Advanced Metal Oxide Semiconductors for Solar Energy Harvesting and Solar Fuel Production

LICENTIATE T H E SIS

Department of Engineering Sciences and Mathematics Division of Material Science Pedram Ghamgosar Advanced for Solar Energy Harvesting and Solar Fuel Production Metal Oxide Semiconductors Advanced Metal Oxide Semiconductors

ISSN 1402-1757 for Solar Energy Harvesting ISBN 978-91-7583-979-0 (print) ISBN 978-91-7583-980-6 (pdf) and Solar Fuel Production Luleå University of Technology 2017

e- e-

2H 2O 2H2 h+

+ + O2 + 4H 4H

MOSs PEM

Pedram Ghamgosar

Experimental physics

Advanced Metal Oxide Semiconductors for Solar Energy Harvesting and Solar Fuel Production

Pedram Ghamgosar

Luleå University of Technology Department of Engineering Sciences and Mathematics Division of Materials Science

Printed by Luleå University of Technology, Graphic Production 2017

ISSN 1402-1757 ISBN 978-91-7583-979-0 (print) ISBN 978-91-7583-980-(pdf) Luleå 2017 www.ltu.se

ABSTRACT

Increasing energy consumption and its environmental impacts make it necessary to look for alternative energy sources. Solar energy as huge energy source which is able to cover the terms sustainability is considered as a favorable alternative. Solar cells and solar fuels are two kinds of technologies, which make us able to harness solar energy and convert it to electricity and/or store it chemically.

Metal oxide semiconductors (MOSs) have a major role in these devices and optimization of their properties (composition, morphology, dimensions, crystal structure) makes it possible to increase the performance of the devices. The light absorption, charge carriers mobility, the time scale between charge injection, regeneration and recombination processes are some of the properties critical to exploitation of MOSs in solar cells and solar fuel technology.

In this thesis, we explore two different systems. The first system is a NiO mesoporous semiconductor photocathode sensitized with a biomimetic FeFe-catalyst and a coumarin C343 dye, which was tested in a solar fuel device to produce hydrogen. This system is the first solar fuel device based on a biomimetic FeFe-catalyst and it shows a Faradic efficiency of 50% in hydrogen production. Cobalt catalysts have higher Faradic efficiency but their performance due to hydrolysis in low pH condition is limited. The second system is a photoanode based on the nanostructured hematite/magnetite film, which was tested in a photoelectrochemical cell. This hybrid electrode improved the photoactivity of the photoelectrochemical cell for water splitting. The main mechanism for the improvement of the functional properties relies with the role of the magnetite phase, which improves the charge carrier mobility of the composite system, compared to pure hematite, which acts as good light absorber semiconductor.

By optimizing the charge separation and mobility of charge carriers of MOSs, they can be a promising active material in solar cells and solar fuel devices due to their abundance, stability, non-toxicity, and low-cost. The future work will be focused on the use of nanostructured MOSs in all-oxide solar cell devices. We have already obtained some preliminary results on 1-

1 dimensional heterojunctions, which we report in section 3.3. While they are not conclusive, they give an idea about the future direction of the present research.

Acknowledgment

I would like to acknowledge and appreciate all the people who helped me and supported me and without them it was not easy to perform this work. The special thanks go to my supervisor, Alberto Vomiero, for giving the opportunity to work on this interesting project. I appreciate his scientific guidelines, all helpful discussions, critical ques- tions and also for being really supportive and motivational. I acknowledge Nils Almqvist, my co-supervisor, for his helpful discussions, encouragement and sharing his knowledge. Thanks to Johnny Grahn and Lars Frisk, the research engineers at LTU, for always being around and helping with instrumental work even during the summer period and also sharing their experi- ences. I am very thankful to all co-authors of appended papers, Leif Hammarström, Sascha Ott, Haining Tian, Somnath Maji, Liisa J. Antila, Sanjay Mathur, Yakup Gönüllüa, Jennifer Leduc, Shujie You and Johanne Mouzon. Also thanks to Graziella Malandrino and Anna Lucia Pellegrino, our collabo- rators in the MOSs for all-oxide solar cells project that we obtained some preliminary results and we hope to publish it soon. I am grateful to my friends and colleagues for their encouragement, support and motivations. You made it easier to me to continue my way and have a nice time in this period and enjoy my life. Last but not least, I would like to tell big and special thanks to my family for supporting me in all my life and made my journey in life so fun and amazing.

Pedram Ghamgosar September 2017, Luleå

Appended Papers

This thesis is based on the following contributed publications:

1. Liisa J. Antila, Pedram Ghamgosar, Somnath Maji, Haining Tian, Sascha Ott, and Leif Hammarström Dynamics and Photochemical H2 Evolution of Dye-NiO Photocathodes with a Biomi- metic FeFe-Catalyst ACS Energy Lett. 2016, 1, 1106-1111

2. Jennifer Leduc, Yakup Gönüllüa, Pedram Ghamgosar, Shujie You, Johanne Mouzon, Al- berto Vomiero, Sanjay Mathur The Role of Heterojunctions on the Water Oxidation Performance in Mixed α- Fe2O3/Fe3O4 Hybrid Film Structures To be submitted to: ACS Appl. Mater. Interface

The following paper is not included in this thesis:

1. Wenxing Yang, Yan Hao, Pedram Ghamgosar, Gerrit Boschloo Thermal Stability Study of Dye-Sensitized Solar Cells with Cobalt Bipyridyl–based Electrolytes Electrochimica Acta 2016, 213, 879–886

Conference contributions

1. H. Tian, Y. Chen, D. Shevchenko, P. Ghamgosar, S. Maji, S. Ott, and L. Hammarström Unpredictable Chemical Bath Deposition of Catalyst on NiO Photocathode for Proton Reduction, Poster presentation at Solar Fuels: Moving from Materials to Devices International Discus- sion Meeting 2015, 7th & 8th July at the Royal Society, London

2. P. Ghamgosar, G. S. Selopal, R. Milan, I. Dobryden, A. L. Pellegrino, I. Concina, N. Almqvist, G. Malandrino, A. Vomiero Composite nanowires (NWs) for all-oxide solar cells Poster presentation at 5th Internatinal Workshop on Advances Materials Challenges for Health and Alternative Energy Solution 2016, 31st August -3rd September, Cologne, Germa- ny

3. F. Enrichi, C. Armellini, S. Belmokhtar, A. Bouajaj, E. Cattaruzza, E. Colusso, M. Ferrari, P. Ghamgosar, F. Gonella, A. Martucci, R. Ottini, P. Riello, G. C. Righini, E. Trave, A. Vomiero, S. You, L. Zur Downconversion enhancement by Ag nanoaggregates in Tb3+/Yb3+ doped sol-gel glass and glass-ceramic films for solar cell applications Oral presentation at Shift 2017, 13th -17th November, Tenerife

4. F. Rigoni, P. Ghamgosar, I. Dobryden, A. L. Pellegrino, R. Borgani, I. Concina, G. Malandrino, P. M. Claesson, N. Almqvist, A. Vomiero Probing local electrical properties of ZnO NW-based all-oxide solar cells with advanced AFM methods Oral presentation at e.MRS Fall Meeting 2017, 18th -21st September, Warsaw

Contents

1. INTRODUCTION ...... 11 1.1. Energy ...... 11 1.2. Solar radiation ...... 12 1.3. Photovoltaic effect ...... 16 1.4. Photoelectrochemical cell ...... 17 1.5. Dye-sensitized solar cell ...... 21 1.6. Objectives of the thesis ...... 22 2. EXPERIMENTAL ...... 23 2.1. Electrodes preparation ...... 23 2.1.1. Tape casting ...... 23 2.1.2. Chemical vapor deposition ...... 23 2.2. Material characterization ...... 25 2.2.1. Fourier transform infrared spectroscopy ...... 25 2.2.2. Ultraviolet-visible spectroscopy ...... 26 2.2.3. Optical microscopy ...... 27 2.2.4. Scanning electron microscopy ...... 27 2.2.5. Energy-dispersive x-ray spectroscopy ...... 28 2.2.6. X-ray diffraction ...... 30 2.2.7. X-ray photoelectron spectroscopy ...... 31 2.2.8. Raman spectroscopy ...... 33 2.3. Device characterization ...... 34 2.3.1. (Photo) electrochemical measurement ...... 34 3. RESULTS and DISCUSSION ...... 36

3.1. Dynamics and photochemical H2 evolution of dye-NiO photocathodes with a biomimetic FeFe-catalyst ...... 36

3.2. The role of heterojunctions on the water oxidation performance in mixed α-Fe2O3/Fe3O4 hybrid film structures ...... 40 3.3. 1-Dimensional heterostructures for all-oxide solar cells ...... 44 4. CONCLUSION and FUTURE OUTLOOK ...... 46 5. REFERENCES ...... 47

Abbreviations and Symbols

λ Wavelength ߭ҧ Wavenumber A Absorbance AES Auger electron spectroscopy Ag Silver AgCl Silver chloride AM Air Mass APS Artificial photosyntheis BiVO4 Bismuth vanadate CRT Cathode ray tube CB Conduction Band CdS Cadmium sulfide CE Counter electrode CEI Commission Electrotechnique Internationale CO2 Carbon dioxide Cu2O Cuprous oxide CV Cyclic voltammetry CVD Chemical vapor deposition DSSC Dye-sensitized solar cell DSSFD Dye-sensitized solar fuel device EBSD Electron backscattering diffraction EDS Energy-dispersive x-ray spectroscopy EIA U.S. Energy Information Administration e- Electron eV Electron Volts Fe Iron Fe2O3 Hematite Fe3O4 Magnetite FT-IR Fourier transform infrared FTO Fluorine doped tin oxide GC-MS Gas chromatography-Mass spectrometry H+ Proton H2 Hydrogen H2O Water HER Hydrogen evolution reaction HOMO Highest occupied molecular orbital I- Iodide ି ଷ triiodide ICDD International Center of Diffraction Data IEA International Energy Agency IEC International Electrotechnical Commission IR Infrared ITO Indium tin oxide 9

KPFM Kelvine probe force microscopy LUMO Lowest unoccupied molecular orbital Mtoe Million tonnes of oil equivalent NaOH Sodium hydroxide NHE Normal Hydrogen Electrode NiO Nickel oxide NPS Natural photosynthesis O2 Oxygen O3 Ozone OER Oxygen evolution reaction PEC Photoelectrochemical PEM Proton Exchange Membrane Pt Platinum PV Photovoltaics RE Reference electrode REN21 Renewable Energy Policy Network for the 21st Century SAM Scanning auger microscopy SEM Scanning electron microscopy SFD Solar Fuel Device SiC Silicon carbide STEM Scanning transmission electron microscopy SSRM Scanning spreading resistance microscopy T Transmittance TA Transient Absorption TCO Transparent conductive oxide TEM Transmission electron microscopy TiO2 Titanium dioxide UV Ultraviolet VB Valence Band Vis Visible WDS Wavelength-dispersive x-ray spectroscopy WE Working electrode XRD X-ray diffraction XRF X-ray fluorescence ZnO Zinc oxide

1. INTRODUCTION

1.1. Energy

The human’s modern life is thankful to learn how to harness the energy. Energy as an ability to perform work can be classified into two main groups based on its sources: renewable and nonrenewable. Energy carriers (also called secondary energy sources): electricity and hydrogen are produced from conversion of renewable and nonrenewable energy sources which are called primary energy sources. 1

Nonrenewable energy sources, like oil, natural gas, and coal form our main energy sources. These energy sources which also can be called fossil fuels have some major problems. The first problem is addressed to the replenishment of them. They are formed by natural processes from natural resources over millions of years. The second issue is back to the pollution. The green- house gases which are a reason for global warming are released from combustion of these carbon based energy sources. 2

Sustainable energy sources can be a key to overcome these problems. The sustainable energy is a source of energy that generates less destructive impact on nature like air pollution and global warming. And in the same time it has enough energy supplies. 3, 4 Renewable energy sources like solar energy, wind energy, biomass, hydropower, wave power, geothermal, and tidal energy have the properties to be sustainable energy sources. Above mentioned energy sources, except geothermal and tidal energy which are non-solar renewable energy sources, are called solar renewable sources. 4 Electricity, heat, and fuel for transportation can be generated from renewable energy sources. 1

1.2. Solar radiation

In the bulk of the sun, hydrogen atoms convert to helium by fusion reactions. This nuclear energy in the core of the sun is almost equal to 3.89 × 1026 joules per second. It transfers from core to the surface of the sun by rapid transformation of nuclear energy to the thermal energy. The thermal energy leaves the surface of the star as an electromagnetic radiation called solar radiation.5

The earth/sun distance and solar activity is not the same all over the year. The changes in these conditions affect the irradiance of the sun, which reaches to the atmosphere of the earth and that can vary from 1% to 6.6% in irradiance within a year. The solar spectrum at the top of the atmosphere is called extraterrestrial spectrum (Figure 1.1). This spectrum is the same all over from the surface of the sun to the atmosphere of the earth, since there is no absorption and scattering effects in this region.6

Figure 1.1. Solar spectrum out of earth's atmosphere, extraterrestrial (dotted line), and solar spectrum at the sea level, terrestrial (solid line) 6

The travel of the solar light from earth’s atmosphere to the surface of the earth causes changes in the extraterrestrial spectrum due to absorption and scattering (Figure 1.2).

The ozone (O3) layer, carbon dioxide (CO2), water vapor (H2O), and oxygen (O2) are responsible for the absorption of the solar radiation. The ozone layer absorbs the light at the shorter wavelength in the ultraviolet spectral region, while the other substances above absorb light in the near infrared region. Scattering from particles such as water drops and aerosols, and Ray- leigh scattering also cause some variations in solar spectrum in addition to absorption. The scattering and absorption in the atmosphere cause some changes in the extraterrestrial spectrum and form another spectrum that is called global radiation spectrum. 4, 6

Figure 1.2. Contribution of the various components in global radiation (adapted from ref 6)

One definition that is important for solar spectrum characterization is air mass (AM). The air mass can be defined in two ways: the first according to path length of the solar radiation through the earth atmosphere and the second according to the sun angle. The relative path length of the solar light when the sun is perpendicular to the earth and when the sun is not perpendicular to the earth is one of the AM definitions. The second definition is coming from the zenith angle when it passes through the atmosphere relative to the overhead air mass (Figure 1.3). 4, 6

Figure 1.3. The air mass according to the zenith angle variation (adapted from ref 6) The air mass coefficient is also important for characterization of solar cell performance. The AM is followed by a number which will be calculated by following equations according to the model when the atmosphere is assumed to be plain (equation 1) and when the curvature of atmosphere is considered (equation 2). 7

௅ ଵ ܣܯ ൌ ൎ , (1) ௅బ ୡ୭ୱ ௭

L: The path length through the atmosphere

L0: The zenith path length Z: zenith angle in degrees

ଵ ܣܯ ൌ . (2) ୡ୭ୱ ௭ା଴Ǥହ଴ହ଻ଶሺଽ଺Ǥ଴଻ଽଽହି௭ሻషభǤలయలర

Some of the different existing standards for air mass are presented in table 1.

The irradiance of the sun has the SI-unit of watt per square meter (W/m2). 1000 W/m2 defines another unit for irradiance which is called 1 sun. 8

Table 1. Existing standards for air mass and their power densities 6

Solar condition Standard Power Density (Wm-2)

Total 250 – 2500 nm 250 – 1100 nm WMO Spectrum 1367 AM 0 ASTM E 490 1353 1302.6 1006.9

AM 1 CIE Publication 85, Table 2 969.7 779.4 AM 1.5 D ASTM E 891 768.3 756.5 584.7

AM 1.5 G ASTM E 892 963.8 951.5 768.6

AM 1.5 G CEI/IEC* 904-3 1000 987.2 797.5 * Integration by modified trapezoidal technique CEI: Commission Electrotechnique Internationale IEC: International Electrotechnical Commission

As mentioned previously, some parts of the solar energy, which is released from the surface of the sun, after absorption and reflection hits the surface of the earth with a total power equal to 100,000 TW. 5 According to the International Energy Agency’s (IEA) report, the annual global energy consumption in 2014 was equal to 9,425 Mtoe, which can be translated to 12.5 TW. 9 This means that solar energy has the ability to cover amply all human’s need for energy.

Figure 1.4 shows the contribution of the various energy sources in annual global energy consumption in 2015, which was reported by Renewable Energy Policy Network for the 21st Centu- ry (REN21). 10 The total renewable energy contribution is 19.3% and solar energy has less than 5% contribution. This means that the existing technologies for solar energy harvesting, which had a great growth during the recent years, still is not enough effective. The photovoltaics gained a major role for electricity production in the market but still it cannot compete with cheap fossil fuel. For this reason, the low-cost and high efficient photovoltaic devices are the golden key for the future energy contribution. 5

Figure 1.4. Contribution of Renewable Energies in annual energy consumption in 2015 10

1.3. Photovoltaic effect

The solar energy, which passes through the earth’s atmosphere can be indirectly used by transformation to wind energy, wave energy or bioenergy or can be directly used by solar thermal or photovoltaics. The photovoltaics (PV) is a technique, which converts the solar energy directly into the electricity. The term photovoltaics consists of two words: photos and volt. “Photos” means light in Greek and “volt” derives from an Italian physicist’s name, Count Alessandro Volta, and it is the electromotive force’s unit. The photovoltaic cells are based on the voltage difference between two different types of semiconductor. A pure semiconductor (intrinsic) is an undoped semiconductor. In this structure, the electrons in conduction band (CB) are a result of thermal excitation and they leave a hole in valence band (VB). By doping semiconductors with some other elements, p-type and n-type semiconductors can be formed. For instance, taking into account silicon, doping it with boron will result in an excess of holes (B has one electron less than silicon and it removes an electron from the VB of the silicon). Because of that, B-doped silicon is called a p-type semiconductor. Since phosphorus (P) has an excess electron compared to silicon, it adds an electron to the CB of the silicon. P-doped silicon is called an n-type semiconductor. The interface between p- and n-type semiconductors in contact is called a p-n junction. An illus- tration of a p-n junction is shown in top-left panel of Figure 1.5. Holes are considered as positive charges and diffuse from the p-type semiconductor to the n-type semiconductor. The electrons diffuse oppositely from n-type to p-type. Hence, a depletion layer with only few mobile holes and electrons is formed in the near-junction region and an electric field appears in the interface (Figure 1.5, top right). Electron-hole pairs will be generated due to photon absorption while solar light shines on the depletion region. Then, electrons and holes migrate to the n-type and p-type side, respectively (Figure 1.5, down left). A potential difference occurs in the depletion region and if these two type semiconductors connect through an external circuit, current will be generated (Figure 1.5, down right). 3, 4

Figure 1.5. Photovoltaics’ principles of a p-n junction. There are three generations of photovoltaic devices. The first generation, with high efficiency and high production cost, consists of mono- and poly-crystalline silicon. The second generation which is also called “thin film photovoltaics” uses cheaper materials with lower photoconver- sion efficiency. In third generation photovoltaics, good properties from the first two generations are kept to increase the efficiency but keeping material and production cost low.4, 11

1.4. Photoelectrochemical cell

In the previous part it was shown how to convert solar energy to electricity through the photovoltaic effect. Another attractive way of using solar energy is to convert it to storable energy carriers like carbohydrates and hydrogen, which are called solar fuels.

In nature, during natural photosynthesis (NPS), carbon dioxide and water is converted to carbo- hydrate and molecular oxygen by using solar energy (equation 3) and this energy stores in the plants, algae and cyanobacteria in the form of chemical bonds.1217

(ଵଶܱ଺ ൅͸ܱଶ. (3ܪ଺ܥଶܱ൅݄ݒ՜ܪଶ ൅͸ܱܥ͸

Artificial photosynthesis (APS) is another technique, in which solar energy is stored in the form of chemical bonds in hydrogen. In APS, by photocatalytic or photoelectrochemical (PEC) water 17 splitting, which is shown in equation 4, it is possible to convert the solar energy to hydrogen. 12, 18

ʹܪଶܱ൅݄ߴ՜ʹܪଶ ൅ܱଶ. (4)

To obtain this conversion, the basic requirements which regulate the NPS should be considered. They consist of a material to absorb solar light, generate an electron-hole pair, separate these two charges, and at the end to have enough energy to operate the catalytic water splitting.12 The first report of demonstration of photoelectrochemical water splitting was by Fujishima and

Honda in 1972. They used a wide band gap transparent semiconductor (anatase TiO2) photoanode coupled with a platinum (Pt) cathode in a pH gradient. 19 The schematic diagram of this photoelectrochemical cell is shown in Figure 1.6. This report was the starting point for this research area and after that the development on photoelectrochemical water splitting was very quick and a massive progress came in hand for mater ials and device design. 2030 Pt

Figure 1.6. Schematic diagram of a photoelectrochemical (PEC) cell uisng TiO2 as a photoanode, Pt as a cathode and a proton exchange membrane (PEM) (adapted from ref 12)

There are several properties that a semiconductor material should cover to be a suitable photocatalyst for water splitting, in which the solar light is the driving force. The band gap and band position of these materials are two main factors.

As a favored semiconductor material, not only the minimum band gap should be at least larger than 1.229 eV to be able to do water splitting according to equation 4, but also the CB and VB 18 should be in a suitable position. The VB position should be more positive compare to the redox potential of water oxidation half reaction (equation 5).

ଵ ܪ ܱ ՜ ʹܪା ൅ ܱ ൅ʹ݁ି ܧ଴ᇱ ൌ ൅ͲǤͺͳͷܸ. (5) ଶ ଶ ଶ ௢௫௜ௗ௔௧௜௢௡

The CB position also should be more negative compared to the redox potential of proton reduction reaction (equation 6).

ା ି ଴ᇱ ʹܪ ൅ʹ݁ ՜ܪଶ ܧ௥௘ௗ௨௖௧௜௢௡ ൌ െͲǤͶͳͶܸ . (6)

Both potentials are reported versus Normal Hydrogen Electrode (NHE) at 25 oC and pH = 7.

Another property to consider is the light absorption of these semiconductor materials. Devices based on semiconductors with photocatalytic activity under ultraviolet (UV) light are more efficient per photon energy absorption for hydrogen production via water splitting, than the visible light-based devices. Still the development of better materials for visible light is needed. The reason is that the UV light contributes just approximately 4% to the total solar spectrum, while the contribution for visible light is around 53%. 3135 Figure 1.7 shows the band gap and band position of some semiconductor materials.

The band gap of all semiconductors in Figure 1.7 is more than 1.23 eV and therefore fulfills one of the requirements for water splitting. Some of them, such as titanium dioxide (TiO2) and bismuth vanadate (BiVO4) are considered as a suitable photocatalyst candidate for water splitting. 31 Silicon carbide (SiC), zinc oxide (ZnO), and cadmium sulfide (CdS) have a suitable band gap, but are not good for water splitting due to their photocorrosion properties. 31 Cuprous oxide

(Cu2O) and hematite (Fe2O3) have favorable band gap for visible light absorption but their band position and their low electronic conductivity limit the possibility to be highly efficient photo- catalysts.

19 Figure 1.7. Diagram of band gap and band position of some semiconductor materials, oxygen evolution reaction (OER) and hydrogen evolution reaction (HER) redox potential at pH = 7. 36

Finding a semiconductor or developing the properties of an existing semiconductor to fulfill the narrow band gap for light absorption and at the same time having the desired energy level for photocatalytic reaction is a challenge in the photocatalytic water splitting research area.

To overcome the limitations described above, several strategies have been tested by different groups. One of the successful techniques was to merge a dye-sensitized solar cell (DSSC) with a suitable photocatalyst and make a dye-sensitized solar fuel device (DSSFD) for water splitting. In this way by using a molecular photosensitizer, it is possible to extend the light absorption in the range of visible light spectrum. 12 In a dye-sensitized solar cell, a molecular photosensitizer 37, 38, 39 able to absorb visible light is anchored on a large band gap semiconductor like TiO2. The energy levels of these molecular photosensitizers, highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO), are easier to tune than the VB and CB of semiconductors.

Doping and creating defects could be another strategy to increase light harvesting of semiconductors. 31, 40 In order to overcome the low conductivity of metal oxide semiconductors of late transition metals like nickel oxide (NiO) and Fe2O3, which have good catalytic activity, it is possible to synthesize multicomponent metal oxides. 31, 41, 42

20 1.5. Dye-sensitized solar cell

In 1991, a different type of photoelectrochemical cell was introduced by Brian O’Regan and Michael Grätzel, which is called dye-sensitized solar cell. 37 Despite of that, the DSSC is also working by photovoltaic effects, it has some structural and functional differences with conventional photovoltaic devices. In DSSC, a molecular photosensitizer (dye molecule) is anchored on a wide band gap semiconductor and it has the responsibility of light absorption. The charge separation happens at the interface of dye/semiconductor. The semiconductor transfers the separated charge to the external circuit. 4, 12, 43 In conventional photovoltaic devices the semiconductor has both responsibility of light harvesting and charge transfer. Also the charge separation is based on the electric field in the p-n junction interface (section 1.3). Figure 1.8 demonstrates a schematic diagram of the main components of a DSSC and the opera- tional mechanism of the cell. It is made of three main components: a working electrode (WE), a redox electrolyte and a counter electrode (CE).

Figure 1.8. Schematic diagram of DSSC components and its operation principles. The red particles are dye molecules sensitized on the TiO2 mesoporous particles (blue particles) (adapted from ref 12)

The working electrode is composed of a wide band gap mesoporous semiconductor deposited on a glass or plastic substrate which is covered by a transparent conductive oxide (TCO) such as indium tin oxide (ITO) or fluorine doped tin oxide (FTO). Dye molecules are anchored on the mesoporous film. The working electrodes, based on the doping of semiconductors, are divided 37, 4447 48, 49, 50 to n-type DSSC (like TiO2, ZnO, …) or p-type DSSC (like NiO, Cu2O, …). The - - electrolyte consists of two components: redox mediator and solvent. Iodide/triiodide (I /I3 ) was the first redox couple used by O’Regan and Grätzel. 37 Later, various types of redox mediaWRUV 21 were tested, 39 and the cobalt redox mediator was introduced as a promising redox coupleby Feldt and coworkers in 2010. 51 As the solvent, also different materials were used in DSSCs such as water, 52, 53 acetonitrile, 54, 55 ionic liquids, 56, 57 etc. The conventional counter electrode is platinized FTO or ITO but also other electrodes using different catalysts like conductive polymers, carbon materials, and graphene have been used in DSSCs. 39, 58, 59 The operating principles of an n-type DSSC start by photon absorption and photoexcitation of dye molecules (process 1 in Figure 1.8). The excited dye injects the electron to the conduction band of the semiconductor, here TiO2 (process 2). The excited dye will be regenerated by reduced species of redox couple (process 3). Electron in the conduction band migrates to the conductive substrate by diffusion (process 4). The collected electron at the conductive substrate performs work via transferring to the counter electrode through the external circuit (process 5). The cycle will be completed by reduction of oxidized species of redox couple at the counter electrode (process 6).

1.6. Objectives of the thesis

In this work, the aim is to improve the photochemical hydrogen production and water oxidation using solar fuel devices (SFDs). In a DSSFD, a biomimetic FeFe-catalyst attached to a photocathode is tested to produce hydrogen. This is the first time that a DSSFD based on a biomimetic Fe-Fe catalyst is examined. Iron is an abundant material and this catalyst is more resistant in the acidic environment compare to the cobalt catalysts that are mostly used in photolysis. Also a hematite/magnetite, α-Fe2O3/Fe3O4, photoanode was used as hybrid electrode to improve the water oxidation performance. Hematite has a suitable band gap for water splitting and good light absorption but it is poor in electrical conductivity. This hybrid structure could improve the electrical conductivity using magnetite which is a better conductor compare to hematite.

2. EXPERIMENTAL

2.1. Electrodes preparation

2.1.1. Tape casting

Doctor blade (also called tape casting) is a simple and undemanding technique which is used for preparing thin films in scalable and inexpensive way on large surface areas. This method which was developed in 1940’s, in the beginning gave the opportunity of preparing piezoelectric materials and ceramic capacitors with better dielectric characteristics. 60 Later on, the doctor blade method was developed for some other none-ceramic materials like: polymers and metal applications. 61 The speed of films production by this method can be scale up to several meters per minutes by a thickness variety between 20 and several hundred micrometers. 60 By recent pro- gresses in the method, it can be used for new applications like polymer batteries and photovoltaics, since it is possible to prepare samples by thickness of 5 microns. 62, 63

Figure 2.1. Doctor blade steps for thin film deposition

2.1.2. Chemical vapor deposition

The use of chemical vapor deposition (CVD) which is a technique for preparing powders, films or coatings by chemical reactions of gaseous reactant close to a preheated substrate backs to the prehistoric time. Burning of firewood which causes to form soot because of incomplete oxidation is an example of CVD. 64 Deposition of tungsten (W) on carbon filaments using CVD was a 23 beginning of the industrial utilization for this technique in 1893. 64, 65 Despite this long history for CVD, the impressive progress in the field of electronics around 40 years ago made it a popu- lar technology and brought the usage of CVD to other fields like ceramics, composites and solar cells. 64 The principles of the CVD consist of following steps: The substrate will be placed in a vacuum chamber which will be activated later by heat, light, or plasma. The precursor will be vaporized either by reducing the pressure or increasing the temperature. The vaporized precursor will be transferred to the vacuum chamber. In the chamber the chemical reactions or dissociation of precursors take place in gaseous phase. The reacted precursor will be deposited on a preheated substrate. The thickness of the coating will be controlled by duration of the deposition and the temperature. The schematic diagram of CVD is shown in Figure 2.2.

Figure 2.2. Diagram of Chemical Vapor Deposition (CVD) system (adapted from ref 64)

CVD like every other technique has both advantages and drawbacks. Some advantages are: uni- formity and reproducibility of deposited film with high deposition rate, ability to prepare good conformal coating on complex structures, pure and compact coating, easy control of thin or thick deposition by adjusting the deposition rate, controlling the surface morphology and crystal structure by adjusting the coating parameters, wide range of products due to large range of suitable precursors, relatively low production cost, and production of refractory materials at appro- priate low temperature compare to the conventional deposition methods. 64

In contrast to these favorable properties, the disadvantages are: Safety problem due to flamma- ble, toxic, and corrosive precursors, relatively high production cost in terms of need of ultrahigh vacuum and low pressure for deposition of some special products, and difficulty in stoichiome- try control for deposition of multicomponent products. 64

2.2. Material characterization

2.2.1. Fourier transform infrared spectroscopy

Infrared (IR) spectroscopy is a technique to determine the structure of molecules using infrared light. The infrared light represents the wavenumber (߭ҧ) between 12800-10 cm-1 in solar spectrum but the main region which is used for IR spectroscopy is between 4000-400 cm-1. This technique works based on the changes in the vibrational energy in the molecule. The sample absorbs the IR radiation, the dipole moment of the molecular bond is changed and its vibrational energy level changes from ground state to the excited state. The vibrational energy gap determines the absorption frequency. Since each molecule has a unique structure of connected atoms to each other, the IR spectrum is unique for each molecule. This makes possible qualitative and quantitative analysis of various types of samples. The unique peak frequency determines the functional groups in the molecule and the intensity of the peak shows the amount of that group in the sample. The ability to analyze solid, liquid, and gas samples is another advantage of IR spectroscopy. 66 The Fourier Transform Infrared (FT-IR) is the third generation of IR spectrometers. In the first generation, a prism made of NaCl was used as an optical splitter in the system. The water content samples, the size of particles, the scan range, and the reproducibility of the data were the limitation of this generation. In the second generation, the prism was replaced by a grating as a monochromator. The accuracy, the scan speed, and the sensitivity were still the limitations of the second generation. 66 In the FT-IR, a Michelson interferometer is used instead of the monochromator. The interferometer has a beam splitter with ability to transmit 50% of the incoming light and reflect the remainder to a stationary and movable mirror, respectively. The beams are reflected back to the splitter and since they travel different paths due to moving of the movable mirror, they will interfere with each other and make an interferogram. The signal passes through the sample and reaches to the detector. The spectrum represents the energy of all wavelengths (λ) versus time. To be able to use these data, the Fourier transform algorithm is used to translate this spectrum in a way that correlates the intensity of the signals to the wavelength. 67

Figure 2.3. Schematic diagram of a Michelson interferometer (adapted from ref 66)

2.2.2. Ultraviolet-visible spectroscopy

In contrast to IR radiation which causes a vibrational transition because of low energy incoming radiation, the ultraviolet (UV) and visible (Vis) radiations have higher energy and they cause electronic transition. The UV and Vis light in the solar spectrum represent the wavelength 200- 400 nm and 400-700 nm, respectively. These short wavelength radiations have enough energy to exit electrons from HOMO to LUMO. 68 The working principle of UV-Vis spectroscopy is based on the variation in light intensity. The light intensity from an UV-Vis source is measured before and after passing the sample and the ratio between them is called transmittance (equation 7):

ூ ܶൌ , (7) ூబ where T is the transmittance, I is the intensity after sample and I0 is the intensity before sample.

The absorbance (A) will be defined based on the transmittance using equation 8:

(8) ܣൌെଵ଴ ܶ, where A is unit less value. 26

A can be correlated to the concentration (C) of the sample and path length (L) in which light passes through the sample using Beer-Lambert low (equation 9) and it will be used mostly for quantitative analysis:

ܣൌߝȉܮȉܥ, (9) where ε is called molar extinction coefficient and is a constant and characteristic value for each specific compound. The unit for ε is M-1cm-1. UV-Vis spectroscopy can be used as a qualitative method using the maximum absorption wavelength (λmax). Also it can be used to determine the thickness and band gap of semiconductors of which the latter helps to gain information about electrical properties of semiconductors. 69

2.2.3. Optical microscopy

In optical microscopy, visible light is used to magnify small samples with the help of different lenses. Optical microscopes have been used more than one and half century in the research and they help to find information about microstructural properties and analyze the surface of the samples.

2.2.4. Scanning electron microscopy

Scanning electron microscopy (SEM) is an alternative microscopy technique which gives the opportunity to observe and analyze the microstructural characteristics of solid samples more in detail compared to optical microscopy. In SEM, a finely focused electron beam is used to obtain images instead of visible light, which is used in optical microscopy. Heterogeneous materials could be analyzed at high resolution, down to nanometers (nm). It is also possible to capture images in three-dimensional-like and topographical forms. 70 From the interaction between electron beam and specimen, various kinds of signals arise which can be used in different analysis techniques (Figure 2.4). The high resolution instrumental analysis, in the range of 1-5 nm, when bulk samples are tested and large depth of field are two major reasons that make SEM a versatile technique. In addition, the structural and elemental analyses are other advantages of SEM. The structural analysis gives the ability to determine the crystal structure and grain orientation of crystals. This capability is

27 called electron backscattering diffraction (EBSD). The X-radiation signal gives the opportunity to do quantitative and qualitative elemental analysis and makes it possible to obtain information of surface chemical composition. Wavelength-dispersive spectroscopy (WDS) and energy- dispersive spectroscopy (EDS) are two techniques that use x-ray characteristic signal to analyze the samples. 70

Figure 2.4. Diagram of signals from electron beam interaction with specimen and possible analysis techniques Besides the control console which controls the electron beam and do signal processing using a cathode ray tube (CRT) screen, knobs and computer, the electron column is the main important part in a SEM instrument. The electron column consists of an electron gun which generates and accelerates electrons and electron lenses which demagnify the electron beam and cause a smaller focused electron beam to impinge on the sample. The schematic diagram of an electron column is shown in Figure 2.5 and more detailed information of SEM can be found in a textbook. 70

2.2.5. Energy-dispersive x-ray spectroscopy

Two kinds of x-ray signals will arise from the electron beam and specimen interaction in a SEM instrument. The continuum x-ray is a result of changes in the energy of the incident electron beam by deceleration in the Coulomb filed of specimen atoms. Since the interactions are ran- dom, the created radiation can gain any energy between zero and incident beam energy and is called bremsstrahlung or braking radiation. Even though the continuum radiation carries information about the average atomic number and overall composition, it is usually considered as a nuisance since it forms a background under the characteristic peak. 70

Figure 2.5. Schematic diagram of an electron column in a scanning electron microscope instrument (adapted from ref 70)

The incident electron beam can also excite the atom by kicking off an inner shell electron. The excited atom will be back to its ground state energy level by transition of an outer shell electron to an inner shell electron within 1 ps. Since the energy difference between the shells is specific for each element, the created photon from this transition is a characteristic x-ray beam. 70 In EDS analysis, these characteristic photon energies are used for qualitative analysis to determine the existing elements in an unknown specimen. Quantitative analysis with EDS is also possible at an analytical precision and accuracy of 1-2%. The small requirement of specimen volume and the nondestructive analysis method are some of the advantages with quantitative EDS analysis. It should be mentioned that the collected x-ray signals for quantitative analysis need to be corrected for the ZAF factor, where Z represents the atomic number, A represents the x-ray absorption, and F represents the x-ray fluorescence. The formula which can be used for this correction is shown in equation 10:

(10) ܥ௜Τܥሺ௜ሻ ൌሾܼܣܨሿ௜ ȉܫ௜Τܫሺ௜ሻ,

where Ci and C(i) are the weight fraction of the element of interest in the sample and in the standard respectively and Ii and I(i) are the measured intensity of the element of interest of the sample and standard respectively. More information can be found in textbook. 70

2.2.6. X-ray diffraction

X-ray diffraction (XRD) is a non-destructive and versatile method which helps to determine structure, phase composition and texture of different kind of the samples. 71 In XRD, the obtained x-ray diffraction pattern from the target sample will be compared by the patterns of a reference database to identify the existing phases in the sample. The reference database can be prepared by experimental diffraction measurement of pure phases/or patterns, using the published results in the scientific literature or using the existing databases which are prepared by centers which professionally work with XRD like International Center of Diffraction Data (ICDD). 71 Qualitative and quantitative phase analysis, analysis of the influence of applied pressure, temperature and humidity on materials, analysis of material properties like crystallite size, preferred orientation and residual stress in polycrystalline materials are some of the most important topics in XRD techniques. 71 X-ray radiation is an electromagnetic wave which is located between UV light and γ-rays. The wavelengths between 0.01 nm and 10 nm represent the x-ray radiation among the electromagnetic radiations. The x-ray radiation can be generated by deflection of high-energy electrons or by bombarding an anode with a focused electron beam. The interaction of generated x-ray and target sample goes through different processes like attenuation, absorption, pair-building and scattering. In the scattering process two main interactions are considered. Elastic scattering which is also called coherent or Rayleigh scattering is a phenomenon which the incoming x-ray scatters without changing its energy. The second process is incoherent scattering and it consists of fluorescence and Compton scattering. The fluorescence is used mostly in x-ray fluorescence (XRF) and the Compton scattering because of low energy of scattered photon is ignored. 71 The coherent scattered x-rays at well-defined angles have constructive interference and this effect is used in XRD analysis and it is described by Bragg’s law:

ʹ݀ݏ݅݊ߠ ൌ ݊ߣ, (11)

30 where d is the interplanar spacing between different lattice planes, θ is the Bragg angle, n is order of interfaces and λ is the wavelength. The derivation of Bragg’s law is shown in Figure 2.6. More explanation can be found in some reference bookes. 72, 73

Figure 2.6. Diagram of Bragg's law derivation The diffracted x-ray will reach to the detector and the output will be shown as diffractogram which is a diagram of variation of the intensity of the diffracted radiation vs. the angle of diffraction.

2.2.7. X-ray photoelectron spectroscopy

In photoelectron spectroscopy, absorption of a photon with energy of hυ and the ejection of an electron, mainly an electron from inner shell (K shell), are two important processes which are considered. The inner shell electron of the sample has no kinetic energy due to the binding energy in atomic structure. This electron, after the collision of incoming photon, gains energy from incoming photon and leaves the atom. The kinetic energy of this electron is used for elemental identification. 74

Magnesium (Mg) and aluminum (Al) are two conventional sources for producing characteristic x-ray. One advantage of these sources is that the production of the disturbing continuum x-rays is less than the other sources. For these two sources around 50 % of the generated x-rays are 74 from Kα transition (Figure 2.7).

Figure 2.7. a) The components of K alpha for Al b) the intensity and energy (position) contribution of these components 74 In photoelectron spectroscopy, the kinetic energy variation of an inner shell electron due to bombardment by photons with energy equal h߭ is analyzed. Equation 12 shows the relation between theses energies:

௜ ௙ (12) ݄߭ ൅ܧ௧௢௧ ൌܧ௞௜௡ ൅ܧ௧௢௧ሺ݇ሻ,

i f where Ekin is the kinetic energy of photoelectron, Etot and Etot(k) are the initial state and final energy of the system (when the electron ejects from k shell) respectively. 74 The binding energy referred to the vacuum level could be defined by following equation:

௏ ௙ ௜ ܧ஻ ሺ݇ሻ ൌܧ௧௢௧ െܧ௧௢௧. (13)

And from equation 12 and 13 we will have:

௏ ݄߭ ൌ ܧ௞௜௡ ൅ܧ஻ ሺ݇ሻ. (14)

Binding energy is a relative value which will be addressed to a reference level. This reference for gas system is the vacuum level and for solid sample is the Fermi level of the sample. For solid sample measuring, the sample and spectrometer will be connected by an electrical contact and since they are thermodynamically in equilibrium, their electrochemical potential are equal to Fermi level of the sample. By looking at Figure 2.8, the equation 15 could be derived to determine the binding energy relative to the Fermi level:

ி (௞௜௡ ൅׎௦௣௘௖, (15ܧ஻ ሺ݇ሻ ൅ܧ ߭ ൌ݄ 32

.where ׎spec is the work function of the spectrometer

Figure 2.8. Diagram of sample and spectrometer geometry and energy levels for a system to measure binding energy 74

2.2.8. Raman spectroscopy

In Raman spectroscopy, as in IR spectroscopy, the variation in vibration energy is used to investigate chemical structure, physical forms, qualify the component of a sample using characteristic spectral pattern and also quantify the component in a sample. 75

An incident photon into a matter could go through 3 different processes: absorption, scattering and transmission. The scattered radiation can be detected by putting a detector at a certain angle in contrast to the incident photon and its energy is different from any possible electronic transition. Raman spectroscopy is one of the main scattering techniques and it is used for molecular identification. 75

In Raman spectroscopy, monochromatic radiation will hit the sample and the energy spectrum of scattered radiation will be detected. The incident light polarizes the electronic clouds around the nuclei of the molecule and the molecule will be excited from ground state to the virtual state energy. In most cases, there is no energy variation in scatter beam and it is called Rayleigh scattering. But there is small probability that the energy of the beam decreases or increases after scattering. This scattering it is called Raman scattering. The decrease in energy happens when the molecule excites from ground vibrational level and it ends up to a higher vibrational level. This kind of scattering is called Stokes scattering. On the other hand when the molecule excites 33 from a higher vibrational level and back to the ground state energy level, its energy increases and this is called anti-Stokes scattering. The schematic diagram of the different energy levels involved in these scattering processes is shown in Figure 2.9.

A Raman spectrum shows scattered intensity versus wavenumber. Each peak in the spectrum represents a given Raman shift. Stokes and anti-Stokes shift are symmetrically placed on both sides of Rayleigh scattering peak with different intensity.

Figure 2.9. Schematic diagram of different scattering processes. υ0 is the frequency of incident beam and υ is the frequency of scattered beam. The ground state has the lowest vibrational energy level and the energy for other vibrational levels at top of the ground state is increasing (adapted from ref 75)

The peaks in Raman spectrum could be characterized by their position, intensity, and polarization. The position gives information about the frequency of the vibration mode, the intensity gives information about the number of diffusing molecules and vibrational modes, and the polarization gives information about the symmetry of the corresponding mode. 76

2.3. Device characterization

2.3.1. (Photo) electrochemical measurement

Electrochemical measurement techniques are commonly used to study chemical reactions. The fundamental parameters in electrochemistry consist of potential, current, concentration and time. The relation between these parameters depends on which of them are being controlled, the

34 kinetics of both the homogeneous chemical reaction and the heterogeneous electron transfer reaction, and the geometry of working electrode and the cell. 77

The standard configuration for potentiostats and galvanostats is a three-electrode setup consists of the working electrode (WE), the counter electrode (CE), and the reference electrode (RE). The WE is the electrode which the redox process happens at its geometry, the CE is the second electrode which the current goes through, and the RE is the electrode which no current pass through and it just referred the potential of the WE.

Cyclic voltammetry (CV) is the most common electrochemical techniques. In CV measurement, the generated current by sweeping potential over the time is measured. Photoelectrochemical measurements study the interaction of light with the electrochemical system and it is a combina- tion of photochemistry and electrochemistry. In PEC, the driving force for the measurement is the incoming photon energy which generates an electron-hole pair in semiconductor materials. But it should be mentioned that the light is not the only driving force and it is required to apply a differential potential which is called bias voltage. The reason is that the life time of generated electron-hole pair is short and the bias potential helps to decrease the recombination rate and increase the photocatalysis rate.

3. RESULTS and DISCUSSION

The “Results and discussion” chapter is divided into three sections based on the contributed publications in this thesis work, and on the ongoing activities on all-oxide solar cells:

1. Dynamics and Photochemical H2 Evolution of Dye-NiO Photocathodes with a Biomimetic FeFe-Catalyst

2. The Role of Heterojunctions on the Water Oxidation Performance in Mixed α-Fe2O3/Fe3O4 Hybrid Film Structures

3. 1-Dimensional Heterostructures for All-oxide Solar Cells

3.1. Dynamics and photochemical H2 evolution of dye-NiO photocathodes with a biomimetic FeFe-catalyst

In this section, the electron-transfer processes between three compartment of a photocathode in a dye-sensitized solar fuel device (DSSFD) and its photochemical H2 production will be discussed. The DSSFD is based on the coumarin C343-sensitized mesoporous NiO photocathode, with a biomimetic FeFe-catalyst anchored on the photocathode. In this system, coumarin is the light absorber and after excitation will be reduced by hole injection from NiO. The reduced coumarin injects electron to catalyst and this electron will be used in reduction of proton to hydrogen. The prepared electrode is investigated by Ultraviolet-Visible (UV-Vis) absorption spectroscopy and Fourier transform infrared (FTIR) spectroscopy to confirm successful co- sensitization of NiO by the dye and the catalyst. The transient absorption (TA) spectroscopy and the decay associated spectrum (DAS) are used to study the charge injection and recombination. Finally, the photoelectrochemical (PEC) experiment is performed, to evaluate the light response and the proton reduction capacity of the device.

The NiO photocathode was prepared by tape casting in three steps (Figure 2.1):

In the first step, a well-mixed solution consisting of nickel particles was applied on a FTO glass substrate. Then the paste was spread on the FTO surface by a razor blade. Finally, the electrode was cooked in an oven to film be dried.

FTIR was used to confirm the successful co-sensitization of coumarin and catalyst on NiO electrode, to determine the relative concentration of them on the electrode, and to relate the decrease of the efficiency to the degradation of the catalyst after PEC measurement.

The carbonyl peaks of catalyst show up at the region 2000-2100 cm-1 and it shows the presence of catalyst on the electrode. The comparison of the intensity of the peaks on the co-sensitized sample and the catalyst/NiO sample could give an estimated ratio between coumarin and catalyst (Figure 3.1. left). The intensity comparison of the carbonyl peaks of catalyst before and after PEC measurement confirms that the catalyst is degraded since the intensity is decreased after PEC measurement (Figure 3.1. right).

Figure 3.1. FT-IR spectra of co-sensitized C343/catalyst on NiO (black), catalyst on NiO (red), C343 on NiO (blue) (left Figure) and ATR-FTIR spectra of C343/catalyst before(black) and after (red) PEC measurement (right Figure)

Figure 3.2. UV-Vis spectra of bare NiO (gray), co-sensitized C343/catalyst on NiO (black), catalyst on NiO (red), C343 on NiO (blue)

UV-Vis was used to study the loading of materials on the NiO electrode (Figure 3.2). The gray line belongs to NiO bare film and it shows no peak in the visible spectral region. The red line represents the catalyst on the NiO surface and it shows two peaks at 450 nm and 600 nm. The blue line shows a peak around 430 nm that belongs to coumarin. The black line, which shows both the properties of the catalyst and the coumarin belongs to the co-sensitized sample.

Femtosecond transient absorption (TA) was carried out to study the evolution/decay of different species at various wavelengths and also the number and rate of decay processes at a specific wavelength.

Figure 3.3. TA spectra after excitation at 440 nm of C343 on NiO (A-C) and C343/catalyst on NiO (D-F) at the time scale of 100 fs - 1 ps (A,B,D,E) and 1 ps -1.8 ns (C,F) Figure 3.3 compares the TA spectra of coumarin/NiO system and coumarin/NiO co-sensitized with catalyst. It can be seen that the hole injection dynamics for both systems is almost the same, but after that they differ in the decay process. For coumarin/NiO system (reduced coumarin species absorbs at 600 nm), the decay is slower than the coumarin/NiO co-sensitized with catalyst. Figure 3.4 shows the time evolution of TA femtosecond excitation at 470 nm. The decay for coumarin/NiO co-sensitized with catalyst is multiexponential and faster than coumarin/NiO system. At this wavelength, the quenching of the stimulated emission happens at 1 ps and 800 fs for the coumarin/NiO system and coumarin/NiO co-sensitized with catalyst, respectively. In addition, for the latter system there is no decay at positive TA signal. A control measurement on the catalyst/NiO system shows a weak positive TA signal, which decays quick- ly (red data in Figure 3.4). This is not related to the reduced catalyst. Therefore, it can be point-

38 ed out that the reduction of the catalyst is due to the presence of the coumarin at the surface and it is not possible to happen due to the excitation of the catalyst alone at the surface.

Figure 3.4. Time evolution of transient absorption after fs excitation at 470 nm for C343 on NiO (blue line), C343/catalyst on NiO (black line), and catalyst on NiO (red line)

To examine the device performance, PEC measurement with a three-electrode setup was performed. The PEC cell was a two compartment cell, connected through a membrane. The working electrode and reference electrode (Ag/AgCl) were in the same compartment and the counter electrode (Pt sheet) was on the other compartment. All the electrodes were places in acetate buffer (pH 4.5). The photocurrent response of the coumarin/catalyst system was much higher than the coumarin alone. This photocurrent difference can be attributed to the hydrogen generation. This was also confirmed by the gas chromatography – mass spectroscopy (GC-MS). Fig- ure 3.5 shows the photocurrent density vs. time and GC-MS data.

Figure 3.5. A) Photocurrent density response of the photocathode in pH 4.5 and at bias voltage -0.3 V vs. Ag/AgCl reference electrode under white chopped light B) GC-MS under continuous photolysis It should be mentioned that the decrease in photocurrent is due to the degradation of the catalyst.

3.2. The role of heterojunctions on the water oxidation performance in

mixed α-Fe2O3/Fe3O4 hybrid film structures

In this part, the preparation and characterization of an α-Fe2O3/Fe3O4 photoanode will be discussed. The hematite photoanode is prepared by CVD. Through a post-deposition plasma- chemical reduction at controlled temperature, different ratio between iron oxide phases in the photoanode is prepared. The electrode is examined by light optical microscopy, SEM, and EDS to investigate microstructure and morphology of the electrode. XRD, XPS, and Raman spectroscopy is carried to a full study of prepared photoanode. At the end, PEC measurement is performed to evaluate the water oxidation performance of the solar fuel device (SFD).

The hematite/magnetite photoanode was prepared by CVD deposition of hematite coupled with the post-deposition plasma reduction of hematite to form magnetite.

The optical microscopy was used to analyze the surface of iron oxide samples. It shows that the grain size of the particles increases by increasing the deposition time (Figure 3.6).

Figure 3.6. Optical images of iron oxide samples with deposition time of 5 min (left) and 45 min (right)

SEM was used to investigate the surface and bulk structure of CVD prepared iron oxide electrodes by 5 min and 45 min deposition time. Figure 3.7 shows the top view and cross-section images of these two samples.

Figures 3.7a,d represent the cross-section of 5 min and 45 min samples, respectively. The largest grain size in the structure is less than 500 nm for the 5 min sample and less than 8 μm for the 45 min sample. The Figures 3.7b,e represent the top view of the samples and the spindle-like grains which cover most of the surface are the hematite structure which is prepared by CVD. The 50 nm fine equiaxed grains which are shown by white arrows are the hematite which is formed by oxidation of magnetite. 40

(a) (b) (c)

500 nm 400 nm 200 nm

(d) (e) (f)

5 μm 400 nm 200 nm

Figure 3.7. Cross-section and top view SEM micrographs of iron oxide photoelectrodes prepared by CVD on FTO substrate. (a-c) represent 5 min and (d-f) represent 45 min samples. Images (c) and (f) display the morphology of the top layer (100-350 nm) for both deposition times.

The EDS was carried out to investigate morphological and crystallographic changes by variation of the deposition time. EDS spectra of the 5 min and 45 min samples are shown in Figure 3.8.

Figure 3.8. EDS spectra of samples with the different deposition time: 5 min (left) 45 min (right)

It was previously shown by Figueroa G. et al. that it is possible to distinguish hematite and magnetite by EDS. 78 In the 5 min sample, the Fe/O ratio is almost 70/30 weight percentage, which fits with the value for hematite. The ratio for magnetite was shown by Figueroa to be 72/28. The measured value for the 45 min sample is 74/26, which is closer to the value for magnetite than hematite. The deviation could be due to existence of both species in the 45 min sample. 41

The XRD was used to examine the bulk composition of the samples. Figure 3.9 shows the diffractogram of the 5 samples which differ in the deposition time. In all the samples, there are peaks at the position (2θ=33.1°, 35.7°, 54.1° and 57.6°) which belongs to hematite, but the samples with higher deposition time show two extra peaks (2θ=30° and 43.1°) which are matched with magnetite.

Figure 3.9. Diffractogram of iron oxide samples prepared with different deposition time

By comparison of XPS spectra of the samples with 5 min and 45 min deposition time, the formation of the magnetite in the latter sample was proved (Figure 3.10).

Figure 3.10. The XPS spectra of iron oxide samples with different deposition time (left) 5 min (right) 45 min Raman spectroscopy was performed to investigate the local distribution of iron oxide species in the samples. Two different lasers with 532 nm and 785 nm wavelength have been used. The first one is more surface sensitive, while the latter has larger penetration depth and gives more information about the bulk composition. The laser source with wavelength of 532 nm only shows hematite in all samples with different deposition time, which means the surface of the films, is covered by hematite. When the laser source with wavelength of 732 nm is used a new

42 peak appears in samples with higher deposition time (20 min and higher) which is representa- tive for magnetite (Figure 3.11).

(a) Fe O / Si, O = 785 nm (b) Fe O / Si, O = 532 nm x y ex x y ex +: hematite 45min + + O: magnetite 30min O 20min *: Si + + 10min + 5min + + + + + + Intensity Intensity

200 400 600 800 1000 1200 1400 200 400 600 800 1000 1200 1400 Raman shift (cm-1) Raman shift (cm-1)

Figure 3.11. Raman spectra of iron oxides with different deposition time and two different excitation laser source a) 785 nm b) 532 nm

The thickness of the sample (11 μm) with highest deposition time (45 min) made it possible to perform a cross-section Raman study to investigate better the geometry of the species in the sample. The Raman spectra show that by moving from the substrate to the surface of the sample, the magnetite peak intensity decreases and the hematite peak intensity increases. This means that the hematite is more pronounced at the surface and the magnetite is more on the bottom of the sample (Figure 3.12).

(a) 45min, O = 785 nm (b) 45min, O = 532 nm ex ex

+ + + + *: Si + o: magnetite + +: hematite + + + + o + *: Si +

o: magnetite +: hematite + +o Intensity Intensity o + o o + o + o o o o x2 x2 * * 200 400 600 800 1000 1200 1400 200 400 600 800 1000 1200 1400 Raman shift (cm-1) Raman shift (cm-1)

Figure 3.12. Cross-section Raman spectra of the sample with 45 min deposition time a) 785 nm b) 532 nm excitation laser. The red arrows show moving from substrate to the surface

To examine the device performance regards to the different phases in photoanode, PEC measurement with a three-electrode setup was performed. The current-voltage (I-V) curve was recorded in a NaOH solution with 1 M concentration and under a chopped white light with 1 sun intensity. The generated photocurrent for the samples with 5 min and 45 min deposition time was 0.28 mA/cm2 and 0.48 mA/cm2 respectively (Figure 3.13).

Figure 3.13. a) Photocurrent density response vs. potential b) The comparison of photocurrent density at 1.23 V (vs. RHE) and onset potential for iron oxide samples with different phases and thicknesses It should be mentioned that the photocurrent from lower deposition time to higher deposition time is not increasing linearly. The photocurrent from 5 min to 10 min sample decreases due to increasing thickness of the hematite layer and absence of magnetite, since even if the hematite is a good light absorber, it is not a good electric conductor. After that by formation of magnetite and increasing the thickness of magnetite the photocurrent increases from 10 min to 45 min sample. The better performance could be due to better electrical conductivity of magnetite and also to the fewer defects in the structure of the films.

3.3. 1-Dimensional heterostructures for all-oxide solar cells

The charge carrier mobility and charge separation are two major problems in the metal oxide semiconductors. Overcoming these issues could bring a big improvement in the all-oxide solar cells and alongside with the stability, abundance, non-toxicity, and low cost could make them competitive with existing solar cell devices like silicon solar cells and perovskite solar cells.

Figure 3.14 compares light absorption and charge separation based on the core-shell nanowire structure and planar thin film structure. Nanowires improve the performance through the light scattering and reducing the charge carriers’ path length.

Figure 3.14. Comparison of light absorption and charge separation (left) in ZnO nanowires coated with Cu2O as light absorber (right) and in planar ZnO coated with Cu2O As it is shown in the scheme, the core-shell consists of ZnO nanowires with the copper oxide coating. The SEM images from the samples are shown in Figure 3.15. To examine the p-n junction functioning, the macroscopic IV curve was recorded and it is shown in Figure 3.16. The normal rectifying behavior for a p-n junction was observed that proves the p-n junction is functioning.

Figure 3.15. Bare ZnO NWs a) cross-section b) plane view and after deposition of copper oxide c) at 250 oC d) 300 oC

0.000000

-0.000005 copper oxide 70 nm thick copper oxide 45 nm thick -0.000010

-0.000015 i [A]

-0.000020

-0.000025

-0.000030

-4 -2 0 2 4 V [V]

Figure 3.16. Macroscopic I-V curve of ZnO/copper oxide p-n junction with different copper oxide thickness

4. CONCLUSION and FUTURE OUTLOOK

In this work, two different solar fuel devices were explored for water splitting.

The first system performed water splitting based on the dye-sensitized solar cell device. The DSSFD was consisting of mesoporous NiO semiconductor photocathode co-sensitized with C343 organic dye and a biomimetic Fe-Fe catalyst as working electrode, Ag/AgCl as reference electrode and a Pt foil (1 cm2) as reference electrode. Previously some other catalysts like cobalt catalysts have been tested in this kind of device but they suffer from instability in acidic pH condition for hydrolysis. In this work for the first time a biomimetic FeFe-catalyst was tested and the faradic efficiency of 50% was achieved.

In the second system, a photoanode based on the different iron oxide phases was tested in SFD. A CVD deposited hematite with thickness of 1.5 μm was tested as a reference. Then we tried to engineer the photoanode by making a hybrid electrode consist of hematite and magnetite and we were able to increase the photocurrent density more than 50% due to better electrical conductivity of magnetite and less defects in the film structure.

The future work, based on the preliminary results illustrated in section 3.3, will be focused on the engineering the metal oxide nanostructures, like using nanowire structures to improve light absorption and charge separation. Optimization of p-type layer thickness, using some passive layer between core-shell structures, the direct/inverse geometry, using some p-type hybrid structures are some of the other plans for the future work. In addition to the previous characterization techniques, new characterization methods like characterization with scanning probe microscopy based methods such as Kelvin probe force microscopy (KPFM) and scanning spreading resistance microscopy (SSRM) and impedance spectroscopy will be applied on the samples.

5. REFERENCES

1 U.S. Energy Information Administration (eia), energy explained, sources of energy, 2017 2 International Energy Agency (IEA), Energy Statistics Manual, 2004 3 Godfrey Boyle, Renewable Energy, power for a sustainable future, Oxford University Press, Oxford, 2012 4 Ghamgosar P., Independent thesis Advanced level, Uppsala University, 2013 5 An Assessment of Solar Energy Conversion Technologies and Research Opportunities, Stanford University, Technical Assessment Report, 2006 6 Introduction to solar radiation, Newport Corporation, 2017 7 Kasten F. and Young A. T., Applied Optics, 1989, 28, 4735–4738 8 U.S. Solar Radiation Resource Maps, October 2012 9 International Energy Agency (IEA), Key World Energy Statistics, 2016 10 REN21, Renewables 2017 Global Status Report (GSR), 2017, ISBN 978-3-9818107-6-9 11 George W. Crabtree and Nathan S. Lewis, American Institute of Physics, 2007, S-0031-9228- 0703-010- 4 12 Yu Z., Fei L., and Sun L., Energy Environ. Sci., 2015, 8, 760−775 13 Tachibana Y., Vayssieres L., and Durrant J. R., Nat. Photonics, 2012, 6, 511–518 14 Duan L., Tong L., Xu Y., and Sun L., Energy Environ. Sci., 2011, 4, 3296–3313 15 Barber J., Chem. Soc. Rev., 2009, 38, 185–196 16 Dau H., and Zaharieva I., Acc. Chem. Res., 2009, 42, 1861–1870 17 Lubitz W., Reijerse E. J., and Messinger J., Energy Environ. Sci., 2008, 1, 15–31 18 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., and Meyer T. J., J. Am. Chem. Soc., 2016, 138, 13085−13102 19 Fujishima A., and Honda K., Nature, 1972, 238, 37−38 20 Wen F., and Li C., Acc. Chem. Res., 2013, 46, 2355–2364 21 Young K. M. H., Klahr B. M., Zandi O., and Hamann T. W., Catal. Sci. Technol., 2013, 3, 1660−1671 22 Katz M. J., Riha S. C., Jeong N. C., Martinson A. B. F., Farha O. K., and Hupp, J. T., Coord. Chem. Rev., 2012, 256, 2521−2529 23 Hong S. J., Lee S., Jang J. S., and Lee J. S., Energy Environ. Sci., 2011, 4, 1781−1787 24 Ling Y., Wang G., Wheeler D. A., Zhang J. Z., and Li Y., Nano Lett., 2011, 11, 2119−2125 25 Navarro R. M., Alvarez-Galvan M. C., Villoria de la Mano J. A., Al-Zahrani S. M. , and Fierro J. L. G., Energy Environ. Sci., 2010,3, 1865–1882 26 Maeda K., and Domen K., J. Phys. Chem. Lett., 2010, 1, 2655–2661 27 Chen X., Shen S., Guo L., and Mao S. S., Chem. Rev. 2010, 110, 6503−6570 28 Kudo A., and Miseki Y., Chem. Soc. Rev., 2009, 38, 253–278 29 Kay A., Cesar I., and Grätzel M., J. Am. Chem. Soc., 2006, 128, 15714−15721 30 Butler M. A. J., Appl. Phys., 1977, 48, 1914−1920 31 Jafari T., Moharreri E., Shirazi Amin A., Miao R., Song W., Suib S. L., Molecules, 2016, 21, 900:1−900:29 32 Tee S. Y., Win K. Y., Teo W. S., Koh L-D., Liu S., Teng C. P., and Han M-Y., Adv. Sci., 2017, 4, 1600337:1−1600337:24 33 Kudo A., and Miseki Y., Chem. Soc. Rev. 2009, 38, 253–278 34 Linsebigler A.L., Lu G., Yates J.T., Chem. Rev., 1995, 95, 735–758 35 Zou Z., Ye J., Sayama K., and Arakawa H., Nature, 2001, 414, 625–627 36 Ong W-J., Tan L-L., Hau Ng Y., Yong S-T., and Chai S-P., Chem. Rev., 2016, 116, 7159−7329 47

37 O'Regan B., and Grätzel M., Nature, 1991, 353, 737–740 38 Hagfeldt A., and Grätzel M., Acc. Chem. Res., 2000, 33, 269–277 39 Hagfeldt A., Boschloo G., Sun L., Kloo L., and Pettersson H., Chem. Rev., 2010, 110, 6595– 6663 40 Zhou W., Li W., Wang J.-Q., Qu Y., Yang Y., Xie Y., Zhang K., Wang L., Fu H., and Zhao D., J. Am. Chem. Soc. 2014, 136, 9280–9283 41 Kay A., Cesar I., and Grätzel M., J. Am. Chem. Soc., 2006, 128, 15714–15721 42 Kim J.Y., Magesh G., Youn D.H., Jang J.-W., Kubota J., Domen K., and Lee J.S., Sci. Rep., 2013, 3, 2681:1–2681:8 43 Grätzel M., J. photochem. Photobioal., C, 2003, 4, 145−153 44 Oku T., Kakuta N., Kobayashi K., Suzuki A., and Kikuchi K., Progress in Natural Science: Ma- terials International, 2011, 21, 122−126 45 Rensmo H., Keis K., Lindstrom H., Sodergren S., Solbrand A., Hagfeldt A., Lindquist S. E., Wang L. N., and Muhammed M., J. Phys. Chem. B, 1997, 101, 2598−2601 46 Bauer C., Boschloo G., Mukhtar E., and Hagfeldt A., J. Phys. Chem. B, 2001, 105, 5585−5588 47 Roh S.-J., Mane R. S., Min S.-K., Lee W.-J., Lokhande C. D., and Han S.-H., Appl. Phys. Lett., 2006, 89, 253512:1−253512:3 48 Odobel F., Le Pleux L., Pellegrin Y., and Blart E., Acc. Chem. Res., 2010, 43 (8), 1063–1071 49 Huang Z., Natu G., Ji Z., Hasin P., and Wu Y., Phys. Chem. C, 2011, 115 (50), 25109–25114 50 Du S., Cheng P., Sun P., Wang B., Cai Y., Liu F., Zheng J., and Lu G., Chem. Res. Chin. Univ., 2014, 30(4), 661−665 51 Feldt S. M., Gibson E. A., Gabrielsson E., Sun L., Boschloo G., and Hagfeldt A., J. Am. Chem. Soc., 2010, 132, 16714−16724 52 Tsubomura H., Matsumura M., Nomura Y., and Amamiya T., Nature, 1976, 261, 402−403 53 Leandri V., Ellis H., Gabrielsson E., Sun L., Boschloo G., Hagfeldt A., Phys. Chem. Chem. Phys., 2014, 16, 19964−19971 54 Yella A., Lee HW., Tsao HN., Yi C., Chandiran AK., Nazeeruddin M. K., Diau EW., Yeh CY., Zakeeruddin SM., and Grätzel M., Science, 2011, 334(6056), 629−634 55 Mathew S., Yella A., Gao P., Humphry-Baker R., Curchod B. F. E., Ashari-Astani N., Tavernelli I., Rothlisberger U., Nazeeruddin M. K., and Grätzel M., Nature chemistry, 2014, 6, 242−247 56 Yu Z., PhD thesis, Royal Institute of Technology, 2012, ISBN 978-91-7501-231-5 57 Gorlov M., and Kloo L., Dalton Trans., 2008, 20, 2655−2666 58 Ahmad S., Bessho T., Kessler F., Baranoff E., Frey J., Yi C., Gr Grätzel M., and Nazeeruddin M. K., Phys. Chem. Chem. Phys. 2012, 14, 10631−10639 59 Roy Mayhew J. D., Boschloo G., Hagfeldt A., and Aksay I. A., ACS Appl. Mater. Interfaces, 2012, 4, 2794−2800 60 Aegerter M. A. et al. (eds.), Sol-Gel Technologies for Glass Producers and Users, Springer Sci- ence+Business Media New York, 2004 61 Dennis J. W., Tape casting advanced Materials, Ceramic Industry, 2007 62 Mistler R. E., Twiname E. R., Tape Casting: Theory and Practice, 2000 63 Mitchell B. S., An Introduction to Materials Engineering and Science for Chemical and Materi- als Engineers, 2004 64 Choy K. L., Progress in Materials Science, 2003, 48, 57–170 65 De Lodyguine J. S., Illuminant for incandescent lamps, US patent 575002, 1893 66 Birkner N., and Wang Q., LibreTexts, Chemistry, How an FTIR spectrometer works, 2015 67 Newport, Introduction to FTIR Spectroscopy, 2017 68 Soderberg T., Libretexts, Chemistry, Ultraviolet and visible spectroscopy, 2017 69 Grolik B., and Kopp J., Optical properties of Thin Semiconductor Films, 2003 70 Goldstein J., Newbury D., Joy D., Lyman C., Echlin P., Lifshin E., Sawyer L., and Michael J., Scanning Electron Microscopy and X-Ray Microanalysis, Springer, New York, 2003 48

71 Ermrich M., and Opper D., X-ray powder diffraction, PANalytical, Kassel, 2011 72 West A. R., Basic Solid State Chemistry, John Wiley & Sons Ltd, Chichester, 1999 73 Cullity B. D. and Stock S. R., Elements of X-Ray Diffraction, Prentice Hall, New York, 2001 74 Feldman L. C. and Mayer J. W., Fundamentals of Surface and Thin Film Analysis, North- Holland, New York, 1986 75 Smith E. and Dent G., Modern Raman Spectroscopy – A Practical Approach, John Wiley & Sons Ltd, Chichester, 2005 76 Noël M., Doctoral Thesis, Luleå University of Technology, 2014, ISBN 978-91-7583-203-6 77 Rossiter B. W., and Hamilton J. F., Physical methods of chemistry – Electrochemical methods, Wiley-Interscience publication, New York, 1986 78 Figueroa G., Moeller K., Buhot M., Gloy G., and Haberlash D., 10th international congress for applied mineralogy (ICAM), 2011

Paper I

This is an open access article published under an ACS AuthorChoice License, which permits copying and redistribution of the article or any adaptations for non-commercial purposes. Letter http://pubs.acs.org/journal/aelccp

Dynamics and Photochemical H2 Evolution of Dye−NiO Photocathodes with a Biomimetic FeFe-Catalyst † † ‡ § Liisa J. Antila, Pedram Ghamgosar, , Somnath Maji, Haining Tian, Sascha Ott, and Leif Hammarström* Department of ChemistryÅngström Laboratory, Uppsala University, Box 523, SE75120 Uppsala, Sweden

*S Supporting Information

ABSTRACT: Mesoporous NiO ﬁlms were cosensitized with a coumarin 343 dye and a proton reduction catalyst of the [Fe2(CO)6(bdt)] (bdt = benzene-1,2-dithiolate) family. Femto- second ultraviolet−visible transient absorption experiments directly demonstrated subpicosecond hole injection into NiO t ∼ from excited dyes followed by rapid ( 50% 6 ps) reduction of the catalyst on the surface with a ∼70% yield. The reduced catalyst was long-lived (2 μs to 20 ms), which may allow protonation and a second reduction step of the catalyst to occur. A photoelectrochemical device based on this photocathode produced H2 with a Faradaic eﬃciency of ∼50%. Fourier transform infrared spectroscopy and gas chromatography experiments demonstrated that the observed device deterioration with time was mainly due to catalyst degradation and desorption from the NiO surface. The insights gained from these mechanistic studies, regarding development of dye−catalyst cosensitized photocathodes, are discussed.

olecular catalysts are attractive for solar fuels production. We give direct spectroscopic evidence that the production because they can be both rapid and catalyst is attached to the NiO surface and could monitor both M selective and require little material per active site. its photoreduction and charge recombination. Their properties and interaction with the substrate can be DSSFDs based on [2] showed poor NiO binding and fi tuned to a great extent via changes in their rst and second negligible photocurrents in aqueous media (not shown). ff coordination sphere, which o ers great potential for rational Therefore, [1] was prepared with a phosphonate group that design. To incorporate molecular catalysts into a device, one was expected to allow for more stable binding to NiO and a approach is to combine it with a dye-sensitized, mesoporous − longer linker that may slow charge recombination between the semiconductor film.1 6 Such dye-sensitized solar fuel devices reduced catalyst and the NiO holes (Figure 1). The electronic (DSSFDs) have attracted rapidly increasing interest, and several − photoanodes based on TiO or nano-ITO films and molecular properties should nevertheless be similar, with E1/2 = 1.18 V 2 − − +/0 28 water oxidation catalysts have been reported.1,3 5,7 9 For the vs Fc for [2] in acetonitrile. C343 absorbs mainly in the corresponding photocathodes for H2 production or CO2 blue part of the spectrum but was chosen as dye for these reduction, there are fewer examples, and the most common − material is mesoporous NiO.10 23 In most cases the catalyst was not anchored to the NiO surface or was likely to detach upon reduction during the catalytic cycle. The photocathodes show poorer performance than the photoanodes, but so far very little detail on these systems has been given. From studies of solar cell materials, it is known that dye−NiO charge recombination is typically very rapid, often on a subnanosecond − time scale,24 26 which is likely to hamper also the performance of DSSFDs. Our group has previously shown that potential Figure 1. Structures of coumarin 343 (C343) and catalysts [1] and molecular catalysts, inspired by the active site of [FeFe]- [2]. hydrogenases, can be photoreduced by a dye on NiO.27,28 In the present work, we were able to make the first DSSFD based Received: October 6, 2016 on a biomimetic FeFe catalyst, on coumarin C343-sensitized Accepted: November 2, 2016 NiO photoanodes, and demonstrate photochemical H2 Published: November 2, 2016

Figure 2. (A) UV−vis spectra and (B) FTIR spectra of the C343:[1] cosensitized NiO ﬁlm (black line), [1]/NiO reference (red line), C343/ NiO reference (blue line), and NiO ﬁlm before sensitization (gray line, not shown in panel B).

Figure 3. TA spectra of C343:[1]-cosensitized NiO after excitation at 440 nm (110 nJ, absorbed photon density 1.7 × 1015 photons/(cm2 × pulse) (A) 500 fs to 1.8 ns after excitation. (B) Magnified view of 1 ps to 1.8 ns after excitation. Inset in panel A shows time evolution of TA signal at 415 nm. studies because it gives rapid hole injection into NiO; it has a with a 1:1 C343:[1] solution (cf. ref 28), assuming that suitable redox potential (E0(C3430/−) ∼−1.6 V vs Fc+/0)in extinction coefficients for these bands are similar in solution DMF29 (see comment in ref 28); and its transient absorption and on NiO, we estimate that the relative concentrations of the spectra do not cover the band of the reduced catalyst (see dye and catalyst are between 1:2 and 2:1 in the C343:[1]/NiO below). sample. Figure 2A shows the ultraviolet−visible (UV−vis) absorption Femtosecond transient absorption (TA) spectra obtained spectra of the NiO films sensitized with C343 dye and catalyst from the control samples, C343/NiO and [1]/NiO films, after [1]. The C343-sensitized NiO (denoted C343/NiO) shows a exciting at 440 nm are presented in the Supporting dye absorption maximum around 400 nm, which is blue-shifted Information. The 440 nm pump wavelength mainly excites compared to the 422 nm absorption maximum reported earlier the surface-bound C343 dye, although [1] also exhibits some for C343 on NiO.30 This is probably due to the different absorption at this wavelength (ε(C343)= 15 100 M−1 cm−1 vs background absorption from the double-layered NiO film ε([2]) = 1500 M−1 cm−1 at 440 nm).29,31 For the C343/NiO (Figure 2A), used in this study to increase loading of C343 and reference (Figures S2 and S3), the obtained results agree well [1], compared to the single-layered film in ref 30. [1] on NiO with those reported by Morandeira et al.,30 and a global fit with ([1]/NiO) shows weak absorption bands centered at 450 and four exponentials gives time constants of <150 fs, 1 ps, 9 ps, 600 nm, similar to those reported for the [Fe2(bdt) (CO)6] and 1.2 ns. The amplitude spectrum of each lifetime complex in solution.31 The spectrum of the C343:[1]- component of the global fit constitutes a decay associated cosensitized NiO film (C343:[1]/NiO) shows mixed features spectrum (DAS) that shows the transient absorption changes from the spectra of both C343/NiO and [1]/NiO references, related to that component. On the basis of the DAS (Figure namely, a broad band centered at ∼450 nm and a weak band S3), the time constants are assigned to the hole injection from around 600 nm. Fourier transform infrared (FTIR) spectra excited C343 dye to the valence band of NiO (<150 fs, measured for the samples are shown in Figure 2B. The formation of the C343•−/NiO(+) charge separated state) and to vibrational peaks of the carbonyl groups31 of [1] at 2005, 2042, the recombination of the injected hole with the reduced and 2079 cm−1 in the spectra of the C343:[1]/NiO sample and C343•− radical (9 ps). The DAS corresponding to the 1 ps time the [1]/NiO reference verify the presence of [1] on the surface constant is assigned to the hole injection and recombination of the NiO film. By comparing the intensities of the vibrational taking place in overlapping time scales, reflecting the peaks (at 2000−2100 cm−1 for [1] and ca. 1300 cm−1 for heterogeneity of the sensitized nanoparticle film and the C343) in the cosensitized sample and in the two references resulting complexity of the charge-transfer dynamics.30 The

1107 DOI: 10.1021/acsenergylett.6b00506 ACS Energy Lett. 2016, 1, 1106−1111 ACS Energy Letters Letter

Figure 4. Time evolution of transient absorption after fs excitation (A) at 600 nm and (B) at 470 nm for C343:[1]/NiO (black squares and lines), C343/NiO (blue circles and lines), and [1]/NiO (red triangles and lines). (C) Time evolution of transient absorption after a 10 ns excitation pulse for C343:[1]/NiO at 470 nm. The red line is the fit with four exponentials to the experimental data: 1.9 μs (38%), 52 μs (24%), 230 μs (20%), and >10 ms (17%). spectrum at 1.2 ns shows a weak positive signal over the whole of [1]− and C343•−, we estimate a yield of ∼70% for electron spectral range, with weak peaks around 350 and 475 nm, and transfer from C343•− to [1] (see the Supporting Information represents a fraction of triplet excited C343.30 for details). This means that surface electron transfer from The femtosecond TA spectra for the C343:[1]/NiO film in C343•− to [1] competes favorably with recombination between air are presented in Figure 3 (for more details, see Figures S4 C343•− and NiO holes, which has ultrafast components (τ ∼ 2 and S5). The initial hole injection dynamics is rather similar to ps;30 see also above). The surface electron-transfer half-time is that of the C343/NiO reference, but the subsequent reaction is •− somewhat shorter and the yield somewhat higher than for the quite different. At 600 nm, where mainly the C343 radical previously reported system C343:[2]/NiO for which t ∼ 10 30 50% absorbs (excited C343 has an isosbestic point here ), the TA ps and a yield of 40−80% was reported.28 However, by direct signal decays faster in the cosensitized sample than in the comparison of the TA signals in ref 28 with those in Figure 3,a reference without the [1] catalyst (see Figure 4A). The decay is yield of ca. 40% in that paper seems most consistent with the ∼ multiexponential with important contributions already at 1ps data. For both C343:[1]/NiO and C343:[2]/NiO, surface and continuing up to 1000 ps. In parallel, a new positive band electron transfer is very rapid and there is no indication of dye− between 410 and 500 nm and a negative band between 350 and catalyst segregation or of isolated C343 molecules. It is ∼ 420 nm are observed to form 1 ps after the excitation and interesting to note that the related process of hole hopping persist beyond the time window of the measurement (Figure between dyes on mesoporous TiO is typically several orders of − 2 3B). The positive band is very similar in shape and position to magnitude slower.32 34 We propose that essentially each C343 the flash−quench generated spectrum of singly reduced 31 dye on NiO is in close contact with a catalyst molecule in the [Fe2(bdt) (CO)6] and is therefore assigned to the reduced τ cosensitized samples, so that surface electron transfer is a [1] that persists beyond the time scale of this experiment ( >2 downhill, single-step process with a short electron-transfer ns). distance. The flexible linking group of [1] may further facilitate The DAS (Figures S5 and S6) and time evolution trace at close intramolecular contact. In contrast, many of the hole 470 nm (Figure 4B) gives further insight into the formation hopping studies on TiO have involved multiple, self-exchange dynamics of reduced [1]. In the C343/NiO reference, the TA 2 steps and may have been limited by steps on the heterogeneous signal at this wavelength corresponds to the quenching of the stimulated emission within 1 ps, which is followed by the decay surface where the transfer distance is longer. To assess whether the reduced [1] catalyst is sufficiently of relatively weak positive absorption that is assigned to the fi reduced C343•− radical (blue data). In the C343:[1]/NiO long-lived for the rst step of the water reduction process, namely, the protonation of the reduced catalyst, the lifetime of sample, the stimulated emission is quenched faster and the •− signal turns positive at 800 fs after excitation (black data). This [1] was recorded on a nano- to millisecond time scale. At 200 fi ns delay after laser flash excitation at 440 nm (10 ns, 2.6 mJ/ positive TA signal does not decay signi cantly in the 2 ns time 2 window of the measurement. A control experiment with the cm ), the C343:[1]/NiO sample shows a strong transient [1]/NiO reference shows only a weak positive signal at this absorption band between 400 and 510 nm (Figure S9), similar wavelength (red data) that decays quite rapidly and is unrelated to what was observed at >10 ps in Figure 3. The decay of this to the reduction of the catalyst (see Figure S7). Therefore, we signal at 470 nm clearly extends over several orders of can conclude that the reduction of the [1] catalyst requires the magnitude (Figure 4c). Fitting of the TA signal required four presence of the C343 dye on the surface of the NiO and cannot exponentials, with roughly equal contributions, with time be carried out by direct excitation of [1] itself. As is clear from constants ranging from ∼2 μs to >10 ms (see Figure 4C the TA spectra in Figure 3 and the traces in Figure 4, reduction caption). We suggest that the slower charge recombination •− ∼ 28 of [1] by C343 is not single exponential. A half-time of t50% compared to the case of [2] on NiO is due to the longer 6 ps was estimated for this process from a plot of the 410−500 linker group in [1]. According to Mirmohades et al.,31 − nm band maximum as a function of the delay time (see Figure protonation of the [Fe2(bdt) (CO)6] complex in acidic S8). This is in agreement with the time evolution of the TA solutions may take place on the microsecond time scale; signal at 470 nm, which suggests that the catalyst reduction therefore, with a lifetime extending to the millisecond time takes place mainly on the time scale of 1−10 ps. The yield of scale, the reduced [1] should be able to undergo protonation [1]− generation is therefore high: from the relative TA signals with high yield. Scheme 1 illustrates the sequence of reactions

1108 DOI: 10.1021/acsenergylett.6b00506 ACS Energy Lett. 2016, 1, 1106−1111 ACS Energy Letters Letter induced by single-photon absorption and suggests a mechanism a photocurrent response that is significantly higher than that of for the catalytic cycle of [1]. the C343/NiO reference, indicating that the photocurrent is due to proton reduction by the hydrogen evolution catalyst. − Scheme 1. (a) Sequence of Photoinduced Electron-Transfer Upon photoelectrolysis of the photocathode under 0.3 V vs Reactions Observed and the Subsequent Protonation Ag/AgCl and light illumination for 1100 s, produced hydrogen Suggested; (b) Suggested Mechanism of Proton Reduction (H2) could be detected in situ with a gas sensor and by gas by [1] under Photochemical Conditions, Supported by Data chromatography−mass spectrometry (GC-MS). The corre- a in Ref 31 sponding data is shown in Figures 5BandS11.The photocathode based on C343:[1] produced 12.4 nmol of hydrogen, and the charge that passed through the circuit was 4.8 mC (Figure S10), which renders a Faradaic efficiency of ca. 50%. Thus, although the catalyst undergoes at most a few (≤3) turnovers during 1100 s, the initial photoelectrochemical H2 evolution occurs with a decent Faradaic efficiency. The photocurrent density is, however, only ca. 10 μAcm−2, which is 3 orders of magnitude smaller than for an optimal system under full sun (ca. 10 mA cm−2). Poor light-harvesting of C343 (Figure 2A) can be expected to account for at least 1 order of magnitude losses. Another large loss factor is probably that charge recombination of the catalyst and NiO competes with protonation and second reduction of the catalyst. Each dye is excited at best only once per second under full sun, and for catalytic turnover to be more rapid than that, each catalyst must be reduced by several dyes in cooperation. Increasing the rate ratio of catalyst turnover and charge recombination is a major feature to consider in further design of dye-sensitized solar fuel devices. Obviously, a dye with a broader absorption spectrum in the visible would also be advantageous. Notably, the photocurrent from the C343:[1]/NiO sample gradually decreased, which could be caused by catalyst decomposition. Rapid performance deterioration is still typical − for dye-sensitized solar fuel devices.1 5 GC-MS experiments (Figure S11) confirm that CO ligands are released from C343: [1]/NiO during photoelectrolysis, which could be responsible for the gradual deactivation of the photocathode. Attenuated a Disproportionation of [FeFeH] (dashed line) is a conceivable total reflection infrared (ATR-IR) spectra of C343:[1]/NiO alternative to the more intuitive, sequential ECEC mechanism (solid before and after photoelectrolysis (Figure 6) further prove the lines). degradation of catalyst by releasing CO ligands. The intensity of the three carbonyl bands of [1] around 1980−2100 cm−1 To evaluate the light response and proton reduction capacity decreased to ca. one-third of their initial values during of the sensitized NiO photocathodes, photoelectrochemical photoelectrolysis, while for example the carbonyl band of − (PEC) experiments (Figure 5) were performed with a three- C343 (∼1740 cm 1) was relatively unaffected. The decrease of electrode system as described in the Supporting Information. the catalyst bands corresponds to the decrease of photocurrent The C343:[1]/NiO electrode in acetate buffer (pH 4.5) shows by the end of the experiment (Figures 5A and S10). This

Figure 5. (A) Amperometric photocurrent density vs time recorded from the photocathodes at a bias of −0.3 V vs Ag/AgCl under chopped 2 ﬀ white light at an intensity of 100 mW/cm in pH 4.5 acetate bu er and (B) H2 evolution under continuous photolysis. Black curves, C343/ NiO; red curves, C343:[1]/NiO. Inset: magniﬁed view of the data for C343/NiO.

1109 DOI: 10.1021/acsenergylett.6b00506 ACS Energy Lett. 2016, 1, 1106−1111 ACS Energy Letters Letter ■ AUTHOR INFORMATION Corresponding Author *E-mail: [email protected]. Present Addresses ‡ P.G.: Department of Engineering Sciences and Mathematics, LuleåUniversity of Technology, Sweden. § S.M.: Department of Chemistry, Indian Institute of Technol- ogy, Hyderabad, India Author Contributions † L.J.A. and P.G. contributed equally to the work. Notes Figure 6. ATR-IR spectra of C343:[1]/NiO before (black) and after The authors declare no competing financial interest. (red) the PEC experiments. ■ ACKNOWLEDGMENTS This work was supported by the Swedish Research Council, the Knut and Alice Wallenberg Foundation, and the Swedish suggests that catalyst degradation is the main reason for the Energy Agency. decrease in PEC activity of the photocathode. To get more insight into the degradation of the catalyst ■ REFERENCES during the PEC measurements, a similar C343:[1]/NiO sample ff fi (1) Youngblood, W. J.; Lee, S.-H. A.; Kobayashi, Y.; Hernandez- on nonconducting CaF2 was soaked in the pH 4.5 bu er rst 1 Pagan, E. A.; Hoertz, P. G.; Moore, T. A.; Moore, A. L.; Gust, D.; h in the dark and then another hour under illumination. An Mallouk, T. E. Photoassisted Overall Water Splitting in a Visible Light- FTIR spectrum of the sample was recorded before and after the Absorbing Dye-Sensitized Photoelectrochemical Cell. J. Am. Chem. soaking (Figure S12). From Figure S12 it is obvious that the Soc. 2009, 131, 926−927. buffer has little effect on the catalyst integrity as only small (2) Tian, H. Molecular Catalyst Immobilized Photocathodes for changes can be seen in the intensities of the IR peaks before Water/Proton and Carbon Dioxide Reduction. ChemSusChem 2015, 8, and after soaking in the dark. In contrast, after soaking under 3746−3759. illumination, the intensities of the three CO peaks at ∼2000 (3) Hammarström, L. Accumulative Charge Separation for Solar −1 fi Fuels Production: Coupling Light-Induced Single Electron Transfer to cm decreased signi cantly. Also, the feature around 1400 − cm−1 (associated only with [1], cf. Figure 2B) loses intensity Multielectron Catalysis. Acc. Chem. Res. 2015, 48, 840 850. (4) Ashford, D. L.; Gish, M. K.; Vannucci, A. K.; Brennaman, M. K.; upon soaking under illumination. This suggests that the catalyst Templeton, J. L.; Papanikolas, J. M.; Meyer, T. J. Molecular is also desorbing from the NiO surface. In contrast, the C343 Chromophore−Catalyst Assemblies for Solar Fuel Applications. dye binding to the surface is not affected upon soaking and Chem. Rev. 2015, 115, 13006−13049. illumination as can been seen by the constant intensity of the (5) Yu, Z.; Li, F.; Sun, L. Recent Advances in Dye-Sensitized − C343 IR peaks at ∼1300 cm 1. These experiments strongly Photoelectrochemical Cells for Solar Hydrogen Production Based on suggest that the catalyst [1] on C343-sensitized NiO is unstable Molecular Components. Energy Environ. Sci. 2015, 8, 760−775. under irradiation. (6) Willkomm, J.; Orchard, K. L.; Reynal, A.; Pastor, E.; Durrant, J. To conclude, we could directly demonstrate for the first time R.; Reisner, E. Dye-Sensitised Semiconductors Modified with Molecular Catalysts for Light-Driven H2 Production. Chem. Soc. Rev. the critical electron-transfer processes between dye, catalyst, − and NiO in a molecularly sensitized photocathode that is active 2016, 45,9 23. (7) Gao, Y.; Ding, X.; Liu, J.; Wang, L.; Lu, Z.; Li, L.; Sun, L. Visible for photoelectrochemical proton reduction to H2.Hole Light Driven Water Splitting in a Molecular Device with injection into NiO followed by ultrafast surface electron Unprecedentedly High Photocurrent Density. J. Am. Chem. Soc. transfer led to rapid and efficient reduction of the catalyst − − 2013, 135, 4219 4222. [1]. Charge recombination of [1] was slow, on the time scale (8) Ashford, D. L.; Lapides, A. M.; Vannucci, A. K.; Hanson, K.; of 2 μs−20 ms, which may allow protonation and a second Torelli, D. A.; Harrison, D. P.; Templeton, J. L.; Meyer, T. J. Water reduction step of the catalyst to occur. Thus, a device with Oxidation by an Electropolymerized Catalyst on Derivatized C343:[1]/NiO as photocathode produced H with a Faradaic Mesoporous Metal Oxide Electrodes. J. Am. Chem. Soc. 2014, 136, 2 − efficiency of ∼50%. FTIR and GC experiments demonstrated 6578 6581. that electrode degradation was mainly due to catalyst (9) Norris, M. R.; Concepcion, J. J.; Fang, Z.; Templeton, J. L.; degradation. The mechanistic studies of sensitized photo- Meyer, T. J. Low-Overpotential Water Oxidation by a Surface-Bound Ruthenium-Chromophore−Ruthenium-Catalyst Assembly. Angew. cathodes provide understanding for the processes and bottle- Chem., Int. Ed. 2013, 52, 13580−13583. necks involved in this kind of system and should guide further (10) Li, L.; Duan, L.; Wen, F.; Li, C.; Wang, M.; Hagfeldt, A.; Sun, L. work to improve their performance. Visible Light Driven Hydrogen Production from a Photo-Active Cathode Based on a Molecular Catalyst and Organic Dye-Sensitized p- ■ ASSOCIATED CONTENT Type Nanostructured NiO. Chem. Commun. 2012, 48, 988−990. * (11) Ji, Z.; He, M.; Huang, Z.; Ozkan, U.; Wu, Y. Photostable p-Type S Supporting Information Dye-Sensitized Photoelectrochemical Cells for Water Reduction. J. The Supporting Information is available free of charge on the Am. Chem. Soc. 2013, 135, 11696−11699. ACS Publications website at DOI: 10.1021/acsenergy- (12) Li, F.; Fan, K.; Xu, B.; Gabrielsson, E.; Daniel, Q.; Li, L.; Sun, L. lett.6b00506. Organic Dye-Sensitized Tandem Photoelectrochemical Cell for Light Driven Total Water Splitting. J. Am. Chem. Soc. 2015, 137, 9153−9159. Experimental details, synthesis of [1],additional (13) Fan, K.; Li, F.; Wang, L.; Daniel, Q.; Gabrielsson, E.; Sun, L. Pt- experimental data, and analysis (PDF) Free Tandem Molecular Photoelectrochemical Cells for Water

1110 DOI: 10.1021/acsenergylett.6b00506 ACS Energy Lett. 2016, 1, 1106−1111 ACS Energy Letters Letter

Splitting Driven by Visible Light. Phys. Chem. Chem. Phys. 2014, 16, (31) Mirmohades, M.; Pullen, S.; Stein, M.; Maji, S.; Ott, S.; 25234−25240. Hammarström, L.; Lomoth, R. Direct Observation of Key Catalytic (14) Click, K. A.; Beauchamp, D. R.; Huang, Z.; Chen, W.; Wu, Y. Intermediates in a Photoinduced Proton Reduction Cycle with a Membrane-Inspired Acidically Stable Dye-Sensitized Photocathode for Diiron Carbonyl Complex. J. Am. Chem. Soc. 2014, 136, 17366− Solar Fuel Production. J. Am. Chem. Soc. 2016, 138, 1174−1179. 17369. (15) Kamire, R. J.; Majewski, M. B.; Hoffeditz, W. L.; Phelan, B. T.; (32) Hu, K.; Robson, K. C. D.; Beauvilliers, E. E.; Schott, E.; Zarate, Farha, O. K.; Hupp, J. T.; Wasielewski, M. R. Photodriven Hydrogen X.; Arratia-Perez, R.; Berlinguette, C. P.; Meyer, G. J. Intramolecular Evolution by Molecular Catalysts Using Al2O3-Protected Perylene-3,4- and Lateral Intermolecular Hole Transfer at the Sensitized TiO2 Dicarboximide on NiO Electrodes. Chem. Sci. 2016, DOI: 10.1039/ Interface. J. Am. Chem. Soc. 2014, 136, 1034−1046. C6SC02477G. (33) Brennan, B. J.; Durrell, A. C.; Koepf, M.; Crabtree, R. H.; (16) Kaeffer, N.; Massin, J.; Lebrun, C.; Renault, O.; Chavarot- Brudvig, G. W. Towards Multielectron Photocatalysis: A Porphyrin Kerlidou, M.; Artero, V. Covalent Design for Dye-Sensitized H2- Array for Lateral Hole Transfer and Capture on a Metal Oxide Surface. Evolving Photocathodes Based on a Cobalt Diimine−Dioxime Phys. Chem. Chem. Phys. 2015, 17, 12728−12734. Catalyst. J. Am. Chem. Soc. 2016, 138, 12308−12311. (34) Ardo, S.; Meyer, G. J. Direct Observation of Photodriven (17) Sahara, G.; Abe, R.; Higashi, M.; Morikawa, T.; Maeda, K.; Intermolecular Hole Transfer across TiO2 Nanocrystallites: Lateral Ueda, Y.; Ishitani, O. Photoelectrochemical CO2 Reduction Using a Self-Exchange Reactions and Catalyst Oxidation. J. Am. Chem. Soc. Ru(II)-Re(I) Multinuclear Metal Complex on a p-Type Semi- 2010, 132, 9283−9285. conducting NiO Electrode. Chem. Commun. 2015, 51, 10722−10725. (18) Kou, Y.; et al. Visible Light-Induced Reduction of Carbon Dioxide Sensitized by a Porphyrin−Rhenium Dyad Metal Complex on p-Type Semiconducting NiO as the Reduction Terminal End of an Artificial Photosynthetic System. J. Catal. 2014, 310,57−66. (19) Sahara, G.; Kumagai, H.; Maeda, K.; Kaeffer, N.; Artero, V.; Higashi, M.; Abe, R.; Ishitani, O. Photoelectrochemical Reduction of CO2 Coupled to Water Oxidation Using a Photocathode with a Ru(II)−Re(I) Complex Photocatalyst and a CoOx/TaON Photo- anode. J. Am. Chem. Soc. 2016, 138, 14152−14158. (20) Castillo, C. E.; et al. Visible Light-Driven Electron Transfer from a Dye-Sensitized p-Type NiO Photocathode to a Molecular Catalyst in Solution: Toward NiO-Based Photoelectrochemical Devices for Solar Hydrogen Production. J. Phys. Chem. C 2015, 119, 5806−5818. (21) Gross, M. A.; Creissen, C. E.; Orchard, K. L.; Reisner, E. Photoelectrochemical Hydrogen Production in Water Using a Layer- by-Layer Assembly of a Ru Dye and Ni Catalyst on NiO. Chem. Sci. 2016, 7, 5537−5546. (22) Bachmeier, A.; Hall, S.; Ragsdale, S. W.; Armstrong, F. A. Selective Visible-Light-Driven CO2 Reduction on a p-Type Dye- Sensitized NiO Photocathode. J. Am. Chem. Soc. 2014, 136, 13518− 13521. (23) van den Bosch, B.; Rombouts, J. A.; Orru, R. V. A.; Reek, J. N. H.; Detz, R. J. Nickel-Based Dye-Sensitized Photocathode: Towards Proton Reduction Using a Molecular Nickel Catalyst and an Organic Dye. ChemCatChem 2016, 8, 1392−1398. (24) Odobel, F.; Pellegrin, Y.; Gibson, E. A.; Hagfeldt, A.; Smeigh, A. L.; Hammarström, L. Recent Advances and Future Directions to Optimize the Performances of p-Type Dye-Sensitized Solar Cells. Coord. Chem. Rev. 2012, 256, 2414−2423. (25) Gibson, E. A.; Smeigh, A. L.; Le Pleux, L.; Fortage, J.; Boschloo, G.; Blart, E.; Pellegrin, Y.; Odobel, F.; Hagfeldt, A.; Hammarström, L. A p-Type NiO-Based Dye-Sensitized Solar Cell with an Open-Circuit Voltage of 0.35 V. Angew. Chem., Int. Ed. 2009, 48, 4402−4405. (26) Daeneke, T.; et al. Dominating Energy Losses in NiO p-Type Dye-Sensitized Solar Cells. Adv. Energy Mater. 2015, 5, 1401387. (27) Gardner, J. M.; Beyler, M.; Karnahl, M.; Tschierlei, S.; Ott, S.; Hammarström, L. Light-Driven Electron Transfer between a Photo- sensitizer and a Proton-Reducing Catalyst Co-Adsorbed to NiO. J. Am. Chem. Soc. 2012, 134, 19322−19325. (28) Brown, A. M.; Antila, L. J.; Mirmohades, M.; Pullen, S.; Ott, S.; Hammarström, L. Ultrafast Electron Transfer between Dye and Catalyst on a Mesoporous NiO Surface. J. Am. Chem. Soc. 2016, 138, 8060−8063. (29) Hara, K.; Sato, T.; Katoh, R.; Furube, A.; Ohga, Y.; Shinpo, A.; Suga, S.; Sayama, K.; Sugihara, H.; Arakawa, H. Molecular Design of Coumarin Dyes for Efficient Dye-Sensitized Solar Cells. J. Phys. Chem. B 2003, 107, 597−606. (30) Morandeira, A.; Boschloo, G.; Hagfeldt, A.; Hammarström, L. Photoinduced Ultrafast Dynamics of Coumarin 343 Sensitized p- Type-Nanostructured NiO Films. J. Phys. Chem. B 2005, 109, 19403− 19410.

1111 DOI: 10.1021/acsenergylett.6b00506 ACS Energy Lett. 2016, 1, 1106−1111 Supporting information for

Dynamics and photochemical H2 evolution of dye/NiO

photocathodes with a biomimetic FeFe-catalyst.

Liisa J. Antila, Pedram Ghamgosar, Somnath Maji, Haining Tian, Sascha Ott, Leif

Hammarström*

Department of Chemistry – Ångström Laboratory, Uppsala University, Box 523, SE75120

Uppsala, Sweden.

Contents: p. S2-S3 Experimental details and synthesis of [1]. p. S3-S4: Calculation of the yield of electron transfer from C343•– to [1]. Figure S1: FTIR and NMR spectra of [1]. Figure S2-S3: fs-TA spectra and DAS of the C343/NiO reference. Figure S4-S5: fs-TA spectra and DAS of C343:[1]/NiO. Figure S6: comparison of the fs DAS for C343:[1]/NiO and C343/NiO. Figure S7: fs-TA spectra of the [1]/NiO reference. Figure S8: TA band maximum vs. time for C343:[1]/NiO. Figure S9: TA spectra at 200 ns for C343:[1]/NiO, C343/NiO and [1]/NiO. Figure S10: Photocurrent and accumulated charge during light irradiation of the DSSFDs. Figure S11: GC-MS data for the DSSFDs. Figure S12: FTIR spectra for C343:[1]/NiO before and after soaking in buffer, w/o light irradiation.

Experimental details

Materials. All solvents and commercially supplied chemicals were reagent grade and used as received without further purification. Triethylamine (TEA), (benzotriazol-1-yloxy) tripyrrolidinophosphonium hexafluorophosphate (PyBOP), 2-Aminoethylphosphonic acid and triirondodecacarbonyl were purchased from Aldrich. [FeFe](mcbdt)(CO)6 was synthesized according to our previously reported procedures.S1 Proton nuclear magnetic resonance spectra (1H NMR) were recorded on a Varian FT-NMR spectrometer (400 MHz for 1HNMR); IR spectra recorded on a Perkin Elmer Spectrum One (Figure S1).

Synthesis of [1] 100 mg [FeFe](mcbdt)(CO)6 (0.215 mmol) and 2-aminoethylphosphonic acid 27 mg (0.216 mmol) and 115 mg (0.220 mmol) (benzotriazol-1- yloxy)tripyrrolidinophosphonium hexafluorophosphate (PyBOP) were stirred in acetonitrile

(CH3CN) in a Schlenk flask in the presence of triethylamine (1 eqv.) for 5 hours. After that, the mixture was evaporated under reduced pressure and then extracted with dichloromethane

(CH2Cl2). The solvent was removed under reduced pressure and the crude product was purified by column chromatography on silica using a mixture CH2Cl2: CH3OH (3:2) to afford -1 orange red powder. Yield 57 mg (46%). νmax/cm (KBr) 2081, 2045, 2004, 1632; δH (400

MHz; CD3CN) 7.23-8.29 (3H, CHarom), 6.75 (1H, NH), 3.37 (4H, CH2). Photocathode preparation and steady-state spectroscopy The double layers NiO films (ca. 1 µm thickness) were prepared on substrates according to the procedure reported in literature.S2 For the photoelectrochemical (PEC) measurements, FTO glass was used as 2 substrate (NiO active area 1 cm ) while CaF2 windows were used in samples for transient absorption spectroscopy. Two of the films were immersed in saturated coumarin 343 (C343) ethanolic solutions overnight. The films were rinsed and their UV–vis absorption spectra recorded (Figure 1A). Subsequently, one of the C343-sensitized films and one non-sensitized film were immersed in a solution of [1] (1.7 mg of [1] in 2 ml of methanol) for two days followed by rinsing with methanol and recording of the UV–vis (Agilent 8453 UV-vis spectrophotometer) and FTIR

(for CaF2 samples, Bruker 66v/S FTIR spectrometer) or ATR-IR (for FTO samples, Nicolet Avatar 370 DTGS, Thermo Electron Corporation) absorption spectra. Before sensitization the NiO film shows a gray color which is related to the partial oxidation of the film surface but upon soaking in the dye bath, the surface is reduced decreasing the background absorption in the sensitized films. After correcting for the NiO background absorption, the optical densities of the C343:[1]/NiO sample and of the C343/NiO reference were ~0.7 at 440 nm pump wavelength for laser experiments and that of the [1]/NiO reference was ~0.2. Femtosecond transient absorption experiments For a detailed description of the femtosecond transient absorption setup, see Petersson et al.S3 and for a more thorough description of the fs TA experiment for cosensitized NiO films, see Brown et al.S4 Here, a 440 nm pump beam was used to excite the dry sensitized NiO films with excitation energy set to 110 nJ (2·1015 photons/cm2/pulse) with a neutral density filter. White-light continuum was used for probing. Polarization of the pump was set at the magic angle, 54.7°, relative to the

S2 probe. The transient absorption spectra at different times were recorded for the sample and the references by scanning the delay of the probe beam relative to the pump from -5 ps to 1.8 ns with the help of an optical delay line. The time scan was repeated 5 times for each film and the films were translated between the individual scans. Global analysis of the time-resolved data was performed using multiple exponential decay schemes in a least-squares fitting program (PyGSpec, http://butler.cc.tut.fi/~tkatchen/prog/pygspec.html). Nanosecond transient absorption spectroscopy Frequency tripled Nd:YAG laser (Continuum Surelight II, repetition rate: 10 Hz, pulse length: 10 ns) was used to pump an Optical Parametrical Oscillator (Continuum Surelite OPO Plus) to obtain the pulsed 440 nm pump light. The pump laser was coupled to a LP920 detection system (Edinburgh Instruments) equipped with a pulsed Xe arc lamp to provide the white light for probing, and a CCD camera (Andor Technology) and a photomultiplier tube for detection. For more details, see ref.S5. Excitation energy density was 2.6 mJ/cm2. Photoelectrochemical (PEC) experiments PEC experiments were carried out in a three- electrode system using acetate buffer (pH 4.5) as electrolyte. Pt foil (1 cm2) was used as counter electrode (CE), Ag/AgCl (3 M KCl) was employed as reference electrode (RE) and C343 or C343:[1] sensitized NiO photocathodes (1.5 cm2, 3.5 µm) was adopted as working electrode (WE). LED lamp (cool white, 5000K) was used as light source. The whole system was degased with high pure Ar (99.999%) for 30 min under dark before applying a bias potential -0.3 V vs. Ag/AgCl. With the bias potential, the whole system was also allowed to stabilize for 200 s under dark. Then the chopped light was given to evaluate the photo- response. Subsequently, fresh samples were also photoelectrolyzed under the same condition for a certain time. The hydrogen production was monitored by Unisense microsensor. The CO analysis was carried out with HPR-20 benchtop gas analysis system (HIDEN Analytical) using Ar as carrier.

Calculation of the yield of electron transfer from C343•– to [1].

At 600 nm, the excited dye gives little transient changes while the reduced dye can be monitored as it is formed and then decays. At 470 nm instead, the signal after 100 ps is exclusively from the reduced catalyst [1]-. The following extinction coefficients were used to estimate the electron transfer yield, where we assume that the values for [1] and [2] are very •– -1 -1 S6 -1 -1 - similar: ε600(C343 ) = 4200 M cm ; ε600([2]) = 1000 M cm and ε470([2] - [2]) = 4000 M-1cm-1.S4

In C343:[1]/NiO the maximum signal at 600 nm is 0.8 mOD (around 200 fs) while the 470 nm signal at 100 ps is 1.0 mOD (data in Figure 4). This would suggest an (unphysical) electron transfer yield of Φ130%:

, , . Φ 1.30 (S1) , , .

However, this result is probably affected by the overlapping and heterogeneous kinetics of formation and subsequent reaction of C343•–, where rapid electron transfer to [1] reduces the concentration of C343•– at the maximum and leads to an underestimate of the total amount of C343•– formed. Indeed, in the C343/NiO reference sample, the maximum 600 nm signal is almost twice that of C343:[1]/NiO (a factor 13/7), when normalized to the same initial ground state bleach at 445 nm. Thus, we assume that the formation of C343•– is equally efficient in the two samples, and multiply the yield in equation S1 by the factor 7/13, to obtain an electron transfer yield of Φ70%:

Φ 1.30 0.70 (S2)

While this estimate based in TA amplitudes includes significant errors, it is in good agreement with what is expected from the relative kinetics of C343•– decay in the two samples C343/NiO and C343:[1]/NiO.

Additional data

Characterization of [1]:

IR Spectra

Fe[2]MCBDT 2 SM[1] 109

2200 2100 2000 1900 1800 1700 1600 1500 Wavenumber (cm-1)

1HNMR Spectrum

1 Figure S1. (top): IR spectra of [2] and [1] in KBr, and (bottom) H NMR spectrum of [1] in CD3CN.

Femtosecond transient absorption experiments

Figure S2. Experimental TA spectra of C343/NiO after excitation at 440 nm (110 nJ, absorbed photon density is 1.7 ·1015 photons/cm2/pulse) A) and B) 100 fs – 1 ps after excitation C) 1 ps – 1.8 ns after excitation.

Figure S3. A) Decay associated spectra (DAS) of C343/NiO in the range of 360-660 nm obtained by global analysis of the decay kinetics. B) Zoom in of the DAS.

Figure S4. TA spectra of C343:[1]/NiO after excitation at 440 nm (110 nJ, absorbed photon density 1.7 ·1015 photons/cm2/pulse) A) and B) 100 fs – 1 ps after excitation, C) and D) 1 ps – 1.8 ns after excitation. The spectra have been corrected for background absorption.

Figure S5. Decay associated spectra (DAS) of C343:[1]/ NiO after excitation at 440 nm obtained by a global fit of 4 exponential decays of the experimental data in Figure 3 of the main paper. Inset: Zoom in to the DAS associated with the long time constants.

The decay associated spectra (DAS) obtained by global analysis of the experimental TA data of C343:[1]/NiO (Figure S4) are shown in Figure S5 and aid analysis of the processes. The DAS corresponding to the shorter time constants of <150 fs and 700 fs are rather similar in shape compared to corresponding <150 fs and 1 ps DAS in the C343/NiO reference, meaning that hole injection and early C343•–recombination dominate on these time scales (see Figure S3 for C343/NiO DAS and Figure S6 for comparison of the two sets of DAS). For the 9 ps DAS, there is some difference, however: in the 410-480 nm region the amplitude in the DAS of C343:[1]/NiO is more negative (rising signal) than in the corresponding region in the C343/NiO DAS. This reflects the formation of a new species in this spectral region in the sample with both the dye and the catalyst. This is confirmed by the DAS corresponding to the > 2 ns time constant (blue line in Figure S6) which shows a new band at 410-500 nm while the corresponding 1.3 ns C343/NiO DAS is flat in this region (blue dashed line in Figure S6). This band is very similar in shape to the flash-quench generated spectrum of singly reduced S5 - [Fe2(bdt)(CO)6] and therefore we assign the band to the reduced species [1] that persists beyond the time scale of this experiment ( > 2 ns).

Figure S6. DAS of the C343:[1]/NiO sample (solid lines), and C343/NiO (dashed lines) and [1]/NiO (dotted line) references. The C343/SM109/NiO spectra at 800 fs and 9 ps and C343/NiO spectra at 1 ps and 9 ps have been normalized at their minima (maximum bleach) to facilitate comparison. The 150 fs spectra have been omitted for clarity.

S10

Figure S7. A) TA spectra of [1]/NiO after excitation at 440 nm (110 nJ, 2·1015 photons/(cm2 × pulse). The absorbed photon density at 440 nm is 9 ·1014 photons/cm2/pulse. B) Decay associated spectra.

The [1]/NiO reference shows a very weak, broad absorption band that extends over the spectral window of the measurement (Figure S7A). Considering that the excited state reduction potential of [1] (estimated as +1.2 V vs. NHE from the ground state reduction potential of -1.2 V vs. NHE and absorption onset of 2.3 eV) is more positive than the VB edge of NiO (+0.5 V vs. NHES7), it is possible that the excited [1] undergoes hole injection to NiO VB. However, the reduced state has a clear spectroscopic fingerprint (cf. Figure S6) that cannot be observed in the experimental spectrum nor in the DAS of the [1]/NiO reference (Figure S7B). Therefore, we tentatively assign the observed feature to the excited state of [1].

S11

Figure S8. Transient absorption band maximum position in the region 470-510 nm during 1- 50 ps after the excitation for C343:[1]/NiO.

Figure S9. Transient absorption at 200 ns after excitation at 440 nm in C343:[1]/NiO sample (black line), [1]/NiO reference (red line) and C343/ NiO reference (blue line). The kinetic trace showing decay of the signal at 470 nm is shown in Figure 4c of the main paper.

S12

Figure S10. Photocurrent and accumulated charge during light irradiation of the DSSFDs.

Figure S11. GC-MS curve of C343:[1]/NiO during photoelectrolysis.

S13

Figure S12. FTIR spectrum of a C343:[1]/NiO sample on CaF2 before (black line) and after soaking in buffer pH 4.5 for 1 hour in the dark (orange line) and 1 hour under illumination

(green line).

References

S1. Pullen, S.; Fei, H.; Orthaber, A.; Cohen, S. M.; Ott, S. J. Am. Chem. Soc. 2013, 135, 16997. S2. Li, L.; Gibson, E. A.; Qin, P.; Boschloo, G.; Gorlov, M.; Hagfeldt, A.; Sun, L. Adv. Mater. 2010, 22, 1759. S3. Petersson, J.; Eklund, M.; Davidsson, J.; Hammarström, L. J. Phys. Chem. B 2010, 114, 14329. S4. Brown, A. M.; Antila, L. J.; Mirmohades, M.; Pullen, S.; Ott, S.; Hammarström, L.; J. Am. Chem. Soc., 2016, 138, 8060. S5. Mirmohades, M.; Pullen, S.; Stein, M.; Maji, S.; Ott, S.; Hammarström, L.; Lomoth, R. J. Am. Chem. Soc. 2014, 136, 17366. S6. Nad, S.; Pal, H. The Journal of Physical Chemistry A 2002, 106, 6823. S7 D’Amario, L.; Antila, L. J.; Rimgard, B. P.; Boschloo, G.; Hammarström, L.; J. Phys. Chem. C, 2014, 118, 19556.

S14

Paper II

The Role of Heterojunctions on the Water Oxidation Performance in Mixed

-Fe2O3/Fe3O4 Hybrid Film Structures

Jennifer Leduca,ǂ, Yakup Gönüllüa,ǂ, Pedram Ghamgosarb, Shujie Youb, Johanne Mouzonb, Alberto Vomierob, Sanjay Mathura,*

a University of Cologne, Institute of Inorganic Chemistry, University of Cologne, 50939, Cologne, Germany. b Division of Materials Science, Department of Engineering Sciences and Mathematics, Luleå University of Technology, 97187, Luleå, Sweden. ǂ both authors contributed equally * Corresponding author. Email: [email protected]

Keywords: solar water splitting, valence dynamics, magnetite, Raman, single-source CVD, heterostructures ACS Appl. Mater. Interfaces

Abstract: Photoelectrochemical water splitting represents a sustainable method for the production of energy with its potential to turn sunlight directly into hydrogen, the most prominent non-fossil energy storage and transport medium. In this article, we report the preparation of mixed -Fe2O3/Fe3O4 films via a single-step chemical vapor deposition of t [Fe(O Bu)3]2 and their use in photoelectrochemical water splitting. The film thickness as well as phase composition and allocation can simply be controlled by varying the deposition time and was investigated in detail using (cross-section) Raman spectroscopy and Scanning Electron Microscopy. The highest water oxidation activity was found for hybrid films with a thickness of 11 µm. This phenomenon is attributed to an improved electron transport resulting from the higher magnetite content towards the bottom of the layer and the increased light absorption of the hematite layer mainly located at the top surface of the film. These results suggest the synergistic interplay among film microstructure, electrochemical catalysts and

1 heterojunctions that are important parameters to configure superior and highly efficient anode materials for photoelectrochemical water splitting applications.

1. Introduction

Hematite (-Fe2O3) has been intensively studied as photoanode material for photoelectrochemical water splitting applications due to its high photochemical stability, non- toxicity, abundance, suitable band gap (~2.1 eV) and high theoretical maximum solar-to hydrogen (STH) conversion efficiency (15%) which makes it a promising candidate for cost- effective PEC hydrogen production. However, due to critical limitations (low carrier mobility, short carrier lifetime, high density of surface states, slow kinetics for oxygen evolution) its experimentally observed maximum STH efficiency is still far away from the theoretical value. Therefore, several attempts were undertaken, among those nanostructuring, doping, formation of composite materials and usage of co-catalysts, to circumvent these limitations. Nanostructured hematite electrodes in the form of nanoparticles1, nanowires2, nanotubes3, nanoflowers4 or nanocones5 have been generated to overcome the drawback of a small hole diffusion length (2-4 nm) as well as a short excited-state lifetime (3-10 ps). Doping represents a well-established method for the enhancement of PEC performance as it can be used to increase the carrier concentration and to fine-tune the band gap width and band positions, 6 Thus, the incorporation of various dopants (e.g. Ti7, Zr8, Si9, Sn10, Ni11, Ta12 or Cr13) into the hematite lattice was investigated. The effect of Sn doping of hematite films deposited on

F:SnO2 substrates (FTO) has been examined by Annamalai who found that Sn ions can diffuse from the FTO substrate to the upper hematite layer and change the morphology, physical properties and PEC performance of the material. The resulting enhancement in PEC performance was explained by an increase in conductivity due to the reduction of Fe3+ to Fe2+ triggered by the neighboring electron donor dopant Sn. In our recent work, we have examined the effect of directed post-deposition plasma-chemical reduction of hematite.14 Via variation of the temperature used during hydrogen plasma treatment, different proportions of Fe3O4 +2 +3 (mixed valent Fe /Fe ) in the resulting Fe3O4/-Fe2O3 nanomaterials could be achieved. For temperatures lower than 300 °C and higher than 500 °C, either pristine -Fe2O3 or Fe3O4 could be selectively generated. Notably for this work, even though Fe3O4 had been reported to 15 be photo-inactive , the mixed Fe3O4:-Fe2O3 photoanodes showed a better PEC performance (photocurrent density of 3.5 mA/cm² at 1.8 V vs. RHE and onset potential of 1.28 V vs. RHE) than the respective pristine hematite samples. This enhancement in PEC response can be ascribed to a reduced band gap energy and higher carrier density resulting from the generation of additional charge carriers upon reduction of Fe3+.15-17

In this work, we report on a simple method to significantly improve the water oxidation performance of -Fe2O3 photoanodes by the directed growth of -Fe2O3/Fe3O4 hybrid film t structures via thermal CVD of [Fe(O Bu)3]2. The film thickness as well as the local distribution of magnetite in the composite material increase the charge carrier generation and collection without decreasing the carrier transport and can simply be controlled by varying the deposition time. The interplay of phase composition, allocation and film thickness was investigated in detail using (cross-section) Raman spectroscopy and correlated with the water oxidation efficiency.

2. Experimental Part

2.1 Material Synthesis

t The gas phase deposition of Fe2O3 layers was performed using [Fe(O Bu)3]2 as precursor in a horizontal cold-wall CVD reactor.18,19 The precursor temperature was kept constant at 100 °C to ensure a homogenous gas flow. The precursor flux was guided to an inductively heated (500 °C) substrate using low pressure (~ 3x10-6 mbar). The deposition time was varied between 5-45 min in order to scan for morphological and crystallographic changes. Iron oxide layers were deposited onto conductive F:SnO2 (FTO) substrates (Sigma Aldrich, ~8 Ω/sq) for photoelectrochemical measurements and silicon substrates for XRD analysis. The substrates were cleaned before deposition by ultrasonication in water and isopropanol. A post deposition heat treatment was conducted at 500 °C for 5 h in air.

2.2 Materials Characterization

To investigate the microstructural properties of the film, a Nikon Eclipse MA200 was used. This light optical microscope is equipped with Nikon's DS-U2 Camera Control Unit and NIS- Elements tracer. The film morphology was analyzed by field emission-scanning electron microscopy (FE-SEM) using a SUPRA 40VP instrument (Zeiss) operated at 10.0 kV, while cross-sections of the film was investigated using a Magellan 400 (the FEI company) at 3kV. AFM topography measurements were performed using a Park Systems XE-100 in true non- contact mode with a cantilever (PPP-NCHR-20, Nanosensors). Energy dispersive X-ray spectroscopy (EDS) was carried out in parallel to SEM observations. The powder X-ray diffraction (XRD, STOE-STADI MP) patterns were measured in reflection mode using Cu Kα (λ = 0.15406 nm) radiation. X-ray photoelectron spectroscopy (XPS) measurements were carried out using a spectrometer (Ф 5600ci, Perkin Elmer) at a pressure below 10-8 mbar under a non-monochromatic Al Kα excitation source (186.6 eV) with an ESCA M-Probe 4 spectrometer (Surface Science Instruments SSI) and a pressure under 10-8 mbar. Correction to the binding energies was done in reference to the C1s-signal (284.8 eV). Spectral corrections and composition calculations were done with CasaXPS. Raman spectra were recorded using a Senterra Raman microscope with 1200 groove/mm grating under 532 nm and/or 785 nm laser excitation. Electrochemical measurements were performed in a flat three-electrode electrochemical cell using a saturated calomel electrode (SCE) as the reference, a Pt wire as the counter electrode and 1 M NaOH as the electrolyte (pH = 13.6). Linear sweep voltammetry (10 mV/s) was carried out in a potential range from -1 to 1 V vs. SCE using a potentiostat (PAR, Versa state IV) in the dark and under simulated solar illumination (Xe lamp, 150 W, Oriel with a Schott KG-3 filter). Potentials with respect to the reversible hydrogen electrode (RHE) scale were calculated using the Nernst equation (1).

(1)

3. Results and Discussion

Using a precursor charge of 300 mg, a non-linear film growth from 1.5 m (5 min) to 11 µm (45 min) was observed (figure 1a). The 5 to 20 min depositions exhibit transparent dark red to black colors whereas the 30 and 45 min samples appear opaque and grey. Figures 1b-e display optical and top view SEM images of 5 and 45 min depositions on FTO. The surfaces of both samples are rough as determined by AFM analysis (Ra(5min)=20 nm;

Ra(45min)=168 nm). 45 min growth time resulted in the formation of micrometer sized large grains, whereas for shorter deposition times the grain sizes came out significantly smaller and the surfaces were much smoother.

(a) (b) (c)

12 10 m 10 m

m) 8  50 m 50 m

(d) (e) Thickness ( Thickness 4

0 0 10 20 30 40 50 5 m Deposition time (min) 5 m

Figure 1: (a) Film thickness as a function of CVD process time. Optical (b, c) and SEM (d, e) images of the top surface of films grown for (b, d) 5 min and (c, e) 45 min, unveiling the different grain size distribution in the two samples. Insets in (b, c) represent higher resolution optical images of typical large grains. 5

To evaluate the surface and bulk structure of the films, SEM was carried out on freshly cleaved cross sections of both the 5 and 45 min depositions (figure 2). The bulk layers were composed of densely intergrown grains and exhibited a microstructure typical for competitive growth. The size of the largest grains on the top of the film was <500 nm and <8 m in the 5 min (figure 2a) and 45 min (figure 2d) films, respectively. The top surfaces of both samples (figures 2b,e) were mostly covered by an irregular layer consisting of spindle-like grains which are characteristic for CVD prepared hematite layers.18,20 However, smaller areas were uncovered and exhibited fine equiaxed grains in the range of ~50 nm, as indicated by the white arrows in figures 2b,e. These features are characteristic for hematite obtained from the oxidation of magnetite films and were more pronounced with increasing deposition time.20 The top irregular hematite layer also increased with deposition time, i.e. from 100 to 350 nm for the 5 min (figure 2c) and 45 min (figure 2f) films, respectively.

Figure 2: Cross section and top view SEM micrographs of CVD prepared iron oxide films deposited on FTO for (a-c) 5 min and (d-f) 45 min. Images (c) and (f) display the morphology of the top layer (100-350 nm) for both deposition times.

To investigate the surface composition of the mixed iron oxides, XPS measurements were performed (figure 3). Both, the 5 min and the 45 min samples showed a remaining carbon content of ~6-8% in the generated iron oxide films. This is either due to the incorporation of precursor based side products (equation 2) into the iron oxide film or to an incomplete decomposition. Similar concentrations for C contaminations were detected via EDS analysis (Supporting information, figure S1, table S1,S2).

[Fe(OtBu) ] ‘Fe-O’ + x iso-C H + y tBuOH + z H O (2) 3 2 4 8 2

The binding energies of Fe 2p3/2 (710.6 eV) and Fe 2p1/2 (724.0 eV) signals in the samples for

5 min and 45 min depositions were consistent with values reported for hematite (Fe 2p3/2 ~ 21 710.6-711.6 eV). Moreover, the Fe 2p3/2 peak of hematite exhibits a satellite signal at ~717 eV, which was also detected in the 5 min sample (figure 3a). In the 45 min layer, this satellite signal was less pronounced (around 717 eV), due to Fe2+ satellites, which are characteristic for magnetite (figure 3b). Although the presence of overlapping multiplets made it difficult to separate the two phases, the intensities and less pronounced shoulders, which appear around 730 eV and 717 eV proved the formation of Fe3O4 in the 45 min deposition.20 These results indicate the presence of hematite and a mixture of hematite and magnetite on the surfaces of the 5 and 45 min depositions, respectively.

a) b)

Figure 3: XPS spectra of iron oxide films after (a) 5 and (b) 45 min depositions.

The phase compositions in the bulk of the depositions were examined using XRD. The patterns (figure 4) of all samples showed reflections (2θ=33.1°, 35.7°, 54.1° and 57.6°) consistent with a hematite phase. The patterns of the 20, 30 and 45 min depositions displayed an additional reflection at 30° and 43.1°, which can be attributed to Fe3O4 (magnetite) as confirmed by Rietveld refinement (figure 4b). As described in our previous work, the as- obtained CVD films deposited at 500 °C are composed of the magnetite phase.18 Post- annealing in air resulted in oxidation of the magnetite phase to -Fe2O3 (hematite) with a favored growth of crystallites along the (110) direction, which was reported to be the most active hematite surface for photoelectrochemical water oxidation. However, with longer deposition time (>20 min) and resulting increased thickness of the film (>5.6 m), oxidation of the magnetite film via post-annealing in air (500 °C, 5 h) did not proceed to completion. As a result, mixed Fe3O4/-Fe2O3 hybrid films were formed with a higher magnetite content

7 towards the bottom of the layer. Indeed, even after annealing for 48 h in air the magnetite phase did not further oxidize pointing towards the stability of the hybrid film structure.

a) b)

Figure 3: (a) XRD patterns of iron oxide films prepared using different deposition times (5, 10, 20, 30 and 45 min) on silicon substrate and (b) Rietveld refinement of the 45 min sample using a magnetite fitting.

In order to investigate the local distribution of magnetite throughout the film, we have performed (cross section) Raman spectroscopy using different laser excitations (532 and 785 nm) and thus different penetration depths. Whereas the 532 nm laser is more surface sensitive (figure 5b), the 785 nm laser penetrates deeper and is therefore able to detect the bulk composition (figure 5a). Using both 532 and 785 nm, only hematite and silicon signals (~520 cm-1) were detected at the surface and in the bulk of the 5 and 10 min samples. In contrast, the 20 and 30 min depositions exhibited hematite at the surface and both hematite and magnetite in the bulk.

(a) Fe O / Si,  = 785 nm (b) Fe O / Si,  = 532 nm x y ex x y ex +: hematite 45min + + O: magnetite 30min O 20min *: Si + + 10min + 5min + + + +

+ +

Intensity Intensity

200 400 600 800 1000 1200 1400 200 400 600 800 1000 1200 1400 Raman shift (cm-1) Raman shift (cm-1)

Figure 5: Raman spectra of CVD prepared iron oxide films using (a) 785 nm and (b) 532 nm laser excitations for 5-45 min depositions.

Due to the thickness of the 45 min deposition (11 m) we were able to perform cross section Raman analysis (figure 6a,b). This method permits to investigate the phase distribution from the substrate to the film surface without using different laser wavelengths. Both hematite and magnetite were detected throughout the 45 min deposition with both laser excitations. The line scans in figure 6a,b showed that magnetite was the dominant phase at the bottom of the film, whereas a higher amount of hematite was detected at the surface. In the bulk, a dispersed heterojunction is formed with different fractions of magnetite and hematite.

(a) 45min,  = 785 nm (b) 45min,  = 532 nm ex ex

+ + + + *: Si + o: magnetite + +: hematite + + + + o + *: Si +

o: magnetite +: hematite + +o Intensity Intensity Intensity o + o o + o + o o o o x2 x2 * * 200 400 600 800 1000 1200 1400 200 400 600 800 1000 1200 1400 Raman shift (cm-1) Raman shift (cm-1)

Figure 6: Raman spectra of CVD prepared iron oxide films using a) 785 nm and b) 532 nm laser excitations for the 45 min deposition. The spectra were recorded on the cross section (inset in (a)) of a freshly fractured sample on a silicon substrate; the red arrow in the inset of (a) and in the main panels (a) and (b) from the bottom to the top refers to the line scan direction from the substrate to the surface of the sample.

To evaluate the water splitting performance of these FeOx layers with regard to their phase composition, we recorded current–voltage curves in 1 M NaOH solution using intermittent AM 1.5 G (100 mW/cm2) standard illumination (figure 7a). No significant change in the onset potential of the samples (figure 7b, blue line) was observed, except for a slight cathodic shift in case of the 10 and 30 min depositions, which was ascribed to a reduced amount of hematite surface states. Recently, a series of studies pointed out a strong correlation between surface states and the photocurrent onset potential.22 Shinde et al. observed either an anodic or a cathodic shift depending on the annealing conditions, when they replaced Fe3+ with Sn4+ at 23 the surface of Fe2O3 photoanodes. The 5 min deposition exhibited a photocurrent density of ~0.28 mA/cm² (at 1.23 V vs. RHE), which is the highest reported value for a pristine hematite layer with a thickness of ~1.5 µm so far (table 1). An increase in thickness without changing the chemical composition (10 min, ~4 µm) resulted in a lower PEC activity, due to the inhibited charge carrier transport over a longer distance to the anode surface. The opposite trend was observed when the as deposited magnetite layers were only partly oxidized to give 9

15 -Fe2O3/Fe3O4 hybrid film structures. Although magnetite is considered as photoinactive, an increase in deposition time and thus thickness and magnetite content resulted in an enhanced photocurrent density of 0.48 mA/cm2 (at 1.23 V vs. RHE) for the 45 min sample. This improvement can be attributed to two effects, an increase in electrical conductivity and a reduction of defects. The hybrid structure facilitates the charge collection upon light absorption and the charge separation due to the distribution of the phases throughout the film. Lower concentrations of magnetite are present at the surface, which is beneficial for the polaron hopping conduction mechanism24 of the mixed valence Fe3+/Fe2+ couple and the light absorption of the photoactive hematite layer. The higher magnetite content towards the bottom of the layer facilitates the electron transport to the counter electrode by collecting from the top α-Fe2O3. In addition, the very dense layer of intergrown grains reduces the amount of defects resulting in a decreased recombination rate.17 In order to corroborate this hypothesis, we have performed post-deposition oxygen and hydrogen plasma treatments to test the PEC performances of phase pure α-Fe2O3 and Fe3O4 layers with film thicknesses of 1.5 µm and 11 µm. Using oxygen plasma, the 45 min sample was completely oxidized to give pure -Fe2O3 as confirmed by the absence of the characteristic magnetite reflexes at 30° and 43.1° in the XRD pattern (Supporting information, figure S4). As expected, the photocurrent density of the 11 µm -Fe2O3 layer decreased to a value of 0.06 mA/cm² (at 1.23 V vs. RHE) resulting from the higher charge carrier recombination rate (Supporting information, figure S5). Even though the phase composition of the 5 min sample was not influenced by the oxygen plasma, the PEC performance also decreased which was ascribed to the reduced crystallinity of the layer and thus higher amount of defects (Supporting information, figure S6). Hydrogen plasma treatment of both the 5 and 45 min samples resulted in the formation of magnetite at the film surfaces. However, dark measurements in alkaline solution (Supporting information, figure S7) showed that these photoanodes were not stable when a potential of higher than 0.3 V was applied. From these experiments it can be concluded that the formation of hybrid -Fe2O3/Fe3O4 film structures as well as the phase distribution throughout the layer are of importance to give reasonable PEC performances for hematite photoanodes with a film thickness of higher than 1.5 µm.

10 a) b)

Figure 7: (a) Photocurrent density-voltage curves of FeOx layers and (b) photocurrent density and onset

potential of the FeOx layers at 1.23 V vs RHE depending on the deposition time.

4. Conclusions

t To summarize, mixed -Fe2O3/Fe3O4 heterostructures generated via CVD of [Fe(O Bu)3]2 and subsequent thermal oxidation showed improved photoactivity for water oxidation. The 2 -Fe2O3/Fe3O4 anode yielded a photocurrent density of 0.48 mA/cm at 1.23 V (vs RHE) despite a film thickness of 11 µm. The enhancement in photoactivity is attributed to the improved electron transport resulting from the higher magnetite content towards the bottom of the layer and the increased light absorption of the hematite layer mainly located at the top of the film. To the best of knowledge, this is the first report on efficient hematite-based water oxidation catalysts with a layer thickness of higher than 10 µm.

Table 1: Thickness evaluation of iron oxide photoanodes for photoelectrochemical water splitting.

Type of structure Thickness (µm) Synthesis methods J at 1.23 V vs RHE (mA/cm2) Ref

Nanoparticles 0.2 Anodization 0.05 25 Dendrites 30 Electrodeposition 0.018 26 Mesoporous 0.9 Colloidal Synthesis 0.12 27 Bulk 0.4-0.5 Electrodeposition No photoactivity 28 Mesoporous single crystal 3 Hydrothermal 0.61 29 Solid 3 Hydrothermal 0.01 29 Bulk 13 CVD 0.48 Present work

Acknowledgements Authors are thankful to the University of Cologne for providing the infrastructural support. J.L. is thankful to Fonds der chemischen Industrie for a PhD fellowship. The financial support in the framework of the DFG priority program (SPP 1613; “Fuels Produced Regeneratively Through Light-Driven Water Splitting: Clarification of the Elemental Processes Involved and Prospects for Implementation in Technological Concepts”) and the Framework Program of the European Commission (FP7) that funded the European Project SOLAROGENIX (www.solarogenix.eu) is gratefully acknowledged. A.V. acknowledges Knut & Alice Wallenberg Foundation, the Swedish Foundations Consolidator Fellowship, LTU Labfund program and Kempe Foundation for partial funding. A. V. acknowledges the European Commission through grant agreement GA299490 (F-LIGHT) for financial support. S.J. acknowledges Kempe Foundation for a postdoctoral fellowship.

References

1 Emin, S. et al. Photoelectrochemical water splitting with porous alpha-Fe2O3 thin films prepared from Fe/Fe-oxide nanoparticles. Appl Catal a-Gen 523, 130-138, doi:10.1016/j.apcata.2016.06.007 (2016). 2 Li, L. S. et al. Facile Solution Synthesis of alpha-FeF3 center dot 3H(2)O Nanowires and Their Conversion to alpha-Fe2O3 Nanowires for Photoelectrochemical Application. Nano Lett 12, 724-731, doi:10.1021/nl2036854 (2012). 3 Cho, S., Jang, J. W., Lee, K. H. & Lee, J. S. Research Update: Strategies for efficient photoelectrochemical water splitting using metal oxide photoanodes. Apl Mater 2, doi:Artn 010703 10.1063/1.4861798 (2014). 4 Zhong, D. K., Cornuz, M., Sivula, K., Graetzel, M. & Gamelin, D. R. Photo-assisted electrodeposition of cobalt- phosphate (Co-Pi) catalyst on hematite photoanodes for solar water oxidation. Energ Environ Sci 4, 1759-1764, doi:10.1039/c1ee01034d (2011). 5 Qiu, Y. C. et al. Efficient Photoelectrochemical Water Splitting with Ultrathin films of Hematite on Three- Dimensional Nanophotonic Structures. Nano Lett 14, 2123-2129, doi:10.1021/nl500359e (2014). 6 Sivula, K. et al. J. Am. Chem. Soc. 132, 7436 (2010). 7 Fu, Z. W. et al. Surface treatment with Al3+ on a Ti-doped alpha-Fe2O3 nanorod array photoanode for efficient photoelectrochemical water splitting. J Mater Chem A 2, 13705-13712 (2014). 8 Kumar, P., Sharma, P., Shrivastav, R., Doss, S. & Satsangi, V. R. Electrodeposited zirconium-doped alpha-Fe2O3 thin film for photoelectrochemical water splitting. International Journal of Hydrogen Energy 36, 2777-2784, doi:10.1016/j.ijhydene.2010.11.107 (2011). 9 Dias, P., Lopes, T., Andrade, L. & Mendes, A. Temperature effect on water splitting using a Si-doped hematite photoanode. J Power Sources 272, 567-580, doi:10.1016/j.jpowsour.2014.08.108 (2014). 10 Annamalai, A. et al. Sn/Be Sequentially co-doped Hematite Photoanodes for Enhanced Photoelectrochemical Water Oxidation: Effect of Be2+ as co-dopant. Scientific Reports 6, doi:ARTN 23183 10.1038/srep23183 (2016). 11 Liu, Y., Yu, Y. X. & Zhang, W. D. Photoelectrochemical properties of Ni-doped Fe2O3 thin films prepared by electrodeposition. Electrochimica Acta 59, 121-127, doi:10.1016/j.electacta.2011.10.051 (2012). 12 Fu, Y. et al. Solution growth of Ta-doped hematite nanorods for efficient photoelectrochemical water splitting: a tradeoff between electronic structure and nanostructure evolution. Physical Chemistry Chemical Physics 18, 3846- 3853, doi:10.1039/C5CP07479G (2016). 13 Kleiman-Shwarsctein, A., Hu, Y. S., Forman, A. J., Stucky, G. D. & McFarland, E. W. Electrodeposition of alpha- Fe2O3 doped with Mo or Cr as photoanodes for photocatalytic water splitting. J Phys Chem C 112, 15900-15907, doi:10.1021/jp803775j (2008). 14 Mettenborger, A. et al. Plasma-chemical reduction of iron oxide photoanodes for efficient solar hydrogen production. International Journal of Hydrogen Energy 39, 4828-4835, doi:10.1016/j.ijhydene.2014.01.080 (2014). 15 Ling, Y. C. et al. The Influence of Oxygen Content on the Thermal Activation of Hematite Nanowires. Angew Chem Int Edit 51, 4074-4079, doi:10.1002/anie.201107467 (2012). 16 Chou, J. C., Lin, S. A., Lee, C. Y. & Gan, J. Y. Effect of bulk doping and surface-trapped states on water splitting with hematite photoanodes. J Mater Chem A 1, 5908-5914, doi:10.1039/c3ta00087g (2013). 17 Grigorescu, S. et al. Thermal air oxidation of Fe: rapid hematite nanowire growth and photoelectrochemical water splitting performance. Electrochem Commun 23, 59-62, doi:10.1016/j.elecom.2012.06.038 (2012). 12

18 Mathur, S. et al. Phase-Selective Deposition and Microstructure Control in Iron Oxide Films Obtained by Single- Source CVD. Chem. Vap. Deposition 8, 277-283, doi:10.1002/1521-3862(20021203)8:6<277::AID- CVDE277>3.0.CO;2-8 (2002). 19 Mathur, S., Veith, M., Sivakov, V., Shen, H. & Gao, H. B. Composition, morphology and particle size control in nanocrystalline iron oxide films grown by single-source CVD. Le Journal de Physique IV 11, Pr3-487-Pr483-494, doi:10.1051/jp4:2001362 (2001). 20 Mathur, S. et al. Phase selective deposition and microstructure control in iron oxide films obtained by single-source CVD. Chem Vapor Depos 8, 277-283, doi:Doi 10.1002/1521-3862(20021203)8:6<277::Aid-Cvde277>3.0.Co;2-8 (2002). 21 Li, X. Y., Lin, E. H., Zhang, C. Y. & Li, S. B. Preparation of Gamma-Fe2o3-Fe2o3 Ultrafine Powders by Laser Vapor-Phase Reaction. J Mater Sci Lett 14, 1335-1337, doi:Doi 10.1007/Bf00270719 (1995). 22 Fu, Z. et al. Surface treatment with Al3+on a Ti-doped α-Fe2O3nanorod array photoanode for efficient photoelectrochemical water splitting. Journal of Materials Chemistry A 2, 13705, doi:10.1039/c4ta02527j (2014). 23 Shinde, P. S., Choi, S. H., Kim, Y., Ryu, J. & Jang, J. S. Onset potential behavior in alpha-Fe2O3 photoanodes: the influence of surface and diffusion Sn doping on the surface states. Phys. Chem. Chem. Phys. 18, 2495-2509, doi:10.1039/c5cp06669g (2016). 24 Wang, J. et al. A Facile Electrochemical Reduction Method for Improving Photocatalytic Performance of alpha- Fe2O3 Photoanode for Solar Water Splitting. ACS Appl Mater Interfaces 9, 381-390, doi:10.1021/acsami.6b11057 (2017). 25 Fu, L. et al. Ethylene glycol adjusted nanorod hematite film for active photoelectrochemical water splitting. Phys. Chem. Chem. Phys. 16, 4284-4290, doi:10.1039/c3cp54240h (2014). 26 Zheng, J. Y. et al. Morphology evolution of dendritic Fe wire array by electrodeposition, and photoelectrochemical properties of α-Fe2O3 dendritic wire array. CrystEngComm 14, 6957, doi:10.1039/c2ce26046h (2012). 27 Goncalves, R. H., Lima, B. H. & Leite, E. R. Magnetite colloidal nanocrystals: a facile pathway to prepare mesoporous hematite thin films for photoelectrochemical water splitting. J. Am. Chem. Soc. 133, 6012-6019, doi:10.1021/ja111454f (2011). 28 Chou, J.-C., Lin, S.-A., Lee, C.-Y. & Gan, J.-Y. Effect of bulk doping and surface-trapped states on water splitting with hematite photoanodes. Journal of Materials Chemistry A 1, 5908, doi:10.1039/c3ta00087g (2013). 29 Wang, C. W. et al. Engineered Hematite Mesoporous Single Crystals Drive Drastic Enhancement in Solar Water Splitting. Nano Lett. 16, 427-433, doi:10.1021/acs.nanolett.5b04059 (2016).

The Role of Heterojunctions on the Water Oxidation Performance in Mixed

-Fe2O3/Fe3O4 Hybrid Film Structures

Jennifer Leduca,ǂ, Yakup Gönüllüa,ǂ, Pedram Ghamgosarb, Shujie Youb, Johanne Mouzonb, Alberto Vomierob, Sanjay Mathura,*

SUPPORTING INFORMATION

Figure S1: EDS spectra for sample 5 min (left) and 45 min (right) using a 10 kV electron beam.

Table S1: Weight and atomic composition of sample 5 min as obtained from EDS analysis using a 10 kV electron beam.

Element Line Wt% Atomic Type % C K series 2.4 5.9 O K series 28.5 55.5 Fe K series 69.1 38.6 Total: 100.0 100.0

Table S2: Weight and atomic composition of sample 45 min as obtained from EDS analysis using a 10 kV electron beam.

Element Line Wt% Atomic Type % C K series 2.4 6.3 O K series 26.3 52.8 Fe K series 71.3 40.9 Total: 100.0 100.0

FeO_45min on FTO 532nm 785nm +

+ + +: hematite

O: magnetite +

Intensity +

+ +

200 400 600 800 1000 1200 1400 Raman shift (cm-1)

Figure S2: Raman spectra of FeOx films after 45 minutes deposition under 785 nm and 532 nm laser excitation.

5 min 20 min

10 min 30 min

Figure S3: Top view SEM images of iron oxide films deposited on silicon for deposition times of a) 5 min, b) 10 min, c) 20 min, d) 30 min and e) 45 min.

Figure S4: XRD patterns of iron oxide films deposited on FTO with subsequent oxygen or hydrogen plasma treatment for deposition times of 5 and 45 min.

Figure S5: Photocurrent density-voltage curves of hydrogen and oxygen plasma treated and reference FeOx layers deposited for 45 min.

Figure S6: Photocurrent density-voltage curves of hydrogen and oxygen plasma treated and reference FeOx layers deposited for 5 min.

Figure S7: Photocurrent density-voltage curves of hydrogen and oxygen plasma treated FeOx layers deposited for 5 and 45 min. 17