Towards High Solar to Fuel Efficiency: From Photonic Design, Interface Study, to Device Integration - CaltechTHESIS

Towards High Solar to Fuel Efficiency: From Photonic Design, Interface Study, to Device Integration - CaltechTHESIS
CaltechTHESIS
A Caltech Library Service
About
Browse
Deposit an Item
Instructions for Students
Towards High Solar to Fuel Efficiency: From Photonic Design, Interface Study, to Device Integration
Citation
Cheng, Wen-Hui (Sophia)
(2020)
Towards High Solar to Fuel Efficiency: From Photonic Design, Interface Study, to Device Integration.
Dissertation (Ph.D.), California Institute of Technology.
doi:10.7907/kd6a-xt88.
Abstract
Efficient unassisted solar fuel generation, a pathway to storable renewable energy in the form of chemical bonds, requires optimization of a photoelectrochemical device based on photonic design and interface study. We first focused on enhancing absorption via nanophotonic design of light absorbers. Near-unity, broadband absorption in sparse InP nanowire arrays with multi-radii and tapered nanowire array designs are simulated and experimentally demonstrated. Later, a few strategies are introduced to achieved high solar-to-fuel efficiency.
Optically, photoelectrochemical device would require the catalyst ensembles to be highly transparent. We report a record solar-to-hydrogen efficiency by integrating Rh nanoparticle catalysts onto photocathodes with minimal parasitic absorption and reflection losses in the visible range. The other two light management strategies have been developed and experimentally verified to create highly active and effectively transparent catalyst structures: i) arrays of mesophotonic dielectric cone structures that serve as tapered waveguide light couplers to efficiently guide incident light through apertures in an opaque catalyst into the light absorber, and ii) an effectively transparent catalyst consisting of arrays of micron-scale triangular cross-sectional metal grid fingers, which are capable of redirecting the incoming light to the open areas of the PEC cell without shadow loss.
The electronic properties of the surface films exposed to the electrolyte are also critical. The anatase TiO₂ protection layer on the photocathode creates a favorable internal band alignment for hydrogen evolution, promoting the transport of the excess electrons and inhibiting voltage drops. The interfacial conduction mechanism between the defected TiO₂ and metal catalysts is investigated. A combinatorial approach of electrochemistry, X-ray photoelectron spectroscopy, and resonant X-ray spectroscopy reveals the correlation between the interfacial quasi-metal phase with TiO₂ properties. By careful control of gas diffusion electrode assembling to maintain appropriate wetted catalyst interface, another record solar-to-CO efficiency with extended stability can be realized.
Item Type:
Thesis (Dissertation (Ph.D.))
Subject Keywords:
Photoelectrochemistry, Nanophotonics, Solar fuels, Water splitting, CO2 reduction, Surface analysis
Degree Grantor:
California Institute of Technology
Division:
Engineering and Applied Science
Major Option:
Materials Science
Thesis Availability:
Public (worldwide access)
Research Advisor(s):
Atwater, Harry Albert
Group:
JCAP
Thesis Committee:
Goddard, William A., III (chair)
Atwater, Harry Albert
Johnson, William Lewis
Houle, Frances A.
Defense Date:
20 April 2020
Non-Caltech Author Email:
sophiasophia0701 (AT) gmail.com
Funders:
Funding Agency
Grant Number
Department of Energy (DOE)
DE-SC0004993
NSF
EEC-1041895
Record Number:
CaltechTHESIS:05282020-183139057
Persistent URL:
DOI:
10.7907/kd6a-xt88
Related URLs:
URL
URL Type
Description
DOI
Article adapted for Chapter 2.
DOI
Article adapted for Chapter 3.
DOI
Article adapted for Chapter 4.
DOI
Article adapted for Chapter 5.
ORCID:
Author
ORCID
Cheng, Wen-Hui (Sophia)
0000-0003-3233-4606
Default Usage Policy:
No commercial reproduction, distribution, display or performance rights in this work are provided.
ID Code:
13731
Collection:
CaltechTHESIS
Deposited By:
Wen Hui Cheng
Deposited On:
01 Jun 2020 22:29
Last Modified:
08 Nov 2023 18:46
Thesis Files
Preview
PDF
- Final Version
See Usage Policy.
45MB
Repository Staff Only:
item control page
CaltechTHESIS is powered by
EPrints 3.3
which is developed by the
School of Electronics and Computer Science
at the University of Southampton.
More information and software credits
Towards High Solar to Fuel Efficiency: From Photonic
Design, Interface Study, to Device Integration

Thesis by

Wen-Hui (Sophia) Cheng

In Partial Fulfillment of the Requirements for the Degree of
Doctor of Philosophy

CALIFORNIA INSTITUTE OF TECHNOLOGY
Pasadena, California

2020
(Defended April 20, 2020)

ii

ã 2020
Wen-Hui (Sophia) Cheng
ORCID: 0000-0003-3233-4606

iii

ACKNOWLEDGEMENTS
First of all, I want to share my immense gratitude to my advisor, Prof. Harry Atwater, for
being supportive and always inspiring me with his passion toward research and science. Also, I
would like to thank my committee members, Dr. Frances Houle, Prof. Bill Johnson, and Prof. Bill
Goddard, for helpful suggestions to improve the thesis and set my vision. In memory to Prof. Hans
Joachim (Achim) Lewerenz, it was great honor to work with him. Prof. Thomas Hannappel and
Dr. Bruce Brunschwig, I appreciate their guidance and kind support when I needed.
Special thanks to Dr. Katherine Fountaine, for being my mentor when I first joined the group,
I have learned a lot from her in both science and life. Special shout-out to Dr. Matthias Richter,
for being one of the best cooperators and friends, I won’t be able to break through all these records
without his support, also want to thank him for introducing me the power of synchrotron work. Dr.
Sisir Yalamanchili and Dr. Erik Verlage, thank them for involving me in the TiO2 project, I am
glad that we finally have it work out. Thanks to Prof. Rengui Li, our photocatalysis work is not
included in this thesis, but I look forward to the fruitful outcome.
Other cooperators who either help me with my project or give me chance to learn from their
works including: Prof. Frank Dimroth, Prof. Nate Lewis, Prof. Yun-Jung Lu, Prof. Pin Chieh Wu,
Prof. Rebecca Saive, Prof. Dieter Schmeißer, Prof. Giulia Tabliabue, Prof. Joeseph DuChene, Dr.
Chengxiang Xiang, Dr. Ian Sullivan, Dr. David Larson, Dr. Colton Bukowsky, Dr. Matthias May,
Dr. Jens Ohlmann, Dr. David Lackner, Dr. Michael Kelzenberg, Dr. Michael Lichterman, Dr.
Ethan Crumlin, Dr. Walter Drisdell, Dr. Yi-Rung Lin, Dr. Ibadillah Digdaya, Philip Jahelka, Paul
Kempler, Alex Welch, Madeline Meier, and Azhar Carim. Thank them all.
I also want to take this chance to thank the best admin team ever, Christy Jenstad, Jennifer
Blankenship, Tiffany Kimoto, Kam flower, Mabel Chik, Jonathan Gross, Lyann Lau, Elizabeth
Jennings, Mabby Howard, and Tiffani Walker. Thank KNI heroes for maintaining the best
cleanroom for research, Guy DeRose, Melissa Melendes, Nils Asplund, Nathan Lee, Bert Mendoza,
Steven Martinez, Alex Wertheim, and Matthew Hunt. I am also grateful to Ali Ghaffari for
technical support and providing teaching opportunity in APh109.

iv
The whole Ateam is just wonderful, I wouldn’t enjoy the past six years as much without
everyone in the group. Girls’ night in Amigo, summer softball game, group retreat, and Christmas
party are all unforgettable memories. Thank the midnight team, Souvik, Joeson, Wei-Hsiang, YaoWei, and Pin Chieh, for rescuing me from problems when it was late in the night. Thank Phil for
his patient and effort as the amazing IT personnel. Also thank Rebecca for always trying to involve
people for different activities and caring our safety. Photocatalysis subgroup (Alex, Lucy, Aiden,
Rengui, Giulia, Joe, Eowyn, Yi-Rung, and Chuen-Yo) are being the great team for all the valuable
discussion.
JCAP as a friendly hub has the best research environment, but also being a nice place for food
and alcohol gathering. Special thanks to Azhar for inviting me to his well-planned parties.
Additional note to Kevin, thank him for his help with everything. The best “Dim Sum” group
(Anya, Yury the strong, Ruzan, Jonathan, Yuri the beautiful, Ghazaleh, Amir, Giulia, Pablo, Benji,
Kelly, Nina, and Pilar), “Craft and Tea” group (Wei-Lin, Yu-Xian, Michelle, Claire, Marianna,
Samuel, Xiaomei, and Shiori), Materials Science buddies (Daryl, Harpreet, Wei-Lin, Louis, Jin,
Max, Mike, Xiaolin, Jong Hun, and Danielle), thank them for all the fun time together. I would
also like to thank my surf student Andrea Perry for trusting me and giving me the chance to be an
independent mentor.
Association of Caltech Taiwanese (especially Sandy, Xavier, Hsiao-Yi, Ting, Louis, Jake,
Hunter, Allen, Yu-Li, Lucas, I-Hui, Yen-Yung, and Tina), thank them for their company that I
feel not lonely even away from home. Gathering events for Lunar New Year, dragon boat
festival, moon festival, and birthday parties are wonderful moment to be remembered, and I hope
all these tradition can be continued. Friends in Taiwan (especially Jian-Yang, Scott, Yi-Ting,
Jerry, Yi-Shan, Peka, Windy, Shiue-Lan, Pin Chieh, Yuri, Hsiu-Ming, Chuen-Yo, Rita, Gloria,
Batista, Xuan, Rex, Yun-Man, Ching, Yun, Jiao-Yu, Jay, Ming, Young, and Tzu-Chieh), even I
didn’t get to see them often, I know they are always there when I need. Special thanks to my
master advisor, Prof. Jen-Sue Chen, for her emotional support.
Lastly, I want to thank all my family members, I wouldn’t be able to go these far and finish
one of the most important challenge in my life without their love and encouragement. I feel
really lucky to get all the help and support during my journey in Caltech, and it would for sure

become the powerful fuels for my future. Please forgive me if I missed any of your name in the
list of acknowledgment, I indeed appreciate whoever interacted with me and made me a better
person. I love you all.

vi

ABSTRACT
Efficient unassisted solar fuel generation, a pathway to storable renewable energy in the form
of chemical bonds, requires optimization of a photoelectrochemical device based on photonic
design and interface study. We first focused on enhancing absorption via nanophotonic design of
light absorbers. Near-unity, broadband absorption in sparse InP nanowire arrays with multi-radii
and tapered nanowire array designs are simulated and experimentally demonstrated. Later, a few
strategies are introduced to achieved high solar-to-fuel efficiency.
Optically, photoelectrochemical device would require the catalyst ensembles to be highly
transparent. We report a record solar-to-hydrogen efficiency by integrating Rh nanoparticle
catalysts onto photocathodes with minimal parasitic absorption and reflection losses in the visible
range. The other two light management strategies have been developed and experimentally verified
to create highly active and effectively transparent catalyst structures: i) arrays of mesophotonic
dielectric cone structures that serve as tapered waveguide light couplers to efficiently guide
incident light through apertures in an opaque catalyst into the light absorber, and ii) an effectively
transparent catalyst consisting of arrays of micron-scale triangular cross-sectional metal grid
fingers, which are capable of redirecting the incoming light to the open areas of the PEC cell
without shadow loss.
The electronic properties of the surface films exposed to the electrolyte are also critical. The
anatase TiO2 protection layer on the photocathode creates a favorable internal band alignment for
hydrogen evolution, promoting the transport of the excess electrons and inhibiting voltage drops.
The interfacial conduction mechanism between the defected TiO2 and metal catalysts is
investigated. A combinatorial approach of electrochemistry, X-ray photoelectron spectroscopy,
and resonant X-ray spectroscopy reveals the correlation between the interfacial quasi-metal phase
with TiO2 properties. By careful control of gas diffusion electrode assembling to maintain
appropriate wetted catalyst interface, another record solar-to-CO efficiency with extended stability
can be realized.

vii

PUBLISHED CONTENT AND CONTRIBUTIONS
Portions of this thesis have been drawn from the following publications:
Katherine T. Fountaine, Wen-Hui Cheng, Colton R. Bukowsky, and Harry A. Atwater, Nearunity unselective absorption in sparse InP nanowire arrays. ACS Photonics, 3 (10), 1826-1832,
(2016). DOI: 10.1021/acsphotonics.6b00341
Contribution: fabrication and data collection.

Wen-Hui Cheng, Matthias H. Richter, Matthias M. May, Jens Ohlmann, David Lackner, Frank
Dimroth, Thomas Hannappel, Harry A. Atwater, and Hans-Joachim Lewerenz, Monolithic
photoelectrochemical device for direct water splitting with 19% efficiency. ACS Energy Letters 3
(8), 1795-1800, (2018). DOI: 10.1021/acsenergylett.8b00920
Contribution: conception and design, fabrication and data collection, simulation and analysis,
manuscript writing, and revision.

Wen-Hui Cheng, Matthias H. Richter, Ian Sullivan, David M. Larson, Chengxiang Xiang, Bruce
Brunschwig, Harry A. Atwater, CO2 reduction to CO with 19% efficiency in a solar-drive gas
diffusion electrode flow cell under outdoor solar illumination. ACS Energy Letters 5 (2), 470-476,
(2020). DOI: 10.1021/acsenergylett.9b02576

Contribution: conception and design, fabrication and data collection, manuscript writing and
revision.

Sisir Yalamanchili*, Erik Verlage*, Wen-Hui Cheng*, Katherine T. Fountaine, Philip R. Jahelka,
Paul A. Kempler, Rebecca Saive, Nathan S. Lewis, Harry A. Atwater, High broadband light
transmission for solar fuels production using dielectric optical waveguides in TiO2 nanocone
arrays. Nano Letters 20 (1), 502-508, (2020). DOI: 10.1021/acs.nanolett.9b04225 (*equal
contribution)

viii
Contribution: fabrication and data collection, simulation and analysis, manuscript writing, and
revision.

Wen-Hui Cheng, Matthias H. Richter, Michael Kelzenberg, Sisir Yalamanchili, Phillip R. Jahelka,
Pin Chieh Wu, Rebecca Saive, Bruce S. Brunschwig, Thomas Hannappel, and Harry A. Atwater.
Incorporating effective transparent catalysts for photoelectrochemical application. (Manuscript
in preparation)
Contribution: conception and design, fabrication and data collection, simulation and analysis,
manuscript writing, and revision.

Matthias H. Richter*, Wen-Hui Cheng*, Michael F. Lichterman, Ethan J. Crumlin, Walter Drisdell,
Harry A. Atwater, Dieter Schmeißer, Nathan S. Lewis, and Bruce S. Brunschwig, Interaction of
metal catalyst thin films with ALD titanium dioxide. (Manuscript in preparation) (*equal
contribution)
Contribution: fabrication, data collection, and manuscript writing.

ix

TABLE OF CONTENTS
ACKNOWLEDGEMENTS ....................................................................................................... iii
ABSTRACT ................................................................................................................................ vi
PUBLISHED CONTENT AND CONTRIBUTIONS.............................................................. vii
TABLE OF CONTENTS ........................................................................................................... ix
LIST OF FIGURES .................................................................................................................... xi
LIST OF TABLES ...................................................................................................................... xxii
CHAPTER 1 Introduction of Solar Fuels ................................................................................ 1
1.1 Fundamental of Photoelectrochemistry ............................................................................ 1
1.2 Photoelectrochemical Device ........................................................................................... 3
1.3 Solar-to-Fuel Efficiency ................................................................................................... 4
1.4 Thesis Outline ................................................................................................................... 5
CHAPTER 2 Broadband Adsorption InP NW ........................................................................ 7
2.1 Introduction ...................................................................................................................... 7
2.2 Experimental and Numerical Method .............................................................................. 8
2.3 Results and Discussion ..................................................................................................... 10
2.5 Conclusion and Outlook ................................................................................................... 21
CHAPTER 3 High Efficiency Solar to H2 PEC Device ........................................................... 22
3.1 Introduction ...................................................................................................................... 22
3.2 Experimental Method ....................................................................................................... 23
3.3 Results and Discussion ..................................................................................................... 28
3.4 Conclusion and Outlook ................................................................................................... 54
CHAPTER 4 High Efficiency Solar to CO PV-GDE Device .................................................. 56
4.1 Introduction ...................................................................................................................... 56
4.2 Experimental Method ....................................................................................................... 59
4.3 Results and Discussion ..................................................................................................... 62
4.4 Conclusion and Outlook ................................................................................................... 80
CHAPTER 5 Broadband Transmission TiO2 Nanocone ........................................................ 82
5.1 Introduction ...................................................................................................................... 82
5.2 Numerical and Experimental Method .............................................................................. 84

5.3 Results and Discussion ..................................................................................................... 87
5.4 Conclusion and Outlook ................................................................................................... 100
CHAPTER 6 Effectively Transparent Catalysts for PEC Device.......................................... 101
6.1 Introduction ...................................................................................................................... 101
6.2 Numerical and Experimental Method .............................................................................. 102
6.3 Results and Discussion ..................................................................................................... 106
6.4 Conclusion and Outlook ................................................................................................... 119
CHAPTER 7 Interface Analysis of Catalysts and Protection Layer ..................................... 121
7.1 Introduction ...................................................................................................................... 121
7.2 Experimental Method ....................................................................................................... 123
7.3 Results and Discussion ..................................................................................................... 124
7.4 Conclusion and Outlook ................................................................................................... 146
CHAPTER 8 Summary .............................................................................................................. 147
References ................................................................................................................................... 151

xi

LIST OF FIGURES
Figure 1.1: Schematic illustrations of the general operating principles for PEC device. ... 2
Figure 1.2: Schematic illustrations of device architectures for Type 1 and Type 2 particle based
PEC device, Tpye 3 and Type 4 planar catalyst coated semiconductor PEC device.4
Figure 2.1: Schematic of two different nanowire geometries used in simulation. ........... 10
Figure 2.2: Characterization of uniform radius array. (a) Absorption, reflection, and transmission
spectra for the PDMS-embedded array shown as inset of (d), measured using an integrating
sphere; (b) Reflection spectra for the array shown as inset of (d) in multiple configurations;
(c,d) Simulated spectra corresponding to the experimental results in (a,b) with solid thick
lines representing an average of three slight geometric variations (thin, dashed, dotted);
inset of (d) is scanning electron micrograph of the uniform radius array; see Table 2.1 for
array dimensions used in simulation. .................................................................... 12
Figure 2.3: Comparison of uniform, tapered, and multi-radii arrays. (a,b,c) Scanning electron
micrographs of the uniform, tapered, and multi-radii array, respectively; (d) Absorption
spectra for the PDMS-embedded arrays shown in (a,b,c), measured using an integrating
sphere; (e) Simulated absorption spectra corresponding to the experimental results in (d)
with solid thick lines representing an average of three slight geometric variations (thin,
dashed, dotted); see Table 2.2 for array dimensions used in simulation. Colors are
coordinated throughout the figure. ........................................................................ 14
Figure 2.4: Characterization of tapered array. (a) Absorption spectra at various incident angles for
the PDMS-embedded array (SEM image shown as inset), measured using an integrating
sphere; (b) Simulated spectra corresponding to the experimental results in (a) with solid
thick lines representing an average of three slight geometric variations (thin, dashed,
dotted), overlaid with the planar equivalent absorption spectra (108 nm thin film, black
dashed); see Table 2.3 for array dimensions used in simulation. .......................... 16
Figure 2.5: Characterization of multi-radii array. (a,b) Absorption spectra at various incident
angles for the PDMS-embedded array, shown as inset of (b), without and with a silver back
reflector, respectively, measured using an integrating sphere; (c,d) Simulated spectra
corresponding to the experimental results in (a,b) with solid thick lines representing an
average of three slight geometric variations (thin, dashed, dotted), overlaid with the planar
equivalent absorption spectra (109 and 218 nm thin films, black dashed); see Table 2.4 for
array dimensions used in simulation. .................................................................... 18
Figure 2.6: Simulated absorption vs. wavelength of the median tapered array with a back reflector,
separated by material. ............................................................................................ 20
Figure 3.1: Process flow for preparing the PEC device: (I) Chemical etching of the GaAs/GaInAs
cap layer stopping at the AlInP window layer. (II) Deposition of the TiO2 protection and
antireflection coating with ALD. (III) Photoelectrochemical deposition of a closed layer of
Rh nanoparticles onto the tandem. ........................................................................ 24

xii
Figure 3.2: (a) Cell used for high efficiency benchmarking with WE and CE in close vicinity.
The distance WE to window is 10 mm and the distance WE to CE is < 10 mm. (b) Front
view and (c) side view of the double glass cell used for gas collection. The distance WE to
window is 10 mm, the distance WE to membrane is 40 mm and the distance membrane to
CE is 20 mm. The membrane has an area of 5 cm2. Each compartment has a gas bubbler
for pre-saturation of the electrolyte with H2/O2 purging and gas outlets which are connected
to inverted water filled burette for gas collection. For both cells (a) and (b/c) the quartz
window is covered with black tape having an opening with Æ 20 mm. ............... 26
Figure 3.3: (a) Light spectrum of the solar simulator (ABET Sun 3000 Solar Simulator) and
AM1.5G spectrum. (b) Light spectrum of the solar simulator and AM1.5G with water filter.
(c) Uniformity map of the solar simulator illumination area. The band gaps of the dualjunction light absorber are indicated in (a) and (b). (Yellow color for top cell and orange
color for bottom cell.) ............................................................................................ 26
Figure 3.4: Illustration of the photoelectrochemical water splitting device structure after
functionalization with interfacial films and electrocatalysts. Band alignment at the
operation point is depicted on the side and zoomed in to gain the visibility......... 30
Figure 3.5: (a) Optical properties (A: absorption, T: transmission, R: reflection) of TiO2 (TTiP
ALD) in air. (b) Tauc plot of ALD grown TiO2. The intersection with the horizontal axis
indicates an indirect optical gap around 3.3 eV. ................................................... 30
Figure 3.6: (a) Work function measurements by UPS for the tandem, for TiO2 on the tandem, and
for Rh metal. The increase of work function from 4.1 eV to 4.5 eV was observed after
applying TiO2 protection layer on tandem. The Rh metal spectrum is measured on the foil
as a reference instead of the photoelectrochemical deposited nanoparticles. (b) Core level
shift of Ti 2p3/2 indicating ~0.3 eV downward band bending at the tandem/TiO2 interface
and nearly no band bending at the TiO2/Rh interface. The tandem/TiO2 sample was made
with 40 ALD cycles TiO2 on top of the tandem. The Rh has originally high metal work
function of 5.1 eV but does not create band bending at the junction with TiO2. This can be
explained by the pinch-off effect when the metal NPs are small enough that the Fermi level
would directly attach to the semiconductor Fermi level without creating a barrier 71. 31
Figure 3.7: Surface band alignment of the electrolyte interface layers (a) without and (b) with
TiO2. ...................................................................................................................... 31
Figure 3.8: SEM images and AFM microtopographs of the dual-junction PEC device with TiO2
coating with and without Rh catalyst nanoparticles. The scale bar is 500 nm. The AFM
images are scaled to the same 50 nm z-axis dynamic range. The surface roughness (RMS)
is 3.6 nm without Rh and 6.3 nm with Rh. ............................................................ 32
Figure 3.9: (a) Fine control of particle size d ranging from 10 nm to 100 nm is achievable by
appropriate adjustment of the potential during catalyst electrochemical deposition.
Stroboscopic deposition under white light illumination as shown in the upper left insert.
The three images on the right inset are SEM images with scale bar 2 µm. (b-d) Particle size

xiii
histograms correspond to each SEM image depicted from top to bottom in (a) with the
most frequent particle size indicated by d. ............................................................ 33
Figure 3.10: (a) X-ray diffraction data of ALD deposited TiO2 from TTiP or TDMAT precursor.
The TTiP TiO2 shows anatase crystalline phase while the TDMAT TiO2 is amorphous. (b)
XPS valance band spectra of TTiP and TDMAT TiO2. A defect band in TDMAT TiO2 can
be observed at -1 eV which facilitates hole transport in photoanodes 74. Instead, TTiP TiO2
exhibits an XPS spectrum without a defect band and would be more suited to prevent
recombination in photocathodes. ........................................................................... 34
Figure 3.11: (a) Reflectance, measured in air, of the dual-junction tandem solar cell with different
thicknesses of the TiO2 coating by changing the ALD deposited cycles. (b) Reflectance,
simulated by Lumerical FDTD, with different thicknesses of TiO2 for correlation with the
experimental results. .............................................................................................. 35
Figure 3.12: FDTD simulated (a) reflectance, (b) transmittance, and (c) absorption defined as
A = 1 - R - T of different Rh particle sizes on 30nmTiO2/AlInP (window layer of the
tandem). ................................................................................................................. 36
Figure 3.13: Reflectance, measured in air, of samples with different Rh NPs size on the dualjunction tandem solar cell with 1500 ALD cycles TiO2 corresponding to a layer thickness
of 30 nm. ................................................................................................................ 37
Figure 3.14: Optoelectronic properties of the surface functionalized electrolyte / Rh / TiO2 / oxide
/ AlInP - GaInP / GaInAs / GaAs water splitting device. (a) Reflectivity, measured in air,
of the dual-junction tandem solar cell without ARC (black curve), secondly reflectivity
obtained after TiO2 coating (blue curve) and after photoelectrochemically deposited Rh
NPs (yellow curve). Reflectivity is larger than under operation in the electrolyte due to the
different refractive indices of air and water. (b) Comparison of the output characteristics of
the tandem device after cap layer etching and of the fully surface functionalized
photoelectrode. The orange arrows indicate the improvement after incorporation of the
TiO2 layer. ............................................................................................................. 38
Figure 3.15: The enhancement of absorption based on the reduction of the reflectivity for the PEC
device due to employment of TiO2 layer. (Absorption = 1-Reflection) The 15% average
increase of absorption can directly contribute to the enhancement of photocurrent.38
Figure 3.16: Relative EQE of a fully processed PEC tandem device: the bandgap combination is
determined to be around 1.78 eV for the top cell and 1.26 eV for the bottom cell.40
Figure 3.17: (a) Calculated optical concentration ratio of the non-parallel light-beam of solar
simulator illumination in PEC cells for plane wavefront and spherical wavefront as a
function of water path length. (b) Illustration of the spherical wavefront case. The
concentration ratio (CR = A0/ACR) depends on the exact sample area A0. (c) Illustration of
the plane wavefront case. An opening aperture in front of the quartz window of the PEC
cell with a diameter of 2 cm was used in this study. The beam divergence was
experimentally determined to be QV = 1.8 ° vertically and QH = 2.5 ° horizontally by

xiv
measuring the size increase of the light beam through a 2 cm aperture at specific
distances (10 cm to 30 cm). ................................................................................... 40
Figure 3.18: Calculated maximum STH efficiency as function of (a) Tafel slope A and exchange
current density J0, (b) ohmic drop. Maximum obtainable efficiencies for the given tandem
absorber are shown in the detailed-balance scheme as a function of the catalyst parameters
and resistivity loss. The maximum photocurrent density was scaled to the experimentally
determined current density under strong cathodic bias. The blue star indicates our device
under pH 0 condition and the red star indicates our device under pH 7 condition.42
Figure 3.19: Tafel plots of (a) Rh and (b) RuO2 catalysts under pH 0 and pH 7 conditions. The
Tafel slopes are 34, 38, 83, and 100 mV·dec-1 for Rh-pH0, Rh-pH7, RuO2-pH0, and RuO2pH7 respectively. ................................................................................................... 42
Figure 3.20: Output characteristics of the RuO2-Ge/GaInAs/GaInP/AlInP/anatase TiO2/Rhelectrolyte dual junction tandem structure. (a) Photocurrent-voltage characteristics in
acidic (pH 0), neutral (pH 7) electrolyte, and in neutral electrolyte including an AEM
membrane. (b) Chronoamperometric data of the initial temporal regime. (c) Stability
measurements at -0.4 V vs. RuO2 counter electrode for acidic and neutral pH. (d) Hydrogen
and oxygen gas collection for operation in acidic (open spheres) and neutral (full spheres)
electrolyte. The measured gas volume for oxygen (blue symbols) and hydrogen (red
symbols) is overlaid with the expected produced gas volume, as calculated from charge
passed through the anode and cathode. ................................................................. 44
Figure 3.21: Potential-pH equilibrium diagram for the system titanium-water system at 25 °C,
adapted from ref. 77. For pH 0, the stable region is small. Upon overpotential to hydrogen
evolution, corrosion sets in, which ultimately leads to the degradation of the device and its
efficiency. .............................................................................................................. 45
Figure 3.22: Comparison of realized limiting STH efficiencies and historic development. The
analysis refers to a theoretical benchmarking value htheo and takes into account the top and
bottom cell band gaps for the respective photolysis cells; also shown are the institutions of
the contributing research teams. Abbreviations: NREL - National Renewable Energy
Laboratory, USA; ISE - Institute for Solar Energy, Germany; JCAP - Joint Center for
Artificial Photosynthesis, Caltech; TU-I - Ilmenau University of Technology, Germany;
HZB - Helmholtz Zentrum Berlin, Germany. The bar chart on the right indicates the
achieved efficiency with respect to the respective theoretical limit (!theo ∗). ..... 47
Figure 3.23: (a) Stability measurements at 0 V vs. RuO2 counter electrode for acidic and neutral
pH. The result from ref. 5 are adapted and included for comparison as the black curve. (b)
Chronoamperometric measurements at -0.4 V vs. RuO2 counter electrode for acidic and
neutral pH. The results from ref. 5 at 0.6 V vs. RHE are adapted and included for
comparison in black. Currents rescaled based on the reported efficiency in ref. 5 are shown
in blue. ................................................................................................................... 49
Figure 3.24: Contact angle measurement for pH0 1 M HClO4 (a, c) or pH7 0.5 M phosphate buffer
(b, d) on the tandem (a, b) or on the TiO2/tandem (c, d) sample. The image was analyzed

xv
with ImageJ with the help of the “Drop Analysis” plugin developed at the École
polytechnique fédérale de Lausanne (EPFL) (http://bigwww.epfl.ch/demo/dropanalysis/).
The larger contact angle of phosphate buffer indicates higher surface tension, which can
lead to more severe bubble accumulation and larger photocurrent density fluctuations. 50
Figure 3.25: Chronoamperometric measurements at 0 V vs. RHE for Rh/Tandem device without
protection layer in acidic environment. ................................................................. 51
Figure 3.26: X-ray photoelectron spectra of tandem samples after each step in the PEC device
production process: after removing the GaAs/GaInAs cap layer by chemical etching (black
curve, indicated as Tandem etched), after deposition of the TiO2 layer by ALD (green
curves, indicated as +TTiP TiO2); and after photoelectrochemical deposition of Rh
nanoparticle catalysts (blue curve, indicated as +Rh). As a reference, spectra of metallic
Rh electrode are included (red curve, indicated as Rh metal). (a) In 3d core levels; (b) P 2p,
In 4s and Al 2s core levels, the peak of POx is indicated; (c) Ti 2p core level; and (d) Rh 3d
core level. .............................................................................................................. 52
Figure 3.27: X-ray photoelectron spectra of a pristine Rh/TiO2/Tandem sample (black) and after
degradation in acidic environment (red): (a) In 3d core levels; (b) P 2p and In 4s core levels,
the peak of POx is indicated; (c) Ti 2p core level; and (d) Rh 3d core level. ........ 53
Figure 3.28: X-ray photoelectron spectra for the study of Rh catalyst poisoning by POx groups in
pH 7: (a) In 3d core levels; (b) P 2p core levels, the peak of POx is indicated; (c) Ti 2p core
level; and (d) Rh 3d core level. The black curve indicates the pristine (p) sample before
any photoelectrochemical measurement. The red curve indicates the aged (a) sample,
which was taken out from the electrolyte under light illumination after operation. The blue
curve indicates the recovered (r) sample, which was taken out from the electrolyte under
dark condition to simulate the diurnal cycle.......................................................... 54
Figure 4.1: Cell configuration composed of 1 NiOx or Pt anode, 2 Ag-NPs on Sigracet 29BC
carbon paper cathode, 3 anion exchange membrane, 4 CO2 gas inlet and CO/CO2 outlet, 5
Acrylic backplate, 6 catholyte chamber, 7 anolyte chamber, 8 reference electrode. Black
arrows indicate the gas flow, and white arrows indicate the electrolyte flow. Note that the
backplate, 5, is designed to use an interdigitated wire electrode flow field to enhance the
interaction between gas and catalysts and improve CO2 utilization (see also Figure 4.2).
............................................................................................................................... 60
Figure 4.2: Backplate as shown in Figure 4.1 item 5 with an interdigitated flow field. .. 60
Figure 4.3 (a) Illustration of the solar tracker. (b) With the addition of C = 3.25 Suns solar
concentrator. The PV element is located in the left with a silicon reference photodiode
mounted on the right. Above the PV element is the light-dependent resistor sensor array
for determining and tracking the position of the sun. For concentrator operation, a Fresnel
lens with 51 mm focal length was placed in front of the solar cell to provide a concentration
of 3.25x. ................................................................................................................. 62
Figure 4.4: Scanning electron microscopy images of carbon paper without (top) and with (bottom)
Ag-NP catalyst, secondary electrons image (left row) backscattered electrons image (right

xvi
row). (b)Illustration of the reverse-assembled GDE cathode cross-section with wetted
catalyst and operation for CO2 reduction. ............................................................. 63
Figure 4.5: Contact angle Q measurement of water on (a) pristine Sigracet 29 BC carbon paper
and (b) with Ag-NPs on Sigracet 29BC carbon paper after electrolysis. The contact angle
is 175° for (a) and 105° for (b). Optical micrographs of water pushing through the back of
the Sigracet 29 BC carbon paper (c) without Ag-NPs and (d) with Ag-NPs. The formation
of small liquid bubbles is observed in (c) while a thin water layer is shown in (d) indicating
the catalyst surface is wetted during operation as proposed.................................. 64
Figure 4.6: Illustration of the reverse-assembled GDE cathode cross-section with wetted catalyst
and operation for CO2 reduction............................................................................ 65
Figure 4.7: Dark catalysis three-electrode measurement of Ag-NPs GDE. Faradaic efficiency
versus GDE potential operated in 1 M KHCO3 (left half of graph) or 1 M KOH (right half
of graph) of (a) the reserve-assembled Ag-NP GDE and (b) a standard-assembled Ag-NP
GDE. ...................................................................................................................... 66
Figure 4.8: Overpotential versus CO partial current of Ag-NPs GDE for CO2 reduction to CO.
Overpotential=#$%&, ()& + 0.11., /01 ≡ /$%& × 45&, 01. ......................... 67
Figure 4.9: (a) GDE potential vs. NHE, (b) GDE potential vs. RHE versus CO partial current of
Ag-NP GDE (rev. indicating reserve-assembled, std. indicating standard-assembled) for
CO2 reduction to CO in 1 M KHCO3 and 1 M KOH. ........................................... 68
Figure 4.10: Stability of reserve-assembled and standard-assembled Ag-NPs GDE operated at 0.6 V vs. RHE in 1 M KOH. ................................................................................. 69
Figure 4.11: Silver 3d (Ag 3d) and carbon 1s (C 1s) X-ray photoelectron spectra of Ag-NP GDE
before/after electrocatalysis with an electrolyte of (a,c) 1 M KHCO3; (b,d) 1 M KOH. 69
Figure 4.12: J-V characteristic of the GaInP/GaInAs/Ge triple junction cell. Uoc is the open circuit
voltage, Jsc the short circuit current, Ump/Jmp the current and voltage at the maximum power
point, and FF the fill factor. ................................................................................... 70
Figure 4.13: Intensity (left axis) of AM 1.5G 1 sun reference spectrum (gold) and solar simulator
spectrum (black), external quantum efficiency (right axis) of the GaInP/GaInAs/Ge (blue,
green, red) triple junction cell. .............................................................................. 71
Figure 4.14: Light driven PV-GDE measurement (APV = AGDE = 0.31 cm2). (a) Illustration of wire
connection between the triple junction cell and GDE cell. (b) J-U characteristic of Ni anode,
solar cell with Ni anode, and Ag-NP gas diffusion cathode under 1 Sun. (c) Current, GDE
potential vs RHE, and cell voltage measurement over 20 h duration. (d) The corresponding
CO Faradaic efficiency and solar to fuel efficiency over the same 20 h duration. 73
Figure 4.15: Efficiency and stability assessment at a solar concentration 3.25 Suns. (C = 3.25,
AGDE = 1 cm2, APV = 0.31 cm2) (a) J-U characteristic of Ni anode, solar cell with Ni anode,
and Ag-NP gas diffusion cathode under 3.25 Suns. (b) Current and cell voltage
measurement over 150 h duration. (c) The corresponding CO Faradaic efficiency and solar
to fuel efficiency over the same 150 h duration. ................................................... 75

xvii
Figure 4.16: Outdoor assessments of solar driven PV-GDE in Pasadena, CA
(APV = AGDE = 0.31 cm2). The solar irradiance was monitored with a calibrated silicon
photodiode. Operation current density J (= JGDE = JPV), cell voltage Ucell, GDE potential
UGDE vs. RHE, CO Faradaic efficiency fFE,CO, and solar to fuel efficiency hSTF were
recorded for a 24h day cycle. ................................................................................ 77
Figure 4.17: Outdoor tests of solar-driven PV-GDE in Pasadena, CA. The solar irradiance was
monitored with a calibrated silicon photodiode. PV operation current JPV, cell voltage Ucell,
working electrode potential UGDE, CO Faradaic efficiency fFE,CO, and solar to fuel
efficiency hSTF were recorded for a 24h day cycle with 3.25x solar concentrator (C = 3.25,
AGDE = 1 cm2, APV = 0.31 cm2). ............................................................................ 78
Figure 4.18: J-U characteristic of the GaInP/GaInAs/Ge triple junction cell under 1 Sun (yellow
solid line) with combined load curve of Ni anode and Ag GDE cathode (red dot-dashed
line) in addition to DC-DC converter output curves (solid PV curve as input) with converter
efficiency of 90 % (black dashed line) and 95 % (black dotted line). The PV curves for
lower illumination conditions are included on the right figure. ........................... 79
Figure 5.1: Process flow diagram for fabrication of Si photoanodes with TiO2 nanocones and Ni
catalysts. ................................................................................................................ 86
Figure 5.2: Schematics of the three configurations that were simulated. (a) TiO2 nanocones on Si
substrate; (b) Ni between the TiO2 nanocones on Si substrate; (c) Ni hole arrays on Si
substrate. ................................................................................................................ 87
Figure 5.3: Simulated transmission (T), absorption (A), and reflection (R) spectra of the three
configurations of the TiO2 nanocone array and Ni layer in Figure 5.2. (a), (b), and (c) plot
the spectra in air for an array of TiO2 cones, a TiO2 cone array with Ni, and a Ni hole array,
respectively. (d), (e), and (f) plot the same structures but in water. The optical effects of
the Si substrate are not shown. .............................................................................. 88
Figure 5.4: Simulated transmission (T), absorption (A), and reflection (R) spectra of the TiO2 film
with thickness of 2.3 µm. (a) plot the spectra in air (b) plot the spectra in water. . 89
Figure 5.5: Area plot of simulated transmitted spectral photon flux in air and water for the three
structures in Figure 1. Blue represents the AM 1.5G spectral photon flux. Orange, yellow,
and purple depict the transmitted spectral photon flux into Si for: nanocones on Si,
nanocones with Ni on Si, and Ni hole array on Si, respectively. .......................... 90
Figure 5.6: Simulated electric field profiles along the cross section of a TiO2 nanocone on a Si
substrate. (a) Cross section and (b) scale for the relative electric field intensity for the
profile plots. (c-f) Profiles for wavelengths of 484 nm, 552 nm, 628 nm, and 770 nm,
respectively, which correspond to the maxima in the transmission spectra shown in Figure
5.3(b). (g-j) Profiles for wavelengths of 442 nm, 584 nm, 738 nm, and 940 nm, respectively,
which correspond to minima in the transmission spectrum in Figure 5.3(b). ....... 91
Figure 5.7: Transmitted light intensities |E|2 versus Y(µm) at interfaces of Si/Ni (indicated as on
Si) and Ni/air (indicated as on Ni). (a-d) Profiles for wavelengths of 484 nm, 552 nm, 628
nm, and 770 nm, respectively, which correspond to the maxima in the transmission spectra

xviii
shown in Figure 5.3(b). (e-h) Profiles for wavelengths of 442 nm, 584 nm, 738 nm,
and 940 nm, respectively, which correspond to minima in the transmission spectrum in
Figure 5.3(b). ......................................................................................................... 92
Figure 5.8: Dimension variation effect of Ni film on transmission to Si through TiO2 nanocone
waveguides in air. Hexagonal array (variant pitches) of 2.5 μm tall cones with a 200 nm
base radius and a 50 nm tip with a 200 nm radius Ni hole array. ......................... 93
Figure 5.9: Scanning-electron micrographs of dry-etched TiO2 nanocones on p+n-Si substrates
before (a,c) and after (b,d) electrodeposition of Ni. .............................................. 94
Figure 5.10: EDS mappings of elements Ti, O, Ni, Si (a) with top view (b) with 30o tilt view of
the Ni/TiO2 nanocones/p+n-Si sample. .................................................................. 95
Figure 5.11: SEM image of the Ni hole array fabricated via electron-beam patterning and dry
etching of a 50 nm thick Ni layer. The average diameter of the holes was ~ 500 nm. 95
Figure 5.12: Real component of the refractive index for (a) an ideal rutile TiO2 standard, and (b)
measured for a sample of electron-beam-evaporated amorphous TiO2. ............... 96
Figure 5.13: Transmission (T), absorption (A), and reflection (R) plots for Si with TiO2 nanocones
and 50 nm thick Ni calculated with evaporated TiO2 refractive index data are shown in (a).
(b) shows the area plot overlapped over the AM 1.5G spectrum for the three different cases,
as shown in Figure 5.2, using the refractive index data for amorphous TiO2 deposited by ebeam evaporation................................................................................................... 97
Figure 5.14: Reflection spectra (a) simulated and (b) measured for samples consisting of an array
of TiO2 nanocones on Si (blue), an array of TiO2 nanocones with Ni on Si (red), an array
of holes in a Ni layer on Si (yellow)...................................................................... 98
Figure 5.15: Current-density versus potential behavior, in the dark and under 100 mW cm-2 of
simulated AM1.5 solar illumination, respectively, for a p+n-Si sample covered by an array
of TiO2 nanocones and 300 mC cm-2 of electrodeposited Ni while in contact with 1.0 M
KOH(aq). The scan rate was 50 mV s-1. ................................................................ 99
Figure 6.1: Schematic of triangular metal gird geometries used in simulation. ............. 102
Figure 6.2: Illustration of the fabrication process for metal catalyst triangles. SEM image shows
an example of printed Ag metal triangle grid with 35% coverage. ..................... 104
Figure 6.3: Illustration of the metal deposition process on top of metal catalyst triangular grids.
............................................................................................................................. 104
Figure 6.4: Cell configuration composed of 1 NiOx anode, 2 ETC-PEC assembled cathode, 3 anion
exchange membrane, 4 quartz window, 5 reference electrode, 6 catholyte chamber, 7
anolyte chamber, 8 CO2 gas inlet, and 9 gas product/CO2 outlet. White arrows indicate the
gas flow. .............................................................................................................. 105
Figure 6.5: Schematic illustration of light management with metal triangle catalyst on top of a
semiconductor photoelectrochemical cell. .......................................................... 106
Figure 6.6: Simulated (a) absorbance spectra, (b) reflectance spectra, and (c) transmittance spectra
respectively of different Ag catalyst triangle coverages of 0%, 10%, 25%, 50%, and 85%
with w = 2.5 μm and h = 7 μm on GaAs substrate. ............................................. 107

xix
Figure 6.7: Simulated field profile of a Ag catalyst triangle with w = 2.5 μm, h = 7 μm, and
50% coverages on GaAs substrate at λ=500nm. ................................................. 109
Figure 6.8: EDS mappings of elements O, Si, Ag, Au, and Cu of the (a) Ag triangle (b) Au on Ag
triangle (c) Cu on Ag triangle on glass substrate. ............................................... 110
Figure 6.9: Experimental measurements of (a) absorption spectra, (b) reflection spectra, and (c)
transmission spectra respectively for different metal catalyst triangles with coverages of
35% on glass substrate. 0% coverage is shown as glass. Calculated transmission spectrum
for 200nm thick Ag film with 35% coverage is included for comparison. ......... 111
Figure 6.10: SEM images of (a) Ag triangle, (b) Au on Ag triangle, and (c) Cu on Ag triangle on
glass substrate with 35% coverage. ..................................................................... 112
Figure 6.11: Current density and product distribution at different potentials vs RHE for (a)(d) Ag
triangle, (b)(e) Au on Ag triangle, and (c)(f) Cu on Ag triangle on glass substrate with 35%
coverage. .............................................................................................................. 113
Figure 6.12: Current density and product distribution at different cell potentials with NiOx anode
for (a)(c) Ag triangle, and (b)(d) Cu on Ag triangle on glass substrate with 35% coverage.
............................................................................................................................. 114
Figure 6.13: External quantum efficiency of (a) Spectrolab GaInP/GaInAs/Ge triple junction cell,
and (b) ISE GaInP/GaAs/Si triple junction cell. ................................................. 115
Figure 6.14: Current density at different cell potentials of light absorber and cathode/anode
catalysts defining the operation point for (a) Spectrolab 3J + Ag triangle/NiOx, (b)
Spectrolab 3J + Cu on Ag triangle/NiOx, (c) ISE 3J + Ag triangle/NiOx, and (d) ISE 3J +
Cu on Ag triangle/NiOx. ...................................................................................... 117
Figure 6.15: (a) Current density and (b) product distribution at different potentials vs RHE for Ag
triangle, and integrated ETC-PEC device with Spectrolab 3J plus Ag triangle on glass
substrate with 35% coverage. .............................................................................. 118
Figure 6.16: (a) Current density, (b) solar to CO efficiency, and (c) product distribution at 0 V vs
NiOx counter electrode of the integrated ETC-PEC device with Spectrolab 3J + Ag triangle
on glass substrate with 35% coverage over 20 h stability test. ........................... 119
Figure 7.1: (a) J-U measurements for p+-Si/a-TiO2/Ni and p+-Si/a-TiO2/Ir sample in 1.0 M
KOH(aq). (b) J-U measurements for p+-Si/a-TiO2/M (M=nickel, iridium, gold) in
50/350 mM Fe(CN)63-/4-(aq) solution. .................................................................. 125
Figure 7.2: J-U measurements for (a) p+-Si/a-TiO2/Ni, (b) p+-Si/a-TiO2/Ir, and (c) p+-Si/a-TiO2/Au
in 50/350 mM Fe(CN)63-/4-(aq) solution with different thicknesses of the metal layer. 126
Figure 7.3: Scanning electron microscopy images of a-TiO2/M. M= (a) nickel, (b) iridium, and (c)
gold for different metal thicknesses. The scale bar is 200 nm. ........................... 127
Figure 7.4: XRD spectra for a-TiO2/M (M=Ni, Ir, Au) for different metal layer thicknesses. 128
Figure 7.5: XPS spectra for a-TiO2/metal systems for (a-d) nickel, (e-h) iridium and (i-l) gold
showing core level peaks for the metal, (a) Ni 2p, (e) Ir 4f, (i) Au 4f; (b, f, j) Ti 2p; (c, g,
k) O 1s, and (d, h, l) of the valence band. The metal overlayer thicknesses are shown in the
graph. The black dashed line shown in the Ti 2p core level plots indicates the position of

xx
bulk Ti 2p3/2 core level peaks, whereas the dashed and solid lines in (a) and (e) indicate
the metallic and oxide peak position, respectively. ............................................. 130
Figure 7.6: XPS spectra for TDMAT ALD a-TiO2 of the (a) Ti 2p and (b) O 1s core levels and (c)
of the valence band for different emission angles from θ = 0° (bulk sensitive) to θ = 70°
(surface sensitive) relative to the surface normal. With increased surface sensitivity
(increased θ), an increase in the oxygen shoulder at 532.5 eV was observed. .... 131
Figure 7.7: Valence band spectra of for different nickel thicknesses: (a) 0 nm, (b) 0.3 nm, (c)1.3
nm, and (d) 10 nm. The spectra were recorded at three different photon energies: 150 eV,
640 eV, and 1100 eV corresponding also to the kinetic energy of electron from the upper
valence band. Hence, the inelastic mean-free path (IMFP) of the photoelectrons
corresponds to l = 4.72 Å (Ni) to 6.28 Å (a-TiO2) for EK = 150 eV, l = 11.19 Å (Ni) to
14.96 Å (a-TiO2) for EK = 640 eV, and l = 16.64 Å (Ni) to 22.39 Å (a-TiO2) for
EK = 1100 eV. Inelastic Mean-Free Path for elements under investigation for relevant
photoelectron energies are calculated by IMFP-TPP2M. 180 ............................... 131
Figure 7.8: He I ultraviolet photoelectron spectra (UPS) of TDMAT a-TiO2/metal, with nickel,
iridium, and gold showing (a) the work function and (b) valence band maximum. The
metals were sputter-cleaned until no contamination or carbon was detectable. (c) Energy
position of Ti 2p3/2 core level depending on contact metal and metal thickness. The values
were extracted from Figure 7.5. .......................................................................... 132
Figure 7.9: Illustration of possible X-ray spectroscopy excitation and decay channels: I. Excitation
of core level electron above the vacuum level (EVac); II. Excitation of valence band (VB)
electron above the vacuum level; III. Resonant excitation of a core level electron into
unoccupied states of the conduction band (CB); IV. After process III, an electron from the
valence band refills the core hole transferring the energy to the initial excited core level
electron thus exciting it above EVac (participator decay); V. After process III, an electron
from the valence band refills the core hole with the emission of a photon. ........ 134
Figure 7.10: Resonant photoemission maps at the Ti L3,2 edge of (a) p+-Si/a-TiO2, (b)
p+-Si/a-TiO2/Ni(0.3 nm), and (c) p+-Si/a-TiO2/Ni(1.3 nm). The insets on the top show the
total electron yield (TEY) mode XAS spectrum, and the inset on the right the valence band
spectrum at 464 eV. In all spectra, the off-resonant contributions were subtracted using the
off-resonant VB spectra at 452 eV. The black arrows on the right panel indicate the position
of the gap state. .................................................................................................... 136
Figure 7.11: Resonant photoemission maps at the Oxygen K edge of (a) p+-Si/a-TiO2, (b)
p+-Si/a-TiO2/Ni(0.3 nm) and (c) p+-Si/a-TiO2/Ni(1.3 nm). The insets on the top show the
TEY XAS spectrum and the inset on the left the valence band spectrum at 531 eV. In all
spectra, the off-resonant contributions were subtracted using the off-resonant VB spectra
at 525 eV.............................................................................................................. 137
Figure 7.12: Resonant photoemission maps at the Ni L3 edge of (a) p+-Si/a-TiO2/Ni(0.3 nm) and
(b) p+-Si/a-TiO2/Ni(1.3 nm). The insets on the top show the TEY XAS spectrum, and the

xxi
insets on the left the valence band spectrum at 853 eV. In all spectra, the off-resonant
contributions were subtracted using the off-resonant VB spectra at 846 eV. ..... 138
Figure 7.13: (a) XPS valence band spectra and (b) RIXS spectra at the Ti L3 eg resonance for
pristine and annealed a-TiO2. .............................................................................. 139
Figure 7.14: RIXS spectra of (a) a-TiO2/Ir and (b) a-TiO2/Au at the Ti L3 t2g resonance. The
position of the characteristic dd transition is indicated. (c) Intensity ratio of dd transition to
elastic peak. A change in this normalized dd intensity gives evidence of changes in the gap
state. ..................................................................................................................... 140
Figure 7.15: Partial density of valence band states for titanium (red) and nickel (grey) derived
states at the a-TiO2-Ni interface for pristine (a) a-TiO2, (b) a-TiO2/Ni(0.3 nm), and (c)
a-TiO2/Ni(1.3 nm). The partial density of states (pDOS) is obtained by calculating the
difference between on and off-resonance valence band spectra a the Ti 2p and Ni 2p X-ray
absorption edge. ................................................................................................... 141
Figure 7.16: Band-energy diagrams for a-TiO2/M with M= (a) nickel, (b) iridium, and (c) gold.
All numeric values are in eV. d is the interface dipole energy difference between TDMAT
a-TiO2 and the metal. The hashed region between the VBM and CBM in the a-TiO2
indicates the position of the gap state with the FWHM taken as its width. The values can
also be found in Table 7.1. .................................................................................. 144
Figure 8.1: State-of-the-art solar fuel device for water splitting and CO2 reduction with category
of PV-Electrolyser, photoanode, photocathode, and PV plus photocathode. ...... 150

xxii

LIST OF TABLES
Table 2.1: Summary of array geometries used in simulation of Figure 2.2(c). UR indicates uniform
radii. ....................................................................................................................... 12
Table 2.2: Summary of array geometries used in simulation of Figure 2.3(e). UR indicates uniform
radii. T indicates tapered. MR indicates multi-radii. ............................................. 15
Table 2.3: Summary of array geometries used in simulation of Figure 2.4(b). T indicates tapered.
............................................................................................................................... 17
Table 2.4: Summary of array geometries used in simulation of Figure 2.5(c,d). MR indicates
multi-radii. ............................................................................................................. 19
Table 3.1: Approaches to theoretical limitation of light-induced photoelectrochemical water
splitting; ideal, only exchange current density limited and devices that are optically and
electrochemically limited are displayed, respectively. For the used band gap combination
and only catalytic exchange current density (JXC) limitation, htheo = 22.8 % at AM 1.5G
irradiation. ............................................................................................................. 46
Table 3.2: Reported STH benchmarks from literature with employed bandgaps, achieved STH
efficiency ( !STH ), theoretical limit for realistic water splitting ( !theo ) and ratio of
achieved STH to !theo as !theo ∗. ....................................................................... 48
Table 4.1: Comparison of the CO2 reduction performance of our Ag-NP catalyst with previously
reported Ag and Au electrodes. (U as overpotential vs. E0CO/CO2 (-0.11 V vs. RHE), mass
activity define as / ∙ 45&>?@ABCD) ........................................................................ 57
Table 4.2: Currents calculated for the individual sub-cells of the of the GaInP/GaInAs/Ge triple
junction PV cell under 1.5G 1 sun illumination assuming the standard reference sunlight
spectrum (AM1.5G ASTM G-173 reference spectrum was taken from the Renewable
Resource Data Center (RReDC) of the National Renewable Energy Laboratory (NREL))
or the solar simulator spectrum and measured short circuit photocurrent Jsc under respective
1 sun conditions. .................................................................................................... 71
Table 4.3: Comparison of the performance of the PV-GDE studied herein with the current state
of the art PV-electrolyzer for CO2 reduction to CO.80 .......................................... 75
Table 4.4: Comparison of the PV-GDE performance studied herein with different measurement
conditions and calculations.................................................................................... 80
Table 6.1: Change of absorbance, reflectance, and transmittance of of different Ag catalyst
triangle coverages of 10%, 25%, 50%, and 85% with w = 2.5 μm and h = 7 μm on GaAs
substrate related to 0% metal coverage. .............................................................. 108
Table 6.2: Curren densities calculated for the individual sub-cells of the Spectrolab
GaInP/GaInAs/Ge triple junction cell and ISE GaInP/GaAs/Si triple junction cell in
conjunction with optical response of different catalysts under AM1.5G 1 sun illumination
assuming the standard reference sunlight spectrum (AM1.5G ASTM G-173). The short

xxiii
circuit photocurrent Jsc is obtained from the minimum current density of the individual
sub-cells. .............................................................................................................. 116
Table 7.1: Parameters used for band-energy diagrams of a-TiO2/M with M=nickel, M=iridium
and M=gold as shown in Figure 7.16. EB is the binding energy. EBB is the band bending at
the interface. The band gap for TiO2 was taken from previous studies. 74,166 ..... 143

CHAPTER 1
Introduction of Solar Fuels
1.1 Fundamentals of Photoelectrochemistry
To achieve a sustainable future with a carbon neutral environment, the world’s reliance on
renewable energy has dramatically increased. Solar photovoltaic and wind energy conversion are
rapidly growing sources of low-carbon electric power due to the size of the resource and its wide
geographical deployment potential.1 However, intermittency of the solar and wind resources over
wide time scales ranging from minutes to months means solar electricity is not a dispatchable
power source. Thus, efficient and inexpensive approaches for energy storage are needed for wide
penetration of renewable energy into the power grid.2 While electrical energy storage in batteries
may be important for short-term storage and grid power management, seasonal energy storage is
unlikely to rely on batteries. Transformation of solar energy into chemical bonds provides a longterm energy storage strategy that opens a path for the synthesis of fuels and chemicals.3 One
approach to chemical energy storage is via solar-driven fuels generation, where i) photovoltaics
supply carbon free electricity to the grid that is used to generate fuels by chemical reaction at high
current densities;56 ii) photovoltaics are used to directly drive electrolysis at low current densities,4
or iii) an integrated photoelectrochemical device that performs unassisted direct fuels
production.5,6
Photoelectrochemical (PEC) devices integrate multiple functional materials and couple
various PEC processes to produce fuels from sunlight and water. Figure 1.1 illustrates key
photoelectrochemical processes in a typical device. First, the incident sunlight is absorbed by the
semiconductor materials. Any materials or components in the optical pathway between the sun
and the semiconductors could potentially modulate and alter the light absorption. Therefore,
understanding light-matter interaction in the PEC system becomes an important objective. Second,
the absorbed photons in the semiconductor material generate energetic electrons and holes, which
then transport to electrocatalysts via interfacial charge transfer. Third, the electrocatalysts perform
the chemical reaction, in which products are produced simultaneously at the catalytic sites. For
example, the two half reactions involved in the water splitting process are shown below:

Half reaction at cathode (reduction): 2H F + 2eG → HI (E = 0 V vs RHE)

Half reaction at anode (oxidation): HI O + 2hF → I OI + 2H F (E = 1.23 V vs RHE)

Net Reaction: HI O → HI + I OI
Where RHE is the reversible hydrogen electrode potential. Other cathode reactions include but
not limit to CO2 reduction reaction (CO2RR):
COI + 2H F + 2eG → CO + HI O (E = -0.11 V vs RHE)
COI + 2H F + 2eG → HCOOH (E = -0.17 V vs RHE)
COI + 8H F + 8eG → CHN + 2HI O (E = 0.17 V vs RHE)
2COI + 8H F + 8eG → CHO COOH + 2HI O (E = 0.13 V vs RHE)
2COI + 12H F + 12eG → CI HN + 4HI O (E = 0.08 V vs RHE)
2COI + 12H F + 12eG → CI HQ OH + 3HI O (E = 0.08 V vs RHE)
3COI + 18H F + 18eG → CO HS O + 5HI O (E = 0.1 V vs RHE)

Figure 1.1: Schematic illustrations of the general operating principles for PEC device.

In the meantime, ionic transport between the cathode and anode chamber and product
separation are required to maintain the efficient and safe operation of the cell. Note that all these

processes need to be fully coupled to produce a single rate of reaction for cathode and anode
reaction in the PEC device. To overcome the thermodynamic potential (ΔUrxn) , the difference
between the anode oxygen evolution (1.23 V vs. RHE) and cathode reduction reaction, the total
voltage (Fermi level splitting) of cathode and anode needs to be large enough to sustain the full
reaction. For example, ΔUrxn = 1.23V for H2/O2, 1.34V for CO/O2. As illustrated in Figure 1.1, by
employing semiconductor/electrolyte junctions, the conduction band edge (ECB) of the cathode
should be positioned at a higher energy level than the cathode reaction, while the valence band
edge (EVB) of the anode should be positioned at lower energy level than the anode reaction. 7,8 In
addition, solid-state, buried junctions using traditional photovoltaic materials, such as Si, GaAs,
etc., are often used to circumvent the stringent requirements for the band edge positions and to
achieve high efficiency solar to fuel performances. 5,9,10

1.2 Photoelectrochemical Device
Four general types of PEC water-splitting device architectures, as shown in Figure 1.2, have
been modeled and experimentally demonstrated in the laboratory scale. 11 Type 1 and Type 2
device indicates a system where the catalyst on a light absorber is in the form of particles
suspended in the electrolyte. Either a single-chamber device (Type 1), in which hydrogen and
oxygen co-evolve where product separation would be necessary afterward, or dual-chamber device
(Type 2), in which a bridge or membrane for ion transport is required in the Z scheme reaction,
has been proposed and studied for particle-based systems. While the Type 1 and Type 2 device
architecture shows great promise in many technoeconomic analyses (TEA), the solar-to-fuel (STF)
conversion efficiency is currently limited to <2%. 12,13 Type 3 and Type 4 devices indicate systems
where catalyst coated planar semiconductor materials and membrane separators are configured to
maximize the light absorption and to minimize the transport loss in the device. Even though all
architectures can operate under un-concentrated sunlight or sunlight with low concentrations, e.g.,
~10x concentration, Type 4 indicates a PEC couple with a solar concentrator specifically.

Type 1

Type 4
Type 3

concentrator

H2/O2
H2O
H2O
H2

Type 2

O2
H2O

membrane

H2O

H2

O2

H2O
H2
O2
membrane

membrane

H2O

H2O
bridge

Figure 1.2: Schematic illustrations of device architectures for Type 1 and Type 2 particle based
PEC device, Tpye 3 and Type 4 planar catalyst coated semiconductor PEC device.

1.3 Solar-to-Fuel Efficiency
Among various performance metrics, the solar-to-fuel (STF) conversion efficiency is one of
the most important parameters in determining the levelized fuel production cost. 11 In particular,
high STF conversion efficiency levitates the land requirements for a given capacity of fuel
production and lowers the balance of system cost. We calculated the solar to fuel efficiency (hSTF)
using equation below.

∙abcde ∙fgh

!UVW = YZ[\ = Z`i
]^

jklmn ∙opq

∙abxy^ ∙fz{ ∙ost\tuvw\

= st\tuvw\ i

u]|}\ ∙opq

∙ab

∙f

= pq i cde gh
jklmn

(Equation 1.1)

Where Pin is input power, Pout is output power, Iop = Jcatalyst×Acatalyst = JPV×APV , Iop is the operation
current of the system, Jcatalyst and Acatalyst are the current density and area of the catalyst, JPV and
APV are the current density and area of PV, ΔUrxn is the thermodynamic potential difference
between the oxygen evolution half reaction (OER) and the cathode reduction reaction, fFE is the
reaction Faradaic efficiency, and Plight is the incident light irradiance (mW×cm-2) on the
photovoltaic.
The energy efficiency for the cell (hcell) was defined as follows:

!~ÄÄ = YÅÇn =

abxy^ ∙rst\tuvw\ ∙ost\tuvw\ ∙fgh
bÉÑjj ∙rpq ∙opq

ke

abxy^ ∙fgh
bÉÑjj

(Equation 1.2)

Where Ucell is the total operating voltage of the cell.
For the integrated photoelectrochemical device, the catalyst and light absorber are coupled
together. The solar fuel generator efficiency hSTH is given by:

∙ab

∙f

!UVÖ = Ühá i cde gh
jklmn

(Equation 1.3)

Where JPEC is the PEC operating current density at 0 V vs counter electrode potential.

1.4 Thesis Outline
This thesis provides the research pathways to achieve a high efficiency solar fuel device from
a better understanding of light matter interaction to carrier transport dynamic through device level
integration. Below is the brief overview of each chapter in this thesis:
Chapter 2 focused on enhancing absorption via nanophotonic design of III-V based light absorber.
Both multi-radii and tapered nanowire arrays are introduced and experimentally evaluated.
Chapter 3 realized high solar-to-H2 efficiency in PEC devices, consisting of a III-V based tandem
light absorber and RuOx/Rh NP catalysts for OER and HER. Minimizing parasitic light absorption
and reflection losses with favorable band alignment further reduces the efficiency gap to the
theoretical limit.
Chapter 4 developed a solar-driven CO2 reduction device using a gas diffusion electrode (GDE)
with Ag nanoparticle catalyst directly powered by a III-V based triple junction solar cell. Device
geometry was studied to extend the operation stability.
Chapter 5 demonstrated light management strategies to create highly active and effectively
transparent catalyst structures with high index TiO2 nanocones. It allowed incident broadband
illumination couples to multiple waveguide modes, reducing interactions of the light with the metal
catalysts.

Chapter 6 introduced another approach with an effectively transparent catalyst consisting of
arrays of micron-scale triangular cross-sectional metal grid fingers. It redirected the incoming light
to the open areas of the PEC cell to reduce the overall shadow loss.
Chapter 7 investigated the interfacial conduction mechanism between the TiO2 protection layer
and metal catalysts with a combinatorial approach of electrochemistry, XPS, resPEX, and RiXS.
Chapter 8 summarized the main outcome and contribution of this thesis.

CHAPTER 2
Broadband Adsorption InP NW
2.1 Introduction
Design of “perfect” absorbers and emitters is of considerable current interest and research
in the nanophotonics and metamaterials fields. 14-16 Perfect absorbers and emitters can find
applications in numerous fields across the electromagnetic spectrum, including light and thermal
sources, 17-19 sensing, 20,21 and energy conversion. 18,22,23 Two types of near-unity or “perfect”
absorption are straightforward to achieve: (1) a wavelength-sized resonator can be used for
selective “perfect” absorption — absorption at a single frequency, polarization and incidence angle,
24-27

and (2) an optically thick layer of lossy material can be used for unselective, “perfect”

absorption — absorption over a large range of frequencies, angles, and polarizations. 28 However,
many applications would benefit from a more comprehensive ability to tailor perfect absorber
characteristics, such as achieving directional, spectrally broadband thermal emission of infrared
radiation sources 17,29 and broadband, angle-insensitive thin film perfect absorbers for high
efficiency, lightweight photovoltaics. 28,30,31 To this end, recent work in the field has focused on
the realization of selective perfect absorbers that are extremely thin, 14,26,32 actively tunable, 33,34
and wavelength, angle, or polarization-insensitive, 16,35,36 as well as unselective perfect absorbers
with small form factors that are insensitive to angle, wavelength, or polarization. 22,23
In this work, we focus on the design and fabrication of unselective perfect absorbers with
small form factors using semiconductor nanowire arrays. Specifically, we examine sparse arrays
of InP nanowires fabricated using a top-down lithographic pattern and etch procedures, followed
by embedding in polydimethylsiloxane (PDMS) and mechanical removal from the substrate. These
sparse arrays of vertically-oriented, semiconductor nanowires represent a promising approach to
flexible, lightweight, high efficiency, and low-cost optoelectronic devices in sensing and energy
applications, such as photodetectors,

37-39

bolometers,

40,41

solar cells,

23,42-44

and

photoelectrochemical devices. 45,46 Currently in the photovoltaics field, the greatest challenge for
use of III-V absorbers and cells is reducing cost/Watt. 47 Therefore, it is highly desirable to develop
fabrication methods that reduce or eliminate costs associated with epitaxial growth and
consumption of III-V compound semiconductor substrates. The wire-array fabrication process

described herein holds considerable promise for achieving these goals because many layers can
be fabricated from a single compound semiconductor substrate. InP is of particular interest for
wire array photovoltaics, and more broadly, InP has garnered much interest for nanowire-based
optoelectronic devices due to its direct bandgap and low intrinsic surface recombination velocity,
48

which is critical to high performance in high surface area devices.
In recent years, the optical properties of semiconductor nanowire arrays have been the subject

of great interest and intensive research. 49-52 Even at very low area fill fractions, semiconductor
nanowire arrays exhibit strong optical absorption due to robust coupling into the waveguide modes
of individual nanowires. 23,53 These arrays of essentially independent optical antennas have an
optical response that is polarization-independent and angle-insensitive. 45 In a standard nanowire
array with a uniform wire radius, strong absorption is observed over a relatively narrow spectral
region in which end-mediated coupling into guided modes is favorable. 45,54 Array geometry, 55-57
nanowire shape, 23,58-60 and order61,62 have previously been shown both experimentally and
theoretically to control the spectral position of this region by others. In previous work, 23 we
theoretically studied different nanowire motifs within sparse arrays and, by optical design of the
waveguide modes, achieved broadband absorption enhancements for constant material usage.
Specifically, we optimized arrays with multiple nanowire radii and tapered nanowires and
predicted > 90 % broadband absorption (Vis-NIR) in 150 nm planar equivalence. Herein, we
validate those predictions and experimentally demonstrate broadband, polarization-independent,
angle-insensitive, near-unity (> 90 %) absorption in sparse arrays of InP nanowires.

2.2 Experimental and Numerical Method
InP nanowire arrays were fabricated top-down from InP wafers (2’’ AXT, n-type, sulfurdoped, (100) orientation), using inductively coupled plasma reactive ion etching (ICP-RIE).
Wafers were cleaned via sequential sonication for 10 min in water, IPA, acetone, and water. A
piranha clean step was omitted because of the existence of the native InPxOy facilitated adhesion
of the hard mask layers.
Two different hard masks were used in this work: SiO2 and Cr. The SiO2 mask was
approximately 400 nm thick and deposited via RF sputtering (SiO2 target, 3 mTorr Ar atmosphere,

200 W forward power). The pattern was defined in the mask via a direct electron beam
lithography written into negative tone, MaN-2403 resist, followed by a pseudo-Bosch etching step
(5 mTorr, 1000 W ICP forward power, 25 W RF forward power, 10°C, 26 sccm SF6, 35 sccm C4F8)
to transfer the pattern from the resist into the SiO2. For the Cr hard mask, the pattern was defined
via direct electron beam lithography written into a bilayer of positive tone resist (PMMA495A4/PMMA95-A4), followed by electron beam deposition of a 50 nm Cr layer (0.5 A∙s-1,10-6 Torr)
and lift-off of the resist.
A Cl2/H2/CH4 etch (4 mTorr, 2200 W ICP forward power, 200 W RF forward power, 28 sccm
H2, 32 sccm Cl2) was used to transfer the pattern into the InP for the creation of nanowire arrays.
The table temperature was set to 60°C and no thermal contact between the sample and the carrier
wafer was omitted to achieve etch temperatures of ~300°C to ensure sufficient volatility of the In
etch products. Because CH4 provides sidewall passivation, its flow was varied between 24 and 30
sccm to control the degree of sidewall taper from normal taper to inverse taper, respectively.
To exfoliate the nanowire arrays from the InP substrate, the samples were covered with a
thick layer of 10:1 PDMS solution, degassed for 20 min, baked overnight at 80°C, and
mechanically peeled-off the substrate after cooling. All SEM images were taken on wafer, prior to
embedding in PDMS. All experimental optical spectra were obtained using a Fianium laser as a
tunable source and a home-built integrating sphere setup with silicon photodiode detectors. 63
Reflection and transmission spectra (when relevant) are composed of 276 points, spaced linearly
in wavelength from 450 to 1000 nm. Absorption was calculated from 1-R-T.
Full-field, 3D simulations were performed using Lumerical FDTD, a commercial
electromagnetics software package. At normal incidence, the infinite periodicity of the nanowire
arrays was rendered using periodic boundary conditions and, at non-normal incidence, Bloch
boundary conditions to enforce phase continuity. Symmetric boundary conditions were applied
when relevant to reduce computation time. In the axial direction, perfectly matched layers were
used to emulate infinite space above and below the array. All nanowire structures were modeled
using Palik data for InP, Ag, Cr, and SiO2, and ellipsometric data for PDMS. Image analysis of
SEM images was used to determine the nanowire dimensions for simulation. The top-down
fabrication process resulted in significant geometric variation within a single array; therefore, in

10
order to capture the optical behavior of the arrays, multiple array geometries were simulated.
Three representative geometries were chosen for each sample, and results contain each of these
three spectra as well as their average. Figure 2.1 displays the two geometries that were used to
render the fabricated nanowires in simulation.

Figure 2.1: Schematic of two different nanowire geometries used in simulation.

The remaining SiO2 and Cr masks were modeled as truncated cones with the bottom radius
matching the top radius of the nanowire and a top radius of 10 nm. A finer mesh was used around
the nanowire, with a mesh cell width of one tenth the smallest radius.
All simulations used a single wavelength, polarized infinite plane wave source, and each
spectrum is composed of 276n simulations, spaced linearly in wavelength from 450 to 1000 nm,
where n is the number of polarizations necessary to capture the unpolarized optical response of the
array. Planar power monitors were used to extract the array absorption and also to distinguish
between absorption in the InP and the other materials.

2.3 Results and Discussion
In this section, we begin with a brief discussion of InP nanowire array design. Next, we
validate our experimental and theoretical methods via a uniform InP nanowire array. Subsequently,
we prove that nanowire taper induces peak broadening, and multiple wire radii generate multiple

11
peaks via characterization of very sparse InP nanowire arrays. Finally, we demonstrate nearunity broadband absorption in sparse arrays of tapered and multi-radii InP nanowire arrays.
As previously stated, vertically-oriented semiconductor nanowires act as cylindrical dielectric
waveguides with high absorption loss. In sparse arrays, the nanowires are essentially noninteracting and, thus, their optical behavior is well-described by traditional waveguide theory. 64
The nanowires exhibit a primary absorption peak, which is due to end-mediated coupling into the
HE11 waveguide mode; the spectral region of strong coupling occurs in the moderately guided
portion of the HE11 modal dispersion. 53 The dominant role of the waveguide modes in the
absorption enhancement translates to a strong correlation between nanowire radius and the spectral
region of absorption enhancement, and is used herein to guide the design of InP nanowire arrays.
A radius range of 40 to 100 nm is needed to observe absorption enhancement in the visible up to
the band edge of InP (450-925 nm).
Initially, we examine a uniform nanowire array (r = 90 nm, h = 1.2 µm, a = 750 nm) to
validate our theoretical framework as well as our experimental and simulation methods. The three
different geometries used are summarized in Table 2.1. The variable labels correspond to Figure
2.1. The fill fractions, ff, and effective InP thicknesses, teff, are also calculated and included in the
table. This array, shown as the inset to Figure 2.2(d), has a fill fraction of 4.5% and a planar
equivalent thickness of ~54 nm. Figure 2.2(a) displays the absorption (blue), reflection (green),
and transmission (red) spectra for the array, after it was embedded in PDMS and peeled off of the
substrate; Figure 2.2(b) displays reflection spectra for the array in three different configurations –
on substrate, embedded in PDMS and on the substrate, and embedded in PDMS and peeled off of
the substrate. These experimental spectra give the expected results – we observe low reflectivity
for the array in all cases and a primary absorption peak around 850 nm, corresponding to coupling
into the HE11 waveguide mode. Figure 2.2(c-d) displays the corresponding simulated spectra, with
thick solid lines representing the average of three slightly different nanowire dimensions (thin,
dashed, dotted lines). We found good agreement between experiment and simulation.

12
Table 2.1: Summary of array geometries used in simulation of Figure 2.2(c). UR indicates
uniform radii.
rb
(nm)

rm
(nm)

rt
(nm)

UR-1

90

87.5

90

UR-2

90

90

95

UR-3

85

85

90

(µm)

1.578

hb
(µm)

1.214

ht
(µm)

0.122

(µm)

0.75

ff

teff
(nm)

0.044

69

0.046

72

0.041

65

Figure 2.2: Characterization of uniform radius array. (a) Absorption, reflection, and transmission
spectra for the PDMS-embedded array shown as inset of (d), measured using an integrating sphere;
(b) Reflection spectra for the array shown as inset of (d) in multiple configurations; (c,d) Simulated
spectra corresponding to the experimental results in (a,b) with solid thick lines representing an

13
average of three slight geometric variations (thin, dashed, dotted); inset of (d) is scanning
electron micrograph of the uniform radius array; see Table 2.1 for array dimensions used in
simulation.

The initial broadband absorber arrays presented herein were designed with the intention of
demonstrating that tapered nanowires broaden the primary HE11 absorption peak and that multiple
radii in a single array generate multiple absorption peaks. Therefore, sparse arrays of short
nanowires were fabricated, and the multi-radii sub-cell consisted of only two different radii. The
sparseness of these arrays results in more well-defined peak features, which is particularly essential
to the distinction of multiple peaks in the multi-radii arrays that might otherwise blend into one
extended peak due to experimental variation in the fabricated dimensions.
Figure 2.3(a-c) displays SEM images of the sparse (a = 750nm), uniform, tapered, and multiradii arrays, which have fill fractions (planar equivalent thicknesses) of 4.5% (54 nm), 2.7% (29
nm), and 2.1% (33 nm), respectively. The fill fractions and planar equivalent thicknesses were
determined from image analysis, and the precise dimensions used in simulation are listed in Table
2.2. Figure 2.3(d-e) displays the experimental and simulated absorption spectra for these arrays,
demonstrating good qualitative agreement between simulation and experiment. The absorption
spectrum of the tapered nanowire array (green) exhibits a broadened absorption peak in
comparison to that of the uniform array (blue) and the absorption spectrum of the multi-radii
nanowire array (red) exhibits two peaks, corresponding to the HE11 waveguide modes of the two
different nanowire radii. It is important to note that while the uniform array may appear to be
outperforming the tapered and multi-radii array designs in terms of integrated absorption, it is only
because the uniform array contains nearly twice the amount of InP.

14

Figure 2.3: Comparison of uniform, tapered, and multi-radii arrays. (a,b,c) Scanning electron
micrographs of the uniform, tapered, and multi-radii array, respectively; (d) Absorption spectra
for the PDMS-embedded arrays shown in (a,b,c), measured using an integrating sphere; (e)
Simulated absorption spectra corresponding to the experimental results in (d) with solid thick lines
representing an average of three slight geometric variations (thin, dashed, dotted); see Table 2.2
for array dimensions used in simulation. Colors are coordinated throughout the figure.

15
Table 2.2: Summary of array geometries used in simulation of Figure 2.3(e). UR indicates
uniform radii. T indicates tapered. MR indicates multi-radii.

rb
(nm)

rm
(nm)

rt
(nm)

UR-1

90

87.5

90

UR-2

90

90

95

UR-3

85

85

T-1

57.5

T-2

60

T-3

62.5

MR-1

(µm)

hb
(µm)

ht
(µm)

(µm)

ff

teff
(nm)

0.044

69

0.046

72

90

0.041

65

75

0.025

27

0.027

29

0.029

31

0.024

38

0.022

35

0.024

38

77.5

1.578

1.08

1.214

0.122

0.12

0.75

0.75

80

{72.5,37.5} {82.5,50} {92.5,57.5}

MR-2

{75,30}

{80,45)

MR-3

{75,42.5}

{80,50}

{92.5,57.5} 1.587
{90,60}

{1.247,
0.113
1.304}

0.75

To achieve near-unity broadband absorption, slightly denser arrays of taller nanowires were
fabricated for both the tapered nanowire and multi-radii nanowire motifs. Additionally, the multiradii array consisted of a 4 wire sub-cell (2x2) to distribute the strong in-coupling region of the
HE11 mode more evenly across the visible spectrum.
Figure 2.4 contains the results for an exemplary tapered nanowire array, with ~1.6 µm tall
nanowires with radii ranging from 30 to 110 nm, spaced 450 nm apart (SEM image in the inset of
Figure 2.4(a)). A detailed accounting of array dimensions are listed in Table 2.3. The
experimentally-measured and simulated absorption spectra for the PDMS-embedded, tapered

16
nanowire array are displayed in Figure 2.4(a-b), respectively, as a function of incidence angle.
The experimental data includes incidence angles from 0° to 30°, in 5° increments. The simulated
data includes 0°, 10°, and 20° for three slight geometric variations (thin, dashed, dotted lines) and
their averages (thick solid lines); the absorption of the planar equivalence (108 nm) is also overlaid
for reference (black, dashed line). In all spectra, we find good qualitative agreement between
simulation and experiment.

Figure 2.4: Characterization of tapered array. (a) Absorption spectra at various incident angles
for the PDMS-embedded array (SEM image shown as inset), measured using an integrating sphere;
(b) Simulated spectra corresponding to the experimental results in (a) with solid thick lines
representing an average of three slight geometric variations (thin, dashed, dotted), overlaid with
the planar equivalent absorption spectra (108 nm thin film, black dashed); see Table 2.3 for array
dimensions used in simulation.

17
Table 2.3: Summary of array geometries used in simulation of Figure 2.4(b). T indicates
tapered.
rb
(nm)

rm
(nm)

rt
(nm)

(µm)

hb
(µm)

T-4

100

50

32.5

1.55

0.775

T-5

107.5

52.5

35

1.6

0.8

T-6

112.5

57.5

37.5

1.65

0.825

ht
(µm)

0.05

(µm)

0.45

ff

teff
(nm)

0.057

91

0.067

107

0.078

125

The tapered array demonstrates angle-insensitive broadband absorption, approaching 90%
absorption across the visible. The PDMS front surface results in approximately a 5% broadband
reflectivity, and the remainder is transmission losses due to incomplete absorption. Compared to
the planar equivalent thin film of 108 nm (black, dashed line in Figure 2.4(b)), the nanowire array
exhibits broadband absorption enhancements across the visible spectrum. These absorption
enhancements occur due to strong coupling into the HE11 waveguide mode, which has been
engineered to occur over a broad spectral region via wire taper. Absorption drops off beyond 800
nm because there is insufficient wire length at the larger wire radii to observe significant absorption
enhancements. Additionally, the noise in both the experimental and simulated spectra is attributed
to residual pieces of the Cr etch mask.
Figure 2.5 displays the results for an exemplary multi-radii nanowire array, with ~ 1.75 µm
tall nanowires with radii ranging from 35 to 115 nm, spaced 520 nm apart (shown as the inset of
Figure 2.5(a)). A detailed account of array dimensions can be found in Table 2.4. The
experimentally-measured and simulated absorption spectra for the PDMS-embedded, multi-radii
nanowire array are displayed in Figure 2.5(a,c), respectively, as a function of incidence angle. To
further push the array absorption towards near-unity, a silver back reflector was deposited to
achieve 2-pass absorption, and the resulting experimental and simulated absorption spectra are
shown in Figure 2.5(b,d). The experimental data includes incidence angles from 0° to 30°, in 5°
increments. The simulated data includes 0°, 10°, and 20° for three slight geometric variations (thin,

18
dashed, dotted lines) and their averages (thick solid lines); the absorption of the planar
equivalence (109 nm for single pass, 218 nm for 2-pass) is also overlaid for reference (black,
dashed line). In all spectra, we find good qualitative agreement between simulation and experiment.

Figure 2.5: Characterization of multi-radii array. (a,b) Absorption spectra at various incident
angles for the PDMS-embedded array, shown as inset of (b), without and with a silver back
reflector, respectively, measured using an integrating sphere; (c,d) Simulated spectra
corresponding to the experimental results in (a,b) with solid thick lines representing an average
of three slight geometric variations (thin, dashed, dotted), overlaid with the planar equivalent
absorption spectra (109 and 218 nm thin films, black dashed); see Table 2.4 for array dimensions
used in simulation.

19
Table 2.4: Summary of array geometries used in simulation of Figure 2.5(c,d). MR indicates
multi-radii.
rb
(nm)
MR-4

{70,80,
87.5,110}

MR-5

{70,82.5,
92.5,115}

MR-6

{75,82.5,
97.5,115}

rm
(nm)

rt
(nm)

(µm)

{35,42.5,
55,67.5}

1.7

{37.5,45,
60,72.5}

1.75

{42.5,50,
62.5,75}

1.8

hb
(µm)

ht
(µm)

0.05

(µm)

0.52

ff

teff
(nm)

0.0
58

98

0.0
64

111

0.0
68

122

The multi-radii array demonstrates angle-insensitive broadband absorption, approaching 90
% absorption in a single pass and exceeding 90% with a silver back reflector up to the band edge
of InP (λ ~ 925 nm). Again, the PDMS front surface results in approximately a 5 % broadband
reflectivity, which accounts for nearly all of the loss when the array has a back reflector. In the
single pass case, the remainder is transmission losses due to incomplete absorption. Compared to
the planar equivalent thin films (109 nm for single pass, 218 nm for double pass; black, dashed
lines in Figure 2.5(b,d)), the nanowire array exhibits broadband absorption enhancements across
the visible spectrum. These absorption enhancements occur due to strong coupling into the HE11
waveguide mode, which has been engineered to occur over a broad spectral region via multiple
wire radii that span the appropriate range.
In the single pass case, absorption is slightly lower in the blue, exhibiting the inverse problem
to the tapered array (absorption dropped off in the red); this discrepancy occurs for two reasons:
(1) the radii range of the multi-radii array is slightly larger than the tapered array, resulting in a
red-shift of the absorption, and (2) the largest wire radius of the sub-cell targets the red region of
the spectrum, which is more effective than the very edge of the wire taper. In the double pass case,
the absorption is nearly flat at > 90% absorption, up to the band edge of InP, but does not drop to
zero beyond the band edge.

20
Figure 2.6 displays simulated absorption spectra for each material in the median tapered
array. Note that a broadband source was used for this simulation, which requires a polynomial fit
to the refractive index. InP absorption beyond the band edge is a result of fitting errors. Up to the
band edge of InP (λ ~ 925nm), the majority of the absorption occurs in the InP nanowires, and
absorption contributions from the silver and chromium are negligible. Beyond the band edge,
significant absorption occurs in the silver layer. At the silver interface, the nanowire radius is at a
maximum, and because the radius was designed to achieve field enhancements up to the band edge,
there are still significant field enhancements at slightly longer wavelengths, which, in the absence
of InP absorption, enhance absorption in the silver in this region. Conversely, the Cr mask absorbs
minimally because it is adjacent to regions of small nanowire radius, corresponding to field
enhancements at shorter wavelengths where InP is strongly absorbing. In multi-radii arrays, both
the Cr and the Ag absorb significantly beyond the band edge because the largest radius nanowire
enhances the field near both the Cr and the Ag.

Figure 2.6: Simulated absorption vs. wavelength of the median tapered array with a back reflector,
separated by material.

21
2.5 Conclusion and Outlook
In this work, we experimentally demonstrated near-unity, unselective absorption —
absorption that is broadband, polarization-independent, and angle-insensitive — in sparse arrays
of InP nanowire arrays, enabled by nanowire motif design. We explored two motifs — tapered
nanowires and arrays of nanowires with varying radii — that aim to enhance absorption across a
wide spectral range by optical design of the HE11 waveguide mode dispersion. Initially, we
demonstrated that wire taper broadens the HE11 mode absorption peak and that incorporation of
multiple wire radii results in multiple HE11 mode absorption peaks. Subsequently, we designed
and fabricated sparse InP nanowire arrays using both wire taper and multiple wire radii that
achieved near-unity, unselective absorption. Specifically, we achieved greater than 90%
absorption up to the band edge of InP in a multi-radii nanowire array with a back reflector that
contained approximately 100 nm planar equivalence of InP.
The cost of light absorbing material for high efficiency devices still creates big issues for real
world application. Future direction involves developing large area nanowire arrays fabrication
through a scalable, epitaxy-free fabrication method, for example, using nanoimprint lithography
and wet etching. Polymer-embedded wires are removed from the bulk InP substrate by a
mechanical method that facilitates extensive reuse of a single bulk InP wafer. The repeatable
process of imprinting, etching, and peeling to obtain many nanowire arrays from one single wafer
represents an economical manufacturing route for high efficiency III-V solar device. These
semiconductor-based, spectrally, and angularly-unselective absorbers have great potential for
flexible, high efficiency, and low-cost optoelectronic devices in energy and sensing applications.

22

CHAPTER 3
High Efficiency Solar to H2 PEC Device
3.1 Introduction
Electrochemical water splitting was achieved by van Trostwijk and Deiman in 1789 and,
about a decade later, by Nicholsen and Carlisle 65, whereas light-induced unassisted water splitting
with rutile as a photoanode was reported in 1972, resulting in a small but measurable efficiency 54.
Efficient solar water splitting was first achieved using a dual junction tandem photoelectrode 66
under a light intensity equivalent to 11 suns.
In 2015, several devices with solar-to-hydrogen efficiency (STH) greater than 10 % at 1 sun
illumination were reported 10, and in 2017 an efficiency of 16.2 % was achieved 5. Overall,
advances in solar water splitting 65 have led to a number of functional prototypes of
photoelectrochemical and photoelectrosynthetic cells in recent years 58, featuring improved
photoelectrode stability through the use of corrosion protection layers 67,68. However, comparison
of solar-to-hydrogen efficiencies realized so far with theoretical limiting efficiencies 69 shows
considerable room for further improvement; at present, the highest efficiency systems reach about
2/3 of the theoretical limiting value for a given photoelectrode. To enable solar-to-hydrogen
efficiencies approaching theoretical limits, the photovoltage has to be as large as possible, which
requires a minimized photoelectrode dark current. This in turn dictates that the charge carrier
recombination at interfaces must be prevented. To maximize the photocurrent, a reduction of the
photoelectrode surface reflectivity under operating conditions is also required, as is mitigation of
light absorption in the catalyst layer applied to the photoelectrode surface 70.
If one utilizes the band gap combination of a given tandem photoelectrode and the best
reported exchange current densities for the HER and OER, omitting losses due to ERE and solution
resistance, the realistic limiting STH efficiencies can be calculated 69. For the tandem
photoelectrode used here (Ga0.41In0.59P/Ga0.89In0.11As with 1.78 eV and 1.26 eV), this value is
22.8 %. Approaching such limiting efficiencies provides a clear objective for a renewable fuels
technology, since inclusion of hydrogen in the existing worldwide fuel generation infrastructure
could enable direct and widespread application of renewable fuels in the transportation sector and
for electricity generation 71.

23
Here, we demonstrate an approach to achieving efficiencies near the theoretical limits for
the photoelectrode energy bandgaps employed. A key aspect of our approach is (i) the use of a
crystalline anatase TiO2 photocathode interfacial layer, deposited by atomic layer deposition
(ALD), to facilitate reduced reflectivity and interface recombination velocity, and (ii) a size
distribution and spatial arrangement of Rh catalyst nanoparticles tailored to achieve ultralow light
attenuation. The crystalline anatase TiO2 interlayer shows excellent energy band alignment with
the tandem window layer and its interfacial ultrathin oxidized surface part and with the electrolyte.
In addition, it serves as an efficient antireflection coating and as a support for the catalyst
nanoparticles, with enhanced adhesion relative to III-V compound semiconductor surfaces.

3.2 Experimental Method
The dual-junction light absorber (Ga0.41In0.59P/Ga0.89In0.11As with 1.78 eV and 1.26 eV) was
grown by metal-organic vapor phase epitaxy in an Aixtron 2800-G4-TM reactor 72,73 on a
4’’ p-GaAs (100) wafer with 6° offcut to [011] using a GaInAs metamorphic step-graded buffer
layer to overcome the difference in lattice-constant between the substrate and the solar cell layers.
The threading dislocation density after the metamorphic buffer is below 1x106 cm-2. Further details
(layer composition and thickness) are given in the reference 72,73.
The native oxide on the back of the GaAs substrate was removed prior to metal ohmic contact
deposition by (1) rinsing in acetone; (2) isopropanol; (3) 30 sec NH4OH (10 %); (4) H2O:N2 and
(5) drying in N2. Immediately afterwards, 70 nm Pd, 70 nm Ti, and 200 nm Au were deposited by
electron beam evaporation followed by rapid thermal annealing at 400 °C for 60 s under N2
atmosphere 66. Prior to the TiO2 layer deposition, the front GaAs/GaInAs cap layer was removed
in a chemical etch bath. The sample was (1) degreased by 15 s rinsing in 2-propanol, (2) 15 s in
H2O:N2 followed by (3) a 60 s etch step in 25 % NH4OH:30 % H2O2:H2O (1:1:10), finishing with
(4) a 20 s rinse in H2O:N2 and (5) drying under N2 (Figure 3.1, step 1). Directly afterwards (a
desiccator was used for sample transfer between systems), TiO2 was deposited by atomic layer
deposition (ALD) in an Ultratech Fiji F200/G2 ALD system using a titanium tetraisopropoxide
(TTiP) precursor (STREM Chemical Inc.) and water as the oxidizer. The deposition temperature
was set to 250 °C, and a total of 1500 ALD cycles were carried out (Figure 3.1, step 2). No high

24
temperature post annealing was required. Note that the edge of the sample had been carefully
removed to prevent shunting of the front and back surfaces. Ag paste was applied to attach an
ohmic contact to a coiled, tin-plated Cu wire which was then threaded through a glass tube. The
sample was encapsulated and sealed to the glass tube using black epoxy (Electrolube ER2162).
The exposed electrode surface area was precisely determined using an optical scanner and the open
source software ImageJ. The steep edge of the high-viscosity epoxy was used as a borderline.
Hence, the spill-out area (~20 µm, see reference 70) was fully included in the area measurement.
In this study the electrodes had different areas of 0.1 – 0.3 cm2.

Figure 3.1: Process flow for preparing the PEC device: (I) Chemical etching of the GaAs/GaInAs
cap layer stopping at the AlInP window layer. (II) Deposition of the TiO2 protection and
antireflection coating with ALD. (III) Photoelectrochemical deposition of a closed layer of Rh
nanoparticles onto the tandem.

The Rh catalyst was photoeletrodeposited (Figure 3.1, step 3) in an aqueous solution of
0.5 mM Rh(III) chloride trihydrate (99.98%, Sigma Aldrich) + 0.5 M KCl (99.5%, Alfa Aesar) at
+0.3 V vs. an SCE reference electrode under pulsed illumination. White light was provided by an
Oriel Instruments Solar Simulator using a 1000 W Mercury-Xenon arc lamp. The frequency of the
stroboscopic illumination resulted from the optical chopper frequency and the double structure of
the chopper wheel.

25
Counter electrodes were prepared by sputtering ruthenium for 60 min on titanium foil
(0.125 mm, 98 %, Sigma Aldrich) using an AJA sputtering system with a forward RF power of
200 W, 5 mTorr Ar atmosphere and a base pressure of 2x10-8 mTorr. Then, prepared electrodes
were cut into 1 cm2 pieces and attached with Ag paste to a tin-plated Cu wire which was then
threaded through a glass tube. The counter electrode sample was encapsulated and sealed to the
glass tube using black epoxy (Electrolube ER2162).
All photoelectrochemical measurements were performed using Biologic SP-200 potentiostats.
1 M HClO4 was used as the electrolyte for pH 0 and 0.5 M KH2PO4/K2HPO4 phosphate buffer for
pH 7. All electrolytes were purged with N2 (4N) for minimum 1 h before usage. A saturated
calomel electrode (SCE) was used as the reference electrode for three-electrode measurements.
Custom-made three-necked cell Glass cells with a quartz window with a volume of 35 mL were
used as the vessel for the experiments, allowing them to be easily cleaned in Aqua Regia. To avoid
internal reflections in the cell, a black mask was directly attached in front of the quartz window so
that only the sample itself was illuminated. The tandem device with areas ranging from
0.1 – 0.3 cm2 was positioned 10 mm away from the quartz window with the counter electrode of
a size of 0.6 cm2 being placed in close vicinity to the working electrode. The photograph of the
custom-built cell is shown in Figure 3.2. The electrolyte was vigorously agitated with a magnetic
stir bar to minimize the diffusion losses. J-V measurements were performed with a scan velocity
of 50 mV/s. To prevent the degradation from running at anodic condition where the dark current
occurred, we only recorded J-V curves until 0 V vs. the counter electrode. Stability and efficiency
tests were carried out in a two electrode configuration using a calibrated Class AAA AM1.5G solar
spectrum provide by an ABET Technologies Sun 3000 Solar Simulator (Figure 3.3). The light
intensity was set to 100 mW/cm2 using a calibrated silicon reference solar cell.

26

Figure 3.2: (a) Cell used for high efficiency benchmarking with WE and CE in close vicinity. The
distance WE to window is 10 mm and the distance WE to CE is < 10 mm. (b) Front view and (c)
side view of the double glass cell used for gas collection. The distance WE to window is 10 mm,
the distance WE to membrane is 40 mm and the distance membrane to CE is 20 mm. The membrane
has an area of 5 cm2. Each compartment has a gas bubbler for pre-saturation of the electrolyte
with H2/O2 purging and gas outlets which are connected to inverted water filled burette for gas
collection. For both cells (a) and (b/c) the quartz window is covered with black tape having an
opening with Æ 20 mm.

Figure 3.3: (a) Light spectrum of the solar simulator (ABET Sun 3000 Solar Simulator) and
AM1.5G spectrum. (b) Light spectrum of the solar simulator and AM1.5G with water filter. (c)
Uniformity map of the solar simulator illumination area. The band gaps of the dual-junction light
absorber are indicated in (a) and (b). (Yellow color for top cell and orange color for bottom cell.)

27
External quantum efficiency (EQE) measurements were performed on fully processed
tandem devices solely to calculate the spectral correction factor to account for the difference
between artificial and solar illumination. Hence, to avoid hydrogen evolution and H2 bubble
formation during EQE measurements, a 50 mM methyl viologen hydrate (98%, ACROS Organics),
dissolved in ultrapure water, was used as the electrolyte. For continuous light biasing of each
individual tandem sub-cell during EQE measurements of the complementary sub-cell, a 780 nm
high-power LED (Thorlabs M780L2) was used to bias the bottom cell and a 455 nm high-power
LED (Thorlabs M455L2) was used to bias the top cell. Monochromatic illumination was delivered
by an Oriel Solar Simulator with a 150 W Mercury-Xenon arc lamp attached to a Newport
monochromator (1200 lines/mm). The monochromatic light was chopped at 10 Hz. The modulated
photocurrent was amplified by an SRS model SR570 low noise current preamplifier. The current
preamplifier was also used to supply a -1 V bias to the tandem working electrode to ensure
measurement in the light limiting current regime. A coiled Pt wire was used as the counter
electrode for this two-electrode measurement. The output from the preamplifier was then measured
by a SRS model SR830 lock-in amplifier which was phase locked to the frequency of the optical
chopper yielding the photocurrent for the individual sub-cell Jtop/bottom(l).
To measure the absolute light intensity (W×nm-1×cm-2) as delivered by the monochromator, a
certified calibrated silicon diode (biased at -1 V) was positioned in the light path inside the
photoelectrochemical cell filled with the electrolyte (to exclude effects of the electrolyte and quartz
window on the measured light intensity), and the photocurrent density was measured (the LEDs
for light biasing of the tandem were switched off during this reference scan). The photocurrent
density could then be converted to the light intensity I(l) by the known spectral response of the
silicon diode. The EQE for each sub-cell is then given by
&à&âäã/çäââäé (ê) =

ìnÅî/ïÅnnÅñ (ó) ò~

∙  = nÅî/ïÅnnÅñ
_(ó)

(ó)

ò~

∙ ó ∙  (Equation 3.1)

Jtop/bottom(l) is the photocurrent density of the corresponding sub-cell in A×nm-1×cm-2, I(l) the light
intensity delivered by the monochromator in W×nm-1×cm-2, λ is the wavelength in nm, h is the
Planck constant, c is the speed of light in a vacuum, and e is the elementary charge. Rtop/bottom is
the spectral response for each sub-cell.

28
To obtain the Faraday efficiency, hydrogen and oxygen gas collection were performed
using a eudiometric gas collection setup. A SELEMINON ion exchange membrane with an area
of 5 cm2 in size was utilized to separate the cathode and anode chamber. Electrolytes were purged
with ultrapure N2 (4N), the cathode side was pre-sutured with H2, and the anode side was presaturated with O2 by means of H2 and O2 gas bubbling through a fine gas dispersion frit for an
hour. Each side was sealed against the ambient but connected via a short thin tubing to an inverted
water filled burette (purged and pre-saturated). The change in pressure in each burette upon H2
and O2 gas collection due to photoelectrochemical water splitting in the PEC cell was monitored
by pressure transducers (EXTECH HD755). The change in pressure over time was then converted
to a gas volume under consideration of the reduced pressure in the inverted burette. For constant

temperature, ∆.ö = i ∙ ∆. with ú = (407.2 − ℎ) and úö = 407.2, using inWC (inches of water

column) as pressure units.

iõ

is the volumetric correction factor necessary to account for reduced

pressure in the inverted burette. The expected produced volume of hydrogen and oxygen gas for
the cathodic and anodic reaction was calculated by the transferred electrical charge as measured
by the potentiostat.
Optical measurements were performed to obtain reflectivity spectra for different surface layer
stacks in air. A Cary 5000 UV/vis/NIR with an integrating sphere that includes a diffuse reflectivity
measurement was used. For surface topography studies, a Bruker Dimension Icon AFM in
Peakforce mode was used. Scanning electron microscopy images were obtained with a FEI Nova
NanoSEM 450 microscope. XPS measurements were performed using a Kratos Axis Ultra and
Surface Science M-Probe system with a base pressure of < 1x10-9 mTorr. A monochromatic AlKa
(hK = 1486.69 eV) source with a power of 150 W was used for all measurements. He I ultraviolet
photoelectron spectroscopy (UPS) was performed on the Kratos Axis Ultra system using a Helium
gas discharge lamp.

3.3 Results and Discussion
We employed a dual-junction tandem photoelectrode where the high band gap subcell
thickness had been increased for better current matching, and the transparency of the tunnel diode

29
was improved 70,72,73. To further increase the STH efficiency, interfacial layers were designed
to reduce charge carrier recombination and to increase optical light coupling into the
photoelectrode absorber layers. The surface conditioning sequence resulted in etching of the GaAs
cap layer by a NH4OH-H2O2-H2O solution, leaving an oxidized surface layer (AlInPOx) on top of
the n+-doped AlInP window layer. A crystalline anatase TiO2 film with an effective thickness of
30 nm was deposited to act as a corrosion protection layer and an antireflection coating, as well as
serving as conducting substrate surface for photoelectrodeposition of Rh nanoparticle (NP)
electrocatalysts. The Rh NPs exhibited large surface areas and thus high exchange current and,
simultaneously, particularly low light attenuation. The photocathode device configuration
employed is generally less prone to photodecomposition than photoanode devices, where charge
carriers with high oxidation potential are present at the semiconductor surface.
Figure 3.4 shows a schematic of the resulting device: the photoelectrode consisting of GaInP
and GaInAs subcells on a GaAs substrate, an anatase TiO2 protective layer, the Rh NPs catalyst
layer, and a sputtered RuO2 counter electrode (OER) are depicted. Also depicted on the side of the
layer structure is an energy band diagram under illumination where the quasi Fermi levels show
the splitting for electrons, and holes necessary to achieve unassisted water splitting. Measurement
of band gap, work function, and band bending are included in Figure 3.5 and Figure 3.6. The
corresponding energy band relations can be inferred from surface characterization using ultraviolet
and X-ray photoelectron spectroscopy. While the simplest approach to assessment of band
alignment follows Anderson’s idealized model 67 for planar contacts and does not consider energy
band shifts due to surface and/or interface dipoles, this approach certainly does not apply here, as
the junctions formed at the AlInP/oxide, oxide/TiO2 and TiO2/Rh/electrolyte interfaces are
complex. Thus the energy band diagram of the heterojunction structure was inferred from
ultraviolet and X-ray photoelectron spectroscopy measurements. It should be noted that
equilibrium formation between small metallic catalyst nanoparticles and semiconductors appears
to depend on the substrate doping level 10 and does obviously not follow a Schottky thermionic
emission model, in particular in contact with an electrolyte 68,69. In addition, metal work functions
depend on NP size 70, so comparison of the energy levels of NP catalyst layers with planar thin
films is notably challenging; therefore, only an estimate of the NP catalyst layer energy level can
be given, supported by the device operating data. The resulting surface band alignment as

30
displayed in Figure 3.7 can be obtained after aligning the Fermi level of the solid state device
to the hydrogen evolution potential (HER) at 4.6 eV. Due to the higher work function of 4.5 eV
for TiO2 than 4.1 eV for the tandem, a lower barrier from 0.5 eV to 0.1 eV for hydrogen evolution
(HER) will be expected.

Figure 3.4: Illustration of the photoelectrochemical water splitting device structure after
functionalization with interfacial films and electrocatalysts. Band alignment at the operation point
is depicted on the side and zoomed in to gain the visibility.

Figure 3.5: (a) Optical properties (A: absorption, T: transmission, R: reflection) of TiO2 (TTiP
ALD) in air. (b) Tauc plot of ALD grown TiO2. The intersection with the horizontal axis indicates
an indirect optical gap around 3.3 eV.

31

Figure 3.6: (a) Work function measurements by UPS for the tandem, for TiO2 on the tandem, and
for Rh metal. The increase of work function from 4.1 eV to 4.5 eV was observed after applying
TiO2 protection layer on tandem. The Rh metal spectrum is measured on the foil as a reference
instead of the photoelectrochemical deposited nanoparticles. (b) Core level shift of Ti 2p3/2
indicating ~0.3 eV downward band bending at the tandem/TiO2 interface and nearly no band
bending at the TiO2/Rh interface. The tandem/TiO2 sample was made with 40 ALD cycles TiO2 on
top of the tandem. The Rh has originally high metal work function of 5.1 eV but does not create
band bending at the junction with TiO2. This can be explained by the pinch-off effect when the
metal NPs are small enough that the Fermi level would directly attach to the semiconductor Fermi
level without creating a barrier 71.

Figure 3.7: Surface band alignment of the electrolyte interface layers (a) without and (b) with
TiO2.

32
The surface of the crystalline TiO2 film illustrated in Figure 3.8 indicates a continuous film
with height variations, seen by AFM, that give it a flake-like appearance. The TiO2 is then
decorated with a uniformly dense layer of ca. 10 nm Rh nanoparticles. The catalyst distribution of
the optimized devices is also shown in Figure 3.8. Figure 3.9 illustrates the protocol for pulsed
photoelectrodeposition of Rh catalyst nanoparticles. The potential choices made are indicated by
black dots; the best result was obtained for E = +0.3 V vs. SCE. Fine control of particle size smaller
than 20 nm was achieved by careful adjustment of the electrode potential, enabling considerably
higher catalyst loading compared to a dense film of equivalent catalyst loading deposited by
conventional vapor phase or electrochemical reduction. This procedure facilitates photocathodes
with high transparency catalysts, which maintain the high photocurrent densities and result in
increased efficiency, which is determined from the relation

!UVÖ = Åî

∙abcde ∙fgh
ike

(Equation 3.2).

The solar fuel generator efficiency hSTH is given by the operating current at the counter electrode
potential, the thermodynamic value for the reaction (ΔUrxn = 1.23 V for water splitting under
standard conditions), and by the reaction Faradaic efficiency fFE, determined by gas product
analysis measurements.

Figure 3.8: SEM images and AFM microtopographs of the dual-junction PEC device with TiO2
coating with and without Rh catalyst nanoparticles. The scale bar is 500 nm. The AFM images are

33
scaled to the same 50 nm z-axis dynamic range. The surface roughness (RMS) is 3.6 nm without
Rh and 6.3 nm with Rh.

Figure 3.9: (a) Fine control of particle size d ranging from 10 nm to 100 nm is achievable by
appropriate adjustment of the potential during catalyst electrochemical deposition. Stroboscopic
deposition under white light illumination as shown in the upper left insert. The three images on the
right inset are SEM images with scale bar 2 µm. (b-d) Particle size histograms correspond to each
SEM image depicted from top to bottom in (a) with the most frequent particle size indicated by d† .

34
Electronically, the photoelectrode configuration used here facilitates alignment of the
conduction bands of the AlInP window layer of the tandem photoelectrode to the indium oxide
and indium phosphate layers (created by the cap layer etching process) and the anatase TiO2
protection/antireflective layer. We note that photogenerated electrons, which are minority carriers
in the main part of the tandem subcells, become majority carriers in the AlInP and TiO2 layers,
reducing recombination losses in carrier transport. In addition, the large valence band offset
between AlInP and TiO2 blocks interfacial hole transport, resulting in a small overall reverse
saturation current, improving the photovoltage. This feature is important for achieving high STH
efficiencies.
Amorphous defective TiO2 coatings have been commonly applied using ALD with
Tetrakis(dimethylamino)titanium (TDMAT) precursors to yield protected photoanodes, a concept
designed to facilitate transport of holes through a defect band in the TiO2 68,74. However, the
dominant process that limits the photoelectrochemical performance of high-quality
semiconductors is not transport but interface recombination. Thus, we instead utilize a defect band
free, microcrystalline anatase phase TiO2 coating formed by ALD with titanium tetraisopropoxide
(TTiP) precursors as an electron-selective contact to protect the surface of photocathode from
photocorrosion (Figure 3.10).

Figure 3.10: (a) X-ray diffraction data of ALD deposited TiO2 from TTiP or TDMAT precursor.
The TTiP TiO2 shows anatase crystalline phase while the TDMAT TiO2 is amorphous. (b) XPS
valance band spectra of TTiP and TDMAT TiO2. A defect band in TDMAT TiO2 can be observed

35
at -1 eV which facilitates hole transport in photoanodes 74. Instead, TTiP TiO2 exhibits an XPS
spectrum without a defect band and would be more suited to prevent recombination in
photocathodes.

To assess the optimal thickness, we performed a series of optical reflectivity measurements
on TiO2 films deposited by atomic layer deposition (ALD) on the tandem photoelectrodes with
various thicknesses in the range of interest. The results are shown in Figure 3.11(a). We find the
30nm thickness TiO2 layer reported in the manuscript has the lowest reflection in the relevant
spectral range. Figure 3.11(b) shows results of calculations using full wave electromagnetic
simulations performed using finite-difference time-domain methods (Lumerical FDTD) assessing
the effect of thickness variations on optical reflectivity similar to the experimental results. For the
TiO2, we find a very low reflectivity over a wide spectral range for a nominal thickness of 30 nm.

Figure 3.11: (a) Reflectance, measured in air, of the dual-junction tandem solar cell with different
thicknesses of the TiO2 coating by changing the ALD deposited cycles. (b) Reflectance, simulated
by Lumerical FDTD, with different thicknesses of TiO2 for correlation with the experimental
results.

To optimize light coupling, we also carefully tailored the optical properties of the Rh
nanoparticles to work in combination with an optimum TiO2 thickness of 30nm, determining the
reflectance, absorbance, and transmittance. To determine the influence of the Rh particle size on

36
reflectance, absorbance, and transmittance, we modeled three particle sizes using full wave
electromagnetic simulations. The optical transmission modeling in Figure 3.12(b) shows that for
a 10nm Rh particle size, an optimum is reached. We show that the transmittance and
photoelectrode light coupling for the entire structure consisting of (tandem photoelectrode/30nm
TiO2/10nm Rh particles) is almost identical to a bare surface without Rh. Although a 40 nm Rh
particle size shows lower reflectivity in the relevant spectral range in Figure 3.12(a), the Rh particle
absorption increases substantially, resulting in an overall lowered transmittance into the cell, as
seen in Figure 3.12(c). The transmittance is therefore reduced and fewer photons reach the
photoactive part of the cell (see Figure 3.12(b)).

Figure 3.12: FDTD simulated (a) reflectance, (b) transmittance, and (c) absorption defined as
A = 1 - R - T of different Rh particle sizes on 30nmTiO2/AlInP (window layer of the tandem).

A similar trend is observed in the experimental results shown in Figure 3.13. We measured
the reflectance of samples with different particle size. The blue curve shows the reflectance of the
tandem with a TiO2 layer, but without Rh catalysts. The red curve shows the system with ~10nm
Rh particles added, as used in our record device. As can be seen from the simulations in Figure
3.12(b), we expect the 10nm Rh particles to be effectively transparent. The yellow and purple
curves show the reflectance with medium and large sized Rh particles, which would lead to a larger
loss due to parasitic absorption and reflection.

37

Figure 3.13: Reflectance, measured in air, of samples with different Rh NPs size on the dualjunction tandem solar cell with 1500 ALD cycles TiO2 corresponding to a layer thickness of 30 nm.

The influence of the surface modifications on optical properties and on the photocurrent of
the optimized device is shown in Figure 3.14. A reduction of the reflectivity by ~15 % is achieved
by use of the TiO2 interlayer (Figure 3.15) whereas the Rh NPs in Figure 3.14(a) show negligible
additional absorption, which is attributed to the blue-shifted plasmonic resonance of the Rh
nanoparticles. For particle sizes below 20 nm, a shift from the visible region into the ultraviolet
one occurs, making the Rh layer almost fully transparent 75,76. The output data shown in Figure
3.14(b) were obtained in an acidic electrolyte of pH 0. The corresponding photocurrent-voltage
characteristics in acidic electrolyte demonstrate a pronounced increase in the current and, as
expected, also a shift of the bend of the photocurrent characteristic towards more anodic potentials,
thereby additionally increasing the photocurrent at the RuO2 counter electrode (OER) potential.
The result with incorporation of TiO2 is a relative increase of 28 % of the tandem cell output. An
STH efficiency of 19.3 % is obtained at 0 V, with an operating current of 15.7 mA/cm2, assuming
an initial Faradaic efficiency of unity, which is supported by the gas evolution measurements.
These data represent a 20 % increase in efficiency above the previously reported one sun
photoelectrosynthetic cell efficiency benchmark 5. The high photocurrent at 0 V vs. RuO2 indicates
that electron transport is virtually uninhibited from the absorber layer through the indium and
phosphorus oxide and TiO2 interfacial layers to the electrolyte, matching well with the surface
band alignment analysis.

38

Figure 3.14: Optoelectronic properties of the surface functionalized electrolyte / Rh / TiO2 / oxide
/ AlInP - GaInP / GaInAs / GaAs water splitting device. (a) Reflectivity, measured in air, of the
dual-junction tandem solar cell without ARC (black curve), secondly reflectivity obtained after
TiO2 coating (blue curve) and after photoelectrochemically deposited Rh NPs (yellow curve).
Reflectivity is larger than under operation in the electrolyte due to the different refractive indices
of air and water. (b) Comparison of the output characteristics of the tandem device after cap layer
etching and of the fully surface functionalized photoelectrode. The orange arrows indicate the
improvement after incorporation of the TiO2 layer.

Figure 3.15: The enhancement of absorption based on the reduction of the reflectivity for the PEC
device due to employment of TiO2 layer. (Absorption = 1-Reflection) The 15% average increase
of absorption can directly contribute to the enhancement of photocurrent.

39
The observed unassisted water splitting efficiencies critically depend on the experimental
conditions. In order to consider the influence of the spectral mismatch of the irradiance between
our solar simulator and the AM1.5G spectrum, a spectral correction factor (SCF) was calculated.
It is based on the relative EQE of the device, the irradiance of the solar simulator {Imeas(l)} and
the AM1.5G reference spectrum {IAM1.5G(l)} (Figure 3.3(a)). The influence of the water filter
{Fwater(l)} on the spectra was considered for the calculations (Figure 3.3(b)). The index j denotes
to the individual sub cell.
1200nm

/o°K.Q¢ = /飧 ∙

•¶ß ∫280nm _™´¨.≠Æ (ó)∙WØ∞nÑc (ó)∙±≤±®,≥Ñ¥kÉÑ (ó)µó
1200nm

•¶ß ∫280nm _ñÑ∞∂ (ó)∙WØ∞nÑc (ó)∙±≤±®,≥Ñ¥kÉÑ (ó)µó

= /飧 ∙ ∑05 (Equation 3.3).

For illumination under AM1.5G conditions, the AM1.5G ASTM G-173 reference spectrum was
taken from the Renewable Resource Data Center (RReDC) of the National Renewable Energy
Laboratory (NREL). Although the SCF is a simple correction between spectra, we still need to be
careful to prevent artificial inflation or underestimation of efficiency. We note that the SCF will
be unsuitable to apply with large deviation from 1 since the current correction will become
unrealistic due to severe difference of overpotential and bubble formation.
To calculate the spectral correction factor (SCF) between solar simulator and the AM1.5G
spectrum, the EQE measurements were performed in 50 mM methyl viologen where no bubble
formation would deteriorate the accuracy. The bias light was 780 nm and 455 nm for the bottom
and top sub-cell, respectively. The result is shown in Figure 3.16. To correct for nonparallel
illumination in the solar simulator that results in focusing of the light by the quartz window, the
beam divergence in each axis was experimentally determined, and a concentration ratio (CR) was

calculated (Figure 3.17). The corrected photocurrent is given by /ö = ñÑ∞∂
. The total correction
∏ì
factor for each sample is then given by /o°K.Q¢ = /飧 ∙ ∑05/0(, e.g. for the 19.3 % efficient
cell reported in this study, the values are SCF = 1.024 and CR = 1.028.

40

Figure 3.16: Relative EQE of a fully processed PEC tandem device: the bandgap combination is
determined to be around 1.78 eV for the top cell and 1.26 eV for the bottom cell.

Figure 3.17: (a) Calculated optical concentration ratio of the non-parallel light-beam of solar
simulator illumination in PEC cells for plane wavefront and spherical wavefront as a function of
water path length. (b) Illustration of the spherical wavefront case. The concentration ratio
(CR = A0/ACR) depends on the exact sample area A0. (c) Illustration of the plane wavefront case.
An opening aperture in front of the quartz window of the PEC cell with a diameter of 2 cm was
used in this study. The beam divergence was experimentally determined to be QV = 1.8 ° vertically
and QH = 2.5 ° horizontally by measuring the size increase of the light beam through a 2 cm
aperture at specific distances (10 cm to 30 cm).

41
For an evaluation of the influence of polarization losses including ohmic losses, diffusion
losses, and kinetic losses 77, we simulated the maximum obtainable efficiencies in the detailedbalance scheme with the program YaSoFo 70. Since the electrolyte was vigorously agitated and
buffered, we are able to minimize the diffusion losses. Optical losses were assumed to be the cause
for the difference in theoretical and practically obtained limiting photocurrents. In the simulation,
these were taken into account by scaling the AM 1.5G spectrum with a constant factor of 0.89.
The diode current-voltage curve (ideality factor ni=1) intersected with the characteristics of the
catalyst following a Tafel behavior described by exchange current density and Tafel slope under
the assumption of an additional ohmic drop of 2 Ω. Fig. Figure 3.18(a) shows the maximum STH
efficiency as a function of exchange current density and Tafel slope. We observe that in the regime
of our OER catalyst (exchange current density of ~10-3 mA·cm-2 for RuO2, Tafel slope for RuO2
as 83 mV·dec-1 at pH 0 and 100 mV·dec-1 at pH 7, see Figure 3.19), the exchange current density
and Tafel slope are still in the plateau of the maximum efficiency; the corresponding points are
indicated in Figure 3.18(a). As our case catalysis is dominated by the OER, it is not an efficiencylimiting factor of our setup. We also show in Figure 3.18(b) the efficiency as a function of the
ohmic drop. For this analysis, we fixed exchange current density and Tafel slope to typical values
of IrO2 and varied the ohmic resistivity. One notices that the efficiency only starts to drop at high
values beyond 40 Ω, which is why the resistive overpotential is not limiting in our setup, either.
Combining the ohmic loss that can be induced from the electrolyte (~4 Ω for pH 0 and ~15 Ω for
pH 7), and interface loss from imperfect surface band alignment (~40 Ω without TiO2 and ~6 Ω
with TiO2), our record device photocurrent is still located at the linear region which indicates that
the optical losses are prevailing in our system. However, the system, in principle, reacts sensitively
to the polarization losses, emphasizing the importance of judiciously combining interface and
catalyst.

42

Figure 3.18: Calculated maximum STH efficiency as function of (a) Tafel slope A and exchange
current density J0, (b) ohmic drop. Maximum obtainable efficiencies for the given tandem absorber
are shown in the detailed-balance scheme as a function of the catalyst parameters and resistivity
loss. The maximum photocurrent density was scaled to the experimentally determined current
density under strong cathodic bias. The blue star indicates our device under pH 0 condition and
the red star indicates our device under pH 7 condition.

Figure 3.19: Tafel plots of (a) Rh and (b) RuO2 catalysts under pH 0 and pH 7 conditions. The
Tafel slopes are 34, 38, 83, and 100 mV·dec-1 for Rh-pH0, Rh-pH7, RuO2-pH0, and RuO2-pH7
respectively.

43
Figure 3.20 summarizes the main performance characteristics. Figure 3.20(a) illustrates the
photocurrent-voltage characteristics under three conditions: i) at pH 0 with 19.3% STH, ii) at
neutral pH with 18.5 % STH, and iii) using an anion exchange membrane (AEM) with an STH of
14.8 %. Figure 3.20(b) gives the unassisted two-electrode photocurrent density vs. time for the
initial operation regime, showing that while the photocurrent density decreases with time for acidic
pH, it remains more stable in neutral pH solutions. Chronoamperometric tests (at -0.4 V vs. counter
electrode as shown in Figure 3.20(c) show that the device photocurrent density decreases in an
acidic electrolyte to low values within 3 h. However, in neutral pH electrolyte, stability over 20 h
was demonstrated, with the photocurrent density remaining at 83 % of its initial value. In both
cases, for pH 0 and pH 7, near unity Faradaic efficiency is confirmed through the agreement
between the expected (solid line) and measured gas volumes (symbols) in Figure 3.20(d). However,
whereas the curves for pH 7 stay linear with a constant gas production rate for H2/O2, as expected
from the stability measurements, the curves for pH 0 show a deviation from linearity due to the
decreasing photocurrent.

44

Figure 3.20: Output characteristics of the RuO2-Ge/GaInAs/GaInP/AlInP/anatase TiO2/Rhelectrolyte dual junction tandem structure. (a) Photocurrent-voltage characteristics in acidic (pH
0), neutral (pH 7) electrolyte, and in neutral electrolyte including an AEM membrane. (b)
Chronoamperometric data of the initial temporal regime. (c) Stability measurements at -0.4 V vs.
RuO2 counter electrode for acidic and neutral pH. (d) Hydrogen and oxygen gas collection for
operation in acidic (open spheres) and neutral (full spheres) electrolyte. The measured gas volume
for oxygen (blue symbols) and hydrogen (red symbols) is overlaid with the expected produced gas
volume, as calculated from charge passed through the anode and cathode.

Etching of TiO2 is expected to occur at pH 0 but not at pH 7, as can be seen in the TiO2
Pourbaix diagram in Figure 3.21. Corrosion reactions can degrade the junction photovoltage, as
well as lead to undercutting and removal of catalyst particles, thus reducing the exchange current
of the Rh NP arrangement and a slowing of the HER kinetics. The system reacts also sensitively

45
to series resistance changes, as illustrated by characteristics for devices employing an anion
exchange membrane. The bend of the J-V curve is shifted to cathodic potentials. However, device
operation at pH 7 still yields a high STH efficiency of 18.5 %, and the device appears to be stable
for a more extended period in accordance with predictions of TiO2 stability from thermodynamics.
Even a slower reduction of the photocurrent is observed, we found that this photocurrent reduction
could be partially reversed by emersion of the device from the electrolyte solution and applying a
soft cleansing procedure. The observation that the photocurrent can be partially restored appears
to rule out loss of Rh catalyst particles, or even partial removal of the anatase interfacial layer, as
causes of photocurrent reduction.

Figure 3.21: Potential-pH equilibrium diagram for the system titanium-water system at 25 °C,
adapted from ref. 77. For pH 0, the stable region is small. Upon overpotential to hydrogen evolution,
corrosion sets in, which ultimately leads to the degradation of the device and its efficiency.

Increasing the efficiency of a photoelectrosynthetic device from already high values towards
theoretical limits is especially challenging. We have used a series of surface conditioning steps
that have a twofold function: light management was drastically improved, and the electronic
properties were at least maintained. Compared to our earlier results 10, we see an increase in the
available cell voltage that is related to the increase in photocurrent at the counter electrode
operation potential. Junction formation between the etched AlInP layer, TiO2 layer, and Rh NPs

46
suggests that the Fermi level alignment is nearly ideal. Using the parameters shown in Table
3.1, our photoelectrosynthetic device reaches 0.85 of the theoretical limiting efficiency. It should
be noted that the theoretical efficiency determined from the data in Table 3.1 is based on the best
presently known electrocatalysts, a unity photoelectrode radiative efficiency, and an absence of
absorption losses 69. Figure 3.22 shows a summary to date of selected STH efficiencies realized
for monolithic integrated photoelectrosynthetic devices capable of unassisted water splitting. (The
performance of each device are listed in Table 3.2.)

Table 3.1: Approaches to theoretical limitation of light-induced photoelectrochemical water
splitting; ideal, only exchange current density limited and devices that are optically and
electrochemically limited are displayed, respectively. For the used band gap combination and only
catalytic exchange current density (JXC) limitation, htheo = 22.8 % at AM 1.5G irradiation.

J0, anode

J0,cathode
(mAcm-2)

(mAcm-2)

ideal

JXC limited
JXC and optically
limited

fabs

ERE

Rs (W)

Rsh (W)

10-3

10-3

0.9

0.03

47

Figure 3.22: Comparison of realized limiting STH efficiencies and historic development. The
analysis refers to a theoretical benchmarking value htheo and takes into account the top and bottom
cell band gaps for the respective photolysis cells; also shown are the institutions of the contributing
research teams. Abbreviations: NREL - National Renewable Energy Laboratory, USA; ISE Institute for Solar Energy, Germany; JCAP - Joint Center for Artificial Photosynthesis, Caltech;
TU-I - Ilmenau University of Technology, Germany; HZB - Helmholtz Zentrum Berlin, Germany.
The bar chart on the right indicates the achieved efficiency with respect to the respective
theoretical limit (!theo
).

48
Table 3.2: Reported STH benchmarks from literature with employed bandgaps, achieved STH
efficiency (!STH ), theoretical limit for realistic water splitting (!theo ), and ratio of achieved STH to
!theo as !theo

Bandgaps

∫STH (%)

∫theo (%)

∫∗theo (%)

Reference

JCAP/TU-I/ISE

1.78/1.26

19.3

22.8

85

This work

NREL

1.8/1.2

16.2

24.2

67

TU-I/HZB/JCAP/ISE

1.78/1.26

14

22.8

61

10

JCAP

1.84/1.42

10.5

19.7

53

78

NREL

1.83/1.42

10

19.7

51

5,66

Compared to an earlier reported record photoelectrosynthetic cell with inverted metamorphic
multi-junction semiconductor architecture 5, the tandem device employed in the present study has
a less ideal bandgap combination of 1.78 V/1.26 V, compared to 1.8 V/1.2 V, which leads to the
reduction of theoretical efficiency from 24.2 % to 22.8 %. However, due to better light
management (antireflection layer, optimized catalyst loading, thinning, and transparency of the
tunnel junction), a higher current density and higher STH efficiency of 19.3 % are observed. This
efficiency corresponds to an enhancement from 67 % to 85 % of the ratio of achieved efficiency
to theoretical efficiency for the employed bandgaps. The stability tests using a two-electrode
configuration at 0 V vs. RuO2 counter electrode to demonstrate unassisted water splitting for the
two devices are shown in Figure 3.23(a). The result from ref. 5 was corrected by scaling the initial
current density (including spectral calibration errors) to 13.17 mA/cm2 calculated from their
reported highest efficiency of 16.2%. At an acidic pH, our device current density drops from
15 mA/cm2 to less than 5 mA/cm2 within an hour. By contrast, for pH 7, the photocurrent density
and device performance are more stable and show similar chronoamperometric performance to
that reported in ref. 5; our device is stable for the first 20 min, then the photocurrent density slowly
decreases. The spikes in current density indicate the influence of bubble formation and subsequent
detachment. The dynamics are different due to the change in the reduction mechanism (proton

49
reduction at pH 0, water reduction at pH 7) and the surface tension of the electrolyte. The surface
tension of the phosphate buffer is higher than for acidic electrolyte (see Figure 3.24) and exhibits
more severe bubble accumulation that induces greater photocurrent density fluctuations.

Figure 3.23: (a) Stability measurements at 0 V vs. RuO2 counter electrode for acidic and neutral
pH. The result from ref. 5 are adapted and included for comparison as the black curve. (b)
Chronoamperometric measurements at -0.4 V vs. RuO2 counter electrode for acidic and neutral
pH. The results from ref. 5 at 0.6 V vs. RHE are adapted and included for comparison in black.
Currents rescaled based on the reported efficiency in ref. 5 are shown in blue.

50

Figure 3.24: Contact angle measurement for pH0 1 M HClO4 (a, c) or pH7 0.5 M phosphate buffer
(b, d) on the tandem (a, b) or on the TiO2/tandem (c, d) sample. The image was analyzed with
ImageJ with the help of the “Drop Analysis” plugin developed at the École polytechnique fédérale
de Lausanne (EPFL) (http://bigwww.epfl.ch/demo/dropanalysis/). The larger contact angle of
phosphate buffer indicates higher surface tension, which can lead to more severe bubble
accumulation and larger photocurrent density fluctuations.

It is well known that photoelectrochemical devices can be better stabilized in a three-electrode
configuration at the RHE potential. We were also able to demonstrate long-term stability of 50 hrs
under these conditions without the existence of a protection layer in an acidic environment (see
Figure 3.25). However, these conditions are not comparable to operation at 0 V vs. CE in a twoelectrode configuration, as operation at RHE potential diminishes the effect of corrosion. For
comparison and to understand the intrinsic differences between operation at different pH
conditions, chronoamperometric tests were conducted at -0.4 V vs. counter electrode (equivalent
to approximately +1.1 V with respect to RHE (estimated to be 1.23 V plus catalysts overpotential
minus 0.4 V) as shown in Figure 3.23(b). The stability data of the device reported in ref. 5, which
was operated at +0.6 V vs. RHE with and without current rescaling, is included for comparison. It
shows that the device photocurrent density decreases in acidic electrolyte to low values within 3 h.
In neutral pH electrolyte, stability over 20 h was demonstrated, with the photocurrent density

51
remaining at 83 % of its initial value. At 12 h into the test, a diurnal cycle was simulated by
emersion of the sample in the dark for a few minutes. This step resulted in a substantial current
density enhancement. Overall, our device operated in neutral pH conditions shows similar stability
characteristics compared to that reported in ref. 5, while exhibiting higher efficiency. Our device,
operated in a biased two-electrode configuration exhibits an extended stability with respect to the
three-electrode measurements reported in ref. 5. We conducted XPS measurements to further
understand the degradation mechanism.

Current density(mA/cm )

-2
-4
-6
-8
-10
-12
-14

10

20
30
Time(h)

40

50

Figure 3.25: Chronoamperometric measurements at 0 V vs. RHE for Rh/Tandem device without
protection layer in acidic environment.

X-ray photoelectron spectra of tandem samples after each step in the PEC device production
process are shown in Figure 3.26. The tandem etched spectra show the exposed AlInPOx window
layer. Upon deposition of TiO2 with no visible In and P signal, we infer full coverage of the
protection layer without pinhole formation. Photoelectrochemical deposited Rh with the similar
intensity to the Rh film indicates a sufficient amount of catalysts. However, the small remaining
Ti peaks present the existence of the TiO2 exposed area. After photoelectrolysis in an acidic
environment, the TiO2 peak enhancement indicates more exposed areas upon local detachment of
catalysts (see Figure 3.27). The maintained prominent Rh peak implies the loss of catalyst is not
the limiting factor of degradation. Instead, the appearance of underlying In and POx peaks supports
the scenario of tandem corrosion due to local TiO2 etching. After operation in neutral pH, we
observed the enhancement of the phosphate peak for the aged sample and reduction again after the

52
recovery as shown in Figure 3.28. Note that no In signal is detected in all samples, which
indicates the source of phosphate species is the buffer electrolyte rather than tandem corrosion.
We thus deduce that the current reduction during the stability test is mainly contributed by the
poisoning of Rh catalysts by the phosphates group. The photoelectrode regeneration procedure
results in a 50 % recovery of the photocurrent lost during the first 12 hrs, suggesting that the high
porosity of the Rh NP layer inhibits full recovery by a short intermediate treatment. The residual
loss remaining after dark recovery can be attributed to the POx groups still remaining on the surface.
Employing a different electrolyte for pH 7 conditions might therefore benefit longterm activity of
the device.

Figure 3.26: X-ray photoelectron spectra of tandem samples after each step in the PEC device
production process: after removing the GaAs/GaInAs cap layer by chemical etching (black curve,
indicated as Tandem etched), after deposition of the TiO2 layer by ALD (green curves, indicated
as +TTiP TiO2); and after photoelectrochemical deposition of Rh nanoparticle catalysts (blue
curve, indicated as +Rh). As a reference, spectra of metallic Rh electrode are included (red curve,

53
indicated as Rh metal). (a) In 3d core levels; (b) P 2p, In 4s and Al 2s core levels, the peak of
POx is indicated; (c) Ti 2p core level; and (d) Rh 3d core level.

Figure 3.27: X-ray photoelectron spectra of a pristine Rh/TiO2/Tandem sample (black) and after
degradation in acidic environment (red): (a) In 3d core levels; (b) P 2p and In 4s core levels, the
peak of POx is indicated; (c) Ti 2p core level; and (d) Rh 3d core level.

54

Figure 3.28: X-ray photoelectron spectra for the study of Rh catalyst poisoning by POx groups in
pH 7: (a) In 3d core levels; (b) P 2p core levels, the peak of POx is indicated; (c) Ti 2p core level;
and (d) Rh 3d core level. The black curve indicates the pristine (p) sample before any
photoelectrochemical measurement. The red curve indicates the aged (a) sample, which was taken
out from the electrolyte under light illumination after operation. The blue curve indicates the
recovered (r) sample, which was taken out from the electrolyte under dark condition to simulate
the diurnal cycle.

3.4 Conclusion and Outlook
Stability appears to remain an issue of this photocathode device configuration, but we have
demonstrated high efficiency in neutral electrolytes, and that extended operation of photocathode
devices becomes possible if one can control the Rh surface chemistry. The use of Rh NPs with
tailored size and shape distributions enables ultralow absorption. The future design of even more
optimized tandem photoelectrodes appears to be possible, enabling solar fuel generation (water
splitting, as well as CO2 or N2 reduction) efficiencies to be even higher than reported here, for

55
example with STH champion device efficiencies of >20 % for integrated direct water photolysis
being a realistic goal.
In PEC system, the cost of raw materials and system level deployment are critical issues for
real world application. As a general guidance, STH efficiency of 15% and extended lifetime of 20
years would make PEC device a practical solution. Stability is currently the most important
challenge to overcome. Although many demonstrations show the photoelectrochemical devices
can be stable for long period of time at 0 V vs RHE potential, the device can barely survive at real
operation conditions for a short time. New approaches and strategies of the protection scheme
would need to be developed.

56

CHAPTER 4
High Efficiency Solar to CO PV-GDE Device
4.1 Introduction
Parallel to solar hydrogen generation approaches, pathways for solar-driven reduction of
carbon dioxide to fuels have used i) direct electrolysis,79 ii) photovoltaic directly driven
electrolysis 80, and iii) integrated photoelectrochemical conversion.81,82 Of particular interest is
solar-driven reduction of carbon dioxide using a high efficiency photovoltaic (PV) device directly
coupled to an electrochemical cell tailored for reduction of CO2 to CO.83 Mixtures of solargenerated CO and H2,84 could be used as syngas precursors in a future Fischer-Tropsch chemical
synthesis process 85 to produce high molecular weight hydrocarbon fuels, or chemicals as
products.68 Carbon dioxide reduction to CO is generally more energy efficient and kinetically
easier than direct reduction of CO2 to multicarbon products.83,86
Among the most efficient heterogeneous solid state catalysts for CO2 reduction to CO are
gold,87,88 silver,89 WSe2,90 and MoS2.91 The use of high surface area morphology structures such
as nanoparticles can improve catalytic activity.92 Other factors that impact catalytic performance
include catalyst morphology,87 cations present in the electrolyte solution,93 electrolyte
concentration 94, and local pH.95 The state-of-the-art CO2 to CO conversion using a Au needle
catalyst 94 showed an operating current of 15 mA×cm-2 and 95 % Faradic efficiency at -0.35 V vs.
RHE. However, the current record efficiency device for solar conversion of CO2 to CO using a
solution-based electrochemical cell suffered from low current density (0.33 mA×cm-2 at -0.6 V vs.
RHE) due to limited catalyst activity. This required the use of large-area electrodes to match the
photovoltaic device area.80 Table 4.1 shows overpotential and Faradic efficiency data at current
densities close to 15 mA×cm-2 along with the electrolyte conditions and catalyst loading for
various Ag and Au electrodes. The catalytic activities of the catalysts indicate that in many cases,
nanoparticles of Ag have a similar activity to that of Au while costing significantly less.

57
Table 4.1: Comparison of the CO2 reduction performance of our Ag-NP catalyst with
previously reported Ag and Au electrodes. (U as overpotential vs. E0CO/CO2 (-0.11 V vs. RHE),
r∙f

gh
mass activity define as Ä䣵ªºΩ
.)

Catalysts

Electrolyte

pH

U (V) at
~15 mA×cm-2

fFE,CO
(%)

Loading
(mg×cm-2)

Mass
activity
(mA×mg-1)

Ref.

Nanoporous
Au

CO2-sat
0.2 M KHCO3

6.8

1.24

N/A

0.39
(200 nm film)

N/A

96

Nanoporous
Au

CO2-sat
0.1 M KHCO3

6.8

0.6

60

0.47
(eq. 240 nm film)

19.15

87

OD-Au

CO2-sat
0.5 M NaHCO3

7.2

0.39

100

193
(0.1 mm foil)

0.08

88

Au-NWs

CO2-sat
0.5 M KHCO3

7.2

0.32

90

4.43

3.05

76

Au nanoneedle

CO2-sat
0.5 M KHCO3

7.4

0.24

95

N/A

N/A

94

Bilayer Au/PE

CO2-sat
0.5 M KHCO3
flow cell

7.2

0.39

85

0.15

85

97

Au-NPs

2M KOH
flow cell

13.
77

0.2

70

0.18

58.33

98

Polycrystalline
Ag

CO2-sat
0.1 M KHCO3

6.8

1.4

30

105
(0.1 mm foil)

0.04

99

Polycrystalline
Ag

CO2-sat
0.5 M KHCO3

7.0

1.04

60

N/A
(foil)

N/A

100

5 nm Ag/C

CO2-sat
0.5 M KHCO3

7.0

0.84

40

0.09

66.67

100

Ag-NPs

CO2-sat
0.1 M KHCO3

6.8

0.84

83

N/A

N/A

101

cysteaminecapped AgNPs

CO2-sat
0.5 M NaHCO3

7.2

0.69

66

0.08

123.75

102

Ag
Nano-coarals

CO2-sat
0.1 M KHCO3

6.8

0.61

95

N/A
(on foil)

N/A

103

58
Nanoporous
Ag

CO2-sat
0.5 M KHCO3

7.2

0.49

92

40

0.35

104

Ag/PTFE

1 M KHCO3
flow cell

8.5

0.64

70

0.52
(eq. 500 nm film)

20.19

105

Ag-NPs

1M KOH
flow cell

14

0.39

95

14.25

106

Ag/PTFE

1M KOH
flow cell

14

0.34

90

0.52
(eq. 500 nm film)

25.96

105

Ag-NPs

0.5M KOH
flow cell

13.
23

0.34

95

7.125

107

Ag-NPs

1M KOH
flow cell

14

0.49

99

0.12

124

This
work

Bulk aqueous electrolyte cells can exhibit high catalyst overpotentials due to the limited
solubility of CO2 (33.4 mM) in the electrolyte, a limited pH operating range of ~6 - 10, and slow
ionic transport in the solution. In contrast, gas diffusion electrode (GDE) assemblies do not suffer
these same restrictions.97,107-111 In a GDE using 1 atm CO2 vapor, CO2 is transported in the vapor
phase and reacts at a thin (<100 nm) solid-liquid-gas phase interface. In this configuration, liquidstate concentration and diffusion do not limit the conversion rate, resulting in lower overpotentials
and higher current densities for CO2 reduction.108 Simulations have also shown that a cell using a
thin (10 nm) layer of electrolyte on the catalysts (wetted catalyst) outperform cells with either a
completely dry or a completely flooded catalyst configuration.112 These insights have led to the
development of gas diffusion electrodes113 and membrane electrode assemblies (MEA) 114 with a
humidified gas supply to facilitate ion conduction and water balance.
Although membrane electrode assemblies systems are more scalable, they often suffer from
short-term stability due to salt precipitation or membrane dehydration at high current densities.90
Hence, we chose to work with an aqueous GDE cell configuration. In this work, we employ a triple
junction photovoltaic (PV) device directly coupled with a gas diffusion electrode (GDE) as the
first demonstration of an electrolyte flow type PV-GDE reactor that provides both high selectivity
and long-term stability. For a directly driven PV-GDE system, the power generated by the PV is
directly supplied to the GDE. In our device, the areas of the PV photo-absorber (APV) and GDE
(AGDE) were both 0.31 cm2. To match the lower current density of the PV cell with the operating

59
conditions of the anode, a relatively low catalyst loading of GDE was chosen. A Ag nanoparticle
catalyst was used owing to its relatively high activity and relatively low cost.

4.2 Experimental Method
GDEs were prepared with diluted Ag-NPs (Sigma Aldrich 736481, particle diameter ≤50 nm,
30-35 wt. % in triethylene glycol monoethyl ether) by drop-casting on carbon paper
(Sigracet 29 BC). 50 μL of Ag-NPs ink was diluted with 15 mL of methanol and sonicated prior
to use. 200 μL of the diluted Ag-NPs were drop-casted on 25 mm by 25 mm size carbon paper
(masked to AGDE = 0.31 cm2 later for operation). The GDE was baked at ~100 °C for 10 min on a
hot plate to remove the remaining solvent and was post annealed at 200 oC for 1 h in a muffle
furnace in air. Surface chemical analysis was conducted using X-ray photoelectron spectroscopy
(XPS) with the catalyst primarily in a metallic phase.
A PEEK compression cell (Figure 4.1, Figure 4.2) was used as the vessel for the measurement
with anode and cathode chamber volumes of 2 mL. The anode and cathode electrode working
areas were 0.31 cm2, and the membrane area was 2.4 cm2 as constrained by the design of the
compression cell. 1 M potassium hydroxide (KOH) was used as the catholyte for experiments at
pH 14, while 1 M potassium bicarbonate (KHCO3) buffer was used for pH 8.5. The anolyte was 1
M KOH. The corresponding anion exchange membrane (AEM) was a Fumasep FAA-3-50 for
alkaline environments and a Selemion AMV for neutral environments. The anode was a Pt foil
under three-electrode operation, while for full cell operation in KOH electrolyte we used Ni foam
to reduce the overpotential for oxygen evolution reaction. A leakless Ag/AgCl reference electrode
was used for three-electrode measurements and to determine the GDE potential in the twoelectrode measurements. All electrochemical measurements were performed using a Biologic
VSP-300 potentiostat. Scan rates were set to 50 mV×s-1. A Keithley 2000 multimeter was used to
record the cell voltage and a Keithley 2182A nanovoltmeter for recording the voltage between the
cathode and reference electrode. Gas flow rates of the flow controller (Gas inlet) and flow meter
(Gas outlet) were recorded by the external device inputs of the potentiostat. Gas was delivered to
the GDE through an interdigitated electrode flow field against which the GDE is compressed to

60
maximize the interaction of CO2 with the catalyst and gas utilization.91 Current to the GDE was
supplied through the interdigitated electrode to the Ag-NP/carbon paper substrate.

Figure 4.1: Cell configuration composed of 1 NiOx or Pt anode, 2 Ag-NPs on Sigracet 29BC
carbon paper cathode, 3 anion exchange membrane, 4 CO2 gas inlet and CO/CO2 outlet, 5 Acrylic
backplate, 6 catholyte chamber, 7 anolyte chamber, 8 reference electrode. Black arrows indicate
the gas flow, and white arrows indicate the electrolyte flow. Note that the backplate, 5, is designed
to use an interdigitated wire electrode flow field to enhance the interaction between gas and
catalysts and improve CO2 utilization (see also Figure 4.2).

Figure 4.2: Backplate as shown in Figure 4.1 item 5 with an interdigitated flow field.

61
The electrochemical setup was operated in a continuous flow mode. Carbon dioxide was
provided to the electrochemical cell and its flow rate was controlled with an Alicat flow controller.
The carbon dioxide stream could be supplied either as dry gas or humidified CO2 with a gas
bubbler between the cell and flow controller. The exhaust gasses went through a liquid trap than
an Alicat flow meter, and finally to a gas chromatograph (SRI-8610) using a Hayesep D column
and a Molsieve 5A column with N2 as the carrier gas. The gaseous products were detected using a
thermal conductivity detector (TCD) and a flame ionization detector (FID) equipped with a
methanizer. Quantitative analysis of gaseous products was based on calibration with several gas
standards over many orders of magnitude in concentration.
The GaInP/GaInAs/Ge triple junction cell is commercially available from Spectrolab (C4MJ)
with a geometric area of 0.31 cm2. For illumination during laboratory tests, an Oriel Instruments
75 W Solar Simulator was used and matched with AM 1.5G. The response to natural sunlight of
the triple junction at short circuit was calibrated by measuring the outdoor sunlight irradiance with
a calibrated Si photodiode. The light intensity of the solar simulator was set to provide the same
short circuit current from the GaInP/GaInAs/Ge triple junction cell as it would under AM 1.5G
outdoor sunlight. While this is not expected to yield a simulated solar irradiance of 100 mW×cm-2
due to the different solar irradiance in the 800–1000 and 1150–1800 nm regions, it does produce
a response of the triple junction PV that is the same as it would be in actual AM 1.5G sunlight.
For outdoor tests, the triple junction solar cell was mounted on a solar tracker, see illustration
in Figure 4.3. An Arduino microcontroller was used to control the solar tracker and measure the
sun light intensity through a calibrated (350 to 1100 nm, 1 cm2) NIST traceable Si photodiode
(Thorlabs FDS1010-CAL).

62

Figure 4.3 (a) Illustration of the solar tracker. (b) With the addition of C = 3.25 Suns solar
concentrator. The PV element is located in the left with a silicon reference photodiode mounted on
the right. Above the PV element is the light-dependent resistor sensor array for determining and
tracking the position of the sun. For concentrator operation, a Fresnel lens with 51 mm focal
length was placed in front of the solar cell to provide a concentration of 3.25x.

Kratos Axis Ultra was used to perform X-ray photoelectron spectroscopy (XPS)
measurements with base pressure under 1x10-9 Torr. A monochromatic Al Kα (ħω = 1486.69 eV)
source with a power of 150 W was used for all measurements. A ramé-hart contact angle
goniometer was used for surface angle measurement. The images were analyzed with ImageJ with
the help of the Drop Analysis plugin developed at the École polytechnique fédérale de Lausanne
(EPFL).

4.3 Results and Discussion
The gas diffusion electrode catalytic performance was evaluated with the compression flow
cell. Dilute silver nanoparticles (Ag-NPs) with diameters of ≤50 nm were drop cast onto the
microporous side of the GDE substrate (Sigracet 29BC). The loading of Ag-NPs in this work was
measured to be 0.12 mg⋅cm-2. Scanning electron microscopy (SEM) images of the microporous
layer with and without Ag-NPs are shown in Figure 4.4.

63

Figure 4.4: Scanning electron microscopy images of carbon paper without (top) and with (bottom)
Ag-NP catalyst, secondary electrons image (left row) backscattered electrons image (right row).
(b)Illustration of the reverse-assembled GDE cathode cross-section with wetted catalyst and
operation for CO2 reduction.

An issue for aqueous GDEs is flooding or saturation of the porous catalyst layer with
electrolyte or water during operation. This results in a thick (>1 μm) electrolyte layer and a
diffusion-limited supply of CO2 to the electrode.115 To maintain the catalyst in a wetted but not
flooded condition that minimizes losses of CO2 to the electrolyte and extends the operational
lifetime, we assembled our aqueous GDE in a nontraditional manner with the catalyst coating of
Ag-NPs facing away from the electrolyte and towards the CO2 gas supply. We denoted this
configuration as a reverse-assembled GDE. The microporous layer of the GDE was treated with
polytetrafluoroethylene (PTFE), which helped to prevent flooding. Needle valves in the gas and
liquid output streams allowed separation of the liquid and gas phases as well as control of the
pressure difference between the aqueous electrolyte and the CO2 stream. Contact angle analysis
indicated that the Ag-NP coated surface was significantly less hydrophobic than the surface
without Ag-NPs. Contact angle and optical microscope images of the GDE are shown in Figure
4.5.

64

Figure 4.5: Contact angle Q measurement of water on (a) pristine Sigracet 29 BC carbon paper
and (b) with Ag-NPs on Sigracet 29BC carbon paper after electrolysis. The contact angle is 175°
for (a) and 105° for (b). Optical micrographs of water pushing through the back of the
Sigracet 29 BC carbon paper (c) without Ag-NPs and (d) with Ag-NPs. The formation of small
liquid bubbles is observed in (c) while a thin water layer is shown in (d) indicating the catalyst
surface is wetted during operation as proposed.

With both the gas inlet and outlet on the same side of the GDE, the device operates in a “flowby” GDE configuration. The Ag-NP catalyst side of the electrode was facing the CO2 gas channel
as illustrated in Figure 4.6. This orientation of the Ag-NPs maintained a thin electrolyte layer on
the catalyst and enhanced the rate of CO2 reduction.112 The turnover frequency of the Ag-NP
catalyst for the reverse-assembled GDE at -0.6 V vs RHE was calculated as ~9 × 103 h-1. Turnover
frequency (TOF) was defined as the CO production rate (in moles cm-1 h-1) divided by the number
of moles of active site catalyst. Consider the 0.26 mmol⋅h-1⋅cm-2 (7.4 mg⋅h-1⋅cm-2) CO production
rate per catalyst area at -0.6 V vs RHE. The total catalyst loading was 0.001 mmol⋅cm-2 (0.12
mg⋅cm-2) that gave a TOF based on the total amount of Ag-NPs as 260 h-1. Since only the surface
atoms of the nanoparticle can contribute to active sites, we estimate the fraction of surface atoms

65
of a 50 nm diameter Ag nanoparticle to be ~3 %, then the moles of active sites are ~3 × 10-5
mmol cm-2, and the TOF based on the number of surface atoms is ~9 × 103 h-1.

Figure 4.6: Illustration of the reverse-assembled GDE cathode cross-section with wetted catalyst
and operation for CO2 reduction.

The anode was made from either Pt or an electrochemically activated Ni foam for three- and
two-electrode measurements, respectively. An aqueous catholyte of 1 M aqueous potassium
bicarbonate (KHCO3) or potassium hydroxide (KOH) was used under near neutral or basic
conditions, respectively. In all cases, 1M KOH was the anolyte. The anion exchange membrane
(AEM) was Selemion for neutral environment or Fumasep FAA-3-50 for alkaline environment.
Electrolyte (500 ml) was continuously pumped through the cathode chamber in a closed loop at a
rate of 2 mL/min. A change of pH (from 14 to 13.7) was observed for the 1 M KOH catholyte after
150 h of continuous operation, corresponding to irreversible loss of 0.25 mol KOH (50% of the
electrolyte). The volume of the catholyte was 0.5 L 1 M KOH (0.5 mol KOH) with an initial pH
of 14 which changed to 13.7 after 150 h of continuous operation. A pH of 13.7 is
10KO.øGKN M KOH = 0.5 M KOH which for 0.5 L is 0.25 mol KOH. The reaction of KOH and CO2
is given:
2KOH + CO2 = K2CO3 + H2O (Equation 4.1).

66
A loss of 0.25 mol KOH corresponds to a loss of 0.125 mol CO2 and formation of an equal
amount of (0.125 mol) K2CO3. The CO2 flowrate during the experiment was 10 sccm, which over
150 h corresponds to 4.043 mol CO2. The total percentage of CO2 lost to KOH neutralization and
ö.KIQ

carbonate formation is then N.öNO = 0.031 ≡ 3.1 %. Further improvement to reduce CO2 loss or
regenerate the electrolyte would be necessary for fully sustainable operation. The neutralized
carbonate electrolyte can possibly be utilized in carbonate-to-syngas system to compensate the
loss of CO2 in a gas-fed MEA cell with bipolar membrane. 116
Results from three-electrode measurements for reverse- and standard-assembled GDEs are
shown in Figure 4.7(a-b), respectively, for 1 M KHCO3 (bulk pH of 8.5) and 1 M KOH (bulk pH
of 14). Current densities are substantially lower than earlier reported GDE devices due to the low
catalyst loading used to match the current from the PV (current matching). For the reverseassembled GDE, both the Faradaic efficiency (fFE,CO) for CO and current density (JGDE) increased
with increasing potential with fFE,CO close to 100 % at -0.6 V vs. RHE in 1 M KOH. Similar trends
of current density and Faradaic efficiency versus applied potential were found for the standardassembled GDE.

Figure 4.7: Dark catalysis three-electrode measurement of Ag-NPs GDE. Faradaic efficiency
versus GDE potential operated in 1 M KHCO3 (left half of graph) or 1 M KOH (right half of graph)
of (a) the reserve-assembled Ag-NP GDE and (b) a standard-assembled Ag-NP GDE.

67
To compare the activity of the Ag-NPs in different orientations and pH, overpotential
analysis for CO2 reduction to CO was preformed, Figure 4.8. The overpotential of the GDE is
defined as the difference of applied potential to the thermodynamic potential of CO2 to CO on the
RHE scale (–0.11 V vs RHE). The comparable Tafel slopes (~0.23V/dec) in KHCO3 and KOH for
either orientation indicate a similar catalytic pathway regardless of the operating conditions. The
Tafel behavior plotted with potentials vs NHE falls on a rough single line (Figure 4.9) and suggests
that the rate-determining step for the reduction on our Ag-NP GDE is not proton limited. The
achievable current density and Faradaic efficiency (fFE,CO) for CO are higher in 1 M KOH than in
1M KHCO3 at the same overpotential, likely due to a pH independent rate determining step. All
subsequent measurements were, therefore, performed using 1 M KOH for the PV-GDE integrated
device.

Figure 4.8: Overpotential versus CO partial current of Ag-NPs GDE for CO2 reduction to CO.
Overpotential=¡#¢¬±,ìÖ± + 0.11.¡, /∏√ ≡ /¢¬± × 4W±,∏√ .

68

Figure 4.9: (a) GDE potential vs. NHE, (b) GDE potential vs. RHE versus CO partial current of
Ag-NP GDE (rev. indicating reserve-assembled, std. indicating standard-assembled) for CO2
reduction to CO in 1 M KHCO3 and 1 M KOH.

Figure 4.10 shows the Faradic efficiency for CO vs. time at -0.6 V vs. RHE for the two GDE
orientations in KOH. For the standard configuration, the fFE,CO decreasing to ~75% after 1 h and
to 50 % after 2 h, while for the reverse configuration, the fFE,CO was ~97% for 3 h. Though similar
in initial current density and fFE,CO, the standard assembly, with the Ag-NP catalyst facing the
electrolyte, became flooded during the first hour of operation resulting in a reduction of the Faradic
efficiency. Surface chemical analysis was conducted using X-ray photoelectron spectroscopy
(XPS) with the catalyst primarily in a metallic phase, see Figure 4.11. No obvious changes were
observed other than the absorption of potassium after operation.

69

Figure 4.10: Stability of reserve-assembled and standard-assembled Ag-NPs GDE operated at 0.6 V vs. RHE in 1 M KOH.

Figure 4.11: Silver 3d (Ag 3d) and carbon 1s (C 1s) X-ray photoelectron spectra of Ag-NP GDE
before/after electrocatalysis with an electrolyte of (a,c) 1 M KHCO3; (b,d) 1 M KOH.

70
We performed two-electrode measurements for the GDE using an electrochemically
activated nickel foam anode coupled to the GaInP/GaInAs/Ge triple junction cell. The solid-state
J-V characteristic and performance parameters of the solar cell are shown in Figure 4.12. For
illumination during laboratory tests, an Oriel Instruments 75 W Solar Simulator was used, the lamp
spectrum matching with AM 1.5G is presented in Figure 4.13. The corresponding sub-cell currents
with integration of external quantum efficiency and short circuit current over the two illumination
spectra are shown in Table 4.2.

Figure 4.12: J-V characteristic of the GaInP/GaInAs/Ge triple junction cell. Uoc is the open circuit
voltage, Jsc the short circuit current, Ump/Jmp the current and voltage at the maximum power point,
and FF the fill factor.

71

Figure 4.13: Intensity (left axis) of AM 1.5G 1 sun reference spectrum (gold) and solar simulator
spectrum (black), external quantum efficiency (right axis) of the GaInP/GaInAs/Ge (blue, green,
red) triple junction cell.

Table 4.2: Currents calculated for the individual sub-cells of the of the GaInP/GaInAs/Ge triple
junction PV cell under 1.5G 1 sun illumination assuming the standard reference sunlight spectrum
(AM1.5G ASTM G-173 reference spectrum was taken from the Renewable Resource Data Center
(RReDC) of the National Renewable Energy Laboratory (NREL)) or the solar simulator spectrum
and measured short circuit photocurrent Jsc under respective 1 sun conditions.

Illumination
Source

JGaInP

JGaInAs

JGe

Jsc

(mA×cm-2)

(mA×cm-2)

(mA×cm-2)

(mA×cm-2)

AM 1.5G

15.56

14.83

18.53

14.6

Solar
simulator

15.32

29.73

14.83

14.6

72
A schematic of the cell is shown in Figure 4.14(a) with 1M KOH as electrolyte using a
Fumasep FAA-3-50 membrane. Both the cell potential (Ucell) and the cathode to reference
electrode potential (UGDE) were monitored during the operation. We calculated the solar to fuel
efficiency (hSTF) for CO2 reduction using equation below.

!UVW = YZ[\ = ƒ≈{
]^

∙abxy^ ∙fz{,∆« ∙oƒ≈{
iu]|}\ ∙opq

r∙abcde ∙fgh,á».
ijklmn

(Equation 4.2)

Where ΔUrxn is the thermodynamic potential difference between the oxygen evolution half reaction
(OER) and the CO2 reduction half reaction of 1.34 V, A is the area of the GDE or PV with
AGDE = APV = 0.31 cm2, J (= JGDE = JPV) is the operation current density of the system, and Plight is
the incident light irradiance (mW×cm-2) on the photovoltaic. The energy efficiency for the GDE
cell (hGDE) was defined as follows:

!… À = YÅÇn =
ke

abxy^ ∙rƒ≈{ ∙oƒ≈{ ∙fgh,á».
bÉÑjj ∙rpq ∙opq

abxy^ ∙fgh,á».
bÉÑjj

(Equation 4.3)

where JGDE×AGDE = JPV×APV , and Ucell is the total operating voltage of the cell.

73

Figure 4.14: Light driven PV-GDE measurement (APV = AGDE = 0.31 cm2). (a) Illustration of wire
connection between the triple junction cell and GDE cell. (b) J-U characteristic of Ni anode, solar
cell with Ni anode, and Ag-NP gas diffusion cathode under 1 Sun. (c) Current, GDE potential vs
RHE, and cell voltage measurement over 20 h duration. (d) The corresponding CO Faradaic
efficiency and solar to fuel efficiency over the same 20 h duration.

To evaluate efficiency and stability, we measured cell parameters using simulated AM 1.5G
sun illumination at 1 Sun in the laboratory, as shown in Figure 4.14(b-d). The blue curve in Figure
4.14(b) represents the performance of the electrochemically activated Ni foam anode alone, while
the yellow curve indicates the behavior of PV plus anode. The red curve shows the catalytic current
of the Ag-NPs GDE. The intersection between the red and yellow curves defines the operation
point, located at -0.6 V vs. RHE and 14.4 mA×cm-2 with a cell voltage of 2.23 V. Figure 4.14(c-d)
illustrates the cell performance over 20 hours with an average Faradic efficiency for CO of

74
99 ± 2 % and an average CO production rate of 2.3 mg/h. No degradation in performance was
observed. From the experimental results, we calculated the average solar to CO efficiency for the
20 h operation as 19.1 ± 0.2 %, with an average energy efficiency hGDE of 59.4 ± 0.6 %. The error
bars were obtained as the variation within the 20 h of operation.
The solar to CO efficiency of 19.1 % represents a new record efficiency. A performance
comparison with the current state-of-the-art PV-electrolyzer for CO2 reduction to CO is shown in
Table 4.3. The PV-GDE device had a CO production rate per projected cathode area 50 times
higher than for the bulk electrolyte device (7.4 mg×h-1×cm-2 versus 0.145 mg×h-1×cm-2) with greatly
improved stability (20 h with no degradation versus 15 % loss in 5 h).80 A similar PV-GDE device
operated under 3.25 Suns illumination with AGDE = 1 cm2, APV = 0.31 cm2, (3.25
≈ AGDE/APV)
showed over 150 hours of stability, with an average Faradic efficiency of 96 ± 2 %, an average
solar to CO efficiency of 18.9 ± 0.5 %, and an average energy efficiency hGDE of 53.7 ± 1.2 %, as
in Figure 4.15.

75
Table 4.3: Comparison of the performance of the PV-GDE studied herein with the current state
of the art PV-electrolyzer for CO2 reduction to CO.80

Current record 80

This work

Solar-to-CO (%)

13.4

19.1

PV size (cm2)

0.563

0.31

Cathode / Anode

both SnO2 / CuO

Ag GDE / Ni foil

Cathode & Anode
size (cm2)

20

0.31

Catholyte / Membrane /
Anolyte

0.25 M CsOH / BPM /
CO2-sat 0.1 M
CsHCO3

1 M KOH /AEM /
1 M KOH

Operation current (mA)

6.6

4.5

fFE,CO (%)

86

99

CO production rate
(mg×h-1×cm-2)

0.145

7.4

CO production rate
outdoor (mg/day)

15
(50 at 3.25 Suns)

5 (with 15% loss)

20
(150 h at 3.25 Suns)

Stability (h)

Figure 4.15: Efficiency and stability assessment at a solar concentration 3.25 Suns. (C = 3.25,
AGDE = 1 cm2, APV = 0.31 cm2) (a) J-U characteristic of Ni anode, solar cell with Ni anode, and
Ag-NP gas diffusion cathode under 3.25 Suns. (b) Current and cell voltage measurement over

76
150 h duration. (c) The corresponding CO Faradaic efficiency and solar to fuel efficiency over
the same 150 h duration.

Full day outdoor tests were conducted with online gas product analysis in order to obtain the
solar to fuel efficiency over the entire day. The triple junction cell and a calibrated silicon
photodiode were mounted on a solar tracker to maintain optimum orientation toward the Sun.
Results are shown in Figure 4.16. The dips in sun intensity at 7:00am - 9:00am and 4:00 - 6:00
p.m. in the data were the result of trees blocking the sunlight. The system operated at a cell voltage
of 2.20 V and GDE potential of -0.57 V vs. RHE under natural full sun illumination. A Faradaic
efficiency of 96 ± 8 % and solar to fuel conversion efficiency of 18.7 ± 1.7 % was observed over
an optimal 6 h period within the day. The diurnal-averaged solar to fuel conversion efficiency was
5.8 %. The CO production rate for one day under actual outdoor sun conditions was calculated to
be 15 mg/day of CO. Another outdoor demonstration used a lens to concentrate the sunlight
producing an irradiance of 3.25 Suns (C = 3.25, AGDE = 1 cm2, APV = 0.31 cm2) with data included
in Figure 4.17 with a CO generation rate of 50 mg/day. Using this calculated rate, a system scale
up to 1 m2 would result in a CO production rate of 0.5 kg/day.

77

Figure

4.16:

Outdoor

assessments

of

solar

driven

PV-GDE

in

Pasadena,

CA

(APV = AGDE = 0.31 cm2). The solar irradiance was monitored with a calibrated silicon photodiode.
Operation current density J (= JGDE = JPV), cell voltage Ucell, GDE potential UGDE vs. RHE, CO
Faradaic efficiency fFE,CO, and solar to fuel efficiency hSTF were recorded for a 24h day cycle.

78

Figure 4.17: Outdoor tests of solar-driven PV-GDE in Pasadena, CA. The solar irradiance was
monitored with a calibrated silicon photodiode. PV operation current JPV, cell voltage Ucell,
working electrode potential UGDE, CO Faradaic efficiency fFE,CO, and solar to fuel efficiency hSTF
were recorded for a 24h day cycle with 3.25x solar concentrator (C = 3.25, AGDE = 1 cm2,
APV = 0.31 cm2).

The performance of our directly coupled PV-GDE device was compared to a DC-DC
converter coupled PV and GDE with power-matching electronics. We simulate DC-DC converter
output curves with the input of our solid-state PV curve as shown in Figure 4.18. Though the
DC-DC converter can track the maximum power point (MPP) of the PV, a practical loss of 5-10 %
is expected.117 The operating point for the directly driven PV-GDE cell is Ucell = 2.23 V,
J = 14.4 mA×cm-2 with a maximum efficiency of 19.3 %. With a 95 % efficient DC-DC converter,
the operation point would be Ucell = 2.22 V, J = 13.8 mA×cm-2 with a maximum efficiency of
18.5 %. For a 90 % efficient DC-DC converter, the operation point would be Ucell = 2.20 V,
J = 13.2 mA×cm-2 with a maximum efficiency of 17.7 %. The maximum efficiencies are calculated

79
assuming 100 % CO Faradic efficiency. The slightly higher efficiency of our directly driven
PV-GDE device, compared to the same setup with integrated DC-DC converter and power
matching electronics, reveals the potential of developing a directly coupled PV-GDE device with
its reduced complexity. All the experimental results (including simulated/natural sun light,
with/without concentrator) and calculated systems in this work are summarized in Table 4.4.

Figure 4.18: J-U characteristic of the GaInP/GaInAs/Ge triple junction cell under 1 Sun (yellow
solid line) with combined load curve of Ni anode and Ag GDE cathode (red dot-dashed line) in
addition to DC-DC converter output curves (solid PV curve as input) with converter efficiency of
90 % (black dashed line) and 95 % (black dotted line). The PV curves for lower illumination
conditions are included on the right figure.

80
Table 4.4: Comparison of the PV-GDE performance studied herein with different measurement
conditions and calculations.

APV AGDE
(cm2) (cm2)

fFE,CO
(%)

hGDE hSTF
(%) (%)

Simulated

0.31

0.31

99

59.4

19.1

14.4

2.23

32.1

325

Simulated

0.31

96

53.7

18.9

47.5

2.39

113.5

91

Natural

0.31

0.31

96

58.5

18.7

13.0

2.20

28.6

296

Natural

0.31

96

55.9

18.9

43.0

2.30

98.9

100

Calculated

0.31

0.31

100

60.0

19.3

14.4

2.23

32.1

100

Calculated

0.31

0.31

100

60.4

18.5

13.8

2.22

30.6

100

Calculated

0.31

0.31

100

60.9

17.7

13.2

2.20

29.0

Irradiance
(mW×cm-2)

Source

100

JPV
(mA×cm-2)

Ucell
Power
(V) (mW×cm-2)

4.4 Conclusion and Outlook
In summary, we have demonstrated a highly efficient solar-driven CO2 reduction device for
CO generation using a flow-by reverse-assembled gas diffusion electrode cell directly coupled to
a triple junction solar cell. The reverse-assembled GDE is designed to minimize parasitic CO2
losses, utilizing a high CO2 concentration and low overpotential catalysts for the CO2 reduction
reaction. The Ag-NPs based catalysts exhibited near unity Faradic efficiency towards CO
generation at approximately -0.6 V vs. RHE in 1 M KOH electrolyte. The PV-GDE system was
evaluated under both laboratory AM 1.5G simulated solar irradiation and outdoor real sun
conditions. Near-unity Faradic efficiency was observed for CO2-to-CO conversion and an average
solar-to-CO energy efficiency of 19.1 % was achieved with AM 1.5G illumination at 1 Sun,
leading to over 50 times higher CO production rate per catalyst area than the current record

81
photovoltaic-driven electrolysis device. The GDE was demonstrated to be stable for over 150
hours without degradation, supporting our hypothesis that, by using a reverse-assembled GDE
device configuration with the catalyst layer facing towards the CO2 gas supply, we could extend
the system operation time without suffering a transition from a wetted to a flooded gas diffusion
layer. Under outdoor sun conditions, the PV-GDE system exhibited a solar to CO conversion
efficiency of 18.7 % during noontime, and yielded a CO production rate of 15 mg×cm-2 per day.
This reverse-assembled PV-GDE establishes a new efficiency record for directly solar-driven CO2
reduction, and offers an example of a very high efficiency, stable device for solar CO2 conversion.
Preventing CO2 neutralization of the basic electrolyte for sustainable operation would be the
practical direction to pursue. It is proven that by catalyst engineering, near unity Faradaic
efficiency toward CO production under low overpotential in bicarbonate electrolyte can be
achieved. Possible strategies involve introducing defects or surface ligand, facet or morphology
control, oxidation state manipulation, and utilizing co-catalysts or alloy catalysts. Direct reduction
from bicarbonate electrolyte eliminates the high energy input of normal CO2 capture process, and
can possibly enhance the CO2 utilization rate.

82
CHAPTER 5

Broadband Transmission TiO2 Nanocone
5.1 Introduction
High efficiency conversion of solar energy to electricity or fuels using photovoltaic (PV) 118122

or photoelectrochemical (PEC) 65,69,123 cells requires optimization of the broadband absorption

of sunlight. Anti-reflective coatings (ARCs) 124, surface textures, 125-128 and high-index lighttrapping structures129,130 are among the many strategies that have been shown to increase
broadband absorption relative to unmodified planar light absorbers. PECs and multijunction PVs
generally require front contacts and/or electrocatalytic films that substantially reflect or absorb
light, thereby reducing their photocurrent densities. 131 In integrated PECs used to affect solardriven water splitting or CO2 reduction, 6,78,82 the front contact is made to an electrolyte, and a
catalyst located in the optical path increases the efficiency of the cell by reducing the kinetic barrier
for the electrochemical half-reaction occurring at the top contact. Depending on the orientation of
the design, as well as the chemical inputs and desired products, the catalyst may be optically
opaque (such as CoP for water reduction or Cu for CO2 reduction) or may be electrochromic (such
as NiFeOx for water oxidation). 132 Although an all-back-contact design133,134 can prevent contact
shading losses in crystalline Si solar cells, such a design is not compatible with all of the
solid/liquid interfaces either with integrated PECs for highly efficient fuel production or with
multijunction PV device structures. 135
Nanostructuring the semiconductor is one approach that has been developed to
enhance broadband absorption. Nanowires, 43,56,136 inverted pyramids, 137 nanodomes, 138
nanoshells, 139 nanopillars, 140,141 and nanocones142 have been explored for use in many
optoelectronic devices. For example, for wavelengths ranging between 400 – 1100 nm and anglesof-incidence between 0° – 50°, Si microcone arrays exhibit nearly perfect angularly and spectrally
averaged reflectivity (< 1 %) as well as high (89.1 %) absorption. 143 In Chapter 2, we also
demonstrate sparse arrays of InP nanocones exhibiting angle-insensitive, near-unity (>90%),
broadband (450−900 nm) optical absorption. 144 Cones inherently possess a continuous range of
radii that provides a range of waveguide modes accessible for coupling with incident light. 23
Hence, cones are intrinsically favorable structures for enabling enhanced broadband absorption.

83
Moreover, the radius at the base of the cone and the radius of any truncation can be chosen
specifically to select a spectral range of interest.
Alternatively, broadband absorption can be enhanced by decorating the light-facing surface
of the semiconductor with dielectric nanostructures, such as nanospheres, 145 that serve as
waveguides. This approach requires a high-index dielectric that can be deposited on planar
surfaces using scalable methods. For integrated PEC devices, the dielectric also must be stable in
the chosen aqueous electrochemical environment. TiO2 has been used in waveguides for nearvisible and telecommunication wavelengths, 146-151 and has a higher index of refraction (n~ 2.5)
than many other dielectric materials commonly used in solar photovoltaic devices, including SiO2
(n = 1.5) and Al2O3 (n = 1.77). TiO2 is relatively inert electrochemically and has a wide band gap
that allows transmission of incident solar illumination, and has been utilized extensively as a
protective coating in efficient PEC devices. 6,74,78,82,152-156 TiO2 is therefore a promising candidate
material for nanostructured waveguides in PEC devices.
Devices that make use of nanostructured surfaces nevertheless require front contacts, either
to a conductor for PVs or to a catalyst for PECs, but the front contact can block light. In devices
that require connections between external circuits and nanostructured optical surfaces, front
contacts typically are formed by coating the surface with a transparent and conductive material,
such as indium tin oxide (ITO). However, efficient integrated PECs for fuel production
additionally require catalysts on at least one of the optical surfaces. Very high (> 90%) absorption
and high front-surface conductivity have been demonstrated using both simulation and experiment
in a device consisting of SiNx-coated Si nanopillars protruding from a crystalline Si substrate
coated with an opaque Au front-side contact that covered 65% of the Si surface. 141 Although the
selective etching process used in this approach limits its applicability to a few specific interfaces
such as Au/Si, this work shows that nanostructures can direct light around opaque metallic front
contacts deposited onto the optical surfaces of PV devices, and suggests a strategy for guiding light
around catalyst layers in PECs.
Considered together, the prior work in the areas of antireflective nanostructures and dielectric
waveguides underscores the potential value and developmental feasibility of modular
antireflective coatings that promote broadband absorption over a spectral range above the

84
photoelectrode bandgap without requiring modification of the underlying semiconductor or
contact interfaces. Herein, we combine simulations and proof-of-concept experiments to evaluate
and demonstrate light management by an array of TiO2 nanocones placed on the surface of a p+nSi photoanode with a metallic Ni contact covering the exposed Si surface.

5.2 Numerical and Experimental Method
3D full-field electromagnetic wave finite-difference time-domain (FDTD) simulations of
TiO2 nanocone arrays with or without Ni films, with hexagonal arrays of holes on Si, were
performed using a commercial software package, Lumerical FDTD. The nanocone arrays on Si
were constructed using the 3D simulation region with periodic boundary conditions along the xand y-axes, and infinite boundary conditions were rendered as perfectly matched layers along the
z-axis. Palik materials data were used for Si and Ni. Material data from the Ioffe Institute and
ellipsometry measurements (J.A. Woollam Co. model VASE) were both used for simulation of the
optical properties of the TiO2 nanocones. A plane-wave source of illumination was applied to
simulate the steady-state behavior of TiO2 nanocones with or without Ni arrays on the Si substrate.
Broadband simulations were performed in the 400 – 1100 nm spectral range for two orthogonal
polarizations. Transmission spectra for unpolarized light were obtained by averaging the
transmission spectra for the two orthogonal polarizations. These transmission spectra, along with
the standard AM 1.5G spectra, were used to calculate the fraction of the spectral photon flux that
was transmitted into the Si. The expected light-limited photocurrent density (Jph,max) can be
estimated by integration of the transmitted spectral photon flux with the corresponding wavelength.
Frequency-domain field and power monitors were applied in the simulation to produce the steadystate electric-field data for plots of electric-field profiles.
Czochralski-grown n-type Si wafers with <100> orientation and a resistivity of 0.1-1 ohmcm (Addison Engineering Inc.) were cleaned via a modified RCA standard clean 1 (5:1:1 by
volume of H2O:NH4OH:H2O2 at 70 °C), then 1 min immersions in 10% (v/v) HF, and followed by
an RCA standard clean 2 (6:1:1 H2O:HCl:H2O2 (v/v) at 70 °C). The cleaned n-Si samples were
thermally doped with boron using a BN-975 (Saint-Gobain) wafer at 950 °C for 30 min to produce
a p+n-Si homojunction. The doped p+n-Si samples were immersed in 10% (v/v) HF(aq) for 2 min,

85
oxidized in a tube furnace at 750 °C for 20 min, and then dipped in 10% (v/v) HF for 2 min to
remove any defective layers on the Si surface.
A layer of SiO2 (5 – 10 nm) was deposited onto the p+n Si wafers via electron-beam
evaporation, and 2.3 µm of TiO2 was then deposited onto the samples via electron-beam
evaporation. Electron-beam evaporation depletes the source of oxygen and results in TiO2 with
higher conductivity than perfectly stoichiometric TiO2, so the deposition was performed in 3-4
steps, refilling the TiO2 source between steps to maintain higher oxygen content in the film. The
TiO2-coated p+n-Si samples were spin-coated at 4000 rpm for 60 s with 495 PMMA A4 and baked
at 80 °C for 5 min, then spin-coated at 4000 rpm for 60 s with 950 PMMA A4 and baked again at
80 °C for 5 min to form a bilayer of positive tone resist to facilitate lift-off. The samples were
patterned with a hexagonal array of 100 nm diameter circles on a 700 nm pitch, using direct
electron-beam lithography (VISTEC electron-beam pattern generator (EBPG) 5000+) with an
acceleration voltage of 100 keV and a current of 5 nA. After electron-beam writing, the pattern
was developed in a 1:3 MIBK (methyl isobutyl ketone):isopropanol for 60 s at room temperature,
resulting in a hexagonal array of 100 nm diameter holes with a 700 nm pitch in the PMMA layers.
A 200 nm layer of Cr was evaporated over these samples (rate 1 A·s−1 at 10−6 Torr), and lift-off
was performed in acetone, leaving a hexagonal array of Cr that served as a hard mask for TiO2 dry
etching. Dry etching was then conducted using an Oxford Instruments Plasma Lab System 100
ICP-RIE by using SF6 / C4F8 etching chemistry, in which SF6 was the etching gas and C4F8 was
the passivating gas. Etching was performed at a capacitively coupled power of 150 W, an
inductively coupled power of 2500 W, a SF6 / C4F8 gas ratio of 23.5 sccm / 40 sccm, a chamber
pressure of 7 mTorr, and a table temperature of 0 °C for 15 min.
Electrodes were prepared from the n+p-Si samples with etched TiO2 cones by first cleaving
the samples to remove the edges, thus avoiding shorts due to doped layers. In-Ga eutectic was
applied on the back side of the samples to form an ohmic contact to the p+n-Si homojunction. Ag
paste (Ted Pella) was used to attach a Sn-plated Cu wire to the In-Ga on the back side of the sample.
The wire was run through a glass tube, and the samples were sealed to the glass tube using epoxy
(Loctite 9460) and annealed at 80 °C for ~ 6 h. The active area of the electrodes was determined
using a high-resolution image taken using a commercial scanner and image-processing software
(ImageJ). Typical electrode areas were ~ 0.04 cm2.

86
The samples were then dipped in buffered HF (40% NH4F to 49% HF volume ratio 6:1)
for 10 s to remove the remaining SiO2 between the TiO2 cones and to fully remove the Cr mask.
The Ni layer was subsequently electrodeposited on the areas of the surface of the p+n-Si between
the TiO2 nanocones using a commercially available Ni plating solution (Clean Earth Nickel Mirror,
Grobet USA) at a potential of -0.956 V vs Ag /AgCl (using a Biologic SP-200 potentiostat) until
~ 300 mC cm-2 of cathodic charge was passed. An illustration of the process flow for fabrication
of the desired structures is shown in Figure 5.1.

TiO2

SiO2

Si

SiO2
deposition

TiO2
deposition

Cr mask
patterning

TiO2

Dry etching

HF dip

Ni
electroplating

Figure 5.1: Process flow diagram for fabrication of Si photoanodes with TiO2 nanocones and Ni
catalysts.

An array of holes in a Ni layer on p+n-Si was fabricated as a comparing sample. Electronbeam evaporation was used to deposit a layer of Ni on p+n-Si samples. The samples were then
covered with ZEP 520A by spin-coating at 4000 rpm for 60 s and were baked at 180 °C for 3 min.
A hexagonal array of circles 500 nm in diameter with a 700 nm pitch was written onto the samples
by direct electron-beam lithography using an acceleration voltage of 100 keV and a current of 50
nA. The pattern was then developed by immersing the samples in a ZED N50 solution for 90 s.
The patterned resist was used as a mask for ICP-RIE etching of the Ni film. Etching was performed
using the following parameters: 4 mTorr, 600 W ICP forward power, 150 W RF forward power,
20 °C, 30 sccm Ar. The final removal of ZEP 520A was performed using remover PG.
A three-necked glass cell with a quartz window was used as a vessel for the
photoelectrochemical oxygen-evolution reaction (OER). The OER was performed in aqueous 1.0
M KOH (Sigma-Aldrich) using a three-electrode setup, with a saturated calomel electrode (SCE)
as the reference electrode, a carbon electrode as the counter electrode, and the p+n-Si sample with

87
TiO2 cones and Ni as the working electrode. Measurements were conducted under simulated
sunlight (Oriel Instruments Solar Simulator equipped with a 1000 W Mercury Xenon lamp
calibrated to 100 mW cm-2 (AM1.5) illumination using a Si photodiode). For current-density
versus voltage (J-V) measurements, the voltage was swept at a scan rate of 50 mV s-1 from -0.5 V
to 1.5 V vs SCE.

5.3 Results and Discussion
Figure 5.2 shows schematics for three device configurations that were compared using
simulations to understand the optical properties of TiO2 nanocones. In the first configuration,
Figure 5.2(a), TiO2 nanocones with a height of 2300 nm and a base radius of 250 nm were placed
on the Si substrate in a 2D hexagonal array with a 700 nm pitch. This arrangement left 54% of the
Si surface uncovered by TiO2 nanocones. In the second configuration, Figure 5.2(b), 50 nm of Ni
covered the area of Si that remained exposed in the first configuration. The third configuration,
Figure 5.2(c) was the same as the second, but the TiO2 nanocones were removed, leaving a
hexagonal array of circles of exposed Si in the Ni layer.

Figure 5.2: Schematics of the three configurations that were simulated. (a) TiO2 nanocones on Si
substrate; (b) Ni between the TiO2 nanocones on Si substrate; (c) Ni hole arrays on Si substrate.

Figure 5.3 compares the simulated transmission, absorption, and reflection spectra for the
TiO2 and Ni components of the three structures, either in air (n = 1.0) or in water (n = 1.33). Figure

88
5.3 (a,d) shows that the TiO2 nanocones neither absorb nor reflect substantially in the 400 –
1100 nm spectral range, with the nanocones allowing transmission of 97.5% of the total incident
photons in air or 96.9% in water. The planar TiO2 film simulations, with higher reflection losses,
are shown in Figure 5.4, leading to transmission of 74.7% in air and 84% in water. When 50 nm
of Ni was added into the spaces between the nanocones, the simulated transmitted photon flux was
reduced to 86.2% in air or 84.7% in water (Figure 5.3(b,e)). The minima in transmission primarily
result from absorption by Ni, and the wavelengths of the minima shift depending on the index of
refraction of the surrounding environment.

Ni hole array

TiO2 nanocones only

TiO2 nanocones with Ni

Ni hole array

TiO2 nanocones only

TiO2 nanocones with Ni

Figure 5.3: Simulated transmission (T), absorption (A), and reflection (R) spectra of the three
configurations of the TiO2 nanocone array and Ni layer in Figure 5.2. (a), (b), and (c) plot the
spectra in air for an array of TiO2 cones, a TiO2 cone array with Ni, and a Ni hole array,
respectively. (d), (e), and (f) plot the same structures but in water. The optical effects of the Si
substrate are not shown.

89

TiO2 film only

0.8

0.6

Water (n=1.33)

Air (n=1)

0.8

T (74.7 %)

0.4

0.2

(a)

400

TiO2 film only

0.6

T (84.0 %)

0.4

0.2

500

600

700

800

Wavelength (nm)

900

1000

1100

(b)

400

500

600

700

800

900

1000

1100

Wavelength (nm)

Figure 5.4: Simulated transmission (T), absorption (A), and reflection (R) spectra of the TiO2 film
with thickness of 2.3 µm. (a) plot the spectra in air (b) plot the spectra in water.

When the TiO2 nanocones were removed from the simulation, leaving just the hexagonal
array of circular holes in a 50 nm layer of Ni that covered 54% of the optical plane, the transmitted
photon fluxes were reduced to 23.3% in air or 24.8% in water, with reflection and parasitic
absorption accounting for ≥ 75% of the optical losses (Figure 5.3(c,f)). The size and pitch of the
holes in the Ni layer were particularly unfavorable for transmission of light through the Ni layer
in the absence of the TiO2 cones, because over a large fraction of the incident solar spectrum, the
diameter of the holes was less than the wavelength of the incident light. In the specific
configuration shown in Figure 5.2(b), the TiO2 nanocones minimized the interaction between the
light and the Ni layer, enabling > 3 times the amount of light to be transmitted than for the Ni hole
array that did not also contain the TiO2 cones. For the TiO2 nanocone array with Ni, the simulated
transmission at the minima was ≥ 60%, Figure 5.3(b,e), whereas for the Ni hole array without TiO2
nanocones, the simulated transmission of was 20-30%, Figure 5.3(c,f)). The simulations thus
indicated that incident light is expected to couple efficiently to the TiO2 nanocones that guide the
light around the Ni layer.
Figure 5.5 plots the transmitted photon flux for each structure along with the Air Mass (AM)
1.5G solar spectrum. Using the transmitted spectral photon flux, the maximum photocurrent
densities, Jph,max, in air estimated from the simulations for a Si solar cell covered by either the TiO2
nanocone array, the TiO2 nanocone array with Ni or the Ni hole array were Jph,max = 42.9 mA cm2

, 37.9 mA cm-2, and 10.9 mA cm-2 respectively. In water, the corresponding estimated simulated

90
maximum photocurrent densities were Jph,max = 41.8 mA cm-2, 36.5 mA cm-2, and 10.7 mA cm2

, respectively.

Figure 5.5: Area plot of simulated transmitted spectral photon flux in air and water for the three
structures in Figure 1. Blue represents the AM 1.5G spectral photon flux. Orange, yellow, and
purple depict the transmitted spectral photon flux into Si for: nanocones on Si, nanocones with Ni
on Si, and Ni hole array on Si, respectively.

Figure 5.6 shows the simulated profiles of the electric field along the central cross section of
a nanocone. Figure 5.6(c-f) shows the field profiles for wavelengths of 484 nm, 552 nm, 628 nm,
and 770 nm, respectively, which correspond to the maxima in the transmission spectra shown in
Figure 5.3(b). The electric field was predominantly confined to the waveguide modes in the
nanocone, with strong coupling of incident light occurring at different radii for the different
wavelengths, as expected for a conical nanostructure. 23,143,144 In the simulation, the light
propagated through the nanocone and was transmitted into the Si substrate, where the field
intensity decreased due to absorption by the Si. Figure 5.6(g-j) shows the field profiles for
wavelengths of 442 nm, 584 nm, 738 nm, and 940 nm, respectively, which correspond to the
minima in the transmission spectra shown in Figure 5.3(b). Compared to the field profiles shown
in Figure 5.6(c-f), the profiles in Figure 5.6(g-j) showed an increased intensity of the electric field
in the space adjacent to the nanocone. The corresponding plots for transmitted light intensities |E|2

Z (µm)

Z (µm)

91

versus Y(µm) at the interfaces of Si/Ni (indicated as on Si) and Ni/air (indicated as on Ni) are
shown in Figure 5.7.

-0.5

0.5

-0.5

Y (µm)

-0.5

(c)

0.5

Y (µm)

-0.5

(d)

628
442nm
nm

0.5

Y (µm)

-0.5

(e)

0.5

-0.5

(f)

Y (µm)
738 nm

770
584nm
nm

33

33

Z (µm)
(µm)

Z (µm)

Z (µm)
(µm)

44

22

11

00

00

-0.5
-0.5

00

0.5
0.5

-0.5
-0.5

(h)

YY(µm)
(µm)
738 nm

00

0.5
0.5

-0.5

(i)

YY(µm)
(µm)

Y (µm)

0.5

11

(g)

Y (µm)
940 nm

44

22

(b)

Z (µm)

Z (µm)

0.5

770
584 nm

Z (µm)

(a)

628
442 nm

552 nm

Z (µm)

Z (µm)

484 nm

Y (µm)

0.5

-0.5

(j)

0.5

Y (µm)

940 nm

Figure 5.6: Simulated electric field profiles along the cross section of a TiO2 nanocone on a Si

Z (µm)

Z (µm)

substrate. (a) Cross section and (b) scale for the relative electric field intensity for the profile plots.

(c-f) Profiles for wavelengths of 484 nm, 552 nm, 628 nm, and 770 nm, respectively, which

correspond to the maxima in the transmission spectra shown in Figure 5.3(b). (g-j) Profiles for
wavelengths of 442 nm, 584 nm, 738 nm, and 940 nm, respectively, which correspond to minima
-0.5

0.5

-0.5

in the transmission spectrum Yin(µm)
Figure 5.3(b). Y (µm)

0.5

92

0.6

(b)

Y (μm)

442nm on Si
442nm on Ni

0.6

(e)

Y (μm)

0.6

|E|2

0.6

Y (μm)

(f)

Y (μm)

0.6

0.6

Y (μm)

940nm on Si
940nm on Ni

-0.6

-0.6

(d)

738nm on Si
738nm on Ni

-0.6

(c)

Y (μm)

-0.6

584nm on Si
584nm on Ni

|E|2

-0.6

|E|2

(a)

770nm on Si
770nm on Ni

-0.6

|E|2

628nm on Si
628nm on Ni

|E|2

|E|2

|E|2

552nm on Si
552nm on Ni

|E|2

484nm on Si
484nm on Ni

-0.6

(g)

Y (μm)

0.6

-0.6

(h)

0.6

Y (μm)

Figure 5.7: Transmitted light intensities |E|2 versus Y(µm) at interfaces of Si/Ni (indicated as on
Si) and Ni/air (indicated as on Ni). (a-d) Profiles for wavelengths of 484 nm, 552 nm, 628 nm, and
770 nm, respectively, which correspond to the maxima in the transmission spectra shown in Figure
5.3(b). (e-h) Profiles for wavelengths of 442 nm, 584 nm, 738 nm, and 940 nm, respectively, which
correspond to minima in the transmission spectrum in Figure 5.3(b).

Depending on the dimensions of the nanocones and the background index of refraction,
simulations indicated that the electric fields associated with some wavelengths of light were highly
confined inside the TiO2 nanocone, while the electric fields associated with other wavelengths
were only partially confined. Simulation results with varying dimensions of cones are presented
in Figure 5.8. In contrast, the optical absorption in the Ni was enhanced for the wavelengths of
light that were not completely confined within the nanocones. The wavelength-dependent variation
in the confinement of the electric field within TiO2 nanocones cannot be explained using effective
medium theory. Instead, wave-optic simulations showed that the nanocones acted as antennae for
the incoming radiation, coupling the light to waveguide modes, and providing a route for the light
to reach the underlying Si substrate even though 54 % of the surface was covered by Ni.

93

Figure 5.8: Dimension variation effect of Ni film on transmission to Si through TiO2 nanocone
waveguides in air. Hexagonal array (variant pitches) of 2.5 μm tall cones with a 200 nm base
radius and a 50 nm tip with a 200 nm radius Ni hole array.

The optimal structures revealed by simulation were fabricated and investigated in detail. To
experimentally demonstrate the enhancement in photocurrent density obtainable by utilizing TiO2
nanocone arrays, planar p+n Si homojunction photoanodes were prepared by doping n-type Si with
boron (B). Briefly, electron-beam evaporation was used to deposit 5 - 10 nm of SiO2 over the Si
photoanodes, prior to deposition of 2.3 µm of TiO2. Electron-beam evaporation depletes the TiO2
source of oxygen and thus increases the conductivity of the resulting films, so the 5 – 10 nm thick
SiO2 was deliberately incorporated to electrically isolate the TiO2 from making an electrical
contact to the highly doped p+-Si surface, while minimally affecting the optical behavior.
Figure 5.9 shows scanning-electron micrographs (SEMs), before and after electrodeposition
of Ni, for samples of dry-etched TiO2 nancone arrays on planar p+n-Si substrates. The EDS
mappings are included in Figure 5.10. The TiO2 nanocones were ~ 2.3 µm tall and had base radii
of ~ 250 nm. Discontinuities in the taper of the nanocones were evident, particularly near the vertex
of each cone. The radii at the vertices of the nanocones were < 50 nm. Figure 5.9(b,d) show that
the Ni predominantly deposited onto the Si surface in the spaces between the TiO2 nanocones, as

94
expected because the insulating 5-10 nm layer of SiO2 beneath the base of the cones should
prevent electrodeposition onto the TiO2. The Ni layer was ~ 70 nm thick as estimated based on the
charge passed during electrodeposition. The SEM image of a 50 nm thick Ni hole array fabricated
via electron-beam patterning and dry etching is shown in Figure 5.11.

Figure 5.9: Scanning-electron micrographs of dry-etched TiO2 nanocones on p+n-Si substrates
before (a,c) and after (b,d) electrodeposition of Ni.

95

Figure 5.10: EDS mappings of elements Ti, O, Ni, Si (a) with top view (b) with 30o tilt view of the
Ni/TiO2 nanocones/p+n-Si sample.

Figure 5.11: SEM image of the Ni hole array fabricated via electron-beam patterning and dry
etching of a 50 nm thick Ni layer. The average diameter of the holes was ~ 500 nm.

Figure 5.12 compares the real component of the complex refractive index measured by
ellipsometry for the electron-beam-evaporated amorphous TiO2 used to make nanocones in this
work relative to the real component of the index of refraction of ideal rutile TiO2 tabulated in

96
standard reference data. 157 The real component of the refractive index for the electron-beamevaporated TiO2 was substantially lower (n = 2.05 – 2.3) than for the standard value (n = 2.7 –
3.3), presumably due to oxygen depletion during evaporation and the amorphous phase of the TiO2
film.

Figure 5.12: Real component of the refractive index for (a) an ideal rutile TiO2 standard, and (b)
measured for a sample of electron-beam-evaporated amorphous TiO2.

Figure 5.13(a) shows the reflection, transmission, and absorption spectra calculated for TiO2
nanocones with 50 nm Ni using the experimentally measured refractive index data for electronbeam-evaporated TiO2, while Figure 5.13(b) shows the simulated transmitted photon flux along
with the AM 1.5G spectrum. The estimated attainable photocurrent density calculated from these
revised simulations for a Si solar cell covered with the TiO2 nanocone array and Ni was Jph,max =
29.8 mA cm-2, after correcting for losses due to reflection at the air/glass/water interfaces that are
unavoidable in an electrochemical cell configuration. This estimated maximum photocurrent
density (Jph,max = 29.8 mA cm-2) was substantially lower than the current density that could be
obtained with ideal TiO2 (Jph,max = 36.5 mA cm-2), but is still larger than the value expected for a
bare, planar Si surface (Jph,max ~ 28 mA cm-2).

97

Figure 5.13: Transmission (T), absorption (A), and reflection (R) plots for Si with TiO2 nanocones
and 50 nm thick Ni calculated with evaporated TiO2 refractive index data are shown in (a). (b)
shows the area plot overlapped over the AM 1.5G spectrum for the three different cases, as shown
in Figure 5.2, using the refractive index data for amorphous TiO2 deposited by e-beam evaporation.

Figure 5.14 compares simulated and experimentally measured reflectance spectra for a TiO2
nanocone array on Si, a TiO2 nanocone array with Ni on Si, and for a 50 nm layer of Ni with an
array of holes. The experimental and theoretical spectra are in good mutual agreement, with certain
differences readily ascribed to technical differences between the simulation and experimental
conditions. For example, the simulations were performed using a coherent illumination source,
whereas experimental measurements were not. Moreover, the simulations used smoothly tapering
nanocones and a flat layer of Ni, whereas the samples did not have either smoothly tapering cones
or a perfectly flat Ni layer (Figure 5.9).

98

Figure 5.14: Reflection spectra (a) simulated and (b) measured for samples consisting of an array
of TiO2 nanocones on Si (blue), an array of TiO2 nanocones with Ni on Si (red), an array of holes
in a Ni layer on Si (yellow).

Figure 5.15 shows the current density versus potential behavior, in the dark and under 100
mW cm-2 of simulated AM1.5 illumination, respectively, while in contact with 1.0 M KOH(aq),
for a photoanode made from a p+n-Si substrate covered with an array of TiO2 nanocones and a
layer of Ni (Figure 5.9(b,d)). The light-limited photocurrent density, obtained by subtracting the
current density measured in the dark at ~ 1 V vs the saturated calomel electrode (SCE) from the
current density measured in the light at the same potential, was ~ Jph,max = 28 mA cm-2, and
matched well with the value estimated (Jph,max = 29.8 mA cm-2) from the simulations. The current
density at the formal potential for water oxidation, E°′(O2/H2O), was 2 - 7 mA cm-2. The observed
light-limited photocurrent density was comparable to the photocurrent density normally measured
for a bare planar Si surface, consistent with the TiO2 nanocones serving as antireflective structures
that can couple to incoming light to enable transmission of light into the Si substrate even when ~
54 % of the Si surface was covered with ~70 nm of Ni.

E°′(O2/H2O)

99

Figure 5.15: Current-density versus potential behavior, in the dark and under 100 mW cm-2 of
simulated AM1.5 solar illumination, respectively, for a p+n-Si sample covered by an array of TiO2
nanocones and 300 mC cm-2 of electrodeposited Ni while in contact with 1.0 M KOH(aq). The scan
rate was 50 mV s-1.

Si photoanodes with a uniform 2 nm layer of Ni have been reported previously to exhibit a
light-limited photocurrent density of Jph,max ~55 mA cm-2 under ~2.25 Suns equivalent of
illumination, whereas increasing the Ni thickness to 20 nm reduced Jph,max to ~32 mA cm-2. 158
These results translate at 1 Sun intensity into Jph,max < 25 mA cm-2 for 2 nm Ni and < 15 mA cm-2
with 20 nm Ni, whereas for comparison, the p+n-Si photoanodes investigated herein exhibited
Jph,max ~ 28 mA cm-2. Although thick Ni catalysts are not required to lower the overpotential for
water oxidation, 159 the use of thick electrocatalyst layers can minimize performance degradation
associated with catalyst detachment.
A wide range of alternate approaches to efficient photoelectrode performance have been
demonstrated, especially when active catalysts for the desired water-splitting half-reactions are
used. 132 For example, p+n-Si(111) photoanodes decorated with 15nm thick Ni islands that covered
18% of the photoelectrode surface have demonstrated a light-limited current density of 20.4 mA
cm-2. 159 Modeling has shown that the optimal efficiency of a water-splitting system using a lightfacing photocathode patterned with Pt islands covering 5% of the optical plane closely approaches
that of a system using a photocathode patterned with a hypothetical transparent catalyst with the

100
same activity as Pt. 160 Furthermore, photoanodes do not require reflective metallic catalyst
coatings; indeed, p+n-Si(100) photoanodes coated with 75 nm of sputtered NiOx have been shown
to be stable for 1200 h of continuous oxygen evolution in contact with 1.0 M KOH(aq) and exhibit
a light-limited photocurrent density > 30 mA cm-2. The TiO2 nanocone array structure can
effectively optimize light transmission and catalysis simultaneously in a variety of possible
materials systems, and would be especially beneficial for chemical reactions that require very high
mass loadings of catalysts, such as O2 evolution or CO2 reduction using earth-abundant
electrocatalyst materials.

5.4 Conclusion and Outlook
Dielectric nanocone arrays provide an additional option to a growing toolbox of strategies for
directing broadband light around opaque top contacts to PV or PEC cells. The approach ought to
be generally applicable for any combination of semiconductor and metal, and in principle is
scalable; however, arbitrary combinations of semiconductors and metals may not be compatible
with electrodeposition of the metal onto the semiconductor as used in the fabrication process
described herein. Furthermore, the fabrication process developed herein for a proof-of-concept
experimental demonstration of a device that makes use of a TiO2 nanocone array for light
management was complex relative to other known options for light management, such as
antireflective catalyst coatings or deposition of a controlled density and diameter of catalyst islands.
The value of the dielectric nanocone approach to light management can be increased by developing
a simplified fabrication process and by developing synthetic methods that yield TiO2 nanocones
with a refractive index that approaches the index for ideal TiO2. Modeling and simulation efforts
that compare attainable efficiencies for application-specific devices across relevant lightmanagement strategies will prove valuable for identifying the strategies that are most promising
for a given application.

101

CHAPTER 6
Effectively Transparent Catalysts for PEC Device
6.1 Introduction
Direct solar-to-fuel generation using a photocathode-based PEC cell requires a light absorber
which can provide the photovoltage necessary to overcome the thermodynamic potential (1.23V
for H2/O2, 1.33V for CO/O2) as well as the catalyst overpotentials for both cathode and anode
reactions. To realize high solar-to-fuel efficiency, it is necessary to maintain a catalytic current
density close to the light limiting photocurrent density for a solar-driven light absorber, which can
be fulfilled when catalyst ensembles are highly transparent. In Chapter 3, we have successfully
achieved a record for solar-to-hydrogen PEC conversion efficiency of 19.3% (under simulated
sunlight) in acid electrolytes by integrating Rh nanoparticle catalysts onto photocathodes with
minimal parasitic absorption and reflection losses in the visible range. However, for CO2 reduction,
a different approach is required, given the opaque nature and limited activity of most CO2R
catalysts. The complexity of the CO2R kinetic landscape makes it harder to control than the
competing HER at lower overpotentials. A large geometric filling fraction of opaque
electrocatalysts on the electrode surface and therefore a high active catalyst area will help to
enhance the catalytic activity and reduce the overpotential. Thus, strategies for design and
fabrication of front illuminated photocathode PECs need to be developed.
Earlier demonstration on Si photocathode using metal catalyst hole arrays as the catalyst still
block the majority of light96, and it would be even sensitive to apply such approach on tandem or
triple junction solar cells since the current matching between each subcell is so critical that
broadband transmission through catalysts layer is necessary. The earlier work with catalysts
loading on high aspect ratio wire to prevent light blocking effect is promising161,162, nevertheless,
is only suitable for single junction cells and can’t be applied as a general approach. Here, we
propose to use light management strategies to create highly active and effectively transparent
catalyst (ETC) structures for photocathodic CO2 reduction. An effectively transparent catalyst
consisting of arrays of micron-scale triangular cross-sectional metal grid fingers is capable of
redirecting the incoming light to the open areas of the PEC cell without shadow loss. Broadband

102
high transmission in the visible range enables the high photocurrent, further realizing
renewable fuel production from sunlight.

6.2 Numerical and Experimental Method
All of the simulation spectra and field distributions are carried out by solving the Maxwell
equation with the commercial COMSOL Multiphysics software based on the finite element
method (FEM). The triangular metal grids were constructed using the 2D simulation with periodic
boundary conditions along the x-axis, and infinite boundary conditions were rendered as perfectly
matched layers (PMLs) along the y-axis. Figure 6.1 displays the geometry parameters that were
used in simulation. Coverage is defined as width (w) divided by pitch (p), and height is indicated
as h. A plane-wave source of illumination in the wavelength range from 350 nm to 1350 nm was
utilized. Spectra for unpolarized light were obtained by averaging the spectra for the two
orthogonal polarizations. All materials refractive index were modeled using tabulated data
provided in the software.

Figure 6.1: Schematic of triangular metal gird geometries used in simulation.

The GaInP/GaInAs/Ge triple junction cell from Spectrolab (C4MJ) and GaInP/GaAs/Si triple
junction cell from ISE are considered in this study. For illumination during laboratory tests, an
Oriel Instruments 75 W Solar Simulator was used and matched with AM 1.5G. The corresponding

103
sub-cell currents with integration of external quantum efficiency determined the expected short
circuit current density under AM 1.5G illumination. The light intensity of the solar simulator was
set to provide the expected short circuit current density from the specific triple junction cell. While
this is not expected to yield a simulated solar irradiance of 100 mW×cm-2 due to the different solar
irradiance in the 800–1000 and 1150–1800 nm regions, it does produce a response of the triple
junction PV that is the same as it would be in actual AM 1.5G sunlight.
The fabrication processes of metal triangular grid on glass to be placed onto triple junction
photocathode is illustrated in Figure 6.2. First, a master was fabricated on Si substrate using a twophoton lithography technique. Second, a PDMS stamp was formed with the lithography master.
The PDMS stamp has an inverse structure to the lithography master. Next, the PDMS stamp was
stamped onto a glass substrate to fill the metal ink where a pre-clean process of the glass substrate
is preferred. The metal grid structures were printed onto the glass substrate after baking on a hot
plate at ~100 °C for 10 min to remove the remaining solvent, then the PDMS stamp could be taken
off. Afterward, the metal triangular grid on glass sample was post annealed at 200 oC for 1 h in a
muffle furnace in air. Scanning electron microscope (SEM) images show the metal grids with a
cross section of a triangle structure. Additional metal catalysts could be electrodeposited (Figure
6.3) in an aqueous solution at 0.1 mA for 2-4 min. A Cary 5000 UV/vis/NIR with integrating
sphere was used to obtain reflection, transmission, and absorption spectra in air.

104
Making PDMS stamp
Metal ink infill/printing
PDMS

Two photon
lithography master
Metal grid

glass
20 μm

Figure 6.2: Illustration of the fabrication process for metal catalyst triangles. SEM image shows
an example of printed Ag metal triangle grid with 35% coverage.

Figure 6.3: Illustration of the metal deposition process on top of metal catalyst triangular grids.

A modified PEEK compression cell (Figure 6.4) with lateral off-set between cathode and
anode to allow illumination on the cathode was used as the vessel for the measurement. Anode
chamber volumes is 2 mL, and cathode chamber volume is 4 mL. The anode and cathode electrode
working areas were 6 cm2 and 0.2 cm2, and the membrane area was 2.4 cm2. The activated NiOx
foam anode size is enlarged to reduce overpotential from OER, and can be folded to reduce the

105
geometric area. 100 mM potassium bicarbonate (KHCO3) saturated with CO2 was used as the
electrolyte for experiments at pH 6.8. The anion exchange membrane (AEM) was chosen as
Fumasep FAA-3-50 for lower resistance. A leakless Ag/AgCl reference electrode was used to
determine the electrode potential versus RHE. All electrochemical measurements were performed
using a Biologic VSP-300 potentiostat. Scan rates were set to 50 mV×s-1.

Figure 6.4: Cell configuration composed of 1 NiOx anode, 2 ETC-PEC assembled cathode, 3 anion
exchange membrane, 4 quartz window, 5 reference electrode, 6 catholyte chamber, 7 anolyte
chamber, 8 CO2 gas inlet, and 9 gas product/CO2 outlet. White arrows indicate the gas flow.

The electrochemical setup was operated in a continuous flow mode. Humidified carbon
dioxide was provided to the electrochemical cell and its flow rate was controlled with an Alicat
flow controller. The exhaust gasses went through a mixing volumn, then an Alicat flow meter, and
finally to a gas chromatograph (SRI-8610) using a Hayesep D column and a Molsieve 5A column
with N2 as the carrier gas. The gaseous products were detected using a thermal conductivity
detector (TCD) and a flame ionization detector (FID) equipped with a methanizer. The high
performance liquid chromatography (HPLC) is used to analyze liquid products after 1 h
accumulation at each operation condition. Both anolyte and catholyte were sampled to capture

106
possible product cross-over. Quantitative analysis of gaseous/liquid products was based on
calibration with several standards over many orders of magnitude in concentration.

6.3 Results and Discussion
PEC device with an effective transparent catalyst (ETC) is illustrated in Figure 6.5. It
describes a schematic of light management implementing metal triangles on top of a
semiconductor photoelectrochemical cell. The micro-scale triangle grid arrays can redirect light to
photoabsorbing surfaces and reduce reflection loss that is normally expected from metal catalysts.
Once light can be efficiently absorbed, electrons generated from the semiconductor PEC cell
transfer to the metal triangle and initiate cathode reduction reaction. To catalyze CO2 reduction
reactions and generate CO and/or higher energy density hydrocarbon product, silver (Ag), gold
(Au), and copper (Cu) are chosen specifically for the higher activity. Even though the process in
not limited to the photoabsorbing materials, triple junction cells are required to provide enough
photovoltage for the thermodynamic potential and overpotential of CO2R.

Figure 6.5: Schematic illustration of light management with metal triangle catalyst on top of a
semiconductor photoelectrochemical cell.

107
The metal triangle catalysts are constructed to have heights (h) greater than the base width
(w) of the triangles for enhancing actual catalytic surface area. The base to height ratio (w/h) of
the triangle can range from about a 1:1 ratio to about 1:3. The base width of the triangle is supposed
to be greater than the wavelength of incoming light, preventing coupling with the metal structure
where optical loss can be introduced. For visible wavelength range, the base width of the triangle
should be larger than 2 μm. Numerical calculations were used to investigate the optical response
and determine the optimal geometry. In general, we would need larger geometric filling fraction
of the metal catalysts on the electrode surface and therefore a high active catalyst area to help
enhance the catalytic activity and reduce the overpotential. Figure 6.6 presents the simulated
absorbance, reflectance, and transmittance spectra respectively of metal catalyst triangle coverage
ranging from 0%, 10%, 25%, 50%, and 83% with base width of about 2.5 μm and a height of about
7 μm on GaAs substrate. The coverage can be defined as the occluded area by the metal triangle
to the total surface area of the photoelectrochemical cell (w/p).
(a)

(b)

(c)

Figure 6.6: Simulated (a) absorbance spectra, (b) reflectance spectra, and (c) transmittance
spectra respectively of different Ag catalyst triangle coverages of 0%, 10%, 25%, 50%, and 85%
with w = 2.5 μm and h = 7 μm on GaAs substrate.

108
Table 6.1 lists the change of absorbance, reflectance, transmittance, and current of
different metal catalyst triangle coverage ranging from 10%, 25%, 50%, to 83% as compared to
0% coverage. The catalyst surface area is defined as the sum of both the metal triangle slope
surface area. While attaining high catalytic current, it is still required to maintain high transmission
and high photocurrent. In the simulation, we found that a mesoscale Ag grid array with triangular
cross-section lines, metal coverage of 50%, and catalyst surface area of 284.4% exhibits negligible
additional reflection and absorption loss. Figure 6.7 exhibits the simulated field profile of a metal
catalyst triangle with 50% coverage at λ = 500nm. The plane wave is reflected on the slope of the
metal triangle and redirected to the wave to the photoelectrochemical cell. Metal catalyst grids
with coverage ranging from 25% to 50% can be utilized as appropriate solution for balancing light
management and catalytic current.
Table 6.1: Change of absorbance, reflectance, and transmittance of of different Ag catalyst
triangle coverages of 10%, 25%, 50%, and 85% with w = 2.5 μm and h = 7 μm on GaAs substrate
related to 0% metal coverage.
Metal
coverage

Ave ΔR

Ave ΔA

Ave ΔT

Catalyst
surface
area

10%

0.027

0.005

-0.032

56.9%

25%

-0.030

0.012

0.018

142.2%

50%

-0.019

0.031

-0.012

284.4%

83%

0.194

0.098

-0.292

472.2%

109

50% coverage

Max

|E|

Ag

III-V

Min

Figure 6.7: Simulated field profile of a Ag catalyst triangle with w = 2.5 μm, h = 7 μm, and 50%
coverages on GaAs substrate at λ=500nm.

As proof of concept, Ag catalyst triangle catalysts with 2.5 μm width, 5 μm height (w/h = 1:2 ),
and 35% coverage were fabricated with the printing process described in method section. The
catalyst surface area of such a structure is calculated to be 144.3%. Additional catalyst layers
including Au and Cu are selectively electrodeposited on the surface of the printed Ag triangle grids.
Figure 6.8(a)-(c) shows elementary mapping through EDX (Energy-dispersive X-ray spectroscopy)
of three types of metal catalyst triangle grid arrays on glass substrate, including Ag triangle, Au
on Ag triangle, and Cu on Ag triangle. It is observed that after electrodeposition, the Ag catalysts
are conformally covered by the other metal catalysts where signals from Ag are hindered. We note
here that the catalysts are printed on an insulating glass substrate and later wire connected to the
bottom PEC cell instead of directly printed on the PEC cell. It passivates the light absorber surface
and reduces competing reactions of CO2R reactions, hence improving the solar-to-fuel conversion
efficiency.

110
(a)

Ag Triangle

(b)

Au on Ag Triangle

(c) Cu on Ag Triangle

Figure 6.8: EDS mappings of elements O, Si, Ag, Au, and Cu of the (a) Ag triangle (b) Au on Ag
triangle (c) Cu on Ag triangle on glass substrate.

Figure 6.9 includes the optical measurements in air with different wavelengths of light for the
three types of metal catalyst triangle grid arrays on glass substrate with coverages of 35%. Near
10% reflection loss can be introduced with bare glass substrate. Figure 6.9(b) shows additional 10
% reflection loss from the 35% Ag triangle grids while the Au on Ag triangle and Cu on Ag triangle
cases exhibit similar reflection loss at longer wavelengths but lower at shorter wavelengths. The
overall reflections are higher than expected from the simulation, which can possibly be attributed
to the non-ideal surface roughness. The morphology of the bare printed Ag triangle and after
electrodeposition of Au and Cu are shown in SEM images of Figure 6.10. Even though rougher
surfaces can be observed for both electrodeposited Cu and Au, the two samples do not show higher
reflection loss due to the contribution of absorption loss, as shown in Figure 6.9(a). In particular,
the needle structure Au catalysts exhibit the largest absorption loss. Figure 6.9(c) indicates the
transmission spectra for the three metal catalyst triangle grids, where 35% Ag triangle grids show

111
the best transparency than the Cu on Ag triangle, with Au on Ag triangle being the worst.
Although reflection and absorption loss are introduced from non-perfect geometry, the
transmissions for the three samples are still higher than the calculated transmission spectrum for
200nm thick Ag film with 35% coverage, shown as the blue dashed line in Figure 6.9(c). While
35% Ag triangle maintained relatively higher transparency than the 35% Ag film, the catalyst
surface area of 144.3% is 4 times larger, which is one of the critical benefits of the triangle grid

(a)

(b)

Absorbance (%)

Reflectance (%)

design.

Transmittance (%)

(c)

Figure 6.9: Experimental measurements of (a) absorption spectra, (b) reflection spectra, and (c)
transmission spectra respectively for different metal catalyst triangles with coverages of 35% on

112
glass substrate. 0% coverage is shown as glass. Calculated transmission spectrum for 200nm
thick Ag film with 35% coverage is included for comparison.

(b)

(a)

2 μm

(c)

2 μm

2 μm

Figure 6.10: SEM images of (a) Ag triangle, (b) Au on Ag triangle, and (c) Cu on Ag triangle on
glass substrate with 35% coverage.

To evaluate the catalytic behavior of the three different metal catalyst triangle grids, threeelectrode measurements in CO2 saturated 0.1 M KHCO3 (bulk pH of 6.8) are conducted with
results shown in Figure 6.11. For Ag triangle grids, both the Faradaic efficiency for CO and current
density increased with increasing potential vs RHE with fFE,CO close to 90 % at -1.2 V vs. RHE,
see Figure 6.11(a)(d). Comparable current density can be observed for both Au on Ag triangle
grids (see Figure 6.11(b)) and Cu on Ag triangle grids (see Figure 6.11(c)) while Au on Ag triangle
grids show higher current density at lower potential. The Au catalysts also shift the optimized
faraday efficiency for CO to lower potential at -0.8V vs RHE though with lower fFE,CO of 65 %,
see Figure 6.11(e). Higher value hydrocarbon products including acetate, ethylene, ethanol, and npropanol can be generated with additional Cu catalysts as shown in Figure 6.11(f). At -1.4 V vs
RHE, a total Faradaic efficiency for CO2R can reach 90% with C2+ product over 50%. The product
with highest Faradaic efficiency is ethanol with 32%. The selectivity is still a concern that effort
for product separation would be required for practical application. In the later study, the better
performing Ag triangle grids and Cu on Ag triangle grids are further evaluated.

113

J / mA cm -2

J / mA cm -2

(b)

-1.5

(d) 100

-1

U vs. RHE / V

(e) 100

-1

U vs. RHE / V

60
40

(f) 100

-1

U vs. RHE / V

-0.5

60
40

-1.5

-1

-0.5

U vs. RHE / V

80
60
40
20

20

20

Cu on Ag triangle

-1.5

-0.5

80

F.E. / %

F.E. / %

(c) 10

-1.5

-0.5

80

-1.5

Au on Ag triangle
10

J / mA cm -2

Ag triangle
10

F.E. / %

(a)

-1

U vs. RHE / V

-0.5

-1.5

-1

-0.5

U vs. RHE / V

Figure 6.11: Current density and product distribution at different potentials vs RHE for (a)(d) Ag
triangle, (b)(e) Au on Ag triangle, and (c)(f) Cu on Ag triangle on glass substrate with 35%
coverage.

Two electrode measurements of metal triangular grid cathode and NiOx mesh anode with a
range of applying cell potentials were conducted for better understanding of full cell operation,
and the resulting current density and product distribution are shown in Figure 6.12. For the case
of Ag triangle grid cathode with NiOx anode, see Figure 6.12(a), optimized faraday efficiency for
CO can be achieved with cell voltage larger than 2.5 V. The high fFE,CO of 80 % can be extended
to 2.9 V. For the case of Cu on Ag triangle grids cathode with NiOx anode, see Figure 6.12(b),
Faradaic efficiency of ethanol ~ 30% was achieved with cell voltage ranging from 2.5 V to 2.9 V.
Other higher value hydrocarbon products start to generate with cell voltage larger than 2.7 V,
realizing C2+ product over 50% and a total CO2R Faradaic efficiency > 80%.

114

(a) 12

Ag triangle + NiO

(b) 12

2.2

2.4

2.6

2.8

Ucell / V

(c) 100

2.2

2.4

80

80

60

60

40

2.6

2.8

2.8

Ucell / V

(d) 100

F.E. / %

F.E. / %

10

J / mA cm -2

J / mA cm -2

10

Cu on Ag triangle + NiO

40
20

20

2.2

2.4

2.6

Ucell / V

2.8

2.2

2.4

2.6

Ucell / V

Figure 6.12: Current density and product distribution at different cell potentials with NiOx anode
for (a)(c) Ag triangle, and (b)(d) Cu on Ag triangle on glass substrate with 35% coverage.

For a fully integrated device with catalysts on the illuminated side of the light absorber, the
optical response of the catalysts needs to be considered. We calculated the short circuit
photocurrent of two different triple junction cells; one from Spectrolab with an open circuit voltage
of 2.6 V and sub-cells of GaInP/GaInAs/Ge, the other from ISE with an open circuit voltage of
3.125 V and sub-cells of GaInP/GaAs/Si. By applying the transmission spectra of the two different
metal triangle grids in conjunction with the external quantum efficiency (EQE, see Figure 6.13) of
the two triple junction cells, the catalysts light blocking effect on each of the sub-cells can be
evaluated. The short circuit photocurrent Jsc is then obtained from the minimum current density of
the individual sub-cells. The calculation results are listed in Table 6.2. We found that the current
limiting cell remains to be the middle cell for the Spectrolab 3J cell, and remains to be the bottom
cell for ISE 3J cell no matter which catalyst is applied. The light limiting current density of the

115
Spectrolab 3J cell decreases from 14.83 mA/cm2 to 11.46 mA/cm2 for Ag triangle grids, to
10.93 mA/cm2 for Cu on Ag triangle grids. Similarly, the light limiting current density of ISE 3J
cell decrease from 11.66 mA/cm2 to 9.15 mA/cm2 for Ag triangle grids, to 8.73 mA/cm2 for Cu on
Ag triangle grids.
Spectrolab 3J

0.8

0.8

0.6

0.6

0.4

0.4

0.2

0.2

500

1000

1500

ISE 3J

(b) 1

EQE

EQE

(a) 1

2000

Wavelength / nm

400

600

800

1000

1200

Wavelength / nm

Figure 6.13: External quantum efficiency of (a) Spectrolab GaInP/GaInAs/Ge triple junction cell,
and (b) ISE GaInP/GaAs/Si triple junction cell.

116
Table 6.2: Current densities calculated for the individual sub-cells of the Spectrolab
GaInP/GaInAs/Ge triple junction cell and ISE GaInP/GaAs/Si triple junction cell in conjunction
with optical response of different catalysts under AM1.5G 1 sun illumination assuming the
standard reference sunlight spectrum (AM1.5G ASTM G-173). The short circuit photocurrent Jsc
is obtained from the minimum current density of the individual sub-cells.

Device

Spectrolab
3J

ISE 3J

Catalyst

Jtop
(mA×cm-2)

Jmiddle
(mA×cm-2)

Jbottom
Jsc
(mA×cm-2) (mA×cm-2)

none

15.56

14.83

18.53

14.83

Ag triangle 35%
on glass

11.57

11.46

14.93

11.46

Cu on Ag triangle
35% on glass

11.02

10.93

13.90

10.93

none

12.75

13.13

11.66

11.66

Ag triangle 35%
on glass

9.46

10.05

9.15

9.15

Cu on Ag triangle
35% on glass

9.00

9.59

8.73

8.73

Figure 6.14 plots the J-V curves of the two 3J cells after modifying with catalyst transmission
together with catalyst curve for the two metal triangle grid cathodes plus NiOx anode. The intersect
of the two curves defines the operation condition for the integrated ETC-PEC device. For Figure
6.14(a), Spectrolab 3J and Ag triangle/NiOx curves intersect at 2.57 V and 2.7 mA/cm2, so a solarto-CO efficiency of 3.1 % can be expected. For Figure 6.14(b), Spectrolab 3J and Cu on Ag
triangle/NiOx curves intersect at a similar point 2.57 V and 2.7 mA/cm2, and a total solar-to-fuel
efficiency from CO2R of 2.7 % and solar-to-C2+ efficiency of 1.1 % can be expected. For Figure
6.14(c), ISE 3J and Ag triangle/NiOx curves can be interpreted to intersect at 3.0 V and 7.0 mA/cm2,
a solar-to-CO efficiency of 6.6 % can be estimated. For Figure 6.14(d), ISE 3J and Cu on Ag
triangle/NiOx curves intersect at 2.86 V and 8.6 mA/cm2, a total solar-to-fuel efficiency from
CO2R of 8.7 % and solar-to-C2+ efficiency of 5.4 % can be expected.

117

Figure 6.14: Current density at different cell potentials of light absorber and cathode/anode
catalysts defining the operation point for (a) Spectrolab 3J + Ag triangle/NiOx, (b) Spectrolab 3J
+ Cu on Ag triangle/NiOx, (c) ISE 3J + Ag triangle/NiOx, and (d) ISE 3J + Cu on Ag triangle/NiOx.

We also experimentally realized the integrated ETC-PEC device with Spectrolab 3J cell and
Ag triangle catalysts with 35% coverage. J-V characteristics and product distribution with threeelectrode measurements performing CO2RR in CO2 saturated 0.1 M KHCO3 under 1 Sun
illuminations are shown in Figure 6.15. As comparison to bare Ag triangle catalysts, the integrated
device with Ag triangular catalysts stacking on photovoltaic shows a positive shift of the J-V curve
vs RHE. The trends of product distribution over applying potential are almost identical. No
external bias operation can be achieved with 20 h stability as shown in Figure 6.16. Starting from
~3% solar-to-CO efficiency, the integrated device efficiency slowly drops to ~2% possibility due
to the insufficient adhesion of the Ag triangle grids to the glass substrate during electrolysis.

118
(a)

100
Ag triangle
+Spectrolab 3J

Ag triangle

80

F.E. / %

J / mA cm -2

60
40
20

-1.5

-1

-0.5

(c)

0.5

1.5

-1.5

-1

U vs. RHE / V

F.E. / %

80
60

CO
H2

40
20

0.5

1.5

s. RHE / V

-1.5

-1

-0.5

0.5

1.5

U vs. RHE / V

Figure 6.15: (a) Current density and (b) product distribution at different potentials vs RHE for Ag
triangle, and integrated ETC-PEC device with Spectrolab 3J plus Ag triangle on glass substrate
with 35% coverage.

U vs. R

100
Ag triangle
+Spectrolab 3J

-0.5

119

(b)
Solar-to-CO efficiency / %

(a)
J / mA cm -2

10

20

time / h

(c)100

10

20

time / h

F.E. / %

80
60
40
20

10

15

20

time / h

Figure 6.16: (a) Current density, (b) solar to CO efficiency, and (c) product distribution at 0 V vs
NiOx counter electrode of the integrated ETC-PEC device with Spectrolab 3J + Ag triangle on
glass substrate with 35% coverage over 20 h stability test.

6.4 Conclusion and Outlook
A solar-driven photoelectrochemical device capable of reducing CO2 without externally
applied bias is achieved with ~3% solar-to-CO efficiency using Ag ETC-PEC. Over 6% solar-toCO efficiency with Ag triangle catalysts or a total solar-to-fuel efficiency from CO2R > 8 % and
solar-to-C2+ efficiency > 5 % with Cu on Ag triangle catalysts can be expected with triple junction
cell providing larger photovoltage. Our designs featuring photonic structures to enable high
absorption light absorbers and effectively transparent catalyst layers for PEC cell are general
approaches not limited to single reaction or specific photovoltaic system. Different metal catalysts
can possibly be applied to reveal different chemical reactions and fuels. Together with the
capability of scalable processes through ink printing and electroplating, it will be a critical step to
advances in field of solar fuel generation and all other related optical applications.

120
Future directions involve extending device stability and new catalyst design for higher
value product with high selectivity. The solar-to-fuel efficiency can be further increased with
slightly higher metal coverage, and crossed grid structures with similar triangular cross section can
also be considered. A new protection scheme allowing efficient passivation of the PEC cell from
undesired competing reactions would be valuable for simpler integration devices.

121

CHAPTER 7
Interface Analysis of Catalysts and Protection Layer
7.1 Introduction
Band gap tailored dual junction light absorbers typically utilizing silicon and/or III-V
semiconductors have the potential to provide the highest efficiency for unassisted water splitting
in an integrated photoelectrochemical (PEC) device. 69 Depending on the desired product, similar
devices optimized for CO2 reduction consists of triple junction light absorbers 70,163 combined with
catalysts needed for the desired reactions. However, these devices employ semiconductors,
especially the widely used III-V compounds, that are typically unstable in aqueous media and will
corrode and/or passivate unless protected by transparent, conducting, and chemically robust
protective layers. 74,164,165 Amorphous defective TiO2 (a-TiO2) coatings have been found to yield
protective layers that can pass charge from the semiconductor to the catalysts, are stable in highly
basic solution, and can offer antireflective properties to PEC electrodes. The a-TiO2 can be applied
using atomic layer deposition (ALD) with tetrakis(dimethylamino)titanium (TDMAT) precursors
at relatively low temperatures to yield protected photoanodes that facilitate the transport of holes.
It has been proposed that a gap states in the TDMAT a-TiO2 protection layer is responsible
for hole conduction. 74,166,167 Temperature dependent measurements showed a decreasing
activation energy with decreasing temperature, which indicated that neither valence nor
conduction band are the primary pathways of conduction. 167 The data was consistent with a
hopping mechanism between Ti3+ and Ti4+ states and conductivity as well as size of the gap states
correlated with the Ti3+ concentration. Solid liquid junctions of these protective a-TiO2 coatings
alone are shown not to be conductive and/or provide the necessary low overpotential for water
oxidation. 74,166,167 However, in conjunction with the appropriate choice of metal catalyst, high
catalytic exchange current densities and thus low overpotentials can be observed. Metals with a
work function less than a-TiO2 generally provided higher conductivities across the a-TiO2/metal
interface while high work function metals depleted the Ti3+ sites. 167 However, nickel as a high
work function metal acted similar to low work function metals. It was suggested that properties
other than the work function itself must contribute to the different conductivities.

122
This made nickel the material of choice for making TDMAT a-TiO2 conductive
and enabling water oxidation. However, a minimum thickness of nickel is required to observe
significant catalytic current for water oxidation or when using a fast redox couple (Fe(CN)63−/4−).
168

It has been shown in the past that ultrathin metal films and nanoparticles on oxide surfaces

exhibit unique properties for catalytic systems. 169 For example, the deposition of thin reactive
metal films Cr resulted in strong interaction with the interfacial TiO2, leading to a change of the
gap states in crystalline TiO2 within the band gap at the metal interface. The ALD a-TiO2 has a
distorted octahedral symmetry (Oh) with a valence configuration of O 2p6 Ti 3d04s0 which results
in a pure O 2p derived valence band and pure Ti 3d and 4s derived conduction band. 170 The six
oxygen ligands in Oh symmetry break the 3d-orbitals (AÃ Õ , AŒ Õ Gœ Õ , AŒœ , AŒÃ , AœÃ ) into eg (pointing
towards ligands) states and t2g states (pointing between ligands). 170 The gap state in a-TiO2 is a
partial populated t2g derived band (Ti3+) which was originally unoccupied due to the d0 character
of TiO2. 171 Successive dd transitions are possible after excitation of electrons from Ti 2p core
level into eg conduction band states followed by deexcitation from the t2g valence band states into
the core level hole, effectively creating an intermediate state with an electron in the eg and a hole
in the t2g band (dd transition).
In this study we investigate the properties of an a-TiO2/metal interface that controls the
transfer of holes through the a-TiO2 to Ni, Au and Ir contacts to understand why the observed
catalytic currents of these systems differ. Resonant photoemission spectroscopy (resPES) and
resonant inelastic X-ray scattering (RIXS) at the Ti 2p edge were used to investigate the a-TiO2
gap state at the buried interface between the metal and the TiO2. In resonance X-ray spectroscopy,
the ground state is initially excited to an intermediate state, and in our case, this is a conduction
band (CB) excited state. The lifetime of this CB excited state defines the full width at half
maximum (FWHM) of the resonance. This intermediate state then decays by either emission of an
Auger electron for resPES or of an X-Ray for RIXS. The final state is still an excited state (for
RIXS and some decay channels for resPES) with a hole in the valence band, and its lifetime defines
the FWHM of the valence band (VB) resonances. 172 Resonant X-ray spectroscopy is a twodimensional technique using photoelectron/exciting photon energy in resonant photoelectron
spectroscopy, resPES, or emitted X-Ray/exciting photon energy in resonant inelastic X-Ray
scattering, RIXS. By this technique, the X-ray absorption spectrum (XAS) is split into individual

123
X-Ray photoelectron spectra or X-ray emission spectra for each excitation energy value. At
the energy for a core level absorption of a particular element, the cross section for that element
increases dramatically. It is several orders of magnitude higher than for any elements that have a
different core level absorption energy. Due to the increased absorption cross section and the decay
processes of the resonantly created core hole, the density of states of that element can be
determined. Hence, valence bands measured by resonant X-ray spectroscopy give evidence of the
partial density of valence band states (occupied states). For a complete set of VB spectra, the
excitation energy is scanned over the core level absorption edges energies of the elements of
interest. Scanning over the absorption edge an XAS spectrum can be recorded allowing to probe
the CB states in addition, e.g. total electron yield (TEY) or total fluorescence yield (TFY).

7.2 Experimental Method
Films of a-TiO2, prepared by atomic-layer deposition (ALD), 74,153,173,174 were deposited on
degenerately doped p+-type silicon (resistivity ρ < 0.005 Ω×cm) substrates. Si (100) wafers were
first cleaned via an oxidizing etch by soaking of a 3:1 (volume ratio) “piranha” solution of
concentrated H2SO4 (98 %) to 30 % H2O2 for 2 min, followed by a 10 s in a 10 % (by volume)
solution of HF(aq). The wafers were then immediately etched in a 5:1:1 (volume ratio) solution of
H2O, 36 % hydrochloric acid, and 30 % hydrogen peroxide for 10 min at 75 °C before being
moved into the ALD chamber. The a-TiO2 was deposited from a tetrakis(dimethylamido)titanium
(TDMAT) precursor in a Cambridge Nanotech Savannah ALD reactor. A 0.1 s pulse of TDMAT
was followed by 15 s purge of N2 at 20 sccm, followed by a 0.015 s pulse of H2O before another
15 s purge with N2. This process was repeated for 1500 cycles to reach ~ 70 nm in thickness.
Metals were deposited via electron-beam evaporator (System 02520, Angstrom Engineering) with
base pressure of 1x10-7 Torr at rate of 0.1 to 1 Å/s. Where desired, nickel was deposited at a RF
sputtering power of 150 W for 20-300 s in an AJA-International sputtering system, at a rate of
approximately 2 nm/min.
Resonant XPS experiments have been performed at BESSY II, Berlin at the soft X-ray
beamline U49/2-PGM2. 175 The photon energy resolution used was around 30 meV. All spectra
are corrected for the incoming photon flux. For measuring the resPES valence band spectra at the

124
Ti 2p, O 1s, and Ni 2p edge, the photon energy was varied over a wide range
(452 eV – 475 eV; 525 eV – 550 eV; 845 eV – 865 eV, respectively) by sweeping the undulator
gap and the monochromator in parallel. A SPECS Phoibos 150 was used as the electron analyzer
and a 1D delay-line detector as the electron detector. The resolution of the spectrometer used was
70 meV. Together with the photon resolution, we have a resolution for XAS (CB) measurements
of 30 meV and for PES (VB) of around 80 meV. To compensate for polarization dependence of
the resonances a sample position with a magic incidence angle of 54.7 ° was chosen. 176 RIXS
experiments were performed at the Advanced Light Source, Berkeley at beamline 8.0.1, using the
iRIXS endstation equipped with two slit-less variable line-spacing (VLS) grating spectrographs.
177

Laboratory XPS measurements were performed using a Kratos Axis Ultra system with a base
pressure of < 1x10-9 Torr. A monochromatic AlKα (hν=1486.69 eV) source with a power of 150 W
was used for all measurements. For ultraviolet photoelectron spectroscopy (UPS) a Helium gas
discharge lamp was used to provide HeI excitation (hν=21.21 eV) for measurement of the material
work function. Pure metal samples (99.99%) were sputter cleaned until no contamination or carbon
was detectable prior to UPS measurement.
Electrochemical characterization was performed in a three-electrode configuration with a
Ag/AgCl reference electrode and carbon counter electrode using a Biologic SP-200 potentiostat.
To assess the conductivity of the samples, 50/350 mM Fe(CN)63-/4- was used as the electrolyte,
which was vigorously stirred during the experiments. 1 M KOH was used as the electrolyte to
determine the performance for the oxygen evolution reaction (OER). X-Ray diffraction (XRD)
data was collected on a Bruker Discover D8 XRD with a micro-focus X-ray source (Cu) and
VÅNTEC-500 large microgap detector. Scanning electron microscopy images were obtained with
a FEI Nova NanoSEM 450 microscope.

7.3 Results and Discussion
Figure 7.1(a) shows the current voltage curves (J-U) in 1 M KOH for p+-Si/a-TiO2/Ni(20 nm)
and p+-Si/a-TiO2/Ir(20 nm) samples. Redox peaks are observed prior to the onset to OER for the

125
p+-Si/a-TiO2/Ni sample at U = +1.44 V and +1.35 V vs. RHE corresponding to Ni(II)/Ni(III)
oxidations and reductions, respectively. No redox peaks are observed for p+-Si/a-TiO2/Ir. Both
OER catalyst iridium and nickel have comparable overpotentials in basic solution on inert
conductive glassy carbon. 178 However, Figure 7.1(a) shows that p+-Si/a-TiO2/Ir(20 nm) had an
overpotential 0.2 – 0.3 V higher than that of p+-Si/a-TiO2/Ni(20 nm) at a current density of
10 mA/cm2. In contact with a fast one-electron redox couple, ferric ferrocyanide (Fe(CN)63-/4-),
Figure 7.1(b), p+-Si/a-TiO2/Ir(20 nm) showed slower kinetics than p+-Si/a-TiO2/Ni(20 nm). The
current voltage plot for a p+-Si/a-TiO2/Au(20 nm) electrode also exhibited slow kinetics. It is noted
that both gold and nickel have similar work functions, while the work function of iridium is larger.
178,179

Thus, the trend in conductivity observed was Ni>Ir>Au, and does not directly correlate with

metal nobility or work function differences.

Figure 7.1: (a) J-U measurements for p+-Si/a-TiO2/Ni and p+-Si/a-TiO2/Ir sample in 1.0 M
KOH(aq). (b) J-U measurements for p+-Si/a-TiO2/M (M=nickel, iridium, gold) in 50/350 mM
Fe(CN)63-/4-(aq) solution.

The dependence of the conductivity on metal thickness for p+-Si/a-TiO2/metal was also
measured, as in Figure 7.2. Samples with thin nickel layers showed very little conductivity. A
metal thickness of >2 nm was needed to produce significant conductivity. For Iridium, charge
conduction was observed after a nominal metal thickness of 0.5 nm which gradually increased and
saturated after 10 nm. For Gold, no hole conduction could be detected up to 5 nm, and only above

126
10 nm the a-TiO2/Au layers were conductive. These results should be compared to the SEM
images of the TiO2/metal surfaces (Figure 7.3). Both nickel and iridium covered surfaces showed
no distinctive surface morphology changes, indicating a smooth surface coverage of the thin metal
layer. However, for gold deposition on TiO2, the formation of nanoparticles could be observed up
to a thickness of 5 nm. For 10 nm nominal gold layer thickness, these nanoparticles started to
interconnect and form a continuous but still porous layer.

Figure 7.2: J-U measurements for (a) p+-Si/a-TiO2/Ni, (b) p+-Si/a-TiO2/Ir, and (c)
p+-Si/a-TiO2/Au in 50/350 mM Fe(CN)63-/4-(aq) solution with different thicknesses of the metal
layer.

127

Figure 7.3: Scanning electron microscopy images of a-TiO2/M. M= (a) nickel, (b) iridium, and (c)
gold for different metal thicknesses. The scale bar is 200 nm.

128
Performance towards OER is a function of the effectiveness of the catalysts and the
conductivity of the system. The conductivity is strongly dependent of the energetic alignment of
the conduction and valence bands in the various materials. The energy band alignment between Si
and-TiO2 has been studied in detail previously 166, and we note that the p+-Si/a-TiO2 interface
formed an ohmic contact as indicated by the absence of band bending in TiO2 or p+-Si after
equilibration of the Fermi levels of both materials. Therefore, this interface should not significantly
affect the conductivity from the Si to the metal catalyst, and thus we have focused on the
a-TiO2/metal interface.
Reference samples of TDMAT a-TiO2 with and without metal contacts of nickel, iridium,
and gold with metal thicknesses ranging from 0.1 Å to 200 Å were prepared. X-ray diffraction
(XRD) spectra of these samples are shown in Figure 7.4. With increased metal layer thickness, the
(111) metal diffraction peak becomes more pronounced for all metals with additional (200), (220),
and (311) peaks for gold. No XRD signal could be observed for TDMAT a-TiO2 on any of the
samples.

Figure 7.4: XRD spectra for a-TiO2/M (M=Ni, Ir, Au) for different metal layer thicknesses.

Figure 7.5 shows high resolution XPS spectra of a-TiO2/M devices for Ni 2p, Ir 4f, Au 4f,
Ti 2p, and O 1s core levels with the valence band region. The valence band spectra for the bare

129
a-TiO2 sample showed the characteristic gap state at a binding energy of ~1 eV, which shows
no difference in intensity for bulk (Figure 7.6, θ = 0°) or surface sensitive (Fig. S4, θ = 70°) XPS.
Upon deposition of less than one Å of metal, the gap state could not be detected by laboratory XPS
due to the high intensity of the valence band states of the metal (Figure 7.5(d,l,h)). Thin layers of
nickel and iridium showed metal oxide peaks for nickel (Figure 7.5(a)) and iridium (Figure 7.5(e))
as indicated by solid lines. Metallic peaks (dashed lines) at a lower binding energy were observed
with increasing nickel and iridium thickness that did not obscure the metal oxide peaks, which
indicated that the metal oxide was on the metal surface. Depth profiling of-TiO2/Ni samples by
changing the photon energy and thus the inelastic mean free path (IMFP) of the emitted electrons
using synchrotron XPS revealed that the valence band nickel oxide peak decreased in intensity
with decreasing surface sensitivity (increasing photon energy), supporting the conclusion that the
metal oxide was on the surface of the metal layer. The opposite trend was observed for the metallic
nickel contributions evident by states at the Fermi energy (EF) (see Figure 7.7). Previous
investigations using Ambient Pressure XPS also showed the presence of a metallic nickel phase
for the samples in 1 M KOH under applied potential. 168 While no core level shifts for the metal
peaks (Ni 2p, Ir 4f, Au 4f) were observed (only changes in oxidation state), binding energy shifts
for the Ti 2p core level were observed for a-TiO2/Ni and a-TiO2/Ir samples. No shoulder or change
in peak shape (FWHM) for the Ti 2p core level was evident in Figure 7.5(b,f,j), indicating that
there was no oxidation state change for the a-TiO2, i.e., and no transition from Ti4+ to Ti3+ or lower
oxidation states. Thus, the single component for Ti 2p3/2 and Ti 2p1/2 core levels was assigned to
Ti4+, indicating that the shift in the Ti 2p core level was due to band bending and not to changes in
oxidation state.

130

Figure 7.5: XPS spectra for a-TiO2/metal systems for (a-d) nickel, (e-h) iridium and (i-l) gold
showing core level peaks for the metal, (a) Ni 2p, (e) Ir 4f, (i) Au 4f; (b, f, j) Ti 2p; (c, g, k) O 1s,
and (d, h, l) of the valence band. The metal overlayer thicknesses are shown in the graph. The
black dashed line shown in the Ti 2p core level plots indicates the position of bulk Ti 2p3/2 core
level peaks, whereas the dashed and solid lines in (a) and (e) indicate the metallic and oxide peak
position, respectively.

131

Figure 7.6: XPS spectra for TDMAT ALD a-TiO2 of the (a) Ti 2p and (b) O 1s core levels and (c)
of the valence band for different emission angles from θ = 0° (bulk sensitive) to θ = 70° (surface
sensitive) relative to the surface normal. With increased surface sensitivity (increased θ), an
increase in the oxygen shoulder at 532.5 eV was observed.

Figure 7.7: Valence band spectra of for different nickel thicknesses: (a) 0 nm, (b) 0.3 nm, (c)1.3
nm, and (d) 10 nm. The spectra were recorded at three different photon energies: 150 eV, 640 eV,
and 1100 eV corresponding also to the kinetic energy of electron from the upper valence band.
Hence, the inelastic mean-free path (IMFP) of the photoelectrons corresponds to l = 4.72 Å (Ni)
to 6.28 Å (a-TiO2) for EK = 150 eV, l = 11.19 Å (Ni) to 14.96 Å (a-TiO2) for EK = 640 eV, and

l = 16.64 Å (Ni) to 22.39 Å (a-TiO2) for EK = 1100 eV. Inelastic Mean-Free Path for elements
under investigation for relevant photoelectron energies are calculated by IMFP-TPP2M. 180

132
For pristine TDMAT a-TiO2 and each metal, the work function (Wf) was determined by
ultraviolet photoelectron spectroscopy (UPS) using He I excitation (Figure 7.8(a)). The values
were Wfa-TiO2 = 4.7±0.10 eV, WfAu = 5.05±0.22 eV, WfNi = 5.09±0.19 eV, WfIr = 5.32±0.12 eV.
The binding energy for the valence band minimum maximum for TDMAT a-TiO2 was
VBM = 2.94±0.10 eV .The band bending of the a-TiO2 inferred from the changes in Ti 2p3/2 peak
position shown in Figure 7.8(b) follows the trend of the differences in work functions between the
a-TiO2 and the metal, ΔWf. Differences between the magnitude of band bending and ΔWf were
attributed to factors such as the interface dipole and fermi level pinning, and will be discussed later
on.

Figure 7.8: He I ultraviolet photoelectron spectra (UPS) of TDMAT a-TiO2/metal, with nickel,
iridium, and gold showing (a) the work function and (b) valence band maximum. The metals were

133
sputter-cleaned until no contamination or carbon was detectable. (c) Energy position of
Ti 2p3/2 core level depending on contact metal and metal thickness. The values were extracted from
Figure 7.5.

To further investigate the a-TiO2/metal interface, resonant photoemission (resPES) and
resonant inelastic X-ray scattering (RIXS) measurements were employed, which use an elementspecific probe to look through the relatively thick metal layer and visualize changes in the valence
band states of a-TiO2. Standard X-ray photoelectron spectroscopy involves the measurement of
core level (CL, Figure 7.9, process I) and valence band (VB) states (Figure 7.9, process II) of a
specimen. Due to the distinguishable characteristic binding energies (EB) between elements for
core levels (binding energy difference is sufficiently large), different element and oxidations states
can be identified without difficulty. However, for valence band spectroscopy, this is not the case,
as every element has electronic states close to the fermi energy (EF), making identification of
different elemental contributions extremely challenging with standard laboratory X-ray
spectroscopy. Instead, tunable synchrotron radiation makes it possible to excite valence band states
in resonance and thus gives the possibility to distinguish between valence band states of different
elements. With titanium as an example, these processes are described by equations below.
Ti 2p– + ℏ“′′ → Ti 2pQ + ‘’£~
(Equation 7.1)
[.◊]º + ℏ“′′ → [.◊]ºGK + ‘’£~
(Equation 7.2)

Ti 2p– + [.◊]º + [0◊]ö + ℏ“ → Ti 2pQ + [.◊ ]º + [0◊]K (Equation 7.3)
Ti 2pQ + [.◊]º + [0◊]K → Ti 2p– + [.◊]ºGK + [0◊ ]ö + ‘’£~
(Equation 7.4)

Ti 2pQ + [.◊]º + [0◊]K → Ti 2p– + [.◊]ºGK + [0◊]K + ℏ“′ (Equation 7.5)
For regular XPS with a fixed photon energy (e.g. monochromatic AlKa radiation of 1486.69 eV),
Equation 7.1 depicts the excitation of a Ti 2p core level electron above the vacuum level (Evac).
Equation 7.2 shows the excitation of a valence band electron above Evac. Since the photon energy
is high enough to excite core levels above Evac (ℏw’’>EB,Ti 2p+Wf), two different final states can
be found.

134

Figure 7.9: Illustration of possible X-ray spectroscopy excitation and decay channels: I.
Excitation of core level electron above the vacuum level (EVac); II. Excitation of valence band (VB)
electron above the vacuum level; III. Resonant excitation of a core level electron into unoccupied
states of the conduction band (CB); IV. After process III, an electron from the valence band refills
the core hole transferring the energy to the initial excited core level electron thus exciting it above
EVac (participator decay); V. After process III, an electron from the valence band refills the core
hole with the emission of a photon.

If we consider tunable photon energy with specific value of ℏw that electron in CL can be
excited into empty states of the conduction band, we will obtain an intermediate state shown in
Figure 7.9 process III (resonant excitation). It has to be noted that this (optical) process is governed
by dipole selection rules, indicating that all states (CL, VB and CB) need to be reached within the
same atom. 175 High elemental sensitivity can thus be realized with this technique. The
intermediate state can relax into different final states through several routes with only two
discussed here (Figure 7.9, processes IV and V). Both have commonalities with the core hole
(created through Figure 7.9 process III) refilled by a valence electron. The energy released can

135
either be transferred to the initial excited core electron in the conduction band so it can be
excited above Evac (Figure 7.9, process IV, participator Auger decay) or leave the atom as a photon
with a characteristic energy. (Figure 7.9, process V)
It is evident that when the tunable photon energy is high enough (ℏw>EVB+Wf), Figure 7.9
process II also happens that the same final state for photoelectron spectroscopy channel can be
reached as shown in Equation 7.2 and Equation 7.4. Though the initial excitation in Equation 7.3
is performed by resonant excitation, the valence band contribution from the corresponding element
is thus enhanced. By directly subtracting the off-resonant valence band spectrum as a background
from the on-resonant excitation, the valence band spectrum from single elemental species can be
obtained. As comparison, the background subtraction is not necessary for the optical channel in
Equation 7.5 as the emitted photons have a characteristic energy ℏw’ for each element.
Usually, the photon energy is scanned across the respective absorption edge (Ti L2/3 edge for
excitation from Ti 2p core levels) during X-ray absorption experiments. At each individual photon
energy, a valence band or X-ray emission spectrum can be recorded, which results in resonant
Photoelectron spectroscopy (resPES) or resonant inelastic X-ray spectroscopy (RIXS) maps.
Traditionally, the photoelectron intensity is plotted versus the binding energy (or sometimes
kinetic energy EK = ℏw-EB), whereas in RIXS spectra, the photon yield is shown versus the loss
energy (Eloss = ℏw-ℏw’) or emission energy ℏw’.
The resPES color contour plot for-a-TiO2 at the titanium L2/3 edge for the a-TiO2/Ni system
is shown in Figure 7.10. It shows valence band XPS spectra recorded as a function of the exciting
X-ray energy as scanned across the Ti 2p absorption edge. The ordinate is the binding energy of
the initially formed excited valence state (-1 eV to 20 eV) and the abscissa reflects the excitation
energy (452 eV – 475 eV). Off-resonant contributions to the spectra were subtracted using an offresonant reference spectrum for the same sample at an excitation energy of 452 eV.

136

Figure 7.10: Resonant photoemission maps at the Ti L3,2 edge of (a) p+-Si/a-TiO2, (b)
p+-Si/a-TiO2/Ni(0.3 nm), and (c) p+-Si/a-TiO2/Ni(1.3 nm). The insets on the top show the total
electron yield (TEY) mode XAS spectrum, and the inset on the right the valence band spectrum at
464 eV. In all spectra, the off-resonant contributions were subtracted using the off-resonant VB
spectra at 452 eV. The black arrows on the right panel indicate the position of the gap state.

Figure 7.10(a) shows the valence band resPES map for TDMAT a-TiO2 at the Ti L3,2 edges.
The characteristic gap state is clearly evident in the resPES map at a binding energy of 1 eV for
both the L3 (459.6 eV photon energy) and L2 edges (465.2 eV photon energy), as well as in the
valence band spectra shown in the righthand side panel. The XPS of the pristine a-TiO2 showed
no difference between bulk and surface Ti 2p core and valence band levels (Figure 7.6), suggesting
that the gap staes was the same throughout the bulk of the sample. After deposition of 0.3 nm of
Nickel, no gap state was observable in the resPES map, Figure 7.10(b). When the nickel thickness
was further increased to 1.3 nm, two gap states could now be observed with one peak at a slightly
reduced binding energy of 0.7 eV and a second at 2.6 eV, as in Figure 7.10(c) righthand side panel.
For this sample, the valence band on the right was measured separately with increased dwell time
to improve signal to noise ratio of the spectra after recording of the initial map.
The resPES maps for the Oxygen K and Ni L3 edges are shown in Figure 7.11and Figure 7.12.
Figure 7.11(a) represents the Oxygen K edge map for pristine TDMAT a-TiO2. The oxygen
valence band states observed between binding energies of 5 and 10 eV and photon energies
between 530 eV and 535 eV for the pristine a-TiO2 were absent in the maps with increased nickel
thickness. While for nickel layers <0.3 nm thick, the nickel was completely oxidized (Figure 7.5(a),

137
Figure 7.11(b)), and for thicker layers, the nickel was primarily metallic at the TiO2 surface,
with NiOx at the nickel surface. Because of the limited information depth and element sensitivity
of resPES, the oxide peaks observed for thicker layers should be attributed to oxygen of NiOx
rather than to oxygen in a-TiO2. The corresponding Ni L3 edge maps are shown in Figure 7.12.
The two prominent features at photon energies of 853 and 854.7 eV were attributed to Ni2+ (NiO)
and Ni3+ (NiOOH) states. 181,182 The valence band maximum shifted toward the Fermi energy with
increasing nickel thickness from 0.3 nm to 1.3 nm.

Figure 7.11: Resonant photoemission maps at the Oxygen K edge of (a) p+-Si/a-TiO2,
(b) p+-Si/a-TiO2/Ni(0.3 nm) and (c) p+-Si/a-TiO2/Ni(1.3 nm). The insets on the top show the TEY
XAS spectrum and the inset on the left the valence band spectrum at 531 eV. In all spectra, the offresonant contributions were subtracted using the off-resonant VB spectra at 525 eV.

138

Figure 7.12: Resonant photoemission maps at the Ni L3 edge of (a) p+-Si/a-TiO2/Ni(0.3 nm) and
(b) p+-Si/a-TiO2/Ni(1.3 nm). The insets on the top show the TEY XAS spectrum, and the insets on
the left the valence band spectrum at 853 eV. In all spectra, the off-resonant contributions were
subtracted using the off-resonant VB spectra at 846 eV.

Oxygen vacancies in a-TiO2 result in n doping. The localization of the mobile electron
produces a partially populated t2g band which is visible by XPS in the band gap at 1 eV binding
energy. In RIXS, this is visible due to resonant excitation into the eg band followed by de-excitation
from the t2g band resulting in a dd loss feature. 171 To illustrate the correlation between valence
band gap state (XPS) and dd loss feature in RIXS, both measurements were performed on a sample
with and without gap state. Figure 7.13 shows the valence band and RIXS spectra for pristine and
annealed TDMAT a-TiO2. For pristine a-TiO2, the gap state is visible in the XPS valence band,
and a dd transition peak can be observed in the RIXS spectrum. Upon annealing, the gap state in
the XPS spectra disappeared and no dd transition in RIXS was visible.

139

Figure 7.13: (a) XPS valence band spectra and (b) RIXS spectra at the Ti L3 eg resonance for
pristine and annealed a-TiO2.

For a-TiO2/Ir and a-TiO2/Au RIXS spectra (Figure 7.14(a-b)) the gap state was visible as a
characteristic dd transition at an energy loss of 1.6 to 1.8 eV with the elastic X-ray scattering peak
at 0 eV. To obtain a signal directly proportional to the intensity of the gap state, the intensity of
the dd transition must be normalized to that of the elastic peak. In Figure 7.14(c), the normalized
gap state intensity and its position are given as a function of metal coverage for both iridium and
gold. No clear change was evident, suggesting that the gap state was unaffected by the deposition
of iridium or gold.

140

Figure 7.14: RIXS spectra of (a) a-TiO2/Ir and (b) a-TiO2/Au at the Ti L3 t2g resonance. The
position of the characteristic dd transition is indicated. (c) Intensity ratio of dd transition to elastic
peak. A change in this normalized dd intensity gives evidence of changes in the gap state.

The resPES results make it possible to determine the partial density of states for titanium,
oxygen, and nickel of the valence band of TiO2/Ni layer. Figure 7.15 summarized the density of
states at valence band for the-TiO2/Ni interface with varying nickel thickness. Pristine TDMAT
a-TiO2 shows a clear titanium derived gap state in the valence band at 1 eV (Figure 7.15(a)). The
gap state of the pristine a-TiO2 exhibits no change in intensity between the bulk and surface by
XPS measurements (Figure 7.6(c)). Thus, the gap state extends throughout the bulk of the a-TiO2.
The resPES data showed that the gap state was not present at the TiO2/Ni interface after deposition
of a thin nickel layer (0.3 nm, Figure 7.15(b)), which was shown to be completely oxidized to
NiOx, as in Figure 7.5(a) and Figure 7.7(b). This change of the gap state upon nickel oxide
deposition is attributed to a chemical change only at the a-TiO2 surface. We propose that for thin
layers of nickel, the NiOx layer at the TiO2 surface provides additional oxygen species to oxidize
Ti interstitials (Ti3+) in the a-TiO2 at the surface and thus creates a thin interface layer of a-TiO2
without the gap state. NiOx also possesses a different work function than nickel which can be rather

141
large (6.7 eV) but also drops rapidly upon exposure to H2O, O2, and CO2. 183 High work
function materials can lead to a depletion of Ti3+ states and increase the junction resistance through
band bending in a-TiO2. 167 In a classical picture – for no reaction between metal and a-TiO2 – a
high work function of the contact material the Fermi energy would pass through the gap state in
a-TiO2 resulting in a reduction of charge carrier population of the gap state at the interface and
reduction or complete blocking of hole conduction. The bulk properties of a-TiO2 would not be
affected. For thin layers of gold or iridium on a-TiO2, no change in the gap state intensity was
observed (Figure 7.14).

Figure 7.15: Partial density of valence band states for titanium (red) and nickel (grey) derived
states at the a-TiO2-Ni interface for pristine (a) a-TiO2, (b) a-TiO2/Ni(0.3 nm), and (c)
a-TiO2/Ni(1.3 nm). The partial density of states (pDOS) is obtained by calculating the difference
between on and off-resonance valence band spectra a the Ti 2p and Ni 2p X-ray absorption edge.

For nickel thickness greater than 0.3 nm, the nickel at the a-TiO2 interface becomes metallic.
Depth profiling revealed the presence of this buried metallic nickel phase below the NiOx surface
layer. For this case, resPES showed the existence of two gap states at the a-TiO2/Ni interface with
binding energies of 2.6 eV and 0.7 eV (Figure 7.15(c)). We attribute the first gap state to be Ti

142
related and arise from partial reduction of the a-TiO2 surface by Ni. The different binding
energy can be related to possible change in Ti 3d density or change in local symmetry. Chemical
reaction and hybridization of Ni with a-TiO2 gives rise to the second gap state at 2.6 eV attributing
it to Ni-Ti bond formation. Both gap states essentially fill the energy range from VBM to the Fermi
energy. This effect was only observed for the a-TiO2/metal interface, with the bulk a-TiO2
unaffected. With the increase in the density of states in the band gap between VBM and Fermi
energy, the a-TiO2 will exhibit metallic character at the a-TiO2/metal interface and thus enhance
the conduction across the a-TiO2/M interface. For a-TiO2/Ir and a-TiO2/Au, a similar behavior was
not detected as the gap state showed no change over the range of metal deposition (Figure 7.14).
With the help of XPS and UPS, the band alignment between metal layer and TDMAT a-TiO2
was also determined. Buried junctions are, in principle, accessible by XPS if they are not too
remote from the top surface. At heterojunctions of semiconductors, thermodynamic equilibration
of the electrochemical potentials of electrons in the semiconductors (Fermi level), Ÿ, is established
by the exchange of delocalized charges, inducing band bending due to non-compensated ionized
doping atoms in the space charge regions, and/or by formation of an interface dipole ⁄ at the
interface. In general, the thermodynamic equilibration requires difference in electrochemical
potential to be equal to the change in energy due to the band bending and surface dipole in the
materials. The energy band relationships in semiconductor junctions can be determined from core
level and valence band spectroscopy using X-ray photoelectron and the secondary electron
emission cut-off (i.e. work function measurements) from ultraviolet photoemission spectroscopy.
184

Because of the relatively small sampling depth of XPS, a general procedure for energy band
profile determination is to measure the valence band maximum to core level binding energy
differences 185,186 and then monitor variations of the well-defined substrate and adsorbate core level
binding energy positions during step-by-step deposition of the contact material with typical
thicknesses from 5 to 30 Å. 184,186 The individual band bending in a-TiO2/metal was extracted from
the core level shift of the Ti 2p core level and is summarized in Figure 7.8(b). The exact values
which were extracted from XPS and UPS measurements (Figure 7.5, Figure 7.8) and resonant XRay spectroscopy (Figure 7.10, Figure 7.14) are consolidated in Table 7.1. Figure 7.16 shows the
summary of band alignments for a-TiO2/Ni (Figure 7.16(a)), a-TiO2/Ir (Figure 7.16(b)), and

143
a-TiO2/Au (Figure 7.16(c)). A detailed description on the calculation of the band-energy
diagrams can be found in a previous study. 166 Both a-TiO2/Ni and a-TiO2/Ir show upward band
bending in the a-TiO2 close to the metal interface while a-TiO2/Au shows no bending. In all three
cases, the gap state in the a-TiO2 extends to the TiO2/M interface. No change in intensity was
observed from bulk to surface for pristine a-TiO2 without metal layer. In a previous study, it was
shown that the band bending in the a-TiO2 at the interface for a a-TiO2/M device is fixed and
independent from the applied potential in electrolyte. 168
Table 7.1: Parameters used for band-energy diagrams of a-TiO2/M with M=nickel, M=iridium,
and M=gold as shown in Figure 7.16. EB is the binding energy. EBB is the band bending at the
interface. The band gap for TiO2 was taken from previous studies. 74,166

Unit (eV)

a-TiO2/Ni

a-TiO2/Ir

a-TiO2/Au

F_TiO2

4.70

4.70

4.70

F_M

5.09

5.32

5.05

Eg_TiO2

3.34*

3.34*

3.34*

EB_Ti 2p, bulk

459.25

459.25

459.25

EB_M, bulk

852.6

60.9

84

EBB_TiO2

0.19

0.92

0.05

d (dipole)

0.20

-0.30

0.30

EB_defect, bulk

1.0

1.0

1.0

FWHM_defect,
bulk

0.88

0.88

0.88

144
EB_defect,
interface

0.7, 2.6

0.08

0.95

FWHM_defect,
interface

1.4, 1.2

0.88

0.88

Figure 7.16: Band-energy diagrams for a-TiO2/M with M= (a) nickel, (b) iridium, and (c) gold.
All numeric values are in eV. d is the interface dipole energy difference between TDMAT a-TiO2
and the metal. The hashed region between the VBM and CBM in the a-TiO2 indicates the position
of the gap state with the FWHM taken as its width. The values can also be found in Table 7.1.

For a-TiO2/Ni, the gap state is below the Fermi level with the gap between VBM and Fermi
level completely filled with gap states at the TiO2/Ni interface. This is shown in Figure 7.16(a) by
extending the gap state (dashed region) to the VBM. For thin layers of nickel where the nickel was
completely oxidized, no interfacial gap state was detected (Figure 7.15(b)), which results in a thin
interfacial layer where the transport channel (the gap state) for holes from the substrate to the
catalyst was not continuous. In agreement with this observation, conductivity measurements
showed no conduction for a-TiO2/NiOx samples (Figure 7.2(a), up to 0.7 nm).

145
For the a-TiO2/Ir, no change of the gap state at the interface was detected. Also, the Fermi
energy passes through the top half of the gap state at the a-TiO2/Ir interface, resulting in a thin
layer where the gap state has reduced charge carrier population. Without the presence of the quasimetallic channel and lower charge carrier density, the lower conducting behavior of a-TiO2/Ir than
a-TiO2/Ni can be understood. The a-TiO2/Au showed no significant change in binding energy of
the Ti 2p core levels which signifies that there is no upward band bending. Thus, for a-TiO2/Au,
the band-energy diagram depicts a flat band condition for hole transport through the gap state to
the metal (i.e. there is not an electric field that will propel holes through the a-TiO2 toward the Au).
This leads to the poorer performance of the a-TiO2/Au system compared to the other two.
Furthermore, it was observed that for thin layers of gold (up to 5 nm) in a-TiO2/Au showed the
presence of gold nanoparticles and no current was observed in these situations. This can be
understood if in the a-TiO2/Au system the gold nanoparticles exhibit a nonuniform barrier height
contact to the electrolyte (“pinch-off”) where the Au/electrolyte interface energetics are not
determined by the work function of the gold but rather by the a-TiO2. 187 This explanation is
supported by the fact that for a uniform, but porous thin gold layer, the a-TiO2/Au system then
conducts current to the electrolyte, as in Figure 7.2. In the presence of gold nanoparticles, the
contact behavior of the a-TiO2/Au “pinch-off” system will be dominated the a-TiO2/electrolyte
interface and the observable current is dictated by the energetics imposed by the a-TiO2/electrolyte
junction. In our previous studies we have shown that for a-TiO2/electrolyte systems an applied bias
will be compensated by a-TiO2 bend bending close to the electrolyte interface and will lead to the
gap state getting pushed above the Fermi energy at the interface 168,188 This results in complete
depopulation of the gap state at the interface and formation of a thin depletion region with holeblocking properties. No current was observed under these conditions.
Although the difference in work function between a-TiO2 and metal provided a general
guideline for band alignment and conductivity evaluation for a-TiO2/M interface, e.g. metals with
a work function less than a TiO2 provided higher conductivities, it cannot describe the difference
in behavior for metals with similar work function. Here we showed that further effects have to be
taken in account, i.e. reactivity of the metal and its ability to oxidize or reduce interfacial TiO2 (in
the presence or absence of additional oxygen sources) and local energy effects of nanoparticles.

146
7.4 Conclusion and Outlook
We investigated the conduction behavior between a “leaky” a-TiO2 protection layer for
photoanodes and metal catalysts (iridium, nickel, and gold). Resonant X-ray spectroscopy revealed
a critical interfacial reaction between a-TiO2 and nickel catalyst which can create either a nonconductive or a quasi-metallic surface layer on a-TiO2 depending on the chemical state of the
catalysts (nickel oxide or metallic nickel respectively). For a photoanode protection and catalyst
layer system of a-TiO2/Ni, it is required to have a minimum thickness of ~2 nm to prevent complete
oxidation of the a-TiO2 surface. Else a barrier for hole transport through the gap state will be
created by elimination of the gap state at the surface of a-TiO2. The superior performance of the
a-TiO2/Ni system over the a-TiO2/Ir system is the result of the intrinsic formation of a quasimetallic interface layer by gap steates between a-TiO2 and nickel. It allows holes to conduct
between the semiconductor through the thick but “leaky” a-TiO2 protection layer to the nickel
catalyst. This results in an ohmic contact between a-TiO2/Ni. Non-reactive metals which form
nanoparticles can even lead to complete loss of catalytic current due to “pinch-off” effect and hole
barrier formation under anodic bias. This study suggests that a “reductive” metal on top of “leaky”
a-TiO2 is necessary for formation of quasi metallic interface layer between a-TiO2 and catalyst.
Future direction involves applying the resPES and RiXS to understand the degradation
mechanism when regular XPS cannot explain. Operando study is a powerful tool which would
help to examine the surface chemistry during operation. With resonance, we can also distinguish
the different valence and conduction band contributions (pDOS), as catalysis involves conduction
and valence band for reduction and oxidation processes where regular XPS cannot distinguish. It
would be valuable when a complicated catalyst with more than one kind of atom (for example,
alloy, bimetallic, MoS2...) or for OER (which involves oxides) that contribution from different
atomic character can be examined.

147

CHAPTER 8
Summary
This thesis focuses on understanding comprehensive aspects from photonic design, interface
study, to device integration. We first focused on enhancing absorption via nanophotonic design of
a light absorber. III-V compound semiconductor nanowire arrays are promising candidates for
energy harvesting due to their high volumetric absorption. Uniform nanowire arrays exhibit high
absorption at certain wavelengths due to strong coupling into resonant waveguide modes. We have
simulated and experimentally demonstrated near-unity, broadband absorption in sparse InP
nanowire arrays (< 5% fill fraction) with multi-radii and tapered nanowire array designs, where
incident visible light can be coupled into continuous waveguide modes in taper cone structures.
For both designs, the polymer-embedded arrays achieved ~ 90% broadband absorption (λ = 400900 nm) in less than 100 nm planar equivalence of InP. The addition of a silver back reflector
increased this broadband absorption to ~ 95%.
To realize high solar-to-fuel efficiency in PEC devices, it is necessary to maintain a catalytic
current density close to the light limiting photocurrent density for a solar-driven light absorber,
which can be fulfilled when catalyst ensembles are highly transparent. We report a monolithic
photocathode device architecture that exhibits significantly reduced surface reflectivity,
minimizing parasitic light absorption and reflection losses. A tailored multifunctional crystalline
titania interphase layer acts as a corrosion protection layer, with favorable band alignment between
the semiconductor conduction band and the energy level for water reduction, facilitating electron
transport at the cathode−electrolyte interface. It also provides a favorable substrate for adhesion of
high-activity Rh catalyst nanoparticles. Under simulated AM 1.5G irradiation, solar-to-hydrogen
efficiencies of 19.3 and 18.5% are obtained in acidic and neutral electrolytes, respectively. The
system reaches a value of 0.85 of the theoretical limit for photoelectrochemical water splitting for
the energy gap combination employed in the tandem-junction photoelectrode structure.
Solar-driven reduction of carbon dioxide represents a carbon neutral pathway for the synthesis
of fuels and chemicals. We report here results for solar-driven CO2 reduction using a gas diffusion
electrode (GDE) directly powered by a photovoltaic cell. A GaInP/GaInAs/Ge triple junction
photovoltaic cell was used to power a reverse-assembled gas diffusion electrode employing a Ag

148
nanoparticle catalyst layer. The device had a solar-to-CO energy conversion efficiency of 19.1
% under simulated AM 1.5G illumination at 1 Sun. The use of a reverse-assembled GDE prevented
transition from a wetted to a flooded catalyst bed and allowed the device to operate stably for >150
h with no loss in efficiency. Outdoor measurements were performed under ambient solar
illumination in Pasadena, CA, resulting in a peak solar-to-CO efficiency 18.7 % with a CO
production rate of 47 mg⋅cm-2 per day and a diurnal-averaged solar to fuel conversion efficiency
of 5.8 %. The high efficiency and stability of the system suggests that reverse-assembled GDEs
are a promising path to producing chemical fuels from CO2 and sunlight.
We further demonstrated light management strategies to create highly active and effectively
transparent catalyst structures. Covering over 50% of the surface of a light absorber with an array
of high-refractive-index TiO2 nanocones imparted antireflective behavior (< 5% reflectance) to the
surface and allowed > 85% transmission of broadband light to the underlying Si, even when thick
metal contacts or opaque catalyst coatings were deposited on areas of the light-facing surface that
were not directly beneath a nanocone. Three-dimensional full-field electromagnetic simulations
for the 400 – 1100 nm spectral range showed that incident broadband illumination couples to
multiple waveguide modes in the TiO2 nanocones, reducing interactions of the light with the metal
layer. A proof-of-concept experimental demonstration of light-driven water oxidation was
performed using a p+n-Si photoanode decorated with an array of TiO2 nanocones additionally
having a Ni catalyst layer electrodeposited onto the areas of the p+n-Si surface left uncovered by
the TiO2 nanocones. This photoanode produced a light-limited photocurrent density of ~ 28 mA
cm-2 under 100 mW cm-2 of simulated Air Mass 1.5 illumination, equivalent to the photocurrent
density expected for a bare planar Si surface even though 54% of the front surface of the Si was
covered by an ~ 70 nm thick Ni metal layer.
Another approach is developed with effectively transparent catalyst consisting of arrays of
micron-scale triangular cross-sectional metal grid fingers. The capability of redirecting the
incoming light to the open areas of the PEC cell reduces the overall shadow loss. Numerical
calculations using full wave electromagnetic simulations were used to investigate the optical
response and determine the optimal geometry and length scale. We found that a mesoscale Ag grid
array with triangular cross-section lines and metal coverage of > 50% exhibits negligible additional
reflection loss. Our designs feature photonic structures to enable high absorption light absorbers,

149
and effectively transparent catalyst layers for PEC cell are general approaches not limited to
single reaction or specific photovoltaic system. Together with the capability of scalable processes
through printing technology for catalysts using conductive inks and electroplating, it will be a
critical step to advances in the field of solar fuel generation and all other related optical applications.
It also opens up a new route for photoelectrochemical applications toward large-scale
manufacturing.
Last but not least, the interfacial conduction mechanism between the commonly employed
semiconductor protection layer titanium dioxide and metal catalysts is investigated. While iridium
and nickel exhibit similar overpotential for oxygen evolution reaction in alkaline media, a-TiO2/Ir
requires higher overpotential than a-TiO2/Ni to achieve the same current density. A combinatorial
approach of electrochemistry, X-ray photoelectron spectroscopy, and resonant X-ray spectroscopy
reveals the correlation between interfacial metal-TiO2 properties and conduction. While both
nickel and iridium metals impose band bending upon a-TiO2, only nickel creates an interfacial
quasi metallic a-TiO2 surface due to creation of additional interface gap states. The use of noble
metal catalysts (gold, iridium) will result in band bending or formation of a barrier while nonnoble catalysts (nickel) create an ohmic contact after deposition of a minimum metal layer
thickness.
For visualization, Figure 8.1 summarizes the state-of-the-art solar fuel devices with
contribution from this thesis for both water splitting and CO2 reduction with four main categories
of PV-Electrolyser, photoanode, photocathode, and PV plus photocathode.

150
PV-Electrolyser

InGaP/GaAs/GaInNAs + Pt/Ir

III-V/a-Si + CuAg/Ir

GaInP/GaInAs/Ge + Ag GDE/Ni

PV

5.6% to
Ethylene,
Ethanol and
Propanol

19% to CO

COCatalysts
2R catalyst
OER catalyst

30% to H2

Photoanode
Ni/InGaP/GaAs + Pt

Ir/3J SiGe + Ru-organic complex

Ni/InGaP/GaAs + Pd/C/Ni

OER catalyst
photoanode
COCatalysts
2R catalyst

Photocathode

10% to H2

4.6% to formate

Rh/GaInP/GaInAs + Ru

Ag/GaInP/GaInAs/Ge + Ni

10% to formate

Cu/GaInP/GaInAs/Ge + Ni
1.1% to Ethylene,
Ethanol

COCatalysts
2R catalyst
photocathode
OER catalyst

PV + Photocathode

19.3% to H2

3% to CO

perovskite + Si/CuAg + Ir

PV

CO
photocathode
CO22RR catalyst
catalyst
photocathode
catalysts
OER catalyst
OER catalyst

1.5% to hydrocarbons
and oxygenates

Figure 8.1: State-of-the-art solar fuel device for water splitting and CO2 reduction with category
of PV-Electrolyser, photoanode, photocathode, and PV plus photocathode.

151

References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]

Davis, S. J.; Lewis, N. S.; Lewis, N. S.; Shaner, M. R.; Shaner, M.; Aggarwal, S.; Arent, D.;
Azevedo, I. L.; Benson, S. M.; Bradley, T.; et al. Net-Zero Emissions Energy Systems. Science
2018, 360 (6396), eaas9793.
Armstrong, R. C.; Wolfram, C.; de Jong, K. P.; Gross, R.; Lewis, N. S.; Lewis, N. S.;
Boardman, B.; Ragauskas, A. J.; Ehrhardt-Martinez, K.; Crabtree, G.; et al. The Frontiers of
Energy. Nat. Energy 2016, 1 (1), 15020.
Lewis, N. S.; Lewis, N. S. Toward Cost-Effective Solar Energy Use. Science 2007, 315
(5813), 798–801.
Jia, J.; Seitz, L. C.; Benck, J. D.; Huo, Y.; Chen, Y.; Ng, J. W. D.; Bilir, T.; Harris, J. S.;
Jaramillo, T. F. Solar Water Splitting by Photovoltaic-Electrolysis with a Solar-to-Hydrogen
Efficiency Over 30%. Nature Communications 2016, 7, 13237.
Young, J. L.; Steiner, M. A.; Döscher, H.; France, R. M.; Turner, J. A.; Deutsch, T. G. Direct
Solar-to-Hydrogen Conversion via Inverted Metamorphic Multi-Junction Semiconductor
Architectures. Nat. Energy 2017, 2, 17028.
Cheng, W.-H.; Richter, M. H.; May, M. M.; Ohlmann, J.; Lackner, D.; Dimroth, F.;
Hannappel, T.; Atwater, H. A.; Lewerenz, H.-J. Monolithic Photoelectrochemical Device for
Direct Water Splitting with 19% Efficiency. ACS Energy Lett. 2018, 1795–1800.
Qiu, Y.; Liu, W.; Chen, W.; Zhou, G.; Hsu, P.-C.; Zhang, R.; Liang, Z.; Fan, S.; Zhang, Y.;
Cui, Y. Efficient Solar-Driven Water Splitting by Nanocone BiVO4-Perovskite Tandem Cells.
Science Advances 2016, 2 (6), e1501764–e1506473.
Huang, D.; Wang, K.; Yu, L.; Nguyen, T. H.; Ikeda, S.; Jiang, F. Over 1% Efficient Unbiased
Stable Solar Water Splitting Based on a Sprayed Cu 2ZnSnS 4Photocathode Protected by a
HfO 2Photocorrosion-Resistant Film. ACS Energy Lett. 2018, 1875–1881.
Reece, S. Y.; Hamel, J. A.; Sung, K.; Jarvi, T. D.; Esswein, A. J.; Pijpers, J. J. H.; Nocera, D.
G. Wireless Solar Water Splitting Using Silicon-Based Semiconductors and Earth-Abundant
Catalysts. Science 2011, 334 (6056), 645–648.
May, M. M.; Lewerenz, H.-J.; Lackner, D.; Dimroth, F.; Hannappel, T. Efficient Direct Solarto-Hydrogen Conversion by in Situ Interface Transformation of a Tandem Structure. Nature
Communications 2015, 6, 8286.
Shaner, M. R.; Atwater, H. A.; Lewis, N. S.; McFarland, E. W. A Comparative
Technoeconomic Analysis of Renewable Hydrogen Production Using Solar Energy. Energy
Environ. Sci. 2016, 9 (7), 2354–2371.
Liu, J.; Liu, Y.; Liu, N.; Han, Y.; Zhang, X.; Huang, H.; Lifshitz, Y.; Lee, S.-T.; Zhong, J.;
Kang, Z. Water Splitting. Metal-Free Efficient Photocatalyst for Stable Visible Water Splitting
via a Two-Electron Pathway. - PubMed - NCBI. Science 2015, 347 (6225), 970–974.
Hisatomi, T.; Domen, K. Reaction Systems for Solar Hydrogen Production via Water Splitting
with Particulate Semiconductor Photocatalysts. Nature Catalysis 2018 1:8 2019, 2 (5), 387–
399.
Ra’di, Y.; Simovski, C. R.; Tretyakov, S. A. Thin Perfect Absorbers for Electromagnetic
Waves: Theory, Design, and Realizations. Phys. Rev. Applied 2015, 3 (3), 037001.
Soukoulis, C. M.; Wegener, M. Past Achievements and Future Challenges in the Development
of Three-Dimensional Photonic Metamaterials. Nature Photonics 2011, 5 (9), 523–530.
Landy, N. I.; Sajuyigbe, S.; Mock, J. J.; Smith, D. R.; Padilla, W. J. Perfect Metamaterial
Absorber. Phys. Rev. Lett. 2008, 100 (20), 207402.
Sakr, E.; Dhaka, S.; Bermel, P. Asymmetric Angular-Selective Thermal Emission; 2016; Vol.
9743, pp 97431D–97431D–8.

152
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]

[31]
[32]
[33]
[34]
[35]
[36]
[37]

Argyropoulos, C.; Le, K. Q.; Mattiucci, N.; D'Aguanno, G.; Alu, A. Broadband Absorbers
and Selective Emitters Based on Plasmonic Brewster Metasurfaces. Phys. Rev. B 2013, 87
(20), 6.
Liu, X. L.; Tyler, T.; Starr, T.; Starr, A. F.; Jokerst, N. M.; Padilla, W. J. Taming the
Blackbody with Infrared Metamaterials as Selective Thermal Emitters. Phys. Rev. Lett. 2011,
107 (4), 4.
Khanikaev, A. B.; Wu, C. H.; Shvets, G. Fano-Resonant Metamaterials and Their
Applications. nanoph 2013, 2 (4), 247–264.
Liu, N.; Mesch, M.; Weiss, T.; Hentschel, M.; Giessen, H. Infrared Perfect Absorber and Its
Application as Plasmonic Sensor. Nano Lett. 2010, 10 (7), 2342–2348.
Zhu, J.; Yu, Z. F.; Burkhard, G. F.; Hsu, C. M.; Connor, S. T.; Xu, Y. Q.; Wang, Q.; McGehee,
M.; Fan, S. H.; Cui, Y. Optical Absorption Enhancement in Amorphous Silicon Nanowire and
Nanocone Arrays. Nano Lett. 2009, 9 (1), 279–282.
Fountaine, K. T.; Kendall, C. G.; Atwater, H. A. Near-Unity Broadband Absorption Designs
for Semiconducting Nanowire Arrays via Localized Radial Mode Excitation. Opt. Express
2014, 22 (S3), A930–A940.
Salisbury, W. W. Absorbent Body for Electromagnetic Waves. 1952.
de Abajo, F. J. G. Colloquium: Light Scattering by Particle and Hole Arrays. Rev. Mod. Phys.
2007, 79 (4), 1267–1290.
Kats, M. A.; Capasso, F. Ultra-Thin Optical Interference Coatings on Rough and Flexible
Substrates. Applied Physics Letters 2014, 105 (13), 3.
Laroche, M.; Albaladejo, S.; Gomez-Medina, R.; Saenz, J. J. Tuning the Optical Response of
Nanocylinder Arrays: an Analytical Study. Phys. Rev. B 2006, 74 (24), 5.
Diedenhofen, S. L.; Grzela, G.; Haverkamp, E.; Bauhuis, G.; Schermer, J. J.; Rivas, J. G.
Broadband and Omnidirectional Anti-Reflection Layer for III/v Multi-Junction Solar Cells.
Solar Energy Materials and Solar Cells 2012, 101, 308–314.
Basu, S.; Zhang, Z. M.; Fu, C. J. Review of Near-Field Thermal Radiation and Its Application
to Energy Conversion. Int. J. Energy Res. 2009, 33 (13), 1203–1232.
Atwater, H.; Polman, A.; Kosten, E.; Callahan, D.; Spinelli, P.; Eisler, C.; Escarra, M.;
Warmann, E.; Flowers, C. Nanophotonic Design Principles for Ultrahigh Efficiency
Photovoltaics; Linke, H., Borgstrom, M., Pullerits, T., Samuelson, L., Sundstrom, V., Inganas,
O., Eds.; Amer Inst Physics: Melville, 2013; Vol. 1519, pp 17–21.
Callahan, D. M.; Munday, J. N.; Atwater, H. A. Solar Cell Light Trapping Beyond the Ray
Optic Limit. Nano Lett. 2012, 12 (1), 214–218.
de Abajo, F. J. G. Graphene Plasmonics: Challenges and Opportunities. ACS Photonics 2014,
1 (3), 135–152.
Brar, V. W.; Jang, M. S.; Sherrott, M.; Lopez, J. J.; Atwater, H. A. Highly Confined Tunable
Mid-Infrared Plasmonics in Graphene Nanoresonators. Nano Lett. 2013, 13 (6), 2541–2547.
Kats, M. A.; Sharma, D.; Lin, J.; Genevet, P.; Blanchard, R.; Yang, Z.; Qazilbash, M. M.;
Basov, D. N.; Ramanathan, S.; Capasso, F. Ultra-Thin Perfect Absorber Employing a Tunable
Phase Change Material. Applied Physics Letters 2012, 101 (22), 5.
Liu, X. L.; Starr, T.; Starr, A. F.; Padilla, W. J. Infrared Spatial and Frequency Selective
Metamaterial with Near-Unity Absorbance. Phys. Rev. Lett. 2010, 104 (20), 4.
Chou, J. B.; Yeng, Y. X.; Lenert, A.; Rinnerbauer, V.; Celanovic, I.; Soljacic, M.; Wang, E.
N.; Kim, S. G. Design of Wide-Angle Selective Absorbers/Emitters with Dielectric Filled
Metallic Photonic Crystals for Energy Applications. Opt. Express 2014, 22 (1), A144–A154.
Gudiksen, M. S.; Lauhon, L. J.; Wang, J.; Smith, D. C.; Lieber, C. M. Growth of Nanowire
Superlattice Structures for Nanoscale Photonics and Electronics. Nature 2002, 415 (6872),
617–620.

153
[38]
[39]
[40]
[41]
[42]
[43]
[44]
[45]
[46]
[47]
[48]
[49]
[50]
[51]
[52]
[53]
[54]
[55]
[56]

Park, H.; Dan, Y.; Seo, K.; Yu, Y. J.; Duane, P. K.; Wober, M.; Crozier, K. B. Filter-Free
Image Sensor Pixels Comprising Silicon Nanowires with Selective Color Absorption. Nano
Lett. 2014, 14 (4), 1804–1809.
Soci, C.; Zhang, A.; Bao, X. Y.; Kim, H.; Lo, Y.; Wang, D. L. Nanowire Photodetectors. J.
Nanosci. Nanotechnol. 2010, 10 (3), 1430–1449.
Liu, X. L.; Wang, L. P.; Zhang, Z. M. Wideband Tunable Omnidirectional Infrared Absorbers
Based on Doped-Silicon Nanowire Arrays. J. Heat Transf.-Trans. ASME 2013, 135 (6), 8.
Tilke, A. T.; Pescini, L.; Lorenz, H.; Blick, R. H. Fabrication and Transport Characterization
of a Primary Thermometer Formed by Coulomb Islands in a Suspended Silicon Nanowire.
Applied Physics Letters 2003, 82 (21), 3773–3775.
Anttu, N.; Xu, H. Q. Coupling of Light Into Nanowire Arrays and Subsequent Absorption. J.
Nanosci. Nanotechnol. 2010, 10 (11), 7183–7187.
Garnett, E.; Yang, P. D. Light Trapping in Silicon Nanowire Solar Cells. Nano Lett. 2010, 10
(3), 1082–1087.
Cao, L.; Fan, P.; Vasudev, A. P.; White, J. S.; Yu, Z.; Cai, W.; Schuller, J. A.; Fan, S.;
Brongersma, M. L. Semiconductor Nanowire Optical Antenna Solar Absorbers. Nano Lett.
2010, 10 (2), 439–445.
Hu, S.; Chi, C. Y.; Fountaine, K. T.; Yao, M. Q.; Atwater, H. A.; Dapkus, P. D.; Lewis, N. S.;
Zhou, C. W. Optical, Electrical, and Solar Energy-Conversion Properties of Gallium Arsenide
Nanowire-Array Photoanodes. Energy Environ. Sci. 2013, 6 (6), 1879–1890.
Liu, C.; Tang, J. Y.; Chen, H. M.; Liu, B.; Yang, P. D. A Fully Integrated Nanosystem of
Semiconductor Nanowires for Direct Solar Water Splitting. Nano Lett. 2013, 13 (6), 2989–
2992.
Woodhouse, M.; Goodrich, A. Manufacturing Cost Analysis Relevant to Single-and DualJunction Photovoltaic Cells Fabricated with III-vs and III-vs Grown on Czochralski Silicon
(Presentation); National Renewable Energy Laboratory (NREL), Golden, CO., 2014.
Rosenwaks, Y.; Shapira, Y.; Huppert, D. Evidence for Low Intrinsic Surface-Recombination
Velocity on P-Type InP. Phys. Rev. B 1991, 44 (23), 13097–13100.
Grzela, G.; Paniagua-Domínguez, R.; Barten, T.; Fontana, Y.; Sánchez-Gil, J. A.; Gómez
Rivas, J. Nanowire Antenna Emission. Nano Lett. 2012, 12 (11), 5481–5486.
Muskens, O. L.; Rivas, J. G.; Algra, R. E.; Bakkers, E.; Lagendijk, A. Design of Light
Scattering in Nanowire Materials for Photovoltaic Applications. Nano Lett. 2008, 8 (9), 2638–
2642.
Fan, Z. Y.; Kapadia, R.; Leu, P. W.; Zhang, X. B.; Chueh, Y. L.; Takei, K.; Yu, K.; Jamshidi,
A.; Rathore, A. A.; Ruebusch, D. J.; et al. Ordered Arrays of Dual-Diameter Nanopillars for
Maximized Optical Absorption. Nano Lett. 2010, 10 (10), 3823–3827.
Lin, C. X.; Povinelli, M. L. Optical Absorption Enhancement in Silicon Nanowire Arrays with
a Large Lattice Constant for Photovoltaic Applications. Opt. Express 2009, 17 (22), 19371–
19381.
Abujetas, D. R.; Paniagua-Dominguez, R.; Sanchez-Gil, J. A. Unraveling the Janus Role of
Mie Resonances and Leaky/Guided Modes in Semiconductor Nanowire Absorption for
Enhanced Light Harvesting. ACS Photonics 2015, 2 (7), 921–929.
Seo, K.; Wober, M.; Steinvurzel, P.; Schonbrun, E.; Dan, Y. P.; Ellenbogen, T.; Crozier, K. B.
Multicolored Vertical Silicon Nanowires. Nano Lett. 2011, 11 (4), 1851–1856.
Heiss, M.; Russo-Averchi, E.; Dalmau-Mallorqui, A.; Tutuncuoglu, G.; Matteini, F.; Ruffer,
D.; Conesa-Boj, S.; Demichel, O.; Alarcon-Llado, E.; Morral, A. F. I. III-v Nanowire Arrays:
Growth and Light Interaction. Nanotechnology 2014, 25 (1), 9.
Wallentin, J.; Anttu, N.; Asoli, D.; Huffman, M.; Åberg, I.; Magnusson, M. H.; Siefer, G.;
Fuss-Kailuweit, P.; Dimroth, F.; Witzigmann, B.; et al. InP Nanowire Array Solar Cells

154

[57]
[58]
[59]
[60]
[61]
[62]
[63]
[64]
[65]
[66]
[67]

[68]
[69]
[70]
[71]

[72]
[73]
[74]
[75]

Achieving 13.8% Efficiency by Exceeding the Ray Optics Limit. Science 2013, 339
(6123), 1057–1060.
Diedenhofen, S. L.; Janssen, O. T. A.; Grzela, G.; Bakkers, E. P. A. M.; Gómez Rivas, J.
Strong Geometrical Dependence of the Absorption of Light in Arrays of Semiconductor
Nanowires. ACS Nano 2011, 5 (3), 2316–2323.
Deinega, A.; John, S. Solar Power Conversion Efficiency in Modulated Silicon Nanowire
Photonic Crystals. Journal of Applied Physics 2012, 112 (7), 7.
Hong, L.; Rusli; Wang, X. C.; Zheng, H. Y.; Wang, H.; Yu, H. Y. Design Guidelines for
Slanting Silicon Nanowire Arrays for Solar Cell Application. Journal of Applied Physics 2013,
114 (8), 6.
Cansizoglu, H.; Cansizoglu, M. F.; Finckenor, M.; Karabacak, T. Optical Absorption
Properties of Semiconducting Nanostructures with Different Shapes. Advanced Optical
Materials 2013, 1 (2), 158–166.
Lin, C. X.; Huang, N. F.; Povinelli, M. L. Effect of Aperiodicity on the Broadband Reflection
of Silicon Nanorod Structures for Photovoltaics. Opt. Express 2012, 20 (1), A125–A132.
Vynck, K.; Burresi, M.; Riboli, F.; Wiersma, D. S. Photon Management in Two-Dimensional
Disordered Media. Nat Mater 2012, 11 (12), 1017–1022.
Kelzenberg, M. Silicon Microwire Photovoltaics; California Institute of Technology, 2010.
Snyder, A. W.; Love, J. Optical Waveguide Theory; Springer, 1983.
Walter, M. G.; Warren, E. L.; McKone, J. R.; Boettcher, S. W.; Mi, Q.; Santori, E. A.; Lewis,
N. S. Solar Water Splitting Cells. Chem. Rev. 2010, 110 (11), 6446–6473.
Khaselev, O.; Turner, J. A. A Monolithic Photovoltaic-Photoelectrochemical Device for
Hydrogen Production via Water Splitting. Science 1998, 280 (5362), 425–427.
Lichterman, M. F.; Sun, K.; Hu, S.; Zhou, X.; McDowell, M. T.; Shaner, M. R.; Richter, M.
H.; Crumlin, E. J.; Carim, A. I.; Saadi, F. H.; et al. Protection of Inorganic Semiconductors for
Sustained, Efficient Photoelectrochemical Water Oxidation. Catalysis Today 2016, 262, 11–
23.
Bae, D.; Pedersen, T.; Seger, B.; Iandolo, B.; Hansen, O.; Vesborg, P. C. K.; Chorkendorff, I.
Carrier-Selective P- and N-Contacts for Efficient and Stable Photocatalytic Water Reduction.
Catalysis Today 2017, 290, 59–64.
Fountaine, K. T.; Lewerenz, H.-J.; Atwater, H. A. Efficiency Limits for Photoelectrochemical
Water-Splitting. Nature Communications 2016, 7, 13706.
May, M. M.; Lackner, D.; Ohlmann, J.; Dimroth, F.; van de Krol, R.; Hannappel, T.;
Schwarzburg, K. On the Benchmarking of Multi-Junction Photoelectrochemical Fuel
Generating Devices. Sustainable Energy Fuels 2017, 1, 15010.
Sathre, R.; Scown, C. D.; Morrow, W. R.; Stevens, J. C.; Sharp, I. D.; Ager, J. W., III;
Walczak, K.; Houle, F. A.; Greenblatt, J. B. Life-Cycle Net Energy Assessment of Large-Scale
Hydrogen Production via Photoelectrochemical Water Splitting. Energy Environ. Sci. 2014, 7
(10), 3264–3278.
Dimroth, F.; Beckert, R.; Meusel, M.; Schubert, U.; Bett, A. W. Metamorphic
GayIn1−yP/Ga1−xInxAs Tandem Solar Cells for Space and for Terrestrial Concentrator
Applications at C ≫ 1000 Suns. Prog. Photovolt: Res. Appl. 2001, 9 (3), 165–178.
Ohlmann, J.; Sanchez, J. F. M.; Lackner, D.; Förster, P.; Steiner, M.; Fallisch, A.; Dimroth, F.
Recent Development in Direct Generation of Hydrogen Using Multi-Junction Solar Cells;
Author(s), 2016; Vol. 1766, p 080004.
Hu, S.; Shaner, M. R.; Beardslee, J. A.; Lichterman, M. F.; Brunschwig, B. S.; Lewis, N. S.
Amorphous TiO2 Coatings Stabilize Si, GaAs, and GaP Photoanodes for Efficient Water
Oxidation. Science 2014, 344 (6187), 1005–1009.
Porter, J. D.; Heller, A.; Aspnes, D. E. Experiment and Theory of “Transparent” Metal Films.
Nature 1985, 313 (6004), 664–666.

155
[76]

[77]
[78]

[79]
[80]

[81]
[82]

[83]
[84]
[85]
[86]
[87]
[88]
[89]
[90]
[91]

Sanz, J. M.; Ortiz, D.; Alcaraz de la Osa, R.; Saiz, J. M.; González, F.; Brown, A. S.;
Losurdo, M.; Everitt, H. O.; Moreno, F. UV Plasmonic Behavior of Various Metal
Nanoparticles in the Near- and Far-Field Regimes: Geometry and Substrate Effects. J. Phys.
Chem. C 2013, 117 (38), 19606–19615.
Singh, M. R.; Clark, E. L.; Bell, A. T. Effects of Electrolyte, Catalyst, and Membrane
Composition and Operating Conditions on the Performance of Solar-Driven Electrochemical
Reduction of Carbon Dioxide. Phys. Chem. Chem. Phys. 2015, 17 (29), 18924–18936.
Verlage, E.; Hu, S.; Liu, R.; Jones, R. J. R.; Sun, K.; Xiang, C.; Lewis, N. S.; Atwater, H. A. A
Monolithically Integrated, Intrinsically Safe, 10% Efficient, Solar-Driven Water-Splitting
System Based on Active, Stable Earth-Abundant Electrocatalysts in Conjunction with Tandem
III-v Light Absorbers Protected by Amorphous TiO2 Films. Energy Environ. Sci. 2015, 8 (11),
3166–3172.
Seh, Z. W.; Kibsgaard, J.; Dickens, C. F.; Chorkendorff, I.; Nørskov, J. K.; Jaramillo, T. F.
Combining Theory and Experiment in Electrocatalysis: Insights Into Materials Design. Science
2017, 355 (6321), eaad4998.
Schreier, M.; Héroguel, F.; Steier, L.; Ahmad, S.; Luterbacher, J. S.; Mayer, M. T.; Luo, J.;
Grätzel, M. Solar Conversion of CO₂ to CO Using Earth-Abundant
Electrocatalysts Prepared by Atomic Layer Modification of CuO. Nat. Energy 2017, 2,
nenergy201787.
Gurudayal; Beeman, J. W.; Bullock, J.; Wang, H.; Eichhorn, J.; Towle, C.; Javey, A.; Toma, F.
M.; Mathews, N.; Ager, J. W., III. Si Photocathode with Ag-Supported Dendritic Cu Catalyst
for CO2 Reduction. Energy Environ. Sci. 2019, 12 (3), 1068–1077.
Zhou, X.; Liu, R.; Sun, K.; Chen, Y.; Verlage, E.; Francis, S. A.; Lewis, N. S.; Xiang, C.
Solar-Driven Reduction of 1 Atm of CO2 To Formate at 10% Energy-Conversion Efficiency
by Use of a TiO2-Protected III–v Tandem Photoanode in Conjunction with a Bipolar
Membrane and a Pd/C Cathode. ACS Energy Lett. 2016, 1 (4), 764–770.
Zhou, X.; Xiang, C. Comparative Analysis of Solar-to-Fuel Conversion Efficiency: a Direct,
One-Step Electrochemical CO 2Reduction Reactor Versus a Two-Step, Cascade
Electrochemical CO 2Reduction Reactor. ACS Energy Lett. 2018, 1892–1897.
Delacourt, C.; Ridgway, P. L.; Kerr, J. B.; Newman, J. Design of an Electrochemical Cell
Making Syngas (CO+H[Sub 2]) From CO[Sub 2] and H[Sub 2]O Reduction at Room
Temperature. J. Electrochem. Soc. 2008, 155 (1), B42–B49.
Schulz, H. Short History and Present Trends of Fischer–Tropsch Synthesis. Applied Catalysis
A: General 1999, 186 (1-2), 3–12.
Lum, Y.; Ager, J. W., III. Sequential Catalysis Controls Selectivity in Electrochemical CO
2reduction on Cu. Energy Environ. Sci. 2018, 11 (10), 2935–2944.
Welch, A. J.; DuChene, J. S.; Tagliabue, G.; Davoyan, A.; Cheng, W.-H.; Atwater, H. A.;
Atwater, H. A. Nanoporous Gold as a Highly Selective and Active Carbon Dioxide Reduction
Catalyst. ACS Appl. Energy Mater. 2018, 2 (1), 164–170.
Chen, Y.; Li, C. W.; Kanan, M. W. Aqueous CO2 Reduction at Very Low Overpotential on
Oxide-Derived Au Nanoparticles. J. Am. Chem. Soc. 2012, 134 (49), 19969–19972.
Hatsukade, T.; Kuhl, K. P.; Cave, E. R.; Abram, D. N.; Jaramillo, T. F. Insights Into the
Electrocatalytic Reduction of CO2 on Metallic Silver Surfaces. Phys. Chem. Chem. Phys.
2014, 16 (27), 13814–13819.
Asadi, M.; Kim, K.; Liu, C.; Addepalli, A. V.; Abbasi, P.; Yasaei, P.; Phillips, P.; Behranginia,
A.; Cerrato, J. M.; Haasch, R.; et al. Nanostructured Transition Metal Dichalcogenide
Electrocatalysts for CO2 Reduction in Ionic Liquid. Science 2016, 353 (6298), 467–470.
Asadi, M.; Kumar, B.; Behranginia, A.; Rosen, B. A.; Baskin, A.; Repnin, N.; Pisasale, D.;
Phillips, P.; Zhu, W.; Haasch, R.; et al. Robust Carbon Dioxide Reduction on Molybdenum
Disulphide Edges. Nature Communications 2014, 5 (1), 1330.

156
[92]
[93]
[94]
[95]
[96]
[97]
[98]

[99]
[100]
[101]
[102]
[103]
[104]
[105]
[106]
[107]

Cheng, T.; Huang, Y.; Xiao, H.; Goddard, W. A., III. Predicted Structures of the Active
Sites Responsible for the Improved Reduction of Carbon Dioxide by Gold Nanoparticles. J.
Phys. Chem. Lett. 2017, 8 (14), 3317–3320.
Singh, M. R.; Kwon, Y.; Lum, Y.; Ager, J. W., III; Bell, A. T. Hydrolysis of Electrolyte
Cations Enhances the Electrochemical Reduction of CO 2over Ag and Cu. J. Am. Chem. Soc.
2016, 138 (39), 13006–13012.
Liu, M.; Pang, Y.; Zhang, B.; De Luna, P.; Voznyy, O.; Xu, J.; Zheng, X.; Dinh, C. T.; Fan, F.;
Cao, C.; et al. Enhanced Electrocatalytic CO₂ Reduction via Field-Induced
Reagent Concentration. Nature 2016, 537 (7620), 382–386.
Varela, A. S.; Kroschel, M.; Reier, T.; Strasser, P. Controlling the Selectivity of CO2
Electroreduction on Copper: the Effect of the Electrolyte Concentration and the Importance of
the Local pH. Catalysis Today 2016, 260, 8–13.
Song, J. T.; Ryoo, H.; Cho, M.; Kim, J.; Kim, J.-G.; Chung, S.-Y.; Oh, J. Nanoporous Au Thin
Films on Si Photoelectrodes for Selective and Efficient Photoelectrochemical CO2 Reduction.
Advanced Energy Materials 2017, 7 (3), 1601103.
Li, J.; Chen, G.; Zhu, Y.; Liang, Z.; Pei, A.; Wu, C.-L.; Wang, H.; Lee, H. R.; Liu, K.; Chu, S.;
et al. Efficient Electrocatalytic CO 2 Reduction on a Three-Phase Interface. Nature Catalysis
2018 1:8 2018, 1 (8), 592–600.
Verma, S.; Hamasaki, Y.; Kim, C.; Huang, W.; Lu, S.; Jhong, H.-R. M.; Gewirth, A. A.;
Fujigaya, T.; Nakashima, N.; Kenis, P. J. A. Insights Into the Low Overpotential
Electroreduction of CO 2to CO on a Supported Gold Catalyst in an Alkaline Flow
Electrolyzer. ACS Energy Lett. 2017, 3 (1), 193–198.
Clark, E. L.; Resasco, J.; Landers, A.; Lin, J.; Chung, L.-T.; Walton, A.; Hahn, C.; Jaramillo,
T. F.; Bell, A. T. Standards and Protocols for Data Acquisition and Reporting for Studies of
the Electrochemical Reduction of Carbon Dioxide. ACS Catal. 2018, 6560–6570.
Kim, C.; Jeon, H. S.; Eom, T.; Jee, M. S.; Kim, H.; Friend, C. M.; Min, B. K.; Hwang, Y. J.
Achieving Selective and Efficient Electrocatalytic Activity for CO2 Reduction Using
Immobilized Silver Nanoparticles. J. Am. Chem. Soc. 2015, 137 (43), 13844–13850.
Jiang, K.; Kharel, P.; Peng, Y.; Gangishetty, M. K.; Lin, H.-Y. G.; Stavitski, E.; Attenkofer,
K.; Wang, H. Silver Nanoparticles with Surface-Bonded Oxygen for Highly Selective CO2
Reduction. ACS Sustainable Chem. Eng. 2017, 5 (10), 8529–8534.
Wang, Z.; Wu, L.; Sun, K.; Chen, T.; Jiang, Z.; Cheng, T.; William A Goddard, I. Surface
Ligand Promotion of Carbon Dioxide Reduction Through Stabilizing Chemisorbed Reactive
Intermediates. J. Phys. Chem. Lett. 2018, 9 (11), 3057–3061.
Hsieh, Y.-C.; Senanayake, S. D.; Zhang, Y.; Xu, W.; Polyansky, D. E. Effect of Chloride
Anions on the Synthesis and Enhanced Catalytic Activity of Silver Nanocoral Electrodes for
CO 2Electroreduction. ACS Catal. 2015, 5 (9), 5349–5356.
Lu, Q.; Rosen, J.; Zhou, Y.; Hutchings, G. S.; Kimmel, Y. C.; Chen, J. G.; Jiao, F. A Selective
and Efficient Electrocatalyst for Carbon Dioxide Reduction. Nature Communications 2014, 5
(1), 1–6.
Dinh, C. T.; García de Arquer, F. P.; Sinton, D.; Sargent, E. H. High Rate, Selective, and
Stable Electroreduction of CO 2to CO in Basic and Neutral Media. ACS Energy Lett. 2018, 3
(11), 2835–2840.
Tornow, C. E.; Thorson, M. R.; Ma, S.; Gewirth, A. A.; Kenis, P. J. A. Nitrogen-Based
Catalysts for the Electrochemical Reduction of CO2 to CO. J. Am. Chem. Soc. 2012, 134 (48),
19520–19523.
Verma, S.; Lu, X.; Ma, S.; Masel, R. I.; Kenis, P. J. A. The Effect of Electrolyte Composition
on the Electroreduction of CO2 to CO on Ag Based Gas Diffusion Electrodes. Phys. Chem.
Chem. Phys. 2016, 18 (10), 7075–7084.

157
[108]
[109]
[110]

[111]
[112]
[113]
[114]
[115]

[116]
[117]
[118]
[119]

[120]
[121]
[122]
[123]
[124]
[125]

Higgins, D.; Hahn, C.; Xiang, C.; Jaramillo, T. F.; Weber, A. Z. Gas-Diffusion Electrodes
for Carbon Dioxide Reduction: a New Paradigm. ACS Energy Lett. 2018, 4 (1), 317–324.
Singh, M. R.; Papadantonakis, K.; Xiang, C.; Lewis, N. S. An Electrochemical Engineering
Assessment of the Operational Conditions and Constraints for Solar-Driven Water-Splitting
Systems at Near-Neutral pH. Energy Environ. Sci. 2015, 8 (9), 2760–2767.
Lobaccaro, P.; Singh, M. R.; Clark, E. L.; Kwon, Y.; Bell, A. T.; Ager, J. W., III. Effects of
Temperature and Gas–Liquid Mass Transfer on the Operation of Small Electrochemical Cells
for the Quantitative Evaluation of CO2 Reduction Electrocatalysts. Phys. Chem. Chem. Phys.
2016, 18 (38), 26777–26785.
Song, J. T.; Song, H.; Kim, B.; Oh, J. Towards Higher Rate Electrochemical CO2 Conversion:
From Liquid-Phase to Gas-Phase Systems. Catalysts 2019, Vol. 9, Page 224 2019, 9 (3), 224.
Weng, L.-C.; Bell, A. T.; Weber, A. Z. Modeling Gas-Diffusion Electrodes for CO 2
Reduction. Phys. Chem. Chem. Phys. 2018, 20 (25), 16973–16984.
Cook, R. L.; MacDuff, R. C.; Sammells, A. F. High Rate Gas Phase CO[Sub 2] Reduction to
Ethylene and Methane Using Gas Diffusion Electrodes. J. Electrochem. Soc. 1990, 137 (2),
607–608.
Hori, Y.; Ito, H.; Okano, K.; Nagasu, K.; Sato, S. Silver-Coated Ion Exchange Membrane
Electrode Applied to Electrochemical Reduction of Carbon Dioxide. Electrochimica Acta
2003, 48 (18), 2651–2657.
Dinh, C. T.; Burdyny, T.; Kibria, M. G.; Seifitokaldani, A.; Gabardo, C. M.; de Arquer, F. P.
G.; Kiani, A.; Edwards, J. P.; De Luna, P.; Bushuyev, O. S.; et al. CO2electroreduction to
Ethylene via Hydroxide-Mediated Copper Catalysis at an Abrupt Interface. Science 2018, 360
(6390), 783–787.
Li, Y. C.; Lee, G.; Yuan, T.; Wang, Y.; Nam, D.-H.; Wang, Z.; García de Arquer, F. P.; Lum,
Y.; Dinh, C. T.; Voznyy, O.; et al. CO 2Electroreduction From Carbonate Electrolyte. ACS
Energy Lett. 2019, 4 (6), 1427–1431.
Hossain, M. Z.; Rahim, N. A.; Selvaraj, J. A. L. Recent Progress and Development on Power
DC-DC Converter Topology, Control, Design and Applications: a Review. Renewable and
Sustainable Energy Reviews 2018, 81, 205–230.
Holman, Z.; Boccard, M. Light Management in Silicon Solar Cells. In Photovoltaic Solar
Energy: From Fundamentals to Applications; John Wiley & Sons, Ltd.: Hoboken, New Jersey,
2016; pp 136–149.
Manzoor, S.; Yu, Z. J.; Ali, A.; Ali, W.; Bush, K. A.; Palmstrom, A. F.; Bent, S. F.; McGehee,
M. D.; Holman, Z. C. Improved Light Management in Planar Silicon and Perovskite Solar
Cells Using PDMS Scattering Layer. Solar Energy Materials and Solar Cells 2017, 173, 59–
65.
Haug, F. J.; Ballif, C. Light Management in Thin Film Silicon Solar Cells. Energy Environ.
Sci. 2015, 8 (3), 824–837.
Schmid, M. Review on Light Management by Nanostructures in Chalcopyrite Solar Cells.
Semicond. Sci. Technol. 2017, 32 (4), 043003.
Peer, A.; Biswas, R.; Park, J.-M.; Shinar, R.; Shinar, J. Light Management in Perovskite Solar
Cells and Organic LEDs with Microlens Arrays. Opt. Express 2017, 25 (9), 10704–10709.
Hu, S.; Xiang, C.; Haussener, S.; Berger, A. D.; Lewis, N. S. An Analysis of the Optimal Band
Gaps of Light Absorbers in Integrated Tandem Photoelectrochemical Water-Splitting Systems.
Energy Environ. Sci. 2013, 6 (10), 2984–2993.
Raut, H. K.; Ganesh, V. A.; Nair, A. S.; Ramakrishna, S. Anti-Reflective Coatings: a Critical,
in-Depth Review. Energy Environ. Sci. 2011, 4 (10), 3779–3804.
Krč, J.; Smole, F.; Topič, M. Potential of Light Trapping in Microcrystalline Silicon Solar
Cells with Textured Substrates. Prog. Photovoltaics: Res. Appl. 2003, 11 (7), 429–436.

158
[126]
[127]
[128]
[129]
[130]
[131]
[132]

[133]
[134]
[135]
[136]
[137]
[138]
[139]
[140]
[141]
[142]
[143]
[144]
[145]

Campbell, P.; Green, M. A. Light Trapping Properties of Pyramidally Textured Surfaces.
J. Appl. Phys. 1987, 62 (1), 243–249.
Yablonovitch, E.; Cody, G. D. Intensity Enhancement in Textured Optical Sheets for Solar
Cells. IEEE Trans. Electron Devices 1982, 29 (2), 300–305.
Cho, S. J.; An, T.; Lim, G. Three-Dimensionally Designed Anti-Reflective Silicon Surfaces for
Perfect Absorption of Light. Chem. Commun. 2014, 50 (99), 15710–15713.
Brongersma, M. L.; Cui, Y.; Fan, S. Light Management for Photovoltaics Using High-Index
Nanostructures. Nat Mater 2014, 13, 451.
Mokkapati, S.; Catchpole, K. R. Nanophotonic Light Trapping in Solar Cells. J. Appl. Phys.
2012, 112 (10), 101101.
Wilson, G. The Multi-TW Scale Future for Photovoltaics. AIP Conf. Proc. 2018, 1924 (1),
020003.
Sun, K.; Moreno-Hernandez, I. A.; Schmidt, W. C.; Zhou, X.; Crompton, J. C.; Liu, R.; Saadi,
F. H.; Chen, Y.; Papadantonakis, K. M.; Lewis, N. S. A Comparison of the Chemical, Optical
and Electrocatalytic Properties of Water-Oxidation Catalysts for Use in Integrated Solar-Fuel
Generators. Energy Environ. Sci. 2017, 10 (4), 987–1002.
Yoshikawa, K.; Kawasaki, H.; Yoshida, W.; Irie, T.; Konishi, K.; Nakano, K.; Uto, T.; Adachi,
D.; Kanematsu, M.; Uzu, H.; et al. Silicon Heterojunction Solar Cell with Interdigitated Back
Contacts for a Photoconversion Efficiency Over 26%. Nat. Energy 2017, 2, 17032.
Lammert, M. D.; Schwartz, R. J. The Interdigitated Back Contact Solar Cell: a Silicon Solar
Cell for Use in Concentrated Sunlight. IEEE Trans. Electron Devices 1977, 24 (4), 337–342.
Spitzer, M. B.; Fan, J. C. C. Multijunction Cells for Space Applications. Sol. Cells 1990, 29
(2), 183–203.
Tian, B. Z.; Zheng, X. L.; Kempa, T. J.; Fang, Y.; Yu, N. F.; Yu, G. H.; Huang, J. L.; Lieber,
C. M. Coaxial Silicon Nanowires as Solar Cells and Nanoelectronic Power Sources. Nature
2007, 449 (7164), 885–U888.
Mavrokefalos, A.; Han, S. E.; Yerci, S.; Branham, M. S.; Chen, G. Efficient Light Trapping in
Inverted Nanopyramid Thin Crystalline Silicon Membranes for Solar Cell Applications. Nano
Lett. 2012, 12 (6), 2792–2796.
Zhu, J.; Hsu, C. M.; Yu, Z. F.; Fan, S. H.; Cui, Y. Nanodome Solar Cells with Efficient Light
Management and Self-Cleaning. Nano Lett. 2010, 10 (6), 1979–1984.
Yao, Y.; Yao, J.; Narasimhan, V. K.; Ruan, Z. C.; Xie, C.; Fan, S. H.; Cui, Y. Broadband Light
Management Using Low-Q Whispering Gallery Modes in Spherical Nanoshells. Nature
Communications 2012, 3.
Kapadia, R.; Fan, Z. Y.; Takei, K.; Javey, A. Nanopillar Photovoltaics: Materials, Processes,
and Devices. Nano Energy 2012, 1 (1), 132–144.
Narasimhan, V. K.; Hymel, T. M.; Lai, R. A.; Cui, Y. Hybrid Metal–Semiconductor
Nanostructure for Ultrahigh Optical Absorption and Low Electrical Resistance at
Optoelectronic Interfaces. ACS Nano 2015, 9 (11), 10590–10597.
Jeong, S.; Garnett, E. C.; Wang, S.; Yu, Z. G.; Fan, S. H.; Brongersma, M. L.; McGehee, M.
D.; Cui, Y. Hybrid Silicon Nanocone-Polymer Solar Cells. Nano Lett. 2012, 12 (6), 2971–
2976.
Yalamanchili, S.; Emmer, H. S.; Fountaine, K. T.; Chen, C. T.; Lewis, N. S.; Atwater, H. A.
Enhanced Absorption and <1% Spectrum-and-Angle-Averaged Reflection in Tapered
Microwire Arrays. ACS Photonics 2016, 3 (10), 1854–1861.
Fountaine, K. T.; Cheng, W.-H.; Bukowsky, C. R.; Atwater, H. A. Near-Unity Unselective
Absorption in Sparse InP Nanowire Arrays. ACS Photonics 2016, 3 (10), 1826–1832.
Grandidier, J.; Callahan, D. M.; Munday, J. N.; Atwater, H. A. Light Absorption Enhancement
in Thin-Film Solar Cells Using Whispering Gallery Modes in Dielectric Nanospheres. Adv
Mater 2011, 23 (10), 1272.

159
[146]
[147]
[148]
[149]
[150]
[151]
[152]
[153]

[154]

[155]
[156]
[157]
[158]
[159]

[160]

[161]

Choy, J. T.; Bradley, J. D. B.; Deotare, P. B.; Burgess, I. B.; Evans, C. C.; Mazur, E.;
Lončar, M. Integrated TiO2resonators for Visible Photonics. Opt. Lett. 2012, 37 (4), 539–541.
Evans, C. C.; Liu, C.; Suntivich, J. Low-Loss Titanium Dioxide Waveguides and Resonators
Using a Dielectric Lift-Off Fabrication Process. Opt. Express 2015, 23 (9), 11160–11169.
Evans, C. C.; Shtyrkova, K.; Reshef, O.; Moebius, M.; Bradley, J. D. B.; Griesse-Nascimento,
S.; Ippen, E.; Mazur, E. Multimode Phase-Matched Third-Harmonic Generation in SubMicrometer-Wide Anatase TiO2waveguides. Opt. Express 2015, 23 (6), 7832–7841.
Evans, C. C.; Shtyrkova, K.; Bradley, J. D. B.; Reshef, O.; Ippen, E.; Mazur, E. Spectral
Broadening in Anatase Titanium Dioxide Waveguides at Telecommunication and Near-Visible
Wavelengths. Opt. Express 2013, 21 (15), 18582–18591.
Guan, X.; Hu, H.; Oxenløwe, L. K.; Frandsen, L. H. Compact Titanium Dioxide Waveguides
with High Nonlinearity at Telecommunication Wavelengths. Opt. Express 2018, 26 (2), 1055–
1063.
Bradley, J. D. B.; Evans, C. C.; Choy, J. T.; Reshef, O.; Deotare, P. B.; Parsy, F.; Phillips, K.
C.; Lončar, M.; Mazur, E. Submicrometer-Wide Amorphous and Polycrystalline Anatase
TiO2waveguides for Microphotonic Devices. Opt. Express 2012, 20 (21), 23821–23831.
Sun, K.; Kuang, Y.; Verlage, E.; Brunschwig, B. S.; Tu, C. W.; Lewis, N. S. Sputtered NiOX
Films for Stabilization of P+N‐InP Photoanodes for Solar‐Driven Water Oxidation. Advanced
Energy Materials 2015, 5 (11), 1402276.
Lichterman, M. F.; Carim, A. I.; McDowell, M. T.; Hu, S.; Gray, H. B.; Brunschwig, B. S.;
Lewis, N. S. Stabilization of N-Cadmium Telluride Photoanodes for Water Oxidation to O2(G)
in Aqueous Alkaline Electrolytes Using Amorphous TiO2films Formed by Atomic-Layer
Deposition. Energy Environ. Sci. 2014, 7 (10), 3334–3337.
McDowell, M. T.; Lichterman, M. F.; Carim, A. I.; Liu, R.; Hu, S.; Brunschwig, B. S.; Lewis,
N. S. The Influence of Structure and Processing on the Behavior of TiO2 Protective Layers for
Stabilization of N-Si/TiO2/Ni Photoanodes for Water Oxidation. ACS Appl. Mater. Interfaces
2015, 7 (28), 15189–15199.
McDowell, M. T.; Lichterman, M. F.; Spurgeon, J. M.; Hu, S.; Sharp, I. D.; Brunschwig, B. S.;
Lewis, N. S. Improved Stability of Polycrystalline Bismuth Vanadate Photoanodes by Use of
Dual-Layer Thin TiO2/Ni Coatings. J. Phys. Chem. C 2014, 118 (34), 19618–19624.
Shaner, M. R.; Hu, S.; Sun, K.; Lewis, N. S. Stabilization of Si Microwire Arrays for SolarDriven H2O Oxidation to O2(G) in 1.0 M KOH(Aq) Using Conformal Coatings of Amorphous
TiO2. Energy Environ. Sci. 2015, 8 (1), 203–207.
Siklitsky, V. New Semiconductor Materials. Characteristics and Properties. Available:
Kenney, M. J.; Gong, M.; Li, Y.; Wu, J. Z.; Feng, J.; Lanza, M.; Dai, H. High-Performance
Silicon Photoanodes Passivated with Ultrathin Nickel Films for Water Oxidation. Science
2013, 342 (6160), 836–840.
Sun, K.; Ritzert, N. L.; John, J.; Tan, H.; Hale, W. G.; Jiang, J.; Moreno-Hernandez, I.;
Papadantonakis, K. M.; Moffat, T. P.; Brunschwig, B. S.; et al. Performance and Failure
Modes of Si Anodes Patterned with Thin-Film Ni Catalyst Islands for Water Oxidation.
Sustainable Energy Fuels 2018, 2 (5), 983–998.
Chen, Y.; Sun, K.; Audesirk, H.; Xiang, C.; Lewis, N. S. A Quantitative Analysis of the
Efficiency of Solar-Driven Water-Splitting Device Designs Based on Tandem Photoabsorbers
Patterned with Islands of Metallic Electrocatalysts. Energy Environ. Sci. 2015, 8 (6), 1736–
1747.
Choi, S. K.; Kang, U.; Lee, S.; Ham, D. J.; Ji, S. M.; Park, H. Sn‐Coupled P‐Si Nanowire
Arrays for Solar Formate Production From CO2. Advanced Energy Materials 2014, 4 (11),
1301614.

160
[162]
[163]
[164]

[165]
[166]

[167]

[168]

[169]
[170]
[171]

[172]
[173]

[174]

[175]
[176]
[177]

Kong, Q.; Kim, D.; Liu, C.; Yu, Y.; Su, Y.; Li, Y.; Yang, P. Directed Assembly of
Nanoparticle Catalysts on Nanowire Photoelectrodes for Photoelectrochemical CO
2Reduction. Nano Lett. 2016, 16 (9), 5675–5680.
Fountaine, K. T.; Lewerenz, H.-J. Efficiency Limits for Hydrogen and Formate Production via
Fully-Integrated Photoelectrochemical Devices. ECS Trans. 2017, 77 (4), 15–23.
Sun, K.; Saadi, F. H.; Lichterman, M. F.; Hale, W. G.; Wang, H.-P.; Zhou, X.; Plymale, N. T.;
Omelchenko, S. T.; He, J.-H.; Papadantonakis, K. M.; et al. Stable Solar-Driven Oxidation of
Water by Semiconducting Photoanodes Protected by Transparent Catalytic Nickel Oxide
Films. PNAS 2015, 112 (12), 3612–3617.
Bae, D.; Seger, B.; Vesborg, P. C. K.; Hansen, O.; Chorkendorff, I. Strategies for Stable Water
Splitting via Protected Photoelectrodes. Chem. Soc. Rev. 2017, 46 (7), 1933–1954.
Hu, S.; Richter, M. H.; Lichterman, M. F.; Beardslee, J.; Mayer, T.; Brunschwig, B. S.; Lewis,
N. S. Electrical, Photoelectrochemical, and Photoelectron Spectroscopic Investigation of the
Interfacial Transport and Energetics of Amorphous TiO2/Si Heterojunctions. J. Phys. Chem. C
2016, 120 (6), 3117–3129.
Nunez, P.; Richter, M. H.; Piercy, B. D.; Roske, C. W.; Cabán-Acevedo, M.; Losego, M. D.;
Konezny, S. J.; Fermin, D. J.; Hu, S.; Brunschwig, B. S.; et al. Characterization of Electronic
Transport Through Amorphous TiO2 Produced by Atomic Layer Deposition. J. Phys. Chem. C
2019, 123 (33), 20116–20129.
Lichterman, M. F.; Richter, M. H.; Hu, S.; Crumlin, E. J.; Axnanda, S.; Favaro, M.; Drisdell,
W.; Hussain, Z.; Brunschwig, B. S.; Lewis, N. S.; et al. An Electrochemical,
Microtopographical and Ambient Pressure X-Ray Photoelectron Spectroscopic Investigation of
Si/TiO2/Ni/Electrolyte Interfaces. J. Electrochem. Soc. 2016, 163 (2), H139–H146.
Morales, F.; de Groot, F. M. F.; Glatzel, P.; Kleimenov, E.; Bluhm, H.; Hävecker, M.; KnopGericke, A.; Weckhuysen, B. M. In Situ X-Ray Absorption of Co/Mn/TiO 2Catalysts for
Fischer−Tropsch Synthesis. J. Phys. Chem. B 2004, 108 (41), 16201–16207.
Diebold, U. The Surface Science of Titanium Dioxide. Surface Science Reports 2003, 48 (5-8),
53–229.
Moser, S.; Fatale, S.; Krüger, P.; Berger, H.; Bugnon, P.; Magrez, A.; Niwa, H.; Miyawaki, J.;
Harada, Y.; Grioni, M. Electron-Phonon Coupling in the Bulk of Anatase
Measured by Resonant Inelastic X-Ray
Spectroscopy. Phys. Rev. Lett. 2015, 115 (9), 096404.
Glatzel, P.; Sikora, M.; Fernández-García, M. Resonant X-Ray Spectroscopy to Study K
Absorption Pre-Edges in 3d Transition Metal Compounds. Eur. Phys. J. Spec. Top. 2009, 169
(1), 207–214.
Lichterman, M. F.; Richter, M. H.; Hu, S.; Crumlin, E. J.; Axnanda, S.; Favaro, M.; Drisdell,
W.; Hussain, Z.; Mayer, T.; Brunschwig, B.; et al. Investigation of the Si/TiO2/Electrolyte
Interface Using Operando Tender X-Ray Photoelectron Spectroscopy. ECS Trans. 2015, 66
(6), 97–103.
Richter, M. H.; Lichterman, M. F.; Hu, S.; Crumlin, E. J.; Mayer, T.; Axnanda, S.; Favaro, M.;
Drisdell, W.; Hussain, Z.; Brunschwig, B.; et al. Measurement of the Energy-Band Relations
of Stabilized Si Photoanodes Using Operando Ambient Pressure X-Ray Photoelectron
Spectroscopy. ECS Trans. 2015, 66 (6), 105–113.
Schmeißer, D.; Hoffmann, P.; Beuckert, G. Electronic Properties of the Interface Formed by
Pr2O3 Growth. In Materials for Information Technology; Springer: London, 2005; pp 449–460.
Stöhr, J. NEXAFS Spectroscopy, 2nd ed.; Springer: Heidelberg, 2003; Vol. 25.
Qiao, R.; Li, Q.; Zhuo, Z.; Sallis, S.; Fuchs, O.; Blum, M.; Weinhardt, L.; Heske, C.; Pepper,
J.; Jones, M.; et al. High-Efficiency in Situ Resonant Inelastic X-Ray Scattering (iRIXS)

161

[178]
[179]
[180]
[181]
[182]
[183]
[184]
[185]
[186]
[187]
[188]

Endstation at the Advanced Light Source. Review of Scientific Instruments 2017, 88 (3),
033106.
McCrory, C. C. L.; Jung, S.; Ferrer, I. M.; Chatman, S. M.; Peters, J. C.; Jaramillo, T. F.
Benchmarking Hydrogen Evolving Reaction and Oxygen Evolving Reaction Electrocatalysts
for Solar Water Splitting Devices. J. Am. Chem. Soc. 2015, 137 (13), 4347–4357.
Michaelson, H. B. The Work Function of the Elements and Its Periodicity. Journal of Applied
Physics 2008, 48 (11), 4729–4733.
Tanuma, S.; Powell, C. J.; Penn, D. R. Calculations of Electron Inelastic Mean Free Paths. v.
Data for 14 Organic Compounds Over the 50–2000 eV Range. Surface and Interface Analysis
1994, 21 (3), 165–176.
Cho, D.-Y.; Song, S. J.; Kim, U. K.; Kim, K. M.; Lee, H.-K.; Hwang, C. S. Spectroscopic
Investigation of the Hole States in Ni-Deficient NiO Films. J. Mater. Chem. C 2013, 1 (28),
4334–4338.
Mossanek, R. J. O.; Domínguez-Cañizares, G.; Gutiérrez, A.; Abbate, M.; Díaz-Fernández, D.;
Soriano, L. Effects of Ni Vacancies and Crystallite Size on the O 1s and Ni 2p X-Ray
Absorption Spectra of Nanocrystalline NiO. J. Phys.: Condens. Matter 2013, 25 (49), 495506.
Greiner, M. T.; Helander, M. G.; Wang, Z.-B.; Tang, W.-M.; Lu, Z.-H. Effects of Processing
Conditions on the Work Function and Energy-Level Alignment of NiO Thin Films. J. Phys.
Chem. C 2010, 114 (46), 19777–19781.
Mankel, E.; Hein, C.; Kühn, M.; Mayer, T. Electric Potential Distributions in Space Charge
Regions of Molecular Organic Adsorbates Using a Simplified Distributed States Model.
physica status solidi (a) 2014, 211 (9), 2040–2048.
Horn, K. Semiconductor Interface Studies Using Core and Valence Level Photoemission.
Appl. Phys. A 1990, 51 (4), 289–304.
Franciosi, A.; Van de Walle, C. G. Heterojunction Band Offset Engineering. Surface Science
Reports 1996, 25 (1-4), 1–140.
Rossi, R. C.; Lewis, N. S. Investigation of the Size-Scaling Behavior of Spatially Nonuniform
Barrier Height Contacts to Semiconductor Surfaces Using Ordered Nanometer-Scale Nickel
Arrays on Silicon Electrodes. J. Phys. Chem. B 2001, 105 (49), 12303–12318.
Lichterman, M. F.; Hu, S.; Richter, M. H.; Crumlin, E. J.; Axnanda, S.; Favaro, M.; Drisdell,
W.; Hussain, Z.; Mayer, T.; Brunschwig, B. S.; et al. Direct Observation of the Energetics at a
Semiconductor/Liquid Junction by Operando X-Ray Photoelectron Spectroscopy. Energy
Environ. Sci. 2015, 8 (8), 2409–2416.