FUNDAMENTAL STUDIES OF WATER
OXIDATION AT MODEL HEMATITE ELECTRODES
PREPARED BY ATOMIC LAYER DEPOSITION

By
Benjamin M. Klahr

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Chemistry – Doctor of Philosophy
2013

ABSTRACT
FUNDAMENTAL STUDIES OF WATER
OXIDATION AT MODEL HEMATITE ELECTRODES
PREPARED BY ATOMIC LAYER DEPOSITION

By
Benjamin M. Klahr

An increasing global demand for energy, combined with an awareness of anthropogenic climate
change, has recently fueled the search for abundant, carbon neutral energy sources.

The sun

offers an enormous amount of energy that is practically inexhaustible and well distributed across
Earth. Thus, it is an ideal source for meeting our future energy needs in a carbon neutral fashion.
This work focuses on using hematite and sunlight to oxidize water, which is the rate limiting step
of splitting water into the energy dense fuel, hydrogen, and the byproduct, oxygen. Hematite is
abundant, absorbs a large fraction of the solar spectrum and has an appropriately placed valence
band for water oxidation. However, the often cited poor bulk properties, and slow charge
transfer kinetics require large applied potentials to oxidize water.

Atomic layer deposition

(ALD) was utilized to deposit uniform thin films of hematite on transparent conductive
substrates as model electrodes to better understand the nature of the limitations in the bulk and at
the surface. Comparison of the oxidation of water to the oxidation of fast redox shuttles allowed
for the separation of bulk and surface processes. A combination of electrochemical impedance
spectroscopy, photoelectrochemical and electrochemical measurements were employed to
determine the cause of the large required applied potential. It was found that photogenerated
holes initially oxidize the electrode surface under water oxidation conditions, which is attributed
to the first step in water oxidation. A critical number of these surface intermediates need to be

generated in order for subsequent hole-transfer steps to proceed. At low applied potentials, these
intermediates are subject to recombination from the large concentration of electrons in the
conduction band due to low band bending. At higher applied potentials, high band bending
eliminates surface recombination and the charge collection efficiency of the electrolyte reaches
unity.

A water oxidation mechanism is proposed to interpret these results.

In addition,

fundamental studies of hematite electrodes coated with the phosphate mediated cobalt oxide
catalyst were performed. The catalyst was found to reduce the surface recombination mentioned
above. However, oxidized cobalt oxide was still subject to electron recombination at low applied
potentials. This recombination was reduced with the use of an alumina blocking layer, which
resulted in a modified hematite electrode capable of oxidizing water with a near unity charge
collection efficiency at low applied potentials.

ACKNOWLEDGEMENTS

I would like to thank my advisor, Professor Tom Hamann, for his patience, wisdom and
guidance throughout my PhD. Tom has provided abundant support and opportunities, for which
I am extremely grateful. Perhaps most importantly, Tom’s critical insight and integrity has
provided a great model which has influenced how I approach scientific research. In addition,
Tom was able to secure funding which allowed me to perform the research presented herein.
Specifically, I would like to thank Michigan State University, the donors of the American
Chemical Society Petroleum Research Fund (#51099-DNI10) and the National Science
Foundation (CHE-1150378) for funding Tom who could in turn fund me.
I would like to thank my colleagues Alex Martinson and Shannon Riha from Argonne
National Lab as well as Sixto Gimenez, Francisco Fabregat-Santiago and Juan Bisquert from
Universitat Jaume I. The fruitful collaborations I have had with these highly capable people
have helped shape me into the scientist I am today. I would also like to thank my lab-mates for
contributing to a fun and productive working environment. Specifically, I am grateful for the
fruitful discussions with Jesse Ondersma and collaborations with Kelley Young and Omid Zandi.
Finally, I would like to thank my friends, family, parents and brothers for their endless
support. I cannot count the ways these people have contributed to my ongoing personal and
professional development. Specifically, I would like to thank my wife and best friend, Ashlea
Klahr. As part of the goal to contribute equally to our partnership, this dissertation represents an
attempt to keep up with her.

iv

TABLE OF CONTENTS

LIST OF TABLES .....................................................................................................................viii
LIST OF FIGURES ...................................................................................................................ix
Chapter 1: Introduction ............................................................................................................1
1.1 Motivation ..................................................................................................................2
1.2 Approach ....................................................................................................................2
1.3 Objectives ..................................................................................................................7
REFERENCES ................................................................................................................8
Chapter 2: Thickness Dependence of Thin Film Hematite Deposited by Atomic Layer
Deposition ..............................................................................................................................12
2.1 Abstract ......................................................................................................................13
2.2 Introduction ................................................................................................................14
2.3 Experimental ..............................................................................................................15
2.3.1 Atomic layer deposition of hematite ...........................................................15
2.3.2 Absorbance Calculation ..............................................................................16
2.3.3 Photoelectrochemical Measurements..........................................................19
2.4 Results and Discussion ..............................................................................................20
2.3.1 Thin Film Characterization .........................................................................20
2.3.2 Photoelectrochemical Measurements..........................................................25
2.5 Conclusions ................................................................................................................38
2.6 Acknowledgements ....................................................................................................39
APPENDIX ......................................................................................................................40
REFERENCES ................................................................................................................42
Chapter 3: Current and Voltage Limiting Processes in Thin Film Hematite Electrodes ...46
3.1 Abstract ......................................................................................................................47
3.2 Introduction ................................................................................................................48
3.3 Experimental ..............................................................................................................50
3.4 Results and Discussion ..............................................................................................51
3.4.1 [Donor] Dependence ...................................................................................51
3.4.2 [Acceptor] Dependence ..............................................................................55
3.4.3 pH Dependence ...........................................................................................57
3.4.4 Redox Potential Dependence ......................................................................61
3.5 Conclusions ................................................................................................................64
3.6 Acknowledgements ....................................................................................................66
APPENDIX ......................................................................................................................67
REFERENCES ................................................................................................................69
Chapter 4: Voltage Dependent Photocurrent of Thin Film Hematite Electrodes ...............72
4.1 Abstract ......................................................................................................................73
4.2 Introduction ................................................................................................................74
v

4.3 Experimental ..............................................................................................................75
4.4 Results and Discussion ..............................................................................................75
4.5 Conclusions ................................................................................................................83
4.6 Acknowledgements ....................................................................................................84
REFERENCES ................................................................................................................85
Chapter 5: Photoelectrochemical Impedance Spectroscopy of Hematite Electrodes: The
Role of Surface States ................................................................................................................87
5.1 Abstract ......................................................................................................................88
5.2 Introduction ................................................................................................................89
5.2.1 Impedance Spectroscopy Background ........................................................89
5.3 Experimental ..............................................................................................................95
5.4 Results and Discussion ..............................................................................................96
5.4.1 Photoelectrochemical Results .....................................................................96
5.4.2 Equivalent Circuit Justification...................................................................100
5.4.3 Fitting Analysis ...........................................................................................101
5.5 Conclusions ................................................................................................................108
5.6 Acknowledgements ....................................................................................................109
APPENDIX ......................................................................................................................111
REFERENCES ................................................................................................................113
Chapter 6: The Nature of Surface States of Water Oxidation at Hematite Electrodes
through the Comparison with a Fast Redox Shuttle ..............................................................116
6.1 Abstract ......................................................................................................................117
6.2 Introduction ................................................................................................................118
6.3Experimental ...............................................................................................................119
6.4 Results ........................................................................................................................120
6.4.1 Steady State Behavior .................................................................................120
6.4.2 Electrochemical Impedance Spectroscopy .................................................121
6.4.3Transient Behavior .......................................................................................127
6.4.4 Cyclic Voltammetry ....................................................................................135
6.5 Discussion ..................................................................................................................137
6.6 Conclusions ................................................................................................................143
6.7 Acknowledgements ....................................................................................................145
APPENDIX ......................................................................................................................146
REFERENCES ................................................................................................................149
Chapter 7: Methanol Oxidation at Hematite Electrodes and the Implications for Water
Oxidation
..............................................................................................................................153
7.1 Abstract ......................................................................................................................154
7.2 Introduction ................................................................................................................155
7.3 Experimental ..............................................................................................................155
7.4 Results ........................................................................................................................157
7.4.1 Steady State Oxygen Evolution ..................................................................157
7.4.2 Electrochemical Impedance Spectroscopy .................................................162
7.4.3 Cyclic Voltammetry ....................................................................................167
vi

7.5 Discussion ..................................................................................................................173
7.6 Conclusions ................................................................................................................174
7.7 Acknowledgements ....................................................................................................175
APPENDIX ......................................................................................................................176
REFERENCES ................................................................................................................182
Chapter 8: Photoelectrochemical and Impedance Spectroscopic Investigation of Water
Oxidation with “Co-Pi” coated Hematite Electrodes .............................................................185
8.1 Abstract ......................................................................................................................186
8.2 Introduction ................................................................................................................187
8.3 Experimental ..............................................................................................................188
8.4 Results ........................................................................................................................191
8.4.1 Steady State Behavior .................................................................................192
8.4.2 Transient Behavior ......................................................................................195
8.4.3 Electrochemical Impedance Spectroscopy .................................................200
8.5 Discussion ..................................................................................................................208
8.6 Conclusions ................................................................................................................212
8.7 Acknowledgements ....................................................................................................213
APPENDIX ......................................................................................................................214
REFERENCES ................................................................................................................218
Chapter 9: Water Oxidation at Co-Pi coated Hematite Electrodes: Reducing
Recombination to Co-Pi with an Al2O3 Blocking Layer ........................................................222
9.1 Abstract ......................................................................................................................223
9.2 Introduction ................................................................................................................224
9.3 Experimental ..............................................................................................................225
9.4 Results and Discussion ..............................................................................................227
9.5 Conclusions ................................................................................................................237
APPENDIX ......................................................................................................................239
REFERENCES ................................................................................................................241
Chapter 10 Conclusions and Future Directions ......................................................................243
10.1 Conclusions ..............................................................................................................244
10.2 Future Directions .....................................................................................................246
REFERENCES ................................................................................................................249

vii

LIST OF TABLES

Table 3-1 Formal potentials of outer-sphere redox shuttles employed ......................................62
Table 5-1 Parameters derived from Mott-Schottky plot under illumination ..............................108
Table 5-2 Correlation of surface state charge and flat band shift ..............................................108
Table 7-1 Methanol and water mixture oxygen evolution faradaic efficiency ..........................160

viii

LIST OF FIGURES

Figure 1-1 Tandem photoelectrochemical cell configuration ....................................................5
Figure 1-2 Solar irradiance and hematite absorption as a function of wavelength ....................6
Figure 2-1 Diagram of transmitted and reflected light used
to calculate hematite absorbance .....................................................................................17
Figure 2-2 XRD and Raman spectra of hematite of FTO ..........................................................21
Figure 2-3 Absorption and reflection spectra;
Absorption as a function of thickness ..............................................................................24
Figure 2-4 SEM images of hematite surface on FTO ................................................................25
Figure 2-5 Open circuit voltage and short circuit current
density as a function of thickness ....................................................................................27
Figure 2-6 APCE calculated as a function of hematite thickness ..............................................29
Figure 2-7 Nyquist plot and Mott-Schottky plot of a hematite electrode ..................................32
Figure 2-8 Normalized APCE for light absorbed adjacent to electrolyte interface ...................34
Figure 2-9 APCE measured with front and backside illumination ............................................35
Figure 2-10 Qualitative energy diagram for hematite-electrolyte
junction under illumination ..............................................................................................37
Figure A2-1 IPCE vs. wavelength for hematite electrodes of varying thickness ......................41
Figure A2-2 APCE vs. wavelength for hematite electrodes of varying thickness .....................41
Figure 3-1 Qualitative energy diagram for hematite-electrolyte
junction under illumination ..............................................................................................49
Figure 3-2 Electron donor concentration dependence ...............................................................54
Figure 3-3 Electron acceptor concentration dependence ...........................................................57
Figure 3-4 pH dependence with and without redox shuttle .......................................................60
Figure 3-5 Redox shuttle dependence ........................................................................................63

ix

Figure A3-1 Open circuit voltage as a function of light intensity and ferricyanide ..................68
Figure A3-2 Photocurrent vs. light intensity..............................................................................68
Figure 4-1 Characteristic light and dark J-V for hematite-ferricyanide junction .......................76
Figure 4-2 Series resistance calculation and measurement of hematite.....................................77
Figure 4-3 Experimental dark J-V curve fit to the diode equation.............................................79
Figure 4-4 Qualitative energy diagram showing
bandbending with built in potential .................................................................................80
Figure 4-5 Experimental J-V with model fits with varying light intensity ................................81
Figure 4-6 Calculation of total charge collection efficiency......................................................83
Figure 5-1 Possible equivalent circuits for interpreting
water oxidation at hematite electrodes .............................................................................92
Figure 5-2 Qualitative energy diagram of quasi- Fermi
levels in the dark and light ...............................................................................................94
Figure 5-3 Water oxidation J-V under varying illumination and pH ..........................................97
Figure 5-4 Characteristic Nyquist plot .......................................................................................98
Figure 5-5 Equivalent circuit parameters fit to 2 models...........................................................99
Figure 5-6 Total resistance measured by impedance and J-V curve ..........................................102
Figure 5-7 Correlation of J-V curve, surface state
capacitance and charge transfer resistance ......................................................................103
Figure 5-8 Calculated density of states from surface state capacitance .....................................104
Figure 5-9 Mott-Schottky plots under varying illumination and pH .........................................106
Figure A5-1 Equivalent circuit parameters for hematite in pH 13.3 .........................................220
Figure 6-1 J-V curve of hematite in contact with ferricyande and water electrolyte ..................121
Figure 6-2 Nyquist plots of hematite in contact with
ferricyanide and water electrolytes ..................................................................................122
Figure 6-3 Equivalent circuits used to fit impedance spectra ....................................................123
x

Figure 6-4 Equivalent circuit parameter fits and Mott-Schottky ...............................................125
Figure 6-5 Total resistance measured by impedance and J-V curve ..........................................127
Figure 6-6 J-V measured under chopped-light illumination ......................................................129
Figure 6-7 Constant voltage current transients ..........................................................................131
Figure 6-8 Charge measured from current transients.................................................................132
Figure 6-9 Lifetime measured from impedance
spectroscopy and current transients .................................................................................135
Figure 6-10 Cyclic voltammetry measured immediately after illumination ..............................137
Figure 6-11 Measuring and correcting for Fermi-level pinning ................................................142
Figure A6-1 Bulk capacitance and trap resistance of
hematite in contact with water .........................................................................................147
Figure A6-2 Bulk capacitance of hematite in contact
with ferri/ferrocyanide electrolyte ...................................................................................147
Figure A6-3 Comparison of single and exponential fits to transient decay ...............................148
Figure A6-4 CV of hematite in contact with ferri-ferrocyanide
electrolyte in the dark.......................................................................................................148
Figure 7-1 Methanol and water mixture J-V and oxygen measurements ..................................159
Figure 7-2 Mott-Schottky and flat band normalization .............................................................162
Figure 7-3 Methanol and water Nyquist plots ...........................................................................164
Figure 7-4 Equivalent circuits used to fit impedance data .........................................................165
Figure 7-5 Equivalent circuit parameter fits ..............................................................................167
Figure 7-6 Cyclic voltammetry measured immediately after illumination ................................169
Figure 7-7 Surface state charge and lifetime by transient analysis ............................................172
Figure A7-1 CV of methanol and methanol/water mixtures......................................................177
Figure A7-2 Scan rate dependence of methanol/water mixtures ...............................................178
xi

Figure A7-3 Steady state vs. chopped light J-V of mixtures .....................................................179
Figure A7-4 Cathodic and anodic transients of hematite in water .............................................180
Figure A7-5 Cathodic and anodic transients of hematite
in water/methanol mixture ...............................................................................................180
Figure A7-6 Cathodic and anodic transients of hematite in methanol .......................................181
Figure A7-7 Comparison of transients in different water/methanol mixtures ...........................181
Figure 8-1 SEM image of Co-Pi coated hematite ......................................................................192
Figure 8-2 Steady state J-V characteristics of varying
Co-Pi thickness on hematite.............................................................................................194
Figure 8-3 Oxygen measurements of Co-Pi coated hematite ....................................................195
Figure 8-4 Co-Pi thickness dependent current transients of Co-Pi coated hematite ..................198
Figure 8-5 Cathodic current transients. Co-Pi thickness and potential dependence ..................199
Figure 8-6 Steady state vs. chopped light for Co-Pi coated hematite ........................................200
Figure 8-7 Nyquist plot of bare and Co-Pi coated hematite ......................................................201
Figure 8-8 Equivalent circuits used in EIS
interpretation for Co-Pi coated electrodes .......................................................................203
Figure 8-9 Capacitance and charge transfer resistance of the Co-Pi .........................................205
Figure 8-10 Comparison of resistances from EIS of Co-Pi coated electrodes...........................206
Figure 8-11 Hematite bulk capacitance and Mott-Schottky
for Co-Pi coated electrodes ..............................................................................................208
Figure 8-12 Mechanistic scheme of Co-Pi coated hematite electrodes ....................................210
Figure A8-1 Current density as a function of Co-Pi
thickness on Co-Pi coated hematite .................................................................................215
Figure A8-2 Chopped light vs. steady state J-V for varying thicknesses of Co-Pi ....................216
Figure A8-3 Light and dark Mott-Schottky plot of Co-Pi coated hematite ...............................217

xii

Figure 9-1 Effect and stability of alumina on hematite electrodes ............................................228
Figure 9-2 Effect of alumina blocking layer thickness
between Co-Pi and hematite ............................................................................................230
Figure 9-3 Steady state oxygen production of
hematite-alumina-Co-Pi electrodes ..................................................................................231
Figure 9-4 Steady state vs. chopped light for Co-Pi coated hematite ........................................232
Figure 9-5 Nyquist and equivalent circuit for hematite-alumina-Co-Pi electrodes ....................234
Figure 9-6 Capacitance and charge transfer resistance
of Co-Pi in composite electrodes ....................................................................................236
Figure 9-7 Mechanistic scheme of hematite-alumina-Co-Pi electrodes ...................................237
Figure A9-1 Full J-V curve for hematite-alumina-Co-Pi composite electrodes ........................240

xiii

Chapter 1:
Introduction

1

1.1 Motivation
The total average worldwide primary power demand was approximately 15 TW in 2010
(corresponding to the use of 484 x 10

18

1

J).

Due to the world’s increasing population and

industrialization, the total primary power demand is expected to increase to 23 TW by the year
1

2035. Approximately 70 % of the supplied power is, and is expected to continue to be, derived
from the combustion of fossil fuels.

1

This combustion is the primary component of huge levels

of carbon dioxide emissions of approximately 30 billion metric tons a year, which is expected to
1-3

increase to 40 billion metric tons by 2030.

It is now widely accepted that that the release of
3

carbon dioxide from this combustion is a major contributor to global climate change. There are
many negative environmental consequences associated with climate change, including increasing
droughts, increasing population experiencing water stress, widespread coral mortality, decreases
in crop productivity, and a rise in sea-level.

2,4

In addition to environmental hazards, the reliance

on imported fossil fuels can lead to political and economic instability. Therefore, it is imperative
to develop easily distributable, carbon-neutral sources of energy in order to minimize these
2,4

consequences.

1.2 Approach
Of all of the carbon-neutral power options that currently exist (solar, wind, biofuels,
hydro, geothermal), solar is the only one that can realistically be implemented on a large enough
5-7

scale to supply the world’s energy demands.

The sun illuminates the earth with ~120,000 TW

of power, where utilizing ~0.01% of it has the potential of satisfying the world’s energy demands
in a carbon neutral fashion.

6

Solar energy is well-distributed across the most populous areas of
2

the globe, which provides the resource accessible where it is needed most. This makes solar
energy an ideal method of powering the planet.
Efficient solar energy conversion has been around for decades through the use of
crystalline-silicon photovoltaics.

However, these devices require highly pure materials which
-1

increase the manufacturing costs. This results in a high price of $0.35 (kW-hr)

which is not
5,6,8

competitive with the cost of energy derived from fossil fuels of 0.02 to 0.05 (kW-hr)-1.

The

high cost of electricity production from solar conversion continues to thwart wide spread use of
this technology. In addition, solar energy is diurnal and intermittent, while our energy demands
are not. Thus, if the vast solar energy resource is ever to be used to supply a significant fraction
of the world’s ever-increasing energy demand, cheap and efficient methods of energy collection
and storage need to be developed.
Analogous to photosynthesis, storing solar energy in chemical bonds (i.e. solar fuels)
would be an ideal method.

This is preferable for certain applications (e.g. transportation)

compared to storing electricity in batteries which are still expensive and have limited capacities.
A particularly attractive example of this approach is using solar energy to split water into H 2 and
O2, where the H2 could subsequently be used in a fuel cell or other energy conversion scheme.
This process can be described by the simplified equations:
4H  4e  2H2

(1-1)

2H2O  O2  4H  4e

(1-2)

Forty years ago, the breakthrough report by Fujishima and Honda demonstrated the feasibility of
this approach with phtoelectrochemical (PEC) water splitting on TiO2 photoelectrodes.

3

9

Unfortunately, TiO2 has a large band gap of 3.2 eV and absorbs only UV light making it
impractical for economic water splitting. Since then, there has been significant effort to develop
efficient, unassisted water splitting systems. In order to achieve this, a given material must
satisfy a number of requirements including strong visible light absorption, efficient charge
carrier separation and transport, facile interfacial charge-transfer kinetics, appropriate positions
of the conduction and valence band energy levels with respect to required reaction potentials and
good stability in contact with aqueous solutions.

10

While such systems were heavily investigated

several decades ago, no material so far has fulfilled all the required conditions.

11,12

Realistically, it is expected to be very challenging to synthesize or discover a material
that is capable of absorbing a sufficient fraction of the solar spectrum, as well as providing
enough driving force for both reactions (1-1) and (1-2).

An alternative approach would be to

design a system that contains two absorbers, one absorbing long wavelengths, and one absorbing
shorter wavelengths. In addition to being capable of absorbing a greater percentage of the solar
spectrum, this has the benefit of being able to separately tune the potential of holes capable of
oxidizing water (which come from the conduction band of the semiconductor on the left in
Figure 1-1), and the potential of electrons capable of reducing protons (which come from the
valence band of the semiconductor on the right in Figure 1-1).

13

Also, this design takes into

account the requisite practical consideration of being able to separate the evolved H2 and O2.

4

Figure 1-1.

Relative energy diagram of a photolectrochecmial cell employing an oxygen

evolving (orange) and a hydrogen evolving (gray) semiconductor in tandem. For interpretation
of the references to color in this and all other figures, the reader is referred to the electronic
version of this dissertation.

Hematite (α-Fe2O3) is one promising candidate for the water oxidation half reaction
(semiconductor on the left of Figure 1-1) of PEC water splitting due to its adequate absorption of
visible light up to 590 nm through its ~2.1 eV band gap, and its valence band energy which,
14-20

through the absorption of light, generates energetic holes capable of oxidizing water.

Importantly, for practical reasons, hematite is nontoxic, abundant and extremely stable under
17,21

caustic conditions.

This would ensure the economic viability for long term, large scale

applications. Attempts to utilize hematite electrodes for solar energy conversion, however, have
resulted in very modest energy conversion efficiencies.

22,23

The combination of a low minority

charge carrier mobility and short lifetimes, results in a very short charge collection length of

5

13,15

between 2 and 20 nm.

The minority charge mobility mechanism is best described by small

polaron hopping, thus the short charge collection length is inherent to the chemical structure of
hematite.

24,25

This short charge collection length is especially problematic since the light

absorption depth can be up to 375nm for 550 nm light. Figure 1-2 shows the value 3/α , which
according to Beer’s law is the thickness required to absorb 95% of the light at that wavelength.

Figure 1-2. AM 1.5 Solar Irradiance (black line) and the thickness required for hematite to
absorb 95% of the light, 3/ (orange line).

An early attempt to overcome this short collection length was performed by Itoh and
Bockris who stacked thin film hematite-substrate layers which effectively absorbed more light
while collecting a higher fraction of minority carriers as more charge carriers were generated
within a short distance of the semiconductor-electrolyte interface.

26,27

Subsequently, many

techniques including colloidal solution deposition, reactive ballistic deposition, electrochemical
anodization, spray pyrolysis, and chemical vapor deposition have been used to create new
nanostructures of hematite in attempts to maximize light absorption while minimizing the length

6

minority carriers must travel before being collected.

14,19,28-35

Despite these attempts, the best

electrodes only produce current densities on the order of five times lower than the possible 12.5
mA cm

-2

of a semiconductor with a 2.1 eV band gap.

36

In addition, a large applied potential is

required to achieve this currents, which has thus far been attributed to slow water oxidation
kinetics at the surface of hematite.

13,37

1.3 Objectives
It is the purpose of this dissertation to deeply understand the limitations of hematite in the
bulk, and the nature of the limited charge transfer at the surface. Recently, such attempts to
study these limitations were performed with nanostructured electrodes. While these have the
benefit of generating a significant amount of charge near the surface and are able to achieve
relatively high photocurrent densities, identifying the precise structure-function relationship is
difficult utilizing such a variety of nanostructured electrode morphologies. In addition many
nanostructured electrodes are difficult to reproduce and often have inconsistent real surface
areas. In this work we use atomic layer deposition (ALD) to deposit hematite electrodes in order
to investigate the nature of both bulk and charge transfer limitations of water oxidation. ALD
allows the reproducible deposition of pinhole-free, conformal films with a controllable thickness
at angstrom resolution.

38-40

Importantly, the self-limiting gas phase mechanism of ALD is an

ideal technique for depositing thin films on high aspect ratio scaffolds which are deposited
identically to the thin films used herein.

14

Thus, the lessons learned on the model thin films can

be directly applied to nanostructured electrodes prepared by ALD. The thin, planar films used
herein provide an ideal electrode for fundamental studies as they avoid the complexity and
irregularities of most nanostructured surfaces.
7

REFERENCES

8

REFERENCES

1. **** Energy Information Administration: www.eia.doe.gov.
2. Pachauri, R. K. a. R., A. IPCC 2007: Climate Change 2007: Synthesis Report. Contribution
of Working Groups I, II, and III to the Fourth Assessment Report to the Intergovernmental Panel
on Climate Change, 2008.
3. Moomow, W.; Yamba, F.; Kamimoto, M.; Maurice, L.; Nyboer, J.; Urama, K.; Weir, T. In
IPCC Special Report on Renewable Energy Sources and Climate change Mitigation; Edenhofer,
O., Pichs-Madruga, R., Sokona, Y., Seyboth, K., Matschoss, P., Kadner, S., Zwickel, T.,
Eickemeier, P., Hansen, G., Schlömer, S., von Stechow, C., Eds.; Cambridge University Press:
United Kingdom and New York, NY, USA, 2011.
4. Kerr, R. A. Science, 2007, 318, 1230-1231.
5. Lewis, N. S. MRS Bull. 2007, 32, 808-820.
6. Crabtree, G. W.; Lewis, N. S. Phys. Today 2007, 60, 37-42.
7. Lewis, N. S.; Nocera, D. G. PNAS, 2006, 103, 15729-15735.
8. Green, M. A.; Emery, K.; Hishikawa, Y.; Warta, W. Prog. Photovoltaics, 2009, 17, 85-94.
9. Fujishima, A.; Honda, K. Nature, 1972, 238, 37-38.
10. Walter, M. G.; Warren, E. L.; McKone, J. R.; Boettcher, S. W.; Mi, Q. X.; Santori, E. A.;
Lewis, N. S. Chem Rev. 2010, 110, 6446-6473.
11. Bolts, J.; Wrighton, M. J. Phys. Chem. 1976, 80, 2641-2645.
12. Butler, M. A.; Ginley, D. S. J. Mat. Sci. 1980, 15, 1-19.
13. Dareedwards, M. P.; Goodenough, J. B.; Hamnett, A.; Trevellick, P. R. J. Chem. Soc.,
Faraday Trans. 1 F 1983, 79, 2027-2041.
14. Lin, Y. J.; Zhou, S.; Sheehan, S. W.; Wang, D. W. J. Am. Chem. Soc. 2011, 133, 2398-2401.
15. Kennedy, J. H.; Frese, K. W. J. Electrochem. Soc. 1978, 125, 709-714.
16. Marusak, L. A.; Messier, R.; White, W. B. J. Phys. Chem. Solids, 1980, 41, 981-984.
17. Hardee, K. L.; Bard, A. J. J. Electrochem. Soc. 1977, 124, 215-224.

9

18. Hamann, T. W. Dalton Trans. 2012, 41, 7830-7834.
19. Duret, A.; Gratzel, M. J. Phys. Chem. B 2005, 109, 17184-17191.
20. Bindaar, G.; Dareedwards, M. P.; Goodenough, J. B.; Hamnett, A. J. Chem. Soc. - Far.
Trans. 1983, 79, 1199-1213.
21. Finklea, H. O. Semiconductor Electrodes; Elsevier: Amsterdam, 1988.
22. Lindgren, T.; Wang, H. L.; Beermann, N.; Vayssieres, L.; Hagfeldt, A.; Lindquist, S. E. Sol.
Energy Mater. Sol. Cells, 2002, 71, 231-243.
23. Sivula, K.; Le Formal, F.; Grätzel, M. ChemSusChem, 2011, 4, 417-417.
24. Goodenough, J. B. Prog. Solid State Chem. 1971, 5, 145-399.
25. Kerisit, S.; Rosso, K. M. J. Chem. Phys. 2007, 127, 124706-124710.
26. Itoh, K.; Bockris, J. O. J. Appl. Phys. 1984, 56, 874-876.
27. Itoh, K.; Bockris, J. O. J. Electrochem. Soc. 1984, 131, 1266-1271.
28. Bjorksten, U.; Moser, J.; Gratzel, M. Chem. Mat. 1994, 6, 858-863.
29. Hahn, N. T.; Ye, H. C.; Flaherty, D. W.; Bard, A. J.; Mullins, C. B. ACS Nano 2010, 4,
1977-1986.
30. Mohapatra, S. K.; John, S. E.; Banerjee, S.; Misra, M. Chem. Mater. 2009, 21, 3048-3055.
31. Sivula, K.; Zboril, R.; Le Formal, F.; Robert, R.; Weidenkaff, A.; Tucek, J.; Frydrych, J.;
Gratzel, M. J. Am. Chem. Soc. 2010, 132, 7436-7444.
32. Cherepy, N. J.; Liston, D. B.; Lovejoy, J. A.; Deng, H. M.; Zhang, J. Z. J. Phys. Chem. B,
1998, 102, 770-776.
33. Cesar, I.; Sivula, K.; Kay, A.; Zboril, R.; Graetzel, M. J. Phys. Chem. C, 2009, 113, 772782.
34. Kay, A.; Cesar, I.; Gratzel, M. J. Am. Chem. Soc. 2006, 128, 15714-15721.
35. Beermann, N.; Vayssieres, L.; Lindquist, S. E.; Hagfeldt, A. J. Electrochem. Soc. 2000, 147,
2456-2461.
36. Tilley, S. D.; Cornuz, M.; Sivula, K.; Grätzel, M. Angew. Chem. Int. Ed. 2010, 49, 64056408.

10

37. Upul Wijayantha, K. G.; Saremi-Yarahmadi, S.; Peter, L. M. Phys. Chem. Chem. Phys.
2011, 13, 5264-5270.
38. George, S. M. Chem. Rev. 2010, 110, 111-131.
39. Hamann, T. W.; Martinson, A. B. F.; Elam, J. W.; Pellin, M. J.; Hupp, J. T. Adv. Mater.
2008, 20, 1560–1564.
40. Hamann, T. W.; Martinson, A. B. F.; Elam, J. W.; Pellin, M. J.; Hupp, J. T. J. Phys. Chem.
C 2008, 112, 10303-10307.

11

Chapter 2:
Thickness Dependence of Thin Film Hematite
Deposited by Atomic Layer Deposition

Adapted with permission from:
Photoelectrochemical Investigation of Ultrathin Film Iron Oxide Solar Cells Prepared by Atomic
Layer Deposition, Benjamin M. Klahr, Alex B.F. Martinson, and Thomas W. Hamann,
Langmuir, 2011, 27, (1), 461-468. Copyright 2011 American Chemical Society.

12

2.1 Abstract
Atomic layer deposition was used to grow conformal thin films of hematite with
controlled thickness on transparent conductive oxide substrates.

The hematite films were

incorporated as photoelectrodes in regenerative photoelectrochemical cells employing an aqueous
[Fe(CN)6]

3-/4-

electrolyte. Steady state current density vs. applied potential measurements under

monochromatic and simulated solar illumination were used to probe the photoelectrochemical
properties of the hematite electrodes as a function of film thickness.

Combining the

photoelectrochemical results with careful optical measurements allowed us to determine an
optimal thickness for a hematite electrode of ~20 nm. Mott-Schottky analysis of differential
capacitance measurements indicated a depletion region of ~17 nm. Thus, only charge carriers
generated in the depletion region were found to contribute to the photocurrent.

13

2.2 Introduction
The water oxidation efficiency of hematite electrodes suffers due to a large amount of
recombination within the bulk. This is often attributed to the very low minority collection
lengths of between 2 and 20 nm.

1,2

The studies often cited, have calculated these values by

fitting experimental photoelectrochemical measurements of water oxidation to the Gartner
equation. This derivation is made under the assumption that all minority carriers generated
within the depletion layer are able to reach the surface. While this may be the case of many
semiconductor electrodes, the uniquely low minority carrier lifetime of hematite prevents this
from being true (see below and chapter 4).

3-5

This indirect method for calculating the minority

collection length therefore presents a significant opportunity for error.
Another way of measuring these lengths is to synthesize thin films that have a thickness
on the order of the charge collection length. In addition, thin films should serve as excellent
model system to determine the feasibility of overcoming the poor bulk properties of hematite by
controlling the dimensions. To this end spray pyrolysis was used to prepare thin films of
hematite for water oxidation photocatalysis.

6,7

While spray pyrolysis is a convenient method for

preparing thin films, the films are not especially uniform, the thickness is difficult to control, and
the reproducibility is questionable.

Other techniques, including chemical vapor deposition,

electrodeposition, sputtering, and more recently reactive ballistic deposition, have also been used
to prepare films of hematite, however they all suffer from similar lack of precise dimensional
control and uniformity.

8-13

Further, since water is generally the redox species, the overall

performance may depend on the water oxidation step, not strictly the hematite electrode, which
can complicate interpretation of the bulk parameters.

14

In this work we use atomic layer deposition, ALD, to deposit conformal thin films of iron
oxide to investigate the photoelectrochemical properties as a function of film thickness. ALD is
a self-limiting process which deposits at most a single atomic layer at a time by alternating pulses
of a metal precursor and an oxidizer.

14

Hematite has been deposited previously via ALD

employing ferrocene and oxygen as the metal precursor and the oxidizer, respectively, however
recent work performed by Martinson et. al. has demonstrated high growth rates at low deposition
15-19

temperatures can be achieved by employing ozone as the oxidizer.

Ferrocene and ozone

were therefore selected as the precursors for the ALD of iron oxide in this work. The unique
self-limiting property of ALD results in uniform films with highly controllable thicknesses.

20,21

Importantly, since ALD is not a line-of-sight technique, it can be used to make vertically oriented
‘thin films’ which are prepared in an identical fashion to the model thin films.

22

Thus lessons

learned with our model system thin films can be directly applied to nanostructured architectures
with precisely controlled dimensions fabricated via ALD.

16

2.3 Experimental
2.3.1 Atomic Layer Deposition
Atomic layer deposition was performed using a Savannah 100 instrument (Cambridge
Nanotech Inc.). Thin films of iron oxide were deposited on fluorine doped tin oxide (FTO)
-2

coated glass (Hartford Glass, 12 Ω cm ) by ALD using ferrocene and O3 as precursors.
Cleaned FTO substrates were treated with 0.1 M HCl, rinsed with isopropanol and blown dry
with nitrogen prior to deposition. Ferrocene (Aldrich) was purified by sublimation by heating to
70°C on a hotplate prior to use. The O3 was made by a Yanco Industries ozone generator (10

15

wt% in 62 sccm ultra high purity O2). The nitrogen carrier gas was kept at 5 sccm. The
ferrocene precursor cylinder was heated to 70°C, however the vapor pressure was still low and
not detectable by our pressure gauge.

The substrate chamber was maintained at 200°C.

Exposure times of 20 and 60 seconds were used for ferrocene and O3, respectively, with an 8
second nitrogen purge time between pulses. After deposition, iron oxide electrodes were heated
to 500 °C at a rate of 17 °C / min., sintered at 500 °C for 30 minutes and allowed to cool to room
temperature over 2 hours. Sintering was found to increase the stability of the electrodes. The
ALD of iron oxide was found to be very sensitive to the substrate surface (e.g. growth rate and
uniformity), therefore all measurements presented herein were performed on films deposited on
nominally identical FTO substrates. Three series of electrodes were made by varying cycles from
100 up to 1,200 cycles. All values reported are the average of the data sets acquired with the
three series of electrodes with errors bars representing the standard deviation.
Films were characterized by X-ray diffraction, XRD, using Cu Κα radiation (Rigaku Dmax operating at 45 KV and 100 mA) and Raman spectroscopy.

Small angle XRD was

performed using a Bruker D2 Phaser. Raman spectra were recorded using a 100 mW argon ion
laser (Melles Griot CW) at 514.5 nm, an Olympus BH-2 microscope assembly, and a Spex 1250
spectrograph with a 600 grooves/mm holographic grating. The detector was a Symphony 2000 x
800 charge-coupled device (Horiba Jobin-Yvon). Morphology and thickness of the films were
determined by field emission scanning electron microscopy using a Hitachi S-4700 FESEM.
Absorbance and reflectance measurements were made using a Perkin Elmer Lambda 35 UV-Vis
spectrometer with a Labsphere integrating sphere.

16

2.3.2 Absorbance Calculation
The absorbance spectra of the films were measured by illuminating from the substrateelectrode (SE) interface such that the incident light had to be corrected for passing through and
being reflected by the substrate as illustrated in Figure 2-1.

a)

b)

Figure 2-1. Diagram indicating the light reflectance and transmittance of (a) the FTO substrate
(S) and (b) the substrate plus hematite film (H).

All parameters are a function of wavelength

and are described in the text. Parameters in blue are directly measurable and terms in red are
calculated.

The Napierian absorbance of the hematite thin films, AH, at a given wavelength, λ, is given by
the equation:


I  
AH      ln 

 I '    1  RH     



(2-1)

where I(λ) is the intensity of light that is transmitted through the substrate and hematite film, I’(λ)
is the intensity of light that is transmitted through the substrate, and RH(λ) is the reflectance of
the substrate/hematite interface (Figure 2-1b). The measured transmittance of the substrate,
TS(λ), is given by

17

TS    

I ' 
I0   

(2-2)

where I0(λ) is the incident light intensity (Figure 2-1a). The measured transmittance of the
substrate and hematite film, TS+H(λ), is given by

TS  H    

I  
I0   

(2-3)

(Figure 2-1b). The ratio of the measured transmittances is therefore:

TS  H    I   

TS   
I ' 

(2-4)

Substituting equation (2-4) into equation (2-1) yields:

T
   / TS    
AH   ln  S  H

 1  RH     



(2-5)

The value of RH(λ) that is needed is the fraction of light incident on the substrate/hematite
interface, I’(λ), that is reflected. The measured reflectance of the substrate/hematite interface is
given by the difference of the measured reflectance of the substrate and the measured reflectance
of the substrate and hematite: RS  H     RS    . This expression, however, does not take into
account the light that is absorbed by the substrate. The reflectance in terms of I’(λ) is therefore
obtained by dividing the measured reflectance ( RS  H     RS    ) by the transmitted light,
TS(λ) incident on the interface. The light reflected by the substrate/hematite interface actually
passes through the substrate twice (once to the interface, then again to the detector), therefore the
reflected light needs to be divided by TS again. The corrected expression for RH(λ) is then given
by:

18

RH    

 RS  H     RS    
TS    2

(2-6)

Substituting equation (2-6) into (2-5) yields the final equation for the absorbance of the hematite
film, fully corrected for reflectance and absorbance of the substrate where all parameters are
measureable, which is employed herein:






TS  H    / TS   

AH   ln 
   RS  H     RS      

 1 
2


TS   



(2-7)

A similar equation was presented by Beerman et. al. to find the absorbance of nanorod hematite
23

on a FTO substrate.

2.2.3 Photoelectrochemical Measurements
Iron oxide electrodes were masked with a 60 μm Surlyn film (Solaronix) to prevent
scratching of the very thin films. Surlyn films were first adhered to the surface at 85°C and then
melted to the electrode at 115°C.

Electrodes were clamped to a custom made glass

electrochemical cell. The solution was referenced to a homemade Ag/AgCl electrode which was
frequently compared to a Radiometer Analytical saturated calomel electrode (SCE).

High

surface area platinum mesh was used as a counter electrode. All experiments shown used an
aqueous electrolyte containing 200 mM KCl, 20 mM K3Fe(CN)6 (Sigma-Aldrich), and 200 mM
K4Fe(CN)6·3H2O (Mallinckrodt Baker) which were purified via re-crystallization. Aqueous
solutions were maintained at pH 12 (KOH). Sandwich cells were made by applying light
pressure to 120 m Surlyn film sandwiched between a hematite electrode and a cleaned FTO
counter. Electrochemical impedance measurements were made using an Eco Chemie Autolab
19

potentiostat coupled with Nova electrochemical software using a 10 mV amplitude perturbation
5

and sweeping frequencies from 10 to 10 Hz. Capacitance data were fit using ZView software
(Scribner Associates).

Photoelectrochemical measurements were made with a Gamry

Instruments Reference 600 potentiostat. The light source was a Xe arc lamp interfaced with a
Horiba Jobin Yvon MicroHR monochromator. An AM 1.5 solar filter and neutral density filters
2

were used to simulate sunlight at 100 mW/cm .
2.4 Results and Discussion
2.4.1 Thin Film Characterization
While growing samples for this study, additional thick-film samples were grown
concurrently in our ALD reactor, in order to facilitate characterization. The powder XRD
spectrum for an annealed thick film (consisting of approximately 3,700 ALD cycles and
estimated to be ~200 nm by SEM) of iron oxide deposited on an FTO substrate is displayed in
Figure 2-2a. The XRD spectrum is dominated by peaks due to the underlying FTO substrate.
There are additional well-defined peaks present, however, which can be assigned to hematite (αFe2O3). Peaks were assigned using the International Union of Crystallography (IUCr) database,
in agreement with films using different deposition methods including spin coating and spray
pyrolysis.

24,25

XRD spectra obtained from thinner films had very low signal which could not be

interpreted unambiguously. Low angle XRD was also performed on films 10-60 nm and no
peaks could be seen down to 1 degree.
Raman spectroscopy is often used to analyze the structure of iron oxides due to the very
different spectra of the common forms: hematite (α-Fe2O3), maghemite (γ-Fe2O3) and magnetite
(Fe3O4). Figure 2-2b shows a Raman spectrum gathered from the same thick film (~200 nm) of

20

iron oxide deposited on FTO as the XRD measurements. All peaks correspond with Raman
scattering from the crystal structure of α-Fe2O3 previously reported except for the peaks at 660
-1 26

and 1320 cm .

Although magnetite has a very strong characteristic peak at 660 cm

-1

no other

signals of magnetite were found in the XRD spectrum. This peak has been attributed to disorder
in the crystal lattice and the breaking of symmetry upon crystallization.

27

The peak at1320 cm

-1

-1

has been found to be related to the peak at 660 cm as an overtone peak. Both peaks at 660 and
1320 cm

-1

are expected to be Raman inactive in perfect crystals.

around 1100 cm

-1

28

A hump can also be seen

which is likely due to background scattering from the FTO. The large

scattering of the substrate thwarted attempts to measure thinner films ( 60 nm) on FTO with
Raman spectroscopy; no distinguishable peaks were evident, even when integrating the signal for
up to 6 hours. The combination of XRD and Raman data show that the only observable
crystallographic phase in the iron oxide deposited with ALD using ferrocene and ozone is αFe2O3 (hematite).

21

Figure 2-2. (a) X-ray diffraction pattern and (b) Micro Raman spectrum of iron oxide film
prepared by ALD on FTO. Peaks are labeled for -Fe2O3(*) and the SnO2 substrate (•).
Figure 2-3a shows the absorbance spectra, which have been corrected for reflectance and
the FTO substrate absorbance, of hematite films grown with an increasing number of ALD
cycles. The absorption onset at around 560 nm for all samples is consistent with a 2.1-2.2 eV
bandgap of hematite reported in many other studies.

29-33

Figure 2-3b shows the absorbance at

a single wavelength (410 nm) as a function of ALD cycles from 100 to 1,200 cycles. The data
represent the averaged value from measurements of three independently prepared series of thin
film hematite electrodes with error bars representing the standard deviation. Ideal ALD behavior
predicts that thickness should increase linearly with cycles, however the absorbance increases
linearly with ALD cycles only over the first 500 cycles. A recent report suggested that strain
could produce increased absorption coefficients, , in thin (~25 nm) hematite films.

24

In that

work, the reflectance of the films was not accounted for which can be significant as shown in
Figure 2-3b. In addition, the thickness was determined by SEM, which is not a particularly
accurate method. Therefore we regard conclusions about strain induced changes in α with
skepticism. A recent investigation of the ALD mechanism of hematite using ferrocene and ozone
as precursors indicated that the thinnest (6.7 nm) films had slightly higher  compared to films
19

12 nm; films thicker than 12 nm had essentially constant .

We note that a different substrate

was employed in that work (quartz) which can have a significant effect on the very thinnest films.
The growth rate, however, was reported to be linear. In any case, the initial linear increase in
absorbance with ALD cycles is either due to linear growth with an approximately constant , or
coincidentally offsetting growth-rates and . We apply Ockham's razor and assume the former:

22

approximately linear growth for films grown with up to 500 ALD cycles and constant . Using
5

an absorption coefficient of 2.2 × 10 cm

-1

at 410 nm (this value is in agreement with several

other reports and is used to calculate thickness for other films throughout this paper

8,19,29

), the

growth rate is calculated to be approximately 0.62 Å per cycle by analyzing the slope of the best
29

fit line for the first 500 cycles in Figure 2-3c.

After 500 ALD cycles, the growth rate decreases

and becomes slightly non-linear. A detailed analysis of the ALD mechanism of hematite using
19

ferrocene and ozone as precursors is presented elsewhere.

From here on, the electrode

thicknesses will be referenced in terms of nanometers instead of ALD cycles, taken from the data
presented in Figure 2-3c. The thicknesses given represent average thicknesses resulting from a
given number of ALD cycles from at least three independently prepared samples. For example, 3
electrodes made in separate batches using 300 ALD cycles have thicknesses of 20 ± 2.2 nm,
which are all referred to as 20 nm electrodes for the sake of ease of discussion.

23

Figure 2-3. (a) Calculated absorbance spectra and (b) measured reflectance spectra of hematite
thin films prepared with 100, 200, 300, 500, 700, 900, 1000, 1200 ALD cycles. (c) Relationship
of the absorbance at 410 nm with ALD cycles. The line of best fit only represents the first 5 data
points (orange circles).

SEM images can be seen in Figure 2-4. Figures 2-4a, 2-4b and 2-4c are top down views
of a blank FTO substrate, a 22 nm film and a 50 nm film respectively. The texture of the blank
FTO is maintained with a 22 nm film deposited on top of it suggesting deposition is uniform and
conformal as expected. However, the 50 nm films show the formation of spherical particles

24

suggesting that at some point, the ALD process of hematite changes. This change in morphology
with thicker films is consistent with the deviation from linear growth shown in Figure 2-3c. The
cross section image of a 1,200 cycle electrode, shown in Figure 2-4d, indicates that the thickness
is approximately 50 nm. The thickness, d = 50 nm, is in excellent agreement with the absorbance
5

of this electrode, AH = 1.1, and the absorption coefficient,  = 2.2 × 10 cm

-1

at 410 nm,

according to Beer’s Law (AH = d).

Figure 2-4. Top down SEM images of (a) blank FTO (b) 22 nm film (c) 50 nm film. (d) A cross
section SEM image of a 50 nm film.

2.4.2 Photoelectrochemical Measurements
Steady state current density, J, vs. applied potential, E, measurements were performed on
three series’ of thin film hematite electrodes ranging from 7 to 56 nm in contact with an aqueous
ferricyanide/ferrocyanide electrolyte.

34,35

Figure 2-5a shows representative J-E curves measured

25

-2

in the dark and in response to simulated AM 1.5 illumination (100 mW cm ). Figure 2-5b
shows the average short circuit current densities, JSC, as a function of thickness. The JSC values
were taken at the Nernstian potential, ES, of the ferricyanide/ferrocyanide electrolyte (190 mV vs.
SCE) since this potential best approximates the short circuit current density of a 2-electrode cell.
Initially, for the ~3 thinnest electrodes, JSC increases linearly with thickness. The increased
photocurrent is partially due to the thicker electrodes absorbing more photons. The JSC increases
faster than can be accounted for just by an increase in light absorption however; the 20 nm
electrodes absorb approximately twice as many photons as the 7 nm electrodes, but has over four
times the photocurrent.

In other words, the absorbed photon to current efficiency, APCE,

appears to increase for these electrodes. The photocurrent is essentially constant for electrodes
thicker than 20 nm, despite the fact that these electrodes absorb an increasing fraction of incident
light, indicating the APCE decreases for these electrodes.

26

Figure 2-5. (a) Current density vs. applied potential curves measured in the dark (solid orange
line) and under simulated AM 1.5 illumination (dashed red line) for a 20 nm thick electrode. (b)
Photocurrent density and (c) open circuit voltage extracted from J vs. E curves vs. film thickness.

The APCE is calculated by the equation: APCE ( ) 

IPCE ( )
, where IPCE(λ) is
Absorptance( )

the incident photon-to-current efficiency at a given wavelength and the absorptance is the
percentage of incident photons absorbed by the hematite at that wavelength. Therefore, IPCE
measurements were performed for all of the α-Fe2O3 electrodes. IPCE measurements were also

27

measured at ES. Figure 2-6a shows a representative IPCE spectrum for a 20 nm electrode
(additional IPCE spectra are provided in the supporting information). The dominant trends of the
IPCE spectra are consistent with the JSC results, as expected since the photocurrent under white
light is just the sum of photocurrent response from monochromatic light. The reflectance of
hematite varies with respect to number of ALD cycles/thickness (see Figure 2-3b), which is
therefore convoluted in the trends of photocurrent and IPCE response to increasing film
thickness. In order to correct for any reflectance dependence, the absorptance was calculated by

Absorptance  1  e AH where AH is calculated in equation (2-7). Figure 2-6a shows the
spectrum of the percentage of incident photons that are absorbed by the 20 nm hematite film
overlaid with the IPCE spectrum. The APCE’s were then calculated by dividing the IPCE
spectra by the percentage of photons absorbed at each wavelength (spectra shown in appendix 2).
Figure 2-6b shows the APCE at different wavelengths as a function of thickness. The overall
trend shows an increase in APCE with thickness for electrodes up to 20 nm thick, consistent with
the large initial increase in JSC. The APCE’s decrease for all thicker electrodes as expected for a
constant JSC and increasing light absorption.

28

Figure 2-6. (a) The IPCE (orange circles) and absorptance spectra (solid black line) for a 20 nm
thick α-Fe2O3 film (b) The trend in APCE at several wavelengths for increasing α-Fe2O3 film
thickness. The lines drawn are only to guide the viewer’s eye and do not represent a fit. (c) The
APCE plotted as a function of the incident photon energy for a 20 nm electrode.

The trends in the APCE with thickness can be understood in terms of drift (charge
carriers moving in response to an electric field) competing with recombination (which is
reflected in the lifetime, τ) described by the hole drift length, L p drift :

29

36

L p drift   E
2

-1 -1 37

where µ is the hole mobility (0.2 cm V

s )

(2-8)

and E is the electrostatic field. The field is a

result of the depletion region which develops from the equilibration of the semiconductor Fermi
38

level (EF) with the solution potential (ES).

The maximum field (Emax) which exists at the

electrode / electrolyte interface is given by
Emax 

qN D

 0

W

(2-9)

where ND is the dopant density, W is width of the depletion layer, κ is the dielectric constant of
the semiconductor (32 for hematite) and ε0 is the permittivity of a vacuum (8.854 × 10
−1 38-40

m ).

−12

−1

CV

The depletion width depends on the voltage drop across the semiconductor depletion

region, Vsc, according to:

38

W

2 0Vsc
qN D

(2-10)

The dopant density and voltage drop were determined through differential capacitance, C,
vs. applied potential measurements. We note that Mott-Schottky analysis is only valid for planer
surfaces (or when the surface roughness is either much larger or much smaller than the depletion
width) not dominated by surface state capacitance (which would result in a frequency
dependence); the films measured herein fulfill these criteria. The capacitances were determined
by fitting the remarkably ideal impedance data (displayed in Figure 2-7a) to a simple Rs(C-Rp)
equivalent circuit over a frequency range spanning five orders of magnitude. The resultant
capacitance values, shown in Figure 2-7b, were fit to the Mott-Schottky equation:

30

2


2
kT 
 AS 
 C   q N  E  EFB  q 



0 D

(2-11)

where As is the surface area of the electrode. The roughness factor of FTO has been estimated to
be 1.5 when correcting the Mott Schottky calculations for the real surface area.
produced a dopant density of 7.14 × 10
(EFB) of -0.40 vs. SCE.

43

18

cm

-3

41,42

The slope

and the intercept produced a flat-band potential

Since only high purity ferrocene and ozone were introduced to the

ALD chamber, we speculate that the dopants must be a residual carbon impurity or oxygen
vacancies in the lattice. Accounting for the dependence of the flat band potential with dopant
density, our experimental determination of EFB is in good agreement with literature
2,44,45

values.

If the entire potential drops across the semiconductor (such that Vsc = Vbi = EFB –

Es = 0.59 V), the depletion width would be ~17 nm thick.

31

Figure 2-7. (a) Impedance spectra of a hematite electrode prepared with 1,200 ALD cycles for
applied potentials of -0.034 (red circles), -0.059 (orange downward pointing triangles), -0.084
(yellow squares), -0.109 (green diamonds), -0.134 (teal upward pointing triangles) V vs SCE.
Capacitances were determined by fitting data to a simple Rs(C-Rp) circuit shown in inset. (b)
Mott-Schottky plot with the errors of those fits represented in the error bars shown (generally
smaller than the symbols).

The thickness of the optimal electrodes is ~20 nm, which is approximately equal to the
length of the maximum depletion region.

The increasing electric field that develops for

electrodes from 7 to 20 nm would result in an increased L p drift , which can account for the
increase in APCE for these electrodes. It has also been hypothesized that a “dead layer” exists as
46

a result of the substrate-hematite junction which renders very thin films inactive.

This has

been at least partially overcome more recently by adding various metal oxide under-layers and
46-48

through doping of the hematite.

However, it is still unclear how these modifications fix this

dead layer, thus we continue our discussion of the results in terms of an increasing L p drift .

32

Electrodes thicker than ~20 nm are larger than the depletion region and no additional
photocurrent is produced. These data are consistent with the explanation that only a fraction of
the holes that are produced within the depletion region are collected, and holes generated in the
bulk do not contribute to the photocurrent.
The L p drift can be estimated using literature values for the minority carrier lifetimes.
The lifetime of holes in both thin film and single crystal hematite was found to be 4.2 ps by
ultrafast transient absorption and reflectance measurements.

3

The Emax determined from

equations 2-9 and 2-10, with Vsc = 0.59 V, produces a hole drift length of 5.8 nm. The electric
field varies as a function of distance in the space charge region.

38

Since the charge carrier

concentrations are spatially dependent, the lifetime should also be a function of position in the
space charge region.

49

Thus, the calculated drift length of 5.8 nm should be considered a

maximum value. Nevertheless, a drift length of 5.8 nm is in reasonable agreement with our
results, where we found electrodes with an optimal film thickness of ~20 nm to have APCEs of
10 – 30 %.
In the analysis presented above, we assumed that the photoelectrochemical properties are
constant for all of the material layers deposited, and only the thickness of the hematite was
changing. In order to test this hypothesis, the APCE response due only to the last 20 nm
deposited (corresponding roughly to the depletion region) was calculated for all electrodes. The
calculation consisted of multiplying the APCEs by the percentage of incident photons absorbed
only in the 20 nm adjacent to the solution. The percentages of photons absorbed only in the
depletion region were calculated by subtracting the percentage of photons absorbed in the rest of
the film from unity as if it was an optical filter. Aside from the very thin electrodes, the APCEs

33

of the depletion region were found to be essentially constant for all electrodes (Figure 2-8). This
is consistent with our assumption that material properties, when thicker than the depletion region,
are essentially invariant with thickness, and that only charge carriers absorbed near the electrolyte
interface contribute to the photocurrent.

Figure 2-8. APCEs for several wavelengths normalized for the light absorbed only in the ~20 nm
adjacent to the electrode / electrolyte interface as a function of thickness. Lines are drawn only to
guide the viewer’s eye and do not represent a fit.

In addition to the APCE dependence on electrode thickness, there is a striking
dependence of APCE with wavelength of incident light (Figure 2-6c).

The wavelength

dependence is inversely related to the absorption profile; a slightly larger fraction of higher
energy photons will be absorbed farther away from the electrolyte interface than lower energy
photons since illumination is through the back substrate. This trend continues even as electrodes
are made thicker and fewer low energy photons are absorbed in the depletion region. APCEs
were also measured on sandwich cells from both the substrate-electrode (SE) and the electrolyteelectrode (EE) interfaces. After correcting for the absorption of the counter electrode and

34

electrolyte, the APCEs show the same wavelength dependence (Figure 2-9). Thus we discount
differences in light penetration depths as a significant contributing factor. The trend in APCE
with thickness for all wavelengths appear to be the same – i.e. all wavelengths have a maximum
value for electrodes 20 nm thick – which indicates they are all limited by drift. We therefore
suggest that the wavelength (energy) dependence of the APCE is a consequence of hot charge
carriers (generated with higher energy photons) having higher mobilities and/or longer lifetimes
(slower recombination) which produce longer drift lengths. The relaxation of hot carriers is
generally on the femtosecond time regime, which could increase lifetimes by a similar amount,
however there is no evidence the lifetime would increase linearly with photon energy.

3

Lacking

direct evidence, we speculate that higher energy holes have linearly increasingly higher
mobilities enabling a higher fraction to reach the interface and be collected. A more detailed
understanding of this phenomenon is currently under investigation.

Figure 2-9. APCEs calculated when shining light from the substrate-electrode interface (SE)
and the electrolyte-electrode interface (EE). EE APCE values were corrected for light absorption
of the FTO counter electrode and electrolyte.

35

The open circuit voltage, Voc, (seen in figure 2-5c) was measured as the difference
between the potential at which no current density flows relative to Es. The Voc also increases
sharply with thickness until 20 nm then remains essentially constant. The increasing Voc with
thickness can be explained by the diode equation:

VOC 

 kT

J
ln SC
q
J0

(2-12)

where γ is the diode quality factor, k is the Boltzmann constant, T is temperature, q is the charge
of an electron, and J0 is the recombination current density.

49

The increasing photocurrent

accounts for only a minor fraction of the ~250 mV increase in Voc from samples ~20 nm thick,
compared to samples ~7 nm thick. Therefore, J0, which consists of the sum of all recombination
including surface state recombination, Jss, bulk recombination, Jbr, depletion region
recombination, Jdr, and electron transfer from the semiconductor to the electrolyte, Jet, shown in
Figure 2-10, must decrease over the first ~20 nm.

36

Figure 2-10. Qualitative energy diagram showing the favorable pathways of minority carriers
being generated (G) and being collected as current (Jphoto) as well as the unfavorable
recombination pathways including bulk recombination (Jbr) depletion region recombination (Jdr)
surface state recombination (Jss) and electron transfer (Jet).

The development of the space charge region can also explain the large initial increase in
VOC for electrodes from 7 to 20 nm thick. As the electrode thickness increases, the ratio of
surface area to bulk decreases, and a field develops to drive electrons from the surface and holes
towards it. We therefore attribute the trend of decreasing J0 (increasing VOC) with thickness to
decreasing recombination at the surface: Jss and/or Jet. As discussed in chapter 6, Jss is shown to
be insignificant when a fast redox shuttle is used to collect holes. On the other hand, an
increasing thickness for films smaller than the depletion region would increase the electric field
within the semiconductor for an a given applied potential since very thin films would be

37

completely depleted.

This would cause lower amounts of band bending, and a higher

concentration of electrons at the surface and a higher Jet. The constant voltage of electrodes ≥ 20
nm is thus attributed to a fully developed electric field and a constant Jet.
The measurements reported herein were performed very carefully; however a note should
be made regarding possible error involved with the analysis presented above due to uncertainties
5

in literature constants. Absorption coefficients reported in the literature range from 1.3 × 10 to
5

2.5 × 10 cm

-1

at 410 nm.

1,29,40,50

5

Thus our use of 2.2 × 10 cm

-1

at 410 nm, albeit consistent

with our SEM measurements, may result in a slight error in film thickness and growth rate.

29

A

bigger issue is the uncertainty in the dielectric constant. Literature values for the dielectric
39,40,51-53

constant of hematite range from 12-120.

believe to be most relevant) reported value of 32.

40

Again, we used a moderate (and what we
A larger dielectric constant would produce a

smaller dopant density, larger depletion region, smaller electric field, and shortened drift length.
Our results taken together, however, are very consistent with the model of an optimal electrode
thickness that is slightly smaller than the maximum depletion width, and a drift length that is
shorter than the depletion width.
2.5 Conclusions
Use of atomic layer deposition allowed for precise structure-function relationships to be
made for thin film hematite electrodes in regenerative photoelectrochemical cells. We found that
there is an optimal film thickness of 20 ± 2.2 nm which balances the development of an
electrostatic field with having a short collection distance. Use of films thicker than ~20 nm may
have comparable photovoltaic parameters (JSC and VOC), but come at a cost of wasted material

38

and absorbed photons.

The maximum APCE values are unfortunately still rather low, in

agreement with the calculated drift length being shorter than the film thickness. Low APCEs
result from recombination being kinetically competitive with charge extraction.

Our data,

especially the wavelength dependence of the APCEs, is consistent with bulk recombination being
the dominant recombination mechanism which limits the overall performance of hematite
electrodes. The thinnest films have an unexpectedly low APCE and VOC which is attributed to
the fact that they are thinner than the depletion region, which results in under-devolved electric
field, causing a large electron recombination from the conduction band to solution. Bulk region
recombination follows Shockley-Reed-Hall formalism where the hole lifetime is inversely
proportional to the trap state density.

49

Thus, either increasing the hole mobility or minimization

of mid-bandgap traps is necessary in order to improve performance. Use of intentional dopants
such as titanium and silicon have been shown to result in improved IPCEs, attributed to
increasing electrical conductivity, passivation of grain boundaries, or morphological
modification.

8,54

Alternatively, optimization of the ALD process (temperature, precursors, etc.)

could result in more pristine films that exhibit longer lifetimes. Achieving the near unity APCEs
required for efficient energy conversion is a very challenging task; however the potential benefits
are great enough to warrant further study. We believe the systematic approach employed here,
and by others, can eventually achieve this goal.

8

2.6 Acknowledgements
T.W.H. would like to thank MSU for providing a generous start-up package in support of
this work. A.B.F.M. acknowledges the support of the Energy Frontier Research center (DESC0001059) at the Argonne-Northwestern Solar Energy Research (ANSER) Center.

39

APPENDIX

40

Figure A2-1 IPCE spectra for hematite electrodes prepared by 100 (red circles), 200 (orange
downward facing triangles), 300 (yellow squares), 800 (teal upward facing triangles), and 1200
ALD cycles (green diamonds).

Figure A2-2. (a) APCE spectra for hematite electrodes prepared by 100 (red circles), 200 (orange
downward facing triangles), 300 (yellow squares), 800 (teal upward facing triangles), and 1200
ALD cycles (green diamonds).

41

REFERENCES

42

REFERENCES

1. Kennedy, J. H.; Frese, K. W. J. Electrochem. Soc. 1978, 125, 6.
2. Dareedwards, M. P.; Goodenough, J. B.; Hamnett, A.; Trevellick, P. R. J. Chem. Soc.,
Faraday Trans. 1 F, 1983, 79, 2027-2041.
3. Joly, A. G.; Williams, J. R.; Chambers, S. A.; Xiong, G.; Hess, W. P.; Laman, D. M. J. Appl.
Phys. 2006, 99, 6.
4. Gartner, W. W. Phys. Rev. 1959, 116, 84-87.
5. Wilson, R. H. J. Appl. Phys. 1977, 48, 4292-4297.
6. Cesar, I.; Sivula, K.; Kay, A.; Zboril, R.; Graetzel, M. J. Phys. Chem. C, 2009, 113, 772-782.
7. Le Formal, F.; Gratzel, M.; Sivula, K. Adv. Funct. Mater. 2010, 20, 1099-1107.
8. Glasscock, J. A.; Barnes, P. R. F.; Plumb, I. C.; Savvides, N. J. Phys. Chem. C, 2007, 111,
16477-16488.
9. Spray, R. L.; Choi, K. S. Chem. Mat. 2009, 21, 3701-3709.
10. Fu, Y. Y.; Wang, R. M.; Xu, J.; Chen, J.; Yan, Y.; Narlikar, A.; Zhang, H. Chem. Phys. Lett.
2003, 379, 373-379.
11. Prakasam, H. E.; Varghese, O. K.; Paulose, M.; Mor, G. K.; Grimes, C. A. Nanotechnology,
2006, 17, 4285-4291.
12. Sartoretti, C. J.; Alexander, B. D.; Solarska, R.; Rutkowska, W. A.; Augustynski, J.; Cerny,
R. J. Phys. Chem. B, 2005, 109, 13685-13692.
13. Hahn, N. T.; Ye, H. C.; Flaherty, D. W.; Bard, A. J.; Mullins, C. B. ACS Nano, 2010, 4,
1977-1986.
14. Ritala, M.; Leskela, M. In Handbook of Thin Film Materials; Nalwa, H. S., Ed.; Academic
Press: San Diego, 2001; Vol. 1, p 103.
15. Scheffe, J. R.; Frances, A.; King, D. M.; Liang, X. H.; Branch, B. A.; Cavanagh, A. S.;
George, S. M.; Weimer, A. W. Thin Solid Films, 2009, 517, 1874-1879.
16. Rooth, M.; Johansson, A.; Kukli, K.; Aarik, J.; Boman, M.; Harsta, A. Chem. Vap.
Deposition, 2008, 14, 67-70.
17. Lie, M.; Fjellvag, H.; Kjekshus, A. Thin Solid Films, 2005, 488, 74-81.
43

18. Bachmann, J.; Jing, J.; Knez, M.; Barth, S.; Shen, H.; Mathur, S.; Gosele, U.; Nielsch, K. J.
Am. Chem. Soc. 2007, 129, 9554-9555.
19. Martinson, A. B. F.; DeVries, M. J.; Libera, J. A.; Christensen, S. T.; Hupp, J. T.; Pellin, M.
J.; Elam, J. W. J. Phys. Chem. C 2011, 115, 4333-4339.
20. Hamann, T. W.; Martinson, A. B. F.; Elam, J. W.; Pellin, M. J.; Hupp, J. T. J. Phys. Chem.
C, 2008, 112, 10303-10307.
21. Hamann, T. W.; Martinson, A. B. F.; Elam, J. W.; Pellin, M. J.; Hupp, J. T. Adv. Mater.
2008, 20, 1560–1564.
22. Martinson, A. B. F.; Elam, J. W.; Liu, J.; Pellin, M. J.; Marks, T. J.; Hupp, J. T. Nano Lett.
2008, 8, 2862-2866.
23. Beermann, N.; Vayssieres, L.; Lindquist, S. E.; Hagfeldt, A. J. Electrochem. Soc. 2000, 147,
2456-2461.
24. Souza, F. L.; Lopes, K. P.; Longo, E.; Leite, E. R. Phys. Chem. Chem. Phys. 2009, 11, 12151219.
25. Duret, A.; Gratzel, M. J. Phys. Chem. B 2005, 109, 7.
26. deFaria, D. L. A.; Silva, S. V.; deOliveira, M. T. J. Raman Spectrosc. 1997, 28, 873-878.
27. Bersani, D.; Lottici, P. P.; Montenero, A. J. Raman Spectrosc. 1999, 30, 355-360.
28. Leon, C. P.; Kador, L.; Zhang, M.; Muller, A. H. E. J. Raman Spectrosc. 2004, 35, 165-169.
29. Marusak, L. A.; Messier, R.; White, W. B. J. Phys. Chem. Solids, 1980, 41, 981-984.
30. Kung, H. H.; Jarrett, H. S.; Sleight, A. W.; Ferretti, A. J. Appl. Phys. 1977, 48, 2463-2469.
31. Gilbert, B.; Frandsen, C.; Maxey, E. R.; Sherman, D. M. Phys. Rev. B, 2009, 79.
32. Mohanty, S.; Ghose, J. J. Phys. Chem. Solids, 1992, 53, 81-91.
33. Tahir, A. A.; Wijayantha, K. G. U.; Saremi-Yarahmadi, S.; Mazhar, M.; McKee, V. Chem.
Mater. 2009, 21, 3763-3772.
34. Finklea, H. O. Semiconductor Electrodes; Elsevier: Amsterdam, 1988.
35. Anderman, M.; Kennedy, J. H. J. Electrochem. Soc. 1984, 131, 1565-1570.
36. Mihailetchi, V. D.; Wildeman, J.; Blom, P. W. M. Phys. Rev. Lett. 2005, 94, 4.

44

37. Bosman, A. J.; Vandaal, H. J. Adv. Phys. 1970, 19, 1-&.
38. Tan, M. X.; Laibinis, P. E.; Nguyen, S. T.; Kesselman, J. M.; Stanton, C. E.; Lewis, N. S.
Prog. Inorg. Chem. 1994, 41, 21-144.
39. CRC Handbook of Chemistry and Physics; 81 ed.; Lide, D. R., Ed.; CRC Press, 2001.
40. Glasscock, J. A.; Barnes, P. R. F.; Plumb, I. C.; Bendavid, A.; Martin, P. J. Thin Solid Films,
2008, 516, 1716-1724.
41. Cameron, P. J.; Peter, L. M. J. Phys. Chem. B 2003, 107, 14394-14400.
42. Fattori, A.; Peter, L. M.; McCall, K. L.; Robertson, N.; Marken, F. J. Solid State
Electrochem. 2010, 14, 1929-1936.
43. ****, the errors of these calculated values propagated from fitting the capacitance data were
insignificant.
44. Lindgren, T.; Wang, H. L.; Beermann, N.; Vayssieres, L.; Hagfeldt, A.; Lindquist, S. E. Sol.
Energy Mater. Sol. Cells, 2002, 71, 231-243.
45. Kennedy, J. H.; Frese, K. W. J. Electrochem. Soc. 1978, 125, 723-726.
46. Hisatomi, T.; Brillet, J.; Cornuz, M.; Le Formal, F.; Tetreault, N.; Sivula, K.; Gratzel, M.
Faraday Discuss. 2012.
47. Zandi, O.; Klahr, B. M.; Hamann, T. W. Energy Environ. Sci. 2013, 6, 634-642.
48. Hisatomi, T.; Le Formal, F.; Cornuz, M.; Brillet, J.; Tetreault, N.; Sivula, K.; Gratzel, M.
Energy Environ. Sci. 2011.
49. Nelson, J. The Physics of Solar Cells; Imperial College Press: London, 2003.
50. Souza, F. L.; Lopes, K. P.; Longo, E.; Leite, E. R. Phys. Chem. Chem. Phys. 2009, 11, 12151219.
51. Papaioannou, J. C.; Patermarakis, G. S.; Karayianni, H. S. J. Phys. Chem. Solids 2005, 66,
839-844.
52. Kulkarni, S. S.; Lokhande, C. D. Mater. Chem. Phys. 2003, 82, 151-156.
53. Quinn, R. K.; Nasby, R. D.; Baughman, R. J. Mater. Res. Bull. 1976, 11, 1011-1017.
54. Kay, A.; Cesar, I.; Gratzel, M. J. Am. Chem. Soc. 2006, 128, 15714-15721.

45

Chapter 3:
Current and Voltage Limiting Processes in Thin
Film Hematite Electrodes

Adapted with permission from:
Current and Voltage Limiting Processes in Thin Film Hematite Electrodes, Benjamin M. Klahr
and Thomas W. Hamann, Jounral of Physical Chemistry C, 2011, 115, (16), 8393-8399.
Copyright 2011 American Chemical Society.

46

3.1 Abstract
Atomic layer deposition was used to grow conformal thin films of hematite on
transparent

conductive

oxide

substrates

and

used

as

electrodes

in

regenerative

photoelectrochemical cells (PECs). Systematic variations of the contacting electrolyte and redox
shuttles were performed to determine the rate limiting processes of hematite PECs. Steady state
current density vs. applied potential measurements were employed to show that when the fast,
one electron redox shuttles are used, hole collection is not the rate limiting step; however, hole
collection is the rate limiting step when hematite is employed for water oxidation. Electron
transfer was found to be significant which limits the open circuit photovoltage. We further
demonstrate that with variations of the pH and redox potentials of the contacting electrolyte,
increases in both the photocurrent density and photovoltage can be achieved, which is attributed
to increasing the built-in voltage.

47

3.2 Introduction
In the previous chapter, we used atomic layer deposition (ALD) to create a series of
conformal hematite thin films, with thickness from ~5 – 60 nm, to determine structure-function
1

relationships in regenerative PECs.

ALD is a unique technique, which is capable of producing

very thin films which are conformal and pinhole free, such as those employed herein.

2,3

We

found that a film thickness of 20 ± 2.2 nm maximizes the internal quantum yield. This optimal
thickness balances the development of an electrostatic field with having a short collection
distance. Even with ~20 nm thick films, however, the maximum APCE values are unfortunately
significantly less than unity, which results in suboptimal photocurrents.

In addition, the

photovoltage was less than the maximal value. The photocurrent is a result of the generation of
charge carriers by light absorption, G, separation of the charge carriers by the electrostatic field,
and collection of the minority carriers (holes for a n-type semiconductor) by the contacting
electrolyte, Jhc, shown in Figure 3-1. The photovoltage is limited by the primary recombination
processes, including bulk recombination, Jbr, depletion region recombination, Jdr, surface state
recombination, Jss, and electron transfer, Jet, shown in Figure 3-1 as red dashed arrows. While
we were preparing this manuscript, Dotan et. al. reported using H2O2 as a fast hole collector to
deconvolute the performance of the hematite films from the rate limiting process of water
oxidation.

4

In this work we employ a similar strategy by using fast, one-electron outer-sphere

redox shuttles. In addition, we perform systematic variations of the contacting electrolyte to
elucidate the dominant recombination mechanism which limits the photovoltage of hematite
electrodes in PECs. The resultant framework allows for interpretation of the pH dependence of
48

the photovoltaic performance. We also investigate a series of redox shuttles to demonstrate a
new strategy to overcome the processes limiting the performance of hematite PECs. If hematite,
or a similar material, will ever be used effectively for solar energy conversion, it is crucial to
identify and overcome these bottlenecks. The work presented herein represents an important
step in achieving that goal.

Figure 3-1. Qualitative energy diagram showing the favorable pathways (green solid arrows) of
charge being generated by light absorption, G, driven to the electrolyte interface by the electric
+

field in the depletion region (drift) and collected, Jhc, by a redox couple, D /D, in solution. Also
shown are the deleterious recombination pathways (red dashed arrows) including bulk
recombination, Jbr, depletion region recombination, Jdr, surface state recombination, Jss, and
electron transfer, Jet.

49

3.3 Experimental
The synthesis and characterization of hematite films deposited by atomic layer deposition
1

(ALD) was described in detail in Chapter 2.

The thin films were all deposited on fluorine-

doped tin oxide (FTO) substrates. Multiple films approximately 30 nm thick were prepared and
employed in the measurements described herein, since this thickness has previously produced the
highest combination of photocurrent density and photovoltage, and thus overall efficiency.
While the magnitude of results between various electrodes exhibited some variance, the trends
were all consistent. For clarity of presentation, the results reported herein are for a single 33 nm
thick electrode. Hematite electrodes were masked with a 60 μm Surlyn film (Solaronix) before
being clamped to a custom made glass electrochemical cell to prevent scratching of the thin
films. K3Fe(CN)6 (Sigma-Aldrich) and K4Fe(CN)6·3H2O (Mallinckrodt Baker) were purified
by recrystallization. K4IrCl6 and K3IrCl6 were purchased from Sigma-Aldrich and used as
received. K5Mo(CN)8·2H2O and K5W(CN)8·2H2O were prepared by literature procedures.

5,6

Unless stated otherwise in the text, 200 mM of K4Fe(CN)6·3H2O and 20 mM of K3Fe(CN)6
were used. For alternative redox shuttles, 100 mM of the reduced form and 10 mM of the
oxidized form were used. Bulk electrolysis was employed to prepare the oxidized form of
molybdenum and tungsten species until an amount of charge had been passed to produce a
desired concentration. All electrolytes contained 200 mM KCl. Solutions were all maintained at
pH 12 (KOH) unless otherwise noted. Otherwise, phosphate buffers were used to maintain
solutions at approximately pH 7 and pH 8, and borate buffers were used to maintain solutions at
approximately pH 9 and pH 10. Electrochemical measurements were referenced to a homemade
Ag/AgCl electrode which was frequently compared to a Radiometer Analytical saturated calomel
50

electrode (SCE).

High surface area platinum mesh was used as a counter electrode.

Photoelectrochemical measurements were made with a Gamry Instruments Reference 600
potentiostat. The light source was a Xe arc lamp interfaced with a Horiba Jobin Yvon MicroHR
monochromator. An AM 1.5 solar filter and neutral density filters were used to simulate sunlight
2

at 100 mW/cm . All photoelectrochemical measurements were performed by shining light from
the substrate-electrode (SE) interface.
3.4 Results and Discussion
3.4.1 [Donor] Dependence
Steady state current density, J, vs. applied potential, E, curves were measured for a 33 nm
thin film hematite electrode in contact with an aqueous pH 12 [Fe(CN)6]
typical J-E curve measured under 100 mW cm

-2

3-/4-

electrolyte. A

illumination can be seen in Figure 3-2a. Since

the photocurrent density is not constant with potential, for purposes of comparison we define the
photocurrent density, Jphoto, at the formal potential, E°, of the contacting redox couple (E° =
0.250 V vs. SCE for [Fe(CN)6]

3-/4-

). Jphoto was also determined at a constant potential to

isolate the performance of the hematite electrode from changes in the solution potential which
varies with donor/acceptor ratio in accord with the Nernst equation. The open circuit voltage,
VOC, is defined as the difference between the applied potential which results in zero measured
current and E°. The concentrations of the reduced form (electron donor) of the [Fe(CN)6]

3-/4-

redox couple was varied from 1 mM to 400 mM while maintaining a constant concentration (20
mM) of the oxidized form (electron acceptor). Dark J-E curves were found to be invariant with
respect to variations of donor concentration as expected. Figure 3-2b shows the Jphoto extracted
51

from J-E curves taken in response to varying light intensities a function of donor concentration.
It can be seen that Jphoto initially increases with the concentration of donor, but quickly reaches
a saturation current at relatively low concentrations (~20 mM). Jphoto remained constant for
donor concentrations 20 – 400 mM, shown in Figure 3-2c. Figure 3-2d shows results of current
density transients, also measured at the formal potential, in response to 100 mW cm

-2

illumination. For very low concentrations of donor (< 20 mM), there was an initial current spike
which quickly decayed into a lower steady state value. This behavior is consistent with Jphoto
7,8

being limited by diffusion of the redox couple to the electrode surface.

For donor

concentrations higher than 20 mM, the instantaneous current density remained constant at a
value equal to the value of photocurrent density extracted from the J-E curves. As long as
sufficient redox shuttle is present to support diffusion to the photoelectrode surface, the current
density is independent of donor concentration.

In addition, the measured VOC’s were

approximately constant with donor concentration as expected (see Appendix). The rate of hole
collection at the hematite electrode surface should depend on the concentration of donor (hole
collecting species). Since no increase in photovoltaic performance is observed with increasing
donor concentration (i.e. increased rate of hole collection), we conclude that the interfacial hole
collection kinetics are not the rate limiting process controlling the photovoltaic performance of
hematite electrodes in contact with the [Fe(CN)6]

3-/4-

electrolyte. This result is in agreement

with conclusions reported by Dotan et. al. who employed H2O2 as a hole collector.

4

J-E curves were also measured for each concentration of donor in response to different
light intensities. Figure 3-2b shows that when the Jphoto is not limited by diffusion (at high
52

donor concentrations), and Jphoto increases proportionally with light intensity. A linear increase
in Jphoto with incident light intensity is observed for intensities from 1 – 400 mW cm

-2

(see

Appendix). The small deviations from linear behavior were determined to be due to wavelength
variations of the neutral density filters employed to adjust the light intensity. Very linear
behavior is seen with monochromatic light (see Appendix), however in this case, we are
restricted to lower light intensities.

The independence of Jphoto with donor concentration

coupled with the linear increase of Jphoto with light intensity are consistent with Jphoto only
being limited by the charge collection length which limits the number of holes that reach the
1

electrode interface, in agreement with our previous conclusions.

53

Figure 3-2: (a) A J-E curve for a 33 nm hematite film under 100 mW cm
contact with a pH 12 [Fe(CN)6]

3-/4-

-2

illumination in

electrolyte containing 200 mM reduced and 20 oxidized

species. (b - c) Plots of the photocurrent density measured as a function of donor concentration
4-

([Fe(CN)6] ) for light intensities of 50 mW cm
triangles) and 160 mW cm

-2

-2

(yellow squares), 100 mW cm

-2

(orange

(red circles). (d) Current transients measured in response to 100

-2

mW cm illumination for donor concentrations from 2 – 400 mM.

54

3.4.2[Acceptor] Dependence
J-E curves were also measured in contact with an aqueous pH 12 [Fe(CN)6]

3-/4-

electrolyte with a constant concentration of the reduced form (donor) of the redox couple (200
mM) while varying the concentration of the oxidized form (acceptor) from 20 – 700 mM. Figure
3-3a shows a semi-logarithmic plot of the dark J-E curves for the varying concentrations. The
dark current density, at a given potential, is proportional to the concentration of acceptor,
indicating first order behavior. The only recombination mechanism that depends on acceptor
concentration is back electron transfer, Jet (see Figure 3-1). Analysis of the slope of the dark ln(J) vs. E curves, which is equal to q/γkT, indicates that γ = 2, and thus recombination is not first
order with respect to surface electron concentration.

9,10

Under one sun illumination, the photocurrent density was found to be essentially
independent of acceptor concentration. This behavior is expected since we demonstrated that the
photocurrent density is limited by the rate of holes reaching the electrode interface to be
collected, and acceptors in solution should therefore not contribute to the photocurrent
generation processes (see Figure 3-1). A semi-logarithmic plot of VOC as a function of acceptor
concentration can be seen in figure 3-3b. The VOC was found to decrease at a rate of 147 mV
per 10-fold increase in [A] at pH 12. According to the diode equation

VOC 

J 
ln  SC 
q
 J0 

 kT

9

(3-1)

where k is the Boltzmann constant, T is temperature, JSC is the short circuit photocurrent density,
and J0 is recombination exchange current density, the VOC should vary linearly with the
55

logarithm of J0. The 147 mV per decade slope indicates that γ ~2, in good agreement with the
dark J-E measurements. Also shown is a plot of the applied potential required to produce a dark
-2

current density of 500 µA cm . The VOC changes with acceptor concentration in identical
fashion (~147 mV per decade) as the dark current density predicts, demonstrating that
superposition holds. In addition, since J0 is the sum of all recombination current densities

9,10

J 0  Jbr  J ss  J dr  J et

(3-2)

the dependence of VOC on [A] indicates that Jet is the dominant pathway such that J0 ≈ Jet.
Thus, interfacial electron transfer is the recombination mechanism which limits the photovoltage.
4

We note that this contradicts the conclusions of recent work using H2O2 as a hole scavenger. In
that work, however, the effects of recombination were inferred from measurements of the
photocurrent density, which is not always affected by recombination.
Back electron transfer from conduction band electrons, including surface-state capture as
an intermediate parallel pathway, to acceptors in solution can be described by the rate law:

J et  qket  A ns  ksol  A N ss f ss

11

(3-3)

where A is the electron acceptor concentration in solution, ns is the concentration of electrons in
the conduction band, ket is the electron-transfer rate constant from conduction band electrons,
Nss is the density of surface states, fss is the fraction of surface states that are filled with
electrons, and ksol is the electron-transfer rate constant from surface traps. This rate law would
thus produce first order behavior with respect to [A], but not ns, which is consistent with our

56

observation. We therefore tentatively conclude that surface-state mediated electron transfer is a
contributing pathway of recombination, in addition to direct electron transfer from conduction
band electrons. Work is ongoing in our laboratory to address the mechanistic details of this
reaction.

Figure 3-3: a) Semi-logarithmic plot of dark J-E curves of PECs as a function of acceptor
3-

concentration ([Fe(CN)6] ), from 20 to 700 mM. b) Plot of VOC (red circles) and the applied
potential to achieve a given dark current density (orange triangles) as a function of acceptor
3-

concentration ([Fe(CN)6] ).

3.4.3 pH Dependence
It has previously been shown that increasing the pH of the electrolyte improves the
photovoltaic performance of hematite electrodes.

12,13

To the best of our knowledge, however,

the cause of the improvement has not been established. In order to address this point, J-E curves
were measured in aqueous electrolytes with the pH adjusted from 7 – 12, both with and without
the [Fe(CN)6]

4-/3-

redox shuttle. Dark J-E curves can be seen in figure 3-4a without the
57

[Fe(CN)]6

4-/3-

redox shuttle. Anodic current is seen at very positive potentials attributed to the

oxidation of water. The current onset potential shifts negatively by approximately ~63 mV per
pH (sampled at 500 µA). The water oxidation potential should increase by -59 mV per pH in
accord with the Nernst equation. Therefore, assuming the overpotential required to oxidize
water with valence band holes is pH independent, the shift in the current onset potential is due to
a concurrent -59 mV per pH shift in the valence band potential, EVB/q. This is in excellent
agreement with literature, where the EFB of hematite has been shown to increase at the Nernstian
rate of ~59 mV per pH for sintered polycrystalline,
curves measured under 100 mW cm

-2

14,15

and single crystal hematite.

16,17

J-E

illumination with no redox couple can be seen in figure 3-

4b. Even at pH 12, significant photocurrent only exists at potentials more positive than 0.4 V vs
SCE, in agreement with literature reports. Hematite is often employed as a water oxidation
photocatalyst, and a majority of recent papers studying the photoelectrochemistry of hematite
focused on water oxidation.

18-21

Water oxidation at hematite electrodes has been calculated to
22

be a very slow process, however.

Since the photocurrent density is much lower at a given

potential and pH compared with the results above when [Fe(CN)6]

4-/3-

is employed as a redox

couple, water oxidation appears to be the rate limiting step of the photocurrent density
production.

The measurements performed herein with the [Fe(CN)6]

4-/3-

redox couple,

however, are all performed at potentials negative of 0.3 V vs SCE, and thus are not significantly
affected by the side reaction of water oxidation.

58

-2

Figure 3-4c shows J-E curves measured under 100 mW cm illumination for solution pH
values from 7 – 12 in the presence of the [Fe(CN)6]

4-/3-

redox shuttle. Clearly, the photocurrent

density at more negative potentials increases with pH. Plots of the Jphoto and VOC extracted
from J-E curves as a function of pH can be seen in figure 3-4d. Recently, we have suggested
that the photocurrent of thin film hematite electrodes is controlled by the electric field that
develops in the depletion region.

1

The width of the depletion region in turn depends on the

voltage drop across the semiconductor, VSC, (or the difference between the flat band potential,
23

EFB, and the solution potential, ES where VSC = EFB - ES).

Thus, the increasing Jphoto with

pH can be attributed to the larger depletion region that develops as the flat band potential shifts
negatively. The VOC increases by ~42 mV per pH, which is somewhat less than expected due to
the ~59 mV per pH shift in the conduction band edge. This can be explained by the fact that
although increasing pH raises the conduction band, it also increases the driving force, G , for
back electron transfer, which we showed above to be the voltage limiting recombination
24,25

mechanism (see Jet in figure 3-1).

The increase in G should result in an increase in ket,

as well as ksol, in equation (3-3), assuming normal region behavior.

24,25

Unfortunately, since

first order behavior in ns was not observed, and we were not able to quantify the product Nssfss in
equation (3-3), we cannot calculate rate constants to verify this. Nevertheless, an increase in
recombination rate constants due to an increasing G with pH can readily account for a ~12
mV decrease in VOC per pH unit.

59

Figure 3-4: Plots of J-E curves in aqueous electrolytes, where current is attributed to H2O
oxidation, in pH 6.8 (red solid line), pH 8 (orange dotted line), pH 9 (yellow short dashed line),
pH 10 (green dashed double dotted line), pH 11 (light blue long dashed line), pH 12 (dark blue
dashed dotted line) in the (a) dark and (b) under 100 mW cm
aqueous electrolytes containing the [Fe(CN)6]

3-/4-

-2

illumination. c) J-E curves in

redox couple at pH 6.8 (red solid line), pH 8

(orange dotted line), pH 9 (yellow short dashed line), pH 10 (green dashed double dotted line),
pH 11 (light blue long dashed line), pH 12 (dark blue dashed dotted line) d ) Plots of Jphoto and
VOC, extracted from the J-E plots in (c), as a function of pH.

60

3.4.4 Redox Potential Dependence
The built-in voltage is the difference between the Fermi level in the semiconductor and
the solution potential. The pH increases the voltage drop by increasing the Fermi level (flat band
potential) of the semiconductor electrode. Decreasing the solution potential should likewise
increase the voltage drop, and thus the photocurrent and photovoltage. To test this hypothesis, a
homologous series of four different redox shuttles with formal potentials spanning 500 mV were
prepared. The redox couples and their formal potentials are listed in Table 3-1. Each solution
contained 100 mM of the electron donor species and 10 mM of the electron acceptor species in a
pH 6.8 buffered 0.05M solution of KH2PO4/K2HPO4. Solutions were kept at a neutral pH to
maximize solubility and stability for all the redox couples investigated. Concentrations were
limited by the least soluble redox couple, [IrCl6]

3-/4-

. Plots of J-E curves measured in the dark,

referenced to their respective E°, can be seen in figure 3-5a. The dark current density decreases
in response to the solution potential being made more positive. Since superposition was shown
to hold, this decrease in dark current density is expected to translate into an increase in VOC.
The light J-E curves for cells employing [Fe(CN)6]

3-/4-

and [IrCl6]

3-/4-

as redox shuttles are

shown in figure 4b. In this configuration, the J-E curves correspond to a relative measure of the
photovoltaic efficiency. The light curves for [Mo(CN)8]

3-/4-

and [W(CN)8]

3-/4-

are not shown

because these couples are not very stable when exposed to light, however preliminary results
were found to be consistent with the trends reported for the other couples.
The Jphoto approximately triples when [IrCl6]

3-/4-

is used compared with [Fe(CN)6]

3-/4-

,

in addition to an increase in VOC of ~330 mV. This value is less than the difference between the

61

490 mV solution potentials of the redox couples. Similar to the pH dependence studies above,
the entire change in VSC caused by a positive shift in E° is not gained in the VOC, most likely due
to an increased driving force for electron transfer from the conduction band of the semiconductor
to the solution. The effects of an increasing Jet (Jsol) can be seen in the dark J-E curves plotted
against a fixed reference in supporting information. This plot shows that as VSC is increased the
dark current at a given potential vs a fixed reference also increases simply because the increased
driving force as no other recombination pathways exist in the dark.

Similarly to the pH

dependence study, as the E° becomes more positive and VSC increases, the photocurrent and
VOC increase. Again, this is because an increasing VSC results in a larger depletion region and a
higher electric field, driving holes to the electrolyte interface to be collected.

Table 3-1. Formal Potentials of Outer-sphere Redox Shuttles Employed
Redox shuttle
Fe(CN)6
Mo(CN)8
W(CN)8
IrCl6

E° / V vs SCE

3-/4-

0.25

3-/4-

0.33

3-/4-

0.60

3-/4-

0.74

62

Figure 3-5. Dark J-E curves for a hematite electrode in contact with the redox couples
[Fe(CN)6]

3-/4-

(red solid line), [W(CN)8]

dashed line) and [IrCl6]

3-/4-

potentials and (b) SCE.

3-/4-

(orange dotted line), [Mo(CN)8]

3-/4-

(yellow

(green dash-dot-dash line) measured vs. (a) their respective solution

(c) Dark J-E where J is plotted on a log scale.

measurements of J-V curves for [Fe(CN)6]

3-/4-

(red) and [IrCl6]
-2

and under AM 1.5 simulated light, 100 mW cm (solid).

63

3-/4-

(d) Two-electrode

(green) in the dark (dashed)

3.5 Conclusions
Through variations in the concentration of reduced species and incident light intensity,
we have shown that photocurrent density of hematite electrodes is not limited by the rate of hole
collection by the [Fe(CN)6]

4-/3-

redox shuttle. Instead, the photocurrent is strictly dependant on

the fraction of photogenerated holes that reach the semiconductor-electrolyte interface. We
performed analogous measurements employing [IrCl6]

3-/4-

as the redox shuttle and verified that

hole collection is not the rate limiting step for this system either. The substantially lower
photocurrent densities when an aqueous electrolyte was used without redox shuttle, however,
indicate that the water oxidation step is rate limiting. This can be at least partially overcome
through introducing water oxidation catalysts at the hematite surface.
oxidation with use of redox shuttles such as [Fe(CN)6]

4-/3-

26

Comparisons of water

should therefore provide a

convenient framework for determining the effectiveness of a catalyst.
Electron transfer from the semiconductor to the redox couple was identified to be the
dominant (i.e. VOC limiting) recombination mechanism. As a consequence, it is not viable to
obtain the entire possible increase in VOC though increases in the pH, since this produces a
concomitant increase in the driving force for back electron transfer. Nevertheless, an increase in
pH does lead to an increase in VOC. We further demonstrated that, analogous to increasing the
pH, the VOC can be increased through use of more positive redox shuttles. For example, use of
[IrCl6]

3-/4-

as the redox couple at pH 7 produced a VOC of 0.61 V. To the best of our

knowledge, this is the highest VOC reported using a hematite electrode. If a higher pH could be

64

used in conjunction with such a redox shuttle with a positive potential, we expect even further
significant improvements in VOC, and thus efficiency, to be possible. Unfortunately, such
positive redox shuttles are generally not stable in basic media. An alternative way to increase the
VOC would be through use of a barrier layer at the electrode surface to slow recombination.
Such a strategy has been demonstrated on dye sensitized solar cells by employing thin alumina
layers on the TiO2 surface to curb the back electron transfer.

7,27-29

Recently, this strategy has

been explored by the deposition of oxides such as alumina and silica on nanostructured
hematite.

30,31

To date, the highest efficiency regenerative PEC employing a hematite electrode was
reported by Sanchez et. al. who combined a Ge-doped hematite electrode and a [Fe(CN)6]
electrolyte at pH 9.1 to achieve an efficiency of 0.06%.

32

3-/4-

We estimate an potential efficiency of
-2

0.04% by multiplying the current density measured at the solution potential (0.31 mA cm ), the
open circuit potential relative to the solution potential (0.55 V) and the fill factor (0.24) and
-2

dividing by the incident power (100 mW cm ). While we acknowledge that this is not the true
energy conversion efficiency, since our measurements were performed with a 3-electrode setup,
this value represents a reasonable comparison of performance. Although our goal in this work
was not to achieve maximum efficiency, the nearly record efficiency using a very thin film
hematite electrode in contact with an [IrCl6]

3-/4-

redox couple at a pH 6.8 is significant. These

efficiencies, however, are still substantially below the ~15% maximum efficiency for a 2.2 eV
bandgap material. In our case, since only very thin films were examined to minimize losses due

65

to bulk recombination, only a relatively small fraction of incident light is absorbed. Therefore,
the most straightforward way to increase the overall efficiency is to increase the light absorption
while maintaining a short charge carrier collection length. This strategy is currently being
pursued by many research groups. We point out, however, that use of such nanostructued
electrodes in contact with aqueous electrolytes make it very challenging to identify the
bottlenecks preventing substantially higher efficiencies from being achieved. We believe that
systematic investigations on controllable model systems, such as those presented here, will
ultimately allow for the design of a system with superior efficiencies. Finally, one issue not
addressed so far is the very poor fill factors (typically less than 0.3) which are generally observed
with hematite electrodes. This problem is the subject of chapter 4.
3.6 Acknowledgments
TWH would like to thank Michigan State University for providing a generous start-up package
in support of this work.

66

APPENDIX

67

Figure A3-1: VOC as function of [Fe(CN)6]
-2

4-

for under 50 mWcm

-2

(yellow squares), 100 mW

-2

cm (orange triangles), and 160 mW cm (red circles) of white light.

Figure A3-2: a) Plot of short-circuit photocurrent density vs. (a) white light intensity (b)
monochromatic (410 nm) light intensity.

68

REFERENCES

69

REFERENCES

1. Klahr, B. M.; Martinson, A. B. F.; Hamann, T. W. Langmuir, 2011, 27, 461-468.
2. George, S. M. Chem. Rev. 2010, 110, 111-131.
3. Standridge, S. D.; Schatz, G. C.; Hupp, J. T. J. Am. Chem. Soc. 2009, 131, 8407-8409.
4. Dotan, H.; Sivula, K.; Gratzel, M.; Rothschild, A.; Warren, S. C. Energy Environ. Sci. 2011,
4, 958-964.
5. Leipoldt, J. G.; Bok, L. D. C.; Cilliers, P. J. Z. Anorg. Alleg. Chem. 1974, 409, 343-344.
6. Leipoldt, J. G.; Bok, L. D. C.; Cilliers, P. J. Z. Anorg. Alleg. Chem. 1974, 407, 350-352.
7. Klahr, B. M.; Hamann, T. W. J. Phys. Chem. C, 2009, 113, 14040-14045.
8. Nelson, J. J.; Amick, T. J.; Elliott, C. M. J. Phys. Chem. C, 2008, 112, 18255-18263.
9. Nelson, J. The Physics of Solar Cells; Imperial College Press: London, 2003.
10. Lewis, N. S.; Rosenbluth, M. L. In Photocatalysis: Fundamentals and Applications;
Serpone, N., Pelizzetti, E., Eds.; John Wiley & Sons: New York, 1989.
11. Fajardo, A. M.; Lewis, N. S. J. Phys. Chem. B, 1997, 101, 11136-11151.
12. Kennedy, J. H.; Frese, K. W. J. Electrochem. Soc. 1978, 125, 709-714.
13. Beermann, N.; Vayssieres, L.; Lindquist, S. E.; Hagfeldt, A. J. Electrochem. Soc. 2000, 147,
2456-2461.
14. Launay, J. C.; Horowitz, G. J. Cryst. Growth 1982, 57, 118-124.
15. Anderman, M.; Kennedy, J. H. J. Electrochem. Soc. 1984, 131, 21-26.
16. McGregor, K. G.; Calvin, M.; Otvos, J. W. J. Appl. Phys. 1979, 50, 369-373.
17. Kennedy, J. H.; Frese, K. W. J. Electrochem. Soc. 1978, 125, 723-726.
18. Souza, F. L.; Lopes, K. P.; Longo, E.; Leite, E. R. Phys. Chem. Chem. Phys. 2009, 11, 12151219.
19. Sivula, K.; Zboril, R.; Le Formal, F.; Robert, R.; Weidenkaff, A.; Tucek, J.; Frydrych, J.;
Gratzel, M. J. Am. Chem. Soc. 2010, 132, 7436-7444.
70

20. Tahir, A. A.; Wijayantha, K. G. U.; Saremi-Yarahmadi, S.; Mazhar, M.; McKee, V. Chem.
Mater. 2009, 21, 3763-3772.
21. Cesar, I.; Sivula, K.; Kay, A.; Zboril, R.; Graetzel, M. J. Phys. Chem. C, 2009, 113, 772782.
22. Dareedwards, M. P.; Goodenough, J. B.; Hamnett, A.; Trevellick, P. R. J. Chem. Soc.,
Faraday Trans. 1 F, 1983, 79, 2027-2041.
23. Tan, M. X.; Laibinis, P. E.; Nguyen, S. T.; Kesselman, J. M.; Stanton, C. E.; Lewis, N. S.
Prog. Inorg. Chem. 1994, 41, 21-144.
24. Hamann, T. W.; Gstrein, F.; Brunschwig, B. S.; Lewis, N. S. J. Am. Chem. Soc. 2005, 127,
13949-13954.
25. Hamann, T. W.; Gstrein, F.; Brunschwig, B. S.; Lewis, N. S. Chem. Phys. 2006, 326, 15-23.
26. Tilley, S. D.; Cornuz, M.; Sivula, K.; Grätzel, M. Angew. Chem. Int. Ed. 2010, 49, 64056408.
27. Ondersma, J. W.; Hamann, T. W. J. Phys. Chem. C, 2010, 114, 638-645.
28. Hamann, T. W.; Farha, O. K.; Hupp, J. T. J. Phys. Chem. C, 2008, 112, 19756-19764.
29. DeVries, M. J.; Pellin, M. J.; Hupp, J. T. Langmuir, 2010, 26, 9082-9087.
30. Spray, R. L.; McDonald, K. J.; Choi, K.-S. J. Phys. Chem. C 2011, 115, 3497-3506.
31. Le Formal, F.; Tetreault, N.; Cornuz, M.; Moehl, T.; Gratzel, M.; Sivula, K. Chem. Sci.
2011.
32. Sanchez, H. L.; Steinfink, H.; White, H. S. J. Solid State Chem. 1982, 41, 90-96.

71

Chapter 4:
Voltage Dependent Photocurrent of Thin Film
Hematite Electrodes

Adapted with permission from:
Voltage dependent photocurrent of thin film hematite electrodes, Benjamin M. Klahr and
Thomas W. Hamann, Applied Physics Letters, 2011, 99, 063508. Copyright 2011 American
Institute of Physics.

72

4.1 Abstract
The current density (J) vs. applied voltage (V) curves of thin-film hematite electrodes
under illumination exhibit non-ideal behavior. As a result very poor fill factors and photocurrent
densities are generally observed. A simple model is presented to describe the photocurrent
density behavior of hematite photoelectrodes, which assumes only drift collection of holes in a
uniform electric field.

Excellent agreement is found between the model and experimental

results. Use of this model provides important insight into the limitations of hematite electrodes
as well as strategies to achieve improved efficiency.

73

4.2 Introduction
The previous two chapters demonstrated that the use of outersphere redox shuttles or
1,2

H2O2 as hole scavengers, can collect essentially all of the holes that reach the surface.

This

3

has also been shown by Dotan et. al. with the use of H2O2. The use of such hole scavengers is
then able to de-convolute the rate limiting process of water oxidation at the surface from the
other processes in the bulk of the hematite. We again employ this strategy of using a fast redox
couple in order to elucidate the fundamental limitations of this material in the bulk.
The power conversion efficiency, η, of a regenerative photoelectrochemical cell, PEC, is
given by the fraction of the maximum power output, Pmax, to the incident solar power, Pin. The
maximum power output is the product of the photocurrent density and the voltage at the power
point, usually expressed as



Pmax J sc Voc  ff

P
P
in
in

(4-1)

where Jsc is the photocurrent density at short-circuit, Voc is the open-circuit photovoltage, and ff
is the fill factor. Hematite-based PEC’s suffer from concomitantly sub-optimal Jsc’s, Voc’s and
ff ’s. Recent efforts have focused on developing nanostructures and introducing high levels of
4-9

impurities (such as Si, Ti, Al, Ge, and Nb), primarily in order to improve Jsc.

These strategies

have produced some promising results, however the overall power conversion efficiency is still
very modest. We recently employed a series of straightforward systematic studies to determine
several limitations on the performance of thin-film hematite in terms of both the photocurrent
density and the photovoltage in regenerative PEC’s.

74

1,2

While these lessons point to several

straightforward routes of increasing the photocurrent and photovoltage, the low fill factor, which
has been typically less than 0.3, has not been addressed in detail. In this work we aim to explain
cause of the poor fill factors which limit the power conversion efficiency of hematite electrodes.
These results provide additional insights into the operational principles of hematite PEC’s and
further identifies the physical parameters which must be addressed if hematite, or analogous
materials, will ever be used effectively.
4.3 Experimental
Hematite electrodes were deposited on fluorine doped tin oxide (FTO) coated glass
1

substrates by atomic layer deposition, ALD, as described in chapter 2.

The results reported

herein were from a single 33 nm hematite electrode, however additional nominally identical
electrodes were also measured and produced nominally identical results.

The electrolyte

consisted of 200 mM of K4Fe(CN)6·3H2O, 20 mM of K3Fe(CN)6 and 200 mM KCl in a pH 12
aqueous solution (KOH).
4.4 Results and Discussion
Figure 4-1 shows typical plots of the current density, J, measured as a function of
applied voltage, V, of a thin-film hematite electrode in the dark and under 100 mW cm
1.5) illumination.

-2

(AM

The dark J-V curve is highly rectifying, however the J-V curve under

illumination is sloped. This non-ideal light J-V behavior results in a diminished fill factor, and
thus energy conversion efficiency. We note that similar looking J-V curves have been reported
elsewhere when fast hole collecting redox species are employed, including the highest reported
9

power conversion efficiency of a regenerative, hematite photoelectochemical cell.
9

consequence, no fill factor greater than 0.31 has been reported.
75

As a

This is in contrast to the

maximum fill factor of 0.8 that can be achieved with a VOC of 0.5 V, according to the diode
equation (below).

Figure 4-1. Current density vs. applied potential curves for PEC employing a 33 nm hematite
electrode in conjunction with a pH 12 [Fe(CN)6]

3-/4

electrolyte under 1 sun illumination (orange

dashed line) and in the dark (solid black line).

In general a J-V curve can be described using the diode equation:
 
 q V  IRs     V  IRs
J  J photo   J 0  exp 

  1  
 kT
 

   Rsh
 






(4-2)

where Jphoto is the photocurrent density, J0 is the exchange current density, k is the Boltzmann
constant, and T is temperature,  is the diode quality factor, Rs is a series resistance and Rsh is the
shunt resistance.

10

In principle, , Rs, and/or Rsh can cause a sloped J-V curve and diminished

fill factor as seen in figure 4-1. We have previously shown that the diode quality factor is
approximately 2 for these thin-film hematite electrodes, which cannot account for the severe
2

slope of the light J-V curves.

The series resistance of these films were measured by several
76

methods including electrochemical impedance spectroscopy (EIS), which produced Rs 110
11

(figure 4-2a) and by plotting dV/dJ vs 1/J, which produced Rs 23 (figure 4-2b).

Figure 4-2. Series resistance of the hematite photoelectrochemical cell measured by plotting (a)
dV/dJ vs 1/J, where the y-intercept is the series resistance (~23 ), and (b) Nyquist plots
showing the intercept at Z’ as the series resistance (~110 with a zoomed in view as the insert.

Because the voltage drop due to a resistance of ~100  with the small currents produced is
insignificant, the effects of series resistance on the light J-V curves is negligible. Another
possible cause of the non-ideal light J-V curves is a low shunting resistance. Shunting – for
example pinholes in the film allowing charge transfer from the FTO to the solution – would be
especially problematic when fast, reversible redox couples are employed, such as [Fe(CN)6]

3-/4-

.

ALD is a unique deposition technique, however, which is known to produce conformal pinhole
free films, even when much thinner films than the 33 nm film measured here are deposited.

12,13

In addition, we note that if there were a low shunting resistance, it would be reflected in the dark

77

J-V curve as well as the light.

10

As mentioned above, however, the dark J-V curve is highly

rectifying, which allows us to tentatively rule out shunting as the cause of the non-ideal light J-V
curves.

The Gartner model is often used to interpret voltage dependent photocurrent
14,15

densities.

The Gartner model neglects recombination in the depletion region; however we

have previously shown that depletion region recombination is significant in hematite electrodes.

2

1/2

In addition, the Gartner model was derived under reverse bias conditions which results in a V
dependence of Jphoto, which cannot reasonably fit or explain our data.

We have recently suggested that the photocurrent of hematite electrodes is strictly limited
by the drift length of holes, which in turn is dependent on the voltage drop across the
1,2

electrode.

For example, we showed that increasing the pH and/or decreasing the solution

potential, which increases the built in voltage, Vbi, results in a concomitant increase in the short
circuit photocurrent density.

2

This situation is analogous to amorphous silicon p-i-n junction

photovoltaic cells, PV’s, whose performance is largely dominated by the voltage drop across the
intrinsic region.

10

Amorphous silicon PV’s can also suffer from sub-optimal fill factors arising

from non-ideal light J-V curves. Hegedus has developed a model to describe such J-V curves of
amorphous Si PV’s, which we adopt here to explain the light J-V curves of hematite
11

electrodes.

In this model, the dark J-V is described by the bracketed term in equation 4-1.

This is in excellent agreement with the measured dark J-V behavior of hematite electrodes with γ
-2

= 2, Rs = 0 Ω, Rsh = ∞, and J0 = 3 nA cm (shown in figure 4-3).

78

Figure 4-3. Experimental dark J-V curve (orange circles) fit to the diode equation (black line).
-2

Fitting yields γ = 2 and J0 = 3 nA cm assuming Rs = 0 Ω, and Rsh = ∞. Correcting for the ~100
 series resistance has a negligible effect.

The photocurrent density is described as the product of the optically saturated photocurrent
density, given by the flux of photons that are absorbed by the material, Jabs, and the collection
efficiency, (V), which is a function of voltage:

J photo (V )  J abs  V )

(4-3)

The collection efficiency is described by the equation:

 V ) 

 Vbi 

 V 
1  

D 2   Vbi  



(4-4)

where  is the minority carrier mobility,  is the minority carrier lifetime, and D is the electrode
thickness. Note that equation (4-4) only includes the contribution of charge that is generated
within an electric field. The original equation derived by Hegedus to describe amorphous silicon
also included a term which accounted for the diffusion of charge through a concentration

79

gradient.

11

However, based on our results from chapter 2, we estimate that diffusion can be

neglected. The result is an equation where linear photocurrent density vs. applied voltage curves
are expected. The quantity Vbi / D is the hole drift length, assuming a uniform field across
the electrode, at no applied voltage. The collection efficiency is thus expected to increase with
increasing hole lifetime and mobility, and decrease with an increasing electrode thickness. This
is in agreement with our previously reported electrode thickness dependence of Jphoto; a constant
Jphoto with increasing electrode thickness (thus increasing Jabs) was attributed to a
1

concomitantly decreasing (V).

The collection efficiency is also linearly dependent on the

applied voltage relative to the Vbi, going to zero when the applied voltage offsets the built-in
voltage. An illustration of these two extreme conditions, when V = 0 and when V = Vbi can be
seen in Figure 4-4.

Figure 4-4. Qualitative energy diagram showing the conduction bands and valence bands when
V = 0 (red lines) and when V = Vbi (blue lines).
80

In order to test this model of the photocurrent density, Jphoto(V) was extracted by
subtracting the experimental dark current density from the J-V curve under illumination (the
black solid line from orange dashed line in figure 4-1). In addition, the incident light intensity
-2

was varied from 35 to 390 mW cm , which is only expected to affect Jabs. The resultant plots
of Jphoto(V) are shown in figure 4-5 for varying light intensities. In all cases, the plots of
Jphoto(V) are linear over the entire voltage range measured, consistent with equations (4-3) and
(4-4). The slope of the Jphoto(V) increases with increasing light intensity, as expected for
increasing Jabs. We note that any shunting resistance would not be sensitive to incident light
intensity, thus supporting our above claim that shunting does not significantly contribute to the
poor fill factors observed. In addition, we found that Jsc increases linearly with increasing light
intensity (see Appendix of Chapter 3).

Figure 4-5. The experimental photocurrent for varying light intensity (colored shapes) and the
linear fit (solid line) of equation (4-3) and (4-4).

81

A Jabs value was calculated at each light intensity by multiplying the absorbtance and the
incident photon spectral flux. Using this method, Jabs values were calculated to be between 0.4
-2

-2

to 13 mA cm for incident white light intensities ranging from 35 to 390 mW cm . Figure 4-6a
shows a very linear plot of Jsc vs. Jabs. The slope of this plot should be equal to the collection
efficiency at V = 0, which is ~7%. This is in very close agreement with the collection efficiency
of 8.8% calculated by integrating the absorbed photon to current efficiency shown in figure 4-6b.
This small collection efficiency accounts for the modest current densities generally observed
2

with hematite photoelectrodes. The product of /D was determined by fitting the Jphoto(V)
data to equations (4-3) and (4-4), with fixed calculated values of Jabs. The fit yielded a value of
2

-1

2

/D = 0.14 V . Using literature values for  (0.2 cm V
2

-1 -1 16

s )

and  (5 ps)

17

and our

-1

measured values for D (33 nm), the product of /D is calculated to be 0.092 V , in excellent
agreement with the fitted value. We note that the lifetime of 5 ps was measured under high
incident light intensities and therefore may underestimate the lifetime in response to the light
intensities employed in our measurements since (at least in part) additional recombination
mechanisms such as Auger recombination can diminish the lifetime under such high injection
conditions. In addition, the Vbi can be determined by the intercept of the Jphoto(V) plots (when
Jphoto = 0), which was found to be 0.49 V. Ideally, the Vbi is the difference between the flat
band potential, EFB, and the solution potential, ES; the calculated value for Vbi based on Mott
Schottky analysis of nominally identical films is 0.59 V, also in reasonable agreement with the
1

fit value produced here.

82

Figure 4-6. (a) Plot of Jsc vs. Jabs; the slope of this plot is equal to the collection efficiency at V
= 0, which is ~7%. (b) The absorbed photon to current efficiency (APCE). Integrating between
350 and 560 nm gives a collection efficiency of 8.8%.

4.5 Conclusion
In conclusion, we were able to accurately describe the total photovoltaic behavior of
hematite electrodes using a modified version of the diode equation, with a linearly dependent
photocurrent on voltage. This analysis allowed us to identify a very short hole drift length as the
cause of the generally observed poor fill factors, in addition to the sub-optimal photocurrent
densities. An additional feature of the model presented is the ability to extract the product of 
(assuming the electrode thickness is known) as well as Vbi, which we showed to be in good
agreement with literature values, from simple J-V measurements. This model thus provides a
useful framework to guide improvement of this enticing material.

83

4.6 Acknowledgements
TWH would like to thank Michigan State University for a generous start-up package in support
of this work.

84

REFERENCES

85

REFERENCES

1. Klahr, B. M.; Martinson, A. B. F.; Hamann, T. W. Langmuir, 2011, 27, 461-468.
2. Klahr, B. M.; Hamann, T. W. J. Phys. Chem. C, 2011, 115, 8393-8399.
3. Dotan, H.; Sivula, K.; Gratzel, M.; Rothschild, A.; Warren, S. C. Energy Environ. Sci. 2011,
4, 958-964.
4. Cesar, I.; Sivula, K.; Kay, A.; Zboril, R.; Graetzel, M. J. Phys. Chem. C, 2009, 113, 772-782.
5. Glasscock, J. A.; Barnes, P. R. F.; Plumb, I. C.; Savvides, N. J. Phys. Chem. C, 2007, 111,
16477-16488.
6. Sartoretti, C. J.; Alexander, B. D.; Solarska, R.; Rutkowska, W. A.; Augustynski, J.; Cerny,
R. J. Phys. Chem. B, 2005, 109, 13685-13692.
7. Kleiman-Shwarsctein, A.; Huda, M. N.; Walsh, A.; Yan, Y.; Stucky, G. D.; Hu, Y.-S.; AlJassim, M. M.; McFarland, E. W. Chem. Mater. 2009, 22, 510-517.
8. Sanchez, C.; Sieber, K. D.; Somorjai, G. A. J. Electroanal. Chem., 1988, 252, 269-290.
9. Sanchez, H. L.; Steinfink, H.; White, H. S. J. Solid State Chem. 1982, 41, 90-96.
10. Nelson, J. The Physics of Solar Cells; Imperial College Press: London, 2003.
11. Hegedus, S. S. Prog. Photovoltaics Res. Appl. 1997, 5, 151-168.
12. Ritala, M.; Leskela, M. In Handbook of Thin Film Materials; Nalwa, H. S., Ed.; Academic
Press: San Diego, 2001; Vol. 1, p 103.
13. George, S. M. Chem. Rev. 2010, 110, 111-131.
14. Gartner, W. W. Phys. Rev. 1959, 116, 84-87.
15. Reichman, J. Appl. Phys. Lett. 1980, 36, 574-577.
16. Bosman, A. J.; Vandaal, H. J. Adv. Phys. 1970, 19, 78.
17. Joly, A. G.; Williams, J. R.; Chambers, S. A.; Xiong, G.; Hess, W. P.; Laman, D. M. J. Appl.
Phys. 2006, 99, 6.

86

Chapter 5:
Photoelectrochemical Impedance
Spectroscopy of Hematite Electrodes: The
Role of Surface States

Adapted with permission from:
Water Oxidation at Hematite Electrodes: The Role of Surface States, Benjamin M. Klahr,
Sixto Gimenez, Francisco Fabregat-Santiago, Thomas W. Hamann, and Juan Bisquert.
Journal of the American Chemical Society, 2012, 134, 4294-4302. Copyright 2012 American
Chemical Society.
87

5.1 Abstract
Thin film hematite electrodes prepared by Atomic Layer Deposition were used to
study the photoelectrochemical properties of this material under water splitting conditions.
We employed impedance spectroscopy to analyze the general mechanism of water oxidation
at different conditions of voltage, light intensity and electrolyte pH. A general physical
model is proposed, which includes the existence of a surface state at the semiconductor/liquid
interface where holes accumulate. The strong correlation between the charging of this state
with the reduction in charge transfer resistance and the photocurrent onset provides new
evidence that the accumulation of holes in surface states is responsible for water oxidation.
The charging of this surface state under illumination is also related to the shift of the
measured flat band potential.

88

5.2 Introduction
Thus far, the discussed experiments have focused on elucidating the limitations of
bulk hematite by eliminating the rate limiting step of water oxidation at hematite electrodes
through the use of a fast redox shuttle. Focus is now shifted to the reaction of interest - using
photogenerated holes to oxidize water at the surface of hematite. Recently, a series of studies
based on different characterization techniques have improved our understanding of the factors
controlling the water splitting performance of hematite photoelectrodes.

1-4

These works have

pointed out the crucial role of the accumulation of holes at the surface of hematite electrodes
under visible light which is responsible for a highly anodic current onset potential. However,
the nature of this accumulated charge, and the role it plays in the water oxidation mechanism
is still under debate.
We employed phtoelectrochemical (PEC) impedance spectroscopy (IS) to determine
the general mechanism of water oxidation. Our results are based on the analysis of surface
5-8

state capacitance,

a concept developed in the 1980s, which provides evidence of the

accumulation of holes in surface states at the semiconductor/electrolyte interface. We have
found that the charge-transfer reaction to oxidize water occurs from these surface states rather
than from holes from the valence band; the photocurrent onset appears only after holes start to
accumulate in these surface sites.
5.2.2 Impedance Spectroscopy Background
Impedance Spectroscopy is a common tool used in electrochemistry for the analysis of
electrical properties of materials and interfaces. A small amplitude AC potential perturbation,
V,

is applied around a constant bias potential. The resulting oscillating current, i , is

measured from which an impedance can be calculated by Z   

V
i

89

. In the presence of a

capacitor, Z includes a real (Z’) and imaginary (Z”) component which is represented as a
single point for a single frequency on a plot of Z” vs. Z’ (often termed “Nyquist plots”). The
frequency of the AC signal is swept over the relevant range of AC frequencies,  , (typically
from MHz to mHz) to build the Nyquist plots such as those shown in figure 5-4. The
spectroscopic scan can be resolved into a combination of resistances and capacitors in a given
arrangement, called the equivalent circuit (EC). The EC is a useful tool for the interpretation
of experimental results, provided that the different elements and their particular arrangement
possess physical meaning. It is sometimes claimed that the interpretation of the results based
on an EC is ambiguous, based on the fact that different equivalent representations of a given
Z() function are possible. While there is a variety of possible representations of a single
physical-chemical model, it is also true that the EC allows for the conveyance of a great
amount of information if a model is sufficiently simplified that the resulting fits contain only
one solution, yet not oversimplified so that the EC parameters can be related to the relevant
physical processes in a useful way. Thus, while IS is straightforward to measure, the main
problem is the extraction of information contained in the data. It is important, therefore, to
clarify the basis for a given approach to treat the experimental results, and we describe ours in
the following.
The main purpose for employing IS in this case is to understand the characteristic
plots of current density with respect to applied bias voltage under steady state conditions, J-V.
Indeed, the J-V characteristics are used to derive the efficiency, considering the number of
photons impinging on the semiconductor.

9

The total resistance, Rtot, of a point on a J-V

curve can be calculated by the equation

J 

Rtot   AS

 V 

1

(5-1)

where As is the surface area of the electrode. Similarly, the resistance can be measured with
90

IS, however Rtot is able to be split into components which is made possible by the frequency
dependent impedance of a capacitor, Z   iC 

1

.

For a sufficiently thick and doped semiconductor material, there are three capacitances
which are well established in the literature. First, there is the depletion layer capacitance, Csc,
which is described by the Mott-Schottky (MS) equation:
2

 AS 
2 
k T

 
V  E fb  B 
 C  q N 
q 

0 D 
 sc 

(5-2)

where V is the applied voltage, Efb is the flat band potential, ND is the dopant density, kB is
Boltzmann’s constant, T the absolute temperature, q is elementary charge, κ is the dielectric
10

constant of the semiconductor (taken to be 32 for the hematite)
permittivity (8.854 × 10

−12

−1

CV

and ε0 is the vacuum

−1

m ). In addition, there is the series connection of the

Helmholtz capacitance, CH, where both elements have been lumped into Cbulk. These two
capacitances are dielectric in origin. In contrast to this, there is a third, chemical
11,12

capacitance,

5-8,13,14

and is termed the surface state capacitance, Css.

This capacitance

can be expressed by the equation:

C ss  AS qN ss

f ss
N ss q 2
 AS
f ss (1  f ss )
E Fn
k BT

(5-3)

where Nss is the surface density of the surface states, fss is the fractional occupancy of the
state, and EFn is the electron Fermi level of the state. For the interpretation of the IS
measurements of illuminated and dark hematite electrodes we adopt a classical model which
is depicted in Figure 5-1a.

15

From this physical model, and the capacitances discussed above,

we suggest the EC shown in Figure 5-1b. A formal derivation of the interfacial impedance is
91

given in Ref.

16

This EC can also be traced back to a large body of literature on IS of

photoelectrochemical systems.

6-8,17

Figure 5-1. (a) Proposed physical model for the charge carrier dynamics in hematite
electrodes, showing their generation, G, by light absorption, surface state trapping, and
interfacial charge-transfer reactions. (b) Equivalent circuit corresponding to physical model in
(a). (c) Simplified model used for IS interpretation created by removing Rct,ss (d) Simplified
model used for IS interpretation created by removing Rct,bulk. (e) Randles circuit

According to Equation (5-3) the surface state capacitance traces a peak with respect to
Fermi level variation at the point EFn = Ess, where Ess is the energy level of the surface state.
Therefore, a voltage scan of the capacitance should allow for a direct spectroscopic
92

measurement of the surface states of a semiconductor in the energy axis. This method has
been applied over many years, especially using an inert electrolyte that totally blocks the
current flow.

5,14

However, in general one must be careful when establishing the relationship

between the applied voltage and the energy of the surface state. In a system containing an ntype semiconductor in contact with a well-defined redox couple, the voltage in a dark
measurement relates to the difference of the Fermi level of electrons in the semiconductor,
EFn, and the redox potential of the redox couple, as indicated in Figure 5-2a. When irradiated
with suprabandgap illumination from the electrolyte side, electron–hole pairs are generated in
the region where the light is absorbed, plus one diffusion length of the minority carriers,

15,18

which are the holes in this case. Excess minority carriers are therefore created close to the
interface, with the concomitant lowering of their own Fermi level, as indicated in Figure 5-2b.
The split of the Fermi levels produces a photovoltage, and for infinitely fast exchange of
holes with the redox species across the interface, equilibrium would be achieved, where
surface
EFp
 Eredox . This means that the voltage will be given by  qV  E Fn  Eredox , just as in

surface
Figure 5-2a. The usual case, however, is that E Fp
 Eredox , due to sluggish exchange,

particularly when intermediate steps for the reaction or surface states are involved.

19

Therefore, the Fermi level of holes at the surface becomes effectively disconnected from the
applied voltage.

The concepts discussed above will be used to interpret the following

experimental results.

93

Figure 5-2. Scheme of a semiconductor with conduction band and valence band edges E c
and E v , and the quasi-Fermi levels of the respective states. (a) Applied voltage V in the
dark. (b) Photogeneration of electron hole pairs and charge transfer of hole to redox couple in
solution.
94

5.3 Experimental
Thin films of hematite were deposited on fluorine-doped tin oxide (FTO) coated glass
-2

substrates (Hartford Glass, 12 Ω cm ) by atomic layer deposition (Savannah 100, Cambridge
Nanotech Inc.) using a modified procedure to that described previously.

20

The modification

consisted of using both ozone and water as the oxidation source instead of just ozone, which
results in increased growth rate and uniformity of the hematite films compared to those made
using only ozone as the oxygen source.

21

A single precursor-oxidation cycle consisted of a

20 second ferrocene pulse followed by an oxidation sub-cycle which included 10 cycles of a
0.015 s H2O pulse followed by a 6 s ozone pulse, where each sub-cycle was separated by a 5
s purge. This oxidation cycle is thus essentially a 60 s “wet” ozone pulse. All films in this
experiment were prepared by 1,200 ALD cycles and measured to be ~60 nm by absorption
measurements (Perkin Elmer, Lambda 35 with a Labsphere integrating sphere) corrected for
reflection as described previously, as well as ellipsometric measurements (Horiba Jobin
20

Yvon, Smart-SE).

Electrolytes were prepared at pH 6.9 (employing a 0.1 M phosphate buffer) and pH
13.3 (0.1 M KOH). The pH was determined with Fisher Scientific Accumet pH meter. All
aqueous solutions contained 200 mM KCl as a supporting electrolyte. Hematite electrodes
2

were masked with a 60 μm Surlyn film (Solaronix) with a 0.25 cm hole to define the active
area and to prevent scratching of the thin films which were clamped to a custom made glass
electrochemical cell. Surlyn films were adhered to the electrodes by heating to 120° C. A
homemade saturated Ag/AgCl electrode was used as a reference electrode and high surface
area platinum mesh was used as the counter electrode. Impedance spectroscopic and
photoelectrochemical measurements were made with an Eco Chemie Autolab potentiostat
95

coupled with Nova electrochemical software. IS data were gathered using a 10 mV amplitude
perturbation of between 10,000 and 0.01 Hz. The IS data over this six order of magnitude
variation of frequencies were simultaneously fit to the equivalent circuits described in the text
using Zview software (Scribner Associates). The light source was a 450 W Xe arc lamp. An
-2

AM 1.5 solar filter and neutral density filters were used to simulate sunlight at 100 mW cm ;
-2

neutral density filters were also employed to reduce the light intensity to 33 and 10 mW cm .
All photoelectrochemical measurements were performed by shining light from the substrateelectrode (SE) interface. While all experiments shown in this work were performed multiple
times, the data shown herein are from a single hematite electrode.
5.4 Results and Discussion
5.4.1 Photoelectrochemical Results
Plots of steady state photocurrent density, J, vs. applied voltage, V, curves in response
-2

to varying light intensities – 10, 33 and 100 mW cm

– are shown in Figure 5-3a for pH 6.9

and Figure 5-3b for pH 13.3. Since the water oxidation potential and the hematite bands both
shift at the Nernstian rate of 59 mV / pH, the potentials were normalized to the reversible
22-24

hydrogen electrode reference (RHE).

The J-V curves under 100 mW cm

-2

illumination

(1 sun) are plotted vs. RHE in Figure 5-3c. The curves show remarkable overlap, however the
performance of the electrodes at pH 13.3 does show a somewhat improved current onset
potential of about 100mV.

96

Figure 5-3. J-V curves for a 60 nm thick Fe2O3 electrode at pH 6.9 (a) and 13.3 (b) under
different illumination intensities; dark J-V curves are indistinguishable from the x axis on this
scale. (c) J-V curves measured under 1 sun illumination for electrodes measured at pH 6.9 and
pH 13.3 plotted on the RHE scale.

IS measurements were performed over the same potential range as the J-V curves,
under each light intensity and in the dark, at pH 6.9 and pH 13.3. Representative Nyquist
plots under illumination are shown in Figure 5-4. Clearly there are two semicircles apparent
in the impedance spectrum. Similar looking impedance spectra have recently been reported
for hematite electrodes, and a variety of equivalent circuits put forth interpret these
3,25

spectra.

In these analyses, the low frequency semicircle is generally attributed to the
97

series arrangement of the depletion capacitance of the semiconductor and the Helmholtz
capacitance at the electrode surface, and the role of surface states has largely been ignored.
The general EC proposed in Figure 5-1b, cannot unambiguously fit the impedance spectra
since it cannot discriminate between Rct,bulk and Rct,ss. Consequently, two simplifications of
this general equivalent circuit have been employed, as shown in Figure 5-1c and 5-1d, which
are labeled as ‘model 1’ and ‘model 2’ respectively for discussion purposes.

In these

simplifications, either Rct,ss or Rct,bulk is eliminated. The simplified equivalent circuits are
excellent approximations if charge transfer (water oxidation) is dominated by one route,
either from the valence band (Figure 5-1c, model 1) or surface states (Figure 5-1d, model 2).

Figure 5-4. Nyquist plots measured at pH 6.9 at 0.65 V (red circles) and 0.70 V (orange
triangles) vs. Ag/AgCl under 1 sun conditions. Select frequencies (black symbols) labeled.

98

Figure 5-5a. Equivalent circuit parameters obtained from fitting IS data measured at pH 6.9
in the dark (red circles), at 0.1 sun (orange triangles), 0.33 sun (yellow squares), and 1 sun
(green diamonds) light intensities measured to model shown in figure 5-1c.

99

Figure 5-5b. Equivalent circuit parameters obtained from fitting IS data measured at pH 6.9
in the dark (red circles), at 0.1 sun (orange triangles), 0.33 sun (yellow squares), and 1 sun
(green diamonds) light intensities measured to model shown in figure 5-1c.

5.4.2 Equivalent Circuit Justification
Both models 1 and 2 were used to fit the EIS data for which the fits can be viewed in
figures 5-5a and 5-5b respectively. The fit results using both models produce very similar
values for Cbulk, and Css. The evidence for a particular model can be found when comparing
100

the Css and charge transfer resistances for both models. Model 1 contains a peak in Css with
a corresponding decrease in Rct,bulk. This implies that as more charge is stored in the surface
states, charge transfer from the bulk becomes more facile. This is counterintuitive and is in
contrast to conclusions made by those who propose direct hole transfer from the valence band
as general mechanism since an increased level of trapping in the surface states would be
26

expected to decrease charge transfer from the bulk (increase Rct,bulk).

A more reasonable

mechanism then, is the one described by model 2. As charge is stored in the surface states, as
indicated by a peak in Css, charge transfer from those surface states becomes more facile and
the charge transfer resistance from those surface states, Rct,ss, is reduced. The physical
correlation of surface states and a charge transfer resistance from those surface states agree
with the mathematical correlation of Css and Rct,ss suggesting that model 2 better fits the
physical processes. The simplified model, model 2, is therefore the only model employed and
discussed in interpreting the IS measurements below. Fits for data measured in pH 13.3 using
model 2 can be found in the Appendix which show a similar trend as for pH 6.9.
5.4.3 Fitting Analysis
Figure 5-6 shows a plot of the total resistance, Rtot ( Rtot  RS  Rtrap  Rct,trap ),
directly measured by IS compared to that obtained by the derivation of Rtot from the J-V
curves as shown in Equation (5-1). It is observed that both quantities are coincident, within
experimental error. The excellent agreement between the Rtot calculated from the J-V curve
and the IS parameters constitutes strong evidence that the factors determining the J-V are
represented in the IS measurements.

101

Figure 5-6. Rtot values determined by calculating dV/dJ from the J-V curves in Figure 5-4
(lines) as well as the calculated Rtot values by adding the resistances associated with charge
transfer from IS (symbols) in pH 6.9 (a) and pH 13.3 (b) electrolytes. The colors correspond
to different light intensities: 0.1 sun (orange triangles), 0.33 sun (yellow squares), and 1 sun
(green diamonds).

The series resistance, Rs, was essentially constant and small, which is consistent with
ohmic behavior at the FTO/hematite interface.
the decrease of Rct,ss , where both Css

27-29

As mentioned above, Css is correlated to

and Rct,ss

shift in concert for the different

illumination conditions tested. This clearly indicates that hole transfer for the water splitting
reaction takes place through the surface state. We note that similar behavior was obtained for
multiple photoelectrodes, including those prepared with different thicknesses. Moreover, the
photocurrent onset is also coincident with the charging of this surface state and the decrease
of Rct,ss, further confirming that charge transfer takes place from this surface state (Figure 57).

102

Figure 5-7. J-V curve (green solid line), Css (orange triangles) and Rct,ss (red circles) values
obtained for a 60 nm hematite electrode under 1 sun illumination and pH 6.9.

The values for Css displayed in Figure 5-5 and 5-7 show Gaussian behavior when the
sample is illuminated, in good agreement with the existence of a trap state described by
equation 5-3.

The peak of Css shifts to more negative potentials with increasing light

intensity at pH 6.9. At pH 13.3, however, there is no clear trend in the Css peak with light
intensity. The total magnitude of Css increases with light intensity at pH 6.9, however at pH
13.3 the magnitude is lower compared to pH 6.9 and essentially constant for the different
illumination intensities. Comparing the position of the Css peaks at both pH values, there is a
shift by approximately 59 mV/pH unit, demonstrating Nernstian behavior. The Css values
can be used to calculate the energetic distribution of the density of states, g  EFn  or DOS, by
11,30

Css  E   q g  EFn 

103

(5-4)

The calculated DOS for hematite at pH 6.9 and 13.3 can be seen in Figure 5-8a plotted vs.
Ag/AgCl under different light intensities. A plot of the DOS vs. RHE under 1 sun is also
shown in Figure 5-8b. The formal potential for water oxidation (1.23 V vs. RHE) is also
included. It is worth noting that the energetic distribution of the density of trap states peaks
are very close to the formal potential of the oxygen evolution reaction, particularly at pH
13.3. This result suggests an equilibration of trap state energy and hole accepting species in
the electrolyte. Since the water oxidation reaction involves the participation of four holes,

31

some complications in interpretation arise from the fact that holes have to be stored in
intermediate states.

3

Figure 5-8. a) DOS as a function of potential for pH 6.9 (solid lines) and pH 13.3 (dashed
lines) for 0.1 sun (orange triangles), 0.33 sun (yellow squares), and 1 sun (green diamonds)
light intensities. b) DOS curves under 1 sun illumination vs. the RHE reference.

The values for Cbulk do not show a significant trend in the logarithmic representation
of Figure 6. The values of Cbulk were fit using the Mott Schottky equation (eq. 5-2). The
resultant Mott-Schottky plots are displayed in Figures 5-9a for pH 6.9 and 5-9b for pH 13.3.

104

The calculation of the flatband potential, Efb, and the dopant density, ND, for different light
intensities, can be seen in Table 5-1. The dopant density, ND is practically constant for all the
18

tested conditions, in the range 3×10
20

values reported for this material.

18

- 6×10

-3

cm

in good correspondence with previous

The identity of these dopants has not been confirmed, due

to the very small amount of impurity needed to produce these modest doping levels in the thin
films studied. The typical assignment of oxygen vacancies for metal oxide electrodes, which
can be related to the annealing and cooling of iron oxide independent of preparation method,
29,32

may be applicable.

A recent report on the mechanism of the ALD of hematite suggests

the dopants are due to a trace Nb impurity in the ALD reactor; thus the n-type doping may be
33

due to Nb or other residual metal contamination in our ALD reactor.

In any case, we

consistently obtain these doping levels, which is typical of hematite and other metal oxide
28,29

electrodes.
a)

105

Figure 5-9. (a) MS plot at pH 6.9 under 0.1 (orange triangles), 0.33 (yellow squares), and 1
(green diamonds) sun illumination and in the dark (red circles). (b) Same as in (a) but for pH
13.3. (c) Mott Schottky plots at pH 6.9 in the dark (red circles) and under 1 sun (green
diamonds). A plot of the trap state capacitance, Css, (orange squares) is superimposed to show
the effect of Fermi level pinning.

At pH 6.9, there is an obvious positive shift in the Mott Schottky plot under
illumination; the shift increases with light intensity. This behavior is consistent with surface
state charging as described by Memming et. al.

8

This flat region in the Mott Schottky plot

under illumination from approximately 0.6 V to 0.9 V vs. Ag/AgCl at pH 6.9 corresponds to

106

the peak of the Css measured under illumination. Figure 5-9c shows a plot of Css
superimposed on the Mott Schottky plots at pH 6.9. This is a clear example of Fermi level
pinning. At pH 13, the magnitude of the surface state capacitance is lower compared to pH
6.9 (Appendix). Consequently, a lower positive shift of the flat band potential can be
anticipated by charging the surface state. This is indeed observed in Figure 5-9b. In
quantitative terms, the shift of the flat band potential (Efb) can be related to that derived
from

the

charging

of

the

surface

states,

which

is

calculated

from

Vch arg ing  Qtot / CH The total charge Qtot is obtained by integration of Css with voltage
as:

Qtot   Css dV

(5-5)

In Table 5-2, the values of Qtot are listed for the different pH and illumination intensities
tested. While there is good qualitative agreement between Efb and Vch arg ing , good
quantitative agreement is obtained only when the value of Helmholtz capacitance is
-4

CH=2×10

-2

F cm

(Table 5-2). We note that this is an extremely high value for a Helmholtz

capacitance.

107

Table 5-1. Parameters derived from the Mott Schottky plots under illumination. Calculations
assume k=32.
pH 7
Illumination
/ Sun

Efb
(V vs Ag/AgCl)

pH 13.3
ND

Efb
(V vs Ag/AgCl)

-3

(cm )
18

0
(Randle
circuit)
0.1

0.25

3×10

0.33

3.6×10

0.33

0.38

3.4×10

1

0.45

3.8×10

ND

-3

(cm )

-0.22

4.9×10

18

-0.20

5.8×10

18

-0.13

4.7×10

18

-0.07

4.8×10

18
18
18
18

Table 5-2. Correlation of the voltage shift in flat-band potential VFB and that calculated
-4

-2

from the charging of the surface state Vcharging (with CH=2×10 F·cm ).

pH 7
Qtot

-2

(C·cm )
0.1 sun

1.97×10

0.33 sun

3.05×10

1 sun

4.08×10

VFB
(mV)

pH 13.3
Vcharging
(mV)

Qtot

-2

(C·cm )

-5

76

98

2.42×10

-5

130

153

1.23×10

-5

197

204

2.42×10

VFB
(mV)

Vcharging
(mV)

-5

14

121

-5

82

61.4

-5

147

121

5.5 Conclusions
Impedance Spectroscopy was employed to investigate the photoelectrochemical
behavior of hematite electrodes under water splitting conditions. The impedance spectra are
characterized by the existence of a prominent surface state, which follows classical behavior
6

in terms of capacitive features and the dependence on voltage and illumination intensity. The
strong correlation between the Css peak with the decrease in Rct,ss, along with the
photocurrent onset as illustrated in Figure 5-7 clearly indicates that the hole-transfer step
108

leading to water oxidation takes place from surface states, and not directly from valence band
holes. This result represents an important step in understanding the mechanism of water
oxidation at metal oxide electrodes. This conclusion is however in stark contrast to
conclusions made by those who see an improvement in the water oxidation performance as a
result of adding surface state passivating layers. According to our study, true surface state
25

passivation would hinder water oxidation.

It is a goal of ongoing work in our lab to

reconcile these different interpretations.
The details of the mechanism of water oxidation with metal oxide electrodes are still
not clear, however it is generally believed to proceed from one or more surface hydroxyl (MOHx) intermediate states formed from hole transfer to a surface coordinated water and a
concomitant deprotonation step.

34-37

We therefore suggest that the surface state capacitance

that builds up during water photo-oxidation is due to a M-OHx intermediate. It is well known
that metal oxide electrodes in contact with aqueous solutions have hydroxy-terminated
surfaces; it is the protonation/deprotonation equilibrium of these M-OHx sites which gives
rise to the pH-dependent variation of the band energies of metal oxide electrodes

28,38

in

good agreement with the Nernstian behavior of the photoelectrodes observed in the present
study. This could additionally account for the somewhat different IS behavior observed for
water oxidation in pH 6.9 and 13.3 electrolytes since the surface termination chemistry would
be different. More work is needed to clarify these points, however, which is the subject of
ongoing research in our labs.
5.6 Acknowledgements
TWH acknowledges the Donors of the American Chemical Society Petroleum Research Fund
(#51099-DNI10) for support of this research. JB acknowledges support by projects from
109

Ministerio de Ciencia e Innovación (MICINN) of Spain (Consolider HOPE CSD200700007), and Generalitat Valenciana (PROMETEO/2009/058). SG acknowledges support by
MICINN of Spain under the Ramon y Cajal programme.

110

APPENDIX

111

.
Figure A5-1. Equivalent circuit parameters obtained from fitting IS data measured at pH
13.3 in the dark (red circles), at 0.1 sun (orange triangles), 0.33 sun (yellow squares), and 1
sun (green diamonds) light intensities.

112

REFERENCES

113

REFERENCES

1. Barroso, M.; Cowan, A. J.; Pendlebury, S. R.; Grätzel, M.; Klug, D. R.; Durrant, J. R. J.
Am. Chem. Soc. 2011, 133, 14868-14871.
2. Pendlebury, S. R.; Barroso, M.; Cowan, A. J.; Sivula, K.; Tang, J. W.; Gratzel, M.; Klug,
D.; Durrant, J. R. Chem. Commun. 2011, 47, 716-718.
3. Wijayantha, K. G. U.; Saremi-Yarahmadi, S.; Peter, L. M. Phys. Chem. Chem. Phys. 2011,
13, 5264-5270.
4. Dotan, H.; Sivula, K.; Gratzel, M.; Rothschild, A.; Warren, S. C. Energy Environ. Sci.
2011, 4, 958-964.
5. Frese, K. W.; Morrison, S. R. J. Electrochem. Soc. 1979, 126, 1235-1241.
6. Allongue, P.; Cachet, H. J. Electrochem. Soc. 1985, 132, 45-52.
7. Lincot, D.; Vedel, J. J. Electroanal. Chem. 1987, 220, 179-200.
8. Kelly, J. J.; Memming, R. J. Electrochem. Soc. 1982, 192, 730-738.
9. Murphy, A. B.; Barnes, P. R. F.; Randeniya, L. K.; Plumb, I. C.; Grey, I. E.; Horne, M. D.;
Glasscock, J. A. Int. J. Hydrogen Energy, 2006, 31, 1999-2017.
10. Glasscock, J. A.; Barnes, P. R. F.; Plumb, I. C.; Bendavid, A.; Martin, P. J. Thin Solid
Films, 2008, 516, 1716-1724.
11. Bisquert, J. Phys. Chem. Chem. Phys. 2003, 5, 5360-5364.
12. Bisquert, J. Phys. Chem. Chem. Phys. 2008, 10, 49-72.
13. Hoffmann, P. M.; Oskam, G.; Searson, P. C. J. Appl. Phys. 1998, 83, 4309.
14. Scurtu, R.; Ionescu, N. I.; Lazarescu, M.; Lazarescu, V. Phys. Chem. Chem. Phys. 2009,
11, 1765-1770.
15. Reichman, J. Appl. Phys. Lett. 1980, 36, 574-577.
16. Bisquert, J. J. Electroanal. Chem. 2010, 646, 43–51.
17. Vanmaekelbergh, D.; Cardon, F. J. Phys. D: Appl. Phys. 1986, 19, 643-656.
18. Gartner, W. W. Phys. Rev. 1959, 116, 84-87.
19. Salvador, P. J. Phys. Chem. B, 2001, 105, 6128-6141.
114

20. Klahr, B. M.; Martinson, A. B. F.; Hamann, T. W. Langmuir, 2011, 27, 461-468.
21. Libera, J. A.; Hryn, J. N.; Elam, J. W. Chem. Mater. 2011, 23, 2150-2158.
22. Anderman, M.; Kennedy, J. H. J. Electrochem. Soc. 1984, 131, 1565-1570.
23. Klahr, B. M.; Hamann, T. W. J. Phys. Chem. C, 2011, 115, 8393-8399.
24. Launay, J. C.; Horowitz, G. J. Cryst. Growth, 1982, 57, 118-124.
25. Le Formal, F.; Tetreault, N.; Cornuz, M.; Moehl, T.; Gratzel, M.; Sivula, K. Chem. Sci.
2011, 2, 737-743.
26. Le Formal, F.; Sivula, K.; Grätzel, M. J. Phys Chem. C, 2012, 116, 26707-26720.
27. Glasscock, J. A.; Barnes, P. R. F.; Plumb, I. C.; Savvides, N. J. Phys. Chem. C, 2007,
111, 16477-16488.
28. Morrison, S. R. Electrochemistry at Semiconductor and Oxidized Metal Electrodes;
Plenum: New York, 1980.
29. Finklea, H. O. Semiconductor Electrodes; Elsevier: Amsterdam, 1988.
30. Fabregat-Santiago, F.; Garcia-Belmonte, G.; Mora-Sero, I.; Bisquert, J. Phys. Chem.
Chem. Phys. 2011, 13, 9083-9118.
31. Tang, J. W.; Durrant, J. R.; Klug, D. R. J. Am. Chem. Soc. 2008, 130, 13885-13891.
32. Gardner, R. F. G.; Sweett, F.; Tanner, D. W. J. Phys. Chem. Solids 1963, 24, 1183-1196.
33. Martinson, A. B. F.; DeVries, M. J.; Libera, J. A.; Christensen, S. T.; Hupp, J. T.; Pellin,
M. J.; Elam, J. W. J. Phys. Chem. C, 2011, 115, 4333-4339.
34. Cook, T. R.; Dogutan, D. K.; Reece, S. Y.; Surendranath, Y.; Teets, T. S.; Nocera, D. G.
Chem. Rev. 2010, 110, 6474-6502.
35. Conway, B. E.; Tilak, B. V. Advances in Catalysis, 1992, 38, 1-147.
36. Montoya, J. F.; Peral, J.; Salvador, P. ChemPhysChem, 2011, 12, 901-907.
37. Salvador, P. J. Electrochem. Soc. 1981, 128, 1895-1900.
38. Hamann, T. W.; Lewis, N. S. J. Phys. Chem. B, 2006, 110, 22291-22294.

115

Chapter 6:
The Nature of Surface States of Water
Oxidation at Hematite Electrodes through the
Comparison with a Fast Redox Shuttle

The results in this chapter were published in:
Electrochemical and photoelectrochemical investigation of water oxidation with hematite
electrodes, Benjamin M. Klahr, Sixto Gimenez, Francisco Fabregat-Santiago, Juan Bisquert and
Thomas W. Hamann. Energy & Environmental Science, 2012, 5, 7626-7636. Reproduced by
permission of The Royal Society of Chemitry.

116

6.1 Abstract
Atomic layer deposition (ALD) was utilized to deposit uniform thin films of hematite (αFe2O3) on transparent conductive substrates for photocatalytic water oxidation studies.
Comparison of the oxidation of water to the oxidation of a fast redox shuttle allowed for new
insight in determining the rate limiting processes of water oxidation at hematite electrodes. It
was found that an additional overpotential is needed to initiate water oxidation compared to the
fast

redox

shuttle.

A

combination

of

electrochemical

impedance

spectroscopy,

photoelectrochemical and electrochemical measurements were employed to investigate the cause
of the additional overpotential. It was found that photogenerated holes initially oxidize the
electrode surface under water oxidation conditions, which is attributed to the first step in water
oxidation. A critical number of these surface intermediates need to be generated in order for the
subsequent hole-transfer steps to proceed. At higher applied potentials, the behavior of the
electrode is virtually identical while oxidizing either water or the fast redox shuttle; the slight
discrepancy is attributed to a shift in potential associated with Fermi level pinning by the surface
states in the absence of a redox shuttle. A water oxidation mechanism is proposed to interpret
these results.

117

6.2 Introduction
There have been several recent studies, employing a variety of techniques, to understand
the limited nature of the water splitting reaction with hematite photoelectrodes. For example,
Durrant and co-workers used transient absorption (TA) spectroscopy to monitor the evolution of
surface-trapped holes and acorrelated a long-lived trapped hole lifetime with the onset of
1-4

photocurrent.

One limitation of TA, however, is the difficulty with assigning the energetics of

the system, e.g. the position of the relevant surface states. Peter and co-workers recently used IS
and intensity modulated photocurrent spectroscopy (IMPS) to study the kinetics of water
oxidation at hematite electrodes which also pointed out the crucial role of surface-trapped
5-7

holes.

We also recently employed IS at different conditions of voltage, light intensity and

electrolyte pH to investigate the main steps involved in water oxidation.

8

Importantly, a general

physical model, which includes the existence of a surface state at the semiconductor/liquid
interface where holes accumulate, was established. These surface states were proposed to be
intimately involved in the water oxidation mechanism, indicated by a build-up in charge which is
concomitant with a decrease in charge transfer resistance as applied potentials become
increasingly anodic. To investigate these surface states further, we compare the PEC behavior of
3-/4-

hematite with ideally-behaving hole collector, [Fe(CN)6]

, with the behavior under water

oxidation conditions. This comparison allows for isolation of the water oxidation steps from the
behavior of the bulk hematite electrode.

We employed IS and photoelectrochemical

measurements to determine the main steps involved in water oxidation. From these combined
results we propose a mechanism of water oxidation on hematite electrode surfaces.

118

6.3 Experimental
Thin films of iron oxide were deposited on fluorine-doped tin oxide (FTO) coated glass
-2

substrates (Hartford Glass, 12 Ω cm ) by atomic layer deposition (Savannah 100, Cambridge
Nanotech Inc.) using a procedure described previously.

8

Films were prepared by 1,200 ALD

cycles and measured to be ~60 nm by absorption measurements (Perkin Elmer, Lambda 35 with
a Labsphere integrating sphere) corrected for reflection as described previously, as well as
ellipsometry measurements (Horiba Jobin Yvon, Smart-SE). Films were determined to be
9

hematite by Raman Spectroscopy and XRD measurements in a previous publication.

Hematite electrodes were masked with a 60 μm Surlyn film (Solaronix) with a 0.25 cm

2

hole to define the active area and to prevent scratching of the thin films. Surlyn films were
adhered to the electrodes by heating to 120° C. The hematite electrodes were examined in
contact with aqueous solutions buffered to pH 6.9 using a 0.1 M phosophate buffer, with 200
mM KCl as a supporting electrolyte. The pH was determined with Fisher Scientific Accumet pH
3-/4-

meter. Electrolytes containing the [Fe(CN)6]

redox couple consisted of 200 mM K4Fe(CN)6

and 20 mM K3Fe(CN)6. The protected hematite electrodes were clamped to a custom made
glass electrochemical cell. A homemade saturated Ag/AgCl electrode was used as a reference
electrode and high surface area platinum mesh was used as the counter electrode. All potentials
were converted to the reversible hydrogen electrode (RHE) scale by the equation
VRHE  VAg / AgCl  0.197  pH( 0.059 ) .

Impedance spectroscopy and photoelectrochemical

measurements were made with an Eco Chemie Autolab potentiostat coupled with Nova
electrochemical software. Impedance data were gathered using a 10 mV amplitude perturbation

119

of between 10,000 and 0.01 Hz. Data were fit using Zview software (Scribner Associates). The
light source was a 450 W Xe arc lamp (Horiba Jobin Yvon). An AM 1.5 solar filter was used to
simulate sunlight at 100 mW cm
intensity to 33 mW cm

-2

-2

(1 sun). Neutral density filters were used to adjust the

(0.33 Sun) and 10 mW cm

-2

(0.1 Sun). All photoelectrochemical

measurements were performed by shining light on the hematite electrode through the FTO
substrate such that there was no competitive light absorption from the adjacent electrolyte. Light
chopped J-V curves were measured at a rate of 75 mV/s. The light was chopped using a
computer controlled ThorLabs solenoid shutter which was set to activate every 266 ms such that
the light was turned on or off every 20 mV. Steady state J-V curves were measured at a scan rate
of 20 mV/s. Cyclic voltammogram surface state measurements were performed by applying a
potential of 2.0 V vs RHE under 1 sun illumination for 60 seconds, and then scanning over a
potential range in the dark.
6.4 Results
6.4.1 Steady State Behavior
Steady state J-V curves of a typical thin film hematite electrode immersed in an aqueous
3-/4-

electrolyte (H2O) and an aqueous electrolyte containing the [Fe(CN)6]
seen in figure 6-1.

redox couple can be

We have recently shown that the J-V response of hematite thin film
3-/4-

electrodes in contact with the [Fe(CN)6]

redox couple is limited by the fraction of holes that

reach the semiconductor/liquid interface which is determined by the potential drop across the
9-11

electrode.
unity.

10

4-

In other words, the surface-hole collection efficiency by [Fe(CN)6]

is essentially

Since the same film is used in the J-V measurements shown in figure 6-1, in aqueous

120

electrolytes buffered to the same pH under the same illumination conditions, the discrepancy in
photocurrent density and onset potential can be attributed to differences in the surface-hole
collection efficiency under H2O oxidation conditions (in the absence of a fast hole collector such
4-

as [Fe(CN)6] ). The details underlying this discrepancy are discussed below.

Figure 6-1. Steady state J-V curve of a hematite electrode in contact with a H2O (red curve) and
3-/4-

[Fe(CN)6]

(orange curve) electrolyte.

6.4.2 Electrochemical Impedance Spectroscopy
IS measurements were also performed for nominally identical hematite electrodes under
the same conditions as the J-V measurements with varying illumination intensities. Nyquist plots
resulting from measurements in each electrolyte under 1 sun illumination at 1.3 V vs RHE and at
1.6 V vs RHE are displayed in figure 6-2a and 6-2b, respectively. At 1.3 V vs RHE, two clear
semicircles are visible for the H2O electrolyte.

A full equivalent circuit representing the

hematite/liquid junction is shown in figure 6-3a. The equivalent circuit elements include a
space-charge capacitance of the bulk hematite, Cbulk, surface state capactitance, Css; the
121

resistances consist of a series resistance, Rs, a resistance which represents the trapping of charge
in the surface states, Rtrap, a charge transfer resistance from the surface states to solution, Rct,ss,
and a charge transfer resistance from the valence band to solution, Rct,bulk. To allow
unambiguous fitting of the impedance data to the equivalent circuit in figure 6-3a, Rct,bulk was
removed for the case of water oxidation, which yielded the simplified equivalent circuit in figure
8

6-3b, for reasons that have been discussed in chapter 5. At potentials positive of ~1.45 V, the
impedance spectra for water oxidation loses the low frequency semicircle and becomes a single
semicircle; in this case the simple Randle’s circuit, shown in figure 6-3c, is used to fit these
impedance spectra. Only one semicircle is visible at all potentials and light conditions measured
3-/4-

with the [Fe(CN)6]

electrolyte, therefore a simple Randle’s circuit is also used to fit these

impedance spectra.

Figure 6-2. Nyquist plots for IS data measured under 1 sun illumination for H2O (red circles)
3-/4-

and [Fe(CN)6]

(orange triangles) electrolytes at (a) 1.3 V vs. RHE and (b) 1.6 V vs. RHE.

122

Figure 6-3. a)

Full equivalent circuit used to conceptualize hematite-liquid junctions. b)

Simplified model used to fit IS data for Nyquist plots that show 2 semicircles. c) Simplified
model used to fit IS data for Nyquist plots that show 1 semicircle.

Results from fitting the impedance spectra to the appropriate equivalent circuits described
above can be seen in figure 6-4 (additional fitting results not discussed are provided in the
Appendix). Figure 6-4a shows a Gaussian distribution of Css in the H2O electrolyte, which shifts
to more negative potentials with increasing light intensity; this behavior was described
8

previously. The Css peak corresponds with a decrease in Rct,ss (Fig 6-4b), which indicates that
interfacial charge transfer (water oxidation) occurs as a result of charging of these surface
8

3-/4-

states. With [Fe(CN)6]

in solution, however, the charge transfer resistance is essentially

constant over the measured potential range for a given light intensity as shown in figure 6-4c.
123

The absence of a measureable surface state capacitance and absence of a dip in the charge
4-

transfer resistance suggests that photooxidaiton of [Fe(CN)6]

does not involve surface states

and can be described as a simple outersphere hole collection process from the valence band. In
both electrolytes, there is an obvious decrease of the charge transfer resistance with increasing
light intensity which is attributed to a concomitant increase in the number of photogenerated
valence band holes with light intensity.
Mott Schottky plots were created from the derived Cbulk values from both H2O and
3-/4-

[Fe(CN)6]

electrolytes in the dark and under 1 sun illumination conditions, which are shown

in figure 6-4d.

The Mott Schottky plots measured under illumination for water oxidation

conditions show a significant horizontal shift compared to the dark, which is an example of
8,12

Fermi level pinning.

3-/4-

The Mott Schottky plots with the [Fe(CN)6]

electrolyte are

identical in the dark and under illumination, however, further indicating that the surface states do
4-

not play a role in [Fe(CN)6]

oxidation. In addition, the impedance spectra and Mott Schottky
3-/4-

plots are essentially identical for the H2O and [Fe(CN)6]

electrolytes in the dark, which

indicates the surface states are only active under illumination. The details of the surface states are
described further below.

124

Figure 6-4. Equivalent circuit parameters obtained from fitting IS data for a hematite electrode
3-/4-

in contact with H2O (a and b) and [Fe(CN)6]

(c) electrolytes under 0.1 sun (orange

triangles), 0.33 sun (yellow squares) and 1 sun (green diamonds) illumination. d) Mott Schottky
plots calculated from Cbulk values determined from fitting the impedance spectra for H2O
3-/4-

(circles) and [Fe(CN)6]

(triangles) electrolytes in the dark (grey) and under 1 sun

illumination (colored).

125

The total resistance, Rtot was calculated from the IS fittings (Rtot = RS + Rtrap + Rct,ss for
4-

H2O oxidation and Rtot = RS + Rct,bulk for [Fe(CN)6]

oxidation) and was compared to the
8

resistance calculated from the J-V curves where Rtot=dV/dJ. A plot of Rtot calculated from both
3-/4-

methods for both H2O and [Fe(CN)6]

electrolytes can be seen in figure 6-5 under 1 sun

illumination. The agreement between the calculated Rtot from the IS and J-V curves shows that
all of the resistances which determin the J-V curve are represented in the impedance
measurements. At potentials negative of the photocurrent onset potential for H2O oxidation, Rtot
is found to be approximately two orders of magnitude higher compared to Rtot for the oxidation
4-

of [Fe(CN)6] . This large resistance is consistent with the lack of photocurrent density at these
potentials. At around 1.2 V, the Rtot due to water oxidation drops coincidently with the initiation
3-/4-

of photocurrent. Interestingly, Rtot for both H2O and [Fe(CN)6]

electrolytes become equal

at about 1.6 V vs RHE, suggesting that this hole transfer reaction is not the rate limiting step in
water oxidation at positive potentials.

126

Figure 6-5. Rtot calculated from dV/dJ of the J-V curves (lines) and from IS data (symbols) for
3-/4-

H2O and [Fe(CN)6]

electrolytes under 1 sun illumination.

6.4.3 Transient Behavior
In order to gain further insight into the differences in behavior of hematite electrodes in
3-/4-

contact with H2O and [Fe(CN)6]

electrolytes, J-V curves were also measured under light

chopping conditions and compared with J-V curves with constant illumination. Figure 6-6a
3-/4-

shows the superimposed chopped and constant light J-V curves when [Fe(CN)6]

is present in

the solution. The instantaneous photocurrent density with chopped light reaches the constant
light photocurrent density and remains constant until the light is turned off, where the current
11

immediately decays to J0.

This ideal behavior is consistent with the expectations of a fast hole

collector efficiently scavenging valence band holes at the surface of a hematite photo10,13,14

electrode.

In the case of water oxidation, however, the behavior is quite different: see

figure 6-6b. At potentials from 0.8 – 1.2 V vs. RHE, there is a significant instantaneous
photocurrent density when the chopped light is turned on, which quickly decays to the near zero
127

steady-state photocurrent density under constant illumination. Interestingly, the peak of the H2O
anodic transient photocurrent density closely matches the steady state photocurrent density
3-/4-

measured with the [Fe(CN)6]

electrolyte (also shown in 6-6b). This is attributed to the

identical flux of photogenerated holes which initially reach the surface. When the holes that
4-

reach the surface are efficiently collected, as in the case of [Fe(CN)6]

oxidation, this flux

determines the steady state photocurrent density. Under water oxidation conditions, however,
the photocurrent density decay indicates that a faradaic current cannot be sustained and the hole
transfer efficiency is therefore minimal. In this case, the flux of holes that reach the surface must
either be trapped in surface states or recombine with conduction band electrons. When the light
is switched off at potentials where transient anodic peaks are observed, transient cathodic peaks
are also visible. The cathodic transients are thus attributed to reducing the surface-trapped holes.
This behavior is consistent with previous reports where current transients were also attributed to
the charging and discharging of surface states of -Fe2O3,

14-16

CdTe,

17

18

and TiO2.

Indeed,

measuring transient peaks through light chopping has become a prevalent method of qualitatively
probing the surface states of hematite.

13,19-23

All additional photogenerated holes, which are

not stored in surface states, therefore undergo recombination.

128

Figure 6-6. (a) Chopped light (grey line) and steady state (orange line) J-V measured under 1
sun illumination for a hematite electrode in contact with a Fe(CN)6

3-/4-

electrolyte. (b) Chopped

light (grey line) and steady state (red line) J-V measured under 1 sun illumination for a hematite
electrode in contact with a H2O electrolyte (the steady state J-V curve measured in a
Fe(CN)6

3-/4-

electrolyte (yellow line) is also shown for comparison).

In order to deconvolute the change in potential with change in current over time as in the
data shown in figure 6-6b, current transients at a constant potential were also measured. These
measurements were performed by measuring the current as a function of time at a fixed potential
in response to turning on (anodic) and off (cathodic) 1 sun illumination. Plots of the anodic and
cathodic current transients measured at various constant potentials can be seen in figures 6-7a
and 6-7b, respectively. Clearly, there is a dependence in the magnitude of the current transient
with potential as well as the rate of decay. In order to determine the amount of charge stored in
the surface state at a given potential, the current vs. time plots in figure 6-7 were integrated. A
plot of the anodic and cathodic charge stored as a function of potential is shown in figure 6-8a.

129

The anodic charge was not calculated for potentials above 1.2 V vs RHE because the faradaic
and charging current cannot be separated accurately.

The cathodic and anodic charge is

approximately equal, at least for potentials up to 1.2 V vs RHE, suggesting that the charge
measured is related to the amount of charge trapped in surface states on the hematite electrode.
There is a clear peak in the stored charge at approximately 1.3 V vs RHE. Interestingly, this is
near the potential which yields the peak CSS from IS measurements. The same experiments
were also performed in response to varying illumination intensities; the results are displayed in
figure 6-8b which were calculated from cathodic current transients. The magnitude of the peak
clearly increases with light intensity and the peak shifts to more negative potentials with
increasing light intensity. This is the same trend found with the peak of Css as a function of light
intensity measured by IS. The charge measured by the current transients can be related to Css by
the equation CSS 

Q
where ∆V is the change in the quasi-Fermi level as a result of turning the
V

light off at a constant potential. Assuming the surface states probed by IS and current transients
are the same, ∆V can be calculated by plotting CSS obtained from IS vs. ∆Q measured from
current transients (see Figure 6-8d). From the slope of these plots, ∆V was found to be 0.08 ,
0.13 and 0.17 V for 0.1, 0.33 and 1 sun respectively. This corresponds to an increase of 92 mV
per 10 fold increase in light intensity, in good agreement with ideal behavior. The Css were
calculated from the current transients and compared to the Css measured from IS and can be seen
in figure 6-8c. Taken together, these results confirm that the CSS measured with IS is the same
entity probed by measuring the stored charge of the current transients.

130

Figure 6-7. J measured for a constant V vs. RHE plotted vs. time for when 1 sun illumination is
turned on (a) and off (b). Time is corrected so that the all of the peaks occur at 10 ms.
Measurements are shown for 0.8 V (red solid line), 1.0 V (orange dotted line), 1.2 V (yellow
small dashed line), 1.4 V (green dashed dotted line), and 1.6V (blue large dashed line) vs. RHE.

131

Figure 6-8. (a) Transient anodic (red circles) and cathodic (orange triangles) charge passed as a
function of potential. (b) Transient cathodic charge vs. potential (c) ) Css from EIS vs Q from
transients for a given potential measured. d) Css determined from IS, (open shapes) and CSS
determined from current transients (filled in shapes). (b) (c) and (d) show data measured under
0.1 sun (orange triangles), 0.33 sun (yellow squares) and 1 sun (green diamonds) illumination.

The potential dependence of the transient photocurrent decay was also examined.
Abrantes and Peter examined the decay of current transients of passive iron oxide electrodes in

132

an H2O electrolyte and were able to gain kinetic insight on the back reaction of electrons
reducing surface states (i.e. filling surface trapped holes) by fitting the decays to a single
exponential function.

14

We found, however, that a double exponential decay is necessary to

achieve good fits for the decays measured (See Appendix). Fitting the decays to a double
exponential decay function ( J  t   J 0 1 et /  1  J 0 2 et /  2  J    where J(0) is the
maximum current for the decay process at time = 0,  is the lifetime of the process and J(∞) is
the steady state current reached after both processes have fully decayed) produced two lifetimes
that differ by approximately an order of magnitude, which are displayed in figure 9. Two
lifetimes were also calculated from current transients for both doped and undoped hematite
electrodes in water oxidation conditions by Glasscock et. al.

24

This result is also in reasonable

agreement with a recent report by Pendlebury et. al. of transient absorption spectroscopic
measurements on hematite electrodes under water oxidation conditions.

1

Both lifetimes are

dependent on potential where the lifetimes increase by approximately an order of magnitude
from when there is no faradaic current (1.0 V vs RHE) to when there is signification faradaic
current (>1.4 V vs RHE). This also agrees with a recent paper correlating long lived holes and
photocurrent on hematite electrodes.

1

The origin of the two lifetimes found actually result from a single process: the
discharging of the surface state. As mentioned above, under illumination, charge is accumulated
in the surface state. At short times after switching off the light, the conditions of the trap state
(amount of charge accumulated, trap transfer resistance) are similar as when under illumination.
Therefore, the short lifetimes obtained from the transients in figure 7-7b are very close to the

133

relaxation times of the surface state at 1 sun: 1  ( RS  Rtrap )  CSS . As the surface state
discharges (is reduced), both Rtrap and Css change. In fact, the amount of charge accumulated in
the surface states sweeps the states produced at lower light intensities until reaching dark
conditions. Thus, the discharging time rises providing the second, slower process,  2 . Figure 9
shows the two lifetimes fit from the double exponential decay of the transients (figure 7-7b) as
well as the relaxation time of the trapped electrons measured from IS data under 1 sun and 0.1
sun illumination. It is observed that there is a good match between the fast time constant of the
transient discharge time and the relaxation time measured under 1 sun illumination. As the light
intensity diminishes, conditions approach those in the dark thus the surface state discharging
time obtained at 0.1 sun approaches the slow time constant found from the cathodic discharge.
The increase in the time constant obtained is due to the increase in Rtrap during the discharge of
the surface state.

134

Figure 6-9. A plot of the 2 different lifetimes (black and grey circles) fit from a double
exponential of the cathodic decays shown in figure 6-7b. The lifetime of the surface state
charging-discharging process at 1 sun (green squares) and at 0.1 sun (orange squares) are also
shown.
6.4.4 Cyclic Voltammetry
The anodic and cathodic current density transients measured above are attributed to the
reversible oxidation and reduction, respectively, of surface species. In other words, surface
trapping of holes means the photooxidation of a chemical species on the surface (discussed
below), and surface recombination is the reduction of that oxidized surface species.

We

therefore reasoned that we should also be able to measure the light-generated surface trapped
holes using cyclic voltammetry in the dark. In this experiment, holes accumulate on the surface
25

by applying a positive potential under illumination,

followed by measuring the cathodic
26,27

current in the dark as the potential is scanned negatively and the surface states are reduced.
A representative experiment measured at 200 mV s

-1

is shown in figure 6-10a. On the first

cycle, shown as a red solid line, a peak in current appears at around 1.3 V and 0.8 V vs RHE.
On the second cycle, both of these peaks are gone. The fact that the peaks cannot be seen on the
135

second cycle is consistent with the idea that these peaks represent the reduction of surface
trapped holes as the potential is scanned negative on the first cycle. These holes are not reoxidized by the application of a positive potential (at least on the relatively fast timescale of the
experiment) which indeed means that the holes are generated by photoexcitation of hematite and
not by the potential. Thus there are no longer oxidized surface species to be reduced on the
second cycle. The peak at 1.3 V is very close to the peak of CSS fit by the IS data, shown in
figures 6-4a, as well as the potential where the maximum transient charge is found, shown in
figure 6-8. This suggests the peak at 1.3 V is due to filling the surface states which are
associated with water oxidation. In order to probe the timescale of the transient peaks, a scan
rate dependence of the CV measurements was performed.

Capacitance plots were then

calculated by dividing the current density by the scan rate, which can be seen in figure 6-10b for
varying scan rates. It is apparent that the species probed at 1.3 V vs RHE decreases with scan
rate which means the states occupancy decays over time. The time constant(s) of this decay
process cannot be confirmed quantitatively because of the constantly changing potential,
however the timescale is in general agreement with the transient lifetimes displayed in Fig 6-9.
This further suggests that the state measured is the same measured in the current transients. The
charge passed from the surface states corresponding to the peak at 1.3 V vs RHE was calculated
-2

to be 25 µC cm

-1

when measured at a fast scan rate of 1 V s . The similar values between the

charge measured through the CV peak, the CSS peak from IS and the charge passed through
current transients, as well as the lifetimes measured from the decay of the CV peak and current
transients, all suggest that the species being measured around 1.3 V vs RHE is the same for all
3-/4-

measurements. Similar CV measurements were also performed with [Fe(CN)6]

136

in solution,

3-

(see Appendix) however no cathodic current was measured until the reduction of the [Fe(CN)6]

began, again suggesting that the surface states are associated with water oxidation intermediates
and not intrinsic to the hematite electrode.

Figure 6-10. (a) Cyclic voltametry of hematite in H2O scanned at 200 mV s

-1

after holding the

electrode at a potential of 2 V vs RHE for 60 seconds under 1 sun illumination. The first cycle is
a solid red line and the second cycle is a dashed orange line. (b) CVs measured with the same
-1

conditions as (a) at different scan rates: 20 mV s
-1

200 mV s

(yellow, medium dash), 500 mV s

-1

(red, solid), 50 mV s

-1

(orange, long dash),

(green, short dash) and 1,000 mV s

-1

(blue,

dots).

6.5 Discussion
The J-V curves of hematite electrodes immersed in an aqueous electrolyte containing
3-/4-

[Fe(CN)6]

can be described by the potential dependent fraction of photogenerated holes that
11

reach the electrode surface.

These valence band holes at the surface are efficiently collected
137

4-

10

by [Fe(CN)6] , and this faradaic charge transfer results in a steady-state photocurrent density.

Under H2O oxidation conditions the flux of holes to reach the surface (which is just a property of
the same bulk semiconductor) is the same, however a steady-state photocurrent is not produced
at low applied potentials. Instead of the surface holes undergoing interfacial charge transfer to
solution, a fraction of the holes are initially (at low potentials) trapped in surface states; the
remainder recombine.

This charging of surface states was observed through IS, transient

photocurrent and cyclic voltammetric measurements. These results provoke questions regarding
the identity of these surface states and their role in the water oxidation process. Instead of being
what is generically termed a surface state, we believe the surface states presented herein actually
consist of intermediates in the overall water oxidation reaction. One piece of (albeit indirect)
evidence of this is that the surface state is only, and always, observed under water oxidation
conditions – not in the dark or in the presence of a competitive hole scavenger. In addition, the
3

pH-dependant surface state is at the same energy as the water oxidation potential, which we
believe is not just a coincidence.
We therefore propose that the charging of the surface state is actually the first oxidative
step in the overall water oxidation reaction. The fully hydroxylated species, (OH)3–Fe–H3O3–R
where R represents the bulk hematite crystal, is the thermodynamically most stable surface
28

configuration of the (0001) surface.

Hellman and Pala recently suggested that the first step of

water oxidation on hematite electrodes is the proton coupled oxidation of such surface hydroxide
29

species.

This is also in agreement with one of the general mechanisms of water oxidation at
30-32

metal oxide electrodes previously reviewed by others.

138

The initial step is thus given by:

h



Fe( III )  OH  Fe( IV )  O  H 


(1)

e

We believe this reaction is what is referred to as “hole trapping” or “surface state charging.”
Hellman and Pala further calculated that the most stable configuration during water oxidation (at
higher applied potentials) is the oxidized (OH)3-Fe-O3-R surface, also consistent with this
assignment.

29

Since the oxidized species on the surface, Fe=O, can be reduced, as demonstrated

by the cathodic current transients (figure 6-7) and the CV measurements (figure 10), the reaction
is written as reversible.
While it is very challenging to directly determine the identity of surface states on a bulk
electrode, there are molecular iron catalysts which can serve as analogs. There are several
examples of Fe(IV)=O complex intermediates which are known to be involved in catalytic
33,34

cycles.

For example, a Fe-TAML complex was recently shown to catalyze the oxidation of

water to dioxygen.

35

Further, Collins and co-workers have shown that the Fe(IV)-oxo species

has a reduction potential of 0.9 V vs NHE in basic conditions, according to the reaction
 Fe  IV  L   O  OH2 



2

2
 e  H2O  Fe  III  L  OH  OH 2   OH 



(2)

which is essentially the recombination reaction (reduction of the surface intermediate species)
proposed herein.

34

This behavior of the Fe(III)-OH / Fe(IV)=O molecular complexes is

therefore in good agreement with the proposed first water oxidation step on hematite electrodes.
In addition, it was shown that an intermediate Fe(IV)=O complex has a broad absorption
spectrum centered at 820 nm.

33

This absorbance is also consistent with the broad near IR

absorption observed during water oxidation with hematite electrodes.

139

1,7

There are two important features of the surface state behavior that provide further insight
into the water oxidation mechanism. The first is that at potentials where the surface oxidation
begins, the formation of the proposed intermediate Fe=O doesn’t result in subsequent water
oxidation steps as indicated by the lack of steady-state photocurrent; i.e. the transient
photocurrent at these potentials decays to zero. Attack of the surface Fe=O by water molecules,
leading to a surface bound peroxide followed by decomposition to O2, should be
29

thermodynamically downhill.

In this case, it is expected that the steady-state concentration of

surface Fe=O should lead to photocurrent production. Since no photocurrent is observed under
these conditions, this is not the predominant mechanism. The other feature is the fact that
steady-state concentration of surface Fe=O reaches a peak that is coincident with a sharp drop in
the charge transfer resistance and an increase of photocurrent. This indicates that a critical
concentration of the surface intermediate species is necessary before the further water oxidation
steps can proceed.

The necessary number of surface intermediates can be estimated by

integrating the capacitance as a function of potential. The CSS peaks determined from IS data
were fit to a Gaussian curve and integrated to calculate the total charge of the surface states to be
24.3, 31.6 and 38.8µC cm

-2

for 0.1, 0.33, and 1 sun illumination respectively. Using the value
14

for 1 sun illumination as an example, this corresponds with a surface coverage of 2.4 × 10
2

-

cm

. The surface coverage is then calculated to be ~53%, by using the unit cell parameters for the
29,36-39

0001 surface of hematite, and estimating 1 active site per unit cell.

This behavior is

consistent with the surface intermediate decomposing bimolecularly, such as:
2 Fe( IV )  O  2 Fe( III )  O2


140

(3)

We note that such an oxygen formation mechanism has been proposed extensively before,
although not applied to hematite photoelectrodes.

30-32

A water molecule, or hydroxide, could

then quickly coordinate to the open iron site to complete the photocatalytic cycle.
At positive potentials (≥1.45 V vs. RHE) the current transients become negligible, the
low frequency semicircle disappears from the IS spectra and high steady state photocurrent
densities are achieved. This behavior is essentially indistinguishable from when the fast hole
4-

collector [Fe(CN)6]

is present in solution (figure 6-6b). Further, these two systems have the

same charge transfer resistance at positive potentials. Therefore, the surface hole collection
efficiency must also be essentially unity under water oxidation conditions at such positive
potentials.

Thus, once a sufficient concentration of intermediate is built up, allowing the

subsequent steps of water oxidation to proceed, water oxidation is not the rate limiting step in the
photoelectrochemical behavior of hematite electrodes; the flux of holes to the electrode surface
limits the photocurrent density at these potentials.
3-/4-

We note that the J-V curves of the H2O and [Fe(CN)6]

electrolytes, as seen in figure

6-1, do not overlap exactly, even at positive potentials; the H2O electrolyte always produces
3-/4-

somewhat lower photocurrents at a given positive potential compared to [Fe(CN)6]

. Since

we have demonstrated the surface hole collection efficiency is essentially unity for both of these
systems, and the surface hole concentration is determined by the potential drop of the electrode,
the discrepancy can be attributed to the extra potential needed to compensate for Fermi level
pinning in the H2O electrolyte. According to the Mott-Schottky plots, the bands under water
3-/4-

oxidation are shifted positive compared to [Fe(CN)6]

141

, or compared to the MS at low applied

potentials. The shift in the band is calculated to be ~140 mV, as shown in figure 6-11a. Shifting
the J-V curve in the H2O electrolyte by 200 mV produces excellent agreement with the
3-/4-

[Fe(CN)6]

J-V curve at positive potentials (Figure 6-11b). These plots show that the lower

current measured at high applied potential in water oxidation compared to [Fe(CN)6]

3-/4-

can be

almost entirely attributed to Fermi-level pinning, and not decreased hole transfer kinetics.

Figure 6-11. a) Mott Schotty plot made from Cbulk values measured under 1 sun illumination.
The lines represent lines of best fit of the points measured at potentials negative (red circles) and
positive (green circles) of the shift in the flat band. The white circles were not used in either fit.
Both sets of data were made to share the same slope in the fit. Using this method, the shift in flat
band potential is calculated to be 140 mV. Also included is the Css measured under one sun
illumination (green diamonds) and the Gaussian fit (green line). b) J-V curve measured for a
3-/4-

hematite electrode in a [Fe(CN)6]

electrolyte (orange curve) and in a H2O electrolyte (red
3-/4-

curve). The H2O curve has been shifted by 200 mV to overlap with the [Fe(CN)6]

142

curve.

Finally, the above discussion of the intrinsic role of the surface state seen at 1.3 V does
not include the other state that we observed around 0.8 V in the CV measurements. Since this
state’s energy is significantly higher than the water oxidation potential, and it is not observed in
the impedance spectra at photocurrent onset, we do not think it plays an active role in water
oxidation. There are surface terminations of hematite, in addition to the (OH)3–Fe–H3O3–R
termination discussed above, that are stable. According to Trainor et. al., the other most relevant
surface is the (OH)3–R termination.

28

This surface can also be oxidized, i.e. trap a hole, but it is

thermodynamically uphill for subsequent water oxidation.

Importantly, this termination’s

oxidation potential is approximately 0.5 V more negative than the (OH)3–Fe–H3O3–R
termination, which is in excellent agreement with the differences in potential of the two surface
states we observed.
6.6 Conclusions
The photoelectrochemical behavior of hematite electrodes in contact with H2O and
3-/4-

[Fe(CN)6]

electrolytes were compared. The steady state photocurrent onset potential was
3-/4-

approximately 0.6 V higher in H2O compared to [Fe(CN)6]
10,13,40

literature reports.

, which is in accord with

The much more positive onset potential represents a loss of energy

which must be minimized for any practical use of hematite in water splitting applications. A
combination of impedance spectroscopy, photocurrent transient and cyclic voltammetry
measurements were therefore employed to determine the cause of the large overpotential for
water splitting. We found that at moderate potentials, prior to the photocurrent onset potential,
photogenerated holes became trapped in surface states. We deduce that the physical meaning of
143

photogenerated holes being trapped at the surface is that this is the first step in the water
oxidation process, specifically the reversible oxidation of surface hydroxide species. It is only
after a buildup of these oxidized intermediates that water oxidation proceeds. This calls into
question the strategy of passivating surface states on hematite electrodes which has been pursued
recently since they might be an inherent feature of oxidizing water, and not necessarily
21,23

deleterious.

It is well known that water oxidation proceeds by an innersphere mechanism,

thus a proposed mechanisms should account for this;

32

describing this reaction as a charge

transfer from a valence band hole is misleading. In addition, the negligible faradaic, steady state
photocurrent densities at these potentials is often attributed to slow hole transfer kinetics of water
oxidation.

2,3,5,6,15

We suggest that this may be an intrinsic part of the bimolecular mechanism

of water oxidation on the hematite surface.

At high applied potentials, however, the
3-/4-

photocurrent density is essentially the same for the H2O and [Fe(CN)6]

systems; the minor

difference is attributed to Fermi level pinning in the H2O electrolyte. This ideal behavior at high
potentials is consistent with the idea that water oxidation is not the rate limiting step, once a
critical concentration of surface intermediates is formed. While the addition of catalysts have
been shown to improve performance, the cause of the improvement has not been
established.

4,19,41-44

Insight into the mechanism of water oxidation on hematite, and other

surfaces, is critical for the design of better catalysts or other systems aimed at improving the
efficiency of water splitting. Work is ongoing in our lab to clarify these points.

144

6.7 Acknowledgements
TWH thanks the National Science Foundation (CHE-1150378) and the Donors of the American
Chemical Society Petroleum Research Fund (51099-DNI10) for support of this research. TWH
and BK also thank Jesse Ondersma for helpful conversations. JB & FFS acknowledge support by
projects from Minis- terio de Economia y Competitividad of Spain (MINECO) (Consolider
HOPE

CSD2007-00007),

and

Generalitat

Valenciana

(PROMETEO/2009/058).

acknowledges support by MINECO of Spain under the Ramon y Cajal programme.

145

SG

APPENDIX

146

Figure A6-1. a) Cbulk and b) Rtrap values fit from IS data of a hematite electrode in contact with
a H2O electrolyte in the dark (red circles), under 0.1 sun (orange triangles), 0.33 sun (yellow
squares) and 1 sun (green diamonds) illumination.

Figure A6-2. Cbulk values fit from IS data of a hematite electrode in contact with a
3-/4-

[Fe(CN)6]

electrolytes in the dark (red circles), under 0.1 sun (orange triangles), 0.33 sun

(yellow squares) and 1 sun (green diamonds) illumination.

147

Figure A6-3. Experimentally measured decay measured at 1.2 V vs RHE (black line) and fits
for a single exponential decay (blue small dashed line) and double exponential decay (red long
dashed line)

3-/4-

Figure A6-4. Cyclic voltammetry of hematite in [Fe(CN)6]

electrolyte scanned at 200 mV

-1

s after holding the electrode at a potential of 2 V vs RHE for 60 seconds under 1 sun
illumination. The first cycle (solid red line) and the second cycle (dashed orange line) are
overlapping.
148

REFERENCES

149

REFERENCES

1. Pendlebury, S. R.; Cowan, A. J.; Barroso, M.; Sivula, K.; Ye, J.; Gratzel, M.; Klug, D. R.;
Tang, J.; Durrant, J. R. Energy Environ. Sci. 2012, 5, 6304-6312.
2. Pendlebury, S. R.; Barroso, M.; Cowan, A. J.; Sivula, K.; Tang, J. W.; Gratzel, M.; Klug, D.;
Durrant, J. R. Chem. Commun. 2011, 47, 716-718.

J. R.; Klug, D. R. J. Am. Chem. Soc. 2011, 133, 10134-10140.
J. Am.
Chem. Soc. 2011, 133, 14868-14871.
5. Upul Wijayantha, K. G.; Saremi-Yarahmadi, S.; Peter, L. M. Phys. Chem. Chem. Phys. 2011,
13, 5264-5270.
6. Peter, L. M.; Wijayantha, K. G. U.; Tahir, A. A. Faraday Discuss. 2011, 155, 309-322.
7. Cummings, C. Y.; Marken, F.; Peter, L. M.; Upul Wijayantha, K. G.; Tahir, A. A. J. Am.
Chem. Soc. 2012, 134, 1228-1234.
8. Klahr, B.; Gimenez, S.; Fabregat-Santiago, F.; Bisquert, J.; Hamann, T. W. J. Am. Chem. Soc.
2012, 134, 4294-4302.
9. Klahr, B. M.; Martinson, A. B. F.; Hamann, T. W. Langmuir, 2011, 27, 461-468.
10. Klahr, B. M.; Hamann, T. W. J. Phys. Chem. C, 2011, 115, 8393-8399.
11. Klahr, B. M.; Hamann, T. W. Appl. Phys. Lett. 2011, 99, 063508.
12. Morrison, S. R. Electrochemistry at Semiconductor and Oxidized Metal Electrodes; Plenum:
New York, 1980.
13. Dotan, H.; Sivula, K.; Gratzel, M.; Rothschild, A.; Warren, S. C. Energy Environ. Sci. 2011,
4, 958-964.
14. Abrantes, L. M.; Peter, L. M. J. Electroanal. Chem. 1983, 150, 593-601.
15. Dareedwards, M. P.; Goodenough, J. B.; Hamnett, A.; Trevellick, P. R. J. Chem. Soc.,
Faraday Trans. 1 F, 1983, 79, 2027-2041.
16. Hardee, K. L.; Bard, A. J. J. Electrochem. Soc. 1977, 124, 215-224.

150

17. Lincot, D.; Vedel, J. J. Electroanal. Chem. 1987, 220, 179-200.
18. Salvador, P. J. Phys. Chem. 1985, 89, 3863-3869.
19. McDonald, K. J.; Choi, K. S. Chem. Mater. 2011, 23, 1686-1693.
20. Le Formal, F.; Graetzel, M.; Sivula, K. Adv. Funct. Mater. 2010, 20, 1099-1107.
21. Le Formal, F.; Tetreault, N.; Cornuz, M.; Moehl, T.; Gratzel, M.; Sivula, K. Chem. Sci.
2011, 2, 737-743.
22. Sivula, K.; Zboril, R.; Le Formal, F.; Robert, R.; Weidenkaff, A.; Tucek, J.; Frydrych, J.;
Grätzel, M. J. Am. Chem. Soc. 2010, 132, 7436-7444.
23. Hisatomi, T.; Le Formal, F.; Cornuz, M.; Brillet, J.; Tetreault, N.; Sivula, K.; Gratzel, M.
Energy Environ. Sci. 2011.
24. Glasscock, J. A.; Barnes, P. R. F.; Plumb, I. C.; Savvides, N. J. Phys. Chem. C ,2007, 111,
16477-16488.
25. ****, We found the addition of light was necessary to obtain reproducible results by fully
oxidizing the surface species.
26. Bisquert, J.; Fabregat-Santiago, F.; Mora-Sero, I.; Garcia-Belmonte, G.; Barea, E. M.;
Palomares, E. Inorg. Chim. Acta, 2008, 361, 684-698.
27. Boschloo, G.; Fitzmaurice, D. J. Phys. Chem. B. 1999, 103, 2228-2231.
28. Trainor, T. P.; Chaka, A. M.; Eng, P. J.; Newville, M.; Waychunas, G. A.; Catalano, J. G.;
Brown Jr, G. E. Surf. Sci. 2004, 573, 204-224.
29. Hellman, A.; Pala, R. G. S. J. Phys. Chem. C, 2011, 115, 12901-12907.
30. Dau, H.; Limberg, C.; Reier, T.; Risch, M.; Roggan, S.; Strasser, P. ChemCatChem, 2010, 2,
724-761.
31. Cook, T. R.; Dogutan, D. K.; Reece, S. Y.; Surendranath, Y.; Teets, T. S.; Nocera, D. G.
Chem. Rev. 2010, 110, 6474-6502.
32. Conway, B. E.; Tilak, B. V. Advances in Catalysis, 1992, 38, 1-147.
33. Rohde, J.-U.; In, J.-H.; Lim, M. H.; Brennessel, W. W.; Bukowski, M. R.; Stubna, A.;
Münck, E.; Nam, W.; Que, L. Science, 2003, 299, 1037-1039.
34. Popescu, D.-L.; Vrabel, M.; Brausam, A.; Madsen, P.; Lente, G.; Fabian, I.; Ryabov, A. D.;
van Eldik, R.; Collins, T. J. Inorg. Chem. 2010, 49, 11439-11448.
151

35. Ellis, W. C.; McDaniel, N. D.; Bernhard, S.; Collins, T. J. J. Am. Chem. Soc. 2010, 132,
10990-10991.
36. Blake, R. L.; Hessevick, R. E. Am. Mineralogist, 1966, 51, 123-129.
37. Eggleston, C. M.; Hochella, M. F. Am. Mineralogist, 1992, 77, 911-922.
38. Eggleston, C. M.; Stack, A. G.; Rosso, K. M.; Higgins, S. R.; Bice, A. M.; Boese, S. W.;
Pribyl, R. D.; Nichols, J. J. Geochim. Cosmochim. Acta, 2003, 67, 985-1000.
39. ****, The number of active sites per unit cell is estimated by assuming one Fe atom in a
given plane per unit cell. Each atom is estimated to have only 1 active site because the formation
of 2 oxo groups on a single Fe atom is energetically unfavorable.
40. Sanchez, H. L.; Steinfink, H.; White, H. S. J. Solid State Chem. 1982, 41, 90-96.
41. Zhong, D. K.; Cornuz, M.; Sivula, K.; Graetzel, M.; Gamelin, D. R. Energy Environ. Sci.
2011, 4, 1759-1764.
42. Cummings, C. Y.; Marken, F.; Peter, L. M.; Tahir, A. A.; Wijayantha, K. G. U. Chem.
Commun. 2012, 48, 2027-2029.
43. Zhong, D. K.; Gamelin, D. R. J. Am. Chem. Soc. 2010, 132, 4202-4207.
44. Zhong, D. K.; Sun, J. W.; Inumaru, H.; Gamelin, D. R. J. Am. Chem. Soc. 2009, 131, 60866087.

152

Chapter 7:
Methanol

Photo-Oxidation

at

Hematite

Electrodes and the Implications for Water
Oxidation

153

7.1 Abstract
Fe2O3

photoanodes

synthesized

by

Atomic

Layer

Deposition

(ALD)

were

photoelectrochemically characterized in the presence of CH3OH in H2O. Systematic
photoelectrochemical characterization along with O2 evolution measurements, were carried out
in order to better understand the mechanisms of both H2O and methanol oxidation at hematite
electrodes. It was found that while methanol oxidation is preferred in methanol-water mixtures,
hole transfer to methanol is not unity and therefore, cannot be used to probe the properties of the
bulk hematite electrode. Photoelectrochemical and impedance spectroscopic measurements show
two different reaction pathways for methanol oxidation at hematite electrodes. In addition, the
oxidation of water and methanol mixtures show that increasing the methanol concentration
decreases the surface state capacitance and oxygen evolution proportionally, further confirming
that the surface states are an integral part of the water oxidation mechanism.

154

7.2 Introduction
Methanol (CH3OH) oxidation in the presence of H2O is a well-studied process,
1

particularly for its relevance for the development of methanol based fuel cells. Methanol has
also attracted considerable attention for fundamental studies oriented to the photocatalytic
2

elimination of organic compounds in polluted water and air. The oxidation of CH3OH results in
several different intermediates and products, including formate and formaldehye under varied
conditions which are then able to be qualitatively analyzed.

3-6

The identity of methanol

oxidation intermediates and products thus provide deep insight into the mechanism of oxidiation
at a metal oxide electrode. The oxidation of this small molecule is also expected to contain
analogies with the mechanism of water oxidation due to the similar chemical structure and bond
angles of the molecules. The present study aims to further understand the mechanisms of hole
transfer for water and methanol photo-oxidation with hematite electrodes.
7.3 Experimental
Hematite electrodes (~60 nm thick) were measured in contact with aqueous electrolytes
containing a 0.1 M phosphate buffer (pH 6.9) with 0.2 M KCl as supporting electrolyte. This
aqueous electrolyte contained varying amounts CH3OH. Electrolytes containing only CH3OH as
the solvent also contained 0.2 M tetrabutylammonia chloride as supporting electrolyte. The pure
CH3OH solution was prepared and sealed in an electrochemical cell under nitrogen in a glove
box to minimize the H2O content. Experiments were also performed with pH 13.6 (1M KOH)
solutions and with different electrodes which all showed the same trends.

155

Photoelectrochemical and impedance measurements made in aqueous electrolytes were
made using a homemade saturated Ag/AgCl reference electrode.

Those made in the pure

CH3OH electrolyte were made using a commercial Ag/AgCl electrode (ESA 66-EE009 “no
leak”). Both electrodes were calibrated to a saturated calomel electrode (Koslow Scientic) and
tested regularly in a Fe(CN)6

3-/4-

solution and to ensure consistency. A platinum mesh electrode

was used as the counter electrode. All measurements were made under 1 sun illumination ( AM
-2

1.5 , 100 mW cm ). Cyclic voltammetry (CV) measurements were carried out immediately after
the electrode was held at a positive potential ( of 1.6 V vs Ag/AgCl for aqueous electrolytes and
1.8 V vs Ag/AgCl for the CH3OH electrolyte) and 1 sun illumination for 60 seconds. This is
expected to fully oxidize the surface intermediates. Immediately before scanning the CV, the
light is turned off and the CV measurement is made in the dark. Steady state current vs. voltage
,J-V, curves were measured at 20 mV/s. Chopped light J-V curves were measured at a scan rate
of 75 mV/s and the light was chopped every 266 ms. The current was sampled at a rate of 500
Hz. In order to deconvolute the chopped light J-V curve from potential, constant potential
transients were measured after turning the light on (where the anodic current is measured) and
turning the light off (where the cathodic current is measured). These transients are labeled the
anodic and cathodic transients respectively. The current was sampled at a rate of 1000 Hz.
Impedance measurements were measured at different applied biases using a perturbation
amplitude of 10 mV. The frequency range was 10 kHz-10 mHz. Data were fit using Zview
software (Scribner Associates). The light source was a 450 W Xe arc lamp (Horiba Jobin Yvon).
An AM 1.5 solar filter (Sciencetech Inc.) was used to simulate sunlight at 100 mW cm-2 (1
sun). The pH was determined with an Accumet pH meter.
156

Oxygen measurements were made with an Ocean Optics spectrophotometer using a
FOSPOR fluorescent patch. The electrode was illuminated at 2 suns to produce more oxygen
and to reduce noise. The theoretical amount of oxygen produced was calculated by assuming
that 4 holes are required to produce 1 molecule of oxygen.
7.4 Results
7.4.1 Steady State Oxygen Evolution
Figure 7-1a shows the current density, J, vs. applied voltage, V, curves of Fe2O3 in
contact with electrolytes with varying concentrations of CH3OH. When CH3OH is added to
H2O, a slight cathodic shift of the photocurrent onset takes place (see Figure 7-1b), in good
7

agreement with observations made by other researchers. For pure methanol, there is a slow
photocurrent onset observed at a similar potential (0.7 V vs Ag/AgCl), followed by a steeper
current onset which mimics the shape measured with water oxidation at hematite electrodes.
Surprisingly, there was an anodic shift of approximately 300 mV which was needed to sustain a
-2

given photocurrent density of 200A cm compared to water based electrolytes. This result is in
contrast to literature results of hematite photoelectrodes in contact with good hole scavengers,
3-,4-

such as [Fe(CN)6]

8,9

and H2O2.

A fluorescence probe was used to measure the amount of O2 produced, which can be
used to evaluate the faradaic efficiency of H2O oxidation. O2 concentration was measured at an
applied potential of 1.4 V vs Ag/AgCl for H2O and CH3OH-H2O mixture electrolytes and 1.6 V
vs Ag/AgCl for the CH3OH electrolyte such that the currents were approximately equal. This

157

also accounts for the difference in the EFB potentials between aqueous and methanol based
electrolytes, vide infra. In the case of the aqueous electrolyte, the faradaic efficiency due to O2
production is approximately unity as shown in Figure 7-1c and Table 7-1. As increasing
amounts of CH3OH added to H2O, the faradaic efficiency due to O2 generation decreased where
no O2 generation could be detected in pure CH3OH. This is expected since the oxidation of
10-12

CH3OH should produce formaldehyde, not O2.

Interestingly, however, a significant amount

of O2 was generated even with large concentrations of CH3OH. A summary of the amount of O2
measured for different CH3OH concentrations at applied potentials of 0.75 and 1.4 V vs Ag/Agl
can be seen in Table 1. Relative rates, vR, for CH3OH oxidation compared to water oxidation

 1  FE  FE 
was calculated by the equation vR  

 CH OH    H O  

3

 2 

1

, where FE is the faradaic

efficiency for O2 generation, and can also be found in Table 7-1. This equation describes the
ratio of the FE of methanol oxidation over the FE of water oxidation, both of which are
normalized for their concentration.

Despite an obvious preference for CH3OH oxidation

compared to H2O oxidation based on the relative rates shown in Table 7-1, the total hole transfer
rate (current) is not significantly altered by adding CH3OH to H2O as shown in figure 7-1a.
Poor current onset potentials still exist with CH3OH in solution, showing that it is not acting as a
fast and model hole collector.

158

Figure 7-1. (a) J-V curves of a hematite electrode under 1 sun illumination in contact with H2O
(red), 0.2M MeOH (orange), 2M (yellow), 5M (green), 10M (teal), and pure MeOH (blue) vs
Ag/AgCl. (b) Same data as (a), on a semi-logarithmic plot. (c) Theoretical [O2] produced under 2
suns (lines) and measured [O2] (shapes) for a hematite electrode in contact with H2O (solid line
and circles), 0.2M MeOH (dotted line and downward triangles), 2M (short dashed line and
squares), 5M (green double dotted dashed line and diamonds), 10M (long dashed line and
upward triangles), and pure MeOH (single dotted line and hexagons). Measured at 1.4 and 1.6 V
vs Ag/AgCl for aqueous and the MeOH electrolyte respectively. (d) Same as (c) but for 0.75 V
vs Ag/AgCl (CH3OH not shown).
159

Table 7-1. Summary of faradaic efficiency due to O2 generation for varying concentrations of
CH3OH in H2O
0.75 V vs Ag/AgCl
[O2] / moles L

-1

Relative
Rate*

11.11

O2 Faradaic
Efficiency
99.42%

7.90

5.41

68.48%

133.48

51.0

12.90

4.01

31.09%

56.52

5.0

44.0

10.77

1.89

17.55%

41.35

10.0

33.0

7.29

0.59

8.06%

37.65

O2 Faradaic

Relative

Efficiency

Rate*

[CH3OH]/
M
0.0

[H2O] / M

Theoretical

Measured

55.0

11.17

0.2

58.0

2.0

1.4 V vs Ag/AgCl
[O2] / moles L
[CH3OH]/

[H2O] / M

Theoretical

-1

Measured

M
0.0

55.0

50.13

49.81

99.36%

0.2

58.0

49.20

36.00

73.17%

106.33

2.0

51.0

55.15

28.35

51.40%

24.11

5.0

44.0

53.15

22.80

42.90%

11.71

10.0

33.0

49.86

15.94

31.97%

7.02

*Ratio of CH3OH to H2O oxidation

160

Mott Schottky plots were generated for pure H2O, CH3OH and 5M CH3OH electrolytes
by fitting impedance spectra measured under dark conditions to a Randle’s circuit. Since the
electrode, and therefore dopant density, is the same for all electrolytes, the slope of the Mott
Schottky is expected to be independent of the electrolyte. This is indeed the case, apparent from
inspection of figure 7-2a.

Therefore, to fit the Mott Schottky plots, the slopes of all 3
18

electrolytes were fit globally. The extracted dopant density, ND, was calculated to be 5.3×10
-3

cm , in good correspondence with previous results.

13,14

The flat band potential was also

calculated from the Mott Schottky plots to be 0.137, 0.143 and 0.330 V vs. Ag/AgCl for H2O,
5M CH3OH and pure CH3OH respectively. The measured EFB for the aqueous electrolytes are
nearly equal. In pure CH3OH however, the EFB is moved ~200 mV positive. Part of the
difference in J-V curves can therefore be attributed to a shift in the band energies. In order to
carry out an adequate comparison between the different electrolytes, potentials were normalized
to their respective EFB’s. Normalized J-V curves are shown in Figure 7-2b. At high positive
applied potentials, the current achieved in all electrolytes is approximately equal. We have
suggested previously that in these thin films, at high applied potential, all of the holes that reach
the surface oxidize water. Thus, adding a fast hole collector (e.g. H2O2, CH3OH, Fe(CN)6

3-/4-

)

should not increase the current at a given positive Fermi level. Even when correcting for the
change in the EFB, the steep current onset of pure CH3OH oxidation occurs at more positive
potentials compared to H2O oxidation suggesting that photoelectrochemical differences cannot
be solely attributed to differences in the position of the bands of Fe2O3 in different media.

161

Figure 7-2. (a) Mott Schottky plot of Cbulk measured in the dark for H2O (red circles), 5M
CH3OH (green diamonds) and CH3OH electrolytes (blue hexagons). (b)

J-V curves of a

hematite electrode under 1 sun illumination in contact with H2O (red), 5M CH3OH (green) and
CH3OH (blue) plotted vs EFB.

7.3.2 Electrochemical Impedance Spectroscopy
In order to get more detailed mechanistic information of water and methanol oxidation,
impedance spectroscopy (IS) was employed under illuminated conditions. Aqueous electrolytes
produced Nyquist plots with two clearly distinguishable semicircles near the current onset
potential which are shown in Figure 7-3a.

A physical model which has been previously

developed for the interpretation of charge transfer from hematite electrodes to water can be seen
13,14

in the form of an equivalent circuit, EC, shown in Figure 7-4a.

This general model includes

the presence of a surface state able to capture conduction band electrons and valence band holes;
the trapping-detrapping kinetics are related to Rtrap, and the occupancy of the surface state

162

produces a capacitance, Css. Charge transfer takes place through this surface state with a charge
transfer resistance, Rct,ss. Additionally, a bulk capacitance, Cbulk, accounts for the space charge
capacitance. This EC was used to fit IS data gathered in H2O and 5M CH3OH electrolytes. For
the pure H2O electrolyte, a clear peak in Css develops with voltage which is coincident with the
current onset potential (Figure 7-5a).

This peak highlights the requirement for charge

accumulation at the surface for water oxidation to occur, and has been pointed out by other
15

researchers.

When CH3OH is added to the aqueous solution, a Css peak is also observed, at

the same potential vs EFB as shown in Figure 7-5a. The magnitude of this peak is lower,
however, reflecting that fewer holes are accumulated at the surface. This result is in good
agreement with the O2 evolution measurements displayed in Figures 7-1c and 7-1d, which
indicated that both oxidation of water and methanol lead to the obtained photocurrent. A
corresponding decrease in Css and O2 generation with the addition of CH3OH further shows that
water oxidation occurs through a surface state intermediate. Despite methanol reducing the
measured capacitance of surface states due to water oxidation, two capacitive features are still
observed in impedance spectra measured in the pure methanol electrolyte (Figure 7-3b). Since it
appears that charge transfer to methanol occurs through a different pathway than that of water,
we invoke another possible EC to describe pure methanol oxidation at hematite electrodes shown
in figure 7-4b. This model was chosen due to its corroboration with photoelectrochemical
experiments discussed below. As in the EC used to describe H2O oxidation, this EC still
contains a surface state capacitance, Css and a resistance of transfer to those surface states, Rtrap,

163

based on the observation of 2 clear capacitances in the Nyquist plot. However, instead of charge
transfer through those surface sates, this EC contains a separate charge transfer resistance which
is not directly associated with those surface states, Rct. This model could imply direct charge
transfer from the valence band of the hematite to methanol through an outersphere electron
transfer mechanism. However, a more likely possibility is that a surface state or surface
intermediate is generated, such as a surface adsorbed methoxy species (Fe-O-CH3) which is the
10

analogous adsorbed species responsible for charge transfer for photogenerated holes on TiO2.

This mechanism could be interpreted as a single charge transfer resistance if the steady state
concentration is very low, resulting in a very low or immeasurable capacitance. By using this
model, the obtained surface state capacitance shows a double peak feature, with a significantly
lower intensity compared to water based electrolytes (Figure 7-5a).

Figure 7-3. (a) Nyquist plots of impedance data gathered under 1 sun illumination for a H2O
(red circles) and 5M CH3OH in H2O (green diamonds) electrolytes at 0.66 V vs Ag/AgCl. (b)
Nyquist plot in CH3OH measured at 1.06 V vs Ag/AgCl.

164

a)

b)

c)

Figure 7-4. (a) Equivalent circuit used to analyze impedance data measured for hematite
electrodes in contact with electrolytes containing H2O (b) Equivalent circuit used to analyze
impedance data measured for hematite electrodes in contact with pure CH3OH electrolytes. (c)
Randles circuit.

The values of charge transfer resistance, Rct, are shown in Figure 7-5b for the different
electrolytes. At low bias (0 - 0.4V vs EFB), the addition of CH3OH to H2O reduces the charge
transfer resistance. This is because we are measuring the charge transfer resistance from the
surface states (due to water oxidation) which is reduced with the addition of CH3OH, as shown
by O2 evolution measurements and Css. In other words, for a constant potential, the resistance of
charge transfer to water through water oxidation intermediates is reduced because the current due
to water oxidation is reduced. Compared to aqueous electrolytes, Rct for methanol oxidation is
higher positive of 0.4 V vs EFB, consistent with the lower photocurrents measured at more
anodic bias. The values of Rtrap are compiled in Figure 7-5c, and are identical for both water
based electrolytes and higher for pure methanol between 0 V vs EFB and 0.7 V vs EFB.

165

Confirmation of the fits can be gained by plotting the total resistance from impedance
(Rtot=Rs+Rtrap+Rct,ss for aqueous electrolytes and Rtot=Rs+Rct,bulk for the CH3OH electrolyte)
and comparing that to the resistance calculated from the J-V curves (Rtot=dV/dJ). This plot can
be seen in figure 7-5d. All electrolytes show Rtot from J-V curves and Rtot from IS are in good
general agreement indicating that the main resistances which control the J-V are represented in
the impedance spectra. It is interesting to note that pure methanol oxidation involves two
different Rtot dips, around 0.4 V vs EFB and 0.8 V vs EFB, related to the two small trap Css
peaks shown in Figure 7-5a. This suggests that two different oxidation mechanisms exist; one
between 0.2-0.6 V vs EFB where the slope of the J-V curve is low, and one above 0.6 V vs EFB,
where the J-V curve takes a similar slope as that of aqueous solutions (see Figure 7-2b).

166

Figure 7-5. (a) Css (b)Rct, (c)Rtrap and (d) Rtot vs EFB extracted from IS analysis under
2

illumination (100 mW·cm ) for a hematite electrode in contact with H2O (red circles) 5M
CH3OH in H2O (green diamonds) and CH3OH (blue hexagons).

7.4.3 Cyclic Voltammetry
Cyclic voltammetry experiments were carried out on Fe2O3 films to measure surface
13

states as previously reported.

First, high positive bias (1.6 V vs Ag/AgCl for aqueous

electrolytes and 1.8 V vs Ag/AgCl for methanol electrolyte) was applied under illumination to
167

fill the surface states with holes. The light was then turned off and a CV was immediately
scanned several times (Appendix Figure A7-1).

Cathodic peaks were observed for all

electrolytes on the first scan at potentials similar to the potentials where a Css is observed from
IS measurements (Figure 7-6). These peaks are observed at the same energy and with similar
relative intensity to those observed in the peaks of Css in Figure 7-5a. Additional figures
including data measured in all CH3OH electrolytes can be seen in figure 7-6b which confirms
the trend of decreasing cathodic peak currents with the addition of CH3OH.

On the second

scan, these peaks disappear because the surface states are not reoxidized on the timescale of this
experiment (Appendix, Figure A7-1). The evolution of this peak with scan rate (Appendix,
Figure A7-2) also confirms its capacitive nature. Upon zooming in on the cyclic voltammetry
curve for methanol (Figure 7-6b), two small peaks are evident, which is consistent with the Css
measured by IS using the model in figure 7-4b. It should be noted that these two peaks were not
observed in the IS interpretation using the model in figure 7-4a.

Thus these two peaks

corroborate the selection of the equivalent circuit for interpreting CH3OH oxidation at the
hematite surface.

168

Figure 7-6. (a) Comparison of the first scan of the CV measured at 200 mV/s in H2O (red solid
line), 5M MeOH in H2O (green double dot dashed line) and MeOH (blue single dot dashed line).
(b) A magnified image of the MeOH curve which shows 2 small peaks. (c) First scan of the CVs
measured in the dark for a hematite electrode in contact with H2O (red), 0.2 M (orange), 2 M
(yellow), 5M (green), and 10 M MeOH (teal) measured at 200 mV/s). (d) Css values fit from
model shown in figure 7-3a (orange triangles) and Css values fit from model shown in figure 73b (yellow squares) as well as capacitance calculated of the first scan of CV experiments in
MeOH solution measured at 1000 mV/s . C=J/scan rate.
169

The capacitive nature of the surface states of both H2O and CH3OH surface states was
examined by transient light experiments. Chopped light J-V curves were obtained and compared
to those with constant illumination (Appendix, Figure A7-3). In all tested electrolytes, the
presence of anodic and cathodic transients indicated the presence of surface states.

8,16,17

These

spikes were analyzed quantitatively by switching the light on (anodic) and off (cathodic) at a
constant potential while measuring the current as a function of time. Examples of current
transients can be seen in Figure 7-7a and in Appendix Figures A7-4 through A7-7. Integration of
the cathodic current provides a quantitative measurement of the charge stored in the surface
states, which is Figure 7-7b. The charge measured in the current transients was compared to the
Css measured by impedance and a good agreement in the trend was observed, confirming that the
current transients are primarily a result of charging and discharging of surface states.
Single exponential lifetimes of the cathodic transients were calculated and can be seen in
Figure 7-7c. The lifetimes are approximately equal for H2O and 5M CH3OH electrolytes
throughout the potential range with the exception of potentials greater than ~0.6 V vs EFB where
the mixture has lower lifetimes. These decays are attributed to electrons from the conduction
band reducing the oxidized surface states. Since the dominant surface state is related to water
oxidation, it is not surprising that these measured lifetimes are constant for the two aqueous
electrolytes. We note that this result disagrees with Cowan et. al. who showed a decrease in
lifetime for a 10% CH3OH in H2O solution compared to a H2O solution under no applied bias.

18

It is possible, however, the techniques of transient electrochemistry employed herein and
transient absorption spectroscopy employed by Cowan et. al. are measuring different species. In

170

the CH3OH electrolyte, the lifetimes of these transients are approximately an order of magnitude
lower than those measured in aqueous electrolytes at potentials where photocurrent is measured.
This is attributed to an increased rate of recombination to surface CH3OH oxidation
intermediates compared to surface H2O oxidation intermediates.

171

11

Figure 7-7. (a) Cathodic transients in H2O (red solid line), 5M CH3OH in H2O (green double
dot dashed line) and CH3OH (blue single dot dashed line) measured within 5 mV of 0.71 V vs
EFB.

(b) The calculated charge passed measured by integrating cathodic transients (filled

symbols, solid lines) along with Ctrap measured by IS (hollow symbols, dashed lines) for a
hematite electrode in contact with H2O (red circles) 5M CH3OH in H2O (green diamonds) and
CH3OH (blue hexagons). (c) Lifetimes of transients fit by a single exponential decay for H2O
(red circles), 5M CH3OH (green diamonds), and CH3OH (blue hexagons).

172

7.5 Discussion
From the double peak feature observed for Css from EIS analysis of pure methanol
oxidation, correlated to respective dips in the charge transfer resistance (Rct) and total resistance
(Rtot), it appears that methanol oxidation takes place by two distinct processes, as previously
mentioned. While there is limited information studying the photooxidation of methanol on
hematite electrodes, there is a wealth of information of the photooxidation of methanol on TiO2
which is often considered as a model system for metal oxide surfaces.

3,19

There are two general

mechanisms typically discussed for methanol oxidation at TiO2 photoelectrodes: (1) a direct
charge transfer mechanism where an adsorbed methanol or methoxy species is directly oxidized,
or (2) an indirect charge transfer mechanism where a surface hydroxyl is oxidized which
subsequently oxidizes methanol. The existence of these two mechanisms are said to be
responsible for both formate and formaldehyde intermediates identified during methanol
oxidation. There is, however, discrepancy when assigning the mechanism responsible for a
particular intermediate where several researchers discuss formate and formaldehyde as a result of
direct and indirect mechanism respectively,

4,20

whereas others propose the opposite

6

relationship. When considering a direct mechanism, there has also been significant debate over
whether the relevant photogenerated hole transfer occurs through a surface adsorbed molecular
methanol or through a dissociatively adsorbed methoxy group.

3,19

Both molecular and

dissociatively absorbed methoxy groups are commonly observed on TiO2 which agrees with
DFT calculations which show molecular and dissociated states on a Ti
173

4+

site are nearly

21-23

isoenergetic.

(110) rutile TiO2

More recently, studies employing temperature programmed desorption on

10

12

and FTIR on anatase nanoparticle TiO2

have found that the most relevant

hole transfer is through a dissociatively adsorbed methoxy species. The observation of different
possible reactants (adsorbed methanol, methoxy, hydroxyl species) combined with possible
mechanisms (direct, indirect) and intermediates (formate, formaldehyde), show that methanol
oxidation may occur via different reaction pathways on TiO2. It is not surprising then that the JV curve, surface state capacitance and charge transfer resistance measured for water oxidation at
hematite photoelectrodes shows two oxidation features at different potentials. The molecular
details of this mechanism are currently being investigated in our lab.
7.6 Conclusions
Fe2O3 photoanodes synthesized by atomic layer deposition were electrochemically
characterized with H2O, CH3OH and H2O- CH3OH mixture electrolytes. Compared to aqueous
electrolytes, the photo-oxidation of methanol required a more anodic bias to sustain similar
photocurrent values. The corresponding decrease in the O2 generation and Css with the addition
of CH3OH further confirms that the surface states are actively participating in the hole transfer
13,14

mechanism to H2O, as suggested in previous studies.

This reduction in surface states has

been measured by IS, CV and transient measurements. IS analysis suggests that methanol
oxidation occurs through a single resistor which does not involve the build up of a capacitance
through a surface state. It should be noted, however, that a simple outersphere hole transfer to
CH3OH cannot be assigned with certainty.

An innersphere mechanism is possible if the

174

intermediate had a very low and capacitance, resulting in an equivalent circuit parameter
appearing as a simple charge transfer resistance. More work is necessary to determine the nature
of this charge transfer. A significantly lower surface state capacitance with two distinct peaks
associated with corresponding charge transfer dips was also identified for methanol oxidation,
suggesting two different reaction mechanisms depending on the applied bias.
7.7 Acknowledgements
I would like to thank Sixto Gimenez, Francisco Fabregat-Santiago, Juan Bisquert and Thomas
Hamann for contributing to this work.

175

APPENDIX

176

Figure A7-1. The first (solid blue line) and second (dashed purple line) scan of a CV measured
in the dark of a hematite electrode in contact with (a) H2O (b) 5M CH3OH in H2O (c) and
CH3OH

177

Figure A7-2. Calculated capacitance of the first scan of the CVs measured in the dark for a
hematite electrode in contact with (a) H2O (b) 5M CH3OH in H2O (c) and CH3OH. (d)
Magnified image of (c). The capacitance was calculated by C=J/scan rate. Scan rates were 50
mV/s (red solid line), 100 mV/s (orange long line), 200 mV/s (yellow short line), 500 mV s
(green shorter line) and 1000 mV/s (blue dotted line)

178

Figure A7-3. Chopped light (grey line) and steady state (colored line) J-V measured for a
hematite electrode in contact with (a) H2O (b) 5M CH3OH in H2O and (c) CH3OH.

179

Figure A7-4. (a) cathodic and (b) anodic transients measured for a hematite electrode in contact
with H2O at applied potentials of 0.35 (solid teal line), 0.45 (dotted blue line) and 0.55 (dashed
purple line) V vs Ag/AgCl.

Figure A7-5. (a) cathodic and (b) anodic transients measured for a hematite electrode in contact
with 5M CH3OH in H2O at applied potentials of 0.35 (solid teal line), 0.45 (dotted blue line) and
0.55 (dashed purple line) V vs Ag/AgCl.

180

Figure A7-6. (a) cathodic and (b) anodic transients measured for a hematite electrode in contact
CH3OH at applied potentials of 0.55 (solid teal line), 0.65 (dotted blue line) and 0.75 (dashed
purple line) V vs Ag/AgCl.

Figure A7-7. (a) anodic and (b) cathodic transients in H2O (red solid line), CH3OH in H2O
(orange dotted) and MeOH (yellow dashed line) measured within 5 mV of 0.56 V vs EFB.

181

REFERENCES

182

REFERENCES

1. Iwasita, T. Electrochim. Acta, 2002, 47, 3663-3674.
2. Hoffmann, M. R.; Martin, S. T.; Choi, W.; Bahnemann, D. W. Chem. Rev. 1995, 95, 69-96.
3. Henderson, M. A. Surf. Sci. Rep. 2011, 66, 185-297.
4. Wang, C.-y.; Rabani, J.; Bahnemann, D. W.; Dohrmann, J. K. J. Photochem. Photobiol., A
2002, 148, 169-176.
5. Zhou, C.; Ren, Z.; Tan, S.; Ma, Z.; Mao, X.; Dai, D.; Fan, H.; Yang, X.; LaRue, J.; Cooper,
R.; Wodtke, A. M.; Wang, Z.; Li, Z.; Wang, B.; Yang, J.; Hou, J. Chem. Sci. 2010, 1, 575-580.
6. Wu, W.-C.; Chuang, C.-C.; Lin, J.-L. J. Phys. Chem. B, 2000, 104, 8719-8724.
7. Pendlebury, S. R.; Barroso, M.; Cowan, A. J.; Sivula, K.; Tang, J. W.; Gratzel, M.; Klug, D.;
Durrant, J. R. Chem. Commun. 2011, 47, 716-718.
8. Dotan, H.; Sivula, K.; Gratzel, M.; Rothschild, A.; Warren, S. C. Energy Environ. Sci. 2011,
4, 958-964.
9. Klahr, B. M.; Hamann, T. W. J. Phys. Chem. C, 2011, 115, 8393-8399.
10. Shen, M.; Henderson, M. A. J. Phys. Chem. Lett. 2011, 2, 2707-2710.
11. Villarreal, T. L.; Gomez, R.; Neumann-Spallart, M.; Alonso-Vante, N.; Salvador, P. J. Phys.
Chem. B, 2004, 108, 15172-15181.
12. Panayotov, D. A.; Burrows, S. P.; Morris, J. R. J. Phys. Chem. C, 2012, 116, 6623-6635.
13. Klahr, B.; Gimenez, S.; Fabregat-Santiago, F.; Bisquert, J.; Hamann, T. Energy Environ. Sci.
2012, 5, 7626-7636.
14. Klahr, B.; Gimenez, S.; Fabregat-Santiago, F.; Hamann, T.; Bisquert, J. J. Am. Chem. Soc.
2012, 134, 4294-4302.
15. Cummings, C. Y.; Marken, F.; Peter, L. M.; Tahir, A. A.; Wijayantha, K. G. U. Chem.
Commun. 2012, 48, 2027-2029.
16. McDonald, K. J.; Choi, K. S. Chem. Mater. 2011, 23, 1686-1693.
17. Le Formal, F.; Tetreault, N.; Cornuz, M.; Moehl, T.; Gratzel, M.; Sivula, K. Chem. Sci.
2011, 2, 737-743.

183

.
owan . . Barnett . . endleb r
. . Barroso
.
Durrant, J. R.; Klug, D. R. J. Am. Chem. Soc. 2011, 133, 10134-10140.

i

la

.

r t el

.

.

errera-Melián, J. A.;

19. Diebold, U. Surf. Sci. Rep. 2003, 48, 53-229.
0. ra a . o a- odr e . . abo . . i.
Pérez-Peña, J. Appl. Catal., B 2004, 53, 221-232.

on le - a

21. Guo, Q.; Xu, C.; Ren, Z.; Yang, W.; Ma, Z.; Dai, D.; Fan, H.; Minton, T. K.; Yang, X. J.
Am. Chem. Soc. 2012, 134, 13366-13373.
22. de Armas, R. S.; Oviedo, J.; San Miguel, M. A.; Sanz, J. F. J. Phys. Chem. C, 2007, 111,
10023-10028.
23. Oviedo, J.
nche -de-Armas, R. o.; San Miguel, M. A. n.; Sanz, J. F. J. Phys. Chem. C
2008, 112, 17737-17740.

184

Chapter 8:
Photoelectrochemical

and

Impedance

Spectroscopic Investigation of Water Oxidation
with “Co-Pi” coated Hematite Electrodes

Adapted with permission from:
Photoelectrochemical and Impedance Spectroscopic Investigation of Water Oxidation with “CoPi”-Coated Hematite Electrodes, Benjamin M. Klahr, Sixto Gimenez, Francisco FabregatSantiago, Juan Bisquert and Thomas W. Hamann. Journal of the American Chemical Society,
2012, 134, 16693-16700. Copyright 2012 American Chemical Society.

185

8.1 Abstract
Uniform thin films of hematite (α-Fe2O3) deposited by atomic layer deposition (ALD)
coated with varying amounts of the cobalt phosphate catalyst, “Co-Pi,” were investigated with
steady state and transient photoelectrochemical measurements as well as impedance
spectroscopy. Systematic studies as a function of Co-Pi thickness were performed in order to
clarify the mechanism by which Co-Pi enhances the water splitting performance of hematite
electrodes. It was found that under illumination, the Co-Pi catalyst can efficiently collect and
store photogenerated holes from the hematite electrode. This improved charge separation reduces
surface state recombination which results in increased water oxidation efficiency. It was also
found that thicker Co-Pi films produced increased water oxidation efficiencies which are
attributed to a combination of superior charge separation and increased surface area of the porous
film. These combined results provide important new understanding of the enhancement and
limitations of the Co-Pi catalyst coupled with semiconductor electrodes for water splitting
applications.

186

8.2 Introduction
The previous three chapters discussed how a large applied potential is required to achieve
efficient photoelectrochemical (PEC) water oxidation at hematite electrodes. This large
photocurrent onset potential is generally attributed to slow water oxidation kinetics at the
hematite surface which competes with surface state recombination.

1-5

In order to reduce the

required applied potential, various catalysts have been added to the hematite surface including
6-12

IrO2, cobalt ions and the cobalt phosphate catalyst, “Co-Pi”.

Co-Pi has specifically gained a

lot of recent attention because it uses earth-abundant elements, shows effective water oxidation
characteristics and is stable over time due to its “self-healing” mechanism.

13-15

Consequently, it

has also been applied to many potential photoanodes including ZnO, BiVO4, Si and Fe2O3, and
has shown improvements in both current onset potential and photocurrent density.

7-11,16-22

The

reasons for this improvement, however, are not yet fully understood. Thus far, the increased
7

performance has been attributed to accelerating the oxygen evolution kinetics, increasing band
11,23

bending,

facilitating charge separation

7,8

10

and reducing surface state recombination.

Despite the cathodic shift frequently measured for Co-Pi coated electrodes, Zhong and Gamelin
9

described evidence of a kinetic bottleneck on Co-Pi coated hematite electrodes. Thorough
knowledge of the physical origin of both the improvement and limitation of water oxidation with
Co-Pi coated hematite electrodes is essential to guide further advances in catalyst and
semiconductor design and integration into PEC water splitting systems.
In this work, we employ photoelectrochemical measurements and impedance
spectroscopy to investigate the effect of the Co-Pi catalyst on thin film hematite electrodes. The
187

24-26

thin film hematite electrodes were prepared by atomic layer deposition (ALD),
9,10,21

subsequently coated with Co-Pi films by photoelectrodeposition.

and were

These thin hematite films

have been shown to be a good model system for studying the limitations of water oxidation at the
hematite surface, thus allowing us to separate the effect of the catalyst from the bare
27,28

electrode.

Also, the planar geometry of these films allowed us to perform a controlled

thickness dependence study where increasing the amount of Co-Pi deposited uniformly increases
the Co-Pi thickness in one dimension. The Co-Pi coated hematite films were also analyzed by
impedance spectroscopy and PEC experiments as a function of Co-Pi thickness to elucidate the
factors controlling the enhanced performance of the Co-Pi coated hematite electrode. These
combined results provide new insight and allow for further development of a thorough
mechanistic picture of this important system.
8.3 Experimental
Thin films of hematite were deposited on fluorine-doped tin oxide (FTO) coated glass
-2

substrates (Hartford Glass, 12 Ω cm ) by ALD (Savannah 100, Cambridge Nanotech Inc.) using
ferrocene as the metal precursor and ozone as the oxidation source. ALD is a process which
employs alternating metal precursor and oxidation pulses, each separated by a nitrogen purge
such that chemistry only occurs in a self-limiting fashion where the previous precursor has
adsorbed. The metal precursor was heated to 70°C and pulsed for 20 seconds. After purging the
oxidation pulse was performed. The oxidation pulse consisted of a 0.015 second pulse of H2O,
immediately followed by a 1 second pulse of ozone (~4.5% by weight O3 in ultra-high purity O2
produced by Yanco Industries ozone generator), followed by a 5 second purge time. This cycle

188

was performed 10 times to create 1 oxidation macrocycle. Integrating water with O3 has been
found to be necessary for uniform deposition of hematite in our system. Films were prepared by
1000 ALD cycles and measured to be ~60 nm by absorption measurements (Perkin Elmer,
Lambda 35 with a Labsphere integrating sphere) corrected for reflection as described previously,
26

as well as ellipsometric measurements (Horiba Jobin Yvon, Smart-SE).
characterized by Raman Spectroscopy and XRD of these films previously.

26

Films were

Hematite electrodes

2

were masked with a 60 μm Surlyn film (Solaronix) with a 0.28 cm hole to define the active area
and to prevent scratching of the thin films. Surlyn films were adhered to the electrodes by
heating to 120 °C. The protected hematite films were clamped to a custom made glass
electrochemical cell. A homemade saturated Ag/AgCl electrode was used as a reference and was
frequently calibrated to a commercial saturated calomel reference electrode (Koslow Scientific).
Potentials vs Ag/AgCl were converted to the reversible hydrogen electrode (RHE) scale by the
equation ERHE  E Ag / AgCl  0.197V  0.059V  pH . A high surface area platinum mesh was
used as the counter electrode.
10

Co-Pi catalyst films were deposited onto hematite by photo-assisted electrodeposition.

Hematite electrodes were immersed in a solution containing 0.5 mM Co(NO3)2·6H2O in a 0.1 M
phosphate buffer (pH 6.9). A bias of 0.9 V vs RHE was applied under illumination. The
thickness of the Co-Pi layer was controlled by varying the amount of charge allowed to pass
during the deposition. For the electrodes reported herein, 5 thicknesses were prepared by
allowing 1, 2, 15, 45 and 90 mC cm

-2

to pass. After the catalyst was deposited, the electrodes

were lightly rinsed with DI water to remove any excess cobalt ions. Co-Pi coated hematite
189

electrodes

were

prepared

separately

for

SEM

examination.

After

Co-Pi

was

photoelectrodeposited, electrodes were allowed to air dry. SEM images were taken on an Auriga
CrossBeam FIB-SEM (Carl Zeiss Microscopy).
The water oxidation properties of the catalyst coated hematite films were then examined
in contact with an aqueous solution buffered at pH 6.9 using a 0.1 M phosophate buffer
containing 200 mM KCl as a supporting electrolyte. The pH was determined with Fisher
Scientific Accumet pH meter. Impedance spectroscopic and photoelectrochemical measurements
were made with an Eco Chemie Autolab potentiostat coupled with Nova electrochemical
software. Impedance data were gathered using a 10 mV amplitude perturbation of between
10,000 and 0.01 Hz. Data were fit using Zview software (Scribner Associates). The light source
was a 450 W Xe arc lamp. An AM 1.5 solar filter (Sciencetech Inc.) was used to simulate
-2

sunlight at 100 mW cm . All photoelectrochemical measurements were performed by shining
light from the substrate-electrode (SE) interface which avoids competitive light absorption of the
Co-Pi. Light chopping J-V curves were measured at a rate of 75 mV/s. The light was chopped
using a computer controlled ThorLabs solenoid shutter which was set to activate every 266 ms
such that the light was turned on or off every 20 mV. Steady state J-V curves were measured at a
scan rate of 5 mV/s.
Oxygen was detected by using an Ocean Optics spectrometer which probed the
fluorescent decay of the FOSPOR patch. The FOSPOR patch was placed in solution which filled
an airtight cell. The cell was filled so that very little headspace existed. The solution was stirred
vigorously so that the O2 measurement would be as close to real time as possible. Measurements
were made under 4 sun illumination at 1.25 V vs RHE to increase oxygen production and reduce
noise. In the calculation of the faradaic efficiency, the assumption was made that no oxygen
190

diffused into the very small headspace during the timescale of these experiments.
While these experiments were performed multiple times with many electrodes, only 3
different, but nominally identical electrodes were used to collect the photoelectrochemical data
shown here: one for current transient and J-V curves, one for EIS measurements and one for O2
measurements. The use of one electrode for each experiment allowed us to control for any small
differences between bare electrodes. For experiments with varying amounts of Co-Pi, the Co-Pi
was removed by applying a potential of 0.55 V vs RHE under dark conditions which slowly
removed the Co-Pi film. After the Co-Pi was removed, the electrodes were examined to ensure
that the performance was consistent with a bare electrode before applying Co-Pi for the next
experiment.
8.4 Results
Thin film hematite electrodes were coated with varying thickness of the Co-Pi catalyst
via photoelectrodeposition.

9,10,21

At the potential applied during the deposition (0.9 V vs RHE)

steady state water oxidation does not occur with or without the Co-Pi catalyst (vide infra), thus
any charge passed is assumed to be due to deposition of the catalyst. Film thicknesses can then
be estimated by assuming that a single electron passed deposits one cobalt atom with
3 13

surrounding ligands which occupies a volume of approximately 125 Å .
Co-Pi layers deposited by 1, 2, 15, 45 and 90 mC cm

-2

The thicknesses of the

was calculated to be 8, 15, 113, 337 and

675 nm, respectively. Co-Pi coated hematite films were analyzed by SEM to measure thickness
and morphology; all of the films were found to be flat and uniform. Figure 8-1 shows a hematite
film coated with Co-Pi by allowing 90 mC cm

-2

to pass. This film was allowed to air dry

immediately after the deposition. Assuming a 90 degree plane of the edge of the Co-Pi with
191

respect to the underlying hematite, and a viewing angle of 45 degrees, the thickness of this film
is estimated to be ~425 nm. This thickness is largely the same throughout the film and on other
films prepared by depositing 90 mC cm

-2

Co-Pi, which is considerably less than the calculated

thickness of 675 nm. Therefore, either the size of the cobalt cluster represents an overestimation
or not all the charge passed is productive at depositing the film. In order to avoid ambiguity in
the discussion of the Co-Pi films, the thicknesses will be referred to by the amount of charge
-2

passed (mC cm ) in the deposition, which was found to be roughly proportional to the film
thickness. The morphology of the Co-Pi shown in the SEM image is consistent throughout the
film.

-2

Figure 8-1. SEM image of a Co-Pi film deposited by passing 90 mC cm . Sample was air dried
immediately after deposition. The viewing angle is 45 degrees.

8.4.1 Steady State Behavior
-2

Current density, J, vs. applied voltage, V, curves measured under 1 sun (100 mW cm )
illumination for a bare hematite electrode and the same hematite electrode coated with the
192

varying thicknesses of the Co-Pi catalyst can be seen in figure 8-2a. Steady state current
measurements were also performed where the photocurrent was sampled after stabilizing over
several minutes for individual potentials (Figure 8-2b). It was found that the steady state current
densities are equal to the current densities of the cathodic scan of the J-V curves, thus confirming
that the J-V measurements presented herein represent steady-state behavior (Figure 8-2c). The
faradaic efficiency of the J-V curves was also measured by a fluorescent O2 sensor. The O2
concentration was measured at 1.25 V vs RHE which showed an increase in detected O2 with
-2

increasing Co-Pi thickness for Co-Pi amounts up to 15 mC cm ; the O2 detected was consistent
with ~100% faradaic efficiency when considering the number of coulombs that have passed
(Figure 8-3). Clearly, the addition of Co-Pi to the hematite improves the water oxidation
efficiency. Figure 8-2b shows a semi-logarithmic plot of the cathodic shift of the of the J-V
-2

curves in figure 8-2a by sampling the potential needed to sustain a 200 µA cm

current density.

The shift in photocurrent onset potential increases with Co-Pi thickness, quickly reaching a
-2

saturated shift of ~0.23 V for Co-Pi thicknesses greater than 15 mC cm . This behavior is
consistent with experiments examining Co-Pi on planar FTO coated glass, which was attributed
to an increase in the number of active catalytic sites.

13

This trend, however, was not observed in
10

a recent study examining Co-Pi deposited on high surface area hematite.

Thus, the simple

explanation of increasing the number of active catalytic sites does not adequately describe the
enhanced performance of the Co-Pi–hematite system. An alternative explanation is discussed
below. In addition to the shift of the J-V curve, a slight increase in the photocurrent density at
potentials positive of 1.4 V vs RHE is observed. This enhancement, however, is independent of
193

Co-Pi thickness. The improved PEC performance is generally consistent with previous reports of
7,10,11

Co-Pi coated hematite electrodes.

Figure 8-2. (a) J-V curves measured under 1 sun illumination of a bare electrode (red solid line)
and the same electrode with 1 (orange dotted) 2 (yellow short dashed) 15 (green dashed double
-2

dotted) 45 (teal long dashed) and 90 (blue dashed single dotted) mC cm

Co-Pi. (b) Steady state

photocurrent. (c) Steady state photocurrents (shapes), plotted with J-V curves (lines) to show that
-2

they overlap for a bare electrode (red circles) and one with 15 mC cm

Co-Pi (green diamonds).
-2

(d) Potential shift of the current onset relative to the bare electrode measured at 200 mA cm .

194

Figure 8-3. (a) Oxygen detected over time at 1.25 V vs RHE under 4 sun illumination for a bare
hematite electrode (red circles) and the same electrode with 1 (orange pointing down triangles), 2
(yellow squares), 15 (green diamonds), 45 (teal pointing up triangles) and 90 (blue hexagons)
-2

mC cm Co-Pi. (b) The calculated amount of O2 that should be produced based on the number
of coulombs passed (red line) compared to the O2 that was detected (blue hexagons) for a
-2

hematite electrode coated with 90 mC cm Co-Pi.

8.4.1 Transient Behavior
While performing steady state measurements, it became apparent that increasing the CoPi thickness required an increasing amount of time to reach steady state. Current transients were
therefore measured in response to turning on (anodic) and off (cathodic) 1 sun illumination at a
constant potential. Examples of anodic and cathodic current transients for different thicknesses
of Co-Pi on hematite electrodes can be seen in figures 8-4a and 8-4b, respectively, at an applied
bias of 1.05 V vs RHE. As shown previously for bare hematite electrodes,

27

at potentials

negative of the current onset potential (such as 1.05 V vs RHE), an anodic spike in current is
195

visible when the light is turned on which quickly decays to the steady state current density.
When the light is turned off a cathodic spike in current is observed which quickly decays to J0.
These spikes have been attributed to the charging (trapping of holes) and discharging of surface
27

states, or oxidizing and reducing surface species.

Analogous behavior is observed when Co-Pi

is added to the surface of hematite, however the amount of charge passed in the transients
obviously increases with Co-Pi thickness. This suggests that the processes governing the
transients are controlled by the Co-Pi. The anodic transients are attributed to the oxidation of
Co(III) in the Co-Pi catalyst layer to Co(IV) by photogenerated holes in the valence band. This
assignment is based on the recent observation of Co(IV) by EPR from Co-Pi which was
electrodeposited during water oxidation.

29

Since such a large amount of charge is passed, and

the quantity scales with thickness of the Co-Pi layer, the catalyst film must have the Co(IV)
species distributed throughout. This indicates efficient diffusion of holes through the catalyst
film via charge transfer from/to the cobalt centers, consistent with recent self-exchange
measurements of Co-Pi using a model cubane molecule.

30

The cathodic transient measured after

turning the light off is attributed to the reduction of Co(IV) to Co(III) by electrons from the
conduction band of the hematite (i.e. recombination). The recombination time scale indicated by
the cathodic transients generally scales with Co-Pi thickness, with thicker films requiring a
longer time to reduce all the stored Co(IV), however in a less straightforward manner than the
anodic transients. The initial current peak is the same for the 1 and 2 mC cm

-2

Co-Pi films, then

decreases monotonically with increasing Co-Pi thickness. The peak current should be
proportional to the concentration of electrons in the conduction band and electron acceptors,
[Co(IV)], at the interface, and the total charge passed proportional to the total number of Co(IV)
196

centers stored. At a given applied bias it is not expected that changing the Co-Pi thickness would
change the concentration of electrons in the conduction band immediately after turning off the
light, assuming a constant conduction band position (vide infra). Therefore, the decreasing
cathodic peak heights can be attributed to decreasing concentrations of Co(IV) for Co-Pi
thicknesses larger than 2 mC cm

-2

at an applied bias of 1.05 V vs RHE. This corresponds to the

same Co-Pi thickness that water oxidation occurs at 1.05 V vs RHE. Since water oxidation
represents a parallel pathway that would deplete the Co-Pi film of Co(IV), it makes sense that
increasing water oxidation kinetics would lead to a decreased concentration of Co(IV) in the
film. In order to test this hypothesis, we also measured cathodic current transients at 0.95 V vs
-2

RHE where water oxidation does not occur for 1, 2 or 15 mC cm

films (see figure 8-5a). At

this potential, the initial anodic peak current was the same for all films and the total charge
passed is clearly proportional to film thickness, as expected. In addition, cathodic current
transients were measured as a function of applied potential (see figure 8-5b). The initial peak
current drops rapidly with increasing applied bias as expected since increasing the applied bias
should decrease the conduction band concentration at the electrode surface as well decrease
Co(IV) due to increasing water oxidation.

197

Figure 8-4. (a) anodic and (b) cathodic transients measured for a bare hematite electrode (red
solid line) and the same electrode with 1 (orange dotted line) 2 (yellow short dashed line) 15
(green dashed double dotted line) 45 (teal long dashed line) and 90 (blue dashed single dotted
-2

line) mC cm Co-Pi catalyst at an applied bias of 1.05 V vs RHE.

198

Figure 8-5. (a) Cathodic transients measured after turning the light off at an applied potential of
0.95 V vs RHE for a bare hematite electrode (red solid line), one with 1 (orange dotted line), 2
-2

(yellow dashed line), and 15 mC cm (green dashed, dotted line ) of Co-Pi. (b) Cathodic
-2

transients after turning the light off for a hematite with 15 mC cm Co-Pi measured at 1.05
(green solid line), 1.15 (teal dotted line), 1.25 (blue dashed line), and 1.35 V vs RHE (purple
dashed, dotted line).

Light chopping experiments were also performed for electrodes with varying thicknesses
of Co-Pi. Figure 8-6 shows a J-V curve measured under chopped and constant 1 sun illumination
-2

for a 15 mC cm Co-Pi coated hematite electrode. The maximum current of the chopped light JV is approximately linear which is reached instantaneously upon turning the light on. This
behavior is similar to previous reports of hematite electrodes measured in contact with an
electrolyte containing a fast redox shuttle.

27,31,32

Two different regions are observed when

comparing the chopped light J-V curve to the steady state J-V curve. One is at potentials positive
of ~1.4 V vs RHE where the instantaneous photocurrent measured by chopped light is equal to
199

the steady state J-V curve. The other region is between 0.85 and 1.4 V vs RHE where the
instantaneous photocurrent measured by light chopping is much higher than the steady state
photocurrent. In other words, at these potentials, charge is being transferred to and stored in the
Co-Pi film without steady state water oxidation occurring. This “trapping” in the Co-Pi film
presents an opportunity for recombination of electrons in the conduction band and Co(IV) to
produce Co(III). This balance of charge separation and recombination may account for the
different behavior observed for high aspect ratio electrodes.

10

Chopped light J-V curves

measured for different Co-Pi thicknesses can be seen in the Appendix.

-2

Figure 8-6. J-V curve for a hematite electrode coated with 15 mC cm

Co-Pi measured under

constant 1 sun illumination (green line) and under chopped 1 sun illumination (grey line).

8.4.3 Electrochemical Impedance Spectroscopy
IS measurements were also performed for electrodes with varying thicknesses of Co-Pi.
An example of a Nyquist plot measured under illumination for a bare hematite electrode and a
-2

hematite electrode coated with 15 mC cm of Co-Pi at 1.25 V vs RHE can be seen in figure 8-7.

200

Around the photocurrent onset, two semicircles are clearly visible for both bare and Co-Pi coated
hematite electrodes. The low frequency (high impedance) semicircle is clearly much smaller for
Co-Pi coated electrodes compared to the bare electrodes. At more positive potentials ( > 1.25 V
vs RHE), the low frequency semicircle disappears for Co-Pi coated hematite electrodes.

Figure 8-7. Nyquist plots measured under illumination of a bare hematite electrode (red circles)
-2

and with 15 mC cm Co-Pi catalyst (green triangles) measured at 1.25 V vs RHE.

The complete equivalent circuit, EC, used to interpret the IS data is shown in figure 8-8a.
The proposed circuit includes the circuit established previously for a bare hematite electrode
under illumination which has also recently been modeled in detail.

28,33

This EC consists of the

capacitance of the bulk hematite, Cbulk, charge transfer resistance from the valence band of the
hematite, Rct,bulk, a resistance which is related to the rate of trapping holes in surface states,
Rtrap, a capacitance of the surface states, Css, and a charge transfer resistance from the surface

201

28

states, Rct,ss.

Additional electrical components were added to account for the Co-Pi layer

including the capacitance of the Co-Pi layer, CCo-Pi, and charge transfer resistance from the CoPi layer, Rct,Co-Pi. Clearly, the full EC shown in figure 8-8a cannot be used to unambiguously fit
the IS data for Co-Pi coated hematite electrodes. One thing that allowed us to simplify the
equivalent circuit was independent examination of the low frequency semicircle; the capacitance
of this feature increases approximately linearly with increasing Co-Pi thickness (shown and
discussed below). This allowed assignment of this capacitance to CCo-Pi. In addition, as shown
in Figure 8-9b, charge transfer resistance from Co-Pi (Rct,Co-Pi), which were also found to be
proportional to the Co-Pi film thickness, showed much lower values compared to those found for
bare hematite ( Rct,ss). Thus the complete EC could be simplified to that shown in figure 8-8b.

202

Figure 8-8. a) Proposed full equivalent circuit used for interpretation of hematite electrodes
coated with Co-Pi catalyst. b) Simplified equivalent circuit used for interpretation of or Co-Pi
coated Fe2O3. c) Randles’ circuit used when only one semicircle is visible.

The impedance spectra of hematite electrodes with varying amounts of Co-Pi deposited,
measured under 1 sun illumination, were fit to the EC shown in figure 8-8b. Figure 8-9a shows
plots of CCo-Pi vs. applied potential for the different thicknesses of Co-Pi deposited on the
hematite film. The CCo-Pi increases with thickness which is consistent with the assignment of
this capacitance to the chemical capacitance of Co-Pi. This is also consistent with the trend of
increasing charge passed before reaching a steady state current in the current transient
measurements displayed in figure 8-4. Evidence of a capacitance due to the Co-Pi is also
observed by performing J-V measurements at various scan rates of Co-Pi coated hematite
203

9

electrodes in a recent study by Zhong et. al. The values for Rct,Co-Pi, shown in figure 8-9b, are
also related to the Co-Pi layer thickness; the resistance decreases with increasing Co-Pi
thickness. When deposited on FTO, Co-Pi is known to be a porous material which shows a
decrease in the required applied potential with an increasing amount of the catalyst.

13

The

dependence of the decreasing Rct,Co-Pi with Co-Pi thickness is also consistent with the Co-Pi
being a porous material. For all Co-Pi thicknesses, Rct,Co-Pi decreases exponentially with
increasing potential. Similar to the current transients discussed above, there are two clear regions
that are observed in the Nyquist plots. One is at potentials less than 1.4 V vs RHE which shows
two clear semicircles where the contribution of the Co-Pi can be observed. However at potentials
greater than 1.4 V vs RHE the low frequency capacitive feature disappears, and meaningful
values for CCo-Pi and Rct,Co-Pi cannot be calculated. For these data, a simple Randles’ circuit
(figure 8-8c) is used to fit the impedance spectra and calculate values for Rct,bulk and Cbulk. The
low frequency semicircle also disappears for bare hematite electrodes at more positive potentials
( > 1.5 V vs RHE), which has recently been attributed to a hole transfer from the surface states of
28

iron oxide to solution which is not the rate limiting step of water oxidation.

Similarly, we

propose that at potentials where this low frequency capacitive feature measured on Co-Pi coated
electrodes disappears, charge transfer from the Co-Pi to solution is not the rate limiting step. This
is the cause of the independence of photocurrent measured at potentials positive of ~1.4 V vs
RHE for increasing Co-Pi thickness. At these positive potentials, the photocurrent is controlled
by the number of holes that reach the hematite surface for both bare and Co-Pi coated hematite
32

electrodes.

204

Figure 8-9. (a) CCo-Pi and (b) Rct,Co-Pi values fit from impedance response of hematite
electrodes with 1 (orange triangles pointing down), 2 (yellow squares), 15 (green diamonds) 45
(teal triangles pointing up) and 90 mC cm

-2

(blue hexagons) Co-Pi deposited. Bare hematite

fitting parameters of a) Css, b) Rct,ss are shown for comparison (red open circles)

Values of Rct,bulk, which represent the charge transfer from the hematite to the Co-Pi
2

3

2

catalyst, can be found in figure 8. Rct,bulk is on the order of 10 -10 Ω cm . These resistances
3-/4-

are comparable to when a fast redox shuttle such as [Fe(CN)6]
27

reported previously.

is used as a hole scavenger as

This is consistent with having fast charge transfer of holes from the

valence band of hematite to the Co-Pi. At potentials negative of the current onset potential,
Rct,Co-Pi is much higher than Rct,bulk. This is consistent with light chopping experiments where
at these low applied potentials, charge transfer to Co-Pi is facile, yet water oxidation from the
Co-Pi does not occur according to the steady state J-V curves and O2 measurements. Rct,Co-Pi is

205

also included in the same graph as Rct,bulk emphasizing which resistance is the limiting
resistance for a given potential. The total resistance, Rtot was calculated (Rs+ Rct,Co-Pi +Rct,bulk)
and compared to the resistance derived from the J-V curve (Rtot = As(dV/dJ)). A plot of Rtot
derived from both impedance and J-V results can be seen in Figure 8-10b. The overlap of the Rtot
determined from IS and the J-V curve shows how Rct,bulk and Rct,Co-Pi determine the shape of
the J-V curve.

Figure 8-10. (a)Rct,bulk (solid shapes) and Rct,Co-Pi (open shapes) measured for a hematite
electrode coated with 1 (orange pointing down triangles), 2 (yellow squares), 15 (green
diamonds), 45 (teal pointing up triangles) and 90 mC cm

-2

(blue hexagons) Co-Pi. (b) Rtot

calculated from IS (shapes) and J-V curves (lines) for a bare hematite electrode (red circles) and
the same electrode with Co-Pi.

Plots of Cbulk values can be seen in figure 8-11a. These values are essentially constant
for all thicknesses of Co-Pi and the bare hematite electrode. Mott Schottky (MS) plots were
206

prepared from these Cbulk values and are displayed in figure 8-11b. As shown previously for
bare hematite electrodes, MS plots show a horizontal shift which we attribute to Fermi level
27,28

pinning by surface trapped holes.

This pinning is not observed in dark conditions (see

Figure A8-3 in the Appendix). The MS plots measured with the Co-Pi coated hematite electrode
shows a much more linear behavior compared to the bare electrode suggesting that Fermi level
pinning is reduced. This is consistent with photogenerated holes being transferred to the Co-Pi
layer instead of being trapped in surface states. Also, the flat band potential extrapolated from
both bare and Co-Pi coated hematite electrodes is very similar, within the error of this
measurement, suggesting that Co-Pi does not shift of the band positions. This is in contrast with
a recent report that has suggested the primary role of Co-Pi on Co-Pi coated hematite electrodes
is increased band bending.

11

The flat band potential and dopant density extracted from the MS
26-28

plots are in good agreement with previously reports of thin film hematite electrodes.

207

Figure 8-11. (a) Cbulk values fit from IS data of a bare hematite electrode (red circles) and the
same electrode with 1 (orange pointing down triangles), 2 (yellow squares), 15 (green
-2

diamonds), 45 (teal pointing up triangles) and 90 (blue hexagons) mC cm

Co-Pi deposited. (b)

Mott Schottky plot of Cbulk values shown in (a)

8.5 Discussion
To summarize the results presented above, deposition of the Co-Pi catalyst to planar
hematite electrodes produces a favorable shift in the photocurrent onset potential. This shift
increases with increasing Co-Pi thickness up to a saturated value of ~240 mV. Transient
photocurrent and chopped light measurements showed that valence band holes in hematite
efficiently oxidize Co(III) in the Co-Pi film to Co(IV). The transient anodic and cathodic
photocurrent measurements, as well as the CCo-Pi measured by IS, as a function of Co-Pi
thickness clearly show that photogenerated holes from hematite can be stored as Co(IV)
throughout the Co-Pi layer. Also, the decreasing Rct,Co-Pi with increasing Co-Pi thickness
showed that the Co-Pi is porous and has active sites throughout. The resistances extracted from
208

the IS measurements accurately account for the steady-state J-V behavior of Co-Pi coated
hematite electrodes which we further showed corresponds to unity faradaic efficiency of oxygen
generation. Therefore these results clearly demonstrate that water oxidation occurs
predominately from the Co-Pi film, not the hematite surface. Cathodic current transients show
that recombination of electrons from the hematite conduction band to Co(IV) is also a fast
process, which decreases with increasing Co-Pi thickness. The improved photocurrent onset
potential with increasing thickness of Co-Pi can therefore be accounted for by improved charge
separation since an increasing fraction of Co(IV) is distributed throughout the Co-Pi film away
from the hematite surface. It was also found that at potentials >~1.4 V vs RHE, the photocurrent
does not increase with increasing Co-Pi thickness. We propose that at these potentials, there are
enough adjacent Co(IV) active sites that when a hole reaches the hematite-Co-Pi interface, it can
be immediately turned into faradaic photocurrent (oxygen). This is consistent with current
transients which instantaneously reach the maximum current upon illumination, as well as IS
which shows the disappearance of the low frequency semicircle attributed to trapping of holes,
27,28

be it in surface states for a bare electrode or in the Co-Pi for a Co-Pi coated electrode.

209

Figure 8-12. Diagram of a Co-Pi coated hematite electrode at under illumination and applied
bias; the mechanism of Co-Pi is adopted from reference

34

.

We now are in a position to compare our findings with the literature. The cathodic shifts
in photocurrent onset potential measured in this study are higher than the ~180 mV shifts
9-11

observed when Co-Pi is deposited on hematite electrodes with higher surface area.

Also, for

high surface area hematite, the shift did not increase with Co-Pi thickness, and required the
9,10

optimization of thin Co-Pi layers.

Based on our results, a higher surface area hematite

electrode would provide a larger source of electrons to recombine with the Co(IV) since thicker
layers of Co-Pi may bridge two regions of hematite and thus not produce better charge
210

separation. In addition, it is worth pointing out that measurements on structured hematite films
are performed by shining light from the front (electrode) side, whereas we shine light from the
back (FTO substrate) side. Since Co-Pi films broadly absorb visible light,

11

the measurements of

thick Co-Pi films on structured electrodes are difficult to interpret since the decreasing light
absorption by the hematite with thicker Co-Pi films is convoluted with the effect of the Co-Pi
film. Although the thin films used herein are not practical for achieving the highest
photocurrents, the use of such films allowed us to perform a controlled thickness dependence
which showed an increasing performance with increasing Co-Pi thickness; a trend consistent
13

with Co-Pi on FTO.

Transient absorption measurements have also recently been employed to

determine the role of Co-Pi in enhancing the performance of a variety of hematite
11,23

electrodes.

These authors suggested that Co-Pi acts as an electron acceptor which causes

increased band bending and enhancement of charge separation within the hematite, and not hole
transfer to the Co-Pi. This interpretation implies that water oxidation occurs at the hematite
surface and not through the Co-Pi. Our measurements of Cbulk however, including their
representation in Mott-Schottky plots, indicate that the band bending is essentially constant upon
the addition of Co-Pi. Further, as described above, our measurements clearly show efficient hole
transfer from hematite to Co-Pi and that water oxidation is occurring from the Co-Pi catalyst, not
the hematite surface. Thus, the results presented herein are in stark contrast to several recent
reports. We note that the hematite films being compared were prepared in different fashion and
geometry, and very different techniques were employed to interrogate them, which may account
for these differences in interpretation.

211

8.5 Conclusions
Uniform thin films of hematite were coated with varying amounts of the Co-Pi and
investigated with a variety of steady-state and transient photoelectrochemical measurements and
impedance spectroscopy. These systematic studies allowed for the identification of the
mechanism by which Co-Pi enhances the water splitting performance of hematite electrodes. The
Co-Pi catalyst efficiently collects and stores photogenerated holes from the hematite electrode.
This charge separation reduces recombination which results in lower photocurrent onset
potentials and hence water oxidation efficiency. One way of rationalizing the better water
oxidation efficiency with increasing charge separation is through a bimolecular water oxidation
mechanism which requires the oxidation of two adjacent cobalt atoms before oxidizing water.
This mechanism has been previously proposed for Co-Pi on FTO electrodes.

13

In this case, long

lived Co(IV) species are necessary in order to move to an adjacent Co(IV). This framework
allows several analogies to be drawn between the bare hematite electrode and the Co-Pi coated
electrode. Both require the oxidation of a water oxidation active site to an intermediate species
before facile water oxidation occurs.

27

The accumulation of intermediates at illuminated bare

hematite surface in contact with water has been measured in several recent papers by
photoelectrochemical
12

spectroscopy,

impedance

spectroscopy,

2,27,28

transient absorption spectroscopy,

structure measurements.

36

5,23,35

intensity

modulated

photocurrent

and near edge X-ray absorption fine

In the case of a bare electrode this is possibly the oxidation of an

Fe(III)-hydroxide to an Fe(IV)-oxo intermediate, although the identity of this Fe(IV)-oxo
intermediate has not yet been confirmed.

27,37,38

In the case of Co-Pi, this is likely the oxidation

212

13,29

of a Co(III)-hydroxyl to a Co(IV)-oxo intermediate (figure 8-12).

The water oxidation

intermediates of both bare and Co-Pi coated hematite electrodes are both, however, subject to
recombination or reduction by electrons in the conduction band. For both bare and Co-Pi coated
electrodes, this recombination is turned off with an applied potential and high band bending. CoPi coated hematite electrodes differ from bare hematite electrodes by efficiently separating
charge and thereby reducing recombination and allowing longer lived holes as Co(IV) species.
Despite this improved charge separation, Co-Pi is not immune to recombination from electrons
in the conduction band. For Co-Pi to improve photoanodes further, recombination of electrons
from the conduction band must be reduced.

8.6 Acknowledgements
TWH thanks the National Science Foundation (CHE-1150378) for support of this research. JB &
FFS acknowledge support by projects from Ministerio de Industria y Competitividad (MINECO)
of

Spain

(Consolider

HOPE

CSD2007-00007),

and

Generalitat

Valenciana

(PROMETEO/2009/058 and the “Institute of Nanotechnologies for Clean Energies”, under
project ISIC/2012/008). FFS thanks the funding of University Jaume I- Bancaixa (Grant
P1·1B2011-50). SG acknowledges support by MINECO of Spain under the Ramon y Cajal
programme.

213

APPENDIX

214

Figure A8-1. The current measured at 1.25 V vs RHE under 1 sun illumination. The trend
agrees with O2 production in figure 8-3a.

215

Figure A8-2. J-V curve measured under 1 sun illumination (colored line) and chopped light
(grey line) for a) bare hematite electrode and the same electrode coated with b) 1, c) 2, d) 15, e)
-2

45 and f) 90 mC cm of Co-Pi.

216

Figure A8-3. Mott Schottky plot for a bare hematite electrode (circles) in the dark (black
circles) and under 1 sun illumination (red circles) as well as the same hematite electrode coated
-2

with 15 mC cm Co-Pi (diamonds) measured in the dark (black diamonds) and under 1 sun
illumination (green diamonds).

217

REFERENCES

218

REFERENCES

1. Dareedwards, M. P.; Goodenough, J. B.; Hamnett, A.; Trevellick, P. R. J. Chem. Soc.,
Faraday Trans. 1 F, 1983, 79, 2027-2041.
2. Upul Wijayantha, K. G.; Saremi-Yarahmadi, S.; Peter, L. M. Phys. Chem. Chem. Phys. 2011,
13, 5264-5270.
3. Peter, L. M.; Wijayantha, K. G. U.; Tahir, A. A. Faraday Discuss. 2012, 155, 309-322.
Cowa
ar ett C
Pe dle r
arroso
J. R.; Klug, D. R. J. Am. Chem. Soc. 2011, 133, 10134-10140.

i

la

r t el

rra t

5. Pendlebury, S. R.; Barroso, M.; Cowan, A. J.; Sivula, K.; Tang, J. W.; Gratzel, M.; Klug, D.;
Durrant, J. R. Chem. Commun. 2011, 47, 716-718.
6. Tilley, S. D.; Cornuz, M.; Sivula, K.; Grätzel, M. Angew. Chem. Int. Ed. 2010, 49, 64056408.
7. McDonald, K. J.; Choi, K. S. Chem. Mater. 2011, 23, 1686-1693.
8. Zhong, D. K.; Sun, J. W.; Inumaru, H.; Gamelin, D. R. J. Am. Chem. Soc. 2009, 131, 60866087.
9. Zhong, D. K.; Gamelin, D. R. J. Am. Chem. Soc. 2010, 132, 4202-4207.
10. Zhong, D. K.; Cornuz, M.; Sivula, K.; Graetzel, M.; Gamelin, D. R. Energy Environ. Sci.
2011, 4, 1759-1764.
arroso
Cowa
Pe dle
Chem. Soc. 2011, 133, 14868-14871.

r

r t el

l

rra t

J. Am.

12. Cummings, C. Y.; Marken, F.; Peter, L. M.; Tahir, A. A.; Wijayantha, K. G. U. Chem.
Commun. 2012, 48, 2027-2029.
13. Surendranath, Y.; Kanan, M. W.; Nocera, D. G. J. Am. Chem. Soc. 2010, 132, 16501-16509.
14. Kanan, M. W.; Nocera, D. G. Science, 2008, 321, 1072-1075.
15. Lutterman, D. A.; Surendranath, Y.; Nocera, D. G. J. Am. Chem. Soc. 2009, 131, 3838-3839.
16. Hong, Y.-R.; Liu, Z.; Al-Bukhari, S. F. B. S. A.; Lee, C. J. J.; Yung, D. L.; Chi, D.; Hor, T.
S. A. Chem. Commun. 2011, 47.

219

17. Pilli, S. K.; Deutsch, T. G.; Furtak, T. E.; Turner, J. A.; Brown, L. D.; Herring, A. M. Phys.
Chem. Chem. Phys. 2012, 14, 7032-7039.
18. Abdi, F. F.; van de Krol, R. J. Phys. Chem. C, 2012, 116, 9398-9404.
19. Zhong, D. K.; Choi, S.; Gamelin, D. R. J. Am. Chem. Soc. 2011, 133, 18370-18377.
20. Seabold, J. A.; Choi, K.-S. Chem. Mater. 2011, 23, 1105-1112.
21. Steinmiller, E. M. P.; Choi, K.-S. PNAS, 2009, 106, 20633-20636.
22. Pijpers, J. J. H.; Winkler, M. T.; Surendranath, Y.; Buonassisi, T.; Nocera, D. G. PNAS,
2011.
23. Barroso, M.; Mesa, C. A.; Pendlebury, S. R.; Cowan, A. J.; Hisatomi, T.; Sivula, K.; Grätzel,
M.; Klug, D. R.; Durrant, J. R. PNAS, 2012.
24. Martinson, A. B. F.; DeVries, M. J.; Libera, J. A.; Christensen, S. T.; Hupp, J. T.; Pellin, M.
J.; Elam, J. W. J. Phys. Chem. C, 2011, 115, 4333-4339.
25. George, S. M. Chem. Rev. 2010, 110, 111-131.
26. Klahr, B. M.; Martinson, A. B. F.; Hamann, T. W. Langmuir, 2011, 27, 461-468.
27. Klahr, B.; Giménez, S.; Fabregat-Santiago, F.; Bisquert, J.; Hamann, T. Energy Environ. Sci.
2012, 5, 7626-7636.
28. Klahr, B.; Gimenez, S.; Fabregat-Santiago, F.; Hamann, T.; Bisquert, J. J. Am. Chem. Soc.
2012, 134, 4294-4302.
c l i
re dra ath
i c
tich
toia
Nocera, D. G.; Britt, R. D. J. Am. Chem. Soc. 2010, 132, 6882-6883.

Case

H

30. Symes, M. D.; Surendranath, Y.; Lutterman, D. A.; Nocera, D. G. J. Am. Chem. Soc. 2011,
133, 5174-5177.
31. Sanchez, H. L.; Steinfink, H.; White, H. S. J. Solid State Chem. 1982, 41, 90-96.
32. Klahr, B. M.; Hamann, T. W. Appl. Phys. Lett. 2011, 99, 3.
33. Bertoluzzi, L.; Bisquert, J. J. Phys. Chem. Lett. 2012, 3, 2517-2522.
34. Kanan, M. W.; Surendranath, Y.; Nocera, D. G. Chem. Soc. Rev. 2009, 38, 109-114.
35. Pendlebury, S. R.; Cowan, A. J.; Barroso, M.; Sivula, K.; Ye, J.; Gratzel, M.; Klug, D. R.;
Tang, J.; Durrant, J. R. Energy Environ. Sci. 2012, 5, 6304-6312.
220

36. Braun, A.; Sivula, K.; Bora, D. K.; Zhu, J.; Zhang, L.; Grätzel, M.; Guo, J.; Constable, E. C.
J. Phys. Chem. C, 2012, 116, 16870-16875.
37. Hellman, A.; Pala, R. G. S. J. Phys. Chem. C, 2011, 115, 12901-12907.
38. Trainor, T. P.; Chaka, A. M.; Eng, P. J.; Newville, M.; Waychunas, G. A.; Catalano, J. G.;
Brown Jr, G. E. Surf. Sci. 2004, 573, 204-224.

221

Chapter 9:
Water Oxidation at Co-Pi coated Hematite
Electrodes: Reducing Recombination to Co-Pi
with an Al2O3 Blocking Layer

222

9.1 Abstract
Uniform thin films of hematite (α-Fe2O3) deposited by atomic layer deposition (ALD)
were coated with Al2O3 (alumina) before depositing the cobalt phosphate catalyst, “Co-Pi.”
Compared to a hematite electrode coated only with Co-Pi, the alumina blocking layer was found
to effectively push move the water oxidation current onset potential by an additional 200 mV
cathodically. Steady state and transient photoelectrochemical studies were paired with
impedance spectroscopic analysis in order to clarify the role of the Al2O3 layer. It was found
that adding Al2O3 prevents recombination of electrons in the conduction band of the hematite to
the Co-Pi, resulting in a near quantitative hole collection.

223

9.2 Introduction
In the previous chapter, we performed systematic studies of water oxidation at Co-Pi
coated hematite electrodes. Through photoelectrochemical and impedance spectroscopic
measurements, the Co-Pi was found to shift water oxidation to more cathodic potentials as a
result of storing holes capable of oxidizing water as oxidized intermediates and reducing
recombination of electrons in the conduction band to those oxidized states in the Co-Pi. A
thickness dependence showed that the water oxidation continued to shift cathodically as the
thickness of Co-Pi was increased as a result of separating electrons in the conduction band from
oxidized intermediates in the Co-Pi layer even further. These intermediates are hypothesized to
be Co(IV)-oxo species.

1,2

However, it was observed that even thick layers (~400 nm) of Co-Pi

were still susceptible to electron recombination which prevent steady state water oxidation at the
1

most cathodic potentials. In addition, the very thick films of Co-Pi required to obtain a large
shift in the current onset potential are not useful on practical, high aspect ratio substrates because
of the large non-productive optical absorption. While thin film hematite offers advantages for
studying catalysts on semiconductor electrodes, it is likely that only high aspect ratio hematite
electrodes will be practical. Thus, efficient catalysts will also need to be optically transparent.
In order to reduce recombination an alumina blocking layer was deposited onto the
hematite by atomic layer deposition before the photoelectrodeposition of Co-Pi. A similar
strategy has been used to reduce recombination of electrons in the conduction band of TiO2 to
cobalt based redox shuttles of dye-sensitized solar cells.

3-6

In this work, we employ

photoelectrochemical measurements and impedance spectroscopy to investigate the effect of the
alumina blocking layer on the water oxidation Co-Pi coated hematite.
224

9.3 Experimental
Thin films of hematite were deposited on fluorine-doped tin oxide (FTO) coated glass
-2

substrates (Hartford Glass, 12 Ω cm ) by ALD (Savannah 100, Cambridge Nanotech Inc.) using
ferrocene as the metal precursor and ozone as the oxidation source. Deposition conditions have
been described in detail elsewhere.

7,8

The hematite films were annealed by heating to 500° C

over 30 minutes and soaking at 500° C for 30 minutes. Films were prepared by 800 ALD cycles
and measured to be ~48 nm by absorption measurements (Perkin Elmer, Lambda 35 with a
Labsphere integrating sphere) corrected for reflection as described previously, as well as
9

ellipsometric measurements (Horiba Jobin Yvon, Smart-SE). Films were characterized by
Raman Spectroscopy and XRD as described previously.

9

Al2O3 was deposited on the hematite films by ALD using trimethylaluminum and water
as the precursors. Pulses of both precursors were each 0.015 seconds separated by a 6 second
nitrogen purge time. Precursors were used at room temperature and the substrate was heated to
200° C. The growth rate of the deposition was measured to be 1.1 angstroms per cycle by
ellipsometry analysis on witness silicon wafer chips. The growth rate is assumed to be the same
on the hematite electrodes. Samples were annealed after Al2O3 deposition by heating to 500°C
over 30 minutes and soaking at 500° C for 30 minutes.
2

Electrodes were masked with a 60 μm Surlyn film (Solaronix) with a 0.28 cm hole to
define the active area and to prevent scratching of the thin films. Surlyn films were adhered to
the electrodes by heating to 120° C. The protected hematite films were clamped to a custom
225

made glass electrochemical cell. A homemade saturated Ag/AgCl electrode was used as a
reference where potentials were converted the reversible hydrogen electrode (RHE) scale by the
equation ERHE  E Ag / AgCl  0.197  0.059 pH . A high surface area platinum mesh was used
as the counter electrode.
Co-Pi catalyst films were deposited by photoelectrodeposion which was originally
described on ZnO electrode by Steinmiller and Choi and later adapted to hematite by Zhong and
Gamelin.

10,11

Photoelectrodeposition has been shown to provide a superior deposition at the

photoactive anode which is in contrast to electrodeposition which favors deposition at the
underlying conductive substrate.

10-12

Hematite electrodes were immersed in a solution

containing 0.5 mM Co(NO3)2·6H2O in a 0.1 M phosphate buffer (pH 6.9). A bias of 0.8 V vs.
RHE was applied under illumination. At this potential, steady state water oxidation does not
occur with or without the Co-Pi catalyst, thus any charge passed is assumed to be due to
deposition of the catalyst. In this study the amount of Co-Pi deposited was controlled by
-2

allowing 15 mC cm

of charge to pass. Thicknesses were estimated by assuming that a single
32

electron passed deposits one cobalt atom with surrounding ligands which occupies 125Å .
Using this calculation, the deposition of 15 mC cm

-2

corresponds with a thickness of 113 nm.

However, due to the dynamic nature of the catalyst, which is continuously removed and redeposited during water oxidation, catalyst thicknesses will be identified in terms of mC cm
avoid ambiguity.

-2

to

1,2

After the catalyst was deposited, the electrodes were lightly rinsed with DI water to
remove any excess cobalt ions. The water oxidation properties of the catalyst coated hematite
226

films were then examined in contact with an aqueous solution buffered at pH 6.9 using a 0.1 M
phosophate buffer containing 200 mM KCl as a supporting electrolyte. The pH was determined
with Fisher Scientific Accumet pH meter. Impedance spectroscopic and photoelectrochemical
measurements were made with an Eco Chemie Autolab potentiostat coupled with Nova
electrochemical software. Impedance data were gathered using a 10 mV amplitude perturbation
of between 10,000 and 0.01 Hz. Data were fit using Zview software (Scribner Associates). The
light source was a 450 W Xe arc lamp. An AM 1.5 solar filter was used to simulate sunlight at
-2

100 mW cm . All photoelectrochemical measurements were performed by shining light from
the substrate-electrode (SE) interface which avoids competitive light absorption of the Co-Pi.
Light chopping J-V curves were measured at a rate of 75 mV/s. The light was chopped
using a computer controlled ThorLabs solenoid shutter which was set to activate every 266 ms
such that the light was turned on or off every 20 mV. Steady state J-V curves were measured at a
scan rate of 5 mV/s.
Oxygen was detected by using an Ocean Optics spectrometer which probed the
fluorescent decay of the FOSPOR patch. The FOSPOR patch was placed in solution which filled
an airtight cell. The cell was filled so that very little headspace existed. The solution was stirred
vigorously so that the O2 would quickly diffuse through solution. Measurements were made
under 4 sun illumination at 1.3 V vs RHE to increase oxygen production and reduce noise. In the
calculation of the faradaic efficiency, the assumption was made that no oxygen diffused into the
very small headspace during the timescale of these experiments.
9.4 Results and Discussion
Alumina was deposited on to hematite electrodes by ALD. As a control, the effect of an
alumina layer on hematite films was examined without the addition of Co-Pi in a pH 6.8 aqueous
227

solution. It was observed that 1 cycle of alumina had little to no effect on the characteristics of
the J-V where 5 cycles caused anodic shifts in the current onset potential and lower current
density at positive potentials (Figure 9-1a). Figure 9-1b shows J-V curves of a hematite electrode
with 5 cycles of alumina which shifts cathodically with increasing the number of scans. This
shows the instability of the alumina layer under water oxidation conditions which is expected
since alumina is well known to be unstable in the basic conditions associated with water
oxidation. The lack of at least a temporary cathodic shift in the current onset potential for
alumina coated hematite electrodes with respect to bare electrodes is in contrast to a recent study
which has measured a cathodic shift in the water oxidation of hematite films with the addition of
Al2O3 at pH 13.6.

13

Figure 9-1. (a) J-V curves of a ~48 nm thick bare hematite electrode (red solid line), one coated
with 1 cycle Al2O3 (orange dotted line) and one coated with 5 cycles of Al2O3 (yellow dashed
line) in a phosphate buffered aqueous solution (pH 6.8). (b) J-V curves of a hematite electrode
with 5 cycles of Al2O3. The arrow shows J-V curves taken sequentially.

228

A series of hematite electrodes were prepared with varying cycles of alumina deposited
before the photoelectrodeposition of Co-Pi. To each alumina coated hematite electrode, 15 mC
-2

cm

of Co-Pi was added by photoelectrodepositon by applying a potential of 0.8 V vs RHE

under illumination in a Co(NO3)2 solution. The effect of the thickness of the alumina layer can
be seen in Figure 9-2a. With increasing layers of alumina, the current onset shifts cathodically.
However, the current measured throughout most of the potential range is reduced with increasing
alumina thickness. While thicker layers of alumina shift the current onset marginally, the
decrease in photocurrent detracts from the total power that can be gained from this system.
Therefore, to study this system further, an electrode containing only 1 cycle of alumina
separating the Co-Pi and the hematite was used. Figure 9-2b shows the J-V curves of the only
electrodes used for the remainder of the study: a bare hematite electrode, one coated with 15 mC
-2

cm

of Co-Pi and one with a single alumina layer separating the hematite and the Co-Pi. This

shows that current can be measured with the Al2O3 blocking layer at potentials ~200 mV
cathodic compared to Co-Pi electrodes without the blocking layer.

229

Figure 9-2. (a) J-V curves for a bare hematite electrode (red solid line) an electrode coated with
15 mC cm

-2

Co-Pi (orange dotted line), and hematite films with 15 mC cm

-2

Co-Pi separated by

1 (green short dashed line) 3 (teal double dotted dashed line) and 5 (blue long dashed line) cycles
of alumina. (b) J-V curves for a bare hematite electrode (red solid line) an electrode coated with
15 mC cm

-2

Co-Pi (orange dotted line), and hematite films with 15 mC cm

-2

Co-Pi separated by

1 cycle of alumina (green dashed line). Figures only show cathodic scan for clarity. J-V curves
of both cathode and anodic scan can be seen in the Appendix.

To ensure that the current measured by the J-V curves was stable and non-transient,
steady state current measurements were performed where the photocurrent was sampled after the
current was allowed to stabilize for 1 minute for individual potentials. It was found that the
steady state current densities are equal to the current densities of the cathodic scan of the J-V
curves, confirming that the J-V measurements presented herein represent steady-state behavior
(Figure 9-3a). The faradaic efficiency of the J-V curves was also measured by a fluorescent O2
sensor. The O2 concentration was measured at 1.0 V vs. RHE to see if the current at such
230

cathodic potentials could be attributed to water oxidation. Due to the low currents at this
potential, oxygen measurement were carried out under simulated solar illumination equivalent to
4 suns.

The amount of O2 detected is closely consistent with ~100% faradaic efficiency when

considering the number of coulombs that have passed (figure 2b). The deviation measured after
400 seconds is expected to be an experimental deviation where O2 is diffusing in to the small
volume of headspace. This deviation appears to be large because of the low currents and low O2
concentrations being generated at these very cathodic potentials. Despite this, it is clear that the
addition of the Al2O3 blocking layer in between Co-Pi and hematite makes it possible to oxidize
water at the very cathodic potentials of 1.0 V vs. RHE.

Figure 9-3. (a) Photocurrent measured after one minute at the applied potential to reach steady
state (circles) and J-V measured at 5 mV/s for an Al2O3 modified Co-Pi coated hematite
electrode. (b) Oxygen detected over time at 1.0 V vs RHE under 4 sun illumination for the same
electrodes. The black line shows the theoretical amount of oxygen produced based on 100%
faradaic efficiency

231

Figure 9-4. J-V curve for a hematite electrode coated with 15 mC cm

-2

Co-Pi measured under

constant 1 sun illumiation scanned at 5 mV/ s (green line) and under chopped 1 sun illumination.

In chapters 6 and 8, we used chopped J-V curves to emphasize the difference between
steady state and transient behavior. For hematite electrodes in contact with a fast redox shuttle,
3-/4-

[Fe(CN)6]

, the chopped J-V curves matched closely with the steady state J-V curves showing

that the redox couple is collecting all of the holes that reach the surface without significant
recombination (Figure 6-6a). On the other hand, for hematite electrodes in contact with just an
aqueous electrolyte which excludes a fast redox shuttle, current spikes are measured which
represent the significant recombination at surface states (Figure 6-6b). Thus the difference
between the number of holes that reach the surface can be compared to the number of holes that
perform steady state oxidation can be examined by analyzing the chopped and steady state J-V
respectively. The chopped illumination J-V curve shown in Figure 9-4 (grey line) shows a
transient behavior which is almost consistent with the steady state behavior. This suggests that
almost all of the holes that are being transferred to the Co-Pi are participating in water oxidation.
232

This is in contrast to Co-Pi coated hematite electrodes which show transient behavior
significantly different from steady state measurements, especially at anodic potentials (Figure 86). For Co-Pi coated electrodes without the alumina modification, this difference was attributed
to the large amount of recombination that still occurs in the Co-Pi. The fact that the Al2O3
reduces this difference in the transient behavior suggests that the alumina is reducing
recombination to the Co-Pi and that the Co-Pi is oxidizing water nearly quantitatively with the
holes that it is exposed to.
Electrochemical impedance spectroscopy was employed to further understand the role of
the alumina layer in the composite electrodes. Nyquist plots measured under 1 sun illumination
and at a potential of 1.15 V vs RHE are shown in Figure 9-5. Nyquist plots of all three electrodes
show similar characteristics which include two semicircles. For both Co-Pi coated electrodes, an
equivalent circuit was used which contains the series resistance Rs, the capacitance of the bulk
hematite, Cbulk, the resistance of charge transfer to the Co-Pi, Rct,bulk, a capacitance of the CoPi, CCo-Pi, and a charge transfer resistance from the Co-Pi to solution, Rct,Co-Pi. This equivalent
circuit has been discussed and justified for Co-Pi coated hematite electrodes previously.
7,8

depth discussion on the impedance interpretation of bare electrodes,

1

An in

and the comparison of

1

bare to Co-Pi coated electrodes can be found elsewhere. For clarity, only the impedance results
of Co-Pi coated hematite electrodes and alumina modified Co-Pi coated hematite electrodes will
be discussed.

233

Figure 9-5. (a) Nyquist plots measured at 1.15 V vs RHE under 1 sun illumination for a bare
hematite electrode (red circles), one coated with 15 mC cm

-2

Co-Pi (green diamonds) and one

with a Al2O3 layer separating the hematite and the Co-Pi (orange triangles). (b) Equivalent
circuit used to fit impedance data.

The capacitance measured of the Co-Pi, CCo-Pi as a function of potential under 1 sun
illumination can be found in figure 9-6a. For the Co-Pi coated hematite electrode there is a large
capacitance peak which is consistent with results in the previous chapter for Co-Pi films of
1

similar thickness. The capacitance disappears at a potential of 1.3 V vs RHE which corresponds
to when the J-V curve becomes linear and is limited by charge reaching the surface.

7,14

The

capacitance of Co-Pi measured for the alumina modified electrode is approximately 5 times
lower. One possibility for the lower capacitance could be that the Co-Pi layer is thinner on the
alumina modified film. For example, since the number of coulombs were used to determine the
thickness, it is possible that fewer of the coulombs passed when depositing the Co-Pi film on the
alumina modified could actually be contributed to building the Co-Pi film. However, if the
234

thickness of the Co-Pi is lower for alumina modified Co-Pi coated hematite electrodes, the
charge transfer from the Co-Pi, Rct,Co-Pi, would be expected to be higher. We have recently
shown, as result of increasing the thickness of Co-Pi on bare hematite electrodes, the CCo-pi
increases as more charge can be stored in the Co-Pi, and the Rct,Co-Pi decreases as the porous CoPi increases the surface area able to participate in water oxidation. In light of this, the Rct,Co-Pi of
the alumina modified electrode actually is decreased compared to the Co-Pi hematite electrode.
This suggests that the decreased CCo-Pi and the decreased Rct,Co-Pi are a result of the alumina
layer, and not related to a difference (if any) of the Co-Pi on each electrode. The decreased CCoPi with the addition of an alumina blocking layer suggests that a lower amount of charge is stored

in the Co-Pi film. A decreased Rct,Co-Pi, for an assumed identical (if not smaller) amount of CoPi suggests that there are more sites (oxidized cobalt species) which are capable of oxidizing
water. This is consistent with a reduction in electron recombination from the conduction band of
the hematite to those oxidized cobalt species in the Co-Pi film.

235

Figure 9-6. (a) CCo-Pi and (b) Rct,Co-Pi values fit from impedance response of a hematite
electrode coated with 15 mC cm

-2

Co-Pi (orange triangles), and hematite films with 15 mC cm

Co-Pi separated by 1 cycle of alumina (green diamonds).

236

-2

Figure 9-7.

Scheme of the effect of an alumina blocking layer in Co-Pi coated hematite

electrodes. Figure adapted from figure 8-12 and reference

15

.

9.5 Conclusion
The addition of a Co-Pi layer to planar hematite electrodes was shown to shift the current
onset potential cathodically due to a separation of electrons in the conduction band of the
hematite from holes capable of oxidizing water. However, despite a reduced recombination,
stored holes in the Co-Pi were still found to be vulnerable to recombination from electrons in the
conduction band.

1

In this chapter, this recombination was successfully reduced further by

237

separating the hematite and the Co-Pi with a single layer of alumina which was deposited by
ALD. This allowed for water to be oxidized at potentials ~200 mV cathodic of a Co-Pi coated
hematite electrode without the alumina blocking layer. The comparison of chopped light and
steady state behavior showed that nearly all holes that reach the surface can be used to oxidize
water. The assignment of reduced electron recombination from the conduction band of hematite
to stored holes in the Co-Pi agrees with the impedance spectroscopy analysis. A scheme of the
effect of the alumina blocking layer in Co-Pi coated hematite electrodes can be found in figure 97. This figure shows how the alumina reduces the deleterious recombination of electrons in the
conduction band to the oxidized Co-Pi (likely a Co(IV)-oxo), while allowing the favorable
oxidation of the Co-Pi. It is observed, however, that thicker alumina layers also hinders hole
transfer from the hematite to the Co-Pi. A possible reason for this effect is the fact that alumina
is a wide band gap semiconductor in which the conduction band is at a higher energy than the
conduction band of hematite and the valence band of alumina is at a lower energy than the
valence band of hematite. This would prevent the transfer of both holes and electrons from the
hematite to the Co-Pi by an energetic barrier. However, the description of a conduction band is
not likely relevant for a layer of alumina of only ~1.1 angstroms thick. A more detailed
description of the cause of the reduced recombination by the addition of alumina layer is
currently being developed in our lab.

238

APPENDIX

239

Figure A9-1. J-V curves for a bare hematite electrode (red solid line) an electrode coated with 15
-2

mC cm

Co-Pi (orange dotted line), and hematite films with 15 mC cm

-2

Co-Pi separated by 1

(green short dashed line) 3 (teal double dotted dashed line) and 5 (blue long dashed line) cycles
of alumina. b) J-V curves for a bare hematite electrode (red solid line) an electrode coated with
15 mC cm

-2

Co-Pi (orange dotted line), and hematite films with 15 mC cm

1 cycle of alumina (green dashed line).

240

-2

Co-Pi separated by

REFERENCES

241

REFERENCES

1. Klahr, B.; Gimenez, S.; Fabregat-Santiago, F.; Bisquert, J.; Hamann, T. W. J. Am. Chem. Soc.
2012, 134, 16693-16700.
2. Surendranath, Y.; Kanan, M. W.; Nocera, D. G. J. Am. Chem. Soc. 2010, 132, 16501-16509.
3. Hamann, T. W.; Farha, O. K.; Hupp, J. T. J. Phys. Chem. C, 2008, 112, 19756-19764.
4. Klahr, B. M.; Hamann, T. W. J. Phys. Chem. C, 2009, 113, 14040-14045.
5. Ondersma, J. W.; Hamann, T. W. J. Phys. Chem. C, 2010, 114, 638-645.
6. DeVries, M. J.; Pellin, M. J.; Hupp, J. T. Langmuir, 2010, 26, 9082-9087.
7. Klahr, B.; Giménez, S.; Fabregat-Santiago, F.; Bisquert, J.; Hamann, T. Energy Environ. Sci.
2012, 5, 7626-7636.
8. Klahr, B.; Gimenez, S.; Fabregat-Santiago, F.; Hamann, T.; Bisquert, J. J. Am. Chem. Soc.
2012, 134, 4294-4302.
9. Klahr, B. M.; Martinson, A. B. F.; Hamann, T. W. Langmuir, 2011, 27, 461-468.
10. Steinmiller, E. M. P.; Choi, K.-S. PNAS 2009, 106, 20633-20636.
11. Zhong, D. K.; Cornuz, M.; Sivula, K.; Graetzel, M.; Gamelin, D. R. Energy Environ. Sci.
2011, 4, 1759-1764.
12. Zhong, D. K.; Gamelin, D. R. J. Am. Chem. Soc. 2010, 132, 4202-4207.
13. Le Formal, F.; Tetreault, N.; Cornuz, M.; Moehl, T.; Gratzel, M.; Sivula, K. Chem. Sci.
2011, 2, 737-743.
14. Klahr, B. M.; Hamann, T. W. Appl. Phys. Lett. 2011, 99, 3.
15. Kanan, M. W.; Surendranath, Y.; Nocera, D. G. Chem. Soc. Rev. 2009, 38, 109-114.

242

Chapter 10:
Conclusions and Future Directions

243

10.1 Conclusions
Atomic layer deposition was used to deposit thin films of hematite onto planar substrates.
This process was studied and optimized so that thin, pinhole free, crystalline films could be
prepared in a reproducible fashion. These thin films were used in systematic studies to evaluate
the limitations of hematite for use in the water oxidation half of photoelectrochemical water
splitting. The following conclusions were made from this work:
3-/4-

1. Hematite films were evaluated in contact with a fast redox shuttle, [Fe(CN)6]

. It was

shown that this redox shuttle can collect all of the holes that reach the surface. This redox
shuttle was used to probe the properties of the bulk hematite by removing the limitations
at the surface associated with water oxidation (chapters 3 and 6).

1,2

2. A controlled thickness dependence, made possible by the controlled deposition by ALD,
3-/4-

was performed in contact with [Fe(CN)6]

. It was found that the optimum thickness

of hematite electrodes was approximately the thickness of the depletion region according
to Mott Schottky analysis. This suggests that the diffusion length is approximately zero
3

and charge can only be collected in the presence of an electric field (chapter 2).

This is

confirmed by modifying the electric field through the modification of redox shuttles and
1

pH in the contacting electrolyte (chapter 3).

This conclusion allowed us to model the

linear current-voltage characteristics under illumination for hematite electrodes in contact
3-/4-

with [Fe(CN)6]

4

(chapter 4).

3. Water oxidation was then studied at hematite electrodes under illumination using
electrochemical impedance spectroscopy.

244

An equivalent circuit was developed and

justified in order to interpret the experimental impedance data. This model included a
surface state capacitance and an associated charge transfer resistance. The coincident
observation of a peak in the surface state capacitance, a valley in the charge transfer
resistance from those surface states, and the photocurrent onset for an applied potential
5

suggests that water oxidation occurs from these surface states (chapter 5). The charge of
these surface states were also detected by current transient and cyclic voltammetry
measurements. The requirement of the buildup of charge in the surface states was found
to be responsible for Fermi level pinning which accounted for an additional ~140 mV
applied potential to obtain a similar Fermi potential in the electrode compared when it is
in contact with a fast redox shuttle. The remaining difference of photocurrent between
the oxidation of a fast redox shuttle and water is attributed to recombination from
electrons in the conduction band to oxidized intermediates at the surface (chapter 6).

2

4. The photooxidation of water and methanol was examined independently and in
competition with each other at hematite electrodes.

Oxygen evolution experiments

allowed the measurement of a faradaic efficiency for water oxidation with different
amounts of methanol added to solution. A decreasing faradaic efficiency and measured
surface state capacitance as a result of an increasing concentration of methanol further
confirms that water oxidation occurs from the electrochemically measured surface states
(chapter 7).
5. In order to avoid recombination at the surface, the Co-Pi catalyst was deposited onto the
hematite electrode. Co-Pi was found to efficiently collect photogenerated holes from the
hematite which helped separate them from electrons in the conduction band. This was
confirmed by a systematic Co-Pi thickness dependence which showed the current onset
245

potential shift cathodically with an increasing amount of Co-Pi. Despite this advantage,
Co-Pi was still susceptible to electron recombination as indicated by a comparison of the
steady state and chopped illumination J-V measurements (chapter 8).

6

This

recombination was reduced by adding an alumina blocking layer in between the hematite
and the Co-Pi layer. This allowed for a near quantitative charge collection efficiency
where nearly all of the holes that reach the surface were able to participate in water
oxidation.
10.2 Future Directions
While surface states have been identified as a crucial component in the mechanism for
water oxidation at hematite electrodes, the identity of those surface states has only been
speculated on. For example, in chapter 6, we have hypothesized that a surface Fe(III)-OH is
oxidized Fe(IV)=O where Fe(IV)=O is the charge which is observed in the impedance, cyclic
voltammetry and current transient measurements. Water oxidation may require the accumulation
of two adjacent Fe(IV)=O which would result in the requirement of a buildup of charge or a
measured capacitance. The determination of the chemical identity of these intermediates would
provide insight into the detailed mechanism of water oxidation at hematite electrodes, as well as
guide the discovery/synthesis of more efficient water oxidation photocatalysts. In attempts to
identify these intermediates, in-situ Raman, UV-Vis and IR spectroscopies should be performed.
Depending on the sensitivity of the technique, high surface area electrodes may be required to
increase the signals attributed to the surface.
The Co-Pi catalyst was added to hematite electrodes and nearly quantitative charge
collection was achieved by adding an alumina blocking layer which prevented electron
recombination to the Co-Pi. However, this was only accomplished on planar films which
246

employed a relatively thick layer of Co-Pi (>50 nm). This Co-Pi modification would not be as
effective on nanostructured films due to the competitive light absorption of the Co-Pi film.
Thus, optically transparent catalysts must be developed. To this end, thin layers (less than one
monolayer) of Co(OH)2/Co3O4 were deposited onto planar and high aspect ratio hematite.

7

Although these electrodes were not able to achieve quantitative hole collection, this modification
shifted the current onset potential by ~200 mV while being optically transparent, even on high
aspect ratio substrates.

This enhancement was confirmed to be catalytic in nature by

electrochemical impedance spectroscopy. More work is necessary to achieve quantitative hole
collection at the surface of hematite with optically transparent overlayers.
Assuming a perfect overlayer/catalyst can be deposited onto hematite, the problem still
remains of getting holes, generated deep within the bulk, to the surface so that they may oxidize
water. Aside from the obvious benefits of nanostructuring electrodes in order to maximize light
absorption with minimizing the length charge must travel before being collected, other strategies
which increase the distance charge travels before recombining should be examined.

For

example, adding titanium impurities to hematite has been observed to increase the photocurrent
density for water oxidation. This has been attributed to wide range of explanations including Ti
8

9

acting as an electrical dopant, an increased crystallinity, and the passivation of surface states.

10

More recently, the systematic control of morphology, light absorption and Ti concentration in
hematite films prepared by ALD has ruled out the effect of Ti as an electrical dopant, catalyst, or
a surface passivation layer. Rather, the enhancement of Ti impurities in thin films was attributed
to fixing the “dead layer” which is hypothesized to result from a lattice mismatch of the substrate

247

and the hematite.

11

Further fundamental studies are required to understand the role of other

impurities so that the discovery of beneficial additives can be systematically pursued.
Chapters 2 through 4 have pointed out that charge can only be collected in the presence
of an electric field in the hematite electrodes studied herein. For a given applied potential, the
electric field within the hematite can be increased by creating a p-n junction. One possible
example of this is coating hematite with a nickel oxide layer, a p-type semiconductor with band
positions that should facilitate charge separation. Finding a surface modification which can
increase the electric field within the hematite electrode and combining that with a surface
catalyst capable of quantitative charge collection offers the possibility of an efficient water
oxidation electrode made of inexpensive and stable materials.

248

REFERENCES

249

REFERENCES

1. Klahr, B. M.; Hamann, T. W. J. Phys. Chem. C, 2011, 115, 8393-8399.
2. Klahr, B.; Giménez, S.; Fabregat-Santiago, F.; Bisquert, J.; Hamann, T. Energy Environ. Sci.
2012, 5, 7626-7636.
3. Klahr, B. M.; Martinson, A. B. F.; Hamann, T. W. Langmuir 2011, 27, 461-468.
4. Klahr, B. M.; Hamann, T. W. Appl. Phys. Lett. 2011, 99, 3.
5. Klahr, B.; Gimenez, S.; Fabregat-Santiago, F.; Hamann, T.; Bisquert, J. J. Am. Chem. Soc.
2012, 134, 4294-4302.
6. Klahr, B.; Gimenez, S.; Fabregat-Santiago, F.; Bisquert, J.; Hamann, T. W. J. Am. Chem. Soc.
2012, 134, 16693-16700.
7. Riha, S. C.; Klahr, B. M.; Tyo, E. C.; Seifert, S.; Vajda, S.; Pellin, M. J.; Hamann, T. W.;
Martinson, A. B. F. ACS Nano, 2013, 7, 2396-2405.
8. Wang, G.; Ling, Y.; Wheeler, D. A.; George, K. E. N.; Horsley, K.; Heske, C.; Zhang, J. Z.;
Li, Y. Nano Lett. 2011, 11, 3503-3509.
9. Tang, H.; Matin, M. A.; Wang, H.; Deutsch, T.; Al-Jassim, M.; Turner, J.; Yan, Y. J. Appl.
Phys. 2011, 110, 123511-123517.
10. Chemelewski, W. D.; Hahn, N. T.; Mullins, C. B. J. Phys. Chem. C, 2012, 116, 5255-5261.
11. Zandi, O.; Klahr, B. M.; Hamann, T. W. Energy Environ. Sci. 2013, 6, 634-642.

250