570–3577.
Octadecyl Thiol Self-Assembled Monolayer: 92. M. J. Natan, J. W. Thackeray, M. S. Wrigh-
Structural and Electrical Properties, Jap. ton, Interaction of Thiols with n-Type
J. Appl. Phys. 1991, 30, 3759–376. Cadmium Sulfide and n-Type Cadmium
81. K. Asai et al., Electronic Passivation of Selenide in Aqueous Solutions: Adsorption
GaAs Surfaces by Electrodeposition of of Thiolate Anion and Efficient Photoelec-
Organic Molecules Containing Reactive trochemical Oxidation to Disulfides, J. Phys.
Sulfur, J. Appl. Phys. 1995, 77, 1582–1586. Chem. 1986, 90, 4089–4098.
82. E. D. Lu et al., A Sulfur Passivation of 93. J. J. Hickman, M. S. Wrighton, Face-Speci-
GaAs Surface by an Organic Molecular, fic Interactions of Anionic Sulfur Donors
CH3 CSNH2 Treatment, Appl. Phys. Lett. with Oriented Crystals of (0001) CdX (X =
1996, 69, 2282–2284. Se, S) and Correlation with Electrochemical
83. M. Tabib-Azar, A. S. Dewa, W. H. Ko, Properties, J. Am. Chem. Soc. 1991, 113,
Langmuir-Blodgett Film Passivation of 4440–4448.
Unpinned n-Type Gallium Arsenide 94. T. Uchihara et al., Effect of Ethylenedi-
Surfaces, Appl. Phys. Lett. 1988, 52, aminetetraacetic Acid on the Photocatalytic
206–208. Activities and Flat-Band Potentials of Cad-
84. V. J. Rao, V. S. Kulkarni, Interface Barrier mium Sulfide and Cadmium Selenide,
Height in n-GaAs(100)/Langmuir-Blodgett J. Phys. Chem. 1990, 94, 415–418.
Film Structures, Thin Solid Films 1991, 198, 95. J. K. Lorenz, T. F. Kuech, A. B. Ellis, Cad-
357–362. mium Selenide Photoluminescence as a
85. N. K. Singh et al., Etching Reactions of Probe for the Surface Adsorption of Di-
C2 H5 I on GaAs(100), Sur. Sci. 1998, 409, alkyl Chalcogenides, Langmuir 1998, 14,
272–282. 1680–1683.
86. P. J. Lasky et al., The Adsorption and 96. G. J. Meyer et al., Semiconductor-Olefin
Thermal Reaction of Dimethylcadmim, Adducts. Photoluminescent Properties of
Dimethylzinc and Trimethylgallium on Cadmium Sulfide and Cadmium Selenide
GaAs(110), Surf. Sci. 1996, 364, 312–324. in the Presence of Butenes, J. Am. Chem.
87. H. T. Lam, N. Venkateswaran, J. M. Vohs, Soc. 1989, 111, 5146–5148.
The Reactions of Diethylzinc on Gallium- 97. C. J. Murphy et al., Photoluminescence-
Rich and Arsenic-Rich Reconstructions of Based Correlation of Semiconductor Elec-
GaAs(100), Surf. Sci. 1998, 401, 34–46. tric Field Thickness with Adsorbate Ham-
88. T. Hou et al., Passivation of GaAs (100) mett Substituent Constants. Adsorption
with an Adhesion Promoting Self-Assemb- of Aniline Derivatives onto Cadmium
led Monolayer, Chem. Mater. 1997, 9, Selenide, J. Am. Chem. Soc. 1990, 112,
3181–3186. 8344–8348.
89. Y. Gu, D. H. Waldeck, Electron Tunneling 98. G. C. Lisensky et al., Electro-Optical Evi-
at the Semiconductor-Insulator-Electrolyte dence for Chelate Effect at Semiconductor
Interface. Photocurrent Studies of the Surfaces, Science 1990, 248, 840–843.
n-InP-Alkanethiol-Ferrocyanide System, 99. G. J. Meyer, G. C. Lisensky, A. B. Ellis, Ev-
J. Phys. Chem. B 1998, 102, 9015–9028. idence for Adduct Formation at the
2.3 Grafting Molecular Properties onto Semiconductor Surfaces 151
Photovoltaic Structures, Electron. Lett. 1980, tions by a New GaAs-Based Sensor, Chem-
16(6), 201–202. istry Eur. J. 2001, 7, 1743–1749.
123. I. H. Campbell et al., Controlling Charge 130. D. G. Wu et al., Tuning of Au/n-GaAs
Injection in Organic Electronic Devices Diodes with Highly Conjugated Molecules,
Using Self-Assembled Monolayers, Appl. J. Phys. Chem. B 2001, in press.
Phys. Lett. 1997, 71, 3528–3530. 131. G. Ashkenasy et al., Assemblies of
124. H. Sirringhaus, N. Tessler, R. H. Friend, ‘‘Hinged’’ Iron-Porphyrins as Potential
Integrated Optoelectronic Devices Based on Oxygen Sensors, J. Amer. Chem. Soc. 2000,
Conjugated Polymers, Science 1998, 280, 122, 1116–1122.
1741–1744. 132. A. Ivanisevic et al., Linker-enhanced Bind-
125. L. Kronik et al., Surface States and Photo- ing of Metalloporphyrins to CdSe and
voltaic Effects in CdSe Quantum Dot Films, Implications for O2 detection, Langmuir
J. Electrochem. Soc. 1998, 145, 1748–1755. 2000, 16, 7852–7858.
126. A. Hagfeldt, M. Grätzel, Molecular Photo- 133. T. Vondrak et al., Interfacial Electronic
voltaics, Acc. Chem. Res. 2000, 33, 269–277. Structure in Thiolate Self-assembled Mono-
127. D. Cahen et al., Nature of Photovoltaic Ac- layers; Implications for Molecular Elec-
tion in Dye-Sensitized Solar Cells, J. Phys. tronics, J. Amer. Chem. Soc. 2000, 122,
Chem. B 2000, 104, 2053–2059. 4700–4707.
128. Y. Selzer, D. Cahen, Fine-tuning of 134. M. Carrara, F. Nüesch, L. Zuppiroli, Car-
Au/SiO2 /Si diodes by varying interfacial boxylic acid anchoring groups for the
dipoles using molecular monolayers, Adv. construction of self-assembled monolayers
Mater. 2001, 13, 508–511. on organic device electrodes, Synth. Metals
129. D. G. Wu et al., Direct Detection of Low 2001, 121, 1633–1634.
Concentration NO in Physiological Solu-
2.4 Capacitance, Luminescence, and Related Optical Techniques 153
0 X
0
(b)
Electric field
0 W
0 X
(c)
Potential
X
0
(d)
− eUfb
Energy
EC
EF = − eU
Eredox
EV
(e)
This equation can be rewritten as follows: charges and surface dipoles cause a shift
! " of the semiconductor band positions at the
1 2 kT surface, as shown schematically in Fig. 2.
= U − U fb − (6)
Cd2 qε0 εs ND q The shift is expressed as a change in φ(0) or
Ufb in the foregoing equations. However,
The plot of 1/Cd2 measured against U is the preceding equations themselves can
called the Mott–Schottky plot, which can be be applied to such cases with the changed
used to determine Ufb and ND (or NA ), as φ(0) or Ufb .
discussed later.
It has been assumed thus far that no sur- 2.4.1.3 Measurement of Differential
face charge nor surface dipole is present at Capacitance
the semiconductor. In general cases, both The capacitance of the semiconduc-
surface charges and surface dipoles are tor–electrolyte interface can be measured
present in the semiconductor owing to ad- by use of a semiconductor electrode, in
sorption equilibria of various ions between which the front side of the semiconductor
the electrolyte and the semiconductor sur- is in contact with the electrolyte and the
face as well as formation of polar bonds at rear side is electrically connected with a
the semiconductor surface. Such surface metallic leading wire via an ohmic contact.
−
− +
Ions at the + − +
− +
Helmholtz layer + − + +
− + Ionized donor
Surface charge −
+ − + +
n -type semiconductor
Electrolyte
(a)
Depletion n
region Region
Charge density
0 X
0
(b)
Potential
0
X
(c)
Fig. 2 Schematic illustration of
(a) charge distribution, − eUfb
(b) charge-density distribution,
Energy
C.E. W.E.
The experimental setup for capacitance hydrogen iodide [5]. It should be noted that
measurements is schematically shown in the slope of the straight line depends not
Fig. 3. The electrode potential U is regu- only on ND (or NA ) and εs but also on the
lated with a potentiostat. The differential true surface area (or surface roughness) of
capacitance is measured by superimposing the semiconductor electrode.
an AC voltage with a small amplitude of The Ufb is one of the most important
about 10 mV and a frequency of a few Hz quantities for semiconductor electrodes
to 1 MHz on the electrode potential. One because it determines the band edge
can use a commercial impedance analyzer positions at the semiconductor–electrolyte
to measure the differential capacitance, to- interface, which in turn, determine the
gether with a personal computer to analyze energies of conduction-band electrons
obtained data automatically. and valence-band holes reacting with the
Recently, scanning capacitance micro- electrolyte solution. It is known that Ufb
scopy (SCM) using a fine tip has been for most semiconductors, such as n- and
developed to investigate two-dimensional p-GaAs, GaP, InP, n-ZnO, n-TiO2 , and
dopant profiling of semiconductor sur- n-SnO2 , in aqueous electrolytes is solely
faces [4]. determined by the solution pH and shifts
in proportion to pH with a ratio of
2.4.1.4 Mott–Schottky Plots and Flat Band −0.059 V/pH [1, 2]. This is explained by an
Potentials adsorption equilibrium for H+ or OH− at
Equation 6 indicates that a plot of 1/Cd2 the semiconductor–electrolyte interface,
against U gives a straight line with a for example,
slope of (2/qε0 εs ND ), which is termed the
Ss -OH + Haq + &
−−
−−−
%− Ss -OH2
+
(7)
Mott–Schottky plot, as mentioned earlier.
The extrapolation of the straight line to where Ss -OH refers to surface OH group
1/Cd2 = 0 gives (Ufb + kT /q). Therefore, at the semiconductor.
the plot can be used to determine the The Ufb for n- and p-Si [6] and metal
flat band potential Ufb . The donor density calcogenide semiconductors such as n-
ND (or the acceptor density NA ) can CdS, n-CdSe, and CdTe [2, 7] does not
also be determined from the slopes obey the foregoing rule, remaining nearly
of the plots. Figure 4 shows examples constant in a range of pH lower than about
of Mott–Schottky plots, obtained for 6 for Si and about 10 for n-CdS. This is
n-Si(111) and n-Si(100) electrodes in 7.1 M most probably because the semiconductor
2.4 Capacitance, Luminescence, and Related Optical Techniques 157
Fig. 4 Examples of 6
Mott–Schottky plots obtained n- Si (111)
for n-Si(111) and n-Si(100) 5 10.0 ∼ 15.0 Ωcm
(1/C2) × 1017
electrodes in aqueous 7.1 M HI. 4
[F−2cm4]
3
2 n- Si (100)
0.8 ∼ 1.16 Ωcm
1
0
−0.8 −0.4 0 0.4 0.8 1.2 1.6
Potential versus SCE
[V]
surface has no OH group in this pH range the radiative recombination may occur ei-
and no adsorption equilibrium for H+ or ther in the semiconductor bulk or at the
OH− is attained. For metal calcogenide semiconductor surface (Fig. 5a).
semiconductors such as n-CdS, the Ufb On the other hand, EL is emitted when
shifts by adsorption of HS− and Cd2+ an n-type semiconductor electrode, for ex-
ions. It is also known that Ufb for ample, is negatively biased in an electrolyte
some semiconductor electrodes shifts by a solution containing a strong oxidant. Un-
change in the surface termination bond [8, der this condition, holes are injected into
9], as well as electrode illumination [10, the valence band by the oxidant, some
11] and the presence of a redox couple [10, of which recombine with electrons in the
12]. The Ufb in nonaqueous electrolytes conduction band existing as the majority
has been reported for some semiconductor carrier in the n-type semiconductor, result-
electrodes [13, 14]. ing in emission of luminescence called EL,
as illustrated in Fig. 5(b). Similarly, EL is
2.4.2 emitted when a p-type semiconductor elec-
Luminescence from Semiconductor trode is positively biased in an electrolyte
Electrodes
solution containing a strong reductant be-
2.4.2.1 Photoluminescence and cause some of the electrons, injected into
Electroluminescence the conduction band by the reductant, re-
Illumination of a semiconductor electrode combine with holes existing as the majority
generates excited electrons in the con- carrier in the p-type semiconductor. In
duction band and holes in the valence both cases, EL may be emitted either via
band. Some of them recombine with each direct recombination between the conduc-
other radiatively, resulting in emission of tion and valence bands or via an impurity
luminescence called photoluminescence level(s), and either via recombination in
(PL). The radiative recombination may the semiconductor bulk or at the semi-
occur directly between the conduction conductor surface. The situation is quite
and valence bands (inter-band transition) the same as that in PL emission (Fig. 5).
or via certain impurity or defect levels Figure 6 shows examples of PL and
within the band gap at which either elec- EL spectra from some semiconductor
trons or holes, or both are trapped, as electrodes [15]. It is to be noted, however,
schematically illustrated in Fig. 5(a). Also, that such spectra strongly depend on
158 2 Experimental Techniques
Conduction band
−
− − − − −
Surface
luminescence
Illumination
Bulk luminescence
Surface
states
+ +
+
Valence band
(a)
Conduction band
− − − −
Surface luminescence
Bulk luminescence
Surface
states
+ +
e Oxidant
+ (Fe3+, HO , or SO4 −)
• •
Valence band
(b)
Fig. 5 Mechanisms of emission of (a) PL and (b) electroluminescence
(EL) from a semiconductor electrode in an electrolyte solution.
Eg
n -ZnO
Eg
n -TiO2
Eg
25 20 ∼
v 15 10
[103 cm−1]
Monochromater
Mirror
Hg lamp
Filter
Monochromater
Photomultiplier
Pt
Semiconductor V
SCE
Photodiode Multichannel
pulse height
Laser analyzer
Start
Stop
Time
Picosecond single-photon timing spectroscopy
Fig. 8 Experimental setup for a time-correlated single-photon counting
method or picosecond single-photon timing method, used for
time-resolved luminescence measurements in a range of picoseconds.
PM
T
1000 C343/MeOH
( τ ET > 1.5 psec)
Counts
D-1421/MeOH
D-1421/water ( τ ET = 30 fsec)
C343/water ( τ ET < 20 fsec)
0
−0.5 0 0.5 1.0 1.5
Time
[ps]
Fig. 10 Examples of luminescence decays obtained by femtosecond up-conversion
spectroscopy. The luminescence is emitted from a cumarin dye adsorbed on nanocrystalline
TiO2 particles suspended in water or methanol (MeOH). C343: cumarin 343 and D-1421:
7-diethylaminocumarin-3-carboxylic acid.
2.4 Capacitance, Luminescence, and Related Optical Techniques 163
H2Q/Q
X1.47 e
Br − / Br2
1.0 hν
(Illumination) CI− / CI2
2.0
OH−
3.0 +
Valence band
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 .
164 2 Experimental Techniques
on surface structures and processes. The reaction intermediate (X1.47 , probably va-
most important is to measure surface PL cant 2p-level of surface Ti-O· radical) of
that is emitted via surface states connected water photooxidation, as shown in Fig. 11.
with surface intermediates of electrode re-
actions. An example of such luminescence 2.4.3
is a luminescence band peaked at 840 nm Other Optical Techniques
(Fig. 6), observed for a rutile-type n-TiO2
electrode, which was activated beforehand 2.4.3.1 Time-resolved Laser Spectroscopy
by photoetching in aqueous H2 SO4 un-
der anodic bias. In situ PL measurements, 2.4.3.1.1 Transient Absorption Spectro-
combined with in situ photocurrent mea- scopy Measurements of transient ab-
surements, can lead to detailed studies on sorption spectra are much more difficult
molecular mechanisms of electrode reac- than the measurements of luminescence
tions. It is concluded [25] that the 840-nm (PL) described thus far, especially for solid
band for rutile n-TiO2 is emitted via a samples. However, time-resolved tran-
transition of an electron in the conduc- sient absorption spectroscopy has been
tion band to a vacant level of a surface used to study carrier dynamics in the
SHG
2ω
ω
H2O
Optical delay
(in 1 cm cell)
fs white-light
continuum
Pump pulse
(390 nm)
Monochro- MCPD2
mator
Powder sample
(in 2 mm cell)
Iris
Monochro-
mator
; Mirror
MCPD1
; Dichroic mirror
; Filter
semiconductor bulk and at the semicon- conduction band of TiO2 particles. The rate
ductor surface. Figure 12 shows an exam- of photoelectron transfer from adsorbed
ple of experimental setup for femtosecond dyes to the conduction band of TiO2 in
(pump probe) transient absorption spec- vacuum and solution is also determined by
troscopy [26]. The principle is to generate use of measurements of transient absorp-
an excited state of the sample (semiconduc- tion of injected electrons. It is reported
tor electrode) by a laser pulse (pump pulse) that the photoelectron transfer occurs in
and then measure an absorption spectrum a time range of 150–25 femtoseconds
of the excited state by another laser pulse for chemically bound, strongly interacting
(probe pulse), with varied delays in time dyes [28–31].
from the pump pulse. A femtosecond
laser, such as a Ti-sapphire laser, is used 2.4.3.1.2 Transient Grating Spectroscopy
as the light source. In Fig. 12, a laser pulse Transient grating spectroscopy is rela-
with a frequency ω from the light source tively easily handled compared with the
is converted with BBO to a light pulse with transient absorption spectroscopy, and is
2ω for use as the pump pulse. The original often used to study carrier dynamics at
pulse with ω is separated with a mirror semiconductor electrodes [32]. Figure 14
and used for the probe pulse. The latter schematically shows the principle of tran-
pulse is led through an optical delay cir- sient grating spectroscopy. A femtosecond
cuit to an H2 O cell to be converted into laser pulse for sample excitation is split
white light continuum for measurements into two beams, which are crossed again
of absorption spectra. The spectral inten- at the semiconductor surface to produce
sity of the probe pulse is measured with a an optical striped interference pattern. The
multichannel photodiode (MCPD). interference pattern produces a striped pat-
Figure 13 shows an example of tran- tern of the densities of photo-generated
sient absorption spectra observed by direct electrons and holes near the semicon-
excitation of nanocrystalline TiO2 parti- ductor surface. The latter striped pattern
cles suspended in vacuum [27]. The broad gives rise to a striped pattern of optical
absorption band in the red to near in- refractive index near the semiconductor
frared region is assigned to light absorp- surface, which is monitored by measuring
tion of photo-generated electrons in the a diffraction pattern of a second probe laser
20 0.2 ps
2 ps
10 ps
15 100 ps
% Absorption
1 ns
5 ns
[%]
10
5
Fig. 13 Transient absorption spectra of
nanocrystalline TiO2 particles in
vacuum after excitation by a 0
femtosecond laser pulse (390 nm, 400 450 500 550 600 650 700 750
170 fs, 2 mJ cm−2 ). The delay time after Wavelength
excitation is indicated in the figure. [nm]
166 2 Experimental Techniques
White light
continuum Λ Diffraction
spectrum
θ
Wavelength / nm
Probe
Transient grating
Reflection
Transmission
Fig. 14 A schematic view of a sample region in transient grating spectroscopy to
explain its principle.
b
a
0 10 20 30 40 50
Time
[ps]
pulse (white light continuum). The carrier surface structures at semiconductor elec-
dynamics near the semiconductor surface trodes. In situ multiple internal reflectance
is observed as rise and decay curves for Fourier Transform Infrared (FTIR) spec-
the diffraction pattern of the second laser troscopy is widely used to investigate
probe pulse. surface termination bonds and their re-
Figure 15 shows examples of decays of actions. Figure 16 shows a semiconductor
transient grating signal (the intensity of the (Si) wafer and an electrochemical cell
diffraction pattern) observed for an n-TiO2 for in situ multiple internal reflectance
(100) electrode by excitation at 360 nm spectroscopy. By this method, various
and probing at 670 nm [33]. The decays hydrogen termination bonds (Si−H, SiH2 ,
are related with the rate of electron-hole and SiH3 ) at terraces and steps for hy-
recombination near the n-TiO2 surface. drogen fluoride–etched or ammonium
fluoride–etched Si (111) and (100) are
2.4.3.2 In Situ Spectroscopic Investigation clearly detected with good spectral reso-
of Semiconductor Surfaces lution [34, 35].
Various spectroscopic techniques have In situ investigations of semiconduc-
been used for in situ investigations of tor surfaces are also done by means
2.4 Capacitance, Luminescence, and Related Optical Techniques 167
IR
Si prism
of second harmonic generation (SHG) related with the density of free carriers,
spectroscopy, ultraviolet and infrared light and modulated illumination enables us to
mixing spectroscopy, (surface enhanced) investigate minority carrier dynamics.
laser Raman spectroscopy, X-ray diffrac- Surface photovoltage (SPV) measure-
tion (XRD) method, electron paramag- ments have been used to investigate
netic resonance (EPR) spectroscopy, and electronic structures at the semiconduc-
so forth. Space-resolved investigations are tor surfaces for semiconductor–vacuum
also done by means of surface near-field and semiconductor–gas interfaces. This
optical microscopy (SNOM), photoconduc- method is applied to semiconductor–
tive atomic force microscopy, and so forth. liquid interfaces in case of insulating liq-
uids [39].
2.4.3.3 Other Miscellaneous Techniques
2.4.4
The electrolyte electroreflectance (EER) Summary
method is successfully used [36, 37] to
determine the flat band potential (Ufb ). Various methods to measure the capac-
In this method, the optical reflectance itance and luminescence (PL and EL)
at the semiconductor electrode surface at the semiconductor–electrolyte inter-
is measured under modulation of the face have been reviewed together with
electrode potential by superimposition of related optical techniques to investigate
a small AC voltage. The modulation of surface structures, optical properties, and
the electrode potential causes modulation carrier dynamics at the interfaces. Re-
of the density of majority carriers near cent rapid progress in laser spectroscopy
the semiconductor surface, which in turn, has enabled us to investigate very fast
causes modulation in light reflectance. interfacial processes with a high time
In situ microwave photoconductivity resolution of a few ten femtoseconds.
measurements, combined with photo- Moreover, rapid progress in scanning
current measurements, are also success- probe microscopy, combined with optical
fully used [38] to investigate dynamics techniques, has enabled us to investigate
of photo-generated minority carriers and interfacial structures and processes with
interfacial kinetics at semiconductor elec- a high space resolution of an atomic
trodes. The microwave signal is closely (subnanometer) scale. Progress in these
168 2 Experimental Techniques
fields continues and it is thus highly prob- 18. M. Oyama, M. Mitani, M. Washida et al., J.
able that other new powerful methods will Electroanal. Chem. 1999, 473, 166–172.
19. D. V. O’Connor, D. Phillips in Time-correla-
appear in the near future.
ted Single Photon Counting, Academic Press,
New York, 1984.
References 20. I. Yamazaki, N. Tamai, H. Kume et al., Rev.
Sci. Instrum. 1985, 56, 1187–1194.
1. S. R. Morrison in Electrochemistry at Semicon- 21. N. Tamai, M. Ishikawa, N. Kitamura et al.,
ductor and Oxidized Metal Electrodes, Plenum Chem. Phys. Lett. 1991, 184, 398–403.
Press, New York, 1980. 22. G. S. Beddard, T. Dout, G. Porter, Chem.
2. Y. V. Pleskow, Y. Y. Gurevich, in Semicon- Phys. 1981, 61, 17.
ductor Photoelectrochemistry (Translated by 23. K. Murakoshi, S. Yanagida, M. Capel et al.,
P. N. Bartlett), Consultants Bureau, New ACS Symp. Series 679 (Nanostructured
York, 1986. Materials-Clusters, Thin Films, and Compos-
3. S. M. Sze in Physics of Semiconductor Devices, ites), Amer. Chem. Soc. 1997, Chap. 17,
2nd edition, John Wiley & Sons, New York, pp. 221–238.
1981. 24. K. Mecker, A. B. Ellis, J. Phys. Chem. B 1999,
4. V. V. Zavyalov, J. S. McMurray, C. C. 103, 995–1001.
Williams, Rev. Sci. Instrum. 1999, 70, 25. Y. Nakato, H. Akanuma, Y. Magari et al., J.
158–164. Phys. Chem. B 1997, 101, 4934–4939.
5. Unpublished result in a laboratory of 26. T. Asahi, A. Furube, M. Ichikawa et al., Rev.
Prof. Y. Nakato, Osaka University. Sci. Instrum. 1998, 69, 361.
6. Y. Nakato, T. Ueda, Y. Egi et al., J. Electro- 27. A. Furube, T. Asahi, H. Masuhara, J. Phys.
chem. Soc. 1987, 134, 353–358. Chem. B 1999, 103, 3120.
7. H. Minoura, T. Watanabe, T. Oki et al., Jpn. 28. Y. Tachibana, J. E. Moser, M. Graetzel et al.,
J. Appl. Phys. 1977, 16, 865. J. Phys. Chem. 1996, 100, 20056–20062.
8. J. N. Chazalviel, J. Electroanal. Chem. 1987, 29. T. Hannappel, B. Burfeindt, W. Storck et al.,
233, 37. J. Phys. Chem. B 1997, 101, 6799–6802.
9. M. Fujitani, R. Hinogami, J. G. Jia et al., 30. R. J. Ellingson, J. B. Asbury, S. Ferrere et al.,
Chem. Lett. 1997, 1041–1042. J. Phys. Chem. B 1998, 102, 6455–6458.
10. A. J. Nozik, R. Memming, J. Phys. Chem. 31. J. B. Asbury, R. J. Ellingson, H. N. Ghosh
1996, 100, 13061–13078. et al., J. Phys. Chem. B 1999, 103, 3110–3119.
11. A. J. McEvoy, M. Etman, R. Memming, J. 32. J. J. Kasinski, L. A. Comez-Jahn, K. J. Faran
Electroanal. Chem. 1985, 190, 225. et al., J. Chem. Phys. 1989, 90, 1253–1269.
12. Y. Nakato, A. Tsumura, H. Tsubomura, J. 33. Unpublished result in a laboratory of
Electrochem. Soc. 1981, 128, 1300–1304. Prof. N. Tamai, Kwansei-Gakuin University.
13. B. L. Wheeler, A. J. Bard, J. Electrochem. Soc. 34. P. Jakob, Y. J. Chabal, J. Chem. Phys. 1991,
1983, 130, 1680. 95, 2897–2909.
14. W. Jaegermann in Modern Aspects of Electro- 35. M. Niwano, T. Miura, N. Miyamoto, J. Elec-
chemistry (Eds.: R. E. White, B. E. Conway, trochem. Soc. 1998, 145, 659–661.
J. O’M. Bockris), Plenum, New York, 1996, 36. M. Pujadas, P. Salvador, J. Electrochem. Soc.
pp. 136. 1989, 136, 716–723.
15. Y. Nakato, A. Tsumura, H. Tsubomura, 37. L. J. Ferrer, H. Muraki, P. Salvador, J. Phys.
Chem. Phys. Lett. 1982, 85, 387–390. Chem. 1986, 90, 2805–2807.
16. ftp://ftp.batnet.com/pub/wombats/srsys/ 38. A. M. Chaparro, H. Tributsch, J. Phys. Chem.
ftp/lia.pdf B 1997, 101, 7428–7434.
17. ftp://ftp.batnet.com/pub/wombats/srsys/ 39. S. Bastide, D. Gal, D. Cahen, Rev. Sci. In-
ftp/sr.pdf strum. 1999, 70, 4032–4036.
173
they were deposited directly on graphite [6, ‘‘wires’’, was used to improve thin film
8, 14]. The graphite, in contrast to metals, CdS/CuInS2 photovoltaic cells [20].
apparently does not efficiently quench the The nature of the anion of the metal
luminescence. Of particular importance is salt is important in this technique. Salts of
the observation that, under suitable prepa- many anions are either insoluble or unsta-
ration conditions, all the semiconductors ble in hot dimethyl sulfoxide (DMSO) con-
emitted essentially band gap luminescence taining dissolved chalcogen. Perchlorate
at room temperature with little or no and chloride are the most commonly used
deep sub–band gap band response that anions although other halides, methylsul-
is normally characteristic of these materi- fonate, and borofluoride have also been
als. This implies that the nanocrystals were used. However, another reason that the
of very high quality. anion is important, particularly relevant
to the present discussion, is that it af-
3.1.3 fects the crystal size and therefore the
Electrodeposition of Nanocrystals from band gap. Films deposited from Cl− so-
Nonaqueous Solution lutions exhibit band gaps between 0.1 and
0.2 eV higher than those of the same com-
One of the simplest techniques used to pound deposited from ClO4 − [21]. More
electrodeposit semiconductors is cathodic recent and detailed studies indicate that
deposition from nonaqueous solutions that this difference is due to adsorption
containing elemental chalcogen (S, Se) and of the more strongly adsorbed Cl− on the
a metal salt, first described by Baranski and growing CdX surface, preventing further
Fawcett [17]. Two main mechanisms have crystal growth [22].
been considered: deposition of metal (e.g. Although both CdSe and CdS can be
Cd) followed by chemical reaction with deposited in nanocrystalline form by this
elemental chalcogen in solution and re- nonaqueous deposition, the essentially
duction of chalcogen to (poly)chalcogenide total insolubility of Te in DMSO prevents
followed by ionic reaction between chalco- the use of the method for deposition of
genide and metal cations. Which mech- nanocrystalline tellurides (a small amount
anism will dominate depends on the of Te can be codissolved with Se and mixed
specific system (substrate, semiconductor, selenide-tellurides with small amounts of
deposition conditions) and may change in Te can be deposited; see following section).
the same deposition. However, a related method to deposit
This technique was found to give CdTe has been described by Cocivera and
nanocrystalline semiconductors which ex- associates [23, 24]. They reacted elemental
hibited quantum size effects with typical Te with tri-n-butyl phosphine (TBP), which
lateral crystal size of 5 nm [18, 19]. The reacts with Te to form TBP telluride.
height of the crystals [measured by X-ray This compound, together with a Cd salt
diffraction (XRD)] was often up to several dissolved in propylene carbonate, allowed
times larger than the lateral dimensions cathodic electrodeposition of CdTe. The
(measured by TEM). In fact, the increased as-deposited films were reported to be X-
transparency to shorter wavelengths be- ray amorphous, a fact that suggested that
cause of size quantization, together with they might in fact be nanocrystalline [26].
good (photo)conductivity expected from (Cd,Hg)Te films grown by the same
3.1 Preparation of Nanocrystalline Semiconductor Materials 177
technique exhibited a crystal size of about and {110}Au ||{11.0}Cdse orientation rela-
5 nm [25]. tionship. The epitaxy arises because of
This technique was modified and sim- the good lattice match between the CdSe
plified, in particular, by using a one- and Au in a 2 : 3 relationship (−0.6% mis-
step technique to prepare the solution in match). Beyond the first layer of crystals,
DMSO [26]. The CdTe was indeed found to the epitaxy is gradually lost and the crystal
be nanocrystalline with a wide size distri- size grows for the perchlorate bath; for a
bution varying from several nm up to tens CdCl2 bath, the crystal size does not grow
of nm. In addition, the films were gen- much because of capping, as discussed ear-
erally nonstoichiometric with excess Te or lier. The mismatch strain, which gradually
Cd, depending on the deposition potential. increases as the crystal grows, eventually
It was difficult to deposit close to stoichio- leads to termination of growth. This can ex-
metric film. To improve the stoichiometry, plain the relatively narrow size distribution
reverse pulse deposition was used to strip obtained.
excess Cd or Te during the anodic pulse. This hypothesis of strain-determined
The pulse regime also decreased the crystal growth termination suggested that crys-
size to an extent depending on the pulse tal size should be controllable by choice
parameters (crystal size typically several of the semiconductor and substrate lat-
nm) and improved the size distribution. tice parameters: the larger the mismatch,
The main factors limiting crystal growth the smaller the crystal size, and vice
were short pulse on-times and capping of versa. A range of different electrode-
the crystals with strongly adsorbing phos- posited semiconductor-substrate combi-
phine, mainly during the pulse off-time. nations has been investigated to test this.
The lattice parameter of the CdSe was
3.1.4 varied with (assumed) little change in
Size Control Using Semiconductor- chemical interaction with the substrate
substrate Lattice Mismatch by depositing an alloy of CdSex Te1−x [29].
From Vegard’s law, a value of x = 0.88
An important factor that can influence (12% Te) should result in an increased
crystal size, particularly for the first layer lattice with a perfect match to Au. Although
of crystals, is the nature of the substrate. Te could not be dissolved in DMSO,
Deposition of CdSe from the above DMSO small amounts of Te could be dissolved
electrolyte using Cd(ClO4 )2 onto films of in the presence of dissolved Se. The
evaporated Au on glass or mica resulted concentration of Te in the electrolyte
in crystals of about 4–5 nm in size. was not known (it was almost certainly
The distribution of the nanocrystals on <1 mM), but the amount of Te in
the Au depended on current density and the deposit was quantitatively varied by
deposition temperature: high currents and varying the deposition current density
low temperatures favored isolated crystals, and temperature. Relatively high current
whereas increasing aggregation occurred densities resulted in severe depletion of the
with higher temperature and decreased Te near the cathode, thereby favoring the
current [27]. more concentrated Se in the deposit. In the
ED [27] and HTEM [28] showed that same way, higher deposition temperatures
the CdSe crystals were epitaxially de- increased the concentration of Te at the
posited on the Au in a {111}Au ||{00.2}Cdse cathode and therefore in the deposit.
178 3 Semiconductor Nanostructure
The maximum Te concentration in the forms fairly large particles (10–20 nm),
deposits was measured to be only about which become larger with increasing de-
5%. However, this was sufficient to see position time. Even more important, the
a major increase (by up to several times) CdSe deposit on Pd is both XRD and ED
in nanocrystal size with increase in Te amorphous. However, careful examina-
concentration, whereas the epitaxy was, as tion of a HTEM image reveals the presence
expected, retained. of many regions showing short-range or-
The lattice parameter was decreased der, often with hexagonal symmetry. An
(with expectation of reduced crystal size) by in-depth analysis of this short-range order
alloying the Cd with Zn [30]. Cdx Zn1−x Se showed the presence of irregular struc-
was epitaxially deposited using very low tures of ordered CdSe, on the order of 1 nm
concentrations of Cd in the deposition in size, surrounded by disordered CdSe.
solution. The crystal size for x = 0.78 was Furthermore, Fourier analysis of HTEM
between 2 and 3 nm – half the size of the images showed some degree of preferen-
corresponding CdSe deposits. The crystal tial orientation of the ordered structures
height, measured by XRD, was 5 nm. relative to the Pd lattice, but rotated 30◦
The crystals are therefore short quantum to the Pd [32]. Subsequent modeling of
columns rather than QDs. the superimposed CdSe-Pd lattices showed
CdS has a 4.5% mismatch with Au (2 : 3 that this 30◦ rotation indeed resulted in
ratio). If epitaxy was to occur for such a an improved lattice match compared with
large mismatch, the crystal size should the aligned lattices [33]. The very small
be very small. In this case, the proper- ordered structures are in qualitative agree-
ties of the deposit were dependent on the ment with the mismatch strain argument
amount deposited [31]. For small amounts that predicts very small crystals. However,
of deposit, the crystals are predominantly it is likely that differences in chemical re-
epitaxial (although not perfectly so as for activity between CdSe and Au or Pd also
CdSe on Au). The average crystal size play a role.
grows, and the degree of orientation de- CdS has only a small mismatch with
creases as the amount of deposit increases. Pd – +0.17% (2 : 3 ratio). However, at-
The crystal height for a nominally 5-nm tempts to obtain thin electrodeposits of
thick (in reality, ≤3 nm) deposit was found CdS on Pd were not successful. Small par-
to be 3 nm – similar to the lateral ones. ticles (approximately 2 nm) were obtained;
The mismatch may also be varied by however, the identity of these particles was
changing the material of the substrate. Pd not known – XPS analysis showed the al-
has the same fcc cubic structure as Au; most total absence of Cd. In this case,
the lattice spacing, however, is almost 5% chemical interaction between S and Pd
smaller than that of Au, with the result probably dominates the surface.
that the mismatch between CdSe and Pd In an attempt to separate chemical
is +4.1%. This is close to the mismatch effects from lattice mismatch, we studied
between CdS and Au but in the opposite an alloy of 3% Cd with Au. Cd was chosen
direction, that is, the CdSe should be com- because of its strong interaction with Se.
pressed instead of stretched. In contrast The change in lattice-spacing owing to
to CdSe on Au, which grows as isolated the 3% of Cd is very small (+0.07%) and
crystals on Pd, the CdSe covers the en- therefore any change in CdSe properties
tire surface and, with further deposition, can be attributed to differences in chemical
3.1 Preparation of Nanocrystalline Semiconductor Materials 179
interaction. The lateral crystal size is current or potential [35]. The deposition is
approximately 5 nm – similar to that on based on oxidation of Tl(I) to Tl(III) and
pure Au. However, the crystal height is Pb(II) to Pb(IV); the higher oxidation states
considerably smaller than this, as seen of these ions are readily hydrolyzed to the
qualitatively by TEM contrast and surface oxides. Because Tl2 O3 is deposited prefer-
coverage. Most surprising, however, is entially (owing to a less anodic potential)
that the deposit is not only a Wurtzite compared to PbO2 , low current densities
structure but also a mixture of Wurtzite, (or low anodic potentials) form essentially
Zincblende, and Rocksalt. Rocksalt CdSe only Tl2 O3 from mixed Tl/Pb solutions.
is normally stable only at high pressures. It If a solution is made with a higher Pb
was suggested that surface tension forces, than Tl concentration, then as the cur-
together with (or maybe because of) the rent density (anodic potential) is increased
strong interaction between the CdSe and the concentration of Pb in the resulting
the substrate, stabilized this high-pressure oxide increases as the low concentration
structure [34]. Tl becomes diffusion-limited. The inter-
play of relative cation concentrations, mass
3.1.5 transport conditions, and pulse character-
Electrodeposition of Layered istics (current/potential and time scales)
Nanostructures
imparts control of the stoichiometry, as
well as thickness of the individual layers.
Many modern semiconductor devices
The sharpness of the interface between
comprise alternating layers of different
layers can also be controlled: deposition
materials forming superlattices and mul-
where one of the constituent ions is in
tiple quantum wells. One well-known
the diffusion-controlled regime results in a
example of such structures is the diode
graded composition, whereas if deposition
laser, a mass-produced device. This device
depends on confinement of charges in the takes place in the kinetic control regime
two-dimensional structures for enhanced for both metal ions, a sharp compositional
laser output at lowered current thresholds. difference between layers is obtained [36].
Such alternating semiconductor layers are STM was used to identify the individ-
usually manufactured either by chemical ual layers [37, 38]. Both Tl2 O3 and PbO2
vapor deposition or by molecular beam are degenerate semiconductors with band
epitaxy. The thickness of the layers can be gaps of about 1.4 and 1.8 eV, respectively.
closely controlled in both techniques. As Therefore, such layers could be used as
mentioned earlier, electrodeposition also quantum wells with electrons confined
allows good control of thickness. preferentially in the lower band gap Tl-rich
The group of Switzer has been active layers.
in electrodeposition of alternating layers Layers of Tl2 O3 or Pb-Tl-O could be de-
of different metal oxides or metal/metal posited on steel with strong (100) or (210)
oxides. Ceramic superlattices of Pba -Tlb - texturing, respectively, and superlattices
Oc /Pbd -Tle -Of with individual thickness of Pb-Tl-O could then be electrodeposited
of 5–10 nm were electrodeposited onto epitaxially on these prelayers with these
stainless steel from an aqueous solution same orientations [39].
of TlNO3 and Pb(NO3 )2 in 5 M (NaOH) Tl2 O3 can be electrodeposited with a
at room temperature using either pulsed defect chemistry controllable through the
180 3 Semiconductor Nanostructure
although the transport in the lateral di- forming a rectifying contact. AC deposi-
rection was always ohmic, the transport in tion then resulted in formation of CdS
the vertical direction exhibited pronounced only in the pores and not in cracks or de-
negative differential resistance, attributed fects in the film. The CdS wire dimensions
to resonant tunneling from Cu into hole closely followed those of the pores. For the
states in Cu2 O [45, 46]. smaller pores, the wires were apparently
made up of single crystals joined in the ax-
3.1.6 ial direction, whereas for the larger pores,
Template-directed Electrodeposition the coherence length, measured by XRD,
was less than the pore diameter, suggest-
There are a number of studies on electrode- ing that the wires were composed of several
position of semiconductors in the pores of crystals also in the radial direction (or pos-
various membranes forming nanowires of sibly, defected single crystals). It should be
the semiconductor. Klein and associates noted that even for films of CdS deposited
electrodeposited CdSe and CdTe into the by this technique the crystal size is very
pores of Anopore membranes [47]. These small (see preceding section). The CdS
membranes are 50-µm thick with closely was hexagonal (wurtzite) with its c-axis
spaced 200-nm-sized pores. One side of aligned predominantly along the mem-
the membrane was sputter-coated with Au brane thickness. HTEM studies of these
followed by Ni electrodeposition, which nanowires confirmed the well-ordered and
partially filled the pores, as a substrate. basal plane-textured CdS crystallinity [50].
CdSe and CdTe were electrodeposited Small quantum size effects were seen for
from acidic aqueous solutions containing the smallest wires, with band gaps varying
CdSO4 and SeO2 or TeO2 . The semicon- from 2.36 to 2.42 eV [51].
ductors grow as compact wires that are The same technique to form CdS
many µm long and 200–300 nm in di- nanowires, only employing dc electrode-
ameter. By electrodepositing first CdSe position, has been used by Xu and asso-
and then CdTe, nanowires were formed, ciates [52]. In this case, the Al was etched
which were compositionally graded over away and a silver film evaporated on the
their length. membrane. The wires in this case were
Chakarvarti and Vetter electrodeposited textured with the (10.1) plane parallel
Se in the pores of nuclear track mem- to the membrane surface [in contrast to
branes [48]. These pores were relatively the ac deposition where the (00.1) plane
large (2.5 µm). grew parallel to the surface]. This tech-
Using porous anodic aluminium ox- nique was extended to CdSe and CdTe
ide films, Routkevitchand associates elec- nanowires [53].
trodeposited very thin CdS nanowires [49]. This group also deposited CdS nano-
The alumina membranes were typically wires in porous alumina templates using
1–3-µm thick with pore sizes ranging from an aqueous solution containing CdCl2 and
9 to 35 nm. The CdS was electrodeposited, thioacetamide at pH = 4.6 [54]. This de-
using ac electrodeposition, from a DMSO position is actually an electrochemically
solution containing CdCl2 and elemental induced chemical solution deposition ini-
S. The porous alumina, as anodized, was tiated by local pH changes caused by
separated from the Al by a dense oxide electrolysis [55]. In this case, the wires
182 3 Semiconductor Nanostructure
grew with the (00.1) plane parallel to the photoelectrochemical response) of TiO2
membrane surface. was dependent on the electrodeposited
metal matrix: deposition of Ni, Cu, Ag, and
3.1.7 In matrices resulted in different disper-
Occlusion Electrodeposition of Composites sion geometries of the TiO2 particles [62].
In the case of Cu, the photoelectrochem-
Apart from the layered Cu/Cu2 O com- ical response at a high pH showed both
posites described earlier, composites of n-type and p-type polarities because of the
nanocrystalline semiconductors with non- formation of p-Cu2 O. The incorporation
semiconductors (metals or polymers) have of alumina into electrodeposited Ni films
been electrodeposited by incorporation was shown to be controlled by diffusion
of the semiconductor phase from so- of the alumina particles to the Ni surface
lution into the electrodepositing metal and their residence time at the surface [63].
(also known as occlusion). WO3 particles CdS, as well as TiO2 particles occluded into
suspended in solution can be incorpo- electrodeposited Ni, were shown to exhibit
rated in electrodeposited polypyrrole [56] increased photoelectrochemical activity af-
or polyaniline [57] films, and the result- ter potential cycling owing to passivation
ing films exhibit electrochromism, both of of the Ni matrix [64].
the polymer and of the WO3 . Although this
technique worked only for oxides with very
3.1.8
low isoelectric points, the isoelectric point Sonoelectrochemical Formation of
of oxides with relatively high points of zero Nanocrystalline Semiconductors
charge could be increased by adsorption of
anions, such as sulfate or iodide, from
Sonoelectrochemical synthesis has re-
solution. In that manner TiO2 was incor-
cently been used for the preparation of
porated into electrodeposited polypyrrole
films, and the resulting films showed a semiconductor nanocrystalline powders.
photoanodic photocurrent response be- In the sonoelectrochemical method, the
cause of the incorporated TiO2 [58]. Very ultrasound horn acts as both cathode and
high concentrations of WO3 – up to 53 ultrasound emitter. This technique was
wt% – were incorporated into electrode- used for preparing metal powders [65]
posited polypyrrole [59]. High stirring and was extended to CdTe, although de-
rates and low current densities of plating tails of the CdTe particle size were not
favored high rates of incorporation. In- given [66]. CdSe nanocrystalline powders
corporation of TiO2 into polypyrrole was have been prepared by pulsed sonoelec-
attributed to mechanical entrapment of the trochemical reduction from an aqueous
TiO2 particles in the rough, soft polypyr- selenosulfate solution. The crystal size
role. In addition, the electric field at the could be varied from X-ray amorphous up
surface of the growing film was also as- to 9 nm (sphalerite phase) by controlling
cribed a role [60]. the various electrodeposition and sonic pa-
In a similar manner, TiO2 particles in rameters [67]. Crystal size was smaller for
suspension were incorporated into grow- lower preparation temperatures, higher ul-
ing electrodeposited Ni films and the trasound intensity, and shorter current
resulting films were photoelectrochem- pulse width. These dependencies could
ically active [61]. The morphology (and be explained based on a pulse of electric
3.1 Preparation of Nanocrystalline Semiconductor Materials 183
31. D. Behar, I. Rubinstein, G. Hodes et al., Su- 52. D. S. Xu, Y. J. Xu, D. P. Chen et al., Chem.
perlattices Microstruct. 1999, 25, 601. Phys. Lett. 2000, 325, 340.
32. Y. Golan, E. Ter-Ovanesyan, Y. Manassen 53. D. S. Xu, D. P. Chen, Y. J. Xu et al., Pure
et al., Surf. Sci. 1996, 350, 277. Appl. Chem. 2000, 72, 127.
33. Y. Golan, G. Hodes, I. Rubinstein, J. Phys. 54. D. S. Xu, Y. J. Xu, D. P. Chen et al., Adv.
Chem. 1996, 100, 2220. Mater. 2000, 12, 520.
34. Y. Zhang, G. Hodes, I. Rubinstein et al., Adv. 55. K. Yamaguchi, T. Yoshida, T. Sugiura et al.,
Mater. 1999, 11, 1437. J. Phys. Chem. B 1998, 102, 9677.
35. J. A. Switzer, M. J. Shane, R. J. Phillips, Sci- 56. H. Yoneyama, Y. Shoji, J. Electrochem. Soc.
ence 1990, 247, 444. 1990, 137, 3826.
36. J. A. Switzer, R. J. Phillips, T. D. Golden, 57. H. Yoneyama, S. Hirao, S. Kuwabata, J. Elec-
Appl. Phys. Lett. 1995, 66, 819. trochem. Soc. 1992, 139, 3141.
37. J. A. Switzer, R. P. Raffaelle, R. J. Phillips 58. K. Kawai, N. Mihara, S. Kuwabata et al., J.
et al., Science 1992, 258, 1918. Electrochem. Soc. 1990, 137, 1793.
38. T. D. Golden, R. P. Raffaelle, J. A. Switzer, 59. F. Beck, M. Dahlhaus, J. Appl. Electrochem.
Appl. Phys. Lett. 1993, 63, 1501. 1993, 23, 781.
39. R. J. Phillips, T. D. Golden, M. G. Shumsky 60. F. Beck, M. Dahlhaus, N. Zahedi, Electro-
et al., Chem. Mat. 1997, 9, 1670. chim. Acta 1992, 37, 1265.
40. J. A. Switzer, C. J. Hung, B. E. Breyfogle 61. M. Zhou, N. R. deTacconi, K. Rajeshwar, J.
et al., Science 1994, 264, 1573. Electroanal. Chem. 1997, 421, 111.
41. R. A. Vanleeuwen, C. J. Hung, D. R. Kamm- 62. N. R. de Tacconi, C. A. Boyles, K. Rajeshwar,
ler et al., J. Phys. Chem. 1995, 99, 15 247. Langmuir 2000, 16, 5665.
42. J. A. Switzer, C. J. Hung, E. W. Bohannan 63. P. M. Vereecken, I. Shao, P. C. Searson, J.
et al., Adv. Mat. 1997, 9, 334. Electrochem. Soc. 2000, 147, 2572.
43. J. A. Switzer, C. J. Hung, L. Y. Huang et al., 64. N. R. de Tacconi, H. Wenren, K. Raje-
J. Mat. Res. 1998, 13, 909. shwar, J. Electrochem. Soc. 1997, 144,
44. E. W. Bohannan, L. Y. Huang, F. S. Miller 3159.
et al., Langmuir 1999, 15, 813. 65. J. Reisse, H. Francois, J. Vandercam-
45. J. A. Switzer, C. J. Hung, L. Y. Huang et al., men et al., Electrochim. Acta 1994, 39,
J. Am. Chem. Soc. 1998, 120, 3530. 37.
46. J. A. Switzer, B. M. Maune, E. R. Raub et al., 66. J. Reisse, T. Caulier, C. Deckerheer et al.,
J. Phys. Chem. B 1999, 103, 395. Ultrason. Sonochem. 1996, 3, S147.
47. J. D. Klein, R. D. Herrick, D. Palmer et al., 67. Y. Mastai, R. Polsky, Y. Koltypin et al., J. Am.
Chem. Mat. 1993, 5, 902. Chem. Soc. 1999, 121, 10 047.
48. S. K. Chakarvarti, J. Vetter, J. Micromech. 68. J. J. Zhu, S. T. Aruna, Y. Koltypin et al.,
Microeng. 1993, 3, 57. Chem. Mat. 2000, 12, 143.
49. D. Routkevitch, T. Bigioni, M. Moskovits 69. H. J. Byker, A. E. Austin, J. Electrochem. Soc.
et al., J. Phys. Chem. 1996, 100, 14 037. 1981, 128, 381 (CA no. 536).
50. J. L. Hutchison, D. Routkevitch, A. Albu- 70. E. A. Ponomarev, M. NeumannSpallart,
Yaron et al., Microscopy of Semiconducting G. Hodes et al., Thin Solid Films 1996, 280,
Materials 1997, 1997, 389. 86.
51. D. Routkevitch, T. L. Haslett, L. Ryan et al., 71. Y. Mastai, M. Homyonfer, A. Gedanken
Chem. Phys. 1996, 210, 343. et al., Adv. Mat. 1999, 11, 1010.
3.2 Macroporous Microstructures Including Silicon 185
reaction dependent on the overlap of the decomposition potential is below the con-
density of states. duction band minimum (Ec < n Ed ) i.e.
When photoelectrochemical solar cells more positive than the conduction band.
became popular in the 1970s, many reports A similar statement can be given for the
appeared concerning the stability, dissolu- holes. The semiconductor electrode is un-
tion, and flat-band potential of semicon- stable against anodic (photo)corrosion if its
ductors in solutions. These papers investi- decomposition potential is higher (more
gated parameters such as the energy level negative) than the valence band maximum
of the band edges, which is critical for the (Ev > p Ed ) of the semiconductor. In this
thermodynamic stability of the semicon- case, the hole arriving at the interface car-
ductor and how to determine the potential ries out an oxidation reaction with the
for the onset of the (photo)electrochemical semiconductor or undergoes fast recombi-
etching [38–40]. The criterion for ther- nation with the electrons. The conditions
modynamic stability of a semiconductor of corrosion in an electrolyte are illustrated
electrode in an electrolyte solution is deter- in Fig. 1.
mined by the position of the Fermi level EF Generally, the semiconductors are pro-
with respect to the decomposition potential tected against cathodic (photo)corrosion in
of the electrode with either the conduction indifferent electrolytes because of the for-
band electrons n Ed or valence band holes mation of a thin metallic film produced
p E . Under illumination, the quasi-Fermi at the first instant of (photo)corrosion by
d
level replaces the Fermi level. The Fermi the (photo)electrons, which for GaAs, for
example, can be written as
level is usually found within the band gap
of the semiconductor and its position is GaAs + 3e− −−−→ Ga0 + As3− (2)
not easily evaluated (especially the quasi-
Fermi level of minority carriers). Therefore The situation is different in the an-
it was found more practical to use the con- odic regime. Gerischer [38] and Heller [42]
duction band minimum (Ec ) and valence pointed out that the majority of semi-
band maximum (Ev ) as criteria for elec- conductors (in the dark for p-type or
trode corrosion. Thus, a semiconductor under illumination for n-type) are unsta-
will be corroded in a certain electrolyte ble against corrosion with valence band
by the conduction band electrons if its (photo)holes, as it is illustrated in Figs. 1(c)
nE n
E d E Ed
Ec Ec Ec Ec
n
nE
d
Ed
p
pE
d
Ed
Ev Ev Ev Ev
pE pE
d d
h+
(a) (b)
and 2. The oxidation reaction induced by which case a rough or porous morphol-
the holes in the valence band is the oxida- ogy is developed at the semiconductor
tion of the semiconductor itself, which for surface. Under high current densities, a
GaAs, for example, can be written as: solid oxide is formed at the semiconduc-
tor surface because of mass transport of
GaAs + 6h+ −−−→ Ga3+ + As3+ (3) a limited supply of reactants or prod-
ucts. The reaction is diffusion-controlled
Depending on the composition of the elec- and a smooth surface is consequently ob-
trolyte, the ions from the semiconductor tained. This is best demonstrated for Si in
are either soluble in the electrolyte or are the presence of a solution of hydrofluoric
insoluble, and a new phase can form, acid (HF) [44, 45]. Current–voltage studies
generally, an oxide. By choosing an ap- (Fig. 3) performed on Si-HF electrolyte
propriate electrolyte, potential bias, and junctions show that the main requirement
anodization time, the corrosion can be for electrochemical porous Si formation
controlled. is that anodic current densities below a
critical value, icrit , must be used. For a low-
3.2.2.2 Electrochemical Reactions doped n-type semiconductor, light must be
As for many other electrochemical reac- supplied.
tions, two kinetic regimes are distinguish- The exact dissolution chemistry of sil-
able in the anodic dissolution reaction [43]: icon is still a matter of debate; however,
(1) reaction kinetics is rate-limiting (ki- ignoring the intermediate steps involved in
netic control) and (2) the rate-limiting step the dissolution mechanism, the overall dis-
is the diffusion of one of the reactants to (or solution reaction for porous Si formation
products from) the electrode surface. The is as follows:
morphology and properties of the semi-
Si + 6HF + (2 − λ)h+ ===⇒ SiF6 2− + H2
conductor surface are totally different in
the two regimes. Generally, if the elec- + 4H+ + λe− , with λ ≤ 1 (4)
trochemical corrosion reaction is carried
out at sufficiently low current densities, Porous Si formation is characterized by
the electrode kinetics is rate-limiting, in the fact that two charges are required to
188 3 Semiconductor Nanostructures
14
Electropolishing
12
10 PS formation
icrit
[mA cm–2 ]
8
I
6
Anodic dissolution
4
0
−1 0 1 2
E
[V/SCE]
as a result of the crystal reactivity at the word hole always refers to a defect electron,
different planes. whereas a pore is understood as an etchpit
whose depth exceeds its width.
3.2.2.4 Terminology Anodization under galvanostatic condi-
Beside the diffusion-rate-limited elec- tions is generally the preferred approach
tropolishing resulting in flat surfaces, two for reproducibility attaining a wide range
types of corrosion with kinetically con- of porosity or thickness, but studies under
trolled etching rate can be distinguished. potentiostatic conditions have been per-
When the corrosion slightly modifies the formed. Porous Si films have also been
surface properties leading to the for- generated by stain-etching without apply-
mation of surface microrelief structures, ing light or electrical bias. Strong oxidation
improvement of the surface chemistry and agents added to the electrolyte are used
optoelectronic properties is observed. This to remove electrons from the conduction
corrosion process is named electrochem- band and lead to chemical-etching by the
ical etching (EC etching) or photoelectro- resulting holes.
chemical etching (PEC etching) when it is
additionally light-controlled. On the other 3.2.3
hand, the corrosion type that transforms Porous Silicon
the semiconductor into a porous material
exhibiting different properties than those 3.2.3.1 Anodic Dissolution
of the bulk material is referred to as porous Several remarkable features characterize
etching or porous photoetching when it is the anodic oxidation of Si in fluoride-
light-assisted. containing solutions. First of all, the
Following the IUPAC guidelines, porous anodization is a multistep process that
media in general should be classified as involves charge-transfer reaction both via
the following: microporous, mesoporous, the valence band (hole capture) and via the
and macroporous, when the dominant conduction band (electron injection) [8,
pore sizes are <2 nm, in the 2–50- 50–58]. The dissolution valence number
nm range and >50 nm, respectively [49]. n = 2. Hydrogen evolution is observed
In the porous Si literature, the term at potentials far more positive than the
microporous has often been too loosely Nernst potential of the H/H2 couple [5, 45,
applied and bears no resemblance to 55, 59]. SiF6 2− is the final dissolution prod-
the criteria accepted by the community uct and it is well established (from Fourier-
studying porous carbons, glasses, and so transform infrared measurements) that
on. Part of the problem originated from during pore formation, the Si surface
the fact that the optoelectronic properties is mainly hydride-terminated (Si−H) [60,
of porous Si are largely determined by 61]. The two models proposed by Lehmann
the skeleton size and not pore size. New and Gösele [62] and Gerischer and cowork-
terminology such as nanoporous Si has ers [57, 58] take into account all the facts
been frequently used to indicate a solid mentioned earlier. Both models include
skeleton with dimensions small enough two electrochemical reactions (hole cap-
to constitute a nanostructure. The term ture and electron injection) followed by a
nanoporous generally comprises those of chemical oxidation reaction. It is generally
microporous and mesoporous Si. To use accepted that the formation or decompo-
consistent terminology in this paper, the sition of a crystal occurs preferentially at
190 3 Semiconductor Nanostructures
weakly coordinated sites, such as kink sites by injection of an electron into the con-
(present in steps on the surface or at the in- duction band. The reaction steps at a
tersection of dislocations), where Si atoms kink site (in which the Si atom has two
have two or three ‘‘dangling’’ bonds ex- bonds linked to the lattice and two other
posed to the solution. Si atoms on kink site bonds saturated by H-ligands) for the dis-
atoms have a binding energy, that is, on an solution in aqueous HF are described in
average, half of that which they would have Fig. 4, at a molecular level with a two-
inside the crystal. In both models it is as- dimensional representation, following the
sumed that a valence band hole is captured Gerischer and coworkers’ model. The elec-
in a Si−H surface species, resulting in an trochemical reaction is initialized by a hole
electron-deficient Si−H bond. The main capture at a Si−H bond, which is thus
difference between the two descriptions of weakened. The H atom is captured by an
the surface chemistry is the position of the F− ion as a proton H+ , leaving its electron
•
positive charge in the electron-deficient at the Si atom (≡Si radical). By interac-
Si−H bond. Lehmann and Gösele as- tion with F− ions, the unpaired electron
sume that the positive charge is located can be excited up to the conduction band
on the silicon side of the bond, resulting edge. The excited unpaired electron is in-
in a hydrogen radical [62]. Gerischer and jected into the conduction band, whereas
coworkers assume that hole-trapping in the electrons of the F− ion form a new
the Si−H bond results in a silicon sur- very stable Si−F bond. The large elec-
face radical species and a proton [57]. In tronegativity difference between Si and F
both cases the hole-capture step is followed is responsible for the polarization of the
A B C
Si Si Si
H H
Si Si H Si OH
H
+h+ + +H2O +H+
(I) Si Si Si Si +H Si Si
H −e− H
Si Si H Si
H H H
C′ A′
Si Si
Si H Si H OH H
F +2H2O
+F− + Si
(II) Si Si Si H
+OH− H F
Si H OH
Si
H H
+ H2O, HF X X
(III) Si + H2 (X = OH, F)
(In solution) X F
Fig. 4 Two-dimensional description of the reaction steps at a kink site on a silicon
surface for the oxidation initiated by a hole (from Ref. 58).
3.2 Macroporous Microstructures Including Silicon 191
two Si−Si back bonds underneath. This intensity, Q is almost 4 but it drops to 2
increases their chemical reactivity and fa- as the photon flux is increased. An expla-
vors the chemical disruption of one back nation for the current-quadrupling effect
bond. The splitting of the back bond may has been sought in terms of a model in-
well occur by HF or by H2 O. Two new volving four consecutive electrochemical
bonds are formed: Si−H at a back Si, and dissolution steps [53, 55, 56, 59]. In the
Si−F at the kink site (or Si−OH with fur- initial step, a photogenerated hole is cap-
ther exchange of the OH ligand by F− ). tured in a surface bond and a surface
The remaining Si−Si back bond is further intermediate is formed. This intermediate
weakened (and even more polarized) and is further oxidized to a Si(IV) species by
will be quickly split by another HF (or three consecutive electrochemical steps,
H2 O) molecule. The Si atom at the kink in which electrons are thermally excited
site leaves the surface with a divalent oxi- to the conduction band. The variation of
dation state as a HSiF3 molecule and the Q with the light intensity is explained by
kink site situation is restituted at the next assuming that electron injection from sur-
atom of the step (situation A& ). The HSiF3 face intermediates and hole capture by
molecule reacts chemically with water to surface intermediates are competing pro-
give H2 gas, as shown in the following cesses in two of the three charge-transfer
reaction: steps.
j max
30
b
20
j psi
10 c
d
0
−2 0 2 4 6 8 10 12
Applied voltage
[V]
[mA cm−2 ]
20
j
15
10
In the dark
5
0
−1.0 −0.5 0 0.5 1.0
E vs SCE
[V]
region because the less rate-limiting re- ranging from nanoporous silicon with
action creates the greatest potential drop pores and crystallites in the nanometer
and therefore dictates the specific disso- scale up to macroporous silicon with pore
lution step. Impedance spectroscopy and and pillar dimensions in the microme-
capacity measurements of the silicon/HF ter scale. Each of these variations has its
system have helped to clarify part of the origin in different physical effects. The
controversy in the potential distribution main pores grow in (100) directions. Most
at the silicon–electrolyte interface [81–84]. pores extend from the surface down to
The applied potential is partitioned across the porous Si/bulk Si interface. Although
the electrochemical double-layer and the side pore formations and branching may
SCR in the silicon. The fraction of the be extensive, pores are never seen to inter-
potential dropped across the SCR is depen- connect. Porous Si is macroscopically and
dent on dopant concentration and etching roughly characterized by its porosity, fvoid ,
conditions although n-type Si represents a which is the fraction of voids in the porous
limiting case wherein the applied poten- layer.
tial is in the major part dropped across the
SCR when no pores are present (flat sur-
3.2.3.5.1 Nanoporous Silicon Including Mi-
face). This means that the potential drop is
cro and Mesoporous Si It consists of a
equal to the difference between the applied
potential, V , and the flat band potential, spongelike randomly distributed array of
Vfb (VSCR = V − Vfb ). For p-type and de- nanometer-sized silicon features. When
generately doped Si, the process is more the size of the features is less than 5 nm,
difficult to analyze quantitatively because quantum size effect caused by the confine-
of the partitioning of the potential and the ment of the excitons is assumed, which
nonuniform current distribution at the in- gives rise to room temperature photolumi-
terface between the Si substrate and the nescence and electroluminescence in the
porous layer. Under actual pore-formation visible range (1.6–1.8 eV). These proper-
conditions for all types of Si samples, one ties have been intensively investigated for
emerging opinion is that both the SCR and applications of light-emitting devices [6,
the Helmholtz regions control the anodic 88], although porous Si is highly resistive
behavior of silicon. The Helmholtz region in air (ρ = 105 –107 # cm) [89].
controls the charge-transfer reaction rate,
whereas the semiconductor SCR controls Microporous Si It originates from low-
the specific morphology [83, 84]. doped p-type silicon (1 # cm) [5–7, 85,
89, 90]. By employing modified transmis-
3.2.3.5 Many Different Morphologies sion electron microscopy (TEM) specimen
The formation of porous Si critically de- preparation such as direct specimen cleav-
pends on the type of conductivity and age, it has been observed that the coral-like
doping level of the Si electrode, HF concen- structure of anodized p-type Si is com-
tration in the electrolyte, nature of solvent posed of highly interconnected micropores
in which HF is diluted, and applied current in the size range ≤2 nm (Fig. 7). High res-
density (or voltage) [5, 7, 10, 11, 71, 85–87]. olution transmission electron microscopy
Porous Si exhibits distinctive appear- (HRTEM) images showed that strongly
ances (Table 1). The size of the structures luminescent samples (>70% porosity) ex-
varies by three orders of magnitude, hibit narrow undulating Si columns with
3.2 Macroporous Microstructures Including Silicon 195
diameter size down to <3 nm (Fig. 8). The As a result, the photoluminescence output
porous material is completely crystalline. rises dramatically and the peak position
This material also presents a very large is blue-shifted toward 780 nm (Fig. 9).
internal surface area (1000 m2 cm−3 ) and Because the removal of Si constitutes
is therefore suitable for catalytic surface an overall decrease in Si skeleton, this
reactions. Freshly prepared microporous experience was considered as an indirect
Si luminesces at room temperature. The evidence that visible room temperature
position and intensity of the maximum photoluminescence could be attributed to
of the luminescence peak depends on the quantum size effect [6].
porosity. It varies from 1000 nm to 880 nm
for 64% and 77% porosity, respectively. Mesoporous Silicon This kind of porous
Porosity of microporous Si, further sub- silicon forms on heavily (degenerated)
jected to a continuous chemical etching doped silicon substrates, either n+ -
in 40% HF, increases from 70 to 80%. type or p + -type (0.01–0.001 # cm), under
300K
Photoluminescence intensity
[arbitrary units]
1hr
2hr
6hr
conditions that would produce microp- in aqueous HF electrolyte [101, 103]. Low-
ores in low and moderately doped Si. and medium-doped p-type (5–0.2 # cm) is
The pore and pillar sizes are in the made macroporous in HF-containing or-
10–50 nm range [5, 11, 85, 86, 89, 91, ganic electrolyte [79, 80, 03]. Macroporous
92] (Fig. 10). The macroscopic surface is Si consists of micrometer-sized pores and
slightly darker-looking in comparison to silicon pillars. The internal surface area is
bulk silicon, and of excellent smoothness. only 1 m2 cm−3 .
The internal surface area is important
(100 m2 cm−3 ). Mesoporous Si (like mi- Macropore on n and n+ -Type Si Un-
croporous Si) can be easily etched in der Illumination Macropore formation
low-concentration alkaline solutions. The on n-type in aqueous HF under il-
mesoporous Si has been exploited for lumination was studied in detail in
the fabrication of silicon on insulator the 1990s [70–72, 99, 100]. Macropore
(SOI) structures [94], using variations in formation is anisotropic and occurs
porosity in buried layers with altered dop- preferentially along the (100), (010), and
ing. It is also used as a sacrificial layer (001) directions. Figure 11 shows that
in microelectronic and thin film micro- anisotropy should result in the formation
crystalline solar cells [95]. Mesoporous Si of square-shaped tubular pores on a (100)-
layers, depending on the porosity and layer oriented substrate and triangular pores on
width, present remarkable antireflective a (111)-oriented substrate [71]. This has in-
properties and as such can be used as an deed been experimentally confirmed and
antireflective coating in crystalline Si so- macropores formed on (100) substrates are
lar cells [96–98]. Mesoporous Si exhibits tubular with a square-sectional form and a
a weaker luminescence than microporous round bottom (Fig. 12a) [70]. In the case of
Si. This result suggests that quantum size (111) substrates, they have an almost trian-
Si particles are present within the meso- gular section (Fig. 12b). In both cases the
porous structure. macropores delineate a regular network.
The pore-to-pore distance equals twice the
3.2.3.5.2 Macroporous Silicon n-Type value of the SCR, which can be adjusted
silicon (low, medium, and high) [67, by doping level and applied voltage. The
69–71, 99, 100] and very low-doped p- pore size can be controlled by doping, HF
type (<5 # cm) are made macroporous concentration, and applied current density.
198 3 Semiconductor Nanostructures
SC lk
R
bu
SCR
bulk
(a) (b)
The active state of the pore tips is due to the tubular and triangular macropores formed
collection efficiency required for the holes on (100)- and (111)-oriented substrates
initiating the dissolution process. The pas- (Fig. 14). The nanoporous Si is charac-
sivation of the pore walls is ascribed to terized through a network of tangled
a depletion of holes in the walls between filaments of nanometer width, some
the pores. This has been confirmed by of them exhibiting quantum size effect
spectroscopy impedance measurements (Fig. 14) [10, 91]. This nanoporous Si that
on macroporous n+ -type Si [104]. appears fibrous was shown to be the
When the silicon–electrolyte interface source of the visible luminescence [92].
is front-side illuminated during the an- This is not the case when the interface
odization, the macropore Si walls and the is illuminated from the backside of the
top of the Si pillars are covered by a silicon substrate leading to macroporous
nanoporous silicon layer (Fig. 13) [70, 91, Si formation only [72]. The possibility of
92, 99, 105, 106]. Cross-sections of elec- controlling pore location with a standard
trochemically detached nanoporous layers photolithography technique and produc-
well illustrate the morphology of the ing (with rear-light-controlled corrosion)
3.2 Macroporous Microstructures Including Silicon 199
(a) (b)
500 nm
500 nm
Fig. 14 TEM micrographs (cross-section) revealing the typical internal structure through
the thickness of the nanoporous Si formed on highly doped n-type Si (Nd = 1018 cm−3 )
with a charge of Q = 10 C cm−2 : (a) (100) oriented; (b) (111) oriented (from Refs. 91).
order of 0.3 µm [93]. The mesoporous Si and pore wall thicknesses are two times
layer can be dissolved in 0.1 M NaOH, smaller (Fig. 15). When porous-etching
whereas the macroporous Si was only is done in an organic electrolyte, the
dissolved in HF (48%): HNO3 (69%) (1 : 9 macropore shape is better defined and the
ratio volume). This permits measuring the diameter is larger (0.3–1 µm) [113].
thickness of each layer by profilometry.
Depending on the current density and Macropore on p-Type Si Macropore for-
amount of charge used during porous- mation in p-type Si was recently evi-
etching, thicknesses up to 5 µm and 4 µm denced [77–80, 101–103, 108]. Macrop-
for nanoporous and macroporous layers orous p-type Si was first observed on
respectively, have been measured [112]. low-doped (>10 # cm) substrates. It
The morphology resembles that shown in can be formed in HF contained aque-
Fig. 13 for macroporous Si obtained under ous and organic solutions (Fig. 16a).
illumination, except that pore diameter Using specific HF/organic electrolytes
(a) (b)
20 nm 35 nm
a b
the donor charges are not shielded and at the etchpit initiation sites or at the bot-
that the donors are expected to have an in- tom of the pores). Electrochemical kinetics
fluence on the etchpit pattern. The first determines how fast holes react with the
layer of surface at sites when they are available.
# dopant
1/3 $
(with a mean distance
a = 1/ 2Nd from the semiconductor Specific surface chemistries can make
surface) contributes to a nonuniform pho- the reactivity for holes strongly differ-
tocurrent, which is high near the dopant ent from one semiconductor–electrolyte
atoms and can be ascribed to a photoin- system to another. The disadvantages of
duced avalanche effect with a multiplica- the most popular models proposed to ex-
tion factor M(Ea ) where Ea is the electric plain porous Si formation are that they
field near the dopant atom [100, 122]. An do not take into account the chemical as-
important consequence of this model is pects that may influence the kinetics at the
that the surface layer becomes dopant-poor Si/electrolyte interface.
after PEC-etching. This assumption was For pores to propagate, the pore walls
supported by the reduced surface recombi- have to be passivated and the pore tips
nation velocity, as indicated by photovoltaic to be active in the dissolution reaction.
measurements, which show substantial Consequently, a surface, which is depleted
gain in the short wavelength part of the of holes, is passivated to porous-etching,
photocurrent spectrum and increased fill which means that anodization is self-
factor, following PEC etching [123]. Once limiting.
the initiation process sets in, other mecha- One must emphasize the distinct role
nisms, such as photocurrent focussing in of the SCR for n- and p-type Si. In the
the tip, may take place. A redistribution case of p-Si, it acts as a barrier for the
holes, and only near the pore tips is the
process starts. Some pores stop growing
field sufficiently large for opening gaps
and terminate, whereas other pores con-
at the top of the barrier. The reaction
tinue to grow with increasing diameters.
may proceed entirely over the valence
Lehmann and Föll [71, 72] showed that
band (n = 2). In the case of n-type Si,
pore initiation at the surface is determined
the SCR rather acts as a sink channeling
by the doping density of the Si substrate.
the holes to the pore tips. The porous Si
From the initial 1010 cm−2 PEC-etched
may be formed on n-type Si under front
pits formed, only one in a hundred or
and back illumination or under strong
one in a thousand survive and become
polarization (avalanche breakdown). On
pore tips. The density of the resulting pore
heavily doped n- and p-type Si, porous
tips is determined by the SCR width and
Si may be obtained by hole-tunneling.
therefore by the doping density.
Eventual presence of intermediate Si
oxide may favor the electron-tunneling
3.2.3.6.2 Pore Propagation process through the conduction band.
Models Based on the Role of the Space In summary, the effects responsible for
Charge Region Two fundamental aspects pore wall passivation (top row) and for
of porous-etching are the electrostatics and passivation breakdown at the pore tip
electrochemical kinetics. Electrostatics is (middle row) as well as the resulting kind of
crucial in determining the sites wherein porous Si structure together with substrate
semiconductor valence band holes are doping type (bottom row) are presented in
available for the dissolution reactions (e.g. Fig. 17 [86].
3.2 Macroporous Microstructures Including Silicon 203
Field effect
(a)
(SCR)
Models Relevant to Micropore Formation field. As a pore grows, the radius of curva-
The mechanism of microporous Si for- ture increases and the local field decreases
mation from p-type is controversial. A and the pore growth decreases. To main-
puzzling question remains as to the un- tain the total current density, new pores
usually high stability of thin fibers or start to grow. As soon as the fiber/particle
nanoparticles of Si to electrolytic attack or, diameter in microporous Si gets thin, its
in other words, what prevents nanometer- resistivity is sharply increased by the Fermi
thick fibers/particles from being dissolved level pinning to surface states. Because the
altogether as the etching front moves mi- electrochemical dissolution is essentially a
crons into the crystal bulk. current-controlled process, further etching
The first model, proposed by Beale and is impossible and a thin fiber/nanoparticle
coworkers [85, 89], assumes that at po- is stable. This model is often disregarded
tentials corresponding to pore growth the because Fermi level pinning by surface
Fermi level of Si is pinned near the middle states must be assumed. Unfortunately
of the band gap due to of a high number electron paramagnetic resonance (EPR)
of surface states. This results in a potential measurements indicate that there is a low
drop at the interface, which restricts the density of surface states in the porous
hole transfer from the substrate. Anodic layer [124, 125]. This has been confirmed
currents may flow as a result of thermionic by Searson and coworkers, using in situ
emission across the potential barrier. The impedance spectroscopy to measure inter-
local current flow is then determined by face states at [(111), Nd = 1013 cm−3 ] Si
the potential height, which in turn is de- surfaces in fluoride solutions [126, 127].
pendent on the local electric field. High They found, at pH between 3 and 6, a very
electric field favors hole transfer. Hence, low density of electrically active surface
the electric field strength being greatest at states equal to 2 × 1010 cm−2 correspond-
the base pore, this results in dissolution ing to about one in every 105 surface states.
only at the pore tips. Therefore, following Nevertheless, the idea of barrier-lowe-
pore initiation, the pores propagate as a re- ring may remain relevant. For example,
sult of the focussing effect of the electrical if part of the applied potential appears in
204 3 Semiconductor Nanostructures
the Helmholtz layer [83, 84], the barrier current [128]. In addition, the ‘‘quantum
will be decreased by the corresponding wire model’’ has nothing to say about the
amount. If the Helmholtz potential drop formation of all other types of pores.
is proportional to the surface electric field,
the barrier lowering will be inversely Models Relevant to Mesopore Formation
proportional to the SCR thickness. The Beale and coworkers’ model has been
A very often cited model is the so-called extended to explain mesopore formation
‘‘quantum wire model’’ [62]. The aim of on degenerately doped p- and n-type Si (in
the model developed by Lehmann and the dark) for which hole-tunneling is the
Gosële was to explain the morphology ob- major charge-transfer mechanism [85, 89].
served for porous Si, which luminesces as The probability of a hole-tunneling across
a result of quantum confinement. As the the Si-electrolyte interface depends on both
dimensions of the particle decrease, the the potential barrier height and SCR width.
band gap increases (from 1.1 to 1.5 eV). The total electric field at the pore tip results
Because of quantum confinement, the in both a lowering of the potential barrier
energy of valence band states is low- and a decrease of the SCR width. Hence,
ered in the walls between the micropores. as mentioned earlier, the propagation of
The holes are transferred to the interface pores is favored. As the pores grow, the
only at the pore bottoms. The quantum depletion layers of adjacent pores may
confinement model may be regarded as overlap. This results in the probability of
chemical passivation of pore wall disso- tunneling becoming negligible at the pore
lution, where the nanocrystalline silicon walls; that is, the reaction is passivated
particles are the passivating species. How- at the pore walls. Overlap of the depletion
ever, this model seems to be contradictory layers prevents branching and explains the
to the well-established properties of iso- observed columnar structure.
type heterojunctions. A barrier height of a Zhang [66] and Searson [9] confirmed
few hundred meV is known to be largely that model and evoked a ‘‘field-enhance-
insufficient for blocking majority carrier ment effect’’ at the tips of the pores
transport in a heterojunction at room tem- wherein the current flow is controlled
perature for both the directions of electric by a tunneling mechanism (Fig. 18). Their
model predicts the formation of highly on n+ -Si as well as on low- and medium-
oriented parallel pores for (100)-oriented doped p-type Si.
substrates and gives a reasonable expla- 1. n-type porous Si under illumination:
nation for the regular pore distribution There is a reasonable agreement on the
and spacing by considering the depletion formation of macroporous Si from n-
layer width. The calculation of the current type Si [71, 72]. The formation mechanism
distribution around the pore front using of n-type macroporous Si is ruled by
a two-dimensional solution to Poisson’s the reversed-biased SCR at the solution
equation shows that the current is consid- interface.
erably greater at the pore tip as compared Front-side illumination: The condition
with the pore wall [9], accounting for the at the pore tip of an illuminated n-type
unidirectional growth of the pores. In the electrode is different from that in the
case of multiple adjacent pores, the equipo- dark, because the presence of a breakdown
tential lines in the region between the pore field is not necessary to generate charge
walls are shifted further away from the carriers. Every depression or pit in the
surface and into the bulk, effectively de- surface of the n-type silicon anode bends
pleting the pore wall regions. The pore the electric field in the SCR in a way
spacing is determined by the distance at that the concave surface regions become
which the region between the pore walls more efficient in collecting holes than the
becomes completely depleted and is com- convex ones. Concave regions are etched
parable to the calculated SCR width at a preferentially and the pores start to grow,
planar surface. consequently enhancing the local current
In 1972, Theunissen [67] observed for- density [68].
mation of wide-etched channels for donor After initiation of microetchpits, macro-
concentrations of less than Nd = 2 × pore formation occurs. The density of the
1017 cm−3 under high voltage (>10 V), in resulting pore tips is determined by the
2.5–5% HF, when anodization was per- SCR width and therefore by the doping
formed in the dark. He proposed that at density. The thickness of the remaining sil-
some pore tips the electrical field strength icon walls between the pores is two times
was sufficient to allow avalanche break- the SCR width and as a result the pore walls
down at the depletion layer, which then are charge carrier–depleted (Fig. 19). This
generated the necessary carriers for fur- finding has been confirmed by impedance
ther pore growth. This was corroborated spectroscopy studies [104].
recently by Lehmann et al. [86] show- Until now, no model has been proposed
ing experimentally and by simulations to explain the simultaneous formation
based on the electrical field distribution of the duplex nanoporous–macroporous
present at the pore tip and pore walls structure, when the n-Si and n+ -type Si
that the mesopore formation is domi- are illuminated from the front side.
nated by charge carrier tunneling, whereas The striking property of the orienta-
avalanche breakdown is responsible for the tion dependence of the nanoporous and
formation of large etchpits. macroporous structures may be correlated
with the observation that the critical cur-
Models Relevant to Macropore Formation rent icrit is greatest for the (100)-oriented
Macropores can be formed under illumi- substrate [72]. One possible explanation
nation on n- and n+ -Si and in the dark considers the process in which one Si−H
206 3 Semiconductor Nanostructures
bond is substituted by a Si−F bond and suitable wavelength from the rear of the
may well be an indication that the surface Si wafer favors the collection of the pho-
chemistry is an extremely sensitive func- togenerated holes in the bulk and their
tion of the silicon crystallographic orienta- migration toward the tip of the initiated
tion. For instance, the unreconstructed Si pits. Such an experiment is only possible
(111) surface is ideally terminated by one for lightly doped n-type Si, because the dif-
H atom per Si, forming a monohydride fusion length of the holes is comparable
(≡SiH) [129], and the Si (100) surface by to the sample thickness. Stable macropore
two H atoms per Si, forming a dihydride growth occurs when the current density
(=SiH2 ) [130]. It has been proposed that is limited by hole generation and not by
because of the presence of the two H atoms the applied bias. In addition, if the local
per Si the dihydride-terminated (100) sur- current density at the pore tip, itip , exceeds
face is sterically hindered [77, 78, 31]. The the critical current density (itip ≥ icrit ), no
steric hindrance causes bond strain, en- nanoporous silicon will form at the pore
hances the chemical reactivity, and the tip because of the presence of the elec-
dissolution will occur faster along the (100) tropolishing regime. For a given initiation
planes [5, 72]. pattern, the pore diameter, dp , and wall
Rear illumination: When anodization is thickness, W , are determined by the ratio
performed under illumination from the between the applied current, i, and the
backside, holes move to the interface by critical current, icrit [Eqs. (8) and (9)].
diffusion until they are captured by the % &
i 0.5
SCR, which accelerates them toward the dp = p (8)
icrit
pore bottoms. Lehmann [72] has clearly
demonstrated the validity of the macro- and
' % &0.5 (
pore propagation model by studying, in i
aqueous HF solution, the formation of w =p 1− (9)
icrit
a regular orthogonal pattern of cylin-
drical macropores on a rear-illuminated The diameter of the macropores etched at
substrate. He initiated the pores by us- different current densities, i, is a linear
ing standard lithography to produce a function of the square root of i/icrit
predetermined homogeneous pattern of following Eq. (8). It was found that the wall
pits and subsequently developed the pores thickness could be up to a factor 10 smaller
by alkaline etching. Illumination with a or wider than the SCR width. The rate of
3.2 Macroporous Microstructures Including Silicon 207
pore growth is a function only of the critical macropore arrays (Fig. 20). It has been
current density icrit (which is a function of shown recently that 100-µm wide pores
the HF concentration and temperature). with 2-µm wall thickness and 200-µm
The stable macropore formation ob- depth (Fig. 21) can be formed on highly re-
tained for an applied bias is sufficient sistive (2000–5000 # cm) n-type Si [132].
to generate the critical current density The demonstration that there is no restric-
(about 1 V). This understanding of the tion concerning the wall spacing opens
formation mechanism has allowed good the route to form vertical structure by
control of the geometrical parameters of (photo)electrochemistry.
Fig. 20 Cross-sectional SEM images and 45◦ array. Pore growth was induced by regular
bevel of n-type Si samples (Nd = 1015 cm−3 ) patterns of etchpits produced by standard
showing the predetermined patterns of lithography and subsequent alkaline etching
macropores. (a) orthogonal array, (b) hexagonal (inset upper right) (from Ref. 72).
A recent systematic study of macropore They suggest that the contrast between
formation performed on various doped n- aqueous and nonaqueous etching can be
type Si substrates with rear illumination, attributed to two factors, the competition
by Föll and coworkers [106] showed that a of OH− with F− for complexing Si, and
strong influence of the SCR on the average the kinetically slow dissolution of oxide
macropore density is indeed observed in (or hydroxide) species formed in aqueous
accordance with the Lehmann model [72] solutions.
(i.e. an increased anodic bias decreases the
density of pores), except for highly doped Computer Simulations These have been
Si. It was observed that an increasing attempted to obtain morphologies similar
anodic bias increases the pore density, in to those observed in porous Si. The models
contrast to the prediction. The pore growth proposed to explain porous Si formation
seems to be dominated by the chemical- are similar in spirit to those previously
transfer rate and most likely calls for a used to understand the complex crystal-
chemical passivation mechanism of the growth phenomena. Although porous Si
macropore walls. formation is a dissolution process, similar-
2. p-type Si: Formation of macroporous ity is found with the growth phenomenon.
Si is surprising at first sight because in The models fall in two categories. The first
contrast to n-type Si, p-type electrodes type corresponds to the popular diffusion-
are under forward bias conditions. An ex- limited aggregation (DLA) model, which is
tended SCR is therefore not expected for a based on the diffusion of an electrostatic
p-type electrode in the anodic regime. To species such as hole (electron) to (from)
explain macropore formation process on the interface [133]. The second type is the
p-type Si, Lehmann and Rönnebeck [102] Mullins-Sekerta instability model [134]. It
put forward the key role of diffusion consists of analyzing the linear instability
across the space charge region. Especially, of the Si/electrolyte interface by taking into
the field enhancement and the associated account both the transport of holes in the
narrowing of the SCR at the pore tips semiconductor and ions in the electrolyte
are assumed to be responsible for the together with the surface tension of Si. The
pore tip dissolution and pore wall pas- DLA model has first been used by Smith
sivation. This model predicts macropore and coworkers to explain pore formation
formation up to Na = 1017 cm−3 doping as a result of the diffusion of holes to the
level, but fails to explain why macroporous interface [135, 136]. Yan and Hu adopted a
formation on moderately doped p-type different approach by modeling the inter-
Si (1–0.2 # cm) is only observed when facial dynamics governing the formation
HF-contained organic protophilic solvents of porous Si using a two-dimensional two-
(DMF and DMSO) are used [80]. component resistor network model [137].
Another model proposed by Kohl The ratio of the resistances of the two
and coworkers [77, 78] is based upon networks is used as a control parameter
the strain – induced preferential etching to simulate the advance of the interface
described earlier. The model accounts for according to stochastic dynamics, which
the formation of macropores and highly involves local current. John and Singh
branched micropores when the silicon developed a diffusion-induced nucleation
is rendered porous in either nonaque- model for the formation of porous Si
ous or aqueous HF solutions, respectively. based on two primary processes [138]. The
3.2 Macroporous Microstructures Including Silicon 209
diffusion of holes from the bulk to the distribution of structure sizes (from mi-
surface is controlled mainly by (1) the SCR croporous to macroporous) as the layer
width (w) and (2) the drift-diffusion length thickened, and the dependence of the
(l) of holes inside the lattice. The theoreti- pore sizes on the resistivity of the starting
cal models permit obtaining the porous Si material.
morphology, which looks very similar to
what is observed in practice but generally 3.2.4
lack physical substantiation and serve as Porous Semiconductors: A Review
an illustrative facility rather than an ana-
3.2.4.1 PEC and EC Etching (Table 2)
lytical tool relevant for both scientific and
The first observation of intentional local-
practical applications.
ized corrosion on semiconductors, other
The concept of the instability model
than Si, was purposely performed to corru-
used to explain porous Si formation was
gate the semiconductor surface, decrease
first introduced by Kang and Jorné [139,
the reflectivity, improve the optoelectronic
140]. The model considers the pore nucle-
properties, and consequently increase the
ation at the Si surface as a mathematical
performance of photoelectrochemical so-
problem of the instability of a planar in-
lar cells.
terface toward small perturbations. The
interface can be destabilized for an opti-
mal deformation wavelength as a result 3.2.4.1.1 Group II–VI Semiconductors
of a competition between the destabilizing Beside the porous Si, II–VI semicon-
effect of hole diffusion and the stabiliz- ductors are those for which the origin
ing ones because of ion diffusion in the of deep microrelief (micro-etchpit) forma-
electrolyte and surface tension of Si. The tion was studied first. Long ago, n-type
optimal wavelength is expected to give an CdS [144], CdSe [145, 146, 47], n- and
order of magnitude of the interpore spac- p-type CdTe [119, 120, 148, 149], n-type
ing. The analytical and numerical stability CdSe0.65 Te0.35 [150], ZnSe [121], and Cd-
HgTe [151] have been found to undergo
analysis of the Si/electrolyte interface for
an extreme surface roughening under EC
PEC-etching of n-type Si was performed
and PEC etching. The II–VI electrodes
by Valance [141, 142]. This model allowed
were exposed in aqua regia to a reverse
expression of the dissolution speed and
bias of over 1 V/SCE accompanied by high-
derivation of the scaling laws for interpore
intensity illumination. The duration of the
spacing as a function of the doping level
PEC-etching generally did not exceed 5 s.
of Si and applied potential. Another im-
The anodic dissolution valency is two,
portant result obtained by Kang and Jorné
which implies the formation of elemental
was the relationship between the intensity
tellurium forming an insulating layer that
of the rear illumination and the pore di-
hampers further transfer of photogener-
ameter: the higher the intensity, the larger ated holes.
the photocurrent per pore and the larger
the pore diameter. Chazalviel and cowork- CdTe + 2h+ −−−→ Cd(aq) 2+ + Te0 (10)
ers recently addressed the linear stability
analysis of the interface to the case of After removal of the insulating (photo)-
p-type Si [13, 143]. It allowed an under- corrosion products obtained by dissolution
standing of the observed changes in the in a suitable medium, the surface contains
210
Tab. 2 PEC-etched semiconductors.
+1 V, light,
147] H2 O (1/4/20) 3–4.5 s
1 M Na2 SO3
0.2 M FeSO4 +
0.1 M H2 SO4
1016 –1017 cm 0.2 µm etchpits
(1 # cm)
CdTe (111) d 1.44 n, 1000 # cm HNO3 , HCl, 1 µm triangular etchpits
[119, 120, H2 O (1/4/20)
148, 149]
1016 –1017 cm−3 Idem 0.15 µm
(5–0.05 # cm)
ZnSe [121] d 2.7 n, 4 × 1017 cm−3 HNO3 , HCl, +1 V, light Etchpits λ laser (nm)
H2 O (1/4/20) 1010 cm−2 457.9
5 × 108 cm−2 472.7
0 488
InP (149) i 1.35 n, 1 M HCl 10 mA cm−2 , Elongated etchpits µm. Groove
1–2 × 1018 cm−3 light, structure along (011) axis
1–120 min.
WSe2 [158] i 1.22 n 0.4–1 M +1 V, light Microetchpits on van der Waals
d 1.37 HClO4 planes on initiated punctures
3.2 Macroporous Microstructures Including Silicon 211
(a) (b)
Fig. 22 Plan view SEM images of photoetched (111) n-type CdSe. Influence of
the doping concentration on the density of the micro-etchpits.
(a) Nd = 1016 cm−3 ; (b) Nd = 1017 cm−3 (from Refs. 145, 146).
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).
Semi- Type Band gap Conduction type Electrolyte Potential (V/SCE) or Morphology
conductor size (eV) current density
(continued overleaf)
3.2 Macroporous Microstructures Including Silicon
213
214
Tab. 3 (continued)
Semi- Type Band gap Conduction type Electrolyte Potential (V/SCE) or Morphology
3 Semiconductor Nanostructures
porous p-type SiC (1.5 # cm) was very different morphology were observed [186].
fragile and looked very dendritic. SiC The 3C, 6H, and 4H n-type porous SiC
structures down to 10 nm could be exhibited identical luminescent emission
seen [186]. Dramatic difference in pore spectra with a broad peak around 2.6 eV
wall size was observed between the (except for an additional broad shoulder
structure of porous n-type materi- at 1.9–2.1 eV for the 3C polytype) demon-
als produced under light (10–30 nm strating a clear impedance of the band gap
pore size, 10–100 nm pore walls) and energy of a particular SiC polytype [182].
in the dark (0.1–1 µm pore walls) The photoluminescence of porous SiC
(Fig. 24) [184]. with different microstructures was found
to become enhanced as the microporous
Pores in 4H-SiC first propagate nearly
SiC changed from fibrous to porous den-
parallel to the substrate surface and gradu-
dritic. A weak blue electroluminescence
ally change direction toward the direction
was also observed on porous n-type 6H-SiC
of the c-axis [195]. Similar anisotropic
porous-etching was found in some 6H- (0.1 # cm) [180, 183].
porous SiC [189]. The electrical properties measured with
Intense blue–green luminescence (two the use of a metal-point probe contact
order of magnitude stronger than that ex- showed for porous SiC obtained in the
hibited by crystalline SiC) were observed dark, current–voltage characteristics qual-
for porous n-type 6H-SiC produced by itatively similar to those of bulk SiC,
porous photoetching in HF-based elec- whereas porous SiC obtained under light
trolyte [176, 177, 181, 182]. The photolumi- was insulating. This was interpreted in
nescence peaked at wavelengths (2.6 eV) terms of Fermi level pinning at the surface
lower than those corresponding to inter- of the 50 nm thick fibers corresponding to
band transition in crystalline SiC, thus the SCR thickness for highly doped SiC.
showing the possibility of a related de- In the case of SiC, the reaction with HF is
fect rather than quantum confinement known to produce a carbon-enriched phase
mechanism of light emission from porous at the surface, which may provoke a high
SiC [181, 182, 186, 190]. Similarity be- density of surface states. A model based
tween photoluminescence properties of n- on the Fermi level pinning was evoked to
and p-type porous SiC layers with highly explain porous SiC formation [182].
218 3 Semiconductor Nanostructures
20 µm
porous morphology resembled that which thickness of 50–100 nm, sufficient to leave
was previously obtained after a long etch- a GaP skeleton completely free of carriers,
ing time and removal of the surface as a consequence of the SCR effect.
layer [220]. The high number of surface Porous-etching of InP done in the dark
defects created by ion implantation pro- under tunneling breakdown conditions
vided a high density of etchpits, and thus a leads to the formation of small pores
high density of primary pores with a wall with a diameter around 50 nm on (111)-
3.2 Macroporous Microstructures Including Silicon 221
60 nm
(a) Vs = 2 V, ta = 3 min
with a bias voltage of 12 V. The hexagonal ratio structures (grooves) of average pe-
hollows exhibited 19 nm average diameter riod of 1 µm are developed (Fig. 31).
with a standard distribution of 200 nm. The reflectance of such highly corru-
Self-organization formation of the hollows gated surfaces drops from 40 to 4% after
in porous GaAs was explained by repulsive photoetching at 20 mA cm−2 current den-
interaction between the hollows [210]. sity during 60 min. in 1.2 M HCl [154].
The first work on porous InP, [(001)- This effect can, in principle, be very useful
oriented, Nd = 1018 cm−3 ], was accom- in the development of new antireflective
plished with anodization in HCl elec- surfaces for solar cells.
trolyte under illumination, for producing Porous photoetching of n-type GaP in
surface relief features in optoelectronic H2 SO4 /H2 O/H2 O2 leads to the formation
of macroporous GaP (random networks
devices [152]. The pore location was con-
of single crystalline GaP with pore sizes
trolled with standard photolithography,
of 150 nm and porosity of 35 to 50% for
and well-defined macroporous structures
porous etched and porous photoetched, re-
with a pore aspect ratio up to 250 were
spectively), which appears to be the most
produced by photoelectrochemistry in an
strongly scattering material in visible light
analogous manner to the macroporous pat- reported to date and localization effects are
terned n-Si. When the initial surface was anticipated [224, 225].
mirror polished, the surface microstruc- 4. Photoluminescence: Porous GaAs
ture varied from 50 nm to several µm obtained in HCl exhibits visible region
depending on the porous-etching duration luminescence (around 540–570 and
and was strongly affected by the crystallog- 850–870 nm) (Fig. 32). The green and
raphy [47]. Using a homogeneous white infrared bands were interpreted by
light beam, Decker and coworkers [153] Schmuki and coworkers as being caused
found that microetchpits develop prefer- by a quantum confinement effect in GaAs
entially along the (011) direction. Upon microcrystallites and nanocrystallites,
longer etching time, high depth/width respectively [201, 207–209]; whereas
120000 80000
(a) Porous GaAS (b)
500× 8× 70000
100000
1 clean 60000
3 GaAs
80000
50000
Intensity
Intensity
[a.u.]
[a.u.]
60000 40000
1
2 2 30000
40000
3 20000
20000
10000
0 0
500 600 700 800 900 500 600 700 800 900
Wavelength Wavelength
[nm] [nm]
Fig. 32 Photoluminescence spectra (room −0.5 to 6 V (SCE); (b) highest ‘‘green’’
temperature) of porous GaAs obtained in 0.1 M photoluminescence intensity observed from a
HCl: (a) for three different surface locations on specific location on a porous GaAs sample
porous GaAs formed by a potential sweep from formed by a potential step (8 V/SCE, 5 min).
Finnie and coworkers identified the Contradictory results have been pub-
source of the green luminescence to be lished on the photoluminescence of
arsenic oxide microcrystals formed during porous InP. Some authors reported blue-
porous-etching [211, 212] and that of the shifted photoluminescence peaks lying
infrared band to the scattered excitation above the band gap energy interpreted
radiation, exciting luminescence from the in terms of quantum confinement [230].
relatively unperturbed outer regions of the Hamamatsu and coworkers reported that
etchpit [211]. contrary to expectation of blue-shifted
Photoluminescence spectra of porous emission caused by quantum confinement
GaP obtained in HF solution exhibits three at pore walls, porous (001) InP samples
strong sharp emission lines in the orange exhibited an intense red-shifted photolu-
spectral region (2.1–2.3 eV, 580 nm) and minescence peak [231]. This was explained
a weak emission between 300 and 700 nm by the formation of a set of well-defined
with maxima in the blue (2.4 eV) and new surface state levels on anodized pore
UV (3.2 eV) spectral regions [216]. The wall surfaces.
sharp lines are characteristic of bulk GaP. 5. GaN: This semiconductor was made
The blue and UV emission has been porous by continuous cycling of potential
explained by Anedda and coworkers [216] between −1 and 2 V in 0.5 M H2 SO4 +
on the basis of charge-carrier confinement 10% ethanol [237, 238].
in crystalline quantum wires of about
25-nm diameter. However, similar results 3.2.4.2.3 Porous II–VI Semiconductors:
observed by Kelly and coworkers when Low-doped (1015 cm−3 ) p-type CdTe [239],
porous GaP is obtained in H2 SO4 solution Cd0.95 Zn0.05 Te [239, 240], and ZnTe [241–
are interpreted, not in terms of quantum 243] were made porous by anodization in
confinement effect, but as caused by the dark.
gallium oxide (Ga2 O3 or GaOx ) formed p-Type CdTe and Cd0.95 Zn0.05 Te: Por-
on the surface of porous GaP [217]. ous-etching was performed in 0.5 M
3.2 Macroporous Microstructures Including Silicon 225
H2 SO4 at +1.2 V (3 mA cm−2 ) for several and p-Cd0.95 Zn0.05 Te, which implies that
hours. A 100 µm thick layer consisting no blocking insulating layer is formed.
of highly porous network (0.5–0.8 µm Indeed when anodic dissolution is slow,
macropore diameter) was observed be- further oxidation of Te to HTeO2(aq) + is
neath a polished surface, which was left in- not rate-limiting [Eq. (12)].
tact [Fig. 33]. The reaction mechanism for p-Type ZnTe: Porous-etching of p-
the anodic dissolution of Cd0.95 Zn0.05 Te type polycrystalline ZnTe has been un-
is expected to be similar to that of n- dertaken to texturize the surface for
CdTe [244, 245]. The reaction proceeds in photovoltaic applications [241, 243]. The
two steps, the formation of elemental tel- goal was to increase the surface of
lurium followed by its oxidation: ZnTe while decreasing the reflectivity.
Low-doped (Na = 1014 − 1015 cm−3 ) large
CdTe + 2h+ −−−→ Cd(aq) 2+ + Te0 (11) grains polycrystalline ZnTe is rendered
porous in acidic solution composed of
Te0 + 4h+ + 2H2 O −−−→
HNO3 : HCl : H2 O (1 : 4 : 20) either poten-
HTeO2(aq) + + 3H(aq) + (12) tiostatically at 1 V vs SCE or galvanostat-
ically at 300 mA cm−2 constant current
Because elemental Te0 is an intermediate density. After a few seconds and subse-
species in the dissolution process, the sur- quent dissolution of the corrosion layer, a
face of the porous layer is at least partially layer of macroporous ZnTe with a colum-
covered with Te0 . This was confirmed by nar structure is formed. Diameter of the
XPS, a 14 nm thin Te porous layer lining pores (3–5 µm deep) varies from 0.1 to
the pore walls [239]. The results obtained 1 µm upon increase of porous-etching
under PEC etching with the formation of duration from 5 to 25 secs. The pore
an insulating Te0 layer are not necessary in wall thickness remains constant around
conflict with the obtaining of porous CdTe 100 nm. The effect of grain orientation
226 3 Semiconductor Nanostructures
results in different pore shapes, for ex- The TiO2 crystals or sintered pellets were
ample, tubular or flat isosceles triangles first treated in hydrogen at 650 ◦ C for
or inverted pyramids with a rectangular 30 min. to obtain n-type semiconductiv-
base [243]. The grain boundaries remain ity. Porous photoetching of 0.2–2 # cm
largely undissolved. Enhanced photore- (100) rutile type TiO2 done potentio-
sponse (30%) of a ZnTe/aqueous poly- statically (1–1.5 V vs SCE) renders the
sulfide photoelectrochemical cell is ob- surface velvety black due to light con-
served for the porous surface because finement arising from the presence of
of decrease in reflectivity [242]. Further square shaped macropores (0.3–1-µm di-
porous-etching results in the formation ameter), whose morphology is similar to
of a needle-like structure with a triangular that of highly doped porous Si obtained
cross-section, the smallest structure reach- by porous photoetching (Fig. 12a) [246,
ing 25 nm typical size. Such porous layer 247]. Oxygen gas is evolved at the TiO2
exhibits blue-shifted luminescence com- surface. Increasing the SO4 2− concen-
pared to bulk ZnTe, which is attributed to tration and pH of the solution (from
quantum-confinement effect [241]. 0.45 to 1.65) results in the formation of
spare macropores on a much finer scale
3.2.4.2.4 Porous Oxides Among the ox- (<0.1 µm). In concentrated 9 M H2 SO4 ,
ides, only titanium dioxide, TiO2 , has been corrosion of TiO2 is inhibited. Addition
made porous. TiO2 is one of the few semi- of Co2+ in 0.5 M H2 SO4 quenches the
conducting materials highly stable against porous-etching, the photogenerated holes
photocorrosion, with the conduction band- oxidizing preferentially SO4 2− ions to
edge located so that photogenerated holes S2 O8 2− ions, whereas addition of Co2+
have sufficient energy to oxidize water in 1 M HClO4 does not prevent porous-
or the reduced member of redox species etching [248].
involved. It plays a leading role in the Porous photoetching was also done on
photoenergy conversion and acts as an effi- sintered pellets of TiO2 . The influence of
cient photocatalyst to decompose harmful applied potential on corrosion mechanism
organic compounds as in the TiO2 dye- is illustrated in Fig. 34. When anodization
sensitized photoelectrochemical cells. A is done at 0.2 V vs SCE, the grains are
highly developed surface morphology of selectively dissolved, whereas the grain
TiO2 is a crucial factor in these applica- boundaries are not dissolved creating a
tions and porous-etching is one technique ‘‘skeleton structure’’. By contrast under
that allows surface enlargement. Porous 1 V bias, the grain boundaries of the
photoetching of n-type TiO2 , under xenon pellet are selectively dissolved and a
lamp illumination, occurs in 0.05–0.5 M characteristic porous pattern appears on
H2 SO4 and 1 M HClO4 (perchloric acid). each grain surface [249].
Porous photoetching reaction was as-
sumed to be a two-hole process following 3.2.5
reaction (13) in H2 SO4 [246]: Photoelectrochemical Properties of Porous
Semiconductors
TiO2 + SO4 2− + 2h+ −−−→
Porous-etching has been applied in the de-
TiO·SO4 + 12 O2 (13) sign of photoelectrochemical devices. The
3.2 Macroporous Microstructures Including Silicon 227
1 µm
(b)
1 µm
porosity and pore wall size of porous semi- intensity close to 40% is also observed. In
conductors have different effects on the addition, the external quantum efficiency
photocurrent depending on the band gap vs light wavelength of the porous elec-
type (indirect or direct) of the crystalline trode is greatly enhanced compared to the
material. flat electrode, without any shift in the on-
Photoelectrochemical cells with indirect set of the photocurrent [27]. Similar trends
band gap porous semiconductors such as are observed for n-type porous GaP and
n-type Si (obtained under illumination), porous SiC [215, 250]. Photocurrent in-
SiC, and GaP (made porous in the reverse creases by a factor 100 after porous-etching
bias range) show spectacular results. Pres- of GaP, whereas the internal quantum ef-
ence of a porous layer on lightly and highly ficiency (corrected by the photon flux) of
doped Si strongly modifies the photocur- a 30 µm thick porous layer electrode is
rent potential curve of porous Si/HI (57%) spectacularly enhanced with respect to a
junctions, whose rising part is shifted flat electrode for light absorbed in the in-
to more negative potentials by −0.25 V direct band gap and for sub–band gap
compared to a flat (nonporous) electrode light [214, 219] (Fig. 36). Porous n-type
(Fig. 35). In the case of lightly doped Si, 6H-SiC electrode exhibits a similar be-
an increase in the plateau photocurrent havior. The enhancement of the quantum
228 3 Semiconductor Nanostructures
−2
−1.0 −0.5 0.0 0.5 1.0
E vs SCE
[V]
1.0
II
0.8
0.6
jph/ef
I
0.4
0.2
0.0
2.0 2.5 3.0 3.5 4.0
hν /eV
Fig. 36 Quantum efficiency (jph /eφ) at 1 V (SCE) as function
of photon energy for a polished n-type (100) GaP electrode
(0 C cm−2 ) and for the same electrode after porous-etching
at +10 V (SCE) in the dark for the time needed to pass
16 C cm−2 . The indirect optical transition is at 2.24 eV and
the direct transition at 2.76 eV (from Ref. 218).
efficiency can be attributed to two ef- the band gap may be very large. Such states
fects related to the nature of the porous can give rise to sub–band gap photocur-
layer. First, light scattering gives rise to a rent as it has been observed for porous
more effective absorption of photons in the GaP and SiC. In addition, charge carriers
mesoporous layer. Second, the size of the in bulk band gap states, which normally
pore walls corresponds to the SCR width do not contribute, can also give rise to
in the photocurrent range, where the band photocurrent in a porous system.
bending is sufficient for effective compe- For materials with a direct optical tran-
tition of charge separation with surface sition (short absorption length) and a
recombination. Because of the large inter- low dopant density (wide SCR) or large
nal surface area of porous semiconductors, diffusion for minority carriers, porous-
the volume density of the interface states in etching is less interesting. This was
3.2 Macroporous Microstructures Including Silicon 229
28. C. Lévy-Clément, Semiconductor Micropro- 46. (a) R. Memming, G. Schwandt, Surf. Sci.
cessing (Eds.: H. J. Lewerenz, S. A. 1966, 4, 109; (b) 1966, 5, 97–110.
Campbell), Wiley and Sons, Chapter 5, 47. P. A. Kohl, C. Wolowodiuk, F. W. Oster-
1998, pp. 177–215. mayer Jr., J. Electrochem. Soc. 1983, 130,
29. C. Lévy-Clément, S. Bastide, Zeitschrift für 2288–2293.
Physikalische Chemie 1999, 212, 123–143. 48. N. J. Hovel, Semiconductors and Semimetals,
30. O. Bisi, Surface Science Reports 1999, 38, Vol. 11, Solar Cells (Eds.: R. K. Willardson,
1–126. A. C. Beer), Academic Press, New York,
31. Optical properties of Low dimensional Sili- 1975, pp. 225.
con Structures, (Eds.: D. C. Benshael, L. T. 49. J. Rouquerol, D. Avnir, C. W. Fairbridge
Canham, S. Ossicini) NATO ASI Series, et al., Pure Appl. Chem. 1994, 66, 1739.
Vol. 244, Kluwer Academic Publishers, Dor- 50. M. Matsumura, S. R. Morrison, J. Elec-
drecht, 1993. troanal. Chem. 1983, 144, 113–120; (b) Ibid
32. Porous Silicon Science and Technology (Eds.: 1983, 147, 157–166.
J. C. Vial, J. Derrien) Les Editions de 51. H. Gerischer, M. Lübke, Ber. Bunsenges.
Physique, Paris, 1995. Phys. Chem. 1987, 91, 394–398.
33. Porous Silicon (Eds.: Z. Chuan Feng, 52. H. Gerischer, M. Lübke, J. Electrochim. Soc.
R. Tsu), World Scientific, New York, 1995. 1988, 135, 2782.
34. Properties of Porous Silicon (Ed.: L. T. Can- 53. H. J. Lewerenz, J. Stumper, L. M. Peter,
ham) IEE INSPEEC, The Institution of Phys. Rev. Lett. 1988, 61, 1989–1992.
Electrical Engineers, London, 1997. 54. J. Stumper, H. J. Lewerenz, C. Pettenkofer,
35. Structural and Optical Properties of Porous Phys. Rev. B. 1990, 41, 1592.
Silicon Nanostructures, (Eds.: G. Amato, 55. L. M. Peter, A. M. Borazio, H. J. Lewerenz,
C. Delerue H. J. von Bardeleben), Gordon J. Stumper, J. Electroanal Chem. 1990, 290,
and Breach, Amsterdam, 1997. 229–248.
36. Light emission in Silicon: from Physics to 56. J. Stumper, L. M. Peter, J. Electroanal.
Devices. Semiconductors and Semimetals, Chem. 1991, 309, 325.
Vol. 49, (Ed.: D. J. Lockwood), Academic 57. H. Gerischer, P. Allongue, V. Costa Kieling,
Press, London, 1998. Ber. Bunsenges. Phys. Chem. 1993, 97,
37. H. Gerischer, A. Mauerer, W. Mindt, Surf. 753–757.
Sci. 1966, 4, 431–439. 58. P. Allongue, V. Kieling, H. Gerischer, Elec-
38. H. Gerischer, Physical Chemistry: Vol. IXA/ trochim. Acta 1995, 40, 1353–1360.
Electrochemistry (Eds.: H. Eyring, D. Hend- 59. D. J. Blackwood, A. M. Borazio, R. Greef
erson, W. Jost) Academic Press, New York, et al., Electrochem. Acta 1992, 37, 889.
1970, pp. 463–542. 60. L. M. Peter, D. J. Blackwood, S. Pons, Phys.
39. H. Gerischer, J. Electroanal. Chem. 1977, 82, Rev. Lett. 1989, 62, 308–311.
133–143. 61. A. Venkateswara Rao, F. Ozanam, J.-N.
40. A. J. Bard, M. S. Wrighton, J. Electrochem. Chazalviel, J. Electrochem. Soc. 1991, 138,
Soc. 1977, 124, 1706–1710. 153–159.
41. R. Tenne, in Semiconductor Micromachin- 62. V. Lehmann, U. Gosele, Appl. Phys. Lett.
ing, Vol. 1: Fundamental Electrochem- 1991, 58, 856.
istry and Physics (Eds.: S. A. Campbell, 63. S. M. Sze, Physics of Semiconductor Devices,
H. J. Lewerenz), J. Wiley and Sons Ltd, New 2nd ed., Wiley and Sons, New York, 1981,
York, 1998, pp. 139–175. pp. 804.
42. A. Heller, B. Miller, F. A. Thiel, Appl. Phys. 64. R. Memming, Comprehensive Treatise of
Lett. 1981, 38, 282. Electrochemistry (Eds.: B. E. Conway, J.
43. A. J. Bard, R. L. Faulkner, Electrochemical O’M. Bockris, E. Yeager et al.,) Plenum,
Methods, Fundamendals and Applications, New York, 1983, Vol. 7, p. 529.
Wiley and Sons, New York, 1980. 65. H. Gerischer, Electrode Process in Solid State
44. M. J. Eddowes, J. Electroanal. Chem. 1990, Ionics (Eds.: H. Kleitz, J. Dupuy), D. Reidel:
280, 297–311. Dordrecht, The Netherlands, 1976, p. 299.
45. M. J. Eddowes, J. Electrochem. Soc. 1990, 66. X. G. Zhang, J. Electrochem. Soc. 1991, 138,
137, 3514–. 3750–3756.
3.2 Macroporous Microstructures Including Silicon 233
117. P. Allongue, M. C. Bernard, J. E. Peou et al., 141. A. Valence, Phys. Rev. B 1995, 52,
Thin Solid Films 1997, 297, 1–4. 8323–8336.
118. J. E. Peou, C. Henry de Villeneuve, 142. A. Valence, Phys. Rev. B 1997, 55,
P. Allongue et al., Pits and Pores: Formation, 9706–9715.
Properties and Significance for Advanced Lu- 143. R. B. Wehrspohn, F. Ozanam, J.-N. Chazal-
minescent Materials, Corrosion Luminescence viel, J. Electrochem. Soc. 1999, 146,
and Display Materials, (Eds.: D. J. Lockwood 3309–3314.
et al.), The Electrochemical Society Pro- 144. R. Tenne, V. Marcu, Y. Prior, Appl. Phys.
ceedings Series, Pennington, NJ, 1997, Lett. 1985, A37, 205–209.
Vol. 97-7, 83–91. 145. R. Tenne, H. Mariette, C. Lévy-Clément
119. R. Tenne, G. Hodes, Appl. Phys. Lett. 1980, et al., Phys. Rev. B 1987, 15, 1204–1207.
37, 428–430. 146. R. Tenne, C. Lévy-Clément, H. Mariette
120. R. Tenne, G. Hodes, Surf. Sci. 1983, 135, et al., J. Cryst. Growth 1987, 86, 826–833.
453. 147. H. Homyonfer, R. Tenne, W. Giriat et al.,
121. R. Tenne, H. Flaisher, R. Triboulet, Phys. Phys. Rev. B 1993, 47, 1244.
Rev. B 1984, 29, 5799. 148. R. Tenne, Appl. Phys. Lett. 1983, 43,
122. V. M. Dubin, Surf. Sci. 1992, 274,82. 201–203.
123. C. Lévy-Clément, A. Lagoubi, M. Neumann- 149. R. Tenne, V. Marcu, N. Yellin, Appl. Phys.
Spallart et al., J. Electrochem. Soc. 1991, 138, Lett. 1984, 45, 1219–1221.
L69–70. 150. C. Lévy-Clément, J. Rioux, S. Litch et al., J.
124. S. M. Prokes, W. E. Carlos, V. M. Bermu- Appl. Phys. 1985, 58, 4703–4708.
dez, Appl. Phys. Lett. 1992, 61, 1447. 151. C. Lévy-Clément, R. Triboulet, R. Tenne,
125. H. J. von Bardeleben, C. Ortega, A. Gros- Sol. Energy Mater. 1988, 17, 201–206.
mann et al., J. Lumin. 1993, 57, 301. 152. P. A. Kohl, F. W. Ostermayer Jr., Ann. Rev.
126. G. Oskam, P. M. Hoffman, J. C. Schmidt Mater. Sci. 1989, 19, 379–399.
et al., J. Phys. Chem. 1996, 100, 1801–1806. 153. F. Decker, D. A. Soltz, L. Cestato, Elec-
127. P. M. Hoffman, I. E. Vermeir, P. Searson, trochim. Acta 1993, 38, 95–99.
J. Electrochem. Soc. 2000, 147, 2999–3002. 154. N. G. Ferreira, D. A. Soltz, F. Decker et al.,
128. S. M. Sze, Physics of Semiconductor Devices, J. Electrochem. Soc. 1995, 142, 1348.
2nd ed., Wiley and Sons, New York, 1981, 155. R. Tenne, B. Theys, J. Rioux et al., J. Appl.
p. 122. Phys. 1985, 57, 141–145.
129. V. A. Burrows, Y. J. Chabal, G. S. Higasi 156. R. Tenne, A. Wold, Appl. Phys. Lett. 1985,
et al., Appl. Phys. Lett. 1988, 53, 998. 47, 707–709.
130. Y. J. Chabal, G. S. Higasi, K. Raghavachari 157. D. Mahalu, A. Wold, R. Tenne et al., J. Phys.
et al., J. Vac. Sci. Technol., A 1988, 7, 2104. Chem. 1990, 94, 8012.
131. J. J. Boland, Surf. Sci. 1992, 261, 17. 158. C. Lévy-Clément, R. Tenne, Photoelectro-
132. P. Kleimann, J. Linnros, S. Petersson, Mat. chemistry and Photovoltaics of Layered
Sci. Eng. 2000, B69–70, 29–33. Semiconductors, Chapter IV, (Ed.: A. Aru-
133. T. A. Witten, L. M. Sander, Phys. Rev. Lett. chamy), Kluwer Academic Publishers, Nor-
1981, 47, 1400–1403. well, MA, 1991, pp. 155–194.
134. W. W. Mullins, R. F. Sekerta, J. Appl. Phys. 159. D. Mahalu, L. Margulis, A. Wold, R. Tenne
1964, 35, 444. Rapid Comm. Phys. Rev. B 1992, 45, 1943.
135. R. L. Smith, S. D. Collins, Phys. Rev. A 1989, 160. F. Ozanam, R. B. Wehrspohn, J. N. Chazal-
39, 5409. viel, I. Solomon, Phys. Stat. Sol. (a) 1998,
136. R. W. Bower, S. D. Collins, Phys. Rev. A. 165, 15.
1991, 43, 3165. 161. K. H. Beckmann, Surf. Sci. 1966, 5,
137. H. Yan, X. Hu, J. Appl. Phys. 1993, 73, 187–196.
4324–4331. 162. M. Sendova-vassileva, N. Tzenov, D. Dim-
138. G. C. John, V. A. Singh, Phys. Rev. B. 1995, ova-malinskova et al., Thin Solid Films 1994,
52, 11125–11131. 255, 282–285.
139. Y. Kang, J. Jorné, J. Electrochem. Soc. 1993, 163. S. Miyazawa, K. Shiba, K. Hirose et al., Thin
140, 2258. Solid Films 1995, 255, 99.
140. Y. Kang, J. Jorné, Appl. Phys. Lett. 1993, 62, 164. S. Bayliss, Q. Zhang, P. Harris, Appl. Surf.
2224–2226. Sci. 1996, 102, 390–394.
3.2 Macroporous Microstructures Including Silicon 235
165. H. C. Choi, J. M. Buriak, Chem. Commun. 191. J. N. Wang, Z. M. Chen, P. W. Woo et al.,
2000, 1669–1670. Appl. Phys. Lett. 1999, 74, 923–925.
166. A. Ksendzov T. George, W. D. Pike et al., 192. J. E. Spanier, I. P. Herman, J. Porous Mater.
Appl. Phys. Lett. 1993, 63, 200–202. 2000, 7, 139–142.
167. S. Gardelis, P. Dawson, J. S. Rimmer et al., 193. J. E. Spanier, I. P. Herman, Phys. Rev. B
Appl. Phys. Lett. 1991, 59, 2118–2120. 2000, 61, 10437–10450.
168. H. Shi, Y. Zeng, Y. Wang et al., Appl. Phys. 194. J. E. Spanier, G. T. Dunne, L. B. Rowland
A 1993, 57, 573–575. et al., Appl. Phys. Lett. 2000, 76, 3879–3881.
169. M. Schoisswohl, J. L. Cantin, M. Chamarro 195. S. Zangooie, H. Arwin, Phys. Status Solidi.
et al., Phys. Rev. B 1995, 52, 11898–11903. (A) 2000, 18, 213–219.
170. K. Kolic, E. Borne, A. Sibai et al., Thin Solid 196. T. Monguchi, K. Ono, Y. Baba et al., J.
Films 1995, 255, 279–281. Electrochem. Soc. 2000, 147, 741–743.
171. M. Schoisswohl, J. L. Cantin, S. Lebib et al., 197. J. P. Krumme, M. E. Straumanis, Trans.
J. Appl. Phys. 1996, 79, 9301–9304. Met. Soc. AIME 1967, 239, 395.
172. D. Buttard, J. L. Cantin, M. Schoisswohl 198. M. M. Faktor, D. G. Fiddyment, M. R. Tay-
et al., Thin Solid Films 1997, 297, 233–236. lor, J. Electrochem. Soc. 1975, 122,
173. B. Ünal, P. Phillips, E. H. C. Parker et al., 1566–1567.
Thin Solid Films 1997, 305, 274–279. 199. D. H. Holt, J. Mater. Sci. 1988, 23, 1131.
174. S. Lebib, J. L. Fave, J. Cernogora et al., J. 200. P. Schmuki, J. Fraser, C. M. Vitrus et al., J.
Lumin. 1999, 80, 153–157. Electrochem. Soc. 1996, 143, 3316–3322.
175. B. Ünal, M. Parkinson, S. C. Bayliss et al., 201. P. Schmuki, D. J. Lockwood, B. F. Mason
J. Porous Mater. 2000, 7, 143–146. et al., Appl. Phys. Lett. 1996, 69, 1620–1622.
176. S. Shor, B. Z. Weiss, A. D. Kurtz et al., 202. M. Hao, C. Shao, T. Soga et al., J. Cryst.
Appl. Phys. Lett. 1993, 62, 2836–2838. Growth 1997, 179, 6661.
177. T. Matsumoto, T. Tamaki, T. Futagi et al., 203. D. N. Goryachev, O. M. Sreseli, Semicon-
Appl. Phys. Lett. 1994, 64, 226–228. ductor 1997, 31, 1192–.
178. J. S. Shor, L. Bernis, A. D. Kurtz et al., J. 204. G. Oskam, A. Natarajan, F. M. Ross et al.,
Appl. Phys. 1994, 76, 4045–4049. Appl. Surf. Sci. 1997, 119, 160–168.
179. A. O. Konstantinov, A. Henry, E. Janzën, 205. F. M. Ross, G. Oskam, P. C. Pearson et al.,
Appl. Phys. Lett. 1994, 65, 2699–2701. Phil. Mag. A 1997, 75, 525–539.
180. H. Mimura, T. Matsumoto, Y. Kanemitsu, 206. I. M. Tiginyaru, G. Irmer, J. Monecke et al.,
Appl. Phys. Lett. 1994, 65, 3350–3352. Semicond. Sci. Technol. 1997, 12, 491–493.
181. V. Petrova-Koch, O. Sreseli, G. Polisski 207. P. Schmuki, L. E. Erickson, D. J. Lockwood
et al., Thin Solid Films 1995, 255, 107–110. et al., Appl. Phys. Lett. 1998, 72, 1039–1041.
182. A. O. Konstantinov, A. Henry, C. I. Harris 208. P. Schmuki, L. E. Erickson, D. J. Lockwood
et al., Appl. Phys. Lett. 1995, 66, 2250–2252. et al., J. Electrochem. Soc. 1999, 146,
183. H. Mimura, T. Matsumoto, Y. Kanemitsu, 735–740.
Solid-State Electron. 1996, 40, 501–504. 209. D. J. Lockwood, P. Schmuki, H. J. Labbé
184. M. F. MacMillan, J. Appl. Phys. 1996, 80, et al., Physica E 1999, 4, 102–110.
2412–2419. 210. Y. Morishita, S. Kawai, J. Sunagawa, Jpn. J.
185. V. P. Parkhutik, F. Namavar, E. Andrade, Appl. Phys. 1999, 38, L1156–L1158.
Thin Solid Films 1997, 297, 229–232. 211. C. M. Finnie, X. Li, P. W. Bohn, J. Appl.
186. O. Jessensky, F. Müller, V. Gösele, Thin Phys. 1999, 86, 4997–5003.
Solid Films 1997, 297, 224–228. 212. X. Li, P. W. Bohn, J. Electrochem. Soc. 2000,
187. H. W. Shim, K. C. Kim, Y. H. Seo et al., 147, 1740–1746.
Appl. Phys. Lett. 1997, 70, 1757–1759. 213. B. D. Chase, D. B. Holt, J. Electrochem. Soc.
188. V. P. Parkhutik, J. Appl. Phys. 1998, 83, 1972, 119, 314–317.
4647–4651. 214. B. H. Erné, D. Vanmaekelbergh, J. J. Kelly,
189. A. M. Danishevskii, V. B. Shuman, A. A. Adv. Mater. 1995, 7, 739–742.
Sitnikova et al., Semicond. Sci. Technol. 1998, 215. D. Vanmaekelbergh et al., Electrochim. Acta
13, 1111–1116. 1995, 10, 689.
190. W. Shin, T. Hikosaka, H. S. Ahn et al., J. 216. A. Anedda, A. Serpi, V. A. Karavanski et al.,
Electrochem. Soc. 1998, 145, 2456–2460. Appl. Phys. Lett. 1995, 67, 3316–3318.
236 3 Semiconductor Nanostructures
217. A. Meijerink, A. A. Bol, J. J. Kelly, Appl. 240. B. H. Erné, A. Million, J. Vigneron et al.,
Phys. Lett. 1996, 69, 2807–2803. Electrochem. Solid Stat. Lett. 1999, 2,
218. B. H. Erné, D. Vanmaekelbergh, J. J. Kelly, 619–621.
J. Electrochem. Soc. 1996, 143, 305–314. 241. F. Zenia, R. Triboulet, C. Lévy-Clément
219. F. Iranzo Marin, M. A. Hamstra, D. Vanm- et al., Appl. Phys. Lett. 1999, 75, 531–533.
aekelbergh, J. Electrochem. Soc. 1996, 143, 242. F. Zenia, C. Lévy-Clément, R. Triboulet
1137–1142. et al., Electrochem. and Solid Stat. Lett. 2000,
220. I. M. Tiginyanu, C. Schwab, J.-J. Grob et al., 3, 73–76.
Appl. Phys. Lett. 1997, 71, 3829–3831. 243. F. Zenia, C. Lévy-Clément, R. Triboulet
221. I. M. Tiginyanu, G. Irmer, J. Monecke et al., et al., Thin Solid Films 2000, 361–362,
Phys. Rev. B 1997, 55, 6739–6742. 49–52.
222. K. Kuriyama, K. Ushiyama, K. Ohbora et al., 244. D. Lincot, J. Vedel, J. Cryst. Growth 1985, 72,
Phys. Rev. B 1998, 58, 1103–1105. 426.
223. A. Sarua, I. M. Tiginyanu, V. V. Ursaki 245. D. Lincot, J. Vedel, J. Electroanal. Chem.
et al., Solid. Stat. Comm. 1999, 112, 1987, 220, 179.
581–585. 246. L. A. Harris, R. H. Wilson, J. Electrochem.
224. F. J. P. Schhurmans, D. Vanmaekelbergh, Soc. 1976, 123, 1010–1015.
J. van de Lagemaat et al., Science 1999, 284, 247. Y. Nakato, H. Akanuma, J. I. Shimizu et al.,
141–143. J. Electroanal. Chem. 1995, 396, 35–39.
225. F. J. P. Schhurmans, M. Megens, D. Vanm- 248. L. A. Harris, D. R. Cross, M. E. Gerstner, J.
aekelbergh et al., Phys. Rev. Lett. 1999, 83, Electrochem. Soc. 1977, 124, 939–844.
2183–2186. 249. T. Sugiura, T. Yoshida, H. Minoura, Elec-
226. W. H. Lubberhuizen, D. Vanmaekelberg, trochem. Solid State Lett. 1998, 1, 175–177.
250. J. van de Lagemaat, M. Plakman, D. Vanm-
E. van Faassen, J. Porous Mater. 2000, 7,
aeckelbergh et al., Appl. Phys. Lett. 1996, 69,
147–152.
2246–2248.
227. I. M. Tiginyanu, I. V. Kravetsky, J. Monecke
251. (Ed.: P. M. Fauchet) IEEE J. Select. Topics
et al., Appl. Phys. Lett. 2000, 77, 2415–2417.
Quantum Electron. 1998, 4, 1020.
228. T. Takizawa, S. Arai, M. Nakahara, Jpn. J.
252. B. Lendt et al. Analyt. Chem. 1997, 69, 2877.
Appl. Phys. 1994, 33, L643–L645.
253. G. Lerondel et al. J. Imaging Sci. Technol.
229. E. Kikuno, M. Amiotti, T. Takizawa et al.,
1997, 41, 468.
Jpn. J. Appl. Phys. 1995, 34, 177–178. 254. H. Shinoda et al. Nature 1999, 400, 853.
230. T. Takizawa, M. Nakahara, E. Kikuno et al., 255. J. H. Song, M. J. Sailor, Comments Inorg.
J. Electron. Mater. 1996, 25, 657–660. Chem. 1999, 21, 69.
231. A. Hamamatsu, C. Kaneshiro, H. Fujikura 256. M. Wainwright et al. Lett. Appl. Microbiol.
et al., J. Electroanal. Chem. 1999, 473, 1999, 29, 224.
223–229. 257. V. G. Zubko et al. Astrophys. J. 1998, 501.
232. H. Fujikura, A. Liu, A. Hamamatsu et al., 258. T. E. Bell et al. J. Micromech. Microeng. 1996,
Jpn. J. Appl. Phys. 2000, 39, 4616–4620. 6, 361.
233. P. Schmuki, L. Santinacci, T. D. Djenizian 259. V. P. Bondarenko et al., IEEE Trans Nucl.
et al., Phys. Status Solidi (A) 2000, 182, Sci. 1997, 44, 1719.
51–61. 260. C. Nam et al. IEEE Microwave Guided Wave
234. A. Sarua, G. Gärtner, G. Irmer et al., Phys. Lett. 1997, 7, 236.
Status Solidi (A) 2000, 182, 207–211. 261. V. Parkhutik, E. Yu. Budnikiv, S. F. Tima-
235. K. Chernoutsan, V. Dneprovskii, O. Shali- shev, Mater. Sci. Eng. B 2000, 67/70, 53–58.
gina et al., Phys. Status. Solidi. (A) 2000, 262. L. T. Canham, Adv. Mater. 1995, 7, 1033.
182, 347–352. 263. N. Koshida, H. Koyama, Appl. Phys. Lett.
236. M. Mynbaeva, D. Tsvetkov, Inst. Phys. Conf. 1992, 60, 347.
Ser. 1997, 155, 365. 264. S. Zangooie, R. Bjorklund, H. Arwin, Sens.
237. J. van de Lagemaat, Thesis, Utrecht, 1998. Actuators B 1997, 43, 168.
238. M. Mynbaeva, A. Titkov, A. Kryganovskii 265. T. Laurell, J. Drott, L. Rosengre et al., Sens.
et al., Appl. Phys. Lett. 2000, 76, 1113–1115. Actuators B 1996, 31, 161.
239. B. H. Erné, C. Mathieu, J. Vigneron et al., J. 266. S. Setzu, G. Lérondel, R. Romestain, J.
Electrochem. Soc. 2000, 147, 3759–3767. Appl. Phys. 1998, 84, 3129.
3.2 Macroporous Microstructures Including Silicon 237
267. Z. Zangooie, R. Janson, H. Arvin, J. Vac. 271. L. Pavesi, P. Dubost, Semicond. Sci. Technol.
Sci. Technol. A 1998, 16, 2901. 1997, 12, 570.
268. S. Fronhoff, M. G. Berger, Adv. Mater. 1994, 272. A. Loni, L. T. Canham, M. G. Ber-
12, 963. ger et al., Thin Solid Films 1996,
269. M. G. Berger, M. Thönissen, R. Arhens- 276, 143.
Fischer et al., Thin Solid Films 1995, 255, 273. V. P. Parkhutik, L. T. Canham, Phys. Status
313. Solidi (A) 2000, 182, 591–598.
270. C. Mazzoleni, L. Pavesi, Appl. Phys. Lett. 274. V. P. Parkhutik, L. T. Canhamll, Phy. Status
1995, 67, 2893. Solidi. (A) 2000, 182, 591–598.
287
led to the discovery of wet PV solar cell. protected by insulation, so that only
This is popularly known as a PEC cell. It one of its surfaces is exposed to the
was believed that a PEC cell might be more redox electrolyte. In a regenerative PEC
economically viable as compared with the solar cell, the metal counter-electrode is
solid-state PV cell. Simplicity in fabricat- expected to perform an electrochemical
ing the cell was the main reason for such reaction that is the opposite of the
an optimistic view. Moreover, as discussed process occurring at the semiconductor
in a later chapter, it has been possible electrode. The counter-electrode should
to make a rechargeable battery with in be electrochemically stable. It would be
situ storage capability by using a PEC desirable that work function of counter-
cell [7–9]. The wet type PEC cell suffers electrode is compatible with that of the
from the instability of semiconductor in Fermi level of the semiconductor. The
aqueous media. It has been realized that matching of work function may not be
in spite of its simplicity and economical logical with metallic electrodes such as
viability, the wet-type PEC cell cannot eas- platinum, but may be important with
ily replace a silicon PV solar cell, unless semiconducting electrode such as SnO2 .
we discover photoelectrochemically stable On illumination of the semiconductor-
semiconductor materials possessing band electrolyte junction with a light having
gap approximately 1.4 eV. Till such time, energy greater than the band gap of
future of (unsensitized) wet-type PEC so- the semiconductor, photogenerated elec-
lar cell appears gloomy. However, research tron/holes are separated in the space
utilizing a standard PEC cell configuration charge region. The photogenerated minor-
is not complex, facilitating testing of the ity carriers (holes for n-type or electron
photoactivity of new semiconductor ma- for p-type semiconductor) arrive at the in-
terials. Current PEC cell research trends terface of the semiconductor–electrolyte.
have also been directed toward sensitized Photogenerated majority carriers (i.e. elec-
semiconductors and applications in the trons for n-type and holes for p-type
field of decontamination of water from semiconductor) accumulate at the back-
pathogenic bacteria [10], detoxification of side of the semiconductor (i.e. the side
water from toxic organic/inorganic ma- that is not illuminated and not in con-
terials, photography [11] or in any other tact with the redox electrolyte). With the
field where instability of the semiconduc- help of a connecting wire, photogener-
tor does not pose a real problem. ated majority carriers are transported to
The present section is devoted to various counter-electrode via a load (Fig. 1). These
aspects that are essential for the develop- carriers at the counter-electrode electro-
ment of a PEC cell. chemically react with the redox electrolyte
(i.e. reduction of redox electrolyte occurs
4.1.2 with n-type or oxidation of redox electrolyte
Description of a PEC Cell with p-type semiconductor).
Conversely, the photogenerated minor-
A PEC cell consists of a photoactive semi- ity carriers generated at the interface of
conductor electrode (either n- or p-type) the semiconductor perform the opposite
and a metal counter-electrode. Both these reaction, which occurred at the metal
electrodes are immersed in a suitable counter-electrode. Thus, photogenerated
redox electrolyte. The semiconductor is carriers are consumed. Because of these
4.1 The Photoelectrochemistry of Semiconductor/Electrolyte Solar Cells 289
Load
A
Light hν
∆G = 0
Photoelectrolytic cells
(Optical energy stored as chemical
energy in endorgic reactions
Photoelectrochemical e.g. H2O H2 + 1/2 O2)
cells
∆G ≠ 0 ∆G > 0
Photoelectrosynthetic cells
(Optical energy used to affect
chemical reactions)
∆G < 0
Photocatalytic cells
(Optical energy provides activation energy
for exorgic reactions
e.g. N2 + 3H2 2NH3)
e e
− −
e e
e Electrochemical
O O photovoltaic cell
R R
∆G = 0
e
+
(a) n -SC soln M p -SC + soln M
e e
− −
e e
O′ Photoelectrolytic
R′ O′ cell
R′ ∆G > 0
O O
R R
e
+ +
e e
− − Photocatalytic cell
O
R ∆G < 0
e O e
O′ R
R′ O′
R′
+ +
(c) n -SC soln M p -SC soln M
+e +e +e +e +e +e
EF
+e +e +e +e +e +e E Fredox
+e +e +e +e +e +e
m
u
cu
Va
O
V +
−Ve EA + +
e +
e+ + +
e e+ +
+e +e +e + + E Fredox
e e +e + + + +
+e +e +e + +
+ ++
e e +e
+e +e + +
(b)
+ wn
e +e + E Fredox
+e +e +e + +
+
e +e +
+e +e +e + e
+
+
e +e + +
+e +e +
(c)
wn
+ +
e + +
+e +e +e + + + E redox
+
Lp +
+e +e +e +e
+
e e
+e + + +e
x
(d) −Lp −Wn O
Fig. 4 Schematic diagrams showing the formation of band bending
at a semiconductor–electrolyte interface. (a) A semiconductor
possessing its charge homogeneously distributed. +/e refers to a
situation in which lattice sites are neutral. A horizontal line of vacuum
scale showing no variation in energy of electrons of semiconductor.
(b) The contact of semiconductor with redox electrolyte disturbs the
neutrality, and nonhomogeneous charge separation occurs. Vacuum
scale takes the same shape as that of semiconductor to maintain its
electron affinity. (c) A hypothetical model showing charges retained in
two parallels planes separated by a distance w. (d) Creation of
diffusion region as a result of accumulation of minority carriers at
plane situated at distance x = −w.
294 4 Solar Energy Conversion without Dye Sensitization
plane at a distance ‘‘w’’ from the interface all charges present within the distance of
(Fig. 4c). We can get a better idea of x = 0 and x = −w are regular and fully
this hypothetical plane by considering ionized. Moreover, within the space charge
various positive charges q1 , q2 , q3 , . . . qn region, there are no free carriers. In other
situated at x1 , x2 , x3 , . . . xn distances from words, entire lattice sites falling within
the interface of the semiconductor. The the space charge region are fully ionized.
total electrical field created between the It is also assumed that electrostatic field
negative charges situated at the interface becomes zero beyond x = −wn .
(i.e. at x = 0) and positive charges situated Should the changes in distribution of
at various distances would be sum of electrons (majority carriers) in n-type be
the product of charges divided by their abrupt at the interface as compared to
respective distances. the bulk or should it change linearly or
That is, total electrical field created exponentially? The most reasonable as-
q1 e − q2 e− qn e − sumption would be that the change in con-
(F) = + + ··· + centration of majority carrier between the
x1 x2 xn
bulk and interface is exponential. This sit-
Q+ Q−
total total uation can be illustrated by an exponential
≈ (7)
w bending in the conduction band (Fig. 4).
where Q+ total is total positive charge of
Similarly, the valence band must also bend
minority carriers, equivalent to the total in a similar fashion to maintain the differ-
charge of majority carriers accumulated ence between the conduction band and va-
at the interface (Q− total ) and ‘‘w’’ is the
lence band, which is equal to the band gap
distance of the hypothetical plane from of the semiconductor. Since difference be-
the interface (x = 0), which contains Q+ total
tween the Fermi level and conduction band
total charge, such that (Q+ −
total Qtotal )/w =
is related to the magnitude of dopant’s con-
F. It is worth noting that ‘‘w’’ is not centration, this difference must be main-
equal to the sum of all distances (i.e. tained even after the formation of band
w ' = x1 + x2 + x3 + · · · + xn ). bending. However, it is difficult to present
This hypothetical visualization of charge it pictorially. Hence, we compromise by
distribution near the interface (i.e. its pres- depicting it as a horizontal broken line.
ence in one plane) reveals the possibility of The position of the conduction band
formation of charges in two parallel planes. on the vacuum scale also reflects the
One plane appears at the interface (which magnitude of electron affinity of the
is populated with the majority carriers at semiconductor. This value is not altered
x = 0) and another at a depth w (which because the semiconductor has come in
is populated with charges equivalent to contact with a redox electrolyte. Assuming
total charges developed at x = −w). The the vacuum level to follow a similar
depth between these two planes, (Fig. 4c), bend as the conduction band solves this
is called the space charge width designated problem (Fig. 4). Considering all these
as wn (for n-type) and wp (for p-type). It factors, the entire energy diagram of the
should be understood that concentration semiconductor after making its contact
of minority carriers at x = −wn would with redox electrolyte may be expressed as
be much greater than the intrinsic con- shown in Fig. 5. The variation in charges
centration of hole at x ≈ −α. In this and potential within the space charge
mathematical model, it is assumed that region are shown in Fig. 6.
4.1 The Photoelectrochemistry of Semiconductor/Electrolyte Solar Cells 295
Vacuum scale
−Ve
Dark Illuminated
E E
−
Ec
E N Light
w nE F
Ec nv
Eredox ∆E N Eredox
EF N
pE F +
Ev
+
Ev Redox Redox
(a) n -type semicond electrolyte n -type semicond electrolyte
E
Ec E Vacuum scale
−
Ec −
EF N
w
Eredox nE F Eredox
Ev
∆E N
EF pE F
Light
Ev
+
Redox Redox
(b) p -type semicond electrolyte p -type semicond electrolyte
Fig. 5 Nature of band bending in the dark and under illumination with
(a) an n-type semiconductor-electrolyte and (b) a p-type
semiconductor–electrolyte interface. Vacuum scale changes its shape to
match the shape of band bending. Variation in Fermi level for minority
carriers is shown by broken line (for illuminated semiconductor).
Space charge
+ layer
Charge
−
(a) Gouy layer
in addition to its bulk region. In the bulk kind to prevent them from recombining.
region, there is no accumulation of charge. Thus, the net result of these illuminations
is that photogenerated minority carriers
4.1.8 accumulate at the interface and photogen-
Effect of Junction Illumination erated majority carriers accumulate at the
bulk of the semiconductor.
We shall now examine the effect of illu-
While photogenerated holes at the
mination of these three regions separately.
interface perform oxidation of the redox
When photon of energy greater than band
electrolyte, photogenerated electrons move
gap (known as band gap light) falls in the
via the load toward the counter-electrode to
space charge region, electrons from the
perform reduction of the redox electrolyte
valence band get excited to the conduction
(Fig. 1). Thus, electrical energy is created
band, leaving behind a positive charge (i.e.
from light energy.
hole) in the valence band. These carriers
experience an electrostatic potential in the 4.1.9
space charge region. The magnitude of the Efficiency of Conversion of Light Energy
electrical field is very high in this region. into Electrical Energy
For example, with a contact potential of
1 V and space charge width of 1000 Å, an After constructing a PEC cell, it is neces-
electrostatic field of around 105 V/cm is sary to find the efficiency of conversion
formed. This field is strong enough to pre- of light energy into electrical energy. For
vent the recombination of photogenerated this purpose, a PEC cell is connected
electron/holes pairs in the space charge with a resistance (as a load), variable
region. As a result, photogenerated holes D.C. power source and an ammeter in
migrate toward the interface and elec- series (Fig. 7a). A current-voltage charac-
trons migrate toward the diffusion region teristic is measured by applying different
(Fig. 5). When illumination is conducted potentials to the PEC cell and measur-
in the diffusion region, electron/hole pairs ing the corresponding current. This is
also may be formed. In this region, there called dark current (Fig. 7b). The PEC cell
is only equilibrium drift current (estab- is then illuminated with a constant source
lished due to flow of minority carrier of light. The current-voltage characteristic
toward the bulk of semiconductor) and is again measured as before. This cur-
no electrostatic field. Therefore, photogen- rent is the photocurrent (Fig. 7b). It is
erated electrons/holes of diffusion region observed that the dark current is constant
experience only the drift current as a driv- in the reverse bias condition. This cur-
ing force to prevent their recombination. rent is known as the saturation current
To maintain the equilibrium condition, (I o ). The magnitude of the saturation cur-
rent (Ipho ) increases when the PEC cell is
electrons are directed to move toward the
bulk and holes toward the interface. Drift illuminated. The difference between the
current being not very strong, the major- two saturation currents is the photocur-
ity photogenerated carriers in this region rent (Iph ) that is produced from the PEC
undergo recombination. Finally, if bulk cell. However, the efficiency of power con-
region is illuminated, photogenerated elec- version by the PEC cell is calculated from
trons/holes would undergo 100% recom- the current-voltage characteristics, plotted
bination, as this region has no force of any in the fourth quadrant. It is observed that
298 4 Solar Energy Conversion without Dye Sensitization
inati
Da
I
Illum
−Ve V max
Io V
V oc
Imax
I Ph P max
I oPh
I sc
−I
(b)
I sc
FF = 1
Vmax P max
Photocurrent
I max
Voc
(c) Photovoltage
the maximum photopotential is obtained from the PEC cell, a square or rectangle
when photocurrent is zero. This poten- with maximum area is drawn (Fig. 7b,c).
tial is known as the open-circuit potential The intercept of the square at the current-
(Voc ). Similarly, a maximum photocurrent voltage curve is the magnitude of maxi-
is observed when photopotential is zero. mum power (Pmax ) that can be drawn from
This photocurrent is known as short-circuit the PEC cell. Similarly, the intercept on the
current (Isc ). One would get zero power potential axis is the magnitude of the max-
by working at these values. To find the imum photopotential (Vmax ) and intercept
maximum power, which can be drawn on current axis is the value of maximum
4.1 The Photoelectrochemistry of Semiconductor/Electrolyte Solar Cells 299
photocurrent (Imax ) that can be drawn space charge width (w). While the mate-
from the PEC cell. These characteristics rial should have large space charge width
are obtained when the PEC cell is con- (≈1/α) to ensure entire absorption of band
nected in series with a load of resistance gap light in the space charge region, the
(R). To determine the maximum power diffusion region should also be large. This
load that can be connected to the PEC cell, is because the magnitude of photocurrent
the current-voltage curve is drawn for var- flowing through the semiconductor de-
ious resistances, and the corresponding pends on the width of the diffusion region.
Pmax is determined. The resistance that It will be seen in the forthcoming section
gives the maximum Pmax is taken as the that concentration of dopant as well as
maximum load resistance (Rmax ) that can be the redox electrolyte forming the junction
connected in series with a PEC to drive the can control the width of the space charge
maximum power. The power efficiency of region. Moreover, the FF should also ap-
the PEC cell is calculated from the ratio of proach to unity. Both the space charge
Pmax to the power input (Pin ) to the cell. width and the FF depend mainly on qual-
The nature of the junction formed between ity of semiconductor-electrolyte junction.
the semiconductor and the electrolyte can However, width of the diffusion region is
also be obtained from the current-voltage the specific property of the material. Fur-
characteristic (Fig. 7b,c). This is obtained thermore, the contribution of redox elec-
from a factor known as fill factor (FF), trolyte in giving a better efficiency of a PEC
which is given by: cell is significant and is later discussed.
Imax Vmax 4.1.11
FF = (8)
Isc Voc Energy Levels in Redox Couples
FF decides how well the curve approxi- 4.1.11.1 How Do We Compare Redox
mates a rectangle. In an ideal condition Potential with Fermi Level of a
FF = 1. Generally, current-voltage curve Semiconductor?
instead of plotting in fourth quadrant is At the gas-solid interface, the vacuum
plotted in the first quadrant (Fig. 7c). Note, level is taken as the reference energy
that the photocurrent, although shown as a level; the energy of an electron in vacuum
positive current in figure, depends whether is taken as zero (i.e. Evaccum = 0). The
an n-type (i.e. negative current) or p-type energy of a bound electron in the solid
(i.e. positive current) semiconductor is uti- is referred as a negative value in this
lized. reference scale. In electrochemistry, the
4.1.10
standard zero level of the energy is the
Factors Considered in Selecting a potential of hydrogen ions (at unit activity),
Semiconductor which is in equilibrium with hydrogen
gas at one atmosphere pressure. The
From the previous discussions, it is clear electrode potential of any other redox
that material for a PEC cell must be able to electrolyte is referenced to this potential.
absorb band gap light in the space charge With this hydrogen scale, the value of
region. In other words, the reciprocal of redox potential can be either positive or
the absorption coefficient (i.e. 1/α) of band negative. Conventionally, positive redox
gap light preferably should be equal to the potentials are shown below zero, whereas
300 4 Solar Energy Conversion without Dye Sensitization
(ev)
Vacuum 0.0
−0.2
−0.2
+0.2
(SCE)
+0.4
Ec −5.1 +0.6
Fe+2/Fe+3
+1.0
−5.5 +1.2
Ev
(a) n -semiconductor Electrolyte
E (ev)
Density of state
+1.0
W
re
d (E
)
+0.5
E ox (Unoccupied level)
E redox 0.0 2λ
)
(E
x
Wo
−1.0
negative redox potentials are shown above semiconductor and back, it would come
the zero (Fig. 8a). back to the same energy level of the
When dealing with semiconductor-elec- electrolyte. Since the charge density on
trolyte junction, we have to consider both ions is different, the extent of hydration
of these energy scales. Therefore, there is differs. Therefore, it has been proposed
a need to find out the relationship between that the energy states of reduced and
them, so that one can draw useful conclu- oxidized species should not be discrete but
sion from the band diagram of the semi- are broadened into thermal distribution
conductor–electrolyte interface (Fig. 8a). of states caused by thermal fluctuation,
Considering the free energy changes interaction with solvents, and interaction
associated with the reduction of silver, with directly bonded ligands. The position
one can derive the relationship between of the localized energy states in a polar
Evacuum and the NHE. The energy involved solvent depends greatly on whether the
with various reactions can be presented by state is occupied or not because of
considering the following four steps [8]: the strong interaction with the solvent.
Ag+ − Fluctuation in the structure of hydrated
(g) + e = Ag(g) −7.64 eV
ions thus extends the energy of the
Ag(g) = Ag(s) −2.60 eV
electronic quantum states in the redox
Ag+ = Ag+ +4.96 eV
(aq) (g) couple around two most probable values.
Ag(s) + H+ (aq) = Ag+ 1
(aq) + 2 H2 +0.80 eV The energy change caused by the variation
H+ + e− = 1
H2 −4.48 eV in solvation shell can be represented in
2
(9) terms of harmonic oscillators. On the basis
In this analysis, ‘‘g’’ indicates the gas of the concepts proposed by Levich [13]
phase, ‘‘s’’ the solid, and ‘‘aq’’ the aqueous and the distribution function proposed by
solution phase. Thus, the relation between Gerischer [14–16] a model for the energy
the two scales, namely, Evacuum and NHE levels in solution is discussed in this
can be represented as: section. In this model, the ground state
vibration level of only the first electronic
o
EF(redox) (eV) = Eredox(NHE) (volt) state is considered relevant to electron
−4.48 (eV) (10) transfer processes at the electrode. The
o
Eredox(NHE) and EF(redox) are the potentials energy change caused by vibrations in
of a redox couple on the electrochemical the solvent shell is represented in terms
NHE scale and vacuum scale, respectively. of harmonic oscillations. The potential
energy is considered to vary around
the ground state to an equal degree.
4.1.11.2 Do the Energy Levels for
Reduction and Oxidation Process Possess
The resulting distribution functions for
the Same Value? oxidized and reduced species are given as:
The energy level of the redox electrolyte
o )2 /4πkT λ}]
[exp{−(E − Eox
shown in Fig. 8(a) is an oversimplified Wox (E) =
model because it indicates that the energy (4πkT λ)1/2
levels for the reduced and oxidized species (11)
of the redox couple, Fe2+/3+ , are the same. o )2 /4πkT λ}]
[exp{−(E − Ered
This also means that if electrons were Wred (E) =
(4πkT λ)1/2
to be transferred from electrolyte to the (12)
302 4 Solar Energy Conversion without Dye Sensitization
where Eox o and E o are the most probable must match with energy level of the redox
red
energies for the unoccupied and occupied electrolyte. For example, an electron trans-
quantum states in solution, respectively. ferred from the valence band to redox
The parameter λ represents the activation electrolyte would be facilitated if the level
energy for the process of transferring the of the former matches the oxidized level
solvation shell structure from equilibrium of the later. Alternatively, if electrons are
condition of one species to the most to be transferred to the conduction band
probable structure of the other. This is from the redox electrolyte, the conduction
also called reorientation energy. The density band must match the reduced level of the
of states for reduced and oxidized species redox electrolyte. Thus, knowledge of such
at a particular energy are given as distribution functions of the redox elec-
trolyte helps in understanding the I–V
Dred (E) = Cred × Wred (E) (13) characteristics of a semiconductor, espe-
Dox (E) = Cox × Wox (E) (14) cially when the λ value of the electrolyte
approaches the material’s band gap. Under
where Cred and Cox are the concentrations such a condition, the rectifying charac-
of the reduced and oxidized species, ter of semiconductor-electrolyte junction
respectively. The energy distribution is does not control the nature of anodic and
a Gaussian bell shaped curve (Fig. 8b). cathodic current, that is, an n-type semi-
In the harmonic oscillator approximation, conductor would give a cathodic current
the reorientation energy of reduced and in addition to the anodic current, and the
oxidized species are equal and are given by reverse would be true for a p-type semi-
o o o conductor. With this type of condition, one
Eredox = Eox − λ = Ered +λ ···
either has to select an electrolyte whose re-
(15)
dox potential (i.e. its Fermi level) is such
or, that only one of the two levels of the redox
o or E o as the requirement
electrolyte (Eox
o o
Eox − Ered = 2λ ··· (16) red
may be) matches the concerned band edge
Bockris and Khan [17], however, have level of the semiconductor. Alternatively,
suggested that energy is not symmetrically bias would have to be applied to the elec-
distributed around the central point (i.e. trode to stop the unwanted current.
the ground state) but in a Maxwell type
distribution. 4.1.11.3 Effect of Charge Distribution
The most interesting concept derived in the Electrolyte
from these views is that like a semicon- What is the role of the constituents of
ductor, the oxidized and reduced species the redox electrolyte present at the in-
of redox electrolyte are linked with a con- terface of the semiconductor-electrolyte
duction band (unoccupied electronic state) junction? Choice and optimization of the
and a valence band (occupied electronic electrolyte has a substantial effect on PEC
state). The energy necessary to transfer an solar cells [18–19]. Before a semiconduc-
electron from reduced to oxidized state is tor is immersed in an electrolyte, anions
2λ, analogous to the band gap of a semi- and cations freely and randomly move in
conductor Eg . the solution. As a result of this movement,
For an effective charge transfer to occur, no specific spatial accumulation of ions
the energy level of the semiconductor occurs in the solution. This situation alters
4.1 The Photoelectrochemistry of Semiconductor/Electrolyte Solar Cells 303
of (1/α) for band gap light, so that of Aw(hν − Eg ) can be made zero except
efforts could be made to select a suitable (hν − Eg ). This would mean that the value
redox electrolyte such that (1/α) ∼ w. of (hν) (i.e. the intercept on (hν)-axis ‘‘s’’)
In addition, we can also calculate Ln should be equal to Eg . If we get the linear
from magnitude of the intercept (i.e. plot with n = 1, the band gap is a direct type
ln(1 + αLn )) of the linear plot on y-axis. and if n = 4, it is an indirect one. Some
From Ln the lifetime of the carrier (τ ) can be semiconductors may give linear plot for
calculated by using the following equation both values of n, indicating simultaneous
# $ presence of both types of band gap.
µn kT τ 1/2 Thus, application of the Gärtner model
Ln = (24)
q to data obtained from photocurrent ver-
sus potential measurement yields several
µn the mobility of the carrier, can be
important parameters of a semiconductor-
obtained from Hall measurement.
electrolyte junctions (e.g. Eg , Vfb , Lmin , τ ,
From these measurements, we can
and w).
thus evaluate the flat band potential
(which is equal to EF for heavily doped
4.1.14
semiconductor), absorption coefficient (α)
Effect of Counter-Electrode
for band gap light, diffusion length (L) and
lifetime of the carrier (τ ).
Besides the considerations on the param-
eters of the semiconductor and the redox
4.1.13.1.3 When ‘w’ is Constant During electrolyte, performance of the counter-
Photocurrent Measurement Using α = electrode also matters for designing an
{A(hν − Eg )n/2 }/ hν and keeping w con- efficient PEC cell. The counter-electrode
stant, Eq. (22) becomes in a PEC cell is a charge carrier collecting
' ( material, and catalyzes the electron trans-
A(hν − Eg )n/2 fer, but is not chemically involved in the
− ln(1 − φ) =
hν electrochemical reaction.
× w + ln(1 + αLmin ) (25) The counter-electrode material should
be electrically conducting, economical,
When (αLmin ) ( 1, Eq. (25) can be rear- and chemically stable in an electrolyte.
ranged to get It should minimize polarization losses
during exchange of electrons at the
−[{ln(1 − φ)}hν]2/n = Aw(hν − Eg ) interface. In other words, polarizability of
(26) the counter-electrode should be small in a
In order to utilize Eq. (26), a semiconduc- given redox electrolyte, indicating that the
tor-electrolyte junction is illuminated by electrode reaction occurring at the counter-
different wavelengths of light, keeping the electrode is reversible and fast. This
applied potential to a PEC cell constant maximizes the potential obtained from the
(i.e. under a constant forward potential). semiconductor–electrolyte interface.
A plot of [{ln(1 − φ)}hν]2/n versus hν For many redox reactions, noble met-
would thus result in a linear graph. Let als such as Pt, Au have been preferred
us assume that the value of the intercept as counter-electrode by electrochemists.
on the x-axis at [{ln(1 − φ)}hν]2/n = 0 is However, those are expensive for ter-
‘s’. At this condition, none of the terms restrial application, and materials such
4.1 The Photoelectrochemistry of Semiconductor/Electrolyte Solar Cells 307
0.006
(a)
0.005
(b)
0.004 Pmax
(c)
0.003
[Amp]
0.002
I
0.001
0.000
−0.001
−0.002
−0.2 0.0 0.2 0.4 0.6 0.8 1.0
V
[Volts]
Fig. 9 Current-voltage characteristic of a PEC cell (α-PbO/Fe (CN)6 4− (0.1M), Fe(CN)6 3−
(0.01M) pH 9.2/(counter-electrode). Three types of counter-electrodes were used:
(a) conducting ITO electrode, (b) surface modified ITO electrode using platinum and
(c) platinum electrode.
308 4 Solar Energy Conversion without Dye Sensitization
−8.4
−8.0 30°C 630
In conductivity × 10−2
21°C
−7.6 12°C 610 0°C
4°C 5°
−7.2 2°C 590 10°
In σ
−9°C 15°
−6.8 570 20°
−6.4 550 25°
−6.0 530
−5.6 510
5.4 5.8 6.2 6.6 7.0 7.4 7.6 8.2 8.6 9.0 9.4 0 1 2 3 4 5 6 7
(a) pH (b) pH
Barrier height (Activation energy)
0.5 0.14
0.12
0.4
0.10
[eV]
[eV]
0.3
0.08
0.2
0.06
0.1
0.04
0.0
0.02
0 1 2 3 4 5 6 7 8 9 5 6 7 8 9 10
(c) pH (d) pH
Fig. 10 Surface conductivity measurements of thin films of α-Fe2 O3 and
p-SnSe21 . Log of surface conductivity versus pH at different temperatures for
(a) n-α-Fe2 O3 (b) p-SnSe. Activation energy of surface conductivity versus
pH for (c) n-α-Fe2 O3 and (d) p-SnSe.
In Fig. 10, typical results of thin films n- because the surface of the semiconductor
α-Fe2 O3 (a,c) and p-SnSe (b,d) are shown. would have no excess charge for pinning
n-α-Fe2 O3 gives a minimum at pH 4.1, the Fermi level. Moreover, photogenerated
which is equivalent to its pzc. A reverse carriers will also not be trapped by the sur-
behavior is obtained with p-SnS showing face states as their charges are neutralized
a maximum at pH 7.5, which is equivalent by pH of the solution.
to its pzc. This is the only nondestructive
method available to get information about 4.1.15
pH that corresponds to pzc for large area Photodecomposition of Semiconductor
thin-film semiconductor material.
A PEC cell operating at the pH equiv- In the dark many semiconductors are sta-
alent to the pzc will experience the least ble in contact with a given electrolyte, but
hindrance caused by the surface states, and start decomposing on illumination of the
may be able to show its anticipated PEC junction. This and energy conversion effi-
behavior. PEC operating at pzc would be ciency are the most serious problems faced
able to show its predicted contact potential in PEC devices. If the photo-generated
310 4 Solar Energy Conversion without Dye Sensitization
minority carriers are not rapidly used in Unfortunately, most of the low band gap
oxidation or reduction by the electrolyte materials studied so far have been found to
immediately, they have a tendency to re- undergo such corrosions. Therefore, there
act with the semiconductor material. The is a need to develop newer materials for
semiconductor electrode therefore starts developing a photoelectrochemically stable
corroding. PEC etching is being dealt semiconductor.
in a greater depth in a separate chap-
ter. However, the easiest way to establish 4.1.15.1 Materials for PEC Cell
PEC stability of the semiconductor is by Because of the limitation of space, it is
measuring the photocurrent of a PEC cell difficult to discuss details of materials
over a longer period under continuous il- studied for various types of PEC cell. Some
lumination. If semiconductor is stable, the of the materials and the redox electrolytes
photocurrent will show a constant value used for studying PEC cells are tabulated
over a period of illumination. in Tables 1 and 2. Nevertheless, Sharon
Tab. 1 Some of the chalcogenide materials studied for developing PEC cell in various redox
electrolytes
Tab. 2 Some of the phosphide/oxide and other materials studied for developing PEC cell in
various redox electrolytes
and coworkers [29, 30] have extensively Table 3. Highest reported conversion ef-
reviewed these materials. ficiencies include more than 9% for dye
Examples of chronological improve- sensitized thin film of n-TiO2 (nanos-
ments in high solar to electrical conver- tructure) in polyiodide electrolyte [97] and
sion efficiency PEC’s are summarized in more than 16% for single-crystal n-CdSe
312 4 Solar Energy Conversion without Dye Sensitization
in cyanide modified K3 Fe(CN)6 3− elec- efforts to extend the lifetimes of the requi-
trolyte [98]. The synthesis of single-crystal site dyes [97].
semiconductor material is not cost effec- But unfortunately none of these mate-
tive, being both time intensive and limited rials show a long-term PEC stability to
to formation of smaller surface area of warrant making of a commercially viable
materials. A variety of techniques have PEC cell. Interestingly, it has been hinted
been developed to provide polycrystalline that phosphides of Co, Fe, Mo, Ni, V,
or amorphous thin-film semiconductors. and W (band gap in vicinity of 1 eV) may
A particular advantage of electrodeposition be photoelectrochemically stable. Among
(plating) techniques is a one-step chemi- the oxide materials, PbO2 (Eg = 1.6 eV) is
cal synthesis of the semiconductor. PV, worth exploring because it is highly stable
as well as PEC solar cells requires the in acidic medium. Unfortunately, due to
use of thin-film semiconductors to be cost high oxygen vapor pressure, PbO2 exits
competitive. Each of these techniques con- as a degenerated semiconductor (i.e. it be-
sumes relatively small amounts of material haves like a metal). If suitable doping could
and can be applied to develop large surface open the band gap, then PbO2 can become
area electrodes. Modes of film preparation one of the best materials for making com-
include slurry techniques, chemical vapor mercially viable PEC cell. Semiconducting
deposition, wackercast, rf sputtering, vac- carbon is another class of material that
uum coevaporation, sol gel, chemical bath needs to be extensively studied for this pur-
deposition, hot pressing, and electroplat- pose. Earlier, diamond was the only known
ing. These techniques have been used semiconductor of carbon. But in the recent
to prepare thin films of n-CdS, n-CdSe, past, it has been possible to synthesize
n-CdTe, n-GaAs, p-InP, n-Si and p-Si, n- semiconducting carbon with a band gap
TiO2 and used in PEC solar cells [97, 105]. in the vicinity of 1.0 to 2.8 eV. Sharon and
Advantageously, thin film PEC’s have coworkers [107, 108] has recently reported
exhibited photocurrent lifetimes signifi- a PEC cell from a semiconducting carbon
cantly greater than single-crystal PEC’s. made from the pyrolysis of camphor. In
This phenomenon has been attributed addition, efforts should be made to reduce
to the high microscopic surface area, the band gap of stable oxides such as TiO2
and resulting lower effective microscopic into a low band gap material by manipu-
photocurrent density, passing through lating to obtain their mixed oxides. Mixed
the thin-film devices. The longest life- oxides of Mo & W also need similar at-
time PEC device to date, slurry de- tention. Sharon and coworkers [109, 110]
posited thin-film n-Cd (Se, Te) electrodes have tried to develop a theoretical model to
immersed in modified polysulfide so- predict the variation in band gap by varying
lutions, has demonstrated outdoor op- the composition of components in mixed
eration for approximately one year. In oxides. In addition to this, there is a need to
this system, photoelectrode and electrolyte develop a theoretical model, to predict PEC
stability was achieved by optimization stability of materials. Such models will go
of the solution alkalinity, using alter- a long way to minimize the number of ex-
native cations, and modifying the ratio periments needed in developing materials
of dissolved sulfur to sulfide in solu- for making a commercially viable PEC cell.
tion [106]. The recent advance in dye- It is necessary to emphasize that in spite
sensitized PEC devices requires future of the simplicity in fabrication, the future
4.1 The Photoelectrochemistry of Semiconductor/Electrolyte Solar Cells 313
of economically viable PEC depends en- beam is scanned over the entire surface of
tirely on the development of suitable low the film, and photocurrent is measured at
band gap, photoelectrochemically stable each illuminated point. A 3D plot of pho-
materials. tocurrent versus (x, y) distances is made. If
the composition of materials formed over
4.1.16 the large surface area electrode contains
Laser Scanning Technique for a Large Area photoactive oxide, a uniform distribution
Electrode of photocurrent would be observed. Fail-
ing to get uniform photocurrent would
There are materials, which can form va-
be an indication of nonuniformity of the
rieties of nonstoichiometric compounds.
surface with the desired material. Though
For example, PbO2 can coexist as PbOx
this technique is not specific about the
where x can have values from 0 to 2. composition, it is a very useful tool to give
Among these oxides, most photoactive ox- a guideline for improving the preparation
ide is PbO0.8 . If PbO0.8 oxide is to be technique especially for a large area elec-
prepared by anodic oxidation of lead, it trode. The resolution of this experiment,
is possible that surface of the lead film however, depends on the scanning size of
might contain other forms of oxides as the laser beam. For example, the 3D laser
well. Thus, a PEC cell prepared by anodiza- scanning map of an anodized lead sheet
tion of lead may give a low photoresponse, (2.5 cm2 ) shows that the photocurrent
not because PbO0.8 is an inappropriate ma- is not as uniform (Fig. 11a) as obtained
terial but because anodized surface may be with film anodized by condition shown in
contaminated with other forms of oxides Fig. 11(b).
as well. There seems to have no nonde-
structive technique available to confirm
4.1.17
the uniformity of surface with photoac- Summary
tive species, especially of a large area
electrode. Sharon and coworkers [111] has In this section, efforts are made to explain
recently developed a laser scanning sys- the basic principles of a regenerative PEC
tem to overcome this problem. A laser solar cell. Greater emphasis is made on
y
y
x
x
[a.u.]
[a.u.]
I ph
I ph
30
7.
30
8.4 8.4
0
7.
87
0 5.6
87
4.
5.6 0
4.
0
43
2.8 2.8
43
2.
0 0
2.
0.0
00
0.0 0
00
0.
(a) 0 (b)
0.
Fig. 11 3D photocurrent map obtained with scanning of He/Ne laser (3 mW) over the surface of lead
sheet anodized in the potential range (a) −0.32 to +0.08 V and (b) −0.22 to +0.08 V versus SCE [26].
314 4 Solar Energy Conversion without Dye Sensitization
76. A. Barrasse, H. Cachat, G. Horowitz et al., 94. A. J. Nozik, Appl. Phys. Lett. 1976, 29, 150.
Rev. Phys. Appl. 1982, 17, 801. 95. A. Heller, M. Robbins, K. J. Bachmann
77. D. S. Ginley, R. M. Bie Feld, B. A. Parkinson et al., J. Electrochem. Soc., 1978, 125, 831.
et al., J. Electrochem. Soc. 1982, 129, 145. 96. S. Licht, O. Khaselev, T. Soga et al., Elec-
78. W. S. Hobson, A. B. Ellis, Appl. Phys. Lett. trochem. Soc. State Lett. 1998, 1, 20.
1982, 41, 891. 97. B. O’Regan, M. Grätzel, Nature 1991, 353,
79. S. Gourgaud, D. Elliott, J. Electrochem. Soc. 737.
1977, 124, 102. 98. S. Licht, D. Peramunage, Nature 1990, 345,
80. A. Heller, B. Miller, B. A. Parkinson, 330.
J. Electrochem. Soc. 1979, 126, 954. 99. S. Licht, D. Peramunage, Nature 1991, 354,
81. M. G. Chris, S. L. Nathan, Nature 1982, 300, 440.
733. 100. B. J. Tufts, I. L. Abrahams, P. G. Santangelo
82. R. Memming, J. Electrochem. Soc. 1978, 125, et al., Nature 1987, 326, 861.
117. 101. S. Licht, R. Tenne, G. Dagan et al., Appl.
83. A. J. Bard, R. E. Malpas, K. Itaya, J. Am. Phys. Lett. 1985, 46, 608.
Chem. Soc. 1979, 101, 2535. 102. S. Menezes, H. J. Lewerenz, K. J. Bach-
84. H. J. Lawerenz, M. Lubke, Appl. Phys. Lett. mann, Nature 1983, 305, 615.
1981, 39, 798. 103. G. Kline, K. Kam Keung, R. Ziegler et al.,
85. W. M. Ayers, J. Electrochem. Soc. 1982, 129, Sol. Energy. Mater. 1982, 6, 337.
1644. 104. G. Kline, K. Kam, D. Canfield et al., Sol.
86. A. J. Bard, F.-R.-F. Fan, G. A. Hope, Energy. Mater. 1981, 4, 301.
J. Electrochem. Soc. 1982, 129, 1647. 105. A. Heller, B. Miller, Appl. Phys. Lett. 1981,
87. T. S. Jayadevaiah, Appl. Phys. Lett. 1974, 25, 38, 282.
399. 106. B. A. Parkinson, A. Heller, B. Miller, Appl.
88. A. J. Bard, F.-R.-F. Fan, H. S. White et al., Phys. Lett. 1978, 33, 521.
J. Am. Chem. Soc. 1980, 102, 5142. 107. S. Licht, J. Phys. Chem. 1986, 90, 1096.
89. M. S. Wrighton, A. B. Bocarsly, E. G. Wal- 108. K. Mukhopadhyay, I. Mukhopadhyay,
ton et al., J. Electroanal. Chem. 1979, 100, M. Sharon et al., Carbon, 1997, 35, 863.
283. 109. M. Sharon, I. Mukhopadhyay, K. Mukho-
90. M. S. Wrighton, K. D. Legg, A. B. Ellis et al., padhyay, Sol. Energy Mater. Sol. Cells 1997,
Proc. Natln. Acad. Sci. USA 1977, 74, 4116. 45(1), 35.
91. M. S. Wrighton, R. G. Austin, A. B. Bocarsly 110. K. Murali Krishna, M. Sharon, M. K.
et al., J. Am. Chem. Soc. 1978, 100, 1602. Mishra, J. Phys. Solids 1996, 57(5), 615.
92. M. S. Wrighton, J. M. Bolts, A. B. Bocarsly 111. K. Murali Krishna, M. Sharon, M. K. Mishra
et al., J. Am. Chem. Soc. 1979, 101, 1378. et al., Electrochem. Acta 1996, 41(13), 1999.
93. R. M. Candea, M. Kastner, R. Goodman 112. M. Sharon, I. Mukhopadhyay, S. Ghosh,
et al., J. Appl. Phys. 1976, 47, 2724. J. Solid State Electrochem. 1999, 3, 141.
4.2 Photoelectrochemical Solar Energy Storage Cells 317
Load Load
E E
Light Light
EC − D∗
− EC − −
EF EF
Energy Energy
− gain gain
Red Ox+ Red Ox+ −
D
EV + EV
+
in the valance band toward the electrolyte, charge is then injected into a wide band
where they participate in an oxidation re- gap semiconductor. This alternative car-
action. Electrons through the bulk drive an rier generation mode can also lead to
external load before they reach the counter effective charge separation as illustrated
electrode or storage electrode, where they in Fig. 1(b). The first high solar to elec-
participate in a reduction process. Under tric conversion efficiency example of such
illumination and open circuit, a negative a device was demonstrated in 1991 [3]
potential is created in a photoanode, and as through the use of a novel high surface area
a result the fermi level for the photoanode (nanostructured thin film) n-TiO2 , coated
shifts in the negative direction, thus reduc- with a well-matched trimeric ruthenium
ing the band-bending. Under illumination complex dye immersed in an aqueous
with increasing intensity, the semiconduc- polyiodide electrolyte. The unusually high
tor fermi level shifts continually toward surface area of the transparent semi-
negative potentials until the band-bending conductor coupled to the well-matched
effectively reduces to zero, which corre- spectral characteristics of the dye leads
sponds to the flat band condition. At this to a device that harvests a high proportion
point, a photoanode exhibits its maximum of insolation.
photovoltage, which is equal to the barrier
height. 4.2.1.2 Photoelectrochemical Storage
Excitation can also occur in molecules PECs can generate not only electrical
directly adsorbed and acting as a medi- but also electrochemical energy. Figure 2
ator at the semiconductor interface. In presents one configuration of a PEC com-
this dye sensitization mode, the func- bining in situ electrochemical storage
tion of light absorption is separated from and solar-conversion capabilities; provid-
charge carrier transport. Photoexcitation ing continuous output insensitive to daily
occurs at the dye and photogenerated variations in illumination. A high solar to
4.2 Photoelectrochemical Solar Energy Storage Cells 319
hn
S
P P= 0
S
POLYSULFIDE
OXIDATION
n -Cd(Se,Te) n -Cd(Se,Te)
POLYSULFIDE
TIN SULFIDE
REDUCTION
REDUCTION
POLYSULFIDE
REDUCTION
OXIDATION
MEMBRANE
MEMBRANE
Sn/SnS
TIN
Sn/SnS
CoS
1.0 m Cs2S4
0.8 m CsOS
0.8 m CsHS
1.8 m CsOH
1.0 m Cs2S4
0.8 m CsOS
1.8 m CsHS
0.8 m CsHS
1.8 m CsOH
1.8 m CsHS
L CoS LOAD
LOAD L
(a) (b)
Fig. 2 Schematic of a photoelectrochemical solar cell combining both solar conversion and storage
capabilities. (a) Under illumination; (b) in the dark.
electric conversion efficiency cell config- upon reversal of the storage process.
uration of this type was demonstrated in To date, the predominant nonbiologic
1987 and used a Cd(Se,Te)/Sx conversion utilization of solar energy is to heat a
half cell and a Sn/SnS storage system, working fluid that is maintained in an
resulting in a solar cell with a contin- insulated enclosure, storing a portion of
uous output [4]. Under illumination, as the incident solar radiation for future use.
seen in Fig. 2(a), the photocurrent drives Limitations of this approach include the
an external load. Simultaneously, a por- low energy available per unit mass of
tion of the photocurrent is used in the the storage medium, and low efficiencies
direct electrochemical reduction of metal of thermal to mechanical and thermal to
cations in the device storage half-cell. In electrical energy conversion. A variety of
darkness or below a certain level of light, passive and dynamic optical concentrators
the storage compartment spontaneously have been studied to compensate for these
delivers power by metal oxidation, as seen limitations.
in Fig. 2(b). The high temperatures generated by
concentrated solar power have been uti-
4.2.2 lized to drive highly endothermic re-
Comparative Solar-Storage Processes actions. The reverse reaction releases
the chemically stored energy as ther-
4.2.2.1 Thermal Conversion and Storage mal energy. Various systems have been
Solar insolation can be used to directly investigated such as in 1985 [5]:
activate a variety of thermal processes; the
enthalpy is stored physically or chemically SO3 −−−→ SO2 + 12 O2 !H = 98.94
and then either directly utilized or released (1)
320 4 Solar Energy Conversion without Dye Sensitization
nν
Decomposition ∗Ru(bpy) 2+
3
∗TEOA
Ru(bpy)3 2+
Mv++
Ru(bpy)33+
Pt H2O
TEOA
Mv+•
TEOA N·(CH3·CH2·OH)3
+ H2
Mv++ CH3 — N N — CH3
+
Scheme 1
processes, these reactions are also gener- principle, this should lead to a higher level
ally driven only by high-energy radiation of photon absorption and more effective
(short wavelength) and cannot efficiently charge separation. Both effects can sub-
convert incident AM1 solar radiation. stantially increase solar photochemical-
H2 or O2 generation from water is conversion efficiency, but these systems
a multielectron process. Optimization of have not yet displayed high efficiencies of
photoredox-splitting of water necessitates fuel generation or long-term stabilities.
the presence of a catalyst to mediate Photoredox processes at semiconductor
this complex multielectron transfer. In electrodes generating fuels or products
one such process, a sacrificial reagent other than hydrogen, including methanol
triethanolamine (TEOA) is consumed ir- and ammonia, have been attempted with
reversibly in the process, as denoted in low overall yields. The photoelectrolysis of
Sch. 1 [9]: HI into H2 and I3 − at p-InP electrodes has
Direct multielectron processes are rare been described [11]. These H2 and I3 − pho-
and instead incorporate one or more rad- togenerated products are prime candidates
ical intermediate steps. These reactive for a fuel cell. Analogous advanced sys-
intermediates are susceptible to unfavor- tems, in which the photoelectrochemically
able side reactions, resulting in substantial generated fuels have been successfully re-
losses in the energy-conversion process. combined to generate electrical energy, are
Kinetically favored back reactions further discussed later in this section.
reduce the overall conversion efficiency. A system exemplifying photoelectro-
The engineering of these complex molec- chemical synthesis to generate hydro-
ular organizations provides a substantial gen is water photoelectrolysis. An early
scientific challenge and have generally re- demonstration of water photoelectroly-
sulted in systems with low conversion sis used TiO2 (band gap 3.0 eV) and
efficiencies. was capable of photoelectrolysis at ∼0.1%
solar to chemical energy–conversion ef-
4.2.2.3 Semiconductor Photoredox ficiency [12]. The semiconductor SrTiO3
Storage was demonstrated to successfully split wa-
Semiconductor surfaces have been used ter in a direct photon-driven process by
as sensitizers to drive photochemical con- Bolts and Wrighton (1976), albeit at low
version and storage of solar energy. In solar energy–conversion efficiencies [13].
322 4 Solar Energy Conversion without Dye Sensitization
L e L
e−
R R′
O R′
hν hν
x−
x−
O R O′
O′
cannot efficiently fulfill the duel role of configuration as shown in Fig. 4. In Fig. 4,
being kinetically sluggish to reduction dur- the switches E and F are generally alter-
ing illumination and yet being kinetically nated during charge and discharge. During
facile to the same reduction during dark the charging, only switch E may be closed,
discharge. The configuration represented facilitating the storage process, and dur-
in Fig. 3 has another disadvantage, the ing discharge, E is kept open while F
disparity between the small surface area is closed. In this case, chemical changes
needed to minimize photocurrent dark that took place during the storage phase
current losses and the large surface area are reversed, and a current flow is main-
necessary to minimize storage polarization tained from the storage electrode to a third
losses to maximize storage capacity [20]. (counter) electrode that is kept in the first
compartment. To minimize polarization
4.2.3.2 Three-Electrode Configurations losses during the discharge, this third elec-
Several of the two-electrode configura- trode should be kinetically fast to the redox
tion disadvantages can be overcome by couple used in the first compartment.
considering a three-electrode storage cell Still an improved situation would be to
E
• •
L
• •
F
R′
R R
hν
EL
O O′
O
P A M S
have both switches closed all the time. In Improvements relating to its stability and
this case, electric current flows from the conversion efficiency are of paramount
photoelectrode to both counter and stor- importance.
age electrodes. The system is energetically
tuned such that when insolation is avail- 4.2.4.1 Improvements of the
able, a significant fraction of the converted Photoresponse of a Photoelectrode
energy flows to the storage electrode. In To improve the solar response of a
the dark or diminished insolation, the photoelectrode, a proper match between
storage electrode begins to discharge, driv- the solar spectrum and the band gap of
ing continued current through the load. the semiconductor should be maintained.
In this system, a proper balance should When a single band gap semiconductor is
be maintained between the potential of used, a band gap in the vicinity of 1.4 eV
the solar energy–conversion process and is most desirable from the standpoint of
the electrochemical potential of the stor- optimum solar-conversion efficiency. An
age process. There may be residual electric important criterion is that the minority
flow through the photoelectrode during carrier that is driven toward the semi-
dark cell discharge, as the photoelectrode conductor–electrolyte interface should not
is sluggish, but not entirely passive, to a participate in a photocorrosion reaction
reduction process. This can be corrected that is detrimental to the long-term stabil-
by inserting a diode between the photo- ity of the photoelectrode. Photocorrosion
electrode and the outer circuit. can be viewed in terms of either kinetic
or thermodynamic considerations and the
4.2.4 real cause may be a mixture of both. From
Optimization of Photoelectrochemical thermodynamic perspective, a photoanode
Storage is susceptible to corrosion if the fermi level
for holes is at a positive potential with re-
The power obtained is the product of volt- spective to the semiconductor corrosion
age and current, and consideration of the potential [21]. The corrosion can be pre-
photocurrent is as important as the photo- vented or at least inhibited by choosing a
voltage. If the band-bending is sufficiently redox couple that has its Eredox more nega-
large, then the minority carrier redox re- tive than that for the corrosion process [22,
action, which is essential to maintain 23]. The kinetic approach has been to allow
the photocurrent, can compete effectively another desired redox process to occur at a
with the recombination of photogenerated much faster rate than the photocorrosion
electron hole pairs. This recombination reaction [13]. Other attempts to minimize
represents a loss of absorbed photo en- the photocorrosion has been to coat the
ergy. Therefore, an objective is to maintain photoelectrode surface with layers such as
a high band-bending and at the same time Se [24] and protective conductive polymer
a significant photovoltage. A photoanode films [25], and to search for alternate low
creates a negative photovoltage under il- band gap semiconductors [26]. Extensive
lumination, which results in reducing the reviews on the performance and stability
band-bending. In principal, one way to ac- of cadmium chalcogenides include those
complish high band-bending is to choose a by Cahen and coworkers, 1980 [27] and
very positive redox couple in the electrolyte. Hodes, 1983 [28]. Etching of photoelec-
The converse is true for a photoanode. trode surface has been recognized and
326 4 Solar Energy Conversion without Dye Sensitization
10. economy and cost effectiveness, and a counter electrode. As an example of chal-
11. reduced toxicity and utilization of lenges that may arise in the combined
environmentally benign materials. photoconversion and storage system, con-
sider an n−CdSe/polysulfide/tin sulfide
A photoelectrochemical solar cell implic- version of Fig. 4 and consisting of Cell 1.
itly contains an electrolytic medium. In Cell: 1. CdSe | HS− , OH− , Sx 2− |
the majority of laboratory PEC configu- Membrane | HS− , OH− | SnS | Sn
rations, incident light travels through the With illumination, the cell exhibits si-
electrolytic medium before illuminating multaneous photoelectrode, counter elec-
the photoelectrode. The resultant light ab- trode and storage reactions, and equilibria
sorbance by the electrolyte is a significant including
loss, which is avoided by use of a back cell
configuration. For example, the substan- Photoanode:
tial absorptivity of dissolved polysenide
HS− + OH− −−−→ S + 2e− + H2 O
species has been avoided in a n-GaAs pho-
(9)
toelectrochemistry through the use of the
Photocompartment equilibria:
back wall cell configuration presented in
Fig. 5 [45]. S + S2−
x −−−→ S( x + 1)
2−
(10)
The photoelectrochemical system shown
Counter electrode:
in Fig. 4 is a combination of a photoelec-
trode, electrolyte, membrane, storage, and S + 2e− −−−→ HS− + OH− (11)
Photoelectrode Counter
contact electrode
contact
Sapphire Cu wire
window
Mounted
hν thin-film
PEC
hν Ag paste
Au grid
hν
1.9 µm Photoelectrode
thick contact
n -GaAs
Polyselenide
(aq)
Window
Epoxy
A: P: IVphoto
B: ISC
Current
Vmax
EB VPh EA
Potential
Fig. 6 Current–voltage curves for power, Pmax , and Isc and Vph are the short-circuit
electrochemical storage processes, A or B. current and open-circuit photopotentials,
Process A may be charged by the photodriven respectively. EA and EB refer to the redox
current–voltage curve P, whereas process B may potentials for redox processes, A and B,
not. In the photodriven IV curve P, Vmax is the respectively.
voltage corresponding to the point of maximum
330 4 Solar Energy Conversion without Dye Sensitization
photovoltage that can be generated. Isc is Ideally, the membrane used to separate
the short-circuit photocurrent correspond- the two-cell compartments, as indicated
ing to maximum band-bending. In Fig. 6, in Fig. 4, must be permeable only to ions
consider the electrochemical process rep- that will transport charge, but that will not
resented by curve B. This process is located chemically react or otherwise impair any
outside the region of potentials gener- electrode. The permeability of membranes
ated by the photoelectrode; it does not is generally less than ideal. Different mem-
represent a potential storage system to branes permit other ions and water to per-
be driven by a single photoelectrode. In meate to a varying degree [46, 47]. Gross
such a case, a serial combination of more mixing of active materials across the mem-
than one photoelectrode would be neces- brane causes them to combine chemically
sary. For a redox process to be a potential and in the process lose energy. Favorable
candidate for a redox storage system, the qualities that a membrane should exhibit
storage and photodriven current–voltage are low permeability toward chemically
curves should intersect. Whereas Vph and reactive ions, low resistivity, mechanical
Isc correspond to zero power, the point integrity, and cost effectiveness.
Pmax shown in Fig. 6 corresponds to the
maximum power point. Solar energy con- 4.2.5
version is accomplished at its maximum High Efficiency Solar Cells with Storage
efficiency only during operation in the po-
4.2.5.1 Multiple Band Gap Cells with
tential vicinity of Pmax . Storage
By adjusting the electrical load L, shown A limited fraction of incident solar photons
in Fig. 4, the system can be constrained have sufficient (greater than band gap) en-
to operate near its maximum power effi- ergy to initiate charge excitation within a
ciency. In this case, if the counter electrode semiconductor. Because of the low frac-
is not polarized, the potential difference tion of short wavelength solar light, wide
between photoelectrode and the counter band gap solar cells generate a high photo-
electrode will be close to Vmax . If one voltage but have low photocurrent. Smaller
chooses a facile redox process for the stor- band gap cells can use a larger fraction of
age electrode, as indicated by the sharply the incident photons, but generate lower
rising IV curve for process A in Fig. 6, photovoltage. Multiple band gap devices
with Eredox in the vicinity of Vmax , then the can overcome these limitations. In stacked
potential during charge and discharge of multijunction systems, the topmost cell
the storage process will remain near Vmax . absorbs (and converts) energetic photons,
As a result, the potential will be a highly but it is transparent to lower energy pho-
invariant current variation through the tons. Subsequent layer(s) absorb the lower
load L, regardless of insolation intensity. energy photons. Conversion efficiencies
This situation represents an ideal match can be enhanced, and calculations predict
between solar energy conversion and stor- that a 1.64-eV and 0.96-eV two–band gap
age processes within a PECS. Nonideality system has an ideal efficiency of 38% and
occurs with poor voltage-matching or ki- 50%, light of 1 and 1000 suns intensity,
netic limitations and polarization losses respectively. The ideal efficiency increases
associated with the counter, storage, or to a limit of 72% for a 36–band gap solar
photoelectrodes. cell [48].
4.2 Photoelectrochemical Solar Energy Storage Cells 331
Recently, high solar conversion and stor- storage cell described by the Fig. 7 en-
age efficiencies have been attained with ergy diagram. The single cell contains
a system that combines efficient mul- both multiple band gap and electrochem-
tiple band gap semiconductors, with a ical storage, which unlike conventional
simultaneous high capacity electrochem- photovoltaics, provides a nearly constant
ical storage [49, 50]. The energy diagram energetic output in illuminated or dark
for one of several multiple band gap cells conditions. The cell combines bipolar Al-
is presented in Fig. 7, and several other GaAs (Eg = 1.6 eV) and Si (Eg = 1.0 eV)
configurations are also feasible [1, 2a, b]. and AB5 metal hydride/NiOOH storage.
In the figure, storage occurs at a potential Appropriate lattice-matching between Al-
of Eredox = EA+/A − EB/B+ . On illumina- GaAs and Si is critical to minimize dark
tion, two photons generate each electron, current, provide ohmic contact without ab-
a fraction of which drives a load, whereas sorption loss, and maximize cell efficiency.
the remainder (1/xe− ) charges the storage The NiOOH/MH metal hydride storage
redox couple. Without light – the poten- process is near ideal for the AlGaAs/Si be-
tial falls below Eredox – the storage couple cause of the excellent match of the storage
spontaneously discharges. This dark dis- and photocharging potentials. The electro-
charge is directed through the load rather chemical storage processes utilizes MH ox-
than through the multijunction semicon- idation and nickel oxyhydroxide reduction:
ductor’s high dark resistance.
MH + OH− −−−→ M + H2 O + e− ;
Cell: 2. In Fig. 8, an operational form
of the solar conversion is presented and a EM/MH = −0.8 V vs SHE (13)
1/xe–
(1–1/x)e–
V = Vw + Vo Rload
Electrocatalyst anode
e– 1/xA → 1/xA+
Electrocatalyst cathode
ECs
ηA
– EFermi(ns)
e EAredox
Vs
FCw
hν E
Gs
EFermi(nw=ps) 1/xB → 1/xB+
+
E Bredox ηB
Vw EGw > hν > EGs ⇒ h E Vs
hν EGw
Electro-
Semiconductor Electrolyte
catalyst
EFermi(pw)
Storage bipolar MPEC
+
hν > EGW ⇒ h E VW
Ohmic
p n p n
junction
Wide gap Small gap
Fig. 7 Energy diagram for a bipolar band gap indirect ohmic storage multiple band
gap photoelectrochemical solar cell (MBPEC).
332 4 Solar Energy Conversion without Dye Sensitization
1/xe–
(1−1/x)e-
Storage
Rload
current
current
current
Photo
Load
e–
350 µm
300 nm
(4 × 1019cm–3) 800 nm
Au-Zn/Au
1.0 µm
1.0 µm
1.7 nm
50 nm
20 nm
10 nm
P+ -GaAs
NiOOH + H2O + e–
MH + OH–
Separator with KOH electrolyte
19cm–3)
(8 × 1015cm–3)
(1 × 1018cm–3)
(0.3-0.15)Ga(0.7-0.85)As (1 × 10 cm )
(2 × 1017cm–3)
(1 × 1018cm–3)
–3
x
18
Nickel cathode
(1 × 10
2hν
Au-Sb/Au
⇒
MHx–1 + H2O + e–
Ni(OH)2 + OH–
GaAs(buffer layer)
Illumination
N+-Al(0.15)Ga(0.85)As
N-Al(0.15)Ga(0.85)As
P+-Al0.8Ga0.2As
N-GaAs
N+-Si
P+-Al
+-Si
N-Si
P
1.5
Load potential
1.0 Generated VCell insensitive to illumination variation
[mA]
0.5
AlGa/Si MH multiple band gap storage solar cell
4
I Photo
0
Illumination Dark Illumination Dark
Discharge Discharge
0
I Storage,
[mA]
Charge Charge
−2
insitu AB5/NiOOH
Metal hydride storage
0 6 12 18 24 30 36 42 48
Time,
[h]
Fig. 9Two days measured conversion and storage characteristics of the
AlGaAs/Si/MH/NiOOH MBPEC solar cell.
product of the measured cell potential and cell, and the remainder consists of the
measured load current. power over load during illumination, as
Under constant 12-hour (AM0) illumi- illustrated in Fig. 10. In the dark, inclusive
nation, the long-term indoor cycling cell of storage losses, the stored energy is spon-
generated a nearly constant photocurrent taneously released and this power over
density of 21.2 (constant to within 1% or load during both 12-hour illumination and
±0.2 mA cm−2 ), and as seen in the top 12-hour dark periods is also summarized.
curves of Fig. 10, a photopower that var- The cell is a single physical–chemical de-
ied by ±3%. The cell’s storage component vice generating load current without any
exhibits the expected increase in charging external switching.
potential with cumulative charging, which
moves the system to a higher photopoten- 4.2.5.2 PECS Driving an External Fuel Cell
tial. The observed increase in photopower In the early 1980s, Texas Instruments,
during 12 hours of illumination is be- Inc. developed an innovative program
cause of this increase in photopotential based on a hybrid photovoltaic storage that
with cumulative charging. A majority of used imbedded multilayer photoanode and
the photogenerated power drives the redox photocathode silicon spheres and was
334 4 Solar Energy Conversion without Dye Sensitization
7 7
6 6
5 5
Power
[mW]
[V]
3 3
1 1
0 0
0 2 40 42 140 142 240 242
Time, days
Fig. 10 Eight-months photopotential and power characteristics of the AlGaAs/Si/MH/NiOOH
solar cell under fully charged AM0 conditions. Each day, the cell is illuminated for 12 hours and
is in the dark for 12 hours.
Glass panel
Separator cover
Br2
Hydrogen HBr
bromide anode
H2 cathode
Br h 1/2 Br2
e− 1/2 H2
H Metal
e h
p n np
e
h Glass p n h− p n
e
Conductive plane
Storage
Hydrogen Fuel cell and heat
storage converter exchanger
are of significant scientific interest and The back reaction between Leucoth-
form a solid basis for further develop- ionene (TH4 2+ ) and Fe3+ is slow. Leucoth-
ment toward future systems. The following ionene is oxidized at the SnO2 electrode.
three sections, covering PECS with either
a solution, solid, or intercalation storage TH4 2+ −−−→ TH+ + 2e− + 3H+ (17)
redox processes, and provides a brief sum- Ferric cation is reduced at a Pt electrode
mary of many of these investigations, in a second compartment.
with a particular emphasis on their per- As a result, the concentration ratio of
formance. Fe3+ /Fe2+ is increased in the first com-
partment and decreases in the second
4.2.6.1 PECS Cells with Solution Phase compartment, which is equivalent to a dif-
Storage ference in chemical potential. The system
Cell: 4 [52]. This is an example of the use returns to its original uncharged state by
of photoexcitable absorbers to promote a discharging in the dark. Only 60 mV of po-
redox process using the following reaction tential difference is equivalent to a decade
sequence: change of the Fe3+ /Fe2+ concentration ra-
tio. Hence, this concentration cell does
not generate a significant potential and
TH+ +2Fe2+ +3H+ −−−→ TH4 2+ +2Fe3+
the power density is low. The cell has the
(TH+ = Thionene acetate) (16) configuration
336 4 Solar Energy Conversion without Dye Sensitization
moderate resistance of 20 ohm cm2 . The anthraquinone redox couple have been
storage reaction and cell configuration are demonstrated in this study. Any H2 evolu-
tion would carry out direct hydrogenation
Light of AQ and associated side reactions, and
2Br− + 3I2 ←
−−−
−−
→− Br2 + 2I3
−
(20)
Dark
therefore a carbon electrode has been se-
lected because of the H2 over potential
n-MoSe2 | 0.1 M HBr, 0.01 M Br2 |
on this electrode. The cell underwent sev-
Nafion 315 | 1 M KI, 0.18 M I2 | Pt eral deep charge and discharge cycles with
Studied cell characteristics: Single reproducible performance. The storage re-
crystal photoelectrode, Pt counter action and cell form are
electrodes, cell configuration as in Fig. 4, Light
Eredox of Br/Br− = 1.087 vs SHE, Eredox 2I− + AQ + 2H+ −
←−−
−−
→− AQH2 + I2 (22)
of I3 − /I− = 0.534 V vs SHE. Illumination Dark
200 mW cm−2 Xe lamp, conversion n-WSe2 | 1 M KI, 0.1 M Na2 SO4 , 0.5 M
efficiency = 6.2%, potential across two Pt H2 SO4 | Saturated KCl bridge | 5 × 10−2
terminals of the charged cell = 0.49, and M AQ, 0.5 M H2 SO4 | C
short-circuit current = 0.5 mA.
Cell: 8 [55]. The next cell also uses a Studied cell characteristics: Single-crys-
Nafion membrane, but makes use of n- tal photoelectrode, C counter electrode
CdSe to drive a polysulfide–polyselenide during charging, Pt during discharg-
storage couple. Low-output power density ing. The cell configuration is similar
is the biggest drawback in this cell. The to that in Fig. 4. Eredox of I3− /I− =
storage reaction and cell configuration are 0.534 V vs SHE, Eredox of AQ/AQH2 , il-
lumination 150 mW cm−2 He–Ne laser
Dark (632.8 nm), conversion efficiency = 9%,
S0 + Se2− ←
−−−
−−
→ 2−
− S + Se
0
(21)
Light discharge across a 10 ohm load produces
a current of 1 mA cm−2 , and open-circuit
n-CdSe | 1 M in Na2 S, S, NaOH | Nafion voltage 200 mV.
315 | 1 M in M Na2 Se, Se, NaOH | Pt Cell: 10 [44]. This study uses a p-WSe2
Studied cell characteristics: Poly- photocathode rather than n-WSe2 . During
crystalline photoelectrode, Pt counter the cell discharge, oxidation of AQH2 at
electrodes, cell configuration as in the surface of p-WSe2 indicates that the
Fig. 4, Eredox of polysulfide electrolyte = electrode has the duel role of being a
−700 mV vs SCE, saturated calomel elec- cathode during the charging and being the
trode, Eredox of Se2 2− /Se2− = −800 mV vs anode during the discharge. As discussed
SCE, illumination 100 mW cm−2 Xenon earlier, this limits the activity and low-
lamp, conversion efficiency 4%, FF = current densities were observed. The
0.45, photovoltage = −400 mV, charged storage reaction and cell configuration are
cell has an open-circuit voltage of 60 mV, Light
and initial current across a 100 ohm across AQ + 2H+ + 2I −
←−−
−−
→− AQH2 + I2 (23)
Pt electrode = 0.5 mA. Dark
Cell: 9 [44]. This study uses organic re- p-WSe2 | 5 × 10−2 M AQ | Saturated |
dox species for energy-storage purposes. 1 M KI, 0.5 M H2 SO4 | Pt
Stability of the n-WSe2 photoanode in Single crystal | 0.5 M H2 SO4 | Salt
iodine electrolyte and the stability of bridge | 0.5 M Na2 SO4 |
338 4 Solar Energy Conversion without Dye Sensitization
Cell: 11 [55]. The theoretical band gap of Studied cell characteristics: Polycrys-
WSe2 provides a near ideal single band gap talline photoelectrode, Pt counter elec-
match for the solar spectrum. But the fol- trodes, cell configuration is similar
lowing cell has some disadvantages. These to Fig. 4, Eredox of O2 , H+ /H2 O
include the low solubility of the storage re- couple = 1.23 V vs NHE at pH = 1,
dox couple employed, MV2+ and MV+• and Eredox of Ag/Ag+ = 0.80 V vs NHE,
the possibility of undesirable side reactions normal hydrogen electrode illumination
of the radical ion MV+• . Using dual (n-type 500 W Hg lamp, conversion efficiency =
and p-type) photoelectrodes expands the 1%, photopotential = 0.28 V vs NHE,
potential regime one can access for the open-circuit voltage of the charged
redox-storage couple. The storage reaction cell = 0.28 V, and short-circuit current =
and cell configuration are 0.3 mA cm−2 .
Light Cell: 13 [57]. Of the four TiO2 Cells
2I− + 2MV2+ −
←−−
−−
→ +•
− 2MV + I2 (24) 10–13, the following cell exhibited the
Dark highest short-circuit discharge current and
−
n-WSe2 | I | MV2+ | p-WSe2 voltage. However, during the charging
process, a stationary concentration of
4.2.6.2 PECS Cells Including a Solid
Ce4+ was observed in the photoanode
Phase–Storage Couple compartment. This suggest the existence
The earlier experimental investigations, of competing process that consumes the
Cells 2–9, use only solution phase redox oxidized species Ce4+ . The later is known
couples. However, as indicated in the to participate in photochemical reactions
following examples, a solid phase–storage under illumination [10]. Considering the
couple may also be employed, which low concentration of the reduced form of
in principle tends to increase the cell’s active materials used with the photoanode,
storage capacity. there is a possibility that the water
Cell: 12 [57]. Having at least one oxidation becomes the dominant process
component in insoluble form may add during charging. In this study, it was
compactness into the cell configuration, observed that with a passage of a charge
although low conductivity of the insoluble of 10 coulomb during charging, Ce4+
active component may cause significant present in the photoanode compartment
polarization losses associated with the accounted for only 22% of the charge. In
storage electrode, as exemplified by the this second TiO2 photoelectrode cell, the
low conductivity of silver (chloride) in one storage reaction and cell configuration are
of the next cells. The next four cells use Light
a TiO2 polycrystalline photoelectrode. In Ce3+ + Ag+ ←
−−−
−−
→ 4+
− Ce + Ag (26)
the first cell, the storage reaction and cell Dark
configuration are TiO2 | 1 M HNO3 , 0.05 M Ce2 (SO4 )3 ,
Light 0.1 M Ce(SO4 )2 | Anion Specific Mem-
2H2 O + 4Ag+ −
←−−
−−
→ +
− 4H + 4Ag + O2 brane | 1 M AgNO3 , 1 M KNO3 | Ag
Dark
(25) Studied cell characteristics: Polycrys-
TiO2 | 1 M HNO3 , 1 M KNO3 | Anion talline photoelectrode, Pt counter elec-
Specific Membrane | 1 M AgNO3 , 1 M trode, cell configuration illumination etc.
KNO3 | Ag are similar to the earlier cell, Eredox
4.2 Photoelectrochemical Solar Energy Storage Cells 339
and Ti change their oxidation state and the n − CdSe | 1 M in NaOH, Na2 S,
charge balance is maintained by the mi- S Nafion-315 | 0.1 M ZnO, 1 M NaOH | C
gration of Na+ ions from one phase to the
other. In the actual cell design, an n-type Studied cell characteristics: Polycrys-
semiconductor is connected to the alu- talline photoelectrode, Ni counter elec-
mina phase containing Ti and p-material trode, basic cell configuration is based
is connected to the phase containing Fe. on Fig. 4, Eredox of Sx 2 /S2− = 0.500 V vs
Limitations are the comparatively slow SHE, Eredox of Zn/Zn(OH)4 2− = −1.25 V
diffusion of ions in the solid electrolyte vs SHE, artificial illumination, conversion
and resistance to ionic movement at var- efficiency = 3%, photovoltage = −0.50 V,
ious phase boundaries, and lower the during discharge through 75-ohm load
energy output during discharge. In this between C and Ni discharge current =
device, back wall illumination demands 10 mA, and voltage = 0.6 V.
the use of very thin semiconductor layers Cell: 22 [61]. This cell takes advantage
to minimize absorption losses and has the of photocorrosion to drive a storage cell.
general form Under illumination, n-CdSe is decom-
posed and p-CdSe is electroplated, and
n-semiconductor | Na2 O.11(AlFeO3 ) | the reverse occurs during cell discharge.
Na2 O.11(Al2 O3 ) | Na2 O.11(AlTiO3 ) | However, photoactivity depends on an op-
p-semiconductor timized semiconductor surface, and in an
environment where the surface is changed
Cell: 21 [47]. In this detailed study, selec- constantly, the surface optimization is lost.
tion of a Nafion-315 membrane was done This and the poor kinetics of the p-type
on the basis of (1) stability in high alkaline photoreduction result in a continual dete-
sulfide solutions, (2) low IR drop, and (3) rioration of the photoactivity and cause low
low permeability to sulfide. Maintaining an photoefficiency and low-discharge power
area ratio of 1 : 8 between photo and storage density. The storage reaction and cell con-
electrodes has minimized polarization at figuration are
the storage electrode. The storage system
was driven by three semiconductor PEC Light
devices connected in series. Charging was CdSe + 2h+ ←
−−−
−−
→ 0
− Se + Cd
2+
(35)
Dark
done up to 90% of the capacity followed
by complete discharge. Overall observed and the other electrode in photoelectro-
charge efficiency was 83%. Although the plated by Cd
system was not fully optimized with re-
Light
spect to photoelectrode, electrolyte, and CdTe + 2e− ←
−−−
−−
→ 0 2−
− Cd + Te (36)
storage, voltage efficiency of 75% was ob- Dark
tained during discharge. Discharge curves n-CdSe | 0.1 M CdSO4 | p-CdTe
were flat until the stored active material
was fully consumed. The storage reaction Cell: 23 [46]. This is a detailed study of a
and cell configuration are thin film cell with moderately high outdoor
solar efficiency, high storage efficiency,
Light
− and an output that is highly invariant de-
S2− + Zn(OH)4 2− ←
−−−
−−
→ 0
− S + Zn + 4OH
Dark spite changing illumination. This study
(34) provides extensive details of the choice
342 4 Solar Energy Conversion without Dye Sensitization
Fig. 12 A bipolar thin film photoelectrochemical solar cell with in situ storage.
Compartments A and A2 contain alkali polysulfide solution and compartment B
contains alkali sulfide solution.
intercalating compounds that are able to because to reach the junction, light has to
exchange guest ions and molecules with an travel several layers. The cell functions in
electrolyte in a reversible manner, and yet the same manner as the earlier cell, and
that is not disruptive to photon absorption. the configuration is given by
Cell: 25 [63]. In this cell, Eredox of copper
thiophosphate is variable depending on Conductive Glass | Cu | Cu+ Conducting
the degree of intercalation. A limitation of solid electrolyte | p-Cu2 Te | n-CdTe | Mo
this system is poor-discharge kinetics and
low-energy density of the discharge. The
4.2.7
cell configuration is given by Summary
Cu3 PS4 | 0.02 M CuCl | CH3 CN | Cu2 S
Conversion and storage of solar energy
Studied cell characteristics: Eredox of is of growing importance as fossil fuel
Cu+ /Cu0 = −0.344 vs NHE, illumination energy sources are depleted and stricter
117 mW cm−2 Xe lamp, photopotential = environmental legislation is implemented.
l00 mV, charging current < 50 µA cm−2 , Although society’s electrical needs are
and discharge current < 10 µA cm−2 . largely continuous, clouds and darkness
Cell: 26 [64]. This cell illustrates another dictate that photovoltaic solar cells have
all solid state design for a thin storage an intermittent output. Photoelectrochem-
cell. p-Cux S changes its electrode potential ical systems have the potential to not
with changes in its composition. During only convert but also store incident so-
charging, Cu is oxidized at n-CdS surface lar energy. Design component and system
while it is reduced at the Cu electrode. considerations and a number of photoelec-
Between the two electrodes Cu+ ion trochemical solar cells with storage have
transport process takes place in the solid been reviewed in this chapter.
state electrolyte. The cell configuration is
given by
Acknowledgment
Cu | n-CdS | p-Cu2 S | RuCl4 I5 Cl3.5 | Cu
S. Licht is grateful to Dharmasena Per-
Cell: 27 [64]. As with the earlier cell, amunage and for support by the BMBF
this final cell requires a very thin design Israel–German Cooperation.
344 4 Solar Energy Conversion without Dye Sensitization
52. G. W. Murphy, Solar Energy 1978, 21, 403. 58. Y. Yonezawa, M. Okai, M. Ishino et al., Bull.
53. M. Sharon, S. Kumar, N. P. Sathe et al., Solar Chem. Soc. Jpn 1983, 56, 2873.
Cells 1984, 12, 353. 59. M. Kaneko, K. Takagashi, E. Tsuchida, J.
54. M. Sharon, A. Singha, Int. J. Hydrogen Energy Electroanal. Chem. 1987, 227, 2512.
1982, 7, 557. 60. A. Sammells, A. Ang, Patent 1980, US
55. P. G. Ang, P. A. F. Sammells, Faraday Dis- 4,235,9512.
cussions of the Chemical Society, Gen- 61. H. J. Gerritsen, W. Ruppel, P. Wurfel, J. Elec-
eral Discussions 1980, No. 70, Photoelectro- trochem. Soc. 1984, 131, 2037.
chemistry, St. Catherine’s College, Oxford, 62. H. Tributsh, Appl. Phys. 1986, 23, 61.
Sept 8–10. 63. G. Betz, S. Fiechter, H. Tributsch, J. Appl.
56. F. -R. F. Fan, H. S. White, B. L. Wheeler Phys. 1987, 62(11), 4597.
et al., J. Am. Chem. Soc. 1980, 102, 5142. 64. T. Tonomura, K. Teratoshi, 1986, JP
57. H. Hada, K. Takaoka, M. Saikawa, Y. Yone- Patent 62,249,366; ibid., JP Patent 62,249,
zawa, Bull. Chem. Soc. Jpn. 1981, 54, 1640. 3612.
346 4 Solar Energy Conversion without Dye Sensitization
ε cm−1 Atm−1
M. Zelikoff, J. Opt. Soc. Am.
1953, 43, 753).
250
0
100 150 200
nm
If we return to Eq. (1) and substitute photoelectrochemical (PEC) solar cell con-
Eqs. (7) and (9), we obtain figuration similar to that described in
the earlier chapter. In brief, a PEC cell
ηphotoelectrolysis = ηphoto × ηelectrolysis
! " consists of two electrodes separated by
#GH2 O 1 + losses a suitable redox electrolyte. Both elec-
= (10)
#Hph #GH2 O trodes could either be n-type and p-type
Two conclusions arise from Eq. (10). semiconductors or one electrode could
First, it is seen that efficiency can be be a semiconductor (either n-type or p-
maximized if electrochemical ‘‘losses’’ can type) and the other electrode could be
be made small as compared with the Gibbs a noncorrosive metal. If both electrodes
free energy change for water electrolysis, are made of semiconductor, then the an-
#GH2 O . Second, it is seen that #HH2 O , the odic electrode should be made from n-type
enthalpy of water electrolysis (with little semiconductor and the cathodic electrode
temperature dependence), is involved in should be made from p-type semicon-
determining the overall efficiency as well ductor. This last configuration adds the
as #Hph , the enthalpy of photogenerated additional requirement of photocurrent
charge carriers. If these two can be matching through the two photoelec-
properly matched, a maximum overall trodes. Regarding n-type semiconductor
efficiency may be accomplished. (Fig. 3), during its illumination photogen-
erated carriers are generated, which assist
4.3.3 in the oxidation of electrolyte at the inter-
Photoelectrochemical Cell for face of the semiconductor and electrolyte.
Photoelectrolysis A reverse phenomenon takes place if a
p-type semiconductor is used, that is,
Conversion of solar energy into electrical photogenerated electrons are generated at
energy can be achieved by using a the semiconductor–electrolyte interface to
RL
0.3 cm
C A
Fig. 3 A schematic representation of a
photoelectrochemical cell. A, B, and C
are counter electrode, electrolyte, and hν
semiconductor electrode, respectively.
RL is used to vary the potential (V) to B
measure photocurrent (A) flowing
across two electrodes.
350 4 Solar Energy Conversion without Dye Sensitization
initiate reduction of the electrolyte. Thus, apply an external bias (Eb ) to facilitate the
in a PEC cell (Fig. 3), if water is used in- photoelectrolysis of water. The energy level
stead of a redox electrolyte, oxidation of of semiconductor and other photoelectro-
water can take place at the illuminated chemical reactions are shown in Fig. 4(b).
n-type semiconductor (C) and reduction The semiconductor anode was illuminated
can take place at counter electrode (A). with UV radiation. On illumination of
However, for photodecomposition of wa- TiO2 electrode, the photogenerated holes
ter rather than a single redox couple, two oxidize water to produce oxygen and the
different redox couples are used at the two photogenerated electrons are transferred
immersed electrodes, with a provision to to counter electrode to perform the reduc-
collect hydrogen and oxygen separately. tion of water at the platinum electrode.
E-bias
E-bias
e−
O2 H2 e−
e−
H2
e− H e− EF
e−
EC
H2O Eb
hν
EF
Pt EG
e− OH + H+
EV P+
NaOH Na2SO4 O2
Frit n -TiO2 Electrolyte
n -TiO2 Cathode Metal Helmholtz Pt
Anode contact layer
(a) (b)
Fig. 4 (a) Fujishima-Honda cell with n-TiO2 photoanode and Pt-cathode. (b) Schematic
energy level diagram of the cell. EV -valence band, EC -conduction band, EF -Fermi level,
EG -energy gap (for n-TiO2 ; EG = 3.0 eV), Eb -bias voltage, p+ -hole (after Nikola Getoff,
Int. J. Hydrogen Energy 1990, 15(6), 407).
4.3 Solar Photoelectrochemical Generation of Hydrogen Fuel 351
−
Conduct band
− −
EF,sc 2− − −
H2 2H+ Fermi-level
nEF
Stored
hν ∆V energy
1 O + 2H+
H2O
pEF 2 2
2+
Valence band +
+
Load
h+
e− EC
e 2H2O + 2e− 2OH− + H2
hν
e
EC ∼1.7 V EF
EF Eν
+
hν
35
25 AM1.5 GaAs
Cu2S a-Si : H
Si a-Si : H : F
Efficiency
[%]
20 AMO
15
T = 300 K
Ge CdS
10
5
0.5 1.0 1.5 2.0 2.5
Semiconductor band gap
[eV]
Fig. 7 Theoretically calculated conversion efficiency of solar cell materials versus band
gap for single junction cells (after Adolf Goetzberger, Christopher Hebling, Sol. Energy
Mater. Sol. cells 2000, 62, 1).
matches with the reduction potential of with photopotential larger than VH2 O .
water), which is not sufficient to electrolyze Sharon and Rao [19] developed a photo-
water. On the contrary, for obtaining the electrochemical cell with a semiconductor
necessary electrolysis photopotential using separating two types of electrolytes: one
a single semiconductor, a semiconductor electrolyte forming an ohmic contact and
of much larger band gap is needed (Fig. 5). the other forming a Schottky-type contact.
With such material, biasing is needed, as It is postulated that like the formation of
is case with n-TiO2 (Fig. 4). Moreover, as a Schottky-type barrier, if the magnitude
a result of high band gap, n-TiO2 absorbs of the Fermi level of the semiconductor
only about 2% of the solar spectra (Fig. 2). and the redox electrolyte is same, then the
Therefore, a semiconductor of large band contact between them should be ohmic.
gap would correspondingly give very low This provided an opportunity to visualize a
photocurrent (hence, lowphotoconversion PEC cell in which the semiconductor acts
efficiency), making the photoelectrolysis like a separator between two types of re-
processes uneconomical. dox electrolytes. The electrolyte is selected
such that the front side of the semicon-
4.3.4.2 Multiple-type PEC Cells ductor (i.e. the side to be illuminated)
The calculations from the earlier section forms a Schottky junction. The backside
suggest that there is a need to introduce of the semiconductor is kept in contact
some improvements in the PEC cell to at- with another redox electrolyte, which gives
tain a sufficient solar energy conversion an ohmic contact. The front side of the
4.3 Solar Photoelectrochemical Generation of Hydrogen Fuel 353
Load
Semi- hν Counter hν hν
conductor electrode h+
electrode
e e e B
A
e− + OX− Red
carbon electrode
Z
h+ h+ h+
PEC cells
e
Red e + OX− Pb3O4
Pb3O4 h+ − Red OX−
Pb + X h+
Lead
Y
Fig. 8 Schematic representation of multiple between the metal and the semiconductor and
PEC cells connected by the help of redox the direction of flow of photogenerated electron
electrolyte. Inset X shows the energy level of and hole. Electron is extracted to the load via
semiconductor Pb3 O4 electrode deposited over metal A and the hole is extracted via inert metal
metal lead. Photoelectrochemical reactions B. Inset Z describes reduction process occurring
occurring at two ends of the semiconductor is at inert carbon electrode-B.
shown. Inset Y shows the ohmic contact
354 4 Solar Energy Conversion without Dye Sensitization
by the backside of the last left side elec- radiation (assuming each semiconductor
trode (A) and metallic electrode on the to be of the same area). Employing this
extreme right side of the cell (B). The total cell as a source of electrical power, it also
photopotential of such system is equal to can be used for electrolyzing water (Fig. 9).
the photopotential of one PEC cell times
the number of semiconductor electrodes 4.3.4.3 Bipolar Cell
used [20]. This arrangement gives an op- The bipolar cell of Sharon-Schottky cell has
portunity to get the required electrolysis been further developed. Bard and cowork-
photopotential, even with the semiconduc- ers [21, 22] developed a bipolar semicon-
tor of band gap ≈1.4 eV. In addition, the ductor photoelectrode array (Fig. 10) and
requirement for matching the conduction studied its application to light-driven water
band edge (or flat band potential) with the splitting and electrical power generation.
electrode potential for hydrogen evolution They used five n-TiO2 electrodes in series.
and the valence band edge with elec- In the subsequent developments [23], they
trode potential of oxygen evolution (Figs. 5 devised a similar bipolar cell (Fig. 11) with
and 6) becomes redundant. Photocurrent, CoS/CdSe. In this cell, also a salt bridge
however, would depend on the area of the is used to complete the electrical circuit.
individual semiconductor exposed to solar CoS is used to make ohmic contact with
Bipolar
n -SC electrode
1/4 O2 O O O 1/2 H2
1/2 H2O R R R H+
Oxygen Hydrogen
collector collector
CoS
E
− CB
CdSe/Electrolyte −
interface *
n EF EF E°
hν H2 / H2O
CdSe Pt Pt / Electrolyte
E ° 2− s2
s / 2 *
pEF
interface
+ VB
+
Direction of electron flow
Fig. 11 (a) Schematic representation of water
photoelectrolysis cell. A, J: Pt; B, D, F, H: CoS; C, E, G, I: CdSe.
Solutions 1,6: KOH (1M); 2–5 Na2 S (1M), S (1M), KOH (1M).
For H2 and O2 generation solutions 1 and 6 are connected
with KOH bridge. (b) Expansion shows energetics of bipolar
panel (after A. J. Bard et al., J. Phys. Chem. 1987, 91, 6).
n-CdSe as well as with the electrolyte. Four for sensitization of semiconductor elec-
CoS/CdSe electrodes are used. Polysulfide trode. Hence, this will not be discussed
is used as an electrolyte in these PEC cells. here any further.
In this arrangement, the PEC cell is formed
among (B, C), (D, E), (F, G), and (H, I). 4.3.5
Sides C, E, G, and I form Schottky-type Recent Developments
barriers while sides B, D, F, and H act
like a counter electrode. The first and the Principal solar water-splitting models had
last electrodes are made of platinum. The predicted similar dual band gap photo-
energy level diagram to show the flow of electrolysis efficiencies of only 16% and
electron is shown in Fig. 11. 10–18% [3, 28], respectively, whereas re-
Several oxides (e.g. TiO2 [7], SrTiO3 [8], cently dual band gap systems were cal-
n-SiC [24], p-GaP [25], Fe2 O3 [26], etc.) and culated to be capable of attaining over
chalcogenides such as CdSe [23] have been 30% solar photoelectrolysis conversion ef-
used as anodes for water photodecompo- ficiency [14]. The physics of the earlier
sition. n-TiO2 , even today seems to be one models were superb, but their analy-
of the most important materials, because sis was influenced by dated technol-
it has been possible to extend its spec- ogy, and underestimated the experimental
tral response into visible portion of the ηphoto attained by contemporary devices
solar spectrum through sensitization with or underestimated the high experimen-
organic dyes. A separate chapter is devoted tal values of ηelectrolysis which can be
356 4 Solar Energy Conversion without Dye Sensitization
attained. For example, Ref. 3 estimates low developed in bipolar cells, amounts to
values of ηphoto (less than 20% conversion) connecting large number of PEC cells
because of the assumed cumulative rel- in series. The advantage of bipolar cells
ative secondary losses that include 10% is that it can operate with low band
reflection loss, 10% quantum-yield loss, gap semiconductor, yet provide desired
and 20% absorption loss. Experimentally, photopotential and high solar to electri-
a cell containing illuminated AlGaAs/Si cal efficiency. Although photopotential of
RuO2 /Ptblack was demonstrated to evolve such device is equal to the photopotential
H2 and O2 at record solar–driven water of one-photoelectrode times the number
electrolysis efficiency. Under illumination, of PEC cells connected in series, photocur-
bipolar configured Al0.15 Ga0.85 As (Eg = rent of system depends on the intensity
1.6 eV) and Si (Eg = 1.1 eV) semiconduc- of solar radiation falling on an individual
tors generate open circuit and maximum photoelectrode. But it is important to re-
power photopotentials of 1.30 and 1.57 V, alize that none of semiconductors so far
respectively, well suited to the water elec- developed can be used for prolonged pho-
trolysis thermodynamic potential: toelectrolysis of water, because of their
inherent instability towards photocorro-
H2 O −−−→ H2 + 12 O2 ;
sion. Success of making bipolar cells as a
o commercial viable system entirely depends
EH 2O
= EO2 − EH2 ;
o ◦ on the development of photoelectrochem-
EH 2O
(25 C) = 1.229 V (11)
ically stable low band gap semiconductor.
o
The EH /photopotential-matched semi-
2O
conductors are combined with effective Acknowledgment
water electrolysis O2 or H2 electrocat-
alysts, RuO2 , or Ptblack . The resultant S. Licht is grateful to Helmut Tributsch for
solar photoelectrolysis cell drives sustained his review and suggestions on theoretical
water-splitting at 18.3% conversion effi- sections of this chapter and thankful for
ciencies [14]. These recent developments the support by the BMBF Israel-German
in hydrogen generation at high solar en- Cooperation and the Berman-Shein Solar
ergy conversion efficiency are detailed in Fund. Maheshwar Sharon is thankful to
this volume in the chapter titled ‘‘Opti- his students and in special to G. Ranga
mizing Photoelectrochemical Solar Energy Rao who contributed toward development
Conversion: Multiple Band Gap and Solu- and modification of PEC cell.
tion Phase Phenomena.’’
References
4.3.6
Conclusion
1. M. A. Green, K. Emery, K. Bucher et al.,
Propgr. Photovolt. 1999, 11, 31.
In this section, efforts are made to discuss 2. S. Licht, B. Wang, T. Soga et al., Appl. Phys.
the thermodynamics of photoelectrolysis Lett. 1999, 74, 4055.
of water using a PEC cell. To facilitate 3. J. R. Bolton, S. J. Strickler, J. S. Connolly,
Nature 1985, 316, 495.
the generation of required potential for
4. K. W. Atanabe, M. Zelikoff, J. Opt. Soc. Am.
the photoelectrolysis of water, discussions 1953, 43, 753.
are made on some modified PEC cells 5. Nikola Getoff, Int. J. Hydrogen Energy 1990,
popularly known as bipolar cell. Potential 15(6), 407.
4.3 Solar Photoelectrochemical Generation of Hydrogen Fuel 357
6. L. J. Heidt, A. F. McMillan, Science 1953, 17. A. Fujishima, K. Honda, Bull. Chem. Soc.
117, 75. Jpn. 1971, 44, 1148.
7. A. Fujishima, K. Honda, Nature 1972, 238, 18. Heinz Gerischer, Pure Appl. Chem. 1980, 52,
37. 2649.
8. J. M. Bolts, M. S. Wrighton, J. Phys. Chem. 19. Maheshwar Sharon, G. Ranga Rao, Indian J.
1976, 80, 2641. Chem. 1986, 25A, 170–172.
9. R. Memming, Top. Curr. Chem. 1988, 143, 20. M. Sharon et al., Electrochim. Acta 1991,
79. 36(7), 1107–1126.
10. A. Nozik, Appl. Phys. Lett. 1976, 29, 150. 21. A. J. Bard et al., J. Phys. Chem. 1986, 90,
11. J. White, F. -R. Fan, A. J. Bard, J. Electrochem. 4606.
Soc. 1985, 132, 544. 22. A. J. Bard et al., J. Electrochem. Soc. 1988, 135,
12. R. C. Kaintala, J. Ö. M. Bockris, J. Int. Hydro- 567.
gen Energy 1988, 13, 375. 23. Tooru Inoue, Toshihiro Yamase, Chem. Soc.
13. O. Khaselev, K. Turner, Science 1998, 280, Japan, Chem. Lett. 1985, 869.
425. 24. H. Honeyama, H. Sakamoto, H. Tamura,
14. S. Licht, B. Wang, S. Mukerji et al., J. Phys. Electrochim. Acta 1979, 277, 637.
Chem. B 2000, 104, 8920. 25. Lynn C. Schumacher, Suzanne Mamiche-
15. R. Memming in Photochemical Conversion Afara, Michael F. Weber et al., J. Electrochem.
and Storage of Solar Energy, (Eds.: E. Pelizzetti, Soc. 1985, 132(12), 2945.
M. Schiavello), Kluwer Academic Publishers, 26. M. F. Weber, M. J. Digman, Int. J. Hydrogen
Netherlands, 1991, pp. 193–212. Energy 1986, 11, 225.
16. F. Gutman, O. J. Murphy in Modern Aspects 27. Ibid, J. Electrochem. Soc. 1984, 131, 1258.
of Electrochemistry (Eds.: White, Bockris, 28. Adolf Goetzberger, Christopher Hebling,
Conway) 1983, P5. Sol. Energy Mater. Sol. cells 2000, 62, 1.
358 4 Solar Energy Conversion without Dye Sensitization
Bipolar gap Bipolar gap Bipolar gap Inverted gap Inverted gap Inverted gap
indirect direct Schottky contact indirect direct Schottky contact
(semiconductor- (semiconductor- photoelectro- (semiconductor- (semiconductor- photoelectro-
metal-solution) solution) chemical metal-solution) solution) chemical
ohmic contact ohmic contact solar cell ohmic contact ohmic contact solar cell
(BGIO PEC) (BGDO PEC) (BGS PEC) (IGIO PEC) (IGDO PEC) (IGS PEC)
Regen- Storage Regen- Storage Regen- Storage Regen- Storage Regen- Storage Regen- Storage
erative BGIO erative BGDO erative BGS erative IGIO erative IGDO erative IGS
BGIO PEC BGDO PEC BGS PEC IGIO PEC IGDO PEC IGS PEC
PEC PEC PEC PEC PEC PEC
Fig. 2 Relation of the twelve representative MPEC configurations comprising regenerative ohmic cells: (1) bipolar gap direct ohmic regenerative,
(2) bipolar gap indirect ohmic regenerative, (3) inverted gap direct ohmic regenerative, (4) inverted gap indirect ohmic regenerative; storage ohmic cells:
(5) bipolar gap direct ohmic storage, (6) bipolar gap indirect ohmic storage, (7) inverted gap direct ohmic storage, (8) inverted gap indirect ohmic
storage; regenerative Schottky cells: (9) bipolar gap Schottky regenerative, (10) inverted gap Schottky regenerative; and storage Schottky cells:
(11) bipolar gap Schottky storage, and the (12) inverted gap Schottky storage configuration.
4.4 Optimizing Photoelectrochemical Solar Energy Conversion: 361
EGw = ECw − EVw ; EGs = ECs − EVs Adjacent band gap layers in a multiple
(4) band gap configuration can also be aligned
362 4 Solar Energy Conversion without Dye Sensitization
V = Vw + Vs − η cathode + η anode
e−
R load
ECs e−
e−
A+ → A → A+
Electrocatalyst anode
A
ηcathode
Semiconductor
ηanode
EFermi (ns) EGs
e− Eredox
hν Vs
ECw
EFermi (nw = ps) EVs
Vw h+
EGw > hν > EGs⇒
h ν EGw
EFermi (pw)
+
h ν > EGw ⇒ h EVw
Ohmic p-Schott.
p n Electrolyte
junction small gap
(a) Wide gap
V = Vw + Vs − η cathode + η anode
e−
R load
e−
A+ → A A → A+
e− E
electrocatalyst cathode
Electrocatalyst anode
Cs
Semiconductor or
ηcathode ηanode
e− EFermi(ns)
Eredox
Ecw Vs
hν
EGs
EFermi (nw=ps)
Vw − ηc+ ηaw
Rload-w
Rload-s
Electrocatalyst cathode
Electrocatalyst anode
Electrocatalyst anode
Semiconductor/metal
ECs Vs− η c+ ηas
e− e− e− e−
ECw ηas ηc ηc ηaw
EFermi (nw=ns) E
Vs hν Gs Eredox Eredox
Vw A → A+
h ν EGw EVs EFermi (ps) A+ ← A
Ohmic
p n n p
Wide gap junction Small gap
(c)
e− e−
Es = V s Ew = Vw
ECs
Electrocatalyst cathode
Electrocatalyst cathode
e− e− + +
electrocatalyst anode
electrocatalyst anode
B →B C →C
ECw
Semiconductor or
Semiconductor or
ηcB ηcC
EFermi (nw=ns) hν EGs EredoxB EredoxC
Vs ηaA
Vw
hνE EredoxA D → D+
Gw EVs EFermi (ps) ηaD
EGw > h ν > EGs ⇒ h+ A→ A+
EFermi (pw) EredoxD
Ohmic
hν > EGw ⇒ h+ EVw junction Electrolyte Electrolyte
p n n p
Fig. 3 (Continued)
The examples presented here utilize interference will occur for the top AlGaAs
combine multijunction solid-state layers layers, and bottom Si layers through this
consisting of a bipolar AlGaAs (EGw = or substantially shorter electrolyte path-
1.6 eV) wide band gap, overlaid on a lengths. The solid-state component in-
Si (EGs = 1.0 eV) small band gap, and cludes a graded band emitter, varying from
used in an electrolytic cell [2]. Light ab- Al(0.3 – 0.15) Ga(0.7 – 0.85) As with overlayers
sorption by the electrolyte can interfere of p + -Alx Gax−1 As on n-Alx Gax−1 As. The
with the cell and should be avoided. growth sequence and graded band emitter
Figure 4 overlays the optical characteris- layer improve collection efficiency [2]. The
tics of the solid and solution phase of Si bottom cell consists of a p + -Si, n-Si,
the AlGaAs/Si solid-state and V3+/2+ elec- and n+ -Si multijunction. The band edges
trolyte optimized components within a observed in the figure at approximately
bipolar gap photoelectrochemical cell. So- 800 nm and 1100 nm are consistent with
lution transmission is measured through a the respective AlGaAs and Si band gaps.
pathlength that is typical (1 mm) of many For efficient electron/hole pair charge
experimental front wall photoelectrochem- generation, incident photons need to be
ical cells. As is evident, light-transmission localized within the multiple band gap
1.0 0
0.8 20
0.6 40
Transmittance
[%]
0.4 60
0.0 100
400 500 600 700 800 900 1000 1100 1200
Wavelength
[nm]
Fig. 4 Overlay of the optical characteristics of measured through a pathlength of 1 mm. As
the solid and solution phase of the AlGaAs/Si described in the text, the Si bottom cell consists
solid-state and V3+/2+ electrolyte constituents of a p+ -Si, n-Si and n+ -Si multijunction. The
within a bipolar gap photoelectrochemical cell. Al(0.3 – 0.15) Ga(0.7 – 0.85) As top cell utilizes a
Transmission of the V3+/2+ electrolyte is graded band emitter.
366 4 Solar Energy Conversion without Dye Sensitization
semiconductor small and wide band gap GaAs layer. Internally, a bridging GaAs
regions, rather than lost through compet- buffer layer provides an ohmic contact be-
itive electrolyte light absorption. As seen tween the wide band gap AlGaAs junctions
in Fig. 4, the vanadium electrolyte can sig- and the lower Si layers. An intermediate
nificantly block light, over a wide range contact layer indicated as ‘‘Au’’ is used
of visible and near infrared wavelengths, only for probing separated characteristics
from entering the wide and small band gap of the wide and small band gap junc-
layers of the multiple band gap photoelec- tions, and is not utilized in the complete
trochemical cell. This deleterious effect is cell. Photo generated charge at the indi-
prevented by use of the back wall mul- cated silicon electrolyte interface induces
tiple band gap photoelectrochemical cell solution phase vanadium reduction, and
presented in Fig. 5. Light does not pass a carbon counter electrode provides an
through the solution. As shown, illumina- effective (low polarization) electrocatalytic
tion enters directly through antireflection surface for the reverse process in a regen-
films of 50 nm ZnS situated on 70 nm erative cell, in accord with:
MgF2 . An evaporated Au-Zn/Au grid pro-
V3+ (+hν) −−−→ V2+ + h+ ;
vides electrical contact to the wide gap
AlGaAs layers through a bridging p + = V2+ −−−→ V3+ + e− (10)
e−
Au
Au-Zn/Au
p+ - GaAs
300 nm
1.0 µm
1.0 µm
1.7 nm
350 µm
800 nm
50 nm
20 nm
10 nm
AR coating [ZnS(50 nm)/MgF2(70 nm)]
V 3+ + e− → V 2+
V 2+ → V 3+ + e−
(1 × 1018 cm−3)
p+ - Al(0.3 − 0.15)Ga(0.7− 0.85)As (1 × 1018 cm−3)
(1 × 1018 cm−3)
(1 × 1019 cm−3)
(8 × 1015 cm−3)
(4 × 1019 cm−3)
(2 × 1017 cm−3)
Illumination ⇒
Carbon
GaAs (Buffer layer)
n - Al(0.15)Ga(0.85)As
n+Al(0.15)Ga(0.85)As
p+ - Al0.8Ga0.2As
E°,V 2+/3+
n - GaAs
Electrolyte
p+ - Si
n - Si
− 0.3 V vs H2
n - Si
+
Heller, Miller and coworkers have previ- concentration of HF to the electrolyte can
ously shown that the p-Si surface is capable remove this passivating layer and permit
of sustaining minority carrier injection sustained photocurrent. This is illustrated
into the solution and stabilizes reduction in the inset of Fig. 6, in which the addi-
of the V3+/2+ redox couple [19]. However, tion of 0.02 and 0.2 M HF improves the
at n-Si this does not appear to be the case photocurrent stability of this bipolar direct
for reduction of the V3+/2+ redox couple by ohmic MPEC. Also as shown, no further
majority carrier injection into the solution. improvement in the photocurrent stability
Thermodynamically silicon can be sponta- is observed when the HF concentration is
neously oxidized in aqueous solution. This further increased from 0.2 to 0.5 M HF. As
process may occur despite the photoin- presented in the main portion of Fig. 6, the
duced generation of reductive charge at 0.2 M HF modified electrolyte stabilizes
the semiconductor–electrolyte interface. the photocurrent at the n-Si interface for
This is observed as the time-dependent the measured period of several hours [2].
decrease of photocurrent under constant As will be subsequently shown, photocur-
illumination, during photoinduced reduc- rent stability is further improved through
tion of V2+ (by majority carrier injection at use of a electrode catalyst bridging the
n-Si) indicated in the inset Fig. 6. This silicon–electrolyte interface.
is consistent with the onset of surface As can be seen in Eq. (6), maximiza-
passivation forming a layer that is pas- tion of the photopower necessitates
sive to charge transfer and diminishes minimization of the anodic and ca-
the photocurrent. The addition of a low thodic polarization losses, ηanode and
14
12
10 No HF
Photocurrent
[mA cm ]
0.02 MHF
−2
10 0.2 MHF
0.5 MHF
8
Photocurrent stability, n-Si photoanode
in direct contact with solution
8
0 10 20 30 40 50 60
Time
[min]
0 1 2 3 4 5
Time
[h]
Fig. 6 Photocurrent stability in several V3+/2+ aqueous electrolytes of the bipolar band gap
direct ohmic AlGaAs/Si-V3+/2+ photoelectrochemical cell (measured indoors using a tungsten
halogen lamp to simulate outdoor AM 1.5 insolation).
368 4 Solar Energy Conversion without Dye Sensitization
15 15
Bipolar gap direct ohmic photoelectrochemistry
AlGaAs/Si - V2+/3+ electroyte solar cell
FF = 0.81
12 Electrolyte: Aq.0.35 M V(II)+V(III), 4 M HCl, 0.2M HF 12
Electrolyte cathode: 0.2 cm2 n+ - Si
Photocurrent density
Efficiency: 19.1%
15 15
6 6
Photopowder density
Photocurrent density
Photopower density
Insolation: 80 mW cm−2 Top cell
Illuminated area: 0.22 cm2 characteristics
[mA cm−2]
[mW cm−2]
Solar/Electrical conversion
10 Efficiency: 19.0% 6
8 3.2
Photocurrent density
Photopower density
Current FF = 0.74
5 Power 4 1.6
3
Bottom cell
characteristics
0 0 Photovoltage 0.0 0
0.0 0.6
0.0 0.2 0.4 0.6 0.8
Photovoltage
[V]
Fig. 8 Measured outdoor characteristics of the portion of the cell. Main figure: top layer cell
inverted band gap direct ohmic GaAs/Si/I3 − , photocurrent/voltage characteristics. Inset:
3/2I− /Pt MPEC. The top cell consists of the bottom layer cell photocurrent/voltage
GaAs/Pti/I3 − , 3/2I− /Pt portion of the cell. The characteristics.
lower cell consists of the Si/I3 − , 3/2I− /Pt
Tab. 1 Comparison of bipolar solid-state or bipolar regenerative ohmic PECs under solar
illumination
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: 371
losses arising at the solid–solution inter- I3 − /3/2I− portion of the cell under solar
faces. In certain indirect MPEC cases, as illumination.
seen in Table 1, the photoelectrochemical Photoelectrochemical characteristics we-
energy conversion efficiency is comparable re determined under 80 mW cm−2 inso-
to the solid-state cell process. In this case, lation and the generated photocurrent is
combined electrolytic polarization losses highly stable. As shown under illumina-
are less than or equal to the potential drop tion, the Si bottom cell exhibits an open-
from resistance losses in the back contact circuit potential, Voc = 0.51V, a short-
of the solid-state device. circuit photocurrent, Jsc = 7.4 mA cm−2
a fill factor, FF = 0.73, and a maximum
power, Pmax = 2.8 mW cm−2 . Simultane-
4.4.2.3 Inverted Band Gap PECs
ous to the photopower generated by the
Inverted photoelectrochemistry permits
bottom portion of the MPEC, the top
efficient energy, and maintained smaller
(GaAs) portion also generates photopower.
photopotentials than in comparable bipo-
The main portion of Fig. 8 presents the
lar MPECs. To probe inverted semicon-
outdoor characteristics of the top cell
ductor direct ohmic regenerative electro-
(GaAs/Pt/I3 − , 3/2I− /Pt) portion during
chemistry, an electrolyte such as iodide
solar illumination of the complete in-
was driven by a GaAs (EGw = 1.4 eV) wide
verted direct ohmic GaAs/Si/I3 − , 3/2I− /Pt
band gap, and an Si (EGs = 1.0 eV) small
MPEC. In this portion of the cell, GaAs
band gap were utilized [3]. In the inverted
is isolated from the iodide electrolyte via
configuration, bottom and top cells are a Pt electrocatalytic anode. The Pt elec-
utilized simultaneously. Simultaneous to trode is stable in this electrolyte, and
the photopower generated by the bottom hence the top cell portion of the pho-
portion (Si) of the MPEC, the top (GaAs) tocurrent appears to be fully stable. Un-
portion also generates photopower. A plat- der the same 80 mW cm−2 insolation,
inum counter anode, provides an effective photopower is generated simultaneous
(stable, low polarization) electrocatalytic to the photopower presented in Fig. 8.
surface for the reverse process in both As shown under illumination, the GaAs
the top (wide band gap) and bottom (small cell exhibits an open-circuit potential,
band gap) driven regenerative cells. At the Voc = 0.81 V, a short-circuit photocurrent,
n-Si interface, photocurrent stability can Jsc = 21.0 mA cm−2 , a fill factor, and a
be improved with the addition of HF to maximum power, Pmax = 12.4 mW cm−2
an acidic iodide electrolyte, and the elec- at FF = 0.74. The total power generated is
trolyte used was 10.4 M HI, 0.01 M I2 , the sum of the simultaneous extractable
0.8 M HF. Previous studies have shown power generated by each component.
that other electrolytes and surface modi- Table 2 compares inverted solid-state
fications will also enhance photocurrent photovoltaics with direct and indirect re-
stability through the (single band gap) generative ohmic PECs, each containing
silicon–electrolyte interface [19–21]. The a GaAs wide band gap and an inverted
inset of Fig. 8 presents the inverted di- aligned Si small band gap, and a variety of
rect ohmic outdoor characteristics of the redox couples. As in the case of the bipolar
(Si/I3 − /3/2I− ) bottom cell portion, and systems summarized in Table 1, the solid-
the main portion of the figure presents state device exhibits a marginally enhanced
the characteristics of the GaAs driven energy conversion efficiency compared to
372 4 Solar Energy Conversion without Dye Sensitization
Tab. 2 Comparison of inverted solid-state or inverted regenerative ohmic PECs under solar
illumination
Inverted solid-state
bottom: Si 0.552 22.8 3.6 0.74
top: GaAs 0.855 8.7 14.2 0.74
GaAs/Si 17.7 90.2 19.6
Inverted direct iodide
Bottom: Si/I3 − /I− /Pt 0.510 7.4 2.8 0.74
Top: GaAs/Pt/I3 − /I− /Pt 0.813 21.0 12.4 0.73
GaAs/Si/I3 − /I− /Pt 15.2 80.1 19.0
Inverted indirect iodide
Bottom: Si/Pt/I3 − /I− /Pt 0.541 9.2 3.7 0.74
Top: GaAs/Pt/I3 − /I− /Pt 0.866 24.5 14.9 0.70
GaAs/Si/Pt/I3 − /I− /Pt 18.5 94.1 19.7
Inverted indirect sulfide
Bottom: Si/CoS/S2 2− /S4 2− /CoS 0.556 8.9 3.8 0.76
Top: GaAs/CoS/S2 2− /S4 2− /CoS 0.855 24.0 14.9 0.73
GaAs/Si/CoS/S2 2− /S4 2− /CoS 18.7 94.1 19.8
Inverted indirect V2+/3+
Bottom: Si/C/V2+ /V3+ /C 0.523 7.7 3.0 0.73
Top: GaAs/C/V2+ /V3+ /C 0.823 20.8 12.3 0.72
GaAs/Si/C/V2+ /V3+ /C 15.3 80.1 19.2
Note: Cells utilize multijunction wide band gap GaAs layers over smaller band gap Si layers and one
of the indicated pairs of electrolyte or electrocatalyst electrodes, indicated as Direct Iodide: 10.4 M
HI, 0.01, 4 M I2 , 0.8 M HF at Pt; Indirect Iodide: 10.4 M HI, 0.01, 4 M I2 at Pt; Indirect Sulfide 1 M
K2 S2 , 1 M KOH at CoS; Indirect V2+/3+ : 0.35 M V(II)+V(III), 4 M HCl at carbon.
the analogous direct regenerative MPEC. the V2+/3+ indirect regenerative MPEC in
This difference may be attributed to dimin- this table is consistent with the higher
ished charge transfer through the semi- polarization losses of 3 mV cm2 mA−1
conductor–electrolyte interface. However that we have measured for V2+/3+ at
these differences are small and are only carbon, compared to those of less than
slightly larger than those caused by the 2 mV cm2 mA−1 previously measured at
uncertainty of ±1 mW cm−2 in the mea- Pt and CoS, respectively, for the iodide
sured insolation level. Each of the sulfide and sulfide MPECs [2, 3].
or iodide indirect regenerative MPECs
probed in this study exhibit conversion 4.4.2.4 Bipolar Band Gap Solar Storage
efficiency comparable to, or better than, Cells
the solid-state device in Table 2. This indi- One limitation to solar cells is that, while
cates that electrolytic polarization losses most of our electrical needs are contin-
in these cells are comparable to resis- uous, clouds and darkness dictate that
tance losses of the semiconductor back solar energy is intermittent in nature.
contact in the solid-state MPEC. The PECs can generate not only electrical but
marginally lower conversion efficiency of also electrochemical energy, and provide
4.4 Optimizing Photoelectrochemical Solar Energy Conversion: 373
a single device to solve the problem of band gap and electrochemical storage,
the intermittent nature of solar energy. which unlike conventional photovoltaics,
In 1987 we presented a highly efficient provides a nearly constant energetic output
single band gap photoelectrochemical cell in illuminated or dark conditions. The cell
combining in situ electrochemical storage combines bipolar AlGaAs (Eg = 1.6 eV)
and solar conversion capabilities, pro- and Si (Eg = 1.0 eV) and AB5 metal hy-
viding continuous output insensitive to dride/NiOOH storage. The NiOOH/MH
daily variations in illumination. The 1987 metal hydride storage process is near ideal
configuration was demonstrated with a for the AlGaAs/Si because of the excellent
n-Cd(Se,Te)/polysulfide electrolyte conver- match of the storage and photocharg-
sion half cell and a Sn/SnS storage system, ing potentials. The electrochemical storage
resulting in a single cell operating contin- processes utilizes MH oxidation and nickel
uously at an overall efficiency of 11% [18]. oxyhydroxide reduction [22]:
Under illumination, photocurrent droves
an external load. Simultaneously, a portion MH + OH− −−−→ M + H2 O + e− ;
of the photocurrent was used in the direct EM/MH = −0.8 V versus SHE (13)
electrochemical reduction to a metal (Sn)
in the device storage half cell. In darkness NiOOH + H2 O + e− −−−→
or below a certain level of light, the stor- Ni(OH)2 + OH− ;
age compartment delivers power by metal
oxidation. ENiOOH/Ni(OH)2 = 0.4 V vs SHE (14)
Recently, high solar conversion and stor- As shown in Fig. 10, the cell gener-
age efficiencies have been attained with ates a light variation insensitive potential
a system that combines efficient mul- of 1.2–1.3 V at total (including storage
tiple band gap semiconductors, with a losses) solar/electrical energy conversion
simultaneous high-capacity electrochem- efficiency of 18%. Over an eight-month pe-
ical storage [4, 5]. The energy diagram for riod of daily cycles, under constant 12-hour
one of several multiple band gap cells is (AM0) illumination, the long-term indoor
presented in Fig. 3(d), and as described in cycling cell generated a nearly constant
Fig. 2, several other configurations are also photocurrent density of 21.2 (constant to
feasible. In the figure, storage occurs at a within one percent or ±0.2 mA cm−2 ), and
potential of Eredox = EA+/A − EB/B+ . On a photopower that varied by ±3% [5]. The
illumination, two photons generate each cell is a single physical/chemical device
electron, a fraction of which drives a load, that generates load current without any
while the remainder (1/xe− ) charges the external switching.
storage redox couple. Without light, the
potential falls below Eredox and the stor- 4.4.2.5 Bipolar Band Gap Solar Hydrolysis
age couple spontaneously discharges. This (hydrogen generation) Cells
dark discharge is directed through the Solar energy–driven water splitting com-
load, rather than through the multijunc- bines several attractive features for energy
tion semiconductor’s high dark resistance. utilization. Both the energy source (sun)
In Fig. 9 is presented an operational and the reactive media (water) are readily
form of the solar conversion and storage available and are renewable, and the re-
cell described by the Fig. 2 energy diagram. sultant fuel (generated hydrogen) and the
The single cell contains both multiple emission with fuel consumption (water)
374
Fig. 9
2hν
Illumination ⇓
In dark
Au-Zn/Au Au-Zn/Au Photo
+ current
AR coating [ZnS(50 nm)/MgF2(70 nm)] p+-GaAs AR coating [ZnS(50 nm)/MgF 2(70 nm)] p -GaAs
18 −3 18 −3
Rload
Rload
+ 18 −3 + 18
n -Al(0.15)Ga(0.85)As (1 × 10 cm ) 1.7 nm n Al(0.15)Ga(0.85)As (1 × 10 cm−3 ) 1.7 nm
n -GaAS 20 nm n-GaAS 20 nm
←1/xe−
←e
19 −3 n+-Si
n+-Si
−
(4 × 10 cm ) 800 nm (4 × 1019 cm−3 ) 800 nm
←1 − 1/xe−
e− →
Metal hydride anode Metal hydride anode current
MHx + OH− → MHx−1 + H2O + e− MHx−1 + H2O + e− → MHx + OH−
28 28
Bipolar gap in situ storage solar cell
AlGaAs/Si-MH/Ni cell
24 24
IV curve at full charge
Solar to electrical conversion efficiency: 19.6%
20 Illumination: Simulated AMO (135.5 mW cm2), Illuminated area 0.22 cm2 20
1.5
Photocurrent density
Load potential
Photopower density
1.0 Generated cell potential is insensitive
16 [V]
to variation in illumination
16
[mA cm−2]
[mA cm−2]
0.5
Illumination Dark Illumination Dark
12 IPhoto 12
4
IDischarge IDischarge
8 2 8
[mA]
0
I
IDark IDark
4 -2 ICharge 4
0 6 12 18 24 30 36 42 48
0 Time 0
[hours]
contact with the photosensitizer, and com- The subsequent challenge is to optimize
prise either an ohmic or Schottky junction. sustained water electrolysis, without con-
Independent of this interface composition, siderable additional energy losses. Effec-
the various parameters in models predict- tive water electrolysis must occur at a
ing solar water–splitting conversion effi- potential, VH2 O , near the photocell point
ciency, may be combined into two general of maximum power. VH2 O is greater than
parameters: (1) related to losses in optical EHo (=−#GH2 O /nF) as a result of over-
2O
energy conversion, ηphoto , or (2) related to potential losses, ζ , in driving an electroly-
losses in redox conversion of water to H2 sis current density, j , through both the O2
and O2 , ηelectrolysis . Combined, these yield and the H2 electrodes:
an overall solar electrolysis efficiency (ex-
VH2 O (j ) = EO2 (j ) − EH2 (j )
cluding storage and utilization losses) as: % o & o
= EO + ζO2 (j ) − (EH + ζH2 (j ))
ηphotoelectrolysis = ηphoto × ηelectrolysis 2 2
(17)
(15)
A cell containing illuminated AlGaAs/Si
The thermodynamic potential, E o H 2O , for
RuO2 /Ptblack is demonstrated to evolve H2
the water splitting reaction is given by:
and O2 at record solar-driven water elec-
H2 O −−−→ H2 + 1/2O2 ; trolysis efficiency [6]. Under illumination,
o o ◦ bipolar configured Al0.15 Ga0.85 As (Eg =
EH 2O
= EO − EH2 ;
2 1.6 eV) and Si (Eg = 1.1 eV) semiconduc-
o ◦
EH 2O
(25 C) = 1.229 V (16) tors generate open-circuit and maximum
376 4 Solar Energy Conversion without Dye Sensitization
power photopotentials of 1.30 and 1.57 V, cell are available [6]. Under illumination,
well suited to the water electrolysis ther- this bipolar cell generates an open-circuit
modynamic potential of EH o
2O
(25 ◦ C) = potential of 1.57 V, which is considerably
o
1.229 V. The EH2 O /photopotential-mat- larger the thermodynamic potential for
ched semiconductors are combined with the water-splitting reaction. Also included
effective water electrolysis O2 or H2 elec- are the photopotential dependence of so-
trocatalysts, RuO2 or Ptblack . The resultant lar to electrical conversion efficiency. A
solar photoelectrolysis cell drives sustained portion of the high ηphoto domain lies at
water splitting at 18.3% conversion effi- a potential above EH o , and in principle
2O
ciencies. Alternate dual band gap systems is sufficiently energetic to drive efficient
are calculated to be capable of attaining photoelectrolysis.
more than 30% solar photoelectrolysis con- Planar platinum and Ptblack are effective
version efficiency. H2 electrocatalysts. At low current densi-
Figure 11 presents the measured cur- ties (0.5 mA cm−2 ), the observed Pt over-
rent/voltage characteristics for the Al- potential is low (ζH2 = −17 mV), and even
GaAs/Si photocell at AM0 illumination. smaller (ζH2 = −13 mV) with the Ptblack
Further details and characterization of the electrocatalyst, and smaller yet (ζH2 =
Jphoto E oH2O,the
24 minimum
water electrolysis
potential
AlGaAs/Si cell characteristics
20
Area: 0.22 sq.cm
AM0 (135 mW/sq.cm) illumination
Open-circuit potential: 1.57 V
16 Short-circuit current: 5.2 mA
Fill factor: 0.772
[mA sq.cm]
Jphoto
12 ηphoto
Photoelectrolysis conversion
efficiency upper limit
4 = η photo × E °H2O/Vphoto
0
0 200 400 600 800 1000 1200 1400 1600
Vphoto
[millivolt]
Fig. 11 AlGaAs/Si I–V characteristics at AM0 light intensity (135 mW cm−2 ).
ηphoto = 100% × (Jphoto ∗ Vphoto )/Pillumination . Using the indicated, measured ηphoto
and Vphoto , the upper limits of photoelectrolysis efficiency are calculated as
ηphoto × 1.229 V/Vphoto .
4.4 Optimizing Photoelectrochemical Solar Energy Conversion: 377
−7 mV) with a low level of convection portion of Fig. 12 presents the water elec-
(stirring) to improve mass transport and trolysis potential, VH2 O , determined under
prevent observed gas buildup on the elec- stirred or quiescent conditions, and mea-
trode surface. Minimization of ζO2 is a sured in 1 M HClO4 using equal areas of
greater challenge. In this case Pt is a the optimized RuO2 and Ptblack electrodes.
poor electrocatalyst. ζO2 may be decreased These measurements of the electrolysis
∼500 mV, utilizing an RuO2 electrode. current, as a function of potential, enable
The kinetics of this effective oxygen elec- us to predict the AlGaAs/Si photocurrent
trocatalyst [23] have been attributed to at which photoelectrolysis will occur.
catalysis by the intervening RuO4 /RuO2 Our photoelectrolysis cell consists of
redox couple [24, 25]. In the absence of illuminated AlGaAs/Si RuO2 /Ptblack elec-
competing redox couples, the faradaic effi- trolysis. With these active electrocatalysts,
ciency of H2 and O2 evolution approaches high values of ηelectrolysis are insured by
100%, and ηelectrolysis is determined by the using large surface areas of the electrolysis
current limited VH2 O (j ): electrodes, compared to the illuminated
area. This is accomplished without in-
o
ηelectrolysis = EH /VH2 O (j ); creasing the illuminated electrode area,
2O
◦
as schematically represented in the lower
ηelectrolysis (25 C) = 1.229V/VH2 O (j ) portion of Fig. 13, by utilizing a large
(18) vertical depth of electrolysis electrodes
The Fig. 12 inset contains the Eq. (18)–de- compared to the cross section of illu-
termined ηelectrolysis . The limiting maxi- mination. Specifically, the 10 cm2 Ptblack
mum ηphotoelectrolysis can be readily de- and RuO2 electrodes utilized are large
termined from the solar to electrical compared to the 0.22 cm2 illuminated
conversion efficiency. Expanding Eq. (15) area. In the Fig. 13(a) is the H2 and O2
with Eq. (18), ηphotoelectrolysis is diminished electrolysis current generated by this cell
from ηphoto by the potential of the stored as a function of time. The average pho-
energy compared to the potential at which tocurrent of 4.42 mA (generating a current
the water electrolysis occurred: density of 20.1 mA cm−2 at the illumi-
nated electrode area, and 0.44 mA cm−2 at
ηphotoelectrolysis (T) the electrolysis electrodes) corresponds to a
ηphoto (T) × EH2 O (T) photopotential of 1.36 V in Fig. 11 compa-
= ; rable to the equivalent electrolysis potential
EO2 (T) − EH2 (T)
at 0.5 mA cm−2 in Fig. 12. The overall effi-
ηphotoelectrolysis (25 ◦ C) = 1.229V × ηphoto ciency is determined by the 1.229 V energy
VH2 O stored as H2 and O2 and the incident pho-
(19) topower (135 mW cm−2 ):
Figure 12 includes these limiting
ηphotoelectrolysis values because the Al- ηphotoelectrolysis = 100% × 20.1 mA cm−2
GaAs/Si system is determined from mea- × 1.229 V/135 mW cm−2 = 18.3% (20)
sured values of ηphoto at various values of
Vphoto . AlGaAs/Si solar photoelectrolysis 4.4.2.6 Higher Solar Production Rates of
at conversion efficiencies exceeding 20% Hydrogen Fuel are Attainable
(Fig. 11) are in principle possible at poten- In Eq. (15), the significance of the elec-
tials approaching the EH2 O limit. The main trolysis compared to photo components
378 4 Solar Energy Conversion without Dye Sensitization
100 20
90
80
ηelectrolysis
16
[%]
70
60
12
[mA sq.cm]
50
Jelectrolysis
1200 1600 2000 2400
Velectrolysis
[mv] 8
4
Electrolysis in 1 M HClO4 Stirred
Anode: Pt black on Pt mesh
Quiescent
Cathode: 2.3 µm RuO2
0
1200 1300 1400 1500 1600
Velectrolysis
[millivolt]
Fig. 12 Measured variation of the VH2 O with current density in 1 M HClO4 , using
equal area 2.3 µm RuO2 and Ptblack . Inset: Calculated ηphoto , in a 100%
faradaic-efficient process, as a function of VH2 O .
hν
Cross Illumination ⇓
hν section AR coating [ZnS(50 nm)/MgF2(70 nm)]
expanded P+-AI0.8Ga0.2As 50 nm
P+-AI(0.3−0.15)Ga(0.7−0.85)As 300 nm
P+-GaAs Au-Zn/Au
[mA]
3 n -AI(0.15)Ga(0.85)As 1.0 µm
n+-AI(0.15)Ga(0.85)As 1.7 nm
n -GaAs 20 nm
GaAs (Buffer Layer) 10 nm
RuO2
Ptblack
2
+
p -Si 1.0 µm
n -Si 350 µm
n+-Si 800 nm
Ru02
Au − Sb/Au
Ptblack
1
1M HCIO4
0
2 4 6 8 10 12
Time
[hours]
Fig. 13 Schematic representation (inset) and measured characteristics of the illuminated
AlGaAs/Si RuO2 /Ptblack photoelectrolysis cell. Further details of the layered AlGaAs/Si structure
are given in Ref. 6.
secondary losses that include: 10% re- such cells in series will efficiently drive
flection loss, 10% quantum-yield loss, three 1.3 to 1.4 V water electrolysis cells
20% absorption loss. As summarized in in series, and as discussed here at a wa-
Table 3 demonstrated ηphoto are substan- ter electrolysis at 1.36 V (Fig. 12 inset)
tially higher than 20%. Many ηphoto that yielded electrolysis efficiencies of more
have been carefully reviewed or monitored than 90%. It is reasonable that with larger
are in excess of 30% [29]. surface area, or more effective electrocatal-
Each of the cells in the Table 3 exhibits ysis, these efficiencies will approach 95%.
an open-circuit photopotential signifi- Using this range of ηelectrolysis = 90–95%:
cantly greater than the minimum potential
ηphotoelectrolysis (predicted maximum)
necessary to split water. The majority of
= ηphoto × 90–95% (21)
these cells can generate a photopotential
in excess of 2 volts. An unnecessary limit Table 3 includes predicted maximum
of one multiple band gap photoexcitation ηphotoelectrolysis using observed ηphoto of var-
per electrolysis would under-utilize Vphoto , ious dual band gap sensitizers. It is seen
diminishing ηelectrolysis . For example, a that solar water splitting efficiencies may
GaInP/GaAs cell has a maximum power be viable at up to double the amount of that
photopotential of 2.0 to 2.1 V and an previously predicted. Efficient, three or
open-circuit potential of 2.3 V [12]. Two more multiple band gap photoelectrolysis
380 4 Solar Energy Conversion without Dye Sensitization
Tab. 3 Predicted and measured photoelectrolysis efficiencies. Calculated ηphotoelectrolysis are from
Eq. (21)
Load
Load
S2− HxFe(CN)6(x − 3)
S3= Se3=
KyFe(CN)6(y − 3) Se4=
S4= HxKyFe(CN)6(y + x − 3)
S5= OH−
CN−
H+
OH− OH−
H+ H+
Counter electrode Counter electrode Counter electrode
(a) Aqueous polysulfide PEC (b) Aqueous ferrocyanide PEC (c) Aqueous polyselenide PEC
Fig. 14Chemical species coexisting in polysulfide (a), ferrocyanide (b) and polyselenide (c)
aqueous electrolytes used in regenerative PECs.
Photocurrent
1.6 bottom curve is in 1 M NaOH, 1 M
7.7% efficiency Na2 S, 1 M sulfur. The photocurrent
[mA]
voltage curves were obtained outdoors,
and solar to electrical conversion
0.8 efficiencies are indicated.
~AM1 illumination
Voltage
[mV]
0.5
0.4
Photovoltage over load
0.3
[Volts]
0.2
0.1
10 Photoelectrochemistry of n -CdSe in
[Fe(CN)5 -L]3−/2− aq
L
a
Photocurrent
a) −CN−
[mA cm−2]
5 b b) −NH3
c) −NO2−
c
d d) −Fe(CN)53−
e e) −NO+
0
15
Current Density
Iphoto (normalized)
[mA cm−2]
1 θ
10
d
b
5 0
0 1 Time 3
[days]
No Illumination
0
19
J, Stable
18
J, Photocurrent density
J, Fluctuating
[mA cm−2]
(Corrosion)
J, Diminishing
17
16
15
0 1 2 3 4
KOH
[M]
Fig. 19 Short-circuit photocurrent density, Jsc , for a 75 mW cm−2
tungsten-halogen illuminated single crystal n-GaAs immersed 1 M K2 Se,
0.01 M Se and in varying KOH concentration.
386 4 Solar Energy Conversion without Dye Sensitization
3.0
0.01 M Se
2.5 0.02 M Se
0.20 M Se
2.0
[mW cm−2]
1.5
Power
1.0
0.5
0.0
has been described in one of two manners: To a lesser extent the solution phase
n-WSe2 species IO3 − , HIO, H2 O I+ , I5 − , and
◦
I− + hn −−−−→ I + e− (28) I6 2− can also occur. Figure 21 presents the
relative variation of polyiodide speciation
or: in solutions containing 0.01 molal iodine
n-WSe2 and 1 to 12 molal iodide and which may
3I− + 2hn −−−−→ I3 − + 2e− (29) be compatible with n-WSe2 regenerative
photoelectrochemistry. As emphasized in
Significant species in aqueous polyiodide the figure inset, in these electrolytes the
solution can include: I− , I3 − , I4 2− , I− , concentration of added iodide dominates
OH− , and H+ . Constrained by pH and the iodine in these solutions. The distribution
equilibria [67, 68]: of species in these solutions is calculated in
accordance with Eqs. (30) and (31). As can
I− + I2 ←−→ I3 − K3 = 723 (30)
be seen in the figure inset, little (less than
I3 − + I− ←−→ I4 2− K4 = 0.20 (31) 0.3%) of the added iodine exists as solution
100%
80%
Total iodine = 0.01 m = [I2] + [I3− ] + [I 4= ]
60%
Percent of total
40%
0.3%
[I2] / Total iodine
% of total
0.0%
0 4 8 12
NaI
[m]
0%
0 2 4 6 8 10 12
NaI
[molal]
−
Fig. 21 The fractions of I3 − and I4 − (compared to total iodine) in aqueous polyiodide
solutions as a function of added iodide concentration. Inset: The fractions of solution
phase iodine (compared to total iodine) and total iodine (compared to total iodide).
388 4 Solar Energy Conversion without Dye Sensitization
phase iodine, and the maximum [I2 ] is These improvements are sustained in the
(1 × 10−4 m. As seen in the figure outset, silver-bearing electrolytes. On returning
the bulk (greater than 99.7%) of the added the electrode to the polyiodide electrolyte
iodine resides as the polyiodide species without silver, the PEC gradually (on the
I3 − and I4 2− . In iodide concentrations less order of hours) returns to the silver free
than 5 m, I3 − is the predominant polyio- PEC behavior. This improvement appears
dide, whereas at higher iodide concentra- to provide a mechanism for long-lasting
tions, I4 2− predominates. Longer chain suppression of n-WSe2 exposed edges and
species, I5 − and I6 2− , if they exist, have recombination sites as further discussed
upper limit concentrations of 1 × 10−6 m by Licht and Myung [67, 68].
and 1 × 10−8 m, respectively. Hence, in
these aqueous polyiodide electrolytes, I2 , 4.4.4
I5 − and I6 2− are not at substantial concen- Concluding Remarks
trations in these photoelectrolytes, and a
primary oxidizable ion is I− and a primary In the 1970s through 1990s health con-
reducible ion is either I3 − or I4 2− . cerns (air pollution) and political concerns
Choice of cation in aqueous polyiodide (localized shortages) were cited in the
solution can effect n-tungsten dichalco- need for development of alternative en-
genide photoelectrochemistry. The alkali ergy sources. Today, in addition to these
cations do not substantially interact with concerns, growing awareness of carbon
solution phase iodide. However, other
dioxide emissions as a green house gas,
metal cations including Ag+ , or Zn2+ or
as well as the economic realization that
Cd2+ , will create a series of complexes
oil and coal are better allocated as raw
with dissolved iodide including: AgI2 − ,
materials for pharmaceuticals and plas-
AgI3 2− , ZnI3 − , ZnI4 2− , CdI4 2 and so
tics, than for electrical energy generation,
on. These complexes can shift rest po-
provide impetus to the technological de-
tentials or either enhance or diminish
velopment of renewable energy sources.
charge transfer from the semiconductor
Solar energy remains the principal sym-
surface. Because of the limitations on sil-
ver iodide solubility (Ksp (AgI) = 10−16 ) bol of a clean, abundant renewable energy
high (molal) level concentrations of silver source. Society’s energy needs are continu-
will dissolve only in solutions that facili- ous. Unlike purely solid-state, photovoltaic
tate complex formation as in concentrated processes, the advantage of combined
NaI or KI solutions. In particular, we have solar/electrochemical processes is that en-
shown that the presence of high concentra- ergy is not only converted but also may be
tions of dissolved silver is advantageous to stored for future use when sunlight is not
n-tungsten diselenide photoelectrochem- so frequently available.
istry. As seen in Fig. 22, silver dissolved as Limiting constraints of multiple band
AgNO3 enhances the voltage at maximum gap photoelectrochemical energy conver-
point. The extent of the improvement sion, as well as practical configurations for
varies with the initial condition of the efficient solar to electrical energy conver-
individual n-WSe2 crystal, and general im- sion have been probed [1–6]. Such systems
provements are a 15 to 45 mV increase are capable of better matching and utiliz-
in the voltage of maximum power, Vmax , ing of incident solar radiation (insolation).
and a 5 to 20% relative increase in power. Efficient solar cells, solar storage cells,
4.4 Optimizing Photoelectrochemical Solar Energy Conversion: 389
Illuminated n-Wse2
in 6 m iodide, 0.01 m iodine
1.0
0.8
Relative power
0.6
0.4
0.2
0.0
in Nal in Kl
Without silver Without silver
With 0.7 m silver nitrate With 1.0 m silver nitrate
27. J. R. Bolton, S. J. Strickler, J. S. Connolly, 48. S. Licht, J. Phys. Chem. 1986, 90, 1096.
Nature 1985, 316, 495. 49. J. Reichman, M. Russak, J. Electrochem. Soc.
28. M. F. Weber, M. J. Digman, Int. J. Hydrogen 1984, 131, 796.
Energy 1986, 11, 225. 50. K. W. Freeze Jr., Appl. Phys. Lett. 1982, 40,
28a. M. F. Weber, J. Electrochem. Soc. 1984, 131, 275.
1258. 51. H. D. Rubin, B. D. Humphrey, A. B. Bo-
29. M. A. Green, K. Emery, K. Bucher et al., carsley, Nature 1984, 308, 339.
Prog. Photovolt. 1999, 11, 31. 52. H. D. Rubin, D. J. Arent, A. B. Boc-
30. R. C. Kainthla, B. Zelenay, JO’M. Bockris, J. arsley, J. Electrochem. Soc. 1985, 132,
Electrochem. Soc 1986, 133, 248. 523.
31. G. Hodes, J. Manassen, D. Cahen, Nature 53. S. Licht, D. Peramunage, Nature 1991, 354,
1976, 261, 403. 440.
32. A. B. Ellis, S. W. Kaiser, M. S. Wrighton, J. 54. S. Licht, D. Peramunage, J. Electrochem. Soc.
Am. Chem. Soc. 1976, 98, 1635. 1992, 139, L23.
33. B. Miller, A. Heller, Nature 1976, 262, 680. 55. S. Licht, D. Peramunage, J. Electrochem. Soc.
34. S. Licht, G. Hodes, J. Manassen, Inorg. 1992, 139, 1791.
Chem. 1986, 25, 2486. 56. S. Licht, D. Peramunage, Sol. Energy 1994,
35. S. Licht, J. Manassen, J. Electrochem. Soc. 52, 197.
1987, 134, 918. 57. G. Seshardi, J. K. M. Chun, A. B. Bocarsly,
36. S. Licht, J. Electrochem. Soc. 1988, 135, 2971. Nature 1991, 352, 508.
37. F. Forouzan, K. Longo, D. Peramunage 58. B. A. Parkinson, A. Heller, B. Miller, J. Elec-
et al., Anal. Chem. 1990, 62, 1356. trochem. Soc. 1979, 126, 954.
38. K. Longo, F. Farouzan, D. Peramunage 59. B. J. Tufts, I. L. Abrahams, P. G. Santengelo
et al., J. Electroanal. Chem. Interfacial Elec- et al., Nature 1987, 326, 861.
trochem. 1991, 318, 111. 60. L. G. Casasgrand, B. J. Tufts, N. S. Lewis, J.
39. D. Peramunage, F. Forouzan, S. Licht, Anal. Phys. Chem. 1991, 95, 1373.
Chem. 1994, 66, 378. 61. D. E. Levy, R. J. Myers, J. Phys. Chem. 1990,
40. W. Giggenbach, Inorg. Chem. 1971, 10, 94, 7842.
1333. 62. S. Licht, F. Forouzan, J. Electrochem. Soc.
41. A. Teder, Sr. Papperstodn 1969, 72, 245. 1995, 142, 1539.
42. S. Licht, Joost Manassen, J. Electrochem. 63. S. Licht, F. Forouzan, J. Electrochem. Soc.
Soc. 1985, 132, 1077. 1995, 142, 1546.
43. S. Licht, G. Hodes, Joost Manassen, J. 64. W. Kautek, H. Gerischer, H. Tributsch, J.
Electrochem. Soc. 1986, 133, 272. Electrochem. Soc. 1981, 128, 2471.
44. S. Licht, J. Manassen, J. Electrochem. Soc. 65. B. A. Parkinson, A. Heller, B. Miller, Appl.
1986, 133, 277. Phys. Lett. 1978, 33, 521.
45. S. Licht, R. Tenne, H. Flaisher et al., J. 66. R. Tenne, A. Wold, Appl. Phys. Lett. 1985,
Electrochem. Soc. 1986, 133, 52. 47, 707.
46. S. Licht, R. Tenne, H. Flaisher et al., J. 67. S. Licht, N. Myung, J. Electrochem. Soc.
Electrochem. Soc. 1984, 131, 5920. 1985, 142, 845.
47. S. Licht, R. Tenne, G. Dagan et al., Appl. 68. S. Licht, N. Myung, J. Electrochem. Soc.
Phys. Lett. 1985, 46, 608. 1995, 142, L129.
397
5.1 5.1.2
Dye-Sensitized Regenerative Solar Cells Device Concepts
..
Augustin J. McEvoy and Michael Gratzel It should be remembered that the first
Ecole Polytechnique Fédérale de Lausanne, observation of the photovoltaic effect
Lausanne, Switzerland by A. -E. Becquerel [1] in 1839 was at
a solid–liquid interface, with a semi-
5.1.1 conductor photoelectrode contacting an
Overview electrolyte, and was technically a pho-
toelectrochemical cell. Nevertheless, the
At present, dye-sensitized photosystems modern photovoltaic industry is estab-
provide the only technically and econom- lished exclusively on solid-state devices,
ically credible alternative to solid-state with semiconductors of appropriate band
photovoltaic devices. The concept recon- gap as optical absorbers and junctions to
ciles the electrochemical stability of wide solids with different conduction mecha-
band gap photoelectrodes with the ef- nisms as the sites of charge separation and
ficiency of photovoltaic devices having photovoltage generation. In commerce are
a more extended absorption spectrum found homojunction devices, in which the
in the visible range. Optical absorption semiconductors are chemically identical,
and charge separation take place on dis- differing only in impurity content and
therefore in free carrier polarity; hetero-
tinct sites within these photovoltaic cells.
junctions, in which the semiconductors
Hence, oppositely charged species are re-
of opposite carrier polarity are chemically
stricted to separate phases, so conventional
different, for example, cadmium sulfide or
recombination losses are inhibited. In copper indium-gallium selenide contacts;
consequence, device photoconversion ef- and finally Schottky junctions in which
ficiency is better maintained at low light the contacting phase is a metal. However,
levels than with conventional semicon- all these are solid-state devices in which
ductor solid-state junction devices. Vari- conduction throughout is by an electronic
ous configurations of the dye-sensitized mechanism. None is a photoelectrochem-
concept are under investigation, includ- ical device that has an electrolyte as the
ing electrochemical and heterojunction contacting phase or that uses an ionic
variants. conduction mechanism. Awareness of
398 5 Dye-Sensitized Photoelectrochemistry
Conducting
glass
TiO2 Dye Electrolyte Cathode
E /V vs NHE
Injection
S*
−0.5
Maximum
0 voltage
hν
0.5 Red ox
Mediator
Interception
Diffusion
1.0 S°/S+
junction analog. A typical wide band gap was not realistic of the visual impres-
semiconductor for photoelectrochemistry sion of the same scene. So many early
with stability against photocorrosion is photographs therefore give an impres-
titanium dioxide. Titanium dioxide has a sion of an unnatural ‘‘cleanliness’’ of
band gap of 3.1 eV and is therefore in- the scene imaged that considerable ef-
sensitive to visible light. The selected dye forts were expended to find emulsions
must have a HOMO-LUMO gap corre- capable of more realistic response. The
sponding to a photon energy in the visible problem is now recognized to be due to
or infrared parts of the spectrum, with an the semiconductor nature of the halide
optimum near the solar spectra maximum grains responsible for optical absorption
of 1.4 eV. This mechanism is illustrated in photography, which are activated es-
schematically in Fig. 1. However, before sentially by the photoelectric effect. They
presenting the relevant characteristics of have band gaps typically in the range 2.7
a semiconductor electrode that is apt for to 3.2 eV, and the photoresponse is ef-
chemisorption of a dye and therefore pho- fectively negligible for wavelengths longer
tovoltaic application, a background of the than 460 nm. It was however noted that the
history and development of the sensitiza- use of gelatine as a support medium for
tion of semiconductors is necessary. the alkali halide grains in the photographic
emulsion could significantly modify the
5.1.3 film spectral response. Only in the last
Photography and Photoelectrochemistry century was it finally demonstrated that
an organo-sulfur compound present in
Although artisanal ‘‘virtual reality’’ by calf-skin gelatine was responsible for this
painting is as old as human culture, extension of sensitivity [3]. The mecha-
synthetic ‘‘virtual reality’’ by technical nism now recognized is the induction
reproduction of images is relatively recent, of a nanostructure of silver sulphide on
with the first photography by Daguerre the halide surface of each grain, giving
in 1837 and the introduction of the sil- the first application of the sensitization
ver halide process by Fox-Talbot in 1839. of a semiconductor heterojunction! Even
Thereafter, the art advanced so rapidly more significant in the history of sensiti-
that the nineteenth century events – the zation is the work of Vogel, Professor of
wars, the explorations, the early factory age photochemistry, spectroscopy and photog-
and the introduction of steam power – are raphy at the then Königliche Technische
vividly documented by photographs. Pho- Hochschule of Berlin, who in the 1870s in-
tography remained an art for a century, vestigated the effect of dyes on emulsions
becoming a science only with the theo- and found that the halide spectral response
retical analysis of the process by Gurney extended through the visible and even into
and Mott [2] in 1938. Nevertheless, an the infrared region [4]. Not only did this
empirical skill in the formulation of pho- facilitate his spectroscopic research and
tographic emulsions had evolved in the contribute to astronomy but also made
interval, even in the absence of a funda- possible the commercial ‘‘panchromatic’’
mental understanding. Early emulsions, broad-spectrum black-and-white film with
referred to as ‘‘orthochromic,’’ had a dis- realistic gray scale reproduction, and later
tinct insensitivity to midspectral and red with spectrally selective dyes – the modern
light, so that their gray scale reproduction era of color photography.
400 5 Dye-Sensitized Photoelectrochemistry
I I
proposed, as illustrated in Fig. 3 taken
from the conference proceedings [9]. The Conduction band e
understanding became definitive with the
hν
work of Gerischer and Tributsch on sen- D
Br −
sitization of zinc oxide [10]. However, the Br+ Dye
fact remained that sensitization efficiency hν
remained low, as research systematically
investigated dyes in solution or those Dye
adsorbed on single-crystal substrates. In Valence band
Energy Electron
BrAg
retrospect, it appears that there was an transfer transfer
excessive influence of solid-state engineer-
ing practice in that the disordered surfaces modifications of certain side groups.
were expected to promote surface recombi- Chlorophyll is in turn similar in struc-
nation and therefore loss of excited charge ture to hemoglobin, the oxygen-carrying
carriers, whereas single-crystal surfaces, iron-based pigment found in blood. Given
being better defined physically and chem- that the development of the dye-sensitized
ically, make the effects of sensitization cell arose out of an interest in artificial pho-
more evident. The results of this presump- tosynthesis, the adoption of porphyrin-like
tion were slow to emerge. organometallic dyes as sensitizers was log-
ical. However, although nature confines
5.1.4 itself to magnesium and iron for its princi-
Sensitization of Powders and Rough pal pigments, the synthetic chemist can
Surfaces access the whole range of metallic el-
ements. The use of ruthenium pyridyl
complexes now has more than twenty years
With the ongoing interest in sensitization
of development history. Clark and Sutin
during the 1970s, and the idea that
had already used a trisbipyridyl ruthenium
photocatalysts could possibly carry out
complex in 1977 to sensitize titanium diox-
electrolytic reactions such as the evolution
ide to sub–band gap illumination [12], but
of hydrogen from water, a phase of experi-
in solution only. Charge transfer could
mental development evolved with the sen- only occur after diffusion of the ion to the
sitization of semiconductor powders [11]. semiconductor, so the efficiency of the sen-
With the ecological awareness of time, sitization was very low. By 1980, the idea of
photochemical production particularly of chemisorption of the dye through an acid
energy-related chemicals was seen as a carboxylate group bonding to the metal ox-
form of ‘‘artificial photosynthesis.’’ The ide surface had also been established [11]
prototype energy-converting dye provided so that the sensitizer was immobilized
by nature is indeed chlorophyll, a molecule and it formed the required monomolecular
consisting of a central magnesium atom film on the semiconductor substrate which
surrounded by a nitrogen-containing por- facilitated charge transfer by electron in-
phyrin ring. Variations are due to minor jection. The carboxylated trisbipyridyl dye
402 5 Dye-Sensitized Photoelectrochemistry
(‘‘RuL3 ’’) therefore became the prototype of nanoporous ceramic films is the basis
sensitizer in photochemistry. The subse- for the understanding of semiconductors.
quent development of sensitizers through The nanoporous structure permits the
to the panchromatic ‘‘black dye’’ that con- specific surface concentration of the sen-
stitutes the current state of the art is dealt sitizing dye to be sufficiently high for the
with in detail in a separate chapter. total absorption of the incident light that is
As a substrate for the chemisorption necessary for efficient solar energy conver-
of sensitizing dyes, titania, TiO2 , par- sion, since the area of the monomolecular
ticularly in the anatase form, gradually distribution of adsorbate is 2 to 3 orders of
became dominant [13]. It has many advan- magnitude higher than the geometric area
tages for sensitized photochemistry and of the substrate. Moreover, in the elec-
photoelectrochemistry, being a low-cost, trochemical system, this high degree of
widely available, nontoxic and biocompat- roughness does not promote charge carrier
ible material, and as such is even used loss by recombination, since the electron
in health care products and domestic
and the positive charge find themselves
applications such as paint pigmentation.
within picoseconds on opposite sides of
The objective had also evolved to concen-
the liquid–solid interface. In addition, the
trate on photovoltaic devices rather than
photoexcited carrier in the n-type TiO2 is
on photosynthesis. Progress thereafter was
an electron, a majority carrier, and there-
incremental until the announcement in
fore not subject to depletion by carriers of
1991 [14] of the sensitized electrochemical
the opposite polarity. Solid-state junction
solar cell, with the remarkable conversion
efficiency at that time of 7.1% under so- cells, on the other hand, are minority car-
lar illumination, a synergy of structure, rier devices. The carrier loss mechanisms
substrate roughness, dye photochemistry, are comparatively slow [16], and although
counterelectrode kinetics and electrolyte conventionally referred to as recombina-
redox chemistry. That evolution has con- tion by analogy with the solid-state process,
tinued progressively since then, with certi- the loss of a photoexcited electron from the
fied efficiency now more than 10% [15]. semiconductor should rather be regarded
There has been a continuous develop- as a recapture of the electron by an oxidized
ment of the electrolyte solvent, aqueous dye species, or a redox capture, when the
systems being eliminated in favor of or- electron reacts directly with the iodine in
ganics and molten salts. However, finding the electrolyte. Either can occur on a mil-
substitutes for the original iodine-iodide lisecond timescale, but much slower than
redox system has proved to be difficult electron injection or transport to the rear
although recent work in our laboratory contact layer.
shows that alternative redox couples are The original substrate structure used
finally at hand. for our early photosensitization experi-
ments was a fractal derived by hydrolysis
5.1.5 of an organo-titanium compound but it
Substrate Development and Fabrication has since been replaced with a nanostruc-
ture deposited from colloidal suspension.
If molecular design and engineering has This evidently provides a much more re-
underpinned the evolution of efficient, producible and controlled porous high
stable sensitizer dyes, the materials science surface area nanotexture. Further, since
5.1 Dye-Sensitized Regenerative Solar Cells 403
it is compatible with screen-printing tech- the electron loss by exchange with the re-
nology, it anticipates future production dox system in solution. In the analogous
requirements. nonliquid electrolyte case, it is not im-
While commercially available titania mediately evident that an interpenetrating
powders, produced by a pyrolysis route network of two conducting solids can be
from a chloride precursor, have been suc- so easily established that an immobilized
cessfully employed, the present optimized molecule at their interface can exchange
material is the result of a procedure de- charge carriers with both. However, ini-
scribed by Brooks and coworkers [17]. A tial results [18,19] are promising. In both
specific advantage of the hydrothermal cases, the charge transport materials are
technique is the ease of control of the par- deposited by spin coating from the liquid
ticle size and hence of the nanostructure phase in order to achieve the necessary
and porosity of the resultant semiconduc- intimate contact. In the latter case, a mix-
tor substrate. The relevant preparation flow ture of polymers was used, which phase-
diagram is given and the product is illus- separates spontaneously on removal of a
trated by the accompanying micrograph solvent, whereas the Lausanne laboratory
(Fig. 4; Table 1). Figure 5 presents data on introduces a solution of the conducting
the control of substrate porosity by the compound into a previously sensitized
powder-preparation parameters. nanostructure. The charge transfer ma-
terial currently used is a spirobifluorene,
5.1.6 proprietary to Hoechst [20] as shown in
Dye Sensitization in Heterojunctions Fig. 6. As a matter of technical precision, if
these materials function in the cell as a hole
Because the sensitizing dye itself does
conductor, the device is a sensitized nanos-
not provide a conducting functionality
tructured heterojunction in which the
but is distributed at an interface in the
charge transport is entirely of electronic
form of immobilized molecular species,
nature (Fig. 7). However, if the molecules
it is evident that for charge transfer,
accept positive charge to become cations,
each molecule must be in intimate con-
for which there is initial spectroscopic evi-
tact with both conducting phases. It is
dence [18], the charge transfer mechanism
clear that this applies to the porous
within this organic phase can be consid-
wide band gap semiconductor substrate
ered a redox equilibration, and the device
into which the photoexcited chemisorbed
molecules inject electrons. It is also ev-
ident that in the photoelectrochemical
form of the sensitized cell, the liquid
electrolyte penetrates into the porosity,
thereby permitting the intimate contact
with the charged dye molecule that is
necessary for charge neutralization after
Tab. 1 Flow diagram for the preparation of TiO2 colloids and mesoporous films [17]
Screen-print films
9000
8000 190 °C
210 °C
7000 230 °C
[(cm3/g)/nm] [10−5]
Differential volume
250 °C
6000 270 °C
5000
4000
3000
2000
1000
0
0 10 20 30 40 50 60 70 80
Pore diameter
[nm]
Fig. 5 Control of semiconductor substrate porosity by temperature of
hydrothermal processing.
5.1 Dye-Sensitized Regenerative Solar Cells 405
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.
properties of these complexes can be tuned However, the photophysical and the
in a predictable way; (3) it possesses sta- electrochemical properties of its coordi-
ble and accessible oxidation states from I nation complexes are difficult to tune in
to IV; (4) it forms very inert bonds with an expected fashion [25]. The other no-
imine nitrogen centers [23, 24]. table disadvantage of this metal is weaker
Iron (Fe) that is in the first row ligand field splitting compared to ruthe-
transition metal of the periodic table, nium and osmium (Os). On the other
and the same group as ruthenium, hand, osmium being in the third row tran-
is an inexpensive and abundant metal. sition metal of the periodic table has a
Load
e− e−
S+/S∗
e− e−
−1.0
cb
e−
e− e− (e−)
−0.6
e−
∆v −0.2
0
(R / R−)
TiO2 0.2
e−
Electrolyte
0.6
S+/S
1.0
HOOC
N
N NCS
= Ru
Anchored dye N NCS
N
HOOC
HOOC
Fig. 1 Operating principles and energy level diagram of dye-sensitized
solar cell. S/S+ = Sensitizer in the ground and oxidized state;
S∗ /S+ = Sensitizer in the excited state R/R− = Redox mediator.
5.2 Dyes for Semiconductor Sensitization 409
stronger ligand field splitting compared conduction band and the redox couple
to ruthenium. Moreover, the spin-orbit because of differences in recombination
coupling in osmium complexes leads to rates [28].
enhanced spectral response in the red re-
gion [26]. Nevertheless, the low abundance IPCE =
of this metal restricts its use for large-scale photocurrent density
(1.25 × 103 ) × −2
applications. (mA cm )
photon flux
wavelength (nm) × −2
(W m )
5.2.1.2 Operating Principles of the
Dye-Sensitized Solar Cell (1)
The details of the operating principles of
the dye-sensitized solar cell are given in 5.2.2
Molecular Sensitizers
Fig. 1. The photo-excitation of the metal-
to-ligand charge transfer (MLCT) of the
5.2.2.1 Formation of Complexes
adsorbed dye leads to injection of elec-
One can view the formation of a complex
trons into the conduction band of the
between metal ion and ligands as an
oxide. The original state of the dye is sub-
electrostatic attraction between positively
sequently restored by electron donation
charged metal and negatively charged ions
from an electrolyte, containing the re-
or the negative ends of the dipoles of
dox system (iodide/triiodide). The injected
neutral ligands. A metal ion or atom can
electron flows through the semiconduc-
act as a discrete center about which a set of
tor network to arrive at the back contact
ligands is arranged in a definite way. The
and then through the external load to the
general concept of a ligand (unidentate) is
counter electrode. At the counter electrode,
donating of an electron pair to the central
reduction of triiodide in turn regenerates
metal atom. The number of ligands per
iodide, which completes the circuit.
metal center is generally either four or six,
with others being rare. The approach of the
5.2.1.3 Incident Photon to Current negative, or the neutral, ligands toward a
Efficiency and Open-Circuit Photovoltage charged metal repels electrons residing in
The incident monochromatic photon-to- d-orbitals and raises their energies both
current conversion efficiency (IPCE), de- with respect to those residing on the
fined as the number of electrons generated ligands and the metal.
by light in the external circuit divided In an octahedral ligand field, the five
by the number of incident photons as a degenerate d-orbitals split into degenerate
function of excitation wavelength is ex- t2g (dxy , dxz , dyz ), and eg (dx2 −y2 , dz2 ) sets
pressed in the Eq. (1) [27]. The open-circuit of orbitals. The splitting value increases
photovoltage is determined by the en- by about 40% as one moves from the first
ergy difference between the fermi level row (3d) to the second (4d) and third (5d)
of the solid under illumination and the row transition metal ions [29]. A schematic
Nernst potential of the redox couple in the representation for splitting of d-orbital
electrolyte (Fig. 1). However, the experi- and the position of ligand orbitals in
mentally observed open-circuit potential an octahedral complex containing three
using various transition metal complexes bidentate 2,2$ -bipyridyl ligands is shown
is smaller than the difference between the in Fig. 2.
410 5 Dye-Sensitized Photoelectrochemistry
π*
π* π*
Energy
eg
d t2g
MC
eg
1
p LC MLCT
π*
ISC
s
3MLCT
d Em
t2g
LMCT
σ
Metal orbitals Molecular orbitals Ligand orbitals
Fig. 3 Schematic presentation of a molecular orbital diagram for an
octahedral d6 metal complex involving 2,2$ -bipyridyl type ligands in which
various possible transitions are indicated.
sustain at least 108 redox turnovers under of ruthenium polypyridyl complexes can
illumination corresponding to about 20 be tuned in two ways. First, by introducing
years of exposure to natural sunlight at Air a ligand with a low-lying π ∗ -molecular or-
Mass (AM) 1.5. bital and second, by the destabilization of
On the basis of extensive screening the metal t2g orbitals with a strong donor
of hundreds of ruthenium complexes, ligand. The limiting conditions for an effi-
it was discovered that the sensitizer ex- cient sensitizer makes these two extreme
cited state oxidation potential should be at situations incompatible. Because, in the
least – 0.9 V versus Saturated calomel ref- former case, the excited state is not suffi-
erence electrode (SCE), to inject electrons ciently energetic to enable electron transfer
efficiently onto TiO2 conduction band. The into the conduction band of the TiO2 semi-
ground state oxidation potential should be conductor, and in the latter case, the easily
about 0.5 V versus SCE, to be regener- oxidizable complex cannot be reduced back
ated rapidly via electron donation from the by a suitable electron relay like I− .
electrolyte (iodide/triiodide redox system) To illustrate the tuning aspects of
or a hole conductor. A significant decrease the MLCT transitions in ruthenium
in electron injection efficiencies will oc- polypyridyl complexes, the well known
cur if the excited and ground state redox [RuL3 ] (L = 4,4$ dicarboxylic acid-2,2$ -bi-
potentials are lower than these values [32]. pyridine) type of complex can be consid-
ered. This complex shows strong visible
5.2.2.5 Absorption Spectral Properties of band at 466 nm, because of CT tran-
Metal Complexes sition from metal t2g highest occupied
molecular orbitals (HOMO) to π ∗ -lowest
5.2.2.5.1 Tuning of MLCT Transitions To unoccupied molecular orbitals (LUMO) of
absorb the light below 920 nm, the sensi- the ligand (Fig. 3). The Ru(II)/(III) oxida-
tizer’s redox level needs to be tuned. The tion potential is at 1.3 V, and the ligand
MLCT transitions and the redox properties based reduction potential is at −1.5 V
Ligand π *t
un
π*orbitals ing
∆E
Energy
∆E ∆E
ing
tun
t 2g
Metal
t2g orbitals
versus SCE [33]. From spectro chemical are tuned from 466 nm to 560 nm by
and electrochemical studies of polypyridyl introducing different donor ligands. The
complexes of ruthenium, it has been con- oxidation potential in these complexes
cluded that the oxidation and reduction were tuned from 1.3 V to 0.3 V ver-
potentials are the best indicators of the en- sus SCE, going from [RuL3 ] to [RuL2 L$ ]
ergy levels of the HOMO and LUMO [34]. L$ = phenylpyridine). Thus, the energy of
The energy between the metal t2g orbitals the HOMO is varied to an extent of 1 V
and π ∗ -orbitals can be reduced either by in this series of complexes [35]. Meyer and
raising the energy of the t2g or by de- coworkers have used the same concept to
creasing the energy of the π ∗ -orbitals synthesize black absorbers by judiciously
with donor acceptor ligands, respectively selecting push and pull ligands [36].
(Fig. 4). In the following sections, ways It is interesting to note the magnitude
to tune HOMO and LUMO energy lev- of the spectral shift for the lowest energy
els by introducing various ligands will be CT transition is ≈0.5 eV, and the shift in
discussed. the oxidation potential ≈1 eV. This clearly
shows that not all the HOMO tuning is
5.2.2.5.2 Spectral Tuning in ‘‘Push-Pull’’ translated into the spectral shift of the
Type of Complexes The lowest energy complex. The apparent 0.5 eV, difference
MLCT transition of ruthenium polypyridyl is involved in raising the energy of π ∗ -
complexes of the type [RuL3 ], (L = orbitals of the pulling ligands caused by
4,4$ dicarboxylic acid-2,2$ -bipyridine), can the pushing ligands [35].
be lowered so that it absorbs more in the
red region of the visible spectrum by sub- 5.2.2.5.3 Spectral Tuning in Complexes
stituting one of the bidentate ligand by a Containing Hybrid Donor Ligands The ab-
donor bidentate ligand (pushing ligand). sorption and luminescence spectral prop-
The remaining two electron-withdrawing erties of complexes of the type [Ru(dmbip)
bidentate ligands act as pull ligands. A list (dcbpy)(X)] and [Ru(dmbip)(dcbiq)(X)],
of complexes is included in Table 1, where where dmbip = 2,6-bis(1-methylbenzimid-
the absorption maxima of the complexes azol-2$ -yl)pyridine, dcbpy = 4,4$ -(COOH)2 -
Tab. 2 Absorption, emission, and redox properties of ruthenium complexes containing ‘‘hybrid
donor’’ ligands
2,2$ -bipyridine, dcbiq = 4,4$ -(COOH)2 -2,2$ - the π-acceptor strength of the ancillary
biquinoline and X = Cl− , NCS− , CN− , ligand, CN− , NCS− , H2 O, NCN− Cl− .
N(CN)2 − , NCN− , and H2 O are summa- The red shift of the absorption maxima
rized in Table 2. The 2,6-bis(1-methyl- in complexes containing dcbiq compared
benzimidazol-2$ -yl)pyridine ligand is a hy- to dcbpy acceptor ligand is because
brid ligand that contains donor groups of the low π ∗ -orbitals of biquinoline.
(benzimidazol-2$ -yl) and an acceptor group The resonance Raman spectra of these
(pyridine). The complexes containing complexes for excitation at 568 nm show
these ligands show intense ultraviolet bands predominantly associated with the
(UV) bands at 362 and 346 due to dcbpy and dcbiq ligands indicating that the
intraligand π –π ∗ transition of 2,6-bis lowest excited state is a metal to dcbpy or
(1-methylbenzimidazol-2$ -yl)pyridine. The dcbiq ligand CT state [32, 37, 38].
bands at 308 and 244 nm are because of
the π –π ∗ transition of dcbpy ligand. On 5.2.2.5.4 Influence of Nonchromophoric
acidification, the 308 nm band shifts to Ligands on MLCT Transitions This section
314 nm as a result of the protonation of compares spectral response of the three
the carboxylate groups, whereas the 362 complexes of the type [Ru(P-terpy)(Me2
and 342 nm bands remain unchanged in bpy)(NCS)] 1, (P-terpy = 4-phosphonato-2,
acidic and basic pH. 2$ : 6$ ,2$$ -terpyridine, Me2 bpy] = 4,4$ -di-
The MLCT bands of these complexes methyl-2,2$ -bipyridine), [Ru(H2 dcbpy)2
are broad and red shifted by approximately (NCS)2 ] 2, (H2 dcbpy = 4,4$ -dicarboxy-2,
140 nm, compared to [Ru(dcbpy)3 ]. The 2$ -bipyridine) and (TBA)2 [RuL(NCS)3 ] 3,
lowest-energy MLCT transitions within (L$ = 4,4$ ,4$$ -tricarboxy-2,2$ : 6$ ,2$$ -terpy-
this series were shifted from 486 to ridine) where the number of nonchro-
608 nm, and the HOMO level varied over mophoric ligands such as NCS− are
an extent of 0.45 V versus SCE. The varied from one to three. The absorption
energy of the MLCT transition in these and luminescence spectral properties of
complexes decreases with the decrease in complex 1–3 are summarized in Table 3.
5.2 Dyes for Semiconductor Sensitization 415
Tab. 3 Absorption, emission, and redox properties of ruthenium complexes containing NCS
ligands
The absorption maxima are listed for carboxylic acid groups at the 4,4$$ -position
the intense lowest-energy MLCT bands in of 2,2-bipyridine ligand the MLCT maxima
ethanol, although the spectra possess ad- red shifted to 535 nm. However, substitu-
ditional absorption features at lower and tion at the 5,5$ -position of 2,2$ -bipyridine
higher energies. shifted the maxima further into the red,
The interesting feature of these com- 580 nm. On the contrary, the MLCT max-
plexes is a broad MLCT absorption bands ima blue shifted (500 nm) by substituting
in the visible region. The most intense at the 6,6$ positions of 2,2$ -bipyridine.
MLCT transition maxima of complexes 1, The enhanced red response of com-
2, and 3 are at 500, 535, and 620 nm, plexes containing the 5,5$ -dicarboxylic
respectively. The 120 nm red shift of 3 acid-2,2$ -bipyridine is because of a de-
compared to 1 is the result of an in- crease in the energy of the π ∗ -orbitals,
crease in the energy of the metal t2g which makes them attractive as sensitizers
orbitals caused by introducing nonchro- for nanocrystalline TiO2 films. Bignozzi
mophoric ligands and a slight decrease and coworkers found that the IPCE of com-
in the LUMO of the 4,4$ ,4$$ -tricarboxy- plexes having the 5,5$ -dicarboxylic acid-
2,2$ : 6$ ,2$$ -terpyridine. The electrochem- 2,2$ -bipyridine ligands were lower than
ical data are consistent with the above the analogous complexes that contain 4,4$ -
assignments [22, 39, 40]. dicarboxylic acid-2,2$ -bipyridine [43]. They
rationalized the low efficiency of the sen-
5.2.2.5.5 Influence of the Position of Car- sitizers containing 5,5$ -dicarboxylic acid-
boxyl Groups on MLCT Transitions Ruthe- 2,2$ -bipyridine ligands in terms of low
nium polypyridyl compounds, of the type excited state redox potentials.
[RuL2 (X)2 ], (where L = 4,4$ -dicarboxylic
acid-2,2$ -bipyridine, 5,5$ -dicarboxylic acid- 5.2.2.5.6 MLCT Transitions in Geometric
2,2$ -bipyridine, and 6,6$ -dicarboxylic acid- Isomers Isomerization is another excit-
2,2$ -bipyridine; X = Cl− , CN− and NCS− ) ing approach toward tuning the spectral
have been reported [41–43]. The lowest properties of metal complexes (Fig. 5)
MLCT absorption maxima of [RuL2 (NCS)2 ] [44–46]. The electronic properties of cis-
complex (L = 2,2$ -bipyridine) is seen at and trans-isomers of [RuL2 (X)2 ], (L = 4,4$ -
510 nm in ethanol. By substituting two dicarboxylic acid-2,2$ -bipyridine; X = Cl− ,
416 5 Dye-Sensitized Photoelectrochemistry
0.8 COOH
HOOC COOH
HOOC
N
0.6 NCS N N
N
Ru SCN Ru NCS
N
NCS
Absorbance
N N
HOOC N
0.4
HOOC COOH
HOOC
0.2
0.0
300 400 500 600 700 800
Energy
[nm]
Fig. 5 UV-visible absorption spectra of cis-, trans-
bis(2,2$ -bipyridine-4,4$ -dicarboxylic acid) ruthenium (II) (NCS)2 complexes.
DMF, dimethylformamide.
a in water at pH 4.5.
b Weak and broad emission maximum.
H2 O and NCS− ) are collected in Table 4. and 440 nm. On the other hand, the
The UV-visible absorption spectrum of the cis-dichloro complex in DMF solution
trans-dichloro complex in DMF solution shows only two distinct broadbands in
shows at least three MLCT absorption the visible region at 590 and 434 nm,
bands in the visible region at 690, 592, which was assigned to a MLCT transitions.
5.2 Dyes for Semiconductor Sensitization 417
The lowest energy MLCT band in the the L3 ligands. In these complexes, L3 be-
trans-dichloro complex is significantly red ing the dithiocarbamate ligand, shifts the
shifted compared to the cis-dichloro com- Ru(II) t2g level upward without signifi-
plex. This red shift is due to stabilization cantly affecting the π ∗ -level of the dcbpy
of the LUMO of the dcbpy ligand in the ligand thereby increasing the MLCT ab-
trans-species relative to the cis-species [47]. sorption in the red portion of the visible
The red shift (108 nm) of the lowest region.
energy MLCT absorption in the spectrum
of the trans-dichloro complex compared to 5.2.3
the spectrum of the trans-dithiocyanato is Hydrophobic Sensitizers
due to the strong σ donor property of the
Cl− compared to the NCS− ligand. The The other important aspect in dye-
chloride ligands cause destabilization of sensitized solar cells is water-induced
the metal t2g orbitals, and raising them desorption of the sensitizer from the
in energy relatively closer to the ligand surface. Extensive efforts have been made
π ∗ -orbitals result in a lower energy MLCT in our laboratory to overcome this problem
transition. According to Lever’s electro- by introducing hydrophobic properties in
chemical parameterization, the chloride
the ligand. The heteroleptic complexes
ligand has a parameter value of −0.24 and
containing hydrophobic ligands of
the thiocyanate ligand has a value of −0.06.
the type [Ru(dcbpy)(mhdbpy)(NCS)2 ] 1,
The electrochemical and the absorption
[Ru(dcbpy)(dtdbpy)(NCS)2 ] 2 [Ru(dcbpy)
data are in good agreement with reported
(mddbpy)(NCS)2 ] 3 (dcbpy = 4,4$ -dicar-
parameterization scale for ligands [48].
boxy-2,2$ -bipyridine, mhdbpy = 4-methyl-
4$ -hexadecyl-2,2$ -bipyridine and dtdbpy =
5.2.2.5.7 Spectral Tuning in Heteroleptic
4,4$ -ditridecyl-2,2$ -bipyridine, mddbpy =
Sensitizers
4-methyl-4$ -didodecyl-2,2$ bipyridine) have
The synthesis and characterization of
been synthesized (Fig. 6). The photo-
heteroleptic complexes of ruthenium
[Ru(L1 )(L2 )(L3 )] starting from a poly- current action spectra of these complexes
meric complex Ru(CO)2 Cl2 have been show broad features covering a large part
reported [49–51]. The synthetic method- of visible spectrum and displays a maxima
ology of such heteroleptic complexes in- around 550 nm, where the incident
volves several steps, resulting in very low monochromatic IPCE exceeds 80%.
yields. Recently, a novel synthetic route The performance of these hydrophobic
has been developed to introduce differ- complexes as CT photosensitizers in
ent ligands on a ruthenium (II) precur- nanocrystalline TiO2 -based solar cell
sor, namely, the RuCl2 (DMSO)4 complex, shows excellent stability toward water-
which is a versatile and well-characterized induced desorption [55].
starting material [52]. A range of mixed The rate of electron transport in dye-
ligand sensitizers of the type [Ru(L1 )(L2 ) sensitized solar cell is a major element of
(L3 )] have been prepared and their spectro- the overall efficiency of the cells. The elec-
scopic, luminescent, and redox properties trons injected into the conduction band
are reported [53, 54]. The absorption spec- from optically excited dye, can traverse the
tral data of these complexes show a red TiO2 network and can be collected at the
shift with increasing donor strength of transparent conducting glass or can react
418 5 Dye-Sensitized Photoelectrochemistry
with a redox mediator. The reaction of in- current conversion yields remaining un-
jected electrons into the conduction band der 1%, which is insufficient for solar cell
with a redox mediator gives undesirable applications [57–60]. One of the reasons
dark currents, significantly reducing the for such low efficiencies is aggregation of
charge-collection efficiency, and thereby the dye on the TiO2 surface. This asso-
decreasing the total efficiency of the cell. ciation often leads to undesirable photo-
This problem is drastically suppressed by physical properties such as self-quenching
using hydrophobic sensitizers of the type and excited state annihilation. However,
shown in Fig. 6. The most likely explana- the advantage of this class of complexes
tion for the dark-current suppression is is the near IR response which is very
that the long chains of the sensitizer inter- strong, having extinction coefficients of
acts laterally to form an aliphatic network, close to 50 000 M−1 cm−1 at 650 nm com-
thereby preventing triiodide from reaching pared to the polypyridyl complexes, which
the TiO2 surface. have small extinction coefficients at this
wavelength.
5.2.4
Near IR Sensitizers
5.2.4.2 Ruthenium Phthalocyanines
5.2.4.1 Phthalocyanines The bis(3,4-dicarboxypyridine)Ru(II)1,4,8,
Phthalocyanines possess intense absorp- 11,15,18,22,25-octamethyl-phthalocyanine
tion bands in the near IR region and are (JM3306) shows a visible absorption band
known for their excellent stability render- at 650 nm (ε 49 000 M−1 cm−1 ) and a
ing them attractive for photovoltaic ap- phosphorescence band located at 895 nm.
plications [56]. They have been repeatedly The triplet state lifetime is 474 ns under
tested in the past as sensitizers of wide anaerobic conditions. The emission is
band gap oxide semiconductors and ob- entirely quenched when JM3306 is ad-
tained poor incident photon to electric sorbed onto a nanocrystalline TiO2 film.
5.2 Dyes for Semiconductor Sensitization 419
100
80
60
IPCE
[%]
40
20
0
400 500 600 700 800
Wavelength
[nm]
Fig. 7 Photocurrent action spectra of nanocrystalline TiO2
films sensitized by bis(3,4-dicarboxypyridine)Ru(II)
1,4,8,11,15,18,22,25-octamethyl-phthalocyanin. The incident
IPCE is plotted as a function of wavelength.
420 5 Dye-Sensitized Photoelectrochemistry
behavior using LiI/LiI3 /propylene car- from the lower energy MLCT band of
bonate electrolyte [62]. Interestingly, the complexes containing 5,5$ -dicarboxylic
zinc (II) 2,9,16,23-tetracarboxyphthalocya- acid-2,2$ -bipyridine ligand because of lack
nine exhibited 45% monochromatic cur- of driving force for electron injection.
rent conversion efficiency at 700 nm. It
is shown that electron injection to TiO2 5.2.5.2 Osmium Complexes
occurs from the excited singlet state Although there are a large number of
of the phthalocyanine derivatives. The osmium polypyridyl complexes, very few
inherent problem of phthalocyanine ag- have been used as sensitizers in dye-
gregation is reduced considerably in 3d sensitized solar cells. The osmium com-
metal phthalocyanines by introducing 4- plexes have several advantages compared
tert-butylpyridine and 3α,7α-dihydroxy-5β- to their ruthenium analogues: (1) osmium
cholic acid (cheno) into the dye solution. has a stronger ligand field splitting com-
The added 4-tert-butylpyridine in all likeli- pared to ruthenium, (2) the spin-orbit
hood coordinates to the metal in the axial coupling leads to excellent response in the
position and thereby prevents aggregation red region because of enhanced singlet-
of the phthalocyanines. triplet CT transitions [26]. Heimer and
This type of sensitizers opens up new av- coworkers have synthesized [OsL2 (CN)2 ]
enues for improving the near IR response complex (L = 4, 4$ -dicarboxylic acid-2,2$ -
of nanocrystalline injection solar cell. In bipyridine) and found that the complex
addition, important applications can be is extremely stable under irradiation in a
foreseen for the development of photo- homogeneous aqueous solution compared
voltaic windows transmitting part of the to the analogous ruthenium complex [64].
visible light. Such devices would remain The greater photo-stability for osmium
transparent to the eye, while absorbing is consistent with a stronger crystal field
enough solar energy photons in the near splitting of metal d-orbital that inhibits ef-
IR to render efficiencies acceptable for ficient population of d-d states. However,
practical applications. a factor of 10 lower injection efficiency for
the osmium complex when anchored on
5.2.5 TiO2 electrode was reported compared to
Mononuclear and Polynuclear Metal the ruthenium analog and rationalized in
Complexes of Group VIII terms of slow regeneration rates for the
oxidized dye [65].
5.2.5.1 Iron Complexes Lewis and coworkers have synthesized
The photoconversion efficiency for iron and characterized two osmium com-
complexes [FeL3 ] and [FeL2 (X)2 ], (L = plexes [OsL2 (CN)2 ] and [OsL3 ] (L = 4,4$ -
4,4$ -dicarboxylic acid-2,2$ -bipyridine and dicarboxylic acid-2,2$ -bipyridine). In their
5,5$ -dicarboxylic acid-2,2$ -bipyridine) has studies they found nearly 80% incident
been reported [63]. The [FeL2 (X)2 ] type monochromatic IPCE [66].
complexes show two broad MLCT bands
in the visible region. In both cases, 5.2.5.3 Polynuclear Complexes
the photoconversion efficiency of the Polynuclear systems in which more than
high-energy band was reported in the one complex is connected by a bridging
range of 40%. Nonetheless, there is ligand have been studied in several labo-
virtually no photocurrent contribution ratories [67–72]. Here studies have been
5.2 Dyes for Semiconductor Sensitization 421
restricted exclusively to the cyano-bridged with the three metal centers, the one with
(CN-bridged) trinuclear ruthenium com- lowest energy is based on the central
plexes of the type shown in Fig. 8. Scandola unit carrying the ligand 4,4$ -dicarboxy-
and coworkers have synthesized and char- 2,2$ -bipyridine (dcbpy). When coated on
acterized several of these complexes with a TiO2 -electrode, the complex anchors
respect to their photophysical and elec- through the carboxyl unit of this central
trochemical properties [73]. In such com- unit. Photons absorbed by the higher en-
plexes, one can invoke possible antenna ergy peripheral units efficiently transfer
effects in the light harvesting process by in- excitation energy to this central unit, ac-
troducing various fragments in which the counting for its excellent performance.
photophysical properties are tuned to have To understand the principles of opera-
directionality in the excited state. The pres- tion of polynuclear complexes and possibly
ence of numerous metal centers should improve their performance, the ‘‘tunabil-
allow for different tailored fragments con- ity’’ of photophysical and redox properties
taining various ligands [72, 74]. has been examined in a number of such
Potential application of these complexes complexes. These complexes are com-
as photosensitizers for light energy har- posed of polypyridyl units of Ru and
vesting relies heavily on demanding syn- Os as chromophores and cyanide as the
thetic schemes. The CN-bridged trinu- bridging ligand [76]. By appropriate syn-
clear complex 1 (Fig. 8) has been found thetic procedures, it is possible to control
to be an excellent photosensitizer for the mode of coordination of the bridg-
spectral sensitization of the wide band ing cyanide to a given chromophoric unit
gap semiconductor TiO2 [75]. Among the as a nitrile (C-bonded) or isonitrile (N-
possible MLCT excited states associated bonded) [74–77]. By varying the nature of
R
COOH
R
R
N
HOOC N N
N Ru
N C C
N N
Ru N
N N
N R
N R
HOOC C
C N
Ru
HOOC N N
N
R
R
R
1 : R = H ; 2 : R = COOH ; 3 : R = CH3 ; 4 : R = C6H5
5 = [(CN)(bby)2 Os-CN-Ru(dcbpy)2-NC-Os(bpy)2(CN)]
Fig. 8 Structure of CN-bridge
trinuclear ruthenium complex . 6 = [(H2O)(bby)2 Ru-CN-Ru(dcbpy)2-NC-Ru(bpy)2(H 2O)]
422 5 Dye-Sensitized Photoelectrochemistry
the metal center, the mode of coordination the complex and the substrate, is an
of the bridging ligand, and the spectator important target in dye-sensitized solar
ligands, it is possible to vary the location cells. Several ruthenium complexes con-
and the direction of excited state electron taining carboxylic acid and phosphonated
transfer process. polypyridyl ligands are described [78–83].
The carboxylic and phosphonate function-
5.2.6 ality serves as an anchoring group to
Surface Chelation of Polypyridyl Complexes immobilize the complex on the nanocrys-
onto the TiO2 Surface talline TiO2 films. The immobilized sen-
sitizer absorbs a photon to produce an
5.2.6.1 Anchoring Groups excited state, which efficiently transfers
The functional groups serve as grafting its electron onto the TiO2 conduction
agents for the oxide surface of the TiO2 band. To achieve high quantum yields
films. The grafting of polypyridyl com- of the excited state electron-transfer pro-
plexes onto the oxide surface, which allows cess the dye ideally needs to be in in-
for electronic communication between timate contact with the semiconductor
HOOC COOH
HOOC COOH
N N N N
4,4′-dicarboxyl-2,2′-bipyridine 5,5′-dicarboxyl-2,2′-bipyridine
HOOC COOH
COOH
HOOC COOH
N
N N
N N
4,4′-(paradicarboxyl)diphenyl-2,2′-bipyridine 4,4′,4′′-tricarboxylic-2,2′;6,2′′-terpyridine
P(O)(OH)2
P(O)(OH)2 P(O)(OH)2
N
N N
4,phosphonate-2,2′;6,2′′-terpyridine 4,4′-diphosphonate-2,2′;bipyridine
HOOC COOH
HOOC COOH
N N
N N
4,4′-dicarboxyl-2,2′-biquinoline 4,7-dicarboxyl-1,10′-phenanthroline
1.2
pH 9.0
1.0 pH 6.2
pH 4.4
pH 3.9
0.8 pH 3.0
pH 2.4
Absorbance
pH 1.0
[OD]
0.6
0.4
0.2
0.0
400 450 500 550 600 650 700
Wavelength
[nm]
Fig. 10 Absorption spectral changes of cis-dithiocyanatobis(2,2$ -bipyridine-4,4$ -dicarboxylic acid)
ruthenium (II) complex as a function of pH.
O 2+ O +
HO OH HO O−
P P
pKa4 < 2
H H H H
N N N N
6 2 N N
5 3
4
pKa3 = 3.55
O
HO O−
P
H
N N
N
pKa2 = 4.65
O 2− O −
−
O O− HO O−
P P
pKa1 = 5.99
N N N N
N N
100 TG 0
98
−5
Deriv. mass
Exo(heat flow)
DTG
Mass
[%]
[%]
96
−10
94
0 −15
DSC
92
100 200 300 400
Temperature
[°C]
Fig. 11 The combined thermogravimetric (TG), derivative
thermogravimetric (DTG), and differential scanning calorimetry (DSC)
curves of complex (Bu4 N)4 [Ru(dcbpy)2 (NCS)2 ] adsorbed on TiO2 in
dynamic air atmosphere. Loss of water, exothermic decomposition of
amine (between 200–330 ◦C) and decarboxylation starting at 360, ending
at 390 ◦ C.
5.2 Dyes for Semiconductor Sensitization 427
80
1
2
60
3
IPCE
40
[%]
the dye-sensitized solar cell. Although well engineering aspect of the sensitizers for
worked-out procedures exist for the effi- enhanced spectral response into the visible
cient purification of metal complexes, it and near IR region is clearly illustrated.
was found that the isolation of complexes The development of efficient dye-
at its isoelectric point and followed by col- sensitized solar cells would expand our
umn purification using Sephadex LH-20 ability to better harvest the sun’s energy
gel resulted in analytically pure samples. and decrease our dependency on fossil
fuels and nuclear power.
5.2.8.3 Characterization
The most commonly used methods for Acknowledgment
characterization of ruthenium sensitizers
We acknowledge financial support of this
are NMR, IR, Raman, UV-vis, emission,
work by the Swiss Federal Office for En-
cyclic voltammetry, high-performance liq-
ergy (OFEN) and the Institute for Applied
uid chromatography (HPLC) and X-ray
Photovoltaics (INAP), Gelsenkirchen, Ger-
crystallography.
many. We thank Drs. R. Humphry-Baker,
5.2.9 S. M. Zakeeruddinn, F. P. Rotzinger and
Conclusion K. Kalyanasundaram for their valuable
discussions. Authors are indebted to
By appropriate synthetic methodology, it Dr. P. Infelta for his assistance in handling
has been possible to prepare a wide range various applications during the prepara-
of complexes and to examine the concept of tion of this chapter.
tunability of their photophysical and redox
properties. Different synthetic methods to References
introduce diverse ligands, which have been
judiciously selected to reconcile the tasks 1. R. D. McConnell, Future Generation Photo-
voltaic Technologies, AIP Conference Proceed-
of the sensitizer, to afford vectorial elec- ings 404, Denver, 1997.
tron injection into the semiconductor and 2. H. W. Vogel, Ber. Dtsch. Chem. Ges. 1873, 6,
efficient solar light harvesting, were dis- 1302.
cussed. Analysis of the acid-base properties 3. R. Gleria, R. Memming, Z. Phys. Chem. 1976,
of transition metal complexes contain- 98, 303.
4. W. D. K. Clark, N. Sutin, J. Am. Chem. Soc.
ing ligands with protonatable functional 1977, 99, 4676.
groups provide important information 5. A. Hamnett, M. P. Dare-Edwards, R. D.
about the nature of binding properties Wright et al., J. Phys. Chem. 1979, 83, 3280.
on the oxide surface. Comparison of the 6. M. Grätzel, Prog. Photovoltaics 2000, 8, 171.
ground state and excited state pKa s provide 7. S. Y. Huang, G. Schlichthörl, A. J. Nozik
et al., J. Phys. Chem. B 1997, 101, 2576.
useful data for the identification of the lig-
8. J. van de Lagemaat, N.-G. Park, A. J. Frank,
and involved in the lowest energy MLCT J. Phys. Chem. 2000, 104, 2044.
transition, which is essential in the design 9. R. Argazzi, C. A. Bignozzi, G. M. Hassel-
of photosensitizers for nanocrystalline so- mann et al., Inorg. Chem. 1998, 37, 4533.
lar cells. The effects exerted by the proton 10. K. Murakoshi, G. Kano, Y. Wada et al., J.
Electroanal. Chem. 1995, 396, 27.
content of the sensitizer on both the short-
11. A. Zaban, S. Ferrere, J. Sprague et al., J. Phys.
circuit photocurrent and open-circuit pho- Chem. B 1997, 101, 55.
tovoltage of dye-sensitized nanocrystalline 12. H. Sughihara, L. P. Sing, K. Sayama et al.,
solar cells were also shown. The molecular Chem. Lett. 1998, 1005.
430 5 Dye-Sensitized Photoelectrochemistry
of the electron and hole conducting com- cell at operating point, light forward biases
ponents; Sect. 5.3.4 reviews some basic the junction to stimulate a reverse current,
theoretical principles and attempts to de- which opposes the photocurrent. In a dye-
velop a ‘‘device physics’’ for dye-sensitized sensitized device, unusually, this reverse
devices; Sect. 5.3.5 reviews the experimen- current contains both light and dark com-
tal technique for studying charge transport ponents. The key charge transfer reactions
and discusses the main results and the are illustrated in Fig. 1.
models that have been proposed; finally
some points of consensus and outstanding Key Processes
questions are summarized in Sect. 5.3.6. Light absorption:
hν 1
8 6 2 3 9
8* 4 7 9*
1
hν
2
Fig. 1 Key electron transfer 5
reactions in a dye-sensitized Electron conductor Hole conductor
device. n p
434 5 Dye-Sensitized Photoelectrochemistry
where D, D∗ , and D+ refer to the dye, the the continual photoionization and regen-
excited state of the dye, and the dye cation; eration of the sensitizer under steady state
e(sc), e(n), and e(p) refer to the electron in illumination. In energetic terms, the Gibbs
the semiconductor (electron conductor), in free energy of the excited sensitizer, D∗ ,
the n-contact and in the p-contact material; plus the vacant electron acceptor state in
and HCRed and HCOx refer to the reduced the semiconductor, should exceed that of
and oxidized state of the hole conductor. the dye cation plus the occupied acceptor
Following charge transfer, electrons and state, D+ + e(TiO2 ). This is usually rep-
holes are transported through the respec- resented on a simplified potential energy
tive mediums by electron conduction (6) diagram as a sensitizer excited state of
and hole conduction (7). We have assumed potential energy higher than the conduc-
that hole transport is more appropriately tion band of the semiconductor. Similarly,
described as a redox process and electron the redox potential µRedOx of the redox
transport as a simple electron conduction. couple in the hole conductor should be
In discussing the net transport of elec- sufficient to reduce the dye cation. This
trons, we need to include processes (4), is usually represented as a dye ground
(5), and (6), whereas net hole transport state of potential energy which is lower
involves (3), (4), and (7). than the redox potential of the electrolyte.
In addition, there is the energy transfer These relationships are illustrated in Fig. 2
process because of relaxation of the excited together with the energetically preferred
state of the dye directions for charge transfer. Such po-
tential energy diagrams do not normally
D∗ −−−→ D + hν or heat (1∗ ) distinguish the entropic contribution to
the driving force, which arises from the
which does not result in a forward electron much greater density of states per unit sur-
transfer and may be counted with the face area in the semiconductor than in the
reverse electron transfers as a loss process. dye. This effect is estimated to contribute
We define forward as the direction of up to 0.1 eV to the driving force [12]. Con-
charge transfer under photovoltaic oper- tact materials should allow facile transfer
ating conditions, and will use ‘‘forward’’ of electrons from electron conductor and
bias to mean the direction of potential dif- positive charge from electrolyte.
ference in those conditions (p terminal
positive with respect to n). Note that this 5.3.2.3 Physical Structure and Materials
condition corresponds to negative applied In the most commonly studied configura-
potentials in electrochemical terms. tion of the DSSC, the electron conductor
is a wide band gap, nanocrystalline, metal
5.3.2.2 Energetic Structure oxide film. Colloidal titanium dioxide is
In a functioning dye-sensitized device, the most often used although other wide band
net current output is the sum of all the gap metal oxides are candidates. The sen-
forward electron transfer processes less all sitizer is a transition metal-based (usually
of the reverse processes. For a net forward ruthenium), organic dye, with excited state
current under (steady state) illumination, free energy sufficient to reduce the semi-
processes (2) and (3) must be kinetically conductor, and containing ligands, such
more favorable than (4), and (3) must be as carboxylates or phosphonates, which
preferred to (5). These conditions enable facilitate bonding to the semiconductor
5.3 Charge Transport in Dye-sensitized Systems 435
D*/D +
E cb
Ef Photovoltage
Energy
µ RedOx
D/D +
E vb
− +
− +
Distance
Fig. 2 Potential energy diagram of DSSC. For net forward electron transfer the
oxidation potential of the dye-excited state, D∗ /D+ should be higher than the
semiconductor conduction band, and the redox potential of the hole conductor
should be higher than the dye-ground state potential D/D+ . Absorption of light by
the dye causes charge separation across the interface, resulting in a splitting of the
electrochemical potential and hence a photovoltage.
surface. The hole conductor is tradition- 5–20 nm in size are sintered together
ally a polar liquid electrolyte, such as into a random porous network, leaving
propylene carbonate or acetonitrile, con- a network of voids of similar volume
taining the iodide–triiodide redox couple, fraction and dimensions. In the DSSC,
although latterly solid-state ionic and hole the dye is adsorbed on to the entire
conductors have been substituted. In many film surface, whereas the voids are filled
cases, the electrolyte contains a lithium with the hole conductor. The enormous
salt, which facilitates electron accumula- microscopic surface area, which is typically
tion in the TiO2 and apparently improves 1000 times the geometrical surface area for
charge separation [13]. In the TiO2 –liquid a 10-µm-thick film, leads to a much higher
electrolyte system, ohmic contact is made optical depth than for a planar interface,
to the oxide through a fluorine doped SnO2 whereas the pores allow redox ions to
conducting glass substrate, and to the elec- travel between all points on the surface
trolyte through a platinum coated counter and the counter electrode [2]. Figure 4
electrode, which catalyzes the regeneration illustrates the typical device structure of
of the redox couple. Material properties the DSSC.
and requirements are described here and
further details are given in Refs. 2, 14. 5.3.2.4 Charge Carriers and Charge
A unique feature of DSSCs is that Transfer Processes
the electron conductor (the oxide film) is The charge transfer processes are sum-
both porous and electronically connected. marized here. In the dark, at equilibrium,
The morphology of a typical TiO2 film there is no net charge transfer. When a
is illustrated in Fig. 3. Crystallites of forward bias (negative potential) is applied
436 5 Dye-Sensitized Photoelectrochemistry
300 nm
0 Seal
− − − − − −
Glass
e−
in the dark, electrons flow from substrate Probable intermediate reaction schemes
into the film and oxidized species from are detailed in Ref. 16 and the solution
counter electrode toward the interface. phase redox couple has been probed in
Electrons and oxidized species combine depth [17]. In the steady state, ion popu-
at the interface through process (4), which lations are maintained by the oxidation of
in the iodide–triiodide redox couple can I3 − at the counter electrode, which deliv-
be summarized as ers the electron to the external circuit. This
1 3 recombination process is responsible for
eTiO2 + I3 − −−−→ I − . (10) the dark current from the junction.
2 2
5.3 Charge Transport in Dye-sensitized Systems 437
as recombination centers for electrons the porous material. This leaves two re-
(a loss process), or may be scavenged gions where the potential can be dropped:
by the hole conductor (a photovoltaic at the interface between the TiO2 film
effect). However, the wide band gap and the SnO2 : F (fluorine doped SnO2 )
means that holes are seldom generated, rear conducting contact; and across the
and except in the case of ultraviolet (UV) interface layer between film and elec-
irradiation, can be ignored. trolyte, which includes outer atomic layers
of film and Helmholtz layer. It is most
5.3.2.5 Potential Distribution likely that the potential drop is shared be-
To sustain the forward electron transfer tween the two interfaces [19, 20]. Bisquert
described earlier, the hole-conducting ma- has shown by solving Poisson’s equation
terial should have a lower effective work for a columnar model of the film that
function than the electron conductor. As in the largest electric fields within the film
any electronic junction, the electrochem- are established at the interface with the
ical potential or Fermi level in the two SnO2 substrate [19]. Large fields cannot
materials aligns at equilibrium, leaving a extend beyond a few nanoparticle radii
potential difference equal to the difference from the rear substrate (Fig. 5). The same
in work functions to be accommodated in conclusion was reached earlier with more
the two materials. In a conventional elec- qualitative arguments in Refs. 21, 22. The
trochemical junction, the charge density proportion of the potential difference that
in the layer of electrolyte close to the inter- can be taken up by the film and rear sub-
face (the Helmholtz layer) is much greater strate is determined by morphology and
than the fixed charge density in the semi- materials properties; the remainder must
conductor, and the potential difference is be taken up by the Helmholtz layer [23,
accommodated mainly in the semiconduc- 12], leading to large electric fields at the in-
tor by band bending. In a lightly doped terface, extending only a few nm into the
semiconductor, this band bending extends electrolyte [24], and a film which is largely
over a depletion region some microns in depleted of electrons in equilibrium. The
width, beyond which the semiconductor is size of the potential drop at the Helmholtz
neutral. layer is believed to be influenced by dye ad-
In the porous geometry of the DSSC, sorption, which polarizes the surface layer
the situation is different. The high charge by introducing protons, and by the redox
density of the electrolyte means that the potential of the electrolyte. Under applied
electrolyte tends to maintain the outer sur- reverse bias, the potential drop increases
face of the metal oxide film at a constant and an increasing proportion is taken up
potential, and individual nanoparticles in the Helmholtz layer. When the junction
making up the film are too small to sustain is forward biased either electrically or by
the junction potential difference. Calcu- illumination, the concentrations of both
lations using the Albery–Bartlett model electrons in the film and oxidized species
indicate that only some 50 mV of poten- (‘‘holes’’) in the electrolyte increase. The
tial can be dropped between the surface applied bias will be taken up, in general, by
and center of a 10-nm radius particle of both the TiO2 –TCO (transparent conduct-
(lightly n-type) TiO2 [18–20]. Therefore, if ing oxide) and the TiO2 –hole conductor
the film surface is an equipotential, large interface layers, reducing the potential dif-
electric fields cannot be sustained within ference and modifying the electric field
5.3 Charge Transport in Dye-sensitized Systems 439
at those interfaces. However, according to charge on the TiO2 electrode will attract
Cahen and coworkers, neither potential cations toward the film surface, where they
drop is sufficient on its own to account act to screen and stabilize electrons in the
for the open circuit voltage of the cell, film. Very small cations such as Li+ , which
and this is instead controlled by the dif- are able to intercalate into the interstices
ference between the redox potential of the of the crystal structure, may be drawn
electrolyte and the conduction band of the in. These processes clearly modify the
TiO2 [12]. This is an important distinction electrostatic potential distribution. In the
from conventional photovoltaic devices, steady state, they appear as changes in the
where the difference in work functions effective transport parameters; in transient
controls Voc . It is supported by studies by studies, currents due to ion movement
Pichot and Gregg of the photovoltage of may be resolved [26].
cells deposited on different work function
substrates, which indicate that the photo- 5.3.2.6 Relevance to Device Performance
voltage is not influenced by the potential Key requirements for good performance in
drop at the TiO2 –TCO interface [25]. any photovoltaic device are high light ab-
A key consequence of the potential sorption, efficient charge separation, and
distribution in the cell is that once efficient charge transport. In the DSSC,
separated, the electrons and holes in their light absorption is ensured by the large
respective media do not experience large optical depth of the sensitized porous film.
electrostatic forces because of the junction, Efficient charge separation in DSSC is
and are not driven primarily by built-in provided by the much faster rate of in-
electric fields. jection (2) relative to excited state decay,
An additional effect of applied bias is the and is discussed elsewhere in this Vol-
reorganization of solvent ions to minimize ume. Efficient charge transport is therefore
the energy of the system. A more negative the key. Having identified the species and
Red
+
Ox Ox 10
+ 15
+ + 8
Red 6
φ 10
+ Red 5 4 z/a
+ Ox
0 2
−0.51
−
0
0
0.5
1
r/a
(a) (b)
Fig. 5 (a) Schematic illustration of film surface, electrostatic potential distribution for a
showing lines of constant potential. columnar electrode with constant surface
Reorganization of cations in the electrolyte potential, showing how significant potential
allows the screening of electrons in the film, drops can be maintained only within a few tens
reducing their electrostatic effect. (b) Calculated of nm from the rear substrate (from Ref. 20).
440 5 Dye-Sensitized Photoelectrochemistry
and, possibly, through space charge rutile, and exploited for applications
induced potential barriers. However in gas sensing [39].
quantum confinement effects are not 5. The enormous surface area leads to a
expected because the typical crystal- high-volume density of surface-related
lite radius greatly exceeds the exciton defects. Apart from oxygen vacancy
radius (∼2 nm [18]). Space charge ef- induced Ti3+ states, these include de-
fects are also expected to be small on fects due to surface adsorption of ions
account of the low doping density of and intercalated species, particularly
the crystallites. in the presence of an electrolyte. Sur-
4. The material is nonstoichiometric and face adsorbed cations may lead to
has a natural tendency to acquire intraband gap states, similar in na-
oxygen vacancies through loss of O2 , ture to the oxygen vacancy [40], and
in an oxygen-poor environment: intercalated lithium leads to a sim-
ilar intraband gap density of states,
TiO2 −−−→ TiO2−x + xVO + 12 xO2 as observed by photoemission spec-
(12) troscopy [41]. Such defects can trap
where VO represents an oxygen va- electrons. The profusion of defects
cancy and x the fraction of oxygen leads to the formation of a surface
vacancies in the film. The rate of O2 ef- layer of different electronic structure
fusion can be controlled by the partial to the bulk.
pressure of oxygen and temperature. 6. Under UV illumination, TiO2 is highly
Considered as a point defect, the oxy- hydrophilic and attracts a surface layer
gen vacancy contains two electrons in of hydrogen bonded H2 O molecules.
its ground state but is readily ionized to In certain conditions [42, 43], the
deliver one or both electrons to empty water dissociates to leave surface
3d orbitals on adjacent Ti sites, chang- adsorbed OH groups – which may
ing the Ti oxidation state from Ti4+ to create electron traps – and itinerant
Ti3+ and increasing the population of protons. Protons intercalate easily and
electrons available for conduction [30]. can strongly influence conductivity
The electron in the Ti3+ defect state by assisting the formation of Ti3+
is localized but easily accessible to states [19, 44].
the conduction band. Calculations [37] 7. Ti3+ defect states have a characteris-
and experiment [38] place its energy tic red-infrared optical absorbance [45,
around 2–3 eV above the valence band 46, 19]. This can be used as a measure
edge in the case of rutile. Function- of the density of electrons accumu-
ally, these states are important because lated in the film [47–50]. Although
electrons trapped in these defect lev- there is some debate about whether
els are readily promoted to conduction the states giving rise to the absorbance
band states, where they contribute to are delocalized conduction band states
electrical conduction, whereas unoc- or intraband trap states, the evidence
cupied states act as traps for conduc- favors the latter [19, 51].
tion electrons. Oxygen vacancies thus 8. The nanocrystalline films are insu-
effectively dope the material n-type. lating in normal conditions, with a
The control of conductivity through dark resistivity of 108 –1010 " cm in
stoichiometry is well established for air. This is substantially larger than
442 5 Dye-Sensitized Photoelectrochemistry
D*/D+
E c.b.
3.2 eV
D/D+
3.2 eV
3.2 eV
3.4 eV ZrO2
SrTiO3
ZnO TiO2
E v.b.
SnO2
films on TCO substrates (usually SnO2 : F edge (Burstein shift) can be detected
or ITO) by similar procedures. It is sim- as an indicator of electron accumula-
ilar in band gap (3.2 eV) and band edge tion [18, 63].
position to TiO2 [2] with similar (13 nm di- 3. The exciton radius is, at around 5 nm,
ameter) or smaller (5 nm) crystallite sizes larger than in TiO2 (on account of
than for typical TiO2 films. As with TiO2 , a smaller effective mass), and may
it may be self doped n-type through oxygen lead to quantum confinement effects
deficiency. in nanocrystalline films [18, 65, 67].
Charge transport in nanocrystalline ZnO These affect the ease of transport
is less well studied than in TiO2 , but we between particles, leading to barriers
may make a few relevant comments: between large and small particles for
polydisperse particle size distributions.
1. Nanocrystalline ZnO is highly nonsto- 4. There is some evidence for a barrier at
ichiometric with a defect density of up the ZnO/ITO contact [67].
to 1018 cm−3 . This leads to a higher 5. ZnO is not capable of intercalating Li
electronic mobility than TiO2 , of over ions.
10−3 cm2 V−1 s−1 [65].
2. The oxygen vacancy defects appear to 5.3.3.1.2 p-type Metal Oxides Apart from
function as deep traps for holes and re- the sensitization of n-type semiconduc-
combination centers for electrons and tors with electron injecting dyes, there
shallow traps for electrons [66] (the is, in principle, the possibility of sensi-
limited dispersion of time constants tization of a p-type semiconductor with
in transient current experiments has a hole-injecting dye. The requirement for
been attributed to the absence of deep hole injection is that the potential of the
electron traps). No optical absorbance ground state of the dye lies below the
is associated with occupied defects but valence band of the semiconductor. There
instead a bleaching of the absorption has been some interest in nanoporous
444 5 Dye-Sensitized Photoelectrochemistry
Mobile cations may or may not be present, complex. Under photoexcitation the charge
depending on the material used. Key density shifts from around the ruthenium
requirements are clearly the rapid regen- ion to around the carboxylated bipyridal
eration of the dye cation, and slow recom- groups (a metal to ligand charge transfer
bination with electrons in the porous film. reaction (MLCT)). This encourages rapid
In addition, there are a number of physical injection into acceptor states in the metal
(permeability into the pores of the film, sta- oxide. It is controversial whether MLCT is
bility, and optical transparency) and chem- essential for rapid charge injection. Sub-
ical (stability, redox potential consistency ps charge injection has also been observed
with TiO2 film and dye) requirements. for porphyrin dyes in which MLCT does
Several approaches have been tried, not occur [82]. Although the basic require-
including polymer gel electrolytes con- ments for electronic structure of the dye
taining I − /I2 [76], inert ionic conduct- are not clear, it has been pointed out [2] that
ing polymers containing I − /I2 [77], hole- electronic coupling between the π ∗ orbital
conducting polymers [78] and amorphous of the dye complex and the conduction
molecular hole conductors such as the band of the semiconductor (a manifold of
chiral organic, OMeTAD [79]. Molecular 3d orbitals in the case of TiO2 ) may enable
hole conductors are preferred to ionic electron injection from dye to semiconduc-
conductors because of the elimination of tor with high yield.
the volatile component, but with penal- Another much less well studied aspect
ties for the charge transport. A peak of charge transport in dyes is the lateral
short-circuit incident photon to current transfer of excitations and charge between
efficiency (IPCE) for OMeTAD of 33% in- dye molecules anchored on the metal
dicates poor charge transport. It is not clear oxide surface. Some evidence for this
whether this results from reduced hole mo- ‘‘hole hopping’’ is provided by studies
bility relative to the ionic mobility of the of quantum efficiency as a function of
liquid electrolyte, or from increased inter- dye coverage, indicating that collection
facial recombination because of reduced probability is affected by the density
charge screening. In general, both are of dye molecules at low coverage [83].
likely to matter. Although hole-transport Hole hopping in triarylamine dyes on
properties are unimportant in the liquid nanocrystalline films has been inferred
cell relative to electron transport, the same in Ref. 84 from studies of the redox
will not be true of solid-state devices. activity of these dye molecules, which
A limited number of inorganic solid- are energetically incapable of transferring
state hole conductors have been consid- charge to the metal oxide films.
ered. These need to be p-type semicon- A limited amount of work has been re-
ductor materials, which can be grown ported on inorganic sensitizers such as
within the pores of the TiO2 film. The II-VI semiconductor QDs in dye-sensitized
most successful materials reported to date systems [10, 85, 86]. QD sensitizers offer
are copper (I) compounds such as copper the advantages of a tunable absorption
iodide and copper thiocyanate [80, 81]. edge and, in principle, tunable excited
state energy level to optimize the matching
5.3.3.3 Dyes of energy levels between sensitizer and
The most commonly used dye is a electron conductor. However, charge sep-
carboxylated ruthenium bipyridal dye aration is evidently poorer than for organic
446 5 Dye-Sensitized Photoelectrochemistry
dyes, possibly due to poorer coupling recombination rate per unit volume. G and
between QD excited state levels and the R may include capture and release of elec-
semiconductor conduction band, and the trons by traps. A similar equation can be
injection yield is correspondingly less. constructed for each type of charge carrier
The remainder of this chapter is con- present in the system. If the form of Jn ,
cerned primarily with electron transport G, and R are known, the continuity equa-
in the porous semiconducting film. This tions can be solved together with Poisson’s
is on the grounds that the high mobili- equation and appropriate boundary condi-
ties and concentrations of charge carriers tions to yield the carrier density functions
in liquid electrolytes make ionic trans- for particular physical situations. Carrier
port a secondary consideration, although populations are related to the physically
it may affect the ease of electronic trans- relevant measurables, current and voltage,
port through effective parameters. Charge through the definition of the current terms
transport within solid-state electrolytes, and the relationship between carrier den-
on the other hand, cannot be neglected sity and electrochemical potential (Fermi
in considering solid-state systems, but, level) at contacts.
at the time of writing, the field is too For a solar cell, the most relevant quan-
young to make a fair review of the mech- tity is the current density at the terminals
anisms and observations. Similarly, hole of the device, J , delivered under steady
transport within the dye layer cannot be state illumination when a given bias V
dismissed, but lies beyond the scope of the is applied between the cell terminals, the
present work. J (V ) characteristic. In the steady state, the
left hand side of Eq. (13) is zero, and equa-
5.3.4 tions are solved to determine J (V ) for a
Theoretical Background given incident spectrum and applied bias.
We can make some general comments
In this section we introduce the equations about G and Jn . For a slab of material
of device physics as normally applied to of uniform absorption oriented normal to
solar cells and outline the ways in which the direction of incidence of the light,
they must be modified for dye sensitized photogeneration at depth x below the
systems. surface is given by
!
5.3.4.1 Basic Equations of Device Physics G(x) = g(E, x) dE (14)
In any electronic device, charge carriers
must obey a set of continuity equations, where
which balance the generation and recom-
bination of carriers [87]. For instance, the g(E, x) = α(E)b0 (E)[1 − r(E)]e−α(E)x
electron density n(r, t) as a function of (15)
position r and time t must obey, where α(E) is the absorption coefficient of
the material at photon energy E, b0 (E) is
∂n 1
= ∇ · Jn + G − R (13) the incident spectral photon flux, and r(E)
∂t q
the surface reflectivity at energy E.
where Jn (r, t) is the electron current den- For a semiconductor, the electron cur-
sity, q the electronic charge, G is the elec- rent can usually be defined in terms of the
tron generation rate, and R is the electron electron quasi Fermi level, EFn , and the
5.3 Charge Transport in Dye-sensitized Systems 447
density, a dark current density that is the system is in general more complicated.
proportional to (eqV /kT − 1), resembling The following factors are important:
that of an ideal diode. More refined
models of p–n junction device physics 1. More carrier types A continuity equa-
allow for recombination in the junction tion should be constructed for each of
region and lead to nonideal J –V behavior. the species involved in charge transfer:
Evidently, the current voltage response electrons, reduced and oxidized ions in
can be described with knowledge of the the hole conductor. In addition, cations
transport parameters Dn and τn for either in the hole conductor are mobile and
carrier, and the device structure. A useful will affect electrostatics, and therefore
derived parameter is the minority carrier they should be included, although these
diffusion length, Ln , do not contribute to charge transfer.
2. Three-dimensional geometry Because of
"
Ln = Dn τn (23) the structure of the junction, the
geometry is three dimensional on the
which determines the thickness of the microscopic scale. The electrochemical
neutral region from which photoexcited junction gives rise to large local electric
minority carriers can be collected effec- fields within the hole conductor at
tively. all points of the interface. These are
The transport parameters, although well typically a few nm in extent in the case
known for common photovoltaic materi- of liquid electrolytes [23, 24].
als, are determined by transient measure- 3. Boltzmann approximation invalid Be-
ments on doped semiconductor samples cause the nanocrystalline film contains
of known composition (for example, Dn a high density of intraband gap states,
from time of flight measurements of the Boltzmann approximation for n
mobility, τn from time resolved photolu- (Eq. 18) may not be valid. Instead, n will
minescence). be determined by some distribution of
Notice that although both electrons and intraband gap states, with distribution
holes are present, current in the neutral re- g(E), and Eq. (17) should apply.
gions is dominated by the minority carrier 4. Nature of photogeneration Photogener-
because light and bias induced carrier gra- ation is a two-stage process that requires
dients are larger. Majority carrier transport absorption of a photon by a sensitizer
is described by similar diffusion equations molecule and injection of an electron
but because the recombination probability from dye to semiconductor.
is so low, the majority carrier density is 5. Nature of current The nature of the
effectively constant across the neutral re- driving force for the electron current is
gion. Gradients in the carrier density may not definitely known. For example, the
arise from resistive losses in low mobility electron mobility is not expected to be
materials. a constant because the relaxation time
approximation, which is used to derive
5.3.4.3 Device Physics for Dye-sensitized Eq. (18), is not necessarily valid.
Systems 6. Nature of recombination The defini-
Now we consider the case of a dye- tion of the electron recombination
sensitized device. On account of the rate is not known. In a dye-sensitized
complex geometry and multiple media, cell, unusually, two routes for electron
5.3 Charge Transport in Dye-sensitized Systems 449
0.6
0.5
Electron concentration
0.4
[arbitrary units]
Light 0.3
where nV (0) is the electron concentration the most relevant to photovoltaic device
in the dark when a bias V is applied performance, but time and frequency
dn
across the terminals, and dx = 0 at the resolved measurements, usually on model
extreme semiconductor–electrolyte inter- systems, are the most revealing about the
face. Equation (29) is more complex than nature of charge transport.
the case of electrons in the p-type region
of a p–n junction solar cell (Eq. 22) be- 5.3.5.1 DC Measurements
cause of the generally nonlinear behavior In the steady state the left-hand side
of the recombination and diffusion, and of Eq. (13) is equal to 0. Measurables
the failure of the Boltzmann approxima- are the steady state photocurrent, pho-
tion. The diffusion coefficient is expressed tovoltage, light and dark current voltage
as a function of n(x) to reflect this. Some characteristics, and the quantum efficiency
approaches to this problem are reviewed spectrum (or IPCE). If the physical origin
later. Elucidation of the mechanisms of of Jn is known, the dependence of such
recombination and diffusion to provide DC measurements on variations in in-
appropriate functional forms for Dn and tensity, wavelength, and bias, deliver the
R is a major experimental challenge, and parameters controlling Jn (for example,
is discussed in the following section. the values of the diffusion length and dif-
fusion constant in the case of diffusion
5.3.5 limited transport).
Experimental Techniques and Results
5.3.5.1.1 Incident Photon to Current
We divide these into three groups: DC, Efficiency IPCE, also called external
time domain, and frequency domain. quantum efficiency, is the ratio of electrons
DC measurements under illumination are collected at the terminals to incident
452 5 Dye-Sensitized Photoelectrochemistry
Rul′(NCS)3
60
Rul2(NCS)2
Rul2[Ru(bpy)2(CN)2]2
40 TiO2
L = 2,2′ − dcbpy
20 L′ = tc − terpy
RuL3
0
400 600 800 1000
Wavelength
[nm]
Fig. 8 IPCE’s of some different sensitizer dyes, including the most
widely used Ru(bpy)2 (NCS)2 , and the recently developed ‘‘black’’ dye,
..
developed by Prof. Gratzel’s group at EPFL [97].
5.3 Charge Transport in Dye-sensitized Systems 453
illumination. In normal operation the de- the operating bias of the solar cell, IPCE
vice is illuminated from the glass substrate at peak wavelength falls to about 45%
side, so that light is absorbed preferen- of its saturation value [93]. This indicates
tially close to the SnO2 : TiO2 interface. If competition between recombination and
the diffusion length is comparable with transport, although it is not trivial to
the absorption depth and collection at the disentangle the contributions from Jn
interface is easy, then high quantum effi- and R.
ciencies are expected for illumination from The IPCE for sensitized ZnO films
this direction. For illumination from the is smaller [64], although it is not clear
rear side (electrolyte side) of a film which whether this is because of poorer electron
is optically thick, most electrons have to transport or poorer electron collection at
cross the film to be collected, and study of the interface with substrate [67].
the IPCE as a function of film thickness
provides information on the electron diffu- 5.3.5.1.2 Dark-current Voltage In the
sion length. Sodergren and coworkers [21, dark, G = 0 and the steady state continuity
98] have applied the diffusion solutions to
Eq. (13) becomes
fit the IPCE of a non-dye-sensitized device
for illumination from electrolyte side and 1 dJn
−R =0 (32)
substrate side, and derive an electron dif- q dx
fusion length of 0.8 µm. Their results also
where R depends (usually strongly) on
confirm, importantly, that electrons gen-
the applied bias V . The dark current
erated close to the SnO2 : TiO2 interface
Jdark (V ) is the current passing though
are collected most efficiently, indicating
the presence of an electric field driving x = 0 when a forward bias V is applied
collection at that interface. Measurements between the terminals. Note that Jdark acts
by Lindstrom and coworkers indicate that in the opposite sense to the short circuit
the electron diffusion length is longer in photocurrent. It is normally this process
dye-sensitized devices [95], as would be ex- which limits the performance of a solar
pected from the absence of photogenerated cell. Analysis of Jdark (V ) can in principle
holes. teach us something about R.
The IPCE at the optimum wavelength In a conventional solar cell, R results
can be interpreted as an indicator of the from the recombination of injected minor-
relative efficiency of electron transport ity carriers. The dark current is rectifying
in variations of the dye-sensitized TiO2 and can be approximated by the form
device. For instance, the addition of
electron acceptors such as O2 and I2 to Jdark = J0 (eqV /mkT − 1) (33)
the electrolyte lead to reduced IPCE at all
wavelengths [95], which can be interpreted where J0 is a constant and the ideality
in terms of a shorter electron lifetime or factor m normally lies in the range 1 ≤ m ≤
shorter diffusion length. 2. m and J0 are related to the mechanism
Under forward bias (electrolyte at a more for charge transport and recombination.
negative potential relative to the electrode The same parameters have been used
potential) the electron density in the to characterize dark currents of dye-
film increases and recombination becomes sensitized cells, although their meaning
significant. At a bias that corresponds to is not clear.
454 5 Dye-Sensitized Photoelectrochemistry
In a dye-sensitized cell in the dark, R re- this number is related to the kinetics of
sults mainly from the scavenging of electri- the reduction reaction, but because inter-
cally injected electrons by oxidized species pretation in terms of transport parameters
(usually I3 − ions) in the hole conductor. depends on the type of junction, for in-
Dark currents typically reach current den- stance, p–n, p–i–n, Schottky barrier, it is
sities of several mA cm−2 at the operating not useful for DSSCs. Interpretations of
bias of the solar cell, and exhibit recti- dark-current behavior in terms of various
fying behavior before being taken over barrier [99] or kinetic [92] models have
by series resistance at current densities been made but are speculative.
of ∼10 mA cm−2 (Fig. 9a). Unlike solid- The magnitude of the dark current
state junctions, the dark current does not can be increased by various treatments
saturate at negative biases because of the that are expected to enhance electron
nonnegligible probability of electron trans- recombination. These include the addition
fer to the film at high reverse biases where of electron scavengers such as oxygen [100]
the film is electron poor. The high series re- to the electrolyte, and pretreatment to
sistances are due to the series resistance of increase the density of defects (oxygen
the conducting glass of some 10–15 ohm- vacancies) in the film [54]. The latter
square. Ideality factors of around 1 [99] effect indicates that oxygen vacancies
and 2 [92] have been reported. In principle act as recombination centers. Any such
107 dark
22 °C
106 40 °C
55 °C
105 70 °C
Current
[mA cm 2]
104
103
102
101
−0.2 −0.1 0.0 0.1 0.2 0.3 0.4
Potential
(a) [V]
0.75
treatment that enhances dark current will short circuit, and that recombination is
reduce the efficiency of charge collection. negligible in this bias range.
At larger forward bias, the net photocur-
5.3.5.1.3 Light-current voltage Although rent drops because of increased electron re-
the most relevant measurement for solar combination. In a conventional solar cell,
energy conversion, steady state J (V ) is the process responsible for the recombina-
not the most useful for analyzing charge tion current is normally the same in light
transport. The J (V ) curve for a typical and dark, and it is common to make the ap-
dye-sensitized cell in standard solar il- proximation that J (V ) = Jsc − Jdark (V ).
lumination is presented in Fig. 10. The Although this may be good for semi-
short-circuit photocurrent, Jsc = J (0), cor- conductor p–n junction devices, it is not
responds to the IPCE integrated over the reliable for dye-sensitized photovoltaic de-
spectrum, vices. Ionization of the sensitizer opens
!
Jsc = η(E)b0 (E) dE (34) up a new path for recombination in light,
which is not active in dark (Fig. 11). There
In a good device, J (V ) remains constant is some evidence [69] that the reverse cur-
into small forward bias, indicating that rent is larger under illumination than in
the photocurrent has indeed saturated at the dark.
X-25 IV System
4.0
Sample PL0710/2
3.5 Oct 30, 1998
tape aperture, black dye
3.0
Area = 0.1863 cm2
2.5 Irradiance = 1000 Wm−2
2.0
Current
VOC = 0.7210 V
[mA]
ISC = 3.824 mA
1.5
Vmax = 0.5465 V
Imax = 3.552 mA
1.0 JSC = 20.53 mA cm−2
Pmax = 1.941 mW
0.5 Fill factor = 70.41%
Efficiency = 10.4%
0.0
−0.5
−0.2 0.0 0.2 0.4 0.6 0.8
Voltage
[V]
Fig. 10 Current voltage response of dye-sensitized cell measured at NREL [101]. This cell
has a short-circuit current density of over 20 mA cm−2 and an open circuit voltage of over
0.7 V. The relatively low slope of I–V near Voc indicates a significant series resistance, largely
due to the sheet resistance of the conducting glass substrate.
456 5 Dye-Sensitized Photoelectrochemistry
Dark Light
Study of the dark current is thus useful R, as does the dark J (V ), but without
for understanding the reaction between the problem of series resistance. For a
electrons and redox couple, but not for dye-sensitized cell it is a more relevant
understanding the limits to photovoltage comparison as it is probing recombination
of the DSSC. Nevertheless, simple models under illumination, and not in the dark.
of J (V ) behavior derived from solid-state Variation of Voc with light intensity has
device physics can been applied to DSSC’s been reported by various authors including
to derive values of effective transport pa- Huang and coworkers [69] and Fisher
rameters. For example, Sodergren and and coworkers [16] for TiO2 and by de
coworkers [21] and Ferber and cowork- Jongh and coworkers [67] for ZnO. For
ers [102] have used a model, based on both materials the relationship tends to
diffusion-controlled current, to fit the J (V ) the form
behavior of DSSCs. mkT
The open circuit photovoltage, Voc , Voc = ln X + Voc (1) (36)
q
corresponds to the point, at which the
recombination current and photocurrent at high light intensities, where X is the
are equal, light concentration factor, Voc (1) is the
open-circuit voltage at unit concentration,
1 dJn and m the apparent ideality factor. Values
− Rn (Voc ) + G(x) = 0 (35)
q dx of m are in the range 1.3 to 1.5, but,
as commented earlier, these values are
There is no net current at Voc . If local not meaningful without a model for the
currents are also small – for instance, if the recombination current.
generation rate is constant so that there are Voc is influenced by surface treatments
no internal diffusion currents – then the and chemical environment. Huang and
first term in Eq. (35) can be neglected and coworkers show that Voc decreases with
the variation of Voc with light intensity increasing concentration of triiodide, as ex-
tells us something about the form of pected for a recombination current which
5.3 Charge Transport in Dye-sensitized Systems 457
is dominated by the reaction of electrons TiO2 ) [12, 103] and with evidence for
with I3 − . Several authors have reported directed photocurrents in dry films [94].
the variation of Voc by surface treatments A comparison of the SPV spectra of an
of the TiO2 . Treatment with pyridine anatase single crystal and nanocrystalline
derivatives [69] and tetrabutylammonium film indicate the presence of a tail of states
phosphate [2] is observed to increase Voc . in the band gap below the conduction band
The effect is attributed to the masking of edge [58].
electrons at the TiO2 surface from elec-
tron scavengers in the electrolyte by inert 5.3.5.2 Time Domain
adsorbed molecules. Treatment of TiO2
Traditionally, time resolved measure-
surface by benzoic acid derivatives is ob-
ments are used to study the mechanisms
served to alter the Voc of solid-state devices
of charge transport and recombination in
in a way that is consistent with modifi-
semiconductor materials.
cation of the surface dipole field by the
adsorbed acids [103].
5.3.5.2.1 Photocurrent Transients Time-
resolved photocurrent at zero or reverse
5.3.5.1.4 Photoconductivity and Photovol-
bias is, like IPCE, a probe of the forces
taic Properties Relevant to the light re-
driving conduction rather than recombi-
sponse of dye-sensitized systems are the
nation. If recombination can be neglected,
photoconductivity and photovoltaic prop-
the transient electron current resulting
erties of nanocrystalline films. Earlier
from an instantaneous or steady state light
it was noted that the conductivity of
stimulus G(x, t), is defined by
nanocrystalline TiO2 is very small in the
dark, around 10−10 –10−8 "−1 cm−1 , ris- ∂n 1 ∂Jn
ing by several orders of magnitude in UV = + G(x, t) (37)
∂t q ∂x
light. Sensitized films are photoconduc-
tive in visible light, even in the absence in one dimension, where the form of J is
of a hole conductor [10]. The relationship not known.
between light intensity and photoconduc- Several configurations have been studied
tivity is superlinear, indicating increased exhibiting some common features. Steady
electron mobility through trap filling [104]. state and square-wave illumination of
It can be shown by simulation that this be- dye-sensitized electrochemical junctions
havior result is consistent with dispersive produces a slow photocurrent, which rises
transport (discussed below) in the pres- to a saturation level [73, 106]. Photocurrent
ence of traps [105]. rise times are extremely long and intensity
Steady state surface photovoltage (SPV) dependent (ms to s, depending on inten-
spectroscopy is useful for determining sity). Both features are attributed to the
the nature of the junction between the trapping and release of photogenerated
nanocrystalline film and substrate. SPV electrons. Schwarzburg and Willig [107]
studies of TiO2 films on SnO2 confirm show with a simple rate equation treatment
that an electric field exists at the interface, that the filling of a single trap level leads
driving electrons into the substrate. This to intensity-dependent photocurrent kinet-
is consistent with the values of the ics, which are qualitatively similar to those
work functions of the two materials in observed. In their case of a single trap level
vacuum (4.85 eV for SnO2 : F, 5.15 eV for Eq. (17) can be resolved into the sum of
458 5 Dye-Sensitized Photoelectrochemistry
conduction (nc ) and trapped (nt ) electrons, filling: at higher injection levels, diffusion
and two continuity equations can be con- should be easier because fewer deep traps
structed for nc and nt with well defined rate are available. Observations [73] of faster
constants for trapping and detrapping. Cao rise times under bias illumination and ki-
and coworkers [106] present transients at netics which depend on the direction of
different light intensities and solve Eq. (37) illumination (effectively, on the intensity
for a light step (G(x, t) = G%0 for t > 0) and& reaching the SnO2 interface) are consis-
dn
diffusive electron current Jn = qDn dx . tent with intensity-dependent diffusion.
Although diffusion leads to a photocur- Transient photocurrents in response to
rent which saturates at a level proportional a laser pulse have been studied in elec-
to the light intensity, it predicts kinetics, trolyte supported and dry films [57, 60,
which are intensity independent. This can 74, 108]. Such transient experiments are
be seen by normalizing n with respect to most informative when the optical depth
G0 in Eq. (37): the solution for n(x, t)/G0 of the film is high, and the light pulse
is independent of G0 . In the data, how- effectively creates a sheet of photogener-
ever, a clear intensity dependence is seen ated carriers, which travels through the
(Fig. 12). The authors propose that the ef- film to the absorbing boundary. This is
fective diffusion coefficient is n-dependent achieved by exciting in the UV where
and demonstrate that the numerical so- absorption of the TiO2 is high. Solbrand
lution of the diffusion equation when and coworkers [74, 108] present laser pulse
D(n) = D0 n, leads to a rise time which induced transient photocurrents for a TiO2
decreases with increasing light intensity. electrochemical junction as a function of
This form of D(n) is somewhat arbitrary, film thickness, electrolyte composition and
but is qualitatively consistent with trap direction of illumination (Fig. 13). The
0.1
0.09
0.08
0.07
I (t) / G 0 / A W1
0.06
0.05
0.04
G 0 = 0.05 mW cm−2
0.03
G 0 = 4 mW cm−2
0.02
0.01
0
0 0.1 0.2 0.3 0.4
Time
[s]
Fig. 12 Transient photocurrent in response to step illumination of different
intensities normalized with respect to the light intensity. At higher intensities,
the photocurrent rise time is faster, indicating an electron density dependent
diffusion coefficient. Replotted from Ref. 106.
5.3 Charge Transport in Dye-sensitized Systems 459
Back-contact
TiO2 –electrolyte interface travel
n(x,t)
through the film to deliver a current at
the collecting interface, from Ref. 74.
t2
The thickness dependence of the t3
transient currents is consistent with
diffusion, but contains some features of
(a) 0 x W
dispersive transport.
7.5 µm
Current 11.3 µm
[a.u.] 14.9 µm
17.4 µm
0 20 40 60 80 100
Time
(b) [ms]
observed film thickness dependence is Solbrand and coworkers [74] is that the
consistent with a diffusion picture, where apparent diffusion coefficient is dependent
the carrier profile generated near the semi- on the conductivity and composition of the
conductor–electrolyte interface at x = W electrolyte. This is consistent with an acti-
spreads out over time to deliver a time- vation energy for electron hopping which
dependent flux of carriers through the depends on the energy required to polar-
absorbing boundary at x = 0. It is evident ize its environment, a phenomenon which
that the transit times are long (tens of ms is familiar from conduction in doped dis-
to s) and the current signal becomes more ordered materials [110]. It is also a clear
dispersive with film thickness. They fit the signal that the electrochemical interface is
slow portion of their transients (an initial very important to transport, and consistent
fast decay is attributed to capacitive effects) with the idea that electrons move along or
to the solution of the diffusion equation for close to the surface.
a delta function stimulus, and derive an ef- Laser pulse photocurrent transients
fective diffusion coefficient, D, of the order in solid-state Schottky barrier junctions
of 10−6 cm2 s−1 . The facts that (1) the ex- between dry TiO2 and a platinum con-
perimental tail in the photocurrent is more tact have been studied by Konenkamp and
persistent than predicted by simulation coworkers [57]. They observe a power law
and (2) the kinetic shape of the nor- dependence of the collected photocurrent
malized current, J (0, t)/G0 , is intensity on time,
dependent, indicate that simple diffusion J (0, t) ∝ t −β (38)
is not a complete description. These fea-
tures are, however, consistent with a model which is characteristic of ‘‘dispersive’’
of ‘‘anomalous diffusion’’, a random transport. Such algebraic photocurrent
walk through energetically distributed trap transients are familiar from the physics
states, as discussed by Nelson [109]. A of disordered semiconductors and have
second and important observation by been shown to result from the anomalous
460 5 Dye-Sensitized Photoelectrochemistry
[A]
0.1 mA
−4
−5 20 µA
−6
−8 −7 −6 −5 −4 −3
Log10 (Time)
[s]
films [63] and others have used it to esti- dye cation [90]. These and subsequent [82,
mate charge stored in TiO2 [113]. Results 91] studies showed that injection occurs in
indicate an accumulated electron density sub picosecond time scales in a wide range
of several electrons per ∼15 nm diameter of chemical environments, supporting the
nanoparticle, and are consistent with – but assertion made earlier that absorption is
do not prove – an exponential density of equivalent to photogeneration in a DSSC.
surface states. A variant of the junction Although transient absorption is not a
recovery technique, where recovered cur- direct probe of transport, it can be used
rent from an illuminated electrochemical to probe the kinetics of recombination
junction at open circuit is measured as a processes, which compete with transport.
function of light intensity, has been ap- Haque and coworkers [115] studied the re-
plied to nanocrystalline TiO2 by Duffy and combination between electrons in the TiO2
coworkers [114]. These authors also infer electrode and photoionized dye molecules
an exponential density of states from their by monitoring the concentration of the
data and find that the recovered charge dye cation following a laser pulse. By
varies algebraically with time, which they using a redox inactive environment, all
attribute to a reaction of electrons with I3 − , other mechanisms to reduce the dye cation
which is second order in n. were suppressed. They found that this
recombination reaction is slow (∼ms) at
5.3.5.2.3 Transient Absorption Time-re- zero applied bias, but it is very strongly
solved absorption can be used to mon- dependent on the bias applied to the
itor the appearance or disappearance of junction, indicating that electrically in-
the reduced or oxidized states of molec- jected electrons are capable of reducing
ular species. Ultra fast time-resolved the dye cation. Importantly, they find that
absorption was first used by Tachibana this recombination reaction becomes suf-
and coworkers to measure the rate of ficiently fast (ns-µs time scale) to compete
injection from dye-excited state into semi- with the reduction of dye cation by iodide
conductor, probing the absorption of the at voltages near open circuit, and so should
462 5 Dye-Sensitized Photoelectrochemistry
0.9
0.8
0.7
−200 mV
∆OD(t) / OD0
0.6
0.5 −300 mV
0.4
−400 mV
0.3
−500 mV
0.2
0.1
0
10 000 100000 1000000 10000000 100000000
Time
(a) [ns]
1000
Electrons per cation
n = At−0.46
100
10
1
10 1000 100000 10000000
t50%
(b) [ns]
Fig. 15 (a) Kinetic decay curves for the rereduction of the dye cation by
electrons in the TiO2 film. The strong bias dependence indicates that
electrically injected electrons are capable of reducing the dye. The smooth
lines are fits using a continuous time random walk model of electron
transport and assuming diffusion limited recombination. (b) Power law
relationship between the electron density and the half life, t50% , for
reduction of the dye cation. In the random walk model, the exponent α is
related to the characteristic temperature of the density of trap states.
known to absorb. This has been used Olson and coworkers to observe the ki-
by Hagfeldt and coworkers to observe netics of electron accumulation following
electron recombination kinetics following the application of a potential step in
injection by a UV laser pulse [112] and by the dark [113]. Another novel probe of
464 5 Dye-Sensitized Photoelectrochemistry
80
0.000 60
Im [∆i (ω)/e Φ(ω )]
ω min/rad s−1
0.1 Hz
1 KHz
1 Hz 40
−0.005 100 Hz
8 Hz
20
−0.010 (c) (c)
0
0.00 0.01 0.02 0 1 2 3
(a) Re [∆i (ω )/e Φ(ω)] (b) (Φ/1016 cm−2s−1)0.5
Fig. 16 (a) Complex plane plot of characteristic frequency, which varies
intensity-modulated photocurrent spectrum for a approximately like the square root of the light
nanocrystalline TiO2 film. A characteristic time is intensity. In a trap-free system this frequency
assigned from the frequency at which the would be independent of light intensity (from
imaginary part of the response function is Ref. 121).
greatest. (b) Intensity dependence of the
to be low; in this case multiple scattering is increases and so does Dn . Later work by
believed to dominate transport. The theory the same authors [50, 16] showed that the
of trap limited IMPS response is devel- diffusion coefficient derived this way varies
oped in Ref. 123, where a distinction is with light intensity approximately like
made between multiple trapping and dis-
Dn ∝ X 0.6 (47)
persive transport in the frequency domain,
and in Ref. 124. Fisher and coworkers combine this ob-
Dloczik and coworkers [92] later used servation with the observed (exponential)
IMPS to obtain an estimate of the dependence of photovoltage on light inten-
electron diffusion coefficient in sensitized sity to derive a relationship between Deff
nanocrystalline TiO2 electrodes. This is and electron Fermi level,
done by transforming the time dependent
Dn ∝ e(1−β)EFn /kT (48)
diffusion equation (Eq. 22) into frequency
space and fitting the result for the which is consistent with an exponential
photocurrent modulation to the measured density of trap states of the form,
values to extract values for an effective
g(E) ∝ eβE/kT (49)
diffusion coefficient, Dn and electron
lifetime, τ . This analysis does not explicitly as Eq. (42), with β = α. They deduce a
account for the trapping and detrapping value of 0.57 for the exponent β from their
of electrons. However, it is pointed out data, which corresponds to a characteristic
that for the case of a single trap level, temperature T0 of 526 K [16].
the effective diffusion coefficient is related
to the true diffusion coefficient, Dc , for 5.3.5.3.2 IMVS In IMVS, the driving
conduction band electrons through force is again a modulated light source and
nc the measured response is the modulation
Dn = Dc (46)
n in open circuit photovoltage. The tech-
in analogy with Eq. (39). As injection level nique was first applied to dye-sensitized de-
increases, the proportion of free electrons vices by Schlichthorl and coworkers [125].
466 5 Dye-Sensitized Photoelectrochemistry
The time constant, 1/ωmin , delivered by The origin of the power law dependence
IMVS, corresponds to an effective lifetime of recombination time on light intensity,
for recombination, τrec , whereas the time consistent with Eq. (52) is not clear. Such
constant delivered by IMPS results from a behavior has been observed elsewhere
combination of recombination and charge but attributed alternatively to a reaction
collection, between electrons and triiodide, which
# $ is second order in n [16], and to charge
1 1 −1 trapping [11, 128].
τIMPS = + (50)
τrec τcoll
although at short circuit the time constant 5.3.5.3.3 Electrical Impedance Spectro-
for collection, τcoll , often dominates. It is scopy (EIS) EIS measures the current
clearly important to be able to relate these response to a modulation in applied
experimentally accessible time constants bias, and can be used in the light
to the functionally relevant quantity, the or dark. Traditionally, EIS results are
quantum efficiency ηcoll for charge collec- interpreted in terms of an equivalent
tion, but that is not trivial in a nonlinear circuit and information extracted about
system. Schlichthorl and coworkers [126] relevant capacitive, resistive, and faradaic
processes. The technique has been applied
showed that although the linear approxi-
to nanocrystalline TiO2 junctions in the
mation,
dark by various groups [19, 22] and a
# $
1 + τcoll −1 film capacitance extracted and interpreted
ηcoll = (51) using a simple equivalent circuit. Results
τrec
confirm that a potential drop does exist
is not adequate, a model based on elec- within the film in the dark. In reverse
tron diffusion with nonlinear relationships bias, where the film is depleted, the film
between electron density and recombina- capacitance appears to be bias independent
tion rate and dominated by the potential drop at
R ∝ nm1 (52) the TiO2 –TCO interface. In forward bias
the film becomes populated and a high
where m1 ≈ 2.2, and between total and and a bias dependent capacitance develops
free electron density at the large TiO2 –electrolyte interface,
suggesting an increased charge storage
ncb ∝ nm2 (53) capacity as surface or conduction band
states become filled.
where m2 ≈ 2.7, is able to explain observed Van de Lagemaat and coworkers [93]
IMVS and IMPS behavior. The two apply EIS to illuminated nanocrystalline
power law relationships are deduced from TiO2 electrodes and again observe a
observations of τrec versus light intensity capacitance which increases, approxi-
using IMVS, and Jsc versus electron mately exponentially, with applied forward
density using IMPS. The strong power bias. They attribute this to the charging of
law dependence of ncb on n implied surface states and show that the behavior
by the comparison is consistent with is consistent with an exponential density
an exponential density of states, again of states. They further conclude that the
of the form (40), but with α = 0.37 or potential drop across the TiO2 –TCO in-
T0 = 810 K [127]. terface extends over a narrow region, and
5.3 Charge Transport in Dye-sensitized Systems 467
are not easily addressable with equivalent is considered only in one dimension.
circuit models for planar interfaces. Some consider transport within a single
The fact that recombination is slow nanoparticle, or other highly simplified
and transport relatively efficient at short geometry. There is clearly a need to move
circuit, is qualitatively attributed to the ef- on to models incorporating the interpene-
fects of screening of electrons by cations trating features of the morphology.
in the electrolyte; and to the absence of Looking to the future, the following is-
mobile photogenerated holes. Recombina- sues remain to be addressed quantitatively:
tion with the oxidized species in the hole the relationship between material proper-
conductor is normally assumed to be the ties and the apparent density of states; the
main loss process for electrons, but re- effect of hole conductor composition and
combination with dye cations cannot be mobility on electron transport; Coulom-
ruled out. bic effects of nanoparticle charging and
Continuity equation approaches involv- electron-hole interactions. An improved
ing nonlinear forms for the electron diffu- description of charge transport is needed to
sion and recombination terms are capable replace the simple diffusion model. Tech-
of reproducing some aspects of observed niques developed to deal with disordered
behavior [123, 126]. The challenge is to conductors in the past may prove use-
predictively relate the forms for recombi- ful here. Multiple-trapping models such
nation and diffusion to known material as the continuous time random walk can
properties of the system. explain certain dispersive features of cur-
Chemical environment strongly influ- rent and recombination kinetics. Another
ences transport. Adsorption and interca- tool that offers the potential to describe
lation influence the potential distribution some features of the behavior of these sys-
at the interface, affecting the tendency for tems is percolation theory. This has already
electrons to accumulate in the film and the been used to describe hole conduction in a
photovoltage. Composition of electrolyte polymeric hole conductor [130]; to describe
influences rate of electron scavenging. charge hopping between quantum sized
Film history influences the density of traps ZnO nanoparticles [65]; and to describe
and recombination centers. hole migration between dyes [83, 84].
In nanoparticulate semiconductors other
than TiO2 , quantum confinement may be- Acknowledgment
come important.
In solid-state systems charge transport The author is grateful to James Durrant,
within the electrolyte is expected to be rate Sven Sodergren, Thierry Lutz, Carol Ol-
limiting. Even in liquid electrolyte systems son, Mili Biswas, and Ellen Moons for
there is some evidence for an influence helpful discussions, and acknowledges the
of electrolyte on electron transport. This financial support of the Engineering and
is expected from the screening and reor- Physical Sciences Research Council and
ganization energy. In solid-state systems the Greenpeace Environmental Trust.
the absence of macroscopic fields cannot
be assumed and drift may be expected to References
become significant. 1. M. Grätzel, Nanocrystalline electronic junc-
Most models applied to date use an effec- tions, Stud. Surf. Sci. Catal. 1997 103,
tive medium in which electron transport 353–375.
5.3 Charge Transport in Dye-sensitized Systems 469
2. A. Hagfeldt, M. Gratzel, Light-induced re- (X = Cl-, Br-, I-, Cn-, and Scn-) on nanocrys-
dox reactions in nanocrystalline systems, talline Tio2 electrodes, J. Am. Chem. Soc.
Chem. Rev. 1995, 95(1), 49–68. 1993, 115(14), 6382–6390.
3. B. O’Regan, M. Gratzel, A low-cost, high- 15. E. Topoglidis, T. Lutz, R. L. Willis et al., Pro-
efficiency solar-cell based on dye-sensitized tein adsorption on nanoporous TiO2 films:
colloidal TiO2 films, Nature 1991, 353(6346), a novel approach to studying photoinduced
737–740. protein/electrode interactions, Faraday Dis-
4. N. Vlachopoulos, P. Liska, J. Augustyn- cuss. 2000, 116, 35–46.
ski et al., Very efficient visible-light energy 16. A. C. Fisher, L. M. Peter, E. A. Ponomarev
harvesting and conversion by spectral sensi- et al., Intensity dependence of the back
tization of high surface-area polycrystalline reaction and transport of electrons in dye-
titanium-dioxide films, J. Am. Chem. Soc. sensitized nanocrystalline TiO2 solar cells,
1988, 110(4), 1216–1220. J. Phys. Chem. B 2000, 104(5), 949–958.
5. C. Bechinger, S. Ferrer, A. Zaban et al., 17. N. Myung, S. Licht, J. Electrochem. Soc.
Photoelectrochromic windows and displays, 1995, 142, L129–L132.
Nature 1996, 383(6601), 608–610. 18. D. Fitzmaurice, Using spectroscopy to
6. J. Sotomayor, G. Will, D. Fitzmaurice, Pho- probe the band energetics of transpar-
toelectrochromic heterosupramolecular as- ent nanocrystalline semiconductor-films,
semblies, J. Mater. Chem. 2000, 10(3), Sol. Energy Mater. Sol. Cells 1994, 32(3),
685–692. 289–305.
7. E. Topoglidis, A. E. G. Cass, G. Gilardi 19. F. Cao, G. Oskam, P. C. Searson et al., Elec-
et al., Protein adsorption on nanocrystalline trical and optical-properties of porous
TiO2 films: an immobilization strategy for nanocrystalline Tio2 films, J. Phys. Chem.
bioanalytical devices, Anal. Chem. 1998,
1995, 99(31), 11974–11980.
70(23), 5111–5113.
20. J. Bisquert, G. GarciaBelmonte, F. Fabregat-
8. S. Y. Huang, L. Kavan, I. Exnar et al., Rock-
Santiago, Modelling the electric potential
ing chair lithium battery based on nanocrys-
distribution in the dark in nanoporous
talline Tio2 (anatase), J. Electrochem. Soc.
semiconductor electrodes, J. Solid State Elec-
1995, 142(9), L142–L144.
trochem. 1999, 3(6), 337–347.
9. C. Bechinger, B. A. Gregg, Development of
21. S. Sodergren, A. Hagfeldt, J. Olsson et al.,
a new self-powered electrochromic device
for light modulation without external power Theoretical-models for the action spectrum
supply, Sol. Energy Mater. Sol. Cells 1998, and the current-voltage characteristics of
54(1–4), 405–410. microporous semiconductor-films in pho-
10. P. Hoyer, R. Konenkamp, Photoconduction toelectrochemical cells, J. Phys. Chem. 1994,
in porous Tio2 sensitized by Pbs quantum 98(21), 5552–5556.
dots, App. Phys. Lett. 1995, 66(3), 349–351. 22. A. Zaban, A. Meier, B. A. Gregg, Elec-
11. J. Nelson, S. A. Haque, D. R. Klug et al., tric potential distribution and short-range
Observation of trap limited recombina- screening in nanoporous TiO2 electrodes,
tion in dye-sensitized nanocrystalline metal J. Phys. Chem. B 1997, 101(40), 7985–7990.
oxide electrodes, Phys. Rev. B 2001, 63, 23. D. Matthews, P. Infelta, M. Gratzel, Calcu-
20 5321. lation of the photocurrent-potential char-
12. D. Cahen, G. Hodes, M. Gratzel et al., Na- acteristic for regenerative, sensitized semi-
ture of photovoltaic action in dye-sensitized conductor electrodes, Sol. Energy Mater. Sol.
solar cells, J. Phys. Chem. B 2000, 104(9), Cells 1996, 44(2), 119–155.
2053–2059. 24. B. A. Gregg, A. Zaban, S. Ferrere, Dye sen-
13. C. A. Kelly, F. Farzad, D. W. Thompson sitized solar cells: energetic considerations
et al., Cation-controlled interfacial charge and applications, Zeitschrift Fur Physikalis-
injection in sensitized nanocrystalline TiO2 , che Chemie-International, J. Res. Phys.
Langmuir 1999, 15(20), 7047–7054. Chem. Chem. Phys. 1999, 212(Pt1), 11–22.
14. M. K. Nazeeruddin, A. Kay, I. Rodicio et al., 25. F. Pichot, B. A. Gregg, The photovoltage-
Conversion of light to electricity by cis- determining mechanism in dye-sensitized
X2bis(2,2* -Bipyridyl-4,4* -Dicarboxylate) solar cells, J. Phys. Chem. B 2000, 104(1),
Ruthenium(Ii) charge-transfer sensitizers 6–10.
470 5 Dye-Sensitized Photoelectrochemistry
interparticle contact was specified as a rea- lithium salts were added to the phase of
son for the poor photovoltaic performance organic p-type conductors [6, 7]. It was
of early sensitized TiO2 /polymer com- postulated that the lithium salt might
posites [3–5]. However, Greenham and screen space charge layers, which are
coworkers realized a CdSe/MEH-PPV expected to build up in the illuminated het-
(polymethoxyethyl hexyloxy-p-phenylene- erojunction. The elimination of the space
vinylene) solar cell made via simple spin charge control of the photocurrent could
coating techniques, with a maximum in- thereby explain the observed increase in
cident photon-to-electron conversion effi- photovoltaic performance.
ciency (IPCE) of 12%, suggesting efficient The contacting electrodes also have to
electron percolation in the CdSe net- be carefully chosen. The cathode mate-
work [4]. Decapping of the nanoparticles rial needs to form an ohmic contact to
before the formation of the composite was the p-type material so as to avoid injec-
essential for the functioning of the de- tion barriers that would impede efficient
vice, a fact that has to be considered in charge collection. The anode needs to
future attempts to form composites with form a blocking contact to the p-type
dye-derivatized nanometer-sized particles. material so as to avoid internal shortcir-
Once a positive charge is injected suc- cuits. Generally, transparent conducting
cessfully into the p-type material, it has to metal oxide layers such as FTO (fluo-
be transported efficiently to the contacting rine doped tin(IV)oxide) or ITO (indium
electrode before interfacial recombination doped tin oxide) are used as the backcon-
can occur. To allow for such efficient tact onto which the nanocrystalline film
charge collection, both the p- and n-type is applied. Iodine is unique, because it
material need sufficient conductivities so shows a very high overpotential at these
as not to impose high internal resis- materials. However, most solid p-type ma-
tance and thereby limit charge collection. terials will form ohmic contacts with these
Screening of charges by the displace- transparent metal oxides causing internal
ment of ions plays an important role short circuits at the points of contact [8].
in the charge collection process in pho- This applies both to organic and inorganic
toelectrochemical solar cells. Solid-state materials. Consequently, whenever trans-
dye-sensitized heterojunctions also exhibit parent conducting metal oxide layers are
increased photovoltaic performance, when used in solid heterojunctions, care has to
ionic species introduced in the form of be taken to avoid direct contact between
5.4 Solid State Dye-sensitized Solar Cells – An Alternative Route Towards Low-Cost Photovoltaic Devices 477
the p-type component of the sensitized metal oxide electrode and the CTM inside
heterojunction and the back contact of the the sensitized heterojunction.
photoanode. The application of organic CTMs in
photovoltaic devices is a new research
5.4.2.1 Organic CTMs Applied to field, which still is in its infancy. Figure 2
Nanoporous Dye-sensitized Metal Oxide overviews the organic charge transfer ma-
Electrodes terials that were scrutinized with regard
Organic electronic materials nowadays are to their application in solid-state dye-
successfully replacing inorganic semicon- sensitized solar cells. A much wider range
ductor components in a growing num- of materials is mentioned in the patent
ber of electronic devices such as xero- literature [13–15].
graphic ink drums or light emitting diodes The charge transfer events taking place
(LEDs) [9, 10]. Further applications will in dye-sensitized heterojunctions are de-
soon follow including transistors [11], or picted in Fig. 1. Visible light is absorbed
entire integrated circuits fabricated on by the sensitizing dye. Electron injection
plastic [13]. Very recently, these materials from the excited state of the dye into the
conduction band of TiO2 (1) is followed
also have been scrutinized with regard to
by subsequent hole transfer from the pho-
their applicability in photovoltaic devices.
tooxidized dye to the organic CTM, regen-
In the following section, the application
erating the dye’s original ground state (2).
of organic materials in conjunction with
dye-sensitized mesoporous metal oxide Dye∗ −−−→ Dye+ + e− (cb TiO2 ) (1)
electrodes is reviewed, whereas alterna- + +
Dye + CTM −−−→ Dye + CTM (2)
tive conceptual approaches toward organic
solar cells that do not directly involve dye- The latter process competes with recap-
sensitization are briefly discussed in the ture of the injected electron by the oxidized
final section of this chapter. dye (3):
Organic synthetic chemistry is a pow-
erful tool used to develop materials with Dye+ + e− (cb TiO2 ) −−−→ Dye (3)
very specific electronic and optical proper-
The double injection process described
ties. A large variety of electronically active
by Eqs. (1) and (2) generates mobile major-
materials with excellent, reproducible ma-
ity carriers in both the TiO2 and the hole
terial properties became available over the conductor phases, which are subsequently
last few years. These organic materials are collected at the contact electrodes. A sec-
still lagging their inorganic counterparts ond mechanism can be postulated that
regarding their semiconducting proper- involves photoinduced hole transfer from
ties such as conductivity and mobility. the electronically excited dye to a neighbor-
Organic materials, however, offer an al- ing charge transport molecule, followed
most unlimited potential to vary their by regeneration of the photoreduced state
specific properties. Furthermore, most of by electron injection into the TiO2 . Both
them can be prepared easily in their amor- mechanisms lead to the same initial charge
phous state. This is important because separated state, which results in an elec-
crystallization would most probably impair tron in the conduction band of the TiO2
the formation of a good contact between and a positive charge, localized on a hole
the mesoporous surface of the sensitized conductor molecule.
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.
attribute the low mobility of electrons in higher than previously reported, but still
the final device responsible for the poor did not exceed 0.1%.
performance. Incomplete pore filling and Significant improvement in efficiency
low conductivities might have been fur- were achieved recently in a novel solar
ther reasons for the low efficiency, because cell device applying the amorphous
thermal evaporation is not likely to trans- organic CTM 2,2$ ,7,7$ -tetrakis(N ,N -di-p-
port organic material deep into the porous methoxyphenylamine)9,9$ -spirobifluorene
structure and because the charge carrier [7] (spiro-OMeTAD; Fig. 2). Photoinduced
density inside the hole conductor matrix charge carrier generation was shown to
was low and merely influenced by acciden- be very efficient. A solar cell based on
tal doping as a result of the presence of spiro-OMeTAD converts photons to elec-
oxygen. tric current with a strikingly high yield
The first solid-state dye-sensitized het- of 50% [20]. The new hole conductor
erojunction of TiO2 and a semiconduct- contained a spiro-center, which was in-
ing polymer was reported by Murakoshi troduced to improve the glass-forming
and coworkers [19, 7]. They formed a properties and prevent crystallization of
solid-state heterojunction by photoelec- the material. Its glass-transition tempera-
trochemical polymerization of pyrrole ture of Tg = 120 ◦ C, measured by differen-
into a nanoporous TiO2 film, sensitized tial scanning calorimetry, is much higher
with Ru(dcbpy)2 (SCN)2 . Direct contact of than that of comparable hole conduc-
polypyrrole and the FTO back contact was tors. The methoxy groups were introduced
suppressed by electrodeposition of a thin to match the oxidation potential of the
TiO2 film before the adsorption of the dye CTM to that of the applied sensitizer
onto the nanocrystalline TiO2 electrode. Ru(dcbpy)2 (SCN)2 .
The doping density of the polymer was Figure 3 shows a cross-section scheme
controlled electrochemically. Solar energy of the sensitized TiO2 /spiro-OMeTAD
conversion efficiencies were somewhat heterojunction mentioned earlier. The
50
40
30
IPCE
[%]
20
Mc2
10
Mc3
Ru(II)L 2(SCN) 2
0
400 500 600 700 800
Wavelength
[nm]
O OH
O
N
S
S S
N
O OH
O
N
S
S S
N S
Mc3 O N
OH
O
Fig. 4 Photocurrent action spectra (IPCE) obtained for solid
TiO2 /spiro-OMeTAD heterojunctions, sensitized with three different
sensitizers. (For device structure, see Fig. 3.) The dyes were absorbed from
above solution for 48 hours at room temperature. All devices are based on
4-µm thick nanocrystalline TiO2 electrodes and spiro-OMeTAD as hole
conducting layer (0.070 mol% spiro-OMeTAD++ (PF6 ) in respect to
spiro-OMeTAD; the spin coating solution contained 15 mM
Li[(CF3 SO2 )2 N]). The counter electrode was a 10-nm thick gold layer.
482 5 Dye-Sensitized Photoelectrochemistry
finding strongly suggests that recombi- sites are represented by Ti (IV) ions on the
nation of photogenerated electrons while surface, which, because of their Lewis acid-
migrating toward the collecting electrode ity, are prone to interact with the pyridine
is a major loss mechanism. Applying a derivative. Similar results were obtained by
simple recombination model, a free elec- Huang and coworkers [28], who attributed
tron diffusion length of about 1 µm could the observed increase in Voc to either a
be determined for the observed junction. charging of surface states or a shift of the
Charge recombination occurring during conduction band edge.
charge collection therefore could be iden- Different techniques have been pro-
tified as major loss mechanism of the cell. posed to fill sensitized nanoporous struc-
A strong increase in energy conversion tures with organic material. Evapora-
efficiency could be observed when ter- tion [19], electrodeposition [7] and spin
tiarybutylpyridine was introduced into the coating [6] techniques were described in
matrix of the organic hole conductor [20]. this respect. Little is known regarding inti-
Although the typical open circuit volt- macy of the contact between the sensitized
age (Voc ) in absence of tertiarybutylpyri- metal oxide surface and the organic CTM
dine reaches approximately 350 mV, it when simple spincoating procedures are
can easily exceed 700 mV in the pres- used. Therefore complete pore filling still
ence of tertiarybutylpyridine. This leads remains a critical problem. Although spin
to an increase in white light conversion coating or dip coating techniques are very
efficiency from 0.74% to 1.75% (AM1.5, simple and inexpensive, electrodeposition
10 mW cm−2 ). Very similar effects were might still be a more controlled technique.
reported for classical sensitized I − /I3 − In situ polymerization of monomeric pre-
electrolyte/TiO2 junctions [2]. The authors cursors appears to be the most reliable
rationalized the effect in terms of 4-tert- way to fill nanoporous electrodes with
butylpyridine adsorption at the TiO2 sur- large-molecular-weight polymers. Very re-
face blocking surface states that are active cently it was shown that low-molecular-
as intermediates in the interfacial charge weight organic semiconductors also can
recombination. They presumed that these be electrodeposited into nanoporous TiO2
5.4 Solid State Dye-sensitized Solar Cells – An Alternative Route Towards Low-Cost Photovoltaic Devices 483
electrodes [29], which thereby represents a combinations of wide band gap p-type and
very powerful technique to deposit poorly n-type semiconductors are energetically
soluble materials sush as perylenes or ph- favorable. Organic and organo-metallic
thalocyanines into porous structures. dyes easily can be adjusted with regard
Many questions regarding the function to their redox potentials and absorption
of these sensitized inorganic-organic hy- characteristics to match the band positions
brid materials still remain unsolved. Most of these materials. In principle, the dye also
importantly, the factors controlling the might be chosen to function as a flexible
electronic properties of the junction have buffer between the two nanocrystalline
to be identified and understood. First steps phases, reducing strain effects, which arise
toward a better control of the interface have as a result of the lattice mismatch of the
been made. The dipole field, present at the two semiconductor materials. A potential
interface because of the self-assembled advantage of inorganic semiconductors
monolayer of organic molecules proved over their novel organic counterparts
to strongly affect the junction’s electronic is their generally high charge mobility.
properties [30]. Further systematic studies However, the very limited choice of
will hopefully help to widen the under- potentially interesting materials is a clear
standing of such devices, leading the way drawback, compared to the wide choice of
to increased device performances. organic CTMs.
Two mechanisms can contribute to the
5.4.2.2 Inorganic Nanoporous formation of the initial charge separated
Dye-sensitized Heterojunctions state. The electronic excitation of the dye
In dye-sensitized inorganic heterojunc- either could be followed by electron injec-
tion solar cells, a monolayer of dye is tion into the n-type semiconductor or by
sandwiched between two wide band gap hole injection into the p-type semiconduc-
semiconductors; one of them exhibits a p- tor. Both mechanisms have been shown
type and the other an n-type conduction to occur and both are likely to contribute
mechanism. Inorganic p-type semicon- to the charge separation mechanism [32].
ductors were successfully applied in an However, no studies were undertaken to
attempt to replace the liquid electrolyte differentiate which of the two possible
in photoelectrochemical solar cells, and mechanisms mainly is responsible for the
in fact the first solid-state dye-sensitized initial formation of the charge separated
photovoltaic device described in the litera- state.
ture was based on a wide band gap p-type Although dye sensitization of wide band
semiconductor material [31]. gap n-type semiconductors has been com-
Such junctions mainly are of inorganic prehensively studied, little is known re-
character, however, the dye present at the garding the sensitization of p-type materi-
junction’s interface still plays the central als. Sensitized electron injection into an n-
role in the initial formation of the charge type semiconductor and hole injection into
separated state. The sensitizer takes on a p-type material are analogous processes.
the role of both the light absorber and The misbalance might be explained simply
acts as the link between the two inorganic in terms of the wider availability of sta-
materials. The choice of inorganic wide ble wide band gap n-type materials, which
band gap semiconductors applicable in allow for the preparation of nanoporous
such junctions is small, and only few films with excellent electronic properties.
484 5 Dye-Sensitized Photoelectrochemistry
Only a few materials were studied con- a columnar morphology, oriented per-
cerning their applicability to dye-sensitized pendicular to the substrate surface could
hole injection processes. Among those be realized, using an electrodeposition
are different copper(I) compounds (e.g. technique. These highly anisotropic dye-
Cu(I)SCN, Cu(I)I, Cu2 (I)O [33–35]) and sensitized films were shown to form a
nickel(II) oxide [36]. Photovoltaic perfor- complete inner surface electrical contact
mances of such devices are orders of when p-type CuSCN was electrodeposited
magnitudes poorer than those of classical into the pores (Fig. 6). These junctions ex-
dye-sensitized photoelectrochemical solar hibited external quantum efficiencies of
cells based on n-type materials. Substan- up to 50%, and incident energy conver-
tial advantages could arise if an efficient sion efficiencies of 1.5%. Dye-sensitized
photo–hole injection process would be heterojunctions made from colloidal ZnO
available. The formation of solid-state tan- particles on the other hand, showed maxi-
dem solar cells would become feasible, mum external quantum efficiencies of only
and a quantum step in device efficiency of 20%. It could thereby be shown that lattice
dye-sensitized solar cells could be at reach. mismatch is not an intrinsic problem of
However, because of the poor performance dye-sensitized inorganic heterojunctions
of all known photocathodes, a combination but can be overcome by choosing ap-
propriate device assembly techniques and
of available photoanodes and photocath-
materials. The sensitizing dye might also
odes to a tandem device always results in a
exhibit a certain buffer function between
device that is photovoltaically less efficient
the two crystal structures, especially if the
than the photoanode on its own. The con-
dye is substituted with two sets of at-
cept for electrolyte-based tandem cells ex-
tachment groups, each of them strongly
ists. However, it contains strong potential
interacting with one of the semiconductor
to improve the photovoltaic performance
components.
in both electrolytic and in solid-state, dye-
Solid-state dye-sensitized TiO2 /CuI het-
sensitized solar cells.
erojunction were recently realized by Ten-
Analogous to their photoelectrochemi- nakone and coworkers [38]. The authors
cal counterparts, efficient dye-sensitized reported on surprisingly high solar energy
solid-state solar cells require a nanoporous conversion efficiencies, showing that in
junction structure. The first solar cell the future it might be feasible to form solid-
of this type was reported in 1995 [33] state dye-sensitized solar cells with similar
in form of a TiO2 /Cu(I)I heterojunction, efficiencies as their photoelectrochemical
sensitized with a cyanidine dye. When counterparts.
forming an inorganic interpenetrating net-
work on the nanometer scale, the crystal 5.4.3
lattice mismatch of the two inorganic Alternative Organic Solid-state Approaches
semiconductors is likely to impair the for-
mation of a good contact between the 5.4.3.1 Organic Solar Cells
two nanocrystalline structures. Recently, Thin films of organic pigments are known
O’Regan and coworkers reported on a sen- to exhibit semiconducting properties from
sitized nanocomposite of ZnO and CuSCN their use in xerography; Morel, Ghosh,
forming a bicontinuous network hetero- Feng and coworkers [39, 40] reported on
junction [37]. Nanoporous ZnO films with a solar cell in 1978 based on the Schottky
5.4 Solid State Dye-sensitized Solar Cells – An Alternative Route Towards Low-Cost Photovoltaic Devices 485
SnO2
nanocomposite of ZnO and CuSCN, as
described by O’Regan and
coworkers [37]. (a) Layered structure of Dense ZnO
the dye-sensitized heterojunction.
Glass support
(b) Fractured edge of a complete DSH, CuSCN
showing the SnO2 layer, the porous
ZnO +
ZnO/dye/CuSCN layer, and some of the dye + +
overlying CuSCN. (c) Fractured edge of CuSCN h+
a porous ZnO layer, without CuSCN, e−
showing the columnar structure
Aluminium
perpendicular to the substrate.
Graphite
(d) Schematic of the ZnO/dye/CuSCN
Light
interface with dye molecules shown to
scale, but at low packing density.
(a)
1 µm
a metal, forming a Schottky type cell such control of the charge carrier density via
as the cell reported by Morel and cowork- chemical doping with electron acceptors
ers. A rectifying contact can also be formed such as iodine for p-type materials [48]
by the formation of a p-n junction between or with donors such as H2 for n-type
the light-absorbing, semiconducting dye materials [49] can improve drastically the
and a second semiconductor of opposite conductivity and device performance.
conduction type. This second semicon- An interesting approach partially over-
ductor either can be an inorganic wide coming the problem of low conductivity
band gap semiconductor or a second light- and limited exciton diffusion widths is
absorbing, semiconducting dye, giving rise to combine two organic p-n junctions in
to an organic/inorganic or organic/organic series, leading to a tandem cell with a dou-
p-n heterojunction. bled photovoltage, compared to the simple
The main limitation of all three device p-n structure device [50].
types is the limited diffusion length of ex-
citons in organic materials, which usually 5.4.3.2 Interpenetrating Networks
does not exceed several tens of nanome- As described in the previous section, the re-
ters. Therefore the photovoltaically active combination of an exciton before it reaches
region in a thin-film solar cell is limited the built-in field region is the main loss
to a very thin layer close to the junc- mechanism in organic thin-film solar cells.
tion. This usually is too thin to assure This problem can be overcome by the
good light harvesting over a wide spectral formation of an interpenetrating p-n het-
range. Very similar to conventional inor- erojunction in which the average width
ganic p-n junction solar cells, the exciton of the p-type and n-type domains are
lifetime strongly decreases in the presence smaller than the exciton diffusion length,
of bulk impurities, demanding highly pu- thereby guaranteeing that every exciton
rified dyes. Heterojunction formation by can reach the interface before it decays.
two different organic dyes can improve the However, the role of the electric field in
light harvesting, especially if their absorp- these devices is questionable because the
tion spectra are complementary to each theoretical depletion layer width generally
other [45, 46]. However, layer thicknesses is much wider than the dimension of phase
have to be optimized to prevent opti- segregation. It might be more exact to as-
cal filter effects caused by photovoltaically sume the totality of the interpenetrating
inactive areas away from the junction. At- network to be depleted with exciton split-
tempts were made to widen the thickness ting occurring via an interfacial electron
of the photovoltaically active area by in- transfer mechanism. Similar to thin-film
corporation of a coevaporated layer of the solar cells, the rectifying junction can be
two organic, semiconducting device com- formed by two organic semiconductors, or
ponents, resulting in a significant increase by a blend of an organic and an inorganic
in device efficiency [47]. semiconductor.
Another limitation of organic thin-film Semiconducting polymers are superior
solar cells is the generally poor conductivity to their low-molecular-weight analogs,
of the organic layer. Organic devices with regard to their applicability in in-
commonly show strong sublinearity of the terpenetrating junctions. This is due to the
photocurrent at illumination intensities fact that charge transport in semiconduct-
close to AM1.5 (100 mW cm−2 ). However, ing polymers mainly proceeds along the
5.4 Solid State Dye-sensitized Solar Cells – An Alternative Route Towards Low-Cost Photovoltaic Devices 487
polymer chain, so that percolation starts al- and the polymer take over the function of
ready at low concentration thresholds (for the light absorber. Among those, nanocrys-
a review on polymer solar cells see Ref. 51). talline heterojunctions of TiO2 either with
Poly(p-phenylenevinylene) derivatives PPV [59] or polythiophene [60] derivatives
(PPVs) are the most prominent examples have received particular interest. Several
of polymers used in organic solar cells. different techniques were applied to form
Depending on their substituents they ei- the heterojunctions. Commonly the first
ther can exhibit p-type or n-type behavior. step was the formation of a nanoporous
The first junction of this type was re- thin film of TiO2 via spray pyrolysis or
ported in 1995, using a phase-segregated via deposition of a colloidal paste followed
thin film of MEH-PPV as p-type material by a sintering step while the polymer was
and CN-PPV (cyano-p-phenylenevinylene) applied via a spincoating technique or (in
as n-type material [52]. The phase segrega- the case of polythiophene) via electrode-
tion was on the scale of 10–100 nm and position. IPCE values reported on these
IPCE values as high as 6% were recorded. systems are still very low (below 3%).
Recently, significantly improved device However, just recently, highly efficient
performances were reported based on a TiO2 /polythiophene solar cells exhibiting
polythiophene/PPV-derivative heterojunc- peak IPCE values of up to 10% were
tion, assembled via a lamination technique accomplished [61]. Photovoltaic cells of
with IPCE values reaching 29% [53]. The comparable efficiency were obtained when
cross-section scheme and the molecular polycrystalline films of low band gap semi-
structures can be seen in Fig. 7. conductors such as CdSe, CdS, or CuInSe2
A different approach to form or- are contacted with a thin film of polythio-
ganic interpenetrating heterojunctions is phene via electrodeposition [62]. Overall
to dope strongly a p-type conducting white light conversion efficiencies of 1.3%
polymer with a low-molecular-weight ac- at 56 mW cm−2 were reported for a nanos-
ceptor until charge percolation occurs tructured CdSe/poly(3-methylethiophene)
inside the acceptor phase. The blends junction [63].
attracting most attention are those of One particularly interesting aspect of
PPV- or thiophene derivatives doped with nanoparticle/polymer solar cells is that
fullerenes as acceptor, reaching 23% IPCE charge ‘‘hopping’’ from one semiconduc-
for polythiophene/C60 junctions [54] and tor particle to another takes place even in
29% for PPV/C60 junctions [55]. Perylene blends prepared from spin coating solu-
nanocrystals were also used as doping ac- tions containing both the polymer and the
ceptor [56]. Different theoretical models colloidal nanoparticles. Usually, those de-
were developed to describe the functioning vices are not exposed to successive heat
principals of these devices [57, 58]. treatments. This is surprising because
nanocrystalline semiconductor films used
5.4.3.3 Nonsensitized Inorganic/Organic in photoelectrochemical solar cells gen-
Junctions erally are believed to need such a heat
Recently, increased interest was raised in treatment to induce interparticle contact
the photovoltaic properties of composites and ‘‘necking.’’ Less expensive, flexible
of semiconducting polymers and nano- dye-sensitized solar cells based on plas-
sized semiconductor particles, in which tic substrates could be realized if this heat
either the polymer or the semiconductor treatment could be made redundant.
488 5 Dye-Sensitized Photoelectrochemistry
O O
CN
n
NC
O O
S n
POPT MEH-CN-PPV
(a)
AI or Ca
Glass substrate
Acceptor layer
Donor layer
ITO or
PEDOT on Au
Glass substrate
(b)
Fig. 7 Chemical structures and device structure for an organic
..
polymer photodiode as described by Granstrom and
coworkers [53]. (a) Chemical structures of the polymers used in the
devices. POPT is a regioregular phenyl-octyl substituted
polythiophene. MEH/CN-PPV is a cyano-derivative of
poly(p-phenylene vinylene), with a large electron affinity as a result
of the electron-withdrawing cyano groups. (b) Device structure.
Both polymers (see (a) for chemical structures) were dissolved in
chloroform or toluene (5 mg ml−1 ). For the top half of the device,
aluminum or calcium contacts were thermally evaporated on glass
substrates and the acceptor material, MEH/CN-PPV (and a small
amount of POPT, usually 5%), was spin-coated on top of the metal
electrode. The other half of the device comprises the donor
material, POPT (and a small amount of MEH/CN-PPV, usually
5%), either spin-coated on ITO substrates or glass-coated with
poly(ethylene dioxythiophene) doped with polystyrene sulfonic
acid. To ensure a low contact resistance, a thin layer of gold
(∼10 nm) was thermally evaporated on the glass slide before
spin-coating the PEDOT film from a water solution. The thickness
of the PEDOT layer was 100 nm. To get the desired structure of the
POPT, this half of the device was heated to 200 ◦ C under vacuum
before the device was laminated together by applying a light
pressure while one half was still at elevated temperature. The total
thickness of the organic semiconductive layer was 70–80 nm, and
the active area 2.5 mm2 .
5.4 Solid State Dye-sensitized Solar Cells – An Alternative Route Towards Low-Cost Photovoltaic Devices 489
44. J. Simon, J. J. André, in Molecular Semicon- 56. J. J. Dittmer et al., Sol. Energy Mater. Sol. Cells
ductors, Springer-Verlag, Berlin, 1985. 2000, 61, 53–61.
45. C. W. Tang, Appl. Phys Lett. 1986, 48, 183. 57. A. Köhler, D. A. D. Santos, D. Beljonne et al.,
46. K. Kudo, T. Moriizumi, Jpn. J. Appl. Phys. Nature 1998, 392, 903–906.
1981, 20, L553–L556. 58. L. A. A. Pettersson, L. S. Roman, O. Inganäs,
47. M. Hiramoto, H. Fujiwara, M. Yokoyama, J. Appl. Phys. 1999, 86, 487–496.
J. Appl. Phys. 1992, 72, 3781–3787. 59. T. J. Savenije, J. M. Warman, A. Goossens,
48. K. Yamashita, Y. Harima, T. Matsubayashi, Chem. Phys. Lett. 1998, 287, 148–153.
J. Phys. Chem. 1989, 93, 5311–5315. 60. K. Kajihara, K. Tanaka, K. Hirao et al.,
49. M. Hiramoto, Y. Kishigami, M. Yokoyama, Jpn. J. Appl. Phys. 1997, 36, 5537–5542.
Chem. Lett. 1990, 119–122. 61. S. Spiekermann, S. Mayer, W. Blau et al.,
50. M. Hiramoto, M. Suezaki, M. Yokoyama, unpublished results.
Chem. Lett. 1990, 327–330. 62. S. A. Gamboa, H. Nguyen-Cong, P. Chartier
51. G. Hadziioannou, P. F. van Hutten, in Semi- et al., Sol. Energy Mater. Sol. Cells 1998, 55,
conducting Polymers, Wiley-VCH, Weinheim, 95–104.
2000. 63. P. Chartier, H. N. Cong, C. Sene, Sol. Energy
52. J. J. M. Halls, C. A. Walsh, N. C. Greenham Mater. Sol. Cells 1998, 52, 413–421.
et al., Nature 1995, 376, 498–500. 64. J. S. Salafsky, W. H. Lubbrthuizen, R. E. I.
53. M. Granström, K. Petrisch, A. C. Arias et al., Schropp, Chem. Phys. Lett. 1998, 290,
Nature 1998, 395, 257–260. 297–303.
54. L. S. Roman, W. Mammo, L. A. A. Pettersson 65. J. H. Schoen, C. Kloc, E. Bucher et al., Nature
et al., Adv. Mater. 1998, 10, 774–777. 2000, 403, 408–410.
55. G. Yu, J. Gao, J. C. Hummelen et al., Science
1995, 270, 1789.
497
of the semiconductor can lead to redox acid solution or −1.2 V (SHE) in alka-
processes in the interfacial region around line solution. Among the oxide semicon-
each particle. These types of systems have ductors, TiO2 , SrTiO3 , CaTiO3 , KTaO3 ,
drawn the attention of a large number of Ta2 O5 , and ZrO2 satisfy this requirement,
investigators over the past twenty years. although in most cases the band gaps are
For example, in the area of photocatalytic relatively large (>3.0 eV), and thus the
water splitting, there has been a very active efficiency of solar energy conversion is
effort, as reviewed recently [11, 12]. very low.
In the early work on PEC solar cells Following early TiO2 -induced water
for water electrolysis, one of the first splitting studies, it was discovered that
types of electrode materials examined was TiO2 particularly excels in photocatalyti-
TiO2 , partly because it has a sufficiently cally breaking down organic compounds
positive valence band edge to oxidize water and other pollutants to innocuous prod-
to oxygen [4–7]. It is also an extremely ucts, neutralizing foul odors, and killing
stable material in the presence of aqueous bacteria. Frank and Bard in 1977 were
electrolyte solutions. The possibility of among the first to examine the possibili-
solar photoelectrolysis was demonstrated ties of using TiO2 to decompose pollutants;
for the first time with a system in which in their case, cyanide, in water [2, 3]. Since
then, there has been an increasing inter-
an immersed n-type TiO2 semiconductor
est in environmental applications. These
electrode, connected through an electrical
authors quite correctly pointed out the
load to an immersed platinum black
implications of their result in the field
counter electrode, was exposed to near-
of environmental purification. Their pre-
UV (ultraviolet) light. When the surface
diction has indeed been borne out, as
of the TiO2 electrode was irradiated with
evidenced by the extensive international
light consisting of wavelengths shorter
efforts in this area [8, 12–23].
than ∼415 nm, photocurrent flowed from
the platinum counter electrode to the
6.1.3
TiO2 electrode through the external circuit. TiO2 Characteristics
The direction of the current revealed that
the oxidation reaction (oxygen evolution) 6.1.3.1 General
occurs at the TiO2 electrode and the Titanium dioxide occurs in nature in
reduction reaction (hydrogen evolution) at three crystalline forms, rutile, anatase,
the Pt electrode. and brookite. The crystal structures are
When the conduction band energy, ECB , shown in Fig. 1 [24]. Some of the physical
is higher (i.e. closer to the vacuum level properties are given in Table 1. Much of
or more negative on the electrochemi- the fundamental work on TiO2 has been
cal scale) than the hydrogen evolution carried out with rutile single crystals be-
potential, photogenerated electrons can cause these are relatively straightforward
flow to the counter electrode and drive to produce. Anatase single crystals are
the reduction of protons, resulting in hy- more difficult to produce, and some work
drogen gas evolution without an applied has been carried out with natural crystals
potential. Hence, for this to occur, ECB but it is possible to grow such crystals
should be at least as negative as −0.4 V with the use of the vapor-phase trans-
[standard hydrogen electrode (SHE)] in port method [25]. Photocatalytic reactions
6.1 Fundamentals of Photocatalysis 499
a b
(a)
a
b
(b)
a
(c) c
have been shown to exhibit larger rates on surface with gas or liquid environment.
anatase compared to rutile, and it will be The electrochemical flat band potentials,
very useful to carry out further comparative Efb , can be used to estimate the band edge
studies with single crystals. energies for titanium dioxide. Efb values
for rutile single crystals have been reported
6.1.3.2 Band Edge Energies by several authors [26]. In acid solution
The energies of the valence and conduction (pH 2.1), a value of −0.05 V versus SHE
bands of TiO2 are determined by the has been proposed by Morrison, based
chemical termination of the material, on the results of several workers [27]. In
together with the interactions of the alkaline solution, values of −0.6 [27] and
500 6 Nonsolar Energy Applications
Tab. 1 Selected properties of TiO2 arises from the fact that the surface oxy-
gens of the oxide interact strongly with
Property Rutile Anatase protons and hydroxide ions [27]. To un-
derstand the processes underlying photo-
Crystal structure Tetragonal Tetragonal catalysis on TiO2 , it is useful to take the
Space group P42/mnm I41/amd
pH dependencies of the band edge ener-
(136) (141)
Density, g cm−3 4.250 3.894 gies into perspective. The relationships of
Band gap, eV 3.02 3.2 these energies to those corresponding to
Type Indirect Indirect various redox couples, for example, those
for various forms of reduced oxygen (see
Note: Data taken from Finklea [24] and next section) or for various organic ox-
Wyckoff [26] idations, then become apparent (Fig. 2).
Here we have adapted the commonly used
−0.76 V [28] have been given. Recently, Pourbaix representation in a modified
Kavan and coworkers have measured Efb form (i.e. with negative potentials upper-
values for both rutile and synthetically most), to be consistent with the prevailing
grown anatase single crystals in 1 M practice in energy band diagrams [30].
H2 SO4 and obtained values of +0.04 and Therefore, the ubiquitous lines for the
−0.16 V against SHE, respectively [25]. H+ /H2 and O2 /H2 O redox couples are
The authors point out that the latter inverted and slope upward with increasing
value is sufficiently negative so that no pH (more negative electrochemical poten-
external bias needs to be applied to tials). Likewise, the valence band (VB) and
photoelectrochemically evolve hydrogen. conduction band (CB) energies for TiO2
On the basis of the flat band potential, slope upward (higher electron energies)
the actual location of the conduction band with increasing pH.
edge energy, ECB , then depends on the The VB energies lie approximately
doping level, according to the following 1.8 eV below those necessary to evolve
equation: molecular oxygen from water over the en-
n tire pH range. Electrons in the CB for
ECB − EF = ξ = −kT ln (1)
NC rutile are very close in potential to that
where EF is the Fermi level energy, n necessary to evolve molecular hydrogen
is the occupation of conduction band from water but not at appreciable rates. In
states with electrons, and NC is the early TiO2 -based PEC solar cells, this prob-
density of conduction band states at the lem was circumvented through the use of
band edge [29]. For normal semiconductor a two-compartment cell. The TiO2 elec-
doping levels, ξ could be in the range of 0.1 trode was immersed in a more alkaline
to 0.2 V [27]. In subsequent discussions we half-cell and a Pt counter electrode was
will refer to the Efb values, which should immersed in a more acid electrolyte in the
then be considered to be lower (energy) second half-cell. The electrons generated
limits for the ECB values. at the photoelectrode diffuse at the back of
As already stated, the band positions de- the electrode and then toward the counter
pend on the environment external to the electrode. In this case, electrons photogen-
solid surface, for example, the hydrogen erated at the alkaline TiO2 electrode have
ion concentration or pH. This is a situation a relatively high electrochemical energy,
common to all oxide semiconductors and which is sufficient to evolve hydrogen in
6.1 Fundamentals of Photocatalysis 501
0 O2/H2O2
O2/HO2−
[V vs. SHE]
Potential
+1 O2/H2O
O−/OH−
+2
OH/H2O
TiO2 VB
+3
0 2 4 6 8 10 12 14
pH
the more acidic solution. In the case of the topic of the actual energetics of various
photocatalysis, the same type of effect may oxidation reactions pertinent to photo-
also be operative: specifically, if the local catalysis is beginning to be explored in
pH is high, for example, at a reducing site detail [35]. For example, because of the
at a particular location on a TiO2 parti- presence of water, the actual reaction could
cle, additional reducing power is available be considered to be an electron transfer
for the reduction of molecular oxygen to followed by an extremely fast chemical
superoxide. reaction with water.
In contrast, the oxidizing power of
photogenerated holes decreases with in- R −−−→ • R+ + e− (2)
creasing pH but the standard potential for •
R+ + H2 O −−−→ • RH + OH− (3)
the one-electron oxidation of any given or-
ganic compound RH to its radical cation The effective redox potential of this re-
•
RH+ is not pH-dependent. For example, action would be shifted significantly in the
the standard potential for this reaction for negative direction because of consumption
n-octane in acetonitrile is +2.71 V against of the oxidized product. Thus, it is quite
Ag/Ag+ [34], which can be converted to possible that trapped holes with redox po-
+2.84 V against SHE. This value may not tentials of +1.5 V [36] to +1.75 [37], have
be sufficiently positive because solvent sufficient oxidizing power to induce the
reactions make elucidation of the re- production of the neutral radical, which
versible potential problematic. At present, would involve Eqs. (2) and (3).
502 6 Nonsolar Energy Applications
Tab. 2 Standard redox potentials for various electrochemical reactions involving molecular oxygen
and its reduced forms
Acid dissociations
HO•2 ←−→ •
O 2 − + H+ 4.88 31
H2 O 2 ←−→ HO2 − + H+ 11.63 31
• −
•
OH ←−→ O + H+ 11.9 32
Redox reactions
O2 + e− ←−→ •
O2 − all −0.284 +0.284 33
O2 + H+ + e− ←−→ HO•2 0 +0.005 −0.005 c
H2 O2 + H+ + e− ←−→ •
OH + H2 O 7 +0.433 −0.991a c
Note: Data were taken from Hoare [31], Buxton and coworkers [32], and Tarasevich and
coworkers [33].
Solely considering Eq. (2), it would be reactions are often higher at higher
expected that the reaction of a hole in pH [18]. For organic oxidation reactions
the TiO2 valence band with the organic in aqueous solution, not involving radi-
compound would become less favorable cal species, standard redox potentials are
at high pH. However, considering Eq. (3), known (Table 3) and many exhibit pH
a normal pH dependence, that is, about dependencies of about 60 mV pH−1 , al-
60 mV/decade at 25 ◦ C should exist. Over- though the behavior can be somewhat
all, it might be expected that there should more complex, with at least one change
not be significant pH effects in terms in slope [38] because of involvement of
of changes in excess oxidizing power of acid-base equilibria.
valence band holes or trapped holes. In The majority of the VB and CB values
fact, the rates of photocatalytic oxidation (the latter being approximated by Efb as
Tab. 3 Selected free energies and standard potentials for CO2 reduction reactions
a Indicatesthat the actual "G0 values are more negative than those given.
b converted from values versus ferrocene/ferrocinium [135].
c converted from values versus Ag/Ag+ (0.1 M) [135].
6.1 Fundamentals of Photocatalysis
503
504 6 Nonsolar Energy Applications
already discussed) shown in Fig. 2 are generally forms of reduced oxygen, in-
for rutile single crystals. It has already cluding superoxide • O2 − , the hydroperoxyl
been pointed out that the crystalline form radical • O2 H, hydrogen peroxide H2 O2 ,
more commonly used in photocatalysis, the peroxide anion HO2 − , the hydroxyl
anatase, has a higher CB energy (more neg- radical • OH, and the oxyl radical anion
• −
ative electrochemical potential). A point O . The relevant reactions involving these
is included for single crystal anatase, ap- species are shown in Table 2. The pH de-
proximately 200 mV above the line drawn pendences of these reactions are shown
through the points for rutile, and a point in the same figure with the TiO2 band
is also shown for a nanocrystalline anatase edge energies (Fig. 2). It is interesting
thin film: its CB energy is elevated approx- that the reducing power of the conduc-
imately 100 mV with respect to that for tion band electrons is barely sufficient to
rutile. reduce O2 to • O2 − at pH values below the
Lying approximately 0.6 to 0.8 eV below pKa for • O2 H (4.88). With increasing pH,
the CB energy, there are surface states however, there is an increasing excess of
on TiO2 ; three experimental points are in- reducing power for this reaction.
cluded in Fig. 2 at specific pH values but The pH dependence of the oxidizing
additional experimental evidence exists for power of the • OH radical parallels that for
such surface states under conditions with valence band holes, except that there is a
no pH control, for example, in the gas discontinuity at pH 11.9, which is the pK
phase, and therefore such data cannot be for the equilibrium
included in the figure. A reasonable char- •
OH ⇐'⇒ H+ + O− (5)
acterization for this type of surface state is
that it involves a trapped electron with an At pH 0, the potential for the one-
accompanying proton, viz.: electron reduction of • OH is +2.683 V [32]
(Table 2). As already mentioned, the redox
H potential for trapped holes has been
estimated to be significantly less positive,
O O +1.50 to +1.75 V against SHE.
Ti (IV) Ti (IV) Ti (IV) Ti (III) The standard potentials for the various
(4) redox processes, involving molecular oxy-
This type of species is important for pho- gen, can be conveniently visualized in
tocatalysis because it provides a surface a Frost diagram, given here for pH 7
sink for photogenerated electrons. Addi- (Fig. 3) [33, 43, 44]. In this type of dia-
tionally, this type of species is particularly gram, the relative state of reduction of the
important with respect to the wetting prop- oxygen is plotted along the horizontal axis,
erties of TiO2 , which is of importance to with O2 itself being taken as zero and H2 O
self-cleaning surfaces. being taken as four. The vertical axis is
the free energy of formation in units of
6.1.4 electron volts. The slopes of the lines be-
Photocatalytic Reactions tween species of differing oxidation states
are the negative of the potential for the cor-
6.1.4.1 Oxygen Electrochemistry responding redox process. The free energy
The active chemical species that are change for the process is the slope multi-
involved in photocatalytic reactions are plied by the number of electrons involved
6.1 Fundamentals of Photocatalysis 505
Free energy
−1.0 OH
[eV]
−2.0
H2O
−3.0
0 1 2 3 4
State of reduction
electrons
to a trapped hole [36]. By reacting col- and hydroxylated products of organic com-
loidal TiO2 (13 nm) in aqueous suspen- pounds. The desorption of free • OH is
sion with radiolytically generated • OH, considered to be unlikely because of the ex-
they observed a broad absorption band tremely large rate constant for the reverse
(ca. 350 nm). The intensity of this signal reaction.
exhibited first-order kinetics, with a rate A later study by Goldstein and cowork-
that was proportional to the • OH concen- ers involved phenol as the reactant [45].
tration as a result of the self-reaction of The photocatalytic reaction products were
•
OH to produce hydrogen peroxide: compared with those of radiolytically gen-
erated • OH on one hand and SO4 − on the
•
•
OH + • OH −−−→ H2 O2 (12)
other. The latter is a strong one-electron
with a time constant of about 50 ms. oxidant and would mimic the effect of
In homogeneous solution, this reaction a hole (see section on mechanisms). At
is second-order (Table 4) but becomes low phenol concentrations, the oxidation
first-order on the two-dimensional sur- products of the photocatalytic reaction are
face, as discussed later. The trapped hole similar to those for • OH, whereas at high
can be represented as TiIV − O− . On phenol concentration, they are similar to
those for reaction with SO4 − , indicating
•
the basis of the reaction with SCN− ,
the E 0 for this species was estimated to involvement of holes. This study shows
be about 1.5 V (pH 3), which is approx- that the trapped hole acts as a • OH trans-
imately 1.3 eV above the valence band. fer agent. At high phenol concentrations,
These workers conclude that the exis- phenol adsorbs on the TiO2 surface and
tence of adsorbed • OH (surface-trapped competes for mobile holes, decreasing the
holes) can explain most of the observed rates of hole localization and electron-hole
results, including the production of H2 O2 recombination.
Tab. 4 Homogeneous reaction rate constants for selected reactions involving hydroxyl
radicals
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 507
Bahnemann and coworkers have dis- with neutral molecular oxygen O2 . The ef-
tinguished between shallow and deeply fect of raising the pH from 3 to 6 was
trapped holes [46]. They point out that the consistent with the second pathway be-
absorption features commonly used to in- cause both the • Q− and • O2 − would be
dicate trapped electrons (650 nm, 1.7 eV) stabilized at the higher pH.
and holes (430 nm, 2.7 eV), could not A different approach was taken by
be the actual ones involved with pho- Kesselman and coworkers, who examined
tocatalytic reactions because they would the oxidation of some organic anions
correspond to energy levels near the mid- (chloroacetates) at heavily Nb-doped TiO2
dle of the band gap, for example, 1.7 eV electrodes. Because the selectivity for
below the CB or 2.7 eV above the VB. reaction of these species with • OH in
The latter is assigned to an excitation homogeneous solution is known, as is
of a surface peroxide. The active oxidiz- the ordering of the electrochemical redox
ing species is thought to be a shallow potentials, it is possible to distinguish
trapped hole, which can equilibrate with between pathways. In addition, the extent
VB holes and would have a similar oxi- to which the oxidation current depends on
dizing power. These authors showed that the reactant concentration is a measure
strongly adsorbing species, such as cit- of the involvement of the direct-hole
rate and acetate, react with free holes, pathway. If there is no dependence on
whereas more weakly adsorbed species, concentration, this indicates that • OH
such as ethanol, could react with deeply is produced and then reacts with the
trapped holes (i.e. surface-bound • OH). organic species. These authors found a
They also showed that dichloroacetate and full range of behavior from 12% to 97%
thiocyanate adsorb strongly and react with direct-hole involvement for formate and
either freely or shallowly trapped holes but 4-chlorocatechol, respectively. The results
not deeply trapped holes. were found to vary significantly from one
Cermenati and coworkers used a very electrode to another (15% vs. 65% direct-
interesting approach to assess the rel- hole involvement for acetate).
ative importance of holes and hydroxyl From the work of Nosaka and coworkers
radicals [47]. For a model reactant, they with electron paramagnetic resonance
employed quinoline, which has two aro- (EPR) spectroscopy, the proportion of the
matic rings with differing electron density, direct hole and • OH-mediated pathways
the nitrogen-containing ring with less. can be determined for various types of
Because • OH is electrophilic, the attack TiO2 samples. These workers used the
should occur on the benzene ring. On photocatalytic oxidation of acetate as a
probe reaction and found mixtures of EPR-
the basis of a detailed analysis of prod-
active products. They have proposed the
ucts as a function of time, the major
following schemes:
mechanistic pathways involving • OH were
discounted. Instead, a mechanism involv-
CH3 COOH + H+ −−−→
ing oxidation with holes, together with
superoxide, was proposed (see section on •
CH3 + CO2 + H+ (13)
mechanisms). A third mechanism was •
CH3 COOH + OH −−−→
proposed in which initial oxidation by ei-
•
ther holes or • OH is followed by reaction CH3 + CO2 + H2 O (14)
508 6 Nonsolar Energy Applications
•
OH, whereas the carboxymethyl radical
•
CH2 COOH can be produced only by re- Ti4+ O2 − Ti4+ OH− + • OH (18)
action with • OH. Thus the ratio of • CH3 to
•
CH2 COOH radical yields the proportion The structural arrangements of these
of the direct h+ path against the • OH path. species are shown in Fig. 4.
It was found that the • OH path was favored From work in our laboratory, it was
for TiO2 samples with higher coverages of found that the yields of fluorescent prod-
OH groups. On such surfaces, the type of ucts from the reactions of • OH with
active surface species has been proposed coumarin and 1,4-phthalate could be mea-
by Howe and Grätzel [48], based on EPR sured as a function of time, with a linear
measurements, to be dependence being found for fluorescence
intensity versus illumination time [51]. It
Ti4+ O− Ti4+ OH−
•
(S1) was concluded that the • OH pathway
predominates for the low reactant concen-
On TiO2 surfaces that contain fewer trations used (10−4 to 10−3 M) in aqueous
OH groups, perhaps as a result of heat
solution.
treatment, the active species may instead
In later work, it was found that the
be [49, 50]:
quantum yield for the reaction of • OH
Ti4+ O2 − Ti4+ O−
•
(S2) with 1,4-phthalate was relatively small (7 ×
10−5 ) under the experimental conditions
Hirakawa and coworkers propose that used (4 × 10−4 M, 2 × 10−3 M NaOH; UV
the reaction intensity, 1 mW cm−2 ) [52]. In contrast,
A B C D E
•
O
OH OH OH −O
•
O O O •
O O +
O O O
Ti Ti Ti Ti Ti Ti Ti Ti Ti
O O O −
O O O O O
O O• O O O O O O O
Ti +Ti Ti Ti Ti Ti Ti Ti
O O O O O O O O− O
O O O O O O O •
O
Ti Ti Ti Ti Ti Ti Ti Ti Ti
O O O O O O O O
O O O O O O O O O
Ti Ti Ti Ti Ti Ti Ti Ti
the quantum yield for the photocatalytic diffusing species and could then be con-
oxidation of iodide I− to iodine I2 was verted to • OH on reaching the organic film,
found to be relatively high (5.7 × 10−2 ) as in the following reactions:
for the conditions used (10−2 M KI; UV
intensity, 1 mW cm−2 ). The quantum 2HO•2 −−−→ H2 O2 + O2 (19)
yields usually observed for photocatalytic hν
reactions at similar light intensities are H2 O2 −−−→ 2• OH (20)
more in line with the latter, and therefore
it was concluded that the hole is the active In experiments with chemically gener-
species in most cases. ated • OH (Fenton reagent), similar effects
As already mentioned, the desorption were observed in water contact angle for
of active oxygen species, particularly • OH, PS and ODS.
from the TiO2 surface, with subsequent In earlier work, it was found that
reaction in homogeneous aqueous solu- Escherichia coli cells were killed in aqueous
tion, has been downplayed as a significant suspensions in contact with illuminated
pathway for TiO2 photocatalysis, although TiO2 films, even when separated by a 50-
not by direct experiments. Nevertheless, µm-thick porous membrane [55]. On the
our group has found it possible to observe basis of the inhibitory effect of catalase,
the oxidative degradation of organic films, it was concluded that the active species
for example, methylene blue, that are sepa- is hydrogen peroxide but there appeared
rated from the illuminated TiO2 surface by to be a cooperative effect, possibly due to
air gaps of up to 125 µm [53]. In an effort to superoxide, which significantly enhanced
identify the active species, various types of the lethal effect.
organic reactant films were examined [54]. Now that many of the important photo-
Changes in mass were also monitored generated species have been identified, it
by use of the quartz-crystal-microbalance is useful to consider the kinetics of forma-
(QCM) technique. In addition to methy- tion and decay of some of these species.
lene blue, a polyaniline derivative, poly(2- The absorption of a photon and creation
methoxyaniline-5-sulfonic acid) (PMAS), of an electron-hole pair in a TiO2 parti-
polystyrene (PS), octadecyltriethoxysilane cle occurs in a time range that has, thus
(ODS)-coated glass, and polyvinyl alcohol far, been too short to measure, that is,
were examined. With the ODS-coated glass a few femtoseconds. These charge carri-
separated from the illuminated TiO2 film ers are initially considered to exist in the
by 50 µm, oxygen-containing functional conduction band and valence band, re-
groups were produced, as observed with spectively but are then quickly trapped.
X-ray photoelectron spectroscopy, creating Skinner and coworkers have determined
a hydrophilic surface. For PMAS, PS, and the trapping time for electrons to approx-
ODS, the QCM results showed decreases imately 180 fs based on measurements
in frequency, that is, increases in mass due on TiO2 sols in acetonitrile, which was
to oxygenation. The fact that aliphatic hy- used to avoid the absorbance of hydrated
drocarbons are oxidized suggests that the electrons [56]. The nanoparticle surface
active oxidizing agent may not be HO•2 , consists of many different types of local ge-
H2 O2 , or 1 O2 . Thus the most likely can- ometries and corresponding trap energies,
didate is • OH. However, it may also be leading to a broad range for the absorbance
possible that HO•2 or H2 O2 could be the spectrum.
510 6 Nonsolar Energy Applications
Holes have been estimated to be trapped greatly increase the lifetime of the trapped
within 1–10 ps by Serpone and coworkers, electrons [58].
and these workers find that 90–100% of all It had already been observed in early
electron-hole (e-h) pairs have undergone work with flash photolysis, albeit on
recombination much longer timescales, that recombi-
nation could be greatly inhibited in the
e− + h+ −−−→ heat (21) presence of either hole scavengers or
electron scavengers [67]. Lifetimes of sec-
within 10 ns for colloidal TiO2 particles onds or even hours have been reported
from 2–27 nm in diameter [57]. However, for trapped electrons in the presence of
Colombo and coworkers estimate the half-
hole scavengers such as 2-propanol [68].
life of an e-h pair to be approximately
Lifetimes of milliseconds (in acid) or mi-
30 ps [58]. This is just one example of
croseconds (in base) have been reported for
the wide variety of kinetic results that
trapped holes in the presence of an electron
have been reported (Table 5). Second-order
scavenger such as methylviologen [67].
kinetics are observed particularly at short
The decay process is usually assumed
times and high light intensities, that is,
to involve recombination of electrons and
when there are multiple e-h pairs existing
holes in various states of trapping or de-
simultaneously in a single particle [57, 59].
trapping. Because at short time intervals
Results for time-resolved microwave
and relatively high light intensities, there
conductivity measurements, which reflect
may be approximately equal numbers
the number of charge carriers, indicate
that the decay kinetics span many or- of each type of carrier in a single
ders of magnitude in time, for example, particle, the recombination process can
from ns to s [60, 61, 63]. Striking differ- be expected to be second-order. When
ences were observed, comparing anatase there is on the order of only a single e-
and rutile, with the latter exhibiting h pair in a particle, the decay process is
much faster decay kinetics [60]. Major dif- known to become first-order. This has
ferences were observed among various been explained in terms of stochastic
types of colloidal TiO2 preparations and processes [59, 69–71], that is, because of
in comparison to Degussa P25 in terms the discrete nature of the species and
of recombination kinetics [63]. However, corresponding events, they must be treated
Colombo and coworkers, using diffuse statistically rather than continuously. The
reflectance flash photolysis experiments, simple fact that reactions are taking
have found the behavior of TiO2 colloidal place on a two-dimensional surface is
solutions versus powder form, rutile ver- also important [72]. However, it is well
sus anatase, and nanoparticle versus P25, known in the field of solid-state electronics
all to be similar [58, 65]. These workers that direct recombination in a doped
point out that the diffuse reflectivity re- semiconductor behaves as a first-order
sults are specific for trapped electrons, process for low injection rates, simply
rather than for carriers in general and because the magnitude of the squared
assign the long-lived species observed term in the rate expression becomes
in this and other studies to such elec- small [73]. This is the analog of pseudo-
trons [65]. They also show that the effect of first-order kinetics. In the case of indirect
a hole scavenger, such as thiocyanate, is to recombination (i.e. mediated by a band gap
Tab. 5 Kinetics of photogenerated charge carrier trapping and decay
Reaction TiO2 Particle Medium pH Gas e-h/ Techniquea Half-lifeb Comments Year Author Refe-
type size, nm particle rences
Reaction TiO2 Particle Medium pH Gas e-h/ Techniquea Half-lifeb Comments Year Author Refe-
type size, nm particle rences
a LFP = laser flash photolysis; TRMC = time-resolved microwave conductivity; PR = pulse radiolysis; time-resolved diffuse reflectance spectroscopy.
b Values in parentheses are times for which the intensities decrease by one-half.
6.1 Fundamentals of Photocatalysis 513
0.5
Simulation
0.4 QY(1000 ppm)
QY(100 ppm)
QY(10 ppm)
0.3 QY(1 ppm)
Φ
0.2
0.1
0
10−7 10−6 10−5 10−4 10−3 10−2 10−1 100
I norm
(a) [s−1]
Simulation
QY(1000 ppm)
1 QY(100 ppm)
QY(10 ppm)
QY(1 ppm)
Φ
0.1
This equation can be rearranged to the that the ratio of absorbed photons to ad-
following form: sorbed molecules is the same, 1 : 3000. We
find that $ is the same in both cases, 0.14,
2 or half the limiting value of 0.28. Hence,
$= ! (29)
4Jhν k12 each • OH is approximately as likely to react
1+ 1+ 2 2 with a 2-propanol molecule, as it is to react
k25 &2−p
with another • OH. Thus, the $ value is
determined only by the probabilities that
It can be seen that this function
the • OH radicals will either react with a
approaches unity at low light intensity and
2-propanol molecule or with each other.
zero at high intensity. This type of curve
However, this simple model requires
was used to fit the data in Fig. 5. Of course,
further modification because the $ values
it is also possible that such a relationship
tend to deviate from the theoretical
could also result from a scheme, based
curve at higher light intensities; this
on e-h recombination instead of the • OH
becomes apparent when log $ is plotted
self-reaction. It is highly intriguing that
(Fig. 5b), which indicates a mixing of first
this type of behavior could be exhibited
and second-order kinetics. Second-order
over such a wide range of initial reactant behavior is approached for the lowest 2-
surface concentrations (three orders of propanol concentration but deviates for
magnitude) and light intensities (seven higher concentrations. For the latter,
orders of magnitude). trapped holes would become increasingly
We have proposed a simple model to favored as the active species. Additionally,
rationalize the fact that the quantum yield the adsorbed molecules may block sites at
versus light intensity behavior tends to re- which holes can self-react, and therefore
main constant for a given ratio of adsorbed another decay channel might become
photons to adsorbed molecules. Basically, important, for example, trapping at a
this model involves the assumption that different type of site at which the h-h
the diffusion of • OH on the surface is fast reaction is favored or e-h recombination.
enough, that it is not a limiting factor, In any case, the results show clearly that 2-
and every • OH generated has an oppor- propanol decomposition can be used as
tunity to react with all of the adsorbed a type of standard for quantum yield,
2-propanol molecules in its vicinity. Let us if high reactant concentrations and low
imagine for a moment that the reactions light intensities are used to estimate the
are taking place on a perfectly flat, non- maximum, limiting value. However, it is
porous surface with a certain number of still necessary to estimate the amount of
adsorbed 2-propanol molecules, say 3000. absorbed light, for example, using the
For a low initial gas-phase 2-propanol con- methodology proposed by Anderson and
centration, this number of molecules will coworkers [77].
occupy a certain surface area Alow . For a The idea of • OH radicals self-reacting,
high initial concentration, the molecules has also been proposed for the decompo-
will occupy a smaller surface area Ahigh , sition of chloroform at illuminated TiO2
simply due to the adsorption equilibrium. aqueous suspension by Kormann and
For each surface area, we now assume coworkers [18, 90]. These authors have
that a certain number of photons is ab- given a limiting form of Eq. (29) at high
sorbed, for example one every second, so light intensity:
6.1 Fundamentals of Photocatalysis 517
where we have inserted • OH as the radical values exceeding unity, as high as 1.4, for
initiator. the photocatalytic dechlorination of carbon
We were able to observe clear evidence tetrachloride [100].
for the chain-type mechanism in experi-
ments, involving acetaldehyde decomposi- 6.1.4.4.4 Hydrocarbons and Related Com-
tion in the gas-phase [98], similar to those pounds Reaction pathways involving su-
already discussed for 2-propanol. With peroxide have been treated in detail
acetaldehyde, the $ values exceeded the by Schwitzgebel and coworkers [101], as
maximum value obtained for a similar film shown in Fig. 7. These reactions can oper-
for 2-propanol oxidation (0.28) (Fig. 6). As ate on alkanes and their oxidation products
already discussed, the latter value may be so that with the exception of aldehydes, as
considered to be an intrinsic maximum $ noted in the earlier section, any of the
value for this particular film. Therefore, if main types of CH and C−H−O com-
$ exceeds the intrinsic maximum value, pounds (i.e. alkanes, alcohols, ketones,
it indicates that radical chain reactions are and carboxylic acids) can be completely
important, that is, a single photon can decomposed in this fashion to CO2 and
cause more than one photodecomposition H2 O. These reactions involve cyclic tetrox-
reaction. ide intermediates, similar to those orig-
Other groups have also obtained rela- inally proposed by Russell [102]. It is
tively high $ values for chain-type re- quite interesting that TiO2 is so well
actions. As already mentioned, Lepore suited to the decomposition of organic
and coworkers found values, approach- compounds, with a combined attack of ox-
ing unity for propanol oxidation [62]. idatively and reductively produced radical
Raupp and Junio found values exceed- species.
ing unity, as large as 3.0, for the ox- In Fig. 7(b), a possible three-dimensional
idation of acetone and methyl-t-butyl structure for the tetroxide intermediate for
ether [99]. Stark and Rabani measured the photocatalytic oxidation of n-octane
1000
Quantum yield (total)
100
10
[%]
0.1
10−6 10−5 10−4 10−3 10−2 10−1 100 101 102
I normal
[s−1]
Fig. 6 Quantum yield as a function of normalized light intensity:
acetaldehyde, taken from Ohko and coworkers [98].
6.1 Fundamentals of Photocatalysis 519
(b)
Fig. 7 (a) Photocatalytic reactions involving tetroxides, taken from Schwitzgebel and
coworkers [101]. (b) Proposed structure for the tetroxide intermediate in the oxidation of
octane, with the hydrogen bond shown as a solid rod. Carbon atoms are gray, hydrogen atoms
are a light stipple, oxygen atoms are a dark stipple, and oxygen lone pairs are white.
is shown. Very similar intermediates can facilitate the breakage of the chain, with
be drawn for the corresponding alcohol the formation of two aldehydes and a CO2
and ketone. For the alcohol (3-octanol), molecule.
the cyclic tetroxide structure can form The pathways delineated in the preced-
at C-2, similar to the structure shown ing paragraph are probably also important
for n-octane but the intermediate given in the decomposition of long-chain hy-
in the equation in Fig. 7(a) actually cor- drocarbons, carboxylic acids, fats, oils,
responds to the cyclic structure, form- and even long-chain polymers [103–105].
ing at C-3. In that case, the alcohol In the latter case, we have direct in-
−OH group becomes part of the ring. frared evidence that the hydrocarbon
For 3-octanone, the terminal oxygen of chains can be attacked at any point along
the tetroxide may hydrogen-bond across their length, with C−C bond scission,
to the hydrogens on C-4, which could presumably with subsequent degradation
520 6 Nonsolar Energy Applications
0.10 4h
Absorbance
0.05
End of reaction
0.00
3000 2950 2900 2850 2800
Wavenumber
[cm−1]
3.0 2.0
Weight of octadecane
2.5
CO2 concentration
[10 µg cm−2 ]
1.5
[10 µmol cm−2]
2.0
1.5 1.0
1.0
0.5
0.5
0 0
2.0 100
Carbon concentration
Carbon proportion
80
[µmol C cm−2 ]
1.5
60
[%]
1.0
40
5.0
20
0 0
0 20 40 60 80 100
Fig. 9Photocatalytic decomposition of octadecane: weight loss
and CO2 generation, taken from Minabe and coworkers [105].
6.1 Fundamentals of Photocatalysis 521
via the resulting aldehydes (Fig. 8). The 6.1.4.5 Kinetic and Mass Transport
aldehydes may be unstable enough to Considerations
account for the fact that they are not
detected by IR on the TiO2 film. Also, 6.1.4.5.1 Adsorption For moderate-in-
it is important to note that no gas-phase tensity illumination, over a wide range
products, other than CO2 , are detected in of experimental conditions, photocatalytic
the decomposition of octadecane, stearic decomposition follows first-order kinetics
acid, and glycerol trioleate, demonstrating in the adsorbed concentration of the
the safety of the photocatalytic approach organic compound and one-half-order in
to self-cleaning technology. For exam- light intensity due to recombination.
ple, a 36-µm-thick film of octadecane
was completely converted to CO2 after rate = koxidation &reactant I 1/2 (37)
approximately 50 hours of illumination at
0.8 mW cm−2 (Fig. 9). Where koxidation is a first-order rate
constant, &reactant is concentration per
unit real surface area, and I is the light
6.1.4.4.5 Aromatics As already briefly intensity. As pointed out by Emeline and
mentioned, the photocatalytic oxidation coworkers, however, this equation is a
of phenol was studied by Goldstein special case of a more general one [106].
and coworkers [45]. On the basis of the
analysis of the possible reactions (Fig. 10), n
rate = koxidation &reactant Im (38)
these authors were able to conclude
that either • OH or h+ can be the where the n and m values can vary with I
principal active species. Also, the use of and &reactant , respectively.
quinoline as a probe molecule was already Many organic compounds and noxious
mentioned. The full scheme developed by gases, such as H2 S, follow Langmuir ad-
Cermenati and coworkers is shown in sorption behavior so that Eq. (37) can
Fig. 11 [47]. This scheme shows that the be converted into the familiar Lang-
combined action of holes and superoxide muir–Hinshelwood form by substituting
is important. the adsorption equilibrium expression for
R
R R
•OH Ox
+ •
a d
H OH
OH
H+ H2O
c
R R
SO4• − + +
•
Side chain fragmentation
b e
OH H HO H OH
OH° O2 − HO2°
•
N N N N
H OO°
OH OH HO OO° OH
OH° O2 − HO2°
•
N N N N
HO H HO H OH
+ OH° − OH°
− H2O − H2O
O° O O
O2 − OH°
N N N
(a) H OO° O
O transfer N
H
O
O
TiO2 CHO H2O CHO
Q,O2 Q+°,O2−°
(h+,e) CHO CHO
N N CH CHO NH2
O2 (2-formyl)phenyliminoethonal
O O°
CHO
O −CO2
O
N NH − CHO
(b) O
O O°
CHO
+
Q+° + O2
+ O2
N H2O NH2
− H+ O − HO2°
O
− CHO − CHO
N OH
O +O2
O2 − HCOO− CHO
Q + OH ° OH
N H NH − CHO
O
+ O2 − HO2°
N O
(c) H
Fig. 11 (Continued)
0.2
0.15
[µmol min−1]
20
r
0.1
[µmol−1 min]
15
1/r
10
0.05 5
0
0 0.1 0.2 0.3 0.4 0.5
1/Ceq
0 [µM−1]
0 5 10 15 20 25 30 35 40
Ceq
[µM]
Fig. 12 Photocatalytic decomposition of acetaldehyde:
Langmuir–Hinshelwood plot, based on Noguchi and
coworkers [107].
can be effective even at low formaldehyde Many groups have obtained results that
concentrations. The adsorption character- are consistent with Langmuir–Hinshel-
istics of TiO2 in this case are similar, if not wood kinetics. However, Emeline and co-
better than those of activated carbon. workers have emphasized the well-known
524 6 Nonsolar Energy Applications
106
1018
Mass-transport
105 controlled region 1017
104
1016
UV-light intensity
[µW • cm−2]
1015
102 Target for outdoor
air purification 1014
101
Target for 1013
1 indoor air purification
1012
Curve B
10−1
1011
Light-limited
10−2 controlled region
1010
Catalase
e− H2O2
there is a tendency for the particles to
develop excess charge when they make e−
contact with the electrode, so that they SOD
e− e− O +
move under the influence of the electric e− 2
field and circulate electrophoretically back e− e−
− − O2
and forth between the electrodes. This is e− e e e−
e−
because the photogenerated carriers can be e− hν
effectively trapped via reactions involving TiO2
h+
hydroxide ions or water. Because of the
absence of bulk water surrounding each
OH−
particle, excess charge cannot be carried OH−
away, for example, by the production of
solution-phase peroxide.
4
(a)
2
[µA cm−2]
0
J
−2
−4 (b)
and (3) chemically homogeneous, physi- the type of situation that could be expected
cally inhomogeneous surfaces. An exam- for a pure polycrystalline TiO2 sample.
ple of case 2 in a metal would be a Zn-Cu A direct comparison of the anodic and
alloy, in which Zn undergoes corrosion cathodic reactions on the various single
but Cu does not. A corresponding exam- crystal faces of either anatase or rutile does
ple in photocatalysis would be TiO2 , with not yet appear to have been studied.
deposited Pt, in which the anodic process
might occur predominantly on TiO2 and 6.1.4.6.3 Scanning Electrochemical Micro-
O2 reduction might occur predominantly scopy Another way in which it can be con-
on Pt. An example of case 3 in terms of cor- firmed that oxidative and reductive pho-
rosion would be any polycrystalline metal, tocatalytic reactions simultaneously occur
in which different crystallographic faces on TiO2 particles is to set up a simple
are exposed to the electrolyte, such that heterogeneous model electrode system.
the relative proportions of the anodic and This can be used to simulate model indi-
cathodic processes are different, although vidual particles and can be compatible with
both reactions occur on both types of sur- microelectrode detection of reaction prod-
faces. In terms of photocatalysis, this is ucts [122–126]. Such a system, shown in
Low viscosity
Epoxy resin Mercury epoxy resin
Cu wire
7 µm Carbon fiber
ca. 20 µm
(a)
Potentiostat
Pt SCE
(CE) (RE)
hν Micromanipulator
Carbon microelectrode (WE)
Pd
TiO2 Microscope
(b)
Fig. 16 Schematic diagram of a scanning electrochemical microelectrode
system, taken from Fujishima and coworkers [12].
528 6 Nonsolar Energy Applications
Fig. 16, involves a film that is half-covered an increase in O2 concentration near the
with TiO2 and half-covered with metallic TiO2 surface and a decrease in the O2
regions, for example, Pd or ITO (In-Sn concentration near the TiO2 surface un-
oxide). A scanning microelectrode can be der illumination (Fig. 17) [124]. Oxygen is
positioned close to the surface, as close as produced via water oxidation on TiO2 and
50 µm. Its potential can be set at values at is then consumed at the metallic electrode
which, for example, either O2 or H2 O2 can (ITO or Pd) via reduction. Principally,
be monitored as a function of time. reduction to peroxide occurs although re-
The oxidation and reduction reactions duction to superoxide or to water may also
can be monitored separately to obtain in- contribute on an actual TiO2 microparticle.
formation regarding the mechanism of the There can also be varying contributions by
photocatalytic reaction. In photocatalytic these three reduction reactions, depending
reactions with metal-deposited TiO2 par- on whether there is a deposited catalyst.
ticles and films, the deposited metal acts As already discussed, some photocatalytic
as a reduction site, thereby increasing the reaction pathways can involve • O2 − ; this
efficiency of photogenerated charge sepa- is a desirable product, and metallic cata-
ration. Thus, a TiO2 -ITO composite film lysts may in fact be counterproductive, if
can act as a simple model for the metal- they promote the two- or four-electron O2
deposited photocatalyst. A carbon micro- reduction reactions.
electrode is employed to detect the reaction In the presence of ethanol, although
products. By positioning this electrode dissolved oxygen was consumed at both
close to either the TiO2 or the ITO surface TiO2 and Pd sites, the consumption of
(50–500 µm), the oxidation and reduction oxygen was larger at the TiO2 surface [122].
reactions, respectively, can be monitored A possible scheme that is consistent with
on these regions. This method was used to O2 consumption at TiO2 is:
detect hydrogen peroxide at the two types •
OH + CH3 CH2 OH −−−→
of regions, separately, by use of a ‘‘wired’’
horseradish peroxidase microsensor fabri- CH3 C• HOH + H2 O (40)
cated on the carbon microelectrode sur- •
CH3 C HOH + O2 −−−→
face [122]. Later, it was found that peroxide
could also be detected adequately by setting CH3 CH(OH)OO• (41)
the microelectrode potential at a very pos-
producing an organoperoxyl radical, which
itive potential (+1.0 V versus SCE) [124].
can either participate in a chain-type
Additionally, the carbon microelectrode is
process
able to detect O2 by setting the potential at
a negative potential (−1.0 V versus SCE). CH3 CH(OH)OO• + CH3 CH2 OH −−−→
In one of the early studies, it was found
CH3 CH(OH)OOH + CH3 C• HOH (42)
that H2 O2 is produced largely at the ITO
portion of the illuminated TiO2 -ITO com- CH3 CH(OH)OOH −−−→
posite surface due to O2 reduction [122].
CH3 CHO + H2 O2 (43)
In a subsequent study on the separate
detection of dissolved oxygen in aque- or can react further with • O2 − or HO•2
ous solution at reducing regions (in this to produce a tetroxide intermediate, as al-
case, Pd) and oxidizing regions (TiO2 ), ready discussed. The importance of radical
using a carbon microelectrode, we found intermediates was found to be great, as
6.1 Fundamentals of Photocatalysis 529
UV irradiation UV irradiation
−500
−500
Current
−400
Current
[pA]
[pA]
−400
−300
−300
0 1 2 3 4 0 1 2 3
Time Time
(a) [min] [min]
UV irradiation UV irradiation
−300 −300
Current
−200 −200
Current
[pA]
[pA]
−100 −100
0 0
0 1 2 3 4 5 0 1 2 3 4 5
Time Time
(b) [min] [min]
UV irradiation UV irradiation
+100 +100
Current
Current
[pA]
[pA]
+50 +50
0 0
0 1 2 3 4 0 1 2 3
Time Time
(c) [min] [min]
Above TiO2 Above Pd
Fig. 17 Microelectrode current versus time curves, taken from Ikeda and
coworkers [124].
Without SOD
Ethanol With SOD
(a)
Acetaldehyde
air or water treatment, it is necessary to use In addition, metal ions can enhance
relatively intense UV light sources. It may oxidation rates by acting as electron
be possible to greatly increase the light- acceptors in themselves. Litter points out
harvesting ability of TiO2 through band that metal ions can be either beneficial
gap engineering. or detrimental [22]. Examples are Cu2+ ,
Fe3+ , and Ag+ , all of which can have
6.1.4.7.2 Metal-modified Photocatalysts beneficial effects in certain concentration
The area of metal-modified TiO2 ranges and detrimental effects at higher
photocatalysts is an extensive one but some concentrations. Metal ions can be present
aspects have been reviewed recently [22]. in water that is being photocatalytically
We will only cite a few examples here. treated and the effects must be taken into
We have already introduced two of account. Finally, Litter points out that the
the main themes: (1) improvement of addition of metal ions is generally too
electron transfer; and (2) improvement of expensive in large-scale applications.
light harvesting. One extremely important In terms of the modification of the light
application, the photocatalytic deposition absorption properties of TiO2 , there have
of metals, has also been treated in depth been several notable efforts. Anpo and
by Litter. coworkers have used the ion implantation
In the area of photocatalytic water split- technique extensively to extend the absorp-
ting, the use of deposited noble metals tion spectrum into the visible range [130].
has been studied for many years [11]. With the use of implanted Cr, the spectrum
The objective in this case is to enhance was shifted by as much as about 80 nm,
the transfer of photogenerated electrons with significant absorption at 450 nm. One
from the TiO2 conduction band to water challenge is to try to devise methods of
to produce hydrogen gas. This approach
preparing such photocatalysts with less
does indeed lead to improvements in H2
technology-intensive approaches.
production due to the well-known elec-
Zang and coworkers have prepared
trocatalytic activity of platinum for this
Pt(IV) chloride-modified amorphous TiO2
reaction. In the area of photocatalytic de-
that is effective at visible wavelengths [131].
composition of organic compounds, as
These authors have shown that the Pt is
already discussed, it is often desirable to
produce superoxide from O2 to assist in present in a complex ion form, still sur-
the formation of tetroxides, for example, rounded by chloride. This constitutes a
Ref. 101. The energetic requirement for highly effective use of the precious metal.
the production of superoxide is substan- Photocatalytic activity was found at wave-
tially less than that for hydrogen, so that in lengths as high as 546 nm. Hamerski and
principle, an electrocatalyst is not needed. coworkers have reported the improvement
In fact, as already mentioned, the use of a of photocatalytic activity for oil decomposi-
metallic catalyst may be counterproductive tion under sunlight with calcium-modified
if it leads to a multielectron product such as TiO2 [132].
H2 O2 or H2 O. In terms of simply increas- Murata and coworkers have thermally
ing the electron transfer rate for oxygen re- prepared rutile films on alloys of Ti with
duction a metallic catalyst, such as Pt or Pd, various 4d and 5d transition metals [133].
could be beneficial as shown by Gerischer, With the Ti-Ta and Ti-W alloys, visible
Heller, and coworkers [113, 129]. light activity was found. These results are
532 6 Nonsolar Energy Applications
54. G. Rothenberger, J. Moser, M. Grätzel et al., 80. S. I. Matsushita, T. Miwa, D. A. Tryk et al.,
J. Am. Chem. Soc. 1985, 107, 8054–8059. Langmuir 1998, 14, 6441–6447.
55. K. -M. Schindler, M. Kunst, J. Phys. Chem. 81. T. A. Egerton, C. J. King, J. Oil Col. Chem.
1990, 94, 8222–8226. Assoc. 1979, 62, 386–391.
56. J. M. Warman, M. P. Haas, P. Pichat et al., 82. C. Kormann, D. W. Bahnemann, M. R.
J. Phys. Chem. 1991, 95, 8858–8861. Hoffmann, Environ. Sci. Technol. 1991, 25,
57. S. T. Martin, H. Herrmann, W. Choi et al., 494–500.
J. Chem. Soc. Faraday Trans. 1994, 90, 83. G. P. Lepore, C. H. Langford, J. Vi-
3315–3322. chova et al., J. Photochem. Photobiol. A:
58. D. P. Colombo, Jr., R. M. Bowman, J. Phys. Chem. 1993, 75, 67–75.
Chem. 1995, 99, 11752–11756. 84. M. Anpo, N. Aikawa, Y. Kubokawa et al.,
59. D. Bahnemann, A. Henglein, J. Lilie et al., J. Phys. Chem. 1985, 89, 5689.
J. Phys. Chem. 1984, 88, 709–711. 85. R. F. Howe, M. Grätzel, J. Phys. Chem. 1985,
60. A. Henglein, Pure & Appl. Chem. 1984, 56, 89, 4495–4499.
1215–1224. 86. R. Cai, R. Baba, K. Hashimoto et al., J. Elec-
61. Y. Nosaka, J. Phys. Chem. 1990, 94, troanal. Chem. 1993, 360, 237–245.
3752–3755. 87. K. U. Ingold, Acc. Chem. Res. 1969, 2, 1–9.
62. M. O’Neil, J. Marohn, G. McLendon, J. Phys. 88. N. A. Clinton, R. A. Kenley, T. G. Traylor,
Chem. 1990, 94, 4356–4363. J. Am. Chem. Soc. 1975, 97, 3746–3751.
63. M. A. Grela, M. E. J. Coronel, A. J. Colussi, 89. N. A. Clinton, R. A. Kenley, T. G. Traylor,
J. Phys. Chem. B 1996, 100, 16940–16946. J. Am. Chem. Soc. 1975, 97, 3752–3757.
64. R. Kopelman, Science 1988, 241, 1620–1626. 90. N. A. Clinton, R. A. Kenley, T. G. Traylor,
65. B. G. Streetman, in Solid State Electronic J. Am. Chem. Soc. 1975, 97, 3757–3762.
Devices, Prentice-Hall International, En- 91. Y. Ohko, D. A. Tryk, K. Hashimoto et al.,
glewood Cliffs, New Jersey, USA, 1995, J. Phys. Chem. B 1998, 102, 2699–2704.
pp. 103–106. 92. G. B. Raupp, C. T. Junio, Appl. Surf. Sci.
66. D. Duonghong, M. Grätzel, J. Chem. Soc., 1993, 72, 321–327.
Chem. Commun., 1984, 1597–1599. 93. J. Stark, J. Rabani, J. Phys. Chem. B 1999,
67. Y. Nosaka, Y. Yamashita, H. Fukuyama, 103, 8524–8531.
J. Phys. Chem. B 1997, 101, 5822–5827. 94. J. Schwitzgebel, J. G. Ekerdt, H. Gerischer
68. V. Augugliaro, L. Palmisano, M. Schiavello, et al., J. Phys. Chem. 1995, 99, 5633–5638.
AIChE J. 1991, 37, 1096–1100. 95. G. A. Russell, J. Am. Chem. Soc. 1957, 79,
69. M. A. Aguado, M. A. Anderson, C. G. 3871–3877.
Hill, Jr., J. Mol. Catal. 1994, 89, 165–178. 96. S. Sitkiewitz, A. Heller, Nouv. J. Chem.
70. N. Serpone, J. Photochem. Photobiol. A: 1996, 20, 233–241.
Chem. 1997, 104, 1–12. 97. T. Minabe, P. Swanyama, Y. Kikuchi et al.,
71. L. Davydov, P. G. Smirniotis, S. E. Pratsinis, Electrochem. 1999, 67, 1132–1134.
Indust. Eng. Chem. Res. 1999, 38, 1376–1383. 98. T. Minabe, D. A. Tryk, P. Sawunyama et al.,
72. R. I. Bickley, J. Catal. 1973, 31, 398–407. J. Photochem. Photobiol. A: Chem. 2000, 137,
73. Y. Ohko, K. Hashimoto, A. Fujishima, 53–62.
J. Phys. Chem. A 1997, 101, 8057–8062. 99. A. V. Emiline, V. Ryabchuk, N. Serpone,
74. R. W. Matthews, J. Phys. Chem. 1987, 91, J. Photochem. Photobiol. A: Chem. 2000, 133,
3328–3333. 89–97.
75. I. Sopyan, S. Murasawa, K. Hashimoto 100. T. Noguchi, A. Fujishima, P. Sawunyama
et al., Chem. Lett. 1994, 723–726. et al., Environ. Sci. Technol. 1998, 32,
76. N. Negishi, T. Iyoda, K. Hashimoto et al., 3831–3833.
Chem. Lett. 1995, 841–842. 101. Y. Ohko, A. Fujishima, K. Hashimoto,
77. I. Sopyan, M. Watanabe, S. Murasawa et al., J. Phys. Chem. B 1998, 102, 1724–1729.
J. Photochem. Photobiol. A: Chem. 1996, 98, 102. W. W. Dunn, Y. Aikawa, A. J. Bard, J. Am.
79–86. Chem. Soc. 1981, 103, 3456–3459.
78. I. Sopyan, M. Watanabe, S. Marasawa et al., 103. W. W. Dunn, Y. Aikawa, A. J. Bard, J. Elec-
J. Electroanal. Chem. 1996, 415, 183–186. trochem. Soc. 1981, 128, 222–224.
79. S. Matsushita, T. Miwa, A. Fujishima, 104. M. D. Ward, A. J. Bard, J. Phys. Chem. 1982,
Chem. Lett. 1997, 925–926. 86, 3599–3605.
6.1 Fundamentals of Photocatalysis 535
Photocatalysis citations
1200
1000
Total citations
800
Patents
Number
600
400
200
0
1969 1973 1977 1981 1985 1989 1993 1997
Year
Fig. 1Numbers of papers and patents related to photocatalysis up to
September 2000 [4].
Self-cleaning Materials for residential Exterior tiles, kitchen and bathroom components,
and office buildings interior furnishings, plastic surfaces, aluminum
siding, building stone and curtains, paper
window blinds
Indoor and outdoor lamps Translucent paper for indoor lamp covers, coatings
and related systems on fluorescent lamps and highway tunnel lamp
cover glass
Materials for roads Tunnel wall, soundproofed wall, traffic signs and
reflectors
Others Tent material, cloth for hospital garments and
uniforms and spray coating for cars
Air cleaning Indoor air cleaners Room air cleaner, photocatalyst-equipped air
conditioners, and interior air cleaner for factories
Outdoor air purifiers Concrete for highways, roadways and footpaths,
tunnel walls, soundproofed walls, and building
walls
Water purification Drinking water River water, groundwater, lakes, and water-storage
tanks
Others Fish feeding tanks, drainage water, and industrial
wastewater
Antitumor activity Cancer therapy Endoscopic-like instruments
Self-sterilizing Hospital Tiles to cover the floor and walls of operating
rooms, silicone rubber for medical catheters,
and hospital garments and uniforms
Others Public rest rooms, bathrooms, and rat-breeding
rooms
538 6 Nonsolar Energy Applications
because it operates in a passive fashion, (UV) illumination [33]. Since then there
that is, without the need for electrical have been several reports and patents ap-
power or chemical reagents, only light and pearing on this topic [5, 34–44]. Blake
oxygen being required. For these reasons and coworkers have published a compre-
photocatalysis recently has attracted the at- hensive review of this, including a list
tention not only of the medical world but of patents on photocatalytic disinfection
also of the construction industry in Japan. and killing of cancer cells using TiO2 [5].
Most of the work reported on photocat- Many of the commercial products avail-
alytic disinfection has been carried out in able in the market at present are based on
aqueous suspensions, in view of applica- self-sterilizing photocatalytic surfaces and
tions in water sterilization. In 1985, for disinfection of indoor air.
the first time, Matsunaga and coworkers Bacteria find places in bathrooms,
reported the possibility of killing mi- kitchens, and hospital operating rooms
crobial cells in water using a TiO2 -Pt where they can multiply. Once they find
(platinum) photocatalyst under ultraviolet a suitable environment, they can increase
6.2 Applications of TiO2 Photocatalysis 539
in number at an exponential rate. The key TiO2 -coated tile and, when the surface
point regarding photocatalysis is that it is irradiated with UV light, the metal
can be very effective in controlling bacte- cations are photocatalytically converted to
ria while they still are relatively small in the metallic form as ultrafine particles that
number, before they start multiplying out are firmly attached to the surface. The
of control. photocatalytic deposition of various metals
on TiO2 is also discussed in a later section.
6.2.2.1 Photocatalytic Antibacterial Tiles The present method of metal deposi-
In general, the effective use of photocat- tion was found to have an advantage over
alytic technology depends heavily on the the conventional method in which the
specific way in which the TiO2 powder metal powders are mixed with the glaze
is used. One of the simplest approaches, and fired on a tile [25]. With the conven-
often used in model studies on water treat- tional method, most of the metal particles
ment, is to mix the powder with the water are submerged in the glaze, only a few
and expose the suspension to UV light or being exposed on the surface. Thus, the
sunlight. Although the separation of the particles in the glaze are not effective for
catalyst powder from the water is not an the antibacterial effect. The new photo-
insurmountable problem, an alternate ap- catalytic approach allows the formation of
proach is to immobilize the powder on a a high-density layer of antibacterial metal
substrate [25]. particles on the TiO2 layer, which then ex-
In view of indoor applications, various hibits a strong antibacterial activity. The
kinds of substrates such as glass plates and photocatalytic antibacterial tiles prepared
tiles have been examined as substrates. this way are highly effective and durable.
It has been found that even very thin
films (less than a few micrometers in 6.2.2.2 Sterilizing Effect of TiO2
thickness) are photocatalytically active [45, As already indicated, one of the promising
46]. Moreover, (∼1 µm) TiO2 films on the applications of TiO2 is photocatalytic ster-
surfaces of ceramic tiles have been found ilization. Escherichia coli has been studied
to be effective for antibacterial functions. widely, but work on other organisms also
These films can be prepared by spraying has been reported. As recently reviewed,
a liquid suspension containing TiO2 on various organisms have been examined for
a glazed ceramic tile, followed by heating photocatalytic sterilization [5]. Although
at 800 ◦ C [25, 47, 48]. Films prepared this the feasibility of sterilization was realized
way are attached to the tiles so strongly from the beginning, the mechanism of
that they are difficult to scrape off. cell death was not understood completely.
In the case of indoor antibacterial Initially, the hydroxyl radical was thought
tiles, one has to keep in mind the to be the main species responsible for
possibility of accumulation of bacteria in cell death. However, peroxide currently is
areas either temporarily or permanently considered to be one of the most active
inaccessible to light. To overcome this agents. Evidence against the involvement
problem, silver or copper metal particles of hydroxyl radicals is that there is lit-
have been deposited photocatalytically on tle effect of mannitol, a hydroxyl radical
catalyst surfaces [25]. These metals are scavenger [39].
widely used as antimicrobial agents. Either A typical experiment involves plac-
of the metal solutions is sprayed on the ing 150 µL of an E. coli suspension
540 6 Nonsolar Energy Applications
Photocatalytic
Ordinary tile
antibacterial tile
Methicillin-
resistant Staph.
aureus (MRSA)
E.coli
Pseudomonas
aeruginosa
Number of airborne
30
bacteria
20 Installation of photocatalytic tile
10
0
Immediately 1 1.5 2 2.5 3
after Month Months Months Months Months
installation later later later later later
The applications of these tiles can be ex- TiO2 is capable of decomposing not only
tended further to public toilet facilities, dead cells but also the toxins released
shower rooms, and biological experimen- during cell death. Jacoby and coworkers
tal facilities such as breeding rooms for have demonstrated the complete oxidation
experimental rats. The tests conducted in of E. coli cells to carbon dioxide (CO2 ) [40].
these facilities have indicated that the an- In the case of E. coli, TiO2 decomposes
tibacterial tiles are effective not only in both the living cells and the endotoxin
killing the bacteria but also in significantly released from these cells at the time
reducing the odor caused by ammonia, of death [49]. This unique function of
which is common in such facilities. The de- TiO2 has attracted tremendous attention in
odorizing functions of TiO2 are discussed Japan. Endotoxin is a cell wall constituent
in a later section. of bacteria that consists of a sugar chain,
an O-antigen, and a complex lipid referred
6.2.2.3 Destruction of Microbial Toxins to as lipid A. This endotoxin is toxic and
A major advantage of photocatalytic an- can cause critical problems in medical
tibacterial tiles over conventional antibac- facilities and in factories manufacturing
terial tiles is that they not only kill bacte- pharmaceuticals and medical devices, as
rial cells but also completely decompose well as in food-processing factories. In the
them. Conventional self-sterilizing sur- summer of 1996 in western Japan, there
faces, such as those with silver coatings, are was a serious outbreak of food poisoning,
not useful for long-term use because dead in which nearly 1800 people, including
cells can eventually cover the surface, ren- many children, were hospitalized, and
dering it ineffective. The advantage of TiO2 twelve died. The poisoning was caused
is that it continues to work well even when by the O-157 endotoxin.
there are cells covering the surface and Figure 5 shows the concentration ch-
while the bacteria are actively propagating. ange of endotoxin and the survival ratio
542 6 Nonsolar Energy Applications
120 120
Remaining endotoxin
(in dark)
100 100
Remaining endotoxin
80 80
Live bacteria
[%]
[%]
60 60
Remaining endotoxin
(under illumination)
40 40
20 20
Live bacteria
0 0
0 1 2 3 4 5
Irradiation time
[hours]
Fig. 5Sterilization of E. coli and endotoxin decomposition (endotoxin
amount: 3–8 EU cc−1 = 100%). Both sterilization and endotoxin
decomposition commence immediately [49].
of E. coli as a function of illumination time new cell growth and old cell death are
on a TiO2 -coated glass plate. It was found kept in balance. In cancer, this balance
that two hours were required to decom- is disrupted. This disruption can result
pose most of the toxin and four hours in the formation of a malignant tu-
were required to decompose it completely mor. The selective killing of cancer cells
under UV illumination (0.4 mW cm−2 ) on in vivo by photocatalysis is a challeng-
a TiO2 film [49]. This can be contrasted ing task.
with other techniques that have been de- The possibility of applying photocatalytic
veloped, for example, thermal treatment at techniques to the photodynamic therapy
250 ◦ C for 30 minutes or chemical treat- (PDT) of cancer led to early efforts in this
ment in ethanol containing NaOH. These area. It was found that the multiplication
results clearly indicate that the antibacte- of human malignant cells was suppressed
rial effects of TiO2 -coated materials involve remarkably both in vitro [50–54] and in
not only the nullification of the viability of vivo [55, 56] in the presence of photoirra-
the bacteria but also the destruction of the diated TiO2 . Initially, TiO2 was found to
bacterial cell. be effective in killing HeLe cells under UV
irradiation [50]. Later, various experimen-
6.2.2.4 Photocatalytic Cancer Therapy tal conditions were examined, including
Cancer treatment is carried out mainly by the effect of superoxide dismutase, which
means of radiation and chemical thera- enhances the effect, caused by the produc-
pies, which may generate a wide range tion of peroxide [51–53]. The possibility of
of side effects. Photocatalysis offers a selectively killing a cancer cell was demon-
new approach in cancer treatment. Can- strated using a polarized, illuminated TiO2
cer arises from a loss of normal growth microelectrode [54]. Photocatalytic killing
control. In normal tissues, the rates of of human U937 monocytic leukemia cells
6.2 Applications of TiO2 Photocatalysis 543
also was reported by Chinese researchers to cause tumors to form. When the size
using colloidal TiO2 [57]. A complete of the tumors grew to about 0.5 cm, a
photocatalytic destruction of these cells solution containing fine particles of TiO2
was observed after 30 minutes of UV was injected. After two or three days, the
illumination. The destruction process in- skin was cut open to expose the tumor,
volved membrane blistering and DNA which then was irradiated. This treatment
fragmentation. All of these effects are char- clearly inhibited tumor growth (Fig. 6).
acteristics of apoptosis (programmed cell After 13 more days, the treatment was
death) that is known to occur when can- repeated with TiO2 photocatalyst and a
cer cells are treated with antitumor agents, further marked antineoplastic effect was
radiation, or high temperature. observed [55].
In collaboration with urologists at Yoko- Although this technique offers some
hama City University, animal experiments promise for cancer therapy, it may not be
also have been conducted. Cancer cells effective in treating large tumors. To make
were implanted under the skin of mice this technique more useful, a device (a
2
2
1
1
(a) (b)
Fig. 6 Animal test of photocatalytic cancer injected into Tumor 1. Tumor 2, which was not
therapy; photograph of nude mouse just after injected with TiO2 particles, was also opened
initial treatment (a) and four weeks after surgically. Both tumors were irradiated with a
treatment (b). TiO2 powder (0.4 mg) was mercury lamp for one hour [25].
Eye
TiO2 powder
UV light
Optical fiber
100
80
Quantum efficiency
60
[%]
40
20
0
Fig. 8 Inverse relation between
100 101 102 103 104 105 quantum efficiency and UV light
UV intensity intensity for low-pollutant
[µW cm−2] concentrations [25].
6.2 Applications of TiO2 Photocatalysis 545
most appropriate applications for the that TiO2 particles should be aggregated
photocatalytic approach are those that before they are added to the paper pulp to
involve low concentrations of pollutants maintain the paper strength. If the TiO2
but that also involve serious risks to health particles are dispersed in the paper, there
or comfort. A malodorous pollutant is a is photocatalytic attack on the fibers and
good example because the actual quantity loss of strength. Acetaldehyde degradation
of the chemical may be very small. The was seen with the TiO2 -containing papers
photocatalytic sterilization discussed in the even under very weak UV irradiation. The
previous section is another example. decomposition rate was found to increase
The Japanese government regulates the with increasing amounts of TiO2 con-
concentrations of major malodorous sub- tained in the paper [61]. Furthermore, the
stances as ammonia, hydrogen sulfide, papers showed high stability under illumi-
methyl mercaptan, and acetaldehyde in the nation for about 600 hours. These papers
environment. Methyl mercaptan, for ex- also were tested for the decomposition of
ample, is regulated to concentrations lower cigarette smoke residue. Smoke from one
than 0.002–0.01 ppm. The human sense of cigarette was adsorbed on a 5 × 5-cm sheet
smell, however, senses it at a concentration of TiO2 -containing paper. TiO2 -free paper
as low as 0.00012 ppm. A conventional flu- was also used for control experiments. The
orescent lamp can be used as a light source
smoke residue on the TiO2 paper disap-
for a photocatalytic reactor to control this
peared after only two hours under black
substance [25]. Regarding acetaldehyde,
light (1 mW cm−2 ) illumination and after
which is a highly malodorous compound,
one week under regular fluorescent light
a fluorescent lamp is also sufficient to
(10 µWcm−2 ).
decompose it at an air concentration of
On the basis of these findings, vari-
5 ppm, which is 10–100 times higher than
ous kinds of commercial photocatalytic air
that at which it becomes offensive. Several
cleaners currently are being produced in
studies have been conducted on acetalde-
hyde decomposition under various light Japan by Daikin Industries and are being
intensities and concentrations to under- marketed worldwide (Fig. 9). The Ishihara
stand the reaction mechanism [45, 58–60]. Techno Corporation has developed pho-
Another interesting application of TiO2 tocatalytic filter elements for air cleaners.
photocatalysis was demonstrated using They contain TiO2 and activated carbon
TiO2 -containing paper. In collaboration (C), the latter trapping pollutants for sub-
with the Gifu Prefectural Paper Re- sequent destruction by the irradiated TiO2 .
search Institute, TiO2 -containing papers These filters feature a honeycomb-type
possessing a high catalytic activity toward construction for minimum pressure drop.
the decomposition of acetaldehyde and The air cleaners are available in various
cigarette smoke residues were prepared. sizes, which have been developed to clean
The TiO2 -containing paper was prepared the air in a car or in an entire factory or
from a mixture of softwood kraft pulp and hospital.
TiO2 aqueous solution. The details of the
preparation were reported earlier [61]. The 6.2.3.2 Treatment of Industrial Gaseous
amount of added TiO2 was varied from 2 Effluents
to 10% based on the weight of the pulp. Because gas-phase photocatalysis is in gen-
The important point in this preparation is eral more than 10 times more effective
Pre-filter Plasma-ionizing unit Double-layer roll Photocatalyst 546
photocatalyst filter
Double-layer construction means
more power. Energy savings by using
an inverter motor.
6 Nonsolar Energy Applications
Fig. 9 Schematic diagram of the components of an indoor room air-purification unit (Courtesy of Daikin Industries).
6.2 Applications of TiO2 Photocatalysis 547
than that in liquid phase [22] the use of high levels of water vapor (>4000 ppm)
photocatalytic purification methods for in- were found to decrease the degradation
dustrial gas effluents has gained much efficiency for acetone [76]. The role of
attention. Almost all kinds of organic com- water vapor in gas-phase photocatalysis
pounds are photocatalytically degradable is not completely understood. Even for
on TiO2 . Gas-phase carbon tetrachloride the same compound, divergent results
(CCl4 ) appears to be an exception, in that can be obtained by different groups [77,
it is resistant to photocatalytic degrada- 78]. This may result from subtle dif-
tion [62, 63]. The presence of methanol ferences in the film microstructure. For
helps in slightly increasing the photo- example, one film may have relatively
catalytic reduction of CCl4 , the effect small pores and may become flooded
being higher in aqueous solutions [63, easily in the presence of high humidity,
64]. There have been several reports on but such a film may be more effective
the photocatalytic oxidation of individual in dry conditions because water that is
gaseous substrates such as halogenated produced during the photocatalytic reac-
organics [62, 65–67], alkanes [68], alco- tion may be trapped more effectively [78].
hols [62, 69], aldehydes [60, 70], aromat- This is an area that requires further
ics [62, 66, 71, 72], and heterocycles [73]. study.
Among these organic substrates, only a
The design of suitable reactors is very
few compounds such as TCE, methanol,
crucial for the achievement of maxi-
and ethanol exhibited quantum efficien-
mum conversion efficiencies. Ollis has
cies between 80 and 100% [22, 62]. In
made comparisons of various types of
contrast, the quantum efficiencies for the
photoreactors for air purification [9]. Al-
degradation of nonhalogenated organic
though monolithic photocatalytic reactors
substrates such as benzene, toluene, and
are well suited for air handling in build-
xylene were found to be below 15% [62].
ings and exhausts, fluidized-bed reactors
However, a drastic enhancement in the
efficiencies for the degradation of some are advantageous for large-scale opera-
of these compounds has been reported tions. The monolithic catalysts allow a
to be possible by mixing with chlorocar- drastic reduction in the pressure drop
bons such as TCE that cause radical chain produced by the passage of the gas
reactions [62]. through the catalyst. Monolithic structures
The presence of water vapor plays an increase the illumination area of the cata-
important role in the gas-phase photo- lyst in a three-dimensional way. Although
catalysis. For example, moderate levels fluidized-bed reactors offer good catalyst-
of water vapor promote the photocatalytic light and catalyst-gas contacts, bubble-
degradation of toluene and m-xylene [66, phase formation is a disadvantage. Re-
74], whereas they inhibit the degradation cently, Vorontsov and coworkers [76], have
of ethylene [75]. In some cases, such as demonstrated photocatalytic oxidation of
TCE [62], deactivation of the photocata- acetone using a vibrofluidized-bed reac-
lyst was observed in the absence of water tor. In this type of reactor, fluidization
vapor, which probably is due to the ex- is achieved by vibration. The advantage
haustion of surface hydroxyls or to the is that it can operate under low-gas flow
formation of an intermediate species that rates compared to conventional fluidized-
blocks the active catalyst sites. However, bed reactors.
548 6 Nonsolar Energy Applications
6.2.3.3 Outdoor Air Purification (NOx gas may be trapped. The accumulation of
Removal) HNO3 on the catalyst surface retards the
Air pollution, especially nitrogen oxides photocatalytic action. Therefore it is nec-
(NOx ) contamination from the combus- essary to wash the catalyst after certain
tion of hydrocarbons, is a particularly time intervals. However, if the photocata-
serious problem in urban areas. Despite lyst is used outdoors, the products may be
serious efforts toward emission control, washed off by rainfall. A similar mecha-
the concentrations of NOx often ex- nism is expected for SO2 removal as well.
ceed the air-quality standard, especially The range of concentrations of air
in large cities. TiO2 photocatalysis ap- pollutants that can be removed efficiently
pears to be a promising technology for is from 0.01 ppm to 10 ppm. These
the removal of low concentrations of NOx concentrations range from those present
from ambient air [79]. Daikin Industries typically in the atmosphere to those
has demonstrated the efficient removal in highway tunnels. However, at the
of NOx from indoor air using a photo- end of an automobile’s exhaust pipe,
catalyst coated on activated carbon [25]. the concentration of pollutants is higher
According to their results, the indoor NOx than 100 ppm. Therefore, attaching TiO2
concentration decreases from 0.1 ppm to directly to the car exhaust system does
0.06 ppm (the air-pollution standard) in not seem to be a good idea. This is better
25 minutes. handled by a catalytic converter.
The mechanism of nitrogen oxide (NO) Further efforts are being made to im-
removal by a photocatalyst is somewhat prove the catalytic response of TiO2 for
complicated. It is assumed that the NO is NOx removal. For example, Nakamura
photocatalytically oxidized to NO2 , which and coworkers [80], have used hydrogen
is finally converted to nitric acid (HNO3 ) plasma–treated TiO2 to sensitize the
(Fig. 10) [25]. At the NO2 stage, part of catalytic response in the visible wave-
the gas may escape from the photocatalyst length region to make use of sunlight
surface but with an adsorbent such as acti- more efficiently. Anpo and coworkers
vated carbon mixed with the catalyst, this have used chromium (Cr) ion–implanted
O2
UV light NOx
(300–400 nm)
NO NO
e−
O2 −
NO2 NO2
O−
TiO2 OH Activated carbon
oxidation h+ HNO3 (adsorption)
H2O Rinse H2 O
Drainage
Neutralization with lime
TiO2 to obtain enhanced response for adsorption period of one hour in the
decomposition in the visible region [81]. dark. After four hours of photocatalytic
Because photocatalytic reactions are not treatment under UV light, the chemical
thermochemical, it is not necessary to oxygen demand (COD) decreased by 95%,
supply heat. They proceed at temperatures the water turning clear and odorless.
available in the ambient environment. The advantage of photocatalysis in water
The required UV light intensity is not purification is the complete mineralization
more than about 0.1 mW cm−2 , which of organics caused by the photogeneration
corresponds to the intensity level of of • OH radicals originating from water
outdoor sunlight that can be found on via the OH groups of the TiO2 surface.
a cloudy winter day. Outdoor experiments In the case of gas-phase catalysis, water
have confirmed that photocatalytic effects vapor must be present to achieve com-
can be observed from sunrise to sunset. plete mineralization for most compounds.
Photocatalysts, however, are by no means However, a disadvantage with regard to the
a cure-all. Further efforts are necessary to photocatalytic purification of water is the
use fully their special characteristics. lower quantum efficiencies in comparison
to gas-phase photocatalysis. The principal
6.2.4 reason for this is the smaller availability of
Photocatalytic Water Purification oxygen in liquid water compared to air.
Several reports have appeared in the lit-
6.2.4.1 Wastewater Purification erature on the photocatalytic treatment
In a recent article, Herrmann has shown of industrial effluents such as textile
the feasibility of photocatalytic treatment effluents [82–84], rinse waters of pes-
of an actual highly loaded industrial ticide containers [85], wastewater from
wastewater (Fig. 11) [26]. In that exper- a phenolic resin factory [86], distillation
iment, the foul-smelling black-colored effluents from a pharmaceutical com-
waste was diluted 1000 times before treat- pany [86], waters that are produced dur-
ment. The treatment started after an ing recovery of natural gas and crude
1000
+UV
800
600
COD
400
200
Catalyst pane
Waste Treated
water Recycle if necessary water
UV lamp
Water
These metallic species can be recovered only metal ions with potentials more
from the photocatalyst by mechanical or positive than ∼0.4 V can be reduced on
chemical methods. TiO2 . Prairie and coworkers demonstrated
Dissolved metals in the form of ions the reduction of CrVI , AuIII , HgII , PdII ,
can affect the photocatalytic reaction rates PtIV , and AgI in deaerated conditions in
and efficiencies of organic degradation the presence of salicylic acid [106]. In-
reactions to a significant extent. Recently, significant metal deposition is observed for
Litter has published an extensive review platinum (Pt), silver (Ag), and gold (Au) in
on this subject [27]. Both increases [101, the absence of a hole scavenger [107]. With
102] and decreases [101, 103] in the pho- an alcohol as a hole scavenger, metal depo-
tocatalytic reaction rates were reported, sition may be an indirect reduction process
depending on the metal ion concentrations involving • ROH radicals produced from al-
in the aqueous solutions. At low metal cohol degradation by OH• radicals, rather
concentrations, trapping of photogener- than a direct capture of conduction band
ated electrons by the metal ions, with electrons. These are the two possibilities
resulting inhibition of electron-hole re- for the direct photocatalytic deposition of
combination, was proposed to be one of metals. A third, indirect pathway involves
the reasons for the enhanced reaction metal species such as Pb2+ , Mn2+ , and
rates [101, 102]. Photo-Fenton-type reac- Tl+ , with relatively negative redox poten-
tions may also take place in the presence tials, which may follow an oxidative route
of metal ions by reaction with photo- through hole attack because of the exis-
generated H2 O2 [104]. Photogeneration of tence of stable higher oxidation states [27].
H2 O2 caused by oxygen reduction has been Litter has summarized the pathways for
demonstrated [105]. the photocatalytic deposition of several
The detrimental effect on the oxidation metals [27].
rate of the presence of high metal
ion concentrations was attributed to the 6.2.5
oxidation of the reduced metal ions, Self-cleaning and Antifogging Surfaces
either by • OH radicals or directly by
photogenerated holes [104]. Other reasons In 1997, it was reported that TiO2 surfaces
include the filter effect as a result of can become highly hydrophilic (super-
absorption of UV light by the metal species hydrophilic) under UV illumination [108,
(e.g. FeII , CuII , and NiII ) and precipitation 109]. Although the mechanism for this
of dissolved metals as hydroxides on the effect still is not completely understood,
photocatalyst [103]. great progress has been made, and it
Photocatalytic removal of metals is an is clear that it involves reactions that
important application in its own right. are distinct from those involved in the
There are three pathways for metal re- photocatalytic decomposition of organic
moval, two direct and one indirect. Direct compounds [109–111]. In fact, the pos-
reduction of metal ions by conduction sibility that the conversion of the TiO2
band electrons is the simplest way. How- surface from hydrophobic to hydrophilic
ever, the redox potential of the metallic was due to the photocatalytic removal of
couple (Mn+ /M0 ) should be more positive a thin layer of organic contamination was
than the conduction band edge (Fig. 14). considered, but mounting evidence has
Experimental results have indicated that shown this not to be true. For example,
6.2 Applications of TiO2 Photocatalysis 553
−1.0 Mn2+/Mn0
Zn2+/Zn0
~
~
e− Cd2+/Cd0
vb
−0.3
Ni2+/Ni0
Pb2+/Pb0
0.0
Cu2+/Cu0
Pt4+/Pt0
+0.5
hν
Ag+/Ag0
Hg2+/Hg0
+1.0 AuCl4−/Au0
Cr2O72−/Cr 3+
PbO2/Pb2+
+1.5
~
~
cb +2.9
h+
the surface wettability, which increases The latter is thought to oxidize reduced
under relatively low intensity UV illumi- surface states (Ti3+ ) by • OH radicals or
nation over the span of about one hour hydrogen peroxide. The surface can then
(Fig. 15), can be converted back to the be reconverted to the hydrophilic state by
hydrophobic state by ultrasonic treatment. illumination [110].
554 6 Nonsolar Energy Applications
20
[°]
15
10
0
0 20 40 60 80 100 120
UV irradiation time
[min]
Water
Oil is spontaneously
removed from the
surface.
Time
Superhydro-
phylic surface
Water
Normal surface
Fig. 16 Series of photographs showing how an oil droplet adhering to a polycrystalline
TiO2 film on glass is gradually removed from the surface by water as a result of a
UV-induced increase in hydrophilicity [25].
6.2 Applications of TiO2 Photocatalysis 555
A B A B
B A B A
Fig. 18 Photograph showing heavy tarpaulin material, noncoated on the left and TiO2 -coated
on the right, after exposure to urban air for several months.
556 6 Nonsolar Energy Applications
Fig. 19 Photograph of a temporary storage structure made from TiO2 -treated tarpaulin
(courtesy of Taiyo Kogyo Corp.).
6.2 Applications of TiO2 Photocatalysis 557
appreciate Dr. D.M. Blake for providing 16. A. L. Linsebigler, G. Lu, J. J. T. Yates, Chem.
copies of his comprehensive bibliography Rev. 1995, 95, 735–758.
17. K. Rajeshwar, J. Appl. Electrochem. 1995, 25,
on photocatalysis. 1067–1082.
18. Y. Parent, D. Blake, K. M. Bair et al., Solar
References Energy 1996, 56, 429–437.
19. U. Stafford, K. A. Gray, P. V. Kamat, Het-
1. D. M. Blake, in Bibliography of Work on erogen. Chem. Rev. 1996, 3, 77–104.
the Photocatalytic Removal of Hazardous 20. K. Vinodgopal, P. V. Kamat, CHEMTECH
Compounds from Water and Air, NREL/TP- 1996, 26, 18–22.
430-6084, National Renewable Energy Lab- 21. A. Fujishima, T. N. Rao, Proc. Indian Acad.
oratory, Golden, Colorado, 1994. Sci. (Chem. Sci.) 1997, 109, 471–486.
2. D. M. Blake, in Bibliography of Work on 22. A. Mills, S. LeHunte, J. Photochem. Photo-
biol. A: Chem. 1997, 108, 1–35.
the Photocatalytic Removal of Hazardous
23. J. Peral, X. Domenech, D. F. Ollis, J. Chem.
Compounds from Water and Air, NREL/TP-
Tech. Biotechnol. 1997, 70, 117–140.
473-20300, Update No. 1 (to June 1995),
24. N. Serpone, J. Photochem. Photobiol. A:
National Renewable Energy Laboratory,
Chem. 1997, 104, 1–12.
Golden, Colorado, 1995.
25. A. Fujishima, K. Hashimoto, T. Watanabe,
3. D. M. Blake, in Bibliography of Work on
in TiO2 Photocatalysis: Fundamentals and
the Photocatalytic Removal of Hazardous
Applications, BKC, Inc., Tokyo, Japan, 1999.
Compounds from Water and Air, NREL/TP-
26. J.-M. Hermann, Catal. Today 1999, 53,
430-22197, Update No. 2 (to October 1996),
115–129.
National Renewable Energy Laboratory,
27. M. I. Litter, Appl. Catal. B: Environ. 1999,
Golden, Colorado, 1996. 23, 89–114.
4. D. M. Blake, in Bibliography of Work on 28. M. Romero, J. Blanco, B. Sanchez et al.,
the Photocatalytic Removal of Hazardous Solar Energy 1999, 66, 169–182.
Compounds from Water and Air, NREL/TP- 29. O. M. Alfano, D. Bahnemann, A. E. Cassano
570-26797, Update No. 3 (to January 1999), et al., Catalysis Today 2000, 58, 199–230.
National Renewable Energy Laboratory, 30. A. Fujishima, T. N. Rao, D. A. Tryk, J. Pho-
Golden, Colorado, 1999, available at: tochem. Photobiol. C 2000, 1, 1–21.
http://www.ott.doe.gov/coolcar/pubs.html. 31. M. Tomkiewicz, Catal. Today 2000, 58,
5. D. M. Blake, P. Maness, Z. Huang et al., 115–123.
Sep. Purif. Methods. 1999, 28, 1–50. 32. D. A. Tryk, A. Fujishima, K. Honda, Elec-
6. M. A. Fox, M. T. Dulay, Chem. Rev. 1993, trochim. Acta 2000, 45, 2363–2376.
93, 341–357. 33. T. Matsunaga, R. Tomada, T. Nakajima
7. P. V. Kamat, Chem. Rev. 1993, 93, 267. et al., FEMS Microbiol. Lett. 1985, 29,
8. A. Mills, R. H. Davies, D. Worsley, Chem. 211–14.
Soc. Rev. 1993, 22, 417. 34. C. Wei, W.-Y. Lin, Z. Zainal et al., Environ
9. D. F. Ollis, H. Al-Ekabi, Eds. in Photocat- Sci. Technol. 1994, 28, 934.
alytic Purification and Treatment of Water 35. T. Matsunaga, M. Okochi, Environ. Sci.
and Air, Elsevier, Amsterdam, 1993. Technol. 1995, 29, 501–505.
10. H. Yoneyama, Crit. Rev. Solid State 1993, 36. R. J. Watts, S. Kong, M. P. Orr et al., Water
18, 69–111. Res. 1995, 29, 95–100.
11. W. A. Zeltner, C. G. Hill, M. A. Anderson, 37. M. Bekbolet, C. V. Araz, Chemosphere 1996,
CHEMTECH 1993, 23, 21–28. 32, 959–965.
12. N. Serpone, Res. Chem. Intermed. 1994, 20, 38. M. Bekbolet, Water Sci. Technol. 1997, 35,
953–992. 95–100.
13. K. I. Zamaraev, M. I. Khramov, V. N. 39. Y. Kikuchi, K. Sunada, T. Iyoda et al., J.
Parmon, Catal. Rev. 1994, 36, 617. Photochem. Photobiol. A: Chem. 1997, 106,
14. A. Heller, Acc. Chem. Res. 1995, 28, 51–56.
503–508. 40. W. A. Jacoby, B. C. Maness, E. J. Wolfrum
15. M. R. Hoffmann, S. T. Martin, W. Choi et al., Environ. Sci. Technol. 1998, 32,
et al., Chem. Rev. 1995, 95, 69–96. 2650–2653.
560 6 Nonsolar Energy Applications
41. K. Sunada, Y. Kikuchi, K. Hashimoto et al., 62. M. L. Sauer, M. A. Hale, D. F. Ollis, J. Pho-
Environ. Sci. Technol. 1998, 32, 726–728. tochem. Photobiol. A: Chem. 1995, 88,
42. P.-C. Maness, S. Smolinski, D. M. Blake 169–178.
et al., Appl. Environ. Microbiol. 1999, 65, 63. R. M. Alberici, W. F. Jardim, Appl. Catal. B:
4094–4098. Environ. 1997, 14, 55–68.
43. A. Vidal, A. I. Diaz, A. E. Hraiki et al., Catal. 64. J. Stark, J. Rabani, J. Phys. Chem. B 1999,
Today 1999, 54, 283–290. 103, 8524–8531.
44. Z. Huang, P.-C. Maness, D. M. Blake et al., 65. W. A. Jacoby, M. R. Nimlos, D. M. Blake
J. Photochem. Photobiol. A: Chem. 2000, 130, et al., Environ. Sci. Technol. 1994, 28,
163–172. 1661–1668.
45. I. Sopyan, M. Watanabe, S. Murasawa et al., 66. Y. Luo, D. F. Ollis, J. Catal. 1996, 163, 1–11.
J. Photochem. Photobiol. A: Chem. 1996, 98, 67. O. Hennezel, D. F. Ollis, J. Catal. 1997, 167,
79–86. 118–126.
46. Y. Ohko, K. Hashimoto, A. Fujishima, J. 68. K. Wada, K. Yoshida, Y. Watanabe et al.,
Phys. Chem. A 1997, 101, 8057–8062. J. Chem. Soc., Faraday Trans. 1996, 92,
47. T. Watanabe, A. Kitamura, E. Kojima et al., 685–691.
U.S. Patent 5,874,701, February 23, 1999, in 69. S. A. Larson, J. A. Widegren, J. L. Falconer,
U. S. Patent & Trademark Office (Patent Full J. Catal. 1995, 157, 611–625.
Text and Image Database), Toto Co., Ltd., 70. N. Takeda, M. Ohtani, T. Torimoto et al., J.
USA, 1999. Phys. Chem. B 1997, 101, 2644–2649.
48. T. Watanabe, A. Kitamura, E. Kojima et al., 71. T. N. Obee, R. T. Brown, Environ. Sci. Tech-
U.S. Patent 6,139,803, October 31, 2000, in nol. 1995, 29, 1223–1231.
U. S. Patent & Trademark Office (Patent Full 72. N. N. Lichtin, M. Sedeghi, J. Photochem.
Photobiol. A: Chem. 1998, 113, 81–88.
Text and Image Database), Toto Co., Ltd.,
73. P. Kopf, E. Gilbert, S. H. Eberle, J. Pho-
USA, 2000.
tochem. Photobiol. A: Chem. 2000, 136,
49. K. Sunada, Y. Kikuchi, K. Hashimoto et al.,
163–168.
Environ. Sci. Technol. 1998, 32, 726–728.
74. J. Peral, D. F. Ollis, J. Catal. 1992, 136,
50. A. Fujishima, J. Ootsuki, T. Yamashita
554–565.
et al., Photomed. Photobiol. 1986, 8, 45–46.
75. S. Yamazaki, S. Tanaka, H. Tsukamoto, J.
51. R. Cai, K. Hashimoto, K. Itoh et al., Bull.
Photochem. Photobiol. A: Chem. 1999, 121,
Chem. Soc. Jpn. 1991, 64, 1268–1273. 55–61.
52. R. Cai, K. Hashimoto, Y. Kubota et al., 76. A. V. Vorontsov, E. N. Savinov, P. G.
Chem. Lett. 1992, 427–430. Smirniotis, Chem. Eng. Sci. 2000, 55,
53. R. Cai, H. Sakai, K. Hashimoto, Y. Kubota 5089–5098.
et al., Denki Kagaku 1992, 60, 314–321. 77. S. Sitkiewitz, A. Heller, Nouv. J. Chim. 1996,
54. H. Sakai, R. Baba, K. Hashimoto et al., 20, 233–241.
Chem. Lett. 1995, 185–186. 78. T. Minabe, D. A. Tryk, P. Sawunyama et al.,
55. R. Cai, Y. Kubota, T. Shuin et al., Cancer J. Photochem. Photobiol. A: Chem. 2000, 137,
Res. 1992, 52, 2346–2348. 53–62.
56. A. Fujishima, R. X. Cai, J. Otsuki et al., Elec- 79. T. Ibusuki, K. Takeuchi, J. Mol. Catal. 1994,
trochim. Acta 1993, 38, 153–157. 88, 93–102.
57. N. Huang, M. Xu, C. Yuan et al., J. Pho- 80. I. Nakamura, N. Negishi, S. Kutsuna et al.,
tochem. Photobiol. A: Chem 1997, 108, J. Mol. Catal. A: Chem. 2000, 162, 205–212.
229–233. 81. M. Anpo, Y. Ichihashi, M. Takeuchi et al.,
58. I. Sopyan, S. Murasawa, K. Hashimoto Res. Chem. Intermed. 1998, 24, 143–149.
et al., Chem. Lett. 1994, 723–726. 82. H. Q. Zhan, H. Tian, Dye Pigment. 1998, 37,
59. T. Noguchi, A. Fujishima, P. Sawunyama 231–239.
et al., Environ. Sci. Technol. 1998, 32, 83. S. G. D. Moraes, R. S. Freire, N. Duran,
3831–3833. Chemosphere 2000, 40, 369–373.
60. Y. Ohko, D. A. Tryk, K. Hashimoto et al., J. 84. K. Tanaka, K. Padermpole, T. Hisanaga,
Phys. Chem. B 1998, 102, 2699–2704. Wat. Res. 2000, 34, 327–333.
61. H. Matsubara, M. Takada, S. Koyama et al., 85. S. Malato, J. Blanco, C. Richter et al., Appl.
Chem. Lett. 1995, 767–768. Catal. B: Environ. 2000, 25, 31–38.
6.2 Applications of TiO2 Photocatalysis 561
highest resolution of the area-selective de- were also reported [13]. The produced
position would be twice as large as the images are easily erased by applying the
mean diffusion length of the photogener- reverse bias to the electrodes in the dark.
ated minority charge carriers. Area-selective photooxidative deposition of
Area-selective photoinduced reductive polypyrrole from a pyrrole solution onto
depositions of copper (Cu), nickel (Ni), TiO2 was used to produce a conducting
palladium (Pd), and gold (Au) on cathod- polymer pattern [14, 15]. The photoelectro-
ically biased p-Si and p-GaAs [9] from chemical oxidation of water-soluble aro-
aqueous solutions containing the corre- matic compounds such as O-toluidine on
sponding metal ions were shown to take TiO2 electrodes resulted in hydrophobic
place at a resolution of less than 10 µm. deposits on the irradiated surface only,
Area-selective reductive deposition of hep- which was developed by applying oily
tylviologen radical cation salt (HV+• Br− ) color ink [16]. Photo-patterning was also
from a heptylviologen bromide solution reported with self-assembled monolayer
onto p-GaAs electrode under a cathodic (SAM) of bis[11-[(4-azidobenzoyl)oxy]-1-
bias was useful in photoinduced elec- undecyl]disulfide [17]. Irradiation of the
trochromic displays [10]. Optical images SAM-covered Au electrode through a
due to the deposited HV+. Br− were easily photomask in the presence of several
erased by applying the reverse bias (i.e. amines such as HNEt2 , HN(n-Bu)2 or
anodic bias) to the electrode. The area- HN(CH2 CH2 OH)2 resulted in the attach-
selective deposition was also reported for ments of the amine, and the amine-
photooxidative deposition of PbO2 from attached regions showed little activity
a Pb2+ solution onto n-type ZnO, TiO2 , for electrochemical deposition of polyani-
and SrTiO3 [11], and of Tl2 O3 on these line from an acidic aniline solution,
semiconductor electrodes from a Tl+ so- giving selective deposition of polyani-
lution [12]. Using these photodeposition line on the nonirradiated SAM surfaces
reactions, the photoelectrochemical imag- on Au. The area-selectivity in the pho-
ing was successfully achieved at the an- toelectrochemical reactions finds another
odically biased semiconductor electrodes. application – the micromachining of ZnO
One example of the photoimages produced surfaces by anodic photoetching [18]. Mi-
is given in Fig. 1. The deposited Tl2 O3 do cropatterns having 130-µm line width
not grow beyond the irradiated part of were successfully produced using this
the electrodes because the photoinduced technique.
oxidation reaction took place on the irradi-
ated surfaces only. For example, the rate of 6.3.2.2 Deposition on Semiconductor
the deposition reaction of Tl2 O3 is propor- Surfaces with no External Bias
tional to the irradiation intensities at the When a photooxidative deposition occurs
electrode surfaces. Here, irradiation of the on semiconductor surfaces without any
n-type semiconductor electrodes through applications of external bias, a reduction
a negative film of photography resulted in reaction takes place simultaneously as the
the deposition of the brown oxide of vari- counterpart reaction (see Chapter 8). Ir-
ous contrasts in reflection to the images of radiation of p-Si wafers in a zinc (Zn)
the negative film, allowing the formation or cadmium (Cd) salt solution resulted
of a portrait. Attempts to form photoim- in area-selective deposition of metals on
ages on n-Si using the Tl2 O3 deposition the irradiated part if the back surface of
564 6 Nonsolar Energy Applications
the wafers was coated by metallic films anodic dissolution of the coated metal in
of Zn or Cd [19]. The photoinduced re- the back surface. The metal film on the
action in that case is thought to consist back surface was useful in enhancing the
of the photoinduced cathodic deposition band bending with which the rate of the
of the metal in the irradiated surface and transfer of photogenerated electrons to the
6.3 Silverless Photography – Optical Image Recordings by Means of Photoelectrochemical Processes 565
6.3.3.3 Combined Systems of film had very poor resolution. With an in-
Electrochromic Materials and crease in the irradiation time, the produced
Semiconductor Electrodes images easily spread beyond the irradiated
Attempts were made to form area-selective area because of high electrical conductivity
images on electrochromic thin films of polypyrrole.
coated on semiconductor electrodes. The Polyacrylamide gel films containing
first reports on this category were on Fe(III) dipyridinium complex as an elec-
Prussian blue film-coated TiO2 [42, 43] trochromic component were useful as
electrodes. Later, similar systems were the substrate for the formation of pho-
constructed using SrTiO3 films as the pho- tographic images when coated on a CdSe
tosensitive substrate [44, 45]. The coating film [48]. Poly-Re(CO)3 (vbpy)Cl-coated n-
with Prussian blue films on semiconductor MoSe2 also allowed photoinduced elec-
electrodes was carried out by electrochemi- trochromic recordings in the coated poly-
cal synthesis of the Prussian blue films us- mer layer under anodic bias [49]. The
ing the semiconductor as the electrode. To images produced are easily erased by
produce area-selective photoimages, the applying the reverse bias to the polymer-
entire Prussian blue-coated film was first coated semiconductor electrodes.
reduced to Prussian white. The irradiation
was then carried out with a xenon lamp 6.3.3.4 Composites Made of
through a photomask under anodic bias to Electrochromic Materials and
allow area-selective photooxidation of the Semiconductor Powders
coated film, resulting in blue images in the Photoimages of high resolution were suc-
white film. The same principle was applied cessfully produced in polyaniline films
to poly(N -methylpyrrole) film–coated n- containing dispersed WO3 [50, 52], and
Si [46] and polypyrrole film–coated Si TiO2 [51, 53, 54] powder particles in the
electrodes of both n-type and p-type con- films. The polyaniline/semiconductor par-
ductivity [47]. These conducting films were ticles composite films were synthesized
deposited on the Si electrode surfaces by by electrochemical polymerization of ani-
electrochemical polymerization of the cor- line in aqueous acidic solution in the
responding monomer using the Si as the presence of the suspended semiconductor
electrode. When n-type Si electrodes are particles. The prepared composite films
used, the entire polypyrrole-coated film showed the color of polyaniline even if a
is first reduced by cathodic polarization high amount of the oxide particles was
to give yellow, and then the polypyrrole- used because WO3 and TiO2 are color-
coated Si electrodes are anodically po- less. One example of a photoimage formed
larized and irradiated through a mask. in a polyaniline/TiO2 composite film is
The area-selective photooxidation of the presented in Fig. 2. As shown in the
polypyrrole films is then induced to take figure, a portrait of high resolution was
place to give brown-black images in the obtained. Complex electrochemical reac-
irradiated surfaces. In the case of using p- tions are involved in the formation of such
type Si electrodes, the as-deposited brown good photoimages. Figure 3 shows the
black polypyrrole films changed in color reaction scheme for the photoimage for-
to yellow at the irradiated part under ca- mation in the composite film, derived from
thodic bias. However, it was found that the analyses of changes in absorption spec-
images produced in the coated polypyrrole tra of polyaniline films obtained during
568 6 Nonsolar Energy Applications
A− A−
N N N+ N+ (B)
H H H H
N N N N (A)
H H
+H + +A−
e−
A− A−
hν N N N+ N+ (B)
H H H H
CH3OH
h+ +e− −A−
HCHO + 2H +
N N N N (C)
H H H H
Fig. 3 Schematic illustration of compositional changes in polyaniline caused by
photoreduction. (A) deprotonated blue film, (B) as-grown and/or green film, and
(C) reduced yellow film. A− denotes a monovalent electrolyte anion [51]. (With
permission by American Chemical Society).
6.3 Silverless Photography – Optical Image Recordings by Means of Photoelectrochemical Processes 569
the course of the image formation [53] the electrochemically inactive insulating
and from studies using a quartz crystal polyaniline is changed into electrochemi-
microbalance [54]. When the composite cally active one at the irradiated parts only.
films are prepared, they exhibited vivid As shown in Fig. 3 for the reaction scheme,
green, which is characteristic of highly if the blue composite film is irradiated,
conducting polyaniline having high elec- electrons and positive holes are generated
trochemical activities. Polymer films of in the TiO2 particles in the film. The
high electrical conductivity and of high generated positive holes cause oxidation
electrochemical reactivity do not allow the of methanol at the irradiated TiO2 sur-
formation of photoimage of high reso- faces, resulting in release of protons there.
lution, as already described earlier for The produced protons are instantly picked
polypyrrole-coated Si electrodes. To pro- up by the deprotonated polyaniline, and
duce photoimage of high resolution, the the electrochemically inactive blue polyani-
electrical conductivity of polyaniline of the
line is then converted to electrochemically
composites has to be changed to be as
active green polyaniline, which is then re-
low as possible. Fortunately, polyaniline
duced by the photogenerated electrons in
films of high electrical conductivity are
the TiO2 to form yellow images. Because
easily changed into electrically insulating
the production of protons and their at-
ones either by anodically polarizing in
tachments to the deprotonated polyaniline
alcohol solutions such as methanol and
ethanol or in aqueous neutral solutions
or by immersion in concentrated alkaline
solutions [55]. With the use of either one Writing
of these procedures, the deprotonation oc- ITO Nafion PB TiO2
curs at polyaniline in an oxidized state, and
the color changes from vivid green to blue. ca. −0.4 V
The polyaniline/TiO2 composite films hav- 0.25 V
ing an electrically insulating blue film pre-
pared in such a manner is then immersed
Light
in ethanol solution, and irradiated under
MeOH
no applied bias through a photomask with
lights whose energy is high enough to
photoexcite the incorporated TiO2 . Yellow
images, which are characteristic of the
reduced polyaniline, are then formed in
Erasing
the blue films with high resolution. An
important phenomenon inherent in the
photoimage formation-processes is that
occurs only in the irradiated area, image of CdS in the strong acidic medium
formations of high resolution take place at in which the composite is prepared,
the irradiated part only. results obtained were not so good as
The same idea was used in the image those obtained with the polyaniline/TiO2
formation in polyaniline/CdS composite composites. The composite films of
films [56], but because of the poor stability Prussian blue and TiO2 fine particles
(a)
(b)
1
e−
E(3Ru(bpy)32+/Ru(bpy)33+)
hν
0 e−
E(MB/l −MB)
E vs. SCE
E Oxascrobic acid
[V]
1 E(Ru(bpy)32+/Ru(bpy)33+)
hν
E(3MB/l −MB)
2
12. T. Inoue, A. Fujishima, K. Honda, J. Elec- 38. K. Ajito, L. A. Nagahara, D. A. Tryk et al.,
trochem. Soc. 1980, 127, 1582–1588. J. Phys. Chem. 1995, 99, 16383–16388.
13. R. J. Phillips, M. J. Shane, J. A. Switzer, 39. J. N. Yao, Y. A. Yang, B. H. Loo, J. Phys.
J. Mater. Res. 1989, 4, 923–929. Chem. B 1998, 102, 1856–1860.
14. M. Okano, K. Itoh, A. Fujishima et al., Chem. 40. J. N. Yao, B. H. Loo, K. Hashimoto et al.,
Lett. 1986, 469–472. Ber. Bunsen-Ges. Phys.Chem. 1991, 95,
15. M. Okano, K. Itoh, A. Fujishima et al., 554–556.
J. Electrochem. Soc. 1987, 134, 837–841. 41. J. N. Yao, B. H. Loo, K. Hashimoto et al.,
16. H. Masuda, N. Shimidzu, S. Ohno, Chem. Ber. Bunsen-Ges. Phys. Chem. 1992, 96,
Lett. 1984, 1701–1704. 699–701.
17. L. F. Rozsnyai, M. S. Wrighton, J. Am. Chem. 42. D. W. DeBerry, A. Viehbeck, J. Electrochem.
Soc. 1994, 116, 5993–5994. Soc. 1983, 130, 249–251.
18. M. Futsuhara, K. Yoshioka, Y. Ishida et al., 43. K. Itaya, I. Uchida, S. Toshima et al., J. Elec-
J.Electrochem. Soc. 1996, 143, 3743–3746. trochem. Soc. 1984, 131, 2086–2091.
19. T. L. Rose, D. H. Longendorfer, R. D. Rauh, 44. J. P. Ziegler, J. C. Hemminger, J. Elec-
Appl. Phys. Lett. 1983, 42, 193–195. trochem. Soc. 1988, 134, 358–363.
20. H. Yoneyama, H. Shiotani, N. Nishimura 45. J. P. Ziegler, E. K. Lesniewski, J. C.
et al., Chem. Lett. 1981, 157–160. Hemminger, J. Appl. Phys. 1987, 61,
21. T. Kobayashi, Y. Taniguchi, H. Yoneyama 3099–3104.
et al., J. Phys. Chem. 1983, 87, 768–775. 46. O. Inganas, I. Lundstrom, J. Electrochem. Soc.
22. H. Yoneyama, M. Kitayama, Chem. Lett. 1986, 1984, 131, 1129–1132.
657–660. 47. H. Yoneyama, K. Wakamoto, H. Tamura,
J. Electrochem. Soc. 1985, 132, 2414–2417.
23. H. Yoneyama, K. Kawai, S. Kuwabata,
48. A. A. Nekrasov, V. F. Ivano, A. V. Vannikov,
J. Electrochem. Soc. 1988, 135, 1699–1702.
J. Photogr. Sci. 1993, 41, 43–47.
24. S. Morishita, K. Suzuki, Bull. Chem. Soc. Jpn.
49. H.-T. Zhang, P. Subaramanian, O. Fussary-
1994, 67, 843–846.
del et al., Sol. Energy Mater. 1992, 25,
25. R. J. Colton, A. M. Guzman, J. W. Rabalais,
315–325.
Acc. Chem. Res. 1978, 11, 170–176.
50. H. Yoneyama, S. Hirao, S. Kuwabata, J. Elec-
26. C. M. Lampert, Sol. Energy Mater. 1984, 11,
trochem. Soc. 1992, 139, 3141–3146.
1–27.
51. H. Yoneyama, N. Takahashi, S. Kuwabata,
27. K. Hyodo, Electrochim. Acta 1994, 39, J. Chem. Soc. Chem. Commun. 1992,
265–272. 716–717.
28. R. J. Mortimer, Electrochim. Acta 1999, 44, 52. S. Kuwabata, S. Hirao, H. Yoneyama, Denki
2971–2981. Kagaku oyobi kogyo Butsuri Kagaku 1992, 60,
29. M. Fujii, T. Kawai, H. Nakamatsu et al., 1097–1100.
J. Chem. Soc. Chem. Commun. 1983, 53. S. Kuwabata, N. Takahashi, S. Hirao et al.,
1428–1429. Chem. Mater. 1993, 5, 437–441.
30. B. Ohtani, T. Atsumi, S. Nishimoto et al., 54. S. Kuwabata, A. Kishimoto, H. Yoneyama,
Chem. Lett. 1988, 295–298. J. Electroanal. Chem. 1994, 377, 261–268.
31. E. Kikuchi, K. Iida, A. Fujishima, J. Elec- 55. S. Kuwabata, N. Kihira, H. Yoneyama, Chem.
troanal. Chem. 1993, 351, 105–114. Mater. 1993, 5, 716–719.
32. B. H. Loo, J. N. Yao, H. D. Coble et al., Appl. 56. H. Yoneyama, M. Tokuda, S. Kuwabata, Elec-
Surf. Sci. 1994, 81, 175–181. trochim. Acta 1994, 39, 1315–1320.
33. E. Kikuchi, N. Hirota, A. Fujishima et al., 57. M. Nishizawa, S. Kuwabata, H. Yone-
J. Electroanal. Chem. 1995, 381, 15–19. yama, J. Electrochem. Soc. 1996, 143,
34. Y. A. Yang, Y. W. Cao, P. Chen et al., J. Phys. 3462–3465.
Chem. Solids 1998, 59, 1667–1670. 58. S. Kuwabata, K. Mitsui, H. Yoneyama,
35. J. N. Yao, B. H. Loo, K. Hashimoto et al., J. Electrochem. Soc. 1992, 139, 1824–1830.
J. Electroanal. Chem. 1990, 290, 263–267. 59. S. Kuwabata, K. Mitsui, H. Yoneyama,
36. J. N. Yao, B. H. Loo, A. Fujishima, Ber. J. Electroanal. Chem. 1990, 281, 97–107.
Bunsen-Ges. Phys. Chem. 1990, 94, 13–17. 60. N. Kobayashi, T. Yano, K. Teshima et al.,
37. J. N. Yao, B. H. Loo, A. Fujishima, Nature Electrochim. Acta 1998, 43, 1645–1649.
1992, 355, 624–626.
6.4 Photoelectrochemical Etching 573
(a)
(b)
(a)
1 µm
(b)
1 µm
(c)
1 µm
Fig. 2 SEM photographs of single crystal CdSe surfaces with
different crystallographic plane exposed after photoetching.
576 6 Nonsolar Energy Applications
(a)
1 µm
(b)
1 µm
(a) 1.5
[µm]
0.5
0
0 0.5 1 1.5
[µm]
(b) 69.02
[nm]
0.0
0 1282.701
[nm]
Fig. 7AFM image and it’s cross-sectional analysis of TiO2
specimen collected from the photoetched surface.
photoetched at +0.1 V versus SCE is is shown in Fig. 11. A grain boundary has
shown in Fig. 10. Interestingly, the grains a uniform thickness of about several tens
are selectively dissolved, whereas the grain of nanometers. Detailed analyses of grain
boundaries were left undissolved. The boundary is described as follows. Figure 12
height of the grain boundary wall is several shows its high-resolution TEM photograph
microns and the thickness is 10 ∼ 20 nm. and corresponding SAD pattern. The in-
Such a potential dependence of the pho- cident direction of the electron beam is
toetching pattern can be explained in the parallel to the (110) zone axis. The well-
same manner shown in Fig. 4. regulated lattice fringe with a distance of
0.32 nm, which corresponds to the inter-
6.4.3 planar spacing of (110) plane (0.3247 nm).
Analyses of Grain Boundaries This result shows that the grain boundary
is not amorphous but crystalline.
Now that one can separate and isolate A TEM photograph and its correspond-
grain boundaries from a photoetched poly- ing SAD pattern of another part of the
crystalline TiO2 surface, these samples same specimen are shown in Fig. 13. The
are subjected to closer characterization. A moiré pattern of the 0.7 nm width is seen
TEM photograph of a grain boundary taken in the TEM image (a). Such a moiré pattern
out from the surface of a photoetched TiO2 is produced by the interference of crystal
6.4 Photoelectrochemical Etching 581
100 nm
(b)
100 nm
1 µm
1 µm
582 6 Nonsolar Energy Applications
1 µm
3.247 Å
10 nm
lattices of two twisted thin crystals. In the diffraction (spot A) arises from the crys-
SAD pattern (b) an unexpected diffraction tal (a) and is superimposed with a twist
spot, C, emerges in addition to two sets of 18◦ . When an incident electron beam
of (111) lattice reflections (A & B) and is reflected two times by each (111) lat-
(110) lattice reflections, D, of a rutile struc- tice planes of crystals (a) and (b), it causes
ture. It was interpreted that the spot C double diffraction of the electron beam to
originates from double electron diffrac- create spot C. Overlapping of two (111)
tion at the grain boundary, as illustrated in lattice planes twisted with a small angle
Fig. 14. Considering that the grain bound- produces a moiré fringe. Its spacing is
ary has a sandwich structure that consists calculated by Eq. (2).
of two thin crystals (a, b) of TiO2, hav-
ing different crystal orientations, is quite d1 d2
dM = (2)
reasonable. One of the (111) lattice diffrac- (d12 + d22 − 2d1 d2 cos α)1/2
tion (spot B in Fig. 13b) and (110) lattice
diffraction (spot D in Fig. 13b) are assigned where dM is the spacing of moiré fringe,
to the diffraction spots, arising from the d1 , d2 refers to the spacing of each lattice,
crystal (b) with the (110) zone axis parallel and α is the angle between the two lattices.
to the electron beam. Another (111) lattice By substituting 0.2188 nm (lattice spacing
6.4 Photoelectrochemical Etching 583
500 nm
(b)
a A
(111)
B (111)
c b
C
D (110)
Grain
boundary
a
b
C′
A′ B′ A
C B
of (111) plane) for d1 and d2 , and 18◦ perpendicular to the vector c of the electron
(the angle between two electron diffraction diffraction spot C, which is the sum of the
spots, A and B) for α, dM is calculated vectors a and b of diffraction spots of two
to be 0.7 nm, which is consistent with (111) planes (A and B).
the spacing of moiré fringe in Fig. 13(a). On the basis of the above information
The direction of the moiré fringe is the structure of the polycrystalline TiO2
584 6 Nonsolar Energy Applications
Grain
Grain
boundary
Grain
6. T. Saitoh, S. Matsubara, S. Minagawa, 33. J. L. Weyher, Int. Phys. Conf. Ser. 1995, 146,
J. Electrochem. Soc. 1975, 122, 670–674. 399–408.
7. M. M. Faktor, J. L. Stevenson, J. Electrochem. 34. D. Rotter, J. Uffmann, J. Ackermann et al.,
Soc. 1978, 125, 621–629. Mater. Res. Soc. Symp. Proc. 1998, 482,
8. E. Galun, G. Hodes, M. Peisach et al., 1003–1008.
J. Cryst. Growth 1992, 117, 666–671. 35. C. Youtsey, G. Bulman, I. Adesida, Mater.
9. R. Khare, D. B. Young, E. L. Hu, J. Elec- Res. Soc. Symp. Proc. 1997, 468, 349–354.
trochem. Soc. 1993, 140, L117–L118. 36. C. Youtsey, I. Adesida, G. Bulman, Electron.
10. H. H. Streckert, J. Tong, A. B. Ellis, J. Am. Lett. 1997, 33, 245–246.
Chem. Soc. 1982, 104, 581–588. 37. D. Soltz, M. A. D. Paoli, L. Cescato, J. Vac.
11. R. Tenne, Appl. Phys. Lett. 1983, 43, 201–203. Sci. Technol. B 1996, 14, 1784–1790.
12. G. Hodes, D. Cahen, H. J. Leamy, J. Appl. 38. V. A. Tyagai, V. A. Sterligov, Ya. Kolbasov,
Phys. 1983, 54, 4676–4678. Electrochim. Acta 1977, 22, 819.
13. M. A. Russak, C. Creter, J. Electrochem. Soc. 39. D. V. Podlesnik, H. H. Gilgen, R. M. Osgood
1985, 132, 1741–1745. et al., Appl. Phys. Lett. 1983, 43, 1083–1085.
14. G. Hodes, J. Manassen, D. Cahen, J. Elec- 40. R. M. Lum, A. M. Glass, F. W. Ostermayer
trochem. Soc. 1981, 128, 2325–2330. et al., J. Appl. Phys. 1985, 57, 39–44.
15. R. Tenne, G. Hodes, Appl. Phys. Lett. 1980, 41. R. Matz, J. Zirrgiebel, J. Appl. Phys. 1988, 64,
37, 428–430. 3402–3406.
16. N. Mueller, R. Tenne, Appl. Phys. Lett. 1981, 42. R. M. Lum, F. W. Ostermayer, P. A. Kohl
39, 283–285. et al., Appl. Phys. Lett. 1985, 47, 269–271.
17. G. Dagan, S. Endo, G. Hodes et al., Sol. 43. F. Decker, D. A. Soltz, L. Cescato, Electro-
Energy. Mater. 1984, 11, 57–74. chim. Acta 1993, 38, 95–99.
18. G. Dagan, D. Cahen, J. Electrochem. Soc. 44. R. David Rauh, in Electrochemistry of
1987, 134, 592–600. Semiconductor and Electronics (Eds.: J. Mc-
19. R. Tenne, B. Theys, J. Rioux et al., J. Appl. Hardy, F. Ludwig), Park Ridge, N.J., 1992,
Phys. 1985, 57, 141–145. pp. 177–216.
20. R. Tenne, A. Wold, Appl. Phys. Lett. 1985, 47, 45. H. Matsumoto, T. Sakata, H. Mori et al.,
707–709. Chem. Lett. 1995, 595–596.
21. A. M. Chaparro, P. Salvador, A. Mir, J. Elec- 46. H. Matsumoto, T. Sakata, H. Mori et al.,
troanal. Chem. 1997, 422, 35–44. J. Phys.Chem. 1996, 100, 13781.
22. V. D. Das, L. Damodare, J. Appl. Phys. 1997, 47. T. Torimoto, H. Nishiyama, T. Sakata et al.,
81, 1522–1530. J. Electrochem. Soc. 1998, 145, 1964–1968.
23. H. Minoura, Y. Ueno, H. Kaigawa et al., 48. M. Miyake, T. Torimoto, T. Sakata et al.,
J. Electrochem. Soc. 1989, 136, 1392–1395. Langmuir 1999, 15, 1503–1507.
24. L. Hollan, J. C. Tranchart, R. Memming, 49. T. Torimoto, H. Kontani, T. Sakata et al.,
J. Electrochem. Soc. 1979, 126, 855–859. Chem. Lett. 1999, 379–380.
25. A. Yamamoto, S. Yano, J. Electrochem. Soc. 50. A. van Dijken, D. Vanmaekelbergh,
1975, 122, 260–267. A. Meijerink, Chem. Phys. Lett. 1997, 269,
26. A. Yamamoto, S. Tohno, C. Umemura, 494–499.
J. Electrochem. Soc. 1981, 128, 1095–1100. 51. A. van Dijken, A. H. Janssen, M. H. P.
27. T. Saito, S. Matsubara, S. Minagawa, J. Elec- Smitsmans et al., Chem. Mater. 1998, 10,
trochem. Soc. 1975, 122, 670–674. 3513–3522.
28. F. Kuhn-Kuhnenfeld, J. Electrochem. Soc. 52. T. Sugiura, H. Minoura, Y. Ueno, Denki Ka-
1972, 119, 1063–1068. gaku 1989, 57, 1133–1136.
29. K. Takahashi, Jap. J. Appl. Phys. 1979, 18, 53. T. Sugiura, H. Tomimatsu, H. Minoura
1741. et al., Denki Kagaku 1993, 61, 661–665.
30. P. A. Kohl, IBM J. Res. Dev. 1998, 42, 54. T. Sugiura, R. Tenpaku, H. Minoura et al.,
629–637. Electrochim. Acta 1993, 38, 593–569.
31. E. A. Miller, G. L. Richmond, J. Phys. Chem. 55. T. Sugiura, M. Hida, H. Minoura et al., Appl.
B 1997, 101, 2669–2677. Phys. Lett. 1990, 56, 1954–1956.
32. S. Mueller, J. L. Weyher, K. Koehler et al., 56. T. Sugiura, M. Hida, M. Ohnishi et al., Elec-
Int. Phys. Conf. Ser. 1996, 145, 279–284. trochim. Acta 1992, 37, 1429–1432.
586 6 Nonsolar Energy Applications
57. H. Iwanaga, T. Yoshiie, T. Yamaguchi et al., 63. T. Sugiura, T. Yoshida, H. Minoura, Elec-
J. Crystal Growth 1979, 47, 703–711. trochem. and Solid-state Lett. 1998, 1, 175–177.
58. T. Yoshiie, H. Iwanaga, T. Yamaguchi et al., 64. T. Sugiura, S. Itoh, T. Ooi et al., J. Elec-
J. Crystal Growth 1981, 51, 624–626. troanal. Chem. 1999, 473, 204–208.
59. L. M. Levinson, S. Hirano, Eds., Grain 65. T. Sugiura, T. Yoshida, H. Minoura, Electro-
Boundaries and Interfacial Phenomena in chemistry 1999, 67, 1234–1236.
Electro Ceramics, in Ceramic Transactions, 66. L. A. Harris, R. H. Wilson, J. Electrochem.
The American Ceramic Soc., Westerville, Soc. 1976, 123, 1010–1015.
Ohio, 1993, Vol. 41. 67. L. A. Harris, D. R. Cross, M. E. Gerstner,
60. T. R. Gupta, J. Am. Ceram. Soc. 1990, 73, J. Electrochem. Soc. 1977, 124, 839–844.
1817–1840. 68. M. E. Gerstner, J. Electrochem. Soc. 1979, 126,
61. J. A. S. Ikeda, Y.-M. Chiang, J. Am. Ceram. 944–949.
Soc. 1993, 76, 2437–2446. 69. Y. Nakato, H. Akanuma, J. Shimizu et al.,
62. J. A. S. Ikeda, Y.-M. Chiang, A. J. Garratt- J. Electroanal. Chem. 1995, 396, 35–39.
Reed et al., J. Am. Ceram. Soc. 1993, 76, 70. Y. Nakato, H. Akanuma, Y. Magari et al.,
2447–2459. J. Phys. Chem. B 1997, 101, 4934.