Novel Concepts for Silicon Based Photovoltaics and Photoelectrochemistry

Lihao HAN 韩李豪

Photovoltaic Materials and Devices (PVMD) Laboratory Electrical Sustainable Energy (ESE) Department Electrical Engineering, Mathematics and Computer Science (EEMCS) Faculty Delft University of Technology, the Netherlands

Proefschrift

ter verkrijging van de graad van doctor aan de Technische Universiteit Delft, op gezag van de Rector Magnificus prof. ir. K.C.A.M. Luyben, voorzitter van het College van Promoties, in het openbaar te verdedigen op donderdag 15 januari 2015 om 10:00 uur

door

Lihao HAN

Master of Microelectronic Engineering, Tsinghua University geboren te Zhejiang, China Dit proefschrift is goedgekeurd door de promotor: Prof. Dr. M. Zeman

Copromotor: Dr. A.H.M. Smets

Samenstelling promotiecommissie:

Rector Magnificus, voorzitter Prof. Dr. M. Zeman, Technische Universiteit Delft, promotor Dr. A.H.M. Smets, Technische Universiteit Delft, copromotor Prof. Dr. B. Dam, Technische Universiteit Delft Prof. Dr. J.A. Ferreira, Technische Universiteit Delft Dr. F. Finger, Forschungszentrum Jülich GmbH, Duitsland Prof. Dr. T. Gregorkiewicz, Universiteit van Amsterdam Prof. Dr. M.C.M. van de Sanden, Dutch Institute for Fundamental Energy Research

This project was financially supported by the VIDI projected granted to Associate Prof. Dr. A.H.M. Smets by NWO-STW (the Netherlands Organization for Scientific Research - Dutch Foundation for Applied Sciences).

L. Han Novel Concepts for Silicon Based Photovoltaics and Photoelectrochemistry Ph.D. thesis, Delft University of Technology, with summary in Dutch

Published and distributed by Lihao Han Email: [email protected]

ISBN: 978-94-6186-413-0

Copyright © 2014 Lihao Han All rights reserved. No part of this thesis may be reproduced, stored in a retrieval system, or transmitted in any form or by any means without the prior written permission of the copyright owner.

Cover design by Lihao Han Printed and bound by CPI Wöhrmann Print Service B.V., Zutphen, the Netherlands A digital copy is available at http://repository.tudelft.nl

三十功名尘与土，

八千里路云和月。

——岳飞

We should consider every day lost on which we have not danced at least once.

And we should call every truth false which was not accompanied by at least one laugh.

—Friedrich Wilhelm Nietzsche

Veni.

Vidi.

Vici.

— Julius Caesar

Contents

1. Introduction ...... 1 1.1 Solar energy ...... 1 1.2 Photovoltaic effect and characteristics of solar cells ...... 4 1.3 Three generations of photovoltaics ...... 7 1.4 Photoelectrochemistry ...... 10 1.5 Outline of this thesis ...... 14

2. Processing and characterization of silicon nanocrystals, solar cells and photoelectrodes . 17 2.1 Chemical vapor deposition techniques ...... 17 2.1.1 Expanding thermal plasma chemical vapor deposition ...... 17 2.1.2 Plasma-enhanced chemical vapor deposition ...... 18 2.1.3 Atomic layer deposition ...... 20 2.1.4 Spray pyrolysis ...... 21 2.2 Physical vapor deposition techniques ...... 23 2.2.1 Sputtering...... 23 2.2.2 Evaporation ...... 24 2.3 Characterization tools...... 25 2.3.1 Scanning electron microscope ...... 25 2.3.2 Transmission electron microscopy ...... 26 2.3.3 Fourier transform infrared spectroscopy ...... 27 2.3.4 Raman spectroscopy ...... 28 2.3.5 X-ray photoelectron spectroscopy ...... 29 2.3.6 Photoelectrochemical measurement ...... 30

3. Raman study of laser induced heating effects in free-standing silicon nanocrystals ...... 33 3.1 Introduction ...... 34 3.2 Experimental ...... 35 3.2.1 Synthesis of Si NCs ...... 35 3.2.2 Morphology of Si NC films ...... 36 3.2.3 Raman laser heating of Si NCs ...... 38 3.3 Results and discussions ...... 38 3.4 Conclusions ...... 46

4. Optimization of double-junction thin-film silicon solar cells for a bismuth vanadate photoanode ...... 47

4.1 PEC-WSDs based on BiVO4 ...... 48 4.2 Why a-Si:H/a-Si:H tandem cells? ...... 49 4.3 Experimental ...... 51 4.4 Solar cell optimization ...... 53 4.5 Performance and stability of PEC-WSDs ...... 60 4.6 Conclusions ...... 64

5. An efficient solar water-splitting device based on a bismuth vanadate photoanode and a thin-film silicon solar cell ...... 65 5.1 Introduction ...... 66 5.2 Experimental ...... 67 5.3 Results and discussions ...... 68 5.3.1 Absorption enhancement by light-trapping in photoanode ...... 68 5.3.2 Doping profiling optimization on the photoanode ...... 71 5.3.3 Spectral matching in the PEC/PV configuration ...... 72 5.4 Conclusions ...... 76

6. A thin-film silicon based monolithic photoelectrochemical/photovoltaic cathode with efficient hydrogen evolution...... 77 6.1 Introduction ...... 78 6.2 Experimental ...... 80 6.2.1 PECVD fabrication of photocathodes ...... 80 6.2.2 Glass with integrated micro-textured photonic structures and high quality nc-Si:H materials ...... 80 6.2.3 PEC characterization ...... 81 6.2.4 ASA simulation ...... 81 6.3 Results and discussions ...... 81 6.3.1 Boron doping profiling in the a-SiC:H photocathode...... 81 6.3.2 Monolithic PEC/PV cathode ...... 83 6.3.3 Analysis of spectral utilization ...... 86 6.3.4 Stability of thin-film silicon based PEC/PV cathode ...... 87 6.4 Conclusions ...... 88

7. Nano-structured platinum synthesized by atomic layer deposition as hydrogen evolution reaction catalysts ...... 91 7.1 Introduction ...... 92 7.2 Experimental ...... 93 7.2.1 Preparation of substrates ...... 93 7.2.2 Atomic-layer deposition of platinum ...... 93 7.2.3 Deposition of platinum films by electron-beam evaporation ...... 94 7.2.4 Characterization of deposited platinum films ...... 94 7.2.5 Electrochemistry ...... 94 7.3 Results ...... 94 7.3.1 Growth rate and film morphology ...... 94 7.3.2 Surface characterization ...... 97 7.3.3 Catalytic activity for the hydrogen evolution reaction ...... 97 7.4 Discussions ...... 98 7.5 Conclusions ...... 99

8. Conclusions and outlook ...... 101 8.1 Conclusions ...... 101 8.2 Recommendations ...... 102

Appendix A. Photoanode characterizations ...... 105 A.1 Spectrum of solar simulator ...... 105 A.2 Current-voltage curves of optimized photoanode ...... 106 A.3 Structures of thin-film silicon solar cells ...... 107

Appendix B. Photocathode characterizations ...... 109 B.1 Material optimization ...... 109 B.2 Glass substrate with integrated micro-textured photonic structures ...... 111 B.3 Electrochemical impedance spectroscopy ...... 112

Appendix C. Platinum characterizations ...... 115 C.1 AFM characterization ...... 115 C.2 XPS characterization ...... 115

Bibliography ...... 119

Summary ...... 129

Samenvatting ...... 131

Publications related to this thesis ...... 133

Acknowledgements ...... 137

Curriculum vitae ...... 141

1. Introduction

From the sun I learned this: when he goes down, over-rich; he pours gold into the sea out of inexhaustible riches so that even the poorest ﬁsherman still rows with golden oars. For this I once saw and I did not tire of my tears as I watched it.

—Friedrich Wilhelm Nietzsche: Thus Spoke Zarathustra

1.1 Solar energy

The Sun is indeed as “rich” as what Nietzsche thought. The Sun is a bright star that locates in the center of our Solar System, and has an average diameter of approximately 1,392,684 km (~109 times that of Earth).1, 2 The Sun has a mass of 1.989×1030 kilograms, ~330,000 times the mass of the Earth,3 consisting of hot plasma interwoven with strong magnetic fields.4-6 The energy source of the plasma is the nuclear fusion of hydrogen atoms, and helium and other elements. According to the Theory of Relativity by Albert Einstein, the Sun is continuously generating a huge amount of energy from these reactions using this mass-energy equivalence:

E mc2 (1.1) where c is the light speed in vacuum (c ≈ 3×108 m s-1), and m is the mass converted into energy. The total amount of power that the Sun irradiates is about 3.6×1018 MW, which is close to the blackbody radiation spectrum of an object having a hot surface with a temperature of ~6000 K (following Planck’s Law, see Figure 1.1). Considering the potential amount of mass the Sun is able to convert into energy, it can be considered as the most sustainable energy source for humankind. The tiny amount of the total amount of power that reaches the Earth has a power density as large as 1353 W m-2. This value refers to the solar intensity at the outside of the Earth’s atmosphere, and is noted as Air Mass (AM) 0. AM 1 is referred as the solar vertical irradiation spectrum at the sea surface on a sunny summer noon. When the Earth is illuminated with a zenith angle of θ, the Air Mass coefficient is defined as7

AM  cos 1 (1.2)

For example, AM 1.5 is the solar spectrum when illuminated with a zenith angle of 48.2°. The solar cells used for space power applications are generally characterized according to AM 0 spectrum.8 The terrestrial solar cells are usually optimized in reference to AM 1.5 spectrum,9 which is therefore the most important reference irradiance spectrum in this thesis. The spectra are illustrated in Figure 1.1.

There are a few major energy sources available to human beings today: fossil fuels, hydro- electricity, nuclear energy, wind energy, biomass and solar energy. Fossil fuels were formed by the long-term natural processes such as anaerobic decomposition and reformation of the buried 2 1. Introduction

dead organisms, which can also be considered as a stored form of underground solar energy for millions of years.10 Hydropower is the conversion of potential or kinetic energy of running water into electricity by mechanical devices, and it can be referred as another form of solar energy. Nuclear energy can be used to generate heat and even electricity via the exothermic nuclear processes, such as atomic fission or fusion.11 Solar energy, is generally basic to man's continued survival on the Earth.12 It is a renewable energy source being utilized by human beings in various ways.13-15

)

-1 2.5

nm 6000 K Blackbody radiation -2 2.0

AM 0 radiation 1.5

1.0 AM 1.5 radiation

0.5

0.0 Spectral irradiance (W m 0 1000 2000 3000 4000 nm Figure 1.1 Blackbody radiation spectrum of an object at the temperature of 6000 K (black curve) and the solar radiation spectra (blue curve: AM 0 radiation, red curve: AM 1.5 radiation).

Figure 1.2 Total energy consumption by human beings towards sustainable energy systems. Data adapted from the German Advisory Council on Global Change.16 1.1 Solar energy 3

The development of the human society is accompanied with the exploration of all available forms of energies. The total energy consumption by human beings is increasing almost linearly every year in the recent decades, as illustrated in Figure 1.2.

Among all the energy sources mentioned in Figure 1.2, wind and geothermal are among the cheapest renewables. There is potential for significant growth but they cannot ultimately solve our energy problem due to the restrictions in natural geography and climate conditions. Biomass has the potential to provide part of transportation energy needs. Solar energy is possible to provide all our energy needs, but it is intermittent as the wind energy and currently relatively expensive (levelized cost of electricity, LCOE: 0.08-0.11 €/kWh for utility PV, compared with 0.04-0.06 €/kWh for lignite coal, according to the data analyzed by Fraunhofer ISE, Germany in November 2013). In addition, effective storage is another big challenge for most of the renewables.

Fossil fuels have excellent energy characteristics and have been intensively explored by humankind. However, the fossil fuels cannot meet the needs of the present without compromising the ability of future generations to meet their needs. The reproduction or regeneration of oil, coal and gas takes thousands of times longer than the life expectation of human beings. This means the total amount of these forms of energy stored in the Earth is limited and can therefore be considered not sustainable. Thus, seeking alternative sustainable energy sources has become super urgent in recent decades. Figure 1.2 shows that with the rapid development of the technology, the amount of fossil fuels consumption is expected to stop increasing after the year of 2030, and a rapid booming of sustainable energy utilization will begin from 2040. Among all the renewable sources, the solar power conversion, including photovoltaics (PV), solar thermal generation and solar fuels, is the most promising option, aiming to reach ~64% of the total annual energy consumption at the end of this century.

(a) 4 1. Introduction

(b) Figure 1.3 (a) Annual CO2 emissions for various energy sources. Coal, oil and gas share 43%, 33% and 18% of the total global emission in 2012 respectively. Data adapted from the Global 17 Carbon Project 2013. (b) Potential of CO2-free energy sources. The red bar shows the total amount of the CO2-free energy consumed in 2005. Data adapted from a DOE report by N.S. Lewis et al.18

Long term concerns about climate change and fossil fuel depletion will require a transition towards energy systems powered by solar radiation or other renewable sources. Carbon dioxide (CO2) excessive emissions from the burning of fossil fuel are the culprit for the climate change among various factors. Figure 1.3 (a) shows the enhancement of the annual CO2 emission in recent five decades monitored by the Carbon Dioxide Information Analysis Center (CDIAC) at U.S. Department of Energy. Among the global emissions in 2012, the energy utilization in the form of coal, oil and gas is responsible for 43%, 33% and 18% of the total amount of CO2 emission, respectively. The numbers on the right side indicate the growth rate of each energy source from 2011 to 2012.

The dramatic enhancement of CO2 concentration in the atmosphere has various negative effects to the Earth’s environment, such as the global warming, sea level rise, species extinction, etc. Seeking alternative methods for energy utilization in a CO2-free procedure has been an urgent issue for our sustainable society. Figure 1.3 (b) illustrates the potential for various CO2-free energy sources and the corresponding potentials. The huge potential of solar energy implies that large scale applications of photovoltaics, solar thermal and solar fuel devices are the most promising solutions for the concerns about both climate change and fossil fuel depletion in long term.

1.2 Photovoltaic effect and characteristics of solar cells

Photovoltaic effect, first discovered by the French physicist Edmond Becquerel in 1839 at age 19, is the basic operation principle for solar cells. It is the generation of a voltage and/or current in a 1.2 Photovoltaic effect and characteristics of solar cells 5

semiconductor device upon the absorption of light. This principle can be described by three steps: generation, separation and collection of photo-generated charge carriers in the junction.19

When a doped semiconductor that contains mostly free holes (p-type) is combined with another doped semiconductor that contains mostly free electrons (n-type), a p/n junction is formed. The electrons in the valence band absorb the energy from the incident light and get excited to the conduction band in the case that the energy of the photon is larger than the bandgap of the semiconductor (1st step: generation).

These excited electrons in the conduction band can freely diffuse and can reach the other side of the junction. In the depletion zone around the interface between the p- and n-layer, a build-in electric filed is formed. The light-excited minority carriers drift across this depletion zone (2nd step: separation).

Finally, the photo-generated holes and electrons that extracted from the solar cell are collected at the terminals of the junction consisting of two metal contacts (3rd step: collection). These carriers flow through the external circuit, and the solar energy is converted into electricity.

The solar-to-electricity energy conversion efficiency η can be defined by the ratio of the -2 maximum power (PMPP) and the average power density of the incident AM 1.5 (1000 W cm , Pin) spectrum under standard test conditions (STC, 25 °C), i.e.

P   MPP (1.3) Pin

The most important external parameters used to describe the performance of the solar cell are open-circuit voltage (VOC), short-circuit current density (JSC) and the power at the maximum power point (PMPP) as illustrated by a typical j-V curve of a single-junction a-Si:H solar cell shown in Figure 1.4 (a). From physical viewpoint, the current density should be negative since the current flows from the cathode to the anode in the solar cell. However, it is generally expressed in its absolute value for the sake of brevity.

In Figure 1.4 (a), the area ratio of the two squares (the area of blue rectangular divided by the area of the dashed purple rectangular) is defined as the fill factor (FF), which is another key external parameter for a solar cell:

P FF  MPP (1.4) VJOC SC

The FF of a commercialized solar cell is typically above 0.7. A device with a high FF value means a low internal loss of the produced photocurrent and voltage, which can be evaluated by a high equivalent shunt resistance and a low equivalent series resistance.

6 1. Introduction

The external quantum efficiency (EQE) describes the ratio of the number of charge carriers collected by the solar cell to the number of photons of a given energy incident on the cell. By integrating the EQE of a solar cell at V = 0V over the whole incident AM 1.5 spectrum, one can calculate the total JSC of a solar cell illuminated by that spectrum. The JSC value of lab-scale devices determined from the j-V measurement is very likely to be slightly overestimated due to the extra current collected from the area without the metal contact or slightly underestimated /overestimated due to the spectral mismatching between the light source and the AM 1.5 spectrum. However, the JSC integrated from the EQE measurement is both contact-size- independent and light-source-independent. Therefore, the JSC should be normalized by the value from the EQE integration to determine a reliable energy-conversion efficiency.

j 20 JSC J MPP MPP 15

)

-2

VOC= 0.86 V

(mA cm j J = 17.9 mA cm-2 5 SC FF = 69.5%  = 10.7% 0 0.0 0.2 0.4 0.6 0.8 1.0 VMPP VOC (a) V (V)

1.0

0.8

0.6

EQE 0.4

J = 17.9 mA cm-2 0.2 SC

0.0 300 400 500 600 700 800 (nm) (b)

Figure 1.4 The j-V curve and external characteristics of a typical single-junction a-Si:H solar cell (a) and its EQE spectrum (b). 1.3 Three generations of photovoltaics 7

A typical EQE spectrum of a single-junction a-Si:H solar cell is shown in Figure 1.4 (b). The a- Si:H layers can absorb up to ~85% of light in the visible range (500-600 nm). The bandgap of a- Si:H is ~1.8 eV, which means that photons with a wavelength longer than 800 nm cannot be absorbed any more. The penetration depth of the light is comparable with the absorber layer thickness in the range of range of 600-800 nm. If a significant fraction of the light coupled into the solar cell is coherent, the interference in the absorber cause the interference in the spectrum. -2 The JSC value integrated from this EQE spectrum is 17.9 mA cm .

There are two dominant loss mechanisms in a single-junction solar cell. If the incident photon energy is lower than the bandgap of the absorber, it cannot be absorbed and is transmitted through the absorber layer. If the photon energy is higher than the bandgap of the absorber, the energy exceeding the bandgap is lost in the thermalization processes. The theoretical maximum solar-to- electricity conversion efficiency is determined by these spectrum losses, blackbody radiation and radiative combination. This single bandgap limit of a solar cell was first calculated by W. Shockley and H. J. Queisser in 1961,20 referred as Shockley-Queisser limit from then on.

1.3 Three generations of photovoltaics

The development of photovoltaic technology can be traditionally categorized in three generations according to the cost and efficiency shown in Figure 1.5.21

Figure 1.5 Illustration of the (expected) cost and (potential) efficiency of the three generations of photovoltaics. Data adapted from G. Conibeer.21

The first generation of photovoltaic technology (1GPV) generally refers to Si wafer based solar cells, specifically noted as single-crystalline Si (mono-c-Si) and multi-crystalline (or polycrystalline) Si (poly-c-Si). This technology was first developed at Bell Labs in USA in 1954, when Daryl Chapin, Calvin Fuller and Gerald Pearson fabricated a prototype silicon-based device to convert solar energy into electricity.22 Bell Telephone Laboratories managed to produce a c-Si 8 1. Introduction

PV device with ~4% efficiency and quickly improved the efficiency to 11%.23 Nowadays, the c-Si based solar device is the most popular PV technology due to the earth-abundance of Si (27% of the Earth crust mass, ranked the second just after oxygen),24 relatively high efficiency and good stability. The Si wafer based panels count for 85-90% in the PV market worldwide.25 After the rapid development in recent decades (especially in the 1990s as shown in the NREL record efficiency chart),26 the state-of-the-art silicon heterojunction with intrinsic thin (HIT) layer solar cells based on the interdigitated back contact (IBC) technology have achieved the highest efficiency of 25.6% (PANASONIC Co.)14 approaching the theoretical maximum conversion efficiency of 29.4% (for the c-Si material with a bandgap of 1.1 eV under AM 1.5 spectrum illumination), known as the Shockley–Queisser limit (including Auger recombination).20, 27 Only little room is left for the further optimization on the loss mechanisms such as preventing contact losses and parasitic absorption losses, minimizing the reflection from the surface and introducing light trapping technique in the absorber layers.23, 28, 29 The majority of the PV products nowadays are based on poly-c-Si and mono-c-Si materials. In October 2014 in South and Southeast Asia, the net price of c-Si modules has become as low as 0.48 €/Wp (euro per Watt peak).30

The second generation of photovoltaic technology (2GPV) is based on the thin-film (TF) technology using less amount of raw material and aiming for a lower cost compared with 1GPV. Cadmium telluride (CdTe),31-33 copper indium gallium (di-)selenide (CIGS)34-36 and amorphous silicon (a-Si:H)37-39 are materials that are most frequently used in 2GPV applications. Other options can be the gallium arsenide (GaAs)40 usually for concentrated photovoltaics (CPV) or organic solar cells such as dye sensitized solar cells (DSSC, or known as Grätzel cells).41 The 2GPV used to have a 10-15% PV market occupancy, but its share in the market reduced to ~9% in 2013.25

The solar cells implemented and studied in Chapter 4-6 in this thesis are based on TF-Si technology, in which hydrogenated amorphous silicon (a-Si:H) and/or hydrogenated nano- crystalline silicon (nc-Si:H, a transition material from a-Si:H to c-Si) are the main absorber layers. Sometimes silicon carbide/nitride/oxide layers are also deposited as the supporting layers. TF-Si technologies are quite cost-effective in 2GPV, but devices based on a-Si:H materials are suffering from the light-induced degradation, referred to as the Staebler–Wronski effect (SWE) discovered in 1977.42 The prolonged illumination (generally the first a few hundreds of hours) leads to the creation of metastable defects in the a-Si:H absorber layer,43 therefore decreasing the initial efficiency value to typically 10-20% relatively.

Two types of TF-Si solar cell configurations are utilized in this thesis. The first type is a superstrate configuration. The illumination light enters through the glass superstrate, a front TCO layer, and then enters the “p/i/n” layers. The second type is a substrate configuration in which the light enters through a front TCO layer and the “n/i/p” layers before reaching the substrate. The advantage of the “n/i/p” structures is that the substrates do not have to be transparent, allowing more freedom for the introduction of light-trapping techniques in the device. TF-Si PV technology is flexible and mature, and therefore it can be easily integrated in various hybrid devices.

1.3 Three generations of photovoltaics 9

The concept of third generation photovoltaics (3GPV) was proposed to overcome the Shockley- Queisser limit of a single-junction cell aiming to achieve a higher efficiency than the 1GPV and 2GPV, but still using earth-abundant cost-effective materials. 3GPV is the main focus of this thesis.21, 44, 45 Many novel concepts can be applied in the 3GPV, such as multi-junctions,46 nanocrystals (NCs, or called quantum dots, QDs),47 intermediate band solar cell,48, 49 solar thermal 15, 50 51-53 54, 55 technology, hot carrier solar cells, perovskite (CaTiO3) or even non-semiconductor technologies including biomimetics and polymers.56, 57

Among all these 3GPV techniques, multi-junction is the only type that has been commercialized for large-scale applications. The thin-film technology allows the single “p/i/n” or “n/i/p” structures stack to flexibly form a multi-junction solar cell (therefore this technology may sometimes be considered as 2GPV as well). The VOC value of a multi-junction device can be close to the sum of the VOC of each single-junction as they are connected in series. The absorption of the solar spectrum is distributed over the various junctions. The JSC value of the multi-junction device is limited by the junction generating the lowest current. Therefore, for an optimal performance of such device, the current generation should be the equally distributed over all the junctions.

NCs are materials whose excitons are confined in all three spatial dimensions.58 Therefore NCs can have promising 3GPV applications because of these quantum confinement effects. Solar cells based on NCs have the potential to break through the Shockley-Queisser limit, using mechanisms like multiple exciton generation (MEG)59 and spectral conversion like up or down conversion by space separated quantum cutting (SSQC).60, 61 In a conventional solar cell, the absorbed photon is only capable to excite one electron from the valence band to the conduction band, where the excess energy is lost in the form of heat. MEG is a carrier multiplication procedure in which more than one electron can be excited in a NC by a photon of high energy.59 This MEG phenomenon has the potential to enhance the efficiency of the solar cell. In 2011, above 100% external quantum efficiency (EQE) value in a certain spectral range has been achieved in a cell based on PbSe NCs.62

The concept of spectral conversion is aimed to modify the incident spectrum, so that a better match can be achieved between the spectral response of the PV absorber and the converted light spectrum offered. In the so-called “up-conversion” mechanism in a NC, two photons of lower energy can be combined to generate one photon of higher energy.63 This procedure can convert the red spectrum transmitted through a solar cell to the blue range. The second mechanism is noted as “down-shifting”, in which a photon of higher energy is transformed into a photon of lower energy in the form of luminescence.64 The blue response of a solar cell based on NCs can be improved by shifting the incident photons to a spectral range where the device has a higher response. The third mechanism is called “down-conversion” or “SSQC”,60, 61 in which one photon of higher energy can be transferred into two photons of lower energy.65 Up- and down-conversion mechanism has the potential to break through the Shockley-Queisser limit. For instance, by using down- and up-conversion mechanism, extra 32% and 35% higher intensity can be absorbed by c- Si material illuminated by the AM 1.5 spectrum respectively.66 10 1. Introduction

1.4 Photoelectrochemistry

The Sun continuously provides 1.2 × 1017 W of power to the Earth,17 and in one hour it delivers approximately the entire amount of energy humans use in a year (4.6× 1020 J).67 Figure 1.6 (a) shows the rapid development of the effective installed sustainable energy worldwide. It is predicated that within a decade the installation of solar and wind energy can be almost the same level of nuclear and hydro energy, if the growth of the last years continues at the same rate. Figure 1.6 (b) shows an example of the solar panels on the roof of Prof. M. Zeman, and it is obvious that this PV system generated 5-6 times more energy in summer time than in the winter time in Delft. However, the Dutch population generally needs more energy in the winter than they do in the summer. Due to these seasonally or even daily fluctuations and intermittency of sustainable energy, effective collection, storage and distribution are critical in the utilization of this largest carbon-neutral energy source for our society.

Solar fuels are fuels produced from sunlight via artificial photosynthesis or thermochemical reaction.68, 69 It is a process that partly mimics what plants do by splitting water into hydrogen fuel 70 and oxygen, or reducing carbon dioxide (CO2) to methane (CH4) or organic components. This fuel forming process is typically performed in a photoelectrochemical (PEC) cell, and generally uses photoelectrodes that are constructed from semiconductor(s) and electrocatalysts.

1000 Hydro CF=0,40

100 Nuclear CF=0,90

1 Wind CF=0,30

Solar CF=0,15 0.1

Effective installed Power (GW) 1980 1990 2000 2010 2020 Year (a) 1.4 Photoelectrochemistry 11

Generated energy (kW h) 0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Month (b) Figure 1.6 The (expected) installation of several renewable energy sources with the corresponding capacity factor (a). An example of the seasonal fluctuation of solar energy in the year of 2003 in Delft (b). Data adapted from Solar Energy, a Massive Open Online Course given by A.H.M. Smets.71

Figure 1.7 illustrates the prototype structure of a PEC device proposed by JCAP at Caltech. The sunlight incidents on the photoanode material which can be a semiconductor having relatively 72 73 74 high bandgap such as BiVO4 (Chapter 4 and 5), Fe2O3, WO3 and is generally coated with O2 evolution reaction (OER) catalysts.75 A photocathode material, such as GaAs,76 Si77 (Chapter 6) that has a relatively low bandgap in reference to that of the photoanode material, is located at the bottom part of the cell and absorbs the photons with longer wavelengths. Hydrogen evolution reaction (HER) catalysts can be metal oxide or metal in the form of thin coatings or nano-particles (NPs) made of materials like Ni,78 Pt79 (Chapter 7). A H+ permeable membrane is inserted 80 between the two photoelectrodes and it can separate the generated O2 and H2 gases.

Figure 1.7 A prototype of a water-splitting device proposed by JCAP.81

12 1. Introduction

The electrolyte of the redox reaction can either be in acid or in base solutions. In the acidic solution, the reactions on the cathode and anode happen as:82

 0 (1.5) 4H 4e  2H2 ; E ox  0.0 V vs . NHE  0 (1.6) 2H2 O 4h  4H +O 2 ; E red   1.229 V vs . NHE

Herein e- and h+ represent the electrons and holes respectively.

In the alkaline solution, the reactions on the cathode and anode happen as:82

0 (1.7) 4H2 O 4e  2H 2 +4OH ; E red   0.828 V vs . NHE  0 4OH 4h  2H2 O+O 2 ; E ox  0.401 V vs . NHE (1.8)

Therefore, the overall water-splitting reaction can be described as:

Sunlight 1 2H2 O 2H 2 +O 2 ; G  237 kg mol (1.9)

A potential higher than 1.23 V between the cathode and anode is required for water-splitting reactions to occur in principle, which corresponds to a Gibbs free energy change of 237 kJ mol- 1.82 In practice, more energy than thermodynamically is expected to drive the redox reaction, and this extra energy is defined as overpotential. This overpotential originates from various reasons such as defects in the semiconductor, surface recombination between the semiconductor and electrolyte, low charge carrier collection (sometimes called as charge carrier separation). Novel techniques are proposed and applied in this thesis to reduce this overpotential and consequently to improve the solar-to-hydrogen efficiency. Particularly, we focus on the integration of the photoelectrode with a solar cell, which provides the potential for water-splitting devices with high efficiencies. Various configurations for PEC/PV devices can be explored. Figs. 1.8 (a) and (b) illustrate a photoanode and a photocathode configuration developed in this thesis. The rear PV junction should work efficiently under the illumination of the spectrum transmitted through the front PEC junction. The spectrum utilization and the j-V curve matching between the PEC and PV junctions will be studied in Chapter 4-6.

1.4 Photoelectrochemistry 13

(a)

(b)

Figure 1.8 PEC/PV monolithic water-splitting photoanode (a) and photocathode (b) configurations investigated in this thesis. These two figures have been highlighted as the cover image of ChemSusChem (issue of October 2014) and Journal of Materials Chemistry A (issue of February 2015) respectively. 14 1. Introduction

1.5 Outline of this thesis

As mentioned in the previous sections, this doctoral thesis introduces some novel concepts for silicon based photovoltaics and photoelectrochemistry. Chapters 2 and 3 focus on the processing and characterization of Si NCs as a potential material for 3GPV applications. Chapter 4-6 focus on the development of PEC/PV devices where the PV part is based on TF-Si technology. Chapter 7 focuses on a high rate deposition approach for water reduction catalysts.

The main scientific questions tackled in this thesis are:

 What temperature can the free-standing Si NCs be heated when they are illuminated by a low intensity but sharply focused laser beam? (Chapter 3)  What are the design rules of a double-junction TF-Si solar cell for a hybrid water- splitting device in reference to the transmitted spectrum through the front photoanode? (Chapter 4)

 Where is room for the further optimization of an efficient hybrid BiVO4/double- junction TF-Si photoanode device? (Chapter 5)  What is the optimal device structure for a photocathode based on TF-Si alloys? (Chapter 6)  What is the optimal high-rate deposition conditions for the nano-structured Pt as the water reduction catalyst by ALD technique? (Chapter 7)

Chapter 2 focuses on the deposition and characterization methodology of Si NCs, TF-Si solar cells, metal-oxide photoanodes, TF-Si photocathodes and nano-structured metal catalysts on photocathodes. A high rate fabrication of Si NCs by a state-of-the-art technique called expanding thermal plasma chemical vapor deposition (ETP-CVD) is introduced. The advantages and challenges for the Si NCs fabrication are discussed. The absorber layers of the TF-Si solar cells in this thesis are deposited by PECVD. The BiVO4 photoanodes are synthesized by spay pyrolysis, and the nano-structured Pt are synthesized by atomic layer deposition (ALD) technique for hydrogen evolution reaction (HER) catalysts. The morphology and size distribution of the Si NCs is characterized by SEM and TEM. Analytical methods are employed to investigate material and device properties, such as Fourier transform infrared spectroscopy (FTIR) to study the oxidation process of this material with large surface-area-to-volume-ratio, as well as the carbon ratio in the a-SiC:H films. X-ray diffraction (XRD) measurement is carried out to analyze the crystallinity of the NCs and the various orientation of the crystals. Raman spectroscopy (RS) is another important tool to observe the crystallinity of the NCs, allowing the identification of species presented in the material. XPS is utilized to study the chemical components of the ALD-grown Pt. The performance of the photoelectrodes are verified by PEC measurements.

In Chapter 3, a fundamental study of heating effects of Si NCs induced by Raman laser beam is presented. The free-standing Si NCs can be easily heated using the low intensity but sharply focused Raman probe beam, because the heat loss mechanism is not dominated by the thermal conduction like Si NCs embedded in a host matrix. This thermal expansion results in a significant 1.5 Outline of this thesis 15

red-shift and broadening of peak width from the initial Si crystalline peak. By analyzing the ratio of Anti-Stokes-to-Stokes peak intensities of the first order Si-Si transverse optical (TO) phonon mode, the temperatures of Si NCs under laser exposure can be determined. It is found that the laser heating effects are reversible to a large extent, however the nature of the free-standing Si NCs is slightly modified after intensive illumination.

In Chapter 4, a PEC water-splitting photoanode based on bismuth vanadate (BiVO4) and an amorphous silicon tandem cell (a-Si:H/a-Si:H) is demonstrated. The front BiVO4 photoanode is optimized first for effective oxygen evolution reactions (OER). An a-Si:H/a-Si:H cell is chosen as the rear power source. The tandem solar cells are then optimized in reference to the spectrum transmitted through the photoanode. Finally, the stability of the photoanode is addressed.

In Chapter 5, the further optimization of the hybrid PEC/PV water-splitting device previously demonstrated in Chapter 4 is presented. The improvement with respect to the cell in Chapter 4 that also used gradient tungsten (W) doped BiVO4 has been realized by simultaneously introducing a textured substrate to enhance light trapping in the BiVO4 photoanode and further optimization of the W gradient doping profile in the photoanode. Various TF-Si PV cells have been studied in combination with this BiVO4 photoanode, such as a-Si:H single-junction, a- Si:H/a-Si:H double-junction and a-Si:H/nc-Si:H double-junction. The highest conversion efficiency - which is also the record efficiency for metal-oxide-based water-splitting devices - is reached for a tandem consisting of the optimized W:BiVO4 photoanode and the micromorph Si (a-Si:H/nc-Si:H) cell.

In Chapter 6, a cost-effective, earth-abundant photocathode based on a-SiC:H for hydrogen evolution reaction (HER) is studied and developed. This monolithic a-SiC:H PEC cathode integrated with an a-Si:H/nc-Si:H double PV junction to achieve a high unbiased photocurrent density. The a-SiC:H photocathode used no HER catalyst and the boron doping profile in the a- SiC:H layers is studied. The growth of high quality nc-Si:H PV junctions in combination with high spectral utilization is achieved using glass substrates with integrated micro-textured photonic structures. Advanced Semiconductor Analysis (ASA) software is employed as well to simulate the performance of the PEC/PV cathode.

In Chapter 7, the ALD technique to fabricate nano-structured Pt onto etched p-type Si(111) wafers and glassy carbon discs as hydrogen-evolution reaction (HER) catalysts is introduced. New types of precursors are utilized to increase the Pt growth rate. X-ray photoelectron spectroscopy (XPS), atomic force microscopy (AFM), ellipsometry and scanning-electron microscopy (SEM) are used to analyze the composition, structure, morphology and thickness of the ALD-grown Pt nanoparticle films. The catalytic activity of the ALD-grown Pt for the HER is compared with electron-beam evaporated Pt on glassy carbon electrodes.

Chapter 8 summarizes the whole thesis.

16 1. Introduction

2. Processing and characterization of silicon nanocrystals, solar cells and photoelectrodes

2.1 Chemical vapor deposition techniques

2.1.1 Expanding thermal plasma chemical vapor deposition Up to date, a few fabrication methods of Si NCs have been proposed. Conibeer et al. sputtered Si rich oxide (SRO) and quartz (SiO2) layer by layer on a wafer. The thickness of each layer ranges from 2-6 nm. After annealing at 900-1100 °C for 1 hour, dense Si NCs were formed in the SiOx 83 matrix. The size of Si NCs can be controlled by the layer thickness of SRO and SiO2. Han et al. employed pulsed laser deposition (PLD) tools with Si rich SiOx target, followed by similar annealing procedure as well.84 Molecular beam epitaxy (MBE),85 non-thermal plasma,86 chemical wet etching87 can also be used for Si, SiGe or Ge NCs fabrications. Most of these techniques suffer from the low deposition rate that limits large scale applications of Si NCs. In addition, the high temperature procedure is not cost-effective, as well as the maintenance cost for some setup such as MBE.

In this chapter, a novel fabrication processing of Si NCs by expanding thermal plasma (ETP- CVD) is proposed. A photo of ETP-CVD setup is shown in Figure 2.1. The plasma creation, plasma transport and film deposition occur in separate regions of the reactor. Therefore, ETP- CVD can be considered as a remote plasma technique.88, 89

Figure 2.1 A photo of ETP-CVD located in the cleanroom class 10000 of PVMD lab. 18 2. Processing and characterization for Si NCs, solar cells & photoelectrodes

At the top of the ETP-CVD reactor, six electrically insulated copper plates are stacked, acting as the cascaded arc. The bottom plate serves as the anode, while the three cathode tips are located at the top of the arc and each tip stands in the centers of the copper plate. Each plate has a circular open space in the center to form the plasma channel (ф =2.5 mm), from which the argon (Ar) gas is injected and is operated by a mass flow controller. The plasma is ignited and sustains due to the high voltage/current (typically 70~100 V / 20~60 A) between the cathode tips and the anode plate at pressures commonly in the 0.5 bar range.

With the help of two stacked root blowers, the deposition chamber is kept at a much lower pressure than the arc, typically around 0.5-2 mbar. The distance from the arc to the substrate holder is as long as 38 cm. The pressure difference between the arc and the vessel results in a supersonic expansion of the plasma and a subsonic expansion after a stationary shock. At ~4.5 cm below the nozzle a stainless steel injection ring is located, which serves as an injection point for silane (SiH4) gas.

The SiH4 gas decomposes and reacts with the plasma to form predominantly silyl radicals. During their transportation to the substrate, these silyl radicals form Si NCs due to the lower temperature of the plasma and larger pressure. These Si NCs “freely stand” in the form of nano-powder on top of the substrate. This substrate is pre-cleaned first in acetone and then in IPA in a heated water bath chest. The substrate is kept at room temperature during the plasma synthesis, and the substrate material does not play an important role in this Si NCs fabrication process, which means either a piece of Corning glass or c-Si wafer is suitable as substrate.

The ETP-CVD is highly efficient and several micrometers of the Si NCs porous layer can be deposited on the substrate in 1 second. Small Si NCs with a narrow size distribution of a few nanometers can be collected on the substrate spots which are exactly below the holes in the SiH4 injection ring. We believe that these small NCs were formed in the plasma beam and directly deposited on the substrate. However, free-standing larger NCs (diameter of dozens of nanometers) are also observed, which were collected from the other areas on the substrate. They have a larger average diameter due to the recycling from the back ground volume into the plasma beam.90, 91 By reducing the diameter of the deposition chamber, the amount of the larger NCs can be reduced.

2.1.2 Plasma-enhanced chemical vapor deposition The thin-film silicon (TF-Si) for the solar cells (Chapter 4-6) and photocathodes (Chapter 6) were deposited by plasma-enhanced chemical vapor deposition (PECVD). PECVD is a frequently used technique for PV and semiconductor industry due to its relatively large areas (1.5 m2 for example), low substrate temperature (180-250 °C) during processing and device-grade deposition quality with a good coverage of textured morphologies, uniformity and adhesion.

A sketch of the PECVD setup is illustrated in Figure 2.2 (a). The plasma is initiated by an oscillating voltage with a radio frequency (RF) of 13.56 MHz applied between the electrode plates. Free electrons start accelerating in the oscillating electric field. Ionized gas molecules are 2.1 Chemical vapor deposition techniques 19

created due to the inelastic collisions between the free electrons and the neutral gas molecules. This results in the production of secondary electrons that in turn cause gas molecules to react or dissociate and maintain the plasma. Thin films (TFs) of various materials can be deposited according to the precursor gas that is used. The deposition rate and the physical properties of the films, such as the refractive index (n), extinction coefficient (k), can be turned by varying the RF power, chamber pressure, substrate temperature, electrode distance, amount of gas flows, etc.

(a) (b)

(c) Figure 2.2 A sketch illustration of the PECVD setup (a),92 together with a sketch (b) and a photo (c) of the AMIGO multi-chamber PECVD setup located in Cleanroom Class 10000 of the PVMD group. LLC means the load lock chamber, DPC 1-5 are the depositions process chambers used to synthesis intrinsic and doped TF-Si layers, and DPC 6 is a sputtering chamber for TCO layers.

20 2. Processing and characterization for Si NCs, solar cells & photoelectrodes

Figures 2.2 (b) and (c) show a sketch and a photo of the so-called AMIGO multi-chamber PECVD utilized in this work. The intrinsic (non-intentionally doped, more scientifically) a-Si:H films are deposited by the silane (SiH4) gas (sometimes together with H2 for a better passivation) using RF power. The p-doped films were deposited in another chamber with SiH4 diluted by diborane (B2H6), and the n-doped films were deposited with SiH4 diluted by phosphine (PH3). CO2 is introduced for the fabrication of oxide materials such as nc-SiOx:H, and CH4 is injected for the a-SiC:H depositions.

To guarantee high quality nc-Si:H films, the deposition requires a higher power, pressure and lower deposition rate compared with the parameters of the a-Si:H. The material of nc-Si:H was deposited in a chamber powered by very high frequency (VHF), which guarantees a larger processing window for high quality nc-Si:H.

2.1.3 Atomic layer deposition Atomic Layer Deposition (ALD) is a CVD thin film (TF) deposition technique that is based on self-limiting or sequentially self-terminating gas phase chemical process to form the TFs or nanoparticles (NPs). Chemical precursors are carried by nitrogen (N2) gas from an injection hole into the chamber. After reacting with the surface of a substrate, the extra precursors are pumped away through the taphole. Through the repeated exposure to the precursors, a thin film is deposited in a slow rate and the thickness can be controlled up to angstrom scale. Both amorphous and crystalline materials are possible depending on substrate and temperature.

Advantages of a self-limiting film coated by ALD include uniform surfaces with low defect density, high conformity to surface features, high control and accuracy of atomic level thickness, wide process windows (no sensitivity to temperature or precursor dose variations) and excellent reproducibility. ALD does have some limitations, including incomplete reactions and slow reaction rates. In addition, some ALD process mechanism has not yet been completed understood.

The Cambridge Nanotech S200 ALD system based in the Noyes Laboratory at Caltech is employed for the deposition of nano-structured platinum (Pt) for the water reduction catalysts in Chapter 7. Figure 2.3 (a) is the photo of the ALD system and (b) is the sketch illustration its deposition chamber and precursor system.

2.1 Chemical vapor deposition techniques 21

(a) (b) Figure 2.3 A photo of Cambridge Nanotech S200 ALD (a) and the sketch of its deposition chamber and precursor system (b).

2.1.4 Spray pyrolysis

The bismuth vanadate (BiVO4) photoanodes discussed in Chapter 4 and 5 were deposited by spray pyrolysis in the Chemistry and Chemical Engineering department in TU Delft, and the recipe was developed by Abdi et al.93-96 A sketch of the spray pyrolysis setup is shown in Figure 2.4.

Before the deposition, 4 mM Bi(NO3)3·5H2O (98%, Alfa Aesar) was dissolved in acetic acid (98%, Sigma Aldrich), and an equimolar amount of vanadium in the form of VO(AcAc)2 (99%, Alfa Aesar) was dissolved in absolute ethanol (Sigma Aldrich). The precursor solution in the reservoir was made by mixing these two solutions. Two types of fluorine-doped tin dioxide (FTO) coated glass substrates are used as the substrate. One is TEC-15 (15 Ω/□, Hartford Glass Co.) that is initially used in Chapter 4, and the other is ASAHI VU-type (8 Ω/□, Asahi Co.) that is optimized in Chapter 5. The substrates were cleaned by ultrasonic rinsing treatments for successive 15 min for three times, in a Triton® solution, acetone and ethanol respectively. The substrate temperature during spraying was kept at 450 °C on a heating element, as measured by a thermocouple pressed to the substrate surface. The Quick-Mist Air Atomizing Spray nozzle was located 20 cm above the substrate surface, and the nozzle was connected by a programmed controller.

22 2. Processing and characterization for Si NCs, solar cells & photoelectrodes

Figure 2.4 A sketch illustration of spray pyrolysis setup.

The precursor solution was placed 20 cm below the nozzle and fed to the nozzle via the siphoning effect induced by the nitrogen gas flow with an overpressure of 0.06 MPa. Each spray cycle was consisted of 5 seconds of spray time and 55 seconds of delay time to allow sufficient solvent evaporation. 100 cycles were needed to deposit the film with a deposition rate of ~1 nm per cycle.

95, 97 The doping profile was investigated in this thesis as well. The 1% W-doped BiVO4 sample was prepared by spraying 200 cycles of the BiVO4 precursor solution containing 1 at% of W. The W:BiVO4 homo-junction was prepared by spraying 100 cycles of the BiVO4 precursor solution containing 1 at% of W, followed by 100 cycles of the BiVO4 precursor solution. This sequence was reversed for the deposition of the W:BiVO4 reverse homo-junction. To deposit the gradient- doped W:BiVO4, the concentration of W in the BiVO4 precursor solution was changed in step every 20 cycles, starting from 1 to 0 at%.

Prior to praying the BiVO4, a SnO2 interfacial layer (~80 nm) was coated onto the FTO substrate 98, 99 to prevent possible recombination at the FTO/BiVO4 interface. A 0.1 M SnCl4 (99%, Acros Organics) solution in ethyl acetate (99.5%, J. T. Baker) was used as the precursor solution. The SnO2 layer was deposited at 425 °C using 5 spray cycles (5 s on, 55 s off) in a gravity-assisted siphoning mode, where the precursor solution was placed 30 cm above the nozzle. After deposition, the SnO2/BiVO4 samples were subjected to an additional 2-hour heat treatment in a tube furnace at 450 °C in air atmosphere to further improve the crystallinity.

A 30-nm CoPi catalyst was electrodeposited according to the recipe developed by Kanan and Nocera.100 The electro-deposition was performed at a constant voltage of 1.7 V versus RHE for 15 min. Care was taken to always keep the electrodeposited CoPi layer wet, as intermediate drying of the CoPi was found to adversely affect the stability. 2.2 Physical vapor deposition techniques 23

2.2 Physical vapor deposition techniques

Physical vapor deposition (PVD) describes a variety of vacuum deposition methods used to deposit TFs by the condensation of a vaporized form of the desired film material onto various substrates. No chemical reaction takes place during the vacuum deposition, but the PVD is normally realized by the material phase change, such as from solid to gas phase during the deposition, and back to solid with a different morphology during the transportation onto the substrate. These PVD techniques include pulsed laser deposition (PVD), sputtering, evaporation, cathodic arc deposition (CAD), etc. In this section, the sputtering and evaporation are introduced for the deposition of transparent conductive oxide (TCO) and metal contacts of the PV and/or PEC devices respectively.

2.2.1 Sputtering Physical sputtering is a popular PVD tool for TF depositions that involves eroding material from a target onto a substrate. The deposition is driven by momentum exchange between the ions and atoms in the materials due to collisions.

Argon (Ar) is introduced as the processing gas in the vacuum chamber with a sputter target, which is powered by radio frequency (RF) source. Electrons from the target surface ionize the Ar atoms, creating a plasma. The target attracts positive Ar ions, which in turn ejects neutrally charged atoms from the sputter target towards the substrate’s surface. Subsequently, a uniform film is deposited on the substrate’s surface.

Figure 2.5 RF magnetron sputtering system (Kurt J. Lesker) for the deposition of ITO and AZO layers in the PVMD lab.

24 2. Processing and characterization for Si NCs, solar cells & photoelectrodes

The RF magnetron sputter employed in this thesis is made by RF magnetron sputtering system Kurt J. Lesker, as shown by Figure 2.5. The samples are placed in the load lock and transferred to the process chamber through a gate valve for depositions. The targets in the sputter are indium tin oxide (ITO, 10% SnO2 and 90% In2O3) or aluminum doped zinc oxide (AZO, 2% Al2O3 and 98% ZnO) deposition. By controlling the RF power intensity, electrode distance, substrate temperature and chamber pressure, ITO or AZO layer with good transparency and conductivity can be sputtered, which are both important transparent conductive oxides (TCO) in TF-Si PV technology.

2.2.2 Evaporation The metal contact of the solar cells and PEC devices in this thesis were deposited using an evaporation PVD technique. In an evaporator, a target anode is bombarded with an electron-beam given off by a charged tungsten (W) filament in a high vacuum chamber. The atoms in the target material is transformed into a gaseous phase when exposed to the intensive electron-beam. These atoms precipitate into solid form after they are evaporated out of the boat and cooled down in the vacuum chamber. A thin layer of the anode material is coated on the superstrate placed above the target. A scheme of the evaporator is depicted in Figure 2.6 (a).

A PROVAC PRO500S evaporator is utilized in this thesis to deposit the metal contacts of the solar cells and PEC devices, as shown in the photo in Figure 2.6 (b). Ag can be deposited by thermal evaporation (or known as resistance heating evaporation). Other materials such as Al, Cr, Ti and ceramics with higher melting points should be deposited by electron-beam evaporation. This system is equipped with a 4 pocket electron-beam gun for evaporation. A few substrates up to 10 cm ×10 cm can be loaded into the evaporator, and large area patterns can be created by using shadow masks. Pumping cycles and deposition recipes are fully programmable.

The front contacts of the PV and PEC devices in this thesis consist of a 300 nm thick aluminum (Al) bar on the edge of the TCO-coated substrate (such as ASAHI VU-type). The back contact of the solar cell is a stack of 100 nm Ag, 30 nm Cr and 300-500 nm Al. Ag can form an alloy with TF-Si after annealing and reduce the Ohmic resistance of the contact. Besides, Ag is “shiny” metal that can reflect the light back to the absorption layers. The Al layer prevents the oxidation of the Ag layer, whereas the Cr interlayer avoids mixing of Ag and Al in post-deposition annealing treatments.

2.3 Characterization tools 25

(a)

(b) Figure 2.6 (a) Schematic overview of Resistance heating evaporation (left) and electron-beam evaporation (right) physical vapor deposition; illustrations adapted from Ref.,101 and (b) a photo of PROVAC PRO500S evaporator located in PVMD group.

2.3 Characterization tools

2.3.1 Scanning electron microscope Scanning electron microscope (SEM) is an electron microscope that frequently utilized for the nano-technology research. Electron-beam generated by the electron gun is focused by more than one condenser lens in a magnetic field, and scans fast over the surface of the specimen in vacuum. Various signals can be generated when the electronic beam interacts with atoms of the sample. These detected signals reflect the topography, morphology and composition of the sample’s 26 2. Processing and characterization for Si NCs, solar cells & photoelectrodes

surface. A real-time image can be produced when the beam's position is combined with the detected signal. This technique can achieve a resolution of a few nanometers minimum.

The SEM setup used in this thesis is FEI Quanta 200 FEG, located in the Charged Particle Optics (CPO) group, Department of Imaging Physics at TU Delft. This technique was employed to observe the morphology of the Si NCs, photoanodes, photocathodes and metal catalysts in Chapter 3-7 at various operating voltages from 3 to 15 eV. A photo is shown in Figure 2.7.

Figure 2.7 The SEM setup located in CPO group at Delft utilized in this thesis.

2.3.2 Transmission electron microscopy Transmission electron microscopy (TEM) is a microscopy technique that operates on similar basic principles with that of optical microscopes but using electron-beams instead of visible light beams. This electron-beam transmitted though the ultra-thin specimen in vacuum, and meanwhile interact with the materials of the specimen. The traveling route of the electrons is changed due to the collisions with the atoms when transmitting though the specimen, and these electrons are scattered in various directions. The scattering angles indicate the density and thickness of the specimen, and can be detected with different brightness and contrast. A real-time image is formed on a CCD camera when these signals are focused and magnified.

2.3 Characterization tools 27

Figure 2.8 A photo of the TEM system utilized in this thesis. Photo adapted from Ref.102

The advantage of this technique is that the de Broglie wavelength of the electrons can be thousands of times smaller than the visible light source. As a result, this technique can observe nanostructures in the scale of 0.1-0.2 nm, and it can operated as high resolution transmission electron microscopy (HR-TEM) when observing specimen with a big magnification.

The HR-TEM setup used in this thesis is TECNAI G2 S-Twin F20, which is located in State Key Laboratory of Inorganic Synthesis & Preparative Chemistry at Jilin University in China. When observing the lattice of the Si NCs, the operating voltage is 200 kV. A photo of that system is shown in Figure 2.8.

2.3.3 Fourier transform infrared spectroscopy Fourier transform infrared (FTIR) spectroscopy is an indirect technique to simultaneously collect a transmission or absorption specimen over a wide infrared spectrum and then convert the measured data into the actual spectrum using a mathematical process known as Fourier transform.

In an FTIR setup, the light from the polychromatic infrared source is equally spitted into two beams. One beam is refracted towards a fixed mirror, and the other beam is transmitted through the specimen and arrives at a dynamic mirror that is linearly moving at a constant speed. These two beams are reflected by the mirrors and return to the splitter. An interference beam is formed due to the optical path difference of the two beams, and this beam is transmitted through the specimen and collected by the detector. The measured interferogram is processed by Fourier transformation, and a transmittance or absorbance spectrum is obtained.

The most frequently studied spectrum range is the mid-infrared (400-4000 cm-1) and the near- infrared (4000-13300 cm-1) region. The advantage of this technique is that an FTIR spectrometer 28 2. Processing and characterization for Si NCs, solar cells & photoelectrodes

measures much faster than the other traditional spectral measurement methods using a dispersive monochromatic light beam at a sample.

The FTIR technique was utilized to analyze the Si NCs oxidation procedure in this thesis, as well as the a-SiC:H films for photocathode applications. A photo of the FTIR setup located in the PVMD cleanroom is shown in Figure 2.9.

Figure 2.9 The FTIR setup employed in this thesis.

2.3.4 Raman spectroscopy Raman spectroscopy is named after Sir C.V. Raman for his great contribution in physics. It is a spectroscopic technique that is frequently used in solid state chemistry and bio-pharmaceutical industry to analyze the vibrational and rotational modes of the molecules by studying the scattering spectrum that have a different frequency from the incident light. Raman scattering occurs when the incident laser light illuminates the molecule and interacts with the bonds in the molecule. The most intensive (about a few thousandths of the intensity of the incident light) signals can be observed are from Rayleigh scattering, which has the same frequency as the incoming light (υ0). The scattering light with higher or lower frequency than υ0 can also be observed and they are designated as the Raman spectrum. The smaller frequency than the incident light is known as the Stokes shift (υ0 - υ1), and the larger frequency than the incident light is defined as the anti-Stokes shift (υ0 + υ1). Stokes shift can be observed when the molecules absorb photons with a frequency of υ0 during the inelastic scattering and then emit photons with a frequency of (υ0 - υ1). The absorbed energy is higher than the emitted energy and meanwhile the molecule transits from the ground electronic state to an excited electronic state. Anti-Stokes is the opposite. The emission energy is higher than the absorption energy and meanwhile the molecule transits from an excited electronic state to the ground electronic state.

The Raman measurements in Chapter 3 in this thesis were performed by Renishaw inVia Micro- Raman microscope. Two same Notch filters (Model NF03-514E-25, Laser 2000 Benelux CV Co.) 2.3 Characterization tools 29

were equipped in our Micro-Raman spectra setup to fully filter out the excitation wavelength, so that both the Stokes and Anti-Stokes shift can be observed at the same time.

The Si NCs were illuminated by the visible light (λ = 514 nm) from an argon (Ar) ion laser, and the intensity of the laser power can be varied by switching lens of various magnifications (e.g., 5×, 10×, 20×, 50×, 100×). A photo of this setup located in the PVMD lab is shown in Figure 2.10.

Figure 2.10 Renishaw inVia Micro-Raman microscope utilized in Chapter 3 of this thesis.

2.3.5 X-ray photoelectron spectroscopy X-ray photoelectron spectroscopy (XPS), sometimes called as Electron Spectroscopy for Chemical Analysis (ESCA), is a spectroscopic technique to characterize the elemental composition at the surface of a material. The basic principle of XPS is that the valence electrons or the inner-shell electrons of an atom or molecular can be emitted due to the excitation by the intensive focused beam of X-ray in high vacuum (10-8 mbar). The emitted electrons are excited by the photons, therefore can be called as photoelectrons. These photoelectrons are focused by electron collection lens and then detected by an electron energy analyzer. A spectrum figure can be plotted using the number of detected photoelectrons per second as the y-axis, and binding energy of the photoelectrons as the x-axis. Binding energy means the energy required to disassemble a whole system into separate parts, and can be determined as

EEEbinding photon () kinetic  (2.1)

Here Ephoton is the energy of photons, i.e., hυ. Ekinetic is the kinetic energy and Φ is the work function.

30 2. Processing and characterization for Si NCs, solar cells & photoelectrodes

XPS was employed to characterize the carbon concentration in the a-SiC:H films in Chapter 5 and the platinum elements in Chapter 7 of this thesis. A photo of the XPS setup based in the Molecular Materials Research Center (MMRC) at Caltech is shown in Figure 2.11.

Figure 2.11 A photo of the XPS setup located in MMRC at Caltech.

2.3.6 Photoelectrochemical measurement Since last decade, there is intensive debate in the solar fuel society about how to define the solar- to-hydrogen efficiency and measure the actual performance of a photoelectrode. The main reason is that the photoelectrochemical (PEC) measurement is sensitive to many relevant parameters that are not fully understood.82

(a) (b) Figure 2.12 Different methods to measure the j-V curves of a photoelectrode. A two-electrode setup with a working electrode (WE) and counter electrode (CE) (a); and a three-electrode setup with a WE, CE and a reference electrode (RE) (b). 2.3 Characterization tools 31

From Chapter 4 to 7, the PEC j-V curve measurements were done in either a two-electrode or three-electrode setup and the sketches are depicted in Figure 2.12.

In the two-electrode setup, as depicted by Figure 2.12 (a), the current flowing through the electrolyte (or between WE and CE) is monitored as a function of the voltage in the EG&G potentiostat. A solar-to-hydrogen conversion efficiency can therefore be determined by measuring the j-V curve in the two-electrode setup, when the WE is immersed in the electrolyte and illuminated by the AM 1.5 simulated solar spectrum (Newport Sol3A Class AAA solar simulator).

The difference between the above two configurations is that a reference electrode (RE) is introduced in the circuit in the three-electrode setup, as shown by Figure 2.12 (b). In our case we utilize the most commonly used RE, which is a silver electrode in silver chloride solution (Ag/AgCl). The working electrode (WE) means the photoelectrode, and the counter electrode (CE) is usually a coiled platinum (Pt) coil for the other half of redox reaction. The influence of the CE can be excluded by the introduction of RE, and only the performance of the WE can be evaluated, which can be designated as the photoelectrode efficiency. The measured potential is V vs. RE (i.e., Ag/AgCl), and can be converted into the V vs. reversible hydrogen electrode (RHE) using Equation (2.2) to exclude the influence of the pH value in the electrolyte:82

0 VVVRHE Ag/AgCl  Ag/AgCl vs . SHE 0.059  pH (2.2)

0 Where VAg/AgCl is the measured voltage, V Ag/AgCl vs. SHE is the potential of the Ag/AgCl RE with respect to the standard hydrogen electrode, and the value is 0.199 V at 25 °C. The pH value factor of 0.059 comes from the Nernst equations.82 The potential of the WE is plotted against the RHE as the x-axis, and the current density is the y-axis, which represents the amount of hydrogen generated by the illuminated photoelectrode under a certain voltage.

32 2. Processing and characterization for Si NCs, solar cells & photoelectrodes

3. Raman study of laser induced heating effects in free-standing silicon nanocrystals1

This chapter demonstrates that free-standing silicon nanocrystals (Si NCs) have significant different thermal conductivity properties compared to Si NCs embedded in a host matrix. The temperatures of Si NCs under laser illumination have been determined by measuring the ratio of the Anti-Stokes-to-Stokes intensities of the first order Si-Si transverse optical (TO) phonon mode. It is found that the large free-standing Si NCs are easily heated up to ~953 K by the laser light. The laser heating effects are reversible to a large extent, however the nature of the free-standing Si NCs is slightly modified after intensive illumination. The free-standing Si NCs can even be easily melted when exposed to a well-focused laser beam. Under these conditions, the blackbody radiation of the heated Si NCs starts to compete with the detected Raman signals. A simplified model of the heating effects is proposed to study the size dependence of the heated free-standing Si NCs with increasing laser power. It is concluded that the huge red-shift of the Si-Si TO mode observed under intensive laser illumination originates from the laser induced heating effects. In contrast, under similar illumination conditions the Si NCs embedded in matrices are hardly heated due to better thermal conductivity.

1 This chapter has been submitted: L. Han, M. Zeman and A.H.M. Smets, Nanoscale, 2014 (under review). 34 3. Raman study of laser induced heating effects in free-standing Si NCs

3.1 Introduction

Nanocrystals (NCs), also called as quantum dots (QDs), exhibit unique physical, mechanical and electrical properties since their excitons are confined in all three spatial dimensions.103, 104 NCs made of variety of direct and indirect semiconductor materials, have promising applications in the novel designs of light emitting diodes (LED),105 batteries,106 solar cells84, 107 and water-splitting devices.108 For example, NCs might open routes to new photovoltaic (PV) concepts conquering the Shockley-Queisser limit20 of single-junction solar cell devices, using mechanisms like multiple exciton generation (MEG)109 and down conversion by space separated quantum cutting (SSQC).61 In this contribution we focus on NCs based on silicon (Si), which is the most dominant material in semiconductor industry due to its abundance, relatively low cost processing and resistance against water.110-113 These Si NCs can be either free-standing or embedded in a host matrix, such as amorphous silicon carbide (a-SiC:H), amorphous silicon (a-Si:H) and amorphous silicon oxide (a-SiOx:H).

The bandgap of the Si NCs has a strong dependence on the size of NCs. However, the characterization of the size distribution of the ensembles of Si NCs is a challenge. For instance, scanning electron microscopy (SEM) is a direct way to observe the morphology of either the surface or the cross-section of ensembles of NCs. However, the extent of magnification is limited to establish a clear observation of a single NC with a diameter of a few nanometers (nm). High resolution transmission electron microscopy (HR-TEM) has a higher resolution and can even be utilized to investigate the lattice in the grains,84, 107 but it’s more expensive in maintenance and more complex in preparing the samples for characterization. Especially when the Si NCs are embedded in a solid matrix, the quality of the final images is very sensitive to the ion-beam- thinning technique during the sample preparation. Photoluminescence (PL) is also a popular way to estimate the diameter (d) of small (2 nm < d < 10 nm) Si NCs.114 This technique monitors the radiative recombination of light-excited charge carriers to determine the bandgap energy and consequently the average size of the NCs. A shortcoming of PL on NCs is that indirect bandgap materials have a poor radiative recombination and that the Shockley-Read-Hall (SRH) recombination at defect-rich surfaces or interfaces competes with radiative recombination.115-117 These effects result in a rather low PL intensity from indirect bandgap materials such as Si, compared to direct bandgap semiconductors in the III-V group.118, 119 X-Ray diffraction (XRD) is another good tool for the size estimation according to Scherrer formula,120 but it only works accurate on highly crystallized small NCs.

In this work, we focus on Raman spectroscopy (RS), which can be assumed as a comparatively cheap, fast and easy-to-operate characterization tool utilized for monitoring the crystallinity of the material. The mono-crystalline silicon (c-Si) wafer has a first-order Si-Si peak at 0 = 520.5 -1 121 cm , and this transverse optical (TO) phonon mode () can red-shift to lower wavenumbers in Si NCs due to two different physical principles. The first principle is based on a red-shift of the TO mode frequency  to about 516-519 cm-1 for the Si NCs due to the quantum confinement effect of small Si NCs (d < 10 nm, defined as “small” in the rest of this chapter).122 The average size of Si NCs can be calculated using this red-shift ∆ = | - 0|. However, large Si NCs (d > 10 3.2 Experimental 35

nm, defined as “large” in the rest of this chapter) do not experience any significant quantum confinement effect and the first-order Si-Si peaks in Raman spectra are close to the bulk value of -1 0 = 520.5 cm . Therefore the first-order TO mode cannot be used to determine the sizes of large NCs. The second physical principle for a red-shift of the 520.5 cm-1 peak is due to lattice expansion, for instance, induced by the thermal heating of Si NCs when the Raman spectroscopy is performed under intensive laser illumination. To reliably determine the size of the NCs, it is important to establish which of these physical principles are dominating the Raman spectrum for Si NCs embedded in a host material and free-standing Si NCs.

A detailed study is presented in this chapter which demonstrates that depending on the measurement conditions (laser power) and types of sample (free-standing versus embedded in host material) both physical principles can play an important role. Laser heating effects are observed in free-standing Si NCs, which have relatively poor thermal conductivity properties. It will be shown that this shift for the first-order Si-Si phonon line can be as huge as ~25.7 cm-1 ( ~494.8 cm-1) and a width broadening as large as ~12.7 cm-1. The temperature of free-standing Si NCs under intense Raman laser illumination conditions is determined using the measured ratio of Anti-Stokes-to-Stokes (ASt-to-St) TO mode intensities. The temperature of Si NCs embedded in various types of matrices are determined as well, and they demonstrate moderate heating due to their good thermal conductivity. When the free-standing Si NCs are further heated, the thermal loss mechanism is dominated by blackbody radiation, as revealed by the detected background signals in the Raman spectra. The temperature dependence of the free-standing Si NCs on the illumination intensity allows us to successfully estimate the average size of free-standing Si NCs, as supported by HR-TEM analysis.

3.2 Experimental

3.2.1 Synthesis of Si NCs The free-standing Si NCs studied in this chapter were synthesized using the expanding thermal plasma chemical vapor deposition (ETP-CVD) technique, which has the advantages of incredible high production yield, high deposition rate, fabrication at room temperature, low cost, high purity and potential to integrate post-surface passivation treatment based on plasma processing.90, 91 The size of NCs is mainly determined by the residence time of Si species in the plasma downstream from the silane (SiH4) injection ring to the substrate, and subsequently depends on parameters such as chamber pressure, gas flow, plasma power, etc.90, 91

In this chapter, three types of samples of Si NCs embedded in a host material have been studied as well: Si NCs in a hydrogenated amorphous silicon carbide (a-SiC:H) matrix, Si NCs in a hydrogenated amorphous silicon (a-Si:H) matrix, generally referred to as hydrogenated nano- crystalline silicon (nc-Si:H), and Si NCs in a hydrogenated silicon-oxide (a-SiOx:H), often referred to as nano-crystalline silicon-oxide (nc-SiOx:H). Plasma-enhanced chemical vapor deposition (PECVD) was utilized to fabricate these three types of samples. Si rich a-SiC:H thin film was deposited and Si NCs were formed in the a-SiC:H matrix after a 1000 °C post annealing treatment for 1 hour in nitrogen (N2). In contrast, nc-Si:H were deposited directly by PECVD 36 3. Raman study of laser induced heating effects in free-standing Si NCs

using very high frequency (VHF, 40.68 MHz) power of 40 W (over an area 10 cm × 10 cm) at a low substrate temperature of 200 °C. The SiH4 gas is typically highly diluted by hydrogen (H2) gas, which initiates the formation of crystalline grains during the film growth. The nc-SiOx:H films were directly deposited by PECVD using radio frequency (RF, 13.56 MHz) power of 10 W at a low substrate temperature of 200 °C. The gas mixture is similar to the nc-Si:H conditions with an additional dilution of carbon dioxide (CO2) gas, which results in an a-SiOx:H tissue around the Si NCs.

3.2.2 Morphology of Si NC films A photograph of the free-standing Si NCs material on a piece of Corning glass substrate is shown in Figure 3.1 (a), from which we can see that the deposited brown-colored material is quite homogeneous on a scale of several centimeters (cm). The porous films of Si NCs were analyzed by SEM. Figure 3.1 (b) shows a cross-sectional image of the Si NCs film as thick as ~10 μm. Note that this layer was deposited by running the ETP-CVD for 1 second (s). Figures 3.1 (c) and (d) are the top-view images of the film with low and high magnification respectively, and show that the brown powder has a high porosity.

(a) (b)

Figure 3.1 A photograph of film of the Si NCs material (a), SEM images of the cross-section (b) for the estimation of deposition rate, and from the top-view with a smaller (c) and a larger (d) magnification.

HR-TEM was employed to observe the size and morphology of the NCs at larger magnifications. The Si powder was peeled off from the substrate and distributed in the ethanol solution. Then the Si NCs suspension was dropped onto a copper grid pre-coated with a holey carbon film. The copper grid was transferred into the JEM-2100F filed emission HR-TEM with an operating voltage of 200 kV. The HR-TEM images show two typical sizes for the Si NCs, i.e. small Si NCs 3.2 Experimental 37

in Figure 3.2 (a) and large Si NCs in Figure 3.2 (b). In the HR-TEM image in Figure 3.2 (a), many dense dots are shown clearly in the field of view. These small grains are no other than Si NCs, as apparently shown by the contrast difference between Si NCs and the background. This shows that small Si NCs were formed and crystalized in the hot plasma. Well-isolated Si NCs with clear lattice fringes can be observed, while some surrounding Si tissue without lattice fringes remain in the amorphous state. It can be estimated that most of the small Si NCs have a diameter of 3-4 nm and the local density in the HR-TEM image is 4.0×1018 cm-3. Notice that these small NCs were collected on the substrate spots which are exactly below the holes in the SiH4 injection ring. Therefore we believe that these small NCs were formed in the plasma beam and directly deposited on the substrate. However, free-standing larger NCs (diameter of dozens of nanometers) are also observed, as shown in Figure 2 (b). These large NCs were collected from the other areas on the substrate. They have a larger average diameter due to the recycling from the back ground volume into the plasma beam.90, 91 The large NCs are regular spheres, with native silicon oxide coating around it. Summarized, this bimodal size distribution from a typical sample is a result of the NCs deposited directly from the plasma beam (2 nm < d < 10 nm) and recycling of NCs from the background volume into the plasma beam (20 nm < d < 50 nm).90, 91 By varying the amount of SiH4 gas flow from 230, 420 to 600 sccm (standard cubic centimeter per minute, other depositions parameters were fixed), 3 samples of free-standing Si NCs were deposited with different size distributions. The HR-TEM analysis confirmed that these sample have an average size for the large NCs of 27.2±6.4 nm, 34.7±16.2 nm and 41.3±19.0 nm, respectively. The bimodal size distribution makes the ensembles of Si NCs studied in this paper unique in reference to typical Si NCs studied to literature that have sizes smaller than 12 nm.118, 123-128

(a) (b)

(a) 5 nm (b) 50 nm

Figure 3.2 HR-TEM images of the small Si NCs embedded in an a-Si:H matrix collected from the substrate below the holes in the SiH4 injection ring in ETP-CVD (a) and the large free- standing Si NCs collected from the other areas of the sample (b).

The morphology of nc-Si:H films deposited by PECVD is similar to the small NCs in Figure 3.2 (a), and these Si NCs are also embedded in an a-Si:H matrix. The Si NCs in an a-SiC:H matrix or 38 3. Raman study of laser induced heating effects in free-standing Si NCs

nc-SiOx:H look similar as the nc-Si:H material, but they are embedded in the a-SiC:H and a- 129 SiOx:H matrix respectively.

3.2.3 Raman laser heating of Si NCs Raman spectroscopy was performed by Renishaw inVia Micro-Raman microscope immediately after unloading the Si NCs sample from the ETP-CVD chamber. The 100× lens sharply focused the visible light (λ = 514 nm) from an argon (Ar) ion laser on the samples. The estimated beam diameter of the spot on the sample is ~4 μm. The intensity of the laser power can be varied by switching between filters having transmission (e.g., 1%, 5%, 10%, 50% and 100%). The laser power was measured using a handheld power meter (±8% accuracy). Under optimal focused conditions, the penetration depth in bulk c-Si of the light from the Ar ion laser is ~762 nm. Consequently, the penetration depth is deeper than 762 nm in porous samples of free-standing NCs.

A conventional Raman setup is equipped with high-pass filters, which only allows the -1 measurement of the Stokes (St) peak and blocks the anti-Stokes (ASt) peak 0,ASt = -520.5 cm . However, the amplitude of the ASt TO mode is sensitive to the temperature of the NCs.126, 130 To determine the temperature of the Si NCs, the ASt peak should be measured as well. Sharp band- stop filters at the excitation wavelength (514 nm in this work) are used to simultaneously measure the ASt and St peaks. Therefore, two same Notch filters (Model NF03-514E-25, Laser 2000 Benelux CV Co.) were installed in our Renishaw Micro-Raman setup to fully filter out the excitation wavelength.

3.3 Results and discussions

Under mild laser illumination conditions, the first-order Si-Si mode peak of small Si NCs (2 nm < d < 10 nm) in the Raman spectrum is lower by a few wavenumbers in reference to the number -1 84, 121, 131-133 observed for bulk c-Si 0 = 520.5 cm , as observed by many others. In general this observation is attributed to phonon confinement, due to the presence of small NCs embedded in the hydrogenated amorphous Si (a-Si:H) environment (Figure 3.2 (a)). Assuming that this shift can be fully attributed to quantum confinement effect, the average small NCs size can be estimated from the anharmonic Raman softening according to this formula134

1 B 2 d  2  (3.1) 

Here, d (nm) stands for the diameter of the average NCs size, and B = 2.24 nm2 cm-1, which is a -1 constant for Si material. ∆ is the absolute wavenumber (cm ) difference between ω0 and the measured ω for Si NCs, i.e., ∆ = |ω0 - ω|. The average size can be determined from Equation 3.1. Tensile stress induced effects were also reported to cause the red-shift.128, 135 Tensile stress can be created if the Si NCs are significantly heated by laser absorption. Therefore, the above 3.3 Results and discussions 39

method works only when Si NCs exposed to a low laser intensity and the materials have a good thermal conductivity to guarantee no significant heating of the NCs.

(a) 69.3 µW, 10 s (b) 346 µW, 10 s

10 µm

Figure 3.3 In-situ photos of Si NCs after 10 s exposure of laser with a power of 69.3 μW (a), the material remains unaffected compared to the as-deposited samples; an affected zone with a diameter of ~5 μm appears in the illumination spot after 10 s exposure of laser with a power of 346 μW (b); the(c) Si 602“liquid” µW, diffuses 10 s larger× 10 after (d)10 s 602 exposure µW, of 100 laser swith a power of 602 μW (c), and the sample keeps unchanged if this is repeated 10 times; a volcano-shaped affected zone is formed after 100 s continuous exposure of laser with a power of 602 μW (d).

In our work, it will be shown that the large free-standing Si NCs can be significantly heated by intense laser illumination, and the quantum confinement effect is no longer the dominant mechanism for the red-shifts of the first-order Si-Si mode peak. The large (20 nm < d < 50 nm) Si NCs are free-standing and therefore they have poor thermal conductivity properties.124 The large NCs are hardly physically interconnected and the heat transfer is ineffective. During the measurements on such free-standing Si NCs with poor thermal conductivity, an interesting huge (∆ω > 10 cm-1) red-shift and even melting of the Si NCs can be observed when they are exposed to an intensive and well-focused laser. Figures 3.3 (a)-(d) are pictures taken by the Philips SPC 900NC PC camera installed above the lens in Renishaw Raman setup. The pictures show a sample after the measurements at four different laser illumination intensities and durations. In the employed measurement procedure, the free-standing Si NCs were intentionally and repeatable heated on the same spot. Figure 3.3 (a) corresponds to the image of the Si NCs after a 10 s exposure to laser with a power of 69.3 μW. The material in the spot size is almost unaffected compared to the as-deposited samples. When the laser power raises to 346 μW, the sample is clearly affected by the exposure to the laser light. An affected zone with a diameter of ~4-6 μm appears in the illumination spot, as illustrated in Figure 3.3 (b). This means that the layer of Si NCs in the spot is heated up to the extent that the Si NCs are melted. This melted Si “liquid” diffuses partly out of the illumination spot. If the laser power further increases to 602 μW, the 40 3. Raman study of laser induced heating effects in free-standing Si NCs

melting zone around the laser spot gets larger for an illumination exposure of 10 s, as shown in Figure 3.3 (c). In the following steps, even if this process is repeated for 10 times, i.e. 100 s in total, the affected zone remains the same size and shape as Figure 3.3 (c). What seems interesting is that if the same spot is continuously illuminated by 602 μW laser for 100 s only once, a volcano-shaped zone with a diameter of ~8 μm circle and a ~15 μm tail are observed, as shown in Figure 3.3 (d). Poborchii et al. reported that this volcano-shape is caused by the “wind” due to the simultaneously presence of a very high temperature gradient and decreasing absorption coefficients.136 However, this does not directly explain the difference between effective intermittent exposure of 100 s resulting in less melting of Si NCs compared to the uninterrupted exposure of 100 s. This indicates that due to the poor thermal conductivity of the free-standing Si NCs, it takes more than 10 s to reach a quasi-thermal equilibrium in the layer of Si NCs. These measurements demonstrate that light absorption by the free-standing Si NCs under the employed laser powers can easily result in temperatures as high as the melting point of Si NCs.

At the first sight, it might be surprising that these Si NCs can melt (melting point of c-Si is 1687 K at standard test conditions) when they are exposed to a power that is less than 1 mW. However, if we assume the laser power of 602 μW is equally distributed in a spot with a dL ≈ 4 μm diameter, the average power density is 4.79 × 108 W m-2, which is equal to 47900 Suns in air mass 1.5 (AM 1.5, 1000 W m-2) conditions!

In order to estimate the temperature of the hot large Si NCs before melting under intensive illumination, we heated the Si NCs sample step by step by increasing the Ar ion laser power from 3.3 to 16.5, 33, 165 and 330 μW. Figure 3.4 (a) shows that St TO mode peak at 520.5 cm-1 red- shifts to 518, 517, 500 and even 495 cm-1 respectively. In addition, the full width at half maximum (FWHM) is widened greatly from 8.8 to 8.9, 9.5, 17.1 and 21.5 cm-1 respectively. The amplitude of the measured Ast TO mode at -520.5 cm-1 increases significantly with the enhancement of the laser power. To be more precise, both the ASt and St peaks increase significantly with laser power, however the ASt peaks relatively increases more than the St peaks. The ASt peaks show a red-shift with the increasing laser power as well. To quantify the heating effects by the absorption of the laser light, we use the ratio of the ASt-to-St TO mode intensities to determine the temperature of the Si NCs.137 According to the studies reported in literature,126, 127, 130 the temperature of Si NCs can be estimated by

IEASt R  exp (3.2) ISt  kT

Here, IASt/ISt is the ratio of the Anti-Stokes-to-Stokes (ASt-to-St) TO mode intensities, k is the -7 Boltzmann constant, ER is the phonon energy (i.e., ER = hc/λexcitation - hc/λemission = hc·ω·10 ), h and c are the Planck constant and light speed, respectively. ω is the Raman peak shift frequency expressed in cm-1. T, the temperature of the Si NCs during the laser heating, can then be determined from Equation 3.2.

3.3 Results and discussions 41

The results in Figure 3.4 (a) reflect that the ratio of IASt/ISt increases with laser power, which means the large Si NCs were heated up by the absorption of the laser energy. The power of the probe laser is indicated on the left side of each curve. On the right side, the calculated temperature using Equation 3.2 is presented. The results show that the free-standing Si NCs can be heated from room temperature to as high as ~953 K, when the laser power increases 100 times from 3.3 to 330 μW (~263-26300 Suns). By further analyzing the data shown in Figure 3.4 (a), we find that the temperature of the illuminated free-standing Si NCs increases with the wavenumber enhancement of the red-shifts ∆ω, and simultaneously with the peak widening of the TO mode St peak as well as shown by the FWHM enhancement in Figure 3.4 (b). This demonstrates that both ∆ω and FWHM are determined by the temperature of the Si NCs, and agrees with the findings in literature.123, 138, 139

Laser power P Temperature T

St 330W ASt 953K

(a.u.)

165W 762K

33W 327K

16.5W 306K  Raman intensity intensity Raman  0,St 3.3W 0,ASt 293K

-600 -550 -500 -450 450 500 550 600 (a) -1 Raman shift (cm )

30 30

25 25

)

20 20 -1

)

-1 15 15

(cm

(cm 10 10   5 5 FWHM

0 0

200 400 600 800 1000 (b) Temperature T (K)

Figure 3.4 Anti-Stokes and Stokes TO mode peaks shift in the Raman spectra of the larger free- standing Si NCs with increasing Raman laser illumination (a). The TO mode red-shift and the full width at half maximum (FWHM) of the TO mode of Stokes peak widening with the temperature enhancement (b). The dashed lines are guides for the eyes. 42 3. Raman study of laser induced heating effects in free-standing Si NCs

The question that remains is whether large Si NCs embedded in a host matrix can be significantly heated under the same illumination conditions as well? Therefore, the same laser heating Raman spectroscopy was done on the samples of Si NCs embedded in a matrix, such as Si NCs in a- SiC:H, nc-Si:H and nc-SiOx:H. Figure 3.5 shows the temperatures of the Si based materials when heated by the Ar ion laser. As shown the free-standing Si NCs can be heated to temperatures up to ~953 K. However, it is not easy to heat the other materials when they are embedded in a matrix. It demonstrates that the interconnection between the surfaces of the large Si NCs and the tissue of the host matrix results in good thermal conductivity. For example, Si NCs in an a-SiC:H matrix material (blue triangles in Figure 3.5) can be heated up to ~375 K illuminated by 330 μW. This temperature is slightly higher than for the Si NCs in nc-Si:H and nc-SiOx:H.

1000 free-standing Si NCs 800

(K)

600

400 Si NCs in a-SiC:H nc-Si:H

Temperature nc-SiO :H x 200 0 100 200 300 400 500 Laser power P (W)

Figure 3.5 Laser heating effects of several Si NCs based materials. Free-standing Si NCs are much easier to be heated than Si NCs embedded in various matrices illuminated by intensive laser. The lines are guides for the eyes.

This comparison shows the unique properties of the large free-standing Si NCs. The fact that they are hardly physically interconnected limits the heat transfer between the particles. Another possible route for the free-standing particle to lose its heat is through the blackbody radiation.125, 140 We have observed signatures of such blackbody radiation from hot free-standing Si NCs. Figure 3.6 shows the Raman spectra of a sample of free-standing Si NCs under large laser powers of 330 or even 660 μW. The presence of the St peaks in the spectra shows that the sample still contains solid Si NCs. However, the sample might contain liquid Si phase as well. Furthermore, the red-shift is enormous ∆ ~30 cm-1 (i.e.,  ~490 cm-1). The background signals in the Raman spectra have a large tilt that even overrules the amplitude of the measured St peaks. Therefore it is not possible any more to accurately determine the temperature using Equation 3.2.

3.3 Results and discussions 43

(a.u.)

I Planckian spectral Laser power P radiance fitting 330 W T = 1015  K

T = 1005 5 K 660 W

Raman intensity

200 400 600 800 1000 Raman shift (cm-1)

Figure 3.6 Raman spectra of the large free-standing Si NCs under intensive (I = 330 and 660 μW) laser power.

We claim that the background tilt reflects blackbody radiation of the large Si NCs. The temperature of the large Si NCs gets so high that the intensity of blackbody radiation in the spectral range around the Raman laser 519 – 542 nm (corresponds to 200 cm-1 ≤  ≤ 1000 cm-1) is similar or higher than that of the St Raman peak. According to Planck’s Law, the blackbody radiation of an object can be described as

1 25 hc BT ( ) 2 hc  exp( ) 1 (3.3) kTB where B denotes the wavelength-dependent blackbody spectral radiance at a certain temperature, h the Planck constant, c the speed of light in the medium and kB the Boltzmann constant. λ is the wavelength (nm) and can be transformed into wavenumber (cm-1) in Figure 3.6. The green and orange dashed curves in Figure 3.6 are the fitting curves of Equation 3.3 multiplied by a coefficient that can compensate the difference between the radiance value and the Raman intensity counts. It is interesting to observe that the blackbody radiation curves fit the slopes in the Raman intensity background (excluding the St TO mode peaks) well. The simulations of the Planck’s Law give temperatures in the order of 1015±5 and 1005±5 K under the 330 and 660 µW laser illumination, respectively. It has to be noted that the laser light has a limited penetration depth in to the porous silicon film. As a result, the large Si NCs at the surface will be heated more than the ones sub-surface within the penetration depth. Therefore the measured temperature should be interpreted as an average temperature over the temperature gradient. Nevertheless, the average temperature is close to the temperature of ~953 K before significant melting of Si NCs occurs as can be concluded from Figures 3.3 and 3.4. This implies that significant fraction of Si NCs can be melted at extreme intensive illumination conditions. The found temperatures of 1005- 1015 K indicate that the melting temperature of free-standing Si NCs is much lower than for bulk c-Si (1687 K at standard test conditions). These melting temperatures of free-standing Si NCs are 44 3. Raman study of laser induced heating effects in free-standing Si NCs

close to values reported in various other studies.141-143 An additional effect is that the melted free- standing Si NCs aggregate and as a consequence they significantly improve the thermal conductivity. In the presence of a liquid Si phase, the NCs cannot be heated up much more. Although the laser intensity is doubled from 330 to 660 µW in Figure 3.6, the temperature of the mixed phase of Si liquid and NCs does not increase anymore.

Another experiment was performed to test whether the Raman laser heating effects for large free- standing Si NCs were reversible and whether the morphology of the NCs was modified after the laser illumination. After the step-by-step increasing of the laser power on the same spot on the sample (results shown in Figure 3.4), the laser power was decreased step by step. The temperature at every laser power (exposure time of 10 s) was determined again by the measured ratio of ASt- to-St TO mode intensities using Equation 3.2. The results are presented in Figure 3.7. The TO mode positions and shapes during the “cooling” phase (data not shown here) are very similar to those illustrated in Figure 3.4 at the same illumination conditions. This means that after reaching the highest temperature not all Si is melted and Si NCs are still present in the layer. It can be noticed that during the “cooling” phase, the Si material can still get reasonably heated. However, it is also exemplified that the NCs left at the same spot in the “cooling” phase cannot reach the same temperature as in the initial “heating” phase. At an illumination power of 165 µW the temperature of the Si NCs in the “heating” phase is ~762 K, while in the “cooling” phase it is significantly lower, ~600 K. This temperature difference implies that the thermal conductivity of the ensemble of free-standing Si NCs has improved after going through the “heating” phase. A possible explanation could be that the aggregation of Si NCs or some melting of Si NCs has improved the physical interconnection of the NCs. We can conclude that the laser heating effects are to a large extent reversible, but the nature of the porous Si NCs based layer gets slightly modified under intensive laser induced heating.

1000

(K) 800

T Cooling

600

400 Heating

Temperature

200 0 100 200 300 400 Laser power P (W)

Figure 3.7 Temperature comparison between the “heating” and “cooling” phases on the same spot of Si NCs materials.

3.3 Results and discussions 45

So far, we have demonstrated that large free-standing Si NCs can be significantly heated by laser illumination due to the poor thermal conductivity of these layers. In addition, we have observed that one of the heat loss mechanisms is by blackbody radiation. If blackbody radiation would be the dominant mechanism for free-standing NCs to lose its thermal energy, the equilibrium temperature under laser illumination should depend on the size. In the final part of this chapter, we will study whether this is indeed the case. A simplified model is established for this study. We assume that thermal conductivity between the neighboring Si NCs can be neglected and the laser penetration depth in c-Si (~762 nm) is much larger than the typical average diameter (d) of the Si NCs (20 < d < 50 nm). The laser intensity I is the laser power P divided by the laser spot size on 2 the sample I ~ P/(πdL /4) with dL ≈ 4 μm the diameter of the spot size. The power absorbed by one single NCs (Pabs) is proportional to the NC volume and can be estimated by Pabs ~ 3 2 3 (πd /6)·α·P/(πdL /4) ~ Pd with α the absorption coefficient of Si at 514 nm. The radiation at thermal equilibrium from one single NC (Pirr) is estimated by the Stefan-Boltzmann Law Pirr ~ σT4·πd2 ~ T4d2, where σ stands for the Stefan-Boltzmann constant. In thermal equilibrium, the thermal generation and thermal losses of the Si NCs are equal, i.e., Pabs ≈ Pirr. Under these conditions T, d, and I are related for each specific sample like:

T~ d1/4 P 1/4 ( T , d ) (3.4)

700 41.3 nm 34.7 nm 27.2 nm

600

(K)

T 500

400

300

Temperature

1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 Laser power P1/4 (W1/4)

Figure 3.8 Temperatures determined from the Raman laser heating effects show a dependence on the average sizes of free-standing Si NCs. T ~ P1/4 fitting of each sample is illustrated by the dashed curves.

So in the case that blackbody radiation is dominant, we can expect a relation between T and P given by Equation 3.4. We have measured this dependence for three samples of free-standing NCs with different average diameters for the large particles d = 27.2±6.4, 34.7±16.2 and 41.3±19.0 nm. This data is presented in Figure 3.8. The results confirm that temperature can be simulated with a T ~ P1/4 dependence (for each specific sample, the average diameter d is a fixed value). This implies that indeed the thermal conductivity of the large Si NCs can be neglected compared to thermal radiation. Furthermore, Equation 3.4 predicts that the slope of the T ~ P1/4 46 3. Raman study of laser induced heating effects in free-standing Si NCs

dependence is increasing with average size of the large Si NCs. This is in agreement with the results shown in Figure 3.8. The larger NCs (sample with d = 41.3 nm) are easier to be heated than the smaller ones (samples with d = 34.7 and 27.2 nm) when the same amount of laser power is used. The trend in Figure 3.8 additionally explains why the small Si NCs (if they would be fully free-standing) do not significantly heat up. Considering small Si NCs with a diameter of 4 nm, the Si NCs would only heat up to ~370 K for the highest laser power depicted in Figure 3.8.

3.4 Conclusions

In this chapter, the laser induced heating effects are studied in bimodal size-distributed Si NCs using Raman spectroscopy. The Si NCs were fabricated by ETP-CVD, and the morphology of NCs were characterized by SEM and HR-TEM. Under low intensity illumination at the wavelength of 514 nm, the typical quantum confinement effects for small Si NCs (diameter of 2- 10 nm) are observed. The TO mode of the first order Si-Si peak red-shifts a few wavenumbers in reference to the bulk c-Si at 520.5 cm-1 due to the quantum confinement effect. The average diameter of these small Si NCs can be estimated according to the amount of this red-shift. However, if the free-standing Si NCs are illuminated by intensive Ar ion laser, the thermal heating effects of the large Si NCs (diameter of 20 -50 nm) become the dominant mechanism for the TO mode shift in the Raman spectrum. A huge red-shift up to ~25.7 cm-1 and a peak widening of ~12.7 cm-1 is observed as a result of the lattice expansion in Si NCs. The melting of large Si NCs illuminated by intensive laser is monitored by the camera installed in the Raman setup. The temperatures of the heated Si NCs are determined using the ratio of Anti-Stokes-to-Stokes TO mode intensities. The large free-standing Si NCs can be heated as hot as ~953 K by a well- focused laser with a power of 330 µW. In contrast, Si NCs in various matrices were fabricated by PECVD. These samples can hardly be heated using the same amount of laser power due to their good thermal conductivity. If the large free-standing Si NCs are further heated, the intensity of the blackbody radiation in Raman spectrum starts to compete with those of the TO mode. When the materials are in the mixed phase of liquid and solid NCs, the temperatures are estimated to be in the range of 1005-1015 K according to Planck’s Law. The laser heating effects are confirmed to be reversible to a large extent, but the nature of the material is slightly modified after intensive laser illumination. A simplified model of the heating effects of large Si NCs is established to study the size dependence of the heated free-standing Si NCs with an increasing laser power.

4. Optimization of double-junction thin-film silicon solar cells for a bismuth vanadate photoanode1

A photoelectrochemical water-splitting device (PEC-WSD) was designed and fabricated based on cobalt-phosphate-catalyzed and tungsten-gradient-doped bismuth vanadate (W:BiVO4) as the photoanode. A simple and cheap hydrogenated amorphous silicon (a-Si:H) double-junction solar cell has been used to provide additional bias. The advantage of using thin-film silicon (TF-Si) based solar cells is that this photovoltaic (PV) technology meets the crucial requirements for the PV component in PEC-WSDs based on W:BiVO4 photoanodes. TF-Si PV devices are stable in aqueous solutions, are manufactured in simple and cheap fabrication processes and their spectral response, voltage and current density show an excellent match with the photoanode. This chapter is mainly focused on the optimization of the TF-Si solar cell in reference to the remaining solar spectrum transmitted through the W:BiVO4 photoanode. The current matching between the top and bottom cells are studied and optimized by varying the thickness of the a-Si:H top cell. We support the experimental optimization of the current balance between the two sub-cells with simulations of the PV devices. In addition, the impact of the light-induced degradation of the a- Si:H double-junction, the so-called Staebler-Wronski effect (SWE), on the performance of the PEC-WSD has been studied. The light soaking experiments on the a-Si:H/a-Si:H double-junctions over 1000 hours show that the efficiency of a stand-alone a-Si:H/a-Si:H double-junction cell is significantly reduced due to the SWE. Nevertheless, the SWE has a significant smaller effect on the performance of the PEC-WSD.

1 This chapter has been published: L. Han, F.F. Abdi, P. Perez Rodriguez, B. Dam, R. van de Krol, M. Zeman and A.H.M. Smets, Phys. Chem. Chem. Phys., 2014, 16, 4220-4229. 48 4. Optimization of 2-j TF-Si solar cells for a BiVO4 photoanode

4.1 PEC-WSDs based on BiVO4

A photoelectrochemical water-splitting device (PEC-WSD) provides an attractive route in utilizing sunlight to split water into hydrogen (H2) and oxygen (O2). H2 can then be directly used as a fuel in a combustion engine, or as an intermediate to form hydrocarbons. Electrodes based on metal-oxide semiconductors have been the focus of many studies as potential PEC-WSDs, mainly due to their stability in aqueous solutions, easy synthesis, and low costs. One of the most 94, 144 promising metal-oxide semiconductors is bismuth vanadate (BiVO4). In the monoclinic 145, 146 phase, BiVO4 is a photoactive n-type semiconductor, and it is stable in aqueous solution 93 with pH values between 3 and 11. Theoretically, BiVO4 can generate a photocurrent of ~7.5 mA cm-2 under an AM 1.5 solar spectrum (1000 W m-2), assuming that all photons with energies 147 higher than the bandgap (2.4 eV ) are absorbed and contribute to the O2 generation at the surface.

Early efforts in employing BiVO4 as the photoanode material, however, have been hampered by low efficiencies, resulting in AM 1.5 photocurrents lower than 1 mA cm-2 at 1.23 V vs. reversible 148-152 hydrogen electrode (RHE). In the last couple of years, BiVO4 performance has been 72, 144, 153-159 improved significantly by modifying the material with O2 evolution catalysts. The remaining technological bottleneck is the substantial additional bias potential that needs to be applied in order to draw a reasonable photocurrent. Reducing or eliminating this bias is a 160 161 challenge for nearly all visible-light absorbing PEC materials, such as BiVO4, α-Fe2O3, WO3 162 -2 and Ta3N5. The demonstration of AM 1.5 photocurrents higher than 2 mA cm for BiVO4 requires an external bias higher than 1 V. This is usually supplied by an external voltage source (e.g. potentiostat), which significantly reduces the overall solar-to-hydrogen conversion efficiency 163 (ηSTH).

An elegant solution to the problem is to employ a Z-scheme configuration (named after the shape of energy diagram), by combining a photoelectrode material with a photovoltaic solar cell.164 In this configuration, the solar cell is placed at the back of the photoelectrode. Using the light transmitted through the photoelectrode, the solar cell can supply the bias voltage needed by the photoelectrode as reported in various publications.165, 166 The solar cells integrated in PEC-WSDs range from dye-sensitized solar cells (DSSCs)167 and GaAs p-n junction solar cells168 to more sophisticated AlGaAs/Si solar cell configurations.169 The choice of PV material is determined by two competing requirements, one related to cost reduction and one related to high conversion efficiencies. Very cheap solar cells usually have lower conversion efficiencies, while the high conversion efficiencies can only be achieved with expensive solar cells.

In a recent publication, we have demonstrated that a W:BiVO4 photoanode powered by an a-Si:H tandem cell could give a ηSTH of 4.9%, which is currently the highest value ever reported for a water-splitting device based on a metal-oxide photoelectrode.170 In this chapter, we will discuss the advantages of using a-Si:H tandem solar cells for PEC-WSDs based on W:BiVO4 photoanodes, and will focus on the optimization of the solar cell in reference to the solar spectrum transmitted to the W:BiVO4 based photoelectrode. We demonstrate that the light-induced degradation of the a-Si:H solar cell, the so-called Staebler-Wronski effect (SWE),171 when 4.2 Why a-Si:H/a-Si:H tandem cells? 49

integrated in a PEC-WSD has a smaller impact on the ηSTH compared to the ηSTH of a stand-alone a-Si:H/a-Si:H double-junction PV device.

4.2 Why a-Si:H/a-Si:H tandem cells?

In this chapter, we are still using platinum (Pt) as the counter electrode in the solution for an efficient H2 evolution. However, in the long term strategy, the expensive Pt counter electrode has to be replaced by a cheaper metal and integrated at the back contact of the solar cell. This requires the photovoltaic cell to be immersed in water, so the first requirement for the PV cell in the PEC- WSD is that it should be stable in aqueous electrolytes. Silicon (Si) has been chosen as it is the most resistant PV material in aqueous environments. This allows the use of cheaper encapsulation materials than more sensitive PV materials, such as chalcogenides and III-V semiconductors. In addition, the same reasons why Si is the most dominant material in the PV industry apply to PEC- WSD as well. Si is earth-abundant, non-toxic, environmentally sustainable, relatively low-cost and widely investigated.

(a) (b) Figure 4.1 The cross-section sketch of the photoelectrochemical water-splitting device (PEC- WSD) demonstrated in this chapter (a). The AM 1.5 spectrum and the photoanode modified spectrum (left axis) and the transmittance spectrum (right axis) (b).

The second requirement is related to the transmittance of the solar spectrum through the photoanode component of the PEC-WSD. The maximum power point (MPP) of a stand-alone PV device is generally optimized in reference to the standard AM 1.5 (1000 W m-2) solar spectrum. In contrast, the PV cell in the PEC-WSD has to be optimized and work efficiently under illumination with AM 1.5 sunlight that is filtered by the components at the front side: the quartz- window of the water tank, electrolyte, catalyst layer, photoanode films and F-doped SnO2 (FTO) substrate of the device (as depicted in Figure 4.1 (a)). This transmitted spectrum (TS), shown by 50 4. Optimization of 2-j TF-Si solar cells for a BiVO4 photoanode

the yellow curve in Figure 4.1 (b), is missing most of the blue spectral region (λ < 450 nm) and contains around 60% of the irradiation of the AM 1.5 spectrum in the red part (λ > 550 nm) (as depicted by the blue dash curve).

The third requirement for the PV cell is that its j-V characteristics should match the j-V curve of the photoanode. In our earlier work, an AM 1.5 photocurrent density (j) of 3.6 mA cm-2 for a 153, cobalt-phosphate-catalyzed (CoPi) BiVO4-based photoanode at 1.23 V vs. RHE was reported. 170 This remarkably high photocurrent for a material deposited using a low-cost and easily scalable spray pyrolysis process was achieved by introducing a gradient dopant profile of tungsten (W) in BiVO4 to form a stack of homo-junctions. This results in an enhanced internal electric field. This field greatly increases the charge separation efficiency, which was shown to be the 94 main limiting factor for the performance of these spray-deposited BiVO4 photoelectrodes. The operating point (OP) of the PEC-WSD is determined by the intersection point of the j-V curve of the solar cells and that of photoanode device. For optimal performance, the OP has to be at a voltage as high as possible in order to get a sufficient bias potential to drive the surface reactions. Since the j of the photoanode levels off at V ≥ 1 V, it is more efficient if the VOP ≥ 1 V. This implies that the PV component needs to deliver an open-circuit voltage VOC ≥ 1.5 V, whereas j has to be matched with that of the photoanode. To obtain PV devices with VOC ≥ 1.5 V, options are limited to the implementation of multi-junction PV devices based on TF-Si or III-V materials. In this chapter we do not consider III-V semiconductor materials as they suffer from both instability in aqueous environments and high costs. Typical materials used in TF-Si junctions are a-Si:H, amorphous silicon-germanium (a-SiGe:H) and nano-crystalline silicon (nc-Si:H). The band gaps of these materials determine the range of VOC that can be achieved in the corresponding junction, i.e. VOC, a-Si:H = 0.8-1.0 V, VOC, a-SiGe:H = 0.5-0.6 V and VOC, nc-Si:H = 0.48-0.6 V. Consequently, in view of the required VOC ≥ 1.5 V for the BiVO4 photoanode, the PV device configurations are limited to an a-Si:/a-Si:H double-junction or a triple-junction devices (e.g., a- Si:H/a-Si:H/a-SiGe:H, a-Si:H/a-SiGe:H/a-SiGe:H, a-Si:H/nc-Si:H/nc-Si:H). The individual junctions in these multi-junction configurations have to deliver a j value of 7.5 mA cm-2 in reference to the TS to achieve the theoretical maximum ηSTH of 9.2%.

In this contribution, we focus on the a-Si:H/a-Si:H double-junction device instead of a triple- junction device. The PV component of PEC-WSD is required to have a VOC as high as possible and a j value as high as possible when illuminated by the photons transmitted through the photoanode. These are competing parameters if we consider the double- and triple-junction. A triple-junction has the advantage of a high VOC, but its short-circuit current density (JSC) is limited when the incident light is filtered by the photoanode. Due to the spectral overlap of the -2 photoanode with the top cell, the triple-junction can only deliver a maximum JSC of 3.8 mA cm theoretically (this value is estimated according to the integration of the transmitted spectrum through the photoanode, assuming all the photons are converted into electron hole pairs and the current density is evenly distributed among all the three junctions). In contrast, the double- junction cell generates a more moderate VOC, but has a much higher maximum theoretical JSC of -2 5.8 mA cm when placed behind the photoanode. Since the VOC is sufficiently high to drive the W:BiVO4 photoanode, the double-junction is the most logical approach. Furthermore, it has several advantages over the triple-junction device. The total layer thickness of the double-junction 4.3 Experimental 51

cell is significantly thinner than that of the triple-junction device based on nc-Si:H material. The total amount of the layers in the double-junction cell is less than that of the triple-junctions and the deposition time is also much shorter. Therefore, it is the cheapest solution for this type of PEC-WSD. In addition, the tandem cell allows much more flexibility than the triple-junction cell for equally distributing the current density over both sub-cells by varying the film thicknesses or bandgap of the both junctions. Furthermore, at applied bias voltages larger than the typical value of redox potential of water ~1.23 V, the current density of the photoanode tends to saturate. The gain in ηSTH by boosting the voltage of the solar cell above 2 V by using an additional thick bottom cell in a triple-junction is therefore minimal. In view of all the above issues, the a-Si:H/a- Si:H device is the most straightforward option to meet the requirement for the PEC-WSD based on CoPi-catalyzed gradient-doped W:BiVO4.

4.3 Experimental

The a-Si:H/a-Si:H solar cells were deposited in a multi-chamber tool equipped with radio frequency plasma-enhanced chemical vapor deposition (RF-PECVD) reactors. Solar cells were deposited on 2.5 cm × 10 cm ASAHI VU-type substrates (~600 nm thick textured FTO on glass), which were kept at temperature of 170 °C during the TF-Si deposition. Before the deposition of TF-Si films, a stripe of 300 nm thick aluminum (Al) was coated near the edge of the ASAHI VU- type substrates using a PROVAC evaporator in a rotation mode. This Al stripe acts as the front contact of the tandem cell.

A boron-doped amorphous silicon carbide (a-SiC:H(B)) layer was deposited as p-layer. A very thin intrinsic a-SiC:H buffer layer was deposited between p- and a-Si:H i-layer to achieve higher values for VOC. The n-layer of the top cell is a single layer of phosphorus-doped nano-crystalline silicon oxide (nc-SiOx:H(P)) whereas the n-layer of the bottom cell is a double layer of nc- SiOx:H(P) and phosphorus-doped hydrogenated amorphous silicon (a-Si:H(P)). The n-type nc- SiOx:H(P) material has the advantage of lower parasitic absorption losses and better reflection performance in reference to the conventional n-type a-Si:H(P) with a higher refractive index (n).

Three layers of Ag/Cr/Al were evaporated on the n-type nc-SiOx:H(P) area as the back contact. The Al layer prevents the oxidation of the Ag layer, whereas the Cr interlayer avoids mixing of Ag and Al in post-deposition anneal treatments. All metal back contacts have an area of 1 cm × 1 cm. The cross-section sketch of the solar cell structure is shown in the bottom part of Figure 4.1 (a).

In this work, the thickness of the a-Si:H i-layer has been varied to optimize the performance of the solar cell. Three a-Si:H/a-Si:H tandem solar cells with different thicknesses of the top i-layer (50 nm, 75 nm, 100 nm) were deposited, while the thickness for the bottom i-layer of 350 nm was kept the same for all three cells. In this chapter we refer to these tandem cells as “thin cell” (i- layer in top cell is 50 nm), “medium cell” (i-layer in top cell is 75 nm) and “thick cell” (i-layer in top cell is 100 nm).

52 4. Optimization of 2-j TF-Si solar cells for a BiVO4 photoanode

The gradient-doped W:BiVO4 was synthesized by spray pyrolysis on a piece of TEC-15 FTO substrate, and the fabrication procedure was reported in our previous publication.170 PEC characterization was performed under a two-electrode configuration in 0.1 M (mol L-1) KPi (pH ~7) aqueous solution as the electrolyte and Pt coil as the counter electrode.

The integrated PEC-WSDs were fabricated with the photoanode on the upper side of the glass substrate and an optimized solar cell at the bottom side of the glass substrate. Using the same substrate for both the spray pyrolysis technique on one side and RF-PECVD on the other side, might lead to cross-contamination of the solar cell and photoanode processing. Therefore, a more practical option is to process the photoanode and PV part on different glass substrates, which are subsequently connected back-to-back using transparent glue at the non-processed sides. The FTO window layer of the solar cell was connected to the FTO of the photoanode using a graphite as illustrated in Figure 4.1 (a). The back contact of the solar cell was connected to the Pt counter electrode.

A PASAN flash AM 1.5 spectrum solar simulator was used to measure the j-V characteristics of solar cells under standard test conditions. Using the flash simulator, a constant temperature of 25 °C on the surface of the solar cells during the measurement was guaranteed. In addition, an accurate illumination area was ensured by using a black mask to rule out possible contribution of contactless areas. During the flash measurement, a reference monitoring photodiode is used to calibrate the average intensity of the AM 1.5 spectrum. The optical thickness and properties of the various films are determined using reflection and transmission measurements on each single layer. These layers were separately processed as single layers on flat glass substrates. The thickness of each film was determined using an optical model based on a Tauc-Lorentz oscillator as defined in SCOUT program.172

The external quantum efficiency (EQE) spectrum of solar cells indicates the fraction of photons at a certain wavelength that results in collected electrons at the contacts. It is not straightforward to measure the EQE of individual top and bottom cells as well as that of the total double-junction solar cells. In this work, the device was illuminated by two different types of LEDs during the EQE measurement, respectively: red (λ = 632 nm) LEDs to saturate the bottom cell and measure the top cell; blue LEDs (λ = 397 nm) to saturate the top cell and measure the bottom cell. A problem in achieving a solely biased top cell or bottom cell is the fact that both junctions have the same absorber layers with in principal the same spectral response. Light-biasing the top cell is straightforward by using blue light, as the blue light is fully absorbed in the top cell. However, the bottom cell has to be biased by red light, which is also partly absorbed by the top cell. The intensity of the red LEDs is increased such that only the bottom cell is close to saturation (so no saturation of the top cell).

In this chapter, we use two types of EQE spectra. The first type is the standard EQEAM 1.5, which is used to determine the JSC, AM 1.5 out of the spectral photon flux ΦAM 1.5 under the standard solar spectrum of AM 1.5 (1000 W m-2):

4.4 Solar cell optimization 53

J e ()()  EQE  d  (4.1) SC, AM 1.5 AM 1.5 AM 1.5

The second type of EQE spectra is referred to as EQETS which represents the spectral response of the PV component in which the optical losses due to the front photoanode and its supporting layers is included:

EQETS()()() T  EQE AM 1.5  (4.2) where T(λ) presents the transmittance of the photoanode component (Figure 4.1 (b)). With

 TS()()() AM 1.5  T  (4.3) this gives the following expression for the short-circuit density JSC, TS of the PV device after the photoanode becomes

J e (  ) EQE (  ) d  SC, TS TS AM 1.5  e  (  ) T (  ) EQE (  ) d  (4.4)  AM 1.5 AM 1.5  e (  ) EQE (  ) d   AM 1.5 TS and it is smaller than the JSC, AM 1.5 value.

The potential of the working electrode (diameter 6 mm) was controlled by a potentiostat (EG&G PAR 283) and an immersed Pt wire was used as the counter electrode. White light photocurrent measurements were performed under simulated AM 1.5 solar illumination (1000 W m-2) with a NEWPORT Sol3A Class AAA solar simulator (type 94023A-SR3). Electrical contact to the sample was made using a silver wire and graphite paste. For the combined PEC-WSD, j was monitored by a digital multimeter (KEITHLEY 2001). It should be pointed out that the photoanode performance is very sensitive to the blue (and UV) spectrum, therefore, the measured j is highly sensitive to the differences in the shape of the blue spectral part between the various solar simulators (xenon vs. tungsten). This requires the solar simulator to be calibrated during the various measurements to minimize the possible spectrum mismatching with AM 1.5 spectrum. During the j-t curve measurement of the PEC-WSD, the external power applied to the Class AAA solar simulator is real-time controlled to guarantee an irradiance equal to standard test conditions (1000 W m-2).

4.4 Solar cell optimization

In this section, we optimize the current matching of the a-Si:H/a-Si:H double-junctions and consequently the ηSTH of the PEC-WSD. First, as a reference, we compare the EQEAM 1.5 of the a- Si:H/a-Si:H double-junctions with a single-junction a-Si:H solar cell with an i-layer thickness of 54 4. Optimization of 2-j TF-Si solar cells for a BiVO4 photoanode

350 nm (corresponding to the bottom cell of the a-Si:H/a-Si:H double-junction). The EQEAM 1.5 of the single-junction a-Si:H solar cell shows that it utilizes the AM 1.5 solar spectrum in the range of 300 nm up to 800 nm (blue curve in Figure 4.2 (a)) better than that of the double-junctions. In the a-Si:H/a-Si:H double-junction structure, the spectral utilization of the AM 1.5 solar spectrum is distributed over the top cell (300 nm < λabs < 750 nm) and the thicker (350 nm) bottom cell (400 nm < λabs < 800 nm). The sum of the top and the bottom cell results in a slightly smaller EQEAM 1.5 in the spectral range of λ > 450 nm in reference to the single-junction solar cell. This lower response is explained as the additional doped layers between the top and bottom junctions, which are logically lacking in the single-junction. These layers result in additional parasitic losses and reflection losses back in to the top cell due to the nc-SiOx:H n-layer with low refractive index values. The double-junction with the highest total spectral utilization is the thick cell (green curves in Figure 4.2 (a)), which shows a slight enhancement of the EQE in the 400-550 nm range. This indicates that a thick top cell leads to a reduction of the parasitic absorption losses in the n- doped and p-doped layer between the top and bottom cell. The blue response (300-450 nm) shows a slightly lower EQEAM 1.5 for the single-junction compared to the double-junctions. Both configurations have the same p-layer, so the difference in blue response is not based on parasitic absorption losses in the supporting layers.

It is important to note that we have ruled out the possibility of an artefact of the EQE measurement approach. In practice, by light biasing one of the sub-cells in the double-junction under V = 0 V conditions, the light-biased sub-cell puts a reverse bias of approximately -0.65 V (an estimated voltage between VOC and VMPP, TS of one a-Si:H sub-cell) on the non-light-biased cell. To study the extent of this effect, the EQE spectral response of the single-junction solar cell was measured under an intentionally reverse bias of -0.65 V as well. As illustrated in Figure 4.2 (a) (magenta colored curve), the EQE curve somewhat increases in the whole measured wavelength range, but its blue response in the λ < 450 nm range is still lower than for the double- junction. Consequently, the origin of the improved collection in the blue spectrum is believed to be a result of the larger internal electric field as a result of the much thinner top cells in the double-junctions in reference to the rather thick single-junction. The thicker single-junction results in a smaller internal electric field and less effective collection of the close to the p/i interface photo-excited charge carriers. The single-junction under reverse bias results in small -2 increase in the current of ΔJSC = 0.3 mA cm . This implies that the determined JSC of sub-cells in the tandem cell using EQE and light-biasing can lead to a maximum systematic overestimation of the current of ΔJSC /JSC = 1.7%, due to mutual voltage biasing between the sub-cells.

4.4 Solar cell optimization 55

1.0 top i-layer AM 1.5 50nm 0.8 75nm sum 100nm

0.6 top single junction

(-) -0.65V biased

0.4 bottom EQE single junction

0.2

0.0 300 400 500 600 700 800  (nm) (a)

1.0 -2 TS Jsc (mA cm ) AM 1.5TS top i-layer top bottom 0.8 50nm 8.172.87 8.114.54 75nm 9.413.42 7.324.15 100nm 10.794.11 6.193.56 0.6

(-) sum

0.4

EQE

bottom 0.2 top

0.0 300 400 500 600 700 800  (nm) (b) Figure 4.2 EQE spectra in the series of the top i-layer thickness (black: thin cell, red: medium cell, green: thick cell), compared with that of single-junction a-Si:H solar cell (blue and magenta) (a); EQE spectra of the three tandem PV cells calculated under the transmitted spectrum through the photoanode (b).

Next, the spectral matching of the double-junctions with the PEC-WSD is analyzed. The EQETS of the three devices in reference to the spectrum transmitted through the photoanode structure CoPi/BiVO4/FTO/glass is determined by measuring EQEAM 1.5(λ) and T(λ) of the photoanode (using Equation 4.2). The resulting EQETS are shown in Figure 4.2 (b). The same EQETS(λ) spectra have been directly measured by placing the photoanode structure as a “high-pass filter” between the monochromator and the solar cell during the EQE measurement. The T(λ) of the photoanode has a relatively low transmission of ~10% in the blue region (λ < 450 nm), whereas it has a moderate transmission of ~60% in the red region. Consequently, the EQETS(λ) of the solar cell is significantly reduced in the blue range and therefore affects the top cell the most.

56 4. Optimization of 2-j TF-Si solar cells for a BiVO4 photoanode

Additional validation of the EQE measurements is obtained by simulating the performance of the a-Si:H/a-Si:H solar cell structures using the in-house developed Advanced Semiconductor Analysis (ASA) software.173 A semiconductor model of the a-Si:H/a-Si:H solar cell is built in ASA, and the measured wavelength-dependent n(λ) and k(λ) (extinction coefficient) values have been used as the optical input data for the simulation. The simulations have been performed for the three tandem cells: the thin cell, the medium cell and the thick cell. The simulated EQEAM 1.5 and EQETS of each cell are depicted in Figure 4.3 (a) and (b), respectively. The simulated EQE spectra closely resembles the measured EQEAM 1.5 and EQETS as shown in Figure 4.2 (a) and (b), implying that our approach of measuring EQEAM 1.5 and EQETS is validated.

1.0 AM 1.5 top i-layer 50nm 0.8 75nm sum 100nm top 0.6

(-)

bottom EQE 0.4

0.2

0.0 300 400 500 600 700 800  (nm) (a)

1.0 -2 TS Jsc (mAcm ) AM 1.5TS top i-layer top bottom 0.8 50nm 8.042.88 8.154.54 75nm 9.343.51 7.214.09 100nm 10.324.01 6.503.73 0.6

(-)

sum EQE 0.4

0.2 top bottom

0.0 300 400 500 600 700 800  (nm) (b) Figure 4.3 ASA simulation of EQE spectra in the series of the top i-layer thickness (black: thin cell, red: medium cell, green: thick cell), compared with that of single-junction a-Si:H solar cell (blue and magenta) (a); and under the transmitted spectrum through the photoanode (b).

The optimum current matching between the top and bottom cell depends on the employed spectrum. Figure 4.4 shows both the short-circuit densities JSC,AM 1.5 and JSC,TS of the top and bottom cell for the three a-Si:H/a-Si:H double-junctions for the AM 1.5 solar spectrum and the TS 4.4 Solar cell optimization 57

through the photoanode structure, respectively. The solid and dashed lines indicate the result from experiment and simulation, respectively, which are in excellent agreement. For the AM 1.5 spectrum the double-junction is limited by the JSC, AM 1.5 of the 350 nm thick bottom cell. The thin cell shows an excellent current matching between the top and bottom cells, whereas the medium and thick cells are bottom-limited. The situation is different when part of the spectrum is filtered by the photoanode. Since the photoanode mainly absorbs the blue light, JSC, TS becomes top- limited for the thin and medium solar cells. The data in Figure 4.4 show that an ~89 nm top cell is perfectly current-matched with a 350 nm thick i-layer in this a-Si:H/a-Si:H device.

12 solid: EQE measurement dash: ASA simulation 10 top

) -2 8 AM1.5

cm  bottom

(mA

J TS top 4 bottom 2 50 75 100 top i-layer thickness (nm)

Figure 4.4 The JSC of the reported a-Si:H/a-Si:H cell under the standard AM 1.5 spectrum illumination (solid black & red) and under the transmitted spectrum (TS) through the photoanode (solid blue & green); compared with those values from the ASA simulation (dash)

Figure 4.5 illustrates the external parameters of the tandem solar cells, VOC, AM 1.5, JSC,AM 1.5, fill factor (FFAM 1.5), efficiency (ηPV, AM 1.5) under standard AM 1.5 illumination (black squares) and VOC, TS, JSC, TS, FFTS, ηTS, under illumination of spectrum transmitted through the photoanode (red circles).

As discussed above, the JSC, TS of the PV device decreases significantly in reference to JSC, AM 1.5 when the cell is positioned after the photoanode material. The various VOC values remain roughly constant for the three devices. Slightly lower VOC, TS values are measured for the transmitted 174 spectrum due to the weak dependence of the VOC on the current density

kT JSC (4.5) VOC  ln(  1) qJ0

When illuminated by the standard AM 1.5 solar spectrum, the current-matched thin solar cell results in the highest JSC, AM 1.5 and consequently the highest VOC,AM 1.5. But when the double- junction is shaded by the photoanode, the JSC, TS of the thin solar cell drops the most due to the heavily current-limited top cell. The highest JSC, TS is obtained for the double-junction with the 58 4. Optimization of 2-j TF-Si solar cells for a BiVO4 photoanode

thickest top cell, as it is the configuration closest to the optimum condition for current matching (i-layer thickness of ~89 nm in top cell). The ηSTH for this device, in which no external bias is applied, is defined as175

1.23 V  J (mA cm2 )   100% (4.6) STH OP 100 mW cm-2

 1.6  1.6 AM1.5

(V) (V) TS



OP OC 1.2 OP 1.2

0.8 0.8

)

-2 -2 8  8 6 6

4 4  (mA cm

(mA cm 

SC 2 2

J  9 9

6 6 (%)

(%) 

STH

PV 3  3   0 0  70  70 

(%)

60 60 OP

FF 50 50 FF

0.0  -0.4

(mW)

P -0.8   -1.2 POP - PMPP

PEC-WSD

 50 75 100

PV PV top i-layer thickness (nm)

Figure 4.5 The external electrical properties of a-Si:H/a-Si:H solar cells in the series of top i- layer thickness illuminated by the standard AM 1.5 solar spectrum (black square) and by the photoanode transmitted spectrum (TS) (red circle), the operating point (OP) for PEC-WSDs (blue triangle), and the power mismatching between the OP and the corresponding maximum power point (MPP) (green diamond).

4.4 Solar cell optimization 59

The value of 1.23 V is the chemical potential needed to split the H2O molecule, JOP is the current density at the operating point, and 100 mW cm-2 (or 1000 W m-2) is the average intensity of AM 1.5 spectrum. At first glance it might be surprising to see that the thin and medium cells have higher ηSTH values than their PV component ηTS values (the efficiency of the tandem solar cell operating in the spectrum which is transmitted from the photocathode). It is important to realized that for ηTS, the majority of the energy in blue part of the AM 1.5 solar spectrum does not contribute to generating voltage and current, whereas the energy in this spectral part does contribute to the performance ηSTH of the PEC-WSD.

In Figure 4.6, the j-V curves of the BiVO4 photoanode structure and the a-Si:H/a-Si:H double- junctions are illustrated. It is obvious that the thick solar cell results in the largest power at its OP, in spite of the fact that its VOC is slightly lower compared to the other two cells. Combination of the PEC cell and the PV solar cell would result in a ηSTH of 4.0%. The voltage VOP, current density JOP and ηSTH at OP are presented by the blue triangles in Figure 4.5 (right-side axis), respectively. The efficiency of a PV device under the illumination of the transmitted spectrum through the photoanode, ηTS, is determined by its MPPTS divided by the average intensity of AM 1.5:

PVJMPP, TS MPP, TS MPP, TS (4.7) TS  -2 100% PIncident 100 mW cm

5 6 TS

4 5 OP

) MPP 4 -2 3 OP MPP OP 3 (%) MPP

STH

2 

(mA cm j 2 top i-layer 1 50 nm 1 photoanode 75 nm 100 nm 0 0 0.0 0.4 0.8 1.2 1.6 V (V)

Figure 4.6 The j-V characteristics of the photoanode measured under the AM 1.5 spectrum illumination (blue) and the j-V characteristics corresponding to the three a-Si:H/a-Si:H double- junction cells with various i-layer thickness measured under the illumination of the transmitted spectrum (TS) through the photoanode. The maximum power point (MPP) of the solar cells and the corresponding operating point (OP) in the PEC-WSDs are shown as well. The right axis reflects the corresponding ηSTH. 60 4. Optimization of 2-j TF-Si solar cells for a BiVO4 photoanode

The MPPTS of the PV part of the PEC-WSD is in general not equal to the OP of the PEC-WSD. The loss due to the power mismatching between the OP (solid circles in Figure 4.6) and the MPPTS (hollow circles in Figure 4.6) is defined as ΔP = POP - PMPP, TS. As shown by the green diamonds in Figure 4.5, the minimum mismatching loss is achieved for the PEC-WSD with the thick solar cell. This allows us to define an effective fill factor for the OP:

VJOP OP (4.8) FFOP 100% VJOC, TS SC, TS

By comparing the OP and MPPTS of each cell in Figure 4.6, the FFOP of thick solar cell turns out to be the highest, as shown in Figure 4.5.

4.5 Performance and stability of PEC-WSDs

In this section, we address the performance and the stability of the PEC cell and the PV cell of the PEC-WSD. First, we focus on the performance and the stability of the PEC part. The gradient- doped W:BiVO4 PEC is connected to the optimized thick solar cell using a graphite configurations as illustrated in Figure 4.1 (a).

The resulting chronoamperometry plot of this combined device is shown in Figure 4.7. In the first -2 minute of the AM 1.5 illumination, the JOP of the PEC-WSD reaches ~4.2 mA cm , corresponding to ηSTH ~5.2%. This j value is slightly higher than the results shown in Figure 4.5 -2 (~3.3 mA cm , ηSTH ~4.0%). The main reason is that the potentiostat should sweep from the lower voltage to higher voltage in the j-V curve measurement of the photoanode. In these few minutes, the generated O2 bubbles that accumulate on the surface of the photoanode reduce the number of active catalysts on the surface. This would result in a lower j value near the OP of photoanode (green solid circle in Figure 4.6) than the j value in the beginning moment of the stability measurement of the hybrid PEC-WSD. This initial decrease can be inferred in Figure 4.7 -2 as well: j value drops and tends to stabilize at ~3.4 mA cm (ηSTH ~4.2%) after ~2.5 minutes in the measurement.

Longer term reduction of the j might be due to the CoPi layer peeling off into the electrolyte. The ~4.2% efficiency we report in this chapter was determined after stabilizing for 10 minutes in the 170 illuminated electrolyte. In addition, in a recent publication, we demonstrated a stable 4.9% ηSTH by optimizing the W:BiVO4 photoanode and careful deposition of the CoPi layer. As a result, no degradation in the performance was observed during the course of one hour measurement. We achieve this highest efficiency for metal-oxide semiconductor photoanode reported to date, using double-junction a-Si:H solar cell which is cheaper and easier to fabricate as compared to triple- 176 junction solar cell used the benchmark structure reported in literature. This 4.9% stable ηSTH value is higher than the 4.2% ηSTH value we have achieved in this chapter, and the reason for that is the difference of the substrate on the photoanode side. The 4.9% ηSTH was obtained by using a textured FTO coated glass (ASAHI VU-type) as the substrate of BiVO4. In contrast, a flat FTO 4.5 Performance and stability of PEC-WSDs 61

substrate (TEC-15) is used in this work, mainly because we are focusing on the optimization of the solar cell. Using a flat substrate, less light is being trapped in the photoanode, and higher spectrum photon flux is available for the solar cell.

4 5

Co-Pi + W:BiVO + a-Si:H/a-Si:H cell 4 3 4

)

-2 3

(%)

2 STH  2

(mA cm

j 1 CoPi + a-Si:H/a-Si:H cell 1

0 0 0 150 300 450 600 t (s)

Figure 4.7 Photocurrent density vs. time for the CoPi coated gradient-doped W:BiVO4 and double-junction a-Si:H/a-Si:H device (black), and the CoPi-coated FTO/a-Si:H/a-Si:H device (grey) under AM 1.5 illumination

To illustrate the contribution of the gradient-doped W:BiVO4, the W:BiVO4 photoanode is replaced by only an electrodeposited layer of CoPi on the same type of FTO substrate. As a result, only ~1.5 mA cm-2 of photocurrent density is observed (grey curve in Figure 4.7), significantly lower than the gradient-doped W:BiVO4 photoanode. This confirms that the CoPi-based electrocatalyst requires significant higher bias voltages as delivered by a triple-junction a-Si:H PV device, in order to obtain a reasonable ηSTH. This demonstrates the advantage of the BiVO4 photoanode in reference to the CoPi-based electrocatalyst which provides a significant overpotential for water oxidation.

Finally, we discuss the stability of the PV part of the PEC-WSDs. The a-Si:H/a-Si:H tandem cell suffers from the SWE, which is the creation of additional metastable defects in the a-Si:H absorber layer when exposed to light. These metastable light-induced defects enhance the charge carrier recombination and deteriorates the FF mainly and the JSC of the a-Si:H solar cell to a less extent. The effect can be reversed by thermally annealing out the metastable defects at moderate temperatures of 120-180 °C. The impact of the SWE metastability on the ηSTH of the PEC-WSD is studied in the following paragraphs as well.

62 4. Optimization of 2-j TF-Si solar cells for a BiVO4 photoanode

1.0 top i-layer 1.0 0.9 50nm 0.9 75nm

0.8 100nm STH 0.8  0.7 PV, TS 0.7

0.6 0.6 1.0 1.0 0.9 0.9 FFOP

0.8 FFTS 0.8 0.7 0.7

0.6 0.6 1.0 1.0

Relative degradation @ OP Relative degradation @ TS 0.9 0.9 JSC, TS

0.8 JOP 0.8 0.7 0.7 TS OP 0.6 0.6 0 200 400 600 800 1000 0 200 400 600 800 1000 t (h) t (h)

Figure 4.8 The relative degradation of the η, FF and JSC of the double-junction PV cells with the light soaking time, normalized according to the as-deposited value of each device under the photoanode transmitted spectrum (TS) (left) and the operating point (OP), respectively.

Light soaking experiments of the solely a-Si:H/a-Si:H devices were carried out under 1 Sun AM 1.5 illumination, while keeping the solar cell at a temperature of 25 °C. The relative degradation of ηTS, ηSTH, JSC, TS, JOP and FFTS measured versus light soaking time are illustrated in Figure 4.8 respectively. After light soaking of ~660 h all three devices are fully degraded and stabilized. Among the external parameters, the FF shows the largest relative decrease after 1000 h illumination, i.e. ΔFFTS /FFTS is 20-30% of the corresponding initial value. The relative degradation of the current density ΔJSC, TS /JSC, TS and open-circuit voltage ΔVOC, TS /VOC, TS (not shown in Figure 4.8) is ranging from 5% up to 10%. As a result the relative decrease of the conversion efficiency ΔηTS /ηTS after stabilization is ranging from 30% up to 35%. Figure 4.8 demonstrates that the largest part of the degradation occurs in the first 20 hours of light soaking, which is in general addressed to as the “fast” degradation phase. The “slow” degradation phase corresponds to the additional slow degradation of external parameters after 20 hours of light soaking until the saturated external values for the external parameters are reached after 660 hours of light soaking.

4.5 Performance and stability of PEC-WSDs 63

4.0 TS top i-layer 50nm 4.5 75nm 3.5 100nm

)

-2 OP 4.0 OP 3.0 OP deg (%) OP 3.5

STH

(mA cm

j  OP 2.5 deg 3.0 OPdeg

2.0 photoanode 2.5 0.50 0.75 1.00 1.25 1.50 V (V)

Figure 4.9 The j-V characteristics of the three a-Si:H/a-Si:H double-junction cells with various i- layer thickness after 1000 h light soaking and measured under the transmitted spectrum (TS) through the photoanode. The operating points of the degraded devices (OPdeg) are indicated by the hollow circles on the inter-crossing points, and as a reference the operating point (OP) of the as- deposited PV cells are depicted by the solid circles with the corresponding color (shown in Figure 4.6 as well).

Figure 4.8 shows that relative degradation of ηSTH ranging from 9% up to 18% is significant smaller compared to ηTS (~32%). This demonstrates the OP is less affected by SWE degradation than the MPPTS of the PV part. Another observation is that the stabilized ηSTH depends on the thickness of the top cell. As illustrated in Figure 4.9, this can be explained by the differences in sensitivity of the OP to the degrading external parameters. If the slope (dj/dV) of the j-V curve of the photoanode in the OP is steep, the OP is more sensitive to degradation of JSC, TS, VOC,TS and FFTS. Whereas if the slope dj/dV of the j-V curve is less steep (at higher voltages), OP becomes less sensitive to the degradation of the JSC, TS as can be seen in Figure 4.9. The OP of thick solar cell has the highest voltage and consequently intersects the j-V curve in a range with the smallest slope dj/dV. Therefore the PEC-WSD with the thick solar cells is less affected by the degradation in JSC, TS. The OP of the PEC-WSD based on the thin solar cell has the smallest voltage and j-V curve in a range with the largest slope dj/dV. Consequently, this PEC-WSD should be the most sensitive to variation in the JSC, TS. However, since the degradation ΔJSC, TS /JSC, TS is the smallest for thin solar cell, the effective drop in ηSTH is the smallest. Therefore, the ηSTH is a competition between the sensitivity of the changes of the external parameters given by the dj/dV at the OP and the degradation of the external parameters itself. For that reason, the PEC-WSD based on the medium solar cell shows the largest degradation in ηSTH (red arrow in Figure 4.9). It is more sensitive to the degradation due to a larger slope dj/dV in reference to the PEC-WSD based on the 64 4. Optimization of 2-j TF-Si solar cells for a BiVO4 photoanode

thick solar cell, and also experience a larger degradation in JSC in reference to PEC-WSD based on the thin solar cell. This effect is illustrated in Figure 4.9 as well.

4.6 Conclusions

CoPi-catalyzed gradient-doped W:BiVO4, deposited by spray pyrolysis, is confirmed as an efficient photoanode to split the water using solar energy. An a-Si:H tandem solar cell is introduced in order to provide the bias power source and is integrated into a wireless PEC-WSD. The advantage of the a-Si:H/a-Si:H double-junction solar cell above other PV technologies is that it is simplest and cheapest PV device that can meet the requirement of stability in aqueous environments, straightforward fabrication process, matching spectral response, voltage and current.

Due to absorption of the W:BiVO4 photoanode in the front, only ~10% of blue and ~60% of red spectral range in the AM 1.5 solar spectrum can be utilized by the a-Si:H/a-Si:H tandem cell. The thickness of i-layer in the top cell has been optimized in term of current matching using both EQE measurements and ASA simulations. The thin solar cell (i-layer of 50 nm in top cell) shows the best current matching under AM 1.5 spectrum, while the thick solar cell (i-layer of 100 nm in top cell) performs the best in reference to the spectrum transmitted through photoanode.

The impact of the Staebler-Wronski effect on the ηSTH of the photoelectrochemical water-splitting devices has been studied. The operating point is fortunately less affected by the SWE compared to the maximum power point of the solar cell component under the transmitted spectrum through the photoanode. The optimized a-Si:H/a-Si:H tandem solar cell combined with BiVO4 can result in a -2 photocurrent of ~3.4 mA cm (stabilize for at least 10 min), which corresponds to a ηSTH of 4.0%.

The ηSTH can be further enhanced provided that the infra-red range of the solar spectrum is better utilized, by using a double-junction cell in which the bottom cell is based on PV materials with a lower bandgap, e.g., the micromorph (a-Si:H/nc-Si:H) solar cells. In addition, further optimization of the photoanode material and device is promising, which might reduce the high current and voltage requirement of the solar cell. Furthermore, by the implementation of some light-trapping techniques at the photoanode, higher current density can be achieved due to absorption in the photoanode. The progress of going from concepts based on a triple-junction solar cell to a double- junction solar cell is already realized in this chapter. Our next objective is to utilize a single- junction solar cell in the PEC-WSD to reach higher values of the ηSTH.

5. An efficient solar water-splitting device based on a bismuth vanadate photoanode and a thin-film silicon solar cell1

A hybrid photovoltaic/photoelectrochemical (PV/PEC) water-splitting device with a benchmark solar-to-hydrogen conversion efficiency of 5.2% under simulated air mass (AM) 1.5 illumination is reported. This cell consists of a gradient-doped tungsten- bismuth vanadate (W:BiVO4) photoanode and a thin-film silicon solar cell. The improvement with respect to an earlier cell that also used gradient-doped W:BiVO4 has been achieved by simultaneously introducing a textured substrate to enhance light-trapping in the BiVO4 photoanode and further optimization of the W gradient doping profile in the photoanode. Various PV cells have been studied in combination with this BiVO4 photoanode, such as an amorphous silicon (a-Si:H) single-junction, an a-Si:H/a- Si:H double-junction, and an a-Si:H/nano-crystalline silicon (nc-Si:H) micromorph double- junction. The highest conversion efficiency, which is also the record efficiency for metal-oxide based water-splitting devices, is reached for a tandem system consisting of the optimized W:BiVO4 photoanode and the micromorph (a-Si:H/nc-Si:H) cell. This record efficiency is attributed to the increased performance of the BiVO4 photoanode, which is the limiting factor in this hybrid PEC/PV device, as well as better spectral matching between BiVO4 and the nc-Si:H cell.

1 This chapter has been published: L. Han, F.F. Abdi, R. van de Krol, R. Liu, Z. Huang, H.-J. Lewerenz, B. Dam, M. Zeman and A.H.M. Smets, ChemSusChem. 7(10), 2832-2838, 2014. 66 5. An efficient PEC-WSD based on a BiVO4 photoanode and a TF-Si solar cell

5.1 Introduction

Solar water splitting utilizes sunlight to produce hydrogen, a chemical fuel with the highest gravimetric energy density and a critical element for reducing carbon dioxide into useful chemical 177 178, 179 180 products. For light-induced water-splitting, transition metal-oxides such as TiO2, WO3, 181-184 95, 185 Fe2O3 and BiVO4 have been widely used as photoanode materials. In particular, BiVO4 is among the most promising photoanode materials for this application, due to its direct energy gap of ~2.4 eV, its environmental friendliness and its long-term stability in neutral and basic 93 electrolytes. The theoretical solar-to-hydrogen conversion efficiency (ηSTH) limit of this material is 9.1% with a maximum photocurrent density of 7.5 mA cm-2 under air mass (AM) 1.5 standard test conditions. In addition, its conduction band is located close to the hydrogen evolution potential.186 The low carrier mobility in the material has recently been counterbalanced by introducing an internal electrical field by gradient doping with tungsten (W).95 Carrier drift then competes favorably with the recombination processes, resulting in a considerably enhanced photocurrent and making BiVO4 the photoelectrode with the hitherto highest photocurrent density within its material class.

In the taxonomy of water-splitting structures, buried junctions, as provided by photovoltaic (PV) tandem structures or metal-semiconductor Schottky barriers, are distinguished from systems where the junction to the electrolyte determines the overall behavior. A third alternative are structures where part of the system’s behavior is governed by the built-in rectifying junctions and, simultaneously, by the contact to the electrolyte. This type of structure has been shown to be very attractive, resulting in devices with relatively high efficiency.95, 111, 167, 168, 187, 188 This is also the case for BiVO4 deposited onto PV tandem solar cells, since the BiVO4/electrolyte contact and the buried PV cell influence the overall behavior of the photoanode. To obtain high efficiency, it is therefore essential to optimize the performance of BiVO4 and investigate various underlying tandem structures.

Scheme 5.1 Cross-section sketch of our PEC/PV device. 5.2 Experimental 67

Herein, we improve the performance of spray-pyrolysed BiVO4 by rationalizing the design of the concentration gradient of tungsten and simultaneously employing light-trapping techniques by depositing the BiVO4 onto textured fluorine-doped tin oxide (FTO) glass substrates. Although a preliminary result of using a textured FTO substrate has been presented in our previous work,95 we here further optimize this light-trapping approach and analyze the enhancement in greater details. This modified photoanode, when functionalized with a cobalt-phosphate (CoPi) catalyst, provides an AM 1.5 photocurrent of 3.0 and 4.0 mA cm-2 at 0.75 and 1.23 V versus (vs.) a reversible hydrogen electrode (RHE), respectively. Subsequently, we combined this optimized BiVO4 photoanode with an amorphous silicon tandem (a-Si:H/a-Si:H) cell, a micromorph (a- Si:H/nc-Si:H) cell, and a single-junction a-Si:H cell, respectively. In this tandem configuration, as shown in Scheme 5.1, the AM 1.5 irradiance for the rear solar cell is filtered by the front BiVO4 photoanode. The current matching conditions in the tandem (a-Si:H/a-Si:H) cell and the micromorph silicon (a-Si:H/nc-Si:H) cell are therefore sensitive to the modified shape of the solar spectrum transmitted through the PEC junction. The three PEC/PV tandem devices are characterized and their performance is compared in this work. The tandem configuration of BiVO4 and the micromorph Si solar cell is found to show the highest solar-to-hydrogen conversion efficiency of 5.2%, which is also the highest reported conversion efficiency for solar water-splitting device based on metal-oxides.

5.2 Experimental

Thin-film silicon solar cells were deposited by a radio frequency plasma-enhanced chemical vapor deposition (RF-PECVD) multi-chamber tool. The 2.5 cm × 10 cm ASAHI VU-type substrate (~600 nm thick textured FTO layer on glass) was heated at 170 °C during the thin-film Si deposition.

The p-layers in the a-Si:H cell and a-Si:H/a-Si:H cell are based on a-SiC:H(B), the i-layers are based on a-Si:H and the n-layers are phosphorus-doped nano-crystalline silicon oxide (nc- SiOx:H(P)). The thickness of the i-layer of single-junction a-Si:H cell is 300 nm. The thicknesses of the i-layers in the a-Si:H/a-Si:H double-junction are 100 nm and 350 nm for the top and bottom junction, respectively, optimized according to Chapter 4.111 In the micromorph Si solar cell, the p- layers are based on nc-SiOx:H(B), and the thicknesses of the i-layers are 300 nm for a-Si:H and 1800 nm for nc-Si:H. After the Si thin-films were synthesized, a stripe of 300 nm Al was coated by rotating PROVAC evaporator on the pre-covered region of the sample as the front contact. Each metal back contact has an area of 1 cm × 1 cm. The structure sketches of the 3 PV cells are illustrated in Schemes A.1-3 in Appendix A.

Gradient-doped W:BiVO4 with a thickness of 200 nm was synthesized by spray pyrolysis as 111 reported previously. In the 250 nm thick gradient-doped W:BiVO4, 50 additional spray cycles of the undoped BiVO4 precursor solution were performed. Two types of TCO-coated glass substrates were used in the spray pyrolysis in this chapter. One is TEC-15 F:doped SnO2 (FTO) glass (HARTFORD Glass Co.), which has a relatively flat surface; and the other is ASAHI VU- type substrate (ASAHI Glass Co.), which is an FTO-coated glass with random sharp features as the texturing layer. Electrical contacts to the BiVO4 photoanode in the tandem devices were 68 5. An efficient PEC-WSD based on a BiVO4 photoanode and a TF-Si solar cell

established using a silver (Ag) wire and a graphite paste. CoPi catalyst was electrodeposited on 176 the surface of BiVO4 according to previous reports.

Three-electrode PEC performance was carried out in 0.1 M KPi (pH~7) aqueous solution with a platinum wire and a Ag/AgCl electrode (XR300, saturated KCl and AgCl solution, Radiometer Analytical) as the counter and reference electrode, respectively. Two-electrode PEC performance was measured in the same electrolyte with the samples as working the electrode and a platinum (Pt) wire as the counter electrode. A simulated AM 1.5 solar illumination (100 mW cm-2) was achieved with a NEWPORT Sol3A Class AAA solar simulator (94023A-SR3 type) as the light source. In all the PEC measurements, a circular area with a diameter of 6 mm on each sample was illuminated, corresponding to a total area of 0.283 cm2. The current density-potential profiles of the working electrode were monitored by a potentiostat (EG&G PAR 283).

5.3 Results and discussions

5.3.1 Absorption enhancement by light-trapping in photoanode Light-trapping techniques have been widely investigated and applied in thin-film PV cells for decades.189 By appropriately deploying transparent conductive oxide (TCO) films with suitable root mean square roughness and average lateral feature size on the TCO films,190 a relatively broad spectral range of sunlight can be scattered into large angles, hence enhancing the average absorption path length in the absorber layer. This significantly improves the absorption in the intrinsic layer, and the overall performance of the PV cell. Considering the similarities to PV applications, we introduce a textured surface on the TCO substrate of a photoanode for water splitting. We compare two different FTO substrates: TEC-15 and ASAHI VU-type. Figure 5.1 (a) and (b) show the atomic force microscopic (AFM) images of both substrates. The root mean square surface roughness of the ASAHI VU-type substrate (42 nm) is much higher than that of the TEC-15 substrate (17 nm). This is better illustrated by the cross-sectional profiles of the samples as shown in Figure 5.1 (c). The textures of randomly distributed micro-sized pyramid grains are adopted by the BiVO4 and CoPi films deposited on top.

5.3 Results and discussions 69

Figure 5.1 Atomic force microscopic (AFM) images of (a) flat TEC-15 FTO surface and (b) textured ASAHI VU-type FTO glass substrates; (c) the cross-sectional profile of the (a) (blue curve) and (b) (red curve); (d) absorption of 250 nm W:BiVO4 films on flat TEC-15 FTO (blue curve) and textured (ASAHI VU-type) FTO-glass substrates (red curve).

To study the potential light trapping of the substrates, two 200 nm thick BiVO4 films doped with continuously decreasing concentration of W were deposited by spray pyrolysis onto the TEC-15 (flat) and ASAHI VU-type (textured) substrates, respectively. The absorption of the gradient- doped W:BiVO4/FTO/glass photoanode on the flat and textured substrates is compared in Figure 5.1 (d), where shows that the light absorption is enhanced in the spectral range of 350 < λ < 490 nm in the textured sample. In fact, this textured surface affects the light management in two ways. First, the texture improves the anti-reflective properties of the photoanode. For photons with λ < 450 nm, the light collection efficiency is close to unity (> 95%) due to enhanced anti-reflection of the textured surface. Second, it enhances scattering of light into large angles, increasing its absorption path length in the BiVO4 layer. This mainly affects the spectral region just above the bandgap of BiVO4 (450 < λ < 500 nm).

To confirm the performance improvement by the integration of textured surfaces, we compare the two-electrode current density-voltage (j-V) characteristics of the photoanode on flat and textured substrates for water splitting under AM 1.5 solar irradiance (simulated spectrum, shown in Figure A.1 in Appendix A). As shown in Figure 5.2, at a given voltage, the current density of the photoanode on the textured substrate is generally about 25-30% higher than that of an equal thick BiVO4 film deposited on a flat substrate. This results from the improved light collection 70 5. An efficient PEC-WSD based on a BiVO4 photoanode and a TF-Si solar cell

efficiency due to the textured interfaces. In addition, different textured TCO substrates, such as tin-doped indium oxide (ITO) and aluminum zinc oxide (AZO), were tested as well (data not shown here). The use of ITO and AZO results in lower performance of the photoanode as compared to when FTO was used. We attribute this to the electrical properties of ITO and AZO being less resistant to the thermal degradation under the high temperatures (450 °C) at which BiVO4 is deposited and annealed.

Figure 5.2 Light trapping in a gradient-doped CoPi-catalyzed W:BiVO4 photoanode enhances the performance of the hybrid PEC/PV device. Solid curves are the j-V curves of the photoanodes for water splitting on flat (blue) and on textured FTO substrates (red); dashed curves are the two- electrode j-V curves of the a-Si:H/a-Si:H tandem cells illuminated by AM 1.5 stimulated sunlight that is filtered by the corresponding photoanode. The blue and red dots are the operating points (OPs) of the a-Si:H/a-Si:H cell in combination with the flat and textured W:BiVO4 photoanodes, respectively.

In the hybrid PEC/PV configuration, the absorption by the PEC junction at the front of the device will decrease the transmitted light available for the PV junctions at the back. Therefore, the next question that arises is how the texturing of the PEC junction affects the performance of the underlying PV junctions. To study this effect, flat and textured BiVO4 samples were placed in front of a previously optimized a-Si:H/a-Si:H double PV junction to act as a filter for the incident AM 1.5 light. Note that the thickness of the a-Si:H top and bottom junctions were optimized for a 111 flat photoanode. Figure 5.2 shows that the open-circuit voltage (VOC), the short-circuit current (JSC) and the fill factor (FF) (dashed lines in Figure 5.2) of the a-Si:H/a-Si:H cell are significantly lower for the textured photoanode. This is expected since the textured photoanode absorbs more light than the flat photoanode, transmitting less light to the a-Si:H/a-Si:H cell. However, the operating point of the textured device, which is given by the intersection of the PEC and PV j-V curves,95, 111, 187, 188 is at a higher photocurrent density than that of the flat device. In a working hybrid device where the PEC and the PV junctions are connected in series, ηSTH is determined by

5.3 Results and discussions 71

1.23 V  J (mA cm2 )   100% (5.1) STH OP 100 mW cm-2

-2 where JOP is operating photocurrent density of the system in mA cm and the value of 1.23 V is -2 the thermodynamic potential for water splitting (H2O (l)  H2 (g) + O2 (g)). The 100 mW cm corresponds to the total power density of AM 1.5 sunlight. It is clear from the equation that ηSTH is only determined by the photocurrent density at the operating point (assuming 100% Faradaic efficiency). Therefore despite the poorer j-V characteristics of the a-Si:H/a-Si:H cell when placed behind the textured photoanode, the textured device does give a higher ηSTH. Specifically, the intersection point of the j-V curve of the photoanode on the textured substrate (red hollow dot in Figure 5.2) shows that light trapping in the photoanode enhances the ηSTH relatively ~12% compared with the photoanode on the flat substrate (blue hollow dot in Figure 5.2).

5.3.2 Doping profiling optimization on the photoanode

Because the optical absorption in BiVO4 plays an important role in improving the ηSTH value of the resulting device, we also studied the effect of increasing the thickness of the BiVO4 film. Initially, we prepared 250 and 300 nm thick BiVO4 films, each with the same 10-step gradient in W dopant concentration. This means that the thickness of each step is 25 and 30 nm for the 250 and 300 nm thick films, respectively. Although these two films obviously absorb more light than the 200 nm thick film with the same 10-step gradient doping, the resulting photocurrents are found to decrease (data not shown). We believe the reason for this decrease is related to the competing effects of enhanced light absorption and poorer carrier collection efficiency (ηcol, 94 sometimes also referred as carrier separation efficiency, ηsep) in thicker films. The performances decrease indicates that the detrimental influence on carrier collection efficiency apparently outweighs the improvement in light absorption when increasing the film thickness from 250 to 300 nm.

Recently, we reported a time-resolved microwave conductivity study on spray-pyrolysed BiVO4 96 films. The carrier diffusion length in an undoped BiVO4 was found to be ~70 nm. Mobility 191 measurement on a BiVO4 single crystal reveals a comparable carrier diffusion length. This means that carriers (electrons and holes) can travel ~70 nm via diffusion in undoped BiVO4 before they eventually recombine. Taking this into account, we prepared a 250 nm thick gradient- doped W:BiVO4 film with a different gradient profile. The thickness of the layer for each dopant step is kept at 20 nm, except for the last step (undoped layer) which is 70 nm. This allows us to increase the thickness of an optimized 200 nm thick gradient-doped film to 250 nm, without affecting the carrier collection efficiency (Figure 5.3 (a) and (b)). Figure 5.3 (c) shows that the AM 1.5 photocurrent density indeed increases after this modification. AM 1.5 photocurrent densities as high as 3 mA cm-2 and 4 mA cm-2 are achieved at potentials of 0.75 and 1.23 V vs. RHE, respectively, with excellent reproducibility (Figure A.2 in Appendix A). This is a 10-15% improvement as compared to the 200 nm thick gradient-doped W:BiVO4 reported by us previously95 and comparable to the photocurrent densities recently reported by Choi et al. for their 192 nano-structured BiVO4-FeOOH-NiOOH system. The observed improvement illustrates the 72 5. An efficient PEC-WSD based on a BiVO4 photoanode and a TF-Si solar cell

significance of rational design of the dopant profile in a metal-oxide photoanode for water splitting.

Figure 5.3 Band diagram schematic of the 200 nm reference gradient-doped W:BiVO4 sample (a) and the 250 nm modified gradient-doped W:BiVO4 sample (b). The electrolyte side is on the right-hand side, and the FTO substrate is on the left. (c) Three-electrode AM 1.5 photocurrent density vs. potential (j-V) curves of the samples depicted in (a) and (b) under front illumination (light enters via the electrolyte). Both samples are catalyzed with CoPi. The dashed curve (black) shows the dark current density of the samples.

5.3.3 Spectral matching in the PEC/PV configuration We have demonstrated in our previous report that the profiles of the external quantum efficiency (EQE) spectra for the thin-film silicon cells in a tandem PEC/PV configuration (Scheme 5.1) are highly sensitive to the presence of a front anode structure.111 Considering the improvement that we made with the BiVO4 front photoanode, further optimization of the rear silicon cell needs to be done, and higher ηSTH may be expected. The dependence of the rear cell to the front photoanode mainly results from the fact that the same current has to flow through all the (series- connected) components in the PEC/PV tandem device and the significant spectral overlap between the absorption of BiVO4 and a-Si:H used. This current-matching requirement suggests 5.3 Results and discussions 73

that further improvements can be made by extending the spectral utilization of the bottom PV cell. Therefore, we expect a higher ηSTH by the introduction of a micromorph Si solar cell. The spectral utilization of the nc-Si:H bottom cell is extended to the near-infrared (NIR) range (λ < 1100 nm, Figure 5.4), further than the a-Si:H bottom cell (λ < 800 nm) in the previously used a- Si:H/a-Si:H tandem cell.

Figure 5.4 The external quantum efficiency (EQE) curves of a-Si:H/nc-Si:H solar cell, a-Si:H top junction (black) and nc-Si:H bottom junction (red). Dashed curves indicate EQE spectra of the PV junctions under full AM 1.5 simulated solar illumination; solid curves indicate the EQE spectra measured with a CoPi coated BiVO4 photoanode placed in front of the solar cells. The blue curve represents the optical absorption spectrum of the CoPi-coated BiVO4 photoanode.

For this new PEC/PV device, the spectral response of each component illuminated by the standard AM 1.5 irradiance was measured as shown by the solid curves in Figure 5.4. The dashed curves represent the j-V curves of the PV cells without the CoPi-catalyzed BiVO4 photoanode in the front. Because of the spectral overlap of the photoanode (Figure 5.4, solid blue line) and the a- Si:H PV junction, the EQE of the a-Si:H is reduced by a factor of ~10 below λ = 450 nm. In contrast, the absorption in nc-Si:H layer is less affected by the absorption in BiVO4 - the reduction of its EQE is primarily caused by reflection and scattering losses - and the shape of the spectral response is relatively unchanged. The reflection losses can be further minimized by optimizing the refractive indices of the front layers, which is beyond the scope of this work. Integrating the modified EQE of the micromorph Si solar cell allows us to predict the short-circuit photocurrent density of the cell, which is 5.91 mA cm-2 (top a-Si:H cell limited). This value is higher than that of the a-Si:H/a-Si:H cell (4.37 mA cm-2), and provide additional room for the performance improvement of our BiVO4.

In addition to the spectral match requirement, the j-V characteristics of the PV cell should match those of the PEC cell to optimize the performance. The j-V characteristics of the latter are mostly determined by the optical absorption, the carrier collection efficiency and the catalytic efficiency 74 5. An efficient PEC-WSD based on a BiVO4 photoanode and a TF-Si solar cell

of the semiconductor photoanode.193 The two-electrode j-V characteristics of various thin-film silicon (TF-Si) PV cells and the CoPi-catalyzed 250 nm thick gradient doped W:BiVO4 photoanode are shown in Figure 5.5. The operating current density, voltage, and the ηSTH values of the various PEC/PV combinations are listed in Table 5.1, together with the performance parameters of the respective PV cells (ηPV is the conversion efficiency of the stand-alone PV junctions illuminated by the specific spectrum). Figure 5.5 (a) shows that the operating photocurrent densities of W:BiVO4/a-Si:H/a-Si:H and W:BiVO4/micromorph (a-Si:H/nc-Si:H) are similar, despite the differences in the characteristics of the PV components. A closer look, however, reveals that the operating photocurrent density of the W:BiVO4/micromorph Si is slightly higher, just above 4.22 mA cm-2 (Figure 5.5 (b)). This higher photocurrent density is due to the smaller spectral overlap between the micromorph Si and the BiVO4, compared with the overlap between the a-Si:H/a-Si:H and the BiVO4.

Figure 5.5 The two-electrode j-V curves match between PV cells and photoanode, indicating the micromorph Si cell can beat the a-Si:H/a-Si:H device when the performance of photoanode is improved (a). A ηSTH of 5.2% record is achieved (b: zoom in of the central part in a). The j-V curves of the PV cells are measured using AM 1.5 illumination filtered with the CoPi catalyzed, 250 nm-thick grad-doped W:BiVO4 photoanode.

Short-circuit photocurrent density measurements of the PEC/PV device confirm the operating photocurrent predicted by the intersection of the j-V curves (Figure A.3 in Appendix A). The 5.3 Results and discussions 75

photocurrent is also relatively stable within the course of an hour, with less than 5% degradation observed. Multiple reports using the same material system (W-doped BiVO4 and CoPi) have confirmed that the Faradaic efficiency is effectively 100%.159 By taking this into account, the 4.22 -2 mA cm photocurrent density that we observe with the combination of BiVO4 and micromorph Si corresponds to an apparent ηSTH of 5.2%, which represents the current benchmark for stand-alone water-splitting devices based on a metal-oxide photoelectrode.95, 176, 185

95 Although the increase in the ηSTH (4.9% to 5.2%) may seem to be incremental, the implication is very significant. Recent techno-economic analyses have shown that efficiency is the most important knob in determining the resulting cost of hydrogen.194, 195 This is of course unsurprising as also demonstrated in the solar cells field.

Table 5.1 External parameters of three thin-film silicon solar cells and the hybrid PEC/PV device when each of them is combined with the CoPi catalyzed, 250 nm-thick grad-doped W:BiVO4 photoanode.

Conditions PV parameters a-Si:H a-Si:H/a-Si:H a-Si:H/nc-Si:H

VOC (V) 0.93 1.68 1.32

-2 JSC (mA cm ) 13.92 6.18 11.63 AM 1.5 illumination FF (%) 76.3 69.8 62.4

ηPV (%) 9.9 7.3 9.6

VOC (V) 0.91 1.63 1.27

-2 Photoanode-filtered JSC (mA cm ) 7.76 4.37 5.91 AM 1.5 illumination FF (%) 76.9 67.8 63.5

ηPV (%) 5.4 4.8 4.8

VOP (V) 0.87 1.03 1.10

-2 OP JOP (mA cm ) 3.77 4.14 4.22

ηSTH (%) 4.6 5.1 5.2

Figure 5.5 also shows that the combination of an optimized W:BiVO4 photoanode and an optimized single-junction a-Si:H can result in a relatively high ηSTH of 4.6%. This is only ~11% lower than the record efficiency we report here, but using less number of layers (p/i/n + buffer layers). This means that the processing time and costs of the device are significantly reduced. Moreover, Figure 5.5 (a) clearly shows that the BiVO4 photocurrent density at around 0.8-0.9 V limits the overall performance of the device. With a theoretical photocurrent density of 7.5 mA 76 5. An efficient PEC-WSD based on a BiVO4 photoanode and a TF-Si solar cell

-2 cm for BiVO4, there is plenty of room for improvement by the combination of BiVO4 and single-junction a-Si:H. In contrast, only little improvements are possible for the BiVO4/double- junction Si cell devices, in which the total photocurrent density is already close to the maximum photocurrent density of the double-junction Si cell.95, 111, 185 A configuration with a single-junction a-Si:H cell therefore seems the most promising route for further development of hybrid PEC/PV devices.

Finally, it is imperative that the platinum counter electrode used herein for convenience needs to be replaced by an earth-abundant alternative. For example, one may consider various earth- abundant hydrogen evolution catalysts, such as the transition metal alloys (NiMo, NiMoZn,176 196 197, 198 199 NiFeMo, CoMo ), metal sulfides (MoS2, WS2 ), and the newly emerged metal 200 phosphides (Co2P, Ni2P ). These materials are already highly developed and demonstrate high activities at relatively low overpotentials.

5.4 Conclusions

In conclusion, a new benchmark of 5.2% solar-to-hydrogen conversion efficiency of a metal- oxide based device has been obtained by optimizing the absorption and carrier collection in a BiVO4 photoanode, and combining it with a micromorph Si tandem cell to form a stand-alone hybrid PEC/PV tandem device. The increased absorption is obtained through the employment of textured substrates and by increasing the film thickness, while retaining a high carrier collection efficiency by rational design of the gradient W-dopant concentration profile in BiVO4. The combination with a micromorph Si cell allows better utilization of the AM 1.5 solar spectrum, as compared to the earlier demonstration of tandem with a-Si:H/a-Si:H cell, due to the extended absorption of the nc-Si:H bottom junction up to ~1100 nm. Further improvement of the performance of the BiVO4 photoanode may be possible through nano-structuring and/or employing metal plasmonic nano-particles to improve the light absorption closed to the band edge. When these photoanodes reach photocurrent densities in excess of ~5 mA cm-2, the use of a single-junction a-Si:H cell may lead to higher efficiencies than double-junction a-Si:H based PV cells. As such, the combination of chemically stable metal-oxide absorbers with efficient silicon- based PV devices offers a realistic pathway towards efficient, simple solar water-splitting devices from earth-abundant elements.

6. A thin-film silicon based monolithic photoelectrochemical/photovoltaic cathode with efficient hydrogen evolution1

A cost-effective and earth-abundant photocathode based on hydrogenated amorphous silicon carbide (a-SiC:H) is demonstrated to split water into hydrogen and oxygen using solar energy. A monolithic a-SiC:H photoelectrochemical (PEC) cathode integrated with a hydrogenated amorphous silicon (a-Si:H)/nano-crystalline silicon (nc-Si:H) double photovoltaic (PV) junction achieved a current density of -5.1 mA cm-2 at 0 V versus the reversible hydrogen electrode. The a- SiC:H photocathode used no hydrogen evolution catalyst and the high current density was obtained using gradient boron doping. The growth of high quality nc-Si:H PV junctions in combination with optimized spectral utilization was achieved using glass substrates with integrated micro-textured photonic structures. The performance of the PEC/PV cathode was analyzed by simulations using Advanced Semiconductor Analysis software.

1 This chapter has been published: L. Han, I.A. Digdaya, T.W.F. Buijs, F.F. Abdi, Z. Huang, R. Liu, B. Dam, M. Zeman, W.A. Smith, and A.H.M. Smets, Journal of Materials Chemistry A, DOI: 10.1039/C4TA05523C, 2015. 78 6. A TF-Si based monolithic PEC/PV cathode with efficient H2 evolution

6.1 Introduction

Solar energy is by far the most abundant and sustainable energy source available for humans on the Earth. However, the implementation of photovoltaic (PV) technologies is limited by the periodic nature of the solar irradiance at a certain location and time. Thus we need to store solar energy in the form of a chemical fuel to be used when little or no solar irradiance is available. One promising method to store solar energy is to utilize sunlight to split water into hydrogen and oxygen using a photoelectrochemical (PEC) water-splitting device. Hydrogen is a carbon-free and high caloric energy carrier that can be used to produce electricity when recombined with O2 in a fuel cell. Thus the search for an ideal semiconductor electrode for PEC water splitting has become an important challenge to many researchers.

Currently, most solar-to-hydrogen conversion devices with efficiencies (ηSTH) > 1%, have been achieved by hybrid device configurations, in which at least one PEC junction is combined with PV junction(s).95, 97, 111, 165, 167, 168, 176, 187, 201-204 In 1998, Turner and Khaselev demonstrated a monolithic PEC/PV solar water-splitting device based on a GaInP2 photocathode and a GaAs PV 168 junction, showing a ηSTH of 12.4%. Three years later, Licht et al. showed a stacked multi- 202 junction solar water-splitting device based on AlGaAs/Si with the reported ηSTH of 18.3%. However, the device used scare and expensive materials, as well as complex device architectures. In addition, some of these high performance devices have a limited stability in their electrolyte. These undesirable features are likely to hinder their large-scale application for water splitting.

In search for earth-abundant alternatives, various PEC/PV junctions have been explored using stable metal-oxides such as doped tungsten trioxide (WO3) or bismuth vanadate (BiVO4) as 95, 97, 111, 166, 187, 201, 205 photoelectrodes with silicon (Si). To date, the highest reported ηSTH of 5.2% for metal-oxide-based devices was achieved using a combination of a gradient-doped W:BiVO4 photoanode with a double-junction micromorph silicon (a-Si:H/nc-Si:H) PV device.97 However, metal-oxides generally have relatively large band gaps (2.3-3 eV), which limits their ηSTH values to below 10%.

Hydrogenated amorphous silicon carbide (a-SiC:H) is a more attractive alternative with a band gap that can be tuned between 1.8 and 2.3 eV by varying the carbon concentration.206 For -2 example, a device with an a-Si0.9C0.1:H absorption layer can theoretically generate 15 mA cm (ηSTH ~18%) of an air mass (AM) 1.5 photocurrent based on its band gap energy of 2.0 eV. Further, since doped a-SiC:H is widely used as a window layer in the solar cell industry,207-210 a full monolithic PEC/PV integration between a-SiC:H and thin-film silicon (TF-Si) PV junctions is practically feasible. The processing conditions of a-SiC:H are fully compatible with that of the TF-Si PV junctions since both can be deposited using plasma-enhanced chemical vapor deposition (PECVD) techniques. No deposition at high temperature (only 170-200 °C) or post- annealing is necessary, in contrast to many metal-oxide photoelectrodes.95 In addition, TF-Si processing is highly scalable as already demonstrated by the TF transistor and PV industry. Finally, a-SiC:H is stable in an aqueous environment (even in a pH 0.3 acid),211, 212 and both Si and C are non-toxic and earth-abundant materials.

6.1 Introduction 79

Despite the advantages mentioned above, the PEC performance of a-SiC:H photocathodes reported in the literature to date is well below their theoretical potential.210, 211, 213-217 In 2007, Zhu et al. demonstrated a simple p/i a-SiC:H photocathode that was free of its native oxide and generated a photocurrent density of ~-0.6 mA cm-2 at 0 V versus the reversible hydrogen electrode (vs. RHE).210 Only at a negative bias of -1.5 V vs. RHE did the photocurrent density of reach ~-7 mA cm-2, showing the large overpotential needed to drive the a-SiC:H photocathode to higher photocurrent densities. One of the main challenges for this material is therefore to reduce this overpotential, which originates from surface recombination, poor material quality and low charge carrier collection efficiency (sometimes noted as carrier separation efficiency). Other important challenges are the low catalytic performance and poor spectral utilization when a- SiC:H is used in a tandem configuration. One attempt to improve the low catalytic performance of a-SiC:H by placing metal or metal-oxide nanoparticles on the electrode surface has been 218 -2 reported. The photocurrent increased to ~-1.8 mA cm at 0 V bias (two-electrode, vs. Ru2O counter electrode) using Ru-nanoparticles to catalyzed the hydrogen evolution reaction (HER) on an a-SiC:H photocathode combined with a double-junction a-Si:H solar cell. However, the current density was still an order of magnitude lower than the theoretical limit.

In this chapter, we apply several methods to reduce the overpotential of an a-SiC:H based photocathode. First, the charge carrier collection is improved by introduction of gradient boron doping in the a-SiC:H PEC junction. We then achieve an anodic shift of the onset potential by the monolithic integration of TF-Si PV junctions. Finally we introduce state-of-the-art glass substrates with integrated micro-textured photonic structures for the hybrid PEC/PV hybrid cathodes to improve light collection. These substrates allow the processing of TF-Si PV junctions exhibiting dense and high electrical-grade nc-Si:H and facilitate high spectral utilization in the PEC and PV junctions.

(a) (b) Figure 6.1 A Sketch illustration (a) and an SEM image (b) of the a-SiC:H photocathode integrated with the integration of nc-Si:H/nc-Si:H PV junctions.

We explore three different monolithic PEC/PV cathode configurations based on TF-Si single and double PV junctions. Figure 6.1 shows a schematic illustration of an a-SiC:H(PEC)/nc- Si:H(PV)/nc-Si:H(PV) cathode together with a cross-sectional scanning electron microscope 80 6. A TF-Si based monolithic PEC/PV cathode with efficient H2 evolution

(SEM) image of the processed cathode. The balance between the anodic shift and the optimized spectral utilization of the PEC/PV configuration is also studied. Simulations of the spectral utilization of the PEC/PV cathodes using an in-house developed Advanced Semiconductor Analysis (ASA) software supports the experimentally observed anodic shifts and current densities. Finally, the stability of the silicon-based photocathodes under operation was demonstrated.

6.2 Experimental

6.2.1 PECVD fabrication of photocathodes The a-SiC:H photocathodes were deposited using a radio frequency (RF-) PECVD multi-chamber tool. A 2.5 cm × 10 cm ASAHI VU-type substrate was heated at 170 °C during the deposition. The a-SiC:H(B) layer was deposited as the p-layer, decomposed from SiH4, CH4 and B2H6 diluted H2 gas flow under a controlled pressure. The gradient B doped “p--” layer was deposited by a programmed 10-step recipe in which a 2 sccm B2H6 diluted H2 gas flow rate was evenly reduced per 36 seconds from the processing conditions used for the initial p-layer until a 0 sccm B2H6 gas flow corresponding to typical i-layer processing conditions. A H2 treatment in a low RF-power intensity was followed in the same chamber in order to improve the p/i interface. The i-type a- SiC:H (optimized as a-Si0.9C0.1:H in the rest of this chapter, as discussed in Appendix B Table B.1, Figures B.1 and B.2) was deposited in another chamber specifically to avoid possible cross- contamination. After the a-SiC:H thin films were synthesized, a stripe of 300 nm Al was coated on the pre-covered region of the sample as the front contact using a PROVAC evaporator in a rotation mode. The purpose of the Al stripe is to ensure the effective collection of the generated photocurrent.

6.2.2 Glass with integrated micro-textured photonic structures and high quality nc-Si:H materials A 100 nm thick layer of ITO film was sputtered on a piece of Corning glass with a size of 2.5 cm × 10 cm. The glass was then dipped into an acid solution (41% HF : 50% H2O2 : H2O = 1:2:10 volume ratio) for 45 min. The ITO acted as the sacrificed layer for glass etching, and micro- textured features were formed. On the non-ITO-coated side, the etching rate was extremely slow and glass surface remained flat. An AZO layer with a thickness of 1.5-1.8 µm was sputtered on the textured glass, followed by 0.5% diluted HCl acid etching for 40 seconds. Small features of a few hundreds of nanometers were created on the AZO layer, as shown by in tiny “bubble” structures in the microscopy image in Figure B.3 (Appendix B). A metal layer consisting of 30 nm Cr and 100 nm Ag was evaporated on the top of the modulated surface textured glass substrate as the back reflector. Before the deposition of TF-Si layers, an 80 nm-thick AZO layer was sputtered on the Ag to improve the Si/metal interface. The sample was then transferred into the same PECVD tool and heated at 170 °C. 40 nm thick boron (B) doped nc-SiOx:H was used as the p-layer and 25 nm thick phosphorus (P) doped nc-SiOx:H was used as the n-layer in all photocathodes (A: a-SiC:H/a-Si:H, B: a-SiC:H/a-Si:H/nc-Si:H and C: a-SiC:H/nc-Si:H/nc-Si:H). In photocathode A, the bottom nc-Si:H i-layer is 3 µm, and the top nc-Si:H i-layer is 1 µm. In 6.3 Results and discussions 81

photocathode B, the bottom nc-Si:H i-layer is also 3 µm, but the top a-Si:H i-layer is 0.3 µm. For the best performance of the double-junction cells, the i-nc-Si:H material deposition requires a high VHF (very high frequency) power (40 W). In photocathode A, the a-Si:H i-layer is 0.3 µm. The 10 nm p-doped, 40 nm gradient reducing (p--) doped and 40 nm i-type a-SiC:H films were deposited immediately onto the top n-nc-SiOx:H layers of the three photocathodes, and the processing details are described in Section 6.2.1.

6.2.3 PEC characterization PEC characterization was carried out in a three-electrode configuration: (i) the a-SiC:H photocathode with an active area of 0.283 cm2 (an illuminated hole of 6 mm in diameter) as the working photocathode, (ii) a coiled Pt wire as the counter electrode, and (iii) a Ag/AgCl electrode (XR300, saturated KCl solution, Radiometer Analytical) as the reference electrode. 0.1 M (mol L- 1 ) sulfamic acid (H3NSO3) was utilized as the electrolyte, buffered to pH ~3.75 by potassium biphthalate (KHP). White light photocurrent measurements were performed under simulated Air Mass (AM) 1.5 solar illumination (100 mW cm-2) with a NEWPORT Sol3A Class AAA solar simulator (type 94023A-SR3).

6.2.4 ASA simulation ASA (Advanced Semiconductor Analysis) is a program developed by the PVMD group in TU Delft using MATLAB and C++. It is designed for the optical and electrical simulation of devices based on amorphous and crystalline semiconductors. The ASA program solves the basic semiconductor equations in one dimension (the Poisson equation and two continuity equations for electrons and holes). In this work, we built optical models in ASA to simulate the absorption spectrum of each PEC or PV junction illuminated by the AM 1.5 spectrum transmitted through the electrolyte and quartz window. The ellipsometry measured wavelength-dependent reflective index, n(λ), extinction coefficient, k(λ) and thicknesses of the layers were the optical input data for the simulation. The glass substrates with integrated micro-textured photonic structures were considered as the back reflector in the model as well. The short-circuit current density of each junction was integrated from the absorption spectrum and the AM 1.5 solar spectrum.

6.3 Results and discussions

6.3.1 Boron doping profiling in the a-SiC:H photocathode An a-SiC:H electrode generally has a fast charge carrier recombination and a short diffusion length due to high defect densities.219 Further Ohmic loss can be significant for thick films. The charge carrier recombination limits the maximum splitting of the quasi-Fermi levels in the a- SiC:H PEC junction, and subsequently results in larger effective overpotentials. The thicker the a- SiC:H film, the more charge carrier collection is limited by charge carrier recombination.

The effect of charge carrier recombination is demonstrated using j-V curves of the a-SiC:H based photocathodes measured under AM 1.5 (100 mW cm-2) standard test conditions in the three- electrode configuration as shown in Figure 6.2. Three PEC electrodes are considered, all involve a 82 6. A TF-Si based monolithic PEC/PV cathode with efficient H2 evolution

p-a-SiC:H (boron (B)-doped) layer interface with an i-a-SiC:H layer with a total thickness of 90 nm, where the intrinsic (not intentionally doped, more strictly speaking) a-SiC:H faces the electrolyte. Cartoons of the three are shown in Figure 6.2. The first junction, Figure 6.2 (a), consists of a thin p-a-SiC:H layer (10 nm) with an 80 nm thick i-a-SiC:H layer. The p/i junction separates the light-excited charge carriers and generates a light-induced potential. While drift is the dominant transport mechanism for the light-excited charge carriers in the depletion zone in the p-a-SiC:H layer, diffusion is the dominant transport mechanism in the i-layer. Since the i-layer is much thicker, the charge carrier transport in the PEC is most likely to be dominated by diffusion. As a result, charge carrier collection under light exposure is limited by recombination in the i- layer for this configuration (Figure 6.2 (a)). Thus a thin p-layer configuration results in a negative photocurrent onset potential (0 V vs. RHE), as seen in the j-V curve in Figure 6.2 (d) (blue curve).

In the second device (Figure 6.2 (b)), the thickness of the depletion zone is increased using a p- type gradient boron doping layer between the p- and i-layers. This enlarges the area where drift is the dominate mechanism of charge transport and enhances charge carrier collection, as fewer charge carriers are lost due to recombination in the diffusion-dominated region of the junction. This junction consists of a 10 nm thick p-a-SiC:H (B) layer, 40 nm thick gradient layer in which the concentration of boron doping (herein called “p--”) was reduced from value in the p-layer to 0% in 10 steps of 4 nm each, and a 40 nm thick i-a-SiC:H layer. Due to improved charge carrier collection, the onset potential shifts anodically by ~200 mV as compared to the thin p-layer configuration (Figure 6.2 (a)).

(d) 0

-2 p-layer

)

-2 (c) thick p-layer -4 p-layer (a) thin

(mA cm j -6 (b) gradient

-8 -1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 E (V) vs. RHE

Figure 6.2 Schematic presentation demonstrating the variation of the doping concentration with thickness in the p/i junction (a-c). Three-electrode AM 1.5 j-V curves with and without gradient B doping (d).

To exclude the possibility that the enhancement is instead caused by a thicker doped layer and not the gradient doping, a “thick p-layer” configuration (Figure 6.2 (c)), which consists of a 50 nm thick p-a-SiC:H (B) layer and a 40 nm thick i-a-SiC:H layer, has been tested as well. For this 6.3 Results and discussions 83

configuration the onset potential shifts cathodically, showing that now the recombination in the more defective p-layer dominates over the recombination in the thinner i-layer. We can therefore conclude that the improvement that we observed is indeed caused by the introduction of gradient doping in our film.

The “gradient p-layer” photocathode has a significant current density in the order of ~-0.8 mA cm-2 at 0 V vs. RHE. This is a 33% improvement compared to the much thicker (~220 nm) a- SiC:H photocathode with photocurrent density of ~-0.6 mA cm-2 at 0 V vs. RHE reported by Zhu et al.216 At -1.0 V vs. RHE the a-SiC:H photocathode in Figure 6.2 has a current density of -7.2 mA cm-2 considering its thickness of only 90 nm. Although the potential of -1.0 V vs. RHE is of little relevance, this photocathode is 100% higher than that of any other reported a-SiC:H photocathodes.216

6.3.2 Monolithic PEC/PV cathode As illustrated in Figure 6.2, although the a-SiC:H photocathode can generate a high photocurrent density at negative potentials (≤ -1 V vs. RHE), it only generates a current density of -0.8 mA cm- 2 at 0 V vs. RHE. The challenge is therefore to anodically shift the onset potential of the j-V curves to more positive potentials. Herein we achieve this shift by the monolithic integration of a TF-Si PV device between the textured substrate at the back contact and the photocathode. The extra photovoltage from the PV junction(s) is achieved to provide the overpotential at the semiconductor/electrolyte interface. To illustrate this mechanism, the corresponding energy band diagram of an a-SiC:H(PEC)/a-Si:H(PV)/nc-Si:H(PV) photocathode is shown in Figure 6.3. The flat band potential is characterized by electrochemical impedance spectroscopy (EIS) as illustrated in Figures B.4 and B.5 (Appendix B).

To successfully integrate a PV cell into a tandem PEC/PV cathode, the PV device needs to meet several requirements. Firstly, the spectral utilization of the photocathode and solar cells should be equally distributed among all the PEC and PV junctions. Secondly, the voltage at the operating point (VOP) of the solar cells should be as large enough to overcome the overpotential. Based on these considerations, we integrated three types of TF-Si solar cells with the a-SiC:H, i.e., a single- junction a-Si:H solar cells (VOC ~ 0.9 V) (photocathode A), a micromorph (a-Si:H/nc-Si:H, VOC ~ 1.4 V) solar cell (photocathode B) and a nano-crystalline double-junction (nc-Si:H/nc-Si:H, VOC ~ 1.0 V) solar cell (photocathode C) . An illustrative sketch of photocathode C is illustrated in Figure 6.1 (a).

To guarantee device-grade materials and high spectral utilization, we introduce a special substrate at the back contact for the monolithically integrated photocathodes. Textured substrates are the conventional approach to enhance light trapping in PV devices.220, 221 However, nc-Si:H is highly sensitive to an extreme texture, as the texture initiates the growth of defect-rich filaments above sharp valleys.222 These filaments significantly reduced the fill factor (FF), open-circuit voltage and current density of the PV junctions.222, 223 To prevent the incorporation of defect-rich filaments and to allow the integration of thick high quality nc-Si:H junctions in our PEC/PV cathodes, we use state-of-the-art glass substrates with integrated micro-textured photonic 84 6. A TF-Si based monolithic PEC/PV cathode with efficient H2 evolution

structures. These photonic structures are obtained by wet-etching a sacrificial ITO layer on Corning glass, resulting in craters with diameters on the order of 10 µm and depths of 2 µm. The back-reflector, PV and PEC junctions are deposited on top and adopt the micro-textured photonic structures. As demonstrated in the SEM image in Figure 6.1 (b), the glass substrate with micro- textured photonic structures allows the growth of highly dense nc-Si:H material, without defect- rich filaments. This structure guarantees high VOC values, current densities and excellent light trapping for the PEC/PV cathodes.

Figure 6.3 A band diagram illustration of the multi-junction a-SiC:H/a-Si:H/nc-Si:H cathode under non-biased illumination conditions. The band edges of the outermost layer (a-SiC:H) are assumed to be pinned at the semiconductor/electrolyte interface.

To determine the performance of the PEC/PV cathodes, the samples are immersed in the H3NSO3 electrolyte buffered at pH 3.75 and are illuminated by a solar simulator (AM 1.5, 100 mW cm-2). In a three-electrode configuration, the ability of the photocathode to drive the hydrogen evolution reaction (HER) is evaluated. The PEC characterizations of the a-SiC:H photocathodes integrated with the three different TF-Si solar cells are shown in Figures 6.4 (a)-(c). The current density in the dark for all the samples is negligible (black dashed curves).

The introduction of an a-Si:H single-junction solar cell (photocathode A) shifts the onset potential anodically by ~0.8 V (green curve in Figure 6.4 (a)). This shift is in agreement with typical values for the operating voltage VOP of a single a-Si:H PV junction. For photocathode B with the a- Si:H/nc-Si:H solar cell (red curve in Figure 6.4 (b)), the anodic shift of the onset potential is ~ 1.2 V, which again corresponds to typical values for the operating voltage VOP of an a-Si:H/nc-Si:H solar cell. The onset potential of the photocathode C with the double-junction nc-Si:H/nc-Si:H cell is approximately 1.0 V (blue curve in Figure 6.4 (c)). In contrast to the other two 6.3 Results and discussions 85

photocathodes, this shift is larger than the typical operating point of a nc-Si:H/nc-Si:H double PV junction and is close to the typical VOC values of ~1.0 V for such cells. This implies that the a- SiC:H is the current limiting junction in photocathode C. Under such operating conditions the PV junctions operate at voltages close to their open-circuit voltage VOC. The agreement between the anodic shifts in the j-V curve and the VOP or VOC of the three solar cell devices shows that no additional potential barriers are created due to the monolithic integration of PV junction(s) with the PEC junction.

0 1.0 dark a a-SiC:H d 0.8V 0.8 -2 6.5 a-Si:H a-SiC:H 7.1 0.6

-4 (A) a-SiC:H/ 0.4 -6 a-Si:H (A) 0.2 0 0.0 ) a-SiC:H e 1.2V b -2 0.8 -2 6.1 nc-Si:H a-Si:H 0.6 4.9 10.7 -4 0.4 -6 (B) a-SiC:H/ (B) 0.2 (mA cm a-Si:H/nc-Si:H

j 0.0 Absorption 0 f 1.0V c a-SiC:H 0.8 -2 6.0 nc-Si:Hnc-Si:H 0.6 8.0

-4 8.1 (C) a-SiC:H/ 0.4 -6 nc-Si:H/nc-Si:H (C) 0.2 0.0 -1.0-0.5 0.0 0.5 1.0 400 600 800 1000 E (V) vs . RHE (nm)

Figure 6.4 AM 1.5 photocurrent density-voltage curves of photocathodes integrated with various PV solar cells (a)-(c). Spectral utilization of various PEC/PV configurations by ASA simulation (d)-(f). The numbers below a-SiC:H indicate the maximum current density (mA cm-2) integrated from the absorption spectrum of the a-SiC:H layers, and the numbers below the solar cell junction (a-Si:H and/or nc-Si:H) indicate the short-circuit photocurrent density (mA cm-2) of each junction.

Since all the current densities of the three photocathodes are close to saturation at -1 V vs. RHE, we define a saturation current density, Jsat, as the j measured at -1 V vs. RHE. Since the PEC and PV junctions in the monolithically integrated photocathodes are in-series-connected, the Jsat is determined by the current density of the limiting junction. The highest Jsat among the three PEC/PV cathodes is -7.3 mA cm-2, achieved by the a-SiC:H/nc-Si:H/nc-Si:H (photocathode C) structure. The Jsat values in the cases of a-SiC:H/a-Si:H/nc-Si:H (photocathode B) and a-SiC:H/a- Si:H (photocathode A) are relatively lower: -5.3 and -5.8 mA cm-2, respectively. 86 6. A TF-Si based monolithic PEC/PV cathode with efficient H2 evolution

6.3.3 Analysis of spectral utilization In a monolithically interconnected photocathode, it is highly challenging to determine the quantum efficiency of each individual junction. Therefore, we simulate the spectral utilization using an in-house developed Advanced Semiconductor Analysis (ASA) software173 to support our measurements and study the potential current densities of these photocathodes. The ASA software is an established tool to model Si-based solar devices and excellent agreement between experiments and simulations has been demonstrated. Based on this we can state that the systematic relative error in the calculated current density is not higher than 10%.224-226 Optical models of multi-layers representing the various PEC/PV configurations are constructed in ASA. Measured wavelength-dependent reflective index, n(λ), extinction coefficient, k(λ) and thicknesses of the layers are used as the optical input data for the simulation. The wavelength- dependent reflectance R(λ) and transmittance T(λ) simulations have been performed for the three PEC/PV cathodes: a-SiC:H/a-Si:H (photocathode A), a-SiC:H/a-Si:H/nc-Si:H (photocathode B), and -SiC:H/nc-Si:H/nc-Si:H (photocathode C). The absorption spectrum A(λ) for each junction can be calculated as ART( ) 1  (  )  (  ) (6.1) and the simulated results are illustrated in Figures 6.4 (d)-(f).

For photocathode A, the ASA simulations show that the a-SiC:H PEC junctions is almost current matched with the a-Si:H PV junction (6.5 vs. 7.1 mA cm-2). The maximum total spectral utilization (sum of the absorption of each individual junction) is determined by the junction with the lowest band gap. In photocathode A, this is the a-Si:H cell. Therefore, photocathode A has a total current density of 13.6 mA cm-2 which is already close to that of a typical state-of-the-art single a-Si:H PV junction with current densities of 15-16 mA cm-2 under 100 mW cm-2 AM 1.5 conditions. The maximum total spectral utilization of photocathodes B and C is limited by the low band gap of the nc-Si:H junction(s). The total spectral utilization of photocathode B is 21.7 mA cm-2 which is a value approaching the current density generated in a state-of-the-art single nc- Si:H PV junctions of 25-31 mA cm-2.227 However, the overlap of the spectral response of the a- Si:H PV junction with that of the a-SiC:H PEC and nc-Si:H PV junction is significant, making the a-Si:H junction the current limiting one (~ 5 mA cm-2). Photocathode C is the only configuration in which the photocathode is strongly current limiting (∆J = 2 mA cm-2). This supports the earlier mentioned observation that the PV junctions of photocathode C are forced to operate at a voltage close to the VOC. This simulation shows that there still is plenty of room to improve the spectral utilization of photocathode C.

We then compare the photocurrent densities of photocathodes measured at 0 V vs. RHE (J0). The -2 J0 at this catalyst-free photocathode is the highest for photocathode B with J0 = -5.1 mA cm , as -2 -2 compared to photocathode A (J0 = -4.0 mA cm ) and photocathode C (J0 = -4.5 mA cm ). The results demonstrate that the value for J0 of these PEC/PV cathodes is a competition between maximum utilization of the band gap energy of the PV device to facilitate the anodic shift and the maximum spectral utilization to facilitate the highest current densities. High band gap materials like a-Si:H provide high voltages but low currents, whereas low band gap materials like nc-Si:H 6.3 Results and discussions 87

provide low voltages and high currents. To quantify the performance of the photocathodes further, 228 we define a second external parameter, i.e. the maximum efficiency of the photocathode PC as :

jmax() V max VRHE PC 100% (6.2) I0 where jmax and Vmax are the current density and voltage at the maximum power point, respectively.

We find PC = 2.1%, 1.3% and 0.8% for photocathode A, B and C, respectively, which are promising results considering that these a-SiC:H photocathodes are not deposited with any HER catalysts or passivation layers. By depositing Ru nanoparticles as the catalysts, an a-SiC:H/a-

Si:H/a-Si:H photocathode can indeed enhance the PC from 0.125% to 1.9%, as reported by Zhu et al.218 Therefore, next to further improving carrier collection, the catalytic activity can be significantly improved by depositing HER catalysts on the a-SiC:H. This will result in a large additional anodic shift and significantly higher values for PC.

Based on these results, the realistic potential of the spectral utilization of these PEC/PV cathodes can be estimated. If only a single a-Si:H PV junction is integrated with the a-SiC:H photocathode, -2 the theoretical maximum Jsat value is 24/2=12 mA cm assuming that all the photons in the solar spectrum (until 800 nm) are converted into electron-hole-pairs and the current is equally distributed between the a-SiC:H photocathode and the a-Si:H solar cell. However, a more realistic approach would be to take the highest spectral utilization for an a-Si:H single-junction of 18 mA -2 229 -2 cm achieved to date, which would imply a Jsat of ~9 mA cm . When the double-junction solar cell with nc-Si:H material is integrated with the a-SiC:H photocathode, the theoretical maximum -2 Jsat value can achieve 42/3 = 14 mA cm . In view of the nc-Si:H PV junction with the highest current utilization achieved up to date (~32 mA cm-2), this overall photocathode photocurrent density would be 10.7 mA cm-2. This simple analysis therefore shows that a 13% efficient solar- to-hydrogen conversion device is well within reach with the combination of an a-SiC:H photocathode and a TF-Si solar cell.

6.3.4 Stability of thin-film silicon based PEC/PV cathode An advantage of the a-SiC:H photocathode is its relatively long-term stability in an acidic medium. As reported by Hu et al., their a-SiC:H photocathode showed a stable photocurrent 212 density after 800 h illumination in pH 0.3 H2SO4 acid. .

Herein we also show the stability of our a-SiC:H photocathode in a pH 3.75 sulfamic acid electrolyte. Figure 6.5 (a) shows the chronoamperometry measurement of photocathode B biased at 0 V vs. RHE. We observe a slight current density enhancement in the first 2 min of illumination, which is a result of the acidic electrolyte etching the thin native oxide layer on the i- a-SiC:H surface. The following “periodical” fluctuation is caused by repetitive accumulation and dislodging of H2 bubbles at the photocathode surface. Nevertheless, the photocurrent is relatively stable over the course of an hour. A photograph of photocathode A before and after the one-hour 88 6. A TF-Si based monolithic PEC/PV cathode with efficient H2 evolution

measurement is shown in Figure 6.5 (b), and the purple surface of the sample remains unchanged after immersing in the acidic electrolyte for one hour.

-1

) -2

-3

-4

(mA/cm

-5

-6 0 10 20 30 40 50 60 time (min) (a)

(b) Figure 6.5 The photocurrent density of the a-SiC:H/a-Si:H/nc-Si:H photocathode monitoring measurement in pH 3.75 acid illuminated by AM 1.5 chopped spectrum for 1 hour (a). The surface of the photocathode remains unchanged before (right) and after (left) 1h measurement (b). The orange dot circle with a diameter of 6 mm indicates the active area of the photocathode (0.283 cm2 in total area).

The photocurrent density of -5.1 mA cm-2 at 0 V vs. RHE is the highest photocurrent reported for any a-SiC:H photocathode. The record photocurrent is achieved without catalysts or passivation layers added to the a-SiC:H photocathode. Combined with its stability in acidic solution, our results demonstrate the high potential of thin-film a-SiC:H as a photocathode.

6.4 Conclusions

A catalyst- and precious-metal-free a-SiC:H(PEC)/a-Si:H(PV)/nc-Si:H(PV) water-splitting photocathode with a current density of -5.1 mA cm-2 at 0 V vs. RHE and a photocathode 6.4 Conclusions 89

efficiency of 2.1% has been achieved. This PEC/PV cathode based on earth-abundant materials is highly stable for over 1 hour in an acidic electrolyte under simulated AM 1.5 illumination. The overpotential of an a-SiC:H PEC photocathode is reduced by the introduction of gradient boron doping in the a-SiC:H PEC junction, enhancing the charge carrier collection. Dense TF-Si layers exhibiting electrical performance and high spectral utilization in the PEC and PV junctions are achieved by using state-of-the-art glass substrates with integrated micro-textured photonic structures. By using three different monolithically PEC/PV integrated photocathode configurations, the extent of the PV-induced anodic shift is demonstrated. The balance between the anodic shift and the optimized spectral utilization of the PEC/PV configuration is further studied using simulations of the spectral utilization of various PEC/PV cathodes. Based on the presented work, a-SiC:H/TF-Si based PEC/PV cathodes with current densities above 10 mA cm-2 are within reach.

90 6. A TF-Si based monolithic PEC/PV cathode with efficient H2 evolution

7. Nano-structured platinum synthesized by atomic layer deposition as hydrogen evolution reaction catalysts1

Atomic layer deposition (ALD) was used to deposit nanoparticles and thin films of Pt onto etched p-type Si(111) wafers and glassy carbon discs. The growth rate was 0.8–1.0 Å/cycle using MeCpPtMe3 and ozone as precursors and a temperature window of 200–300 °C. X-ray photoelectron spectroscopy (XPS), atomic force microscopy (AFM), and scanning-electron microscopy (SEM) were used to analyze the composition, structure, morphology and thickness of the ALD-grown Pt nanoparticle films. The catalytic activity of the ALD-grown Pt for the hydrogen-evolution reaction was shown to be comparable with e-beam evaporated Pt on glassy carbon electrode.

1 This chapter has been submitted: R. Liu, L. Han, Z. Huang, I.M. Ferrer, A.H.M. Smets, M. Zeman, B.S. Brunschwig and N.S. Lewis, Thin Solid Films, 2014 (under review). 92 7. Nano-structured Pt synthesized by ALD as HER catalysts

7.1 Introduction

The efficient operation of a photoelectrochemical solar fuels generator based on water-splitting requires catalysts to facilitate charge-transfer reactions at the interface between the semiconducting light absorbers and the electrolyte. The preferred catalysts of the hydrogen- evolution reaction (HER) are metals and metal alloys, which can be deposited onto the photocathodes of photoelectrochemical cells either as nanoparticles (NPs) or as thin films.176, 230, 231 Platinum is the archetypal catalyst for the HER because Pt exhibits a relatively low overpotential (< 40 mV) to drive the HER at a current density of 10 mA cm-2.230, 232-234 The deposition of catalytically active metals onto a semiconductor electrode surface, such that it preserves the electronic properties of the semiconductor/liquid junction, is an important step in the development of a photoelectrochemical water-splitting device.

Thin films of Pt, with thicknesses from a few nm to tens of nm, can be deposited onto surfaces by sputtering235, electron-beam evaporation236, chemical vapor deposition (CVD)237, or electrochemical deposition230. Nevertheless, a method that allows deposition of highly catalytic Pt films of readily controlled thickness without the use of high temperatures or a high-vacuum environment is desirable. Atomic-layer deposition (ALD) has been widely employed to grow thin-film metal-oxide light absorbers182, 238, 239, to deposit protective layers240, 241 onto surfaces, and to deposit catalysts onto photoelectrodes.185, 242 ALD provides facile control over the thickness of the deposited films, produces films with excellent uniformity, and can be applied to structures with high surface areas. ALD also offers the advantage of being a relatively low- temperature process that does not require a high-vacuum environment or post-deposition annealing. Pt nanoparticles have been deposited onto a TiO2-coated Si nanowire structure used in a water-splitting device by ALD using (trimethyl)methylcyclopentadienylplatinum(IV) 243 (MeCpPtMe3) and oxygen gas as precursors. Similar Pt nanoparticles, grown by ALD using the same precursors, have been deposited on Si nanowires as catalysts for sunlight-driven HER.244 In both cases, 10–50 cycles of growth were required to deposit nanoparticles ~2–3 nm in diameter with surface coverages in the low sub-monolayer range.243, 244 The wider use of ALD for deposition of Pt has been limited by either slow growth rates (< 0.5 Å/cycle) or high temperatures (typically 250-350 °C) required.243-248 Although Pt ALD growths with ozone as oxidizer were achieve at temperatures as low as 120-140 °C, the growth rate was only ~0.3 Å/cycle.247, 249, 250 Faster growth rates at low growth temperature for metallic Pt ALD growth are desired.

Herein we describe an ALD method to deposit Pt NPs or thin films onto p-type Si wafers or carbon discs. This method results in a growth rate of ~ 0.8–1.0 Å/cycle using MeCpPtMe3 and ozone as precursors at temperatures of 200–300 °C. Concurrently with the execution of this work, improved rates of ALD growth (0.7 Å/cycle) of Pt have been reported very recently using the same precursors on oxidized Si(0001) surfaces.251 Our work complements these results by including characterization of the electro-catalytic activity of the ALD Pt films on glassy carbon rotating disk electrodes. X-ray photoelectron spectroscopy (XPS), atomic-force microscopy (AFM), and scanning-electron microscopy (SEM) were used to analyze the composition, structure, morphology and thickness of the ALD-grown Pt materials. 7.2 Experimental 93

7.2 Experimental

7.2.1 Preparation of substrates Boron-doped p-type Si(111) wafers with a resistivity of 20 Ω cm (Silicon Resource Company) were cleaned using an RCA1 procedure by soaking in a 5:1:1 (by volume) solution of deionized water, ammonium hydroxide (27%, Sigma Aldrich), and hydrogen peroxide (30%, Sigma Aldrich) for 10 min. The wafers were rinsed with deionized water, immersed in an RCA2 solution, 5:1:1 (by volume) deionized water, hydrochloric acid (27%, Sigma Aldrich), and hydrogen peroxide (30%, Sigma Aldrich) for 10 min, and then immersed in an aqueous NH4F solution (40%, Sigma Aldrich) for 30-40 min to obtain H-terminated, terraced surfaces. A stream of nitrogen was used to dry the wafers, which were then immediately transferred into the ALD chamber for metal deposition.

5 mm in diameter and 4 mm thick glassy carbon disks (Sigradur G, HTW Hochtemperatur- Westoff GmbH) were polished with 600 grit Carbimet SiC paper (Buehler) at 150 rpm for 1 min using a Stuers LaboPol-5 polisher, and then immersed in 1.5 M HNO3 (69% HNO3, Sigma Aldrich). The disks were then polished at 150 rpm using an MD-Floc synthetic nap polishing pad (Stuers) and a sequence of diamond slurries (9, 6, 3, 1 and 0.1 µm MetaDi Supreme, Buehler) for 1 min each. After polishing, the disks were sonicated sequentially for 10 min each in water, acetone, isopropanol, and water.

7.2.2 Atomic-layer deposition of platinum Pt deposition was performed using a Savannah ALD system (Cambridge Nanotech). The substrates were heated to 200–300 °C in a chamber held at a constant pressure of 0.2–0.3 Torr. The precursor, MeCpPtMe3 (Strem Chemicals, 99% pure), was heated to 75 °C and was swept into the ALD chamber using four consecutive pulses of N2 (g) at a flow rate of 10 sccm. The total duration of exposure to the precursor per cycle was varied across samples, in the range of 0.25– 1.5 s, and was divided evenly among the four consecutive pulses to the dependence of the growth rate on duration of the pulse. The remaining MeCpPtMe3 was purged from the system for 10 s using nitrogen gas, and then ozone (O3) from the ozone generator built into the ALD system was admitted for 0.1 s. Either 15, 20, 25, 35, 50 or 100 cycles for the ALD growth, with each cycle consisting of a set of four pulses of MeCpPtMe3 followed by a pulse of ozone, were used, to obtain films with varied thicknesses and morphologies. Upon completion of ALD, the samples were removed from the chamber and stored under ambient conditions until analyzed.

To determine the rate of growth of the Pt films as a function of the duration of the precursor pulses, the samples were removed from the ALD chamber after every 50 cycles of growth and the thickness of the film was determined using ellipsometry. The thickness of the Pt films was determined using ellipsometry (α-SE, J.A. Woollam Co. Inc.), from data collected with the incident beam and detector located at either 60°, 65° or 70° relative to the surface normal. The samples were then reloaded into the ALD chamber for additional cycles of growth. 94 7. Nano-structured Pt synthesized by ALD as HER catalysts

7.2.3 Deposition of platinum films by electron-beam evaporation Electron-beam evaporation of Pt onto fully assembled electrodes was performed under vacuum using a Denton Explorer evaporator system with a base pressure of < 10−5 torr. The Pt (Kurt J. LESKER 99.99%) was deposited at a rate of 0.1–1 Å s−1 to a thickness of ~ 3 nm.

7.2.4 Characterization of deposited platinum films An atomic-force microscope (Bruker Dimension-Icon, Bruker) with a Nanoscope V controller was used to collect and analyze images of the deposited Pt films and substrates. X-ray photoelectron spectra were collected using a Surface Science M-Probe Spectrometer, controlled by Hawk Data Collection software (Service Physics, Bend OR; V7.04.04). The Al K α-line at 1486.6 eV was used for excitation. Ar-ion etching of each sample was performed in the XPS chamber for 45 s prior to the collection of spectra, to remove any adsorbed moisture, O2 or CO2. Scanning-electron microscopy was performed using a Nova NanoSEM 450 with an accelerating voltage of 15 kV.

7.2.5 Electrochemistry Rotating-disk voltammetry was performed using the samples prepared on the carbon-disk substrates. The carbon-disk samples were mounted in an E-6 series ChangeDisk rotating disk electrode assembly (Pine Instrument Company). A conventional three-electrode configuration was employed in a modified U-cell, based on a previously published experimental procedure. Briefly, the U-cell consisted of two chambers separated by a fine-porosity glass frit. One chamber contained ~120 mL of 1.0 M H2SO4 (aq), the carbon-disk working electrode, and a saturated calomel reference electrode (SCE, CH Instruments) that was calibrated against dilute ferrocene monocarboxylic acid (Sigma Aldrich, 97% pure) in 0.2 M phosphate buffer (pH 7, BioUltra) prior to use. The other chamber contained ~25 mL of 1.0 M H2SO4 and a carbon rod (Alfa Aesar, 99.99% pure) auxiliary electrode. An MSR rotator (Pine Instrument Company) was used to rotate the working electrode at 1600 RPM during the electrochemical measurements. The electrolyte was purged for 30 min with hydrogen (UHP, Air Liquide) prior to measurements, and hydrogen was continuously bubbled through the electrolyte during measurements. The electrodes were controlled by a Bio-Logic SP200 potentiostat-galvanostat, and the data were collected and analyzed using the Bio-Logic EC-Lab software. The surface area of each rotating disk electrode that was exposed to the electrolyte was 0.195 cm2.

7.3 Results

7.3.1 Growth rate and film morphology Boron-doped p-type Si(111) wafers were prepared by cleaning with RCA1 & RCA2, then rinsed 252 and etched with NH4F (see Appendix C for details) to obtain H-terminated, terraced surfaces. The wafers were dried and immediately transferred into the ALD chamber for metal deposition. Figure 7.1 (a) shows the dependence of the growth rate on the total duration of exposure to the Pt precursor per growth cycle for deposition onto substrates held at 250 °C. Pt precursor exposures 7.3 Results 95

of ≤ 0.25 s/cycle only yielded scattered Pt NPs, while exposures of ≥ 1.0 s/cycle yielded films with growth rates exceeding 0.8 Å/cycle. The growth rates were highly reproducible across samples. Figure 7.1 (b) shows the dependence of the growth rate on the temperature of the substrate, for growths using a precursor exposure of 1.0 s/cycle. For temperatures < 180 °C, growth rates decreased significantly and were difficult to estimate due to the low deposition of Pt. The rate of deposition did not increase at temperatures > 250 °C. Figure 7.1 (c) shows the dependence of the thickness of the Pt films on the number of ALD cycles, for a substrate temperature of 250 ⁰C and using a precursor exposure of 1.0 s per growth cycle. There is an initial delay of film growth of ~35 cycles as observed for other noble metal ALD growth processes.253 Under these growth conditions, after the initial delay, the thickness of the film increased at a constant rate of 1.1 ± 0.1 Å/cycle. Figure 7.1 (d) shows a SEM image of a cross-section of a Pt film on a Si(111) substrate. The film was obtained using a precursor exposure of 1.0 s per cycle for 500 cycles of growth at 250 ⁰C, and was about 45 nm thick.

Figure 7.1 The ALD characteristics for Pt deposition. (a) Dependence of the growth rate on Pt precursor exposure time for substrates held at 250 ⁰C. The total duration of exposure to the MeCpPtMe3 precursor per cycle [4 × pulse duration] is shown on the x-axis. (b) Dependence of the growth rate on the temperature of the substrate for growths using a precursor exposure time of 1.0 s/cycle. (c) Dependence of the Pt thickness on the total number of precursor exposure cycles for growths using a substrate temperature of 250 ⁰C and a precursor exposure time of 1.0 s/cycle. An initial delay in film growth of ~35 cycles was observed. (d) Cross-sectional SEM image of a 45 nm thick Pt film on a Si(111) substrate obtained from a growth at 250 ⁰C and a precursor 96 7. Nano-structured Pt synthesized by ALD as HER catalysts

exposure time of 1.0 s/cycle for 50 cycles of growth. For Figure 7.1 (a) and (b) growth rates were measured after 50 cycles.

AFM images showed that ALD growths of Pt films were preceded by the formation of Pt NPs (Figure 7.2). Images of the initial Si(111) surface, before ALD growth, showed smooth terraces with a roughness of < 0.20 nm (Figure 7.2 (a)). The Si terraces were still observable in AFM images after the deposition of Pt NPs (Figure 7.2 (b)-(d)), suggesting that ozone did not significantly disturb the morphology of the surface. Herein, the size/orientation of the Si terraces difference from Figure 7.2 (a)-(f) was not due to the ALD process, but only because they were taken from different parts of the surface. Both the number of the Pt NPs and the coverage of the substrate surface increased with the number of ALD cycles. The density, and height of the Pt NPs grew during the first 35 cycles, as measured by AFM topology scans (from 1.4 ± 0.1 nm high for 10 cycles to 3.5 ± 0.5 nm for 35 cycles). The Pt NPs started as clusters on the surface, and then with increasing growth cycles expanded into larger islands (Scheme 7.1), consistent with recent observations for ALD growth of Pt on highly oriented pyrolytic graphite.253 For growths at 250 °C after 35 cycles, NPs with a diameter of ~ 3-4 nm fully covered the Si surface (Figure 7.2 (f)), and the thickness of the film could be determined by ellipsometry. After 200 cycles of growth, the Pt film uniformly covered the surface and exhibited a surface roughness of ~ 1 nm (Figure C.1 in Appendix C), consistent with the roughness measured by ellipsometry (~1.3 nm). Similar morphology changes were observed for Pt growth at temperatures between 200 and 300 °C.

Figure 7.2 AFM images of the surface morphologies of the NH4F-etched Si(111) substrates with (a) 0, (b) 10, (c) 15, (d) 20, (e) 25 and (f) 35 ALD cycles of Pt deposition.

7.3 Results 97

Scheme 7.1 Schematic illustration of the growth of Pt NPs on a substrate and the formation of a continuous film.

7.3.2 Surface characterization A Pt film deposited on glassy carbon using 50 ALD cycles at 250 °C was characterized by XPS, and compared with a sample prepared using electron-beam evaporation, which was expected to produce a Pt film with no organic material. A glassy carbon substrate was chosen for the XPS analysis, rather than a Si substrate, to avoid binding-energy shifts due to band bending at the Pt-Si interface. Pt 4f7/2 peaks were observed at ~71.2 eV for both the ALD and evaporated Pt samples, while no peaks were observed for PtO (or Pt(OH)2) and PtO2, which are shifted by ~1.2 and ~3 eV, respectively, relative to Pt metal (Figure C.2).254 XPS depth profiling showed metallic Pt composition throughout the ALD film (Figure C.3). No carbon signal was detected on ALD- grown Pt sample (Figure C.4) implying that the Pt film contained only small amounts of residual ligand left from the MeCpPtMe3 precursor.

Underlying layers of the films were examined following argon-ion etching performed inside the XPS chamber. After 45 s of plasma etching, the carbon 2p peak was visible, while the Pt 4f7/2 peak did not shift (Figure C.3). No oxygen was detected on the surface before or after etching, and thus very little or no Pt oxide was present. The position of the Pt 4f7/2 peaks did not change when the deposition temperature was 200 or 300 °C. The Pt complex precursor, MeCpPtMe3 and Pt salts have binding energies ≥ 72.5 eV, and no evidence of these materials was found in the XP spectra of the samples.255 Similarly, no Pt oxide was observed on depth profile analysis of ALD Pt deposited on Si substrates (Figure C.5).

7.3.3 Catalytic activity for the hydrogen evolution reaction

Figure 7.3 shows the representative rotating disk voltammograms in 1 M H2SO4 (aq) saturated with H2 gas, for Pt films deposited by electron-beam evaporation and by ALD onto glassy carbon substrates. Figure 7.3 shows the electrochemical performance of Pt films deposited by ALD and by electron-beam evaporation. 3 nm of films prepared using both ALD and electron-beam evaporation exhibited similar electrochemical behavior and required an overpotential of -30 mV (

 -2 HH2 ) to achieve a current density of 10 mA cm (geometric surface E(1M H24 SO ) 241 mV vs . SCE area). The electrochemistry results for samples prepared using the two deposition methods were within experimental error of each other.

98 7. Nano-structured Pt synthesized by ALD as HER catalysts

Figure 7.3 Electrochemical performance of glassy carbon-disk electrodes with 3 nm of Pt deposited by ALD and by electron-beam evaporation in a H2-saturated 1.0 M H2SO4 solution. The disk electrodes were rotated at a rate of 1600 rpm, and the scan rate was 50 mV s-1.

7.4 Discussions

The growth rate (1.1 Å/cycle) of the Pt films (after initial 50 cycles’ inhabitation) achieved using the ALD method described herein is significantly greater than that observed either for use of 243-248 Pt(acac)2 (0.5 Å/cycle) or MeCpPtMe3 and oxygen (0.1-0.5 Å/cycle). In addition, some of the growth rates for ALD of Pt found in the literature were calculated based on NPs,243-245 and thus represents significantly less deposition than the continuous films grown in this work. The principal difference between the method described herein and most of the methods previously reported is the use of the stronger oxidant ozone instead of oxygen. During the ALD process, the ligands on the MeCpPtMe3 complex are removed by oxidation during the ozone cycle, leaving metallic Pt on the surface,249, 250, 256, 257 in accord with other recently reported observations on similar systems.23 The increased rate of Pt deposition at temperatures at lower temperatures then those used for other high rate deposition methods, and without the need for the plasma treatments of thermal ALD processes,258 should reduce the corrosion or oxidation of substrates that can occur during deposition. Moreover, the ability to deposit a continuous catalyst film that completely covers a substrate could allow photoelectrode materials that are unstable under water-splitting conditions to be protected against electrochemical corrosion by preventing direct physical contact between the unstable semiconductor and the electrolyte.

Optimal growth conditions were achieved using a chamber temperature of 250 °C, and a MeCpPtMe3 exposure duration of 1.0 s (4 pulses of 0.25 s each) followed by an ozone exposure of 0.1 s for each cycle. These conditions yielded continuous thin films of Pt with consistent growth rates for up to 500 cycles (the greatest number of cycles used for this work).

7.5 Conclusions 99

The temperature window of 200–300 °C provides advantages for other applications, such as the fabrication of patterned electrodes and other applications that require compatibility with other materials. For example, the polymer mask material SU-8, commonly used for photolithography, undergoes glass transitions at temperatures above 200 °C, and therefore an ALD process that allows a catalyst to be deposited at a temperature lower than 200 °C would be useful in the fabrication of well patterned catalyst/semiconductor heterojunctions for photoelectrochemical applications. Compared with other deposition methods, ALD gives well-controlled deposition, and is a facile process that requires relatively low temperatures and that has potential for use in large-scale fabrication of materials.

7.5 Conclusions

ALD was used to grow Pt nanoparticles and thin films of Pt NPs were under varied growth conditions. Using the precursors MeCpPtMe3 and ozone and temperatures as low as 200 °C, a growth rate of 1.1 Å/cycle was achieved after 50 cycles of inhibition time. AFM and SEM showed that the Pt films grown using the ALD method were uniform over the surface of the sample, and XPS characterization confirmed the composition of the films as metallic Pt. The electro-catalytic activity of ALD Pt films grown on glassy carbon electrodes showed identical activity for the hydrogen-evolution reaction, within experimental error, to that of films deposited using electron-beam evaporation. The low temperatures and decreased processing time required to produce uniform films of Pt with high catalytic activity for the HER using this ALD method may impact many applications, particularly those that require simultaneous processing of multiple materials and/or well-controlled interfaces.

100 7. Nano-structured Pt synthesized by ALD as HER catalysts

8. Conclusions and outlook

8.1 Conclusions

Various novel concepts of silicon based photovoltaics and photoelectrochemistry are proposed in this doctoral thesis. The major novel contributions are:

 The thermal properties of Si NCs are investigated by intentionally heating the material by the Raman laser beam. If the free-standing Si NCs are illuminated by intensive Ar ion laser, a huge red-shift and a peak widening of the first order Si-Si TO mode are observed as a result of the laser heating induced lattice expansion in Si NCs.

 The free-standing Si NCs can be heated as hot as ~953 K by a well-focused laser with a power of 330 µW, as determined using the ratio of Anti-Stokes-to-Stokes TO mode intensities. If the free-standing Si NCs are further heated, the intensity of the blackbody radiation in Raman spectrum starts to compete with those of the TO mode.

 The laser heating effects are confirmed to be reversible to a large extent, but the nature of the material is slightly modified after intensive laser illumination. A simplified model of the heating effects is established to study the size dependence of the heated free- standing Si NCs with an increasing laser power. In contrast, Si NCs in various matrices can hardly be heated using the same amount of laser power because of their excellent thermal conductivity.

 BiVO4 photoanode was deposited by spray pyrolysis. The carrier collection efficiency is improved by gradient W doping through the layer of BiVO4, and the catalytic performance is boosted by the CoPi layer coating.

 An a-Si:H/a-Si:H double-junction solar cell is integrated as the power source with the BiVO4 photoanode into a wireless water-splitting device. The thickness of i-layer in the top cell has been optimized in terms of current matching using both EQE measurements and ASA simulations. The thin solar cell (i-layer of 50 nm in top cell) shows the best current matching under AM 1.5 spectrum, while the thick solar cell (i-layer of 100 nm in top cell) performs the best in reference to the spectrum transmitted through photoanode.

 The impact of the Staebler-Wronski effect on the ηSTH of the water-splitting devices has been studied. The operating point of the water-splitting device is fortunately less affected by the light-induced effect compared to the maximum power point of the solar cell component. The optimized a-Si:H/a-Si:H tandem solar cell combined with BiVO4 can result in a photocurrent of ~3.4 mA cm-2 (stabilize for at least 10 min), which corresponds to a ηSTH of 4.2%.

102 8. Conclusions and outlook

 To further improve the ηSTH, light-trapping techniques have been introduced in the front W:BiVO4. The increased absorption is obtained through the employment of textured substrates and by increasing the film thickness, while retaining a high carrier collection efficiency by rational design of the gradient W-dopant concentration profile in BiVO4.

 The combination of the improved photoanode with a micromorph Si cell allows better utilization of the AM 1.5 solar spectrum, as compared to the a-Si:H/a-Si:H cell, due to the extended absorption of the nc-Si:H bottom junction up to ~1100 nm. A benchmark ηSTH of 5.2% under simulated AM 1.5 illumination is achieved.

 An a-SiC:H(PEC)/a-Si:H(PV)/nc-Si:H(PV) water-splitting photocathode was fabricated without any precious-metal catalysts. A photocurrent density of -5.1 mA cm-2 at 0 V vs. RHE and photocathode efficiency of 2.1% has been achieved. This PEC/PV cathode based on earth-abundant materials was highly stable for over 1 hour in an acidic electrolyte under simulated AM 1.5 illumination.

 The overpotential of an a-SiC:H PEC photocathode was reduced by enhancing the charge carrier collection in the photocathode by the introduction of gradient boron doping in the a-SiC:H PEC junction.

 Dense TF-Si layers exhibiting electrical performance and high spectral utilization in the PEC and PV junctions are achieved by using state-of-the-art glass substrates with integrated micro-textured photonic structures.

 The balance between the anodic shift and the optimized spectral utilization of the PEC/PV configuration was further studied using simulations of the spectral utilization of the various PEC/PV cathodes.

 The electro-catalytic activity of ALD Pt films grown on glassy carbon electrodes had the same activity for hydrogen-evolution from aqueous solution as films deposited using electron-beam evaporation. This ALD method using MeCpPtMe3 and ozone as the precursors, allows lower substrate temperatures and decreased processing time required to produce uniform films of Pt with high catalytic activity for the HER.

8.2 Recommendations

Besides the fundamental studies and techniques applied in solar cells and water-splitting devices discussed in this thesis, a few recommendations are made for future research:

 Many unique size-dependent physical and electrical properties of Si NCs can be investigated on samples Si NCs of a specific diameter. Therefore, Si NCs of a narrower size distribution should be deposited by ETP-CVD. PV devices containing Si NCs as absorber layer can be fabricated. Multi exciton generation and up- or down-conversion 8.2 Recommendations 103

should be investigated in a specific solar cell, to see their effects to the conversion efficiency.

 In the water-splitting device based on BiVO4 photoanode and TF-Si solar cell, the front photoanode is the limiting factor in the hybrid PEC/PV configuration. Therefore controlling the defect density in the BiVO4 material or techniques using plasmonic effects should be applied. When the photoanode is further optimized, more flexible options of TF-Si solar cells such as single-junction device are possible, reducing the cost and deposition time compared with multi-junction devices.

 In the TF-Si based photocathode device, a higher photocurrent density can be achieved using a-SiC:H integrated with a novel structure of tandem solar cells based on a nc-Si:H top cell and a silicon hetero-junction (SHJ) bottom cell as the PV junction in the hybrid PV/PEC configuration. This tandem structure based on Si wafer has better spectral utilization compared with the TF-Si solar cell discussed in Chapter 6, and a photocurrent density of ~10 mA cm-2 is expected in the a-SiC:H/nc-Si:H/SHJ configuration. Besides, a protection layer such as TiO2 and cost-effective HER catalysts should be coated onto the photocathodes for a more stable performance.

 During the PEC characterization, earth-rear metal Pt is still used as the counter electrode. Therefore, unbiased photoanode/photocathode configurations should be investigated that do not need expensive counter electrodes, and facilitate a direct separation of the generated H2 and O2 bubbles. In this case, the current matching and spectral utilization should be optimized.

 The ALD-grown Pt onto p-Si wafer as a photocathode can be further optimized for effective HER. By coating a protective layer onto the p-Si wafer, we found that the wafer surface would not be oxidized by O3, and a higher current density can be achieved than the ALD-grown Pt-catalyzed photocathodes on the substrate of glassy carbon disk.

104 8. Conclusions and outlook

Appendix A. Photoanode characterizations

A.1 Spectrum of solar simulator

The BiVO4 photoanode has a bandgap of 2.4 eV, therefore the measured j value is highly sensitive to the differences in the shape of the blue (and UV) spectral part between the various solar simulators (xenon vs. tungsten). This requires the solar simulator to be calibrated before, during and after the various measurements to minimize the possible spectrum mismatching with the AM 1.5 spectrum.

To guarantee an accurate spectrum, during the j-t and j-V curve measurements we use a Newport Sol3A solar simulator (94023A-SR3 type), which is rated AAA, i.e., the highest rating for solar simulators. The intensity of the solar simulator is real-time controlled to guarantee an irradiance equal to standard test conditions (100 mW cm-2). The spectral shape of the solar simulator is shown in Figure A.1 (black curve), compared with the standard solar spectrum irradiance AM 1.5 Spectrum (red curve).

3.0 NEWPORT Sol3A Class AAA Solar Simulator

) -1 2.5

-2 2.0

1.5

(W m

1.0 AM 1.5 0.5

Irradiance ASTM-G173-3

0.0 400 600 800 1000 1200  (nm) Figure A.1 The irradiance spectrum measured from the solar simulator used in this work (black curve, NEWPORT Sol3A, type 94023A-SR3), compared with the standard solar spectrum irradiance (red curve, AM 1.5, 100 mW cm-2). The AM 1.5 spectrum data is adapted from NREL.9

106 Appendix A. Photoanode characterizations

A.2 Current-voltage curves of optimized photoanode

Figure A.2 18 different AM 1.5 current-voltage curves of our optimized gradient-doped W:BiVO4 photoanode, indicating its reproducibility. The AM 1.5 photocurrent at 1.23 V vs. RHE is 3.92 ± 0.08 mA cm-2.

5 6

4 5 250 nm grad-doped W:BiVO on Asahi 4

) 4

-2 3 + double-jn a-Si short-circuit AM1.5 photocurrent

(%) 2

STH

(mA cm (mA  j 2

1 1

0 0 0 10 20 30 40 50 60 time (min)

Figure A.3 The short-circuit AM 1.5 photocurrent measurement shows that the photoanode integrated with double-junction thin-film silicon solar cell is relatively stable within the course of an hour, with less than 5% degradation observed. The red horizontal dashed line indicates the predicted operating photocurrent based on the intersection of the j-V curves of the BiVO4 photoanode and the double-junction silicon solar cell.

A.3 Structures of thin-film silicon solar cells 107

A.3 Structures of thin-film silicon solar cells

Scheme A.1 Cross-section sketch of single-junction a-Si:H PV cell. The thickness of each layer is indicated but not to scale.

Scheme A.2 Cross-section sketch of double-junction a-Si:H/a-Si:H PV cell. The thickness of each layer is indicated but not to scale. 108 Appendix A. Photoanode characterizations

Scheme A.3 Cross-section sketch of micromorph a-Si:H/nc-Si:H PV cell. The thickness of each layer is indicated but not to scale.

Appendix B. Photocathode characterizations

B.1 Material optimization

The percentage of carbon atom number (C%) in the bulk material can be tuned through the methane/silane (CH4/SiH4) gas flow ratio. Under typical plasma-enhanced chemical vapor deposition (PECVD) conditions SiH4 molecules decompose much easier than CH4 molecules. Therefore CH4/SiH4 gas flow ratios ranging from 1 to 20 are required to establish C% values in the a-SiC:H of ~10 up to 40%. In order to optimize the C% in our a-SiC:H material, we deposited 100 nm i-layers with different C% on Corning glass. The C% values were ~0%, ~10%, ~20%, ~30% and ~40% as confirmed by XPS (data not shown here), which was obtained by using CH4/SiH4 gas flow ratios of 0, 1, 3, 10 to 20, respectively.

In Table B.1 the optical band gaps (determined from the measured reflectance and transmittance spectra) of the five a-Si1-xCx:H samples are presented. These results demonstrate that the higher the carbon content is, the larger the band gap is.

Table B.1 The CH4/SiH4 gas flow ratio injected in the plasma, the resulting percentage of carbon atom number (C%) and the optical bandgaps of the five a-Si1-xCx:H samples.

CH4/SiH4 Material Bandgap (eV) 0 a-Si:H 1.79

1 a-Si0.9C0.1:H 1.81

3 a-Si0.8C0.2:H 2.11

10 a-Si0.7C0.3:H 2.28

20 a-Si0.6C0.4:H 2.47

In order to study the presence of carbon and hydrogen in the material using Fourier transform infrared (FTIR) spectroscopy, we deposited the same five films (100 nm thick) on p-type crystalline silicon wafers. As shown in Figure B.1, differences of C% in the bulk of the film can be revealed by the marked changes of the spectral signatures on the Fourier transform infrared (FTIR) spectra of this series of samples. The presence of carbon can be demonstrated by the -1 absorption modes at 770 cm corresponding to Si-C bonds and Si-CH3 bonds and the modes at -1 1008 cm corresponding to SiC-Hn bonds. These absorption peaks are indeed increasing with increasing values of C%. The surface silicon hydrides (Siy-Hx) bonds have various stretching modes: Si-H bond low stretching mode (LSM)259 at 2000 cm-1 corresponding to hydrides in small volume deficiencies (vacancies) Si-H bonds Si-H high stretching mode (HSM)260, 261 at 2070- 2100 cm-1 corresponding to hydrides at the surface of nano-sized voids. For increasing C% the HSM increases and the LSM decreases. This shows that the incorporation of carbon results in more porous materials due to the incorporation of more nano-sized voids. Higher densities of nano-sized voids create more defects in the films and reduce the PEC performance of this material. 110 Appendix B. Photocathode characterizations

Si-CH3 Si-C

Si-Hn Si-H Si-H(HSM)

Si-C a-Si C :H

0.6 0.4

a-Si0.7C0.3:H

a-Si0.8C0.2:H

a-Si0.9C0.1:H

FTIR intensity (a.u.) a-Si:H Si-H (BM) Si-H (LSM)

500 1000 1500 2000 2500 3000 wavenumber (cm-1)

Figure B.1 Fourier transform infrared (FTIR) spectra of i-a-SiC:H films with different C% and bandgap.

) -2

-2

-4 a-Si:H

(mA cm a-Si C :H j 0.9 0.1 a-Si C :H -6 0.8 0.2

a-Si0.7C0.3:H

a-Si0.6C0.4:H -8 -1.6 -1.2 -0.8 -0.4 0.0 E (V) vs. RHE

Figure B.2 PEC measurement of the photocathodes in the series of C% in the i-a-SiC:H layer. The inset photo shows less sunlight is absorbed in the C rich samples.

Five photocathodes consisting of 10 nm p-a-SiC:H layer and 100 nm i-a-SiC:H layer were prepared, where C% of i-a-SiC:H layer was varied ~0, ~10, ~20, ~30, and ~40%. The inset photo in Figure B.2 shows how the colors of these electrodes change from dark brown to yellow as C% increases. This is consistent with the subsequent bandgap widening as shown in Table B.1. Figure B.2 also demonstrates the increase of C% deleteriously affects the current densities of these electrodes. For example, the photocurrent density at -1.5 V vs. RHE changed from -7.0 mA cm-2 B.2 Glass substrate with integrated micro-textured photonic structures 111

(10% C) to -0.35 mA cm-2 (40% C). This is a result of both the band gap widening which results in reduced photon absorption and an increased density of defects which serve as charge recombination centers. Due to its ideal band gap and stability the i-a-SiC:H layer with 10% carbon incorporated was selected in the fabrication of the PEC/PV configuration.

B.2 Glass substrate with integrated micro-textured photonic structures

Using textured substrates is the conventional approach to enhance the light trapping in PV devices.220, 221 However, a delicate interplay exists between textured substrates and the growth of device grade PV materials on top. Especially, nc-Si:H is highly sensitive to extreme texture, as the texture initiates the growth of defect-rich filaments above sharp valleys during growth.222 The size and density of these defect-rich filaments grow with thickness. These filaments significantly reduced the fill factor (FF), open-circuit voltage and current density of the PV junctions.222, 223 However, the PEC/PV device needs thick (> 3 µm) nc-Si:H bottom cells to achieve the required high spectral utilization. To prevent the incorporation of defect-rich filaments and to allow the integration of thick high quality nc-Si:H junctions in our PEC/PV devices, state-of-the-art glass substrates with integrated micro-textured photonic structures have been used. These photonic structures have been obtained by wet-etching a sacrificial ITO layer on Corning glass, resulting in craters with typical diameters in the order of 10 µm and depths of 2 µm. A back reflector consisting of aluminum doped zinc-oxide (AZO, 2 µm)/ metal reflector (30 nm Cr + 100 nm Ag)/ AZO (10 nm) has been deposited on top. Besides the micro-texture, the back reflector has nano- textured photonic structures as well (Figure B.3), as a result of the natural growth of the AZO. The PEC and PV junctions deposited on top adopt the nano-textured and micro-textured photonic structures. The nano-scale photonic structures facilitate the light scattering of the high energetic photons in to the a-SiC:H photocathode, while the micro-scale photonic features facilitate an efficient light trapping in the TF-Si PV junctions in the red and near infrared spectral range.223 As demonstrated in the SEM image in Figure 6.1 (b), the glass substrate with micro-textured photonic structures allows the growth of highly dense nc-Si:H material. No defect-rich filaments can be observed in the 4 µm thick nc-Si:H material, which guarantees high voltages and current densities in the PV junction of the PEC/PV device.

112 Appendix B. Photocathode characterizations

Figure B.3 Microscopy image of the nano-textured AZO on the surface of the micro-texture glass as the substrate for the PEC/PV cathode.

B.3 Electrochemical impedance spectroscopy

The flat band potential and the acceptor density of the a-SiC:H photocathode with an active area of 0.283 cm2 were characterized by electrochemical impedance spectroscopy (EIS). In order to obtain an accurate Mott-Schottky analysis, the scanning frequencies should be chosen in the range that the real part of the impedance is constant and only the imaginary part of the impedance is varied. Thereby, the imaginary impedance can be fully interpreted as the contribution from the space charge layer capacitance. Based on the Bode and Nyquist plot, the frequency range for our a-SiC:H photocathode is 500 Hz. The following Mott-Schottky equation then applies:

1 2k T 22 VV fb  (B.1) Csc0 r eN A A e

Based on Equation B.1, the x-axis and the slope of the Mott-Schottky plot can be used to determine the flat-band potential (Vfb) and the acceptor concentration (NA), respectively. Figure B.4 shows the Mott-Schottky plot for our a-SiC:H photocathode at 500 Hz under dark condition.

B.3 Electrochemical impedance spectroscopy 113

4 slope: -18 x 1013 F-2V-1 3

)

-2

13 500 Hz 2

(10

-2

C V 1 FB

0 1.05 1.10 1.15 1.20 1.25 1.30 1.35 V vs. RHE (V)

Figure B.4 Mott-Schottky plot of the a-SiC:H film at 500 Hz, taken under dark condition.

17 -3 The Vfb is estimated to be ~1.25 V vs. RHE, and the NA is approximated to be 2.5 × 10 cm , assuming a dielectric constant of 14. The valence band edge EV could be determined using the following relationship:

NA EFV E   kTln (B.2) NV where the effective density of valence band states

 32 2T mkh 22 (B.3) h

* The effective mass of hole (mh ) of 0.3m0 is assumed to be similar to that c-Si and the effective 18 -3 density of valence band states (NV) is calculated to be 4 × 10 cm . This calculation suggests that the Fermi level is located 80 mV above the valence band edge.

114 Appendix B. Photocathode characterizations

1.15 V vs. RHE 200 300

150

) ) 200



 100

100 - 50

0 0 100 101 102 103 104 105 f (Hz) (a) 1.15 V vs . RHE 5 5

4 4

)



 3 3

Re 2 2

1 1

0 0 102 103 104 f (Hz) (b) Figure B.5 Real and imaginary impedance as a function of frequency in the frequency range of 1- 105 Hz (a), and the magnification part in the frequency range of 102-104 Hz (b).

Appendix C. Platinum characterizations

C.1 AFM characterization

Figure C.1 AFM images of a Pt film deposited on an NH4F-etched Si(111) wafer using 200 cycles of ALD at 250 °C.

C.2 XPS characterization

Figure C.2 X-ray photoelectron spectra in the region of the Pt 4f7/2 peak for Pt films deposited by electron-beam evaporation and by ALD. (a) Pt film of 3 nm thickness deposited on a glassy carbon substrate by electron-beam evaporation; (b) 50 cycles of ALD Pt on a glassy carbon substrate. 116 Appendix C. Platinum characterizations

Figure C.3 Comparison of XPS data for Pt films deposited by electron-beam evaporation and by ALD. (a) 50 ALD cycles of Pt on a glassy carbon substrate; (b) 50 cycles of ALD Pt on glassy carbon substrates after 45 s plasma treatment.

Figure C.4 XPS of the ALD Pt on glassy carbon substrates: carbon signal showed up after 45 s plasma etching. C.2 XPS characterization 117

Figure C.5 XPS of the ALD Pt on glassy carbon surface: Pt 4f7/2 with 50 ALD cycles of Pt under growth temperature of (a) 200°C and (b) 300 °C.

Figure C.6 XPS data of Pt films on Si substrate deposited by ALD. (a) 50 ALD cycles of Pt on a Si substrate; (b) 50 cycles of ALD Pt on Si after 45 s plasma treatment.

118 Appendix C. Platinum characterizations

Bibliography

1. M. Emilio, J. R. Kuhn, R. I. Bush and I. F. Scholl, Astrophys J, 2012, 750. 2. J. H. Parkinson, L. V. Morrison and F. R. Stephenson, Nature, 1980, 288, 548-551. 3. M. Woolfson, Astron Geophys, 2000, 41, 12-19. 4. T. Gold, J Geophys Res, 1959, 64, 1665-1674. 5. E. Marsch, H. Tian, J. Sun, W. Curdt and T. Wiegelmann, Astrophys J, 2008, 685, 1262- 1269. 6. E. N. Parker, Astrophys J, 1958, 128, 664-676. 7. C. A. Gueymard, D. Myers and K. Emery, Sol Energy, 2002, 73, 443-467. 8. http://rredc.nrel.gov/solar/spectra/am0/. 9. http://rredc.nrel.gov/solar/spectra/am1.5/. 10. B. P. Tissot and D. H. Welte, Petroleum formation and occurrence, 1984. 11. M.K. Hubbert, Nuclear energy and the fossil fuels, Shell Development Company, Exploration and Production Research Division Houston, TX, 1956. 12. N. Armaroli and V. Balzani, Angew Chem Int Edit, 2007, 46, 52-66. 13. H.B. Gray, Nat Chem, 2009, 1, 7-7. 14. M.A. Green, K. Emery, Y. Hishikawa, W. Warta and E. D. Dunlop, Prog Photovoltaics, 2014, 22, 701-710. 15. D. Mills, Sol Energy, 2004, 76, 19-31. 16. Flagship Report 2003, www.wbgu.de. 17. C.D.I.A. Center, http://cdiac.ornl.gov/GCP/carbonbudget/2013/. 18. N.S. Lewis, Basic Research Needs for Solar Energy Utilization, 2005. 19. K. Jäger, O. Isabella, A.H.M. Smets, R.A.C.M.M. van Swaaij and M. Zeman, Solar Energy - Fundamentals, Technology, and Systems, Delft University of Technology, 2014. 20. W. Shockley and H.J. Queisser, J Appl Phys, 1961, 32, 510-519. 21. G. Conibeer, Mater Today, 2007, 10, 42-50. 22. http://www.californiasolarcenter.org/history_pv.html. 23. D. M. Chapin, C.S. Fuller and G.L. Pearson, J Appl Phys, 1954, 25, 676-677. 24. http://hyperphysics.phy-astr.gsu.edu/hbase/tables/elabund.html. 25. P. Report, Photovoltaics Report, 2014. 26. http://www.nrel.gov/ncpv/. 27. A. Richter, M. Hermle and S.W. Glunz, Ieee J Photovolt, 2013, 3, 1184-1191. 28. M.A. Green and S.R. Wenham, Google Patents, 1988. 29. S.B. Mallick, M. Agrawal and P. Peumans, Opt Express, 2010, 18, 5691-5706. 30. PV Magazine. 31. A. Gupta and A.D. Compaan, Conference Record of the Thirty-First IEEE Photovoltaic Specialists Conference - 2005, 2005, 235-238. 32. A. Gupta, V. Parikh and A.D. Compaan, Sol Energ Mat Sol C, 2006, 90, 2263-2271. 33. M.G. Panthani, J.M. Kurley, R.W. Crisp, T.C. Dietz, T. Ezzyat, J.M. Luther and D.V. Talapin, Nano Lett, 2014, 14, 670-675. 34. C. Platzer-Bjorkman, A. Hultqvist, J. Pettersson and T. Torndahl, Oxide-Based Materials and Devices, 2010, 7603. 120 Bibliography

35. K. Ramanathan, J. Keane and R. Noufi, Conference Record of the Thirty-First IEEE Photovoltaic Specialists Conference - 2005, 2005, 195-198. 36. T. Wada, Y. Hashimoto, S. Nishiwaki, T. Satoh, S. Hayashi, T. Negami and H. Miyake, Sol Energ Mat Sol C, 2001, 67, 305-310. 37. M.A. Wank, R.A.C.M.M. van Swaaij, R. van de Sanden and M. Zeman, Prog Photovoltaics, 2012, 20, 333-342. 38. G. Yang, R.A,C.M.M. van Swaaij, S. Dobrovolskiy and M. Zeman, J Appl Phys, 2014, 115. 39. M. Zeman, O. Isabella, F.D. Tichelaar and S.L. Luxembourg, Phys Status Solidi C, 2010, 7, 1057-1060. 40. R.R. King, D.C. Law, K.M. Edmondson, C.M. Fetzer, G.S. Kinsey, H. Yoon, R.A. Sherif and N.H. Karam, Appl Phys Lett, 2007, 90. 41. M. Grätzel, Journal of Photochemistry and Photobiology C: Photochemistry Reviews, 2003, 4, 145-153. 42. D.L. Staebler and C.R. Wronski, Appl Phys Lett, 1977, 31, 292-294. 43. J. Melskens, A.H.M. Smets, M. Schouten, S.W.H. Eijt, H. Schut and M. Zeman, IEEE J Photovolt, 2013, 3, 65-71. 44. M.A. Green, Physica E, 2002, 14, 65-70. 45. M.A. Green, Conference Record of the Twenty-Ninth Ieee Photovoltaic Specialists Conference 2002, 2002, 1330-1334. 46. I.M. Dharmadasa, Sol Energ Mat Sol C, 2005, 85, 293-300. 47. G. Conibeer, M.A. Green, R. Corkish, Y. Cho, E.C. Cho, C.W. Jiang, T. Fangsuwannarak, E. Pink, Y.D. Huang, T. Puzzer, T. Trupke, B. Richards, A. Shalav and K.L. Lin, Thin Solid Films, 2006, 511, 654-662. 48. M.Y. Levy and C. Honsberg, Phys Rev B, 2008, 78. 49. A. Marti, E. Antolin, C.R. Stanley, C.D. Farmer, N. Lopez, P. Diaz, E. Canovas, P.G. Linares and A. Luque, Phys Rev Lett, 2006, 97. 50. S.A. Kalogirou, Prog Energ Combust, 2004, 30, 231-295. 51. Y. Feng, R. Patterson, S. Lin, S. Shrestha, S. Huang, M.A. Green and G. Conibeer, Appl Phys Lett, 2013, 102. 52. M.A. Green, G. Conibeer, D. Konig, S. Shrestha, S. J. Huang, P. Aliberti, L. Treiber, R. Patterson, B.P. Veettil, A. Hsieh, Y. Feng, A. Luque, A. Marti, P.G. Unares, E. Canovas, E. Antollin, D.F. Marron, C. Tablero, E. Hernandez, J.F. Guillemoles, L. Huang, A. Le Bris, T. Schmidt, R. Cladl and M. Tayebjee, IEEE Phot Spec Conf, 2010, 57-60. 53. L.C. Hirst, M.P. Lumb, R. Hoheisel, C.G. Bailey, S.P. Philipps, A.W. Bett and R.J. Walters, Sol Energ Mat Sol C, 2014, 120, 610-615. 54. J. Burschka, N. Pellet, S.J. Moon, R. Humphry-Baker, P. Gao, M.K. Nazeeruddin and M. Gratzel, Nature, 2013, 499, 316. 55. M.Z. Liu, M.B. Johnston and H.J. Snaith, Nature, 2013, 501, 395. 56. J.Y. Liu, T. Luo, T.S. Mouli, F.L. Meng, B. Sun, M.Q. Li and J.H. Liu, Chem Commun, 2010, 46, 472-474. 57. G. Li, R. Zhu and Y. Yang, Nat Photonics, 2012, 6, 153-161. 58. D.M. Mittleman, R.W. Schoenlein, J.J. Shiang, V.L. Colvin, A.P. Alivisatos and C.V. Shank, Phys Rev B, 1994, 49, 14435-14447. Bibliography 121

59. R.J. Ellingson, M.C. Beard, J.C. Johnson, P.R. Yu, O.I. Micic, A.J. Nozik, A. Shabaev and A.L. Efros, Nano Lett, 2005, 5, 865-871. 60. D. Timmerman and T. Gregorkiewicz, Mater Sci Eng B-Adv, 2009, 159-60, 87-89. 61. D. Timmerman, I. Izeddin, P. Stallinga, I.N. Yassievich and T. Gregorkiewicz, Nat Photonics, 2008, 2, 105-109. 62. O.E. Semonin, J.M. Luther, S. Choi, H.Y. Chen, J.B. Gao, A.J. Nozik and M.C. Beard, Science, 2011, 334, 1530-1533. 63. P. Gibart, F. Auzel, J.C. Guillaume and K. Zahraman, Jpn J Appl Phys 1, 1996, 35, 4401- 4402. 64. E. Klampaftis, D. Ross, K.R. McIntosh and B.S. Richards, Sol Energ Mat Sol C, 2009, 93, 1182-1194. 65. W.G.J.H.M. van Sark, J. de Wild, J.K. Rath, A. Meijerink and R.E. I. Schropp, Nanoscale Res Lett, 2013, 8. 66. B.S. Richards, Sol Energ Mat Sol C, 2006, 90, 2329-2337. 67. N.S. Lewis, et al., Basic Research Needs for Solar Energy Ulitzation, Report on the Basic Energy Sciences Workshop on Solar Energy Utilization, 2005. 68. A.J. Bard and M.A. Fox, Accounts Chem Res, 1995, 28, 141-145. 69. W.C. Chueh and S.M. Haile, Chemsuschem, 2009, 2, 735-739. 70. A. Kogan, M. Kogan and S. Barak, Int J Hydrogen Energ, 2005, 30, 35-43. 71. A.H.M. Smets, https://www.edx.org/course/delftx/delftx-et-3034tu-solar-energy-1996. 72. J. A. Seabold and K.S. Choi, J Am Chem Soc, 2012, 134, 2186-2192. 73. R.R. Rangaraju, A. Panday, K.S. Raja and M. Misra, J Phys D Appl Phys, 2009, 42. 74. C. Santato, M. Ulmann and J. Augustynski, J Phys Chem B, 2001, 105, 936-940. 75. Y. Surendranath, M.W. Kanan and D.G. Nocera, J Am Chem Soc, 2010, 132, 16501-16509. 76. V.M. Aroutiounian, V.M. Arakelyan and G.E. Shahnazaryan, Sol Energy, 2005, 78, 581- 592. 77. J.H. Oh, T.G. Deutsch, H.C. Yuan and H.M. Branz, Energ Environ Sci, 2011, 4, 1690-1694. 78. S.A.S. Machado and L.A. Avaca, Electrochim Acta, 1994, 39, 1385-1391. 79. W.C. Sheng, H.A. Gasteiger and S.H. Yang, J Electrochem Soc, 2010, 157, B1529-B1536. 80. M.S. Kang, Y.J. Choi, H.J. Lee and S.H. Moon, J Colloid Interf Sci, 2004, 273, 523-532. 81. http://www.nanowerk.com/news/newsid=6970.php. 82. R. van de Krol and M. Grätzel, Photoelectrochemical Hydrogen Production, Springer Science+Business Media, 2012. 83. S. Park, E. Cho, D.Y. Song, G. Conibeer and M.A. Green, Sol Energ Mat Sol C, 2009, 93, 684-690. 84. L. Han, J. Wang and R. Liang, IEEE Phot Spec Conf, 2010, 3338-3342. 85. A. Kanjilal, J.L. Hansen, P. Gaiduk, A.N. Larsen, N. Cherkashin, A. Claverie, P. Normand, E. Kapelanakis, D. Skarlatos and D. Tsoukalas, Appl Phys Lett, 2003, 82, 1212-1214. 86. O. Yasar-Inceoglu, T. Lopez, E. Farshihagro and L. Mangolini, Nanotechnology, 2012, 23. 87. C.F. Lin, W.T. Tseng and M.S. Feng, J Appl Phys, 2000, 87, 2808-2815. 88. B.A. Korevaar, C. Smit, A.M.H.N. Petit, R.A.C.M.M. van Swaaij and M.C.M. van de Sanden, Amorphous and Heterogeneous Silicon-Based Films-2002, 2002, 715, 595-600. 89. C. Smit, E.A.G. Hamers, B.A. Korevaar, R.A.C.M.M. van Swaaij and M.C.M. van de Sanden, J Non-Cryst Solids, 2002, 299, 98-102. 122 Bibliography

90. I. Dogan, N.J. Kramer, R.H.J. Westermann, K. Dohnalova, A.H.M. Smets, M.A. Verheijen, T. Gregorkiewicz and M.C.M. van de Sanden, J Appl Phys, 2013, 113:134306, 1-10. 91. I. Dogan and M.C.M. van de Sanden, J Appl Phys, 2013, 114: 134310, 1-8. 92. M. Zeman, Solar Cells, Delft University of Technology, Delft, 2009. 93. F.F. Abdi, N. Firet, A. Dabirian and R. van de Krol, MRS Proceedings, 2012, 1446, mrss12- 1446-u1402-1405. 94. F.F. Abdi, N. Firet and R. van de Krol, Chemcatchem, 2013, 5, 490-496. 95. F.F. Abdi, L. Han, A.H.M. Smets, M. Zeman, B. Dam and R. van de Krol, Nat Commun, 2013, 4:2195. 96. F.F. Abdi, T.J. Savenije, M.M. May, B. Dam and R. van de Krol, J Phys Chem Lett, 2013, 4, 2752-2757. 97. L. Han, F.F. Abdi, R. van de Krol, R. Liu, Z. Huang, H.-J. Lewerenz, B. Dam, M. Zeman and A.H.M. Smets, ChemSusChem, 2014, 7, 2832-2838. 98. Y.Q. Liang, C.S. Enache and R. van de Krol, Int J Photoenergy, 2008. 99. Y.Q. Liang, T. Tsubota, L.P.A. Mooij and R. van de Krol, J Phys Chem C, 2011, 115, 17594-17598. 100. M.W. Kanan and D.G. Nocera, Science, 2008, 321, 1072-1075. 101. http://hivatec.ca/consulting-design/thin-film-deposition/. 102. http://www.synlab.org.cn/2012/10/320.html. 103. M.A. Reed, J.N. Randall, R.J. Aggarwal, R.J. Matyi, T.M. Moore and A.E. Wetsel, Phys Rev Lett, 1988, 60, 535-537. 104. C.W. Jiang and M.A. Green, J Appl Phys, 2006, 99:114902, 1-7. 105. L.F. Sun, J.J. Choi, D. Stachnik, A.C. Bartnik, B.R. Hyun, G.G. Malliaras, T. Hanrath and F.W. Wise, Nat Nanotechnol, 2012, 7, 369-373. 106. B. Sun, P. Munroe and G.X. Wang, Sci Rep, 2013, 3:2247, 1-7. 107. L. Han, J. Wang and R. Liang, Adv Mater Rez, 2012, 383-390, 6270-6276. 108. Y. Tachibana, L. Vayssieres and J.R. Durrant, Nat Photonics, 2012, 6, 511-518. 109. T. Trupke, M.A. Green and P. Wurfel, J Appl Phys, 2002, 92, 1668-1674. 110. F.F. Abdi, L. Han, A.H.M. Smets, M. Zeman, B. Dam and R. van de Krol, Nat Commun, 2013, 4:2195, 1-7. 111. L. Han, F.F. Abdi, P. Perez Rodriguez, B. Dam, R. van de Krol, M. Zeman and A.H.M. Smets, Phys Chem Chem Phys, 2014, 16, 4220-4229. 112. L. Han, F.F. Abdi, R. van de Krol, R. Liu, Z. Huang, H.-J. Lewerenz, B. Dam, M. Zeman and A.H.M. Smets, ChemSusChem, 2014, 7, 2832–2838. 113. L. Han, I.A. Digdaya, T.W.F. Buijs, F.F. Abdi, Z. Huang, R. Liu, B. Dam, M. Zeman, W.A. Smith and A.H.M. Smets, Journal of Material Chemistry A, 2014, under review. 114. W.D.A.M. de Boer, D. Timmerman, K. Dohnalova, I.N. Yassievich, H. Zhang, W.J. Buma and T. Gregorkiewicz, Nat Nanotechnol, 2010, 5, 878-884. 115. A.G. Aberle, S. Glunz and W. Warta, J Appl Phys, 1992, 71, 4422-4431. 116. W.H. Chang, A.T. Chou, W.Y. Chen, H.S. Chang, T.M. Hsu, Z. Pei, P.S. Chen, S.W. Lee, L.S. Lai, S.C. Lu and M.J. Tsai, Appl Phys Lett, 2003, 83, 2958-2960. 117. B.C. Connelly, G.D. Metcalfe, H.E. Shen and M. Wraback, Appl Phys Lett, 2010, 97:251117, 1-3. Bibliography 123

118. D. Gammon, E.S. Snow, B.V. Shanabrook, D.S. Katzer and D. Park, Phys Rev Lett, 1996, 76, 3005-3008. 119. J.Y. Marzin, J.M. Gerard, A. Izrael, D. Barrier and G. Bastard, Phys Rev Lett, 1994, 73, 716-719. 120. A.L. Patterson, Phys Rev, 1939, 56, 978-982. 121. Z.F. Sui, P.P. Leong, I.P. Herman, G.S. Higashi and H. Temkin, Appl Phys Lett, 1992, 60, 2086-2088. 122. Y. Kanemitsu, H. Uto, Y. Masumoto, T. Matsumoto, T. Futagi and H. Mimura, Phys Rev B, 1993, 48, 2827-2830. 123. G. Faraci, S. Gibilisco and A.R. Pennisi, Phys Rev B, 2009, 80. 124. L. Khriachtchev, J Electrochem Soc, 2012, 159, K21-K26. 125. L. Khriachtchev, M. Rasanen and S. Novikov, Appl Phys Lett, 2005, 86:141911, 1-3. 126. L. Khriachtchev, M. Rasanen and S. Novikov, J Appl Phys, 2006, 100:053502, 1-7. 127. L. Khriachtchev, M. Rasanen and S. Novikov, Appl Phys Lett, 2006, 88:013102, 1-3. 128. T. Nikitin, L. Khriachtchev, M. Rasanen and S. Novikov, Appl Phys Lett, 2009, 94:173116, 1-3. 129. P. Babal, PhD thesis, 2014, TU Delft, in press. 130. M. Balkanski, R.F. Wallis and E. Haro, Phys Rev B, 1983, 28, 1928-1934. 131. G. Conibeer, M.A. Green, E.C. Cho, D. Konig, Y.H. Cho, T. Fangsuwannarak, G. Scardera, E. Pink, Y.D. Huang, T. Puzzer, S.J. Huang, D.Y. Song, C. Flynn, S. Park, X.J. Hao and D. Mansfield, Thin Solid Films, 2008, 516, 6748-6756. 132. S.K. Gupta and P.K. Jha, Solid State Commun, 2009, 149, 1989-1992. 133. L. Sirleto, M.A. Ferrara, T. Nikitin, S. Novikov and L. Khriachtchev, Nat Commun, 2012, 3:1220, 1-6. 134. G.X. Cheng, K.J. Chen, W. Zhang, H. Xia and X.K. Zhang, Phys Status Solidi A, 1992, 129, 421-434. 135. A.A.D.T. Adikaari and S.R.P. Silva, J Appl Phys, 2005, 97: 114305, 1-7. 136. V. Poborchii, T. Tada and T. Kanayama, J Appl Phys, 2005, 97:104323, 1-5. 137. A. Compaan, M.C. Lee and G.J. Trott, Phys Rev B, 1985, 32, 6731-6741. 138. G. Faraci, S. Gibilisco and A.R. Pennisi, Phys Lett A, 2009, 373, 3779-3782. 139. G. Faraci, A.R. Pennisi, A.Alberti, R. Ruggeri and G. Mannino, Sci Rep-Uk, 2013, 3. 140. J. Costa, P. Roura, J.R. Morante and E. Bertran, J Appl Phys, 1998, 83, 7879-7885. 141. S.K. Deb, M. Wilding, M. Somayazulu and P.F. McMillan, Nature, 2001, 414, 528-530. 142. A.N. Goldstein, Appl Phys a-Mater, 1996, 62, 33-37. 143. K. Nishio, T. Morishita, W. Shinoda and M. Mikami, Phys Rev B, 2005, 72:245321, 1-4. 144. W.J. Luo, Z.S. Yang, Z.S. Li, J.Y. Zhang, J.G. Liu, Z.Y. Zhao, Z.Q. Wang, S.C. Yan, T. Yu and Z.G. Zou, Energ Environ Sci, 2011, 4, 4046-4051. 145. A. Kudo, K. Omori and H. Kato, J Am Chem Soc, 1999, 121, 11459-11467. 146. A. Kudo, K. Ueda, H. Kato and I. Mikami, Catal Lett, 1998, 53, 229-230. 147. S. Tokunaga, H. Kato and A. Kudo, Chem Mater, 2001, 13, 4624-4628. 148. S.P. Berglund, D.W. Flaherty, N.T. Hahn, A.J. Bard and C.B. Mullins, J Phys Chem C, 2011, 115, 3794-3802. 149. P. Chatchai, Y. Murakami, S.Y. Kishioka, A.Y. Nosaka and Y. Nosaka, Electrochem Solid St, 2008, 11, H160-H163. 124 Bibliography

150. L.H. Dall'Antonia, N.R. de Tacconi, W. Chanmanee, H. Timmaji, N. Myung and K. Rajeshwar, Electrochem Solid St, 2010, 13, D29-D32. 151. M.T. Li, L.A. Zhao and L.J. Guo, Int J Hydrogen Energ, 2010, 35, 7127-7133. 152. M.C. Long, W.M. Cai and H. Kisch, J Phys Chem C, 2008, 112, 548-554. 153. F.F. Abdi and R. van de Krol, J Phys Chem C, 2012, 116, 9398-9404. 154. T.H. Jeon, W. Choi and H. Park, Phys Chem Chem Phys, 2011, 13, 21392-21401. 155. S.K. Pilli, T.G. Deutsch, T.E. Furtak, J.A. Turner, L.D. Brown and A.M. Herring, Phys Chem Chem Phys, 2012, 14, 7032-7039. 156. S.K. Pilli, T.E. Furtak, L.D. Brown, T.G. Deutsch, J.A. Turner and A.M. Herring, Energ Environ Sci, 2011, 4, 5028-5034. 157. K. Sayama, N.N. Wang, Y. Miseki, H. Kusama, N. Onozawa-Komatsuzaki and H. Sugihara, Chem Lett, 2010, 39, 17-19. 158. H. Ye, H.S. Park and A.J. Bard, J Phys Chem C, 2011, 115, 12464-12470. 159. D.K. Zhong, S. Choi and D.R. Gamelin, J Am Chem Soc, 2011, 133, 18370-18377. 160. W. Hamd, S. Cobo, J. Fize, G. Baldinozzi, W. Schwartz, M. Reymermier, A. Pereira, M. Fontecave, V. Artero, C. Laberty-Robert and C. Sanchez, Phys Chem Chem Phys, 2012, 14, 13224-13232. 161. G.W. Ho, K.J. Chua and D.R. Siow, Chem Eng J, 2012, 181, 661-666. 162. Y.B. Li, L. Zhang, A. Torres-Pardo, J.M. Gonzalez-Calbet, Y.H. Ma, P. Oleynikov, O. Terasaki, S. Asahina, M. Shima, D. Cha, L. Zhao, K. Takanabe, J. Kubota and K. Domen, Nat Commun, 2013, 4. 163. Z.B. Chen, T.F. Jaramillo, T.G. Deutsch, A. Kleiman-Shwarsctein, A.J. Forman, N. Gaillard, R. Garland, K. Takanabe, C. Heske, M. Sunkara, E. W. McFarland, K. Domen, E.L. Miller, J.A. Turner and H.N. Dinh, J Mater Res, 2010, 25, 3-16. 164. M. Gratzel, Nature, 2001, 414, 338-344. 165. A.E. Delahoy, S.C. Gau, O.J. Murphy, M. Kapur and J.O.M. Bockris, Int J Hydrogen Energ, 1985, 10, 113-116. 166. E.L. Miller, R.E. Rocheleau and X.M. Deng, Int J Hydrogen Energ, 2003, 28, 615-623. 167. J. Brillet, J.H. Yum, M. Cornuz, T. Hisatomi, R. Solarska, J. Augustynski, M. Graetzel and K. Sivula, Nat Photonics, 2012, 6, 823-827. 168. O. Khaselev and J.A. Turner, Science, 1998, 280, 425-427. 169. S. Licht, B. Wang, S. Mukerji, T. Soga, M. Umeno and H. Tributsch, J Phys Chem B, 2000, 104, 8920-8924. 170. F. F. Abdi, L. Han, A.H.M. Smets, M. Zeman, B. Dam and R. van de Krol, Nat Commun, 2013, 4. 171. D. L. Staebler and C. R. Wronski, Appl Phys Lett, 1977, 31, 292-294. 172. W. Theiss, SCOUT Thin Film Analysis Software Handbook, Aachen, 2002. 173. M. Zeman, J. van den Heuvel, M. Kroon, J. Willemen, B. Pieters, J. Krč and S. Solntsev, Advanced Semiconductor Analysis User Manual, 2011. 174. T. Kirchartz, 2009, 38, 17. 175. D.E. Scaife, Sol Energy, 1980, 25, 41-54. 176. S.Y. Reece, J.A. Hamel, K. Sung, T.D. Jarvi, A.J. Esswein, J.J.H. Pijpers and D.G. Nocera, Science, 2011, 334, 645-648. Bibliography 125

177. M.G. Walter, E.L. Warren, J.R. McKone, S.W. Boettcher, Q.X. Mi, E.A. Santori and N.S. Lewis, Chem. Rev., 2010, 110, 6446-6473. 178. A. Fujishima and K. Honda, Nature, 1972, 238, 37. 179. S.U.M. Khan, M. Al-Shahry and W.B. Ingler, Science, 2002, 297, 2243-2245. 180. R. Liu, Y. Lin, L.Y. Chou, S.W. Sheehan, W. He, F. Zhang, H.J. Hou and D. Wang, Angewandte Chemie, 2011, 123, 519-522. 181. I. Cesar, A. Kay, J.A. Gonzalez Martinez and M. Grätzel, J Am Chem Soc, 2006, 128, 4582- 4583. 182. L.Y. Chou, R. Liu, W.S. He, N. Geh, Y.J. Lin, E.Y.F. Hou, D.W. Wang and H.J.M. Hou, Int J Hydrogen Energ, 2012, 37, 8889-8896. 183. Y. Ling, G. Wang, J. Reddy, C. Wang, J.Z. Zhang and Y. Li, Angewandte Chemie International Edition, 2012, 51, 4074-4079. 184. X. Yang, C. Du, R. Liu, J. Xie and D. Wang, J Catal, 2013, 304, 86-91. 185. M.F. Lichterman, M.R. Shaner, S.G. Handler, B.S. Brunschwig, H.B. Gray, N.S. Lewis and J.M. Spurgeon, J Phys Chem Lett, 2013, 4, 4188-4191. 186. S.J. Hong, S. Lee, J.S. Jang and J.S. Lee, Energ Environ Sci, 2011, 4, 1781-1787. 187. E.L. Miller, B. Marsen, D. Paluselli and R. Rocheleau, Electrochem Solid St, 2005, 8, A247-A249. 188. N. Gaillard, Y. Chang, J. Kaneshiro, A. Deangelis and E.L. Miller, P Soc Photo-Opt Ins, 2010, 7770. 189. K. Jäger, O. Isabella, L. Zhao and M. Zeman, Phys Status Solidi C, 2010, 7, 945-948. 190. C. Battaglia, C.M. Hsu, K. Söderström, J. Escarre, F.J. Haug, M. Charriere, M. Boccard, M. Despeisse, D.T.L. Alexander, M. Cantoni, Y. Cui and C. Ballif, Acs Nano, 2012, 6, 2790- 2797. 191. A.J.E. Rettie, H.C. Lee, L.G. Marshall, J.F. Lin, C. Capan, J. Lindemuth, J.S. McCloy, J.S. Zhou, A.J. Bard and C.B. Mullins, J Am Chem Soc, 2013, 135, 11389-11396. 192. T.W. Kim and K.-S. Choi, Science, 2014, 343, 990-994. 193. H. Dotan, K. Sivula, M. Gratzel, A. Rothschild and S.C. Warren, Energ Environ Sci, 2011, 4, 958-964. 194. B.D. James, G.N. Baum, J. Perez, K.N. Baum, Technoeconomic Analysis of Photoelectro- chemical (PEC) Hydrogen Production, 2009. 195. B.A. Pinaud, J.D. Benck, L.C. Seitz, A.J. Forman, Z.B. Chen, T.G. Deutsch, B.D. James, K.N. Baum, G.N. Baum, S. Ardo, H.L. Wang, E. Miller and T.F. Jaramillo, Energ Environ Sci, 2013, 6, 1983-2002. 196. A.W. Jeremiasse, J. Bergsma, J.M. Kleijn, M. Saakes, C.J.N. Buisman, M.C. Stuart and H.V.M. Hamelers, Int J Hydrogen Energ, 2011, 36, 10482-10489. 197. T.F. Jaramillo, K.P. Jørgensen, J. Bonde, J.H. Nielsen, S. Horch and I. Chorkendorff, Science, 2007, 317, 100-102. 198. J. Kibsgaard, Z.B. Chen, B.N. Reinecke and T.F. Jaramillo, Nat Mater, 2012, 11, 963-969. 199. D. Voiry, H. Yamaguchi, J.W. Li, R. Silva, D.C.B. Alves, T. Fujita, M.W. Chen, T. Asefa, V.B. Shenoy, G. Eda and M. Chhowalla, Nat Mater, 2013, 12, 850-855. 200. E.J. Popczun, J.R. McKone, C.G. Read, A.J. Biacchi, A.M. Wiltrout, N.S. Lewis and R.E. Schaak, J Am Chem Soc, 2013, 135, 9267-9270. 126 Bibliography

201. N. Gaillard, Y. Chang, J. Kaneshiro, A. Deangelis and E.L. Miller, P Soc Photo-Opt Ins, 2010, 7770, 77700V. 202. S. Licht, J Phys Chem B, 2001, 105, 6281-6294. 203. R.E. Rocheleau, E.L. Miller and A. Misra, Energ Fuel, 1998, 12, 3-10. 204. Y. Yamada, N. Matsuki, T. Ohmori, H. Mametsuka, M. Kondo, A. Matsuda and E. Suzuki, Int J Hydrogen Energ, 2003, 28, 1167-1169. 205. E.L. Miller, R.E. Rocheleau and S. Khan, Int J Hydrogen Energ, 2004, 29, 907-914. 206. J. Madan and M.P. Shaw, The Physics and Applications of Amorphous Semiconductors, Academic Press, Inc., California, 1988. 207. C.M. Fortmann, J. O’Dowd, J. Newton and J. Fischer, AIP Conference Proceedings, 1987, 157, 103-110. 208. L.H. Hihara, A. Iwane, S. Voss and R.E. Rocheleau, Corros Sci, 1999, 41, 1403-1417. 209. Y. Tawada, H. Okamoto and Y. Hamakawa, Appl Phys Lett, 1981, 39, 237-239. 210. F. Zhu, J. Hu, A. Kunrath, I. Matulionis, B. Marsen, B. Cole, E.L. Miller and A. Madan, Solar Hydrogen and Nanotechnology II, 2007, 6650, S6500-S6500. 211. I. Matulionis, F. Zhu, J. Hu, T. Deutsch, A. Kunrath, E.L. Miller, B. Marsen and A. Madan, Solar Hydrogen and Nanotechnology Iii, 2008, 7044. 212. J. Hu, F. Zhu, A. Kunrath, T. Deutsch and N. Gaillard, MRS Spring Meeting, 2014, 167-170. 213. J. Hu, F. Zhu, I. Matulionis, A. Kunrath, T. Deutsch, L. Kuritzky, E.L. Miller and A. Madan, 23rd European Photovoltaic Solar Energy Conference and Exhibition, 2008, 69-72. 214. I. Matulionis, J. Hu, F. Zhu, J. Gallon, N. Gaillard, T. Deutsch, E.L. Miller and A. Madan, P Soc Photo-Opt Ins, 2010, 7770. 215. A. Stavrides, A. Kunrath, J. Hu, R. Treglio, A. Feldman, B. Marsen, B. Cole, E.L. Miller and A. Madan, Solar Hydrogen and Nanotechnology, 2006, 6340, U149-U156. 216. F. Zhu, J. Hu, I. Matulionis, T. Deutsch, N. Gaillard, A. Kunrath, E.L. Miller and A. Madan, Philos Mag, 2009, 89, 2723-2739. 217. J. Hu, F. Zhu, I. Matulionis, T. Deutsch, N. Gaillard, E.L. Miller and A. Madan, MRS Spring Meeting, 2009, 1171, 121-126. 218. F. Zhu, I. Matulionis, N. Gaillard, Y. Chang, J. Hu, J. Gallon and A. Madan, MRS Online Proceedings Library, 2013, 1539. 219. I. Martin, M. Vetter, A. Orpella, J. Puigdollers, A. Cuevas and R. Alcubilla, Appl Phys Lett, 2001, 79, 2199-2201. 220. C. Battaglia, C.M. Hsu, K. Soderstrom, J. Escarre, F.J. Haug, M. Charriere, M. Boccard, M. Despeisse, D.T.L. Alexander, M. Cantoni, Y. Cui and C. Ballif, Acs Nano, 2012, 6, 2790- 2797. 221. K. Jager, O. Isabella, L. Zhao and M. Zeman, Phys Status Solidi C, 2010, 7, 945-948. 222. M. Python, E. Vallat-Sauvain, J. Bailat, D. Domine, L. Fesquet, A. Shah and C. Ballif, J Non-Cryst Solids, 2008, 354, 2258-2262. 223. H. Tan, L. Sivec, B. J. Yan, R. Santbergen, M. Zeman and A.H.M. Smets, Appl Phys Lett, 2013, 102, 153902. 224. K. Jager, M. Fischer, R.A.C.M.M. van Swaaij and M. Zeman, J Appl Phys, 2012, 111. 225. K. Jager, M. Fischer, R.A.C.M.M. van Swaaij and M. Zeman, Opt Express, 2013, 21, A656- A668. Bibliography 127

226. K. Jager, D.N.P. Linssen, O. Isabella and M. Zeman, Photonics for Solar Energy Systems V, 2014, 9140. 227. H. Sai, T. Koida, T. Matsui, I. Yoshida, K. Saito and M. Kondo, Appl Phys Express, 2013, 6, 104101. 228. Y. Lin, C. Battaglia, M. Boccard, M. Hettick, Z.B. Yu, C. Ballif, J W. Ager and A. Javey, Nano Lett, 2013, 13, 5615-5618. 229. J. Meier, J. Spitznagel, U. Kroll, C. Bucher, S. Fay, T. Moriarty and A. Shah, Proceedings of 3rd World Conference on Photovoltaic Energy Conversion, 2003, 2801-2805. 230. I. Oh, J. Kye and S. Hwang, Nano Lett, 2012, 12, 298-302. 231. J.R. McKone, A.P. Pieterick, H.B. Gray and N.S. Lewis, J Am Chem Soc, 2013, 135, 223- 231. 232. Z. Huang, J.R. McKone, C. Xiang, R.L. Grimm, E.L. Warren, J.M. Spurgeon, H.-J. Lewerenz, B.S. Brunschwig, N.S. Lewis, Int J Hydrogen Energ, 2014, In Press. 233. J.R. McKone, E.L. Warren, M.J. Bierman, S.W. Boettcher, B.S. Brunschwig, N.S. Lewis and H.B. Gray, Energ Environ Sci, 2011, 4, 3573-3583. 234. J.O. Bockris, I.A. Ammar and A. Huq, Journal of Physical Chemistry, 1957, 61, 879-886. 235. A. Caillard, P. Brault, J. Mathias, C. Charles, R.W. Boswell and T. Sauvage, Surf Coat Tech, 2005, 200, 391-394. 236. X.M. Yan, S. Kwon, A.M. Contreras, J. Bokor and G.A. Somorjai, Nano Lett, 2005, 5, 745- 748. 237. J.R.V. Garcia and T. Goto, Mater Trans, 2003, 44, 1717-1728. 238. R. Liu, Y. Lin, L.Y. Chou, S.W. Sheehan, W. He, F. Zhang, H.J.M. Hou, D. Wang, Angew. Chem. Int. Ed., 2011, 50, 499-502. 239. X.G. Yang, C. Du, R. Liu, J. Xie and D. Wang, J Catal, 2013, 304, 86-91. 240. M.H. Lee, K. Takei, J.J. Zhang, R. Kapadia, M. Zheng, Y.Z. Chen, J. Nah, T.S. Matthews, Y.L. Chueh, J.W. Ager and A. Javey, Angew Chem Int Edit, 2012, 51, 10760-10764. 241. Y.W. Chen, J.D. Prange, S. Duhnen, Y. Park, M. Gunji, C.E.D. Chidsey and P.C. McIntyre, Nat Mater, 2011, 10, 539-544. 242. N.C. Strandwitz, D.J. Comstock, R.L. Grimm, A.C. Nichols-Nielander, J. Elam and N.S. Lewis, J Phys Chem C, 2013, 117, 4931-4936. 243. N.P. Dasgupta, C. Liu, S. Andrews, FB. Prinz, and P. Yang, J. Am. Chem. Soc., 2013, 135, 12932–12935. 244. P. Dai, J. Xie, M.T. Mayer, X. Yang, J. Zhan and D. Wang, Angew. Chem.-Int. Edit., 2013, 52, 11119-11123. 245. M.J. Weber, A.J.M. Mackus, M.A. Verheijen, C. van der Marel, and W.M.M. Kessels, Chem Mater, 2012, 24, 2973–2977. 246. A.J.M. Mackus, N. Leick, L. Baker, and W.M.M. Kessels, Chem Mater, 2012, 24, 1752– 1761. 247. J. Hamalainen, F. Munnik, M. Ritala and M. Leskela, Chem Mater, 2008, 20, 6840-6846. 248. T. Aaltonen, M. Ritala, T. Sajavaara, J. Keinonen and M. Leskela, Chem Mater, 2003, 15, 1924-1928. 249. J. Hamalainen, E. Puukilainen, T. Sajavaara, M. Ritala and M. Leskela, Thin Solid Films, 2013, 531, 243-250. 250. J. Hamalainen, M. Ritala and M. Leskela, Chem Mater, 2014, 26, 786-801. 128 Bibliography

251. H.B.R. Lee, K.L. Pickrahn and S.F. Bent, J. Phys. Chem. C, 2014, 118, 12325-12332. 252. C.P. Wade and C.E.D. Chidsey, Appl. Phys. Lett., 1997, 71, 1679. 253. H.B.R. Lee, S.H. Baeck, T.F. Jaramillo and S.F. Bent, Nano Lett, 2013, 13, 457-463. 254. K. Qadir, S.H. Kim, S.M. Kim, H. Ha and J.Y. Park, J Phys Chem C, 2012, 116, 24054- 24059. 255. C. Battistoni, A.M. Giuliani, E. Paparazzo and F. Tarli, J Chem Soc Dalton, 1984, 1293- 1299. 256. W.M.M. Kessels, H.C.M. Knoops, S.A.F. Dielissen, A.J.M. Mackus and M.C.M. van de Sanden, Appl Phys Lett, 2009, 95. 257. W. Setthapun, W.D. Williams, S.M. Kim, H. Feng, J.W. Elam, F.A. Rabuffetti, K.R. Poeppelmeier, P.C. Stair, E.A. Stach, F.H. Ribeiro, J.T. Miller and C.L. Marshall, J Phys Chem C, 2010, 114, 9758-9771. 258. L. Baker, A.S. Cavanagh, D. Seghete, S.M. George, A.J.M. Mackus, W.M.M. Kessels, Z.Y. Liu and F.T. Wagner, J Appl Phys, 2011, 109. 259. A.H.M. Smets, W.M.M. Kessels and M.C.M. van de Sanden, Appl Phys Lett, 2003, 82, 1547-1549. 260. G. Lucovsky, R.J. Nemanich and J.C. Knights, Phys Rev B, 1979, 19, 2064-2073. 261. D.V. Tsu, G. Lucovsky and M.J. Mantini, Phys Rev B, 1986, 33, 7069-7076.

Summary

Long term concerns about climate change and fossil fuel depletion will require a transition towards energy systems powered by solar radiation or other renewable sources. Novel concepts based on silicon materials and devices are investigated for applications in the next generation photovoltaic (PV) devices and photoelectrochemical (PEC) water splitting for solar energy conversion and storage.

Expanding thermal plasma chemical vapor deposition (ETP-CVD), a remote plasma synthesis method, is verified as an efficient process for the fabrication of free-standing silicon nanocrystals (Si NCs) on an industrial scale. The unique physical, mechanical and electrical properties of Si NCs might open routes to new PV concepts to breach the so-called Shockley-Queisser limit using mechanisms like multiple exciton generation and up- or down-conversion of the incident spectrum.

Under intensive laser illumination conditions, the thermal heating effects of Si NCs become the dominant mechanism for the transverse optical (TO) mode red-shifts of the first order Si-Si peak in reference to the bulk c-Si in the Raman spectrum. The free-standing Si NCs can be heated to their melting points by a well-focused laser, and the temperature can be determined by the measured ratio of Anti-Stokes-to-Stokes TO mode intensities. In contrast, Si NCs in various matrices can hardly be heated using the same amount of laser power due to good thermal conductivity. If the free-standing Si NCs are further heated, the intensity of the blackbody radiation in Raman spectrum starts to compete with that of the TO mode.

Various PEC/PV configurations for solar water splitting structures are discussed in this thesis to directly store the solar energy in the form of hydrogen fuels. In Chapter 4, the a-Si:H/a-Si:H double-junction solar cell is demonstrated as the simplest and easiest option to meet the requirements for the integration with gradient-doped W:BiVO4 photoanode, by considering the stability in aqueous solutions, simple fabrication process, matching spectral response, voltage and current density. The optimization steps of the a-Si:H/a-Si:H solar cells are carried out in both experiments and simulations, by varying the top i-layer thickness in reference to the AM 1.5 spectrum and the spectrum transmitted through the BiVO4 photoanode respectively. The stability of the a-Si:H/a-Si:H solar cell shows less sensitive light-induced degradation kinetics under the spectrum transmitted through the BiVO4 photoanode from that under the standard AM 1.5 spectrum.

In Chapter 5, the performance of the front BiVO4 photoanode is further improved, comparing with the studies in Chapter 4, by improving photon absorption and carrier collection. Photon absorption is enhanced by the application of light trapping techniques on the BiVO4 photoanode using textured TCO glass substrates. The carrier collection is optimized based on our new findings on diffusion length of the photogenerated charge carriers in an undoped BiVO4. By ingenious design of the gradient W-dopant profile, the thickness of the film is extended without 130 Summary

deteriorating the carrier separation efficiency. The catalytic limitation is overcome by electrodepositing a thin film of cobalt phosphate as water oxidation catalysts on the surface of BiVO4. The optimized front photoanode is combined with three types of solar cells to form a hybrid PEC/PV solar water-splitting. The collaboration of a BiVO4/a-Si:H/nc-Si:H photoanode demonstrates the best performance concerning the better solar spectrum utilization of nc-Si:H up to 1100 nm near-infra-red region. A 5.2% solar-to-hydrogen conversion efficiency, which is the highest ratio of metal-oxide based photoanodes ever reported, has been achieved by this PEC/PV configuration.

Besides the photoanode device for oxygen evolution reaction, a photocathode based on thin-film silicon technology is designed and optimized as well, to form an unbiased photoanode /photocathode aiming to revolutionize solar water splitting. Photon absorption is enhanced by state-of-the-art implementation of light trapping techniques on the a-SiC:H photoanode using a glass substrate with integrated micro-textured photonic structures. The light traveling length is prolonged in the high quality grown nc-Si:H benefiting from the scattering morphology on the glass substrate. The carrier collection is boosted by our unprecedented design of the gradient boron dopant profile from the a-SiC:H p-layer to the i-layer. Novel spectral utilization techniques are applied in the device for the integrated PV junctions, supported by a theoretical optical model. A benchmark photocurrent density of -5.1 mA cm-2 at 0 V vs. RHE is achieved in the a-SiC:H/a- Si:H/nc-Si:H configuration. It is note-worthy to address that this photocathode does not contain a passivation layer nor any catalyst.

The efficient operation of a photocathode also requires metal catalysts to facilitate charge-transfer reactions at the interface between the semiconducting light absorbers and the electrolyte. Atomic layer deposition (ALD) is employed to fabricate the Pt nanoparticles and thin films as the hydrogen-evolution catalysts under varied conditions. Using MeCpPtMe3 and ozone as the precursors and substrate temperatures as low as 200 °C, a growth rate as fast as 1.1 Å/cycle is achieved. The electro-catalytic activity of ALD-grown Pt thin films on glassy carbon electrodes shows comparable performance for the hydrogen-evolution reaction as that of Pt films deposited using electron-beam evaporation.

Samenvatting

De verandering van het klimaat en de uitputting van fossiele-brandstofvoorraden maken een overgang naar energiesystemen gebaseerd op zonne-energie en andere duurzame energiebronnen noodzakelijk. Nieuwe, op silicium gebaseerde materialen en structuren worden nu onderzocht voor toepassing in de volgende generatie van fotovoltaïsche (PV) en foto-elektrochemische (PEC) apparaten die zonne-energie kunnen omzetten en opslaan.

Chemische dampdepositie met behulp van een expanderend thermisch plasma (ETP-CVD) heeft zich bewezen als een efficiënt proces voor het fabriceren van vrijstaande silicium nanokristallen (Si NCs) op industriële schaal. De unieke fysische, mechanische en elektrische eigenschappen van deze Si NCs brengen nieuwe PV concepten binnen handbereik die de zogenaamde “Shockley-Queisser limiet” kunnen doorbreken, bijvoorbeeld door middel van “multiple exciton generation” of de spectrale omvorming van het zonnespectrum.

Onder intense laserbelichting van vrijstaande Si NCs, zijn het de thermische effecten die roodverschuiving van de transversaal optische (TO) modus van de eerste Si-Si piek in het Raman spectrum domineren. Een gefocuste laser kan de vrijstaande Si NCs verhitten tot het smeltpunt en de temperatuur kan worden bepaald uit de intensiteitsverhouding tussen de Stokes en anti-Stokes pieken. Si NCs in een matrix daarentegen worden door de goede warmtegeleiding bij hetzelfde laservermogen nauwelijks verwarmd. Als de vrijstaande Si NCs verder worden verhit gaat de “black body” straling het signaal van de TO modus overheersen.

Naast Si NCs, behandelt dit proefschrift verschillende PEC/PV configuraties voor het ontleden van water met behulp van zonlicht. Deze configuraties maken het mogelijk om zonne-energie om te zetten en op te slaan in de vorm van op waterstof gebaseerde brandstoffen. In hoofdstuk 4 laten we zien dat de a-Si:H/a-Si:H tandem zonnecel, vanwege de stabiliteit in water, de eenvoud van fabricage, de spectrale respons, het voltage en de stroomdichtheid, de meest ideale zonnecel is om eenvoudig met een fotoannode van W:BiVO4 te integreren. De a-Si:H/a-Si:H zonnecellen zijn geoptimaliseerd door gebruik te maken van zowel experimenten als simulaties. De dikte van de bovenste i-laag is geoptimaliseerd voor zowel het AM 1.5 spectrum als voor het resulterende spectrum dat door de BiVO4 fotoanode wordt doorgelaten. Onder dit laatste spectrum toont de a- Si:H/a-Si:H zonnecel minder licht-geïnduceerde degradatie dan onder het standaard AM 1.5 spectrum.

In hoofdstuk 5 worden de prestaties van de BiVO4 fotoanode verder verbeterd door zowel de fotonabsorptie als de collectie van de ladingsdragers te optimaliseren. De fotonabsorptie wordt verhoogd door de lichtopsluiting in de BiVO4 fotoanode te verhogen door middel van getextureerde TCO/glassubstraten. Het collecteren van ladingsdragers is geoptimaliseerd aan de hand van onze nieuwe inzichten aangaande de diffusielengte van ladingsdragers in ongedoteerd BiVO4. Door een ingenieus ontworpen gradiënt in het W-doteringsprofiel aan te brengen kan de BiVO4 laag dikker gemaakt worden zonder dat dit de ladingscollectie vermindert. Een dunne film 132 Summary

van kobaltfosfaat, die door middel van elektrodepositie op het BiVO4 oppervlak is aangebracht, dient als katalysator voor wateroxidatie. Door deze geoptimaliseerde fotoanode te combineren met verschillende typen zonnecellen ontstaat een hybride PEC/PV waterontledingsstructuur. Omdat nc-Si het grootste deel van het zonnespectrum kan benutten (tot een golflengte van 1100 nm in het nabije infrarood) behaalt een BiVO4/a-Si:H/nc-Si:H fotoanode een zon-naar-waterstof- omzettingsrendement van maar liefst 5.2%. Dit is het hoogste rendement ooit behaald met een fotoanode gebaseerd op een metaal-oxide.

Naast de fotoanode is een fotokathode ontwikkeld op basis van dunne-film siliciumtechnologie. Dit maakt een revolutionaire fotoanode/fotokathode-combinatie mogelijk die geen voorspanning nodig heeft. De fotonabsorptie in de a-SiC:H fotoanode is verbeterd door middel van geavanceerde lichtopsluitingstechnieken zoals een glassubstraat met geïntegreerde fotonische micro-texturen. De lichtverstrooiing hieraan vergroot de weglengte van het licht in de nc-Si:H laag. Tevens is de collectie van de ladingsdragers enorm verbeterd door ons unieke ontwerp van het borium-dopingprofiel in de a-SiC:H p-laag. Een stroomdichtheid van -5.1 mA cm-2 bij 0 V vs. RHE is behaald met een a-SiC:H/a-Si:H/nc-Si:H configuratie. Het is noemenswaardig dat deze fotokathode nog een passiveringslaag nog een katalysator bevat.

Tenslotte vergt de efficiënte werking van een fotokathode ook een metaalkatalysator die de ladingsoverdrachtsreacties aan het interface tussen de halfgeleider en het elektrolyt bevordert. Atomaire-laagdepositie (ALD) is gebruikt om Pt nanodeeltjes en dunne films te fabriceren. Door MeCpPtMe3 en ozon als precursors te gebruiken in combinatie met een substraattemperatuur van slechts 200°C is een depositiesnelheid van 1.1 Å/cyclus behaald. Op koolstofelektrodes hebben de met ALD gefabriceerde Pt films een vergelijkbare katalytische werking als de opgedampte Pt films.

(Translated by Dr. Rudi Santbergen)

Publications related to this thesis Patent [1] L. Han, M. Zeman and A.H.M. Smets, Photoelectrochemical Solar Water-Splitting Device based on Amorphous Silicon Alloys, International patent (2013). Peer-reviewed journal articles [2] L. Han, F.F. Abdi, R. van de Krol, R. Liu, Z. Huang, H.-J. Lewerenz, B. Dam, M. Zeman and A.H.M. Smets, Efficient Water-Splitting Device Based on a Bismuth Vanadate Photoanode and Thin-Film Silicon Solar Cells, ChemSusChem, 2014, 7 (10), 2832-2838 (Cover highlight). [3] L. Han, I.A. Digdaya, T.W.F. Buijs, F.F. Abdi, Z. Huang, R. Liu, B. Dam, M. Zeman, W.A. Smith and A.H.M. Smets, Gradient Dopant Profiling and Spectral Utilization of Monolithic Thin-Film Silicon Photoelectrochemical Tandem Devices for Solar Water Splitting, Journal of Material Chemistry A, DOI: 10.1039/C4TA05523C, 2015 (Cover highlight). [4] L. Han, F.F. Abdi, P. Perez Rodriguez, B. Dam, R. van de Krol, M. Zeman and A.H.M. Smets, Optimization of Amorphous Silicon Double-Junction Solar Cells for an Efficient Photoelectrochemical Water-splitting Device Based on a BiVO4 Photoanode, Physical Chemistry Chemical Physics, 2014, 16 (9), 4220 – 4229. [5] F.F. Abdi, L. Han, A.H.M. Smets, M. Zeman, B. Dam and R. van de Krol, Efficient Solar Water Splitting by Enhanced Charge Separation in a Bismuth Vanadate-Silicon Tandem Photoelectrode, Nature Communications, 2013, 4:2195. [6] L. Han, M. Zeman and A.H.M. Smets, Raman Study on the Laser Induced Heating Effects of Free-Standing Silicon Nanocrystals, Nanoscales (2014, under review). [7] L. Han, M. Zeman and A.H.M. Smets, High Rate Synthesis of Silicon Nanocrystals by Expanding Thermal Plasma Chemical Vapor Deposition, Applied Physics Letter (2015, under review). [8] R. Liu, L. Han, Z. Huang, I.M. Ferrer, A.H.M. Smets, M. Zeman, B.S. Brunschwig and N.S. Lewis, A Low Temperature Synthesis of Electrochemical Active Pt Nanoparticles and Thin Films by Atomic Layer Deposition on Si(111) and Glassy Carbon Surfaces, Thin Solid Films (2014, under review). [9] I.A. Digdaya, L. Han, T.W.F. Buijs, B. Dam, M. Zeman, A.H.M. Smets and W.A. Smith, Extracting Large Photovoltages from a-SiC:H Photocathodes with an Amorphous TiO2 Front Surface Field Layer for Solar Hydrogen Evolution, Energy & Environmental Science (2014, under review). [10] F.F. Abdi, L. Han, W.A. Smith, A.H.M. Smets, M. Zeman, B. Dam and R. van de Krol, Design Rules for Hybrid Photoelectrode for Water Splitting (in preparation). 134 Publications related to this thesis

Invited talks in international conference, extended abstracts & proceeding articles [11] R. Vasudevan, Z. Thanawala, T.W.F. Buijs, I.A. Digdaya, D. Deligiannis, L. Han, P. Perez Rodriguez, B. Dam, M. Zeman, W.A. Smith and A.H.M. Smets, Silicon Based Devices for Photoelectrochemical/Photovoltaic Water Splitting, MRS Spring Meeting (2015, San Francisco, USA, submitted). [12] L. Han, I.A. Digdaya, W.A. Smith, B. Dam, M. Zeman and A.H.M. Smets, Photovoltaic /Photoelectrochemical Devices Based on Thin-Film Silicon Alloys for Effective Hydrogen Evolution, The 6th World Conference on Photovoltaic Energy Conversion (2014, Kyoto, Japan). [13] L. Han, I.A. Digdaya, W.A. Smith, B. Dam, M. Zeman and A.H.M. Smets, Spectral Response Study of Earth-Abundant Thin-Film Silicon Alloys for Effective Photoelectrochemical Hydrogen Evolution, 562nd Wilhelm and Else Heraeus-Seminar: From Sunlight to Fuels (2014, Bonn, Germany). [14] L. Han, I.A. Digdaya, F.F. Abdi, B. Dam, W.A. Smith, M. Zeman and A.H.M. Smets, Earth-Abundant Silicon Based and Cost-Effective PV/PEC Devices: The Crucial Role of Silicon Based PV Technology, Material Research Society Spring Conference (2014, San Francisco, USA). [15] I.A. Digdaya, L. Han, F.F. Abdi, B. Dam, M. Zeman, A.H.M. Smets and W.A. Smith, Artificial Photosynthesis from a Silicon Based Monolithic PV/PEC Device, Material Research Society Spring Conference (2014, San Francisco, USA). [16] A.H.M. Smets, P. Perez Rodriguez, L. Han and M. Zeman, Earth Abundant Silicon Based Thin-Film Photovoltaic-Photoelectrochemical Water-Splitting Devices, 23rd International Photovoltaic Science and Engineering Conference (PVSEC-23) Proceeding (2013, Taipei, Taiwan). [17] L. Han, F.F. Abdi, B. Dam, R. van de Krol, A.H.M. Smets and M. Zeman, A Hybrid Photoelectrochemical Water-splitting Configuration Based on Bismuth Vanadate Photoanode with Improved Light-Trapping Technique and Thin-Film Silicon Solar Cells, European Photovoltaic Solar Energy Conference and Exhibition (2013, Paris, France). [18] L. Han, I.A. Digdaya, W.A. Smith, B. Dam, M. Zeman and A.H.M. Smets. Hydrogenated Amorphous Silicon Carbide Photocathode for Photoelectrochemical Water Reduction, 12th ISACS: International Symposia on Advancing the Chemical Sciences, Royal Society of Chemistry (2013, Cambridge, UK). [19] L. Han, A.H.M. Smets and M. Zeman, Intensive Luminescence from Laser Heated Freestanding Silicon Nanocrystals, 25th International Conference on Amorphous and Nano-crystalline Semiconductors (ICANS) (2013, Toronto, Canada). [20] L. Han, A.H.M. Smets and M. Zeman, Expanding Thermal Plasma Chemical Vapor Deposition: an Efficient Process for the Fabrication of Free Standing Silicon Nanocrystals, 22nd International Photovoltaic Science and Engineering Conference (PVSEC-22) Proceeding (2012, Hangzhou, China). Publications related to this thesis 135

[21] L. Han, A.H.M. Smets and M. Zeman, Raman Study of the Properties of Free-Standing Silicon Nanocrystals Using Laser Induced Thermal Heating, 59th American Vacuum Society International Symposium and Exhibition (2012, Tampa, USA). Poster presentations in international conference, extended abstracts & proceeding articles [22] R. Vasudevan, Z. Thanawala, L. Han, H. Tan, M. Zeman and A.H.M. Smets, An a- Si:H/c-Si Heterojunction/nc-Si:H Tandem Solar Cell for All-Silicon Based Photoelectrochemical Water Splitting, The 6th World Conference on Photovoltaic Energy Conversion (2014, Kyoto, Japan). [23] L. Han, F.F. Abdi, B. Dam, R. van de Krol, M. Zeman and A.H.M. Smets, Optical Modeling of an Efficient Water-Splitting Device Based on Bismuth Vanadate Photoanode and Micromorph Silicon Solar Cells. 40th IEEE Photovoltaic Specialists Conference (PVSC), pp. 3083-3086, DOI:10.1109/PVSC.2014.6925589 (2014, Denver, USA). [24] L. Han, F.F. Abdi, R. van de Krol, B. Dam, M. Zeman and A.H.M. Smets, Effective Water-splitting Device Based on Bismuth Vanadate Photoanode and Thin-Film Silicon Solar Cells with Improved Spectrum Utilization, Frontiers of Nano Science and Technology (2014, Pasadena, USA). [25] Z. Huang, L. Han, H. Zhang, C. Xiang, H. Audesirk, K. Whitesell, H.-J. Lewerenz, B.S. Brunschwig, J.M. Spurgeon and N.S. Lewis, Control of Interfacial Energetics at Hydrogen-Evolution Photocathodes through Lithography, Frontiers of Nano Science and Technology (2014, Pasadena, USA). [26] L. Han, I.A. Digdaya, A.H.M. Smets, M. Zeman, B. Dam and W.A. Smith, Photoelectrochemical Water Reduction by Silicon Carbide Thin Films, Solar Fuels Gordon Research Conference, (2014, Ventura, USA). [27] L. Han, I.A. Digdaya, W.A. Smith, B. Dam, M. Zeman and A.H.M. Smets, Hydrogenated Amorphous Silicon Carbide Photocathode for Photoelectrochemical Water Reduction, Post Graduate Solar Fuels Symposium, Royal Society of Chemistry (2013, Cambridge, UK). [28] F.F. Abdi, M. de Respinis, L. Han, A.H.M. Smets, M. Zeman, B. Dam and R. van de Krol, Direct Solar-to-Fuel Conversion Hybrid-Device. Delft Energy Day (2013, Delft, Netherlands).

136 Publications related to this thesis

Acknowledgements

Life has come into a new chapter when I am about to accomplish my Ph.D. thesis. All my excitement, happiness, wonders and sorrows are hushed into peace in my heart like the misty evening among the silent trees. Doing academic research as a Ph.D. candidate for four years seems like a nazaritism or investment in the most energetic period in my life. This long march cannot be achieved without the kind help and contribution from many people. At this moment, it is the right time to express my sincere acknowledgement to all those who have ever offered me a hand.

My first thanks go to my promoters Prof. Miro Zeman and Dr. Arno Smets. Miro, I still remember the first time I met you during the 35th IEEE PVSC in Honolulu in June 2010 and I didn’t imagine half a year later you would offer me this interesting but challenging Ph.D. project in the PVMD group in the Netherlands. It is a great pleasure to work in your group and I am very grateful and delighted for everything you provided. Arno, words fail me when I try to express my gratitude to all your encouragement and support in pursuit of my dreams. I am indebted for all the insightful scientific discussions we had in these four years in your office, in the conference halls abroad, or even via Skype during the midnight. I am impressed by your passion in teaching and finicky attitude towards the errors in my manuscripts. I also appreciate the helps from the other faculty members in the PVMD group, Dr. René van Swaaij and Dr. Olindo Isabella. René, thank you for sharing your rich knowledge on semiconductor physics. Olindo, I am truly grateful for your careful corrections before my conference presentations, and I am always warmly inspired by your earnest manner in daily work.

I also want to express my gratitude to Prof. Bernard Dam and Prof. Braham Ferreira from TU Delft, Prof. Tom Gregorkiewicz from University of van Amsterdam Prof. Richard van de Sanden from Dutch Institute for Fundamental Energy Research, and Dr. Friedhelm Finger from Forschungszentrum Jülich GmbH to spend time reading this thesis, giving feedbacks and travel to Delft as my defense committee members.

This thesis would not have been possible without all the support from all the PVMD colleagues. I would like to thank Martijn Tijssen, Stefaan Heirman, Jan Chris Staalenburg, Remko Koornneef and Kasper Zwetsloot for the technical assistance. Martijn, I am in debt for disturbing you even in the weekends for the Amigo robot crashes and frequent discussions to try crazy modifications on ETP-CVD. Stefaan, I appreciate all your creative and imaginative ideas to help me solve the problems on the characterization setups. Besides, I want to thank our sectary Laura Bruns for your patient support with many paper work and frequent sample delivery. Parts of my experiments were carried out in DIMES Technology Center, the Charged Particle Optics (CPO) group and the Kavli Nanolab. Therefore I acknowledge Cassan Visser, Johan van der Cingel, Wim Wien, Charles de Boer, Carel Heerkens and Hozan Miro for providing me a nice cleanroom environment, maintaining the fancy setups so efficiently, delivering me the safety courses and 138 Acknowledgements

showing me the CHA evaporators. I feel happy about the generous 24-hour access to the Cleanroom Class 10000 in which I was allowed run my PECVD tools overnight after overnight. My Ph.D. project is also benefited directly from the postdoctoral scholars in the PVMD group, including Dr. Serge Solntsev, Dr. Klaus Jäger, Dr. Rudi Santbergen, Dr. Karol Jarolimek, Dr. Do Yun Kim and Dr. Sergiy Dobrovolskiy. Serge, thank you for teaching my MSc. students and myself the solar cell simulation by the fancy ASA software; Karol, thank you for sharing me your own thoughts on Si NCs; Klaus, my best friend not only on Facebook but also in office, my admiration for your intelligence and kindness is beyond words.

The PVMD group feels like a family when I am living abroad. I enjoyed working with my peer Ph.D. candidates and graduated colleagues, including Dr. Michael Wank, Dr. Solomon Agbo, Dr. Pavel Babal, Joke Westra, Dong Zhang, Marinus Fischer, Guangtao Yang, Mirjam Theelen, Wendelin Sprenger, Jimmy Melskens, Mark Workum, Andrea Ingenito, Hairen Tan, Ravi Vasudevan, Dimitris Deligiannis, Martijn van Sebille, Fai Tong Si, Paula Perez Rodriguez, Johan Blanker and Robin Vismara. I thank the PVMD peer colleagues for taking time out of your busy lives to help me learn the ropes, make me smile, and warn me of rocks ahead. I thank you for your warmth and wisdom, and for not laughing at me too hard when I made mistakes. Pavel, I am especially grateful for sharing your recipes and deposition experiences on Amigo; Joke, many thanks for teaching me how to make depositions by ETP-CVD in the beginning of my project; Marinus, thanking for introducing me your high-Voc material deposition recipes; Mirjam, thanks for sharing me the interesting the Dutch history, language and culture. Last but not the least, Guangtao, you are the first one I would like to turn to whenever I have met challenges in my research and daily life. I have received so warm welcome from you and Wei Cui even before my arrival in the Netherlands. I value our brotherliness forever.

It is a great honor and pleasure for me to supervise and co-supervise a few MSc. students for their graduation projects. I appreciate the contribution from Paula Perez Rodriguez, Thom Buijs, Karthik Subramanian, Mathew Alani, Zaid Thanawala and Albert Vullers. Paula, thank you for the careful Amigo depositions even until late midnights; Thom, you are the smartest boy for the PEC simulation of our cells. I am also pleased to share the same office with Gürkan Tevrizci and Araz Mohammed in the beginning of my Ph.D. year.

From 2012 to 2014, I was cooperating and doing PEC measurements almost one day per week in the MECS group in the Chemical Engineering Department in TU Delft and had some publications with the smart researchers from Institute for Solar Fuels in Helmholtz-Zentrum Berlin (HZB). I would like to thank Dr. Fatwa Abdi and Ibadillah Digdaya to cooperate with me for the photoanode and photocathode project respectively. The productive supervisions from Prof. Roel van der Krol, Dr. Wilson Smith and Prof. Bernard Dam are highly appreciated. You took me on the way to become a new chemist from a physicist. These three chapters cannot be done without your contribution.

In the third year of my Ph.D. project, I spent four precious months in the Molecular and Nanoscale Interfaces group of Joint Center for Artificial Photosynthesis at California Institute of Technology (JCAP, Caltech) in USA, where I finished Chapter 7 of this thesis and met many nice Acknowledgements 139

colleagues. Dr. Joshua Spurgeon and Dr. Bruce Brunschwig, I won’t forget how happy I was when I was offered this great opportunity to work in Caltech, which is an experience I would never trade with anything else in all my life. Special thanks to Prof. Nathan Lewis, Prof. Hans- Joachim Lewerenz and Prof. Carl Koval for the supervisions of my pinch-off project and ALD project. Nate, thank you sharing me your broad background in chemistry and inviting me to your house for Christmas party. I also thank Prof. Harry Gray and Prof. Harry Atwater for providing me both scientific ideas and equipment. The assistance in my paper work by the Caltech secretaries Susan Fuhs and Barbara Miralles are highly appreciated as well.

I am particularly indebted to my intelligent postdoctoral cooperators Dr. Teddy Zhuangqun Huang and Dr. Rui Liu for introducing me the magic pinch-off effect and ALD project. Teddy, I appreciate all the nano fabrication knowledge, characterization setups training and busy schedules arranged by you during my stay in Caltech. Rui, thank you for aligning with me in the struggle for truth in science and passion for publications. Many thanks should also go to my great postdoctoral/Ph.D. fellows including Dr. Hang Zhang, Dr. Ke Sun, Heather Audesirk, Kelsey Whitesell, Dr. Lan Zhou and Dr. Chengxiang Xiang. Thanks for all the help in the fabrication, characterization and simulation in my project. Los Angeles won’t be the city of angel in my heart without you. Special thanks to my housemate in Pasadena, Edward Haojiang Zhou, I forever remember these sunny days that you drove me to the most popular Asian restaurants in San Gabriel, Arcadia, Rosemond and Los Angeles downtown. The winter of 2013 in California was warmer than any of the previous ones in my life.

My early years as a graduate student were also benefited greatly from the mentoring by Prof. Jing Wang in Tsinghua National Laboratory for Information Science & Technology, Tsinghua University (THU) in China. It was your encouragement and wise supervision that made me the unique one among the 94 students in the same grade in the Institute of Microelectronic Engineering (IME) to finish the Master program half a year in advance. I also acknowledge Dr. Renrong Liang who introduced me to the PVMD group and useful information about the Dutch academic life in advance. I thank my classmates and labmates in IME. Jiangang Kong, Huijuan Liu and Zhen Tan, you are the treasure of my life. I want to thank Prof. Jing Wang, Prof. Wenjing Wang and the founder of SUNTECH group Dr. Zhengrong Shi to be the referees for my PhD application. Dr. Shi, you are the leading man to inspire me on the way to a sustainable energy career.

I have had the honor and pleasure to discuss the future application of photovoltaics and explore some experimental details with many bright scientists from other institutions and companies all over the world. I owe a deep professional thanks to Dr. İlker Doğan from Eindhoven University of Technology (TU/e) in the Netherlands for sharing your knowledge on the plasma processing of Si NCs by ETP-CVD in the beginning of my research, as well as Dr. Takehiko Nagai from Advanced Industrial Science and Technology (AIST) in Japan in the second year of my Ph.D. I appreciate the photoluminescence experiments and discussions with Dr. Dolf Timmerman, Frank Buters and Prof. Tom Gregorkiewicz from University of Amsterdam (UvA) in the Netherlands. I would also like to thank Dr. Karin Söderström from Ecole polytechnique fédérale de Lausanne (EPFL) in Switzerland for the assistance during the EQE measurement for multi-junction solar 140 Acknowledgements

cells. Many thanks goes to Dr. Leonid Khriachtchev from University of Helsinki in Finland, who shared many experiences on the laser heating for Si NCs with me. I appreciate Dr. Valeria Demontis and Dr. Carla Sanna from Laboratorio Fotovoltaico, Sardegna Ricerche in Italy to teach me depositing the first nc-Si:H layers. I thank my friends Dr. Wayne Zhenyu Wan, Steve Qiushi Li and Prof. Gavin Conibeer from University of New South Wales (UNSW) for discussions on the third generation photovoltaics. I should also acknowledge Dr. Shunsuke Kasashima from Tokyo Institute of Technology, Dr. Qiming Liu from Saitama University, and Kenichiro Takagi from Osaka Prefecture University, for helping me the simulation and accompanying me trips in West Europe, North America and East Asia. The assistance with Zhiqiang Cheng from Jilin Agriculture University in China for the HR-TEM characterizations are in my debt as well. I am thankful to Dr. Feng Zhu from MVSystems Inc. to exchange ideas with me on the a-SiC:H photocathode development. I apologize in advance for the omissions I know I will make in my attempt to enumerate those who have contributed. I would also happy to have my excellent designer friends such as Yawen Diao, Chenchen Shen, Dong Liu and Yao Lu for assisting me with some sketch plotting and photographs.

I am very fortunate to meet my close friends in Delft during these four years, especially the other three members of my Elephant Group, Yang Li, Kang Wang and Leilei Hu. I am so grateful for reminding me there is a world outside the confines of lab. I wish the happy hours we spent together at the table tennis tables, poker tables and the BBQ tables will be sustainable forever. There are days when life rewards us and seems to make amends by granting us a marvelous gift, the precious gift of friends (by Karin Schaefer). I am also happy that in previous four years I shared two apartments with my gentle housemates Dr. Zony Duan Zhao, Nigel Yiyi Yang, Dr. Jia Wei, Shuai Lu, Bo Jiang, Shi Qiu, Dr. Leicheng Guo and Peetikamol Kongsamai. I owe a sincere thanks to Chao His Ken and Eric Lieuw family inviting me to your house for monthly Christian parties and delicious Taiwanese food. God bless your kindness.

The Dutch Technology Foundation (STW) and the Netherlands Organisation for Scientific Research (NWO) granted my supervisor Dr. Arno Smets the VIDI project, which is the very important financial support for my Ph.D. research. I am thankful to the nation's taxpayers for their trust and I hope these are good investments.

I dedicated this thesis to my family members including parents, grandparents, aunts, uncles and cousins. My dear Mom and Dad, I cannot explain in words how grateful I am for your unconditional love and trust. You are my consolation in sorrow, my hope in misery, and my strength in weakness. Grandma, thank you for paying the painstaking care and the sweat to bring me up since my babyhood. Now, you should be glad to witness your masterpiece has grown on his way in becoming a true adult man.

Lihao Han Delft, the Netherlands September, 2014

Curriculum vitae

Lihao Han was born in the late midnight of 20th January 1986 in Zhejiang, China. Since 2004, he spent four years studying telecommunication engineering in Beijing University of Posts and Telecommunications. His bachelor project was about novel technology for cooperative network coding. In September 2008, he began his master study in microelectronic engineering in Tsinghua University. He was supervised by Prof. Jing Wang at Tsinghua National Laboratory for Information Science and Technology (TNLIST), focusing on the growth and characterization of Group IV semiconductor nanocrystals (e.g., Si/SiGe NCs) for third generation photovoltaics.

In February 2011, Lihao joined Prof. Miro Zeman’s Photovoltaic Materials and Devices group in Delft University of Technology in the Netherlands, working on the VIDI project with Associate Prof. Dr. Arno H.M. Smets. His doctoral project included various novel applications of Si materials for solar cells, solar water-splitting devices and solar water treatment. He synthesized Si/Ge/SiGe NCs by the expanding thermal plasma chemical vapor deposition (ETP-CVD) and multi-junction solar cells by plasma-enhanced chemical vapor deposition (PECVD). He developed monolithic photoelectrochemical (PEC) water-splitting device based on BiVO4 photoanodes and thin-film Si alloy photocathodes for effective oxygen and hydrogen evolution reaction, respectively. From November 2013 to March 2014, Lihao was working with Dr. Bruce S. Brunschwig and Prof. Nathan S. Lewis in the Joint Center for Artificial Photosynthesis at California Institute of Technology in USA. He investigated the interfacial kinetics of the Si/metal/electrolyte for water reduction and nano-structured platinum growth by atomic layer deposition on photoelectrodes.

Lihao is a member of American Vacuum Society (AVS) and Royal Society of Chemistry (RSC). He is also an active invited referee who has reviewed ~100 manuscripts for more than 20 types of scientific journals including Energy & Environmental Science, Solar Energy Materials & Solar Cells, Solar Energy, Chemistry Communications, Nanotechnology, etc. His water-splitting devices were widely reported by international media, and his results were highlighted as the cover images in the journal of ChemSusChem and Journal of Materials Chemistry A.