8720
J . Phys. Chem. 1990, 94. 8720-8726
Vectorial Electron 1n)ection into Transparent Semiconductor Membranes and Electric
Field Effects on the Dynamics of Light-Induced Charge Separation
Brian O’Regan,2 Jacques Moser,2 Marc Anderson,’ and Michael Gratzel*?*
Institut de Chimie Physique, Ecole Polytechnique F$d?rale de Lausanne. CH- I015 Lausanne. Switzerland, and
Water Cheniistrj*and Material Science lnstit Ute, Unicersity of Wisconsin, Madison. Wisconsin 53706
(Rereired: Ma), 7 . 1990)
Transparent titanium dioxide membranes (thickness 2.7 pm) were prepared by sintering of 8-nm colloidal anatase particles
on a conducting glass support. The dynamics of charge recombination following electron injection from the excited state
of RuL, ( L = 2,2’-bipyridine-4,4’-dicarboxylicacid) into the conduction band of the semiconductor were examined under
potcntiostatic control of the electric field w i t h i n the space charge layer of the membrane. Biasing the Fermi level of the
TiO, positive of the flat-band potential sharply reduced the recombination rate, a 1000-fold decrease being associated with
;I potcntiul chnngc of only 300 m V . Photoelectrochemical experiments performed with the same RuL,-loaded membrane
in Kal-containing watcr show the onset of anodic photocurrcnt to occur in the same potential domain. Forward biasing
01‘ the nicmbranc potcntial impairs photosensitized charge injection turning on the photoluminescence of the adsorbed sensitizer.
Introduction
Tlic dynmics of heterogeneous photochemical electron-transfer
reactions are frequently controlled by local electrostatic potential
gradients present at the
This plays a crucial role
in molecular devices for light-induced charge separation and solar
energy conversion. In the case of semiconductor-liquid junctions.
the dcpletion laycr ficld present within the solid impairs the recombination of charge carriers formed by light excitation6 With
a convcntional scmiconductor electrodes, these kinetic effects of
the space chargc arc difficult to monitor directly and so far few
timc-resolvcd studies have been reported.’
In the present work. we take advantage of the transparent nature
of a newly developed semiconductor membraneE to examine the
influence of the depletion layer field on the rate of charge carrier
recombination following photoinduced electron injection from
surfxc-adsorbed dye molecules. Thin titanium dioxide membranes
( I ) Water Chemistry and Matcrial Scicncc Institute, University of Wisconsin. Madison. WI 53706.
(2) lnstitut de Chimie Physique. Ecole Polytechnique Fidirale de Lausannc. I O 1 5 I.ausannc. Swit7crland.
( 3 ) (a) Gcrischcr. H. Angew. Chem. 1988, 100, 630. (b) Meier. H. J .
Phvs. Chem. 1%5,69, 724. (c) Hauffe. K.; Range, J. 2.Nuturforsch. B 1968.
238. 736. (d) Watanabe. T.: Fujishima. A , : Honda. K . In Energy Resources
Through Photochemistry and Catal.vsis: Academic Press: New York. 1983.
(c) Tributsch. H.:Calvin. M. Photochem. Photobiol. 1971. 14. 95. (0 Karnet.
P. V . : Fox. M. A . Chem. Phys. Lett. 1983. 102, 379. (8) Memming. R. Prog.
S u r t Sci. 1984. 17. 7. (h) Krishnan. M.: Zhang. X.; Bard, A. J. J . Am. Chem.
Sur. 1984. 106. 7371. ( i ) Gerischer. H.; Willig. F. Top. Curr. Chem. 1976.
61. 31. (k) Hashimoto. K.: Sakata. T . J . Phys. Chem. 1986. 90. 4474.
(4) Spitlcr. M. J . Electroanul. Chem. 1987. 228,69. For related work, cf.:
Blosscy. D. F. P h i Rei,.
~
1974. 139. 5183. Willig. F. Chem. Phyc. Left. 1976.
40. 331
( 5 ) ( a ) Photoinduced Electron 7run.rfer: Fox. M. A,. Chanon, M.. Eds.:
Elsevier: Amsterdam, 1988; Part A-4. (b) Gratzel. M. Heterogeneous
Phut~,[.hc,,,ri[,trlElet,rroti Transfer; CRC Prcss: Boca Raton. FL, 1989.
( 6 ) ( a ) Wrighton. M. S . Arc. Chem. Res. 1979. 12. 303. (b) Gerischer.
H. Pirre App1. Chem. 1980. 52. 2649. (c) Heller. A. A r c . Chem. Res. 1981.
1 4 . 1.24.
(7) (a) Bitterling. K.: Willig. F. J . Electroanal. Chem. 1986. 204. 21 I . (b)
Rlnn. M A , : Fit7gerald. E. C.: Spiller. M. T. J . Phys. Chem. 1989, 93. 61 50.
(8) W e uish to draw aitcntion to studies by Fendler et 31. concerning the
formation. chor~icteri7ntion.and photoelectrochemistry of sulfide semiconductor particles supported by bilayer lipid membranes, e.g.: Zhao. X . K.:
Baral. S.; Rolandi. R.: Fendler. J . H. J . Am. Chem. Soc. 1988. 110. 1012
Bard et a l . invcstigatcd t h i n CdS semiconductor films, c.g.: Finlayson. M .
F.: Wheeler. B. L.: Kakuta. N.: Park. K. H.:Bard, A. J.: Fox. M . A,: Webber.
S. E.. White. J . M. J . P h j ~ Chem.
.
1985. 89. 5676. Liu. C.: Bard. A . J . J .
PhJ.7 Chvnr. 1989. 93. 7749. The TiO? membranes introduced here distinguiah thcinrclvcs by thcir transparcnt and microporous character. High
light-harvcrting cfficicncics arc achicvcd in this fashion at monolayer d l c
covcragc allowing for iipplication of timeresolved optical transmission spectroscupq
have been prepared on a conducting glass support allowing for
potentiostatic control of the potential gradient within the semiconductor. RuL3 ( L = 2.2’-bipyridine-4,4’-dicarboxylicacid)
adheres strongly to the surface of TiO19 and is used as a model
chromophore. Time-resolved absorption and transient current
nicasurcmcnts are applied for the first time in conjunction with
laser photolysis to scrutinize the dynamics of charge carrier
formation and recombination events in this system.
Experimental Section
Prepuration of Transparent TiO, Membranes Supported on
Conducting Glass Sheets. Transparent TiO, membranes were
produced by deposition of colloidal particles on a conducting glass
support. Thc procedure applied was similar to that used for the
preparation of unsupported films.I0
TiOz colloid solutions were prepared by hydrolysis of titanium
isopropoxide, Ti(OCH(CH,),),, as follows: Under a stream of
dry nitrogen, 125 mL of Ti(OCH(CH3)2)4(Aldrich) was added
to a 150-mL dropping funnel containing 20 mL of 2-propanol
(Fisher, ACS reagent grade). The mixture was added over IO
min to 750 mL of distilled deionized water, stirring vigorously.
During the hydrolysis a white precipitate formed. Within 10 min
of the alkoxide addition, 5.3 mL of 70% nitric acid (Fisher, ACS
rcagcnt) was added to the hydrolysis mixture, still stirring vigorously. The mixture was then stirred for 8 h at -80 OC. The
2-propanol (and some water) was allowed to evaporate during this
time. Approximately 700 mL of stable TiOzcolloidal sol resulted
from this procedure. The size of the colloidal particles was ca.
8 nm and X-ray diffraction analysis showed them to consist of
anatase. Crystallization occurred during the refluxing, the initial
T O z precipitate being X-ray amorphous.
A portion of the above sol was concentrated under vacuum at
room temperature until it was visibly viscous. Depending on the
iigc of the sol. the proper viscosity was reached between I50 and
200 g of TiOz per liter. Nonporous Sn02films ( F doped) on glass
ucrc used for electrically conductive supports (provided by
Glasstech Solar, Wheat Ridge CO). Membranes were formed
on these supports by spin coating at 3000 rpm. Ti02layers thinner
than 0.5 pm did not crack when fired directly in a 400 OC oven.
Thicker layers cracked under any firing regime. Membranes up
to 1 pm thick were formed by multiple application and firing of
( 9 ) ( a ) Desilvestro. J.; Gratzel. M.; Kavan. L.: Moser. J.; Augustynski, J.
J . 4 m Chem. Soc. 1985, 107. 2988. (b) Furlong, D.N.; Wells, D.; Sasse,
W . H. F. J . Ph),.r. Chem. 1986, 90. 1107. (c) Vlachopoulos, N.; Liska. P.;
Augustynski. J.: Gratzel, M. J . Am. Chem. Soc. 1988. 110, 1216.
(IO) Anderson. 21 A.: Gieselmann. M . J.; Xu, Q.J. Membr. Sci. 1988,
39, 243.
0022-3654/90~2094-8720$02.50~0
6 1990 American Chemical Society
Light-I nduccd Charge Separation
Thm
e Journal of Ph?*sicalChemistry. Vol. 94. No. 24. 1990 8721
w
POTENTIOSTAT
MONO.
MONO. .
DIGITAL
OSCILLOSCOPE
Figure 2. Experimental setup for time-resolved kinctic spectroscopy using
a Ud:YAG laser pulse to excite the dye-derivatized semiconductor
mcmbr;inc.
Figure 1 . Sciinning clcctron micrographs showing crohs sections of thc
TiOz nicmbrancs dcpositcd on conducting glass. Mngnificntions: ( a )
3000 times: (b) 100000 times.
-0.4-pni la!.crs. After ;i final firing a t 400 " C for 1 h. the
mciiibrancs \rere hcatcd in argon a t 550 "C in ;i Lindbcrg tube
furn:icc under ;in argon flow of 500 niL/min. Thc argon w a s
clcancd i r i t l i ;i \rater absorber and two oxygen traps (Alltcch).
The iiiorpholog\ of iiicnibriinc clcctrodcs w a s cxaniincd by
S E M ( I litachi), X - r i i j diffraction, iind BET analysis of N, ad-
sorption measured by a surface acoustic wave technique.
Thickness and porosity have also been measured by profilometry
and clipsometry. Figure 1 shows a cross section of the Ti02film
obtained by scanning electron microscopy at two different magnifications. Low resolution applied in Figure la confirms the
presence of a three-layer structure, the lowest being the glass
support followed by the 0.5 pm thick fluorine-doped Sn02 and
the 2.7 pni thick TiO, layer. High resolution (Figure I b) reveals
thc TiO, film to be composed of a three-dimensional network of
interconnected particles having an average size of approximately
I6 nm. Apparently, significant particle growth occurs during
sintering.
Methods. The photocurrent potential characteristics were
measured by using a xenon arc light source and a Wenkin potentiostat (Bank Electronic GmbH, F.R.G.). The photocurrent
action spectrum was obtained with a Bausch and Lomb 500-nm
blaze high-intensity monochromator. The monochromatic photon
flux impinging on the cell was determined by a YSI Kettering
Model 65 A radiometer. This agreed within 5% with the values
measured by fcrrioxalate actinometry. The electrochemical system
employed a single-compartment, three-electrode cell, with a
platinum counter electrode and a Ag/AgCI or a H g / H g 2 S 0 4
reference electrode in addition to the T i 0 2 surface under investigation. All potentials are reported against SCE.
Coating of the TiO, surface with dye was carried out by soaking
M, pH
the film for 2 h in an aqueous RuL3 solution ( 2 X
4). The R u L 3 was available from previous w ~ r k . ~All~ .other
~
chemicals were at least reagent grade and were used as received.
The kinetics of photosensitized electron injection and, recombination were examined by laser photolysis using potentiostatic
control of the Fermi level within the transparent TiO, film supported on the conducting glass substrate. A three-electrode cell
was employed where the TiO, surface was kept at a 45" angle
to the laser beam (frequency-doubled Nd:YAG; 20 mJ output
at 530 nm; pulse width at half-height ca. I O ns) and to the analyzing light, Figure 2. The latter was passed through a
monochromator prior to entering the cell. Another monochromator was placed in front of the photomultiplier tube used to
monitor the time course of the optical absorbance change induced
by laser excitation of the film.
Transient photocurrents following sensitized electron injection
into the TiO, membrane were measured with the same setup as
in Figure 2 except that the analyzing light was blocked. The
currents were recorded as a voltage drop over a IO-ohm resistor
iidded in series to the working electrode circuit by using a digital
oscilloscope with a 80-MHz bandwidth. Blank experiments
performed with the T i 0 2 membranes in the absence of sensitizer
gave no transient current signals.
The quantum yield for charge injection from the excited state
of R u L , into the conduction band of the T i 0 2 membrane was
obtained from the number of 530-nm laser photons absorbed by
Ru"L3 and the amount of Ru"'L3 generated during photosensiti7ation. The former was determined from the absorption spectrum
8722
0.6
O'Regan et al.
The Journal of Physical Chemistry, Vol. 94, No. 24, 1990
r
j
7
0.3
sensitizer loaded
0.1
1
I
I
1
I
I
I
400
500
600
WAVELENGTH / nm
Figure 3. Absorption spcctrn of the Ti02 membrane supported on conducting glass with and without monolayer coating of RuL,. At 480 nm,
3770 of the incoming light is absorbed by the RuL,. Spectra are corrected
for specular and diffuse reflectance.
of the RuL,-loaded film (Figure 3) while the latter was derived
from the bleaching of the RuL, absorbance at 480 nm immediately
after thc laser pulsc."
Additional experiments were carried out with an electrochemical
cell placed in the sample compartment of a S L M 500C spectrofluorometer. The sctup allowcd the excitation beam of the
spectrofluorometer to be used to irradiate the Ti02 transparent
elcclrodc from the rear without passing through the solution. Thus,
the clcctrodc luminesccnce could be examined as a function of
potential. Potential control and current measurement were provided by a n IBM EC/225 potentiostat. The area of the electrode
exposcd to thc solution was 2 cm2. With a monochromator
bandwidth of 20 nm. I .3 cm2 of the sample was illuminated with
470-nm light.
Results
Absorption and Emission Characteristics of Dye-Deriuatized
Ti02 Films. Figure 3 shows absorption spectra of the Ti02
membrane supported on conducting glass with and without coating
of RuL,. These were derived by combining transmittance and
reflectance spectra in order to eliminate interference oscillations.
The bare film exhibits apart from the band edge transition of Ti02
below 400 nm n feature rising slowly to the red due to free carrier
absorption in the fluorine-doped tin oxide.', Soaking the electrode
in thc aqucous RuL, solution produces its characteristic band with
a maximum in the visible around 470 nm. At 480 nm the absorption due to surracc bound RuL, is 37%. Using for the extinction cocfficient at this wavelength 2.2 X IO4 M-l cm-',98 and
for thc surfacc rcquircmcnt of one adsorbed RuL, molecule the
expcrimcntally determinedgbvalue of 1 nm2, and assuming complctc monolayer coverage. a roughness factor of 50 is dervied for
thc film. A hexagonal closc packing of 16-nm-sized spheres to
a layer of 2.7-pm thickness is expected to give a 230-fold surface
enlargcnicnt. That the area accessible to RuL, is significantly
smaller is not surprising in view of the necking of the particles
during the sintcring proccss resulting in films with microporous
morphology.
The luminescence of surface adsorbed RuL, was found to be
strongly affcctcd by the bias voltage applied to the TiO, membranc. At 0.2 V practically no emission could be detected due
to oxidative quenching of the excited state by charge injection
in the conduction bandga
Ru"L3
A
-
Ru11L3*
A,",
Ru"'L,
0
700
+ ecb- (TiOz)
(I)
Holding thc potcntial of thc film at -0.7 V induced the typical
( I 1 ) For the data in Figure 5 the laser flux was kept sufficiently high to
excitc a l l thc sensitizer molecules. The recombination kinetics remained
essentially t h e same when the laser power was reduced such that only 25%
of the dye molecules were excited.
( 1 2) Shanthi. E.: Banrjee. A.: Chopra. K . L . Thin Solid Films 1982. 88.
93.
I
1
400
500
I
700
600
WAVELENGTH I nm
Figure 4. Photocurrent action spectrum of a Ti02 membrane coated with
a monolayer of RuL,. The incident photon to current conversion efficiency is plotted as a function of excitation wavelength. The membrane
was immersed in aqueous 0.2 M Nal, pH 3, and a bias voltage of 0.2 V
(SCE) was applied.
a
.
I
2
80 -
K
K
60
-
8
40
-
a
20
-
I-
w
3
5
l
600
l
l
400
l
l
'
200
l
1
0
l
l
-200
1
1
'
1
'
1
3
-400 -600 -800
POTENTIAL / mV ( SCE )
Figure 5. Oscillograms showing the effect of the electrical potential on
the temporal behavior of the Ru"L, absorption at 480 nm after 530-nm
laser flash excitation of R"L,-loaded Ti02 membrane electrodes. The
electrode is immersed in an aqueous 0.2 M LiC104 solution, pH 3 .
Electrode potentials indicated on the oscillograms are referenced against
SCE. In Figure 5d the solution contained apart from the electrolyte 0.2
M N a l as electron donor.
luminescence of Ru"L3 with a maximum a t 640 nm, Figure 8.
The emission grows in gradually reaching a steady state within
a few minutes after applying the polarization. Upon applying a
reverse bias one observes within a few minutes complete quenching
of the RuL, emission.
Steady-State and Time-Resolved Photoelectrochemical Experiments. Electricity is generated with remarkable efficiency
when the RuL3-coated transparent Ti0, electrode is immersed
in an aqueous solution (pH 3) containing 0.2 M Nal and irradiated
by visible light. The incident photon to current conversion efficiency (IPCE) attains 25% at 480 nm, the photocurrent action
spectrum matches the light absorption of the film, Figure 4.
Expressing the IPCE by the relation
IPCE = LHEC#Ji,,qo
(2)
where L H E is the light-harvesting efficiency (0.37 at 480 nm),
i#+
isnthe
j quantum yield for electron injection, and qesfis the charge
separation probability, one obtains C#Jinjqesc
= 0.67. Therefore, at
the applied bias voltage of 0.2 V, at least 67% of the injected
electrons are drawn off as a current, the remainder recombining
with parent cations.
The kinetics of electron injection and charge recombination were
examined by nanosecond laser pulse excitation of the dye-derivatized TiO, film by monitoring the changes of the absorbance at
480 nm in the 10-7-100-s time domain. The temporal behavior
of the 480-nm absorbance is shown in Figure 5. The negative
deflection of the signal within the laser pulse is due to rapid
electron injectionga from the excited state of the RuL, in the
conduction band of the Ti02 film, eq 1. The subsequent recovery
The Journal of Physical Chemistry. Vol. 94, No. 24, 1990 8723
Light-Induced Charge Separation
+ 0.15 V
- 0.45 V
6t I
I
b)
I
I
I
I
I
c)
l
I
1
IJ"
0
1
2
3
4
0
5
1
2
3
TIME / ps
TIME / ps
+ 0.15 V
+ 0.15 V + I -
4
t
10'
lo6 lo5 lo4
IO
3
5
121
1
2
3
4
.
I-
z
Lu
TIME I p
TIME / s
Figure 6. Effect of the membrane electric potential on (0) the rate
of the signal arises from the recapture of the injected charge by
the oxidized sensitizer.
Ru"IL3
+ e,<
(Ti02)
kb
Ru'IL3
(3)
The potential applied to the membrane has a pronounced effect
on the back reaction rate. At a bias of -0.45 V, i.e.. close to the
flat-band potcntial of the Ti02 (see below), the bleaching recovers
within a fcw microseconds. Changing the film potential to +0.15
V induces a striking rctardation in the back reaction. The
bleaching recovers hcrc in two steps. Approximately one-third
of the signal dccreases rapidly in a fashion similar to the field-free
case. This is followcd by a much slower decay extending in the
millisecond time domain. In Figure 5c the 480-nm bleaching
signal is prcscntcd on a logarithmic time scale, the time frame
extcnding from IO-' to I 0-3 s. A straight line is obtained, indicating that the kinetics of reaction 3 follow a logarithmic time
law.
The slow componcnt of the bleaching recovery disappears in
thc prcscncc of iodide. This is illustrated in Figure 5d which
displays the temporal behavior of the 480-nm absorbance of RuL3
in thc prcscnce of 0.2 M Nal. The bleaching signal decays here
with a half-lifetime of 0.5 p . The decay is due to the reduction
of Ru"'L3 by iodide competing with back electron transfer:
Ru"'L, + IRu1'L3 0.51,
(4)
-
+
At a potcntial of 0.15 V and 0.2 M Nal this process is fast enough
to intercept completely the slow part of the back reaction.
However. the rapid initial component of the charge recombination
proccss is hardly affcctcd by the addition of iodide.
Figure 6 provides a summary of the kinetic data derived from
timc-resolvcd laser photolysis studies and compares the results
to photocurrent measurements performed under continuous illumination. A separate plot of the rate constant for the fast and
slow component of the charge recombination process is presented.
Both parts converge at around -0.7 V where a single first-order
decay with a rate constant of 1.5 X IO6 s-I is observed. Polarizing
the Ti02 film in the positive direction leads to the appearance of
the two components in the bleaching recovery as was illustrated
in Figure 5 . The kinetics of the initial fast decay is affected very
little by the film potential, the rate constant decreasing by less
than a factor of 2 over the voltage range from -0.8 to +0.6 V .
By contrast, the rate of the second and major componentI3 of the
( I 3) Thc ratc constants for the slow recombination process were calculated
from thc half-lifctimc of the second and major component of the bleaching
recovery. This process does not follow a simple first-order rate law and a more
detailcd kinctic analysis is under way.
I
E
5
constant for the fast component of the recombination reaction of conduction band electrons with Ru"L, parent sensitizer ions;).( the rate
constant for the slow and major component of the recombination reaction;
( + ) the photocurrent under polychromatic visible light excitation ( A >
420 nm) in the presence of 0.2 M iodide as electron donor. Conditions:
pH 3. 0.2 M I.iCI0, clcctrolyte.
1
a
I
0
,
K
I
h.c.
U
3
0
0
5
10
15
20
TIME f ps
Figure 7. Oscillograms showing the transient photocurrent produced by
photosensitized electron injection from Ru'IL, into the conduction band
of the TiO, membrane. Membrane potential was kept a t 0.4 V. (a) No
iodide added to the electrolyte; (b) 0.1 M N a I added. Other conditions
as in Figure 6 .
bleaching recovery drops by a factor of almost 1000 over the same
domain, the decrease being steepest in the vicinity of the flatband
potential of Ti02.
The photocurrent-potential curve included in Figure 6 was
obtained from photoelectrochemical experiments employing the
same RuL,-loaded Ti02 film as in the laser photolysis work.
Aqueous 0.2 M Lil electrolyte acidified to pH 3 with HCIO, and
polychromatic visible light (A 1 420 nm) were used. The incident
light intensity was deliberately kept low (ca. 5 mW/cm2) to avoid
accumulation of signifiant I< concentration during photolysis. The
latter may affect the shape and position of the photocurrent
potential curve. At the relatively weak light intensity employed
in Figure 6 , the photocurrent onset is around -0.3 V. Applying
a reverse bias produces a steep rise in the photocurrent which
attains a plateau around 0.1 V. This increase occurs in the same
potential region where the major part of bleaching recovery experiences strong retardation.
Photocurrent measurements made with chopped light gave a
more accurate estimate of the photocurrent onset. The value found
for pH 3 was -0.35 V. Previous work has shown the flat-band
potential of colloidal anatase particles to be -0.52 V (SCE) a t
pH 3 and a conduction band electron concentration of I O l 9 cm-3.14
From eq 5 the conduction band edge can be estimated to be about
-0.65 V at pH 3. c,- is the concentration of electrons in the
conduction band and ncbis the concentration of states at the bottom
of the conduction band, about 4 X lo2'cm-3 for Ti02. Assuming
approximately the same band edge for the membrane and a
flat-band potential, based on the photocurrent onset, not negative
of -0.4 V, the maximum doping density implied is on the order
of IO'' ~ m - ~A .doping density at least this low is also supported
by the very low dark conductivity of the membranes. The conductivity, measured by the four-point technique under dry nitrogen,
is less than IO-' S. The conductivity is strongly increased by
exposure to U V light indicating that the low conductivity is due
to a low electron concentration in the conduction band rather than
(14)
Duonghong, D.;Ramsden, J . J.; Gratzel, M. J . Am. Chem. Soc. 1982,
104, 2911.
(15) Albery, W. J.; Bartlett, P. N. J . Elerrrochem. Sot. 1984, 131, 315.
8724
The Journal of Physical Chemistry, Vol. 94, No. 24, 1990
to poor electrical contact between the particles.
Time-resolved photocurrent mcasurcments were also performed
and the results are displayed in Figure 7 . In this case the 530-nm
laser pulse was used to excite the RuL,-coated membranes and
the temporal evolution of the photocurrent was monitored in the
1 O-6-10-2-s time domain. Measurements were performed at
diffcrcnt potentials varying from -0.8 to +0.8 V. The data shown
in Figurc 7a.b were obtained at 0.4 V in the absence and presence
of 0.1 M iodide. respectively. The current signal rises after the
laser pulse. attaining a iiiaxiniuni within a few hundred microseconds and decays thereafter within milliseconds back to the
baseline. Integration of the current-time curve yields the total
charge collccted at the conducting tin dioxide back contact of the
mcmbranc. I n Figure 7a this amounts to 3.2 X I O 1 , electrons.
The number of photoinjected electrons determined from transient
absorption measurements performed under similar conditions was
8X
Hence. only 4% of the injected charge traverses the
membrane and is drawn off as a current, the remainder being
recaptured by parent sensitizer ions. That most of the photoinjected electrons undergo recombination even when an anodic bias
is applied to the membrane could be inferred from the time-resolvcd absorption measurements. For example, in Figure 5c. the
recovcrq of Ruti'L, absorption is almost complete 1 ms after laser
excitation. Howcvcr. i t cannot be ruled out that apart from the
recapture of injected electrons other processes such as the reaction
with impurities and ligand oxidation. contribute to the slow
conversion of Ru"'L1 into Ru"L3. Therefore, the extent to which
thc 480-nm bleach.ing signal recovers cannot be taken as a
quantitativc mcasurc for the fraction of photoinjected electrons
that cscnpc from recombination. Thc transient current measurcments provide such information on the vectorial displacement
of electrons across the membrane and their collection efficiency.
The charge collection yield was found to be strongly dependent
on clcctrode potential. No transient photocurrents were observed
at potentials negative of 4 . 2 5 V. Biasing the membrane to more
positive potentials rcsulted in a steep increase in the charge
collection yield which attained a plateau at around 0.2, similar
to the steady-state photocurrent potential curves observed in the
presence of iodidc i n Figurc 6.
The transient photocurrent was also greatly affected by addition
of electron donors. This is illustrated in Figure 7b which displays
thc currcnt rcsponsc to 530-nm excitation of the RuL,-loaded
mcmbranc in the presence of 0.1 M iodide. The signal is increased
3.5 times and its decay time is 3 times longer than in the absence
of iodide. The fraction of the injected charge collected as a current
was c ; ~ .6 5 Y . in good agrccmcnt with the results from the
s t ea d y -st ;I t c phot ocu r r c n t me a s u re m e n t s i n i od id e-co nt a i ni ng
solutions.
Apart from back electron transfer. the electrical potential was
found to affect the quantum yield for charge injection from the
excitcd state of RuL3 in the conduction band of the TiO, membranc. Results are shown in Figure 8. Within the potential range
of 0.6 to -0.1 V the injection yield remains constant and its value
is close to unity. Polarizing the membrane further in the negative
to 0.5 at -0.8 V. As mentioned above.
dircction dccrcascs
the reduction in
is accompanied by the appearance of RuL,
lumincsccnce at -0.7 V. Beyond this potential the initial bleaching
signal includes some excited states that may decay by luminescence
and the blcaching i h an uppcr limit for charge injection.
It should be noted that the electrode potentials referred to in
Figures 5-8 were mcasurcd in the dark. Laser-induced electron
injection from the sensitizer into the membrane shifts the potential
negativcly. This shift is ca. 20. 140. and 450 mV for dark electrode
potcntinls of -800. -500 and -100 mV. respectively.
&,
&,
Discussion
Thin transparcnt TiOz membrancs have bccn prepared on a
conductive glass support and thcir surface has bcen dcrivatizcd
with LI ruthcniuni complex acting as a charge-transfer sensitizer.
Thc large cffcctivc surface arc2 of the film resulting from its
microporous character affords a high light-harvesting efficiency
by thc scnsiti7cr. cven at monolayer coverage. This has allowed
O'Regan et al.
1 .o
9
>
0.9
I
3
5
0.8
sz
0.7
a
0
I-
o
z
0.6
0.5
600
400
200
0
-200
-400
-600 -800
POTENTIAL / mV ( SCE )
Figure 8. Effect of the electric potential on the quantum yield of photosensitized clcctron injection from Ru"L, in the conduction band of the
TiO, membrane. The inset shows the luminescence emission spetra of
RuL, adsorbed on the surface of the membrane at tNo different potentials. Excitation wavelength 500 nm. The electrode was immersed in 0.2
M LiCIO, electrolyte, p H 3.
us to apply simple time-resolved transmission spectroscopy in
conjunction with laser photolysis to unravel the salient kinetic
fcatures of heterogeneous photochemical electron-transfer reactions
at the semiconductor solution interface.
The primary goal of the present study was to scrutinize the role
of the electric potential applied to the membrane in controlling
interfacial charge-transfer events associated with the photosensitization process. By applying a bias voltage to the conductive
glass electrode one can adjust the Fermi level within the semiconductor membrane. The present results show that this has a
dramatic influence on the yield and dynamics of light-induced
charge separation at the TiO,/solution interface. In the following
discussion these observations are interpreted in terms of the effect
of the applied potential on the formation of a depletion layer field
as well as on the occupation of trapping states for electrons within
the membrane.
Polarizing the electrode positively with respect to the flat-band
potential of TiO, is expected to produce a depletion layer within
the film. Under these conditions, the membrane carries excess
positive space charge and the conduction and valence bands are
bcnt upwards from the semiconductor interior to its surface. An
ideal Schottky junction behaves such that in the depletion regime
the externally applied voltage drops across the semiconductor,
leaving the band-edge position at the surface unchanged. Such
a behavior is observed for compact semiconductor electrodes whose
thickness exceeds significantly that of the space charge layer.
Howcvcr, the Ti02 membranes employed here are microporous,
being composed of a densely packed array of colloidal antase
particles fused together during sintering. Such a morphology
imposes limits on the spatial extension of the depletion layer. For
an individual semiconductor particle, the width of the space charge
lajcr cannot exceed the particle radius. The maximal voltage
difference between the center and the surface of a sphere of radius
r is given bqIs
wherc LD= ( ~ , e k T / 2 e ~ N is~ )the
~ . Debye
~
length. With dielectric
constant and the ionized donor concentration values of 130 and
IO" ~ 1 1 respectively,
7 ~ ~ ~
the Debye length is 30 nm and the voltage
difference between the surface and the center of the 16-nm-sized
Ti02 particle is 0.3 mV under conditions of maximum depletion.
I t is evident that the colloidal particles constituting our membrane arc not isolated from each other. Rather they are in
clcctronic contact, forming a three-dimensional array of interconnected clusters. The clustering and interconnection of the
particles is expected to affect the potential distribution, and more
Light-Induced Charge Separation
significant electric fields are likely to be developed in such an array
as compared to individual colloidal particles. Presently, efforts
are undertaken to solve numerically the Poisson-Boltzmann
equation for semiconductor structures that model the membranes
cmploycd hcre. However, it is unlikely that our results can be
explained on the basis of the local potential gradients alone. In
particular, the small value of the collection efficiency of photoinjected electrons, which is only 4% even at a reverse bias voltage
of more than 1 V. cannot be reconciled with the presence of large
transmcmbranc potential gradients. I f a strong depletion layer
field was developed across the membrane under such a polarization, a significant fraction of the photoinjected electrons would
migratc to the back contact and could be drawn off as a current.
Very recent experiments by Spitler et al.7b using single-crystal
TiOz and multiple reflection evanescent wave optical spectroscopy
to monitor the time course of photosensitized electron injection
have illustrated this behavior. The results obtained with the
present membrane are strikingly different from those obtained
with such single-crystal semiconductor electrodes.
Duc to the high ratio of dye molecules to the volume of TiOz
thc potential scale on Figure 6 must be interpreted with care. The
negative shift of the membrane potential due to the photoinjected
chargc carriers needs to be considered. It was shown above that
this shift increases with the bias voltage compensating the depletion
layer field. Therefore. in the depletion regime the membrane
potential immediately after the laser excitation is much closer to
flat-band conditions than in the equilibrated dark state. From
the dye optical density. the membrane thickness (2.7 pm), and
porosity (30%). it is apparent that the electron concentration in
thc TiOz at the end of the laser pulse is on the order of 1OI9 cnr3.
As the bulk donor density appears to be on the order of IO"
and LIS very little current flows out of the membrane. in the absence
of other electron sinks these electrons will remain in the conduction
band. I n such a situation the Fermi level immediately after the
liiser pulse will be about 100 mV below the conduction band
independent of the depletion bias applied. The particle would be
csscntially in a n accumulation mode equivalent to biasing the
clcctrode to about -0.5 V. i.e.. within about 100 mV of the conduction band edge. This condition would persist during most of
thc rccombination pcriod and the recombination rate should be
therefore independent of applied bias for any pre-laser-pulse Fermi
lcvcl positive of -0.5 V. However, Figure 6 clearly demonstrates
a strong effect on recombination kinetics in this potential domain.
Thc results in Figure 6 can be rationalized in terms of a large
concentration of electron traps that do not act as bulk donor sites.
Interstitial Ti4+,Ti4+sites at grain boundaries, or compensated
oxygen vacancies could all provide the requisite traps. The slow
back reaction occurs between these trapped electrons and the
Ru"'L3 adsorbed at the surface of the TiO?. The trapping sites
arc all filled when the membrane is polarized to potentials negative
of thc conduction band edge. Therefore, the rate constant of 1.5
X IO6 s-' obtained at -0.8 V is attributed to the recombination
of frec conduction band electrons with Ru"'L3 parent ions. Polarizing the electrode positive of the conduction band edge empties
traps and leads to a steep decline in the back electron transfer
rate which decreases by a factor of ca. 100 upon applying a reverse
bias of 0.3 V. Increasing the potential further has little effect
on the charge recombination dynamics. I t is worth noting that
thc argument of the previous two paragraphs is not dependent
on the exact ionizable donor level. Any donor level significantly
less than I O i 9 cm-3 will result in the same conclusion.
A plausible explanation of this observation is that, as the reverse
bi ;IS I S incrcnscd. dcepcr lying traps are emptied. The photoinjected
clcctrons trapped on these sites recombine much more slowly than
thc ones located in shallow traps since their reaction requires
thermal activation or tunneling. The linear decay of the bleaching
signal on a logarithmic time scale in Figure 5c is a further indication that clcctron tunneling is indced involved in the recombination process. Similar logarithmic time laws have been observed
for many charge-transfer reactions involving tunneling of trapped
clcctrons and kinctic models interpreting this behavior have been
publishcd.Ih Thc finding that thcrc is no effect of thc applied
, ' , '
The Journal of Physical Chemistry, Vol. 94, No. 24. 1990 8725
bias on the back electron transfer rate at potentials more positive
than 0.3 V may be attributed to the fact that the density of
trapping levels becomes small as the Fermi level approaches the
middle of the band gap.
I t is noteworthy that the value of the rate constant for the slow
decay when the Fermi level is held at -0.52 V. kb = 3 X IOs s-l,
is very similar to that obtained from colloidal solutions of
RuL3-loadcd IO-nm-sized TiO, particlesqgakb = 4 X IO5 s-!. As
mcntioncd earlier, the flat-band potential of these particles was
determined to be -0.52 V. This suggests that the charge recombination dynamics observed previously with the colloidal dispersions involved also trapped electrons.
I f applying a reverse bias to the TiO, electrode does not generate
a significant net electric field across the porous film, electron
transport from the particles to the back contact should occur by
diffusion rather than migration. In crystalline rutile, the electron
mobility is 0.5 cm2 V-' s-' corresponding to a diffusion coefficient
of 0.02 cm2 SKI. Using the mean square displacement relation d
= (2D1)0.5,
the time required for electrons to diffuse from the center
of the membrane to the tin dioxide contact would be 0.45 ~ s The
.
electron movement in our porous membrane is slower since the
diffusion of the electronc requires hopping between traps and
crossing of grain boundaries, which is expected to reduce greatly
the mobility of the charge carriers.
The steep rise in the photocurrent in Figure 6 occurs in a
potential domain" where the inhibitive effect of the applied potential on the charge recombination approaches its maximum. (It
should be noted that the back reaction is intercepted in these
experiments by iodide reducing the Ru"'L3 to Ru"L3 and assisting
in this way the escape of injected electrons from recombining with
their parent ions.) The photocurrent attains a plateau at around
0.1 V corresponding to an incident 480-nm photon to current
conversion efficiency of 25%. Since LHE is 0.37 at this wavelength
and &j
I , one derives from eq 2 a charge separation yield of
67%. This implies that 67% of the electrons injected into the
membrane reach the back contact and arc drawn off as a current,
the remainder recombining with the Ru"'L3 parent ions. This
is in very good agreement with the electron collection efficiency
of 65% obtained from the transient photocurrent measurement
in Figure 7b under similar conditions.
The portion of the recombination which is not intercepted by
iodide in the steady-state experiments may be related to the first
and fast component of the bleaching recovery in Figure 5b. Since
the rapidly decaying fraction amounts to about one-third of the
total signal, the agreement with the photocurrent and absorption
measurements is practically quantitative. One intriguing feature
of the fast recombination process is that its rate is insensitive to
the applied potential. Upon changing the polarization of the
membrane from -0.8 to 0.6 V, the rate constant for the initial
component changes by less than a factor of 2. This contrasts
sharply with the behavior of the second and major part of the
recombination process whose rate constant decreases by almost
a factor of 1000 within the same potential domain. A convenient
explanation for this observation would be that not enough trap
sites are available to accommodate all the injected electrons.
Excess electrons would thus remain in the conduction band and
in the absence of a strong electric field would recombine with the
dye at a rate similar to that of all the recombination when the
membrane is held negative of flat band. This hypothesis appears
to be ruled out by the observation that the fraction of the injected
charge that is involved in the fast recombination remains relatively
constant when the number of injected electrons is varied by a factor
of 4.
The persistence of a fast rccombination process at positive bias
can be rationalized in terms of shallow trapping sites present on
or near the surface of the TiO, membrane. I f a number of such
=
(16) (a) Inokuti. M.: Hirayama. F. J . Chem. Phys. 1965. 43. 1978. (b)
Tachya, M.; Mozumder, A. Chem. Phys. t e f f . 1974, 28, 87. (c) Milosavljevic,
B. H.; Thomas, J . K. J. Phys. Chem. 1985, 89, 1830.
( 17) I t should be noted that the potential for the photocurrent onset as well
as the rising edge of the i ( c ) curve may be displaced to more negative potentials at higher light intensities.
8726
The Journal of Physical Chemistry, Vol. 94, No. 24, 1990
surface states exist from which an electron recombines with the
dyc with thc w m c ratc as a conduction band electron. the two
processes will be indistinguishable in the present experiments.
Altcrnativcly. the recombination of conduction band electrons may
alwaq3 occur via trapping in such shallow states constituted. e.g..
bq surface Ti4+ ions thnt arc partially coordinatcd by watcr
molecules. Such traps arc on or near the surface, possibly near
a d j c iiiolcculc. Thus. thcy may be located within a region where
no depletion I;iycr fields can be developed under reverse bias. This
would exclude the posbibilitj or intercepting the chargc rcconibination rroiii such traps bq the local clcctrostatic potential gradients foriiicd i n the bulk of the membrane under rcvcrsc bias.
Regarding the probability of populating such surface states, during
the laser pulsc. thc clcctrons may initially occupy traps statistically
rather than ncccssarily thc dccpcst traps first. Surface and bulk
traps thus w i l l bccomc occupied independent of bias. In addition.
an clcctrori i n ;I sliallo~t bulk trap ma} rcturn bricfly to the
conduction band bcforc occupying a deeper trap. In the absence
of a field this clcctron m a y rccombinc with the dyc.
The crfcct of the nicnibranc potential on thc )icld of chargc
injection and luminescence displaycd in Figurc 8 dcscrvcs some
final coninicnts. The blcaching a t the cnd of the laser pulse can
bc cxprwcd b)
whcrc AC is the change in concentration and c the absorption
coefficient of the species in the subscript. The absorption coefficient of the dye ground state is 2.2 X IO4. That of the cation.
thc dyc cxcitcd stiitc. and thc conduction band electron are a11
bclou 1000. Thus the dccrcnsc in the initial bleaching at potentials
negative of -0.2 V indicates that some dye molecules have returned
to the ground state via a process with a half-life less than 50 ns.
This is much faster than the luminescence decay of the dye in
oxygenated solution (-400 ns). The fast decay process could be
cithcr ;I nonradiativc dccay of the dye or a new recombination
process not present at more positive potentials. (A large increase
in lumincsccncc docs not appear until the electrode is biased to
-0.7 V . ) Although with the present experiments it is not possible
to distinguish between these two possibilities. absorption of Ru(bpy), dyes to oxide surfaces where injection should not be possible
has been observed to decrease the luminescence lifetime into the
50-ns time range.I8 In the absence of a new recombination process
thc decrease in bleaching negative of -0.2 V is explained by a
dccrciise i n the ratc of injection relative to nonradiativc decay.
The ratc constant for charge injection is proportional to the
cxtcnt of overlap between occupied states of the excited sensitizer
rcdox system and empty electronic states in the conduction band
+m
k,", =
J-,
I ' ~ * c c ~ u n o c dc E
(8)
whcrc is a frequency factor. The position of the maximum for
thc distribution function of the density of occupied states. DKc.
is obtained by subtracting half of the value of the reorganization
energy A from the excited state redox potential. I n acidic solution.
thc redox potcntinl of the Ru"L3/Ru"'L3 couple in the ground
statc is 1.2 V (SCE).I9 From the energy difference between lowest
cxcitcd and ground state. AE = 1.97 eV. and neglecting entropy
cffccts t h x t of the cxcitcd coniplcx is derived as -0.8 V . Since
A for tris(bipyridy1) complexes of ruthenium is typically close to
0.5 cV.,O the maximum of occupied states of the excited dye i s
prcdicted to bc a t --0.ZZ V . This is some 0.10 cV below the
conduction band edge which doc$ not result in optimal overlap.
IJ
( 1 8 ) Kajiwra. T : Hnshimoto. K . : Kawai. T.: Sakata. T. J. Ph),c. Chem.
1982, 86, 4516.
( I 9) Dcsilvertro. J.: Duonghong. D.: Kleijn. M.: Gratzel. M. Chimia 1982.
4 , 102.
(20) Sutin. N. I n T i r m d i n g in Biological Svs!cmc; Chance. P.. Devault.
D. C.. Schriffer. J . R.. Frauenfelder. H.. Sutin. N.. Eds.: Academic Press:
Nc\r York. 1979.
O'Regan et al.
The overlap should be sensitive to small shifts in the conduction
band edge which could be brought about, for example, by changes
in the pH near the Ti02surface. I f this mechanism is important,
one would also expect the rate constant for charge injection, and
hence
to depend on the bulk solution pH. Due to problems
w i t h the desorption of Ru"L, from the Ti02 surface, the range
of pH accessible is rather small, i.e.. between 3-4.5. Nevertheless,
a distinct reduction of ca. 30% in the injection yield is noted upon
increasing the pH within these limits bearing out qualitatively
the predictions of the kinetic model.
Polarizing the membrane negative of the flatband potential (ca.
-0.35 V ) induces a further decrease in Cbinj which drops to 0.5 at
-0.8 V. This is rationalized in terms of a negative displacement
of the Fermi level of Ti02 which moves into the conduction band
under forward bias. This decreases both the driving force for
electron injection and the density of unoccupied electronic states
available for charge transfer, reducing the ratc of interfacial
electron injection. The electric potential of the membrane could
also affect the nature of linkage between the Ru"L, and the TiO,
surface. The preferred adsorption sites for Ru"L3 are likely to
bc Ti4+ions having high Lewis acidity. These centers serve as
electron traps that would be filled upon polarizing the electrode
negatively to produce accumulation layer conditions. This is
expected to weaken the binding between the sensitizer and the
semiconductor surface, reducing their electronic interaction, and
hence the rate of electron transfer. As the dye interaction with
the surface decreases, the effect of adsorption on the nonradiative
decay rate, postulated above, should decrcase. This expectation
is borne out by the appearance of luminescence when the electrode
is biased to -0.7 V .
&,.
Conclusions
The development of a novel transparent Ti02 membrane supported on conducting glass has allowed application of time-resolved
absorption spectroscopy and amperometry in conjunction with laser
photolysis to scrutinize the effect of an applied bias on the photosensitized electron injection and subsequent charge separation
process. Subtle alterations of the electrode potential lead to
dramatic changes in both interfacial charge-transfer events. Even
without a depletion layer field the electron injection proceeds with
practically 100% quantum yield suppressing the sensitizer luminescence below the detection limit. The luminescence is turned
on by negative polarization where an accumulation layer is produced in the membrane. This intriguing observation is rationalized
in terms of the control by the electric potential of the energetics
and kinetics of charge injection as well as the effect of the surface
charge of the semiconductor on the binding of the sensitizer.
The effect of the applied bias on the back electron transfer is
most pronounced in the vicinity of the flat-band potential where
a potential change of 0.3 V decreases its rate constant by a factor
of almost 1000. The time-resolved and steady-state photoelectrochemical experiments confirm that 2/3 of the injected electrons
reach the back contact and are drawn off as a current. This is
contingent on the presence of iodide as an electron donor. I n the
absence of iodide only 4% of the photoinjected charge is collected
at the back contact of the membrane. Important insight into the
nature of the charge-separation mechanism in the porous membrane consisting of a network of interconnected TiOz particles
was derived from this analysis. Particularly, the importance of
trapping sites on the dynamics of the charge-recombination process
was established. Extension of these time-resolved studies to other
electron- and hole-transfer reactions should yield a wealth of
information on the role of the local electric field in these interfacial
redox events.
Acknowledgment. I t is a pleasure to acknowledge financial
support of this work from the Swiss National Fund of Scientific
Research. We thank Dr. Nazeeruddin for uroviding us with a
sample of RuL,.
Registry No. Ti02, 13463-67-7; Ru"'L,, 129448-52-8; N a l , 768 I 1
X2-5; Ru"L,, 78338-26-8; LiCIO,. 7791 -03-9.