Open Pengtao Xu Thesis V5.Pdf

The Pennsylvania State University

The Graduate School

Eberly College of Science

DYE-SENSITIZED PHOTOELECTROCHEMICAL CELL FOR WATER SPLITTING

A Dissertation in

Chemistry

Pengtao Xu

Submitted in Partial Fulfillment of the Requirements for the Degree of

Doctor of Philosophy

December 2018

The dissertation of Pengtao Xu was reviewed and approved* by the following:

Thomas E. Mallouk Evan Pugh University Professor of Chemistry, Biochemistry and Molecular Biology, Physics, and Engineering Science and Mechanics Dissertation Advisor Chair of Committee

Raymond E. Schaak DuPoint Professor of Materials Chemistry

Benjamin J. Lear Associate Professor of Chemistry

Noel C. Giebink Assistant Professor of Electrical Engineering

Scott A. Showalter Associate Professor of Chemistry, and Biochemistry and Molecular Biology Graduate Program Chair

*Signatures are on file in the Graduate School

Abstract

Artificial photosynthesis mimics the natural processes of converting solar energy, water, and

CO2 into chemical fuels. Water-splitting dye-sensitized photoelectrochemical cells (WS-DSPECs) present a modular approach on a molecular level for solar-to-fuel conversion. WS-DSPECs utilize a high surface area metal-oxide electrode that is sensitized with molecular light absorbers for light harvesting and functionalized with molecular or nanoscopic catalysts for promoting water oxidation and reduction reactions. In an operating photoanode of a WS-DSPEC, light-induced electron injection from sensitizer molecules to the metal oxide support (e.g. TiO2) occurs within one nanosecond. The slow electron recombination from the metal oxide to oxidized sensitizer molecules creates at the semiconductor-sensitizer interface a charge-separated state that persists for microseconds to milliseconds, an adequate time for the subsequent water oxidation reaction.

Despite these promising features, WS-DSPECs reported to date still operate at a low energy conversion efficiency, with the main bottleneck being the undesired back electron transfer process.

This thesis explores in details the fundamental kinetic processes in WS-DSPECs. Chapter 1 introduces the research background and working principles of WS-DSPECs and summarizes recent research progress. Chapter 2 presents the characterization of the flat-band potentials of two- dimensional metal oxide nanosheets, an initial attempt to use thin metal oxide nanosheets as sensitizer scaffolds. In Chapter 3, we characterize the charge recombination process in WS-

DSPECs using a combination of intensity-modulated photovoltage spectroscopy, photoelectrochemical impedance spectroscopy, and time-resolved absorption spectroscopy. The sharp differences in recombination lifetimes as measured by different techniques is rationalized in terms of the experimental conditions. We also formulate the reaction orders for the recombination process at the semiconductor-sensitizer interface. In Chapter 4, we combine numerical modeling with intensity-modulated photocurrent spectroscopy to simulate the charge transport dynamics in

iii

WS-DSPECs. This approach outlines how individual processes (such as electron diffusion, electron recombination, and sensitizer regeneration) influence the electrode performance. Chapter 5 demonstrates a buried-junction design of WS-DSPECs for water oxidation. Mummifying a solid- state dye-sensitized solar cell within an atomically thin protecting layer gives improved power- conversion efficiency and greatly enhanced stability. These results help us understand the fundamental mechanisms of electron transfer and catalysis in WS-DSPECs and enable re-design of the photoanode for the creation of more efficient and durable artificial photosynthetic systems.

Table of Contents

List of Figures ...... viii

List of Tables ...... xv

Acknowledgements ...... xvi

Water-splitting Dye-sensitized Solar Cells ...... 1

1.1. Introduction ...... 2 1.2. Dye sensitized solar cells: regenerative and photosynthetic ...... 5 1.3. Dye-sensitized photoanodes for water-splitting cells ...... 8 1.3.1. Electron and proton transfer at dye-sensitized photoanodes ...... 8 1.3.2. Core-shell photoanode architectures ...... 20 1.4. Photocathodes for water splitting dye cells ...... 29 1.4.1. General principles ...... 29 1.4.2. p-Type semiconductors for water splitting photocathodes ...... 35 1.4.3. Photocathode sensitizers ...... 38 1.4.4. Photocathode catalysts ...... 40 1.4.5. Conclusions ...... 42

Flat-band Potential in Molecularly Thin Metal Oxide Nanosheets ...... 44

2.1. Introduction ...... 45 2.2. Experimental Section ...... 47 2.2.1. Preparation of layered metal oxide nanosheets ...... 47 2.2.2. Layer-by-layer (LBL) assembly of the nanosheets ...... 48 2.2.3. Electrochemical measurements ...... 49 2.2.4. Characterization ...... 50 2.2.5. Electronic structure calculations ...... 50 2.3. Results and Discussion ...... 51 2.3.1. Preparation of nanosheets ...... 51 2.3.2. Layer-by-Layer Assembly ...... 52 2.3.3. Band gap determination ...... 53 2.3.4. Mott-Schottky Experiments ...... 55 2.4. Conclusions ...... 61

Charge Recombination with Fractional Reaction Orders in Water-Splitting Dye- sensitized Photoelectrochemical Cells ...... 63

3.1. Introduction ...... 64 3.2. Theory ...... 67 3.3. Experimental Section ...... 69 3.4. Results and Discussion ...... 71 3.4.1. Photoelectrochemical impedance spectroscopy ...... 72 3.4.2. Intensity-modulated photovoltage spectroscopy...... 74 3.4.3. Transient absorption spectroscopy ...... 80 3.5. Conclusions ...... 83

Charge Transport Dynamics in Dye-sensitized Photoelectrochemical Cells ...... 85

4.1. Introduction ...... 86 4.2. Theory for Numerical Modeling of IMPS ...... 88 4.3. Experimental Section ...... 90 4.4. Results and Discussion ...... 91 4.4.1. Simulation parameters ...... 91 4.4.2. Steady-state concentration profile ...... 94 4.4.3. Influence of 푰ퟎ, 푫풏, 풌ퟐ, and 풌ퟑ on IMPS...... 95 4.5. Conclusions ...... 100

Dye-sensitized Photoelectrochemical Water Oxidation Through a Buried Junction . 102

5.1. Introduction ...... 103 5.2. Experimental section ...... 106 5.2.1. ss-DSSC Fabrication ...... 106 5.2.2. Electrodeposition of iridium oxide films ...... 107 5.2.3. Photoelectrochemical measurements ...... 107

5.2.4. Generator-collector O2 detection ...... 108 5.3. Results and Discussion ...... 110 5.3.1. Preparation of ss-DSSCs ...... 110

5.3.2. Electrochemical deposition of the IrOx water oxidation catalyst ...... 112 5.3.3. Photoelectrochemical properties of catalyzed photoelectrodes ...... 113 5.4. Conclusions ...... 118

Conclusion and Future Outlook ...... 120

Supporting Information for Chapter 2 ...... 124

Supporting Information for Chapter 3 ...... 128

Supporting Information for Chapter 5 ...... 134

Reference… ...... 139

vii

List of Figures

Figure 1-1. Solar fuel production by (a) semiconductor photoelectrochemistry, e.g., in the 16 Fujishima-Honda cell , (b) photocatalysis, e.g., at a Rh-catalyzed Ga1-xZnxN1- 17 xOx particle (adapted with permission from Ref. , Copyright 2006 John Wiley & Sons, Inc.), and (c) artificial photosynthesis in a sensitizer-donor(D)- acceptor(A) system 18...... 3

Figure 1-2. The architecture of regenerative (left) and water-splitting (right) DSSCs. The regenerative DSSC is fabricated as a sandwich cell with a short (~100 µm) electrolyte path between the photoanode and dark cathode. In the water- splitting cell, the anode and cathode are typically separated by a membrane or frit (not shown) in order to enable separation of photogenerated oxygen and hydrogen. Adapted with permission from Ref. 35. Copyright 2009 American Chemical Society...... 6

Figure 1-3. Electrochemical potential diagrams for (a) Fujishima-Honda, (b) Grätzel, and (c) water-splitting dye-sensitized solar cells. Solid lines indicate forward electron transfer pathways. Charge recombination pathways are shown as dashed lines...... 7

Figure 1-4. Sensitizers and sensitizer-catalyst assemblies (catalysts are colored red) that have been used in dye-sensitized photoanodes. See Table 1-1 for references...... 9

Figure 1-5. Molecular OER catalysts used in dye-sensitized water splitting cells. See Table 1-1 for references...... 15

Figure 1-6. Photoelectrochemical transients obtained with different dyes and water oxidation catalysts in water-splitting dye cells. Reproduced from Ref. 44 (with permission of The Royal Society of Chemistry), Ref. 43 (with permission of The Royal Society of Chemistry), Ref. 49 (Copyright 2013 American Chemical Society, adapted with permission), Ref. 51, and Ref. 54 (Copyright 2014 American Chemical Society, adapted with permission)...... 16

Figure 1-7. Kinetic scheme for the photoanode of a water-splitting dye-sensitized solar cell containing an adsorbed molecular sensitizer and a colloidal IrO2 catalyst sintered to the high surface area anatase TiO2 film. Reproduced with permission from Ref. 73. Copyright 2014 American Chemical Society...... 17

Figure 1-8. Left: Photocurrent transients from photoanodes sensitized from DMSO and ethanol solutions. Black lines are fits to (Eq. 1-1). Right: Calculated number density of conduction band electrons (ncb), trapped electrons (nt), and oxidized sensitizer molecules (Ru(III)) from photocurrent model of the ethanol- sensitized electrode. Reproduced with permission from Ref. 73. Copyright 2014 American Chemical Society...... 19

Figure 1-9. (a) Cartoon depicting two types of core-shell architecture. (b) TEM image of a core-shell structure made from 75 ALD cycles of TiO2 grown onto a SnO2 film on FTO glass. Reproduced from Ref. 62...... 21

viii

Figure 1-10. (Left) High resolution energy dispersive X-ray spectroscopy mapping of a SnO2/TiO2 core-shell structure. Ti: red, Sn: green. (Center) TRTS measurement of a RuP1-sensitized SnO2/TiO2 photoanode in 0.1 M HClO4 (pH 1, red) and 100 mM potassium phosphate buffer (pH 6.8, blue). (Right) Early picosecond kinetics of a RuP1-sensitized SnO2/TiO2 photoanode showing rapid 88 recombination in the TiO2 shell. Reproduced with permission from Ref. . Copyright 2016 American Chemical Society...... 23

Figure 1-11. Plot of ln(1/τ1/2) vs TiO2 shell thickness for amorphous SnO2/TiO2 (red) and ZrO2/TiO2 (green) core-shell photoanodes at equal injection yields. 휏1/2 is the time for half the total absorbance change in TAS measurement. The fit models the back electron transfer dynamics with contributions from both tunneling and localized shell recombination. Reproduced with permission from Ref. 84. Copyright 2015 American Chemical Society...... 25

Figure 1-13. Chronoamperometry of RuP7-IrOx nanoparticle-sensitized nanoITO/TiO2 photoanodes with and without an ALD TiO2 coating in pH 5.8 buffer solution (Na2SiF6-NaHCO3) at an applied bias of 300 mV vs Ag/AgCl, illuminated at 455 nm and 14.5 mW/cm2. Reproduced with permission from Ref. 65. Copyright 2015 American Chemical Society...... 27

Figure 1-14. Left: Schematic depictions of an added PMMA coating on a metal-oxide surface (TiO2 or nanoITO) and the results of contact angle measurements on a mesoporous TiO2 film before and after soaking in a PMMA/DCM coating solution. Right: Photocurrent stability comparison for TiO2-RuP1 electrodes with and without a PMMA overlayer, tested in 0.1 M phosphate buffer, pH 7, with 0.4 M NaClO4 and 20 mM hydroquinone as a sacrificial electron donor. Reproduced with permission from Ref. 98. Copyright 2014 American Chemical Society...... 28

Figure 1-15. Structure and energy scheme of the dye-sensitized photocathode for water splitting...... 31

Figure 1-16. Structures of the sensitizers (a) and catalysts (b) from the photocathodes summarized in Table 1-2. (Catalyst is colored red when covalently connected to a sensitizer) ...... 34

Figure 1-17. Left: BH4-sensitized NiO photocathode and Mo-based HER catalyst with the energetics of hole and electron transfer. Right: linear voltammetric sweep with light chopping of a BH4-sensitized NiO electrode with varying electrolyte compositions. Reproduced with permission from Ref. 109. Copyright 2016 American Chemical Society...... 37

Figure 1-18. Schematic drawing and energy level diagram (left) of the photocathode reported by Li et al. 106, and its corresponding chronoamperogram under

Figure 1-19. (a) Supramolecular dye-catalyst assembly on a photocathode: layer-by-layer deposition of RuP4, Zr4+ and Ni3. (b) Energy level diagram of the supramolecular dye-catalyst assembly showing the electron transfer scheme. (c) Chronoamperometry of the assembled photocathode under chopped light irradiation at Eappl = 0.3 V vs. RHE. Reproduced with permission from Ref. 111. Published by The Royal Society of Chemistry...... 41

Figure 2-1. (a) XRD patterns of as-prepared layer perovskites and their corresponding proton-exchanged products. (b) TEM (left) and AFM (right) images of the exfoliated nanosheets...... 52

Figure 2-2. UV-vis absorption spectra of nanosheet multilayers deposited on quartz substrates...... 53

Figure 2-3. (a) Original and scattering-corrected Tauc plots for one to ten layer films of Sr2Nb3O10 nanosheets. (b) Scattering-corrected Tauc plots for one- to-ten-layer films of nanosheets of different compositions...... 55

Figure 2-4. (a) Equivalent circuit (Randles circuit) used for fitting of EIS data. Nyquist and (b) and Bode plots (c) of EIS data from five-layer Ca2Nb3O10 nanosheet films on FTO at -0.7 V in a pH 7.8 aqueous electrolyte...... 57

Figure 2-5. Mott-Schottky plots of five-layer Ca2Nb3O10 nanosheets and a bare FTO substrate at pH = 7.8. Shaded regions 1, 2, and 3 correspond to potential ranges in which the capacitance is dominated by the FTO substrate, the FTO/nanosheet interface, and the semiconducting nanosheet film...... 58

Figure 2-6. Mott-Schottky plots of gold-coated slides with different numbers of Ca2Nb3O10 nanosheet layers at pH 6.7...... 59

Figure 2-7. (a) Flat-band potentials of different nanosheet films at various pHs on FTO substrates. Dots are the measured data; solid lines are estimated Fermi level potentials based on Equation (1). (b) DFT-calculated energy diagram including conduction band minimum, valence band maximum, and band gap values for HCa2Nb3O10, HSr2Nb3O10, and HCa2Ta3O10 nanosheets...... 61

Figure 3-1. (a) UV-vis absorption spectra and surface coverages of sensitizers (Γ, inset) for the three types of electrodes under investigation. (b) Chronoamperometric measurement at an applied bias of 0.2 V. The measurement was conducted using hydroquinone (50 mM) as the electron donor under white light illumination (150 W Xenon lamp, > 410 nm, 100 mW/cm2)...... 72

Figure 3-2. (a) PEIS Nyquist plots for a TiCl4-treated electrode at different illumination intensities. Inset: the equivalent circuit used for data fitting. (b) Semi-log plot of the open-circuit potential as a function of the injected electron concentration. Dashed lines are linear fits using (Eq. 3-6)...... 74

Figure 3-3. IMVS Nyquist (a) and Bode (b) plots for the pristine electrode at various illumination intensities. The four quadrants are indicated by Roman numerals...... 75

Figure 3-4. (a) Open-circuit potentials of the photoelectrodes under illumination at different intensities. (b) Recombination rates determined from IMVS at different open-circuit potentials under illumination. Dashed lines are linear fitting results...... 76

Figure 3-5. Comparison of recombination rates measured by IMVS and PEIS for a TiCl4- treated electrode...... 79

Figure 3-6. (a) bleaching recovery kinetics and (b) the corresponding rate constant distribution at 470 nm for the three types of electrode at a probe light intensity of 11.64 mW/cm2...... 81

Figure 4-1. Proposed electron transport and recombination scheme for the DSPEC photoanode...... 88

Figure 4-2. EIS Nyquist plot for a RuP-sensitized TiO2 electrode measured in 0.1 M NaAc/HAc (pH 4.7) and 3mM HQ at an applied bias of 0.3 V vs Ag/AgCl. The light intensity was 4.1 mW/cm2 (470 nm). Inset shows the equivalent circuit used for fitting, and the fitted values are Rs=177.8 Ω, Rp=30.5 kΩ, and C=5.2 µF ...... 93

Figure 4-3. IMPS Nyquist plot (a) and Bode plots (b) for a RuP-sensitized TiO2 electrode in 0.1 M NaAc/HAc and 3mM HQ at an applied bias of 0.3 V vs Ag/AgCl. The light intensity was 4.1 mW/cm2 (470 nm). Blue and orange represent experimental and simulated data, respectively...... 93

Figure 4-4. Left: IMPS Nyquist plots measured at different NaClO4 concentrations (data normalized to the maximum of real current). Right: simulated IMPS Nyquist plots using different values of 푃...... 94

Figure 4-5. Semi-log plots of concentration profiles for injected electrons ( 푛0) and oxidized sensitizer molecules (푅푢푃0) inside the TiO2 film under steady-state illumination. Incident light is from the FTO substrate side (film thickness 0)...... 95

Figure 4-6. Simulated steady-state concentration profiles (a), steady-state current densities (b), IMPS Nyquist (c) and Bode (d) plots at various light intensities. The blue arrows indicate the direction of increasing light intensity...... 96

Figure 4-7. Simulated steady-state concentration profiles (a), steady-state current densities (b), IMPS Nyquist (c) and Bode (d) plots at various 퐷푛 values. The blue arrows indicate the direction of increasing 퐷푛...... 97

Figure 4-8. Simulated steady-state concentration profiles (a), steady-state current densities (b), IMPS Nyquist (c) and Bode (d) plots at various 푘2 values. The blue arrows indicate the direction of increasing 푘2...... 98

Figure 4-9. Simulated steady-state concentration profiles (a), steady-state current density (b), IMPS Nyquist (c) and Bode (d) plots at various 푘3 values. The blue arrows indicate the direction of increasing 푘3...... 99

Figure 4-10. Experimental IMPS Nyquist plots for a RuP-sensitized TiO2 electrode in 0.1 M NaAc/HAc and HQ of different concentrations at an applied bias of 0.3 V vs Ag/AgCl. The light intensity was 4.5 mW/cm2 (470 nm)...... 100

Figure 5-1. Solid-state dye sensitized solar cell as a buried junction for visible-light photoelectrochemical water oxidation...... 105

Figure 5-2. Cross-sectional SEM image of an as-prepared ss-DSSC...... 109

Figure 5-3. Characterization of ss-DSSCs with and without ALD TiO2 thin film overlayers. a) Two-electrode J-V curves of pristine ss-DSSC (blue) and ss-DSSC/ALD TiO2 (red) electrodes in air (1 and 3) and in contact with aqueous solutions (2 and 4), respectively. Inset digital images show the surface morphologies of the corresponding electrodes after testing. Scan rate: 20 mV/s. b) Chronoamperometric measurements of pristine ss-DSSC and ss-DSSC/ ALD TiO2 in solution under short-circuit photoelectrochemical conditions. Light source: 100 mW cm-2 xenon lamp with a 410 nm long-pass filter and an AM1.5 filter...... 111

Figure 5-4. Characterization of a ss-DSSC/ALD/IrOx electrode as a photoanode. a) Measured (line) and calculated (square) linear sweep voltammetric curves of the ss-DSSC/ALD/IrOx electrode. Scan rate: 20 mV/s. b) three-electrode photocurrent measurement at an applied bias of 0.62 V vs Ag/AgCl. Light source: 100 mW/cm2 xenon lamp with a 410 nm long-pass filter and an AM1.5 filter...... 113

Figure 5-5. a) Left: Energy diagrams at the semiconductor-electrocatalyst (SC-EC) interface under illumination for a buried and an adaptive junction, including the semiconductor conduction band (CB) and valence band (VB), electron quasi-Fermi level (En), hole quasi-Fermi level (Ep), EC energy level (EEC). Right: electrocatalyst potential (VEC) change as a function of the semiconductor potential (VSC). Note that the applied bias (V1 to V2) in a buried junction drops at the Helmholtz layer (VH) and in an adaptive junction, the applied bias changes band bending. b) Catalyst potential change (vs Ag/AgCl) recorded during the LSV measurement shown in Figure 5-4a (red line)...... 114

Figure 5-6. a) Chronoamperometry measurement of an ss-DSSC/ALD/IrOx photoelectrode for one hour at an applied bias of 0.62 V vs. Ag/AgCl. Light source: 100 mW cm-2 xenon lamp with a 410 nm long-pass filter and an AM1.5 filter. b) IPCE as a function of wavelength (blue) for an ss-DSSC/ALD/IrOx electrode at an applied bias of 0.62 V vs. Ag/AgCl. UV-vis absorption spectrum of N3 dye- sensitized TiO2 electrode (red). Error bars indicate the standard deviation based on the average of three measurements...... 117

xii

Figure A-1. XRD patterns of 10-layer PDDA/nanosheets on quartz before (black) and after (red) three-day UV exposure. Note that the reflections close to 6° in Ca2NbxTa3- xO10 samples before UV irradiation are due to the presence of unexfoliated particles...... 124

Figure A-2. Top (a) and side (b) view of the optimized structures in DFT calculations...... 125

Figure A-3. Band structures for different nanosheet compositions. Fermi level (red dashed line) is set at 0 eV...... 126

Figure A-4. Partial density of states plotted relative to the Fermi level for different nanosheets. Oxygen 2p (red) and the d states from B site metal (blue) ions are highlighted...... 127

Figure A-5. Nyquist (left) and Bode (right) plots of five-layer TiOx nanosheets on FTO substrate at -0.7 V in pH = 6.8 electrolyte. Red curves are the data fitted with a Randles circuit. Similar deviations of EIS spectra from the Randles circuit fits were observed for TiNbO5 nanosheets. TiOx and TiNbO5 nanosheets were prepared as reported elsewhere.267,268 ...... 127

Figure B-1. XRD profiles (a) and SEM images (b) for the three types of electrode under investigation...... 128

Figure B-2. Resistance (a) and capacitance (b) of the TiO2 films determined from PEIS at varoius open-circuit potenitals under illumination. Error bars indicate fitting errors...... 128

Figure B-3. PEIS Nyquist plots of (a) an unsensitized TiO2 electrode and (b) a TiCl4-treated electrode measured with 470-nm LED illumination at an intensity of 33.8 mW/cm2 from 1500 to 10 Hz...... 129

Figure B-4. Calculated time-domain (a) and frequency-domain (b) profiles for the modulated light I and the electron concentration n. Calculated Nyquist (c) and Bode (d) plots with different values of 훽. The parameters used in calculation are shown in (b)...... 131

Figure B-5. Bleach recovery kinetics (a) and the corresponding rate constant distribution (b) at 470 nm for the three types of electrodes at different probe light intensities. The fitting quality (adj.-R2) and the KWW time constants (see (Eq. 3-16)) with fitting errors are provided in each TA graph. Note that rate distribution analysis is only applied to data of R2 > 0.9...... 132

Figure C-1. J-V curves of a solid-state N3-sensitized solar cell with different aging time under AM 1.5 G illumination with a 410 nm long pass filter (100 mW cm-2)...... 134

xiii

OMeTAD and TiO2/N3/spiro-OMeTAD, and in spiro-OMeTAD , respectively. b) Calculated series resistances obtained by fitting EIS data to the equivalent circuit shown in the inset. Rct, Rct2, and Rct3 represent electron transfer resistances at the interface of Au/spiro-OMeTAD, the interface of TiO2/N3/spiro-OMeTAD, and in the spiro-OMeTAD hole conductor layer, respectively...... 135

Figure C-3. a) Current density as a function of time during 1.2 V electrodeposition of iridium oxide films on a ss-DSSC/ALD in 0.4 mM hydroxyiridate solution 2- (prepared by alkaline hydrolysis of [IrCl6] ) at pH 8.0. Inset: photographs taken before and after electrodeposition. b) Photograph of the experimental setup for electrodeposition of IrOx onto gold with a mechanic stirrer. WE1 and WE2 are the working electrodes connecting FTO and Au, respectively. RE and CE are the reference electrode and counter electrode. WE2 was used as the working electrode during the deposition process. c) Cross-sectional SEM image of an iridium oxide film deposited on a ss-DSSC/ALD electrode...... 135

Figure C-4. Cyclic voltammogram of a ss-DSSC/ALD/IrOx electrode (the Au layer was used as the working electrode). The measurement was recorded at a scan rate of 5 mV/s...... 136

Figure C-5. Top: Collector-generator measurement of the O2 Faradic efficiency of a ss- DSSC/ALD/IrOx electrode (blue) illuminated with 19.4 mW cm-2 470 nm blue light from 300 s to 600 s at a bias of 0.62 V vs. Ag/AgCl, and a collector electrode (red) biased at -0.3 V vs Ag/AgCl. The experiment was performed in 0.1 M phosphate buffer at pH 6.7 with 0.1 M NaClO4. Blue and red shaded regions represent QG and QC, respectively. Bottom: Collector-generator calibration measurement with a FTO/IrOx generator electrode at an applied bias of 1.12 V vs. Ag/AgCl from 300 s to 600 s (blue solid line). The red solid line is recorded from a Pt oxygen sensing electrode biased at -0.3 V vs Ag/AgCl. The experiment was performed in 0.1 M phosphate buffer at pH 6.7 with 0.1 M NaClO4. The blue and red shaded regions represent charges passing the generator and the collector, respectively...... 137

xiv

List of Tables

Table 1-1. Photoanode sensitizers and assemblies for water-splitting dye cells...... 10

Table 1-2. Photocathode sensitizers and assemblies for water-splitting dye cells...... 32

Table 3-1. Fitting results and calculated parameters...... 78

Table 4-1 Simulation parameters used to calculate IMPS ...... 91

Table A-1 Compositions (in atomic percentages) of Ca2Nb2.25Ta0.75O10 and Ca2Nb1.5Ta1.5O10 nanosheets determined by EDS. Carbon and nitrogen reflect the presence of organic material (such as TBA+ ions), and Si signals come from the substrate...... 124

Table A-2. DFT-calculated work functions for different nanosheets...... 126

Table C-1. Photovoltaic parameters from J-V characteristics of solid-state N3 sensitized solar cell with different aging times ...... 134

Table C-2. Calculated current densities (J) in Figure 5-4a with the measured Au (or IrOx) potential (VAu) vs the FTO potential (VFTO) from Figure 5-5b and the voltage- current curve of Figure 5-3a line 4.(△V= VFTO - VAu) ...... 136

Acknowledgements

"A graduate degree in chemistry in one of the few times in life that you'll be paid to improve yourself!" I came across this encouraging saying when I was applying for the chemistry graduate program of a university at a cold winter night five years ago. Although I would not be surprised to see similar quotes from fortune cookies today, I find it well summarizes my five-year PhD life, because I have found it fruitful: I have gained knowledge in chemistry; I have begun to understand the hearts of people; I may have also had a glimpse of what the future has in store for me.

Being a better me and getting to know what I would like to be would never be possible had I not been associated with all the people here I met at Penn State. I would like to extend my sincere gratitude to Dr. Tom Mallouk for giving me full freedom in research and guiding me towards what a good scientist should be like -- embrace the fundamentals! I would like to thank Dr. Ray Schaak,

Dr. Ben Lear, and Dr. Noel Giebink for serving on my thesis committee and supporting my academic pursuit. I would like to thank the whole Mallouk group, past and present, for the company in my good and bad times and showing me what friends are for. I would also like to express my deepest appreciation and love to my family and friends for all the understanding and mental support.

This has been a wonderful period of my life thanks to all of you.

xvi

Chapter 1.

Water-splitting Dye-sensitized Solar Cells

Pengtao Xu,a Nicholas S. McCool,a Thomas E. Mallouka,b,c

a. Department of Chemistry, The Pennsylvania State University

b. Department of Physics, The Pennsylvania State University

c. Department of Biochemistry and Molecular Biology, The Pennsylvania State University

Published in Nano Today 2017, 14, 42–58.

1.1. Introduction

Solar energy conversion has the potential to provide a practically limitless source of renewable and clean energy. Although in past decades solar power could not compete on a cost basis with electricity generation from fossil fuel, nuclear, or hydroelectric sources, the cost of photovoltaic modules has steadily dropped since the 1970s at an average rate of about 15% per year 1. As inexpensive photovoltaics achieve grid parity in more and more of the developed world, they will bring about sweeping changes in the way electricity is produced, distributed, and used. At present, electricity production accounts for about 39% of primary energy use in the United States, which is mostly derived from fossil fuel resources. The remaining 61% of the U.S. energy economy, which is dominated by the transportation and industrial sectors, relies heavily on fossil fuels. All together the combustion of fossil fuels accounts for about 80% of the world’s energy economy 2. Better energy storage technologies have the potential to shift the balance of energy use towards electricity

3,4, but the need for combustible fuel will be with us for some time for transportation, heating, manufacturing, and many other uses.

The electrolysis of water to produce hydrogen is a mature technology, and hydrogen derived

from photovoltaic power can be used directly or combined catalytically with CO2 to make high energy density liquid fuels. While the energy efficiency of electrolyzers is high, their capital cost limits the commercial viability of power-to-gas technologies 5. Additionally, cost targets for solar fuels are much more challenging than they are for solar electricity. Solar electricity can compete with power generation from fossil fuels because the latter process is inherently inefficient due to

Carnot cycle losses. However, fuel derived from solar energy must compete directly with the heating value of fossil fuel, meaning that cost parity of commodity fuels with hydrogen from photovoltaic-electrolyzer systems would require a substantial lowering of electrolyzer cost.

An alternative pathway to solar fuels involves the direct production of fuel through photocatalysis or photoelectrochemistry (See Figure 1-1). Although none of these schemes are yet technologically viable, they could potentially offer a low-cost alternative to a stand-alone electrolyzer by coupling catalytic fuel generation directly to the light absorber. In these systems, light is absorbed by a molecule or semiconductor, electrons and holes are separated over short distances – typically nanometers to microns – and chemical reactions occur at solid-liquid or solid- vapor interfaces. Most typically, the reaction of interest is water splitting, but there has also been

6–11 growing interest in the photochemical splitting of CO2 to carbon-containing fuels and oxygen ,

12,13 the light-driven reverse water gas shift reaction (CO2 + H2 = CO + H2O) , and the reaction of nitrogen with water to produce ammonia and oxygen 14,15.

Figure 1-1. Solar fuel production by (a) semiconductor photoelectrochemistry, e.g., in the 16 Fujishima-Honda cell , (b) photocatalysis, e.g., at a Rh-catalyzed Ga1-xZnxN1-xOx particle (adapted

In principle, the three kinds of solar fuel systems shown in Figure 1-1 are all subject to the same thermodynamic limit of ~33% efficiency 19,20, and better utilization of the solar spectrum can be achieved if two light absorbers are arranged in series (as they are in natural photosynthesis) to drive the energetically demanding water splitting reaction 21,22. In practice, some energy must be given up for catalytic hydrogen and oxygen evolution to proceed at useful rates, and some energy is lost to series resistance. If these combined losses (along with membrane polarization losses 23,24) do not exceed about 1.0 V, then power conversion efficiencies of 10-15% can in principle be achieved 21, and in fact have been demonstrated in model systems 25–28. A recent techno-economic analysis of photocatalytic and photoelectrochemical solar fuel systems suggests that they can be economically viable at power conversion efficiencies above about 10% 29. It is important to recognize from these analyses however that solar fuel performance targets can be met only if the quantum efficiency (the fraction of photons that produce current or a chemical reaction) is close to unity. Although much progress has been made towards achieving this goal in photovoltaic-electrolyzer systems, most semiconductor-based photoelectrochemical systems (Figure 1-1a) are poorly matched to the solar spectrum or are unstable in water, and photocatalysts (Figure 1-1b) typically have low quantum yields because of uncontrolled charge recombination. Recent reviews describe these challenges and the extensive work that has been done to address them 30,31. Here we focus instead on fledgling molecule-based water-splitting systems (Figure 1-1c). Although they are at a primitive stage of development technologically, in principle molecule-based (or artificial photosynthetic) systems

offer flexibility in the design of molecular light absorbers to cover the full solar spectrum, and they can take advantage of the high quantum yield of light-driven charge separation that has already been achieved at dye-semiconductor interfaces.

1.2. Dye sensitized solar cells: regenerative and photosynthetic

Because molecular water splitting cells are currently based on the architecture of the dye- sensitized solar cell (DSSC), it is important to review the basic operating principles of the latter.

The DSSC (also known as the Grätzel cell) is a regenerative photoelectrochemical cell, i.e., a liquid- junction cell in which the anode and cathode reactions are the reverse of each other. The DSSC

utilizes high surface area electrodes – typically ~10 micron-thick films of anatase TiO2 nanoparticles – to adsorb a monolayer of dye, which sufficiently harvests most of the incident light.

Early designs of DSSCs used aqueous electrolytes 32, but a breakthrough came in 1991 when

O’Regan and Grätzel introduced non-aqueous iodide-based dye cells 33. Subsequent optimization of dyes, light scattering layers, electrode blocking layers, and the redox couple in non-aqueous

DSSCs have resulted in quantum efficiencies close to unity over much of the visible spectrum and champion cell solar-to-electric power conversion efficiencies as high as 14.1% 34.

Figure 1-2. The architecture of regenerative (left) and water-splitting (right) DSSCs. The regenerative DSSC is fabricated as a sandwich cell with a short (~100 µm) electrolyte path between the photoanode and dark cathode. In the water-splitting cell, the anode and cathode are typically separated by a membrane or frit (not shown) in order to enable separation of photogenerated oxygen and hydrogen. Adapted with permission from Ref. 35. Copyright 2009 American Chemical Society.

The DSSC is compared to an early design of the water-splitting dye cell in Figure 1-2. In the water-splitting version, an aqueous electrolyte is used, and a catalyst for the oxygen evolution reaction (OER) – either a catalytic nanoparticle or a molecular catalyst that is bound to or co- adsorbed with the sensitizer – replaces the iodide/triiodide redox shuttle in the DSSC. Oxygen is evolved at the dye-sensitized photoanode and water is reduced to hydrogen at a dark catalytic cathode. Because the photoinjected electrons in the anatase film are not sufficiently reducing to generate hydrogen from water, a minimum cathodic bias of 200-300 mV is needed, as in the original

Fujishima-Honda cell 16 and other water splitting cells based on visible light-absorbing anode

36–40 materials such as Fe2O3, WO3, and BiVO4 . Thus the water splitting dye cell can be thought of as a hybrid of the Fujishima-Honda cell and the Grätzel cell.

Figure 1-3. Electrochemical potential diagrams for (a) Fujishima-Honda, (b) Grätzel, and (c) water- splitting dye-sensitized solar cells. Solid lines indicate forward electron transfer pathways. Charge recombination pathways are shown as dashed lines.

The relevant electrochemical potential diagrams for these three kinds of solar cells are shown schematically in Figure 1-3. In the Fujishima-Honda cell and related cells based on wide bandgap oxide semiconductor electrodes, light absorption creates strongly oxidizing valence-band holes that rapidly oxidize water. The internal quantum yield for water oxidation can be high, especially at high bias where electrons and holes are efficiently separated, but utilization of the solar spectrum is poor. In the DSSC, the quantum yield is also high, a consequence of sub-picosecond electron

injection from the photoexcited dye into the TiO2 conduction band, followed by reduction of the oxidized dye by iodide on the nanosecond timescale. Dye reduction by iodide is orders of

magnitude faster than back electron transfer from TiO2 to the oxidized dye, and because back

- 41 electron transfer to I3 is also slow , the forward electron transfer pathways dominate over the

reverse pathways. In contrast, in the water-splitting DSSC, catalytic water oxidation is typically slow (millisecond timescale) because of the weak driving force for the reaction. Low quantum yields for water splitting – typically 1-2% - are primarily a consequence of the fast kinetics of charge recombination, which competes effectively with the catalytic oxidation of water. Since the introduction of the water splitting dye cell in 2009, much of the effort in re-design of the photoanode has focused on improving the quantum yield by slowing down back electron transfer and/or accelerating the water oxidation process.

1.3. Dye-sensitized photoanodes for water-splitting cells

1.3.1. Electron and proton transfer at dye-sensitized photoanodes

Because of its importance in aqueous electrochemical systems, the catalytic four-electron oxidation of water to oxygen has been well studied and many different OER catalysts have been developed. Following the initial demonstration of visible-light water splitting in a DSSC (using

. 35 colloidal IrOx nH2O covalently bound to the sensitizer molecule, Figure 1-2) , a number of groups have tested different colloidal and molecular water oxidation catalysts in similar electrode architectures. The results are summarized in Table 1-1.

Figure 1-4. Sensitizers and sensitizer-catalyst assemblies (catalysts are colored red) that have been used in dye-sensitized photoanodes. See Table 1-1 for references.

Table 1-1. Photoanode sensitizers and assemblies for water-splitting dye cells.

Faradaic Photocur Steady-State IPCE efficiency of rent Photocurrent Reference Semiconductor Dye OER Catalyst photocurrent (waveleng O2 Stability Test conditions (μA/cm2) th) generation (h) 0.5% 35 Xe lamp (>410 nm), pH TiO2 RuP2 IrO2·nH2O 10-30 (0.58 V) 20% 4 (450 5.75 nm) 1.7% Xe lamp (100 42 + a b 2 TiO2 RuC1 [Mn4O4L6] 3-15 (0 V vs Pt) 93% 2 (430 mW/cm , 290-750 nm) nm), pH 7 (no buffer) White light emitting

43 40 diode array (100 TiO2 RuP1 Ru1 n/a 1 n/a (0 V vs Pt)b mW/cm2), pH 7 (no buffer) white light (~200 44 2 TiO2 ZnPor Ir1 30 (0.91 V) n/a n/a n/a mW/cm , >400 nm pH7 no buffer) Poly- 45 Xe lamp (>420 nm), TiO2 heptazin IrO2·nH2O 120 (1.12 V) n/a 1.5 n/a pH7 e Poly- 5.5% 46 TiO2 heptazin Co-Pi 75 (1.0 V) n/a 3 (350 450 nm, pH7 e nm) Xe lamp (100 47 2 TiO2/Nb2O5 RuP2 IrO2·nH2O 15-20 (0.55 V) 109% n/a n/a mW/cm , >410 nm), pH 5.75

48 IrOx Xe lamp ( >410 nm), TiO2 RuP3 80 (0.55 V) 85% n/a 5% nanoparticles pH 5.8

(450 nm, estimat ed) {Ru4O4(OH)2 49 Xe lamp (420-470 nm, TiO2 RuP1 (H2O)4}(γ- 15 (0.55 V) n/a 0.33 0.20% 2 10- 30 mW/cm ), pH 5.8 SiW10O36)2] 14% 50 Xe lamp(>400 nm, 300 TiO2 RuP1 Ru4 1700 (0.60 V) 83% n/a (450 mW/cm2), pH 6.8 nm)

4.4% 2 51 LED (72.5 mW/cm , nanoITO/TiO2 RuP9-Ru8 80 (0.47 V) n/a 0.5 (445 445 nm), pH 4.5 nm) 2.4%

52 Sintered (410- Xe lamp(>410 nm), pH TiO2 RuP1 80 (0.70 V) 98% n/a IrO2 700 6.8 nm) 1400 with Ru2-3, 53 Ru2-1, Ru2- Xe lamp(>400 nm, 300 TiO2 RuP1 700 with Ru2-3 n/a n/a n/a 3 mW/cm2), pH 6.4 (0.58 V) 4.1% 54 Xe lamp(>400 nm, 300 TiO2 RuP4 Ru3 480 (0.58 V) n/a n/a (450 mW/cm2), pH 6.4 nm) 0.12% Xe lamp(315-710 nm, 55 n/a PMPDI CoOx 150 (1.61 V) 80±15% n/a (475 100 mW/cm2), pH7 nm) 1.85% 56 b Xe lamp(>400 nm, 100 TiO2 RuP1 Ru4 40 (0 V vs Pt) 73.80% 0.23 (455 mW/cm2), pH 6.4 nm)

50 with PPor, white light (~200 57 2 SnO2 PPor Ir1 20 with PPor+Ir1 n/a n/a n/a mW/cm , >400 nm (0.92 V) pH7 no buffer) [{Ru4O4(OH) 58 c LED light (455nm, 33 TiO2 RuP5 2(H2O)4}(γ- 54.8 (0.55 V) 86% 5 0.392% 2 10- mW/cm ), pH 5.8, SiW10O36)2] Poly- 59 Xe lamp(>420 nm), TiO2 heptazin Co(II) ion 90 (1.06 V) 38.60% 1 n/a pH6 e White light (~200 60 2 TiO2 RuP1 Ru9 20 (1.16 V) 16.80% 6 n/a mW/cm , >380 nm), pH 8.8 TMP, DMP, MMP, 61 citrate- Xenon lamp (>410 nm TiO2 TTP, PAP, 15 (0.71 V) 102 ±5% n/a <0.05% capped IrOx or > 590 nm), pH 6.8 DMEP, MDC, MDCE

62 b LED light (445 nm, SnO2/TiO2 RuP9-Ru8 100 (0.6 V vs. Pt) 41% 0.42 n/a 46.2 mW/cm2), pH 7

63 White LED (>400 nm, TiO2 L0 Ru10 300 (0.62 V) 73% 1 n/a 100 mW/cm2), pH 7 3.75% White LED light (100 64 2 SnO2/TiO2 RuP8-Ru1 300 (0.95 V) 22% 0.5 (435 mW/cm , >400 nm), nm) pH 7

65 IrOx LED light (450nm, 14.5 nanoITO/TiO2 RuP7 150 (0.54 V) n/a 2 n/a nanoparticle mW/cm2), pH 5.8

66 White light (> 400 nm, SnO2/TiO2 TPA Ru7 100 (0.85 V) 8.2% n/a n/a 100 mW/cm2), pH 7

9.5% White light (300 67 2 TiO2 RuP1 Ru5 1200 (0.60 V) 81% n/a (450 mW/cm , >400 nm), pH nm) 6.8 17% (424 White light (35 68 nm) 2 TiO2 ZnPor-Ru1 100 (0.23 V) 33% 1 mW/cm , >380 nm), 5.9% pH7.3 (564 nm) 7% Xe lamp (300 69 2 TiO2 RuP6 Ru6 800 (0.61 V) 72% n/a (460 mW/cm , >400 nm), pH nm) 7 with 5% CF3CH2OH White light (100 70 2 SnO2/TiO2 RuPS Ru8 10 (0.85 V) 22% n/a n/a mW/cm , >400 nm), pH7

71 White light (100 SnO2/TiO2 RuP9-Ru1 850 (0.63 V) 74% n/a n/a mW/cm2), pH5.7

72 400 (1.06 V, pH 7) 30% (pH 7) White light (100 SnO2/TiO2 RuP8-Ru11 n/a n/a 500 (1.17 V, pH 9) 45% (pH 9) mW/cm2, >400 nm)

*See Figure 1-4 and Figure 1-5 for structure codes.

**Steady-state current and stability data are from chronoamperometry experiments (typically recorded after 10 to 30s of illumination). Potentials in bracket are the applied potential vs RHE unless otherwise noted. Potential conversion follows these equations except where the reference potential is given in the original paper:

1) E(V vs. RHE) = E(V vs. Ag/AgCl 3M KCl/NaCl)+0.210+0.0592×pH

2) E(V vs. RHE) = E(V vs. Ag/AgCl sat.)+0.197+0.0592×pH

3) E(V vs. RHE) = E(V vs. SCE)+0.241+0.0592×pH

4) E(V vs. RHE) = E(V vs. NHE) +0.0592×pH

Equation 1) was used when Ag/AgCl filling solution is not indicated.

***Stability duration results are tabulated based on the longest test time (>10 min) indicated in the literature. a: L = bis(methoxyphenyl)phosphonate b: Measurement was done in a two-electrode configuration. c: Internal quantum efficiency.

Figure 1-5. Molecular OER catalysts used in dye-sensitized water splitting cells. See Table 1-1 for references.

These experiments revealed interesting similarities in the electrochemical behavior of photoanodes – fabricated from different dyes and catalysts – that were initially difficult to rationalize (Figure 1-6). When the photoanodes were abruptly irradiated with simulated sunlight, the initial photocurrent densities were in the range of 15-200 µA/cm2, but they decayed over a timescale of tens of seconds to a fraction of their initial value. Interestingly, dye decomposition or desorption could be ruled out as the principal source of photocurrent decay because most of the initial photocurrent could be recovered after switching the light off for a few seconds.

Figure 1-6. Photoelectrochemical transients obtained with different dyes and water oxidation catalysts in water-splitting dye cells. Reproduced from Ref. 44 (with permission of The Royal

Society of Chemistry), Ref. 43 (with permission of The Royal Society of Chemistry), Ref. 49 (Copyright 2013 American Chemical Society, adapted with permission), Ref. 51, and Ref. 54 (Copyright 2014 American Chemical Society, adapted with permission).

Generator-collector experiments showed that despite the transient nature of the photocurrent, the

Faradaic efficiency of oxygen evolution was close to unity 48,52. Electrodes that were tested in a regenerative DSSC architecture (i.e., in non-aqueous iodide solutions) gave an order of magnitude higher current density and showed minimal photocurrent decay 47.

The kinetic origin of the low quantum yield and transient nature of the photocurrent in water- splitting dye cells was clarified through a series of spectroelectrochemical and transient spectroscopy experiments 52,73–75. The detailed kinetic framework was adapted from earlier studies of the DSSC 76, and is illustrated schematically in Figure 1-7.

Figure 1-7. Kinetic scheme for the photoanode of a water-splitting dye-sensitized solar cell containing an adsorbed molecular sensitizer and a colloidal IrO2 catalyst sintered to the high surface 73 area anatase TiO2 film. Reproduced with permission from Ref. . Copyright 2014 American Chemical Society.

Electrodes fabricated with sintered IrO2 catalysts (Figure 1-7) enabled systematic studies of electron transfer processes as a function of the dye coverage, catalyst loading, solvents used to adsorb the dye, and incident light intensity. Following photoinjection of electrons by dye molecules,

holes diffuse across the surface to reach the IrO2 catalyst. The regeneration of reduced dye molecules occurs in competition with charge recombination and electron scavenging by Ir(IV). The

cross-surface diffusion coefficient (Dapp) was measured by a spectroelectrochemical potential step

77 method , and charge recombination (krecomb) and electron scavenging (kscav) rate constants were obtained from transient photovoltage measurements. Interestingly, the photocurrent for water splitting was strongly dependent on the solvent and conditions used to adsorb the dye. Exposure of the electrodes to acidic, aqueous solvents resulted in low photocurrents despite relatively high

values of Dapp. Spectroelectrochemical and electrochemical impedance measurements showed that intercalated protons helped to trap photoinjected electrons at sites where they could readily combine with oxidized dye molecules 74, consistent with previous studies of proton intercalation in

78 nanoporous TiO2 films .

The impact of proton-induced trap states on electron transfer at dye-sensitized TiO2 photoanodes was recently studied by McCool et al. 75 using transient absorption spectroscopy (TAS) and time-resolved terahertz spectroscopy (TRTS). Lower electron injection yield from TRTS and faster back electron transfer rates from TAS were found for photoanodes sensitized from perchloric acid, relative to those sensitized from anhydrous ethanol. This contrast implies the formation of trap states during exposure to protic solvents. Although these trap states can be temporarily

removed when the electrodes undergo mild heating (80 ℃) under vacuum 75, proton intercalation can still occur slowly during photoelectrolysis 73.

Using rate constants from transient photovoltage measurements, Swierk et al. were able to construct a kinetic model (equation 1) that could accurately fit the photocurrent transients observed with the water-splitting dye cell.

퐸푐−퐸푡 휕푛푡 − = 퐺 − 푘 푒 푘푇 (푁 − 푛 )푛 − 푘 푛 3+푛 − 푘 푛 4+푛 (Eq. 1-1) 휕푡 푑푒푡푟푎푝 푐푏 푐푏 푡 푟푒푐표푚푏 푅푢 푡 푠푐푎푣 퐼푟 푡

In this equation nt and ncb are the densities of trapped and conduction band electrons, G is the

electron generation rate (which is proportional to the light flux), nRu3+ and nIr4+ are the densities of

oxidized dye molecules and electron scavenger sites, Ncb is the total density of conduction band

states, and Ec and Et are conduction band edge and trap energies, respectively. Representative data

and fits are shown in Figure 1-8, along with a plot of nRu3+ and nt as a function of irradiation time.

The model shows that the photocurrent (which is proportional to ncb) decays over a period of tens

of seconds as nRu3+ and nt increase, eventually reaching a photostationary state.

Figure 1-8. Left: Photocurrent transients from photoanodes sensitized from DMSO and ethanol solutions. Black lines are fits to (Eq. 1-1). Right: Calculated number density of conduction band

electrons (ncb), trapped electrons (nt), and oxidized sensitizer molecules (Ru(III)) from photocurrent model of the ethanol-sensitized electrode. Reproduced with permission from Ref. 73. Copyright 2014 American Chemical Society.

It is clear from this analysis that strategies that can lower krecomb or accelerate Dapp or ktrans should improve the efficiency of water splitting. The recombination rate can be lowered by increasing the distance between the sensitizer and electron trap sites, or by preventing proton intercalation. Indeed, the most efficient cells reported, which have quantum efficiencies of ~15% and photocurrent densities of 2-3 mA/cm2 50,54, embody some of these strategies. Gao et al. demonstrated that the photocurrent increases with increasing pH 50, and Zhao et al. showed that higher steady-state current could be obtained by incorporating an electron transfer mediator between the sensitizer and catalyst sites 48.

1.3.2. Core-shell photoanode architectures

As noted above, lowering the back electron transfer rate from the semiconductor to the oxidized sensitizer molecule on its surface should increase the quantum efficiency of oxygen evolution in water-splitting dye cells. Two kinds of core/shell structures have been fabricated and tested in water-splitting dye cells. In the first type, the shell is applied before the sensitizer is adsorbed, and in the second, the adsorbed sensitizer is covered by the shell material (Figure 1-9). While the first type is primarily targeted towards slowing down back electron transfer, the second type of core- shell structure seeks to address the issue of dye stability.

Figure 1-9. (a) Cartoon depicting two types of core-shell architecture. (b) TEM image of a core- shell structure made from 75 ALD cycles of TiO2 grown onto a SnO2 film on FTO glass. Reproduced from Ref. 62.

Core-shell-sensitizer architectures

In solid-state semiconductor solar cells, an insulating layer between the metal top contact and the semiconductor is often used to inhibit interfacial charge recombination 79. This strategy has been adapted to the DSSC by using a conformal coating of insulating overlayers on the surface of the mesoporous oxide semiconductor 80–82. The difference in conduction band edge potentials at the core-shell interface creates an energy barrier for charge recombination, and thus increases the charge transfer efficiency, although thicker shells may lower the effective surface area for dye loading. The concept of using core-shell electrodes has recently been applied to water splitting dye cells, where improved electron transfer kinetics have been reported 47,51,64–66.

47 Lee et al. reported RuP2-sensitized TiO2/ZrO2 and TiO2/Nb2O5 core-shell photoanodes prepared by surface sol-gel deposition of the wide bandgap oxide. TAS showed that back electron transfer from the semiconductor to the oxidized dye was slowed by a factor of 2-3. Incorporation of the insulting layer was also found to increase the photocurrent of the photoanode. However, the

ZrO2 and Nb2O5 films prepared by this method were not uniform and uncoated areas could be seen using high-resolution transmission electron microscopy (TEM). More conformal core-shell electrode films were later reported by the Meyer group using atomic layer deposition (ALD) 51,62,65.

Instead of using TiO2 as the core material, they used tin-doped indium oxide in its nanoparticle

form (nanoITO) as a conductive mesoporous support, with the TiO2 shell material prepared by

ALD. This nanoITO/TiO2 core-shell electrode, when sensitized with RuP7 and IrOx nanoparticles,

65 gave higher photocurrent relative to mesoporous TiO2 electrodes . Further improvement was

achieved with sensitized SnO2/TiO2 photoanodes. Testing under the same conditions with the

RuP9-Ru8 chromophore-catalyst assembly, Alibabaei et al. 62 reported that the photocurrent from

the nanoITO/TiO2 core-shell photoanode was one-fifth that of the SnO2/TiO2 photoanode. One

explanation is that back-electron transfer tunneling through the TiO2 layer is still a competing

83 66 process with nanoITO/TiO2 electrodes . Using co-sensitization of TPA and Ru7, Wee et al.

compared the performance of mesoporous TiO2 electrodes and SnO2/TiO2 (3 nm) core-shell electrodes. TAS experiments showed that back electron transfer from the core/shell structure was

more than five times slower than from the TiO2 photoanode. A profound difference was also found in photocurrent measurements with hydroquinone as the hole scavenger; the core-shell electrode

sustained a photocurrent of ~2.5 mA/cm2 compared to ~5 µA/cm2 from the TiO2 electrode.

Enabled by the precise control of the film thickness with ALD, research has been done to study the impact of the shell thickness on electron transfer rates in core-shell assemblies 84–86. It is important to note that the ALD thickness indicated in the literature is usually determined by

monitoring film growth on a silicon wafer inside the ALD chamber. This value may not reflect the actual coating thickness on the sample, because the rate of ALD film growth depends strongly on the substrate, especially the nucleation of the precursor on the surface in the first few cycles 87.

Figure 1-10. (Left) High resolution energy dispersive X-ray spectroscopy mapping of a SnO2/TiO2 core-shell structure. Ti: red, Sn: green. (Center) TRTS measurement of a RuP1-sensitized

SnO2/TiO2 photoanode in 0.1 M HClO4 (pH 1, red) and 100 mM potassium phosphate buffer (pH

6.8, blue). (Right) Early picosecond kinetics of a RuP1-sensitized SnO2/TiO2 photoanode showing 88 rapid recombination in the TiO2 shell. Reproduced with permission from Ref. . Copyright 2016 American Chemical Society.

Using time-resolved terahertz spectroscopy (TRTS) to probe the electron injection dynamics in

85,88 a dye-sensitized SnO2/TiO2 core-shell assembly, McCool et al. observed rapid (within 10 ps) and slower (longer than 100 ps) injection behavior from TRTS signals, which were ascribed to fast

injection of electrons into the TiO2 shell, subsequent relaxation into non-mobile trap states, and

finally slow release of electrons from the TiO2 shell into the SnO2 core (Figure 1-10). This two-

step injection mechanism was observed for TiO2 shells of thickness greater than 5 Å. The degree of trapping in the shell increased with increasing shell thickness. When the shell was thinner than

5 Å, direct tunneling of electrons through the TiO2 shell into the SnO2 core was found. Interestingly,

the observed injection amplitude was found to increase with as few as one or two ALD cycles

relative to uncoated SnO2 electrodes. This increase in charge injection amplitude was most probably due to the passivation of non-mobile surface states by the ALD coating, as suggested in other reports on similar systems 89,90.

While TRTS probes the ultrafast electron injection dynamics within one nanosecond, TAS can provide useful information about charge recombination dynamics on timescales longer than

nanoseconds in dye-sensitized core-shell systems. From TAS of RuP1-sensitized SnO2/TiO2 core- shell particles, Dempsey and co-workers 84 found that annealing following the ALD process dramatically changed the interfacial charge recombination dynamics despite the improved crystallinity of the shell evidenced by X-ray diffraction (XRD). While the back electron transfer rate through the amorphous (unannealed) shell showed a strong dependence on the shell thickness, charge recombination in the annealed core-shell structure was independent of shell thickness when the shell was thicker than 0.5 nm. They also modeled the back electron transfer rates through amorphous shells of different thickness to determine the contribution of tunneling and direct shell- mediated recombination mechanisms (Figure 1-11). With amorphous shells thinner than 3.2 nm, the tunneling mechanism dominated back electron transfer; in contrast, back electron transfer occurred primarily from the TiO2 shell when shell thickness was greater than 3.2 nm. The

85,88 SnO2/TiO2 core-shell electrodes prepared by McCool et al. were annealed prior to the dye sensitization process.

Figure 1-11. Plot of ln(1/τ1/2) vs TiO2 shell thickness for amorphous SnO2/TiO2 (red) and ZrO2/TiO2

(green) core-shell photoanodes at equal injection yields. 휏1/2 is the time for half the total absorbance change in TAS measurement. The fit models the back electron transfer dynamics with contributions from both tunneling and localized shell recombination. Reproduced with permission from Ref. 84. Copyright 2015 American Chemical Society.

The ultrafast recombination dynamics of SnO2/TiO2 core-shell photoanodes were recently reported by Gish et al. 86. Comparing the excited state absorption and ground-state bleaching signals from femtosecond TAS, they found that back electron transfer to about 60% of the oxidized sensitizer molecules occurred within 1 ns in the core-shell architecture, but the lifetime of the remaining charge-separated states was extended.

“Mummy” photoanodes

The second type of core-shell assembly is formed by insulating the dye-loaded core material with a nm-thick shell of polymeric or semiconducting material. This method of preventing dye desorption has received increased attention in the literature on both conventional and water-

62,65,91–97 splitting DSSCs . When a sensitized mesoporous TiO2 photoanode is covered with fewer

than ten cycles of ALD (i.e., less than 1 nm of an overlayer), the dye stability in aqueous solution increases dramatically. Such electrodes are an order of magnitude more stable than untreated films,

91,92 even at elevated pH . It was proposed (Figure 1-12) that the precursor for the shell (TiCl4 for

TiO2, for example) reacts first with the terminal hydroxyl groups on the TiO2 surface and the anchoring groups of the sensitizer molecules; then, the water pulse in the subsequent ALD step hydrolyzes the remaining Ti-Cl bonds to give hydroxyl termination. Repetition of the alternating

TiCl4 and water pulses leads to layer-by-layer growth of an amorphous TiO2 shell, mummifying the sensitizer molecules. Because of their reactive nature, care has to be taken when selecting organometallic precursors and the oxygen source (e.g. water, ozone) in order to avoid dye degradation 95.

So far, TiO2 and Al2O3 shells grown by ALD have been used to stabilize core-adsorbed

91 sensitizers, and dye desorption rates were found to be slower with TiO2 shells than with Al2O3 .

The addition of these shells was, however, found to reduce the electron injection yield by TAS. In

experiments with the RuP1-sensitized TiO2 core, the emission intensity of the dye decreased with

thicker TiO2 overlayers but increased with thicker Al2O3. This suggested that ALD TiO2 lowered

the electron injection yield by introducing new electron acceptor sites in the shell, whereas ALD

91 Al2O3 did so by inhibiting excited state electron injection . As a result, the photocurrent measured in short-time tests was decreased relative to uncoated electrodes, but the photocurrent was increased in long-time tests because of increased dye stability (Figure 1-13) 60,65. Alibabaeia et al. 62 compared

the photocurrent from RuP9-Ru8-sensitzied SnO2/TiO2 core-shell photoanodes with ALD TiO2 and

Al2O3 as overlayers for dye stabilization. Higher photocurrent was found for the photoanode with

the ALD Al2O3 coating, which, however, may not suggest higher Faradaic efficiency of oxygen

60 generation, because a previous report showed in a generator-collector experiment that TiO2/RuP1

photoanodes with ALD Al2O3 overlayers did not produce O2 despite their measurable photocurrent.

TiO2-RuP1 electrodes with and without a PMMA overlayer, tested in 0.1 M phosphate buffer, pH

Dip-coating of dye-sensitized photoanodes with poly(methyl methacrylate) (PMMA) has also been studied as a way of stabilizing sensitizer dyes 67,98,99. Unlike the vapor-phase ALD process, a

PMMA coating can be applied by simply soaking the sensitized electrode in PMMA solution for a few seconds and drying in air (Figure 1-14). Conformal PMMA coating throughout the oxide films was confirmed by TEM and FIB (focused ion beam)-SEM images [96], and the coating thickness could be controlled by varying the soaking time and PMMA concentration. Improved photocurrent

stability was observed for samples coated with PMMA (Figure 1-14). TAS of PMMA coated TiO2-

RuP1 at different pH values showed that back electron transfer rates were similar to electrodes without the PMMA treatment, and were independent of pH and PMMA thickness. This is in sharp contrast to ALD coating, in which the thickness has a dramatic influence on the electron recombination dynamics as discussed above. Photostability, determined by monitoring absorption spectra of the photoanode during illumination, was found to be enhanced, especially with thicker

PMMA coatings. TiO2-RuP1 treated with a 3.0 wt% PMMA solution, for example, retained 90% and 75% of its initial absorption after a 16 h soaking in pH 12 buffer solution in the dark and under illumination, respectively. Without the PMMA treatment, the electrodes were bleached completely during the test 98. However, reduced photocurrent and cyclic voltammetry current from sensitized electrodes with thicker PMMA coatings implied that the insulating coating inhibited electron transfer between the electrode and the electrolyte 98,99. In addition, with the PMMA coating, the electrolyte may not have access to all the pores of the mesoporous oxide films, because thicker

PMMA coatings block the pores and the PMMA-coated surfaces are hydrophobic 98. Ding et al. 67 immobilized molecular catalysts in the PMMA layer by mixing the PMMA solution with Ru5

molecules before coating a TiO2-RuP1 photoanode with the mixture. The incorporation of catalyst molecules turned the PMMA coating into an active layer for water oxidation, and higher photocurrent was measured. However, the embedded catalyst molecules may prevent the formation of a conformal PMMA coating and the long-term stability of electrodes made by this technique remains to be explored.

1.4. Photocathodes for water splitting dye cells

1.4.1. General principles

Water-splitting cells based on dye-sensitized photoanodes (Figure 1-2) require a small bias

voltage in order to function because electrons in TiO2, SnO2, ITO, and other typical oxide

semiconductors are not sufficiently reducing to generate hydrogen from water. In principle, a tandem system that substitutes a photocathode for the dark catalytic cathode in Figure 1-2 could split water without an applied bias. The study of such photocathodes has begun recently and is at a less advanced stage of development than the water-splitting photoanode. So far, several different photocathode architectures have been reported by different groups (Table 1-2). Most of these adapt the structure of p-type dye-sensitized solar cells (p-DSSC), in which wide bandgap p-type semiconductor nanoparticles (such as NiO) are deposited as a mesoporous thin film on a transparent conductive substrate for hole transport; molecular dyes adsorbed onto the nanoparticle surface serve as sensitizers for light absorption and excited charge carrier generation. Unlike the p-DSSC, in which a redox shuttle such as iodine/triiodide is used,100 the water-splitting photocathode requires the incorporation of a hydrogen-evolution reaction (HER) catalyst due to the slow kinetics of proton reduction. A schematic drawing of a dye-sensitized photocathode and the energy level alignment of each component are shown in Figure 1-15. The efficiencies of electron transfer and charge collection depend on the energetic arrangement of the individual components. The valence band maximum (VBM) of the semiconductor needs to fall between the HOMO and LUMO of the sensitizer, and the conduction band maximum should be negative of the LUMO of the sensitizer.

The HER catalyst (Ecat) must be energetically capable of accepting an electron from the LUMO of the dye. This energetic arrangement promotes directional electron flow toward the HER catalyst.

Excitons, produced by the sensitizer upon light excitation, dissociate at the dye-semiconductor interface, reductively quenched by NiO within 1 ps.101–103 Electrons transfer from the reduced

sensitizer to the HER catalyst, followed by proton reduction for hydrogen evolution. Typically the measured photocurrents and incident photon-to-current efficiencies (IPCE) of these photocathodes are very low, as they are compromised by several factors. The primary issues are poor hole injection, slow hole transport in the semiconductor, slow catalytic rates of proton reduction, and fast charge carrier recombination.

Figure 1-15. Structure and energy scheme of the dye-sensitized photocathode for water splitting.

Table 1-2. Photocathode sensitizers and assemblies for water-splitting dye cells.

Steady-State Faradaic Photocurrent IPCE Photocurrent Reference Semiconductor Dye HER Catalyst photocurrent efficiency of H2 stability (h) (wavelength) test conditions (μA/cm2) generation PMI- 1.7 - 3.9 ca. 0.4% Xe lamp (>420 104 NiO n/a 97±7% 4 6T-TPA (0.62 V) (400 nm) nm), pH 7. Light emitting diode array (25 105 NiO P1 Co1 5-18 (0.22 V) n/a 0.5 n/a mW/cm2, >400 nm), pH7 Xe lamp (>420 ca. 1.1% 106 NiO RuC2-Co1 8 (0.51 V) 68 % 2.5 nm) with water (400 nm) filter, pH7 Xe lamp (300 mW/cm2, >400 107 NiO RuP1 Co2 13 (0.21 V) n/a n/a n/a nm) with water filter, pH 7 White LED light (100 63 NiO P1 Co3 20 (0.42 V) 68% 1.5 n/a mW/cm2, >400 nm), pH7 MoSx Xe lamp (400- 108 n/a Py-Ru 15 (0.33 V) 98% 4.17 n/a nanoparticles 700 nm), pH0.3

4+ 1.54 Xe lamp (344 Mo3S4 109 NiO BH4 183 ±36 (0 V) 49 ±11 % 16.6 ±0.20 % mW/cm2, >420 cluster (460 nm) nm), pH0 Xe-Hg lamp (>470 nm) with 110 NiO DAT Ni1 n/a n/a n/a n/a water filter, pH2.1 Xe lamp (100 111 NiO RuP4 Ni3 4-8 (0.3 V) 8.6 ±2.3% 2 n/a mW/cm2, >400

nm) with water filter, pH3 White LED light (100 98 ±4% for 15 for Ni2, 25 for mW/cm2, >410 Co1 and Ni2, 112 NiO PMI Co1 (-0.4 V vs 2 n/a nm), 0.1 M Ni2 80 ±10% for Ag/AgCl) H2SO4, 0.1 M Co1 Na2SO4 in 1:1 H2O : MeCN Xe lamp (65 113 NiO TPA-Co4 6-15 (0.14 V) 9.30% 3 n/a mW/cm2, 400 - 800 nm), pH5.5 White LED light 102 NiO C343 Fe1 10 (0.18 V) 50% 0.67 n/a (mW/cm2), pH4.5

*See Figure 1-4 and Figure 1-16 for structure codes.

**Other notes for Table 1-1 also apply here.

Figure 1-16. Structures of the sensitizers (a) and catalysts (b) from the photocathodes summarized in Table 1-2. (Catalyst is colored red when covalently connected to a sensitizer)

1.4.2. p-Type semiconductors for water splitting photocathodes

As shown in Table 1-2, NiO has so far been the only semiconductor used successfully as the hole transporting material in water-splitting photocathodes. NiO is a p-type semiconductor due to its mixed valence Ni2+ and Ni3+ states 114, and has an indirect band gap of 3.47 eV 115. The flat-band potential of NiO, as an indication of the upper limit (in energy) of its valence band, is 0.47 V vs

NHE at pH 7 as measured by photocurrent onset potential 115. A somewhat more positive value

(0.53 V vs NHE) has been measured by the Mott-Schottky method 116. This relatively negative flat band potential is favorable for electron transfer from the semiconductor to photoexcited sensitizer molecules, making NiO a good candidate as a hole quencher. Like other oxide semiconductors,

NiO exhibits a pH-dependent (ca. -60 mV/pH) flat-band potential shift in aqueous solutions because of protonation/deprotonation equilibria at the semiconductor-electrolyte interface. This implies that the driving force for reductive quenching of excited sensitizer molecules will be lower in acidic media 117.

Mesoporous NiO electrodes can be prepared via various routes, including electrodeposition

118, hydrothermal synthesis 119,120, sol-gel deposition 121,122, and block copolymer templating 123,124.

Wood et al. 125 compared the photoelectrochemical properties of dye-sensitized photocathodes made from different NiO sources (commercial and lab-made) as well as different deposition techniques (screen printed and doctor bladed) while using the same sensitizer and catalyst (P1 and

Co1 in Figure 1-16). Despite the use of different NiO sources and electrode assembly techniques, the photocurrents of all electrodes were similar in magnitude (ca. 10 μA/cm2), and differences in current varied consistently with the specific surface area of the electrode, suggesting that NiO electrodes made under different conditions should possess comparable photoelectrochemical performance.

Compared to the n-type metal oxides used in DSSCs and water-splitting anodes, p-type metal oxide semiconductors typically have low charge carrier mobility. For example, the hole

-8 -7 2 diffusion coefficient of mesoporous NiO in p-DSSCs is reported to be 10 - 10 cm /s, which is at least two orders of magnitude lower than the electron diffusion coefficient of TiO2 used in n-DSSCs

126,127. Poor hole transport kinetics lead to fast charge recombination, lowering the photoelectrochemical performance of the cell. The fact that charge recombination is the dominant kinetic pathway at both the photocathode and photoanode of water-splitting dye cells underscores the importance of controlling the architecture of the dye-semiconductor interface 128. Surface quality plays an important role in charge separation and recombination, especially for nanomaterials with a high surface-to-volume ratio. Surface defects, which are often formed during preparation of the film (and are only partially removed by annealing), usually consist of atomic vacancies and dangling bonds. These can serve as trap sites for charge carriers, resulting in low charge separation yields. Kaeffer et al. 113 compared the XPS spectra of their photocathodes before and after two-hour photolysis. They found a change in the Ni3+/Ni2+ ratio and the formation of metallic nickel, which indicated competitive reduction of bulk NiO during water reduction.

Strategies for surface passivation and reducing the density of surface defects have been developed in recent years, such as increasing the crystallinity 129,130 and adding overlayers via ALD or solution- phase deposition 90,131,132. ALD alumina in particular has been shown to be effective for enhancing the photocurrent of water-splitting photocathodes 106,113.

Considerations of ion balance in water-splitting solar cells suggest that the photocathode

(photoanode) should be operated in strong acid (base), with two electrodes separated by a bipolar membrane 23,133. This configuration can avoid the formation of a pH gradient across the membrane, which shifts the water redox potentials during continuous operation. Additionally, there is a greater driving force for proton reduction at lower cathode pH. From this perspective, NiO is not a good candidate as the hole-transport layer in water-splitting devices due to its solubility in acid.

Surprisingly, Click et al. 109 recently reported an acid-stable water-splitting dye cell based on mesoporous NiO (Figure 1-17, left). The cell was capable of 16.6 h continuous operation at pH 0.

The acid stability of NiO in their experiments was achieved by using the sensitizer BH4, which was constructed based on the PMI-6T-TPA dye with one additional π-linker and acceptor unit. The hydrophobic hexyl group from the linker protects NiO against acid dissolution in aqueous electrolytes. The protection mechanism was demonstrated by the high contact angle of the electrode after dye sensitization and by a photocurrent onset potential of 400 mV vs NHE at pH 0. Working

37 at this more favorable water reduction potential, this photocathode assembly produced a photocurrent as high as 254 μA/cm2 (-0.2 V vs NHE).

1.4.3. Photocathode sensitizers

The sensitizers listed in Table 1-2 are inherited from p-DSSCs, since extensive research on

DSSCs has prepared a gallery of sensitizers for potential use in water-splitting cells 134,135. The primary function of dye molecules is similar to that in the water splitting photoanode, namely light absorption, charge carrier injection, and charge transfer to an appropriate catalyst molecule or nanoparticle. At the photocathode, the basic requirements for sensitizers include: 1) strong absorption of visible light to maximize the utilization of solar energy, 2) a LUMO that is sufficiently negative of the water reduction couple, and a HOMO level that is more positive than the Fermi level of the semiconductor, 3) a long-lived charge-separated state for competitive photo-induced charge injection relative to recombination processes, and 4) stable attachment to the semiconductor surface for efficient electron transfer and long-term stability.

While many naturally occurring dye molecules have been discovered that have strong visible absorption (1) and appropriate energetics (2), they generally suffer from fast relaxation of excited states (on the ps timescale) 136. A great deal of research has focused on developing dyes with longer- lived excited and charge-separated states (3) for highly efficient DSSCs. Ruthenium polypyridyl sensitizers are widely used in DSSCs 134,135 because of their intense metal-to-ligand charge transfer

-1 -1 137 3 (MLCT) absorption (extinction coefficient ε452=14600 M cm ) and long-lived MLCT state

(hundreds of ns) 138,139. Their MLCT energies can be tuned by proper design of the substituents 128.

Among metal-free sensitizers, P1 in Figure 1-16a exploits an electron push-pull mechanism, first proposed by Qin et al. 122, to extend the lifetime of its excited state. With a carboxylic acid- derivatized triarylamine core as the pusher and an electron-accepting unit, malononitrile, as the

38 pulling group, the HOMO and LUMO can be spatially separated. Additional work has been done using donor-acceptor and donor-π-acceptor systems (PMI-6T-TPA and BH4), another strategy for long-lived charge-separation. In this mechanism, the chromophore is grafted through covalent bonding (conjugated π-linker group) to an electron donor or acceptor unit. Both mechanisms rely on the fact that the spatial separation of the sensitizer HOMO and LUMO can influence the lifetime of the excited state 140. Several new diketopyrrolopyrrole (DPP) sensitizers for p-DSSCs, recently developed by Farre et al. 141, are promising for increasing the photocathode efficiency. It has been reported that hole injection into NiO occurs within 10 ps. When the sensitizer is grafted to an electron acceptor, naphtalenediimide (NDI), the DPP-NDI assembly exhibits a long-lived charge separated state in LiClO4/propylene carbonate solution, with an average lifetime of about 0.25 ms, up to 7 orders of magnitude longer than the timescale of hole injection.

Although anchoring groups (phosphonate and carboxylate, for example) are used to attach dye molecules onto the NiO surface, dye desorption still occurs under catalytic conditions. Using the

“mummy” approach described above for dye-sensitized photoanodes, Kamire et al. 112 applied an

ALD coating of Al2O3 to a PMI-sensitized NiO photocathode, which effectively prevented dye degradation and desorption at very negative potentials. The ALD Al2O3 coating also served to disaggregate the sensitizer molecules, as evidenced by the loss of excimer bands in steady-state and time-resolved spectra. When co-grafting sensitizers with chenodeoxycholic acid (CDCA), a classical coadsorbent that is used in dye-sensitized solar cells (DSSCs) to inhibit dye aggregation,

Kaeffer et al. 113 also found improved stability of their photocathodes without loss of photocurrent.

1.4.4. Photocathode catalysts

Molecular HER catalysts, which are known for their high activity and tunability, have generally been used with dye-sensitized photocathodes. In the NiO-based photocathode reported by Tong et al. 104, the low photocurrent may be attributed to the lack of a HER catalyst despite the high Faradaic

142,143 efficiency of H2 generation. Cobaloxime, a class of highly active HER catalysts , has been used in several reports of water-splitting photocathodes (Table 1-2). In the first of these 105, Li et al. used the drop-casting method to add Co1 to a P1-sensitized NiO electrode, which resulted in rapid photocurrent decay due to the desorption of Co1. The same group later designed a co- sensitization approach to anchor both the molecular sensitizer (P1) and catalyst (Co1) directly onto

NiO, achieving improved stability 63. Ji et al. 106 improved the stability of this NiO-sensitizer-Co1 assembly by coordinating a Ru-based sensitizer directly to the cobalt metal center (Figure 1-18).

Stable photocurrent was observed for 2.5 h under illumination. More examples of chromophore- cobaloxime supramolecular assemblies are summarized in a recent review by Mulfort et al. 144.

Braumüller et al. 145 recently reported another supramolecular sensitizer-catalyst assembly in which the light-harvesting rutheniumpolypyridyl unit was connected to a Pt-based catalyst through a tetrapyridophenazine bridging ligand. This molecule was successfully immobilized onto a NiO surface, but the photoelectrochemical properties of the electrode have not yet been reported.

Figure 1-18. Schematic drawing and energy level diagram (left) of the photocathode reported by Li et al. 106, and its corresponding chronoamperogram under illumination at an applied bias of 0.1 V versus NHE. Reproduced with permission. Copyright 2013 American Chemical Society

Using a RuP1/NiO photocathode, Castillo et al. 117 studied photo-excited electron transfer in acetonitrile between sensitizers and molecular HER catalysts based on Rh and Co complexes. They attributed the large photocurrent with Rh complexes to be the irreversible RhIII/I reduction process which resulted in slow back-electron transfer. Co complexes did not exhibit any photocurrent under the same experimental conditions because fast back-electron transfer from reduced catalyst molecules to NiO competes with electron transfer from RuP1 to CoII. In this regard, the Rh catalysts may be more favorable for minimizing interfacial electron-hole combination in water-splitting photocathodes.

Although the Co1 catalyst demonstrates efficient hydrogen evolution catalysis, its hydrolysis in acid limits its use under low pH conditions 109,146. In the three recent studies of dye-sensitized

108 photocathodes operating in acid, Lattach et al. used MoSx nanoparticles as the HER catalyst.

The particles were electrodeposited into the Ru complex film prepared by electropolymerization of

Py-Ru sensitizers onto a carbon electrode. Using BH4 dye to introduce a hydrophobic layer on the

109 4- NiO surface, Click et al. used a homogenous [Mo3S4] catalyst in acidic solution. In both reports, the performance of the photocathode improved significantly at lower pH. Van den Bosch et al. 110 recently reported a NiO-based photocathode using a co-immobilization strategy for grafting a Ni-

DuBois-type HER catalyst, which is highly active for water reduction across a broad pH range 147.

However, neither photocurrent nor hydrogen was detected for the full dye/NiO/catalyst assembly, although the simple dye/NiO system was photoactive. The authors suggested that dye (or catalyst) aggregation and fast recombination from the reduced catalyst to oxidized NiO occur.

Gross et al. 111 simplified the connection of sensitizers to molecular HER catalysts by using Zr4+ cations for supramolecular assembly, using the layer-by-layer assembly method previously developed for dielectric and electron donor-acceptor assemblies 148,149 as well as chromophore- catalyst assemblies 54,150. By adjusting the deposition cycles they were able to optimize the sensitizer/catalyst ratio for improved photocathode performance (Figure 1-19).

1.4.5. Conclusions

Water-splitting in dye-sensitized solar cells, first demonstrated in 2009, still faces significant challenges in terms of its development as a useful route to solar fuel production. While much progress has been made on understanding the kinetics and mechanism of interfacial charge separation and recombination, the efficiency of both the photoanode and the photocathode remains low. The stability of molecular components, especially at the highly oxidizing potential of the photoanode, is also a problem in all systems studied to date. Nevertheless, research directed towards the charge separation and stability problems has produced clever approaches to both. The highest- performing molecular photoanodes now generate photocurrents for water oxidation in the range of

42 several mA/cm2, rivaling the best inorganic oxide photoanodes, and new designs of supramolecular sensitizers, molecular catalysts, and core-shell electrodes have significantly impacted both the efficiency and stability of photoanodes and photocathodes. The modular nature of the system, and the vast tunability accessible in the molecular space, offer many degrees of freedom for re- designing the architecture, and for incorporating new light absorbers, catalysts, and protection strategies as they are developed.

Chapter 2.

Flat-band Potential in Molecularly Thin Metal Oxide Nanosheets

Pengtao Xu,a Tyler J. Milstein,a Thomas E. Mallouka,b,c

Department of Chemistry, The Pennsylvania State University

Department of Physics, The Pennsylvania State University

Department of Biochemistry and Molecular Biology, The Pennsylvania State University

Published in ACS Appl. Mater. Interfaces 2016, 8, 11539–11547

2.1. Introduction

Nanosheets that are made by intercalation and exfoliation of layered metal oxides are two- dimensional crystals, which are used as building blocks of materials that exploit their electronic, magnetic, dielectric, optical, catalytic, and ion-exchange properties.151,152 Metal oxide nanosheets that contain d0 transition metal ions such as Ti4+, Nb5+, and Ta5+ have been of special interest as light absorbers, electron acceptors, and electron transfer mediators in solar photochemical applications.153 Domen and coworkers introduced the photochemistry of these materials 30 years ago, showing that layered niobates intercalated with transition metal catalysts could generate hydrogen photochemically from aqueous methanol solutions,154 and split water when excited with

UV light.155 Layered metal oxides intercalated with transition metal catalysts and photosensitized

- 156–158 with dye molecules were later shown to photolyze HI to hydrogen and I3 in visible light. The development of techniques for growing nanosheets layer-by-layer on surfaces,159,160 and for making scrolls and other kinds of structured nanosheet colloids,161,162 has enabled the study of more complex molecule-nanosheet photochemical assemblies.138,139,163–169 In these cases, the nanosheets mediate electron transfer reactions between electron donors and acceptors, and the quantum yield for productive photochemistry represents a kinetic competition between forward and back electron transfer reactions. It is thus of paramount importance to control these rates, which in turn requires knowledge of the energetics of the electron transfer reactions. While it is straightforward to measure the bandgaps of oxide semiconductors (either in bulk form or as exfoliated nanosheets) by optical methods, the band edge potentials, which determine the driving forces of electron transfer reactions, have been more challenging to measure directly.

Metal oxide nanosheets such as titanates, niobates and tantalates are typically n-type semiconductors in which Fermi level (EF) is located close to conduction band minimum. In

(photo)electrochemistry, EF is equivalent to the flat-band potential (VFB) when band bending at the

170 semiconductor surface disappears. Two methods are typically used to characterize VFB of semiconductors in solution, the photocurrent-onset potential method and the Mott-Schottky

171 172 173 method. Using the first method, Sakai et al. and Akatsuka et al. characterized VFB of representative titanate and niobate nanosheets in LiClO4/propylene carbonate solutions. These measurements, unfortunately, do not give VFB of the nanosheets under aqueous conditions, since in

174 non-aqueous solvents VFB can be affected by the type of supporting salts and their concentrations.

Chamousis et al.175 conducted photocurrent-onset potential measurements in methanol to determine

VFB of calcium niobate nanosheets restacked from different solutions, demonstrating that specific ion adsorption can affect the energetics of the nanosheets (and hence their photocatalytic activities).

So far, no systematic study of the flat-band potentials of single- or few-layer metal oxide nanosheets in aqueous solutions has been reported. The challenge for photocurrent measurements in aqueous solutions arises from the fact that the nanosheets are wide band gap oxides that absorb only UV light, and kinetic barriers for catalytic photoreactions can shift the photocurrent onset away from

171 VFB (typically in the anodic direction for n-type mateirals).

VFB values for metal oxide nanosheets have been estimated by using empirical rules developed for three-dimensionally bonded oxide semiconductors. An early study by Butler and Ginley developed a correlation based on band gap energies and the average electronegativities of

176 constituent elements to estimate VFB values of metal oxide semiconductors. Later, by correlating the band edge potentials and band gaps of a large number of oxide semiconductors, Matsumoto177 proposed the empirical correlation between conduction band edge potentials (ECB) and band gap energies (Eg) of metal oxide semiconductor given in (Eq. 2-1):

퐸 (eV) 퐸 (V vs. RHE) = 1.23 − g (Eq. 2-1) CB 2

Although this equation has been widely used to estimate the VFB of metal oxide nanosheets,165,178,179 its reliability for molecularly-thin nanosheets is unknown since it was derived from data on bulk, three-dimensionally bonded oxide semiconductors.

The other method used to characterize the VFB of semiconductors in contact with solutions is the

Mott-Schottky method, which relates the differential capacity of the space charge layer to the applied potential. This method has been widely used for bulk as well as thin film semiconductors.170,180–182 Recently, Maeda and coworkers reported band-edge tunable perovskite nanosheets of HCaxSr2-xNbyTa3-yO10 and explored how the band-edge potential influenced their

178,183 photocatalytic properties. Here, we apply the Mott-Schottky method to measure VFB of few- layer nanosheets of these materials in contact with aqueous solutions as a function of pH. From these measurements we can observe a quantitative correlation between the composition of the nanosheets and VFB. We also used first-principles electronic structure calculations based on density functional theory (DFT) to correlate the composition of the nanosheets with their corresponding band diagrams.

2.2. Experimental Section

2.2.1. Preparation of layered metal oxide nanosheets

Calcium niobate nanosheets (TBA1-xHxCa2Nb3O10, TBA = tetra(n-butylammonium)) and

184,185 strontium niobate nanosheets (TBA1-xHxSr2Nb3O10) were prepared as previously described.

Briefly, the layered metal oxides KCa2Nb3O10 and CsSr2Nb3O10 were synthesized by calcining a mixture of K2CO3 (99%, Alfa Aesar), Sr2CO3 (99.9+%, Aldrich), CaCO3 (99.95+%, Sigma-Aldrich) and Nb2O5 (99.99%, Sigma-Aldrich) with elemental ratios K : Ca : Nb = 1.05 : 2 : 3 and Cs : Sr :

Nb = 1.2 : 2 : 3 at 1373 K for 12 h in air. An excess of the alkali metal carbonate was used to

47 compensate for the volatilization at high temperature. The obtained powders were treated with 2 M

HCl aqueous solution for three days to replace the alkali ions with protons. The proton-exchanged materials were then dispersed in 100 mL of an aqueous solution that contained an equimolar amount of tetra(n-butylammonium) hydroxide (TBAOH, prepared from a 40% solution in water, Fluka) and were shaken for at least one week at ambient temperature. The nanosheet suspensions were collected after the removal of the unexfoliated particles by centrifugation. In the case of TBA1- xHxCa2Nb3-yTayO10 nanosheets, the layered oxides RbCa2Nb3-xTaxO10 were prepared by the polymerized complex method reported by Maeda et al183 Nanosheets with x values of 2.25 and 1.5 were prepared. In a typical synthesis, NbCl5 (99+%, Strem Chemicals, Inc.) and TaCl5 (99.99%,

Acros Organics) powders were dissolved in 100 mL methanol with stirring, followed by the addition of CaCO3, RbCl (99%, Alfa Aesar), citric acid (CA, anhydrous, 99.5%, Alfa Aesar) and ethylene glycol (EG, 99+%, Sigma). The molar ratio of Rb/Ca/(Nb+Ta)/Ca/EG was 1.4/2/3/30/120.

The solution was then heated to about 65 ℃ to allow the solvent to evaporate. The esterification reaction between EG and CA was promoted by heating the mixture to about 130 ℃, which yielded a slightly brown glassy resin after one hour. Pyrolysis of the resin was conducted in air at 450 ℃ for one hour to produce black powders, which were then heated in an alumina crucible at 600 ℃ for four hours in air. The obtained white powders underwent a final calcination in air at 1100 ℃ for 12 hours. The subsequent acid-exchange and exfoliation processes were carried out in a similar fashion to KCa2Nb3O10 and CsSr2Nb3O10.

2.2.2. Layer-by-layer (LBL) assembly of the nanosheets

Substrates, including fluorine-doped tin oxide-coated glass (FTO, 8 Ω/cm2, Hartford Glass

Company), quartz (Chemglass Life Sciences), silicon wafers (University Wafer) and gold-coated

48 glass slides (Deposition Research Lab Inc.) were cleaned by sonication in soapy water, isopropyl alcohol, and Nanopure water (each for 10 min), respectively. 2 wt.%

5 poly(diallyldimethylammonium chloride) solution (PDDA, Mw = 1-2 ×10 , 20 wt.% in water,

Aldrich) and a diluted nanosheet suspension (~ 0.5 mg/mL) were prepared for LBL assembly. The clean substrate was first immersed in PDDA solution for 20 min to anchor the positively charged polymer and was then thoroughly rinsed with water and dried in a compressed air stream before being immersed in the nanosheet solution for another 20 min to produce a PDDA/nanosheet bilayer.

The deposition was repeated n times to construct n layers of PDDA/nanosheets.

2.2.3. Electrochemical measurements

Electrochemical measurements were carried out in a three-electrode cell with Ag/AgCl (3.0 M

NaCl, +0.210 V vs NHE) as the reference electrode and Pt gauze (100 mesh, 1×2 cm2) as the counter electrode. Working electrodes were prepared by depositing five layers of nanosheets onto clean FTO-glass and gold-coated slides using the LBL technique described above. Prior to the electrochemical measurements, the electrodes were heated at 200 ℃ for 30 min to remove solvent and then exposed to UV irradiation (Black-Ray B100-AP lamp, UVP, LLC) for about three days to decompose polycations, exploiting the photocatalytic property of the nanosheets. Aqueous 0.1

M KCl containing 10 mM acetate buffer for pH 3 to 6 or 10 mM phosphate buffer for pH 6 to 8 was used as the electrolyte. Accurate pH values were determined by using a pH meter (VWR sympHony SP70P). The electrolyte was purged with nitrogen for 15 min prior to each set of measurements. Electrochemical impedance spectroscopy (EIS) was conducted using Autolab potentiostat with a FRA32M module. A frequency range of 10-1 ~ 104 Hz and an amplitude of 10 mV were used. EIS measurements at each pH for every material were repeated three times with

49 three different samples to avoid artifacts arising from electrode polarization. All electrode potential values are given with respect to the Ag/AgCl reference electrode unless otherwise noted.

2.2.4. Characterization

Samples were characterized by X-ray powder diffraction (XRD, PANalytical Empyrean, Cu-Kα radiation), transmission electron microscopy (TEM, JEOL 1200 EXII, accelerating voltage 80 kV), field-emission scanning electron microscopy (FESEM, FEI NanoSEM 630, accelerating voltage 10 kV), and atomic force microscopy (AFM, Bruker Icon microscope, PeakForce Tapping mode).

Ultraviolet-visible (UV-vis) absorption spectra of the nanosheets were collected using a Varian

Cary 6000i spectrophotometer.

2.2.5. Electronic structure calculations

All calculations were performed within the framework of DFT as implemented in CASTEP.186

187 The structure of a HCa2Nb3O10 monolayer was created as reported elsewhere. Initial structures of HCa2Ta3O10 and HSr2Nb3O10 monolayers were generated by replacing metal ions at the corresponding sites in the HCa2Nb3O10 structure. For geometry optimization, the general gradient approximation (GGA) in the Perdew-Burke-Ernzerhof (PBE) form188 and a cut-off energy of 400 eV for the plane-wave basis set were employed with a convergence threshold of 2.0×10−5 eV/atom for energy and 0.05 eV/Å for force. Integration over the Brillouin zone was performed by using

Monkhorst–Pack (MP) grid of 2 × 2 × 1. The work function was determined as the potential difference between the vacuum level and the Fermi level. Bond orders were analyzed using the

Mulliken method.189 For electronic structure calculations, we used the hybrid functional HSE06190 with a cut-off energy of 800 eV. Band structures were generated along the following high-

50 symmetry points in k-space with a spacing of 0.08 /Å: G (0, 0, 0), F (0, 0.5, 0) and B (0.5, 0, 0).

The partial density of states (PDOS) was calculated with a MP grid of 5 × 5 × 1.

2.3. Results and Discussion

2.3.1. Preparation of nanosheets

The layered metal oxides examined in this study belong to Dion-Jacobson (D-J) structural family

191 of layer perovskites with a general formula of A’An-1BnO3n-1. A’ is an alkali metal cation that separates An-1BnO3n-1 perovskite blocks, in which the A cations occupy cubooctahedral sites surrounded by eight corner-sharing BO6 octahedra and n denotes the number of BO6 repeat units along the stacking axis in each block. When the parent solids are exfoliated into nanosheets, the A’ cations are replaced by protons and TBA+ cations to preserve neutrality. In subsequent discussion, we denote the n=3 TBA1-xHxA2B3O10 nanosheets by the shorthand notation A2B3O10. XRD patterns

(Figure 2-1a) confirmed the successful synthesis of the layered perovskites as well as the proton- exchanged products. TEM images (Figure 2-1b, left) show that the proton-exchanged products are delaminated into nanosheets. Energy-dispersive X-ray spectroscopy (EDS) was used to determine the compositions of Ca2NbxTa3-xO10 nanosheets (Table A-1), and showed that the experimental

Nb:Ta ratios for Ca2Nb2.25Ta0.75O10 and Ca2Nb1.5Ta1.5O10 nanosheets were very close to the 3:1 and

1:1 target ratios, respectively. These nanosheets were further found by AFM to form patchy monolayers in the first step of LBL assembly on Si substrates, despite the presence of a small fraction of multilayer nanoparticles (Figure 2-1b, right). The thicknesses of the monolayer nanosheets were measured at the edges where the sheets overlapped in order to avoid different tip interaction forces between the substrate and nanosheets. Based on the average five measurements,

the thicknesses of the nanosheets were: Sr2Nb3O10, 1.73 ±0.16 nm; Ca2Nb3O10, 1.60 ±0.07 nm;

Ca2Nb2.25Ta0.75O10, 1.79 ±0.06 nm; Ca2Nb1.5Ta1.5O10, 1.78 ±0.07 nm.

Figure 2-1. (a) XRD patterns of as-prepared layer perovskites and their corresponding proton- exchanged products. (b) TEM (left) and AFM (right) images of the exfoliated nanosheets.

2.3.2. Layer-by-Layer Assembly

Exfoliated nanosheets were assembled onto different substrates by sequential adsorption of

PDDA and nanosheet solutions.192 Layer-by-layer assembly of nanosheets on quartz was monitored by UV-vis spectroscopy (Figure 2-2). The linear growth of the absorbance at 265 nm with increasing layer number indicates that approximately the same amount of nanosheets was

52 deposited in each adsorption cycle. Note that the background absorption (λ > 350 nm) also increased with layer number due to Rayleigh scattering from the nanosheets and unexfoliated particles.

Figure 2-2. UV-vis absorption spectra of nanosheet multilayers deposited on quartz substrates.

2.3.3. Band gap determination

The optical band gaps (Eg) of the nanosheets were determined by converting their UV-vis absorbance spectra into Tauc plots using (Eq. 2-2):

1/푛 (ℎ푣훼) = 퐴(ℎ푣 − 퐸g) (Eq. 2-2)

Here h is Plank’s constant, v is the frequency, α is the absorption coefficient, and A is a proportionality constant. The exponent n denotes the nature of photon absorption, where n = 1/2

53 and n = 2 are used for direct and indirect band gaps, respectively. A linear fit from the point where

1/푛 (ℎ푣훼) begins to increase linearly with hv intercepts Eg on the hv axis. The absorption coefficients of nanosheets used in (Eq. 2-2): were taken as proportional to the absorbance values at

265 nm, since at this wavelength only the nanosheets, and not PDDA nor TBA ions, show strong light absorption. The nanosheets used in this study are indirect band gap semiconductors, because their direct band gap values fitted from n = 1/2 were larger than indirect ones. Direct transformation of the UV-vis absorption spectra gives the Tauc plots shown in Figure 2-3a, where an artifactual decrease of band gap due to particle light scattering is observed. For a more accurate determination of the band gap, the spectra were corrected by subtracting a linear baseline, which was made by extrapolating a tangent line to the absorbance in the ~400-500 nm region across the spectrum. The resulting corrected plots are shown in Figure 2-3a. It is worth noting that this method can only partially remove the scattering contribution from the spectra because the scattering efficiency increases as 1/λ4.193

Scattering-corrected Tauc plots were constructed for other nanosheet compositions as shown in

Figure 2-3b, and band gap values of Sr2Nb3O10, Ca2Nb3O10, Ca2Nb2.25Ta0.75O10 and

Ca2Nb1.5Ta1.5O10 were estimated to be 3.77 ± 0.03, 3.81 ± 0.04, 3.88 ± 0.03 and 3.97 ± 0.04 eV, respectively. The trends we observe are that A-site Sr gives a slightly smaller bandgap than A-site

Ca, and that the substitution of Ta for Nb on the B site increases the band gap. These trends are consistent with earlier reports by Maeda et al.,178,183 although they reported smaller band gaps, possibly because their UV-vis diffuse reflectance measurements were made with restacked nanosheets instead of few-layer nanosheet films.

2.3.4. Mott-Schottky Experiments

The flat-band potentials of nanosheets at various pH values were evaluated by electrochemical impedance spectroscopy using the Mott-Schottky method,180 which is based on (Eq. 2-3):

1 2 푘퐵푇 2 = (푉 − 푉퐹퐵 − ) (Eq. 2-3) 퐶 휀휀0푒푁퐷 푒

Here C is the interfacial capacitance normalized to the electrode area, ND is the donor density, V is the applied voltage, kB is Boltzmann’s constant, T is the absolute temperature, e is the electronic charge, and ε and ε0 are the dielectric constant of the semiconductor and the permittivity of free space, respectively. With an ideal semiconductor-solution interface, a plot of 1/C2 against V yields a straight line and its intercept on the V axis corresponds to VFB.

Working electrodes were prepared by depositing five layers of PDDA/nanosheets onto FTO glass. In order to ensure good contact between nanosheets, the electrodes were heated to 200 °C and then exposed to UV light for three days to remove the solvent and decompose the polycations entrained between the sheets. To evaluate the effectiveness of this process, ten layers of

PDDA/nanosheets were deposited on quartz by the LBL technique. We examined XRD patterns of the heat-treated nanosheet films before and after UV exposure (Figure A-1), from which we can see a shift of the basal plane reflections towards higher angle, indicating a decrease in the interlayer distance in the film.

In the frequency range investigated, a Randles circuit was found to model the response of the system adequately (Figure 2-4): RS represents the series resistance from the electrodes, electrolyte, and contacts; RCT describes the charge transfer resistance at the electrolyte-electrode interface. A

CPE (constant phase element) was used to describe the capacitance behavior of the electrode, which deviates from ideal capacitor due to surface non-homogeneity.182 The impedance of a CPE follows

(Eq. 2-4),

1 푍 = (Eq. 2-4) CPE 푇(푗ω)푛

where ω is the applied frequency, T (0 ≤ T ≤ 1) and n are frequency-independent parameters.

Although a CPE gave good fits to the data, the parameter T represents capacitance only when n =

1. Two models are proposed to determine the effective capacitance from a CPE.194 Considering that nanosheets spread both laterally and vertically on the substrate, the surface distribution model developed by Brug et al.195 was used here with the following expression:

1−푛 1 푅푆푅퐶푇 푛 퐶CPE = 푇푛 ( ) (Eq. 2-5) 푅푆+푅퐶푇 although a normal-distribution model196 was found to give very similar results. The capacitance of the CPE was treated only as the space-charge capacitance of the nanosheets and FTO, because the double layer capacitance was considered to be much larger than space-charge capacitance and thus a small contributor to the series capacitance in the Mott-Schottky equation.181,197

Mott-Schottky plots of five-layer Ca2Nb3O10 nanosheet films and the bare FTO substrate at pH

= 7.8 are shown in Figure 2-5, where three regions of different slopes can be observed. When the applied potential is more positive than -0.7 V (Regions 1 and 2), the space-charge region of the nanosheet film is depleted and the capacitance of the working electrode is similar to that of the bare

FTO electrode. A linear fit of Region 1 gives VFB of FTO glass to be about -0.5 V, which is more positive that the value fitted from a bare FTO sample (~ -0.6 V). The latter is comparable to other reported values when adjusted to an equivalent pH with a pH-dependence of -59.2 mV/pH.181 As the applied potential approaches values more negative than -0.7 V (Region 3), a dramatic increase in the capacitance is observed. Our interpretation of this increase is that the space charge region of the nanosheets begins to contribute to the measured capacitance. We can then determine VFB of the nanosheets via a linear fit in this region to be -0.944±0.004 V based on three measurements. A positive slope in Region 3 indicates that the Ca2Nb3O10 nanosheets are n-type, which was also true of the other nanosheets investigated.

Ideally, if the thin semiconducting film is in good contact with the substrate, one should observe only two regions of different slopes, one from the substrate and one from the film.181,182 The crossing point of the two slopes indicates the potential at which the depletion layer reaches the thin film/substrate interface. However, in our experiments, there is an additional region, Region 2 in

Figure 2-5, which can be interpreted as a capacitive contribution due to specific ion adsorption between the FTO surface and the nanosheets. Although decomposition of PDDA was achieved via

UV exposure, some ion-exchange capacity still remains and intimate contact between the nanosheets and the FTO substrate is not guaranteed.198 The adsorption of cations may also explain the positive shift of the measured VFB of the nanosheet-covered FTO substrate relative to clean FTO.

Based on the Mott-Schottky equation, we can also estimate the thickness of space-charge region,

Lsc, from (Eq. 2-6):

1 2휀휀0 2 푘퐵푇 1/2 퐿sc = ( ) (푉 − 푉퐹퐵 − ) (Eq. 2-6) 푒푁퐷 푒

Using a dielectric constant of 210,199 we calculated the space-charge region thickness of the five- layer Ca2Nb3O10 nanosheet film to be about 15 nm, which is comparable to the value estimated from AFM data (5 × 1.6 = 9 nm).

(Eq. 2-6) also suggests that the potential width of Region 3 scales with the square of the thickness of the nanosheet layer, which, however, is not supported experimentally, because the width of

Region 3 appears to be independent of the number of layers deposited. We repeated the measurements on gold-coated slides (which have a smoother surface than FTO) and no significant correlation between the width of Region 3 and the thickness of the nanosheet film was found

(Figure 2-6). One explanation may be that the measured capacitance arises not only from the nanosheet film, but also from electrical double layer at the electrode surface, especially if the solution can penetrate between the nanosheets.

Figure 2-6. Mott-Schottky plots of gold-coated slides with different numbers of Ca2Nb3O10 nanosheet layers at pH 6.7.

Nanosheets of other compositions were also characterized by EIS and their flat-band potentials at different pH values are summarized in Figure 2-7. The flat-band potentials of all nanosheets show a pH dependence close to ~59 mV per pH, which can be explained by proton adsorption/desorption at the surface oxide groups of the semiconductor.200 Consistent with

expectations from (Eq. 2-1), we find that Sr2Nb3O10 has a more positive VFB value than Ca2Nb3O10, and that the introduction of Ta into Nb sites shifts VFB of the nanosheets to more negative values.

To gain more insight into the correlation of composition with electronic properties, we performed

DFT calculations of the electronic structure of the nanosheets (Figure 2-7b; see Figure A-2, Figure

A-3, and Table A-2 for optimized structures, band structures and work functions, respectively). It is worth noting that although DFT calculations systematically underestimate band gaps even when the hybrid functional HSE06 is adopted, they can provide useful information about trends in energies and about the contributions of atomic orbitals to the energy bands. The calculated variation in band gaps is consistent with our measurements. The partial density of states analysis (Figure A-4) indicates that 2p orbitals from oxygen and 3d orbitals from B-site cations contribute the most to the valence band and conduction bands, respectively. Therefore, substitution on the B site can directly influence the energy of the conduction band edge. A-site cations, although not directly contributing to the frontier orbitals, may alter the electronic structure through changes in A-O bond strength.

We observed that the average Mulliken bond order for A-O bonds decreased from 0.06 for

HCa2Nb3O10 to 0.05 for HSr2Nb3O10 in our calculations. The smaller bond order implies a more ionic (less covalent) bond for Sr-O relative to Ca-O, which is consistent with the results of a study

201 of ABO3 perovskites by Chen et al.

We also applied Matsumoto’s empirical equation (Eq. 2-1) to estimate the conduction band edge potentials of the nanosheets. In this correlation, the energy difference between the conduction band edge and Fermi level was estimated from the conductivity of the semiconductors. For wide band gap semiconductors such as those in this study (>104 Ω·cm resistivity202), 0.4 eV was used. The estimated Fermi level potentials of the nanosheets are plotted in Figure 2-7a (assuming a pH dependence of -59 mV/pH), from which we can see that the empirical equation gives reasonably good predictions.

When extending the EIS measurements to other layered metal oxide nanosheets including

Sr2Ta3O10, Ca2Ta3O10, TiOx and TiNbO5, we found two problems in obtaining useful Mott-Schottky plots. First, tantalate compounds are expected to possess a very negative VFB (<-1.2 V in neutral solution), and EIS measurements in this potential range require the inclusion of a Warburg element in equivalent circuit to account for diffusion controlled phenomena, which complicates the determination of the nanosheet capacitance. Second, for TiOx and TiNbO5 nanosheets, the Randles circuit did not give good fits to the EIS spectra collected at all scanned potentials. We found that as the potential approached the expected value of VFB, the measured EIS spectra deviated from the

Randles circuit fit (Figure A-5). We suspect that electrochemical reduction of TiOx and TiNbO5 nanosheets at negative potentials may complicate the analysis in these cases.

2.4. Conclusions

Molecularly-thin perovskite nanosheets A2B3O10 with different atomic compositions, including

Sr2Nb3O10, Ca2Nb3O10, Ca2Nb2.25Ta0.75O10 and Ca2Nb1.5Ta1.5O10 were grown as multilayer films on

61 different substrates by LBL adsorption of polycations and nanosheets. The band gaps of these nanosheets could be determined by constructing scattering-corrected Tauc plots. Mott-Schottky analysis of electrochemical impedance measurements indicated that all nanosheets in this family were n type semiconductors and had a pH dependence of their flat-band potentials close to -59 mV per pH. DFT calculations indicated that the electronic structure can be affected by the strength of

A-O bonding and by substitution of Ta for Nb on B-sites. As long as the materials are robust and the equivalent circuit is consistent throughout the whole measurement, one can use the Mott-

Schottky analysis to measure the flat-band potentials of the exfoliated metal oxide nanosheets in contact with aqueous solutions. The results of these measurements support Matsumoto’s empirical correlation of conduction band edge positions with band gaps for metal oxide semiconductors.

Chapter 3.

Charge Recombination with Fractional Reaction Orders in Water- Splitting Dye-sensitized Photoelectrochemical Cells

Pengtao Xu,a Christopher L. Gray,a Langqiu Xiao,a and Thomas E. Mallouka,b,c

Department of Chemistry, The Pennsylvania State University

Department of Physics, The Pennsylvania State University

Department of Biochemistry and Molecular Biology, The Pennsylvania State University

Published in J. Am. Chem. Soc. 2018 (DOI: 10.1021/jacs.8b04878)

3.1. Introduction

Water-splitting dye-sensitized photoelectrochemical cells (WS-DSPECs) represent a molecular approach to artificial photosynthesis.203–205 In these cells, photoexcited sensitizer molecules inject electrons into a semiconductor (typically mesoporous TiO2 or a core-shell oxide semiconductor) and are regenerated by accepting electrons from a water oxidation catalyst. Because of the sluggish kinetics of the four-electron oxidation of water, recombination of the injected electrons with the oxidized sensitizer molecules is an important parasitic process in WS-DSPECs.35,206,207

Understanding the kinetics of forward and back electron transfer at the semiconductor-sensitizer interface is therefore important from both fundamental and device efficiency perspectives.

The recombination reaction that takes place at the semiconductor-sensitizer interface can be represented as follows:

− + TiO2(e ) + S → TiO2 + S (Eq. 3-1)

− + where TiO2(e ) and S represent the injected electrons in TiO2 and the oxidized form of the sensitizer, respectively. The corresponding reaction rate law can be expressed for the bimolecular process as:

+ 훼 − 훽 rate = 푘[S ] [TiO2(e )] (Eq. 3-2)

− + where 푘 is the recombination rate coefficient, and the reaction orders of TiO2(e ) and S are β

− + and α, respectively. Under photostationary open-circuit conditions, TiO2(e ) and S are present in equimolar amounts, and given the homogeneous structure of the porous sensitized electrode these amounts are often expressed as local concentrations as in equation (Eq. 3-2). Note that this model does not consider the lateral charge transfer between sensitizer molecules, because the hole- hopping time is on the orders of nanoseconds and therefore fast on the timescale of charge recombination.208

Early studies of this reaction used time-resolved transient absorption spectroscopy (TA) to monitor the bleaching recovery of S+ species.209–211 It was found that there were slow and fast recombination events from the bleaching recovery signals, and they could be adequately fitted to the sum of two second-order equal-concentration processes, suggesting both α and β to be 1.211

However, Haque et al. showed that the recombination rate is strongly bias dependent.212 A 600 mV

7 shift in the TiO2 conduction band could result in a 10 variation in recombination rate. This highly nonlinear dependence of the recombination rate precludes a second-order kinetic process.212

− + Recently, by deliberately controlling the concentrations of TiO2(e ) and S through external bias,

Brigham et al.213 quantified both α and β to be 1 using the Ostwald isolation method. This method

− + requires one of the species, either TiO2(e ) or S , to be in 10-fold excess in concentration, which allows the concentration of the other species to determine the recombination rate in pseudo first- order fashion. The conditions under which one species dominates the recombination reaction, however, strongly deviates from the actual operating conditions in which equal amounts of

− + TiO2(e ) and S are present. The charging/discharging currents under external bias should shift the Fermi energy inside the TiO2 away from the level that injected electrons would occupy, and this can potentially change the recombination mechanism relative to open-circuit conditions where

− + [TiO2(e )] equals [S ]. Therefore, results from the Ostwald isolation method may not be accurate for describing reaction (Eq. 3-2) under conditions relevant to photoelectrochemical water splitting.

In this work, we use intensity-modulated photovoltage spectroscopy (IMVS) to characterize

− reaction (Eq. 3-2) by monitoring the TiO2(e ) concentration through the measured potential.

Pioneered by Peter et al.214, IMVS is a light perturbation technique that is widely used to study charge recombination in photovoltaics and photoelectrochemical electrodes.215–218 In a typical measurement, a small sinusoidal modulation of light intensity is superimposed on the steady-state illumination of the photoelectrode and the modulation of photovoltage is measured simultaneously.

The experiment can be conducted under open-circuit conditions without any external bias, and at

steady-state illumination intensities close to those of the operating water-splitting device. The TiO2

Fermi level and filling of trap states are controlled solely by the injected electrons, allowing the recombination process to be studied under quasi-photostationary conditions.

We apply IMVS to study the charge recombination process in dye-sensitized photoelectrodes that have undergone different surface processing steps to improve their efficiency. Most reported electrodes for WS-DSPECs are sensitized with dye molecules immediately after the preparation of the mesoporous TiO2 films (pristine electrodes). However, research on dye-sensitized solar cells has highlighted the importance of a TiCl4 treatment on the TiO2 films before dye sensitization, because this process can improve the solar cell efficiency.219,220 It begins experimentally by soaking the electrodes coated with a mesoporous TiO2 thin film into an aqueous TiCl4 solution at 70 ℃ and then calcining the electrodes at 500 ℃. This treatment is reported to deposit a thin TiO2 shell on the mesoporous particles, passivating surface trap states, and improving the necking between particles, increasing the electron diffusion coefficient.219 It has also recently been applied to

221–223 perovskite solar cells for improved performance. We also prepare a TiO2/TaOx core-shell structure using atomic layer deposition (ALD). Coating a more insulating material over the mesoporous support can increase the energy barrier for back electron transfer of the injected electron through reaction (Eq. 3-1), extending the charge-separation lifetime, as evidenced from many spectroscopic and photoelectrochemical experiments.64,84,86

Our results suggest that the molecularity of reaction (Eq. 3-1) is far from second order and strongly depends on surface treatments. Unmodified electrodes exhibit a more unimolecular recombination process whereas TiCl4-treated and TiO2/TaOx electrodes show a recombination process that appears bimolecular. In all cases, however, α and β are not unity as previous reports suggest. Importantly, we observe a three order of magnitude difference in the recombination lifetimes as measured by IMVS and TA. Comparing the injected electron concentrations, we conclude that faster recombination detected by TA stems from intense laser excitation, which

66 results in high concentrations of charge-separated states that are inaccessible by injection at solar fluence. We also find from photoelectrochemical impedance spectroscopy that a simple RC time constant correlates well with the electron recombination lifetime.

3.2. Theory

Open-circuit Potential, Light Intensity and Electron Recombination Rate

Adapting the analysis of charge recombination in the dye-sensitized solar cell given by Huang et al224, we can express the recombination current density in a dye-sensitized photoanode as a bimolecular recombination process according to the following equation:

훼 β 푗푟푒푐 = 푞푘푟푐푅푢푃(푛 − 푛푑) (Eq. 3-3)

where 푘푟 , 푐푅푢푃 , and 푛푑 are the recombination rate coefficient, the oxidized sensitizer concentration, and the TiO2 electron concentration in the dark, respectively. The reaction order is expressed as α for oxidized sensitizer molecules and as β for electrons. The electron population in

TiO2 due to light-induced electron injection from sensitizer molecules follows:

푞∆푉/푚푘푇 푛 = 푛푑푒 (Eq. 3-4) where kT is the thermal energy and m is the ideality factor which is unity for an ideal diode.225,226

∆푉 is the photoanode potential difference between dark (푉푑푎푟푘) and light (푉푙푖𝑔ℎ푡) conditions:

∆푉 = 푉푑푎푟푘 − 푉푙푖𝑔ℎ푡 (Eq. 3-5)

Note that because our electrodes were measured in a three-electrode configuration, 푉푙푖𝑔ℎ푡 (the potential of the photoanode relative to a reference electrode) is more cathodic than 푉푑푎푟푘 .

Substituting (Eq. 3-5) into (Eq. 3-4) and taking the derivative of 푉푙푖𝑔ℎ푡 with respect to log 푛 we obtain:

푑푉 푙𝑖𝑔ℎ푡 = −2.3푚푘푇/푞 (Eq. 3-6) 푑log푛

In the absence of electron donors and under open-circuit conditions, the only route by which excited sensitizer molecules can be oxidized is by injecting electrons into TiO2. Therefore, the concentrations of the oxidized sensitizer and injected electrons are equal:

푐푅푢푃 = 푛 − 푛푑 (Eq. 3-7)

β 훽 Comparing (Eq. B-2) and (Eq. 3-3) ((푛 − 푛푑) = 푛 , note the slightly different definition of n in Appendix B), we can express the recombination rate (푘퐼푀푉푆) measured by IMVS as:

훼 푘IMVS = 푘푟푐푅푢푃 (Eq. 3-8) From (Eq. B-1),(Eq. B-3), and (Eq. 3-3)-(Eq. 3-7), we can formulate the relationship between

푑푛 푉 and light intensity (퐼 ) at steady-state and open-circuit conditions ( = 0, 푗 = 0) as 푙푖𝑔ℎ푡 0 푑푡 푒푥푡 follows:

푞∆푉 훼+훽 퐴퐼0 = 푘푟 [푛푑 (푒푚푘푇 − 1)] (Eq. 3-9)

푞∆푉 In a typical experiment, m is between 1 and 2, and ∆푉 is larger than 200 mV, and thus 푒푚푘푇 −

푞∆푉 1 ≈ 푒푚푘푇 . (Eq. 3-9) can then be simplified to obtain the light intensity dependence of 푉푙푖𝑔ℎ푡 according to the following expression:

2.3푚푘푇 퐴퐼 푉 = 푉 − log 0 (Eq. 3-10) 푙푖𝑔ℎ푡 푑푎푟푘 푞(훼+훽) 훼+훽 푘푟푛푑

In a plot of 푉푙푖𝑔ℎ푡 as a function of log 퐼0, the slope is therefore

푑푉 2.3푚푘푇 푙𝑖𝑔ℎ푡 = − (Eq. 3-11) 푑 log 퐼0 푞(훼+훽)

Similarly, from (Eq. 3-4)-(Eq. 3-8), we can express the potential dependence of 푘IMVS as:

2.3푚푘푇 푘IMVS 푉푙푖𝑔ℎ푡 = 푉푑푎r푘 − log 훼 (Eq. 3-12) 푞훼 푘푟푛푑

The slope of log 푘IMVS against 푉푙푖𝑔ℎ푡 is:

푑log푘 αq IMVS = − (Eq. 3-13) 푑푉푙𝑖𝑔ℎ푡 2.3푚푘푇

3.3. Experimental Section

Photoanode Preparation. A colloidal TiO2 nanoparticle paste was prepared by a previously reported method.75 The paste was doctor-bladed onto a clean FTO substrate (3×5 cm2, fluorine- doped SnO2-coated glass, 8 Ω/cm2, Hartford Glass), followed by a sintering process at 300 °C for

20 min, 350 °C for 10 min, and 500 °C for 30 min. After cooling to room temperature, the FTO

2 substrate was cut into five electrodes (3×1 cm ). The thickness of the mesoporous TiO2 film was measured by scanning electron microscopy to be about 3 µm. TiCl4-treated samples were prepared by immersing the electrodes into a 50 mM aqueous TiCl4 solution for 40 min at 70 °C before calcination at 500 °C for 30 min. Core-shell samples were prepared by depositing tantalum oxide over the TiCl4-treated films at 150 °C using atomic layer deposition (ALD, Cambridge Savannah

200). Pentakis(dimethylamino) tantalum(V) (heated at 90 °C , >98%, Strem Chemicals) was used as the tantalum source and water vapor (ambient temperature) as the oxygen source. TaOx films were formed by alternately pulsing each precursor (0.25 s for Ta, 0.015-s for water) into the sample chamber under a N2 carrier gas flowing at 20 sccm. Due to the large surface area of the mesoporous

TiO2 film, we allowed a 60 s exposure time for each precursor to interact with the sample before it was purged by N2. A silicon water was placed inside the sample chamber during the deposition to monitor the growth of the TaOx film. We performed four cycles of ALD for the TiO2 electrode, and the deposited TaOx film was 0.57 nm thick according to ellipsometry measurements. The as- prepared electrodes were stored in a 70 °C oven before being soaked in a 0.1 mM ethanolic solution of bis(2,2′-bipyridine)(4,4′-diphosphonato- 2,2′-bipyridine)ruthenium(II) bromide (RuP) for 20 hrs.

The molecular sensitizer was prepared according to previous literature reports.227 Following dye sensitization, the electrodes were rinsed with ethanol and dried in a compressed air stream before being stored in the dark for future use.

Photoelectrochemical Measurements. All photoelectrochemical measurements was carried out at ambient temperature (23-24 °C) in the three-electrode configuration using Ag/AgCl(3M NaCl) as the reference electrode and Pt wire as the counter electrode. All potentials reported here are referenced to the reference electrode unless otherwise noted. The electrolyte was 10 mM acetic acid/sodium acetate solution (pH 4.7, degassed by purging with Ar). Intensity-modulated photovoltage spectroscopy (IMVS) was conducted using an Autolab potentiostat (PGSTAT128N) in combination with an Autolab LED Driver. A 470 nm LED light (LDC470, Metrohm), driven by the LED Driver, was used as the light source. The light intensity was controlled by changing the

DC level of the LED current, and the light perturbation amplitude was set to be 10% of the DC level with a modulation frequency range between 400 and 1 Hz. IMVS measurements were carried out under open-circuit conditions (output current set to 0). Light intensity was measured by a Si photodiode (Thorlabs, S130C). Electrochemical impedance spectroscopy (EIS) was performed in galvanostatic mode under open-circuit conditions with 470 nm illumination. The applied frequency range was from 1000 to 1 Hz. The current perturbation was set to 5 μA.

Nanosecond Transient Absorption Spectroscopy. A Q-switched Nd:YAG laser (Spectra

Physics GCR-130) pulsed at 10 Hz provided the excitation beam (10 ns, 532 nm). The laser pulse energy (6.5 mJ) was measured by using a pyroelectric energy sensor (ES220C, Thorlabs). A 470 nm LED coupled into an optical fiber was used to provide the probe light. The probe beam, oriented perpendicular to the laser beam, was continuously on in each measurement (100 laser shots). A monochromator (Spectral Products CM110) was placed in front of the detector, a gated photomultiplier tube (Hamamatsu H10304-01-NF). Two notch filters (532 nm) were also placed before the detector to minimize scattered light from the laser. The signal was recorded using an oscilloscope (PicoScope 6404c) that was optically triggered by a Si photodiode (Thorlabs

DET36A). The laser pulse and PMT gate timing were controlled by a pulse generator (Berkeley

Nucleonics Corporation, model 505). The response time of the whole system was about 30 ns.

The sample was positioned at a 45° angle to both the laser beam and the probe beam. We noticed that the PMT output signals became saturated under intense probe light, showing a non-flat baseline.

We thus applied a sequential subtraction strategy228 to record the baseline using a two-blade chopper wheel that was synched with the laser at 5 Hz. The chopper wheel blocked the incoming laser beam from exciting the sample for every other shot. The absorbance change was calculated by sequentially dividing the baseline from the signal.

3.4. Results and Discussion

We prepared three types of electrodes for charge recombination analysis. Pristine electrodes represent mesoporous TiO2 film electrodes sensitized with bis(2,2′-bipyridine)(4,4′- diphosphonato-2,2′-bipyridine)-ruthenium bromide (RuP) without any additional treatment, the most commonly used photoanode in WS-DSSCs. When the pristine electrode undergoes a TiCl4- treatment as described in the Experimental Section, we obtain TiCl4-treated electrodes. Further depositing a thin layer of tantalum oxide over TiCl4-treated electrodes by ALD yields TiO2/TaOx electrodes. With TiCl4-treament and ALD, we did not observe significant changes in crystal structure and any change in TiO2 particle size was too subtle to observe by scanning electron microscopy (SEM). (Figure B-1) However, the sensitizer surface coverages estimated by UV-vis absorption (Figure 3-1a) suggested a 13% and a 33% decrease in surface area after TiCl4-treatment and ALD coating, respectively. In the presence of hydroquinone as the electron donor, we measured the photocurrent-time profiles for the three types of electrodes. (Figure 3-1b) Photocurrents from the TiCl4-treated and TiO2/TaOx electrodes exhibited an increase by a factor of 2.3 and 3.9 relative to the pristine electrodes, respectively. This is consistent with previous reports that TiCl4-treatment and a core-shell structure can effectively improve the performance of dye-sensitized photoelectrodes.

3.4.1. Photoelectrochemical impedance spectroscopy

We used PEIS (photoelectrochemical impedance spectroscopy) to measure the ideality factor m in (Eq. 3-4). PEIS experiments were carried out in galvanostatic mode at open-circuit under illumination. The electrode potential modulation in response to a sinusoidal current perturbation was recorded to construct Nyquist plots.

Typical Nyquist plots from PEIS are shown in Figure 3-2a. In the measured frequency range, the spectrum appeared as semi-circles that shrank with increasing illumination intensity. We identify the frequency response as the electron transport and recombination processes in TiO2 films and the data can be adequately fitted to a simple 푅푠(푅푝퐶) circuit (Figure 3-2a inset), where Rs, Rp, and C are the series resistance, and the resistance and capacitance from TiO2, respectively. The fitting

results for the three types of electrodes are shown in Figure B-2. The TiO2 resistance in all cases decreased exponentially towards cathodic potential under illumination, which is expected as electrons populate more mobile states near the conduction band edge and traps states are filled.

Without dye sensitization, the TiO2 electrode showed a very large charge transfer resistance (a semi-circle of large diameter), as plotted in Figure B-3a. The TiO2 capacitance, however, remained essentially unchanged for pristine electrodes, whereas it increased with more cathodic potentials for the TiCl4-treated and TiO2/TaOx electrodes.

From the area under the capacitance-voltage curve (Figure B-2b), the relationship between the open-circuit potential and electron concentration could be obtained and is plotted in Figure 3-

2Figure 3-b. The integrated charge vs. potential was then fitted to (Eq. 3-6) to obtain values of the ideality factor m, and the results for the three types of electrodes are presented in Table 3-1. Note that the integrated number of electrons for each electrode is referenced to the electron concentration at the most positive open-circuit potential (i.e., at the lowest illumination intensity). The ideality factor for TiCl4-treated and TiO2/TaOx electrodes is between 1 and 2, close to the value commonly reported for DSSCs.225,226,229 The pristine electrode shows an m as high as 4.12, which is very likely an artifact. Since the capacitance of these electrodes was small and approximately constant with potential (Figure B-2b), the integration to obtain charge as a function of potential is close to linear, instead of exponential as expected from (Eq. 3-4). This suggests that the charging involves a high density of surface states for the pristine electrode, and that only after TiCl4 treatment does the dye- sensitized electrode/solution interface behave more like an ideal diode.

3.4.2. Intensity-modulated photovoltage spectroscopy

The fundamental equations governing IMVS are provided in Appendix B. Figure 3-3a shows typical Nyquist plots from IMVS measurement. (Eq. B-7) suggests that the frequency response of the modulated term should appear in the first quadrant of a Nyquist plot (Figure 3-3a I), in which the negative imaginary part is plotted against the real part of the voltage oscillation. The appearance of experimental data in the third quadrant (Figure 3-3a III) is because an increase in the electron population in TiO2 induces a cathodic potential. In the

Bode plot Figure 3-3b), where the magnitude of the imaginary part is plotted against frequency, we can identify the recombination rate (푘퐼푀푉푆) as the frequency (푓퐼푀푉푆) at which the imaginary part reaches a maximum, as discussed in the Appendix B:

푘퐼푀푉푆 = 2휋푓퐼푀푉푆 (Eq. 3-14)

Figure 3-3. IMVS Nyquist (a) and Bode (b) plots for the pristine electrode at various illumination intensities. The four quadrants are indicated by Roman numerals.

We measured the open-circuit potentials (푉푙푖𝑔ℎ푡) of the three types of photoanodes as a function of light intensity that spans two orders of magnitude (Figure 3-4a). The corresponding recombination rates determined from IMVS data are plotted in Figure 3-4b. 푉푙푖𝑔ℎ푡 is observed to shift cathodically with an exponential increase in light intensity, and 푘퐼푀푉푆 scales exponentially with 푉푙푖𝑔ℎ푡. With m available from PEIS experiments and using (Eq. 3-13) to fit Figure 3-4a, we can directly calculate α, and from Figure 3-4b, β can be deduced based on (Eq. 3-11).

We obtained by this analysis fractional reaction orders for the three electrodes and their molecularity differs, as shown in Table 3-1. These fractional orders are probably reflecting the fact that recombination occurs via the multiple trap states that distribute exponentially in energy in TiO2.

The results also showed that recombination in the pristine electrode was close to a unimolecular process with fractional reaction order as large as 2.63 with respect to the oxidized sensitizer concentration (β is close to 0 considering fitting errors). The high reaction order, although it may be an indication of abundant trap states in pristine electrodes, can also result from the application of an inappropriate model. As noted above, the charging of pristine electrodes does not follow the diode equation, and therefore (Eq. 3-4) cannot be used with pristine electrodes for recombination analysis. For TiCl4-treated and core-shell electrodes, IMVS measurements suggested a bimolecular recombination process with fractional orders with respect to both oxidized sensitizer molecules and electrons, indicating that the removal of trap states by TiCl4 treatment altered the recombination mechanism. Moreover, we observed a very similar dependence of 푉푙푖𝑔ℎ푡 on the illumination intensity for TiCl4-treated and TiO2/TaOx electrodes. This is reflected in Figure 3-4a as two parallel

lines offset by about 40 mV in potential at the same light intensity. If we assume the same 푉푑푎푟푘,

푚 퐴, 푛 , and for the TiCl4-treated and TiO2/TaOx electrodes, their potential offset at the same 푑 훼+훽 illumination intensity can be expressed as follows based on (Eq. 3-10):

2.3푚푘푇 푘푟1 ∆푉 = 푉2 − 푉1 = − log (Eq. 3-15) 푞(훼+훽) 푘푟2

where the subscripts 1 and 2 represent TiCl4-treated and TiO2/TaOx electrodes, respectively.

2.3푚푘푇 Using a ∆푉 of -40 mV and a slope of -76 mV/dec (− , the average of slopes in Figure 3-4a 푞(훼+훽) for the two electrodes), we calculated that the recombination rate coefficient (푘푟1) for TiCl4-treated electrodes is about 3.4 times larger than that of the TiO2/TaOx electrodes (푘푟2), indicating that the core-shell structure indeed slows down charge recombination in the dye-sensitized photoelectrode.

According to (Eq. 3-12), at the same 푉푙푖𝑔ℎ푡 , the recombination rate measured from IMVS is proportional to 푘푟 , and thus we can estimate 푘푟1 to be 4.1-to-3.1 times larger than (푘푟2) from

Figure 3-4b , and this agrees well with the value calculated from the potential-light intensity profile.

The fractional reaction orders we report here are in disagreement with the conclusions from earlier studies (mainly by time-resolved absorption spectroscopy) that reaction (Eq. 3-1) is a bimolecular process with unity reaction order with respect to both sensitizer and electron concentrations ( 훼 = 훽 = 1 ).209–211 We believe that this discrepancy stems from different measurement conditions, as will be discussed in detail below.

In PEIS measurement, we can also define an electron recombination rate (푘퐸퐼푆) as the inverse of

푅푝퐶. 푘EIS for a TiCl4-treated electrode was measured at different open-circuit potentials (by tuning the illumination intensity), and compared with 푘IMVS measured under the same conditions (Figure

3-5). Interestingly, 푘EIS and 푘IMVS are in close agreement. This suggests that the recombination process probed by IMVS is actually the discharging of the injected electrons into the TiO2 film, which is limited by the TiO2 electrode resistance and capacitance. This also explains that, however the trap state concentration changes (with TiCl4 treatment and core-shell design), 푚/(훼 + 훽)

remains roughly constant and close to 1, because voltage decay in a 푅퐶 circuit follows first-order

kinetics. The first-order potential decay has also been observed in TPVD experiments,74 and Lin,

et al. reported that this decay time can be modeled as the lifetime of an RC circuit in nanocrystal-

based solar cells.230

Table 3-1. Fitting results and calculated parameters.

푑푉푙푖𝑔ℎ푡/푑 log 푛 푑푉푙푖𝑔ℎ푡/푑 log 퐼0 푑 log 푘퐼푀푉푆 /푑푉푙푖𝑔ℎ푡 Sample m 훼 훽 (mV/dec) (mV/dec) (dec/V)

Pristine -243.5 ±11.8 4.12 ±0.20 -84.5 ±2.5 -10.79 ±0.63 2.63 ±0.20 0.25 ±0.26

TiCl4- -105.0 ±2.9 1.78 ±0.05 -74.6 ±0.4 -7.88 ±0.42 0.83 ±0.05 0.58 ±0.06 treated

TiO2/TaOx -92.6 ±3.3 1.57 ±0.06 -77.5 ±1.3 -6.98 ±0.21 0.65 ±0.03 0.55 ±0.06

The fact that PEIS and IMVS characterize the same recombination process suggests that a similar

kinetic analysis, as shown in Figure 3-4, may also be applied to the PEIS data. PEIS also provides

insight into the slower recombination rate measured by IMVS in the core-shell structure. As shown

in Figure B-2, both the TiO2 resistance and capacitance are increased by TaOx coating, leading to

a longer RC time constant. However, we found that IMVS measurements were advantageous over

PEIS under intense illumination, where the injected electrons raise the potential close to the flat-

band potential of TiO2, because the semi-circle in the PEIS Nyquist plot shrinks with increasing

light intensity and the equivalent circuit cannot accurately fit the deformed spectra at high

illumination intensity. (Figure B-3b)

Figure 3-5. Comparison of recombination rates measured by IMVS and PEIS for a TiCl4-treated electrode.

It is worth noting that 푘IMVS is used as a first-order rate constant throughout the analysis, even though the actual reaction order for electrons is fractional. As discussed in the Appendix B (Figure

B-), the characteristic time constant from IMVS (the frequency at the apex point in the Bode plot)

-1 for 훽<1 is much smaller than the actual electron recombination rate 푘1. When 푘1=100 s , 푘IMVS

-1 for 훽 = 1 and 훽 = 0.8 are 100 and 0.045 s , respectively. Therefore, if we measure a 푘IMVS of 100

-1 s for TiCl4-treated electrodes and use 훽 = 0.58 in (Eq. B-2), the actual 푘1 would be larger than

100 s-1 by several orders of magnitude, which contradicts the times scale measured by PEIS. The observation of first-order decay profiles in TPVD experiments also supports that electron recombination is a first-order process, and thus 푘IMVS can be approximated as a first-order rate constant with small perturbation techniques. The fractional orders we obtained using (Eq. 3-11) and (Eq. 3-13) are from Figure 3-4, in which open-circuit potentials and recombination rates were measured by varying light intensity across two orders of magnitude. Although the individual data points in Figure 3-4, when measured by IMVS, can be treated as first-order rate constants, they collectively revealed the fractional orders over wide range of light intensities.

3.4.3. Transient absorption spectroscopy

In the ideal case in which one electron recombines with one oxidized sensitizer molecule simultaneously, one would expect the same lifetime for both electrons and oxidized sensitizers.

Therefore, we also characterized the three types of photoanodes through the bleaching recovery dynamics of RuP upon excitation. With the photoelectrodes held in galvanostatic mode (i.e., under open-circuit conditions), we conducted these transient absorbance experiments at various probe light intensities, since the IMVS data suggested an intensity dependence of the recombination rate.

Figure 3-6a (raw data plotted in Figure B-5) shows the sensitizer bleaching recovery kinetics at 470 nm within 1 ms after laser excitation. Note that there is an instrumental rise time of about 30 ns in the TA system.

The bleaching recovery data show multiphasic recombination kinetics as commonly reported in the literature for DSPECs. We fitted the bleaching recovery kinetics to the Kolrausch-Williams-

Watts (KWW) stretched exponential function:

푡 훽KWW Δ퐴 = 퐴1 exp [− ( ) ] + 퐴2 (Eq. 3-16) 휏KWW

where Δ퐴 is the absorbance change over time 푡 ; 퐴1 and 퐴2 are constant; 휏KWW is the characteristic time constant; 훽 is between 0 and 1, a parameter describing the distribution width of the first-order processes. The KWW kinetic model has been widely applied to describe the complex charge transfer kinetics at the dye-semiconductor interfaces.231–235 The inverse Laplace transform of (Eq. 3-16) presents the distribution of the underlying rate constants, and the distribution function

236 퐺KWW is shown below:

∞ 휏휏 (−1)푘 휏 푘훽KWW+1 1⁄ KWW 퐺KWW( 휏) = − 2 ∑ sin(휋훽KWW푘) Γ(푘훽KWW + 1) ( ) 휋휏 푘! 휏KWW 푘=0 (Eq. 3-17)

2 We fixed 훽KWW at 0.2, and the fitted curves, 휏KWW and the adj-R are shown in Figure B-5. The data collected at low probe light intensities are not included for further analysis due to low adj-R2

(<0.9), and the rest are fed into (Eq. 3-17) for rate distribution analysis. As shown in Figure 3-

6Figure 3-b, the rate constants for all electrodes distribute over more than six orders of magnitude, implying highly dispersive recombination processes. The most probable rate constant in the pristine

7 -1 electrode is about 10 s , and the corresponding rates for TiCl4-treated and TiO2/TaOx electrodes are lower by about one and 1.5 orders of magnitude, respectively. For the same electrode, the rate distribution did not show a strong correlation with the probe light intensity except for the TiO2/TaOx electrode, where the rate distribution shifted slightly towards faster recombination with higher probe light intensity.

We observed that in all cases, the most probable rate constants given by TA are larger than kIMVS by three orders of magnitude even though the probe light intensities in TA were comparable to those in IMVS. This large discrepancy was noted before when we characterized the recombination process of dye-sensitized photoanodes using transient photovoltage decay (TPVD), and we have previously ascribed it to the different electrolytes used in the experiments.74 With the same electrolyte, the same effect appeared here, and this prompts us to reconcile the different lifetimes obtained by different techniques.

Figure 3-6. (a) bleaching recovery kinetics and (b) the corresponding rate constant distribution at 470 nm for the three types of electrode at a probe light intensity of 11.64 mW/cm2.

Although both IMVS and TA apply light perturbation to dye-sensitized photoanodes, there is a pronounced difference in the perturbation amplitude of the two techniques. The light perturbation used in IMVS is set at 10% of the steady-state illumination intensity, and the perturbation at the highest illumination intensity we used is about 3 mW/cm2 at 470 nm. In the case of TA, a Nd:YAG laser pulsed 6.5 mJ at 532 nm in 10 ns, which, with a laser spot of 0.7 cm2, translates to a light intensity as high as 109 mW/cm2. The high-intensity laser excitation is expected to excite more electrons into TiO2. The initial bleaching amplitude in Figure B-5 was close to 0.1, corresponding to an excited RuP concentration of 1019 cm-3. (See Appendix B for calculation details) From Figure

3-2, an electron concentration of 1019 cm-3 will induce an open-circuit potential much more cathodic than those measured in PEIS and IMVS (as well as those measured under solar illumination), although this transient potential change may be hard to record due to the electrode RC rise time. In contrast, the injected electron concentration estimated from PEIS is around 1017 cm-3. With 100 times more electrons injected, one would expect an accelerated recombination from TA. A stronger perturbation beam than the probe beam in TA can also explain the weak dependence of the recombination rate on probe light intensity. Our observation that TA measurements report a faster recombination process than IMVS because of higher injected electron concentrations is in line with the early report from Haque et al.212 They report a strong dependence of the recombination kinetics on laser excitation intensity, and this is explained by the occupancy of conduction band/trap states: the more of those states that are occupied by electrons, the faster the recombination will be. They also note that the charge recombination cannot be simply modeled with Marcus electron transfer theory, because a more negative bias and a negative shift of the trap states/conduction band energy result in opposite effects on the charge recombination: the former (latter) accelerates (retards) the process.

Therefore, previous TA reports of 훼 = 훽 = 1 in reaction (Eq. 3-1) may be understood as recombination from high electron occupancy conditions, which, however, are not accessible from

82 typical solar illumination. Lowering the laser energy so that the number of laser-induced injected electrons is close to those in IMVS may allow observation of bleaching recovery rate comparable to 푘IMVS, but this may also bring challenges in acquiring data with good signal-to-noise ratio.

3.5. Conclusions

We have explored the charge recombination processes in three types of dye-sensitized photoelectrodes through a combination of techniques based on electrochemistry, photoelectrochemistry, and transient spectroscopy.

When recombination kinetics are measured by TA, they are fast because of the high transient concentration of electrons and oxidized sensitizer molecules, and the process is governed by a distributed rate law. This is consistent with previous conclusions of Haque et al.212 However, the rate equation and average lifetime are not very relevant to the operation of the photoelectrochemical cell (under DC conditions at one sun) because of the high laser fluence.

In PEIS or IMVS measurements, the light intensity is closer to that used in WS-DSPECs, and the sinusoidal perturbation of electron concentration is small relative to the steady-state light intensity. Under these conditions one measures slower charge recombination by three orders of magnitude, consistent with earlier photovoltage decay measurements. Dye-sensitized electrodes in which TiO2 surface states are passivated by TiCl4 treatment or by a Ta2O5 shell follow the diode equation with ideality factors in the range of 1.6-1.8. But analysis of the data at any single illumination intensity gives kinetics that are dominated by the RC time constant of the electrode/solution interface, a fact that has not previously been appreciated. The RC time constant explains the apparent first-order decay that has been observed in transient photovoltage measurements. Although both TA and IMVS measurements can associate the semiconductor surface properties (such as TiCl4-treatment and core/shell structure) with a recombination lifetime,

83 we show that with IMVS rate analysis, the changes in the steady-state open-circuit potential, when the light intensity is varied, reveal that the charge recombination follows a bimolecular rate law with fractional reaction orders.

This study underscores the utility of IMVS as a complementary technique to TA for exploring the recombination dynamics of dye-sensitized photoelectrodes in aqueous media used in water splitting cells. The kinetic analysis described here gives the recombination rate law that is most relevant to DC operation of the photoelectrochemical cells, and it should be possible to extend this analysis to understand the recombination processes in full WS-DSPECs in which nanoparticle and molecular catalysts are involved in water oxidation.

Chapter 4.

Charge Transport Dynamics in Dye-sensitized Photoelectrochemical Cells

4.1. Introduction

Water-splitting dye-sensitized photoelectrochemical cells (WS-DSPECs) integrate wide band gap semiconductors with molecular components to achieve solar-to-fuel conversion.204 The introduction of molecular species offers great flexibility in the design of light absorbers and catalysts, and in principle, the same thermodynamic limit of ~31% efficiency of semiconductor- based solar-fuel systems also applies to WS-DSPECs. However, the best reported WS-DSPECs perform far below this limit.237 Although catalysis of the kinetically demanding four-electron oxidation of water is responsible for part of the energy loss, a more significant energy penalty arises from the low quantum yield of the photoanode, which arises from recombination processes during charge collection.238 A quantitative analysis of the electron transport and recombination processes in WS-DSPECs is therefore imperative for optimizing the device performance.

Through a series of photoelectrochemical and transient spectroscopy experiments, our group has explored and established a kinetic framework for describing the low quantum yield and transient photocurrent behavior of WS-DSPECs. Using rate constants from transient photovoltage decay experiments, Swierk et al73 constructed a kinetic model that was able to fit the transient photocurrent in a typical WS-DSPEC chronoamperometric experiment. The model points out that as the photocurrent decays (over a period of tens of seconds), the concentrations of oxidized dye molecules and electron scavenger sites slowly increase, eventually reaching a photostationary state.

However, this scheme failed to take into account the electron transport and recombination properties in the TiO2 mesoporous network, which are known to govern the performance of conventional power-generating DSSCs. The dynamic response of DSSCs, including photoelectron generation, electron diffusion in TiO2, and electron recombination with the redox shuttles, has been explored extensively by intensity-modulated photocurrent spectroscopy (IMPS).216,239,240

In IMPS, a sinusoidal modulation of light intensity is superimposed on the steady-state light illumination. This induces photocurrent modulation in the photoelectrode. The phase shift caused by charge transport processes conveys information about electron diffusion and recombination rates.

While IMPS has brought significant advances to our fundamental understanding and to the optimization of DSSCs, it has not yet been applied to WS-DSPECs. In DSSCs, the competition between electron diffusion in TiO2 and electron recombination with the oxidized redox mediator

- (such as I3 ) determines the overall efficiency. In contrast, the water oxidation reaction, because of its slow kinetics, is the limiting factor in WS-DSPECs. In this chapter, we seek to explore the electron transfer kinetics in DSPECs in the surface-kinetic limiting case (mimicking water oxidation conditions) using IMPS. We construct a kinetic model describing the interplay between electron diffusion, recombination, and surface oxidation, which is also simulated by using numerical methods.

4.2. Theory for Numerical Modeling of IMPS

Figure 4-1. Proposed electron transport and recombination scheme for the DSPEC photoanode.

Our kinetic scheme for DSPECs is an adaptation of the IMPS model for DSSCs.239 If we consider the dye-sensitized TiO2/electrolyte system as a quasi-homogeneous and non-scattering medium, we can use a one-dimensional electron transfer scheme, as shown in Figure 4-1, to model the electron transport processes in DSPECs. Light-induced excited electron injection from dye molecules to TiO2 is followed by diffusive transport inside the mesoporous TiO2 network towards the FTO contact. The pseudo first-order injection rate is on the order of 109 s-1 and we can assume that electron injection is already complete when electron diffusion and recombination occur.

Charge extraction at the FTO-TiO2 interface is treated as a potential-dependent energy barrier with a rate constant 푘푒푥푡 . Meanwhile, recombination of the injected electrons with the oxidized sensitizer molecules occurs via a bimolecular recombination process with a rate constant of 푘2.

This bimolecular recombination, as we have learnt in Chapter 3 from IMVS experiments, can be

88 modeled with a reaction order of 1 under small light perturbation condition, although it occurs with fractional orders over wide range of light intensities. Oxidized sensitizer molecules are regenerated with a pseudo first-order rate of 푘3 by one-electron donors in the electrolyte. The perturbation of the electrolyte composition is neglected. We can describe the electron generation and recombination in DSPECs by the following continuity equations:

휕푛(푥,푡) 휕2푛(푥,푡) = 퐺(푥, 푡) + 퐷 − 푘 × 푅푢푃(푥, 푡) × [푛(푥, 푡) − 푛 ] (Eq. 4-1) 휕푡 n 휕푥2 2 푑

where n is the electron concentration in TiO2, RuP is the oxidized sensitizer concentration, Dn is the effective diffusion coefficient of electrons, x is the distance from the substrate, nd is the dark electron concentration in TiO2. G(x, t) is the generation term describing electron injection with the following expression:

−훼푥 푖휔푡 퐺(푥, 푡) = 휂훼퐼0푒 (1 + 푀푒 ) (Eq. 4-2)

where 휂 represents injection yield, 훼 is the absorption coefficient of the dye-sensitized TiO2 film,

I0 is the steady-state illumination intensity, 휔 is the angular frequency of light modulation, i is the imaginary unit, and M is the light modulation fraction. Similarly, the rate of change in concentration of the oxidized sensitizer can be described as follows:

휕푅푢푃(푥,푡) = 퐺(푥, 푡) − 푘 × 푅푢푃(푥, 푡) × [푛(푥, 푡) − 푛 ] − 푘 × 푅푢푃(푥, 푡) (Eq. 4-3) 휕푡 2 푑 3

The two boundary conditions are

휕푛(푥,푡) 퐷 | = 푘 [푛(0, 푡) − 푛 ] (Eq. 4-4) n 휕푥 푥=0 푒푥푡 푑

휕푛(푥,푡) | = 0 (Eq. 4-5) 휕푥 푥=푑

휕푛(푥,푡) 휕푅u푃(푥,푡) The steady-state condition can be achieved when = 0, = 0, and 푀 = 0. Under 휕푡 휕푡 small light perturbation conditions, the time-dependent electron and oxidized sensitizer concentrations can be represented by the following forms:

푖휔푡 푛(푥, 푡) = 푛0(푥) + 푢1(푥)푒 (Eq. 4-6)

푖휔푡 푅푢푃(푥, 푡) = 푅푢푃0(푥) + 푢2(푥)푒 (Eq. 4-7)

where 푛0 and 푅푢푃0 are the steady-state concentrations of electrons in TiO2 and oxidized sensitizer molecules, respectively. 푢1 and 푢2 are the modulated components of 푛 and 푅푢푃 , respectively.

The AC component of the photocurrent is

휕푢 (푥,푡) 푗(휔) = 푞퐷 1 | (Eq. 4-8) n 휕푥 푥=0

where q is the elementary charge. In a practical measurement, the frequency dependent photocurrent measured in the external circuit will be attenuated by the RC time constant due to the series resistance (R) and double layer capacitance (C) from electrode and electrolyte, given by

퐴(휔) = (1 + 푖휔푅퐶)−푃 (Eq. 4-9)

where P defines the exponent of the RC attenuation. In DSSCs, P is usually 1, but in our experiments with DSPECs, P shows a strong dependence on the electrolyte ionic strength and a value of 2.5 is necessary for satisfactory data fitting as will be shown later. The measured photocurrent in response to the light perturbation is then defined by

φimps(휔) = 푗(휔)퐴(휔) (Eq. 4-10)

The above equations (from (Eq. 4-1) to (Eq. 4-10)) are solved numerically using COMSOL

Multiphysics.

4.3. Experimental Section

The dye-sensitized electrodes used for IMPS were prepared by the same procedures described in

Chapter 4 for TiCl4-treated electrodes. All photoelectrochemical measurements were carried out using Autolab potentiaostat (PGSTAT128N) in a three-electrode electrochemical cell with a Pt wire as the counter electrode and a Ag/AgCl (3M NaCl) electrode as the reference electrode. A 470 nm

LED light (LDC470, Metrohm) provided the illumination. The light intensity modulation was realized through an Autolab LED driver controlled by the poteniostat with the AC amplitude set to

10% of the DC level.

4.4. Results and Discussion

4.4.1. Simulation parameters

Typical values for parameters used in equations (Eq. 4-1) to (Eq. 4-10) were adopted as shown in Table 4-1. 훼 was calculated by normalizing the UV-vis absorbance of the dye-sensitized TiO2 film to the film thickness. A typical electrode has an absorbance of about 0.66 at 470 nm. With a film thickness of 3.5 µm measured by SEM, we can calculate 훼 to be 0.66/3 µm -1 = 2200 cm-1.

Table 4-1 Simulation parameters used to calculate IMPS

Name Expression Unit Description Absorption coefficient of the sensitized 휶 2200 cm-1 film 휼 0.35 1 Electron injection yield 15 -2 -1 푰ퟎ 9.7×10 cm ·s Incident photon flux (DC) -6 2 -1 푫퐧 7.56×10 cm ·s Electron diffusion coefficient 5 -3 풏풅 10 cm Dark electron density in TiO2 Charge extraction rate at FTO/TiO2 풌 105 cm·s-1 풆풙풕 interface Electron recombination rate to oxidized 풌 2.2×10-15 cm-3·s-1 ퟐ dyes Dye regeneration rate from oxidizing 풌 5.6 s-1 ퟑ electron donors 푴 0.1 1 AC modulation percentage RC 10-3 s RC time constant 푷 2.4 1 RC exponent

The yield for initial electron injection 휂 from excited sensitizer molecules to TiO2 nanoparticles is unity for DSSCs. However, in WS-DSPECs, 휂 shows a strong dependence on the pH of the aqueous electrolyte. Using time-resolved THz spectroscopy, Swierk et al88 found that 휂 dropped to about 1/3 when the pH increased from 1 to 6.8. We thus here use the value of 0.35.

′ The incident photon flux at 470 nm, 퐼0, is calculated from the incident light power 퐼0 according to the following formula:

퐼′ [mW·cm−2] 퐼 [cm−2 · s−1] = 0 × 470[nm] (Eq. 4-11) 0 1240[V·nm]×1.602×10−19[C]

푘푒푥푡 describes the electron extraction rate at the FTO-TiO2 interface. In the experiment, the potential we applied to FTO was 0.3 V vs Ag/AgCl, a sufficiently positive bias to efficiently drive electrons through FTO-TiO2 interface. Therefore, 푘푒푥푡 is given a value large enough to not limit

5 the IMPS spectra. We confirmed that 10 cm/s for 푘푒푥푡 meets this requirement.

The RC time constant is determined by measuring the EIS spectrum under illumination conditions. A typical Nyquist plot is shown in Figure 4-2. The data was fitted to a Randles circuit

(Figure 4-2 inset), where Rs, Rp and C represent the series resistance, TiO2 resistance and capacitance, respectively. The RC time constant used for numerical modeling is Rs × C, with a typical value of 1 ms.

Numerical values of Dn, k2, k3, and P were obtained by manually adjusting the parameters for good overlap of the simulated data with experimental IMPS data. A successful fit (shown in Figure

4-3) produced the values of Dn, k2, k3, and P presented in Table 4-1.

Figure 4-4. Left: IMPS Nyquist plots measured at different NaClO4 concentrations (data normalized to the maximum of real current). Right: simulated IMPS Nyquist plots using different values of 푃.

We noticed that the shape of the measured IMPS Nyquist plots showed a strong dependence on the ionic strength of the electrolyte (Figure 4-4 left). As we deliberately added NaClO4 solution to increase the ionic strength, the IMPS Nyquist plots distorted towards the second quadrant (in all

IMPS plots, the two axes where real and imaginary current are zero divide the graph into four regions, which are numbered from first to fourth in a counter-clockwise direction starting from the upper right quadrant). Although the underlining mechanism is unknown, we can reproduce this effect in simulation by modulating the parameter P (Figure 4-4 right). IMPS measured in a solution of 0.1 M NaAc/HAc and 3mM HQ required a P value of 2.4 for a satisfactory fitting.

4.4.2. Steady-state concentration profile

With the sets of parameters in Table 4-1 known, we also calculated the concentration profiles of injected electrons and oxidized sensitizer molecules inside the TiO2 film. As shown in Figure 4-5, the injected electrons (푛0) were depleted towards the FTO interface, and the concentration decay of the oxidized sensitizers (푅푢푃0) followed the direction from which light penetrated into the film.

Note that the 푅푢푃0 concentration exceeded 푛0 throughout the film. The reason is that the concentration of HQ used in the experiment was limited so that the dye regeneration, instead of electron diffusion in TiO2 for conventional DSSCs, becomes rate-limiting. This mimics the kinetic bottleneck (catalysis of water oxidation) in WS-DSPECs.

Figure 4-5. Semi-log plots of concentration profiles for injected electrons (푛0 ) and oxidized sensitizer molecules (푅푢푃0) inside the TiO2 film under steady-state illumination. Incident light is from the FTO substrate side (film thickness 0).

4.4.3. Influence of 푰ퟎ, 푫풏, 풌ퟐ, and 풌ퟑ on IMPS

The IMPS model allows us to understand how individual processes influence the spectra by manipulating the corresponding parameters.

Keeping all other values unchanged in Table 4-1, we simulated the steady-state concentration profiles, steady-state current density, IMPS Nyquist and Bode plots at different light intensities

(Figure 4-6), diffusion coefficients (Figure 4-7), charge recombination rates (Figure 4-8), and dye regeneration rates (Figure 4-9).

In Figure 4-6, the photocurrent density increased with the light intensity, but not linearly. This nonlinearity can be understood by considering that the rate limiting step in the electron transfer scheme (Figure 4-1) is the dye regeneration step, which is limited by the fixed concentration of HQ.

The concentrations of both 푅푢푃0 and 푛0 increased with the light intensity, suggesting that more

electrons were injected and more oxidized sensitizer molecules resided at the TiO2 surface. Except for scaling the data positively in this way, the light intensity did not affect the general shape of the

IMPS plots.

We observed from Figure 4-7 that the steady-state photocurrent density increased logarithmically with the electron diffusion coefficient in TiO2. Faster diffusion of injected electrons resulted in a lower concentration of 푛0 and a higher concentration of 푅푢푃0 . The IMPS plots also scaled positively with the electron diffusion coefficient.

In Figure 4-8, we observed that increasing the recombination rate 푘2 led to a logarithmic decrease in photocurrent density. Faster recombination also lowered the concentrations of 푛0 and

푅푢푃0, and shrank the modulated photocurrent in IMPS plots.

In Figure 4-9, the dye regeneration rate 푘3 was accelerated, leading to a logarithmic increase in photocurrent density. Faster dye regeneration lowered the concentration of 푅푢푃0 as expected, and more electrons were accumulated in TiO2, as suggested by the 푛0 levels. More dramatic changes were observed in the IMPS simulations. The apex frequency of the lower semi-circle increased

-1 with 푘3, and at 푘3=50 s , the lower semi-circle disappeared and seemed to merge into the upper semi-circle. This one semi-circle feature at large 푘3 resembles the IMPS Nyquist plots of DSSCs, where fast dye regeneration takes place.

The above simulations point out directions for increasing the photocurrent of WS-DSPECs. TiO2 with faster electron diffusion can be designed, such as crystalline TiO2 nanowire arrays. This reduces electron accumulation under steady-state conditions and allows more oxidized sensitizer molecules to persist for the catalytic water oxidation reaction. Lowering the rate at which injected electrons recombine with oxidized sensitizer molecules allows more electrons to be collected, and provides more oxidized sensitizer molecules for the oxidation reaction. This can be achieved experimentally by using a core-shell structure. Accelerating the surface oxidation reaction, by using

99 efficient catalysts, is crucial for the faster regeneration of oxidized sensitizer molecules, decreasing the recombination loss of injected electrons.

However, it needs to be pointed out that this model fails to simulate the IMPS spectra (Figure

4-10) measured at low electron donor concentrations, where the radius of the lower semi-circle is large (relative to that of the upper semi-circle). Lowering 푘3 can enlarge the lower semi-circle, but only to some extent. One possible explanation is that the model proposed here requires a minimum of 푘3 that is able to induce diffusion-limited electron transport in TiO2. Also, we note the differences in the IMPS spectra collected at the same electron donor concentration (3 mM HQ,

Figure 4-3 and Figure 4-10), which implies there may be significant batch-to-batch variation in the kinetic parameters (Table 4-1).

4.5. Conclusions

Understanding the electron transport properties in WS-DSPECs is essential for future optimization towards better performance. This Chapter presents our attempts to characterize the

100 charge carrier dynamics of DSPECs using intensity-modulated photocurrent spectroscopy (IMPS).

Using hydroquinone (HQ) as the electron donor, we simplified the charge transfers in DSPECs to three processes: 1) electron diffusion in TiO2, 2) recombination of injected electrons with oxidized sensitizer molecules, 3) sensitizer regeneration through oxidizing HQ. These processes were then modeled using numerical simulations. This allowed us to examine how individual processes affected charge transport dynamics. We found that the steady-state photocurrent increased logarithmically with increasing electron diffusion coefficients (퐷푛), decreasing recombination rates

(푘2), and faster dye regeneration (푘3), although they contributed differently to the steady-state concentrations profiles of injected electrons and oxidized dye molecules. Specifically, smaller 퐷푛, smaller 푘2 , and larger 푘3 would raise the injected electron concentration, whereas larger 퐷푛 , smaller 푘2 , and smaller 푘3 would raise the oxidized sensitizer concentration. These findings suggest that WS-DSPECs can be improved by using oxide semiconductors with higher electron diffusion coefficients, designing core-shell structures to slow down interfacial charge recombination, and using more efficient catalysts. It is worth pointing out that other possible recombination routes, such as charge recombination between injected electrons and catalysts as well as back electron transfer from the catalyst to the sensitizer, are not accounted for in the present model.

101

Chapter 5.

Dye-sensitized Photoelectrochemical Water Oxidation Through a Buried Junction

Pengtao Xu,a† Tian Huang, a,b,c† Jianbin Huang,b Yun Yan,b* and Thomas E. Mallouk,a*

a. Departments of Chemistry, Biochemistry and Molecular Biology, and Physics, The Pennsylvania State University, University Park, PA 16802, USA

b. Beijing National Laboratory for Molecular Science, College of Chemistry and Molecular Engineering, Peking University, Beijing 100871, China

c. Academy for Advanced Interdisciplinary Studies, Peking University, Beijing 100871, China

†P. Xu and T. Huang contributed equally to this work

Published in PNAS 2018, 115, 6946-6951

5.1. Introduction

Artificial photosynthesis mimics the natural processes of converting solar energy, water, and

CO2 into chemical fuels. One of the key steps involved in artificial photosynthesis is water oxidation, in which an oxygen molecule is generated by the four-electron oxidation of two water molecules. This process is kinetically demanding and occurs at positive potentials where undesired side reactions lead to the oxidation of molecular components in both artificial and natural photosynthetic systems.

Water-splitting dye-sensitized photoelectrochemical cells (WS-DSPEC) are artificial photosynthetic devices that exemplify this problem.203,204 In the WS-DSPEC, a high surface area electrode based on a metal oxide, typically a mesoporous TiO2 or SnO2 film that is several microns thick, adsorbs a monolayer of dye molecules for light harvesting. Coupling these dye molecules with water oxidation catalysts enables light-driven water oxidation in a manner analogous to

Photosystem II. Although early versions of these cells were inefficient.238 the introduction of molecular water oxidation catalysts50 and the use of core-shell electrode architectures has significantly improved the quantum yield for water splitting.47,62,66 However, despite these improvements WS-DSPECs generally suffer from poor stability, with substantial photocurrent decay within minutes. The problem is primarily due to dye and catalyst decomposition and desorption.241,242 When in contact with water, the anchoring groups of the dye and catalyst molecules attached to the metal oxide surface (typically phosphonate groups), are susceptible to both oxidation and hydrolysis.243,244 Dye desorption is accelerated at elevated pH (pH>5).

Increasing the pH also lowers the electron injection yield, because the conduction band edge potential of the metal oxide shifts in the cathodic direction by about 60 mV per pH unit,245 making electron transfer from the photoexcited dye to the metal oxide less favorable. The water oxidation reaction, however, is more thermodynamically favorable at higher pH. Operating WS-DSPECs in

103 a more basic environment can produce more photocurrent if molecular desorption is suppressed and a high electron injection yield is maintained at the same time. This can be achieved, in principle, by decoupling the light absorption and water oxidation reactions in WS-DSPECs, so that each process take places in a favorable local environment.

One strategy for overcoming these difficulties is to create a “mummy” electrode architecture in which the sensitizer molecules are encapsulated by a nanometer-thick metal oxide film prepared by atomic layer deposition (ALD).60,91 Although stabilized photocurrents are attained by this technique, the electron injection yield from the sensitizer to the metal oxide is significantly reduced due to concurrent charge injection into the protecting layer. An alternative approach, which we explore here, is to interpose a solid-state hole conductor between the dye and the water oxidation catalyst. This strategy exploits the architecture of the solid-state dye-sensitized solar cell (ss-DSSC), which is adapted to water splitting with five key components (Figure 5-1): 1) a transparent conductive substrate, 2) a mesoporous metal oxide thin film, 3) a monolayer of sensitizer molecules,

4) a solid-state hole transport material (HTM), and 5) a top metal contact.246 ss-DSSCs of high efficiency and high stability have been reported.247 However, the ss-DSSC is intrinsically vulnerable to water because the spin-coated hole transport layer and the deposited top contact are not pinhole free and are water-permeable.248 Oxidized hole transport molecules may also dissolve in water. These problems can be resolved by using an ALD-grown metal oxide as a conformal coating over the photoanode, a technique that has been widely used to improve the stability of semiconductor photoanodes in harsh environments.249 For example, in either strongly acid or basic environments, Si photoanodes protected by ALD-TiO2 films can operate stably under water oxidation conditions for at least 8 h, whereas the same photoelectrodes without the TiO 2 layer quickly fail.250

104

Figure 5-1. Solid-state dye sensitized solar cell as a buried junction for visible-light photoelectrochemical water oxidation.

Here we explore the adaptation of the ss-DSSC as a buried photovoltaic (PV) junction for use in

WS-DSPECs. (Figure 5-2) We used ALD to coat a 2 nm-thin layer of TiO2 on the top contact (Au) to protect the ss-DSSC against oxidation and hydrolysis reactions in water. Although the ALD coating introduces series resistance that lowers the photocurrent in air, ss-DSSCs show enhanced stability in contact with the aqueous environment. We further electrodeposited an iridium oxide film on the Au layer as a water oxidation catalyst. The as-prepared electrode yields relatively high photocurrent and good stability towards water oxidation. Using the Au contact of the ss-DSSCs as a second working electrode, we are able to monitor the potential drop across the catalyst layer. We find that the as-prepared electrode functions as a buried junction, because under conditions of water oxidation, the buried ss-DSSC shows the same current-voltage profile as the corresponding two- electrode photovoltaic cell. Using this buried junction strategy, we make water oxidation possible with N3 dye molecules, which possess a lower oxidizing potential than commonly used phosphonate-based Ru polypyridyl sensitizers. This demonstrates that dye molecules with

105 unfavorable redox properties towards water oxidation can be effectively utilized in buried junction

WS-DSPECs.

5.2. Experimental section

5.2.1. ss-DSSC Fabrication

2 FTO substrates (fluorine-doped SnO2-coated glass, 8 Ω/cm , Hartford Glass) were etched with zinc powder and HCl (2 M) to prepare the desired patterns for photoelectrodes. The FTO-coated glass was then cleaned sequentially by sonication in soapy water, ethanol, and distilled water, followed by 10 min UV-ozone treatment. A compact layer of TiO2 was deposited onto the FTO surface by spin-coating at a speed of 500 rpm for 30s with a solution of titanium butoxide (≥97.0%,

Fluka) diluted in 2-butanol (1:30 v/v). Films were sintered at 500 ℃ for 30 min. After cooling to room temperature, a mesoporous TiO2 film was doctor-bladed onto the compact layer by using a

75 20 nm TiO2 nanoparticle paste prepared by a previously reported method. The electrode films were then sintered at 300 ℃ for 20 min, 350 ℃ for 10 min, and 500 ℃ for 30 min. The films were further treated with 50 mM aqueous TiCl4 solution for 40 min at 70 ℃ and sintered at 500 ℃ again for 30 min. After cooling to 70 ℃, the films were immersed in an ethanol solution containing 0.1 mM N3 dye (95%, Sigma-Aldrich) for 18 h. After dye adsorption, the films were rinsed with ethanol. Subsequently, a hole transport layer consisting of spiro-OMeTAD was deposited onto the dye-sensitized films. The spiro-OMeTAD solution was prepared by dissolving 130 mg of spiro-

OMeTAD (99%, Aldrich) in 1 ml chlorobenzene (99.8%, Sigma-Aldrich), together with additives including 20 μL of 4-tert-butylpyridine (96%, Aldrich) and 40 μL of a 170 mM lithium bis(trifluoromethane)sulfonamide (Li-TFSI, 99.95%, Aldrich) stock solution in acetonitrile. The as-prepared solutions were filtered through a syringe filter with a pore size of 0.22 μm to remove

106

undissolved particles before use. 20 μL of the sprio-OMeTAD solution was applied to each electrode under ambient conditions and the samples were left to dry for 1 min before spin-coating at 2000 RPM for 30 s. Finally, a 100 nm thick Au was deposited onto the spiro-OMeTAD layer in an electron-beam evaporator (Kurt J. Lesker Lab-18), forming the complete solid-state photovoltaic device. For the fabrication of cells for use in contact with water, a 2 nm thick TiO2 film was deposited onto the gold surface at 75 ℃ by ALD (Cambridge Savannah 200) with tetrakis(dimethylamino)titanium(IV) (99% Strem Chemicals) as the Ti precursor. The growth rate of the TiO2 film under these conditions was 0.8 Å per cycle, as determined by ellipsometry on Si wafers.

5.2.2. Electrodeposition of iridium oxide films

A hydroxyiridate monomer solution was synthesized by alkaline hydrolysis of K2IrO6, as described in our previous papers.251,252 Using the Au layer of the ss-DSSC as the working electrode, we deposited hydrous iridium oxide thin films on the Au surface from 0.4 mM hydroxyiridate(III/IV) solutions at an applied potential of 1.2 V versus Ag/AgCl (3 M NaCl) for

400 s. A home-built mechanical stirrer was used during the electrodeposition process. The thickness of the deposited film was characterized by a field-emission scanning electron microscope (FESEM,

Zeiss Sigma) to be about 75 nm.

5.2.3. Photoelectrochemical measurements

The light source for simulating sunlight was a 150 W xenon lamp equipped with a 410 nm long- pass filter and an AM1.5 filter. The intensity of the light incident on the FTO side of the cells was

100 mW cm-2, as measured by a power probe (Molectron, PM3). The effective area of each cell

107 was 0.54 cm2 (0.6 × 0.9 cm2). All photoelectrochemical data were recorded on a digital potentiostat

(PGSTAT128N, Autolab). The photocurrent measurements in three-electrode configuration were performed in a glass cell (96G20, FireflySci) with a Pt wire as the counter electrode and Ag/AgCl

(3 M NaCl) as the reference electrode. The counter electrode was located in a separate compartment.

The electrolyte was 0.1 M potassium phosphate buffer solution (pH 6.7) with 0.1 M NaClO4 added as the supporting electrolyte. Before each measurement, the electrolyte was deaerated by purging with argon for 20 min. Electrochemical impedance measurements (EIS) were performed in galvanostatic mode. A sinusoidal current was applied at an amplitude of 5 μA over a frequency range of 106 Hz to 5 Hz. All EIS measurements were carried out under open-circuit conditions under illumination of a 470 nm LED light at an intensity of 11 mW cm-2. Incident photon-to-current conversion efficiency (IPCE) spectra were collected with a 500 W xenon lamp and a monochromator (Spectral Products CM110). The data were taken in 10-nm increments and the light intensity at each wavelength was measured by a Si photodiode (Thorlabs, S130C). Photocurrent data were recorded at 0.62 V vs. Ag/AgCl using the same three-electrode setup described above.

5.2.4. Generator-collector O2 detection

253 A generator-collector method was used for O2 detection as described elsewhere. Briefly, experiments were performed in bipotentiostat mode with two working electrodes, a Pt wire as the counter electrode and a Ag/AgCl electrode as the reference electrode. One working electrode (the generator) was the FTO contact of the photoanode. The other one (the collector) was a planar Pt electrode made by sputtering a 50 nm thick Pt film onto a FTO glass substrate (with an adhesion layer of 5 nm Ti). The collector was masked with Kapton tape to create an active area of about 1.5 cm2. The generator-collector electrode pair was assembled by placing the two working electrodes face-to-face with the iridium oxide layer facing the Pt electrode. A piece of a 1 mm-thick glass

108 slide was inserted between the lateral edges as a spacer and a plastic paraffin film was wrapped around the long side of the two electrodes. The space between the two working electrodes was filled with electrolyte by capillary action when the assembly was placed in solution. A 470 nm LED

-2 light of 19.4 mW cm was used as the light source. To measure the Faradic efficiency for O2 production, the generator and collector were held at 0.62 V and -0.3V vs Ag/AgCl, respectively.

The charge passing through the generator under illumination was compared to the total charge through the collector electrode. The Faradic efficiency was calculated by comparison to the collection efficiency determined in the same fashion using an iridium oxide-deposited FTO as the generator, held at 1.12 V vs. Ag/AgCl, which was assumed to have a Faradaic efficiency of unity.

The experiment was performed in 0.1 M phosphate buffer at pH 6.7 with 0.1 M NaClO4. In some of our previously reported generator-collector measurements,61 the collector responded instantaneously upon generation of photocurrent, showing a current mirroring effect in the current- time profile. We note that this is an artifact due to the capacitive interference between the generator and collector electrodes, which can be minimized by increasing the ionic strength of the electrolyte, as pointed out by Sherman et al.253

Figure 5-2. Cross-sectional SEM image of an as-prepared ss-DSSC.

109

5.3. Results and Discussion

5.3.1. Preparation of ss-DSSCs

The cross-sectional structure of an as-prepared ss-DSSC is shown in Figure 5-2. The performance of the ss-DSSCs were found to increase over time after fabrication, reaching a maximum after 5 days (Figure C-1), which can be attributed to partial air oxidation of the spiro-

OMeTAD hole conductor.254 Figure 5-3a (line 1) shows typical current density-voltage (J-V) curves of a ss-DSSC, and the corresponding photovoltaic properties are summarized in Table C-1.

The short-circuit current density (Jsc) and open-circuit voltage (Voc) for the ss-DSSCs are 3.29 mA cm-2 and 712 mV, respectively, which is comparable with other N3-senstized solar cells.255

However, the power conversion efficiency of ss-DSSCs made in our laboratory was only 0.76% due to a low fill-factor (0.32); in principle, a significant performance improvement in fill factor can be achieved by introducing p-type dopants into the hole transport layer.256

When the pristine ss-DSSC was soaked in an aqueous solution, the cell performance dropped dramatically, as shown in Figure 5-3a (line 2). Decolorization of the ss-DSSC was observed during the measurement, which indicates that water penetrates through the Au layer and the spiro-

OMeTAD layer. At pH 6.7, the carboxylate anchoring group of the N3 dye molecules easily desorbs from the TiO2 surface because of hydrolysis of the carboxylate-metal oxide linkage.

Moreover, oxidized spiro-OMeTAD and Li-TFSI molecules may dissolve in the aqueous solution.

To improve the stability of the ss-DSSC in an aqueous environment, we used ALD to grow a conformal 2 nm-thick coating of TiO2 over the exposed surface (ss-DSSC/ALD). The ss-

DSSC/ALD cell shows enhanced stability in contact with an aqueous solution with slightly lower photovoltaic performance (5.6% and 16.4% drops in Voc and Jsc, respectively, Figure 5-3a, line 3) but a steady photocurrent of 2.38 mA cm-2 under short-circuit conditions (Figure 5-3b).

110

Figure 5-3. Characterization of ss-DSSCs with and without ALD TiO2 thin film overlayers. a) Two- electrode J-V curves of pristine ss-DSSC (blue) and ss-DSSC/ALD TiO2 (red) electrodes in air (1 and 3) and in contact with aqueous solutions (2 and 4), respectively. Inset digital images show the surface morphologies of the corresponding electrodes after testing. Scan rate: 20 mV/s. b) Chronoamperometric measurements of pristine ss-DSSC and ss-DSSC/ ALD TiO2 in solution under short-circuit photoelectrochemical conditions. Light source: 100 mW cm-2 xenon lamp with a 410 nm long-pass filter and an AM1.5 filter.

Nyquist plots derived from EIS measurements of the ss-DSSC before and after ALD coating are shown in Appendix C (Figure C-2). The series resistance of the cell increases upon heating to the temperature of the ALD cycles, and more significantly when the TiO2 coating is applied. The largest resistive component can be attributed to charge transfer resistance at the spiro-

OMeTAD/TiO2 interface.

111

5.3.2. Electrochemical deposition of the IrOx water oxidation catalyst

To catalyze water oxidation, we electrochemically deposited an iridium oxide thin film on the

Au layer. The Au layer in the ss-DSSC/ALD electrode was used directly as the working electrode

(WE) at an applied bias of 1.2 V vs Ag/AgCl. As water oxidation produced protons at the electrode surface, iridium oxide nanoparticles were formed in situ by acid condensation of soluble hydroxyiridate anions.251 Consistent with previous observations, we saw an increasing current during the deposition process and the color of the electrode surface turned from gold to dark blue.

We carried out the electrodeposition with a home-built mechanical stirrer operated close to the electrode surface because attempts to deposit a uniform IrOx film by this method failed when only a magnetic stir bar was used. An SEM image shows the thickness of the iridium oxide film deposited on Au to be about 76 nm (Figure C-3). The catalyzed electrode is hereafter referred to as ss-DSSC/ALD/IrOx.

112

The electrocatalytic performance of the deposited iridium oxide films was examined by using the Au layer as the working electrode. A cyclic voltammogram (Figure C-4) shows typical redox peaks or shoulders for the Ir(IV)/Ir(III) and Ir(V)/Ir(IV) couples, and we obtained a surface coverage of 23.6 ±2.1 nmol/cm2 of electroactive Ir sites by integrating the Ir(V)/Ir(IV) wave.

Figure 5-4a (blue line) shows the corresponding linear sweep voltammogram (LSV). The electrode was found to be catalytically active and the overpotential required to drive water oxidation at a current density of 1 mA cm-2 was 0.27 V, which is comparable to reported overpotentials at the same current densities.250,257

5.3.3. Photoelectrochemical properties of catalyzed photoelectrodes

Water oxidation at back-side illuminated ss-DSSC/ALD/IrOx electrodes was characterized by linear sweep voltammetry by connecting the working electrode (WE) lead of a three-electrode cell to the FTO back contact. As shown in Figure 5-4a (red line), the photocurrent onset potential shifted cathodically by 0.67 V, which corresponds to the photovoltage provided by the ss-DSSC. At the formal potential for water oxidation (0.62 V vs Ag/AgCl), the ss-DSSC/ALD/IrOx electrode produced a steady photocurrent density of 1.43 mA cm-2 (Figure 5-4b).

113

Boettcher et al.258,259 have developed a dual WE technique to understand the junction behavior of semiconductor-electrocatalyst (SC-EC) layers in photoelectrochemical cells. By making a direct

114 electrical contact to the EC layer, its potential during operation can be monitored. Two types of junctions can be identified (Figure 5-5a). One is an adaptive SC|EC junction in which the EC is electrolyte/ion permeable. In this case, the extent of band bending at the SC surface depends on the potential of redox couples in the electrolyte and its Fermi level depends on the chemical potential of the EC (VEC). The other extreme is a buried junction in which the EC effectively separates the

SC from the electrolyte and VEC determines both the band bending and the Fermi level of the SC.

By monitoring VEC when scanning VSC during LSV measurements, one would expect a linear relationship between VEC and VSC, with a slope of 1 or 0 for a buried junction or an adaptive junction, respectively.260

The Au layer in ss-DSSC/ALD/IrOx serves as a natural potential probe for the IrOx catalyst layer, and the ss-DSSC in ss-DSSC/ALD/IrOx electrode resembles a semiconductor. By measuring the voltage between the FTO and Au contacts in the ss-DSSC/ALD/IrOx electrode, we are able to monitor the catalyst potential (Vcat) change in situ during linear sweep voltammetry (Figure 5-5b).

Interestingly, the potential profile appears to indicate that the ss-DSSC/ALD/IrOx electrode is an adaptive junction because the applied potential through FTO (VFTO) changes Vcat slightly, with a slope far from the value of 1 expected for a buried junction. However, we can still categorize our electrode as a buried junction towards photoelectrochemical water oxidation, because by relating the FTO potential change in Figure 5-5b to the J-V curve in Figure 5-3a (line 4), we can reconstruct the LSV curves of the catalyst (Figure 5-4a, blue squares) as well as that of the ss-DSSC/ALD/IrOx electrode under illumination (Figure 5-4a, red squares). This suggests that the ss-DSSC in ss-

DSSC/ALD/IrOx electrode retains its photovoltaic performance. The difficulty that arises in interpreting the junction behavior from Figure 5-5b arises from the poor semiconducting property of the ss-DSSC. As a result of the low fill factor of the ss-DSSC, external potential changes applied to the ss-DSSC/ALD/IrOx electrode will drop more in the ss-DSSC than at the IrOx-electrolyte interface so that their current densities match. For photoanodes prepared from Si solar cells with

115 high fill-factor, LSV curves under illumination are observed as shifting those measured with the corresponding catalyst layers in the anodic direction by a value close to the photovoltage of the buried solar cells,250 because a very small drop in photovoltage is adequate to supply the current passing through EC-electrolyte interface.

We tested the stability of the oxygen-evolving photoelectrodes by applying a bias of 0.62 vs

Ag/AgCl for one hour. As shown in Figure 5-6, the photocurrent is relatively stable, with 71.4% retention of the initial photocurrent at the end of the experiment. Slight discoloration of the photoelectrode was observed after the one-hour electrolysis, which suggests, though at a slower rate, that electrolyte can still penetrate into the ss-DSSC with the ALD coating. One explanation is that the pinhole-free ALD layer is disrupted by possible dynamic changes occurring at the ss-

DSSC/ALD interface, as suggested by the low Faradaic efficiency described below.

A collector-generator (C-G) dual working electrode method was applied to determine the Faradic efficiency of O2 production. In this experiment, oxygen produced at the generator surface diffuses steadily towards the collector electrode where it is reduced, and this produces a cathodic current in collector. By comparing the charges passed through generator (QG) and collector (QC), we can calculate the Faradic efficiency (η) using η = QC/(QG×ηc)×100%, where ηc is the charge collection efficiency of C-G used to correct for diffusional losses of oxygen in the C-G cell; the latter was measured by using a iridium oxide/ FTO electrode as the generator (Figure C-5). The η of ss-

DSSC/ALD/IrOx electrode is 44.3±6.5 %. This value indicates that about half of the photocurrent arises from other sources, most likely oxidation of the hole transport layer in an initial charging period.

116

Figure 5-6. a) Chronoamperometry measurement of an ss-DSSC/ALD/IrOx photoelectrode for one hour at an applied bias of 0.62 V vs. Ag/AgCl. Light source: 100 mW cm-2 xenon lamp with a 410 nm long-pass filter and an AM1.5 filter. b) IPCE as a function of wavelength (blue) for an ss- DSSC/ALD/IrOx electrode at an applied bias of 0.62 V vs. Ag/AgCl. UV-vis absorption spectrum of N3 dye-sensitized TiO2 electrode (red). Error bars indicate the standard deviation based on the average of three measurements.

The incident photon to current conversion efficiency (IPCE), which is an important parameter for photoelectrochemical cells, was also measured and is shown in Figure 5-6. An IPCE value as high as 22% was observed at 540 nm, which is among the highest values yet reported for WS-

DSPECs. The spectra closely overlap with the UV-vis absorption spectrum of the N3 dye on a nanocrystallineTiO2 film electrode.

It is worth noting that that for N3 dye molecules, with the Ru(III)/Ru(II) formal potential of 0.88

V vs. Ag/AgCl 261 is negative of the water oxidation potential at pH 7. Sensitizers employed in WS-

DSPECs require careful design of their redox properties in the excited state, such as introducing electron-withdrawing groups in Ru-polypyridyl227 and porphyrin sensitizers262. The use of a buried

117 junction bypasses this requirement by avoiding direct contact of the dye molecules with water. This in principle expands the gallery of sensitizers available for use in WS-DSPECs.

5.4. Conclusions

We have demonstrated that ss-DSSCs can be used in WS-DSPECs in order to decouple the light absorption and water oxidation processes. A thin film TiO2 coating applied to the ss-DSSC by ALD effectively stabilizes the solar cell for operation in aqueous solution, although the performance of the cell in air is not as good because of increased series resistance at the TiO2/hole transport layer interface. Despite this performance loss, by electrodepositing an iridium oxide catalyst over the

-2 Au/TiO2 layer, we realized a photocurrent density of 1.43 mA cm at 1.23 V vs RHE, with relatively good stability over 1 hour. Faradaic efficiency determined by generator-collector method was 43% and an IPCE of 22% at 540 nm was measured. While the catalyst layer potential shifts slightly as the FTO potential changes (as in an adaptive junction), the potential profile under water oxidation conditions is closely related to the current-voltage curve of the ss-DSSC, which suggests a buried photovoltaic junction. The fact that the potential drop occurs mostly within the ss-DSSC, rather than at the catalyst-electrolyte interface as expected from a buried junction, is due to the poor photodiode properties of the solar cell as indicated by its low fill-factor. The buried junction design of ss-DSSCs adds to our understanding of semiconductor-electrocatalyst junction behaviors in the presence of a poor semiconducting material.

This cell design addresses an important stability issue in WS-DSPECs by physically separating the dye from the aqueous environment. Efficient charge separation in the ss-DSSC enables the ss-

DSSC/ALD/IrOx electrode to oxidize water at a relatively high photocurrent density. Further improvements are likely to be obtained by optimizing the ss-DSSC fabrication process to increase the fill-factor, and by doping of the TiO2 layer to lower the series resistance at the hole conductor

118 interface. We have also demonstrated that the buried junction approach enables the use of sensitizers with lower oxidizing power in WS-DSPECs, perhaps opening the door to use of sensitizers that more efficiently harvest sunlight.

119

Chapter 6.

Conclusion and Future Outlook

The energy conversion efficiency and device stability of WS-DSPECs have undergone dramatic improvement over the nearly ten years of development. Despite their inheritance of the initial design —from DSSCs— of sensitizing mesoporous semiconductor substrates with molecular light absorbers, WS-DSPECs have evolved in unique ways. These include engineering chromophore assembly,263–265 tuning the chromophore-catalyst interaction,83 and elaborating the underlying design principle of slowing charge recombination at semiconductor-sensitizer interface to allow an adequate time window for catalysts of the water oxidation reaction.

This thesis has explored in detail the charge transport and recombination processes in WS-

DSPECs by using a combination of electrochemical and spectroscopic techniques. In Chapter 3, the sharp contrast of recombination time constants between IMVS and transient absorption measurements prompted us to consider how the measurement conditions could affect the results.

We argued that IMVS experiments, performed under conditions that are close to those of operating photoelectrochemical cells, should give more reliable kinetic results for analyzing the recombination processes in WS-DSPECs. By combining IMVS with photoelectrochemical impedance spectroscopy, we found that the charge recombination kinetics at the semiconductor- sensitizer interface were dominated by the RC time constant of the electrode, a fact that had previously been little appreciated. A bimolecular rate law with fractional reaction orders was also formulated based on IMVS results.

In Chapter 4, we attempted to understand the charge transfer and recombination processes in

WS-DSPECs through a numerical modeling approach. The proposed model, although it was simplified the dye regeneration process, stands out as a theoretical miniature of an operating WS-

DSPEC, from which charge transport and recombination processes can be formulated.

From these models, three strategies emerge for optimizing the device performance: 1. increase the rate of electron diffusion in the semiconductor support; 2. suppress charge recombination at the

121 sensitizer-semiconductor interface; 3. integrate the device with catalysts that have high turnover frequency.

• Rutile TiO2 nanowire arrays have been reported to possess an electron diffusion coefficient two

240 -4 2 orders of magnitude larger than TiO2 mesoporous nanoparticles. If we use a Dn of 10 cm

s-1 in Table 4-1 while keeping other parameters unchanged, the calculated IPCE (470 nm) can

roughly double, from 6.2% to 12.5%.

• Core/shell structures have been shown to slow down interfacial charge recombination. Varying

86 the TiO2 shell thickness over a SnO2 core, Gish et al successfully extended the charge-

separation lifetime from tens of microseconds to several milliseconds; similar observations

84 were also reported by Knauf et al . If the charge recombination from TiO2 to oxidized

sensitizer molecules can be slowed by just one order of magnitude, the IPCE of the photoanode,

as indicated by our numerical calculation in Chapter 4, will increase to 13.6% (using 10-16 for

k2 in Table 4-1).

• The best molecular catalysts for water oxidation reported in the literature turn over the reaction

-1 266 at a rate greater than 300 s . Plugging this number into the model (k3 in Table 4-1), we

anticipate an IPCE as large as 15.2%.

• When the device is improved with the three above-mentioned strategies simultaneously, we

calculate an IPCE of 16.9%, a number only slightly larger than what we obtain from improving

k3 alone. Note that 16.9% is the highest IPCE the model predicts considering the low injection

yield and non-absorbed photons. This implies that highly efficient catalysts are crucial for

improving the performance of WS-DSPECs.

In addition to quantum efficiency losses from charge recombination, WS-DSPECs are also plagued by inherently low stability, which arises from the detachment of sensitizer molecules from the oxide semiconductor/electrolyte interface at elevated pH. Chapter 5 presented a novel strategy for circumventing this problem. By encapsulating the light absorbing molecules with the solid state

122 spiro-OMeTAD hole conductor and applying an ALD protecting layer over the photoanode, we spatially decoupled light absorption and water oxidation, two processes that originally occurred at the same location on the electrode but can now proceed in more favorable media. This buried- junction design adopted the key features of ss-DSSCs, and we achieved a photocurrent as high as

1.43 mA cm-2 at 1.23 V vs RHE, which rivals some semiconductor photoanodes such as hematite.

The IPCE of this design at 470 nm is about 20%, a number higher even than the best value the above-mentioned model predicts. This improvement mainly benefits from the high electron injection yield and good hole transport property in ss-DSSCs. Further improvements, including optimizing the performance of the buried ss-DSSCs and lowering the protecting layer resistance, will enable the design of more efficient WS-DSPECs.

123

Appendix A. Supporting Information for Chapter 2

Element x=1.5 x=2.25 C 47.26 35.68 N 0.4 2.32 O 34.01 51.19 Si 1.2 7.56 Ca 6.2 1.22 Nb 5.27 1.55 Ta 5.66 0.47 Total 100 100

Figure A-1. XRD patterns of 10-layer PDDA/nanosheets on quartz before (black) and after (red) three-day UV exposure. Note that the reflections close to 6° in Ca2NbxTa3-xO10 samples before UV irradiation are due to the presence of unexfoliated particles.

Figure A-2. Top (a) and side (b) view of the optimized structures in DFT calculations.

125

Figure A-3. Band structures for different nanosheet compositions. Fermi level (red dashed line) is set at 0 eV.

Table A-2. DFT-calculated work functions for different nanosheets.

Composition Ca2Nb3O10 Ca2Ta3O10 Sr2Nb3O10

Work function (eV) 7.22 7.18 7.37

126

Figure A-4. Partial density of states plotted relative to the Fermi level for different nanosheets. Oxygen 2p (red) and the d states from B site metal (blue) ions are highlighted.

Figure A-5. Nyquist (left) and Bode (right) plots of five-layer TiOx nanosheets on FTO substrate at -0.7 V in pH = 6.8 electrolyte. Red curves are the data fitted with a Randles circuit. Similar deviations of EIS spectra from the Randles circuit fits were observed for TiNbO5 nanosheets. TiOx 267,268 and TiNbO5 nanosheets were prepared as reported elsewhere.

127

Appendix B. Supporting Information for Chapter 3

Figure B-1. XRD profiles (a) and SEM images (b) for the three types of electrode under investigation.

Figure B-2. Resistance (a) and capacitance (b) of the TiO2 films determined from PEIS at varoius open-circuit potenitals under illumination. Error bars indicate fitting errors.

128

Figure B-3. PEIS Nyquist plots of (a) an unsensitized TiO2 electrode and (b) a TiCl4-treated electrode measured with 470-nm LED illumination at an intensity of 33.8 mW/cm2 from 1500 to 10 Hz.

Intensity-modulated photovoltage spectroscopy (IMVS)

IMVS is a convenient way of measuring charge recombination rates at semiconductor photoelectrodes and in dye-sensitized solar cells. This method records the modulation of photovoltage in response to a small sinusoidal modulation of light intensity (typically less than 10% of the DC illumination), We consider here a dye-sensitized TiO2 photoelectrode in aqueous solution without any electron donor. Under illumination, electrons from excited sensitizer molecules undergo a fast injection into the TiO2 conduction band. This process is known to occur on a sub-

9 -1 nanosecond time scale (k0>>10 s ). The injection current (푗푖푛푗) can be expressed as

푗푖푛푗 = 푞퐴퐼 (Eq. B-1)

where q is the electron charge, I is incident photon flux, and A is the absorption ratio. The recombination process occurs when the injected electrons (n) recombine with the oxidized sensitizer molecules at TiO2 surface at a rate k1, and thus the recombination current (Jrec) can be represented as

훽 푗푟푒푐 = 푞푘1푛 (Eq. B-2)

129

where 훽 is the recombination reaction order for electrons. Note that for simplicity, the definition of n here is the injected electron concentration, the concentration difference of the total and the dark electrons. The electron concentration change in TiO2 over time can be expressed as follows:

푑푛 푞 = 푗 − 푗 − 푗 (Eq. B-3) 푑푡 푖푛푗 푟푒푐 푒푥푡

where 푗푒푥푡 is the external current, which is 0 under open-circuit conditions. Under constant illumination (I0), (Eq. B-3) becomes

푑푛 0 = 퐴퐼 − 푘 푛훽 = 0 (Eq. B-4) 푑푡 0 1 0 The subscript 0 represents the steady-state condition.

When illumination is modulated at a frequency 휔 with an amplitude M,

푖휔푡 퐼 = 퐼0(1 + 푀푒 ) (Eq. B-5)

we assume that 푛 responds at the same frequency given by

푖휔푡 푛 = 푛0(1 + 푚푒 ) (Eq. B-6)

where m is the modulated component of n. When 훽 = 1, from (Eq. B-2)-(Eq. B-6), we can calculate the analytic expression form as follows:

2 M M푘1 M푘1휔 푚 = 𝑖휔 = 2 2 − 푖 2 2 (Eq. B-7) 1+ 푘1+휔 푘1+휔 푘1 Analyzing the above equation shows that the imaginary part of m reaches a maximum when 휔 =

푘1. Modulation of n further results in modulation of the electrode open-circuit potential under illumination (푉푙푖𝑔ℎ푡). In the case of a small light perturbation, we can assume that the electrode capacitance is unperturbed, and thus 푉푙푖𝑔ℎ푡 changes linearly with m. Therefore, one can read the recombination rate from IMVS spectra by identifying the frequency at which the imaginary part reaches a maximum.

When 훽 < 1, we can only numerically solve for the frequency response of n. We first calculated the time-domain profiles (Figure B-a) for I and n, which is then transformed into frequency domain

(Figure B-b) using fast Fourier transform to obtain the modulated information in n, including

130 amplitude, frequency, and phase angle. These were then used to construct the Nyquist plots (Figure

B-c) and Bode plots (Figure B-d). Data from the Nyquist plots appeared noisy, and this may be improved by using more sampling points per period (104 points were used in Figure B-a).

Nevertheless, the numerical simulation suggests that only at 훽=1, the frequency at apex in Bode plot corresponds to k1. The smaller 훽 is, the lower frequency the apex point shifts to.

131

Transient absorption spectroscopy

Estimation of laser pulse intensity (Ilaser) from energy (Elaser):

퐸Laser 6.5(mJ) 퐼 = = = 9.3 × 109 (mW/cm2 ) laser Area × Time 0.7 (cm2) × 10−8 (s)

Estimation of excited RuP concentration from the initial bleach amplitude at 470 nm:

Absorbance Change [RuP∗] = × Avogadro′s number Extinction coefficient × Film thickness

0.1 × 6.02 × 1023(mol−1) = 9000 (M−1cm−1) × 3 × 10−4 (cm) × 1000(cm3/L)

= 2.2 × 1019 (cm−3)

132

The extinction coefficient of RuP is 9300 M-1cm-1 at 458 nm. We here used 9000 M-1cm-1 as an estimate for the extinction coefficient (at 470 nm) difference of the ground-state and excited-state

RuP species. The calculated value of [RuP*] should be a lower estimate because the actual difference of the extinction coefficients would be smaller.

133

Appendix C. Supporting Information for Chapter 5

Figure C-1. J-V curves of a solid-state N3-sensitized solar cell with different aging time under AM 1.5 G illumination with a 410 nm long pass filter (100 mW cm-2).

Table C-1. Photovoltaic parameters from J-V characteristics of solid-state N3 sensitized solar cell with different aging times

-2 Aging time Jsc/mA cm Voc/V FF PCE(%)

0 day 0.29 -0.65 0.62 0.12

3 days 2.02 -0.77 0.44 0.68

5 days 3.29 -0.71 0.32 0.76

134

Figure C-2. a) Complex plane plots from electrochemical impedance spectra (EIS) of a pristine ss- DSSC, a heated ss-DSSC (heated under the same experimental conditions as used in the ALD process) and a ss-DSSC/ALD, respectively. The three frequency-dependent semicircles from high to low frequency, indicated by arrows, reflect charge transfer processes at the interface of Au/spiro- OMeTAD and TiO2/N3/spiro-OMeTAD, and in spiro-OMeTAD , respectively. b) Calculated series resistances obtained by fitting EIS data to the equivalent circuit shown in the inset. Rct, Rct2, and Rct3 represent electron transfer resistances at the interface of Au/spiro-OMeTAD, the interface of TiO2/N3/spiro-OMeTAD, and in the spiro-OMeTAD hole conductor layer, respectively.

Figure C-3. a) Current density as a function of time during 1.2 V electrodeposition of iridium oxide films on a ss-DSSC/ALD in 0.4 mM hydroxyiridate solution (prepared by alkaline hydrolysis of 2- [IrCl6] ) at pH 8.0. Inset: photographs taken before and after electrodeposition. b) Photograph of the experimental setup for electrodeposition of IrOx onto gold with a mechanic stirrer. WE1 and WE2 are the working electrodes connecting FTO and Au, respectively. RE and CE are the reference electrode and counter electrode. WE2 was used as the working electrode during the deposition process. c) Cross-sectional SEM image of an iridium oxide film deposited on a ss-DSSC/ALD electrode.

135

Figure C-4. Cyclic voltammogram of a ss-DSSC/ALD/IrOx electrode (the Au layer was used as the working electrode). The measurement was recorded at a scan rate of 5 mV/s.

Calculation of J-V values in Figure 5-4

The red and blue squares in Figure 5-4 were obtained by combining photovoltaic data from

Figure 5-3a, curve 4, with the measured difference between VFTO and VAu in Figure 5-5b. For example, at VFTO = 0.20 V, we measured VAu = 0.78 V (Figure 5-5b). The photovoltage of the buried ss-DSSC was △V= VFTO - VAu = -0.58 V, which corresponded to a photocurrent J = 0.39 mA/cm2 in

Figure 5-3a, line 4. We can thus plot the corresponding red and blue squares in Figure 5-4 as J vs

2 2 VFTO (0.39 mA/cm vs 0.20 V) and J vs VAu (0.39 mA/cm vs 0.78 V), respectively. Table C-2 shows the values obtained.

Table C-2. Calculated current densities (J) in Figure 5-4a with the measured Au (or IrOx) potential (VAu) vs the FTO potential (VFTO) from Figure 5-5b and the voltage-current curve of Figure 5-3a line 4.(△V= VFTO - VAu)

VFTO VAu △V J*

(V, vs Ag/AgCl) (V, vs Ag/AgCl) (V) (mA/cm2)

0.00 0.73 -0.73 -0.18

0.10 0.74 -0.64 0.11

0.20 0.78 -0.58 0.39

136

0.30 0.82 -0.52 0.61

0.40 0.86 -0.46 0.86

0.50 0.88 -0.38 1.20

0.60 0.89 -0.29 1.55

0.70 0.90 -0.20 1.84

0.80 0.91 -0.11 2.12

0.90 0.92 -0.02 2.36

1.00 0.93 0.07 2.58

137

red shaded regions represent QG and QC, respectively. Bottom: Collector-generator calibration measurement with a FTO/IrOx generator electrode at an applied bias of 1.12 V vs. Ag/AgCl from 300 s to 600 s (blue solid line). The red solid line is recorded from a Pt oxygen sensing electrode biased at -0.3 V vs Ag/AgCl. The experiment was performed in 0.1 M phosphate buffer at pH 6.7 with 0.1 M NaClO4. The blue and red shaded regions represent charges passing the generator and the collector, respectively.

138

Reference

(1) Feldman, D.; Barbose, G.; Margolis, R.; Darghouth, N.; James, T.; Weaver, S.; Goodrich, A.; Wiser, R. Photovoltaic System Pricing Trends: Historical, Recent, and Near-Term Projections–2014 Edition; Golden, CO (United States), 2014; Vol. 9. (2) International Energy Agency. In World Energy Outlook 2015. (3) Williams, J. H.; Debenedictis, A.; Ghanadan, R.; Mahone, A.; Moore, J.; Iii, W. R. M.; Price, S.; Torn, M. S. Science 2012, 335, 53–60. (4) Jacobson, M. Z.; Delucchi, M. A.; Cameron, M. A.; Frew, B. A. Proc. Natl. Acad. Sci. 2015, 112, 15060–15065. (5) Jensen, S. H.; Graves, C.; Mogensen, M.; Wendel, C.; Braun, R.; Hughes, G.; Gao, Z.; Barnett, S. a. Energy Environ. Sci. 2015, 8, 2471–2479. (6) Barton, E. E.; Rampulla, D. M.; Bocarsly, A. B. J. Am. Chem. Soc. 2008, 130, 6342– 6344. (7) Morris, A. J.; Meyer, G. J.; Fujita, E. Acc. Chem. Res. 2009, 42, 1983–1994. (8) Ramesha, G. K.; Brennecke, J. F.; Kamat, P. V. ACS Catal. 2014, 4, 3249–3254. (9) Hamdy, M. S.; Amrollahi, R.; Sinev, I.; Mei, B.; Mul, G. J. Am. Chem. Soc. 2014, 136, 594–597. (10) Sastre, F.; Puga, A. V; Liu, L.; Corma, A.; García, H. J. Am. Chem. Soc. 2014, 136, 6798–6801. (11) Schreier, M.; Gao, P.; Mayer, M. T.; Luo, J.; Moehl, T.; Nazeeruddin, M. K.; Tilley, S. D.; Grätzel, M. Energy Environ. Sci. 2015, 8, 855–861. (12) Sastre, F.; Oteri, M.; Corma, A.; García, H. Energy Environ. Sci. 2013, 6, 2211–2215. (13) Hoch, L. B.; Wood, T. E.; O’Brien, P. G.; Liao, K.; Reyes, L. M.; Mims, C. A.; Ozin, G. A. Adv. Sci. 2014, 1, 1400013. (14) Li, H.; Shang, J.; Ai, Z.; Zhang, L. J. Am. Chem. Soc. 2015, 137, 6393–6399. (15) Banerjee, A.; Yuhas, B. D.; Margulies, E. A.; Zhang, Y.; Shim, Y.; Wasielewski, M. R.; Kanatzidis, M. G. J. Am. Chem. Soc. 2015, 137, 2030–2034. (16) FUJISHIMA, A.; HONDA, K. Nature 1972, 238, 37–38.

139

(17) Maeda, K.; Teramura, K.; Lu, D.; Saito, N.; Inoue, Y.; Domen, K. Angew. Chemie Int. Ed. 2006, 45, 7806–7809. (18) Alstrum-Acevedo, J. H.; Brennaman, M. K.; Meyer, T. J. Inorg. Chem. 2005, 44, 6802–6827. (19) Blankenship, R. E.; Tiede, D. M.; Barber, J.; Brudvig, G. W.; Fleming, G.; Ghirardi, M.; Gunner, M. R.; Junge, W.; Kramer, D. M.; Melis, A.; et al. Science 2011, 332, 805–809. (20) Rühle, S. Sol. Energy 2016, 130, 139–147. (21) Bolton, J. R.; Strickler, S. J.; Connolly, J. S. Nature 1985, 316, 495–500. (22) Bolton, J. R. Sol. Energy 1996, 57, 37–50. (23) Hernández-Pagán, E. A.; Vargas-Barbosa, N. M.; Wang, T.; Zhao, Y.; Smotkin, E. S.; Mallouk, T. E. Energy Environ. Sci. 2012, 5, 7582. (24) Singh, M. R.; Papadantonakis, K.; Xiang, C.; Lewis, N. S. Energy Environ. Sci. 2015, 8, 2760–2767. (25) Khaselev, O.; Turner, J. A. Science. 1998, 280, 425–427. (26) Licht, S.; Wang, B.; Mukerji, S.; Soga, T.; Umeno, M.; Tributsch, H. Int. J. Hydrogen Energy 2001, 26, 653–659. (27) Reece, S. Y.; Hamel, J. a.; Sung, K.; Jarvi, T. D.; Esswein, a. J.; Pijpers, J. J. H.; Nocera, D. G. Science. 2011, 334, 645–648. (28) Luo, J.; Vermaas, D. A.; Bi, D.; Hagfeldt, A.; Smith, W. A.; Grätzel, M. Adv. Energy Mater. 2016, 6, 1600100. (29) Pinaud, B. A.; Benck, J. D.; Seitz, L. C.; Forman, A. J.; Chen, Z.; Deutsch, T. G.; James, B. D.; Baum, K. N.; Baum, G. N.; Ardo, S.; et al. Energy Environ. Sci. 2013, 6, 1983–2002. (30) Lewis, N. S. Science. 2016, 351, 353. (31) Osterloh, F. E. Chem. Soc. Rev. 2013, 42, 2294–2320. (32) Vlachopoulos, N.; Liska, P.; Augustynski, J.; Grätzel, M. J. Am. Chem. Soc. 1988, 110, 1216–1220. (33) O’Regan, B.; Grätzel, M. Nature 1991, 353, 737–740.

140

(34) Kakiage, K.; Aoyama, Y.; Yano, T.; Oya, K.; Fujisawa, J.; Hanaya, M. Chem. Commun. 2015, 51, 15894–15897. (35) Youngblood, J. W.; Lee, S. H. A.; Kobayashi, Y.; Hernandez-Pagan, E. A.; Hoertz, P. G.; Moore, T. A.; Moore, A. L.; Gust, D.; Mallouk, T. E. J. Am. Chem. Soc. 2009, 131, 926–927. (36) Kudo, A.; Miseki, Y. Chem. Soc. Rev. 2009, 38, 253–278. (37) Kay, A.; Cesar, I.; Grätzel, M. J. Am. Chem. Soc. 2006, 128, 15714–15721. (38) van de Krol, R.; Liang, Y.; Schoonman, J.; Krol, R. Van De; Liang, Y.; Schoonman, J. J. Mater. Chem. 2008, 18, 2311. (39) Kim, T. W.; Choi, K.-S. Science. 2014, 343, 990–994. (40) Kim, T. W.; Ping, Y.; Galli, G. A.; Choi, K.-S. Nat. Commun. 2015, 6, 8769. (41) Rowley, J. G.; Farnum, B. H.; Ardo, S.; Meyer, G. J. J. Phys. Chem. Lett. 2010, 1, 3132–3140. (42) Brimblecombe, R.; Koo, A.; Dismukes, G. C.; Swiegers, G. F.; Spiccia, L. J. Am. Chem. Soc. 2010, 132, 2892–2894. (43) Li, L.; Duan, L.; Xu, Y.; Gorlov, M.; Hagfeldt, A.; Sun, L. Chem. Commun. 2010, 46, 7307. (44) Moore, G. F.; Blakemore, J. D.; Milot, R. L.; Hull, J. F.; Song, H.; Cai, L.; Schmuttenmaer, C. A.; Crabtree, R. H.; Brudvig, G. W. Energy Environ. Sci. 2011, 4, 2389. (45) Bledowski, M.; Wang, L.; Ramakrishnan, A.; Khavryuchenko, O. V.; Khavryuchenko, V. D.; Ricci, P. C.; Strunk, J.; Cremer, T.; Kolbeck, C.; Beranek, R. Phys. Chem. Chem. Phys. 2011, 13, 21511. (46) Bledowski, M.; Wang, L.; Ramakrishnan, A.; Bétard, A.; Khavryuchenko, O. V.; Beranek, R. ChemPhysChem 2012, 13, 3018–3024. (47) Lee, S. A.; Zhao, Y.; Hernandez-Pagan, E. A.; Blasdel, L.; Youngblood, W. J.; Mallouk, T. E. Faraday Discuss. 2012, pp 165–176. (48) Zhao, Y.; Swierk, J. R.; Megiatto, J. D.; Sherman, B.; Youngblood, W. J.; Qin, D.; Lentz, D. M.; Moore, A. L.; Moore, T. A.; Gust, D.; et al. Proc. Natl. Acad. Sci. 2012, 109, 15612–15616.

141

(49) Xiang, X.; Fielden, J.; Rodríguez-Córdoba, W.; Huang, Z.; Zhang, N.; Luo, Z.; Musaev, D. G.; Lian, T.; Hill, C. L. J. Phys. Chem. C 2013, 117, 918–926. (50) Gao, Y.; Ding, X.; Liu, J.; Wang, L.; Lu, Z.; Li, L.; Sun, L. J. Am. Chem. Soc. 2013, 135, 4219–4222. (51) Alibabaei, L.; Brennaman, M. K.; Norris, M. R.; Kalanyan, B.; Song, W.; Losego, M. D.; Concepcion, J. J.; Binstead, R. A.; Parsons, G. N.; Meyer, T. J. Proc. Natl. Acad. Sci. 2013, 110, 20008–20013. (52) Swierk, J. R.; McCool, N. S.; Saunders, T. P.; Barber, G. D.; Strayer, M. E.; Vargas- Barbosa, N. M.; Mallouk, T. E. J. Phys. Chem. C 2014, 118, 17046–17053. (53) Gao, Y.; Zhang, L.; Ding, X.; Sun, L. Phys. Chem. Chem. Phys. 2014, 16, 12008. (54) Ding, X.; Gao, Y.; Zhang, L.; Yu, Z.; Liu, J.; Sun, L. ACS Catal. 2014, 4, 2347–2350. (55) Kirner, J. T.; Stracke, J. J.; Gregg, B. A.; Finke, R. G. ACS Appl. Mater. Interfaces 2014, 6, 13367–13377. (56) Ding, X.; Gao, Y.; Zhang, L.; Yu, Z.; Liu, J.; Sun, L. Electrochim. Acta 2014, 149, 337–340. (57) Poddutoori, P. K.; Thomsen, J. M.; Milot, R. L.; Sheehan, S. W.; Negre, C. F. A.; Garapati, V. K. R.; Schmuttenmaer, C. A.; Batista, V. S.; Brudvig, G. W.; van der Est, A. J. Mater. Chem. A 2015, 3, 3868–3879. (58) Fielden, J.; Sumliner, J. M.; Han, N.; Geletii, Y. V.; Xiang, X.; Musaev, D. G.; Lian, T.; Hill, C. L. Chem. Sci. 2015, 6, 5531–5543. (59) Khavryuchenko, O. V.; Wang, L.; Mitoraj, D.; Peslherbe, G. H.; Beranek, R. J. Coord. Chem. 2015, 68, 3317–3327. (60) Lapides, A. M.; Sherman, B. D.; Brennaman, M. K.; Dares, C. J.; Skinner, K. R.; Templeton, J. L.; Meyer, T. J. Chem. Sci. 2015, 6, 6398–6406. (61) Swierk, J. R.; Méndez-Hernández, D. D.; McCool, N. S.; Liddell, P.; Terazono, Y.; Pahk, I.; Tomlin, J. J.; Oster, N. V; Moore, T. A.; Moore, A. L.; et al. Proc. Natl. Acad. Sci. 2015, 112, 1681–1686. (62) Alibabaei, L.; Sherman, B. D.; Norris, M. R.; Brennaman, M. K.; Meyer, T. J. Proc. Natl. Acad. Sci. 2015, 112, 5899–5902.

142

(63) Li, F.; Fan, K.; Xu, B.; Gabrielsson, E.; Daniel, Q.; Li, L.; Sun, L. J. Am. Chem. Soc. 2015, 137, 9153–9159. (64) Sherman, B. D.; Ashford, D. L.; Lapides, A. M.; Sheridan, M. V.; Wee, K.-R.; Meyer, T. J. J. Phys. Chem. Lett. 2015, 6, 3213–3217. (65) Michaux, K. E.; Gambardella, A. A.; Alibabaei, L.; Ashford, D. L.; Sherman, B. D.; Binstead, R. A.; Meyer, T. J.; Murray, R. W. J. Phys. Chem. C 2015, 119, 17023– 17027. (66) Wee, K.-R.; Sherman, B. D.; Brennaman, M. K.; Sheridan, M. V.; Nayak, A.; Alibabaei, L.; Meyer, T. J. J. Mater. Chem. A 2016, 4, 2969–2975. (67) Ding, X.; Gao, Y.; Ye, L.; Zhang, L.; Sun, L. ChemSusChem 2015, 8, 3992–3995. (68) Yamamoto, M.; Wang, L.; Li, F.; Fukushima, T.; Tanaka, K.; Sun, L.; Imahori, H. Chem. Sci. 2016, 7, 1430–1439. (69) Li, H.; Li, F.; Wang, Y.; Bai, L.; Yu, F.; Sun, L. Chempluschem 2016, 81, 1056–1059. (70) Leem, G.; Sherman, B. D.; Burnett, A. J.; Morseth, Z. A.; Wee, K.-R.; Papanikolas, J. M.; Meyer, T. J.; Schanze, K. S. ACS Energy Lett. 2016, 1, 339–343. (71) Sherman, B. D.; Xie, Y.; Sheridan, M. V.; Wang, D.; Shaffer, D. W.; Meyer, T. J.; Concepcion, J. J. ACS Energy Lett. 2016, 2, 124–128. (72) Sherman, B. D.; Sheridan, M. V.; Wee, K.-R.; Marquard, S. L.; Wang, D.; Alibabaei, L.; Ashford, D. L.; Meyer, T. J. J. Am. Chem. Soc. 2016, 138, 16745–16753. (73) Swierk, J. R.; McCool, N. S.; Saunders, T. P.; Barber, G. D.; Mallouk, T. E. J. Am. Chem. Soc. 2014, 136, 10974–10982. (74) Swierk, J. R.; McCool, N. S.; Mallouk, T. E. J. Phys. Chem. C 2015, 119, 13858– 13867. (75) McCool, N. S.; Swierk, J. R.; Nemes, C. T.; Saunders, T. P.; Schmuttenmaer, C. A.; Mallouk, T. E. ACS Appl. Mater. Interfaces 2016, 8, 16727–16735. (76) Nelson, J.; Eppler, A. M.; Ballard, I. M. J. Photochem. Photobiol. A Chem. 2002, 148, 25–31. (77) Hanson, K.; Brennaman, M. K.; Ito, A.; Luo, H.; Song, W.; Parker, K. A.; Ghosh, R.; Norris, M. R.; Glasson, C. R. K.; Concepcion, J. J.; et al. J. Phys. Chem. C 2012, 116, 14837–14847.

143

(78) Halverson, A. F.; Zhu, K.; Erslev, P. T.; Kim, J. Y.; Neale, N. R.; Frank, A. J. Nano Lett. 2012, 12, 2112–2116. (79) Fonash, S. J. Soalr Cell Device Physics, 2nd ed.; Elsevier Inc., 2010. (80) Chen, S. G.; Chappel, S.; Diamant, Y.; Zaban, A. Chem. Mater. 2001, 13, 4629–4634. (81) Kay, A.; Grätzel, M. Chem. Mater. 2002, 14, 2930–2935. (82) Palomares, E.; Clifford, J. N.; Haque, S. A.; Lutz, T.; Durrant, J. R. J. Am. Chem. Soc. 2003, 125, 475–482. (83) Ashford, D. L.; Gish, M. K.; Vannucci, A. K.; Brennaman, M. K.; Templeton, J. L.; Papanikolas, J. M.; Meyer, T. J. Chem. Rev. 2015, 115, 13006–13049. (84) Knauf, R. R.; Kalanyan, B.; Parsons, G. N.; Dempsey, J. L. J. Phys. Chem. C 2015, 119, 28353–28360. (85) McCool, N. S.; Swierk, J. R.; Nemes, C. T.; Schmuttenmaer, C. A.; Mallouk, T. E. J. Phys. Chem. Lett. 2016, 2930–2934. (86) Gish, M. K.; Lapides, A. M.; Brennaman, M. K.; Templeton, J. L.; Meyer, T. J.; Papanikolas, J. M. J. Phys. Chem. Lett. 2016, 5297–5301. (87) George, S. M. Chem. Rev. 2010, 110, 111–131. (88) Swierk, J. R.; McCool, N. S.; Nemes, C. T.; Mallouk, T. E.; Schmuttenmaer, C. A. J. Phys. Chem. C 2016, 120, 5940–5948. (89) Prasittichai, C.; Avila, J. R.; Farha, O. K.; Hupp, J. T. J. Am. Chem. Soc. 2013, 135, 16328–16331. (90) Flynn, C. J.; McCullough, S. M.; Oh, E.; Li, L.; Mercado, C. C.; Farnum, B. H.; Li, W.; Donley, C. L.; You, W.; Nozik, A. J.; et al. ACS Appl. Mater. Interfaces 2016, 8, 4754–4761. (91) Hanson, K.; Losego, M. D.; Kalanyan, B.; Parsons, G. N.; Meyer, T. J. Nano Lett. 2013, 13, 4802–4809. (92) Hanson, K.; Losego, M. D.; Kalanyan, B.; Ashford, D. L.; Parsons, G. N.; Meyer, T. J. Chem. Mater. 2013, 25, 3–5. (93) Swierk, J. R.; Regan, K. P.; Jiang, J.; Brudvig, G. W.; Schmuttenmaer, C. A. ACS Energy Lett. 2016, 603–606.

144

(94) Son, H.-J.; Wang, X.; Prasittichai, C.; Jeong, N. C.; Aaltonen, T.; Gordon, R. G.; Hupp, J. T. J. Am. Chem. Soc. 2012, 134, 9537–9540. (95) Son, H.-J.; Prasittichai, C.; Mondloch, J. E.; Luo, L.; Wu, J.; Kim, D. W.; Farha, O. K.; Hupp, J. T. J. Am. Chem. Soc. 2013, 135, 11529–11532. (96) Kim, D. H.; Losego, M. D.; Hanson, K.; Alibabaei, L.; Lee, K.; Meyer, T. J.; Parsons, G. N. Phys. Chem. Chem. Phys. 2014, 16, 8615–8622. (97) Son, H.-J.; Kim, C. H.; Kim, D. W.; Jeong, N. C.; Prasittichai, C.; Luo, L.; Wu, J.; Farha, O. K.; Wasielewski, M. R.; Hupp, J. T. ACS Appl. Mater. Interfaces 2015, 7, 5150–5159. (98) Wee, K. R.; Brennaman, M. K.; Alibabaei, L.; Farnum, B. H.; Sherman, B.; Lapides, A. M.; Meyer, T. J. J. Am. Chem. Soc. 2014, 136, 13514–13517. (99) Zhang, L.; Gao, Y.; Ding, X. Electrochim. Acta 2016, 207, 130–134. (100) Odobel, F.; Le Pleux, L.; Pellegrin, Y.; Blart, E. Acc. Chem. Res. 2010, 43, 1063– 1071. (101) Brautigam, M.; Kubel, J.; Schulz, M.; Vos, J. G.; Dietzek, B. Phys. Chem. Chem. Phys. 2015, 17, 7823–7830. (102) Antila, L. J.; Ghamgosar, P.; Maji, S.; Tian, H.; Ott, S.; Hammarström, L. ACS Energy Lett. 2016, 1106–1111. (103) Brown, A. M.; Antila, L. J.; Mirmohades, M.; Pullen, S.; Ott, S.; Hammarstr??m, L. J. Am. Chem. Soc. 2016, 138, 8060–8063. (104) Tong, L.; Iwase, A.; Nattestad, A.; Bach, U.; Weidelener, M.; Götz, G.; Mishra, A.; Bäuerle, P.; Amal, R.; Wallace, G. G.; et al. Energy Environ. Sci. 2012, 5, 9472. (105) Li, L.; Duan, L.; Wen, F.; Li, C.; Wang, M.; Hagfeldt, A.; Sun, L. Chem. Commun. (Camb). 2012, 48, 988–990. (106) Ji, Z.; He, M.; Huang, Z.; Ozkan, U.; Wu, Y. J. Am. Chem. Soc. 2013, 135, 11696– 11699. (107) Fan, K.; Li, F.; Wang, L.; Daniel, Q.; Gabrielsson, E.; Sun, L. Phys. Chem. Chem. Phys. 2014, 16, 25234–25240. (108) Lattach, Y.; Fortage, J.; Deronzier, A.; Moutet, J.-C. ACS Appl. Mater. Interfaces 2015, 7, 4476–4480.

145

(109) Click, K. A.; Beauchamp, D. R.; Huang, Z.; Chen, W.; Wu, Y. J. Am. Chem. Soc. 2016, 138, 1174–1179. (110) van den Bosch, B.; Rombouts, J. A.; Orru, R. V. A.; Reek, J. N. H.; Detz, R. J. ChemCatChem 2016, 8, 1392–1398. (111) Gross, M. A.; Creissen, C. E.; Orchard, K. L.; Reisner, E. Chem. Sci. 2016, 6, 242– 247. (112) Kamire, R. J.; Majewski, M. B.; Hoffeditz, W. L.; Phelan, B. T.; Farha, O. K.; Hupp, J. T.; Wasielewski, M. R. Chem. Sci. 2017, 22, 32–57. (113) Kaeffer, N.; Massin, J.; Lebrun, C.; Renault, O.; Chavarot-Kerlidou, M.; Artero, V. J. Am. Chem. Soc. 2016, 138, 12308–12311. (114) Tench, D. M.; Yeager, E. J. Electrochem. Soc. 1973, 120, 164. (115) Koffyberg, F. P. J. Electrochem. Soc. 1981, 128, 2476. (116) Rouse, T. O.; Weininger, J. L. J. Electrochem. Soc. 1966, 113, 184. (117) Castillo, C. E.; Gennari, M.; Stoll, T.; Fortage, J.; Deronzier, A.; Collomb, M. N.; Sandroni, M.; Légalité, F.; Blart, E.; Pellegrin, Y.; et al. J. Phys. Chem. C 2015, 119, 5806–5818. (118) Vera, F.; Schrebler, R.; Muñoz, E.; Suarez, C.; Cury, P.; Gómez, H.; Córdova, R.; Marotti, R. E.; Dalchiele, E. A. Thin Solid Films 2005, 490, 182–188. (119) Lepleux, L.; Chavillon, B.; Pellegrini, Y.; Blart, E.; Cario, L.; Jobic, S.; Odobel, F. Inorg. Chem. 2009, 48, 8245–8250. (120) Kuang, D.-B.; Lei, B.-X.; Pan, Y.-P.; Yu, X.-Y.; Su, C.-Y. J. Phys. Chem. C 2009, 113, 5508–5513. (121) Zhu, H.; Hagfeldt, A.; Boschloo, G. J. Phys. Chem. C 2007, 111, 17455–17458. (122) Qin, P.; Zhu, H.; Edvinsson, T.; Boschloo, G.; Hagfeldt, A.; Sun, L. J. Am. Chem. Soc. 2008, 130, 8570–8571. (123) Li, L.; Gibson, E. A.; Qin, P.; Boschloo, G.; Gorlov, M.; Hagfeldt, A.; Sun, L. Adv. Mater. 2010, 22, 1759–1762. (124) Sumikura, S.; Mori, S.; Shimizu, S.; Usami, H.; Suzuki, E. J. Photochem. Photobiol. A Chem. 2008, 199, 1–7.

146

(125) Wood, C. J.; Summers, G. H.; Clark, C. A.; Kaeffer, N.; Braeutigam, M.; Carbone, L. R.; D’Amario, L.; Fan, K.; Farré, Y.; Narbey, S.; et al. Phys. Chem. Chem. Phys. 2016, 18, 10727–10738. (126) Mori, S.; Fukuda, S.; Sumikura, S.; Takeda, Y.; Tamaki, Y.; Suzuki, E.; Abe, T. J. Phys. Chem. C 2008, 112, 16134–16139. (127) Nakade, S.; Kanzaki, T.; Wada, Y.; Yanagida, S. Langmuir 2005, 21, 10803–10807. (128) Swierk, J. R.; Mallouk, T. E. Chem. Soc. Rev. 2013, 42, 2357–2387. (129) Zhang, X. L.; Zhang, Z.; Chen, D.; Bäuerle, P.; Bach, U.; Cheng, Y.-B. Chem. Commun. (Camb). 2012, 48, 9885–9887. (130) Gibson, E. A.; Awais, M.; Dini, D.; Dowling, D. P.; Pryce, M. T.; Vos, J. G.; Boschloo, G.; Hagfeldt, A. Phys. Chem. Chem. Phys. 2013, 15, 2411–2420. (131) Uehara, S.; Sumikura, S.; Suzuki, E.; Mori, S. Energy Environ. Sci. 2010, 3, 641. (132) Natu, G.; Huang, Z.; Ji, Z.; Wu, Y. Langmuir 2012, 28, 950–956. (133) Vargas-Barbosa, N. M.; Geise, G. M.; Hickner, M. A.; Mallouk, T. E. ChemSusChem 2014, 7, 3017–3020. (134) Chawla, P.; Tripathi, M. Int. J. Energy Res. 2015, 39, 1579–1596. (135) Lee, C.-P.; Lin, R. Y.-Y.; Lin, L.-Y.; Li, C.-T.; Chu, T.-C.; Sun, S.-S.; Lin, J. T.; Ho, K.-C. RSC Adv. 2015, 5, 23810–23825. (136) Calogero, G.; Bartolotta, A.; Di Marco, G.; Di Carlo, A.; Bonaccorso, F. Chem. Soc. Rev. 2015, 44, 3244–3294. (137) Kalyanasundaram, K. Coord. Chem. Rev. 1982, 46, 159–244. (138) Keller, S. W.; Johnson, S. a.; Brigham, E. S.; Yonemoto, E. H.; Mallouk, T. E. J. Am. Chem. Soc. 1995, 117, 12879–12880. (139) Kaschak, D. M.; Lean, J. T.; Waraksa, C. C.; Saupe, G. B.; Usami, H.; Mallouk, T. E. J. Am. Chem. Soc. 1999, 121, 3435–3445. (140) Odobel, F.; Pellegrin, Y.; Gibson, E. A.; Hagfeldt, A.; Smeigh, A. L.; Hammarström, L. Coord. Chem. Rev. 2012, 256, 2414–2423. (141) Farré, Y.; Zhang, L.; Pellegrin, Y.; Planchat, A.; Blart, E.; Boujtita, M.; Hammarström, L.; Jacquemin, D.; Odobel, F. J. Phys. Chem. C 2016, 120, 7923– 7940.

147

(142) Baffert, C.; Artero, V.; Fontecave, M. Inorg. Chem. 2007, 46, 1817–1824. (143) Dempsey, J. L.; Brunschwig, B. S.; Winkler, J. R.; Gray, H. B. Acc. Chem. Res. 2009, 42, 1995–2004. (144) Mulfort, K. L.; Utschig, L. M. Acc. Chem. Res. 2016, 49, 835–843. (145) Braumüller, M.; Schulz, M.; Sorsche, D.; Pfeffer, M.; Schaub, M.; Popp, J.; Park, B.-W.; Hagfeldt, A.; Dietzek, B.; Rau, S. Dalt. Trans. 2015, 44, 5577–5586. (146) Jacques, P.-A.; Artero, V.; Pécaut, J.; Fontecave, M. Proc. Natl. Acad. Sci. U. S. A. 2009, 106, 20627–20632. (147) Dutta, A.; DuBois, D. L.; Roberts, J. A. S.; Shaw, W. J. Proc. Natl. Acad. Sci. U. S. A. 2014, 111, 16286–16291. (148) Lee, H.; Kepley, L. J.; Hong, H. G.; Akhter, S.; Mallouk, T. E. J. Phys. Chem. 1988, 92, 2597–2601. (149) Ishida, T.; Terada, K.; Hasegawa, K.; Kuwahata, H.; Kusama, K.; Sato, R.; Nakano, M.; Naitoh, Y.; Haga, M. Appl. Surf. Sci. 2009, 255, 8824–8830. (150) Hanson, K.; Torelli, D. A.; Vannucci, A. K.; Brennaman, M. K.; Luo, H.; Alibabaei, L.; Song, W.; Ashford, D. L.; Norris, M. R.; Glasson, C. R. K.; et al. Angew. Chemie Int. Ed. 2012, 51, 12782–12785. (151) Ma, R.; Sasaki, T. Acc. Chem. Res. 2015, 48, 136–143. (152) Osada, M.; Sasaki, T. J. Mater. Chem. 2009, 19, 2503. (153) Osterloh, F. E. Chem. Mater. 2008, 20, 35–54. (154) Domen, K.; Kudo, A.; Shibata, M.; Tanaka, A.; Maruya, K.-I.; Onishi, T. J. Chem. Soc. Chem. Commun. 1986, No. 23, 1706. (155) Sayama, K.; Tanaka, A.; Domen, K.; Maruya, K.; Onishi, T. J. Phys. Chem. 1991, 95, 1345–1348. (156) Kim, Y. Il; Salim, S.; Huq, M. J.; Mallouk, T. E. J. Am. Chem. Soc. 1991, 113, 9561– 9563. (157) Kim, Y. Il; Atherton, S. J.; Brigham, E. S.; Mallouk, T. E. J. Phys. Chem. 1993, 97, 11802–11810. (158) Saupe, G. B.; Mallouk, T. E.; Kim, W.; Schmehl, R. H. J. Phys. Chem. B 1997, 101, 2508–2513.

148

(159) Keller, S. W.; Kim, H.; Mallouk, T. E. J. Am. Chem. Soc. 1994, 116, 8817–8818. (160) Kleinfeld, E. R.; Ferguson, G. S. Science. 1994, 265, 370–373. (161) Saupe, G. B.; Waraksa, C. C.; Kim, H. N.; Han, Y. J.; Kaschak, D. M.; Skinner, D. M.; Mallouk, T. E. Chem. Mater. 2000, 12, 1556–1562. (162) Ma, R.; Bando, Y.; Sasaki, T. J. Phys. Chem. B 2004, 108, 2115–2119. (163) Maeda, K.; Eguchi, M.; Youngblood, W. J.; Mallouk, T. E. Chem. Mater. 2008, 20, 6770–6778. (164) Maeda, K.; Eguchi, M.; Youngblood, J. W.; Mallouk, T. E. Chem. Mater. 2009, 21, 3611–3617. (165) Maeda, K.; Mallouk, T. E. J. Mater. Chem. 2009, 19, 4813. (166) Sabio, E. M.; Chamousis, R. L.; Browning, N. D.; Osterloh, F. E. J. Phys. Chem. C 2012, 116, 3161–3170. (167) Abe, R.; Shinmei, K.; Koumura, N.; Hara, K.; Ohtani, B. J. Am. Chem. Soc. 2013, 135, 16872–16884. (168) Oshima, T.; Lu, D.; Ishitani, O.; Maeda, K. Angew. Chemie - Int. Ed. 2015, 54, 2698– 2702. (169) Osterloh, F. E. Springer, Cham, 2015; pp 105–142. (170) Nozik, A. J.; Memming, R. J. Phys. Chem. 1996, 100, 13061–13078. (171) Chen, Z.; Dinh, H. N.; Miller, E. Photoelectrochemical Water Splitting; SpringerBriefs in Energy; Springer New York: New York, NY, 2013. (172) Sakai, N.; Ebina, Y.; Takada, K.; Sasaki, T. J. Am. Chem. Soc. 2004, 126, 5851–5858. (173) Akatsuka, K.; Takanashi, G.; Ebina, Y.; Haga, M. A.; Sasaki, T. J. Phys. Chem. C 2012, 116, 12426–12433. (174) Redmond, G.; Fitzmaurice, D. J. Phys. Chem. 1993, 97, 1426–1430. (175) Chamousis, R. L.; Osterloh, F. E. Energy Environ. Sci. 2014, 7, 736. (176) Butler, M. A.; Ginley, D. S. J. Electrochem. Soc. 1978, 125, 228. (177) Matsumoto, Y. J. Solid State Chem. 1996, 126, 227–234. (178) Maeda, K.; Sahara, G.; Eguchi, M.; Ishitani, O. ACS Catal. 2015, 5, 1700–1707. (179) Maeda, K.; Eguchi, M.; Lee, S. H. A.; Youngblood, W. J.; Hata, H.; Mallouk, T. E. J. Phys. Chem. C 2009, 113, 7962–7969.

149

(180) Gelderman, K.; Lee, L.; Donne, S. W. J. Chem. Educ. 2007, 84, 685. (181) Kavan, L.; Tétreault, N.; Moehl, T.; Grätzel, M. J. Phys. Chem. C 2014, 118, 16408– 16418. (182) van de Krol, R.; Goossens, A.; Schoonman, J. J. Electrochem. Soc. 1997, 144, 1723. (183) Maeda, K.; Eguchi, M.; Oshima, T. Angew. Chemie - Int. Ed. 2014, 53, 13164–13168. (184) Thangadurai, V.; Schmid-Beurmann, P.; Weppner, W. J. Solid State Chem. 2001, 158, 279–289. (185) Xu, F. F.; Ebina, Y.; Bando, Y.; Sasaki, T. J. Phys. Chem. B 2003, 107, 9638–9645. (186) Clark, S. J.; Segall, M. D.; Pickard, C. J.; Hasnip, P. J.; Probert, M. I. J.; Refson, K.; Payne, M. C. Zeitschrift für Krist. - Cryst. Mater. 2005, 220, 567–570. (187) Strayer, M. E.; Senftle, T. P.; Winterstein, J. P.; Vargas-Barbosa, N. M.; Sharma, R.; Rioux, R. M.; Janik, M. J.; Mallouk, T. E. J. Am. Chem. Soc. 2015, 137, 16216– 16224. (188) Perdew, J. P.; Burke, K.; Ernzerhof, M. Phys. Rev. Lett. 1996, 77, 3865–3868. (189) Segall, M. D.; Shah, R.; Pickard, C. J.; Payne, M. C. Phys. Rev. B 1996, 54, 16317– 16320. (190) Krukau, A. V.; Vydrov, O. A.; Izmaylov, A. F.; Scuseria, G. E. J. Chem. Phys. 2006, 125, 224106. (191) Schaak, R. E.; Mallouk, T. E. Chem. Mater. 2002, 14, 1455–1471. (192) Decher, G. Science. 1997, 277, 1232–1237. (193) Bohren, C. F.; Huffman, D. R. Absorption and Scattering of Light by Small Particles; Bohren, C. F., Huffman, D. R., Eds.; Wiley-VCH Verlag GmbH: Weinheim, Germany, 1998. (194) Hirschorn, B.; Orazem, M. E.; Tribollet, B.; Vivier, V.; Frateur, I.; Musiani, M. Electrochim. Acta 2010, 55, 6218–6227. (195) Brug, G. J.; van den Eeden, a. L. G.; Sluyters-Rehbach, M.; Sluyters, J. H. Journal of Electroanalytical Chemistry and Interfacial Electrochemistry. 1984, pp 275–295. (196) Hsu, C. H.; Mansfeld, F. Corrosion 2001, 57, 747–748. (197) De Gryse, R.; Gomes, W. P.; Cardon, F.; Vennik, J. J. Electrochem. Soc. 1975, 122, 711.

150

(198) Sasaki, T.; Ebina, Y.; Fukuda, K.; Tanaka, T.; Harada, M.; Watanabe, M. Chem. Mater. 2002, 14, 3524–3530. (199) Osada, M.; Akatsuka, K.; Ebina, Y.; Funakubo, H.; Ono, K.; Takada, K.; Sasaki, T. ACS Nano 2010, 4, 5225–5232. (200) Matsumoto, Y.; Yoshikawa, T.; Sato, E. ichi. Journal of the Electrochemical Society. The Electrochemical Society May 1, 1989, pp 1389–1391. (201) Chen, Z. X.; Chen, Y.; Jiang, Y. S. J. Phys. Chem. B 2002, 106, 9986–9992. (202) Draskovic, T. I.; Wang, T.; Henderson, C. N.; Mallouk, T. E. Int. J. Hydrogen Energy 2014, 39, 4576–4580. (203) Brennaman, M. K.; Dillon, R. J.; Alibabaei, L.; Gish, M. K.; Dares, C. J.; Ashford, D. L.; House, R. L.; Meyer, G. J.; Papanikolas, J. M.; Meyer, T. J. J. Am. Chem. Soc. 2016, 138, 13085–13102. (204) Xu, P.; McCool, N. S.; Mallouk, T. E. Nano Today 2017, 14, 42–58. (205) Li, F.; Yang, H.; Li, W.; Sun, L. Joule 2018, 2, 36–60. (206) Chen, H.-Y.; Ardo, S. Nat. Chem. 2017, 10, 17–23. (207) Jiang, J.; Spies, J. A.; Swierk, J. R.; Matula, A. J.; Regan, K. P.; Romano, N.; Brennan, B. J.; Crabtree, R. H.; Batista, V. S.; Schmuttenmaer, C. A.; et al. J. Phys. Chem. C 2018, 122, 13529–13539. (208) Brennan, B. J.; Regan, K. P.; Durrell, A. C.; Schmuttenmaer, C. A.; Brudvig, G. W. ACS Energy Lett. 2017, 2, 168–173. (209) Thompson, D. W.; Kelly, C. A.; Farzad, F.; Meyer, G. J. Langmuir 1999, 15, 650– 653. (210) Kilså, K.; Mayo, E. I.; Kuciauskas, D.; Villahermosa, R.; Lewis, N. S.; Winkler, J. R.; Gray, H. B. J. Phys. Chem. A 2003, 107, 3379–3383. (211) Kelly, C. A.; Farzad, F.; Thompson, D. W.; Stipkala, J. M.; Meyer, G. J. Langmuir 1999, 15, 7047–7054. (212) Haque, S. A.; Tachibana, Y.; Willis, R. L.; Moser, J. E.; Grätzel, M.; Klug, D. R.; Durrant, J. R. J. Phys. Chem. B 2000, 104, 538–547. (213) Brigham, E. C.; Meyer, G. J. J. Phys. Chem. C 2014, 118, 7886–7893. (214) Ponomarev, E. A.; Peter, L. M. J. Electroanal. Chem. 1995, 396, 219–226.

151

(215) Schlichthörl, G.; Huang, S. Y.; Sprague, J.; Frank, A. J. J. Phys. Chem. B 1997, 101, 8141–8155. (216) Krüger, J.; Plass, R.; Grätzel, M.; Cameron, P. J.; Peter, L. M. J. Phys. Chem. B 2003, 107, 7536–7539. (217) Pockett, A.; Eperon, G. E.; Peltola, T.; Snaith, H. J.; Walker, A.; Peter, L. M.; Cameron, P. J. J. Phys. Chem. C 2015, 119, 3456–3465. (218) Thorne, J. E.; Jang, J.-W.; Liu, E. Y.; Wang, D. Chem. Sci. 2016, 7, 3347–3354. (219) Hagfeldt, A.; Boschloo, G.; Sun, L.; Kloo, L.; Pettersson, H. Chem. Rev. 2010, 110, 6595–6663. (220) O’Regan, B. C.; Durrant, J. R.; Sommeling, P. M.; Bakker, N. J. J. Phys. Chem. C 2007, 111, 14001–14010. (221) Murakami, T. N.; Miyadera, T.; Funaki, T.; Cojocaru, L.; Kazaoui, S.; Chikamatsu, M.; Segawa, H. ACS Appl. Mater. Interfaces 2017, 9, 36708–36714. (222) Choe, G.; Kang, J.; Ryu, I.; Song, S. W.; Kim, H. M.; Yim, S. Sol. Energy 2017, 155, 1148–1156. (223) Adli, H. K.; Harada, T.; Nakanishi, S.; Ikeda, S. Phys. Chem. Chem. Phys. 2017, 19, 26898–26905. (224) Huang, S. Y.; Schlichthörl, G.; Nozik, A. J.; Grätzel, M.; Frank, A. J. J. Phys. Chem. B 1997, 101, 2576–2582. (225) Ocakoglu, K.; Yakuphanoglu, F.; Durrant, J. R.; Icli, S. Sol. Energy Mater. Sol. Cells 2008, 92, 1047–1053. (226) Barnes, P. R. F.; Anderson, A. Y.; Durrant, J. R.; O’Regan, B. C. Phys. Chem. Chem. Phys. 2011, 13, 5798–5816. (227) Gillaizeau-Gauthier, I.; Odobel, F.; Alebbi, M.; Argazzi, R.; Costa, E.; Bignozzi, C. A.; Qu, P.; Meyer, G. J. Inorg. Chem. 2001, 40, 6073–6079. (228) Rimshaw, A.; Grieco, C.; Asbury, J. B. Rev. Sci. Instrum. 2015, 86, 066107. (229) Barr, T. J.; Meyer, G. J. ACS Energy Lett. 2017, 2, 2335–2340. (230) Lin, W. M. M.; Bozyigit, D.; Yarema, O.; Wood, V. J. Phys. Chem. C 2016, 120, 12900–12908.

152

(231) Abrahamsson, M.; Johansson, P. G.; Ardo, S.; Kopecky, A.; Galoppini, E.; Meyer, G. J. J. Phys. Chem. Lett. 2010, 1, 1725–1728. (232) Hu, K.; Blair, A. D.; Piechota, E. J.; Schauer, P. A.; Sampaio, R. N.; Parlane, F. G. L.; Meyer, G. J.; Berlinguette, C. P. Nat. Chem. 2016, 8, 853–859. (233) Farnum, B. H.; Wee, K.-R.; Meyer, T. J. Nat. Chem. 2016, 8, 845–852. (234) DiMarco, B. N.; Troian-Gautier, L.; Sampaio, R. N.; Meyer, G. J. Chem. Sci. 2018, 9, 940–949. (235) Troian-Gautier, L.; DiMarco, B. N.; Sampaio, R. N.; Marquard, S. L.; Meyer, G. J. J. Am. Chem. Soc. 2018, 140, 3019–3029. (236) Lindsey, C. P.; Patterson, G. D. J. Chem. Phys. 1980, 73, 3348–3357. (237) Wang, D.; Marquard, S. L.; Troian-Gautier, L.; Sheridan, M. V.; Sherman, B. D.; Wang, Y.; Eberhart, M. S.; Farnum, B. H.; Dares, C. J.; Meyer, T. J. J. Am. Chem. Soc. 2017, jacs.7b10809. (238) Youngblood, W. J.; Lee, S. A.; Maeda, K.; Mallouk, T. E. Acc. Chem. Res. 2009, 42, 1966–1973. (239) Dloczik, L.; Ileperuma, O.; Lauermann, I.; Peter, L. M.; Ponomarev, E. A.; Redmond, G.; Shaw, N. J.; Uhlendorf, I. J. Phys. Chem. B 1997, 101, 10281–10289. (240) Feng, X.; Zhu, K.; Frank, A. J.; Grimes, C. A.; Mallouk, T. E. Angew. Chemie Int. Ed. 2012, 51, 2727–2730. (241) Eberhart, M. S.; Wang, D.; Sampaio, R. N.; Marquard, S. L.; Shan, B.; Brennaman, M. K.; Meyer, G. J.; Dares, C.; Meyer, T. J. J. Am. Chem. Soc. 2017, 139, 16248– 16255. (242) Li, F.; Fan, K.; Wang, L.; Daniel, Q.; Duan, L.; Sun, L. ACS Catal. 2015, 5, 3786– 3790. (243) Materna, K. L.; Crabtree, R. H.; Brudvig, G. W. Chem. Soc. Rev. 2017, 46, 6099– 6110. (244) Hanson, K.; Brennaman, M. K.; Luo, H.; Glasson, C. R. K.; Concepcion, J. J.; Song, W.; Meyer, T. J. ACS Appl. Mater. Interfaces 2012, 4, 1462–1469. (245) Xu, P.; Milstein, T. J.; Mallouk, T. E. ACS Appl. Mater. Interfaces 2016, 8, 11539– 11547.

153

(246) Zhang, J.; Freitag, M.; Hagfeldt, A.; Boschloo, G. In Molecular Devices for Solar Energy Conversion and Storage; Springer, 2018; pp 151–185. (247) Cao, Y.; Saygili, Y.; Ummadisingu, A.; Teuscher, J.; Luo, J.; Pellet, N.; Giordano, F.; Zakeeruddin, S. M.; Moser, J.-E.; Freitag, M.; et al. Nat. Commun. 2017, 8, 15390. (248) Xu, T.; Chen, L.; Guo, Z.; Ma, T. Phys. Chem. Chem. Phys. 2016, 18, 27026–27050. (249) Hu, S.; Lewis, N. S.; Ager, J. W.; Yang, J.; McKone, J. R.; Strandwitz, N. C. J. Phys. Chem. C 2015, 119, 24201–24228. (250) Chen, Y. W.; Prange, J. D.; Dühnen, S.; Park, Y.; Gunji, M.; Chidsey, C. E. D.; McIntyre, P. C. Nat. Mater. 2011, 10, 539–544. (251) Zhao, Y.; Vargas-Barbosa, N. M.; Hernandez-Pagan, E. A.; Mallouk, T. E. Small 2011, 7, 2087–2093. (252) Zhao, Y.; Vargas-Barbosa, N. M.; Strayer, M. E.; McCool, N. S.; Pandelia, M.-E.; Saunders, T. P.; Swierk, J. R.; Callejas, J. F.; Jensen, L.; Mallouk, T. E. J. Am. Chem. Soc. 2015, 137, 8749–8757. (253) Sherman, B. D.; Sheridan, M. V.; Dares, C. J.; Meyer, T. J. Anal. Chem. 2016, 88, 7076–7082. (254) Cappel, U. B.; Daeneke, T.; Bach, U. Nano Lett. 2012, 12, 4925–4931. (255) Bach, U.; Lupo, D.; Comte, P.; Moser, J. E.; Weissörtel, F.; Salbeck, J.; Spreitzer, H.; Grätzel, M. Nature 1998, 395, 583–585. (256) Burschka, J.; Dualeh, A.; Kessler, F.; Baranoff, E.; Cevey-Ha, N.-L.; Yi, C.; Nazeeruddin, M. K.; Grätzel, M. J. Am. Chem. Soc. 2011, 133, 18042–18045. (257) Yagi, M.; Tomita, E.; Sakita, S.; Kuwabara, T.; Nagai, K. J. Phys. Chem. B 2005, 109, 21489–21491. (258) Lin, F.; Boettcher, S. W. Nat. Mater. 2014, 13, 81–86. (259) Lin, F.; Boettcher, S. W. In Photoelectrochemical Solar Fuel Production; Springer International Publishing: Cham, 2016; pp 323–351. (260) Mills, T. J.; Lin, F.; Boettcher, S. W. Phys. Rev. Lett. 2014, 112, 148304. (261) Nazeeruddin, M. K.; Kay, A.; Rodicio, I.; Humphry-Baker, R.; Mueller, E.; Liska, P.; Vlachopoulos, N.; Graetzel, M. J. Am. Chem. Soc. 1993, 115, 6382–6390.

154

(262) Materna, K. L.; Jiang, J.; Regan, K. P.; Schmuttenmaer, C. A.; Crabtree, R. H.; Brudvig, G. W. ChemSusChem 2017, 10, 4526–4534. (263) Eom, Y.-K.; Nhon, L.; Leem, G.; Sherman, B. D.; Wang, D.; Troian-Gautier, L.; Kim, S.; Kim, J.; Meyer, T. J.; Reynolds, J. R.; et al. ACS Energy Lett. 2018, acsenergylett.8b01020. (264) Eberhart, M. S.; Bowers, L. M. R.; Shan, B.; Troian-Gautier, L.; Brennaman, M. K.; Papanikolas, J. M.; Meyer, T. J. J. Am. Chem. Soc. 2018, 140, 9823–9826. (265) Shan, B.; Nayak, A.; Brennaman, M. K.; Liu, M.; Marquard, S. L.; Eberhart, M. S.; Meyer, T. J. J. Am. Chem. Soc. 2018, 140, 6493–6500. (266) Duan, L.; Bozoglian, F.; Mandal, S.; Stewart, B.; Privalov, T.; Llobet, A.; Sun, L. Nat. Chem. 2012, 4, 418–423. (267) Sasaki, T.; Watanabe, M.; Hashizume, H.; Yamada, H.; Nakazawa, H. J. Am. Chem. Soc. 1996, 118, 8329–8335. (268) Fang, M.; Kim, C. H.; Mallouk, T. E. Chem. Mater. 1999, 11, 1519–1525.

155

VITA

Pengtao Xu

Education 2013 – 2018 Ph.D. in Chemistry, Penn State University, University Park, PA, USA 2009 – 2013 B.S. in Chemistry, Nankai University, Tianjin, China

Selected Publications (6) Xu, P.; Gray, C.L.; Xiao, L.; Mallouk, T.E.; Charge Recombination with Fractional Reaction Orders in Water-Splitting Dye-sensitized Photoelectrochemical Cells. J. Am. Chem. Soc. In press. (5) Xu, P.; Huang, T.; Huang, J.; Yan, Y.; Mallouk, T. E. Dye-Sensitized Photoelectrochemical Water Oxidation through a Buried Junction. Proc. Natl. Acad. Sci. 2018, 115, 6946–6951. (4) Xu, P.; McCool, N. S.; Mallouk, T. E. Water Splitting Dye-Sensitized Solar Cells. Nano Today 2017, 14, 42–58. (3) Xu, P.; Milstein, T. J.; Mallouk, T. E. Flat-Band Potentials of Molecularly Thin Metal Oxide Nanosheets. ACS Appl. Mater. Interfaces 2016, 8, 11539–11547. (2) Xu, P.; Tang, Q.; Zhou, Z. Structural and Electronic Properties of Graphene-ZnO Interfaces: Dispersion-Corrected Density Functional Theory Investigations. Nanotechnology 2013, 24, 305401. (1) Xu, P.; Yang, J.; Wang, K.; Zhou, Z.; Shen, P. Porous Graphene: Properties, Preparation, and Potential Applications. Chinese Sci. Bull. 2012, 57, 2948–2955.

Presentations

• “Dynamic response of water-splitting dye-sensitized photoelectrochemical cell”, Lion Lectures, Penn State University, March 2018 (talk)

• “Dynamic Response of Water-splitting Dye-sensitized Photoelectrochemical Cells”, Renewable Energy: Solar Fuels, Gordon Research Conference, Ventura, CA, January 2018 (poster)

• “Atomic Layer-deposited Cobalt Oxide Stabilizes Dye-sensitized Photoanodes for Improved Water Oxidation”, 253rd ACS meeting, San Francisco, CA, April 2017 (talk)