Sunteți pe pagina 1din 34

Characterization of Materials, edited by Elton N. Kaufmann.

Copyright 2012 John Wiley & Sons, Inc.


SEMICONDUCTOR
PHOTOELECTROCHEMISTRY
SAMIR J. ANZ,
1
ARNEL M. FAJARDO,
1
WILLIAM J. ROYEA,
1
NATHAN S. LEWIS,
1
AND AMANDA J. MORRIS
2
1
California Institute of Technology, Pasadena, CA, USA
2
Virginia Polytechnic Institute and State University, USA
INTRODUCTION
This article discusses methods and experimental
protocols in semiconductor electrochemistry. We rst
introduce the basic principles that govern the energetics
and kinetics of charge ow at a semiconductorliquid
contact. The principal electrochemical techniques of
photocurrent and photovoltage measurements used to
obtain important interfacial energetic and kinetic
quantities of such contacts are then described in detail.
After this basic description of concepts and methods
in semiconductor electrochemistry, we describe meth-
ods for characterizing the optical, electrical, and chem-
ical properties of semiconductors through use of the
electrochemical properties of semiconductorliquid
interfaces.
In some cases, the semiconductorliquid junction
provides a convenient method for measuring properties
of the bulksemiconductor that canonlybe accessedwith
great difculty throughother techniques; inother cases,
the semiconductorliquid contact enables measure-
ment of properties that cannot be determined using
other methods. Due to the extensive amount of back-
groundmaterial andthe interdisciplinary nature of work
in this eld, the discussion is not intended to be exhaus-
tive, and the references cited in the various protocols
should be consulted for further information.
This article will cover the following methods in
semiconductor electrochemistry:
.
Photoconductor/photovoltage measurements at
semiconductorliquid interfaces
T Measurement of semiconductor band gaps
T Diffusion length determination
T Determination of kinetic properties of
semiconductorliquid interfaces.
. Differential capacitance measurements of
semiconductorliquid contacts
T Electrochemical photocapacitance spectroscopy
(EPS)
T Flat-band potential measurements of
semiconductorliquid interface
.
Transient growth and decay dynamics of
semiconductorliquid contacts
T Time-resolved photoluminescence spectroscopy
.
Measurement of surface recombination velocity
using time-resolved microwave conductivity
.
Laser spot scanning methods at semiconductor
liquid contacts
. Intensity modulated photocurrent and photovol-
tage spectroscopy
. Laser-induced photovoltage transients
An important consideration in application of semicon-
ductor electrochemistry is the size and shape of the
semiconductor employed. It should be noted that the
majority of the techniques described herein are dis-
cussed from the perspective of planar single-crystal
semiconductor electrodes. Some of the techniques can
also be applied to nanocrystalline semiconductors, but
MottSchottky analysis as discussed in Differential
Capacitance Measurements of SemiconductorLiquid
Contacts, for example, cannot be utilized for these
materials. For a more detailed discussion of the electro-
chemistry of nanomaterials, the readers are referred to
other sources (Bisquert, 2008; Hodes, 2001).
PHOTOCURRENT/PHOTOVOLTAGE MEASUREMENTS
Currentvoltage measurements can provide both ther-
modynamic and kinetic information about semiconduc-
torliquid junctions. Discussed below are methods to
measure the JE characteristics, V
OC
, J
SC
, k
et
, E
gap
, L,
and j at these interfaces, where V
OC
is the open-circuit
voltage, J
SC
isthe short-circuit current, k
et
isthe electron
transfer rate constant, E
gap
is the characteristic band
gap, Listhe minoritycarrier diffusionlength, andjisthe
quantum yield for charge carrier collection.
Principles of the Method
Thermodynamics of SemiconductorLiquid Junctions.
When a semiconductor is placed in contact with a liquid,
interfacial charge transfer occurs until electronic
equilibrium has been reached. The direction and
magnitude of this charge ow are dictated by the
relevant energetic properties of the semiconductorliquid
contact. As showninFigure 1, the key energetic quantities
of the semiconductor are the energy of the bottom of the
conduction band E
cb
, the energy of the top of the valence
band E
vb
, the Fermi level E
F
, and the band gap energy E
g
(=E
vb
E
cb
).
The key energetic quantity of the liquidis its electrode
potential, E(A/A

), dened by the redox couple formed


froman electroactive electron acceptor species A and an
electroactive donor species A

, present in the electrolyte


phase. The electrode potential of the phase containing
specic concentrations of A and A

is related to the
formal electrode potential E
0
/
(A/A

) of this redox system


by the Nernst equation:
E(A=A

) = E
0
/
(A=A

)
kT
n
ln
[A[
[A

[
(1)
where n is the number of electrons transferred, k is the
Boltzmanns constant, T is the absolute temperature, [A]
is the concentration of acceptor species, and [A

] is the
concentration of donor species. It is important to note
that the potentials discussed here are electrode poten-
tials and not electrochemical potentials. Electrochemi-
cal potentials, the partial Gibbs energy of charged
species, are related to the electrode potential by the
electrochemical potential of electrons at equilibrium
through the Galvani potential difference (Bard and
Faulkner, 1980).
The doping level of the semiconductor is also impor-
tant in determining the degree of interfacial charge
transfer, because the dopant density determines the
position of E
F
in the semiconductor before contact with
the electrolyte. For an n-type semiconductor and non-
degenerate doping, the dopant density N
d
is given by the
Boltzmann-type relationship:
E
F
= E
CB
kTe
N
d
=Nc
(2)
where N
c
is the effective density of states in the conduc-
tion band of the semiconductor.
Equilibrium is established when the Fermi level is
equal everywhere in both the semiconductor and solu-
tion phases. Interfacial charge will pass to equilibrate E
F
to E(A/A

). This charge ow will produce a spatially


nonuniform charge density in the semiconductor and
inthe liquid. This nonzero charge density inbothphases
will, in turn, produce an electric eld and an electric
potential in the vicinity of the semiconductorliquid
contact. The potential drop across the semiconductor
at electronic equilibriumis often referred to as the built-
in voltage V
bi
.
Variations inredoxpotentials alsogive risetochanges
in an experimental parameter known as the barrier
height. The barrier height f
b
for an n-type semiconduc-
torliquid contact is the potential difference between the
redox potential of the solution and the conduction band
edge. For a given difference between E
cb
and E(A/A

) at
an n-type semiconductorliquid contact, solution of
Poissons equation leads to the following well-known
expressions for the charge density Qand the magnitude
of the electric eldEandelectric potential Vas afunction
of distance x into the semiconductor:
Q = qN
d
; 0 _ x _ W (3a)
Q = 0; x > W (3b)
E(x) =
qN
d
e
S
(Wx); 0 _ x _ W (4a)
E(x) = 0; x > W (4b)
V(x) =
qN
d
2e
S
(Wx)
2
; 0 _ x _ W (5)
W =

2e
S
[E
cb
qV
n
E(A=A

)[
N
d
q
2

(6)
Here, q is the absolute value of the charge onanelectron,
e
S
is the static dielectric constant of the semiconductor,
W is the depletion width, and V
n
is the potential differ-
ence between the Fermi level and the conduction band
level in the bulk of the semiconductor (Fig. 1).
Equation 3a is reasonable because the dopants are
present in relatively low concentration in the solid
(perhaps 1ppmor less), so essentially all the dopants are
ionized until a sufcient distance has been reached and
that the required electrical charge has been transferred
acrossthesolidliquidinterface. Equations4aand5then
follow directly from the charge density prole of
Equation 3a, once the value of W is known from the
amount of charge transferred (Equation 6). Analogous
equations for p-type semiconductorliquid contacts can
beobtainedtorelatetheenergeticsof thevalencebandE
vb
to E(A/A

). The electric eld near the surface, computed


from the limit of Equation 4a as x approaches zero, is
10
5
V/cmforadifferencebetweenE(A/A

) andE
CB
of1eV.
This junction formation leads to diode-like behavior,
in which charge carriers experience a large barrier to
current ow in one direction across the interface but
display an exponentially increasing current density as a
voltage is applied to the system in order to produce a
current owinthe opposite directionacross the contact.
Understandingthe microscopic originof this rectication
k
et
Ecb
Evb
Solid
(
+
)
(

)
Liquid
E
n
e
r
g
y

(
e
V
)

v
e
r
s
u
s

r
e
f
e
r
e
n
c
e
qV
bi
qV
n
n
b
n
S
E
F E(A/A

)
Figure 1. Energy of an n-type semiconductorliquid junction
under equilibriumconditions. At equilibrium, the Fermi level of
the semiconductor E
F
is equal to the electrode potential of the
solution. The surface electron concentration n
s
is proportional
to the bulk concentration of electrons n
b
and the equilibrium
built-in voltage V
bi
. The energy difference between the Fermi
level and the conduction band inthe bulk is constant and equal
to qV
n
. The rate of charge transfer from the semiconductor to
solution species is governed by the interfacial electron transfer
rate constant k
et
. Note the standard electrochemical sign con-
vention, with positive energies representing more tightly bound
electrons, is used throughout this article.
2 SEMICONDUCTOR PHOTOELECTROCHEMISTRY
andinterpretingitspropertiesintermsof thechemistryof
the semiconductorliquid contact of concern is the topic
covered next.
Charge Transfer at Equilibrium. Interfacial charge
transfer at a semiconductorliquid interface can be
represented by the following balanced chemical
equation:
Electron in solidacceptor in solution
electron vacancy in soliddonor in solution
(7)
When Equation 7 is obeyed, the current should depend
linearly on the concentration of electrons near the
semiconductor surface and on the concentration of
acceptor ions that are available to capture charges at
the semiconductor surface (Fig. 1). If this relationship
holds, the reaction is rst order in both the surface
electron concentration and available acceptor ions. In
a solution containing a redox couple A/A

, the rate of
direct electron transfer froman n-type semiconductor to
the acceptor species A can, therefore, be expressed as
Rate of electron injection into solution = k
et
n
s
A [ [
s
(8)
where k
et
is the rate constant for the electrontransfer, n
S
is the surface concentration of electrons, and [A]
S
is the
concentration of acceptors in the interfacial region near
the semiconductorliquid contact. The units of k
et
are
centimeters to the fourth power per second, because the
rate of charge ow represents a ux of charges crossing
the interface, with units of reciprocal centimeters
squared per second, and the concentrations n
S
and
[A]
S
are expressed in units of reciprocal cubic
centimeters.
The electron ux in the opposite direction, that is,
from the electron donors in the solution to the empty
states in the conduction band of the semiconductor, can
be described as
Rate of electron transfer from solution = k
/
et
A

[ [
s
(9)
In Equation 9, k
/
et
is the reverse reaction rate constant,
withunits of s
1
, and[A

]
S
is the concentrationof donors
in the interfacial region near the semiconductorliquid
contact. In this expression, the concentration of empty
states in the conduction band of the semiconductor has
been incorporated implicitly into the value of k
/
et
.
At equilibrium, the rates of Equations 8 and9 must be
equal. Denotingtheequilibriumelectronconcentrationat
the semiconductor surface by the quantity n
S0
, we obtain
k
et
n
S0
[A[
S
= k
/
et
[A

[
S
(10)
Away from equilibrium, the net rate of electron transfer
into solution (dn/dt) is simply the forward rate minus
the reverse rate. From Equations 810, we then obtain
the general relationship:

dn
dt
= k
et
[A[
S
n
S
k
/
et
[A

[
S
(11)
or

dn
dt
= k
et
[A[
S
n
S
n
S0
( ) (12)
Although a rst-order dependence of J on [A]
s
is concep-
tually simple, it has rarely been observed experimen-
tally. Including surface state capture as an intermediate
pathway for electron transfer fromthe semiconductor to
the acceptor in solution (Fig. 2) produces the following
rate law:
Rate = k
et
[A[
s
n
s
k
sol
[A[
s
N
t;s
f
t;s
(13)
where k
sol
is the electron transfer rate constant from
a surface trap to a solution acceptor. At steady state,
df
t,s
/dt =0, produces the following rate expression:
Rate = k
et
[A[
s
n
s
k
sol
A [ [N
t;s
k
n;s
n
s
k
n;s
n
s
k
sol
[A[
s
_ _
(14)
where k
n,s
, the capture coefcient for electrons by
traps, is the rate constant for transfer of an electron
from the conduction band into a surface state. If an
E
vb
A
e

(
+
)
(

)
E
n
e
r
g
y

(
e
V
)

v
e
r
s
u
s

r
e
f
e
r
e
n
c
e
Solid Liquid
A
k
et
E
cb
k
n,s
N
t,s
k
sol
Figure 2. Two possible mechanisms for interfacial electron
transfer. The upper pathway shows the direct transfer of elec-
trons from the conduction band edge of the semiconductor to
the acceptor in solution, with the rate constant k
et
. The lower
pathway depicts the surface state mediation of interfacial elec-
trontransfer. Here, N
t,s
is the densityof surfacestates, k
n,s
is the
rate constant for electron capture from the conduction band
into surface states, and k
sol
is the rate constant for electron
injection from the surface states into solution.
SEMICONDUCTOR PHOTOELECTROCHEMISTRY 3
electron slowly lls a surface trap and proceeds quickly
to the solution such that k
sol
A [ [
s
k
n;s
n
s
and k
n,s
is
interpreted as a collisional event, then Equation 14
becomes
Rate = k
et
[A[
s
n
s
(sv
n
)N
t;s
n
s
(15)
where k
n,s
=n
n
s.
The thermal velocity of electrons is well knownfor n-Si
to be 10
7
cm/s. Equation 15 indicates that in the
scenario of slow electron trapping by surface states and
fast electron ejection into solution, a large N
t,s
will
eliminate the dependence of J on [A]
s
. In the converse
case of fast electron trapping and a rate-determining
step of electron ejection into solution such that
k
n;s
n
s
k
sol
A [ [
s
, Equation 14 reduces to
Rate = k
et
[A[
s
n
s
k
sol
[A[
s
N
t;s
(16)
For the rate in Equation 16, a large N
t,s
will eliminate
the dependence of J on n
s
. Regardless of which surface
state mediation step dictates the rate, a signicant
density of electrical traps will at best hamper and
at worst prevent steady-state measurements of k
et
.
Gerischer (1991) has suggested that redox species
couple more strongly to surface states than to conduc-
tion band states, and hence, even a small N
t,s
would
generally overwhelm the steady-state, electron trans-
fer current ow under practical experimental
conditions.
Dark CurrentPotential Characteristics of Semiconductor
Liquid Junctions. The current density (i.e., the current I
divided by the exposed area of the electrode A
S
) is merely
the electron transfer rate multiplied by the charge on an
electron; therefore, the interfacial electron transfer
current density can be written as
J = q
dn
dt
_ _
= C n
S
n
S0
( ) (17)
where the constant C=qk
et
[A]
S
. The current density J is
dened to be negative when a reduction occurs at the
electrode surface. Therefore, when n
S
>n
S0
, a negative
(reduction) current will ow, because the electrode will
tend to donate electrons to the solution.
A useful form of this equation is
J = Cn
S0
n
S
n
S0
1
_ _
(18)
To obtain explicitly the dependence of the current den-
sity on the potential applied across the solidliquid
interface, we must relate the electron concentration at
the surface of a semiconductor to the electron concen-
trationinthe bulk. The surface electronconcentrationat
equilibrium is given by
n
S0
= n
b
exp
qV
bi
kT
_ _
(19)
where n
b
is the concentration of electrons in the bulk of
the semiconductor and V
bi
is the built-in voltage, that is,
the potential dropped across the semiconductor at equi-
librium. Equation 19 can also be cast in terms of the
barrier height f
b
. The value of qf
b
reects the free energy
associated with interfacial electron transfer. Using this
parameter, the expression for n
S0
can be rewritten as
n
S0
= N
c
exp
qf
b
kT
_ _
(20)
The value of N
c
is known for most semiconductors and is
generally ~10
19
cm
3
(Sze, 1981).
When a potential E is applied to the semiconductor
relative to the situation at equilibrium, the total voltage
drop inthe semiconductor depletionregionis V
bi
E, so
we obtain an analogous Boltzmann relationship away
from equilibrium:
n
S
= n
b
exp
q V
bi
E ( )
kT
_ _
(21)
These equations represent the physical situation that
the electron concentration at the semiconductor surface
can be either increased or decreased through the use of
anadditional voltage. This applied potential controls the
surface carrier concentration in the same fashion as the
built-in voltage, so the same Boltzmann relationship
applies.
Equations 19 and 21 lead to a simple expression for
the variation in the surface electron concentration as a
function of the applied potential:
n
S
n
S0
= exp
qE
kT
_ _
(22)
This makes sense, because any change in the voltage
dropped across the solid should exponentially
change the electron concentration at the semiconductor
surface relative to its value at equilibrium.
Substituting Equation22into Equation18, we obtain
the desiredrelationshipbetweenthe current density and
the potential of a semiconductorliquid junction:
J = Cn
S0
exp
qE
kT
_ _
1
_ _
(23)
This equation is the simple rate equation (Equation 18),
which has been rewritten to emphasize the explicit
dependence of the current density on E.
Equation 23 is often written with only one constant:
J = J
0
exp
qE
kT
_ _
1
_ _
(24)
where J
0
=Cn
S0
. The parameter J
0
is calledthe exchange
current density, because it is the value of the current
density that is present at equilibrium. For convenience,
J
0
is dened as a positive quantity. The parameter J
0
is
clearly dependent onthe value of the equilibriumsurface
4 SEMICONDUCTOR PHOTOELECTROCHEMISTRY
electronconcentration, because asmaller exchange cur-
rent should ow at equilibrium if there are fewer elec-
trons available to exchange with a particular solution.
We can now explicitly incorporate the barrier height
into the current density expression by substituting
Equation 20 into the expression for J
0
:
J
0
= CN
c
exp
qf
b
kT
_ _
(25)
Equation 25 indicates that a value of J
0
can be predicted
if the values of both f
b
and C are known.
The current densitypotential (or JE) characteristic
described by Equations 22 and 23, where the current
can ow predominantly in only one direction under an
applied potential, is called rectication. The rectica-
tion characteristic is typical of electrical diodes. Equa-
tions that have the form of Equations 22 and 23 are
therefore generally called diode equations.
The dependence of the charge transfer rate on the
solution redox potential is perhaps the most important
experimental property of semiconductor electrodes.
Regardless of the value of the redox potential of the solu-
tion, E(A/A

), the diode behavior of Equation 23 will be


obeyed. Changes inE(A/A

), however, will produce differ-


entvaluesofJ
0
, becauseJ
0
dependsonn
S0
. Thesedifferent
exchange currents will produce a measurable change in
the JE behavior of the semiconductorliquid contact.
For an n-type semiconductor, more positive redox
potentials will yield smaller values of J
0
and will produce
highly rectifying diode behavior. For p-type semiconduc-
tors, the opposite behavior is expected, so that negative
redox potentials should produce highly rectifying con-
tacts while positive redox potentials should produce
poorly rectifying contacts. Rectifying JE behavior is
required for efcient photoelectrochemical devices that
use either n- or p-type semiconductors; thus, one goal in
constructing semiconductorliquid junctions is to
ensure that chemical control is maintained over the
JE properties of the semiconductorliquid contact.
Although we have derived the diode behavior of a
semiconductorliquid junction by assuming that elec-
trontransfer is the important charge owprocess across
the interface, the diode equation is generally applicable
to semiconductorliquid devices even when other pro-
cesses are rate limiting. The JE relationships for other
possible charge ow mechanisms, such as recombina-
tion of carriers at the surface and/or in the bulk of the
semiconductor, almost all adopt the formof Equation19
(Fonash, 1981). The major difference between the vari-
ous mechanisms is the value of J
0
for each system.
Mechanistic studies of semiconductorliquid junctions
therefore generally reduce to investigations of the factors
that control J
0
. Such studies also involve quantitative
comparisons of the magnitude of J
0
with the value
expected for a specic charge transport mechanism.
These types of investigations have yieldedadetailedlevel
of understanding of many semiconductorliquid inter-
faces. Recent reviews describing more details of this
work have been written by Koval and Howard (1992),
Lewis (1990), and Tan et al. (1994b).
Current Potential Characteristics of Illuminated
SemiconductorLiquid Junctions. Illumination of a semi-
conductor with light above its band gap energy
produces excess electronhole pairs, and movement
of these charge carriers produces a photocurrent
and a photovoltage at the semiconductorliquid
contact. The effects of illumination are relatively
simple to incorporate into the JE behavior of a
semiconductorliquid contact. The total current in
such a system can be conceptually partitioned into
two components: one that originates from majority
carriers and one from minority carriers. Absorption
of photons creates both majority carriers and minority
carriers; therefore, increasesinbothcurrent components
are expected under illumination.
The concentration of majority carriers generated by
absorption of moderate intensity light is usually small
compared to the concentration of majority carriers that
is obtained from thermal ionization of dopant atoms in
the solid. This implies that such levels of illumination do
not signicantly perturb the majority carrier behavior
either in the semiconductor or at the semiconductorli-
quid interface. Because the majority carrier concentra-
tions are essentially unchanged, the rate equations that
govern majority carrier charge oware also unchanged.
Majority carriers should thus exhibit a JE characteris-
tic that is well described by the diode equation, regard-
less of whether the semiconductor is in the dark or is
exposed to moderate levels of illumination.
Unlike the situation for majority carriers, illumina-
tion generally effects a substantial change in the con-
centration of minority carriers. Calculation of the
minority carrier current is greatly simplied by consid-
ering the effects of the electric eld at the semiconduc-
torliquid junction. For most semiconductorliquid
junctions in depletion, the electric eld is so strong that
essentially all the photogenerated minority carriers are
separated from the photogenerated majority carriers
andthencollected. Using this approximation, the photo-
generated minority carrier current density J
ph
is simply
equal to the photon ux absorbed by the semiconductor
multiplied by the charge q on an electron.
The total current densitypotential characteristics of
an illuminated semiconductor electrode can thus be
obtained by adding together, with the appropriate sign,
the majority and minority carrier components of the cur-
rent density. The majority carrier current density obeys
the diode equation, while the minority carrier photocur-
rent density is relatedto the absorbedlight intensity. The
expression for the total current density is, therefore,
J = J
ph
J
0
exp
qE
kT
_ _
1
_ _
(26)
The sign of the minority carrier current (photocurrent)
density is opposite to that of the majority carrier current
density, because holes crossing the interface lead to an
oxidation current, while electrons crossing the interface
leadto areductioncurrent. Equation26is obviously just
the diode curve of Equation 23 offset by a constant
amount J
ph
over the voltage range of interest (Fig. 3).
SEMICONDUCTOR PHOTOELECTROCHEMISTRY 5
For J
ph
>J
0
, as is generally the case, 1 in Equation 26
can be neglected. We then obtain
J ~ J
ph
J
0
exp
qE
kT
_ _ _ _
(27)
We can then dene the opencircuit voltage V
oc
as the
absolute valueof the voltage present whennonet current
ows and obtain
V
oc
=
kT
q
ln
J
ph
J
0
_ _
(28)
This voltage is signicant in the eld of solar energy
conversion, as it represents the maximum free energy
that can be extracted from a semiconductorliquid
interface.
Equation 28 brings out several important features of
the open-circuit voltage. First, V
oc
increases logarithmi-
cally with the light intensity, because J
ph
is linearly
proportional to the absorbed photon ux. Second, the
open-circuit voltage of a system increases (logarithmi-
cally) as J
0
decreases. Chemically, such behavior is
reasonable, because J
0
represents the tendency for the
system to return to charge transfer equilibrium. Third,
Equation 28 emphasizes that a mechanistic under-
standing of J
0
is crucial to controlling V
oc
. Only through
changes inJ
0
canasystematic, chemical control of V
oc
be
established for different types of semiconductorliquid
junctions.
Another parameter that is often used to describe
illuminated semiconductorliquid junctions is the
short-circuit photocurrent density J
sc
. Short-circuit
conditions imply V=0. From Equation 26, the net
current density at short circuit (J
sc
) equals J
ph
.
The short-circuit current density provides a measure
of the collection efciency of photogenerated carriers
in a particular photoelectrochemical cell. The reader is
referred to earlier reviews for a more extensive discus-
sionof howthese parameters are relevant insolar energy
research using semiconductor-based photoelectro-
chemical cells (Lewis, 1990; Tan et al., 1994b).
Dependence of Photocurrent on the Wavelength of
Illumination. The wavelength dependence of the
photocurrent produced at a semiconductorliquid
contact can provide a nondestructive, routine method
for determining some important optical properties of a
semiconductor. Specically, the value of the band gap
energy, whether the electronic transition is optically
allowed or forbidden, and the minority carrier
diffusion length can be obtained from measurement of
the spectral response of the photocurrent at a
semiconductorliquid contact.
Two types of optical transitions are commonly
observed for semiconductors: direct gaps and indirect
gaps. Near the absorption edge, the absorption coef-
cient a can be expressed as (Pankove, 1975; Sze, 1981;
Schroder, 1990)
a ~ hvE
g
_ _
b
(29)
where his the Plancks constant, v is the frequency of the
light incident onto the semiconductor, and b is the
coefcient for optical transitions. The absorption coef-
cient is obtained from the BeerLambert law, in which
the ratio of transmitted G to incident G
0
photon ux for a
sample of thickness d is (Pankove, 1975)
G=G
0
= exp(ad) (30)
For optically allowed direct gap transitions, b=1/2,
whereas for indirect optically forbidden transitions,
b =2 (Sze, 1981).
According to the G artner equation, the photocurrent
is given as
J
ph
= qG
0
1R
*
_ _
1
exp aW ( )
1aL
_ _
(31)
where L is the minority carrier diffusion length and R
+
is
the optical reectivity of the solid (Sze, 1981; Lewis and
Rosenbluth, 1989; Schroder, 1990). In a semiconductor
samplewithaveryshort minoritycarrier diffusionlength
and with aW 1, this equation simplies to (Sze, 1981;
Schroder, 1990)
J
ph
= qG
0
1R
*
_ _
aW ( ) (32)
Under these conditions, a canbe measured directly from
the photocurrent at each wavelength. These values can
then be plotted against the photon energy to determine
the band gap energy and transition prole, direct or
indirect, for the semiconductor under study.
The minority carrier diffusion length L is an
extremely important parameter of a semiconductor
(Anodic)
(Cathodic)
C
u
r
r
e
n
t
0
() 0
(+)
Voltage
I
ph
a
b
Figure 3. Ideal currentvoltage behavior of a semiconductor
liquid junction (a) in the dark and (b) under illumination. The
current observed under illumination is offset from the current
observed in the dark by the value of the photocurrent I
ph
.
6 SEMICONDUCTOR PHOTOELECTROCHEMISTRY
sample. This quantity describes the mean length over
which photogenerated minority carriers can diffuse in
the bulk of the solid before they recombine with major-
ity carrier. The value of L affects the bulk diffusion
recombination limited V
oc
and the spectral response
properties of the solid.
The ASTM (American Society for Testing and Materi-
als) method of choice for measurement of diffusion
length is the surface photovoltage method. A conceptu-
ally similar methodology can, however, be used when a
liquid provides the electrical contact to the semiconduc-
tor. Use of a semiconductorliquid contact has the
advantage of allowing a reproducible analysis of the
surface condition as well as control over the potential
across the semiconductor during the experiment. In
either mode, the method works well only for silicon and
other indirect gap materials.
We assume that the semiconductor is n-type, so
we are interested in measuring the diffusion length
of holes L
p
. Simplication of the G artner equation,
assuming that aW 1, yields the following expression
for the wavelength dependence of the open-circuit
photovoltage (Sze, 1981; Lewis and Rosenbluth, 1989;
Schroder, 1990):
V
OC
~ 1 L
p
a
_ _
1
_ _
1
(33)
Thus, provided that L
p
W and a
1
W, a plot of
(V
oc
)
1
versus a
1
will yield a value of the inverse of the
slope that is equal to L
p
. An additional check on the data
is that the x intercept of the plot should equal L
p
(Schroder, 1990).
Measurement of L
p
using the surface photovoltage
method in air requires that the semiconductor be
capacitively coupled through an insulating dielectric
such as mica to an optically transparent conducting
electrode. However, for a semiconductorliquid con-
tact, either the photovoltage or photocurrent can be
determined as a function of wavelength l, and in this
implementation the method is both nondestructive and
convenient.
The methods for determination of L
p
require accurate
measurement of the wavelength-dependent quantum
yields for carrier collection at the semiconductorliquid
contact, also termed the IPCE or incident photon to
current efciency. The quantumyield is the ratio of the
rate at which a specimen forces electrons through an
external circuit, I(l)/q, to the rate at which photons are
incident upon its surface:
F(l) =
I(l)=q
G
0
(l)A
s
(34)
In this equation, G
0
(l) represents the ux of monochro-
matic light, which is assumed to be constant over the
area of the specimen A
S
.
Typically, the quantum yield is measured at short
circuit. Commercial siliconphotodiodes have very stable
quantum yields of >0.7 throughout most of the visible
spectrum, making them a nearly ideal choice for a
calibration reference. The quantum yield of the experi-
mental specimen can be calculated as follows:
F
cell
(l) = F
ref
(l)
I
cell
(l)A
s;ref
I
ref
(l)A
s;cell
(35)
Experimentally, the excitation monochromator is
scannedto recordI
cell
(l) andthenthe experimental spec-
imenis replacedwiththe reference andI
ref
(l) is recorded.
Practical Aspects of the Method
SamplePreparation. Electrodesfor photoelectrochemical
measurements should be constructed to allow exposure
of the front face of the semiconductor to the solution
while providing concealment of the back contact and of
the edges of the electrode. This is readily accomplished
using an insulating material that is inert toward both
the etchant and the working solution of interest. The
area of the electrode should be large enough to allow
ready measurement of the bulk surface area, but should
be small enough to limit the total current owing
through the electrochemical cell (because larger currents
require larger corrections for the cell resistance).
Because of these trade-offs, electrode areas are typically
0.11cm
2
.
Ohmic contacts vary widelybetweensemiconductors,
and several books are available for identifying the ohmic
contact of choice for a given semiconductor (Willardson
and Beer, 1981; Pleskov and Guervich, 1986; Fink-
lea, 1988). Although most ohmic contacts are prepared
by evaporating or sputtering a metal on the back surface
of the semiconductor, some semiconductors are amena-
ble to more convenient methods suchas using a scribe to
rub a galliumindiumeutectic onthe back surface of the
solid. Thislatter procedureiscommonlyusedtomakean
ohmic contact to n-type silicon. The quality of an ohmic
contact can be veried by making two contacts, sepa-
ratedby a contact-free region, onone side of anelectrode
and conrming that there is only a slight resistance
between these contacts as measured by a JE
curve collected between these contact points.
The proper choice of a chemical etch depends on the
semiconductor, its orientation, and the desired surface
properties. Generally, an ideal etchant produces an
atomically smooth surface with no electrical surface
defects. Fluoride-based etches are most commonly
used with silicon: a 40% (w/w) ammonium uoride
solutionis well suitedfor (111)-orientedSi andasolution
of HF is appropriate for (100)-oriented Si (Higashi
et al., 1991,1993). For many IIIV semiconductors, an
etchin0.05%(v/v) Br
2
followedby a rinse ina solutionof
NH
4
OHproduces abrupt discontinuities inthe dielectric
at the solidair interface (Aspnes and Studna, 1981). An
exhaustive literature provides informationonadditional
etches for these and other semiconductors, and the
reader is referred to these references for further infor-
mation (Wilson et al., 1979; Aspnes and Studna, 1981;
Higashi et al., 1993). There are also several published
reports of effective dry etching methods (Higashi
et al., 1993; Gillis et al., 1997).
SEMICONDUCTOR PHOTOELECTROCHEMISTRY 7
Because manysemiconductors are reactive inaerobic
environments, it is often necessary to carry out experi-
ments in anaerobic conditions using nonaqueous sol-
vents. Air-sensitive experiments can be performed in
specialized glassware that is continuously purged with
aninert gas or inaninert atmosphere glove box specially
modied for electrical throughputs. Although outlined
here for currentvoltage measurements, the electrode
preparation techniques are applicable not only to these
measurements but also to most other techniques dis-
cussed in this article.
Basic Electrochemical Cell Design. Current density
potential data for semiconductor electrodes are
typically obtained using a potentiostat (Fig. 4). This
instrument ensures that the measured JE properties
are characteristic of the semiconductorliquid interface
and not of the counter electrodeliquid contact that is
needed to complete the electrical circuit in the
electrochemical cell. A three-electrode arrangement
consisting of a semiconductor (working) electrode, a
counter electrode, and a reference electrode is
typically used to acquire data. The potentiostat uses
feedback circuitry and applies the voltage needed
between the working electrode and counter electrode
to obtain the desired potential difference between the
working and reference electrodes. The potentiostat then
records the current owing through the working
electrodecounter electrode circuit at this specic
applied potential. Nominally, no current ows through
the reference electrode, which only acts as a point of
reference for the system. The scanrate shouldbe 50mV/
s or slower in order to minimize hysteresis arising from
diffusion of species to the electrode surface during the
JE scan. In addition, for semiconductor lms with bulk
defect states and/or grain boundary capacitance,
charge accumulation during JE scans is negligible at
fast scan rates.
The electrochemical data are collected directly as
the current versus the applied potential. Electrode
areas are, of course, needed to obtain current densities
fromthe measured values of the current. The projected
geometric area of the electrode is usually obtained by
photographing the electrode and a microruler simul-
taneously under a microscope and digitally integrating
the area dened by the exposed semiconductor
surface.
Under potentiostatic control, the concentrations of
both forms of the redox species need not be as high as
might be required to sustainidentical performance inan
actual, eld-operating photoelectrochemical cell. This
occurs because a two-electrode photovoltaic-type cell
conguration requires sufciently high concentrations
of both forms of the redox couple dissolved in the solu-
tion to suppress mass transport limitations without
mechanical stirring of the electrolyte. In a three-elec-
trode cell with an n-type semiconductor electrode, the
primary consideration is that sufcient redox donor be
present such that the anodic current is limited by the
light intensity and not by the mass transport of donor to
the electrode surface. A high concentration of redox
acceptor is not required to achieve electrode stability
and often is undesirable when the oxidized form of the
redox material absorbs signicantly in the visible region
of the spectrum. The concentration of overpotential that
results from a low concentration of electron acceptor in
the electrolyte can be assessed and corrected analyti-
cally using Equation 38.
In contrast, the performance of an actual energy
conversion device using a two-electrode cell congura-
tion is so dependent on the properties of the working
electrode, the counter electrode, the electrolyte, the cell
thickness, and the properties of the various optical
interfaces in the device that many design trade-offs are
involvedandare unique toaparticular cell conguration
used in the device assessment. Emphasis here has been
placedondetermining the properties of the semiconduc-
tor electrode inisolation, using a potentiostat, so that a
comparison of electrode to electrode can be performed
without considering the details of the device congura-
tion used in each measurement.
Reference Electrodes. Reference electrodes are
constructed according to conventional electrochemical
protocols. For example, two types of reference electrodes
are an aqueous (or nonaqueous) saturated calomel
electrode (SCE) and a nonaqueous ferrocenium
ferrocene electrode.
A simple SCE can be constructed by rst sealing a
platinum wire through one leg of an H-shaped hollow
glass structure. The platinum wire is then covered with
mercury, and a ground mixture of approximately equal
amounts of mercury andcalomel (Hg
2
Cl
2
) dispersedinto
asmall amount of saturatedpotassiumchloridesolution
is then placed on top of the mercury. The remainder of
the tube is lled with saturated potassium chloride
A
W
E
set
R
Solution
C
Figure 4. Circuit consisting of a simple potentiostat and an
electrochemical cell. A potential is set between the working and
reference electrodes, and the current ow from the counter
electrode to the working electrode is measured.
8 SEMICONDUCTOR PHOTOELECTROCHEMISTRY
solution, and the other leg of the structure, which con-
tacts the solution, is capped with a fritted plug. Prior to
use, the nonaqueous SCEshouldbe calibratedagainst a
reference electrode with a known potential, such as an
aqueous SCE is prepared in the same fashion.
For work in nonaqueous solvents, a convenient ref-
erence electrode is the ferroceniumferrocene reference.
This electrode consists of a glass tube with a fritted plug
at the bottom. The tube is lled with a ferroceneferro-
ceniumelectrolyte solution made using the same sol-
vent and electrolyte that is to be used in the
electrochemical experiment. A platinumwire is inserted
into the top of the solution to provide for a stable refer-
ence potential measurement.
When both forms of the redox couple are present in
the electrochemical cell, an even simpler procedure can
be used to construct a reference electrode. A platinum
wire can be inserted into the electrolyteredox couple
solution or into a Luggin capillary that is lled with the
electrolyteredoxcouple solution(see Luggincapillaries,
below). This wire then provides a stable reference poten-
tial that is equal to the Nernstian potential of the elec-
trochemical cell. At any convenient time, the potential of
this reference can be determined versus another refer-
ence electrode, such as an SCE, through insertion of the
SCE into the cell. This approach is not only convenient
but also useful when water and air exposure is to be
minimized, as is the case for reactive semiconductor
surfaces in contact with deoxygenated nonaqueous
solvents.
Luggin Capillaries. The cell resistance can be reduced
by minimizing the distance between the working and
reference electrodes. These small distances can be
achieved through the use of a Luggin capillary as a
reference electrode. The orice diameter of the
capillary should generally be ~0.1mm. A convenient
method to form such a structure is to pull a
disposable laboratory pipette under a ame and then
to use a caliper to measure and then break the pipette
glass at the point that corresponds to the desired orice
radius. The pipette is then lled with the reference
electrode solution of interest and the ow of electrolyte
out of the pipette is minimized by capping the top of the
pipette with a rubber septum. The contact wire is then
inserted through the septum and into the electrolyte.
Under some conditions, asyringe needle connectedto an
empty syringe can be inserted through the septum to
facilitate manipulation of the pressure in the headspace
of the pipette. This procedure can be used to minimize
mixing between the solution in the pipette and the
solution in the electrochemical cell.
Illumination of SemiconductorLiquid Contacts
Monochromatic Illumination. Low-intensity monochro-
matic illumination can be obtained readily from a white
light source and a monochromator. This is useful for
obtaining spectral response data to measure the
diffusion length or the optical properties of the
semiconductor electrode, as described in more detail
under Measurement of Semiconductor Band Gaps
Using SemiconductorLiquid Interfaces.
Laser illumination can also be used to provide mono-
chromatic illumination. However, care should be taken
to diffuse the beamsuchthat the entire electrode surface
is as uniformly illuminated as possible. Because the
photovoltage is a property of the incident light intensity,
careful measurement of the photovoltage requires main-
taining a uniform light intensity across the entire elec-
trode surface. This protocol has not been adhered to in
numerous measurements of the JE properties of semi-
conductor electrodes, and the photovoltages quoted in
such investigations are, therefore, approximate values
at best. To control the light intensity from the laser,
neutral density lters can be used to attenuate the
incident beam before it strikes the electrode surface.
Regardless of whether the monochromatic light is
obtained from a white lightmonochromator combina-
tionor froma laser, measurement of the incident photon
power is readily achieved with pyranometers, photo-
diodes, thermopiles, or other photon detectors that are
calibrated in their response at the wavelengths of
interest.
Polychromatic Illumination. For polychromatic illumina-
tion, solar simulators provide the most reproducible
laboratory method for measuring JE properties
under standard, solar-simulated illumination. A less
expensive method is to use tungstenhalogen ELH-type
projector bulb lamps. However, their intensity
wavelength prole, like that of almost any laboratory
light source, is not very well matched to the solar
spectrum observed at the surface of the earth.
Calibration of the light intensity produced by this type
of source should not be done with a spectrally at
device such as a thermopile. Since laboratory sources
typically produce more photons in the visible spectral
region than does the sun at the same total illumination
power, maintaining a constant power from both illumi-
nation sources tends to yield higher photocurrents, thus
producing overestimates of efciency of photoelectro-
chemical cells in the laboratory, relative to their true
performance under an actual solar spectral distri-
bution in the eld.
An acceptable measurement method instead involves
calibration of the effective incident power produced by
the laboratory source through use of a photodetector
whose spectral response characteristics are very similar
to that of the photoelectrochemical cell of concern. Pref-
erably, the response properties of the photodetector are
linear with light intensity and the absolute response of
the detector is known under a standard solar spectral
distributionandilluminationpower. Theabsolutedetec-
tor response can be obtained either by measurements
performed under a calibrated solar simulator or by mea-
surement of the output of the detector inactual sunlight.
If sunlight isused, another calibrationisthenrequiredto
determine the actual solar power strikingthe plane of the
detector at the time of the measurement. Useful primary
or secondary reference detectors for this purpose are
silicon cells that have been calibrated on balloon ights
SEMICONDUCTOR PHOTOELECTROCHEMISTRY 9
by NASA. Spectrally at radiometers, such as those
produced by Eppley for the purpose of measuring the
absolute spectral power striking a specic location on
the surface of the earth, are also useful for determining
the solar power under conditions used to calibrate the
detector to be used in the laboratory. If these primary or
secondary reference detectors are routinely available, it
is of course also possible to determine the JEproperties
of the photoelectrochemical cell directly in sunlight, as
opposed to having to establish a reference detector mea-
surement and then recalibrate a laboratory light source
to produce the equivalent effective spectral power for the
purposes of the measurement.
Pulsed Illumination. Pulsed light sources can also be
used, and the most convenient method is to have the
excitation beam interrupted by a mechanical chopper,
with a lock-in amplier used to monitor the
photocurrent. If the response time of the specimen is
short compared to the chopping period, then this
method yields exactly the same information as the DC
experiment. However, lock-in detection offers a
substantial advantage in the signal-to-noise ratio,
which can be important at low light and/or low
current levels. Also, for experiments that are not
carried out at short circuit but at some applied bias,
this methodprovides for anautomatic subtractionof the
dark current. In addition, when a lock-in detection
method is used, it is possible to measure the
differential quantum yield of a signal that is
superimposed on top of a DC illumination source.
This information is very useful for materials that have
quantumyields that are dependent on the incident light
intensity. Finally, when the response time of the system
is onthe same order asthe choppingperiod, variations in
the photocurrent with the chopping period and analysis
of the data in the frequency domain can yield detailed
information about the kinetic process occurring in the
semiconductor bulk and at the semiconductorliquid
interface (Peter, 1990).
Cell Conguration Under Illumination. A preferred
experimental setup is shown in Figure 5. In this
arrangement, the ratio of the photocurrent response of
the experimental specimen to that of an uncalibrated
photodiode, I
cell
(l)/I
uncal
(l), is recorded as the
experimental variable. It is not necessary to know the
quantum yield or area of the uncalibrated photodiode,
which merely acts to calibrate the light intensity at each
wavelength. The geometry of the uncalibrated
photodiode with respect to the light source and the
pick-off plate should be arranged such that the
surface of the uncalibrated diode is illuminated by a
small fraction of the light that is incident on the main
specimen. If t(l) is the ratio of the light diverted to the
uncalibrated photodiode relative to that which reaches
the main specimen, then
I
uncal
(l) = qF
uncal
(l)t(l)G
0
(l)A
s; uncal
(36)
The photocurrent response of the calibrated photodiode
relative to that of the same uncalibrated photodiode,
I
REF
(l)/I
UNCAL
(l), must also be determined. When I
cell
(l)
and I
ref
(l) in Equation 35 are replaced with the ratios
I
CELL
(l)/I
UNCAL
(l) and I
REF
(l)/I
UNCAL
(l), respectively, the
unknown terms F
uncal
and A
s,uncal
divide out. This cell
conguration eliminates error that may arise from the
drift in the light source intensity over time in photocur-
rent/photovoltage measurements as a function of wave-
length, which can affect both I
cell
(l) and I
ref
(l).
Data Analysis and Initial Interpretation
CurrentPotential Data. A representative example of a
currentpotential curve is shown in Figure 6, which
displays data for an n-type silicon electrode in contact
with a redox-active methanol solution. In this gure, the
current has been divided by the surface area of the
electrode to allow quantitative analysis of the data. To
extract meaningful results from the currentpotential
curve, it is also necessary to perform corrections for
concentration overpotential and solution resistance,
as discussed below.
The open-circuit photovoltage and short-circuit pho-
tocurrent canbe estimatedfromacorrectedtime-depen-
dent scanof the JEdata. However, it is preferredif these
values be measured directly using four-digit voltmeters
connected to the photoelectrochemical cell. This steady-
state measurement eliminates any bias that might arise
due tothe presence of hysteresis inthe currentpotential
behavior. Also, in some cases, the light-limited photo-
current is not reached at short circuit; in this case, both
the light-limited photocurrent value and the short-cir-
cuit photocurrent value are of experimental interest and
should be measured separately.
Lamp
Monochromator
and filter
Beam splitter
I
uncal
I
cell
or

I
ref
Figure 5. Spectral response measurement system consisting of a white light source, a mono-
chromator, a beam splitter, and a calibrated photodiode.
10 SEMICONDUCTOR PHOTOELECTROCHEMISTRY
Correction for Concentration Overpotentials. Attention to
electrochemical cell design is critical to minimize
concentration overpotentials, mass transport
restrictions on the available current density, and
uncompensated resistance drops between the working
and reference electrodes. Even with good cell design, in
nonaqueous solvents the JE curves must generally be
corrected for concentration overpotential losses as well
as for uncompensated ohmic resistance losses to obtain
the inherent behavior of the semiconductorliquid
contact.
To minimize mass transport limitations on the cur-
rent, the electrolyte should be vigorously stirred during
the JE measurement. For a given redox solution, the
limiting anodic current density J
1,a
and the limiting
cathodic current density J
1,c
should be determined
usingaplatinumfoil electrode placedinexactlythe same
conguration as the semiconductor working electrode.
The areas of the two electrodes should also be compa-
rable. If the redox couple is known to be electrochemi-
cally reversible, the platinumelectrode datacanthenbe
used to obtain the cell parameters needed to performthe
necessary corrections to the JE data of the semicon-
ductor electrode.
Alternatively, the semiconductor electrode can be
fabricated into a disk conguration, and can be rotated
in the electrolyte. Under these conditions, the mass
transport parameters E can be determined analytically,
and the limiting current density (Bard and Faul-
kner, 1980) is
J
1;c
= 0:620nFD
2=3
0
o
1=2
rde
u
1=6
A [ [
b
(37)
where F is Faradays constant, D
0
is the diffusion coef-
cient, o
rde
is the angular velocity of the electrode, v is
the kinematic velocity of the solution (~0.01cm
2
/s for
dilute aqueous solutions near 20

C), and [A]


b
is the
bulk concentration of oxidized acceptor species. A sim-
ilar equation yields the limiting anodic current density
based on the parameters for the reduced form of the
redox species. This procedure allows control over the
limiting current densities instead of merely measuring
the values in a mechanically stirred electrolyte
solution.
Laminar ow typically ceases to exist above Rey-
nolds numbers (dened as the product of o and the
disk radius of the electrode divided by v) of 210
5
(Bard and Faulkner, 1980); so for electrode radii of
1mm, this corresponds to a typical upper limit on
the rotation velocity of 1 to 210
7
rpm. Beyond this
limit, Equation 37 does not describe the mass trans-
port to the electrode. Smaller electrodes can increase
this limit on o, but use of smaller electrodes is
generally not advisable, because edge effects become
important and can distort the measured electro-
chemical properties of the solidliquid contact by hin-
dering diffusion of minority carriers and allowing
recombination at the edges of the semiconductor
crystal.
Once the limiting current densities and the JE data
are collected for a reversible redox system at a metal
electrode, the concentration overpotential Z
conc
can be
determined (Bard and Faulkner, 1980):
Z
conc
=
kT
nq
ln
J
1;a
J
1;c
_ _
ln
J
1;a
J
JJ
1;c
_ _ _ _
(38)
These values can then be used to correct the data at a
semiconductor electrode to yield the proper JE depen-
dence of the solidliquid contact in the absence of such
concentration overpotentials.
Correction for Series Resistance Overpotentials. Even with
goodcell design, measurement of the cell resistance R
soln
is requiredto performanother correctionto the JEdata.
Values for R
soln
can be extracted from the real
component of the impedance in the high-frequency
limits of Nyquist plots (for further discussion, see
ELECTROCHEMICAL TECHNIQUES FOR CORROSION QUANTIFICATION)
for the semiconductor electrode or can be determined
from steady-state measurements of the ohmic
polarization losses of a known redox couple at the
platinum electrode. In the former method, R
soln
is
simply taken as the real part of the impedance in the
high-frequency limit of the Nyquist plot. In the latter
method, the currentpotential properties of a platinum
electrode are determined under conditions where the
platinumelectrode is inalocationidentical tothat of the
semiconductor electrode. After correction of the data
for concentration polarization, R
soln
can be obtained
from the inverse slope of the platinum current
potential data near the equilibrium potential of the
solution.
0.10
0.60
0.2 0.1 0 0.1
Potential (V versus E(A/A

)/q)
J

(
m
A
/
c
m
2
)
0.2 0.3 0.4 0.5
0.50
0.40
0.30
0.20
0.10
0
Figure 6. Representative example of currentvoltage data for a
semiconductorliquid interface. The systemconsists of a silicon
electrode incontact withamethanol solutioncontaining lithium
chloride and oxidized and reduced forms of benzyl viologen. In
this example, the current has been divided by the surface area
of the electrode, yielding a current density as the ordinate. The
curve has not beencorrectedfor cell resistance or concentration
overpotential.
SEMICONDUCTOR PHOTOELECTROCHEMISTRY 11
The nal corrected potential E
corr
is then calculated
from E, the concentration overpotential Z
conc
, R
soln
, and
the current I using (Bard and Faulkner, 1980)
E
corr
= EZ
conc
IR
soln
(39)
The measured value of I is divided by the projected
geometric area of the electrode and plotted versus E
corr
to obtain a plot of the JE behavior of the desired semi-
conductorliquid contact.
Natural Logarithm of J Versus E. Figure 7 shows plots of
ln(J)E for an n-Si electrode in contact with solutions
having varying ratios of [A]/[A

]. In this example, an
ideal rst-order concentration dependence is evident
from a 59mV shift in the ln(J)E curve for a factor
of 10 increase in [A]/[A

]. A rst-order dependence on
the surface electron concentration can be veried by
tting ln(J)E curves to a standard diode equation:
J = J
0
exp
qE
gkT
_ _ _ _
(40)
Note that this expression is a simplied form of Equa-
tion 23 and is valid when the electrode is sufciently
reverse biased. The diode quality factor g is a oating
parameter that should yield a value of 1.0 for anelectron
transfer-limited system(compare Equations 40 and 23).
Assuming a diode quality factor of unity, the value of the
rate constant can be extracted at any given potential.
However, in practice, it is best to obtain the value of the
rate constant using values of the current for which
resistance effects are still minimal and for which the
simplied form of the diode equation is still valid.
Problems. Toobtainreliablevaluesoftheheterogeneous
rateconstant, thesemiconductorliquidsystemofinterest
must exhibit second-order kinetics, with a rst-order
dependence on both the concentration of majority
carriers at the surface and the concentration of redox
species. However, most systems do not exhibit either
one or both of these dependencies. A lack of
concentration dependence is more likely in systems that
employredoxspeciesorelectrolytesthat canabsorbonthe
surface of the electrode andwithelectrodesthat have high
surface recombination velocities. Such systems are often
dominated by alternative recombination mechanisms
precluding a quantitative kinetic analysis of desired rate
process. Thus, behavior generally needs to be explored,
and veried, for several redox couples that have E(A/A

)
varying over a signicant potential range to ensure
condence in the results of the kinetic measurements. In
addition, it is important to establish that one is not simply
working ina linear regionof anadsorptionisotherm, such
that the electron capture event is entirely proceeding to
adsorbed species, with the concentration of the adsorbate
being linearly related to the concentration of the redox
species in the solution phase.
Signicant effort must go into preparation of nearly
defect-free surfaces inorder to extract values for k
et
from
the steady-state JE data. Recent results have shown
that this is possible for n-Si and n-InP semiconductor-
liquid contacts (Fajardo and Lewis, 1997; Pomykal and
Lewis, 1997), and other systems are currently under
investigation as well. Special care should be taken in
determining the kinetics of semiconductor electrodes
according to Equation 8. Rate constants that do not
meet these criteria are often quoted in units of centi-
meters to the fourth power per second (Meier
et al., 1997), and this is clearly not in accord with
conventional chemical kinetic protocols.
To establish the desired kinetic behavior experimen-
tally, the concentration of the acceptor must be varied
(Rosenbluth and Lewis, 1989). However, in doing so, the
electrochemical potential of the solution will change if
the concentration of the other redox partner is held
constant (Equation 1). One approach is to dilute the
solution, thereby not varying E(A/A

) while changing
the concentration of the desired redox species. This is
useful but often precludes simultaneous differential
capacitance measurements, whichcanrequirehighcon-
centrations of both forms of the redox couple in the
electrolyte to avoid spurious results due to concentra-
tion polarization at the counter electrode of the system.
Another method is to change the concentration of only
one form of the redox species. However, care must then
be taken to ensure that the band edge positions of the
semiconductor do not shift as the redox potential is
changed. If this is not the case, interpretation of the data
is difcult and problematic.
Once the correct kinetic dependencies on [A]
s
and n
s
have been established, it is straightforward to evaluate
k
et
from the measured value of J at a given potential. To
dothis, however, requires anindependent measurement
of the value of n
s
at this potential. Traditionally, C
2
E
methods are used for this purpose, and this
0.50
E (V versus SCE)
I
n

(

J
/
(
A
/
c
m
2
)
)
10.5
9.5
8.5
7.5
0.45 0.40 0.35
Figure 7. Plots of ln(J) versus E for the system described in
Figure 4 for two different ratios of [A]/[A

]. The data on the right


had a ratio of [A]/[A

], 10-fold higher than that for the curve on


the left.
12 SEMICONDUCTOR PHOTOELECTROCHEMISTRY
experimental protocol is discussed in detail under Sec-
tion Differential Capacitance Measurements of Semi-
conductorLiquid Contacts. Care should be taken to
ensure that the bandedge positions donot move versus a
xed reference potential, because then the data are
problematic to interpret. Otherwise, simple arithmetic
manipulationof Equation8yields the desiredvalue of k
et
if n
s
, [A]
s
, and J are known.
Photocurrent Versus Wavelength. Figure 8 shows the
spectral dependence of the photocurrent at an n-type
MoS
2
electrode (Tributsch and Bennett, 1977). The
principal analysis step in determining the band gap
energy or any other intrinsic parameter, such as the
diffusion length, from this spectrum depends on
accurately transforming the wavelength-dependent
photocurrent information to a corresponding
absorption coefcient. With a table of absorption
coefcients, photocurrent densities, and wavelengths,
one can plot the absorption coefcient versus photon
energy and extract the band gap energy.
For a direct band gap semiconductor, a plot of a
2
versus photon energy will give the band gap as the x
intercept. If the semiconductor has anindirect bandgap,
a plot of a
1/2
versus photon energy will have two linear
regions, one corresponding to absorption with phonon
emission and one corresponding to absorption with
phonon capture. The average of the two intercepts is the
energy of the indirect band gap. One-half the difference
between the two intercepts is the phononenergy emitted
or captured during the band gap excitation. The only
parameter that needs to be controlled experimentally is
G
0
at the various wavelengths of concern.
Figure 9 illustrates the procedure for determining
the diffusion length from a plot of F versus 1/a for
silicon (Schroder, 1990). The minority carrier diffusion
length is readily obtained by extrapolating the quan-
tum yield data to the x intercept and taking the value
of 1/a when F=0. Since the quantum yield is mea-
sured at different wavelengths, the photon ux is
adjusted to ensure a constant photovoltage at each
F measurement.
DIFFERENTIAL CAPACITANCE MEASUREMENTS OF
SEMICONDUCTORLIQUID CONTACTS
Principles of the Method
Differential capacitance measurements of semiconduc-
torliquid contacts are very useful in obtaining values
for the dopant density and at-band potential of the
bulk semiconductor. In addition, such measurements
have been found to be of great use in determining
doping proles of heterojunctions (Seabaugh
et al., 1989) and of epitaxial layers of semiconductors
(Leong et al., 1985) fabricated for use in light-emitting
diodes, transistors, solar cells, and other optoelec-
tronic devices.
To obtain an expression for the differential capaci-
tance versus potential properties of a semiconductor
liquid contact, we refer again to Equations 3a, 4a, which
describe the basic electrostatic equilibrium conditions
at a semiconductorliquid interface. Because all the
dopants in the depletion region are assumed to be
0.4
0.3
0.2
0.1
0
350 450 550
Wavelength (nm)
P
h
o
t
o
c
u
r
r
e
n
t

d
e
n
s
i
t
y

(
m
A
/
c
m
2
)
650 750
Figure 8. Spectral response of MoS
2
photocurrents (n-type) in
the anodic saturation region. Reprinted with permission from
Tributsch and Bennett, 1977.
1
0.8
0.6
0.4
0.2
0
100 50 0 50 100 150 200
L
n
(m)
N
o
r
m
a
l
i
z
e
d

p
h
o
t
o
n

f
l
u
x
L
n
= 25 m
6 m
90 m
l/ (m)
Figure 9. Plot of F versus the inverse absorption coefcient for
three Si diodes with different diffusion lengths. The minority
carrier diffusion lengths are obtained from the value of |1/a|
whenthe quantumyieldis zero. Reprintedwithpermissionfrom
Schroder, 1990.
SEMICONDUCTOR PHOTOELECTROCHEMISTRY 13
ionized, the charge inthe depletionregionat apotential E
can be expressed as
Q = qee
0
N
d
V
bi
E
kT
q
_ _ _ _
1=2
(41)
Taking the derivative of Q with respect to E yields an
expression for the differential capacitance C
d
of the
semiconductor:
C
d
=
dQ
dE
=
qee
0
N
d
2 V
bi
EkT=q ( )
_ _
1=2
(42)
where e is the relative permittivity of the semiconductor
and e
0
is the permittivity of free space. A plot of C
d
2
versus E (i.e., a MottSchottky plot) should thus yield a
straight line that has a slope of 2(qe
s
N
d
)
1
(with e
s
=ee
0
)
and an x intercept of kT/qV
bi
.
To determine the barrier height, the at-band poten-
tial of the solidliquid contact (the intercept of the
MottSchottky plot) should yield the value of V
bi
after
inclusion of a small correction term, kT/q, which equals
0.0259V at room temperature. The value of V
bi
can be
related to the barrier height as follows:
f
b
= V
bi

kT
q
ln
N
c
N
d
_ _
(43)
The value for V
bi
should be independent of the mea-
surement frequency, and the slope of the plot of C
2
versus E should be in accord with that expected from
the value of N
d
and the electrode area A
s
, according to
Equation 42. In addition, a useful check on the validity
of the V
bi
value obtained from the data is to ensure that
the experimentally determined value of V
bi
shifts 59mV
for every decade that N
d
is increased. This dependence
occurs because the barrier height f
b
is independent of
the dopant density of the semiconductor, so the value
of V
bi
must change, as given by Equation 43, in
response to a variation in N
d
. If the majority carrier
mobility is known, the value of N
d
can be determined
independently by four-point probe measurements of
the sample resistivity or it can be determined by Hall
effect measurements (for further discussion, see DEEP-
LEVEL TRANSIENT SPECTROSCOPY), which yield both N
d
and
m
n
directly from the measurement on an n-type semi-
conductor sample (Sze, 1981).
To characterize the dopant prole through the
solid, an anodic potential is applied to the semicon-
ductor electrode such that the surface is partially
dissolved. The thickness of material dissolved from
the electrode, W
d
, is given by the time integral of the
current density passed, J (Ambridge and Fak-
tor, 1975):
W
d
=
M
n

Fr
_
Jdt (44)
where M is the molecular weight of the semiconductor,
n

is the number of holes required to oxidize one atom


of the electrode material to a solution-phase ion, and r
is the density of the solid. A separate MottSchottky
plot is then obtained at each depth of etching through
the material. The depth of the dopant density measure-
ment is given by W
/
=W W
d
. Thus, periodic dissolu-
tions and MottSchottky determinations can yield a
complete prole of N
d
as a function of distance into
the solid.
The differential capacitance of the semiconductorli-
quidcontact is dominatedbythe differential capacitance
of the spacecharge layer of the semiconductor. Any
additional charge introducedby injectionof carriers into
the solid, either optically or thermally, will therefore
either populate or depopulate states in the bulk or sur-
face of the semiconductor. This change incharge density
in the solid will affect the thickness of the spacecharge
layer. In turn, the change in spacecharge layer thick-
ness will affect the capacitance of the semiconductor
liquid junction. Measurement of the capacitance as a
function of the wavelength (energy) of subband gap light
incident onto the semiconductorliquid junction as is
done in Electrochemical Photocapacitance Spectros-
copy can, therefore, provide informationonthe energies,
time constants, physical location, and character (i.e.,
either acceptor or donor) of the traps in the semicon-
ducting electrode.
Photocapacitance spectroscopy has been used to
characterize deep-level states in semiconductors since
the mid-1960s. The use of EPS, however, offers many
advantages relative to solid-state photocapacitance
measurements. The presence of an electrolyte in EPS
measurements allows leakage currents to be reduced by
adjusting the availability and reorganization energy of
electronic states inthe electrolyte. Also, EPScanbe used
for depth proling of trap states because layers of the
semiconductor can be etched between subsequent EPS
measurements. The depth proling procedure allows
detailed examination of the inuence of surface oxide
layers or surface passivation layers on the surface
recombination kinetics. If the etching is performed elec-
trochemically, then the entire depth proling and EPS
experiment can be conducted in a single apparatus
(Haak et al., 1982).
Bulk Capacitance Measurements. The density of bulk
states (in reciprocal cubic centimeters) can be readily
derived from the MottSchottky relations of
Equations 41 and 42. Upon algebraic manipulation,
the density of optically active states at a particular
illumination energy is given by (Anderson, 1982; Haak
et al., 1982; Goodman et al., 1984; Haak and
Tench, 1984)
N
t
(l) =
8p[EE(A=A

)[
e
sc
q
(C
2
0
C
2
d
) (45)
where C
d
is the differential capacitance measured at this
particular illumination energy (and corresponds to
photons of energy just above the threshold energy for
trap states of interest) and C
0
is the capacitance mea-
sured just prior to reaching the threshold energy for the
14 SEMICONDUCTOR PHOTOELECTROCHEMISTRY
trap states of interest. At an illumination intensity suf-
cient to saturate all the trap states at the energy under
study, the measured plateau or peak capacitance C
d,sat
corresponds to the total density of bulk trap states, N
t
,
optically excitable at this particular illuminationenergy.
The in Equation 45 takes into consideration the fact
that N
t
(l) can either increase or decrease as the space
charge layer is either populated or depleted of charge
carriers. Changes in N
t
(l) are also related to the charac-
ter of the state, with acceptor states becoming
more positive when emptied and donor states becoming
more negative when emptied.
Surface Capacitance Measurements. The density of
surface states (in reciprocal centimeters squared) can
similarly be expressed from the MottSchottky
relationships. In a derivation analogous to the bulk
state treatment of Equation 42, the density of surface
states at a particular illumination energy can be shown
to be (Anderson, 1982; Haak et al., 1982; Goodman
et al., 1984; Haak and Tench, 1984)
N
t;s
(l) =
e
sc
N
d
C
H
8p
1
C
2
0

1
C
2
d
_ _
(46)
where C
d
is the differential capacitance measured at the
threshold energy for which the surface states under
study are ionized (populated or depopulated) and C
0
is
the capacitance measuredat anilluminationenergy just
prior toinuencingthe surface statesunder study. Asfor
the case of bulk states, the measured capacitance at
saturation is the total number of surface states present
at a particular illumination energy,N
t,s
.
Kinetic Rates. The kinetics of lling and emptying
optically active states can also be deduced using EPS.
Kinetic equations can be derived for both surface and
bulk states, but only surface kinetic processes will be
discussedindetail. Toperformthese measurements, the
sample is subjected to an intense initial illumination
pulse capable of saturating the states of interest. The
capacitance is then measured continuously after the
illumination is terminated. The time rate of change in
the measured capacitance parallels the time rate of
change for the concentration of lled surface states
and is given by (Haak and Tench, 1984)
dN
t;s
dt
= k
s
N
t;s
N
t;s
(l)
_
k
n
n
s
(47)
where k
s
and k
n
are the ionization (capture) and the
neutralization (emission) rate constants, respectively.
Both rate constants are functions of both optical and
thermal emission processes. For surface states that
become more positive when emptied (acceptor states),
theserateconstantsaregivenas(HaakandTench, 1984)
k
s
= c
n;t
c
n;o
k
p;s
p
s
(48)
k
n
= c
p;t
c
p;o
k
n;s
n
s
(49)
For surface states whose character becomes more neg-
ative when emptied (donor states),
k
s
= c
p;t
c
p;o
k
n;s
n
s
(50a)
k
n
= c
n;t
c
n;o
k
p;s
p
s
(50b)
In dening these rate constants, c
p,t
and c
p,o
are the
thermal and optical emission rate constants for holes,
respectively, and c
n,t
and c
n,o
are the thermal and optical
emission rate constants for electrons, respectively. The
quantities k
p,s
and k
n,s
are the capture coefcients for
electrons and holes dened in Equation 54. It is impor-
tant to note that while an increase in positive charge
produces an increase in capacitance for an n-type semi-
conductor, the opposite effect is observed for p-type
semiconductors.
Extracting the Surface RecombinationCapture Coefcient.
Whenthesemiconductor isawidebandgapmaterial with
deep-level states, one can measure the time dependence
of the differential capacitance after interruption of the
illumination, C
t
, and plot
ln
C
2
0
C
2
t
C
2
0
C
2
t=0
; for bulk states (51a)
or
ln
C
2
0
C
2
t
1=C
2
0
1=C
2
t=0
; for bulk states (51b)
versus time, where C
t=0
is the differential capacitance
measured at the time of the interruption of the illumi-
nation. For an n-type material with deep electron trap
states, c
n,t
and k
p,s
are zero, and the resulting slope for
these lots equals c
p,t
k
n,s
n
s
. If the surface or bulk
states that give rise to the plotted capacitance are suf-
ciently above the valance bandandhence not thermally
ionizable by holes, then c
p,t
=0, and the product of k
n,s
and n
s
can be calculated directly for surface states by
EPS measurements (Haak and Tench, 1984).
Practical Aspects of the Method
Electrolyte Selection. Aqueous systems appear to
provide the most stability for oxygen-passivated
semiconductor surfaces. Studies have shown that for
n-CdSe, the interface states associated with oxygen
adsorption remain unchanged in aqueous solution
relative to the situation for these surfaces in contact
with a vacuum (Goodman et al., 1984; Haak and
Tench, 1984).
In aqueous solutions, useful electrolytes are those
that yield anodic dissolution of the semiconductor but
do not etchthe solid or produce a passivating oxide layer
in the absence of such illumination. For Si, the appro-
priate electrolyte is NaF-H
2
SO
4
(Sharpe and Lil-
ley, 1980), whereas for GaAs, suitable electrolytes are
Tiron (dihydroxybenzene-3,5-disulfonic acid disodium
salt) and ethylenediaminetetraacetic acid (EDTA)0.2M
SEMICONDUCTOR PHOTOELECTROCHEMISTRY 15
NaOH (Blood, 1986). A variety of nonaqueous solutions
can also be used to perform differential capacitance
measurements, although more attention needs to be
paid to series resistance effects in such electrolytes.
For EPS, the electrolyte shouldbe transparent to light
ranging from the band gap energy down to ~E
g
. This
range permits the entire bandgapregionto be probed. In
addition, the solvent/electrolyte should be a good elec-
trical conductor. The electrolyte must also be chosen
such that the surface is stable both in the dark and
under illumination (Haak and Tench, 1984).
Experimental Setup. In a typical MottSchottky
experiment, the DC bias to the electrochemical cell is
delivered by a potentiostat, and the AC voltage signal is
supplied by an impedance analyzer that is interfaced to
the potentiostat. The DC potential applied to the
semiconductor junction, which is usually in the range
of 01V versus E(A/A

), is always in the direction of


reverse bias, since forward-bias conditions induce
Faradaic current to ow and effectively force the cell to
respond resistively instead of capacitively. The EPS
spectral features associated with bulk states generally
vary linearly with the spacecharge depletion width,
whereas those associated with surface or interface
states are relatively insensitive to the electrode
potential. Thus, variation of the electrode potential is
a powerful method to distinguish between bulk and
surface states. The electrode potential is also the
experimental parameter that determines the depth
into the semiconductor that is probed by the EPS
experiment. When measuring the effect of passivation
or varying the concentration of electrolyte on surface
state density, the bias voltage should be chosen such
that the semiconductor is strongly reverse biased. Since
bulk states are unchanged in response to surface
modication, the extent of surface state perturbation
induced by surface passivation or electrolyte
dissolution is deduced by the residuals between two
EPS spectra taken at large reverse biases prior to and
just after the surface modication procedure (Haak and
Tench, 1984).
The input AC signal, usually 510mV in amplitude,
must be small to ensure that the output signal behaves
linearly around the DC bias point and is not convoluted
with the electric eld in the semiconductor (Morri-
son, 1980). The voltage perturbation frequency used
in an EPS experiment should be chosen to yield a rela-
tively large value for the phase shift between the voltage
perturbation and the current response. Generally, a
phase shift of 70

90

is acceptable. However, to mini-


mize errors in the EPS spectrum, the semiconductor
solution system should show a limited range of fre-
quency dispersion. Large frequency dispersion is usu-
ally an indication of poor crystalline quality in the
semiconductor and will generally lead to irreproducible
EPS spectra (Haak and Tench, 1984).
The quantities that are measured by the impedance
analyzer are the total impedance Z
tot
andthe phase angle
y between Z
tot
and the input AC signal. Fromthese data,
the real component of the impedance Z
re
, which is
dened by Z
re
=Z
tot
cos y, and the imaginary component
of the impedance Z
im
, which is dened by Z
im
=Z
tot
siny,
can be calculated. For the simplied three-element
equivalent circuit illustrated in Figure 10, the depen-
dencies of Z
tot
versus ACsignal frequency (Bode plot) and
Z
im
versus Z
re
(Nyquist plot) ideally obey patterns that
not only permit extraction of the electrical elements in
the circuit but also indicate the frequency range over
which the cell acts capacitively. Of course, deviations
from these ideal patterns can allow evaluation of the
appropriateness of the theoretical equivalent circuit to
the description of the experimental cell.
Ananalysis of the resulting cell impedance andphase
angle in response to this sinusoidal perturbation yields
the value of the semiconductor differential capacitance
C
d
. This method, of course, requires that C
d
can be
accessed from the experimentally measured electrical
impedance behavior of a semiconductorliquid contact.
In general, an equivalent circuit is required to relate the
measuredimpedance valuestophysical propertiesof the
semiconductorliquid contact.
The conventional description of the equivalent circuit
for a semiconductorliquid junction can be represented
as inFigure 10a(Gerischer, 1975). The subscripts s refer
to the semiconductor bulk, sc to the space charge, ss to
surface states, H to the Helmholtz layer, and soln to
solution. This circuit is reasonable because there will
clearly be a capacitance and resistance for the space
charge region of the semiconductor. Because surface
states represent an alternative pathway for current ow
across the interface, C
ss
and R
ss
are in parallel with the
elements for the spacecharge region. The Helmholtz
elements are in series with both the spacecharge and
surface state components, since current ows through
the Helmholtz layer regardless of the pathway of current
ow through the semiconductor. Finally, the series
(a)
(b)
R
s
R
s
R
sc
R
sc
R
ss
R
soln
R
H
C
sc
C
sc
C
ss
C
H
Figure 10. (a) Circuit of a semiconductorliquid junction and
(b) a simplication of that circuit.
16 SEMICONDUCTOR PHOTOELECTROCHEMISTRY
resistance of the electrode and contact and the solution
series resistance are included as distinct resistive com-
ponents in this circuit.
A possible simplication that can often be justied
combines the resistances of the solution and of the
electrode into one series resistor (Figure 10b). If the
Helmholtz layer resistance is small, the RC portion of
the Helmholtz layer circuit impedance is dominated by
C
H
. Furthermore, C
H
is usually much larger than C
sc
.
Therefore, C
H
can typically be neglected, because the
smallest capacitance value for a set of capacitors con-
nected in series dominates the total capacitance of the
circuit. This simplication is not valid for highly doped
semiconductors where C
H
and C
SC
are comparable. In
addition, if the AC impedance measurement is per-
formed at a high frequency, such that surface states
do not accept or donate charge as rapidly as the space
charge region, thenthe elements associatedwithsurface
state chargingdischarging processes can also be
neglected. Alternatively, if the surface state density of
the semiconductor is low enough, the contributions of
C
ss
and R
ss
are negligible.
Prior to any EPS measurement, it is necessary to
record an optical transmission spectrumof the solution
in a cell with a path length equivalent to the distance
between the EPS cells optical window and the working
electrode to be illuminated. This procedure is necessary
to determine which features in the bias-independent
part of an EPS spectrum might result from solvent
absorption (Haak and Tench, 1984).
Alternatively, AC voltammetry can be performed
using a potentiostat in conjunction with a waveform
generator and a lock-in amplier. In this technique, the
magnitude of the AC component of the current (i.e., Z
tot
)
and the phase angle with respect to the input AC signal
are measured at a given frequency. The real and imag-
inary parts of the impedance can then be extracted as
described above. This powerful technique therefore
requires no additional capabilities beyond those that
are required for basic measurements of the behavior of
semiconductor photoelectrodes.
Data Analysis and Initial Interpretation
Atypical Bodeplot for asiliconelectrode incontact witha
redox-active solution is shown in Figure 11. At high
frequencies f of the input AC voltage signal, the capac-
itive reactance w
c
[i.e., the effective impedance of the
capacitor, given by w
c
=(2pfC
sc
)
1
] is small relative to
R
sc
, so most of the current ows through the C
sc
path-
way, and the observed impedance is therefore simply R
s
.
At lowfrequencies, w
c
is highrelative to R
sc
, somost of the
current ows throughthe R
sc
pathway, andthe observed
impedance is R
s
R
sc
. In the intermediate frequency
range, as is showninFigure 11, increments infrequency
translate directly into changes in w
c
, so the magnitude of
Z
tot
is dictated by C
sc
. The MottSchottky measurements
shouldbe performedinthis capacitive frequency regime,
which ideally has a slope of 1 in a Bode plot.
A Nyquist plot is distinguished from the Bode plot in
that it illustrates explicitly the branching between Z
re
and Z
im
as a function of frequency. A Nyqyist plot for a
silicon electrode in contact with a redox-active solution
is shown in Figure 12. At high frequency, Z
re
R
s
. As the
frequency is incrementally decreased, the contribution
of Z
im
rises until the magnitude of the capacitive reac-
tance is identical to the resistance of R
sc
. At this point,
equal current ows through the circuit elements C
sc
and
R
sc
. As the frequency is decreased further, the system
behaves increasingly resistively, until the impedance is
entirely dictated by R
sc
. In practice, since the frequency
maynot reachapoint wherethesystemispredominantly
resistive, it is generally easier to extract the value of R
sc
from a Nyquist plot than from a Bode plot. Circle-tting
algorithms can provide reasonable estimates of R
sc
if the
Nyquist plot possesses asufcient arcover the frequency
range explored experimentally.
Assuming the three-element equivalent circuit
depicted in Figure 10b,
C
sc
=
1

14Z
2
im
=R
2
sc
_
2oZ
im
(52)
where the angular frequency o=2pf. It is also assumed
that R
s
R
sc
, so that R
s
contributes negligibly to the
measured impedance. Typically, since R
s
is on the order
10
6
10
5
10
4
|
Z
|
,

10
3
10
2
10
2
10
3
Frequency (Hz)
10
4
10
5
Figure 11. Typical Bode plots takenat intermediate frequencies
where the impedance is dominated by capacitive circuit ele-
ments. The electrodesolutionsystemis identical tothat usedto
obtain the data in Figure 4. The open circles represent data
taken at 0.2V versus E(A/A

) and the lled circles represent


data taken at 0.8V versus E(A/A

).
SEMICONDUCTOR PHOTOELECTROCHEMISTRY 17
of 10
2
OandR
sc
is onthe order of 10
5
10
7
O, Equation52
gives an accurate value for C
sc
. Therefore, measurement
of Z
im
at various frequencies and DC biases, as well as
extraction of the R
sc
value at each DC bias from the
circular t of the Nyquist plot at that potential, allows
calculation of C
sc
. The subsequent compilation of C
2
sc
versus E for each frequency yields the MottSchottky
plot, as shown in Figure 13 for a typical siliconsolution
interface. Linear regression of these data points and
extrapolation of these linear ts to the x intercepts of
C
2
sc
= 0 gives the value of V
bi
for each frequency.
Ideal MottSchottky plots should be nondispersive
across all measured voltages and frequencies. Fig-
ure 13 shows a representative MottSchottky plot for
the n-Si/CH
3
OH system, which forms ideal junctions
with numerous redox compounds (Fajardo
et al., 1995; Fajardo and Lewis, 1997). As stated,
the difference between the closed and open circles is
an order-of-magnitude increase in the acceptor con-
centration, 10 versus 100mM, respectively. At each
concentration, data points are plotted at input AC
voltage frequencies of 100 and 2.5kHz. This gure
illustrates the quality of data required to achieve
quantitative results from MottSchottky plots. The
data are not only nondispersive across the frequency
ranges used but also somewhat nondispersive across
an order-of-magnitude change in acceptor
concentration.
After extrapolation, the open-circle data points had x
intercepts of 0.028 and 0.023V, respectively, a variance
of <5mV. After addition of the correction term to these
values to yield corresponding V
bi
values, Equation 43
can be applied and will yield barrier heights of 518 and
513mV, respectively. The variance in the barrier height
between the closed and open circles was measured to be
less than 63mV. It is critical to show the nondispersive
(linear) nature of the MottSchottky plots for a wide
range of frequencies. Should the MottSchottky plots
become dispersive, the data set should be recollected
using longer acquisition times. It is also crucial to have
short cables, all of a consistent length, and solid well-
tted connectors on all the interconnects between the
cell and potentiostat. To help prevent current and/or
ground loops from developing, the cables must be kept
straight. Figure 14 shows somewhat nonideal
MottSchottky plots for the p-Si/CH
3
CN-tetra(n-butyl)
ammonium perchlorate contact, taken at 5kHz both in
the absence of redox couples (open circles) and in the
presence of PhNO
2
(30mM) and PhNO
2

(10mM) (open
triangles) (Nagasubramanian et al., 1982). The slight
nonlinearity of these plots makes determination of the
at-band potential to within better than ~0.1V
problematic.
A typical EPS spectrum will exhibit plateaus and/or
peaks (Fig. 15). A plateau is observed in the spectrum
when a transition involves one of the bands of the semi-
conductor, because once a threshold energy for the
transitionhas beenreached, transitions either toor from
a bandcanproceedover a wide range of photonenergies.
In contrast, the EPS spectrumexhibits a peak whenever
localized state or impurity/defect atoms participate in
the charge injection process. These injections require
two-level transitions and thus are only observable over a
smaller range of illumination wavelengths (energy)
(Anderson 1982; Haak et al., 1982; Goodman
et al., 1984; Haak and Tench, 1984).
The energy of a localized state relative to a band
edge energy is given by the onset energy of the peak in
the EPS spectrum. When all the states at a given
illumination energy have been saturated, populated,
or depopulated, the spectrum will saturate with
respect to capacitance as the illumination intensity
30
25
20
15
0
0.2 0.4 0.6
Potential (V versus E(A/A

))
0.8
C
d

2

(
1
0
1
5

f

2
)
Figure 13. MottSchottky plots for the system described in
Figure 4. The data shown are for two different concentrations
of acceptor speciesandfor twodifferent acquisitionfrequencies,
100 and 2.5kHz.
70,000
60,000
50,000
40,000
30,000
20,000
10,000
0
0 20,000 40,000
Z
real
Z

i
m
a
g
i
n
a
r
y
60,000 80,000
Figure12. Nyquist plot for thesystemdescribedinFigure4. The
data were collected at an applied DC bias of 0.45V.
18 SEMICONDUCTOR PHOTOELECTROCHEMISTRY
is increased. The density of such states can thus be
measured from the magnitude of the capacitance at
saturation. Furthermore, by xing the intensity of the
light and scanning a range of energies around a
plateau or peak, the optical absorption cross section
for trap states can be measured from the rate of
change of the capacitance with time. The correspond-
ing emission rate can then be determined from the
rate of decay of the capacitance with time after the
interruption of the incident illumination (Haak and
Tench, 1984).
WithEPS, contributions to the measuredcapacitance
from bulk and surface states can be distinguished by
taking EPS measurements at constant illumination for
different bias voltages. The bias voltage determines the
thickness of the spacecharge layer and thus allows
measurement of the density of bulk states constituting
the spacecharge layer. The bias voltage has, however,
little inuence on the potential drop at the surface or in
the Helmholtz layer. The rate of change of the EPS
capacitance at a xed illumination energy and intensity
as a function of sample bias thus provides a measure of
the density of bulktrapstates as afunctionof the sample
thickness.
Problems. The most prevalent problem in
MottSchottky determinations is frequency dispersion
in the x intercepts of the linear regressions. Nonlinearity
in these plots may originate from various experimental
artifacts (Fajardo and Lewis, 1997). If the area of the
semiconductor is too small, edge effects may dominate
the observed impedance. The area of the counter
electrode must also be large relative to the working
electrode (as a rule of thumb, by a factor of 10) to
ensure that any electrochemical process occurring at
the counter electrode does not introduce additional
electrical elements into the equivalent circuit. In
addition, the redox component that is oxidized or
reduced at the counter electrode must have a
sufcient concentration in solution (roughly 5mM) or
this redox process may be manifested as an additional
resistance that is a major component of the total cell
impedance. Finally, the measuring resistor within the
potentiostat should be adjusted in response to the
magnitude of the currents passing through the cell in
order to obtain the most accurate impedance value
possible.
Other concerns with EPS solutions are surface deg-
radation and lm formation. While for most experimen-
tal EPS measurements, the effect of surface degradation
andlmformationcausedbythe accumulationof photo-
generated minority carriers at the solidliquid interface
is minimal, it does become a concernwhena sample is to
undergomeasurements over the course of hours or days.
To help counter the effect of surface degradation or lm
formation caused by the accumulation of photogener-
ated minority carriers at the solidliquid interface, an
additional redox species can be added to the electrolyte
in order to scavenge carriers that arrive at the semicon-
ductorliquidcontact. This process can, however, have a
detrimental effect if the redox species is present in high
concentration, because interfacial charge transfer
between the redox species and the semiconductor may
produce an overall dark current that can degrade the
sensitivity of the EPS experiment (Haak and
Tench, 1984).
1.5
1.4
1.3
1.2
1.1
1.0
0.9
0.8
800 1000
A
A
B
D
C
C
Valence band
D
i
f
f
e
r
e
n
t
i
a
l

c
a
p
a
c
i
t
a
n
c
e

(
A
r
b
i
t
r
a
r
y

U
n
i
t
s
)
Conduction band
D
B
1200
Wavelength (nm)
1400 1600 1800 2000
Figure 15. Electrochemical photocapacitancespectrum. Displayed
is the measured differential capacitance as a function of the
wavelength of the incident light. The inset illustrates various types
of electro-optical transitions that can be associated with the
observed plateaus and peaks in the EPS spectrum.
1200
a b
1000
800
600
400
C
d

2
(
m
F
/
c
m
2
)

2
200
0
0.2 0 0.2
E (V versus SCE)
0.4 0.6 0.8
Figure 14. MottSchottky plots of p-Si/CH
3
CN with tetrabuty-
lammonium phosphate (TBAP) electrolyte (a) in the absence of
redox couples and (b) with the redox couples PhNO
2
(30mM)
and PhNO
2

(10mM). Reprinted with permission from


Nagasubramanian et al., 1982.
SEMICONDUCTOR PHOTOELECTROCHEMISTRY 19
TRANSIENT GROWTH AND DECAY DYNAMICS OF
SEMICONDUCTORLIQUID CONTACTS
Principles of the Method
The rate of nonradiative recombination mediated
through surface states can be described under many
experimental conditions by the steady-state Shock-
leyReadHall (SRH) rate equation(Hall, 1952; Shockley
and Read, 1952; Schroder, 1990; Ahrenkiel and Lund-
strom, 1993):
R
SRH
(surface) =
n
s
p
s
n
2
i
(n
s
n
1;s
)=N
t;s
k
p;s
(p
s
p
1;s
)=N
t;s
k
n;s
(53)
where n
I
is the intrinsic electron concentration for
the semiconductor, N
t,s
denes the density of surface
states (in units of reciprocal centimeters squared), P
S
is
the surface hole concentration, and k
n,s
and k
p,s
are
the capture coefcients for electrons and holes, respec-
tively, by surface states. Each capture coefcient is
related to the product of the thermal velocity V of the
carrier in the solid and the capture cross section s
for each kinetic event such that (Many et al., 1965;
Kr uger et al., 1994a)
k
p;s
= v
p
s
p
and k
n;s
= v
n
s
n
(54)
The symbols n
1,s
and p
1,s
in Equation 53 represent the
surface concentrations of electrons in the conduction
band and holes in the valence band, respectively, when
the Fermi level is at the trap energy. Values for N
1,S
and
p
1,s
can be obtained through use of the principle of
detailed balance, which when applied to a system at
equilibrium in the dark yields (Many et al., 1965;
Blakemore, 1987)
n
1;s
= N
c
exp[(E
cb
E
t
)=kT[ (55a)
and
p
1;s
= N
v
exp[(E
t
E
vb
)=kT[ (55b)
where Et is the energy of the surface trapping state.
When other recombination processes are minimal
and the initial carrier concentration proles are uniform
throughout the sample, surface recombination should
dominate the entire decay properties of the sample.
Under these conditions, the observed minority carrier
decay dynamics are given by (Ahrenkiel and Lund-
strom, 1993)
Rate =
dDp
dt
= R
SRH
~
Dp
t
s;1
(56)
Furthermore, under such conditions, the fundamental
lament decay lifetime t
s,1
is given by (Schroder, 1990;
Ahrenkiel and Lundstrom, 1993)
t
s;1
= d=S
low
; S
low
= S
p
= N
t;s
k
p;s
(57)
where d is the thickness of the sample and S
LOW
is the
surface recombination velocity (in units of centimeters
per second) at low-level injection (Dn<N
d
and Dp<N
d
).
The other limiting situation is obtained under uni-
formilluminationat anintensity highenoughtoproduce
high-level injection conditions (i.e., Dn N
d
and
Dp N
d
). By using Equation 53 with injected carrier
concentrations Dn N
d
and Dp N
d
and, because
equal numbers of electrons and holes are created by the
optical injection pulse, taking Dn=Dp, the recombina-
tion rate is
Rate =
dDp
dt
= R
SRH
~
Dp
t
s;h
(58)
Now, however, the lament decay lifetime is (Blake-
more, 1987; Schroder, 1990)
t
s;h
=
d
S
high
; S
high
=
S
p
S
n
S
p
S
n
(59)
For conditions of k
p,s
=k
n,s
, S
high
= (1=2)S
low
, and the
surface recombination decay lifetime under high-level
injection, t
s,h
, is equal to 2t
s,l
.
In the presence of redox donors and acceptors in the
solution phase, measurement of the carrier decay
dynamics in the solid can also yield information on the
rates of interfacial charge transfer from carriers in the
semiconductor to redox species in the solution. Use of
this method to determine these important interfacial
kinetic parameters is described below.
The rate of disappearance of charge carriers at a
semiconductorliquid contact is given by the SRH model
of Equation 53, with additional terms required to
account for the interfacial charge transfer processes.
Thus, the boundary conditions that must be used in
evaluating the carrier concentrationdecay dynamics are
j
p
=q = k
ht
[A

[
s
(p
s
p
0;s
) k
p;s
N
t;s
[f
t;s
p
s
(1f
t;s
)p
1;s
[
(60)
and
j
n
=q = k
et
[A[
s
(n
s
n
0;s
) k
n;s
N
t;s
[(1f
t;s
)n
s
f
t;s
n
1;s
[
(61)
The values of j
p
and j
n
represent the current densities to
andfromthe valence andconductionband, respectively,
and f
t,s
is the fraction of surface states occupied with
electrons. Measurement of the decay dynamics in the
absence of, and then in the presence of, increasing
concentrations of redox species can therefore allow
determination of k
et
and k
ht
for the semiconductor
liquid contacts of interest.
Practical Aspects of the Method
The carrier concentration decays can be monitored
using a number of methods. Luminescence is a conve-
nient and popular probe for direct band gap
20 SEMICONDUCTOR PHOTOELECTROCHEMISTRY
semiconductors, while for materials like silicon, conduc-
tivity methods are often employed (for further discus-
sion, see CARRIER LIFETIME: FREE CARRIER ABSORPTION,
PHOTOCONDUCTIVITY, AND PHOTOLUMINESCENCE). The two meth-
ods are complementary in that the photoluminescence
signal decays when either the minority or majority car-
rier is captured, whereas conductivity signals are
weightedby the individual mobilities of the carrier types.
Thus, if all the holes in silicon were to be trapped, the
band gap luminescence signal would vanish, whereas
the conductivity signal wouldstill retain75%of its initial
amplitude (because of the carrier mobilities for silicon,
m
n
wem
p
). Either method can be used to probe the carrier
concentration dynamics, and both are sometimes used
onthe same sample to ensure internal consistency of the
methodology. A somewhat more specialized method of
monitoring the carrier concentrations has also been
recently developed using ohmic-selective contacts on
silicon, where the photovoltage developed betweenthese
contacts yields a probe of the carrier concentration
decay dynamics in the sample of concern (Tan
et al., 1994a).
One of the most reliable methods for monitoring
minority carrier decay dynamics is the time-resolved
photoluminescence (TRPL) method. The use of time-
correlated single-photon counting has added further
sensitivity to the TRPL technique and has allowed sub-
picosecond lifetime measurements.
To perform time-correlated single-photon TRPL, one
startswithpulsed, short-time monochromaticlaser light
tuned to an energy greater than the band gap energy of
the material under study. This laser source can be
produced by the output of a pulse-compressed Nd-
yttrium-aluminum garnet (YAG) pumped dye laser, if
picosecond timing resolution sufces, or from the fun-
damental mode of a solid-state Ti-sapphire laser, if even
shorter time resolution is desired. The timing character-
istics of the laser pulse canbe determinedthroughuse of
an autocorrelator. Any chirping in the light source must
be removed by passing the light through a few meters of
optical ber.
Prior to reaching the sample surface, the laser light is
split by means of a beam splitter (Fig. 16). The less
intense optical beam emanating from the beam splitter
is directed into a photodiode, while the more intense
beam is directed onto the semiconductor sample. The
photodiode and sample must be positioned at equal
distances fromthe beamsplitter to ensure correct timing
in the experiment. The voltage output of the photodiode
is directly connected to the START input of a time-to-
amplitude converter (TAC). Thus, with each laser pulse,
the sample is illuminated and the TAC is triggered on
(start). Both the beam diameter and the polarization of
the incident light are experimental parameters used for
controlling the light intensity, and thus the injection
level, of the semiconductor.
The light emitted by the sample due to radiative
recombination processes is collected and focused onto
a spectrometer that is tuned to the wavelength of the
transition to be monitored. The light output from this
spectrometer is then directed onto a single-photon
detector. When a single photon is detected, the resulting
voltage output producedbythe detector is usedtotrigger
the STOP input of the TAC. Once triggered into the off
position, the TAC will in turn produce a voltage pulse
whose magnitude is linearly dependent on the time
duration between the START and STOP signals. The
distribution of voltagepulse magnitudes is thus a
distribution in time of the photons emitted from the
sample after initial illumination. By allowing a com-
puter-controlledmultichannel analyzer (MCA) todisplay
a histogram, with respect to magnitude, of the distribu-
tionof the voltage pulses that are producedby the TAC, a
single-photon TRPL spectrum is obtained (Ahrenkiel
and Lundstrom, 1993; Kenyon et al., 1993).
To enhance the signal-to-noise ratio in a TRPL spec-
trum, it is necessary to place a pulse-height discrimina-
tor between the single-photon detector and the STOP
input of the TAC. This discriminator will ensure that
low-voltage pulses that arise from either electronic or
thermal noise and/or high-voltage pulses that can be
produced by multiphoton events do not accidentally
trigger the STOP signal on the TAC. To obtain the true
TRPLspectrumfor minoritycarriers, thisexperimentally
obtained TRPL spectrum must be deconvoluted with
respect to the system response function of the entire
apparatus. The systemresponse functionmust be deter-
mined independently for the setup of concern and is
typically measured by exciting a diffuse reector or
scattering solution instead of the semiconductor. A
variety of numerical methods allow generation of the
corrected photoluminescence decay curve from the
observed TRPL decay and the systemresponse function
(Love and Demas, 1984). Deconvolution is especially
Laser
Lens
Sample
Computer
Start Stop
TAC
MCA
Spectro-
meter
Single-
photon
detector
Discriminator
Photodiode
Beam
splitter
Figure 16. Time-resolved single-photon photoluminescence
spectrometer.
SEMICONDUCTOR PHOTOELECTROCHEMISTRY 21
important if the TRPL decay time is comparable to the
system response time.
The success of the single-photon counting method
is due to the development of detectors with sufcient
gain and time response that a single photon produced
by an electronhole recombination event can be
detected within a system response time of 30ps. Two
varieties of these detectors are photomultiplier tubes
(PMTs) and microchannel plates (MCPs). A limitation
of these detectors, however, is that only photon ener-
gies 1.1eV can be detected. Recently, advancements
in the timing resolution of single-photon avalanche
photodiodes (SPADs) have allowed detection of
photons in the near-infrared region and have raised
the possibility of performing TRPL with materials hav-
ing a band gap of <1.1eV, so that additional semi-
conductor materials can be explored using this
approach.
Data Analysis and Initial Interpretation
Once a TRPL spectrum is recorded and the effective
lifetime t
pl
has been measured, it is necessary to extract
the lifetime for surface recombination or any other dom-
inant recombination mechanism. The rate of radiative
band-to-bandemissionR
r
fromasemiconductor is given
by (Pankove, 1975; Ahrenkiel and Lundstrom, 1993;
Kr uger et al., 1994b)
R
r
= k
r
(n
p
n
2
i
) (62)
where k
r
is the rate constant for radiative recombination.
For low-level injection, Equation 62 reduces to
R
r
= K
r
[(N
d
Dn)(p
0
Dp)n
2
i
[ ~ k
r
N
d
Dp (63)
or
R
r
= d(Dp)=Dp = k
r
N
d
dt (64)
where p
0
is the equilibrium minority carrier concentra-
tion in the dark.
When nonradiative recombination mechanisms such
as surface recombination are active in parallel with
radiative recombination, the measured photolumines-
cence signal I
pl
(t) for a volume element in the solid con-
taining a spatially uniform density of excess minority
carriers is given by (Ahrenkiel and Lundstrom, 1993;
Kr uger et al., 1994a)
I
pl
(t) =
Dp
0
exp (t=t
pl
)
t
r;l
(photons=(cm
3
s
1
)) (65)
where Dp
0
is the initial density of injected minority
species after the laser excitation. The quantity t
pl
is the
measured effective total lifetime for the system, which is
determined from the experimental TRPL spectrum
(Fig. 17) and is expressed as
1
t
pl
=
1
t
s

1
t
r
(66)
where t
r
is the radiative lifetime and t
s
is the surface
recombinationlifetime. Depending onthe injectionlevel,
t
r
is denedas either (Many et al., 1965; Schroder, 1990)
t
r
=
1
k
r
N
d
; lowinjection (67)
or
t
r;h
=
1
k
r
Dp
0
; high injection (68)
Since k
r
, N
d
, and Dp
0
are all measurable parameters, the
decay time for a luminescence spectrum can be easily
deconvoluted through application of Equation 66 to
retrieve the surface recombination lifetime and the
resulting surface recombination velocity for a semicon-
ductor of known thickness.
Problems. One problemin TRPL experiments involves
the fraction of surface area of the semiconductor that is
illuminated by the laser beam. The illumination should
always be uniform across the surface. Frequently, the
injectionlevel is controlledby controlling the diameter to
which the laser light is focused. However, carrier
recombination kinetics in the dark differ from those in
the light. With a nonuniformly illuminated surface,
photogenerated carriers can diffuse into the
nonilluminated regions of the semiconductor. The net
recombination lifetime measured will thus be a
convolution of lifetimes for processes occurring in the
illuminated and nonilluminated regions of the
semiconductor. This is especially of concern with
beam spots ~10mm in diameter. In such situations
one cannot make a quantitative assessment for the
carrier dynamic lifetimes of an illuminated
10
4
10
3
10
2
10
1
0 50 100
Time (ns)
I
n
t
e
n
s
i
t
y

(
n
o
r
m
a
l
i
z
e
d
)
150 200
Figure 17. Fitting example of a time-resolved photolumines-
cence decay.
22 SEMICONDUCTOR PHOTOELECTROCHEMISTRY
semiconductor because the majority of the
recombination occurs in the unilluminated regions of
the sample. Inessence, the desired lifetime is affected by
background recombination lifetimes occurring in the
dark.
The repetitionrate of modernlasers is also of concern,
as is their intensity. The repetitionrate shouldalways be
set such that all carrier concentrations return to their
equilibrium dark values prior to the subsequent laser
pulse. Otherwise, the injection levels and hence recom-
bination lifetimes will vary from pulse to pulse. For a
semiconductor in accumulation, the time to restore
equilibriumcan be as large as a few milliseconds. Thus,
the repetitionrate shouldbe regulatedas the bias across
the semiconductor is varied. One method of increasing
the time between subsequent pulses is to use a Kerr
shutter as a pulse gate.
Caution should also be taken to ensure that the laser
intensity does not induce sample damage. The following
protocols should always be followed to ensure that life-
times derived fromTRPL experiments are reliable. First,
MottSchottky measurements should be performed to
extract the at-band potential and barrier height. Sec-
ond, JEcurves should be acquired both inthe dark and
under the same light intensity conditions that will be
used in the TRPL experiments. Finally, after performing
the TRPL experiments, the MottSchottky and JE
curves shouldbe redeterminedto verify that the integrity
of the sample has not been compromised during the
TRPL measurement.
Use of the method is limited to a rather narrow range
of potentials, light intensities, and interfacial charge
transfer rate constants. In fact, recent simulations show
that the photoluminescence decays are generally insen-
sitive to the value of the minority carrier charge transfer
rate constant k
ht
for an n-type semiconductor. Instead,
diffusional spreading and drift-induced separation of
photogenerated carriers in the spacecharge layer of the
semiconductor dominate the time decay of the observed
luminescence signal under most experimentally acces-
sible conditions (Kr uger et al., 1994).
This behavior can be readily understood. The laser
excitation pulse produces an exponential concentration
prole of injected carriers, so the luminescence is ini-
tially very intense because it involves bimolecular
recombination between electrons and holes that are
both created in the same region of space in the solid
(Many et al., 1965; Ahrenkiel and Lundstrom, 1993).
Diffusional motion of carriers tends to alleviate this
carrier concentration gradient and therefore will even-
tually eliminate the quadratic recombination to instead
produce a simple, exponentially decaying, photolumi-
nescence signal. In the limit of negligible nonradiative
recombination, the long-lifetime photoluminescence
decay asymptote is thus dictated by the carrier mobili-
ties, the optical penetration depth of the light, and the
carrier injection level used in the experiment.
The short-lifetime photoluminescence decay asymp-
tote can also be readily understood. When carriers are
removed nonradiatively from the solid as fast as they
arrive at the solidliquid interface, either by surface
recombination or by interfacial charge transfer pro-
cesses (or a combination thereof), the photolumines-
cence decay will reach a short-lifetime asymptote. Any
additional increases ink
ht
or N
t,s
cannot affect the photo-
luminescence decay dynamics because the carrier
quenching rate is already limited by diffusion of carriers
to the surface. Thus, like the long-lifetime photolumi-
nescence decay asymptote, the short-lifetime photolu-
minescence decay asymptote is dictated only by the
carrier mobilities, the optical penetration depth of
the light, and the carrier injection level used in the
experiment.
Due to these competing processes, generally, values
of k
ht
andof the minority carrier low-level surface recom-
bination velocity S
p
can be obtained from an analysis of
the photoluminescence decays only when the following
restricted sets of conditions are satised simulta-
neously: 10
1
cm/s _S
p
_10
4
cm/s, 10
18
cm
4
/s_k
ht
_
10
15
cm
4
/s, andthe electrode potential Eis inthe region
of 0<E< 0.15V relative to the at-band potential of
the n-type semiconductorliquid interface. The simula-
tions further showthat it is not possible to extract a eld
dependence of the charge transfer rate constant when
the semiconductorliquid contact is maintained in
reverse bias (E_ 0.15V versus the at-band potential)
and is subjected to light pulses that produce low or
moderate carrier injection levels, because under such
conditions the photoluminescence decay dynamics are
dominated by drift-induced charge separation in the
spacecharge layer of the semiconductor. Under high-
level injection conditions, no eld dependence can be
observed because the majority of the photoluminescence
decay dynamics occur near the at-band condition, so
the value of the band bending in the semiconductor
under dark, equilibrium conditions has negligible inu-
ence onthe luminescence transients producedby a high-
intensity laser pulse.
A full discussion of the regimes in which the interfa-
cial charge transfer kinetics canbe accessedexperimen-
tally using this method is beyond the scope of this work,
and the reader is referred to the recent literature for an
extensive evaluation of the limitations of this method for
such purposes. However, it is useful for measurement of
the surface recombination velocity of solidliquid con-
tacts in certain specic cases.
The last concernfor interfacial kinetic measurements
is to ensure that electrode corrosion, passivation,
adsorption, and other surface-related processes are not
signicant for the systemof concern, because otherwise
interpretation of the carrier decay dynamics is problem-
atic. The rate constants should be rst order in the
acceptor concentration, or there is no proof that the
kinetic process of interest is rate limiting. One method
for minimally establishing whether surface damage has
occurred or not is to collect JEdata for the systeminthe
electrolyte of concern and in another electrolyte solution
of interest both before and after the photoluminescence
decay experiments to ensure that no surface chemistry
changes are present that would affect the electrical
behavior of the solidliquid contact. This is necessary,
but not sufcient, to secure a robust rate constant
SEMICONDUCTOR PHOTOELECTROCHEMISTRY 23
determinationandisoftenhelpful inrulingout measure-
ments on clearly unstable electrode surfaces. The TRPL
data should always be buttressed by measurements at
varying concentrations of acceptor in the solution in
order to establish whether or not the expected rate law
has been observed in the system.
MEASUREMENT OF SURFACE RECOMBINATION VELOCITY
USING TIME-RESOLVED MICROWAVE CONDUCTIVITY
Principles of the Method
Another contactless methodfor measuring the lifetime of
minority carriers after an initial excitation pulse is time-
resolved microwave conductivity (TRMC) (for further
discussion, see CARRIER LIFETIME: FREE CARRIER ABSORPTION,
PHOTOCONDUCTIVITY, AND PHOTOLUMINESCENCE). The change in
reected microwave power, DR
M
(t), is proportional to the
change in the conductivity of the semiconductor. This
conductivity change is, to a rst-order approximation, a
measure of the time dependence of excess minority
carriers in the semiconductor (Naber and
Snowden, 1969; Chen, 1988; Forbes and Lewis, 1990;
Schroder, 1990; Ramakrishna and Rangarajan, 1995):
DR
M
(t) Ds(t) = q(m
n
Dn m
p
Dp)
DR
M
(t) Ds(t) = q(m
n
m
p
)Dp (under low-level injection)
(69)
Practical Aspects of the Method
To implement the method, either the semiconductor
under study is connectedto amicrowave source through
a radio frequency (RF) coupler, whichconsists of a series
of coils connectedto the base of the sample mount, or the
sample is connected to the source using a waveguide.
The waveguide system offers numerous advantages.
Since the sample is mounted inside the waveguide and
is illuminated colinearly with the microwave radiation,
better protection from electromagnetic interference and
better-dened and more uniformmicrowave radiation is
available.
A microwave experimental setup consists of four
primary components (Fig. 18): (1) the source, which con-
sists of a microwave generator, anisolator, and an atten-
uator; (2) the sample, which is mounted onto the end of a
microwave waveguide; (3) the detector, which by recti-
cation creates a voltage across a load resistor, with the
voltagebeingproportional totheamplitudeofthereected
microwavepower; and(4) thepower meter (ashort-circuit
plate consisting of silver-plated brass), which is used to
measure the power fromthe source and is mounted on a
separate waveguide at anequal distance fromthe source
and the sample. These four components are positioned
around a microwave circulator (Schroder, 1990).
Microwaves from the source, usually ka radiation
from a Gunn diode operating at ~10GHz, are directed
via the circulator to the waveguide systemthat contains
the sample. The reected microwaves are in turn
directed by the circulator to the detector. Most commer-
cially available circulators allow timing resolutions in
the subnanosecond regime. To measure the source
power, the circulator directs microwaves to the short-
circuit plate rather than to the sample. During the
experiment, the sample is initially illuminated by a laser
pulse or strobe light, whichalso triggers anoscilloscope.
The reected microwaves are amplied and measured,
and the corresponding signal is displayed on the oscil-
loscope as a function of time fromthis illumination. The
reected microwaves not only emanate only from the
surface of the semiconductor but also penetrate a skin
depth into the sample. For silicon at 10GHz, this skin
depth is 350mm for a resistivity of 0.5O-cm and is
2200mmfor a resistivity of 10O-cm. Once the microwave
spectrum is collected by a digital oscilloscope, the life-
time can be measured directly.
Sample Preparation. Measurements of bulk or surface
parameters require two different sample preparation
procedures. If one is interested in bulk recombination
lifetimes, thenthe samples shouldbe made progressively
thicker, on the order of 1mm, as the bulk lifetime to be
measured becomes longer, and the surface recombi-
nation velocity should be made as small as possible.
For measurement of the surface recombination velocity,
the sample should be of the thickness typically used in
wafers, or on the order of 0.1mm. The sample should
also be polished on both ends to ensure consistent
transmission and reectivity of the microwaves. The
electrolyte must also not be too conductive or the
radiation signal will be greatly attenuated.
Microwave conductivity experiments permit a great
deal of versatility as to the sample geometry that can be
used. Of concern, however, is what percentage of the
Amplifier
and
detector
Oscilloscope
Light source
Waveguide
Sample
Power
meter
Microwave
source
Isolator
and
attenuator
Circulator
800
Figure 18. Microwave reectance spectrometer.
24 SEMICONDUCTOR PHOTOELECTROCHEMISTRY
microwave decay signal recorded is indeed that of the
semiconductor and not that of the measurement appa-
ratus. When the geometry of a sample/apparatus is off-
resonance, it has beenfoundthat the systemresponse is
very fast, while an on-resonance cavity results in a large
increase in the system fall time. One method of deter-
mining the extent of resonance in microwave experi-
ments is to use etched silicon wafers as calibration
templates. It has been determined that Si wafers etched
in HF:HNO
3
:CH
3
COOH solutions result in a surface
recombinationvelocityas large as 10
5
cm/s. Conversely,
it has been shown that immersing the sample in HF
during decay measurements gives a measured surface
recombination velocity as low as 0.25cm/s. Both these
protocols are useful when calibrating ones instrument
and/or samples.
Data Analysis and Initial Interpretation
Figure 19 displays the microwave conductivity decay for
n-Si (solidline) incontact withconcentratedsulfuric acid
under high-level injection conditions (10
15
injected car-
riers per square centimeter in a 190mm thick sample)
(Forbes and Lewis, 1990). The data can be t using a
simulationpackage that takes intoaccount all aspects of
drift-diffusion, generation, recombination, and charge
transfer for photogenerated carriers ina semiconductor.
Since most bulk parameters for common semiconduc-
tors are known, simulations are performed by paramet-
rically varying an unknown parameter, such as the
surface recombinationvelocity, until a t witha minimal
residual is obtained. Such a simulated t is shown in
Figure 19(dotted line) for a surface recombinationveloc-
ity of 25cm/s and a bulk lifetime of 2ms.
If no simulation package is available, a spectrumcan
be t with a biexponential of the form
I(t) = A
1
exp(t=t
1
) A
2
exp(t=t
2
) (70)
The t is performed from the time of peak conductivity
intensity to the time at whichthe conductivity signal has
decayed to 1% of its peak value. Subsequently, the
resultant t can be described with an effective lifetime
value t), computed using the relationship
t ) =
A
1
t
1
A
2
t
2
A
1
A
2
(71)
where A
1
and A
2
are the preexponential factors and t
1
and t
2
are the exponential decay times of the t (Equa-
tion 70). This effective lifetime has terms that are linear
rather than quadratic with respect to the decay times t
n
.
Equation 71 hence does not bias a t) value toward the
longer of the two decay times derived from the tting
procedure. The above-mentioned tting procedure is
also applicable to all time-dependent decay phenomena,
such as radiative recombination, discussed under Sec-
tion Transient Growth and Decay Dynamics of Semi-
conductorLiquid Contacts.
LASER SPOT SCANNING METHODS AT
SEMICONDUCTORLIQUID CONTACTS
Principles of the Method
Laser spot scanning (LSS) involves recording the photo-
current while a highly focused laser spot is scanned over
asemiconductor surface. The technique has seenbroad-
ened appeal since the early 1990s, when higher laser
uences and spot sizes on the order of 3mm became
available throughuseof ber optics. Laser spot scanning
offers the opportunity to monitor surface recombination
lifetimes as a function of varying treatments to the semi-
conductor surface as well as other local parameters of
semiconductorliquid contacts (Mathian et al., 1985;
Carlsson and Homstr om, 1988; Carlsson et al., 1988;
Eriksson et al., 1991).
Four steps are thought to be involved in producing
LSS spectra. Assuming a circular laser spot of radius r,
the rst step of an LSS experiment involves the gener-
ation of minority carriers by the laser spot (Eriksson
et al., 1991). The ux of photogenerated carriers is equal
to (Eriksson et al., 1991)
F(l) = G
0
(r)2prdr (72)
where G
0
(r) is the BeerLambert generation function.
Carriers are thus generated in a cylinder whose radius
is equal to the radius of the laser spot andwhose depthis
a
1
, where a is the absorption coefcient. Generally, the
illuminationwavelengthis chosensuchthat a
1
is larger
than the depletion width.
Due to the strong electric elds present in the deple-
tion region, minority carriers generated in the depletion
region should migrate rapidly, with little sideways
Time (ms)
500
400
300
200
100
C
h
a
n
g
e

i
n

r
e
f
l
e
c
t
e
d

p
o
w
e
r

(
A
r
b
i
t
r
a
r
y

U
n
i
t
s
)
0
0 1 2 3 4 5
Figure 19. Microwave conductivity decay (solidline) under high
injection(10
15
carriers/cm
2
) for a190mmthicksampleof n-Si in
contact with concentrated sulfuric acid. A simulation with a
surface recombinationvelocity of 25cm/s and a bulk lifetime of
2 ms is shown (dotted line) for comparison. Reprinted with
permission from Forbes and Lewis, 1990.
SEMICONDUCTOR PHOTOELECTROCHEMISTRY 25
diffusion, to the solidliquid interface. However, minor-
ity carriers that are created belowthe depletionlayer will
diffuse laterally and either contribute a background
photocurrent or undergo nonradiative recombination.
The extent to which these carriers contribute to the
detected photocurrent depends on their lifetime in the
semiconductor bulk. The shorter the lifetime, the less
prominent their diffusion, and hence the smaller the
detected photocurrent.
The minority carriers generated in the depletion
region are thus the primary source of the detected pho-
tocurrent. The equations for the diffusion currents of
these carriers are (Eriksson et al., 1991)
Diffusion into spot = J
D
(r)2prdr (73)
Diffusion out of spot = J
D
(r dr)2prdr (74)
where J
D
(r) is the diffusionux of minority carriers along
the surface. Through Ficks second law, one obtains
J
D
(r)J
D
(r dr) =
Dr
2
P
s
(r)
dr
2
(75)
where D now denes the minority (hole for an n-type
semiconductor) carrier diffusioncoefcient. The value of
D is related to the minority carrier mobility m
p
by m =
Dq/kT, and p
s
(r) is the radial dependence of the minority
carrier concentration measured from the center of the
laser spot (Sze, 1981; Carlsson and Homstrom, 1988;
Eriksson et al., 1991).
Experimentally, LSS spectra record the change in
photocurrent as a function of position on the sample.
The recorded photocurrent becomes lower in regions
where surface recombination is rapid. Thus, the LSS
current is expected to decrease in the vicinity of step
edges or grain boundaries, regions of polycrystalline
growth, or other areas of rapid surface recombination.
This lateral prole of surface recombination is not avail-
able directly in a traditional spectral response experi-
ment, which does not use a tightly focused source of
illumination (Carlsson and Homstr om, 1988; Carlsson
et al., 1988; Eriksson et al., 1991).
Unlike EPS, the lifetime of minority carriers in LSS is
determined by two processes, surface recombination
and charge transfer:
t
1
s;l
p
s
(r)2prdr; surface recombination (76)
k
ht
p
s
(r)2prdr; charge transf er (77)
where k
ht
is the minority carrier charge transfer rate
constant. Recent modeling has shown that the radial
dependence for the concentration of minority carriers
relative to their concentration at the laser spots perim-
eter r
0
is (Eriksson et al., 1991)
p
s
(r)
p
s
(r
0
)
= exp 1k
ht
t
s
( )
1=2
Dr
L
_ _
(78)
where Dr =r r
0
and L is the diffusion length, which,
when only surface recombination processes are
considered, is dened as L=(t
s
D)
1/2
(Many
et al., 1965). If we further dene r
eff
as the radial distance
where the minority carrier concentration is 1/e of its
value at the laser spots perimeter, it can be shown that
r
eff
L
= 1k
ht
t
s
( )
1=2
(79)
Thus, when the surface recombination lifetime is short
or the charge transfer to the electrolyte is slow, r
eff
=L. In
contrast, when either surface recombination or the
charge transfer process is fast, r
eff
=0. This latter situ-
ationcorresponds to the limit in whichessentially all the
minority carriers have been contained and consumed
within the laser spot area (Eriksson et al., 1991).
Clearly, the deciency of LSS is that one cannot
distinguish the contribution of surface recombination
fromthe effect of charge transfer kinetics on the induced
photocurrent. Thus, to obtain quantiable surface
recombination lifetimes, either the interfacial charge
transfer rate must be slowor the exact magnitude of the
charge transfer rate constant must be available inde-
pendently. Nevertheless, recent experiments with LSS
have shown the extent to which layers and step edges
inuence the combined surface lifetime of minority car-
riers. For an InSe sample, LSS has shown that a step
edge, which presents a barrier against lateral diffusion
and hence allows carriers to locally accumulate on a
surface, can increase the rate of surface recombination
by nearly 20-fold (Mathian et al., 1985; Eriksson
et al., 1991). Further experiments on polycrystalline
silicon have shown that the recorded change in light-
induced photocurrent as a function of distance from a
step or grain boundary, l, can be approximated as
(Mathian et al., 1985; Eriksson et al., 1991)
I
ph
(l) = k
LSS
L
2D
1
1k
ht
t
s
L=2D
_ _
exp
l
L
_ _
(80)
where k
LSS
is a constant related to the effective circuit
used in the LSS experiments (Mathian et al., 1985;
Eriksson et al., 1991).
Practical Aspects of the Method
One aspect of the appeal of LSSis the degree of simplicity
associated with the experimental setup. The sample or
electrochemical cell is placed on a precision microma-
nipulator that is equipped with motorized control of the
sample position. For greater control of the sample posi-
tion at the expense of the magnitude of the motion, the
sample can be placed onto a piezoelectric mount. Alter-
natively, aswithscanning probemicroscopy techniques,
the sample can be held in a xed position and a ber-
optic light source canbe mountedonapiezoelectric drive
to provide a means of rastering the light source over the
sample (Mathian et al., 1985; Carlsson and
Homstr om, 1988; Carlsson et al., 1988; Eriksson
et al., 1991).
The mounted sample requires only an ohmic back
contact such that reverse-bias conditions can be
26 SEMICONDUCTOR PHOTOELECTROCHEMISTRY
achieved and photocurrents measured. The light source
in LSS must be monochromatic, but inexpensive light
sources such as a HeNe laser sufce for experiments
with most semiconductors. The only excitation require-
ment is that the penetration depth a
1
at the excitation
wavelength has to be larger than the depletion width of
the semiconductor under reverse bias.
To obtain the LSS spectra, the laser light is passed
into an optical ber and focused onto the sample sur-
face. In electrochemical cells, the ber optic can be
embedded through the cell case. The photocurrent data
are then collected as a function of laser spot position for
either various ambient atmospheres or electrolyte con-
centrations (Carlsson and Homstr om, 1988; Carlsson
et al., 1988; Eriksson et al., 1991).
Ina fashionanalogous to displaying scanning tunnel-
ing microscopy (SCANNING TUNNELING MICROSCOPY) or atomic
force microscopy (AFM) data, the measured photocur-
rents obtained while rastering the laser spot across the
sample canbe represented by shades of color. Usually, a
linear shading of 256 colors is associated with the range
of measured photocurrents. If, as a function of sample
position, the measured photocurrent is represented by
its color shades, thenone will obtainanLSStopographic
spectrumof a sample surface. To increase the signal-to-
noise ratio of an LSS spectrum or topography scan, the
laser light canbe choppedprior toentering the ber optic
and a lock-in amplier can then be used to record and
average the photocurrent that is synchronous with the
chopping rate. With respect to apparatus, LSS experi-
mentation is therefore very similar to spectral response
equipment (Mathian et al., 1985; Carlsson et al., 1988;
Eriksson et al., 1991).
When high-resolution LSS spectra or topographies
are required, the intensity of the light prior to entering
the ber optics is measured and regulated by means of a
beam splitter, a PMT, and a series of computer-con-
trolled light attenuators. While the spatial resolution
of most LSS surface topography is limited by the spatial
prole of the laser spot, the contrast or sensitivity of the
method to topography variations is determined by var-
iations in the intensity of the laser beam.
By bundling a series of ber optics around the central
light-emitting optical ber, one can measure the inten-
sity of light reected from the sample surface. This in
turn provides another channel of data on the reective
characteristics, and hence electronic characteristics, of
the surface as a function of the photovoltage or the
surface recombinationprocesses (Erikssonet al., 1991).
Problems. As stated above, the deciency of LSS is
that one cannot distinguish the contribution of surface
recombination fromthe effect of charge transfer kinetics
on the induced photocurrent. Thus, the interfacial
charge transfer rate for the system under study must
be slowor the exact magnitude of the charge transfer rate
constant must be available independently. One,
mechanical, concern is that the ohmic contact needs
to remain intact during LSS experiments. Hence,
corrosive systems such as Si-HF, which are accessible
with time-resolved methods, are not practical for LSS
experiments. Another concern is ensuring that the
samples dopant density is sufciently small so that
the solution-induced depletion width is <a
1
of the
rastered light beam. Randomly increasing the dopant
density can, however, lead to confounding problems.
Since under the LSS method one cannot determine if a
variation in the photocurrent results from surface
defects or from defects buried within the depletion
region, changes in the photocurrent arising from the
dopant atoms and/or buried defect sites can easily be
attributed to poor surface qualities. One method of
deconvoluting buried effects from surface effects is to
overlay an LSS spectrum onto a conventional
topographic image of the surface, such as those
obtainable from AFM or transmission electron
microscopy (TEM), and to correlate variations in the
LSS signal with visible topographic features. For
moderately doped semiconductors, however, the
inuences of buried defects are minimal.
INTENSITY-MODULATED PHOTOCURRENT AND
PHOTOVOLTAGE SPECTROSCOPY
Principles of the Method
Analogous to differential capacitance measurements
where an AC voltage signal is applied on top of a DC
voltage, IMPS or IMVS involves the response of a cell to a
sinusoidally modulated light source. In fact, the mea-
surements from IMPS and IMVS directly relate to the
solar cell impedance measured by differential capaci-
tance measurements through
Z
tot
=
H
IMVS
H
IMPS
(81)
where H are the IMVS and IMPS transfer functions
dened as
H
IMPS
=
I
AC
(o; t)
qG
AC
(o; t)
(82)
H
IMVS
=
V
AC
(o; t)
qG
AC
(o; t)
(83)
(Halme, 2011; Schefold, 1992).
In the transfer functions as dened, I
AC
(o,t) and
V
AC
(o,t) are the modulated photocurrent and photovol-
tage responses. The advantage of IMPS and IMVS over
differential capacitance measurements is that differen-
tial capacitance measures the total electrode impedance
with contributions from both the space charge and the
double layer included, while intensity-modulated meth-
ods provide a direct probe for surface states at the
semiconductorliquid interface.
IMVS and IMPS have been utilized as complementary
techniques for the determination of the minority carrier
diffusion length. IMVS when applied under open-circuit
conditions yields the effective charge carrier lifetime and
SEMICONDUCTOR PHOTOELECTROCHEMISTRY 27
IMPS can be utilized to determine the electron diffusion
coefcient.
L
n

D
n
t
_
(84)
Although applicable to a wide variety of semiconductor
materials, the techniques have been widely utilized for
the characterization of dye-sensitized solar cells
(Peter, 1990).
The derivation of the applicable photocurrent equa-
tions has been completed for cells under various condi-
tions (Li and Peter, 1985; Peat and Peter, 1987;
Peter, 1990; Ponomarev and Peter, 1995). Here we con-
sider the simple case where photogenerated minority
carriers, holes, can transfer across the interface to a
solutionspecies or canrecombine withelectrons directly
with the rate constants, k
1
and k
2
, respectively. In this
case the capacitance due to surface states is considered
to be negligible. The change in photogenerated charge
across the surface charge layer (Q
sc
, C
sc
) and the double
layer (Q
dl
, C
dl
) can be represented by the following dif-
ferential equations:
dQ
sc
dt
= I
0
(1e
iot
)jk
2
(Q
sc
Q
dl
) (85)
dQ
dl
dt
= k
1
(Q
sc
Q
dl
)j (86)
where o is the angular frequency of light modulation, j
is the current through the external circuit, and I
0
is
the amplitude of the current toward the semiconductor
liquid interface. As can be expected with the addition
of other recombination pathways (more complicated
charge transfer mechanisms), Equations 85 and 86
will, in turn, include additional rate terms for these
reactions.
The differential equations above can be solved under
potential control, where Q
sc
/C
sc
Q
H
/C
H
=jR, where R
is the total resistance of the cell, to yield
j = I
0
k
1
io(C=C
sc
)
k
1
k
2
io
1
1iot
_ _
e
iot
I
0
k
1
k
1
k
2
(87)
where C=(C
sc
C
dl
)/(C
sc
C
dl
) and t =RC, the time con-
stant of the cell.
The corresponding relations for photovoltage mea-
surements have also been derived for a variety of con-
ditions (Schlichth orl et al., 1997; Ponomarev and
Peter, 1995; Kamieniecki, 1983). Under galvanastatic
control, the photopotential across the cell is simply V
(o) =j(o)R. Therefore, fromEquation 6, we get a relation
for the photopotential of
oV(o) =
I
0
C
k
1
io(C=C
sc
)
k
1
k
2
io
(88)
Practical Aspects of the Method
Atypical experimental setupisdisplayedinFigure 20. As
shown, a light source is modulated by the use of an
acousto-optic modulator controlled by a frequency
response analyzer (FRA). There have been many light
sources appliedto IMPSandIMVSspectroscopy, includ-
ing CW lasers, arc lamps, and LEDs. Common CW
lasers utilized are heliumneon (632.8nm) or helium
cadmium lasers (442nm). Arc lamps although applica-
ble provide a smaller range of accessible frequencies.
The amplitude of modulation is typically between 10%
and 50% of the steady-state illumination intensity and
Laser OOM
PD
CPU FRA
Potentiostat
Semiconductor electrode
R
e
f
C
J
ph
I
0
Figure 20. IMVS/IMPS experimental setup consisting of a CWlight source (pictured as a laser)
that is incident on an acousto-optic modulator (OOM) to provide an AC light component. The
light is then incident on a beam splitter that splits the light into a photodiode (PD) and the
semiconductor electrode within an electrochemical cell (reference electrode (REF), counter
electrode (C)) connected to a potentiostat. The data are collected by a frequency response
analyzer (FRA) and manipulated on a computer (CPU).
28 SEMICONDUCTOR PHOTOELECTROCHEMISTRY
the frequency of modulation is ~1Hz100kHz. The
beam is then split into a photodiode and the sample.
IMPS experiments are conducted under potentiostatic
control and IMVS experiments are conducted under
galvanostatic control. Alternatively, IMVS measure-
ments may be carried out at open-circuit conditions.
The signals from the potentiostat (connected to the
photoelectrochemical cell) and the photodiode are col-
lected at the FRA, which measures the complex ratio of
the AC photocurrent to the AC illumination signal.
Although applicable, the use of an equivalent circuit
to describe the recorded complex photocurrent and
photovoltage responses as applied in conventional dif-
ferential capacitance measurements is not needed. In
fact, the differential equations derived above are used to
analyze the data. In general, it is more chemically rele-
vant to leave the differential equation in the form of rate
constants than to convert the information into resistors,
which only have meaning when coupled to real capaci-
tances. However, an equivalent circuit can be applied as
is implied in Equation 81.
Data Analysis and Interpretation
A Nyquist plot of IMPS data for the photocurrent
response of TiO
2
dye-sensitized solar cell is shown in
Figure 21 (Franco et al., 1999). The high-frequency
intercept of data collected at short circuit corresponds
to the ux of minority carriers through the cell in the
absence of recombination as the collection efciency
nears unity. Therefore, according to Equation 31, a plot
of the natural log of the high-frequency intercept versus
the applied bias should result in a straight line from
whichthediffusionlengthandabsorptioncoefcient can
be solved for directly (Peat and Peter, 1987). An alterna-
tive approach for defect-rich materials is to calculate the
effective electron diffusion coefcient D
n
from the fre-
quency minimum via the following relation:
t
IMPS
=
d
s
gD
n
(89)
where the factor g depends on the direction of illumina-
tion, absorption coefcient, and lm thickness d.
The effective minority carrier lifetime can be esti-
mated by the frequency minimum of the IMVS
Nyquist plot collected at open circuit by Equation 90
(Guill en, 2011). Open-circuit conditions ensure that
the only component to the measured lifetime is
recombination as no current is collected in the exter-
nal circuit.
t
IMVS
=
1
o
min
(90)
This is expected from Equation 88, which dictates a
semicircle with a maximumof k
1
k
2
. In addition, from
the high-frequency limit of the IMVS complex plane plot,
it is possible to calculate the spacecharge capacitance.
Again, from Equation 88, we see that as the frequency
approaches innity, the functionapproaches I
0
/C
sc
. If I
0
is known, C
sc
can be calculated. This is an advantage
over conventional impedance measurements where the
contributions from C
sc
and C
H
cannot be distinguished
as the AC potential perturbation is applied to both cell
components in series.
It is important to note that both the IMPS and IMVS
complex plots may exhibit more than one semicircle
maxima. This is determined by the ratio of the involved
rate constants for charge transfer. For example, for the
case where you have two charge transfer pathways,
recombinationandinterfacial charge transfer, if the time
constant for the overall cell is much less than the time
constant associated with these processes, 1/(k
1
k
2
),
then the photocurrent response plot will exhibit two
semicircles as a function of o.
Finally, to use IMPS and IMVS data to calculate the
diffusion length as in Equation 84, it is imperative that
the data be corrected for the difference in the quasi-
Fermi levels between the conditions of the experiment,
for example, short-circuit conditions versus open-cir-
cuit conditions. It has been shown that the difference
between the QFL at short circuit versus that at the open
circuit is typically 200meV.
LASER-INDUCED PHOTOVOLTAGE TRANSIENTS
Principles of the Method
The photovoltage risetime method was originally intro-
duced in 2006 by ORegan et al. The method has proven
useful inquantifyingboththe effective diffusionlengthof
electrons and the electron lifetime for nanoparticulate
lms. Inthe method, the photovoltage rise anddecay of a
nanoparticulate semiconductor lmis measured after a
short laser pulse is incident on the sample. The time
constant for the rise canbe relatedto the charge transfer
resistance andthe capacitance of the substrate onwhich
the semiconductor particles are deposited (ORegan
et al., 2006).
t
rise
= R
trans
C
sub
(91)
0.0
500 Hz
50 Hz
5 Hz
0.20
0.15
0.10
0.05
0.00
0.1
Real ()
I
m
a
g

(

)
0.2 0.3
Figure 21. Nyquist plot of the IMPSresponse of a dye-sensitized
solar cell under short-circuit conditions. The electron diffusion
coefcient obtained from the data is 2.810
5
cm
2
/s.
Reprinted with permission from Franco et al., 1999.
Copyright 1999 American Chemical Society.
SEMICONDUCTOR PHOTOELECTROCHEMISTRY 29
The transport lifetime is similarly related to the resis-
tance to charge transport and the capacitance of the
nanoparticulate semiconductor.
t
trans
= R
trans
C
nsc
(92)
Rearrangement of Equations 91 and 92 shows that the
time constant for transport can be determined if the
capacitance of boththe semiconductor nanoparticle and
the substrate are known.
t
trans
= t
rise

C
nsc
C
sub
(93)
The time constant for transport is then related to the
effective diffusion coefcient by Equation 89 as in the
IMPS measurements. The decay of the photovoltage at
longer times is then directly related to the effective life-
time of the electrons under conditions of recombination
to solution acceptors.
The measured D
n
and t
n
can then be utilized to deter-
mine the electron diffusion length via Equation 84 sim-
ilar to the method in IMPS and IMVS. The main
advantage of the photovoltage transient method over
that of IMPS and IMVS is that both the values are con-
ducted at open circuit. Therefore, there is no need to
correct the data for the difference in quasi-Fermi level as
is required in IMPS/IMVA. Comparative data between
the IMPS/IMVS approach and the photovoltage tran-
sient method have shown consistent results between
the two methods at lowDCbias (Dunn and Peter, 2009).
Practical Aspects of the Method
The experimental setup to measure laser-induced tran-
sients is very similar to that discussed for IMPS or IMVS.
A CW light source (see IMPS and IMVS for CW illumina-
tion sources) provides the DC bias to the cell. On top of
the DC illumination, a pulsed laser source is utilized to
cause asmall perturbationinthe photovoltage. Typically
the perturbation is controlled to be within 10% of the
steady-state DCphotovoltage by modulationof the pulse
energy. The data are typically signal averaged and col-
lected with a digital oscilloscope. Alternatively, the data
can be collected with a potentiostat where no signal
averaging is necessary.
Data Analysis and Interpretation
Once the photocurrent transient is collected, the rise
and decay lifetimes must be extracted. The photovoltage
decay behaves exponentially and, therefore, can be
directly t to yield t
n
. However, the photovoltage rise is
oftenconvoluteddue to the fact that transport withinthe
investigated nanocrystalline thin lms occurs via a hop-
ping mechanism. To obtain the t
trans
, two methods can
be employed. The rise may be modeled with a complex
function, for example, the KohlrauschWilliamsWatts
(KWW) model that has been used previously to accu-
rately describe the transport of electrons through nano-
crystalline TiO
2
(Nelson, 1999). Alternatively, the time
constant has been estimated from the time required to
reach half of the photovoltage peak value. Comparative
studies of bothmethods haveshownminimal differences
between the values of D
n
obtained.
Problems. A problem in applying this approach to
nanocrystallite semiconductors involves the necessary
inclusion of the substrate. At high DC bias intensity
(high laser pulse energy), direct charging of the
substrate is observed in the transient as a sharp jump
in the data. Although corrections can be made to data to
estimate the rise timeconstant, these corrections appear
insufcient to produce consistent measurement of the
diffusion length (Dunn and Peter, 2009).
In addition, when measuring the photovoltage rise on
materials that are highly defected and, therefore, have a
large density of trap states, care must be taken that the
measured values are compared at similar trap densities.
This requires that the density of trap states be measured
independently by either capacitive or other methods that
have beenutilized inthe literature (Bisquert, 2008; Mor-
ris and Meyer, 2008). If the density of states is known,
the photovoltage transient needs to be measured at
equal values of E
cb
E
F
and thus equal value of the ratio
of trapped to free electrons.
APPENDIX: GLOSSARY OF TERMS AND SYMBOLS
A Oxidized acceptor species
A
1
(A
2
) Preexponential oating variables of a bi-
exponential t to a decay curve
A

Reduced acceptor species (donor)


A
s
Surface area of electrode
[A]
b
Bulk concentration of oxidized acceptor
species
[A]
S
Acceptor concentration at semiconductor
surface
[A

]
S
Donor concentration at semiconductor
surface
B Circuit susceptance
C Constant equal to qk
et
[A]
s
C
0
Differential capacitance measured at base
of EPS peak
C
d
Experimentally measured differential
capacitance of a semiconductor
C
d,sat
Plateau or peak capacitance measured
under high illumination intensity
C
H
Differential capacitance of Helmholtz layer
c
n,o
Optical emission rate constant for electrons
c
n,t
Thermal emission rate constant for
electrons
c
p,o
Optical emission rate constant for holes
c
p,t
Thermal emission rate constant for holes
C
sc
Differential capacitance of spacecharge
layer
C
ss
Differential capacitance of semiconductor
surface
C
t
Time-dependent differential capacitance
d Semiconductor sample thickness
D Carrier diffusion constant
D
0
Diffusion coefcient
30 SEMICONDUCTOR PHOTOELECTROCHEMISTRY
dn/dt Net rate of electron transfer into solution
E Electric eld
E Applied potential
E(A/A

) Redox potential
E(A/A

) Electrochemical potential of contacting


solution
E
cb
Energy of conduction band edge
E
corr
Potential corrected for resistance and con-
centration overpotentials
E
F
Fermi level of semiconductor
E
g
Band gap energy
E
0
(A/A

) Formal electrode potential of contacting


solution
E
t
Energy of a trap state
E
vb
Energy of valence band edge
E Frequency of AC input voltage signal
F Faradays constant
F(l) Flux of photogenerated carriers
f
t,s
Fraction of surface states lled by electrons
G Circuit conductance
G
0
(r) BeerLambert generation function
h Plancks constant
I Current
I
ph
Photocurrent
I
pl
Photoluminescence intensity
J Current density
J
0
Exchange current density
J
D
(r) Diffusion ux of minority carriers along
surface
J
l,a
Limiting anodic current density
J
l,c
Limiting cathodic current density
j
n
(j
p
) Charge transfer current to and from con-
duction (valence) band
J
ph
Photocurrent density
J
sc
Short-circuit current density
k Boltzmanns constant
k
et
Interfacial electron transfer rate constant
k
et

Reverse interfacial electron transfer rate


constant
k
ht
Interfacial hole transfer rate constant
K
LSS
Effective circuit constant in an LSS
experiment
k
n
(k
p
) Electron (hole) capture coefcient for traps
k
r
Radiative recombination rate constant
k
soln
Electron-transfer rate constant from a sur-
face trap to a solution acceptor
l Step or grain boundary
L Minority carrier diffusion length
L
P
Hole diffusion length
M Molecular weight of semiconductor
n Number of electrons transferred
n

Number of holes required to oxidize one


atom of electrode material
n
1,s
(p
1,s
) Concentration of electrons (holes) in con-
duction (valence) band when trap energy
is at Fermi energy
n
b
Electron concentration in bulk of
semiconductor
N
c
Effective density of states in conduction
band
N
d
Dopant density
n
i
Intrinsic electron concentration
n
s
(p
s
) Surface electron (hole) concentration
n
s0
Equilibrium surface electron
concentration
N
t
Concentration of bulk states (in cm
3
)
N
t
(l) Density of optically active bulk states in a
semiconductor at a particular illumination
wavelength
N
t
,
s
Concentration of surface states (in cm
2
)
N
t,s
(l) Density of optically active surface states ina
semiconductor at a particular illumination
wavelength
N
v
Effective density of states in valence band
P
0
Equilibrium minority carrier concentration
in the dark
p
s
(r) Radial dependence of minority carriers
about an LSS laser spot
q Electronic charge
Q Charge density
r Radius of laser spot
R Gas constant
R
+
Optical reectivity of solid
r
0
Perimeter of laser spot
r
eff
Distance from center of laser spot where
concentration of minority carriers is 1/e of
its value at perimeter of laser spot
R
H
Resistance of Helmholtz layer
R
r
Rate of radiative band-to-band emission
R
s
Resistance of bulk semiconductor
R
sc
Resistance of spacecharge layer
R
soln
Solution resistance
R
SRH
Rate of ShockleyReadHall (SRH)
nonradiative recombination in
semiconductor
R
ss
Resistance due to semiconductor surface
states
S
high
High-level surface recombination velocity
(in cm/s)
S
low
Low-level surface recombinationvelocity for
electrons (holes)
T Temperature
V Electric potential
V
bi
Built-in voltage
v
n
(v
p
) Thermal velocity of electrons (holes) in a
solid
V
n
Potential difference betweenFermi level and
conduction band in bulk
V
oc
Open-circuit voltage
W Depletion width
W
/
Depth of dopant density measurement
(W W
d
)
W
d
Thickness of material dissolved from an
electrode
x Distance into semiconductor
Z
im
Imaginary component of impedance
Z
re
Real component of impedance
Z
tot
Total impedance
a Absorption coefcient
b Coefcient of optical transitions
g Diode quality factor
G Transmitted photon ux
SEMICONDUCTOR PHOTOELECTROCHEMISTRY 31
G
0
Incident photon ux
Dp (Dn) Optically induced excess minority (major-
ity) carrier concentration
Dp
0
Optically induced excess minority
carrier concentration directly after excita-
tion (t =0)
DR
m
(t) Reected microwave power
e Relative permittivity
e
0
Permittivity of free space
e
s
Static dielectric constant of semiconductor
Z
conc
Concentration overpotential
y Phase angle
k
n
Neutralization (emission) rate constant
k
s
Ionization (capture) rate constant
l Excitation wavelength
m
n
(m
p
) Mobility of electrons (holes) in semicon-
ductor
n Frequency of incident radiation
r Density of solid
s
n
(s
p
) Capture cross section of electrons (holes) in
a trap
t Minority carrier lifetime
t) Weighted lifetime parameter for t to a
decay curve
t
1
(t
2
) Lifetimes of biexponential t to a decay
curve
t
pl
Effective total lifetime for photolumines-
cence spectrum
t
r,h
High-level injection radiative lifetime
t
r,l
Low-level injection radiative lifetime
t
s,h
High-level SRHlifetime associatedwithsur-
face recombination
t
sl
Low-level SRH lifetime associated with sur-
face recombination
v Kinematic velocity of a solution
F Quantum yield
f Barrier height
w
c
Capacitive reactance
o Angular frequency of AC input signal
o
rde
Angular velocity of rotating disk electrode
ACKNOWLEDGMENTS
A. Morris wouldlike to acknowledge Virginia Polytechnic
Institute and State University for their generous sup-
port. We acknowledge the Department of Energy, Ofce
of Basic Energy Sciences, and the National Science
Foundationfor their generous support of semiconductor
electrochemistry that is incorporatedpartly inthis work.
In addition, we acknowledge helpful discussions with G.
Coia, G. Walker, andO. Kr uger during the preparationof
this article.
BIBLIOGRAPHY
Semiconductor Photoelectrochemistry in Characterization of
Materials, 1st ed., Vol. 1, pp. 605636, by Samir J. Anz, Arnel
M. Fajardo, WilliamJ. Royea, andNathanS. Lewis, California
Institute of Technology, Pasadena, California; Published
online: October 15, 2002; DOI: 10.1002/0471266965.
com052.
LITERATURE CITED
Ahrenkiel, R. K., andLundstrom, M. S. 1993. Minoritycarriersin
IIIV semiconductors: physics and applications. In Semicon-
ductors andSemimetals, (R. K. Willardson, A. C. Beer, andE.
R. Weber, eds.), pp. 40150. Academic Press, San Diego, CA.
Ambridge, T. and Faktor, M. M. 1975. An automatic carrier
concentration prole plotter using an electrochemical tech-
nique. J. Appl. Electrochem. 5:319328.
Anderson, J. C. 1982. Theory of photocapacitance in amor-
phous silicon MIS structures. Philos. Mag. B 46:151161.
Aspnes, D. E. and Studna, A. A. 1981. Chemical etching and
cleaning procedures for Si, Ge and some IIIV compound
semiconductors. Appl. Phys. Lett. 39:316318.
Bard, A. J. and Faulkner, L. R. 1980. Electrochemical Methods:
Fundamentals and Applications. John Wiley & Sons, Inc.,
New York.
Bisquert, J. 2008. Physical electrochemistry of nanostructured
devices. Phys. Chem. Chem. Phys. 10:4972.
Blakemore, J. S. 1987. Semiconductor Statistics. Dover Pub-
lications, New York.
Blood, P. 1986. Capacitancevoltage proling and the charac-
terization of IIIV semiconductors using electrolyte barriers.
Semicond. Sci. Technol. 1:727.
Carlsson, P. and Homstr om, B. 1988. Laser spot scanning for
studies of the photoelectrochemical properties of InSe. Finn.
Chem. Lett. 121:5253.
Carlsson, P., Holmstr om, B., Uosaki, K., and Kita, H. 1988.
Fiber optic laser spot microscope: a new concept for photo-
electrochemical characterization of semiconductor electro-
des. Appl. Phys. Lett. 53:965972.
Chen, M. C. 1988. Photoconductivity lifetime measurements on
HgCdTe using a contactless microwave technique. J. Appl.
Phys. 64:945947.
Dunn, H. K., and Peter, L. M. 2009. How efcient is electron
collection in dye-sensitized solar cells? Comparison of dif-
ferent dynamic methods for the determinationof the electron
diffusion length. J. Phys. Chem. C 113:47264731.
Eriksson, S., Carlsson, P., and Holmstr om, B. 1991. Laser spot
scanning inphotoelectrochemical systems, relationbetween
spot size and spatial resolution of the photocurrent. J. Appl.
Phys. 69:23242327.
Fajardo, A. M., Karp, C. D., Kenyon, C. N., Pomykal, K. E.,
Shreve, G. A., Tan, M. X., and Lewis, N. S. 1995. New
approaches to solar energy conversion using Si/liquid junc-
tions. Solar Energy Mater. Solar Cells 38:279303.
Fajardo, A. M., and Lewis, N. S. 1997. Free-energy dependence
of electron-transfer rate constants at Si/liquid interfaces. J.
Phys. Chem. B 101:1113611151.
Finklea, H. O. 1988. Semiconductor Electrodes. Elsevier, New
York.
Fonash, S. J. 1981. Solar Cell Device Physics. Academic Press,
New York.
Forbes, M. D. E. and Lewis, N. S. 1990. Real-time measure-
ments of interfacial charge transfer rates at silicon/liquid
junctions. J. Am. Chem. Soc. 112:36823683.
Franco, G., Gehring, J., Peter, L. M., Ponomarev, E. A., and
Uhlendorf, I. 1999. Frequency-resolved optical detection of
photoinjected electrons in dye-sensitized nanocrystalline
photovoltaic cells. J. Phys. Chem. B 103:692698.
Gerischer, H. 1975. Electrochemical photo and solar cells:
principles and some experiments. J. Electroanal. Chem.
58:263274.
32 SEMICONDUCTOR PHOTOELECTROCHEMISTRY
Gerischer, H. 1991. Electron-transfer kinetics of redox reac-
tions at the semiconductor/electrolyte contact: a new
approach. J. Phys. Chem. 95:13561359.
Gillis, H. P., Choutov, D. A., Martin, K. P., Bremser, M. D., and
Davis, R. F. 1997. Highly anisotropic, ultra-smooth pattern-
ing of GaN/SiC by low-energy-electron enhanced etching in
DC plasma. J. Electron. Mater. 26:301305.
Goodman, C. E., Wessels, B. W., and Ang, P. G. P. 1984.
Photocapacitance spectroscopy of surface states on indium
phosphide photoelectrodes. Appl. Phys. Lett. 45:442444.
Guill en, E., Peter, L. M., and Anta, J. A. 2011. Electron trans-
port and recombination in ZnO-based dye-sensitized solar
cells. J. Phys. Chem. C 115(45):2262222632.
Haak, R., Ogden, C., and Tench, D. 1982. Electrochemical
photocapacitance spectroscopy: a new method for charac-
terization of deep levels in semiconductors. J. Electrochem.
Soc. 129:891893.
Haak, R. and Tench, R. 1984. Electrochemical photocapaci-
tance spectroscopy method for characterization of deep
levels and interface states in semiconductor materials. J.
Electrochem. Soc. 131:275283.
Hall, R. N. 1952. Electronhole recombination in germanium.
Phys. Rev. 87:387388.
Halme, J. 2011. Linking optical and electrical small amplitude
perturbation techniques for dynamic performance charac-
terization of dye solar cells. Phys. Chem. Chem. Phys.
13:1243512446.
Higashi, G. S., Becker, R. S., Chabal, Y. J., and Becker, A. J.
1991. Comparison of Si(111) surface prepared using aque-
ous-solutions of NH
4
F versus HF. Appl. Phys. Lett.
58:16561658.
Higashi, G. S., Irene, E. A., and Ohmi, T. 1993. Surface Chem-
ical CleaningandPassivationfor Semiconductor Processing.
Materials Research Society, Pittsburgh, PA.
Hodes, G. 2001. Electrochemistry of Nanomaterials.
WileyVCH, New York.
Kamieniecki, E. 1983. Surface photovoltage measured capac-
itance: application to semiconductor/electrolyte system. J.
Appl. Phys. 54:64816487.
Kenyon, C. N., Ryba, G. N., and Lewis, N. S. 1993. Analysis of
time-resolved photocurrent transients at semiconductor
liquid interfaces. J. Phys. Chem. 97:1292812936.
Koval, C. A. and Howard, J. N. 1992. Electron transfer at
semiconductor electrodeliquidelectrolyte interfaces. Chem.
Rev. 92:411433.
Kr uger, O., Jung, C., and Gajewski, H. 1994a. Computer sim-
ulation of the photoluminescence decay at the GaAselec-
trolyte junction: 1. The inuence of the excitationintensity at
the at-band condition. J. Phys. Chem. 98:1265312662.
Kr uger, O., Jung, C., and Gajewski, H. 1994b. Computer sim-
ulation of the photoluminescence decay at the GaAselec-
trolyte junction: 2. The inuence of the excitation intensity
under depletion layer conditions. J. Phys. Chem.
98:1266312669.
Leong, W. Y., Kubiak, R. A., and Parker, E. H. C. 1985. Dopant
proling of SiMBE material using the electrochemical CV
technique. Paper presented at the First International Sym-
posium on Silicon MBE, Toronto, ON, Electrochemical
Society.
Lewis, N. S. 1990. Mechanistic studies of light-induced charge
separation at semiconductor/liquid interfaces. Acc. Chem.
Res. 23:176183.
Lewis, N. S. andRosenbluth, M. 1989. Theory of semiconductor
materials. In Photocatalysis: Fundamentals and Applica-
tions, (N. Serpone and E. Pelizzetti, eds.), pp. 4598. John
Wiley & Sons, Inc., New York.
Li, J. and Peter, L. M. 1985. Surface recombination at semi-
conductor electrodes: Part III. Steady-state and intensity
modulated photocurrent response. J. Electroanal. Chem.
193:2747.
Love, J. C. and Demas, J. N. 1984. Phase plane method for
deconvolution of luminescence decay data with a scattered-
light component. Anal. Chem. 56:8285.
Many, A., Goldstein, Y., andGrover, N. B. 1965. Semiconductor
Surfaces. North-Holland Publishing, New York.
Mathian, G., Pasquinelli, M., and Martinuzzi, S. 1985. Photo-
conductance Laser Spot Scanning Applied to the Study of
Polysilicon Defect Passivation. Physica 129B:229233.
Meier, A., Kocha, S. S., Hanna, M. C., Nozik, A. J., Siemoneit, K.,
Reineke-Koch, R., andMemming, R. 1997. Electron-transfer
rate constants for majority electrons at GaAs and
GaInP2 semiconductor/liquid interfaces. J. Phys. Chem. B
101:70387042.
Morris, A. J. and Meyer, G. J. 2008. TiO
2
surface functionaliza-
tion to control the density of states. J. Phys. Chem. C
112 (46):1822418231.
Morrison, S. R. 1980. Electrochemistry at Semiconductor and
Oxidized Metal Electrodes. Plenum, New York.
Naber, J. A. andSnowden, D. P. 1969. Applicationof microwave
reection technique to the measurement of transient and
quiescent electrical conductivity of silicon. Rev. Sci. Instrum.
40:11371141.
Nagasubramanian, G., Wheeler, B. L., Fan, F.-R. F., and Bard,
A. J. 1982. Semiconductor electrodes: XLII. Evidence for
Fermi level pinning from shifts in the atband potential of
p-type silicon in acetonitrile solutions with different redox
couples. J. Eletrochem. Soc. 129:17421745.
Nelson, J. 1999. Continuous-time random-walk model of elec-
trontransport innanocrystalline TiO
2
electrodes. Phys. Rev.
B. 59:1537415380.
ORegan, B. C., Bakker, K., Kroeze, J., Smit, H., Sommeling, P.,
and Durrant, J. R. 2006. Measuring charge transport from
transient photovoltage rise times: a new tool to investigate
electron transport in nanoparticle lms. J. Phys. Chem. B
110:1711517160.
Pankove, J. I. 1975. Optical Processes in Semiconductors.
Dover Publications, New York.
Peat, R. and Peter, L. M. 1987a. Determination of the electron
diffusion length in p-GaP by intensity modulated photocur-
rent measurements with an electrolyte contact. Appl. Phys.
Lett. 51:328330.
Peat, R. and Peter, L. M. 1987b. A study of the passive lm on
iron by intensity-modulated photocurrent spectroscopy. J.
Electroanal. Chem. 228:351364.
Peter, L. M. 1990. Dynamic aspects of semiconductor photo-
electrochemistry. Chem. Rev. 90:753769.
Pleskov, Y. V. and Guervich, Y. Y. 1986. Semiconductor Photo-
electrochemistry. Consultants Bureau, New York.
Pomykal, K. E., and Lewis, N. S. 1997. Measurement of inter-
facial charge-transfer rate constants at n-type InP/CH
3
OH
junctions. J. Phys. Chem. B 101:24762484.
Ponomarev, E. A. and Peter, L. M. 1995. A generalized theory of
intensity modulated photocurrent spectroscopy (IMPS). J.
Electroanal. Chem. 396:219226.
SEMICONDUCTOR PHOTOELECTROCHEMISTRY 33
Ramakrishna, S. and Rangarajan, S. K. 1995. Time-resolved
photoluminescence andmicrowave conductivity at semicon-
ductor electrodes: depletion layer effects. J. Phys. Chem.
99:1263112639.
Rosenbluth, M. L. and Lewis, N. S. 1989. Ideal behavior of the
open circuit voltage of semiconductor/liquid junctions. J.
Phys. Chem. 93:37353740.
Schefold, J. 1992. Impedance and intensity modulated photo-
current spectroscopy as complementary differential meth-
ods in photoelectrochemistry. J. Electroanal. Chem.
341:111136.
Schlichth orl, G., Huang, S. Y., Sprague, J., and Frank, A. J.
1997. Band edge movement and recombination kinetics in
dye-sensitized nanocrystalline TiO
2
solar cells: a study by
intensity modulated photovoltage spectroscopy. J. Phys.
Chem. B 101:81418155.
Schroder, D. K. 1990. Semiconductor Material and Device
Characterization. John Wiley & Sons, Inc., New York.
Seabaugh, A. C., Frensley, W. R., Matyi, R. J., and Cabaniss, G.
E. 1989. Electrochemical CV proling of heterojunction
device structures. IEEE Trans. Electron. Dev. ED-
36:309313.
Sharpe, C. D. and Lilley, P. 1980. The electrolytesilicon inter-
face: anodic dissolution and carrier concentration proling.
J. Electrochem. Soc. 27:19181922.
Shockley, W. and Read, W. T. 1952. Statistics of the recombi-
nation of holes and electrons. Phys. Rev. 87:835842.
Sze, S. M. 1981. The Physics of Semiconductor Devices. John
Wiley & Sons, Inc., New York.
Tan, M. X., Kenyon, C. N., and Lewis, N. S. 1994a. Experi-
mental measurement of quasi-Fermi levels at an
illuminated semiconductor/liquid contact. J. Phys. Chem.
98:49594962.
Tan, M. X., Laibinis, P. E., Nguyen, S. T., Kesselman, J. M.,
Stanton, C. E., and Lewis, N. S. 1994b. Principles and
applications of semiconductor photoelectrochemistry. Prog.
Inorg. Chem. 41:21144.
Tributsch, H. and Bennett, J. C. 1977. Electrochemistry and
photochemistry of MoS
2
layer crystals: I. J. Electroanal.
Chem. 81:97111.
Willardson, R. K. and Beer, A. C. 1981. Semiconductors and
Semimetals: Contacts Junctions. Emitters Academic Press,
New York.
Wilson, R. H., Harris, L. A., and Gerstner, M. E. 1979. Charac-
terized semiconductor electrodes. J. Electrochem. Soc.
126:844850.
KEY REFERENCES
Bard and Faulkner, 1980. See above.
General electrochemistry text with a good introduction to imped-
ance spectroscopy.
Schroder, 1990. See above.
Describes in more detail many of the techniques discussed here,
including photoluminescence and photoconductivity decay.
Shockley and Read, 1952. See above.
Classic reference for carrier recombination.
Tan et al., 1994b.
A detailed introduction to the thermodynamics and kinetics of
semiconductorliquid interface.
34 SEMICONDUCTOR PHOTOELECTROCHEMISTRY

S-ar putea să vă placă și