Jiang Q 2020 PHD Thesis

Imperial College London
Royal School of Mines

Department of Materials
Plasmonic Enhanced Photoelectrochemical

Water Splitting by Hematite-Based
Nanocomposites
Qianfan Jiang
Submitted in partial fulfilment of the requirements

for the degree of Doctor of Philosophy at
Imperial College London, September 2020
Declaration of Originality
I hereby declare that this thesis and the work reported herein was composed by and originated
entirely from me. Information derived from the published and unpublished work of others has
been acknowledged in the text and references are given in the list of sources.
Qianfan Jiang
Declaration of Copyright
The copyright of this thesis rests with the author. Unless otherwise indicated, its contents
are licensed under a Creative Commons Attribution-Non Commercial 4.0 International Licence
(CC BY-NC). Under this licence, you may copy and redistribute the material in any medium
or format. You may also create and distribute modified versions of the work. This is on
the condition that: you credit the author and do not use it, or any derivative works, for a
commercial purpose. When reusing or sharing this work, ensure you make the licence terms
clear to others by naming the licence and linking to the licence text. Where a work has been
adapted, you should indicate that the work has been changed and describe those changes.
Please seek permission from the copyright holder for uses of this work that are not included in
this licence or permitted under UK Copyright Law.
This work is dedicated to my family.
Abstract
As a promising solar energy harvesting technology, artificial solar water splitting enables di-
rect solar-to-hydrogen conversion and thus exhibits huge potential in the field of alternative
energy source. Hematite is one of the most common photoactive semiconductors and can be
utilized as photoanodes in photoelectrochemical water splitting systems. Despite being non-
toxic, cost-efficient and earth-abundant, pristine hematite has poor light harvesting efficiency.
Various strategies to improve the efficiency are essential for hematite photoanodes. Plasmonic
enhancement that employs plasmonic nanostructures for enhanced electromagnetic field is one
of the promising approaches to boost the solar water splitting performance of hematite. This
PhD project focused on the implementation of plasmonic enhancement in hematite photoan-
odes, together with other optimization strategies such as nanostructuring, surface modification
and upconversion. Modelling work was carried out as a theoretical framework to guide plas-
monic nanostructures’ design. In particular, FDTD modelling for plasmonic nanodisk arrays,
nanotriangle arrays and nanohole arrays were carried out to identify their plasmonic resonance
wavelength as well as the spatial distribution of the enhanced field. Plasmonic nanohole arrays
were chosen as the best candidates to be utilized in solar water splitting systems due to the sig-
nificantly enhanced field originated from surface plasmon polaritons. Various novel plasmonic
nanohole arrays including dual-layer nanohole arrays and non-noble, aluminum-based nanohole
arrays are fabricated and applied to hematite photoanodes. Improved photocurrent density can
be observed due to plasmonic enhancement. Furthermore, plasmonic enhanced triplet-triplet
annihilation upconversion nanoparticles that can convert low-energy photons to high-energy
photons were utilized to overcome the limitation of hematite’s bandgap. The concepts of plas-
monic enhanced upconversion and upconversion induced solar water splitting were proven by
observing below-bandgap enhancement in incident photon-to-current efficiency measurements.
This work not only demonstrated the feasibility of implementing plasmonic enhancement in
hematite photoanodes but also paved the way towards employing plasmonic nanostructures in
other solar water splitting devices.
Acknowledgements
I would like to express my sincere gratitude to:
• My supervisor, Dr. Fang Xie, for her support during my PhD research. Her continu-
ous guidance helped me throughout my study at Imperial College. Her creativity and
commitment to science is always giving me inspirations.
• My second supervisor, Prof. Jason Riley for his constructive and expert advice. He has
always provided me with the precise details on fundamental science.
• My colleagues: Chengyu Ji, Xiangyu Xie, Weixin Song, Lukas Malms, Heng Qin, Daniel
Darvill, Jawad ZAR, Ioannis G. Theodorou, Sarah Fothergill, Caoimhe Joyce, Jiamin Xu,
and Sixing Tang. They were always willing to help me and sharing their fantastic ideas
throughout my PhD project. I wouldn’t have had the great days in lab 1.17 without
them. It is my honor to have worked with them.
• My wife: Jingyi Chen, for her accompany. I would have never made this far without her
love and support.
• My parents, for their unconditional support and understanding. Without them, my life
would be totally different.
Finally, special thanks to Fang Xie’s Plasmonic Group at Imperial College London.
Contents
Abstract 3
Acknowledgements 5
List of Abbreviations 10
List of Tables 12
List of Figures 13
1 Introduction 23
2 Literature Review 27
2.1 Plasmonics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.1.1 Principles of Plasmonics . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.1.2 Plasmonic Metals and Nanostructures . . . . . . . . . . . . . . . . . . . . 30
2.1.3 FDTD Modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.2 Solar Water Splitting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.2.1 Photoelectrochemical Water Splitting . . . . . . . . . . . . . . . . . . . . 34
2.2.2 Hematite Based Nanocomposites as Photoanodes . . . . . . . . . . . . . 36
6
CONTENTS 7
2.2.3 Optimization Strategies of Hematite Photoanodes . . . . . . . . . . . . . 38
2.2.4 Plasmonic Enhanced Solar Water Splitting . . . . . . . . . . . . . . . . . 40
2.3 Upconversion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
2.3.1 Upconversion and TTA-based Upconversion . . . . . . . . . . . . . . . . 44
2.3.2 Upconversion Induced Water Splitting . . . . . . . . . . . . . . . . . . . 45
2.3.3 Plasmonic Enhanced Upconversion . . . . . . . . . . . . . . . . . . . . . 46
3 Instrumentation and Characterisation Methods 49
3.1 Scanning Electron Microscopy (SEM) . . . . . . . . . . . . . . . . . . . . . . . . 50
3.2 Transmission Electron Microscopy (TEM) . . . . . . . . . . . . . . . . . . . . . 50
3.3 Energy-dispersive X-Ray Analysis (EDX) . . . . . . . . . . . . . . . . . . . . . . 51
3.4 X-Ray Powder Diffraction (XRD) . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.5 X-Ray Photoelectron Spectroscopy (XPS) . . . . . . . . . . . . . . . . . . . . . 52
3.6 Photoluminescence (PL) Spectrometer . . . . . . . . . . . . . . . . . . . . . . . 53
3.7 Ultraviolet-visible-near-infrared (UV-Vis-NIR) Spectrometer . . . . . . . . . . . 54
3.8 Photoelectrochemical (PEC) Photocurrent Measurements . . . . . . . . . . . . . 55
3.9 Incident Photon-to-Current Efficiency (IPCE) Measurements . . . . . . . . . . . 56
4 FDTD Modelling of Plasmonic Nanoarrays 59
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
4.3 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4.3.1 Nanodisk Arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4.3.2 Nanotriangle Arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

CONTENTS 8
4.3.3 Nanohole Arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
4.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
5 Dual-layer Plasmonic Nanohole Array Enhanced BiVO4-Hematite Photoan-

odes 81
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
5.2 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
5.3.1 Morphology and elemental analysis . . . . . . . . . . . . . . . . . . . . . 85
5.3.2 Optical properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
5.3.3 Water splitting performance . . . . . . . . . . . . . . . . . . . . . . . . . 93
5.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
6 Aluminum Plasmonic Nanohole Array Enhanced Hematite Nanotube Pho-

toanodes 99
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
6.2 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
6.3.1 Characterization of hematite nanotubes . . . . . . . . . . . . . . . . . . . 103
6.3.2 Characterization of the fabricated photoanodes . . . . . . . . . . . . . . 105
6.3.3 Characterization of water splitting performance . . . . . . . . . . . . . . 109
6.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
7 TTA Upconversion Induced Solar Water Splitting by Hematite Photoanodes

with Plasmonic Enhancement 115
7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116

CONTENTS 9
7.2 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
7.3.1 Characterization of TTA@SiO2 upconversion nanoparticles. . . . . . . . . 119
7.3.2 Characterization of plasmonic enhanced TTA upconversion. . . . . . . . 122
7.3.3 Inverse-opal hematite photoanodes with TTA upconversion materials . . 125
7.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
8 Conclusion 130
8.1 Summary of Thesis Achievements . . . . . . . . . . . . . . . . . . . . . . . . . . 131
8.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
A Source Code of the FDTD Models 135
A.1 Nanohole Array and Nanotriangle Array . . . . . . . . . . . . . . . . . . . . . . 136
A.2 Nanohole Array with Oxidation Layer . . . . . . . . . . . . . . . . . . . . . . . . 140
A.3 Nanodisk Array . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
A.4 Nanostars and Nanospheres . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
B Supporting Information 149
Bibliography 151
List of Abbreviations
BEM Boundary Element Method.
DDA Discrete Dipole Approximation.
EDX Energy dispersive X-Ray Analysis.
EELS Electron Energy Loss Spectroscopy.
EIS Electrochemical Impedance Spectroscopy.
EOT Extraordinary Transmission.
FDTD Finite-difference Time-domain Method.
FEM Finite Element Method.
FTO Fluorine-doped tin oxide.
HEI Hot-electron Injection.
HER Hydrogen Evolution Reaction.
IPCE Incident Photon-to-Current Efficiency.
ISC Intersystem Crossing.
LSPR Localized Surface Plasmon Resonance.
NDA Nanodisk Arrays.
NHA Nanohole Arrays.
10
List of Abbreviations 11
NTA Nanotriangle Arrays.
OER Oxygen Evolution Reaction.
PC Photocatalytic.
PEC Photoelectrochemical.
PIRET Plasmon-induced Resonance Energy Transfer.
PL Photoluminescence.
RHE Reversible Hydrogen Electrode.
SEM Scanning Electron Microscopy.
SPP Surface Plasmon Polariton.
TEM Transmission Electron Microscopy.
TTA Triplet-triplet Annihilation.
TTET Triplet-triplet Energy Transfer.
UV-Vis-NIR Ultraviolet-Visible-Near-Infrared.
XPS X-Ray Photoelectron Spectroscopy.
XRD X-Ray Powder Diffraction.

List of Tables
2.1 Selected literature reports of metal-based plasmonic nanostructures and appli-

cations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.2 Selected literature reports of hematite photoanodes with different optimization

methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
2.3 Selected literature reports of plasmonic enhanced PEC water splitting . . . . . . 43
2.4 Selected literature reports of plasmonic enhanced upconversion systems . . . . . 48
4.1 Adopted values for the Drude-Lorentz models used in FDTD modelling[1] . . . . 63
7.1 IPCE enhancement factors of hematite photoanodes with TTA upconversion

materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
B.1 Permissions Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
12
List of Figures
2.1 (a) Schematic illustration of localized surface plasmon resonance. The electron
cloud oscillates when the incident field couples to the LSPR mode. (b) Schematic
illustration of surface plasmon polaritons. The propagating oscillation is at the
metal-dielectric interface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.2 Map of common plasmonic materials. Materials with high interband losses are
represented by elliptical bubbles and those with low interband losses are repre-
sented by spherical bubbles. The other two axes are carrier mobility and carrier
concentration. From [2]. Reprinted with permission from AAAS. . . . . . . . . 30
2.3 Schematic illustration of Yee cell used in FDTD models. The voxel of H com-
ponents (green) has a shift of half unit length from the voxel of E components
(blue). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.4 Schematic illustration of (a) the process of PC water splitting: the photogen-
erated electron-hole pairs move to the surface active site to drive the HER and
the OER[3], (b) the process of photoanode induced PEC water splitting: the
photogenerated holes move to the surface of photoanode to drive the OER while
the photogenerated electrons move to the cathode through external circuit to
drive the HER[4]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.5 (a) Absorption spectra of Fe2O3 and TiO2 along with the solar spectrum. Hematite
covers a larger spectral area in the solar spectrum due to its low band gap. (b)
Reduction potential of the HER and oxidation potential of the OER, as well as
energy bands of common semiconductors. From [5]. Reprinted with permission
from Wiley. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
13
LIST OF FIGURES 14
2.6 Schematic illustration of three mechanisms of plasmonic enhanced solar water

splitting. (i) Light scattering effect. The incident light is scattered to the semi-
conductor by the plasmonic nanoparticles. (ii) Hot-electron injection. The plas-
monic nanoparticles inject hot electrons directly into the conduction band of the
semiconductor when a direct contact exists. (iii) Plasmon-induced resonance en-
ergy transfer. Energy is transferred to the semiconductor through the enhanced
field around the plasmonic nanoparticles. . . . . . . . . . . . . . . . . . . . . . 41
2.7 Schematic illustration of TTA-based upconversion system. S1 is the singlet ex-

cited state while T1 is the triplet excited state. The TTET process occurs
between a sensitizer with populated T1 and an annihilator. The TTA process
happens between two annihilators with populated T1 and results in one singlet
excited annihilator which then emits upconverted photon. . . . . . . . . . . . . . 45
3.1 Schematic diagram of the PL spectrometer. During the measurement, the sample
is excited by the monochromatic light while the emitted photons are captured
by the photon detector after passing through another monochromator. The red
arrows represent the optical path of the excitation light while the blue arrows
represent the optical path of the emission light. . . . . . . . . . . . . . . . . . . 53
3.2 Schematic diagram of the UV-Vis-NIR spectrometer. The photon detector can
capture the transmitted light and thus measure the transmission intensity. . . . 54
3.3 Schematic diagram of the PEC photocurrent measurement circuit. Working

electrode, reference electrode and counter electrode are placed in a PEC cell
and fully immersed in electrolyte. All three electrodes are connected to the
potentiostat and the working electrode is illuminated by solar light from a solar
simulator. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
3.4 Schematic illustration of a self-designed PEC cell ready for 3D printing. An

o-ring and the working electrode will be inserted between the cell body and the
cover plate before measurements. . . . . . . . . . . . . . . . . . . . . . . . . . . 57
LIST OF FIGURES 15
3.5 Schematic diagram of the IPCE measurement setup. The PEC cell is the same as
the one used for PEC photocurrent measurement. Chopped monochromatic light
is used as the incident light. The lock-in amplifier can extract weak photocurrent
signals from the current readings of potentiostat using the chopper’s sine wave
as a reference signal. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4.1 SEM images of experimentally fabricated (a) NDA, (b) NTA and (c) NHA. The
scale bars in (a)(c) are 1000 nm. The scale bar in (b) is 500 nm. . . . . . . . . . 61
4.2 Top view of modelled plasmonic nanoarrays: (a) NDA, (b) NHA, (c) NTA. The
orange rectangles represent the 2D unit cells. . . . . . . . . . . . . . . . . . . . . 62
4.3 Schematic diagram of a FDTD modelling space for plasmonic nanoarrays (not to
scale). The modelling space has PML boundaries in the +Z/-Z directions (red)
and periodic boundaries in other directions (green). The plane wave source
(white) is above the plane of plasmonic nanoarrays with the incident direction
towards the plasmonic structure. . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4.4 FDTD simulated extinction spectra of Ag, Au, Al and Cu NDA with different
diameters. The thickness is fixed at 60 nm while the period is fixed at 500 nm.
The LSPR peak becomes less significant as the diameter gets closer to the period. 65
4.5 FDTD simulated normalized E-field mapping at LSPR wavelength of Au NDA

with diameter of left: 90 nm, right: 400 nm. Both of them have 60 nm thickness
and 500 nm period. Smaller diameter leads to stronger local enhanced field while
larger diameter results in larger enhanced area. . . . . . . . . . . . . . . . . . . 66
thickness. The diameter is fixed at 400 nm while the period is fixed at 500 nm.
Different thickness results in LSPR peaks at similar wavelength position. . . . . 67
period. The thickness is fixed at 60 nm while the diameter is fixed at 80% of the
period. Increasing the period results in significant red shift of the LSPR peak. . 67
LIST OF FIGURES 16
4.8 FDTD simulated extinction spectra of dual-layer NDA with various period. The
row indicates the metal in the bottom layer while the column indicates the metal
in the top layer. The thickness is fixed at 30 nm for each layer while the diameter
is fixed at 80% of the period. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
4.9 Illustration of broadband LSPR effect in dual-layer NDA with distinct LSPR
wavelength of each layer. In the spectral range between the two extinction peaks,
enhanced field exists at both bottom and top layer. . . . . . . . . . . . . . . . . 70
4.10 FDTD simulated normalized E-field mapping at LSPR wavelength of Au NTA

with period of left: 400 nm, right: 900 nm, both of which have 60 nm thickness.
The LSPR enhanced area is between the tips of the neighboring triangles. . . . 71
4.11 FDTD simulated extinction spectra of Ag, Au, Al and Cu NTA with different
The LSPR peaks slightly blue-shift as the thickness increases. . . . . . . . . . . 72
4.12 FDTD simulated extinction spectra of Ag, Au, Al and Cu NTA with different
period. The thickness is fixed at 60 nm. Increasing period results in significant
red shift of the LSPR peak. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
4.13 FDTD simulated extinction spectra of dual-layer NTA with various period. The
in the top layer. The thickness is fixed at 30 nm for each layer. . . . . . . . . . . 74
4.14 FDTD simulated normalized E-field mapping at (a), (b): extinction peak and
(c), (d): extinction dip. (a) and (c) represent top view while (b) and (d) repre-
sent cross-sectional view. The modelled NHA is Au NHA with period of 500 nm,
diameter of 400 nm and thickness of 60 nm. The curve in the middle is its sim-
ulated extinction spectrum. The hot zones in field mapping are inside the holes
at extinction dip while above the continuous metal at extinction peak. . . . . . 75
4.15 FDTD simulated extinction spectra of Ag, Au, Al and Cu NHA with different
diameters. The thickness is fixed at 60 nm while the period is fixed at 500 nm.
The extinction intensity decreases as the diameter decreases. . . . . . . . . . . . 76
LIST OF FIGURES 17
Varying the thickness has negligible influence on the extinction spectrum. . . . . 77
4.17 FDTD simulated extinction spectra of Ag, Au, Al and Cu NHA with different
period. The thickness is fixed at 60 nm while the diameter is fixed at 80% of the
period. Increasing the period results in significant red shift of the SPP wavelength. 78
4.18 FDTD simulated extinction spectra of dual-layer NHA with various period. The
in the top layer. The thickness is fixed at 30 nm for each layer while the diameter
is fixed at 80% of the period. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
5.1 Schematic illustration of the BiVO4-coated hematite nanorods photoanode en-

hanced by Ag/Au dual-layer nanohole array. . . . . . . . . . . . . . . . . . . . . 82
5.2 Step-by-step illustration of the colloidal lithography method used to fabricate

nanohole arrays. The steps are (i) coating self-assembled template nanospheres
onto substrate, (ii) using reactive ion etching to reduce the size of the spheres,
(iii) coating of the plasmonic metal layer and (iv) removing the organic residue
by duct tape. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
5.3 TEM image of polystyrene spheres used as template in the colloidal lithography
process. The diameter of the spheres is measured to be around 400 nm. The
scale bar is 500 nm. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
5.4 SEM images of hexagonally packed polystyrene spheres on the FTO substrate
(a) before the reactive ion etching process and (b) after the reactive ion etching
process. The etching process shrank the polystyrene spheres and thus created
larger spacing between adjacent spheres. The scale bar is 1000 nm. . . . . . . . 87
5.5 SEM images of 400 nm plasmonic nanohole arrays (a) on the glass substrate and
(b) on the FTO substrate. The FTO substrate is observed to have higher surface
roughness compared to glass. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
LIST OF FIGURES 18
5.6 SEM images of (a) hematite nanorods, with length measured to be around
350 nm, (b)(c) BiVO4 coated hematite nanorods. A scratch was made on the
surface of sample in (c) to reveal the hematite nanorods and the FTO surface
underneath the BiVO4 layer. The scale bars are 500 nm. . . . . . . . . . . . . . 89
5.7 Digital photograph of LEFT: blank FTO, MID: FeOOH nanorods on FTO,
RIGHT: hematite nanorods on FTO. The conversion from FeOOH to hematite
requires annealing at 450 ◦C. An area of the FTO substrate is left uncovered to
preserve a highly conductive area to attach wires for PEC measurement. . . . . 90
5.8 SEM image of BiVO4 layer on FTO (a) before annealing and (b) after annealing
at 500 ◦C. The scale bars are 1000 nm. The morphology turned from nanoflakes
to thin film during the annealing process. . . . . . . . . . . . . . . . . . . . . . . 90
5.9 EDX spectra of (a) hematite nanorods on FTO and (b) BiVO4 coated hematite
nanorods on FTO. A clear difference can be noticed at 2.4 eV and 4.9 eV, corre-
sponding to Bi M and V Kα, respectively. . . . . . . . . . . . . . . . . . . . . . 91
5.10 (a) XPS survey spectrum of fabricated BiVO4-coated photoanode and core level
spectra of (b) Bi 4f , (c) V 2p, (d) Fe 2p. . . . . . . . . . . . . . . . . . . . . . 92
5.11 (a) Extinction spectra of pristine hematite and BiVO4-coated hematite. Im-
proved absorption in the range of 450 nm-700 nm can be observed for the sample
with BiVO4. (b) Tauc plot of pristine hematite and BiVO4-coated hematite. The
sample with BiVO4 has a smaller optical band gap. . . . . . . . . . . . . . . . . 94
5.12 Normalized extinction spectra of Au NHA (solid lines) and Ag/Au dual-layer
NHA (dotted lines) with period of 400 nm. The extinction peak of Au NHA below
500 nm is due to the interband absorption instead of plasmonic effect. The EOT
wavelength of Ag/Au dual-layer NHA is at around 500 nm. The above-bandgap
region of hematite is marked in yellow. . . . . . . . . . . . . . . . . . . . . . . . 95
5.13 Measured photocurrent density of the hematite nanorods with different times of
BiVO4 layer spin-coating. The one with 5 times of spin-coating shows the best
performance among the all measured samples. . . . . . . . . . . . . . . . . . . . 96
LIST OF FIGURES 19
5.14 PEC photocurrent density of measured photoanodes under chopped light from
solar simulator. The potential is calibrated vs Ag/AgCl reference electrode and
converted to vs RHE. The BiVO4-coated samples shows much better perfor-
mance compared to pristine hematite. Performance enhancement can be seen
for samples with metal nanoarrays while the sample with dual-layer NHA shows
the best performance. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
5.15 IPCE curve of measured photoanodes at 1.23 V vs RHE. The sample with dual-
layer NHA shows the best performance, which is in confirmation with PEC
photocurrent measurement. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
6.1 Schematic illustration of the hematite nanotube photoanodes based on aluminum

nanohole array. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
6.2 Illustration of the critical steps to fabricate hematite nanotube photoanode on Al

nanohole array. The steps are (a) coating self-assembled template nanospheres
onto the substrate, (b) using reactive ion etching to reduce the size of the spheres,
(c) coating of the Al layer and the silica protective layer, (d) removing the organic
residue by duct tapes and (e) coating as-synthesized hematite nanotubes onto
the nanohole array. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
6.3 (a) TEM image of hematite nanotubes. The scale bar is 100 nm. (b) STEM
image of hematite nanotubes. The scale bar is 200 nm. (c) HRTEM image of
hematite nanotubes. The scale bar is 5 nm. The distance between adjacent
(d) TEM diffraction pattern of hematite
lattice planes is measured to be 2.2 A.
nanotubes. The scale bar is 1 nm−1 . . . . . . . . . . . . . . . . . . . . . . . . . . 104
6.4 XRD pattern of single crystalline hematite nanotubes. . . . . . . . . . . . . . . . 105
6.5 FDTD modelled extinction spectra of Al nanohole array with various thickness
of oxidization layer (Al2O3). With thicker oxidization layer, the extinction curve
becomes flatter. The inset table lists the normalized plasmonic enhancement
factors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
LIST OF FIGURES 20
6.6 SEM images of (a) self-assembled monodispersed polystyrene spheres on FTO

substrate, (b) Al nanohole array on FTO substrate and (c) fabricated photoanode
with hematite nanotubes on Al nanohole array. The scale bars are 500 nm. . . . 107
6.7 Normalized extinction spectra of Al nanohole arrays and hematite nanotubes.

The EOT wavelength is 490 nm for 500 nm Al NHA and 570 nm for 620 nm Al
NHA. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
6.8 PEC photocurrent density of the fabricated hematite nanotube photoanodes.

Best performance is achieved by hematite nanotubes with 500 nm Al NHA. . . . 110
6.9 IPCE spectra of fabricated hematite nanotube photoanodes. The black curve
represents the reference sample of pristine hematite and the grey curve represents
the reference sample of hematite with Al film. Best performance is achieved by
the sample with 500 nm Al NHA, which is in consistent with the photocurrent
density curves. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
6.10 (a), (b) Experimental (red) and fitted (green) electrochemical impedance spec-
troscopy (EIS) plot of reference samples (hematite nanotubes on Al film), (c),
(d) Experimental (red) and fitted (green) EIS plot of hematite nanotubes on Al
NHA, (e) Equivalent circuit used for fitting. Inset table shows the fitted results.
The lower R1 value of reference sample indicates it is more conductive which
is due to the absence of silica layer. The lower Rct value of hematite+Al NHA
shows it has better water splitting performance. . . . . . . . . . . . . . . . . . . 112
6.11 Photoelectrochemical stability measurement of the fabricated hematite nanotube

photoanodes for 240 min. Pristine hematite photoanodes has the most stable
current over the measured time interval. Al NHA sample with protective layer
has more than 90% of the original performance after 240 min. Al film sample
without protective layer has less than 30% of the original performance at the
end of the measurement. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
LIST OF FIGURES 21
7.1 Schematic illustration of the inverse-opal hematite photoanodes with Ag-couple

TTA upconversion materials. The interactions between the components are:
(i) plasmonic enhanced upconversion, (ii) absorption of upconverted photons by
hematite and (iii) plasmonic enhanced water splitting . . . . . . . . . . . . . . . 116
7.2 Illustration of the fabrication process of hematite inverse-opal photoanodes. (a)

Self-assembly of multi-layer polystyrene spheres on FTO substrate forming an
opal structure. (b) Growth of iron by electrodeposition. (c) Annealing process
to remove organic residue and oxidize iron to hematite. . . . . . . . . . . . . . . 119
7.3 TEM image of synthesized TTA@SiO2 core-shell upconversion nanoparticles.

The particle size is 9 nm-12 nm and the thickness of the silica coating is 2 nm-
3 nm. (a) The scale bar is 10 nm. (b) The scale bar is 30 nm. . . . . . . . . . . . 120
7.4 PL intensity of synthesized TTA@SiO2 core-shell upconversion nanoparticles. (a)

Excitation-emission mapping indicating the upconversion process from around
510 nm-545 nm to 410 nm-455 nm. The best upconversion efficiency is achieved
by excitation at around 520 nm and emission at around 440 nm. (b) Normalized
emission spectrum (with excitation at 520 nm) and excitation spectrum (with
emission at 440 nm) of TTA@SiO2 upconversion nanoparticles together with the
extinction spectrum of hematite. The emission spectrum of the TTA@SiO2 up-
conversion nanoparticles overlaps with the extinction peak of hematite. . . . . . 121
7.5 Extinction spectra of Ag nanoparticles with different diameters ranging from

10 nm to 100 nm. Smaller Ag nanoparticles have more significant LSPR peaks.
The emission and excitation peak positions of the TTA@SiO2 upconversion
nanoparticles are marked in green and blue, respectively. The emission peak
of TTA overlaps with the LSPR peaks of all the measured Ag nanoparticles. . . 122
7.6 Measure PL emission intensity of Ag-coupled TTA@SiO2 upconversion nanopar-

ticles with different particle ratio between TTA and (a) 10 nm Ag, (b) 40 nm Ag,
(c) 60 nm Ag, (d) 100 nm Ag. All samples were excited by 520 nm incident light. 123
7.7 PL emission spectrum of TTA@SiO2@Ag particles with 520 nm incident wave-

length. The sample with 10 nm Ag nanoparticles exhibits the best performance. 124
LIST OF FIGURES 22
7.8 TEM images of TTA@SiO2 nanoparticles coupled with Ag nanoparticles. Left:

the scale bar is 50 nm. Right: the scale bar is 200 nm. Larger particles are formed
due to the aggregation of Ag nanoparticles (darker particles) and TTA@SiO2
nanoparticles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
7.9 SEM images of the self-assembled multi-layer polystyrene spheres (a), (b) before
UV-ozone treatment and (c), (d) after UV-ozone treatment. The scale bars are
1000 nm. The gap between adjacent spheres are significantly larger after the
UV-ozone treatment, leaving more space for electrodeposition. . . . . . . . . . . 126
7.10 SEM images of the finally fabricated inverse-opal photoanodes. The high porous
structure has hole size measured to be 150 nm-200 nm. (a) The scale bar is
1000 nm. (b) The scale bar is 300 nm. . . . . . . . . . . . . . . . . . . . . . . . . 126
7.11 Solar water splitting performance of hematite inverse-opal photoanodes with

TTA upconversion materials. (a) PEC photocurrent curves under simulated
solar light. The sample with TTA+Ag(10 nm) has slightly better performance
than other samples. (b) IPCE with monochromatic incident light in wavelength
range of 300 nm-600 nm. Peaks at around 520 nm can be observed for samples
with upconversion nanoparticles. The sample with TTA+Ag(10 nm) has the
highest IPCE at 520 nm. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
Chapter 1
Introduction
23
24
Due to the rise in world energy consumption lead by economic growth and expanding popula-
tion, discovering a reliable and economic energy source has become a critical task. The solar
energy can be regarded as a sustainable and unlimited energy source[6]. Finding a feasible way
to harvest solar energy is a promising solution to the foreseeable energy crisis in the future.
Photovoltaic systems that enables direct solar-to-electrical conversion, although widely studied
and utilized, can only supply up to 35% of our overall energy needs[7]. Moreover, the replace-
ment of fuel driven devices with electricity driven devices could be hindered by the low energy
density of batteries[8]. Analogous to the photosynthesis method employed by plants in the
nature, artificial photosynthesis methods can be utilized as a feasible, reliable and efficient way
to not only harvest but also store the solar energy in the form of chemical bonds where high
energy density can be easily achieved. Among distinct photosynthesis approaches, artificial
solar water splitting is an advanced method that use water as ingredient to generate hydrogen
and oxygen with no harmful side products[9]. As a recyclable and clean fuel, hydrogen can be
used as an ideal alternative to fossil fuels and already has its own developed markets such as
hydrogen driven engine[10] and fuel cells[11].
Various implementations of artificial solar water splitting have been developed including pho-
tocatalyst (PC) and photoelectrochemical (PEC) devices[12]. The PEC solar water splitting
utilizes a photoactive electrode, usually a semiconductor, to drive water splitting reactions.
It exhibits its unique advantage by spatially separating the hydrogen evolution process and
the oxygen evolution process. Hence, no further steps are required for the purification of the
produced gas. Many semiconducting oxides can be used as photoanodes for artificial solar wa-
ter splitting. Among them, hematite is earth-abundant, cost-efficient and non-toxic, making
it a potentially excellent candidate that can be employed in large-scale industrial solar water
splitting devices. Nonetheless, pristine hematite has poor performance and thus strategies to
improve its performance are necessary for highly efficient solar water splitting systems based
on hematite.
Being capable of manipulating and enhancing electromagnetic fields by localized surface plas-
mon resonance (LSPR) and surface plasmon polariton (SPP), plasmonic nanostructures have
been widely explored for chemical sensing[13], biosensing[14] and microscopic imaging[15]. Em-
25
ployed as components for field enhancement in solar water splitting systems, plasmonic nanos-
tructures can improve the overall efficiency via light scattering, plasmon-induced resonance
energy transfer (PIRET) as well as hot electron injection (HEI)[16]. Therefore, utilizing them
in hematite based solar water splitting systems is a feasible approach to tackle the performance
challenge. In this research project, I focused on the plasmonic enhanced PEC water splitting
by hematite-based nanocomposites. A systematic model of plasmonic nanoarrays based on
finite-difference time-domain method (FDTD) were firstly built to get insight of the optimal
plasmonic nanostructures, laying the theoretical framework for plasmonic enhanced water split-
ting systems. Three plasmonic enhanced PEC water splitting systems are established based
on distinct hematite photoanodes and various plasmonic nanostructures, together with other
optimization methods include nanostructuring, surface modification and plasmonic enhanced
upconversion.
This chapter, as the introduction, briefly outlines the background and motivations for the
project as well as the chapters of the PhD thesis. Chapter 2 reviews the principles and the
state of the arts of all the involved components including plasmonic nanostructures, solar water
splitting systems as well as upconversion materials in depth. Some mechanisms utilized in this
project, such as plasmonic enhanced water splitting, upconversion induced water splitting and
plasmonic enhanced upconversion are also reviewed in Chapter 2. Chapter 3 introduces the
instrumentation and characterization methods used in this project.
In Chapter 4, FDTD models of three different plasmonic nanoarrays: nanodisk array (NDA),
nanotriangle array (NTA) and nanohole array (NHA) are established and discussed, laying
the theoretical framework for the following chapters. These nanoarrays can be fabricated ex-
perimentally through similar procedures using self-assembled polystyrene spheres and e-beam
evaporation or sputtering. They are selected as candidates of plasmonic nanostructures to be
employed in the following chapters. Four plasmonic metals (Ag, Al, Au and Cu) are modelled
for the nanoarrays, with both single-layer and dual-layer nanoarrays taken into consideration.
Strong and tunable LSPR peaks in NDA and NTA as well as extraordinary transmission effect
(EOT) in NHA have been examined and the spatial distribution of the enhanced field is also
studied.
26
In Chapter 5, a dual-layer Ag/Au NHA is used to enhance hematite photoanodes based on 1D

nanorods. The hematite nanorods were coated with a layer of BiVO4 and results in improved
performance, compared to pristine hematite nanorods, due to the formation of BiVO4-hematite
heterojunction. With plasmonic nanoarrays applied, further improvement of photocurrent den-
sity was observed.
In Chapter 6, plasmonic enhanced photoanodes with Al NHA and hematite nanotubes were
fabricated and investigated. The enhancement from the non-noble plasmonic nanostructure
based is significant and up to 3.6 times improvement of PEC photocurrent density can be
observed. The enhancement mainly originates from the plasmon-induced resonance energy
transfer by the SPP of the plasmonic nanoarray.
In Chapter 7, triplet-triplet annihilation (TTA) based upconversion nanoparticles, capable of

converting incident photons at around 520 nm, to photons at around 440 nm, were synthesized
to harvesting below-bandgap energy and improved the performance of solar water splitting
systems. Furthermore, Ag nanoparticles of different sizes were attached to the upconversion
nanoparticles for plasmonic enhanced upconversion. Ag nanoparticles with 10 nm in diameter
shows the best performance and a two-fold enhancement can be achieved for plasmonic enhanced
TTA upconversion. The upconversion composites were then incorporated into a highly porous
hematite photoanode with inverse-opal structure. The incident photon-to-current efficiency
(IPCE) was observed to be enhanced by 61 times at 520 nm.
Chapter 8 summarizes the conclusions and proposed future work. The fabricated photoanodes
proved the concept of plasmonic enhance PEC water splitting by hematite-based nanocom-
posites as well as utilizing plasmonic enhanced TTA upconversion materials in water splitting
systems. The established FDTD models of the nanoarrays may help evaluate not only plas-
monic enhanced water splitting systems but also other applications in different spectral range.
The study of this novel concept can help deepen the understanding of the plasmonic nanostruc-
tures and PEC water splitting, paving the way towards the large-scale applications of artificial
solar water splitting systems.
Chapter 2
Literature Review
27
2.1. Plasmonics 28
2.1 Plasmonics
2.1.1 Principles of Plasmonics
Plasmon resonant modes originate from the interaction between electromagnetic radiation and
free charges. Plasmonics takes advantage of plasmons by coupling plasmonic materials and
incident light at specific frequencies to enable significant field enhancements. Materials that
have free charges and thus plasmonic resonant modes are called plasmas. Having freely mobile
electrons, metals are the most typical plasmas. Bulk plasmas like metals have a threshold
frequency, namely the plasma frequency ωp , which is given by[17]:
ne2
ωp2 = (2.1)
0 m
where n is the number density of free carriers (electrons for metal); e is the charge of the carrier;
m is the mass of the carrier; 0 is the relative permittivity of free space.
When the incident photons’ frequency is below the plasma frequency, the free electrons can
move in a way to cancel the incident wave so that the incident photons will be reflected. On
the other hand, if the incident photons’ frequency is above the plasma frequency, they can
transmit through the metal because the free electrons cannot respond quickly enough to filter
out the incident field. The Drude model can be used to describe the plasmonic properties,
which is given by[18]:
ωp2
m (ω) = 1 − 2 (2.2)
ω + iγω
where m (ω) is the relative permittivity of the metal and γ is its relaxation frequency.
Plasmonic effects can be generally divided into two types: the localized surface plasmon reso-
nance (LSPR) and the surface plasmon polaritons (SPP)[19]. In LSPR mode, the free carriers
will redistribute to provide a restoring force against the dipole moment which originates from
2.1. Plasmonics 29
Figure 2.1: (a) Schematic illustration of localized surface plasmon resonance. The electron
cloud oscillates when the incident field couples to the LSPR mode. (b) Schematic illustration
of surface plasmon polaritons. The propagating oscillation is at the metal-dielectric interface.
the external incident field. As illustrated in Figure 2.1(a), the incident field couples with the
free electrons oscillations and thus generate a strong localized electromagnetic field. The plas-
monic resonance wavelength of LSPR mode is significantly related to the feature size and the
morphology of the nanostructure[20]. Usually, larger nanoparticles result in longer plasmonic
resonance wavelength. On the other hand, the SPP mode induces propagating oscillation and
usually happens at the metal-dielectric interface, as shown in Figure2.1(b). For a planar metal
surface, the dispersion relation of SPP is given by[18]:
r
2π 2π m d
= (2.3)
λSP P λ0 m + d
where λSP P is the SPP wavelength; λ0 is the incident wavelength; m and d are the permitivities
of the metal and the dielectric, respectively. Freely propagating incident light is unable to couple
to the SPP mode since the square root term is usually greater than one. Special momentum
matching techniques like prism coupling is necessary in this case. An alternative method is to
use metal films with sub-wavelength periodic structures, such as nanohole arrays[21]. The SPP
wavelength is related to the period of such nanostructures. In contrast to LSPR mode, the
electromagnetic field from SPP mode has a much longer decay length and can thus penetrate
deep into the dielectric[22]. However, LSPR mode is more flexible with a small occupied space
and only related weakly to the incident angle[23]. In some cases, these two modes can exist
2.1. Plasmonics 30
Figure 2.2: Map of common plasmonic materials. Materials with high interband losses are
represented by elliptical bubbles and those with low interband losses are represented by spherical
bubbles. The other two axes are carrier mobility and carrier concentration. From [2]. Reprinted
with permission from AAAS.
simultaneously, resulting in strong field enhancement and highly tunable plasmonic resonance
wavelength.
2.1.2 Plasmonic Metals and Nanostructures
Conventional plasmonic materials include metal and metal-like materials with negative real
permittivity[24]. As the application domain of plasmonic enhancement keeps expanding, the
search for more promising plasmonic materials never stopped. It is discovered that plasmonic
materials can be noble metals, metal nitrides[25], metal alloys[26], transparent conductive
oxides[27], graphene[28] and other semiconductor materials[29]. Natural plasmonic materi-
als usually have many drawbacks such as the difficulty of tuning its spectral range of plasmonic
response. Plasmonic nanostructures has received significant attention from researchers[30].
The careful design and structuring lead to the highly controllable optical properties of plas-
monic nanostructures, which provides the potential of application in many different fields
2.1. Plasmonics 31
such as chemical sensing[31], biological sensing[32], solar energy harvesting[33] and microscopic
imaging[15].
As shown in Figure 2.2, metal-based plasmonic nanostructures takes an important part espe-
cially in the UV-Vis-NIR region. Ag, Au, Cu and Al are the most common and well-studied
plasmonic metals and a tremendous amount of nanostructures based on them have been re-
ported for various applications. Tsai et al. fabricated a plasmonic gold nanorings arrays with
2D square lattice[34]. A LSPR resonance in NIR region around 1300 nm can be observed which
is suitable for biomedical applications. Shen et al. reported plasmonic gold mushroom array
fabricate by two-beam interference lithography and thermal evaporation[35]. The nanoarray
has LSPR peaks at around 1053 nm and 1174 nm and can be used as a label-free biosensor. The
nanoarray also has a high figure-of-merit value measured to be 80–108 which is very close to
the theoretical limit. Compared to Au, Ag benefits from a smaller inter-band loss and results
in a more intense plasmonic resonance. Meanwhile, the plasmonic resonance wavelength of Ag
is shorter than that of Au. Jakab et al. synthesized highly sensitive Ag nanorods with LSPR
peaks from 633 nm to 930 nm controlled by the length of the nanorods[36]. The sensitivity of
Ag nanorods is determined to be 1.2 to 2 times higher than Au.
Non-noble metal based plasmonic nanostructures open up new possibilities in many fields due
to its cost-effectiveness. Cu has the similar plasmonic resonant wavelength as Au while Al can
extend the plasmonic resonance to deep UV region, making them the promising alternatives.
Chan et al. reported the fabrication of oxide-free copper nanotriangle arrays by nanosphere
lithography[37]. The copper-based plasmonic nanostructure has LSPR wavelength tunable from
600 nm to 900 nm. It is also demonstrated that the spectral position of LSPR peak is identical
to gold nanotriangle arrays with the same geometrical parameters although the LSPR intensity
of Cu is weaker. Knight et al. studied the plasmonic properties of Al nanorod antennas[38]. It
is found that plasmonic resonances is highly tunable by controlled the antenna length and range
from deep UV region to the visible region. Li et al. reported a novel super absorber consists of
Ag/SiO2/Al nanohole arrays[39]. The SPP at metal-dielectric-metal interface was utilized at
this system to enhance the absorption at the visible-NIR region. Extraordinary transmission
can also be observed in the system at certain wavelength due to SPP. Some selected literature
2.1. Plasmonics 32
Table 2.1: Selected literature reports of metal-based plasmonic nanostructures and applications
Resonance
Nanostructure Synthesis/Fabrication Application
Wavelength
Fluorescence
Au nanostar[32] Seed-based growth NIR
enhancement
Au nanorings
Electron beam lithography NIR Index sensing
array[34]
Au mushroom
Electron beam lithography NIR Index sensing
array[35]
Au Nanocross[40] Facile chemical synthesis NIR-mid IR Photothermal
Self-assembly in
Au nanorods[41] Visible Imaging
anisotropic fluids
Au nanohole Nanosphere lithography &
Visible Water splitting
array[33] E-beam evaporation
Ag nanorods[36] Seed-based growth Visible-NIR Index sensing
Hydrothermal chemical
Ag nanowires[42] Visible Solar cell
synthesis
Ag nanoflowers[43] Seed-based growth Visible Random laser
Surface-enhanced
Ag nanodomes[44] Electro deposition Visible-NIR
raman scattering
Al nanorods[38] planar lithography UV-Visible N/A
Al/SiO2/Al Colloidal lithography & Infrared
NIR
nanodisks array[45] Dry etching spectroscopy
Ag/SiO2/Al nanohole Nanosphere lithography &
Visible-NIR Super absorber
arrays[39] E-beam evaporation
Cu nanotriangle Nanosphere lithography & Surface-enhanced
Visible-NIR
arrays[37] E-beam evaporation raman scattering
Hydrothermal chemical
Cu nanowire[46] Visible Solar cell
synthesis
reports of metal-based plasmonic nanostructures and their applications are listed in Table 2.1.
2.1.3 FDTD Modelling
In order to match with the system they are attaching to, it is necessary to tailor the resonance
wavelength of the plasmonic nanostructures. It is also helpful to find the spatial distribution
of the plasmonic enhanced electromagnetic field. This can be achieved experimentally by tech-
niques such as UV-Vis-NIR spectroscopy for finding the extinction peak and electron energy
loss spectroscopy (EELS) which can be used to visualize the enhanced field[47]. Nevertheless,
modelling methods that can predict these properties will be beneficial to save budget and time
2.1. Plasmonics 33
Figure 2.3: Schematic illustration of Yee cell used in FDTD models. The voxel of H components
(green) has a shift of half unit length from the voxel of E components (blue).
as well as allow a high-throughput search for desired plasmonic nanostructures. Various meth-
ods have been developed over the last century to provide an accurate numerical solution for
Maxwell equations which can be used to model plasmonic nanostructures, such as finite ele-
ment method (FEM), boundary element method (BEM), discrete dipole approximation (DDA)
and finite difference time domain method (FDTD). As a versatile modeling method for solving
Maxwell equations numerically, FDTD was firstly proposed by Yee in 1966[48] and has become
a well-developed technique now.
The key approximation used in FDTD is using Yee cells in space together with a time-stepping
scheme. The FDTD model utilize the Yee cells as voxels in space while there is a shift of half
unit length between the voxels of E components and the voxels of H components, as illustrate
in Figure 2.3. Due to the Maxwell curl equation, the change of E-field in time depends on the
changing H-field and vice versa. The E-field will thus be updated at each time step based on
the previous H-field and the H-field will also be updated between two successive updates of
E-field[49].
The input of a FDTD model consists of three major parts: the geometry of modelled structures,
2.2. Solar Water Splitting 34
the source and the boundary conditions. The permittivity of plasmonic material is usually de-
fined by Drude-Lorentz model. As a time-domain method, the response of the system can be
obtained over a wide range of wavelengths with the help of discrete Fourier transform if the
source is a broadband pulse. The boundary conditions defines the border of the simulated area.
Commonly used boundary conditions are perfectly matched layer[50] which is used to absorb
any excessive wave, and periodic boundary which is used to simulate periodic structures[51].
The output of a FDTD model is the E-field and H-field at each time step within the com-
putational domain. Other valuable information such as transmission spectrum, absorption
spectrum, and field mapping can also be obtained.
2.2 Solar Water Splitting
2.2.1 Photoelectrochemical Water Splitting
Artificial solar water splitting is a photosynthesis method to convert the solar energy directly to
the energy stored in chemical bonds. It is an elegant, realistic, and more efficient way to produce
clean and recyclable fuel: hydrogen. Any water splitting process involves two half reactions:
the oxygen evolution reaction (OER) and the hydrogen evolution reaction (HER)[52]:
4OH– + 4h+ → 2H2O + O2 (2.4)
2H2O + 2e– → H2 + OH– (2.5)
The water splitting reaction has a potential of 1.23V vs reversible hydrogen electrode (RHE). As
a four-electron process, the water oxidation reaction is both kinetically and thermodynamically
difficult. Besides, multiple intermediates are involved in the reaction, thus a large overpotential
is usually expected which means photons with energy much higher than the semiconductor’s
band gap are necessary to drive the reaction[53]. To enable the solar water splitting process,
a photoelectrode or a photocatalyst is essentially required to provide electrons with enough
energy. On account of this, artificial solar water splitting has two different pathways: via
Figure 2.4: Schematic illustration of (a) the process of PC water splitting: the photogenerated
electron-hole pairs move to the surface active site to drive the HER and the OER[3], (b)
the process of photoanode induced PEC water splitting: the photogenerated holes move to
the surface of photoanode to drive the OER while the photogenerated electrons move to the
cathode through external circuit to drive the HER[4].
photocatalytic (PC) method or via photoelectrochemical (PEC) method[54], as shown in Figure

2.4.
The PC water splitting is a more straightforward process which employs a photoactive semi-
conductor. Upon absorbing photons with a energy higher than its band gap, an electron from
the valence band will be excited to the conduction band, generating an electron-hole pair[55].
The carriers will move to the surface active sites and thus the electrons can drive the HER
while the holes can drive the OER.
The PEC water splitting is similar to PC water splitting in principle. As illustrated in Figure
2.4(b), the PEC process utilizes the photoactive semiconductor as photoanode or photocathode
together with electrolyte and an external circuit as a complete loop. When the anode is the
photoactive electrode, a metal counter electrode will usually be used as the cathode. When
illuminated, the electron-hole pairs will be generated in the photoanode. The electrons will
travel through the external circuit and reach the surface of the cathode to participate in the
HER while the holes will move to the surface active sites of the photoanode to participate in
the OER[4]. During the process, photocurrent can be detected in the external circuit due to
the travelling of electrons which can be used to benchmark the water splitting performance.
One of the advantages of utilizing PEC is that it naturally separate the produced hydrogen and
oxygen since the HER and the OER take place in different locations[56]. Besides, it doesn’t
require the photoactive semiconductor to have both active sites for the HER and the OER,
leaving us more options in the selection of materials[4].
The requirements of the semiconductors that are utilized in PEC and PC solar water splitting
systems are similar as discussed above. The band gap of the photoactive semiconductor limits its
light harvesting efficiency. A semiconductor with lower band gap is usually more favorable as it
can cover larger area in the solar spectrum. As shown in Figure 2.5(a), Fe2O3 can absorb photons
with wavelength up to 600 nm while pristine TiO2 can only absorb photons with wavelength
below 380 nm due to its higher band gap (3.0 eV) compared to hematite’s (2.3 eV). For an
ideal candidate for PC water splitting, it is required that the conduction band and valence
band need to straddle the potential of the HER and the OER so that it can drive both half
reactions[57]. As shown in Figure 2.5(b), pristine Fe2O3 or TiO2 are not possible to drive
PC water splitting. But for PEC method, an external bias voltage can be provided by the
external circuit which makes the band position no longer a barrier. Besides band gap, the
stability of the photoactive semiconductor in electrolyte, its cost and its conversion efficiency
of photogenerated carriers are also critical[58]. Generally, transition metals oxides are excellent
candidates for photoactive material used in solar water splitting[59]. Researchers have done a
vigorous search for a photoactive material matching all these requirements but to date not a
single material has been found perfect to meet all the prerequisites[60]. For instance, TiO2 has
enough stability but visible light energy is mostly wasted due to its large band gap[61] while
some other low-bandgap semiconductors like InP shows poor stability in electrolyte[62] or not
cost-effective for applications in a larger scale.
2.2.2 Hematite Based Nanocomposites as Photoanodes
As an iron ore and the most stable form of iron oxide in nature, hematite (α Fe2O3) has
and c = 13.75 A
lattices parameters of a = 5.03 A in space group R3c. In hematite’s crystalline
structure[63], O2– are in a hexagonal closest packed lattice along the 001 direction and two-
thirds of the octahedral interstices basal planes are occupied by Fe3+[64].
Figure 2.5: (a) Absorption spectra of Fe2O3 and TiO2 along with the solar spectrum. Hematite
covers a larger spectral area in the solar spectrum due to its low band gap. (b) Reduction
potential of the HER and oxidation potential of the OER, as well as energy bands of common
semiconductors. From [5]. Reprinted with permission from Wiley.
Among photoactive semiconductors for solar water splitting, hematite, as one of the most well
researched and attractive candidates, exhibits its unique advantage of being earth-abundant,
cost-efficient, stable in aqueous solution and having a low band gap of 2.3 eV. Due to its stability
in electrolyte with high pH and the capability of converting up to 17% of energy of solar light
into energy of hydrogen[65], hematite is widely used as photoanode in PEC water splitting
systems. Besides, hematite has distinct morphologies such as nanoparticles[66], nanowires[67],
nanorods[68] and nanotubes[69], making it a promising photoactive semiconductor to couple
with different materials.
On the contrary, a wider application of hematite for solar water splitting has been limited
by its poor majority carrier conductivity and low charge diffusion length (< 10 nm). As a
result, the electron–hole pairs are easy to recombine before they can reach the active sites and
participate in the water splitting reactions, which is disadvantageous and will lower the overall
light harvesting efficiency. To overcome the shortcomings, 1D hematite nanostructures such as
nanorods have been developed[70], as they have larger surface area as well as shorter carrier
collection distance. Besides morphology control, more optimization strategies can be applied
to hematite as well as other photoanodes to further boost the performance[71].
Table 2.2: Selected literature reports of hematite photoanodes with different optimization meth-
ods
Technique Dopant Feature size Performance
IPCE 7% at 400 nm,
hematite nanorods[72] none 50 nm × 500 nm
0.9VRHE
Co-Pi & single-crystalline IPCE 50% at 400 nm,
Pt 50 nm × 500 nm
nanorods[70] 1.23VRHE
Potentiostatic anodization
none 6 nm × 1 mm 1.0 mA cm−2 at 1.23VRHE
nanotubes[73]
Magnetic field assisted
Sn 250 nm 2.7 mA cm−2 at 1.23VRHE
deposition[74]
Solvothermal-etched thin film[75] Ti 50 nm film 1.61 mA cm−2 at 1.23VRHE
Ultrasonic spray pyrolysis[76] Si 10 nm × 300 nm 1.5 mA cm−2 at 1.23VRHE
IPCE 8% at 400 nm,
Electro-deposition[77] Mo/Cr 20 nm - 50 nm
1.2VRHE
WO3 & hematite thin film[78] Si 50 nm film 1.3 mA cm−2 at 1.23VRHE
Plasmonic Au nanoholes &
none 50 nm × 500 nm 0.9 mA cm−2 at 1.23VRHE
hematite nanorods[33]
Rare-Earth Upconversion & 125 nA cm−2 at 1.23VRHE
none N/A
hematite film[79] under IR light
2.2.3 Optimization Strategies of Hematite Photoanodes
Various techniques including morphology control[73], doping[80], surface modification[81], uti-

lizing plasmonic nanostructures[33] and utilizing upconversion nanoparticles[79] have been car-
ried out to enhance the overall performance of hematite for solar water splitting. Some reported
results of hematite photoanodes using various optimization techniques has been listed in Table
2.2.
Morphology control for hematite is mainly for the purpose of increasing its absorption effi-
ciency while avoiding stack semiconductors together and result in a bulky system. Simply
increasing the thickness of hematite semiconductor layer seems to improve the absorption of
photons but it will usually lead to worse performance since hematite also has a short hole
diffusion length. To rectify this contradiction, 1D material and porous material have always
been the hot spots. In the work of Mohapatra et al.[73], potentiostatic anodization method
was used to synthesize hematite nanotubes for PEC water oxidation and achieved 1.0 mA cm−2
at 1.23VRHE . Beermann et al.[72] reported oriented nanorod thin films of hematite which has
7% IPCE at 400 nm and 0.9VRHE . Other advanced morphology control techniques utilizing
spray pyrolysis[76], electrochemical nanostructuring[82], magnetic field assisted deposition[74]

and atmospheric pressure chemical vapor deposition[83] for hematite photoanodes have also
been studied.
Adding dopants to pristine hematite can effectively improve its carrier conductivity by increas-
ing the carrier concentration due to:
σ = neµe + peµp (2.6)
where p and n are carrier concentration, σ is the conductivity, µe and µp are carrier mobility.
Appropriate dopants can compensate for the significantly low µe and µp of hematite. Divalent
metal dopants such as Cu and Ni can be replace Fe3+ in hematite’s lattice and form a p-type
semiconductor[84] while doping tetravalent metal such as Sn and Si into hematite can result
in a n-type semiconductor[80]. For photoanodes, n-type hematite is usually desired. It is
reported that by annealing hematite on fluorine-doped tin oxide (FTO) at a relatively high
temperature, the Sn from FTO can diffuse into the hematite layer as a dopant and achieve
five-fold performance improvement in PEC solar water splitting[85]. The Grätzel group also
reported a Si-doped hematite nanostructure that shows appreciable photocurrent related to the
increased Si4+ density[86]. Besides, Ti[87] and Mn[88] have also been investigated as promising
dopants for hematite photoanodes.
Another method of optimizing hematite photoanodes is surface modification. As noted before,

the OER process requires a considerable overpotential while untreated hematite has a slow
carrier transport rate across the photoanode-electrolyte interface. A surface treatment material,
usually another metal oxide, can act as a co-catalyst or a surface passivation layer to enhance
the carrier transportation. A co-catalyst can help transporting the carriers across the interface
while a surface passivation layer can reduce the defects near the interface and thus reduce
the charge recombination rate. For example, as a water oxidation catalyst, Co-Pi can be
attached to hematite photoanodes by photochemical method or electrodeposition and result
in increased photocurrent[89]. As a commonly used surface passivation material, a very thin
layer of Al2O3 can be coated to photoanodes by atomic layer deposition[81]. It is revealed that
utilizing corundum-type passivation layer in hematite water splitting systems can reduce the
surface states density and lattice strain in order to improve its water splitting performance[90].
Moreover, other materials including CoF[91], BiVO4[92] and ZnO[93] as surface passivation
layer as well as materials like IrO2[94] and cobalt oxides[95] as OER catalysts have also been
widely studied.
Utilizing plasmonic nanostructures and upconversion materials are also promising optimization
methods for photoanodes. Unlike the optimization methods mentioned above, upconversion
materials and plasmonic nanostructures can help photoanodes harvesting photons below its
bandgap, which expands the spectral region that photoanodes can cover in the solar spectrum.
Plasmonic enhanced water splitting has multiple complicated mechanisms. As an important
enhancement method utilized widely in this project, its principles and mechanisms will be
reviewed and discussed in the next section. Upconversion materials can also interact with
plasmonic nanostructures when applied together to water splitting systems and the details will
be further reviewed in Section 2.3.
2.2.4 Plasmonic Enhanced Solar Water Splitting
In photocatalytic and photoelectrochemical solar water splitting systems, the prime motivation
of utilizing plasmonic material is to improve the light harvesting efficiency and thus enhance
the the photogeneration of the charge carriers. Although it is usually difficult to identify the en-
hancement mechanisms in plasmonic enhanced solar water splitting system, many theories have
been proposed. Among them, three mechanisms have general acceptance: light scattering[96],
hot-electron injection (HEI) and plasmon-induced resonance energy transfer (PIRET)[16]. The
schematic illustration of these mechanisms is shown in Figure 2.6.
As the most apparent mechanism of plasmonic solar water splitting, the light scattering effect
utilize the enlarged scattering cross-sections of plasmonic nanostructures. In this way, the
plasmonic nanostructures can help concentrate the energy flux and act as mirrors to provide
transport pathways of incident photons towards the semiconductors, results in a improved light
harvesting efficiency. It is reported that the scattered light can be coupled into waveguiding
Figure 2.6: Schematic illustration of three mechanisms of plasmonic enhanced solar water
splitting. (i) Light scattering effect. The incident light is scattered to the semiconductor by the
plasmonic nanoparticles. (ii) Hot-electron injection. The plasmonic nanoparticles inject hot
electrons directly into the conduction band of the semiconductor when a direct contact exists.
(iii) Plasmon-induced resonance energy transfer. Energy is transferred to the semiconductor
through the enhanced field around the plasmonic nanoparticles.
modes or reflection modes of the semiconductor and up to more than 90% of the incident light
can be scattered into the semiconductor[97]. It is also noted that an insulating while optically
transparent layer as a spacer between the plasmonic material and the semiconductor plays an
important role to reduce the damping[98]. Nevertheless, if the incident photons have energy
below the bandgap of the semiconductor, this mechanism cannot help since the light scattering
process does not inject extra energy to the photons.
The HEI effect, also known as LSPR-sensitization effect, was firstly demonstrated by Tian et
al.[99]. The energy states of the electrons in the plasmonic metal can be upshifted by the
incident light coupled to the LSPR mode. When a direct contact between the plasmonic metal
and the semiconductor exists, the excited hot electrons can be injected into the conduction
band of the neighboring semiconductor. In this case, the plasmonic material acts as a sensitizer
working in a similar way as the dye molecules in dye-sensitized solar cells[100]. Furube et al.
utilized femtosecond transient absorption spectroscopy to experimentally provide the evidence
of HEI[101]. Upon illumination, the HEI between Au nanoparticles and titanium dioxide is
observed within 240 fs and has a yield of 40%. The required energy of the incident photons
to drive HEI is determined by the Schottky barrier height between the plasmonic metal and
the semiconductor instead of the band gap of the semiconductor. As a result, it is possible for
the HEI mechanism to work below the band gap of the semiconductor and thus expand the
spectral range of the light harvesting.
The PIRET mechanism induces the transfer of energy between the plasmonic material and a
neighboring absorber via the oscillating electric field originated from the plasmonic resonance.
When the peak plasmonic resonance overlaps with the spectrum of the incident light as well
as the absorption spectrum of the semiconductor, energy can be transferred to the semicon-
ductor through PIRET to drive solar water splitting[16]. Since the spatial distribution of the
enhanced electromagnetic field can be manipulated by adjusting the type and morphology of
the plasmonic nanostructure, it is possible to precisely control where the PIRET and the fol-
lowing charge separation take place. PIRET can mitigate the limitation from the short carrier
transport distance of some semiconductors including hematite. Therefore, it is helpful to find
the spectral and spatial distribution of the plasmonic enhanced electromagnetic field which can
be done by FDTD modelling[102] and is discussed in detail in the next section.
Various studies has been reported for plasmonic enhanced PEC water splitting. For example,
Li et al. reported aluminum based plasmonic enhanced water splitting with TiO2 nanocavity
arrays. Due to the LSPR induced HEI and PIRET, the photocurrent has a drastic improve-
ment from 0.2 mA cm−2 to 0.8 mA cm−2 at 1.23VRHE [103]. Moreover, a Au nanohole array
enhanced hematite water splitting system was reported[33]. The Au nanohole array has both
SPP and LSPR modes and boosts the water splitting performance of hematite nanorods to
0.9 mA cm−2 at 1.23VRHE with a ten-fold improvement. Table 2.3 summarises the reported
results of plasmonic enhanced PEC water splitting using different combinations of photoactive
semiconductors and plasmonic nanostructures.
Table 2.3: Selected literature reports of plasmonic enhanced PEC water splitting
Plasmonic Performance without Performance with
Semiconductor
nanostructure plasmonic plasmonic
0.5 mA cm−2 at 1.5 mA cm−2 at
ZnO nanowire[104] Au nanoparticles
1.23VRHE 1.23VRHE
sandwiched Au 0.25 mA cm−2 at 0.60 mA cm−2 at
WO3 film [105]
film 1.23VRHE 1.23VRHE
TiO2 nanotube
with photonic Au nanocrystals
1.23VRHE 1.23VRHE
crystal [106]
0.03 mA cm−2 at 0.10 mA cm−2 at
CuWO4 film [107] Au nanoparticles
1.23VRHE 1.23VRHE
ZnFe2O4/ZnO 0.3 mA cm−2 at 0.8 mA cm−2 at
Au nanoparticles
nanorod [108] 1.23VRHE 1.23VRHE
TiO2 nanocavity 0.2 mA cm−2 at 0.8 mA cm−2 at
Al coating
arrays [103] 1.23VRHE 1.23VRHE
WO3 nanorod with 0.8 mA cm−2 at 1.3 mA cm−2 at
Ag nanoparticles
CdS [109] 1.23VRHE 1.23VRHE
Fe2O3 nanorod [33] Au nanoholes
1.23VRHE 1.23VRHE
Zr-doped Fe2O3 IPCE 1.5% at IPCE 3.2% at
Au nanoparticles
film [110] 500 nm, 0.8VSCE 500 nm, 0.8VSCE
0.10 mA cm−2 at 0.35 mA cm−2 at
BiVO4 film [111] Pd nanoparticles
1.23VRHE 1.23VRHE
2.3. Upconversion 44
2.3 Upconversion
2.3.1 Upconversion and TTA-based Upconversion
Upconversion refers to the process which produces photons at shorter wavelengths after the se-
quential absorption of two or more high-energy photons. Upconversion process can be achieved
through various mechanisms with both inorganic and organic materials. Inorganic upconversion
materials such as lanthanide-based nanoparticles generally utilize three different mechanisms:
excited state absorption, energy transfer upconversion and photon avalanche[112]. On the other
hand, organic upconversion systems are usually based on the triplet-triplet annihilation (TTA)
mechanism[113]. Compared to other mechanisms, TTA-based upconversion system exhibits
its unique advantage of high quantum yields even with low excitation intensity which has a
state-of-art quantum efficiency of up to 21%[114]. Besides, the capability of visible light driven
green-to-violet and green-to-blue upconversion in TTA-based systems makes them the best
candidates to be utilized in solar energy harvesting systems[115].
The TTA-based upconversion systems consist of two main components known as sensitizer
and annihilator[116]. During the upconversion process, the incident low-energy photons are
firstly absorbed by the sensitizer and thus populate the singlet excited state of the sensitizer.
Subsequently, the electrons in the singlet excited state transfer to a lower energy level, namely
the triplet state, through intersystem crossing (ISC). Afterwards, the sensitizer transfers the
excitation energy to the annihilator and excites the annihilator to its triplet excited state while
the sensitizer itself returns to the ground state. Since this energy transfer induces the exchange
of triplet excited state between the two components, it is named triplet-triplet energy transfer
(TTET). Eventually, triplet-triplet annihilation occurs between two triplet excited annihilators
and results in a ground state annihilator and a singlet excited annihilator which can then emit
an upconverted photon. The schematic illustration of the TTA-based upconversion process is
shown in Figure 2.7.
TTA-based upconversion materials have a wide range of applications especially in the fields
of solar energy harvesting[117]. Nattestad et al. integrated TTA-based upconversion mate-
Figure 2.7: Schematic illustration of TTA-based upconversion system. S1 is the singlet excited
state while T1 is the triplet excited state. The TTET process occurs between a sensitizer
with populated T1 and an annihilator. The TTA process happens between two annihilators
with populated T1 and results in one singlet excited annihilator which then emits upconverted
photon.
rials in dye-sensitized solar cell[118]. This photovoltaic device displays enhanced current in
sub-bandgap wavelength range. It is also reported that TTA-based upconversion materials can
be utilized in organic intermediate band solar cells[119]. Sub-bandgap photons can be con-
verted to charge carriers with the help of TTA-based upconversion materials and thus results
in enhanced photocurrent. Besides, there are reports of the photocatalytic applications of TTA-
based upconversion. Kwon et al. utilized TTA-based upconversion materials in CdS-decorated
SiO2 nanocapsules and sub-bandgap photocatalytic oxidation can be observed[120]. Recently,
Hagstrom et al. reported the utilization of two different TTA-based upconversion materi-
als in layered photocatalytic system and achieved red-to-blue and green-to-blue upconversion
simultaneously[121]. As a promising way of enabling below-bandgap solar-to-fuel conversion,
upconversion induced solar water splitting will be reviewed in detail in the next section.
2.3.2 Upconversion Induced Water Splitting
The motivation of employing upconversion materials in solar water splitting systems is to har-
vest the below-bandgap photons. The most common photoactive semiconductors for solar
water splitting such as WO3, TiO2 and Fe2O3, all have difficulties in harvesting photons with
wavelengths longer than 500 nm, limiting their coverage of the solar spectrum. By utilizing
an appropriate upconversion material with emission wavelength in violet-blue range, the solar
water splitting system can efficiently extends its spectral range of light harvesting. The ap-
plications of both lanthanide-based upconversion and TTA-based upconversion in solar water
splitting systems have been reported.
Zhang et al. reported a water splitting system utilizing NaYF4:Yb,Er nanocrystals coupled
with hematite photoanodes[79]. Those rare-earth nanocrystals have a excitation wavelength at
980 nm and emit at 550 nm and 670 nm. Upon illumination with a 980 nm laser, a photocurrent
can be observed in nA cm−2 range. Lanthanide-based upconversion materials requires excita-
tion light with high intensity so that a laser source with sufficient power is necessary. Moreover,
lanthanide-based upconversion materials usually have excitation wavelength in the NIR region
while the TTA-based upconversion materials can have their excitation wavelength in the vis-
ible region. Since the solar spectrum has its peak located at the UV-Vis region, it is more
ideal to employ TTA-based upconversion materials for solar water splitting. Ye et al. utilized
TTA-based upconversion in CdZnS photoanodes for photocatalytic water splitting[122]. Pho-
tocurrent of 90 nA cm−2 can be achieved under the illumination of 532 nm laser. These reports
have demonstrated the feasibility of employing upconversion materials in solar water splitting
systems. The quantum yield of the upconversion process plays an important role which de-
termines the enhancement factor from the upconversion materials. One of the best ways to
improve the upconversion efficiency is to utilize plasmonic enhancement as it can enhance both
the solar water splitting process and the upconversion process.
2.3.3 Plasmonic Enhanced Upconversion
As mentioned before, the upconversion materials are usually very inefficient for viable applica-
tions in solar water spitting systems. Upconversion efficiency can be significantly improved
by the plasmonic nanostructures in two distinct ways: enhanced excitation and enhanced
emission[123]. When the plasmonic resonance has a spectral overlap with the upconversion
material at its excitation wavelength, the excitation enhancement can be achieved by concen-
trating the incident field and thus increases the incident illumination flux. On the other hand,
if the plasmonic resonance wavelength matches the emission wavelength of the upconversion
material, the radiative decay rate can be improved, which results in enhanced emission. In
this case, however, the plasmonic enhanced field needs to have a precisely tailored spatial and
spectral distribution to avoid increasing the nonradiative decay rate which would quench the
emission[124].
Plasmonic enhancement has been widely studied for both lanthanide-based upconversion as
well as TTA-based upconversion. A large variety of plasmonic nanostructures are proven to be
feasible to work in upconversion systems with both excitation enhancement as well as emission
enhancement being achieved. Feng et al. reported coupling silver nanowires with NaYF4:Yb,Er
upconversion nanoparticles[125]. The nanowires have the plasmonic resonance wavelength at
420 nm which overlaps with the emission spectrum of the upconversion nanoparticles at 540 nm.
The enhancement factor is measured to be 2.3. Paudel et al. used gold pillars with plasmonic
resonance at 998 nm to enhance the excitation of NaYF4:Yb,Er at 980 nm and the enhancement
factor is measured to be 2.2[126]. As for TTA-based systems, Poorkazem et al. implemented
silver nanoplates in the PdOEP/DPA/PMMA system. The silver nanoplates have a broadband
LSPR peak centered at 560 nm which can effectively enhance both the excitation (532 nm) and
emission (430 nm) of the upconversion material. The upconversion emission intensity has a
8.5-fold improvement after applying the plasmonic nanoparticles[127]. Some typical plasmonic
enhanced upconversion systems are listed in Table 2.4.
Table 2.4: Selected literature reports of plasmonic enhanced upconversion systems

Plasmonic Enhancement
Upconversion Material Plasmonic Material
Resonance(nm) Factor
NaYF4:Yb,Er[125] Ag nanowires 420 2.3
NaYF4:Gd,Yb,Er[128] Ag islands 430 3.3
Er-doped PbO–GeO2[129] Ag spheres 470 2.0
Er-doped Al2O3[130] Ag islands 560 105
Er-doped BGO[131] Ag spheres 600 10.1
NaYF4:Yb,Er[132] Au nanoholes 930 32.6
NaYF4:Yb,Er[133] Au pillars 998 2.2
*RB–DPBF[134] Au nanoparticles 520 1.7
*PdOEP/DPA/PMMA[127] Ag nanoplates 560 8.5
*PtOEP/DPA[135] Ag-SiO2 nanoparticles 450 4.6
*PdOEP/DPA[136] Au nanoparticles 525 3.1
*: TTA-based upconversion
Chapter 3
Instrumentation and Characterisation

Methods
49
3.1. Scanning Electron Microscopy (SEM) 50
3.1 Scanning Electron Microscopy (SEM)
SEM utilizes focused electron beam controlled by scan coils to provided detailed information
about the specimen surface. Various signals including X-rays, backscattered electrons and
secondary electrons can be produced during the interaction between the electrons and the
specimen. Secondary electrons are highly localized signals with high space resolution which are
generated at the point of impact on the sample surface. Back-scattered electrons are originated
from elastic scattering and have higher energy compared to secondary electrons. Signals from
back-scattered electrons can be utilized to obtain information from deeper locations and can
also be used for elemental analysis. These signals are collected by the corresponding detectors
and processed by computer to generate images reflecting the morphology and the composition
of the specimen surface[137].
In this project, SEM was mostly used to characterize the morphology of the fabricated nanos-
tructures, for instance, particle size. To prepare SEM samples of the fabricated plasmonic
nanostructures and hematite photoanodes, the FTO substrate was fixed onto the SEM sample
holder using double-sided tape. Subsequently, the conductive side of the FTO substrate was
connected to the sample holder by silver paste to avoid sample charging problem during the
imaging process. All the SEM images presented in this thesis were obtained using the Zeiss
Leo Gemini 1525 SEM unless otherwise specified.
3.2 Transmission Electron Microscopy (TEM)
In TEM, a high energy beam of electrons is generated and shone through a very thin sample.
The microscopic features of the sample such as crystal structure and grain boundaries can be
revealed by the interaction between the electrons and the specimen atoms. A camera is often
used to capture the image or diffraction patterns produced by the TEM for further analysis[138].
Compared to SEM, the electron beam in TEM has much higher energy (60 kV-300 kV) and can
potentially provide a better resolution.
3.3. Energy-dispersive X-Ray Analysis (EDX) 51
In this project, TEM was used to get the finest details about the morphology of the nanoparti-
cles. To prepare samples for TEM, the nanoparticles were firstly dispersed in water, followed by
the drop-coating on copper meshes. The copper meshes were then attached to the TEM sample
holder and transferred to the vacuum chamber for imaging. the All TEM images presented here
were obtained using the JEOL 2100F TEM unless otherwise specified.
3.3 Energy-dispersive X-Ray Analysis (EDX)
EDX is an X-ray technique usually attached to a TEM or SEM for identifying the elemental
composition of materials. The EDX system mainly consists of an excitation source and an
X-ray detector. In both TEM and SEM, the high-energy electron beam can be used as the
excitation source while an X-ray detector is attached to the vacuum chamber to collect the
generated X-ray signals. The peaks in the EDX spectra are originated from the X-rays given
off as excited electrons move back to the lower energy level. Different elements have distinct
peak positions and thus the peaks in the EDX spectra represent the elements in the analyzed
sample. When coupled with SEM or scanning TEM, EDX can also provide the elemental
mapping which reveals the spatial distribution of the elements in the specimen[139]. In this
project, all the EDX data were obtained using the EDX equipment in the Zeiss Leo Gemini
1525 system.
3.4 X-Ray Powder Diffraction (XRD)
XRD is a well-established X-ray technique utilizing Bragg’s law to discover the crystalline
structures of the specimen. According to Bragg’s law, the relation of X-ray scattering angle
and the lattice spacing is
nλ = 2d sin θ (3.1)
where n is the order of reflection; λ is the X-ray wavelength; d is the lattice spacing; θ is the
X-ray scattering angle.
3.5. X-Ray Photoelectron Spectroscopy (XPS) 52
The diffractometer illuminate the sample with X-ray photons generated from a cathode ray
tube and collect the scattered photons. The X-ray photons can be scattered by the atoms
in the periodic lattice in all possible diffraction directions when the sample is scanned with a
wide range of incident angle, resulting in the XRD pattern. Various information can thus be
extracted from the pattern including the lattice spacing and the unit cell structure[140]. By
comparing the XRD pattern of a unknown specimen to the patterns in a standard database,
its composition can be revealed. In this project, XRD was used as a non-destructive method
to find the crystallinity and the crystalline phase of the semiconductor samples.
3.5 X-Ray Photoelectron Spectroscopy (XPS)
XPS is an X-ray based surface analysis technique. Due to the photoelectric effect, the photo-
electrons are emitted from the specimen surface when illuminated by monochromatic X-rays.
The binding energy of these emitted photoelectrons can be determined by
Eb = Ep − (Ek + φ) (3.2)
where Eb is the binding energy; Ep is the energy of the X-ray photon; Ek is the electron kinetic
energy; φ is the surface working function.
By collecting the emitted photoelectrons and measuring their kinetic energy, the count of
detect electrons at each binding energy can thus be plotted as the XPS spectrum. The binding
energy is determined by the atomic orbital and the surrounding environment of the atom, so
not only the elemental composition but also the chemical state of the studied surface can be
measured[141]. In this project, XPS was used to study the quantitative and chemical state
information of the heterojunction between two semiconductors. The site survey spectrum was
firstly measured to find the elements on the sample surface and subsequently core level spectra
were measured for each interested element to confirm their chemical states.
3.6. Photoluminescence (PL) Spectrometer 53
Figure 3.1: Schematic diagram of the PL spectrometer. During the measurement, the sample
is excited by the monochromatic light while the emitted photons are captured by the photon
detector after passing through another monochromator. The red arrows represent the optical
path of the excitation light while the blue arrows represent the optical path of the emission
light.
3.6 Photoluminescence (PL) Spectrometer
As illustrated in Figure 3.1, the PL spectrometer mainly consists of two monochromators, a light
source and a photomultiplier tube as the photon detector. The PL spectrometer uses a xenon
lamp and a monochromator to produce monochromatic incident light. Photo-excitation occurs
when the sample is illuminated by the incident light and thus generates emission photons.
The emission light will pass through another monochromator and then be collected by the
photomultiplier tube for measuring the intensity at each wavelength. PL excitation spectrum
and emission spectrum can thus be obtained by fixing the emission wavelength or the excitation
wavelength, respectively. In this project, the Horiba Fluorolog-3 PL spectrometer was used to
obtain the emission and excitation spectra of the upconversion nanoparticles.
3.7. Ultraviolet-visible-near-infrared (UV-Vis-NIR) Spectrometer 54
Figure 3.2: Schematic diagram of the UV-Vis-NIR spectrometer. The photon detector can
capture the transmitted light and thus measure the transmission intensity.
3.7 Ultraviolet-visible-near-infrared (UV-Vis-NIR) Spec-
trometer
The UV-Vis-NIR spectroscopy is the technique that can be utilized to measure the transmission,
reflection and extinction spectra. As illustrated in Figure 3.2, the UV-Vis-NIR spectrometer
mainly consists of a light source, a monochromator and a photon detector. When the monochro-
matic incident light travels through the sample, a portion of the light will be scattered, reflected
or absorbed by the sample. The transmission spectra can be obtained using the light intensity
measured by the detector compared to the reference transmission intensity without the sample.
In this project, the Cary 5000 UV-Vis-NIR spectrometer was used to obtain the extinction
spectra of semiconductors to estimate its band gap. Besides, it was also applied to plasmonic
nanostructures to identify their plasmonic resonance wavelength. The extinction refers to the
sum of reflection, scattering and absorption. It can be obtained by 1 − T where T is the
transmission intensity.
3.8. Photoelectrochemical (PEC) Photocurrent Measurements 55
3.8 Photoelectrochemical (PEC) Photocurrent Measure-
ments
The PEC photocurrent is measured by a potentiostat and a closed circuit of a working electrode,
a reference electrode and a counter electrode as illustrated in Figure 3.3. When the working
electrode is a photoanode, the counter electrode will act as a cathode and vice versa. All three
electrodes are immersed in the electrolyte contained in a PEC cell. The potentiostat provides
a bias voltage between the working electrode and the reference electrode and measures the
photocurrent generated from the working electrode when it is illuminated by solar light. In
practice, a solar simulator is used as the light source instead since it can provide stable power
and it is easier to operate indoors. The solar simulator usually use a short-arc xenon lamp
together with an AM 1.5G filter to simulate the real solar spectrum. The total light intensity
must be 100 mW cm−2 at the working distance. During a measurement, the potentiostat will
measure the current at different bias voltages. The collected current signal not only contains
the photocurrent, but also contains the dark current which refers to the current contributed by
the bias voltage itself instead of the light illumination. The dark current can be measured by
running a reference measurement without the incident light[4]. A PEC photocurrent curve can
then be plotted with photocurrent density versus bias voltage. The photocurrent density can
be calculated by:
i = (I − Id )/A (3.3)
where I is the current with illumination; Id is the current without illumination; A is the
illuminated area of the working electrode.
In this project, Ag/AgCl electrodes were used as reference electrodes since they are reliable
and cost-efficient. To convert a voltage versus Ag/AgCl to a voltage versus RHE, the following
equation is applied:
0
ERHE = EAg/AgCl + 0.059pH + EAg/AgCl (3.4)
0
where EAg/AgCl is the standard electrode potential of Ag/AgCl and pH is the pH value of the
electrolyte.
3.9. Incident Photon-to-Current Efficiency (IPCE) Measurements 56
Figure 3.3: Schematic diagram of the PEC photocurrent measurement circuit. Working elec-
trode, reference electrode and counter electrode are placed in a PEC cell and fully immersed in
electrolyte. All three electrodes are connected to the potentiostat and the working electrode is
illuminated by solar light from a solar simulator.
To further standardize the PEC photocurrent measurement process, a self-designed 3D-printable

cell is made as shown in Figure 3.4. The working electrode can be placed between the cell body
and the cover plate, together with an o-ring to precisely control the illuminated area. In this
project, all PEC measurements were carried out using the Autolab potentiostat, the Newport
LCS-100 solar simulator, a Pt mesh as the counter electrode and a Ag/AgCl reference electrode
together with this type of PEC cell unless otherwise specified.
3.9 Incident Photon-to-Current Efficiency (IPCE) Mea-
surements
IPCE is defined as the count of photogenerated electron per incident photon. As shown in
Figure 3.5, IPCE measurement utilizes a similar setup as in the PEC measurement. Since in
IPCE measurement the photon-to-current efficiency at each wavelength needs to be measured,
Figure 3.4: Schematic illustration of a self-designed PEC cell ready for 3D printing. An o-ring
and the working electrode will be inserted between the cell body and the cover plate before
measurements.
a monochromatic light is used instead of the simulated solar light. The monochromatic light,
generated by a xenon lamp and a monochromator, has a much lower power intensity compared
to the solar light. Hence, the resulting photocurrent is in µA or even nA range. It is impossible
for a normal potentiostat to pick up such a small signal out from the total current (photocurrent
+ dark current). A light chopper and a lock-in amplifier are employed to solve this problem.
The lock-in amplifier can extract the signal from a extremely noisy input with a known carrier
wave. The chopper will output its driving sine wave to the lock-in amplifier as the reference
wave and the lock-in amplifier will extract any signal with the same frequency from the current
readings collected by the potentiostat. Since the dark current is not affected by the chopped
light, only photocurrent will be extracted by the lock-in amplifier. The IPCE as a function of
incident wavelength can then be calculated using the following equation[4]:
I(A/cm2 ) 1240
IP CE(λ) = 2
· · 100% (3.5)
P (W/cm ) λ(nm)
where I is the photocurrent extracted by the lock-in amplifier; P is the incident light intensity
which can be measured by a light power meter; λ is the incident wavelength.
In this project, the IPCE measurements were carried out using the Autolab potentiostat, the
Figure 3.5: Schematic diagram of the IPCE measurement setup. The PEC cell is the same as
the one used for PEC photocurrent measurement. Chopped monochromatic light is used as the
incident light. The lock-in amplifier can extract weak photocurrent signals from the current
readings of potentiostat using the chopper’s sine wave as a reference signal.
SRS SR830 DSP lock-in amplifier, the Newport monochromator, a 100 W xenon lamp and the
same cell setup as in the PEC photocurrent measurements.
Chapter 4
FDTD Modelling of Plasmonic

Nanoarrays
59
4.1. Introduction 60
4.1 Introduction
Plasmonic nanoarrays are 2D periodic plasmonic nanostructures that can be fabricated through
various approaches, such as e-beam lithography[142] and electrodeposition[143] with the help
of a periodic template or mask. Plasmonic nanoarrays with divergent morphology shows signif-
icantly different plasmonic properties and thus attracted the attention of researchers in many
fields. In this project, spherical polystyrene template based lithography was used in the fabri-
cation of plasmonic nanoarrays. Three different plasmonic nanostructures including nanodisk
array (NDA), nanotriangle array (NTA) and nanohole array (NHA), can be fabricated through
similar procedures. Figure 4.1 shows the SEM images of all the three common plasmonic
nanoarrays based on the spherical polystyrene templates. Nonetheless, no systematical study
has been done covering all of these nanostructures. In this chapter, FDTD method was used to
model these nanoarrays with different geometrical parameters and plasmonic metals. The ex-
tinction spectra of the modelled nanostructures as well as their field mapping can be obtained
by FDTD modelling. The extinction intensity is the sum of absorption intensity, scattering
intensity and reflection intensity. The features in the extinction spectra can be utilized to find
the spectral position of the plasmonic effect. On the other hand, the field mapping of the
nanostructure reveals the spatial distribution of the enhanced field.
As illustrated in Figure 4.2 these nanodisk arrays are periodic plasmonic nanostructures with
2D hexagonal lattice originated from the self-assembly of the polystyrene spheres. For NDA,
the polystyrene spheres are packed on the top of a plasmonic film and ion etching was used to
shrink the spheres. Ar milling are then used to remove any material not cover by the spheres,
resulting in standalone nanodisks. During the fabrication of NHA, the polystyrene spheres are
instead packed on the top of the glass or FTO substrate. The following e-beam evaporation
or sputtering fills the gaps with plasmonic materials, resulting in a continuous film of the
plasmonic metal with many nanoholes. The process of fabricating NTA is almost identical to
that of fabricating NHA, but the difference is that the polystyrene spheres are closely packed
thus no continuous plasmonic film forms.
To systematically study the plasmonic properties of these nanoarrays, four plasmonic metals,
Figure 4.1: SEM images of experimentally fabricated (a) NDA, (b) NTA and (c) NHA. The
scale bars in (a)(c) are 1000 nm. The scale bar in (b) is 500 nm.
4.2. Methods 62
Figure 4.2: Top view of modelled plasmonic nanoarrays: (a) NDA, (b) NHA, (c) NTA. The
orange rectangles represent the 2D unit cells.
namely Ag, Al, Au, Cu are selected. Single-layer nanoarrays with one metal as well as dual-layer
nanoarrays based on two of these metals combined were investigated. The modelled geometrical
parameters include the period, the diameter and the thickness. The period represents the
center-to-center distance of the neighboring holes/disks that can be experimentally controlled
by the diameter of the polystyrene spheres. The modelled range of period is from 200 nm
to 900 nm. The diameter is the actual size of each disk/hole, which can be experimentally
controlled by etching the self-assembled polystyrene spheres. The thickness is the vertical
length of the plasmonic metal layer, which can be experimentally controlled by the time of
metal deposition. The modelled range of thickness is from 40 nm to 100 nm. Unlike single
nanoparticles, the nanodisks and nanotriangles have small spacing between each other which
makes the near-field electrostatic coupling[144] between them significant. Therefore, various
spacing can result in spectral shift of the LSPR peak.
4.2 Methods
Software. All the FDTD modelling in this project were carried out using an open-source
FDTD software package: MEEP[145]. As a popular FDTD simulation software, MEEP has
implemented many advanced algorithms including pervasive interpolation, supporting subpixel
features and symmetry optimizations, which not only increase the accuracy of the model but
4.2. Methods 63
Table 4.1: Adopted values for the Drude-Lorentz models used in FDTD modelling[1]
Ag (ωp = 9.01eV)
j 1 2 3 4 5 6
f 0.845 0.065 0.124 0.011 0.840 5.646
ω0 (eV) 0.000 0.816 4.481 8.185 9.083 20.29
Γ(eV) 0.048 3.886 0.452 0.065 0.916 2.419
Au (ωp = 9.03eV)
j 1 2 3 4 5 6
f 0.760 0.024 0.010 0.071 0.601 4.384
ω0 (eV) 0.000 0.415 0.830 2.969 4.304 13.32
Γ(eV) 0.053 0.241 0.345 0.870 2.494 2.214
Al (ωp = 14.98eV)
j 1 2 3 4 5
f 0.523 0.227 0.050 0.166 0.030
ω0 (eV) 0.000 0.162 1.544 1.808 3.473
Γ(eV) 0.047 0.333 0.312 1.351 3.382
Cu (ωp = 10.83eV)
j 1 2 3 4 5
f 0.575 0.061 0.104 0.723 0.638
ω0 (eV) 0.000 0.291 2.957 5.300 11.18
Γ(eV) 0.030 0.378 1.056 3.213 4.305
also speed up the simulation to allow high-throughput modelling. The models were run on the
cluster in imperial high-performance computing facilities.
Definition of materials. The permittivity of the plasmonic metals involved in this project is
defined by the Drude-Lorentz model:
n
f1 ωp2 X fj ωp2
(ω) = 1 − + 2
(4.1)
ω(ω − iΓ1 ) j=2 (ω0,j − ω 2 ) − iωΓj
where ω0,j is the resonant frequency of each oscillator; fj is the strength of each oscillator; Γj
is the reciprocal of each oscillator’s lifetime; ωp is the plasma frequency. The adopted values[1]
of these parameters are listed in Table 4.1.
Modelling space. As illustrated in Figure 4.3, the FDTD modelling space for the plasmonic
nanoarrays mainly consists of three components: the boundary conditions, the plasmonic struc-
4.3. Results and Discussion 64
Figure 4.3: Schematic diagram of a FDTD modelling space for plasmonic nanoarrays (not to
scale). The modelling space has PML boundaries in the +Z/-Z directions (red) and periodic
boundaries in other directions (green). The plane wave source (white) is above the plane of
plasmonic nanoarrays with the incident direction towards the plasmonic structure.
tures and the incident source. In the +Z/-Z directions, PML layers are used as boundary
condition as it can absorb the excessive wave and thus simulate the free space. Meanwhile in
other directions, periodic boundaries are employed since all the plasmonic arrays studied in
this project are periodic in the X-Y plane. Periodic boundary condition can help accurately
simulating such structures in the small modelling space since only one unit cell is necessary to
be modelled. The modelled unit cells for each type of nanoarray are shown in Figure 4.2. The
incident plane wave is set above the plasmonic structure with the incident direction towards
the plasmonic structure to simulate the solar light.
4.3 Results and Discussion
4.3.1 Nanodisk Arrays
The role of diameter of NDA was firstly studied. With thickness fixed at 60 nm and period fixed
at 500 nm, the simulated extinction spectra are presented in Figure 4.4. It is shown that when
Figure 4.4: FDTD simulated extinction spectra of Ag, Au, Al and Cu NDA with different
diameters. The thickness is fixed at 60 nm while the period is fixed at 500 nm. The LSPR peak
becomes less significant as the diameter gets closer to the period.
the diameter is closed to period, the extinction curve flattens with less significant extinction
peak which indicates the wavelength position of LSPR. Also, as the diameter increases, the
extinction peak is red-shifted. Similar to spherical plasmonic nanoparticles, both of these
phenomena are due to the increasing radius of curvature[146]. Smaller nanoparticles usually
lead to shorter LSPR wavelength and stronger LSPR effect[147]. However, as the field mapping
in Figure 4.5 suggests, when the spacing between the nanodisks gets too large, the near-field
electrostatic coupling[144] disappears and thus the fields between adjacent nanodisks are no
longer enhanced. In this case, the nanodisks act like standalone plasmonic nanoparticles and
show weaker LSPR effect compared to the nanodisks in the periodic plasmonic arrays. When
considering enhanced field intensity as the first priority, the optimized diameter of nanodisks
is 40% to 60% of the period.
The influence of varying the thickness is also studied and the resulting extinction spectra are
shown in Figure 4.6. The height and the spectral position of the LSPR peaks are not affected
much by the thickness of NDA so it is not feasible to manipulate the LSPR by tailoring the
Figure 4.5: FDTD simulated normalized E-field mapping at LSPR wavelength of Au NDA with
diameter of left: 90 nm, right: 400 nm. Both of them have 60 nm thickness and 500 nm period.
Smaller diameter leads to stronger local enhanced field while larger diameter results in larger
enhanced area.
thickness. However, as the thickness increases, the nanodisks will eventually become nanorods
and may have very different plasmonic properties which is is beyond the scope of this study.
The influence of varying the period is investigated by fixing the thickness to 60 nm and fixing
the diameter to 80% of the period. The resulting extinction spectra is shown in Figure 4.7. As
the period increases, the wavelength position red shifts significantly due to the increasing radius
of curvature. Hence, tailoring the period is a viable method to tune the wavelength position
of the LSPR effect. Cu NDA and Au NDA have similar LSPR wavelengths, longer than that
of Ag NDA and Al NDA with the same period. For example, with 500 nm period, the LSPR
peaks of Au NDA and Cu NDA are at around 650 nm while the LSPR peak of Ag NDA is at
550 nm and the LSPR peak of Al NDA is at 480 nm. With longer LSPR wavelength, Au NDA
and Cu NDA are preferred for applications in the NIR region. On the other hand, for Al and
Ag NDA, the LSPR wavelength can be positioned shorter than 400 nm for applications in the
UV-visible region. Since most solar energy are provided in the UV-visible region, Ag NDA and
Al NDA are more feasible nanostructures for plasmonic enhanced solar water splitting.
The LSPR effect of NDA could be further manipulated by depositing an additional layer of
plasmonic metal to obtain dual-layer NDA. The simulated extinction spectra of dual-layer NDA
thickness. The diameter is fixed at 400 nm while the period is fixed at 500 nm. Different
thickness results in LSPR peaks at similar wavelength position.
period. The thickness is fixed at 60 nm while the diameter is fixed at 80% of the period.
Increasing the period results in significant red shift of the LSPR peak.
with various period are shown in Figure 4.8. As discussed before, the LSPR wavelengths of Au
NDA and Cu NDA are usually longer than that of Al NDA and Ag NDA. If an layer of Ag or
Al is deposited to Au NDA or Cu NDA, an additional LSPR peak may be seen at a shorter
wavelength. For instance, 200 nm Ag/Au dual-layer NDA has two LSPR peaks located at
400 nm and 600 nm. Moreover, broadband LSPR effect can be observed as illustrated in Figure
4.9. At the 390 nm extinction peak, LSPR enhanced field can be observed in the top layer. At
the 630 nm extinction peak, the field near the bottom layer is enhanced. At the spectral region
between the these two peaks, both layers show LSPR effect. Dual-layer NDA thus provides a
promising way to extend the spectral range of LSPR effect. It is useful in distinct systems such
as upconversion materials since both emission and excitation enhancement can be achieved on
the same plasmonic nanostructure. Furthermore, in solar water splitting systems, pristine Al
and Ag suffer from chemical stability issues in the electrolyte. The stable Au layer in dual-layer
NDA can act as a protective layer to improve the durability of the photoanodes.
Figure 4.8: FDTD simulated extinction spectra of dual-layer NDA with various period. The
row indicates the metal in the bottom layer while the column indicates the metal in the top
layer. The thickness is fixed at 30 nm for each layer while the diameter is fixed at 80% of the
period.
Figure 4.9: Illustration of broadband LSPR effect in dual-layer NDA with distinct LSPR wave-
length of each layer. In the spectral range between the two extinction peaks, enhanced field
exists at both bottom and top layer.
Figure 4.10: FDTD simulated normalized E-field mapping at LSPR wavelength of Au NTA
with period of left: 400 nm, right: 900 nm, both of which have 60 nm thickness. The LSPR
enhanced area is between the tips of the neighboring triangles.
4.3.2 Nanotriangle Arrays
Although both exhibit LSPR effect, the field distribution of NTA is quite different from that
of NDA. As can be seen in Figure 4.10, the LSPR enhanced area in NTA is mostly confined to
the region between the neighboring tips. With extremely low radius of curvature around the
sharp tips, it shows much higher field intensity compared to the area near the vertices. The
limited enhance area in NTA makes it less ideal for plasmonic enhanced water spitting where
large enhanced area is necessary to achieve high overall efficiency. Similar to NDA, the LSPR
peak of NTA is less affected by thickness, as shown in Figure 4.11. As the thickness increases,
the extinction peak of NTA only shifts to a slightly shorter wavelength. It is thus not feasible
to tune the wavelength position of LSPR for NTA by tailoring its thickness.
Nonetheless, the LSPR peak of NTA is sensitive to its period as shown in Figure 4.12. A drastic
red shift can be expected as the period increases, showing the identical trend as NDA, which
is possibly due to the increasing size of the triangles. For Au NTA and Cu NTA with the
same period, they have similar spectral position of LSPR peak with the tunable range starting
from 700 nm. For Ag NTA, the tunable range starts from 550 nm while Al NTAs are able to
Figure 4.11: FDTD simulated extinction spectra of Ag, Au, Al and Cu NTA with different
thickness. The diameter is fixed at 400 nm while the period is fixed at 500 nm. The LSPR
peaks slightly blue-shift as the thickness increases.
have LSPR peak below 400 nm. In this matter, Al NTA and Ag NTA are more suitable for
applications in the UV-visible region. Furthermore, the LSPR peak position is less material
dependent as the period increases. For instance, the spectral distance between LSPR peaks of
Ag NTA and Au NTA is 150 nm for models with 200 nm period while the difference is less than
100 nm for models with 500 nm period.
For dual-layer NTA as shown in Figure 4.13, two extinction peaks can be observed when the
bottom layer and the top layer has very different LSPR wavelength. For example, the 200 nm
Au/Al dual-layer NTA has one LSPR peak at 400 nm whose position is identical to that of Al
NTA, and another peak at 700 nm which is from the Au layer. In this case, the LSPR peaks
from each layer keeps its original position and no shift happens. On the contrary, the LSPR
peaks from each layer can merge into one if their wavelength position is close. For instance,
the 400 nm Au/Ag NTA only has one LSPR peak at 770 nm, which locates in the middle of
the 900 nm LSPR peak from the Au layer and the 700 nm LSPR peak from the Ag layer. This
phenomenon originates from the near-field electrostatic coupling of the LSPR from different
Figure 4.12: FDTD simulated extinction spectra of Ag, Au, Al and Cu NTA with different
period. The thickness is fixed at 60 nm. Increasing period results in significant red shift of the
LSPR peak.
layer. It can be employed to tune the spectral position of the LSPR effect more precisely.
Figure 4.13: FDTD simulated extinction spectra of dual-layer NTA with various period. The
layer. The thickness is fixed at 30 nm for each layer.
Figure 4.14: FDTD simulated normalized E-field mapping at (a), (b): extinction peak and (c),
(d): extinction dip. (a) and (c) represent top view while (b) and (d) represent cross-sectional
view. The modelled NHA is Au NHA with period of 500 nm, diameter of 400 nm and thickness
of 60 nm. The curve in the middle is its simulated extinction spectrum. The hot zones in field
mapping are inside the holes at extinction dip while above the continuous metal at extinction
peak.
4.3.3 Nanohole Arrays
In plasmonic NHA, the continuous periodic film shows significant SPP effect and extraordinary
transmission (EOT) can be observed at certain wavelength due to SPP, which can be indicated
by the extinction dip[148]. The E-field distribution is also different from those LSPR-dominant
nanoarrays as illustrated in Figure 4.14. At EOT wavelength (the extinction minimum), the
enhanced field is inside the holes, resulting in the the increased transmission intensity. On the
contrary, the field enhancement is above the film surface at the extinction peak. The oscillators
of single holes couple through SPP over the periodic lattice and lead to field enhancement in
large area[149].
As can be seen in Figure 4.15, the diameter of the holes affects the extinction of NHA con-
Figure 4.15: FDTD simulated extinction spectra of Ag, Au, Al and Cu NHA with different
diameters. The thickness is fixed at 60 nm while the period is fixed at 500 nm. The extinction
intensity decreases as the diameter decreases.
siderably. Increasing the diameter reduces the area of the metal film and thus results in lower
extinction intensity. Besides, if the diameter is close to the period or if the nanoholes are too
small, the EOT effect becomes less significant. On the other hand, the spectral position of
EOT shows weak relationship with the diameter. Therefore, it is not practical to tune the
EOT wavelength by varying the diameter of NHA. It is also not feasible to tune the EOT
wavelength by varying the thickness since the EOT wavelengths are identical for NHAs with
different thickness as shown in Figure 4.16.
As can be seen in Figure 4.17, the extinction spectrum of NHA red shifts as the period increases.
Theoretically, the resonance wavelength of SPP generated from grating coupling is related to
the grating wave vector kg = 2π/λg , where λg is the period of the grating[150]. So the period of
NHA affects its SPP wavelength remarkably. The EOT wavelength is also related to the type of
the plasmonic metal. For Au NHA and Cu NHA, the tunable range of EOT wavelength starts
from around 600 nm. On the other hand, Ag NHA and Al NHA are able to exhibit EOT at
much shorter wavelengths. With period higher than 400 nm, the EOT wavelength of Al NHA
thickness. The diameter is fixed at 400 nm while the period is fixed at 500 nm. Varying the
thickness has negligible influence on the extinction spectrum.
is about 50 nm shorter than that of Ag NHA and 100 nm shorter than that of Au NHA and Cu
NHA.
Since the plasmonic film will be fully covered by semiconducting materials in photoanodes,
the enhanced field inside the holes (at extinction dip) as well as the enhanced field above the
metal film (at extinction peak) can both potentially contribute to the solar water splitting
performance. Given the same period, the wavelength of extinction peak of NHA is shorter
compared to that of NDA and NTA, which is favorable for above-bandgap enhancement in
solar water splitting systems. For example, Ag NHA with 400 nm period has the extinction
peak at 390 nm while Ag NDA with the same period has LSPR peak at 600 nm and Ag NTA
has LSPR peak at 550 nm. Most semiconductors used for solar water splitting have absorption
peak below 500 nm. Hence, employing plasmonic NHAs in these systems can enable the PIRET,
which only works above band gap and has the best efficiency among all three mechanisms of
plasmonic enhanced water splitting. On the contrary, NHAs are less feasible for applications
in the infrared region since it is unable to obtain NHAs with EOT wavelength above 1000 nm
using polystyrene templates <1 µm. Besides, due to the interband absorption of Au below
Figure 4.17: FDTD simulated extinction spectra of Ag, Au, Al and Cu NHA with different
period. The thickness is fixed at 60 nm while the diameter is fixed at 80% of the period.
Increasing the period results in significant red shift of the SPP wavelength.
500 nm, Au NHA cannot enable plasmonic enhancement in this wavelength range.
Dual-layer NHAs are also modelled and the resulting extinction spectra are shown in Figure
4.18. Broadband EOT effect can be observed if the EOT wavelength of the bottom layer differs
from that of the top layer. In this case, a wider extinction dip can be seen. For example,
the 400 nm Al/Au dual-layer NHA shows a wide extinction dip from 400 nm to 600 nm. In
order to obtain the broadband EOT effect, the period between 200 nm to 500 nm is optimal
because the EOT wavelength of each layer becomes identical for longer period. Compared to
single-layer NHAs, dual-layer NHAs have field enhancement in a larger spectral region which is
ideal for plasmonic enhanced solar energy harvesting. Similar to dual-layer NDA, the Au layer
in dual-layer NHA can also be employed as a protective layer to improve the chemical stability
of the plasmonic nanostructure.
Figure 4.18: FDTD simulated extinction spectra of dual-layer NHA with various period. The
layer. The thickness is fixed at 30 nm for each layer while the diameter is fixed at 80% of the
period.
4.4. Conclusion 80
4.4 Conclusion
In this chapter, FDTD models of three different plasmonic nanoarrays are systematically stud-
ied. Distinct field enhancement can be observed for different nanoarrays. NDAs are LSPR
dominant and show field enhancement at LSPR wavelength (extinction peak) with the en-
hanced field near the edge of the nanodisks. NTAs are also LSPR dominant but the most
enhanced area is between the tips of the neighboring triangles. NHAs are SPP dominant and
exhibits field enhancement at both the extinction peak with the active area above the con-
tinuous metal film as well as the extinction dip with the active area inside the holes. The
spectral range of the plasmonic resonance is also studied by examining the simulated extinc-
tion spectra. Among all three nanoarrays provided with the same geometrical parameters and
plasmonic metal, NTAs have the longest plasmonic resonance wavelength while NHAs have the
most ideal plasmonic resonance wavelengths for solar water splitting. The wavelength position
of the plasmonic resonance is proven to be manipulative via controlling the period. Increased
period results in red shift of the plasmonic resonance. With short period (<500 nm), the reso-
nance wavelength of NHA is also strongly dependent on the plasmonic metal. Au and Cu are
more suitable for applications in the visible-NIR region while Ag and Al work better in the
UV region. Furthermore, dual-layer nanoarrays are also studied. Broadband field enhancement
can be achieved by combining two plasmonic metal layers with distinct plasmonic resonance
wavelength. The plasmonic resonance wavelengths of all the combinations can be looked up
in Figure 4.8, 4.13 and 4.18, which are used as references to select plasmonic nanoarrays for
plasmonic enhanced water splitting systems in the following chapters. To a greater extent, the
results in this chapter can also be utilized to search for ideal plasmonic nanoarrays that can be
employed in other systems, not limited to the application of plasmonic enhanced solar water
splitting.
Chapter 5
Dual-layer Plasmonic Nanohole Array

Enhanced BiVO4-Hematite
Photoanodes
81
Figure 5.1: Schematic illustration of the BiVO4-coated hematite nanorods photoanode enhanced
by Ag/Au dual-layer nanohole array.
5.1 Introduction
Metal nanoarrays, as promising 2D plasmonic nanostructures, exhibits great compatibility

and high enhancement efficiency in many different systems including biosensors[151], chem-
ical sensors[152], refractive index sensor[153] as well as photoanodes[154]. When employed in
solar water splitting systems, noble metal nanoarrays naturally show better stability and thus
receive more attention. According to the FDTD modelling results in Chapter 4, nanohole ar-
rays exhibit SPP effect and are expected to enable plasmonic enhancement in the UV-visible
region which is ideal for solar light harvesting applications.
In this chapter, a novel plasmonic enhanced photoanode with BiVO4-coated hematite nanorods
on dual-layer Ag/Au nanohole array was fabricated, as illustrated in Figure 5.1. The utilized
5.2. Experimental 83
Figure 5.2: Step-by-step illustration of the colloidal lithography method used to fabricate
nanohole arrays. The steps are (i) coating self-assembled template nanospheres onto substrate,
(ii) using reactive ion etching to reduce the size of the spheres, (iii) coating of the plasmonic
metal layer and (iv) removing the organic residue by duct tape.
hematite nanorods are excellent photoactive semiconductor materials with 1D morphology.

They were thus selected as the base materials in the photoanode. By coating a BiVO4 layer,
the issues of low carrier concentration and high charge recombination rate of pristine hematite
can be mitigated by the formation of the BiVO4-Fe2O3 heterojunction[155][156]. In the exper-
iments described here, a five-fold improvement of photoelectrochemical solar water splitting
performance is achieved by introducing the BiVO4 layer. Moreover, the dual-layer nanohole
arrays can provide enhanced field to further increase the photocurrent by plasmonic enhance-
ment, mainly through the PIRET mechanism. With the plasmonic nanohole array applied, the
solar water splitting performance can be further enhanced by up to two-fold.
5.2 Experimental
Materials. Hydrochloride acid, nitric acid, iron(III) chloride hexahydrate, sodium nitrate, am-
monium metavanadate, bismuth nitrate pentahydrate, sodium dodecylsulphate and anhydrous
citric acid were purchased from Sigma-Aldrich and used as received without further purification.
Polystyrene spheres were purchased from Bangs Laboratories, Inc and used as received. FTO
glass was purchased from Solaronix and washed with acetone, isopropyl alcohol and ethanol
before use.
Fabrication of nanohole arrays. Au nanohole arrays and Ag/Au dual-layer nanohole arrays
were prepared using the colloidal lithography method[157][158]. Briefly, 10% aqueous solution
of monodisperse polystyrene nanoparticles was mixed with an equal volume of ethanol. Clean
silicon wafers and FTO substrates were prepared and treated with UV-ozone for 2 h. 10 µL of
polystyrene solution was deposited onto the onto the silicon wafer by drop-coating to produce a
homogeneous thin layer. After that, the wafer was slowly immersed in a vessel filled with 450 mL
of deionized water and 3 µL of 2% sodium dodecyl sulphate solution, leaving a self-assembled
polystyrene monolayer floating on the surface of the water. The polystyrene monolayer was
then transferred onto the as-prepared FTO substrate and dried in air for 24 h before further use.
Additionally, the polystyrene-covered FTO substrates were treated with reactive ion etching
for 25 s to reduce the diameter of the polystyrene spheres. Next, the substrates were moved
into Mantis evaporation/sputtering system for sputtering to obtain a 40 nm layer of Ag and a
40 nm layer of Au. Subsequently, the polystyrene spheres were removed by duct tapes. The
substrates covered with plasmonic nanohole array were later checked under SEM to reject those
with considerably excessive defects.
Synthesis of BiVO4-coated hematite nanorods. The growth of hematite nanorods on

FTO was carried out with a template-less hydrothermal method[70][159]. In brief, iron(III)
chloride hexahydrate (4.055 g) and sodium nitrate (8.499 g) were dissolved in deionized water
(100 mL). 12 µL of 37% hydrochloride acid was also added to the aqueous solution to set
its pH to 1.5. 15 mL of the solution was then transfered to a 20 mL glass vial and a FTO
substrate (with nanohole array) was immersed in the solution. The vial was kept at 100 ◦C for
4 h. Subsequently, the substrate was taken out, rinsed gently with water and dried in air for
10 h before further use. A muffle oven was used to anneal the substrate at 450 ◦C for 4 h and
at 600 ◦C for 0.5 h. The substrate was cooled to room temperature before the spin-coating of
BiVO4 precursors[160]. To prepare the precursor solution, ammonium metavanadate (0.29 g),
bismuth nitrate pentahydrate (1.21 g), anhydrous citric acid (0.96 g) and 1.65 g of 60% nitric
acid were mixed with 5.85 g of deionized water. The spin-coating was processed firstly at 500
rpm for 10 s then 2000 rpm for 30 s. Finally, the substrate was annealed again at 500 ◦C for 3 h
and cooled to room temperature before further use.
Characterization. The polystyrene templates were checked under TEM (JEOL-2100F) and
SEM (Zeiss LEO 1525). The fabricated plasmonic nanohole arrays, hematite nanorods and
the coated BiVO4 layer were also checked under SEM. EDX and XPS were used for elemental
analysis to confirm the formation of the BiVO4-hematite heterojunction. UV-Vis-NIR spec-
trometer was used to obtain the extinction spectra of the nanohole arrays and the fabricated
photoanodes. PEC photocurrent measurement and IPCE measurement were carried out using
a three-electrode cell with buffer solution (pH=7) as electrolyte to characterize the solar water
splitting performance of the fabricated photoanodes.
5.3.1 Morphology and elemental analysis
The plasmonic nanoarrays are fabricated using the colloidal lithography technique. According
to the FDTD model, the resulting plasmonic nanohole arrays are supposed to have period
smaller than 700 nm to enable above-bandgap plasmonic enhancement for hematite. To achieve
this, 400 nm polystyrene spheres were chosen as templates to be used. The TEM image of these
polystyrene spheres was taken and shown in Figure 5.3. It can be seen that the polystyrene
spheres can self-assemble to form a hexagonal structure. The colloidal lithography process
utilized the self-assembly of these polystyrene spheres to obtain hexagonally packed masks.
The self-assembly process acts as a critical step in the colloidal lithography technique since it
can essentially determine the quality of the final plasmonic nanoarrays. The SEM images of
the polystyrene spheres on the FTO substrate are shown in Figure 5.4. A clear hexagonal 2D
lattice can be seen, suggesting the successful self-assembly of the high-quality nanospheres. The
purpose of the following reactive ion etching process was to shrink the polystyrene spheres so
Figure 5.3: TEM image of polystyrene spheres used as template in the colloidal lithography
process. The diameter of the spheres is measured to be around 400 nm. The scale bar is 500
nm.
that more space can be created between the nanospheres, allowing a continuous metal film to be
deposited onto the substrate. Without the reactive ion etching process, the final product would
be nanotriangle arrays instead of nanohole arrays which are less ideal for large-area plasmonic
enhancement as discussed in the Chapter 4. On the other hand, the polystyrene spheres were
no longer closely packed after the reactive ion etching process, which made them more sensitive
to vibrations generated by handling and transferring the samples and hence defects are possible
to form. As can be seen in the bottom-left corner of Figure 5.4(b), some polystyrene spheres
are no longer in the hexagonal lattice.
After the sputtering of the plasmonic metal and removal of the organic residues, the plasmonic
nanostructure was checked under SEM as shown in Figure 5.5. The plasmonic nanoholes are
in hexagonal 2D lattices. The surface of the substrate can be seen inside the nanoholes. The
center-to-center distance of the nanoholes is measured to be around 400 nm, which is identical
to the diameter of the polystyrene spheres as expected. Both glass and FTO substrates were
originally used and it can be observed that the FTO substrates have rougher surface while the
Figure 5.4: SEM images of hexagonally packed polystyrene spheres on the FTO substrate
(a) before the reactive ion etching process and (b) after the reactive ion etching process. The
etching process shrank the polystyrene spheres and thus created larger spacing between adjacent
spheres. The scale bar is 1000 nm.
Figure 5.5: SEM images of 400 nm plasmonic nanohole arrays (a) on the glass substrate and
(b) on the FTO substrate. The FTO substrate is observed to have higher surface roughness
compared to glass.
glass substrates are much smoother. However, the glass-based plasmonic nanohole arrays were
later deprecated mainly due to two reasons. Firstly, the hydrothermal technique of growing
hematite nanorods is not compatible with glass substrate. Secondly and more importantly, the
glass substrate is not conductive so the overall conductivity of glass-based nanohole arrays are
poor which is not ideal for PEC solar water splitting applications. On the other hand, the
glass-based nanohole arrays can be favorable when utilized in systems such as photocatalytic
water splitting and biosensing where conductivity is not a concern.
The hematite nanorods were grown directly on the plasmonic nanostructure via a template-less
hydrothermal method as described in Section 5.2. The SEM images of the resulting photoanodes
are illustrated in Figure 5.6. Due to the template-less method used, the hematite nanorods are
not vertically aligned as shown in Figure 5.6(a). The disordered alignment of the nanorods is
advantageous since larger surface area of the nanorods are facing the incident light, resulting
in higher photocurrent density. The length of the nanorods are measured to be around 350 nm.
The BiVO4 layer was subsequently coated onto the hematite nanorods via spin-coating, followed
by annealing. As a important step, the annealing process can not only help with the formation
of the hematite and BiVO4 layers but can also contribute to a better contact between the BiVO4
layer and the hematite nanorods.
Upon hydrothermal growth, the nanorods were initially in the form of FeOOH and have poor
PEC performance. Therefore, the annealing process at 450 ◦C is necessary to turn them into
hematite. The color change during this process is illustrated in Figure 5.7. It can be seen
that during the annealing process, the color of the substrate turns to dark brown, indicating
the formation of hematite. It is also reported that by annealing at a much higher temperature
(650 ◦C - 800 ◦C), the water splitting performance of the hematite nanorods can be further
improved[161]. The improved performance from high temperature annealing is mainly due
to the diffusion of Sn4+ ions from the FTO substrate to the hematite nanorods, which will
increase the carrier density and thus mitigate the disadvantage of high charge recombination
rate of pristine hematite[162]. However, the high temperature annealing is avoided in this
project with two reasons taken into consideration. Firstly, the BiVO4 layer in the system can
also lower the charge recombination rate so Sn4+ doping is not necessary. Secondly, the high
temperature annealing process can potentially damage the metal nanoarrays and weaken the
plasmonic effect, which is undesirable for the study of plasmonic enhancement in this system.
The annealing temperature of the hematite nanorods in this project is hence set to 450 ◦C for
4 h and 600 ◦C for 0.5 h without further optimization.
After the spin-coating of the BiVO4 precursors, another annealing process at 500 ◦C is necessary.
This annealing process resulted in a more homogeneous thin layer of BiVO4, to guarantee a
large interface between the BiVO4 layer and the hematite nanorods. As shown in Figure 5.8, the
morphology of the spin-coated BiVO4 layer was like nanoflakes before annealing. The annealing
process later turned it into a homogeneous thin film.
Figure 5.6: SEM images of (a) hematite nanorods, with length measured to be around 350 nm,
(b)(c) BiVO4 coated hematite nanorods. A scratch was made on the surface of sample in (c)
to reveal the hematite nanorods and the FTO surface underneath the BiVO4 layer. The scale
bars are 500 nm.
Figure 5.7: Digital photograph of LEFT: blank FTO, MID: FeOOH nanorods on FTO, RIGHT:
hematite nanorods on FTO. The conversion from FeOOH to hematite requires annealing at
450 ◦C. An area of the FTO substrate is left uncovered to preserve a highly conductive area to
attach wires for PEC measurement.
Figure 5.8: SEM image of BiVO4 layer on FTO (a) before annealing and (b) after annealing
at 500 ◦C. The scale bars are 1000 nm. The morphology turned from nanoflakes to thin film
during the annealing process.
Figure 5.9: EDX spectra of (a) hematite nanorods on FTO and (b) BiVO4 coated hematite
nanorods on FTO. A clear difference can be noticed at 2.4 eV and 4.9 eV, corresponding to Bi
M and V Kα, respectively.
Elemental analysis was carried out using EDX for samples with and without BiVO4 layer
as illustrated in Figure 5.9. The existence of Bi and V can be confirmed by the two extra
EDX peaks at 2.4 eV and 4.9 eV. The peak at 2.4 eV is corresponding to the M emission line
of bismuth (2.419 eV) while the peak at 4.9 eV is corresponding to the Kα emission line of
vanadium (4.949 eV). Other elements such as Fe, O and Sn can be seen in both samples.
XPS analysis was utilized to reveal more information about elemental chemical states and
surface composites of the BiVO4-coated photoanode. XPS peaks of all the major elements have
been marked out in Figure 5.10. The core level spectrum for Bi 4f has peaks at 163.68 eV
and 153.38 eV, which are related to Bi 4f5/2 and Bi 4f7/2 , respectively. The peak position
is in agreement with previous report, showing that the main oxidation state of Bi is +3, as
expected in BiVO4[163][164]. Two XPS peaks in the core level spectrum of V 2p at 523.32 eV
and 519.99 eV can be assigned to V 2p1/2 and V 2p3/2 , indicating that V exists as V5+ in
this system[165]. For Fe 2p core level spectrum, two peaks can be observed at 723.71 eV and
710.16 eV which are consistent with the typical binding energy of Fe 2p1/2 and Fe 2p3/2 in
hematite[166][164]. Furthermore, since no obvious peak is present at 715 eV, most Fe2+ ions
have been successfully oxidized to Fe3+ in the annealing process.
Figure 5.10: (a) XPS survey spectrum of fabricated BiVO4-coated photoanode and core level
spectra of (b) Bi 4f , (c) V 2p, (d) Fe 2p.
5.3.2 Optical properties
The optical properties were studied for both pristine hematite and BiVO4-coated hematite
using UV-Vis spectrometer. The extinction spectra of pristine hematite, and the BiVO4-coated
hematite can be seen in Figure 5.11(a). The absorption edge of pristine hematite and the
BiVO4-coated hematite locate at around 615 nm and 650 nm, respectively. The BiVO4-coated
hematite shows improved absorption in the wavelength range of 450 nm-700 nm which indicates
better light harvesting efficiency for the photoanode with the hematite-BiVO4 heterojunction.
Optical band gap can be calculated from the the absorption coefficient[167], following the
relationship of:
(αE)2/n = K(E − Eg ) (5.1)
where α is the absorption coefficient; E is the photon energy (hν); Eg is the band gap and K is
a constant. For semiconductors with a direct band gap, n = 1. Therefore, a curve (Tauc plot)
of (αE)2 versus E can be plotted. The optical band gap can be indicated by the x intercepts as
illustrated in Figure 5.11(b). The optical band gap of pristine hematite and the BiVO4-coated
hematite are located at 2.14 eV and 2.08 eV, respectively. The narrowed band gap originates
from the the raised Fermi level of hematite as well as the band bending near the BiVO4-hematite
heterojunction[164].
The optical properties of the fabricated plasmonic nanoarrays are also studied using UV-Vis
spectrometer. The extinction spectra of Au NHA and Ag/Au dual-layer NHA with period of
400 nm are shown in Figure 5.12. Since Au shows interband absorption below 500 nm, it cannot
enable above-bandgap plasmonic enhancement for hematite. For Ag/Au dual-layer NHA, the
EOT wavelength is around 500 nm, which is close to the value predicted by FDTD models
(500 nm-600 nm). The second extinction peak which locates at around 700 nm is probably
originated from the LSPR effect which cannot be clearly seen in the simulated extinction spectra
(Figure 4.18). The different results from the model and the experiment could be caused by the
FTO substrate with high surface roughness and the defects of the fabricated nanoarrays. In
another word, the fabricated nanohole arrays are far from perfect periodic arrays so the SPP
effect which requires the periodicity is less significant while the LSPR effect that is not related
to the periodicity becomes relatively stronger. Since the wavelengths of their second extinction
peak are higher than the absorption edge of hematite, only the HEI mechanism will take place
at this region (below-bandgap enhancement). It is considered that the contribution from the
below-bandgap HEI enhancement to photocurrent is limited, so the difference of the LSPR
peak at around 700 nm should not have much impact on the solar water splitting performance.
5.3.3 Water splitting performance
Optimization was firstly done for the BiVO4 spin-coating process. Four samples with 1-10 times
of BiVO4 spin-coating were prepared. Their photocurrent density curves were measured and
shown in Figure 5.13. All the samples with BiVO4 layer showed better performance compared
to pristine hematite nanorods. Among them, the sample with 5 times of BiVO4 spin-coating
Figure 5.11: (a) Extinction spectra of pristine hematite and BiVO4-coated hematite. Improved
absorption in the range of 450 nm-700 nm can be observed for the sample with BiVO4. (b) Tauc
plot of pristine hematite and BiVO4-coated hematite. The sample with BiVO4 has a smaller
optical band gap.
Figure 5.12: Normalized extinction spectra of Au NHA (solid lines) and Ag/Au dual-layer NHA
(dotted lines) with period of 400 nm. The extinction peak of Au NHA below 500 nm is due to
the interband absorption instead of plasmonic effect. The EOT wavelength of Ag/Au dual-layer
NHA is at around 500 nm. The above-bandgap region of hematite is marked in yellow.
showed the best performance of 0.2 mA cm−2 at 1.2 V vs RHE. When the spin-coating time
increased to 10, the photocurrent density became lower. Most photons are absorbed by the
BiVO4 layer instead of the heterojunction if the BiVO4 layer is too thick. Hence, 5 times of
spin-coating was selected as the optimized procedure.
Figure 5.14 shows the measured PEC photocurrent density of the fabricated photoanodes. For
pristine hematite nanorods, the photocurrent density is 0.04 mA cm−2 (at 1.23 V vs RHE) while
for the BiVO4-coated hematite nanorods, it is about 0.2 mA cm−2 . It is worth noting that when
directly coated to the FTO substrate, BiVO4 has negligible photocurrent so it is clear that
the BiVO4 layer does not act as a standalone photoactive semiconductor to drive the water
splitting reactions by itself but instead as a surface modification layer for hematite to improve
its performance through the heterojunction. Since the major drawbacks of pristine hematite
are the low carrier concentration and the high charge recombination rate, the BiVO4 layer can
improve the photocurrent by increasing the carrier density. Furthermore, the site where the
OER happens moves to the interface between the BiVO4 layer and the electrolyte so that the
issue of high charge recombination rate can also be mitigated. The IPCE measurement can
Figure 5.13: Measured photocurrent density of the hematite nanorods with different times of
BiVO4 layer spin-coating. The one with 5 times of spin-coating shows the best performance
among the all measured samples.
provide insight into the enhancement, as demonstrated in Figure 5.15. The IPCE of pristine
hematite quickly dropped to zero after 400 nm while the BiVO4-coated samples still have decent
IPCE at 460 nm, showing improved efficiency at the band edge. Due to the decreased band gap
of the BiVO4-coated samples, more incident photons can be harvested at the band edge and
consequently increase the IPCE.
The samples with metal nanohole arrays shows further improved photocurrent density as well
as IPCE. In the photocurrent density measurement, the enhancement factor is about 1.3 for
single-layer NHA and 2.1 for dual-layer NHA (at 1.23 V vs RHE). In the IPCE measurement,
similar enhancement factor can also be seen for samples with metal nanoarrays. However, the
mechanisms are different for Au NHA and Au/Ag NHA. Since only above-bandgap enhance-
ment was observed while Au NHA shows no plasmonic properties in this wavelength range,
the possible enhancement mechanism for single-layer Au NHA is non-plasmonic direct carrier
injection. On the other hand, the Ag/Au NHA shows SPP effect in this wavelength range,
so the possible enhancement mechanisms for the dual-layer nanohole arrays are PIRET and
light scattering[16]. Compared to the non-plasmonic mechanisms, the enhancement effects of
plasmonic induced mechanisms are expected to be better, which is consistent with the exper-
imentally measured results. Due to the interband absorption of the Au layer, single-layer Ag
5.4. Conclusion 97
Figure 5.14: PEC photocurrent density of measured photoanodes under chopped light from
solar simulator. The potential is calibrated vs Ag/AgCl reference electrode and converted
to vs RHE. The BiVO4-coated samples shows much better performance compared to pristine
hematite. Performance enhancement can be seen for samples with metal nanoarrays while the
sample with dual-layer NHA shows the best performance.
NHA may have even better performance. Nevertheless, Ag is not as stable as Au, so single-layer
Ag NHA can be damaged during the annealing process of the hematite nanorods. As a result,
single-layer Ag NHA is not studied in this project. Besides, the ratio between the thickness
of each metal layer may also influence the enhancement factor but only 1:1 dual-layer NHA is
studied. To further increase the plasmonic enhancement factor, it is necessary to investigate
dual-layer NHAs with different thickness ratio.
5.4 Conclusion
In this chapter, dual-layer Ag/Au nanohole arrays with homogeneous hexagonal structure were
successfully fabricated, showing SPP effect above the band gap of hematite. Hematite nanorods
were grown directly on the nanohole array without using any template. The successful coating of
the BiVO4 layer on hematite nanorods can be confirmed by EDX and XPS analysis. By optical
measurements, the sample with BiVO4 layer shows decreased band gap, which is beneficial to the
solar water splitting performance. The PEC photocurrent measurement also shows enhanced
5.4. Conclusion 98
Figure 5.15: IPCE curve of measured photoanodes at 1.23 V vs RHE. The sample with dual-
layer NHA shows the best performance, which is in confirmation with PEC photocurrent mea-
surement.
photocurrent density for BiVO4-coated samples with best enhancement factor being close to 5.0.
Furthermore, the presence of the metal nanostructure can further boost the PEC performance
by 2.1 times using the dual-layer Ag/Au NHA or 1.3 times using the single-layer Au NHA. The
plasmonic enhancement from the Ag/Au NHA was observed to be more significant compared to
the non-plasmonic enhancement from the Au NHA. The IPCE measurement shows consistent
results with the PEC photocurrent measurement which confirms that dual-layer Ag/Au NHA
has the best enhancement factor among the fabricated photoanodes.
Chapter 6
Aluminum Plasmonic Nanohole Array

Enhanced Hematite Nanotube
Photoanodes
99
Figure 6.1: Schematic illustration of the hematite nanotube photoanodes based on aluminum
nanohole array.
6.1 Introduction
Noble-metal plasmonic nanostructures have been proven to have outstanding performance as

well as great stability in various systems. However, the use of expensive metals such as gold and
silver hinders large-scale commercial applications. Finding a feasible way to employ non-noble
plasmonic nanostructures in such systems is one of the major challenges.
Aluminum and copper as plasmonic metals, have received many attentions as alternatives
to silver and gold[168]. Aluminum, which has plasmonic resonance wavelength in the UV-
visible region, is a promising plasmonic metal to be utilized in solar energy harvesting systems
including photoanodes. In this chapter, aluminum nanohole arrays were used to enhance the
performance of hematite photoanodes. Due to the stability issue of aluminum, a novel hematite
nanostructure, nanotube, was used to avoid the high temperature annealing process which is
usually required for traditional hematite photoanodes. With the help of a silica protective layer,
the utilized aluminum nanohole arrays showed consistent performance in the stability test. The
SPP effect from the periodic plasmonic nanostructure is employed to help increase the solar
water splitting efficiency. It was demonstrated that aluminum, the earth-abundant plasmonic
material, exhibits great performance in our system and can boost the PEC photocurrent density
by up to 3.6 times, compared to pristine hematite photoanodes. The excellent enhancement
from aluminum nanohole arrays can pave the way for water splitting devices with both low cost
and high efficiency.
6.2 Experimental
Materials. Iron(III) chloride hexahydrate, ammonium dihydrogen phosphate and sodium

dodecylsulphate were purchased from Sigma-Aldrich and used as received without further pu-
rification. Polystyrene spheres were purchased from Bangs Laboratories, Inc and used as re-
ceived. FTO glass was purchased from Solaronix and washed with acetone, isopropyl alcohol
and ethanol before use.
Fabrication of periodic aluminum nanohole array on FTO substrate. The aluminum

plasmonic arrays were fabricated using the same colloidal lithography method as in Chapter
5.2[157][158]. Briefly, 10% aqueous solution of monodispersed polystyrene nanoparticles was
mixed with an equal volume of ethanol. Clean silicon wafers and FTO substrates were prepared
and treated with UV-ozone for 2 h. 10 µL of polystyrene solution was deposited onto the
silicon wafer by drop-coating to produce a homogeneous polystyrene thin layer. After that,
the wafer was slowly immersed in a vessel filled with 450 mL of deionized water and 3 µL of
2% sodium dodecyl sulphate solution, leaving a self-assembled polystyrene monolayer floating
on the surface of the water. The polystyrene monolayer was then transferred onto the UV-
ozone treated FTO substrate and dried in air for 24 h before further use. Additionally, the
polystyrene-covered FTO substrates were treated with reactive ion etching for 25 s to reduce
the diameter of the polystyrene spheres. Next, the as-prepared substrates were moved into
Mantis evaporation/sputtering system for sputtering to obtain a 50 nm layer of Al and 10 nm

of SiO2 as protective layer. Subsequently, the polystyrene spheres were removed by a duct tape
and the substrates covered with plasmonic nanohole array were later checked under SEM to
reject those with excessive defects.
Synthesis of hematite nanotubes. The single-crystalline hematite nanotubes were syn-

thesized through a hydrothermal method following a previous report[169]. To produce the
hydrothermal precursor solution, 1.35 g (5 mmol) of iron(III) chloride hexahydrate and 20.7 mg
(0.18 mmol) of ammonium dihydrogen phosphate were weighted and transferred to a 250 mL
volumetric flask. Then 250 mL of DI water was added to the volumetric flask. After that, the
precursor solution was stirred with a magnetic stirrer for 15 min. Meanwhile, a 80 mL PTFE-
lined autoclave reactor was washed with DI water, ethanol and dried for further use. When the
powder was completely dissolved in the precursor solution, 60 mL of the solution was added to
the autoclave. The autoclave was then transferred to an oven and kept at 220 ◦C or 40 h. After
cooled to room temperature, the solution was washed and centrifuged twice with DI water and
then two more times with ethanol. Afterwards, the solution was dried in a vacuum oven at
50 ◦C and the powder product was checked with XRD and kept in plastic sample bags at room
temperature for further use.
Fabrication of hematite nanotubes photoanodes. The as-synthesized hematite nanotubes

were spin-coated[25] onto the FTO substrate to obtain a layer of hematite nanotubes on the
plasmonic nanohole arrays. Briefly, the hematite nanotubes were dispersed into 1.5% chitosan
solution to form a 10 mg mL−1 solution and then ultrasonicated for 15 min to get a homogeneous
spin-coating precursor solution. Before coating, a small area of the FTO substrate was covered
with Kapton tape to preserve a highly conductive area for further PEC measurements. Next,
the hematite nanotubes were spin-coated onto the FTO substrate with spinning speed of 500
rpm for 10 s and 2000 rpm for 25 s. All the fabricated substrates were kept at 280 ◦C for 4 h to
remove any organic residues. The 3D illustration in Figure 6.2 shows all the critical steps in
the fabrication process.
Characterization. The synthesized hematite nanotubes were checked under TEM and ana-
Figure 6.2: Illustration of the critical steps to fabricate hematite nanotube photoanode on
Al nanohole array. The steps are (a) coating self-assembled template nanospheres onto the
substrate, (b) using reactive ion etching to reduce the size of the spheres, (c) coating of the
Al layer and the silica protective layer, (d) removing the organic residue by duct tapes and (e)
coating as-synthesized hematite nanotubes onto the nanohole array.
lyzed by XRD. The fabricated aluminum nanohole arrays were checked under SEM with EDX
mapping. UV-Vis spectrometer was used to obtain the optical properties of the nanohole arrays.
The finally fabricated hematite photoanodes were checked under SEM before PEC measure-
ments. The water splitting performance was benchmarked by PEC photocurrent measurement,
IPCE measurement as well as electrochemical impedance spectroscopy (EIS) measurement.
6.3.1 Characterization of hematite nanotubes
Unlike the nanorods used in the previous chapter which grow directly from the FTO substrate
which requires annealing to be converted to hematite, the hematite nanotubes are dispersed
particles upon synthesis and do not require any further annealing. This way the substrate was
prevented from being damaged during the hydrothermal reaction as well as the high temperature
Figure 6.3: (a) TEM image of hematite nanotubes. The scale bar is 100 nm. (b) STEM image
of hematite nanotubes. The scale bar is 200 nm. (c) HRTEM image of hematite nanotubes.

The scale bar is 5 nm. The distance between adjacent lattice planes is measured to be 2.2 A.
(d) TEM diffraction pattern of hematite nanotubes. The scale bar is 1 nm−1 .
annealing process, enabling the possible application of the aluminum nanohole arrays.
The single-crystalline hematite nanotubes were characterized after the hydrothermal synthesis.
Figure 6.3(a) and 6.3(b) show the TEM and STEM images of the nanotubes, indicating that the
hematite nanotubes have homogeneous morphology of 200 nm in length and 50 nm in diameter.
Hollow structure can be observed and the wall thickness is measured to be around 15 nm.
HRTEM and diffraction pattern are presented in Figure 6.3(c) and Figure 6.3(d), confirming
that the synthesized nanotubes are single-crystalline. Besides, the distance between adjacent
matching the reported value for the {113} plane of
lattice planes is measures to be 2.2 A,
Figure 6.4: XRD pattern of single crystalline hematite nanotubes.
hematite[170]. Figure 6.4 presents the XRD pattern of the as-synthesized hematite nanotubes
which exactly matches the pattern of α-Fe2O3[171], confirming there are no other impurities.
As discussed before, hematite suffers from the disadvantage of high charge recombination rate
which makes it beneficial to have 1D nanostructures such as nanorods and nanotubes in order
to decrease the necessary distance charges must travel[70]. The hollow structure of hematite
nanotubes are not only able to further mitigate the charge recombination issue but also have
larger surface area than that of nanorods, which enables more active reaction site for the OER
reaction. Moreover, by using single-crystalline hematite, the charge trapping effect by grain
boundaries is minimized. Therefore, the photocurrent density can benefit from the increased
maximum charge transfer distance[172].
6.3.2 Characterization of the fabricated photoanodes
Since aluminum is much more active than gold and silver, it can be quickly oxidized when
exposed to air. The oxidization layer of Al2O3, when present, can significantly impact the
plasmonic effect. Figure 6.5 shows the predicted influence from the oxidization layer by FDTD
modelling. For 80 nm-thick Al nanohole array, an oxidation layer of 30 nm Al2O3 can lead to
a 40% loss of plasmonic effect, as a result of the decreased thickness of the plasmonic layer as
well as the loss of field intensity in the oxidation layer. Due to the difficulty of controlling the
thickness of Al2O3 layer and the resulting impact on plasmonic effect when it is too thick, a
Figure 6.5: FDTD modelled extinction spectra of Al nanohole array with various thickness of
oxidization layer (Al2O3). With thicker oxidization layer, the extinction curve becomes flatter.
The inset table lists the normalized plasmonic enhancement factors.
protective layer of silica was deposited to the nanoarray immediately after the sputtering of
aluminum without opening the vacuum chamber of the sputtering equipment to avoid any direct
contact between the aluminum layer and the air. The thickness of the silica layer is controlled to
be 10 nm to minimize the impact on the plasmonic effect. Nevertheless, the applied silica layer
is not conductive and thus an increased resistance can be expected for the protected substrate
which may have negative impacts on the photocurrent. For the same reason, the silica layer is
not applied on reference sample with Al film, otherwise the hematite nanotubes are completely
isolated from the FTO substrate.
Figure 6.6: SEM images of (a) self-assembled monodispersed polystyrene spheres on FTO
substrate, (b) Al nanohole array on FTO substrate and (c) fabricated photoanode with hematite
nanotubes on Al nanohole array. The scale bars are 500 nm.
Figure 6.7: Normalized extinction spectra of Al nanohole arrays and hematite nanotubes. The
EOT wavelength is 490 nm for 500 nm Al NHA and 570 nm for 620 nm Al NHA.
The substrates were characterized under SEM after the lift-off of the monodispersed polystyrene
masks, after the removal of the masks and after the final fabrication. The SEM image of the
polystyrene mask on FTO is presented in Figure 6.6(a). The hexagonal array of the polystyrene
spheres confirms the successful self-assembly of polystyrene spheres at the water-air interface
which was later transferred to the FTO substrate. Defects can still be observed in some areas,
which is mainly due to the rough surface of FTO. Figure 6.6(b) shows the SEM image after
the aluminum and silica deposition, in which the nanoholes keep their hexagonal order with
the period equal to the diameter of the polystyrene spheres. The surface inside the nanoholes
has the typical morphology of FTO substrate with high surface roughness. Figure 6.6(c) shows
the SEM image of the finally fabricated hematite nanotube photoanodes. The direction of
the single-crystalline hematite nanotubes are randomly distributed after the spin-coating. The
hematite layer is highly compact, which can boost the light harvesting efficiency and avoid any
direct contact between the electrolyte and the plasmonic nanostructure.
The optical properties of the nanohole array as well as the hematite nanotubes were examined
by UV-Vis spectrometer. The extinction spectrum of the hematite nanotubes as shown in Fig-
ure 6.7 indicates its absorption edge at around 550 nm. Due to the relatively low band gap of
hematite, it can benefit from the wider range of solar light absorption which leads to a better
light harvesting efficiency and allows a much better overlap with the plasmonic resonance from
the periodic arrays[173][33]. It is also demonstrated in Figure 6.7 that the EOT wavelength
is different between Al nanohole arrays with different period. Nanohole arrays with smaller
holes result in shorter resonance wavelength, showing the feasibility to control the plasmonic
resonance wavelength by manipulating the morphology of the plasmonic nanostructure in order
to match the band gap of the semiconductors. The experimental results showed that 500 nm Al
NHA has EOT at 490 nm and 620 nm Al NHA has EOT at 570 nm, which is in consistent with
the FDTD predicted values. It can also be seen for both NHA, the extinction peak as well as the
extinction dip (EOT wavelength) overlap with the absorption spectrum of the hematite nan-
otubes, potentially enabling SPP-induced plasmonic enhanced water splitting above hematite’s
band gap.
6.3.3 Characterization of water splitting performance
The measured photocurrent density curves are presented in Figure 6.8 with two reference
samples that are pristine hematite nanotubes and hematite nanotubes on Al film (without
nanoholes). The Al film as the reference sample is not coated with the silica protective layer
otherwise the hematite nanotubes are completely isolated from the FTO substrate. It is demon-
strated that hematite nanotubes on Al NHA with 500 nm period has the best performance
among all the samples tested. It has a 3.6 times enhancement compared to the reference
sample with pristine hematite. The sample with 620 nm Al NHA also shows improved pho-
tocurrent, which is about 2.2 times compared to pristine hematite. The curves from the IPCE
measurement (Figure 6.9) shows similar results. The peak of IPCE locates at around 380 nm
and the band edge is at approximately 500 nm. Among all the measured samples, the one
with 500 nm Al NHA has the highest IPCE. A better IPCE performance eventually leads to
higher overall quantum efficiency and hence the enhancement factor in the IPCE measure-
ments corresponds to that in the PEC photocurrent measurement. The IPCE depends on
the efficiency of photon harvesting, charge separation and charge transferring rate across the
interface. Plasmonic nanostructures can considerably improve charge separation with PIRET
Figure 6.8: PEC photocurrent density of the fabricated hematite nanotube photoanodes. Best
performance is achieved by hematite nanotubes with 500 nm Al NHA.
mechanism and improve charge transferring rate with HEI mechanism[16]. However, the HEI
effect usually contributes much less compared to the PIRET effect. Since PIRET mechanism
only takes place when the incident photon energy are above the band gap of the photoactive
semiconductor, sufficient spectral overlap between the plasmonic resonance and the absorption
of hematite is required. The plasmonic resonance wavelength of the 500 nm Al NHA is lower
than that of the 620 nm Al NHA, so a better overlap is expected which explains the superior
enhancement factor of the 500 nm Al NHA. Besides, in both PEC photocurrent measurement
and IPCE measurement, the sample with Al film showed no better performance compared to
pristine hematite. Without the existence of the periodic nanoholes, the SPP effect cannot be
exhibited under the perpendicular incident light and thus no plasmonic enhancement can be
observed.
To further investigate the enhancement effect from the plasmonic nanostructure, EIS mea-
surement was done for the hematite photoanode with Al NHA and the reference hematite
photoanode with Al film. The results of the EIS measurement are shown in Figure 6.10. The
EIS curves were fitted using ZView with a equivalent circuit that consists of two resistance
component: R1 and Rct as well as two constant phase component: CP E1 and CP E2. The
diagram of the equivalent circuit is shown in Figure 6.10(e). The R1 value is corresponding to
Figure 6.9: IPCE spectra of fabricated hematite nanotube photoanodes. The black curve
represents the reference sample of pristine hematite and the grey curve represents the reference
sample of hematite with Al film. Best performance is achieved by the sample with 500 nm Al
NHA, which is in consistent with the photocurrent density curves.
the contact resistance, which is determined by the resistance of the electrolyte, the substrate
and the external circuit. It can be seen from the fitted result that R1 is smaller in the reference
photoanode which is due to the absence of the non-conductive silica protective layer. The R1
value can affect the dark current but should have negligible impact on the photocurrent. On
the other hand, the Rct value is related to the charge transfer process and is thus related to the
photocurrent. The hematite photoanode with the plasmonic nanostructure has Rct of 121.8 Ω
which is much smaller than that of the reference photoanode (2104 Ω). The EIS measurement
shows consistent results with the IPCE and the PEC photocurrent measurement. The plas-
monic nanohole array can improve the charge transfer process and thus results in increased
photocurrent.
Furthermore, the stability of the fabricated photoanodes was also investigated. Chopped light
was utilized to show both the photocurrent and the dark current during the stability test, to help
identify where the performance degradation originates originates from. If increased dark current
is observed, it means that the photoactive semiconductor is detached from the substrate, leading
to a direct contact between the substrate and the electrolyte. On the other hand, if decreased
Figure 6.10: (a), (b) Experimental (red) and fitted (green) electrochemical impedance spec-
troscopy (EIS) plot of reference samples (hematite nanotubes on Al film), (c), (d) Experimental
(red) and fitted (green) EIS plot of hematite nanotubes on Al NHA, (e) Equivalent circuit used
for fitting. Inset table shows the fitted results. The lower R1 value of reference sample indi-
cates it is more conductive which is due to the absence of silica layer. The lower Rct value of
hematite+Al NHA shows it has better water splitting performance.
6.4. Conclusion 113
Figure 6.11: Photoelectrochemical stability measurement of the fabricated hematite nanotube

photoanodes for 240 min. Pristine hematite photoanodes has the most stable current over the
measured time interval. Al NHA sample with protective layer has more than 90% of the original
performance after 240 min. Al film sample without protective layer has less than 30% of the
original performance at the end of the measurement.
photocurrent is observed, it indicates the decreased water splitting efficiency of the photoactive
semiconductor. As shown in Figure 6.11, the pristine hematite sample have stable photocurrent
across the measured time interval. Samples with aluminium show decreased photocurrent after
240 min of solar water splitting. The photoanode with Al film performed worst in stability test
and its photocurrent gradually decreased to 30% of its original performance due to the lack of
the silica protective layer. On the contrary, the sample with Al NHA still has more than 90%
of original performance after the same amount of time. It is hence confirmed that the silica
protective layer can significantly improve the chemical stability of the aluminum plasmonic
nanostructure. The hematite photoanodes with protected Al NHA have comparable stability
to the pristine hematite photoanodes.
6.4 Conclusion
In this chapter, novel hematite nanotube photoanodes enhanced by plasmonic nanotube array
were fabricated using non-noble metal. Better PEC performance as well as higher IPCE in the
6.4. Conclusion 114
UV-visible region were observed under solar light. With hematite nanotubes on Al nanohole
array, enhancement factor of up to 3.6 in PEC photocurrent measurement was achieved under
compared to samples without plasmonic nanoarray. The superior performance mainly results
from the SPP-induced plasmonic enhancement with the PIRET mechanism. With the help of
silica protective layer, the chemical stability issue of Al nanoarrays can also be mitigated. It
proved the concept that Al nanohole array as a non-noble plasmonic nanostructure can be a
promising platform for plasmonic enhanced solar water splitting devices. The low-cost nature of
Al makes it a excellent alternative to noble metals and can thus enable large-scale applications.
The necessity of matching SPP wavelength with the absorption peak of the semiconductor
is also demonstrated. By further tailoring the plasmonic resonance wavelength of the the Al
nanohole array, better enhancement factors can be achieved.
Chapter 7
TTA Upconversion Induced Solar

Water Splitting by Hematite
Photoanodes with Plasmonic
Enhancement
115
Figure 7.1: Schematic illustration of the inverse-opal hematite photoanodes with Ag-couple
TTA upconversion materials. The interactions between the components are: (i) plasmonic
enhanced upconversion, (ii) absorption of upconverted photons by hematite and (iii) plasmonic
enhanced water splitting
7.1 Introduction
In Chapter 5 and Chapter 6, hematite photoanodes with plasmonic nanoarrays have shown
improved photocurrent density. Nevertheless, the plasmonic enhancement through hot electron
injection below the band gap of hematite has not been observed. Using plasmonic enhancement
alone is not sufficient to extend the spectral range of hematite’s absorption of the incident solar
light.
As a photon-management technique, upconversion materials can be utilized in light harvesting

systems to overcome the spectral limitations, by converting photons with long wavelength to
photons with short wavelength. With the help of upconversion materials, it is possible for
hematite photoanodes to absorb more photons with energy below its band gap and thus results
in higher overall solar-to-hydrogen efficiency. A previous joint work by me and Xiangyu Xie
which focused on utilizing lanthanide based upconversion materials for hematite nanorods has
proven the concept of upconversion induced water splitting[174]. However, further application
of such a system in solar light harvesting is hindered by the requirement of NIR laser source
to excite the lanthanide based upconversion materials. On the other hand, the triplet–triplet
annihilation (TTA) upconversion materials, which can be excited by low intensity incident
light, can enable observable photocurrent improvement under solar light[175]. Moreover, TTA
upconversion materials can be excited by visible light and have tunable excitation as well
as emission wavelength, making them better candidates to be utilized in photoanodes and
photocatalysts[176].
In this chapter, TTA@SiO2 core-shell upconversion nanoparticles based on DPA (diphenylan-

thracene) and PdOEP (octaethylporphyrin palladium) were synthesized and the plasmonic
enhanced TTA upconversion by Ag nanoparticles was studied. A promising enhancement fac-
tor of up to 1.8 was achieved. The plasmonic enhanced TTA upconversion material was then
applied to a highly porous hematite photoanodes with inverse-opal structure as illustrated in
Figure 7.1. By measuring the water splitting performance, contributions from both TTA up-
conversion as well as plasmonic enhancement can be observed. An optimal IPCE enhancement
factor of 61.3 was obtained at 520 nm.
7.2 Experimental
Materials. Pluronic F127, 2,3,7,8,12,13,17,18-octaethylporphyrin palladium, toluene, 9,10-

diphenylanthracene, dichloromethane, hydrochloric acid, tetraethyl orthosilicate, polyethylen-
imine, silver nanoparticles, diethoxydimethylsilane, ferrous sulfate, ascorbic acid, amidosulfonic
acid and boric acid were purchased from Sigma-Aldrich and used as received without further
purification. Polystyrene spheres were purchased from Bangs Laboratories, Inc and used as
received. FTO glass was purchased from Solaronix and washed with acetone, isopropyl alcohol
and ethanol before use.
Synthesis of the Ag-coupled TTA@SiO2 core-shell NP. The synthesis of the core-shell
nanoparticles is achieved by the synthesis of TTA upconversion molecules and the growth
of SiO2 following an adapted protocol from literature[177]. Polyethylenimine was later used
to couple the Ag nanoparticles to the synthesized TTA core-shell nanoparticles[178]. Briefly,
0.5 mg of 2,3,7,8,12,13,17,18-octaethylporphyrin palladium, 2.0 g of Pluronic F127 and 10 mg

of 9,10-diphenylanthracene were dissolved in a 20 mL mixture of dichloromethane and toluene
(volume ratio 1:3). The solvent was slowly evaporated under Ar atmosphere at room temper-
ature. After 24 h, the remaining solvent was removed under reduced pressure at 40 ◦C. The
solid residue was then dissolved in 100 mL 0.85 mol L−1 hydrochloric acid. After adding 2 mL of
tetraethyl orthosilicate, the resulting aqueous solution was stirred for 6 h at 25 ◦C. 2.1 mmol of
diethoxydimethylsilane was then added to the solution followed by stirring at room temperature
for another 48 h. The solution was transferred into a dialysis bag and immersed in deionized
water. The dialysis process lasted for 48 h and the water in the bath was replaced every 6 h.
Afterwards, the aqueous solution inside the dialysis bag was transferred to a brown glass bottle
and kept at 4 ◦C for further use. To couple the TTA@SiO2 core-shell nanoparticles with Ag
nanoparticles, 5.0 mg of polyethylenimine was added to the solution (10 mL) followed by stirring
for 2 h. Excess polyethylenimine was removed by centrifugation and washed with deionized wa-
ter. The solution of Ag nanoparticles was added to the aqueous solutions of polyethylenimine
coated TTA@SiO2 core-shell nanoparticles. An optimized particle ratio of 1:4 is controlled be-
tween Ag and TTA@SiO2 core-shell nanoparticles. After 2 h of stirring, the resulting solution
was washed with deionized water to remove unattached Ag nanoparticles.
Synthesis of the hematite inverse-opal photoanodes. The inverse-opal photoanodes[179]

were synthesized via an electrodeposition method using self-assembled multi-layer polystyrene
templates[180]. In brief, a cleaned FTO glass was put into a glass cylinder filled with solution
of 0.15 wt% polystyrene spheres (300 nm). The glass cylinder was then kept at 55 ◦C for a
week to allow the self-assembly of polystyrene spheres on the surface of FTO. The FTO glass
was then put under UV-ozone for 1 h. The resulting polystyrene-covered FTO glass was then
used as templates for the electrodeposition of iron. The electrolyte was prepared by adding
60 g ferrous sulfate, 1.5 g ascorbic acid, 0.5 g amidosulfonic acid and 15 g boric acid to 1.0 L
of deionized water. A constant potential of -0.5 V vs. Ag/AgCl was applied to the FTO
glass as working electrode while a Pt mesh was used as counter electrode. After 300 s of
electrodeposition, the FTO glass was taken out and washed with deionized water. It was
subsequently annealed at 450 ◦C for 4 h to convert iron to hematite and remove any organic
Figure 7.2: Illustration of the fabrication process of hematite inverse-opal photoanodes. (a)
Self-assembly of multi-layer polystyrene spheres on FTO substrate forming an opal structure.
(b) Growth of iron by electrodeposition. (c) Annealing process to remove organic residue and
oxidize iron to hematite.
residue. Finally, the as-prepared solution of TTA@SiO2@Ag nanoparticles were spin-coated

onto the hematite photoanodes. The spinning speed was 500 rpm for the first 20s and 2000
rpm for the following 30s.
Characterization. The purchased Ag nanoparticles and the as-synthesized TTA@SiO2 core-

shell nanoparticles where checked under TEM. The extinction spectra of the Ag nanoparti-
cles were obtained by UV-Vis spectrometer. Photoluminescence measurement was carried out
for TTA@SiO2 core-shell nanoparticles with and without Ag nanoparticles. The multi-layer
polystyrene templates and the hematite inverse-opal photoanodes were checked under SEM.
The IPCE and PEC solar water splitting performance were measured using pH7 buffer solution
as the electrolyte and Ag/AgCl as the reference electrode.
7.3.1 Characterization of TTA@SiO2 upconversion nanoparticles.
The synthesized TTA@SiO2 core-shell upconversion nanoparticles were checked under TEM as
shown in Figure 7.3. The diameter of the core-shell nanoparticles are ranging from 9 nm to
Figure 7.3: TEM image of synthesized TTA@SiO2 core-shell upconversion nanoparticles. The
particle size is 9 nm-12 nm and the thickness of the silica coating is 2 nm-3 nm. (a) The scale
bar is 10 nm. (b) The scale bar is 30 nm.
12 nm. The coated silica shell is 2 nm-3 nm thick. The TTA upconversion material itself is
actually in the form of organic molecules and is thus difficult to be attached directly to the
hematite photoanodes. The silica shell not only packs the TTA upconversion material into
nanoparticles but also acts as a protective layer to prevent any damage from the electrolyte
and oxidation. The PL property of the core-shell upconversion nanoparticles is measured by a
PL spectrometer and the results are shown in Figure 7.4(a). The TTA upconversion process
takes place with excitation wavelength at 510 nm-545 nm and emission wavelength at 410 nm-
455 nm, which is similar to the reported values[177]. The hot zone indicates the region where
the upconversion process shows the highest efficiency. The optimized excitation wavelength
is 520 nm-525 nm while the maximum emission intensity is obtained at 425 nm-445 nm. As
illustrated in Figure7.4(b), the emission peak of the TTA upconversion material overlaps with
the extinction peak of hematite and the excitation peak of the TTA upconversion material
is at the band edge of hematite. The upconversion material can thus improve hematite’s
water splitting performance at its band edge by helping hematite absorb more photons with
wavelength longer than 500 nm.
Figure 7.4: PL intensity of synthesized TTA@SiO2 core-shell upconversion nanoparticles. (a)

Excitation-emission mapping indicating the upconversion process from around 510 nm-545 nm
to 410 nm-455 nm. The best upconversion efficiency is achieved by excitation at around 520 nm
and emission at around 440 nm. (b) Normalized emission spectrum (with excitation at 520 nm)
and excitation spectrum (with emission at 440 nm) of TTA@SiO2 upconversion nanoparticles
together with the extinction spectrum of hematite. The emission spectrum of the TTA@SiO2
upconversion nanoparticles overlaps with the extinction peak of hematite.
Figure 7.5: Extinction spectra of Ag nanoparticles with different diameters ranging from 10 nm
to 100 nm. Smaller Ag nanoparticles have more significant LSPR peaks. The emission and
excitation peak positions of the TTA@SiO2 upconversion nanoparticles are marked in green
and blue, respectively. The emission peak of TTA overlaps with the LSPR peaks of all the
measured Ag nanoparticles.
7.3.2 Characterization of plasmonic enhanced TTA upconversion.
To further improve the efficiency of TTA@SiO2 core-shell upconversion nanoparticles, silver

nanoparticles were attached to the upconversion nanoparticles for plasmonic enhancement.
The extinction spectra of the Ag nanoparticles with diameter from 10 nm to 100 nm were
measured and shown in Figure 7.5. The LSPR peaks of Ag nanoparticles are found to be
red-shifted as the diameter increases. Also, the intensity of the peaks are lower for larger Ag
nanoparticles. For all the samples, the LSPR peaks locate at UV-Vis range and overlap with
the emission spectrum of TTA@SiO2 upconversion nanoparticles. Furthermore, only 100 nm
Ag nanoparticles have LSPR peak covering 500 nm-600 nm that overlaps with the excitation
spectrum of TTA upconversion nanoparticles. For smaller Ag nanoparticles, they can only
enhance the TTA upconversion process by emission enhancement.
To identify the optimal ratio between the TTA upconversion nanoparticles and the Ag nanopar-
Figure 7.6: Measure PL emission intensity of Ag-coupled TTA@SiO2 upconversion nanoparti-

cles with different particle ratio between TTA and (a) 10 nm Ag, (b) 40 nm Ag, (c) 60 nm Ag,
(d) 100 nm Ag. All samples were excited by 520 nm incident light.
ticles, samples were prepared with different ratio and the resulting PL emission intensity curves
were measured as shown in Figure 7.6. For samples with particle ratio (TTA:Ag) smaller than
1:1, quenching effect can be observed. When too many Ag nanoparticles are present, a sub-
stantial amount of upconverted photons emitted by TTA are absorbed by the Ag nanoparticles
and thus a decreased emission intensity is expected. Among all the measured samples, particle
ratio 4:1 shows the best performance and was thus selected as the optimal ratio for further
experiments.
With particle ratio fixed at 4:1, the PL emission spectra in Figure 7.7 shows enhanced perfor-
mance for samples with 10 nm, 40 nm and 60 nm Ag nanoparticles. The maximum enhancement
factor is obtained by sample TTA+Ag(10 nm) which is close to 1.8. The sample with 100 nm
Ag nanoparticles has negligible plasmonic enhancement with no higher emission intensity com-
pared to the reference sample. Other samples have similar enhancement factor in the range
Figure 7.7: PL emission spectrum of TTA@SiO2@Ag particles with 520 nm incident wavelength.
The sample with 10 nm Ag nanoparticles exhibits the best performance.
of 1.3-1.6. Since all the samples with observable enhancement do not have spectral overlaps
between the LSPR peak of Ag and the excitation spectrum of TTA, the effective mechanism
of plasmonic enhanced upconversion here should be emission enhancement. The 10 nm Ag
nanoparticles only have partial spectral overlap with the emission spectrum of TTA, but their
highest LSPR intensity among all the Ag nanoparticles leads to the superior plasmonic en-
hancement factor. On the contrary, the low LSPR intensity of 100 nm Ag nanoparticles results
in the worst performance. The best two samples (with 10 nm and 40 nm Ag nanoparticles) were
selected as the candidates and were spin-coated onto the hematite photoanodes.
The Ag-coupled TTA@SiO2 nanoparticles were also checked under TEM. As shown in Figure
7.8(a), particle aggregation happened after the attachment of Ag nanoparticles and thus larger
particles formed. Besides, tiny dots (3 nm-4 nm) can be found which are solid silica nanopar-
ticles as side products from the silica coating process. It is difficult to remove them from the
solution of TTA@SiO2 nanoparticles and they are not supposed to have any impact on the
upconversion performance so no further purification was taken. On a larger scale, as shown in
Figure 7.8(b), the dark Ag nanoparticles are homogeneously distributed and the TTA@SiO2
nanoparticles aggregated around them.
Figure 7.8: TEM images of TTA@SiO2 nanoparticles coupled with Ag nanoparticles. Left:
the scale bar is 50 nm. Right: the scale bar is 200 nm. Larger particles are formed due to the
aggregation of Ag nanoparticles (darker particles) and TTA@SiO2 nanoparticles.
7.3.3 Inverse-opal hematite photoanodes with TTA upconversion
materials
SEM images were taken before the electrodeposition to confirm the quality of the prepared
templates with self-assembled multi-layer polystyrene spheres as shown in Figure 7.9. The
polystyrene opal template has a multi-layer, hexagonal packed structure. Although cracks can
be found in the template (for example in Figure 7.9(a)), good homogeneity was achieved on
a large scale. As a critical step, the UV-ozone treatment can shrink the polystyrene spheres
to leave extra space for the electrodeposition to get thicker walls in the resulting inverse-opal
structure. Thicker walls can increase the mechanical strength of the inverse-opal structure and
prevent it from getting damaged during further measurements.
The synthesized inverse-opal photoanodes ready for performance measurement were also firstly
checked under SEM. As can be seen in Figure 7.10, the inverse-opal structure is formed by filling
the gaps between the polystyrene spheres. The wall thickness of the inverse-opal hematite is
measured to be around 20 nm. The holes in the inverse-opal structure have the size of 150 nm
providing sufficient space for the Ag-coupled TTA@SiO2 particles to get in. Furthermore, the
hollow structure of the inverse-opal hematite photoanode has a large interface between the
electrolyte and hematite which can lead to excellent solar water splitting performance.
The IPCE measurement results in curves as shown in Figure 7.11(b) with peaks centered at
Figure 7.9: SEM images of the self-assembled multi-layer polystyrene spheres (a), (b) before
UV-ozone treatment and (c), (d) after UV-ozone treatment. The scale bars are 1000 nm. The
gap between adjacent spheres are significantly larger after the UV-ozone treatment, leaving
more space for electrodeposition.
Figure 7.10: SEM images of the finally fabricated inverse-opal photoanodes. The high porous
structure has hole size measured to be 150 nm-200 nm. (a) The scale bar is 1000 nm. (b) The
scale bar is 300 nm.
Table 7.1: IPCE enhancement factors of hematite photoanodes with TTA upconversion mate-
rials
vs hematite at vs hematite +
Sample name
520 nm TTA at 520 nm
hematite + Ag10 1.5 <1
hematite + TTA 18.3 1.0
hematite + TTA + Ag10 61.3 3.4
hematite + TTA + Ag40 36.9 2.0
350 nm. For pristine hematite, the IPCE decreases to zero at around 500 nm which is close
to the theoretical band gap of hematite, 2.2 eV. It can be found in the IPCE curves that
samples with TTA upconversion have additional peaks centered at 520 nm matching the exci-
tation wavelength of the TTA upconversion material. As listed in Table 7.1, the enhancement
factor of the hematite+TTA sample is 18.3 vs pristine hematite at 520 nm. It proves that the
the upconversion process does contribute to the photocurrent in this system. Some incident
photons with wavelength above 500 nm were upconverted to photons with shorter wavelength
of 400 nm-450 nm where pristine hematite still has decent IPCE. Moreover, with Ag nanopar-
ticles attached, an optimal IPCE enhancement factor of 61.3 can be observed at 520 nm. The
sample with TTA+Ag(10 nm) shows better upconversion efficiency compared to the sample
with TTA+Ag(40 nm), which is in conformity with the previous PL measurements. Since the
sample of hematite+Ag only has an enhancement factor of 1.4 at the same wavelength, the en-
hanced IPCE at 520 nm is mainly from plasmonic enhanced upconversion instead of plasmonic
enhanced solar water splitting. The 10 nm and 40 nm Ag nanoparticles have LSPR peaks at
400 nm-450 nm, so the HEI effect at above 500 nm is expected to be weak. On the other hand,
the IPCE of samples with Ag nanoparticles is indeed slightly higher than those without Ag
nanoparticles at around 420 nm where plasmonic enhanced water splitting take place. There-
fore, the overall photocurrent enhancement can be associated with contributions from both
pathways.
As can be noticed in Figure 7.11(a), the overall enhancement of photocurrent is not significant.
The best performance is also achieved by the sample with TTA+Ag(10 nm) but it only has
an enhancement factor of 1.1 at 1.23 V vs RHE. It is considered that the photocurrent is
Figure 7.11: Solar water splitting performance of hematite inverse-opal photoanodes with TTA
upconversion materials. (a) PEC photocurrent curves under simulated solar light. The sam-
ple with TTA+Ag(10 nm) has slightly better performance than other samples. (b) IPCE
with monochromatic incident light in wavelength range of 300 nm-600 nm. Peaks at around
520 nm can be observed for samples with upconversion nanoparticles. The sample with
TTA+Ag(10 nm) has the highest IPCE at 520 nm.
proportional to the area under the IPCE curve. However, the TTA upconversion material
utilized in system has a narrow excitation wavelength range (500 nm-550 nm) and thus the IPCE
is only improved in this range. A large proportion of incident photons at longer wavelengths still
cannot be harvested by hematite, resulting in limited improvement for the overall photocurrent.
A broadband upconversion material or the combination of multiple upconversion materials that
covers a larger spectral range would be more ideal in this scenario. Moreover, the effect of
plasmonic enhanced water splitting is also less significant compared to previous projects using
NHA. This is because, unlike the large enhanced areas in NHA, the LSPR enhanced field of
Ag nanoparticles only exists near their surface and hence only a small amount of hematite can
benefit from the plasmonic enhancement[181]. Although a lot of work are still necessary to be
done before the overall enhancement factor can be further improved for industrial applications,
this project proved the concept of plasmonic enhanced TTA upconversion by Ag nanoparticles
as well as utilizing plasmonic enhanced TTA upconversion in solar water splitting systems.
7.4. Conclusion 129
7.4 Conclusion
In this chapter, TTA@SiO2 core-shell nanoparticles were successfully synthesized and their up-
conversion property was confirmed by PL measurements. Ag-coupled TTA@SiO2 nanoparticles
were synthesized with various sizes of Ag nanoparticles. The best enhancement factor of 1.8
was achieved by using 10 nm Ag nanoparticles. Hematite photoanodes with inverse-opal struc-
ture were fabricated using polystyrene opal templates. After incorporating TTA@SiO2@Ag
upconversion nanoparticles to the hematite photoanodes, the upconversion effect in the solar
water splitting process can be observed. The Ag nanoparticles can enhance both the upconver-
sion process and the solar water splitting process. The sample with TTA@SiO2@Ag (10 nm)
has the best performance and the IPCE enhancement factor at 520 nm was measured to be
61.3 while the enhancement factor of overall photocurrent is only 1.1. The less satisfactory
overall enhancement factor is mainly due to the narrow wavelength range of TTA excitation
and the limited contribution of plasmonic enhancement water splitting by Ag nanoparticles.
This project paved the way for plasmonic enhanced TTA upconversion and the application
of TTA upconversion in solar water splitting systems. Further optimization methods can be
taken, such as replacing Ag nanoparticles with plasmonic nanoarrays for plasmonic enhance-
ment in larger area and utilizing broadband upconversion materials to harvest incident photons
at longer wavelengths.
Chapter 8
Conclusion
130
8.1. Summary of Thesis Achievements 131
8.1 Summary of Thesis Achievements
Artificial solar water splitting is an eco-friendly technique to generate hydrogen fuel as a sustain-
able, abundant and clean energy source. The search for photoactive materials as electrodes in
photoelectrochemical solar water splitting systems has attracted tremendous attention from the
researchers. Hematite has always been a promising photoanode material and exhibits many ad-
vantages such as earth-abundance, cost-efficiency and non-toxicity. However, pristine hematite
photoanodes has poor overall water spitting efficiency due to its low carrier density and high
charge recombination rate. Utilizing enhancement methods is essential in any hematite-based
solar water spitting systems. In this research, I focused on employing plasmonic nanostructures
in the hematite photoanodes for significant field enhancement to improve the water spitting
efficiency.
FDTD modelling was carried out for three distinct 2D plasmonic nanostructures: NDA, NTA
and NHA. These plasmonic nanoarrays can be experimentally fabricated using the colloidal
lithography method. The influence of varying geometrical parameters as well as varying plas-
monic metals have been systematically investigated. It is shown that NDA has LSPR enhanced
field around the edges of the disks while NTA has strong localized field between the tips of the
triangles. NHA exhibits SPP-induced extraordinary transmission and large-area field enhance-
ment inside the holes as well as above the continuous film at different wavelengths. It is also
demonstrated that the most feasible way to tune their plasmonic resonance wavelength is by
controlling their period, which equals to the size of the template spheres used in the colloidal
lithography method. Among all the three nanoarrays, NHA shows its unique advantage of
having plasmonic resonance wavelength in the UV-visible region especially for Ag NHA and
Al NHA with period shorter than 500 nm, making it more favorable in solar water splitting
systems. On the other hand, NDA and NTA are better for applications in the visible-NIR
region. Moreover, the plasmonic peaks of the 2D nanostructures can be further manipulated
by adding another layer of different plasmonic metal. Broadband plasmonic effect can be ob-
served in the dual-layer plasmonic nanoarrays. A systematical database of plasmonic resonance
wavelengths has been established through the modelling framework, making it easier to find
8.1. Summary of Thesis Achievements 132
desired nanoarrays that can be employed in different applications. Based on the modelling
results, two types of plasmonic nanoarrays have been experimentally fabricated: Al nanohole
array and Ag/Au dual-layer nanohole array. The nanoarrays were later utilized in two different
photoanodes based on 1D hematite nanocomposites.
Under the guidance of the FDTD modelling, dual-layer plasmonic nanohole arrays with Ag
and Au were fabricated for plasmonic enhanced hematite photoanodes. A template-less ap-
proach was used to grow hematite nanorods directly on the plasmonic nanoarrays. To tackle
the obstacle of the high charge recombination rate and the low carrier concentration of pristine
hematite, an extra layer of BiVO4 was coated onto the hematite nanorods, forming a hetero-
junction. UV-Vis spectroscopy study indicates a lowered band gap and an extended absorption
region of the BiVO4-coated photoanodes. A five-fold improvement of PEC performance can be
obtained with the help of the heterojunction. Significantly, by applying dual-player plasmonic
nanohole arrays to the photoanode, the performance can be further improved by 2.1 times.
Non-noble plasmonic materials were subsequently investigated. Hematite nanotubes have been
successfully synthesized through a template-less hydrothermal method. XRD and TEM analysis
shows that it has single-crystalline structure. A novel photoanode of hematite nanotubes on
silica-coated Al NHA was then fabricated which has improved performance due the plasmonic
field enhancement from the SPP effect. The Al NHA with 500 nm period shows the highest
enhancement factor with photocurrent density measured to be 3.6 times higher than that of
pristine hematite. Its superior performance is related to the better spectral overlap between its
plasmonic resonance and hematite’s absorption spectrum.
Lastly, in order to further improve the light harvesting efficiency, upconversion materials were
taken into consideration as they can convert multiple low-energy photons to one high-energy
photon. Core-shell TTA@SiO2 upconversion nanoparticles were successfully synthesized and
its upconversion effect from 520 nm to 440 nm has been observed by PL spectrometer with
incident light from xenon lamp. Moreover, Ag nanoparticles of different sizes were coupled
to the upconversion nanoparticles to achieve plasmonic enhanced upconversion. The used Ag
nanoparticles shows LSPR at around 410 nm, overlapping with the emission spectrum of the
8.2. Future Work 133
upconversion nanoparticles. The maximum plasmonic enhancement factor of upconversion

emission is obtained by 10 nm Ag. The TTA@SiO2@Ag nanocomposites were then spin-
coated into a highly porous hematite photoanodes with inverse-opal structure. Benchmarked
by IPCE measurement, the performance of the hematite photoanode with plasmonic enhanced
TTA upconversion nanoparticles is 61.3 times higher than that of pristine hematite at 520 nm.
However, the overall PEC performance has less than 10% improvement due to the limited
excitation wavelength range of the TTA upconversion material.
In summary, this work studied the properties of plasmonic nanodisk arrays, nanotriangle ar-
rays and nanohole arrays. The concept of plasmonic enhanced solar water splitting has been
proved by utilizing plasmonic nanoarrays in various hematite-based photoanodes. Besides, it
is demonstrated that plasmonic enhanced TTA upconversion can provide a promising way to
below-bandgap enhancement for hematite photoanodes. To a greater extent, the potential of
utilizing plasmonic nanostructures as well as plasmonic enhanced TTA upconversion in solar
water splitting systems has been revealed. Plasmonic nanostructures will play a more important
role in designing highly-efficient commercial solar water splitting devices.
8.2 Future Work
The modelling part of this work only investigated three types of 2D plasmonic nanoarrays while
there are much more plasmonic nanostructures that can potentially be employed in solar water
splitting systems. A comprehensive FDTD modelling can be carried out to cover more of these
fascinating materials. Besides, non-metal plasmonic materials have also been an interesting field
that can be potentially explored using similar approaches and may also provide new possibilities
for plasmonic enhanced solar water splitting.
In the experimental part, the solar water splitting performance is clearly enhanced by the
plasmonic nanostructures. However, to gain more insight of the enhancing pathways and the
mechanisms, advanced characterization methods such as femtosecond transient absorption spec-
troscopy can be utilized to discover the details of energy transfer in the plasmonic enhanced
8.2. Future Work 134
solar water splitting process. Besides, high overall solar-to-hydrogen efficiency is essential for
the commercial applications but it is not likely to achieve the state-of-the-art performance only
by plasmonic enhancement. This project utilized nanostructuring, surface modification as well
as upconversion in hematite photoanodes but other enhancement strategies such as doping can
also be synergistically applied to hematite nanocomposites and are definitely worth further
exploring. Furthermore, the search for highly-efficient solar water splitting systems shouldn’t
be limited to hematite-based systems. There are other photoactive semiconductors besides
hematite that can be used as photoanodes, photocathodes as well as photocatalysts. Imple-
menting plasmonic enhancement in those systems allows great opportunities for significant
performance improvement.
For the hematite photoanodes with TTA upconversion materials, only Ag nanoparticles are
employed for plasmonic enhancement. It is worthy to employ plasmonic nanoarrays in the
system and find out how plasmonic nanoarrays works together with TTA upconversion ma-
terials. Moreover, although the below-bandgap IPCE enhancement factor is high, the overall
performance is far from satisfactory. As discussed, the narrow excitation wavelength range of
the utilized TTA upconversion TTA upconversion material makes it difficult to harvest more
photons. Different types of upconversion materials can be employed together to achieve broad-
band upconversion and maximize the performance especially when coupled with plasmonic
nanostructures with broadband enhancement.
Appendix A
Source Code of the FDTD Models
135
A.1. Nanohole Array and Nanotriangle Array 136
A.1 Nanohole Array and Nanotriangle Array

import meep a s mp
import numpy a s np
import m a t p l o t l i b
import m a t p l o t l i b . p y p l o t a s plt
import math
def s a v e T r a n s T o F i l e ( f i l e n a m e , wavelength , transmission ) :

f = open ( f i l e n a m e , ’w ’ )
for i i n range ( l e n ( w a v e l e n g t h ) ) :
f . w r i t e ( str ( wavelength [ i ] ) + ’ , ’ + s t r ( t r a n s m i s s i o n [ i ] ) + ’ \n ’ )
f . close ()
def f r a n g e ( s t a r t , stop , step ) :

i = start
while i < stop :
yield i
i += s t e p
# conversion factor f o r eV t o 1/um [=1/ h c ]
eV um scale = 1/1.23984193
#−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−
Al plasma frq = 14.98∗ eV um scale

Al f0 = 0.523
A l f r q 0 = 1 e −10
Al gam0 = 0 . 0 4 7 ∗ e V u m s c a l e
A l s i g 0 = A l f 0 ∗ A l p l a s m a f r q ∗∗2/ A l f r q 0 ∗∗2
Al f1 = 0.227
A l fr q 1 = 0.162∗ eV um scale # 7 . 6 5 4 um
Al gam1 = 0 . 3 3 3 ∗ e V u m s c a l e
Al f2 = 0.050
Al gam2 = 0 . 3 1 2 ∗ e V u m s c a l e
Al f3 = 0.166
Al gam3 = 1 . 3 5 1 ∗ e V u m s c a l e
Al f4 = 0.030
Al gam4 = 3 . 3 8 2 ∗ e V u m s c a l e
A l s u s c = [ mp . D r u d e S u s c e p t i b i l i t y ( f r e q u e n c y=A l f r q 0 , gamma=Al gam0 , si g m a=A l s i g 0 ) ,

mp . L o r e n t z i a n S u s c e p t i b i l i t y ( f r e q u e n c y=A l f r q 1 , gamma=Al gam1 , si g m a=A l s i g 1 ) ,
mp . L o r e n t z i a n S u s c e p t i b i l i t y ( f r e q u e n c y=A l f r q 4 , gamma=Al gam4 , si g m a=A l s i g 4 ) ]
Al = mp . Medium ( e p s i l o n = 1 . 0 , E s u s c e p t i b i l i t i e s =A l s u s c )
#−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−
def runNHA ( d i s t a n c e , diameter , thickness , resolution , fcen , df , n f r e q , mat , f i l e p r e f i x ) : #um

DISTANCE = d i s t a n c e #um
DIAMETER = d i a m e t e r #um
THICKNESS = t h i c k n e s s #um
PADDING = 1 . 0 #um
DPML = 0 . 4 #um
c e l l = mp . V e c t o r 3 (DISTANCE, DISTANCE ∗ math . s q r t ( 3 ) , THICKNESS + 2 ∗ PADDING + 2 ∗ DPML)
#d e f i n e geometry
s l a b = mp . B l o c k (mp . V e c t o r 3 ( 1 e20 , 1 e20 , THICKNESS ) , c e n t e r=mp . V e c t o r 3 ( 0 , 0, 0) , m a t e r i a l=mat )
h o l e 1 = mp . C y l i n d e r (
r a d i u s=DIAMETER/ 2 ,
c e n t e r=mp . V e c t o r 3 ( 0 , 0, 0) ,
h e i g h t =1e 2 0
)
h o l e 2 = mp . C y l i n d e r ( r a d i u s=DIAMETER/ 2 ,
c e n t e r=mp . V e c t o r 3 (DISTANCE/ 2 , DISTANCE ∗ math . s q r t ( 3 ) / 2 , 0) ,
h e i g h t =1e 2 0
)
c e n t e r=mp . V e c t o r 3 (−DISTANCE/ 2 , DISTANCE ∗ math . s q r t ( 3 ) / 2 , 0) ,
h e i g h t =1e 2 0
)
c e n t e r=mp . V e c t o r 3 (DISTANCE/ 2 , −DISTANCE ∗ math . s q r t ( 3 ) / 2 , 0) ,
h e i g h t =1e 2 0
)
c e n t e r=mp . V e c t o r 3 (−DISTANCE/ 2 , −DISTANCE ∗ math . s q r t ( 3 ) / 2 , 0) ,
h e i g h t =1e 2 0
)
geometry = [ slab , hole1 , hole2 , hole3 , hole4 , hole5 ]
#d e f i n e Gaussian p l a n e wave
sources = [
mp . S o u r c e (
mp . G a u s s i a n S o u r c e ( f c e n , f w i d t h=d f ) ,
component=mp . Ex ,
c e n t e r=mp . V e c t o r 3 ( 0 , 0, −0.7) ,
s i z e=mp . V e c t o r 3 (DISTANCE, DISTANCE ∗ math . s q r t ( 3 ) , 0 ) ,
)
]
t r a n f r = mp . F l u x R e g i o n (
c e n t e r=mp . V e c t o r 3 ( 0 , 0, 0.7) ,
s i z e=mp . V e c t o r 3 (DISTANCE, DISTANCE ∗ math . s q r t ( 3 ) , 0 )
)
#d e f i n e pml layers
p m l l a y e r s = [ mp .PML( t h i c k n e s s=DPML, d i r e c t i o n=mp . Z ) ]
#s e t sim and sim ref and r u n

sim = mp . S i m u l a t i o n ( c e l l s i z e =c e l l ,
b o u n d a r y l a y e r s=p m l l a y e r s ,
g e o m e t r y=geometry ,
s o u r c e s=s o u r c e s ,
r e s o l u t i o n=r e s o l u t i o n ,
k p o i n t=mp . V e c t o r 3 ( 0 , 0 , 0 ) )
dftsXY = sim . a d d d f t f i e l d s ( [ mp . Ex ] , fcen , df , nfreq ,
where=mp . Volume ( c e n t e r=mp . V e c t o r 3 ( 0 , 0 , −THICKNESS/ 2 ) ,
s i z e=mp . V e c t o r 3 (DISTANCE, DISTANCE ∗ math . s q r t ( 3 ) ) )
)
d f t s X Z = sim . a d d d f t f i e l d s ( [ mp . Ex ] , fcen , df , nfreq ,

where=mp . Volume ( c e n t e r=mp . V e c t o r 3 ( 0 , 0 , 0 ) ,
s i z e=mp . V e c t o r 3 (DISTANCE, 0 , THICKNESS + PADDING) )
)
d f t s Y Z = sim . a d d d f t f i e l d s ( [ mp . Ex ] , fcen , df , nfreq ,
s i z e=mp . V e c t o r 3 ( 0 , DISTANCE ∗ math . s q r t ( 3 ) , THICKNESS + PADDING) )
)
t r a n = sim . a d d f l u x ( f c e n , df , nfreq , tran fr )
s i m r e f = mp . S i m u l a t i o n ( c e l l s i z e =c e l l ,
geometry = [ ] ,
k p o i n t=mp . V e c t o r 3 ( 0 , 0 , 0 ) )
t r a n r e f = s i m r e f . add flux ( fcen , df , nfreq , tran fr )
sim . run ( u n t i l a f t e r s o u r c e s =mp . s t o p w h e n f i e l d s d e c a y e d ( 5 ,mp . Ex , mp . V e c t o r 3 ( ) , 1 e −5))
s i m r e f . run ( u n t i l a f t e r s o u r c e s =mp . s t o p w h e n f i e l d s d e c a y e d ( 5 ,mp . Ex , mp . V e c t o r 3 ( ) , 1 e −5))
sim . o u t p u t d f t ( dftsXY ,
fileprefix +
’ XY Pit ’ + s t r ( ’ %.4 f ’ % d i s t a n c e ) +
’ Dia ’ + s t r ( ’ %.4 f ’ % d i a m e t e r ) +
’ Thi ’+ s t r ( t h i c k n e s s )
)
sim . o u t p u t d f t ( dftsXZ ,
fileprefix +
’ XZ Pit ’ + s t r ( ’ %.4 f ’ % d i s t a n c e ) +
)
sim . o u t p u t d f t ( dftsYZ ,
fileprefix +
’ YZ Pit ’ + s t r ( ’ %.4 f ’ % d i s t a n c e ) +
)
t r a n d a t a = mp . g e t f l u x e s ( t r a n )
t r a n r e f d a t a = mp . g e t f l u x e s ( t r a n r e f )
t r a n f r e q = mp . g e t f l u x f r e q s ( t r a n )
wl = [ ]
Ts = [ ]
for i i n range ( n f r e q ) :
wl = np . append ( wl , 1/ t r a n f r e q [ i ] )
Ts = np . append ( Ts , t r a n d a t a [ i ] / t r a n r e f d a t a [ i ] )
plt . c l f ()
p l t . p l o t ( wl , Ts )
plt . savefig ( f i l e p r e f i x +
’ t r a n s P i t ’ + s t r ( ’ %.4 f ’ % d i s t a n c e ) +
’ Thi ’+ s t r ( t h i c k n e s s ) + ’ . png ’
)
saveTransToFile ( f i l e p r e f i x +
’ t r a n s P i t ’ + s t r ( ’ %.4 f ’ % d i s t a n c e ) +
’ Thi ’+ s t r ( t h i c k n e s s ) +
’ . t x t ’ , wl , Ts
)
for pitchlength in f r a n g e ( 0 . 2 , 1.0 , 0.02):

for etchratio in f r a n g e ( 0 . 2 , 0.8 , 0.1):
runNHA ( d i s t a n c e=p i t c h l e n g t h ,
d i a m e t e r=p i t c h l e n g t h ∗ e t c h r a t i o ,
t h i c k n e s s =0.05 ,
r e s o l u t i o n =250 ,
f c e n =2 ,
d f =3 ,
n f r e q =100 ,
mat=Al ,
f i l e p r e f i x = ’ AlNHA ’
)
A.2. Nanohole Array with Oxidation Layer 140
A.2 Nanohole Array with Oxidation Layer

import meep a s mp
import numpy a s np
import math
from m a t e r i a l s import M a t e r i a l s
def s a v e T r a n s T o F i l e ( f i l e n a m e , wavelength , transmission ) :

f = open ( f i l e n a m e , ’w ’ )
for i i n range ( l e n ( w a v e l e n g t h ) ) :
f . w r i t e ( str ( wavelength [ i ] ) + ’ , ’ + s t r ( t r a n s m i s s i o n [ i ] ) + ’ \n ’ )
f . close ()
def f r a n g e ( s t a r t , stop , step ) :

i = start
while i < stop :
yield i
i += s t e p
def runNHA ( d i s t a n c e , diameter , thickness , othickness , resolution , fcen , df , nfreq , f i l e p r e f i x ) : #um

DISTANCE = d i s t a n c e #um
THICKNESS = t h i c k n e s s #um
OTHICKNESS = o t h i c k n e s s #um
PADDING = 1 . 0 #um
DPML = 0 . 7 5 #um
c e l l = mp . V e c t o r 3 (DISTANCE, DISTANCE ∗ math . s q r t ( 3 ) , THICKNESS + 2 ∗ PADDING + 2 ∗ DPML)
f i l m m e t a l = mp . B l o c k (
mp . V e c t o r 3 ( 1 e20 , 1 e20 , THICKNESS ) ,
c e n t e r=mp . V e c t o r 3 ( 0 , 0, 0) ,
m a t e r i a l=M a t e r i a l s . g e t A l ( )
)
s u b s t r a t e = mp . B l o c k (
mp . V e c t o r 3 ( 1 e20 , 1 e20 , 2.0) ,
c e n t e r=mp . V e c t o r 3 ( 0 , 0, 1 . 0 + THICKNESS/ 2 ) ,
m a t e r i a l=M a t e r i a l s . g e t S i O 2 ( )
)
c e n t e r=mp . V e c t o r 3 ( 0 , 0, 0) ,
h e i g h t =1e 2 0
)
h e i g h t =1e 2 0
)
c e n t e r=mp . V e c t o r 3 (−DISTANCE/ 2 ,
DISTANCE ∗ math . s q r t ( 3 ) / 2 , 0) ,
h e i g h t =1e 2 0
)
h e i g h t =1e 2 0
)
h e i g h t =1e 2 0
)
o h o l e 1 = mp . C y l i n d e r (
r a d i u s =(DIAMETER/2)+ o t h i c k n e s s ,
c e n t e r=mp . V e c t o r 3 ( 0 , 0, 0) ,
h e i g h t=THICKNESS ,
m a t e r i a l=M a t e r i a l s . g e t A l 2 O 3 ( )
)
o h o l e 2 = mp . C y l i n d e r ( r a d i u s =(DIAMETER/2)+ o t h i c k n e s s ,
)
c e n t e r=mp . V e c t o r 3 (−DISTANCE/ 2 , DISTANCE ∗ math . s q r t ( 3 ) / 2 , 0) ,
)
)
)
f i l m o x i d e = mp . B l o c k (mp . V e c t o r 3 ( 1 e20 , 1 e20 , OTHICKNESS) ,

c e n t e r=mp . V e c t o r 3 ( 0 , 0, −0.5∗THICKNESS+0.5∗OTHICKNESS) , m a t e r i a l=M a t e r i a l s . g e t A l 2 O 3 ( )
)
geometry = [ f i l m m e t a l , film oxide ,

ohole1 , ohole2 , ohole3 , ohole4 , ohole5 ,
hole1 , hole2 , hole3 , hole4 , hole5
]
sources = [
mp . S o u r c e (
component=mp . Ex ,
c e n t e r=mp . V e c t o r 3 ( 0 , 0, −0.7) ,
s i z e=mp . V e c t o r 3 (DISTANCE, DISTANCE ∗ math . s q r t ( 3 ) , 0 ) ,
)
]
t r a n f r = mp . F l u x R e g i o n (
c e n t e r=mp . V e c t o r 3 ( 0 , 0, 0.7) ,
s i z e=mp . V e c t o r 3 (DISTANCE, DISTANCE ∗ math . s q r t ( 3 ) , 0 )
)
p m l l a y e r s = [ mp .PML( t h i c k n e s s=DPML, d i r e c t i o n=mp . Z ) ]

sim = mp . S i m u l a t i o n ( c e l l s i z e =c e l l ,
k p o i n t=mp . V e c t o r 3 ( 0 , 0 , 0 ) )
dftsXY = sim . a d d d f t f i e l d s ( [ mp . Ex , mp . Ey , mp . Ez ] ,
fcen , df , nfreq ,
where=mp . Volume ( c e n t e r=mp . V e c t o r 3 ( 0 , 0 , −THICKNESS/ 2 ) ,
s i z e=mp . V e c t o r 3 (DISTANCE, DISTANCE ∗ math . s q r t ( 3 ) ) )
)
d f t s X Z = sim . a d d d f t f i e l d s ( [ mp . Ex , mp . Ey , mp . Ez ] ,
fcen , df , nfreq ,
s i z e=mp . V e c t o r 3 (DISTANCE, 0 , THICKNESS + PADDING) )
)
d f t s Y Z = sim . a d d d f t f i e l d s ( [ mp . Ex , mp . Ey , mp . Ez ] ,
fcen , df , nfreq ,
s i z e=mp . V e c t o r 3 ( 0 , DISTANCE ∗ math . s q r t ( 3 ) , THICKNESS + PADDING) )
)
t r a n = sim . a d d f l u x ( f c e n , df , nfreq , tran fr )
s i m r e f = mp . S i m u l a t i o n ( c e l l s i z e =c e l l ,
geometry = [ ] ,
k p o i n t=mp . V e c t o r 3 ( 0 , 0 , 0 )
)
t r a n r e f = s i m r e f . add flux ( fcen , df , nfreq , tran fr )
sim . run ( u n t i l a f t e r s o u r c e s =mp . s t o p w h e n f i e l d s d e c a y e d ( 5 ,mp . Ex , mp . V e c t o r 3 ( ) , 1 e −5))
s i m r e f . run ( u n t i l a f t e r s o u r c e s =mp . s t o p w h e n f i e l d s d e c a y e d ( 5 ,mp . Ex , mp . V e c t o r 3 ( ) , 1 e −5))
sim . o u t p u t d f t ( dftsXY ,
fileprefix +
’ XY ’ + s t r ( ’ %.0 f ’ % ( d i s t a n c e ∗ 1 0 0 0 ) ) +
’ ’ + s t r ( ’ %.0 f ’ % ( d i a m e t e r ∗ 1 0 0 0 ) ) +
’ ’ + s t r ( ’ %.0 f ’ % ( t h i c k n e s s ∗ 1 0 0 0 ) ) +
’ ’ + s t r ( ’ %.0 f ’ % ( o t h i c k n e s s ∗ 1 0 0 0 ) )
)
sim . o u t p u t d f t ( dftsXZ ,
fileprefix +
’ XZ ’ + s t r ( ’ %.0 f ’ % ( d i s t a n c e ∗ 1 0 0 0 ) ) +
’ ’ + s t r ( ’ %.0 f ’ % ( d i a m e t e r ∗ 1 0 0 0 ) ) +
’ ’ + s t r ( ’ %.0 f ’ % ( t h i c k n e s s ∗ 1 0 0 0 ) ) +
’ ’ + s t r ( ’ %.0 f ’ % ( o t h i c k n e s s ∗ 1 0 0 0 ) )
)
sim . o u t p u t d f t ( dftsYZ ,
fileprefix +
’ YZ ’ + s t r ( ’ %.0 f ’ % ( d i s t a n c e ∗ 1 0 0 0 ) ) +
’ ’ + s t r ( ’ %.0 f ’ % ( d i a m e t e r ∗ 1 0 0 0 ) ) +
’ ’ + s t r ( ’ %.0 f ’ % ( t h i c k n e s s ∗ 1 0 0 0 ) ) +
’ ’ + s t r ( ’ %.0 f ’ % ( o t h i c k n e s s ∗ 1 0 0 0 ) )
)
t r a n d a t a = mp . g e t f l u x e s ( t r a n )
t r a n r e f d a t a = mp . g e t f l u x e s ( t r a n r e f )
t r a n f r e q = mp . g e t f l u x f r e q s ( t r a n )
wl = [ ]
Ts = [ ]
for i i n range ( n f r e q ) :
wl = np . append ( wl , 1/ t r a n f r e q [ i ] )
Ts = np . append ( Ts , t r a n d a t a [ i ] / t r a n r e f d a t a [ i ] )
plt . c l f ()
p l t . p l o t ( wl , Ts )
plt . savefig ( f i l e p r e f i x +
’ trans ’ + s t r ( ’ %.0 f ’ % ( d i s t a n c e ∗ 1 0 0 0 ) ) +
’ ’ + s t r ( ’ %.0 f ’ % ( d i a m e t e r ∗ 1 0 0 0 ) ) +
’ ’+ s t r ( ’ %.0 f ’ % ( t h i c k n e s s ∗ 1 0 0 0 ) ) +
’ ’ + s t r ( ’ %.0 f ’ % ( o t h i c k n e s s ∗ 1 0 0 0 ) ) + ’ . png ’
)
saveTransToFile ( f i l e p r e f i x +
’ trans ’ + s t r ( ’ %.0 f ’ % ( d i s t a n c e ∗ 1 0 0 0 ) ) +
’ ’ + s t r ( ’ %.0 f ’ % ( d i a m e t e r ∗ 1 0 0 0 ) ) +
’ ’+ s t r ( ’ %.0 f ’ % ( t h i c k n e s s ∗ 1 0 0 0 ) ) +
’ ’ + s t r ( ’ %.0 f ’ % ( o t h i c k n e s s ∗ 1 0 0 0 ) ) + ’ . t x t ’
, wl , Ts
)
for othickness in f r a n g e ( 0 . 0 1 , 0.09 , 0.01):

runNHA ( d i s t a n c e = 0 . 5 ,
diameter =0.4 ,
t h i c k n e s s =0.08 ,
o t h i c k n e s s=o t h i c k n e s s ,
r e s o l u t i o n =200 , f c e n =2 ,
d f =3 ,
n f r e q =150 ,
f i l e p r e f i x = ’ AlNHAO ’
)
A.3. Nanodisk Array 144
A.3 Nanodisk Array

import meep a s mp
from math import s q r t , pi
import numpy a s np
r e s o l u t i o n =250 # p i x e l /um
dpml =0.5 #um
h =0.05 #um
a i r h e i g h t =0.7 #um
d =0.4 #um
r=d /2
f i n a l d i a m e t e r= 0 . 3 1 5
f c e n =5 #
d f =3
dpml =0.4 #um
#d e f i n e computational cell
x=d
y=d∗ s q r t ( 3 )
z=a i r h e i g h t ∗2+h+2∗dpml
c e l l s =mp . V e c t o r 3 ( x , y , z )
g l a s s=mp . B l o c k (mp . V e c t o r 3 ( 1 e20 , 1 e20 , 1 0 0 ) ,
c e n t e r=mp . V e c t o r 3 ( 0 , 0 , 5 0 + h / 2 ) ,
m a t e r i a l=mp . Medium ( e p s i l o n = 1 . 5 )
)
c e n t e r d i s k=mp . C y l i n d e r ( r a d i u s=f i n a l d i a m e t e r / 2 ,
c e n t e r=mp . V e c t o r 3 ( 0 , 0 , 0 ) ,
h e i g h t=h ,
m a t e r i a l=M a t e r i a l s . g e t A u ( )
)
d i s k 1=mp . C y l i n d e r ( r a d i u s=f i n a l d i a m e t e r / 2 ,
c e n t e r=mp . V e c t o r 3 ( r , s q r t ( 3 ) ∗ r , 0 ) ,
h e i g h t=h ,
)
c e n t e r=mp . V e c t o r 3 (−r , s q r t ( 3 ) ∗ r , 0 ) ,
h e i g h t=h ,
)
c e n t e r=mp . V e c t o r 3 (−r ,− s q r t ( 3 ) ∗ r , 0 ) ,
h e i g h t=h ,
)
c e n t e r=mp . V e c t o r 3 ( r ,− s q r t ( 3 ) ∗ r , 0 ) ,
h e i g h t=h ,
)
g e o m e t r y =[ g l a s s , c e n t e r d i s k , d i s k 1 , d i s k 2 , d i s k 3 , d i s k 4 ]
s o u r c e =[mp . S o u r c e (
A.3. Nanodisk Array 145
component=mp . Ex ,
c e n t e r=mp . V e c t o r 3 ( 0 , 0 , − 0 . 4 ) ,
s i z e=mp . V e c t o r 3 ( x , y , 0 ) ) ]
p m l l a y e r s =[mp .PML( t h i c k n e s s=dpml , d i r e c t i o n=mp . Z ) ]
sim=mp . S i m u l a t i o n ( c e l l s i z e =c e l l s ,
s o u r c e s=s o u r c e ,
k p o i n t=mp . V e c t o r 3 ( 0 , 0 , 0 )
)
sim0=mp . S i m u l a t i o n ( c e l l s i z e =c e l l s ,
g e o m e t r y =[ g l a s s ] ,
s o u r c e s=s o u r c e ,
k p o i n t=mp . V e c t o r 3 ( 0 , 0 , 0 )
)
#T r a n s m i s s i o n
f r e g = mp . F l u x R e g i o n ( c e n t e r=mp . V e c t o r 3 ( 0 , 0 , 0 . 5 ) , s i z e=mp . V e c t o r 3 ( x , y , 0 ) )
n f r e q =300
t r a n s = sim . a d d f l u x ( f c e n , df , nfreq , freg )
t r a n s 0=sim0 . a d d f l u x ( f c e n , df , nfreq , freg )
##r u n simulation
sim . run ( u n t i l =100)
sim0 . run ( u n t i l =100)
e p s d a t a = sim . g e t a r r a y ( c e n t e r=mp . V e c t o r 3 ( 0 , 0 , 0 ) , s i z e=mp . V e c t o r 3 ( x , y , 0 ) , component=mp . D i e l e c t r i c )
e x d a t a = sim . g e t a r r a y ( c e n t e r=mp . V e c t o r 3 ( 0 , 0 , − h / 2 ) , s i z e=mp . V e c t o r 3 ( x , y , 0 ) , component=mp . Ex )
#p l o t
p l t . f i g u r e ( d p i =100)
p l t . imshow ( e p s d a t a . t r a n s p o s e ( ) , i n t e r p o l a t i o n= ’ s p l i n e 3 6 ’ , cmap= ’ b i n a r y ’ )
plt . axis ( ’ off ’ )
p l t . s a v e f i g ( ” . / e p s . png ” )
p l t . f i g u r e ( d p i =100)
p l t . imshow ( e x d a t a . t r a n s p o s e ( ) , i n t e r p o l a t i o n= ’ s p l i n e 3 6 ’ , cmap= ’RdBu ’ , a l p h a = 0 . 9 )
plt . axis ( ’ off ’ )
p l t . s a v e f i g ( ” . / Ex . png ” )
#o u t p u t flux data
sim . d i s p l a y f l u x e s ( t r a n s )
sim0 . d i s p l a y f l u x e s ( t r a n s 0 )
A.4. Nanostars and Nanospheres 146
A.4 Nanostars and Nanospheres

import meep a s mp
import numpy a s np
import math , random
def f i b o n a c c i s p h e r e ( s a m p l e s =1 , r a d i u s =1 , r a n d o m i z e=F a l s e ) :
rnd = 1 .
if randomize :
rnd = random . random ( ) ∗ s a m p l e s
points = [ ]
o f f s e t = 2./ samples
i n c r e m e n t = math . p i ∗ ( 3 . − math . s q r t ( 5 . ) ) ;
for i i n range ( s a m p l e s ) :
y = (( i ∗ o f f s e t ) − 1) + ( o f f s e t / 2);
r = math . s q r t ( 1 − pow( y , 2 ) )
p h i = ( ( i + rnd ) % s a m p l e s ) ∗ i n c r e m e n t
x = math . c o s ( p h i ) ∗ r
z = math . s i n ( p h i ) ∗ r
p o i n t s . append ( [ x ∗ r a d i u s , y ∗ r a d i u s , z∗ radius ] )
return p o i n t s
def runCSNS ( c o r e D i a m e t e r , diameter , tipCount , tipTheta ,

tipLength , resolution , fcen , df , nfreq , fileprefix
) : #um
COREDIAMETER = c o r e D i a m e t e r #um
TIPCOUNT = t i p C o u n t #um
TIPTHETA = t i p T h e t a #r a d
TIPLENGTH = t i p L e n g t h #um
PADDING = 0 . 3 #um
ZPADDING = 0 . 7 #um
DPML = 0 . 7 5 #um
a = DIAMETER + 2 ∗ TIPLENGTH + 2 ∗ PADDING + 2 ∗ DPML #s i z e o f FDTD r e g i o n

c = DIAMETER + 2 ∗ TIPLENGTH + 2 ∗ ZPADDING + 2 ∗ DPML
c e l l = mp . V e c t o r 3 ( a , a , c)
r = DIAMETER / 2
rRoot = math . s q r t ( r ∗∗ 2 − (TIPLENGTH ∗ math . t a n (TIPTHETA) ) ∗∗ 2 )
r o o t s = f i b o n a c c i s p h e r e ( tipCount , rRoot , False )
cones = [ ]
for r o o t in roots :
xRoot = r o o t [ 0 ]
yRoot = r o o t [ 1 ]
zRoot = r o o t [ 2 ]
p r i n t ( s t r ( xRoot ) + ” , ” + s t r ( yRoot ) + ” , ” + s t r ( zRoot ) )
x C e n t e r = ( 1 + TIPLENGTH / ( 2 ∗ rRoot ) ) ∗ xRoot
y C e n t e r = ( 1 + TIPLENGTH / ( 2 ∗ rRoot ) ) ∗ yRoot

z C e n t e r = ( 1 + TIPLENGTH / ( 2 ∗ rRoot ) ) ∗ zRoot
coneToAdd = mp . Cone ( r a d i u s =(TIPLENGTH ∗ math . t a n (TIPTHETA ) ) ,
r a d i u s 2 =0 ,
a x i s=mp . V e c t o r 3 ( xRoot , yRoot , zRoot ) ,
c e n t e r=mp . V e c t o r 3 ( xCenter , yCenter , zCenter ) ,
h e i g h t=TIPLENGTH,
)
c o n e s . append ( coneToAdd )
m e t a l S p h e r e = mp . S p h e r e ( r a d i u s=r ,
c e n t e r=mp . V e c t o r 3 ( 0 , 0, 0) ,
)
geometry = [ metalSphere , ∗ cones ]
sources = [
mp . S o u r c e (
component=mp . Ex ,
c e n t e r=mp . V e c t o r 3 ( 0 , 0 , −1 ∗ ( r + TIPLENGTH + 0 . 5 ∗ ZPADDING) ) ,
s i z e=mp . V e c t o r 3 ( a , a , 0) ,
)
]
p m l l a y e r s = [ mp .PML( t h i c k n e s s=DPML) ]

sim = mp . S i m u l a t i o n ( e p s a v e r a g i n g=F a l s e ,
c e l l s i z e =c e l l ,
r e s o l u t i o n=r e s o l u t i o n
)
d f t S i z e = DIAMETER + 2 ∗ TIPLENGTH + PADDING

dftsXY = sim . a d d d f t f i e l d s ( [ mp . Ex , mp . Ey , mp . Ez ] ,
fcen , df , nfreq ,
c e n t e r=mp . V e c t o r 3 ( 0 , 0 , 0 ) ,
s i z e=mp . V e c t o r 3 ( d f t S i z e , dftSize )
)
d f t s X Z = sim . a d d d f t f i e l d s ( [ mp . Ex , mp . Ey , mp . Ez ] ,
fcen , df , nfreq ,
c e n t e r=mp . V e c t o r 3 ( 0 , 0 , 0 ) ,
s i z e=mp . V e c t o r 3 ( d f t S i z e , 0, dftSize )
)
d f t s Y Z = sim . a d d d f t f i e l d s ( [ mp . Ex , mp . Ey , mp . Ez ] ,
fcen , df , nfreq ,
c e n t e r=mp . V e c t o r 3 ( 0 , 0 , 0 ) ,
s i z e=mp . V e c t o r 3 ( 0 , dftSize , dftSize )
)
d f t s A l l = sim . a d d d f t f i e l d s ( [ mp . Ex , mp . Ey , mp . Ez ] ,
fcen , df , nfreq ,
c e n t e r=mp . V e c t o r 3 ( 0 , 0 , 0 ) ,
s i z e=mp . V e c t o r 3 ( d f t S i z e , dftSize , dftSize )
)
sim . run ( u n t i l a f t e r s o u r c e s =mp . s t o p w h e n f i e l d s d e c a y e d ( 5 ,

mp . Ex ,
mp . V e c t o r 3 ( 0 , 0 , r + TIPLENGTH + 0 . 5 ∗ ZPADDING) , 1 e −5))
p r i n t ( ” S a v i n g DFT s l i c e s ” )
sim . o u t p u t d f t ( dftsXY , f i l e p r e f i x + ’ XY ’ )
sim . o u t p u t d f t ( dftsXZ , f i l e p r e f i x + ’ XZ ’ )
sim . o u t p u t d f t ( dftsYZ , f i l e p r e f i x + ’ YZ ’ )
p r i n t ( ” S a v i n g DFT v o l u m e s ” )
sim . o u t p u t d f t (
dftsAll , f i l e p r e f i x + ’ All ’ ,
c e n t e r=mp . V e c t o r 3 ( 0 , 0 , 0 ) ,
s i z e=mp . V e c t o r 3 ( d f t S i z e , dftSize , dftSize )
)
runCSNS ( c o r e D i a m e t e r = 0 . 0 ,
diameter =0.214 ,
t i p C o u n t =16 ,
tipTheta =0.11 ,
tipLength =0.060 ,
r e s o l u t i o n =100 ,
f c e n =2 ,
d f =3 ,
n f r e q =100 ,
f i l e p r e f i x = ’ SolidAuNS ’
)
Appendix B
Supporting Information
149
Table B.1: Permissions Summary
Copyright Permission
Page Type I have
Name of work Source of work holder requested Note
No. of work permission
and contact on
Alexandra Boltasseva
Map of and Harry A Atwater.
©AAAS for
31 figure common plasmonic Low-loss plasmonic metamaterials. 02.09.20 yes
www.aaas.org thesis/dissertation
materials Science,
331(6015):290–291, 2011
Wenbo Hou and
Absorption spectra Stephen B Cronin.
of Fe2O3 and TiO2. A review of surface plasmon ©Wiley for
38 figure 02.09.20 yes
Energy bands of resonance-enhanced photocatalysis www.wiley.com thesis/dissertation
semiconductors. Advanced Functional Materials,
23(13):1612–1619, 2013.
150
Bibliography
[1] Aleksandar D Rakić, Aleksandra B Djurišić, Jovan M Elazar, and Marian L Majewski.
Optical properties of metallic films for vertical-cavity optoelectronic devices. Applied
optics, 37(22):5271–5283, 1998.
[2] Alexandra Boltasseva and Harry A Atwater. Low-loss plasmonic metamaterials. Science,
331(6015):290–291, 2011.
[3] Kazuhiko Maeda and Kazunari Domen. Photocatalytic water splitting: recent progress
and future challenges. The Journal of Physical Chemistry Letters, 1(18):2655–2661, 2010.
[4] Zhebo Chen, Huyen N. Dinh, and Eric Miller. Photoelectrochemical Water Splitting.
Springer, 2013.
[5] Wenbo Hou and Stephen B Cronin. A review of surface plasmon resonance-enhanced
photocatalysis. Advanced Functional Materials, 23(13):1612–1619, 2013.
[6] O Coddington, JL Lean, P Pilewskie, M Snow, and D Lindholm. A solar irradiance

climate data record. Bulletin of the American Meteorological Society, 97(7):1265–1282,
2016.
[7] Thomas B Johansson, Anand Prabhakar Patwardhan, Nebojša Nakićenović, and Luis
Gomez-Echeverri. Global energy assessment: toward a sustainable future. Cambridge
University Press, 2012.
[8] Nathan S Lewis, George Crabtree, Arthur J Nozik, Michael R Wasielewski, Paul
Alivisatos, Harriet Kung, J Tsao, Elaine Chandler, Wladek Walukiewicz, Mark Spitler,
151
BIBLIOGRAPHY 152
et al. Basic research needs for solar energy utilization. report of the basic energy sci-
ences workshop on solar energy utilization, april 18-21, 2005. Technical report, DOESC
(USDOE Office of Science (SC)), 2005.
[9] Michael G Walter, Emily L Warren, James R McKone, Shannon W Boettcher, Qixi Mi,
Elizabeth A Santori, and Nathan S Lewis. Solar water splitting cells. Chemical reviews,
110(11):6446–6473, 2010.
[10] Sebastian Verhelst and Roger Sierens. Aspects concerning the optimisation of a hydrogen
fueled engine. International journal of hydrogen energy, 26(9):981–985, 2001.
[11] David L Trimm and Z Ilsen Önsan. Onboard fuel conversion for hydrogen-fuel-cell-driven
vehicles. Catalysis Reviews, 43(1-2):31–84, 2001.
[12] Nikola Getoff. Photoelectrochemical and photocatalytic methods of hydrogen production:

a short review. International Journal of Hydrogen Energy, 15(6):407–417, 1990.
[13] Chiara Valsecchi and Alexandre G Brolo. Periodic metallic nanostructures as plasmonic
chemical sensors. Langmuir, 29(19):5638–5649, 2013.
[14] JR Mejı́a-Salazar and Osvaldo N Oliveira Jr. Plasmonic biosensing: focus review. Chem-
ical reviews, 118(20):10617–10625, 2018.
[15] Jing-Liang Li and Min Gu. Surface plasmonic gold nanorods for enhanced two-photon
microscopic imaging and apoptosis induction of cancer cells. Biomaterials, 31(36):9492–
9498, 2010.
[16] Peng Zhang, Tuo Wang, and Jinlong Gong. Mechanistic understanding of the plasmonic
enhancement for solar water splitting. Advanced Materials, 27(36):5328–5342, 2015.
[17] Stefan Alexander Maier. Plasmonics: fundamentals and applications. Springer Science
& Business Media, 2007.
[18] W Andrew Murray and William L Barnes. Plasmonic materials. Advanced materials,
19(22):3771–3782, 2007.
BIBLIOGRAPHY 153
[19] Anatoly V Zayats and Igor I Smolyaninov. Near-field photonics: surface plasmon po-
laritons and localized surface plasmons. Journal of Optics A: Pure and Applied Optics,
5(4):S16, 2003.
[20] Suljo Linic, Phillip Christopher, and David B Ingram. Plasmonic-metal nanostructures
for efficient conversion of solar to chemical energy. Nature materials, 10(12):911–921,
2011.
[21] JB Pendry, L Martin-Moreno, and FJ Garcia-Vidal. Mimicking surface plasmons with

structured surfaces. science, 305(5685):847–848, 2004.
[22] JM Pitarke, VM Silkin, EV Chulkov, and PM Echenique. Theory of surface plasmons

and surface-plasmon polaritons. Reports on progress in physics, 70(1):1, 2006.
[23] Luis M Liz-Marzán. Tailoring surface plasmons through the morphology and assembly of
metal nanoparticles. Langmuir, 22(1):32–41, 2006.
[24] Paul R West, Satoshi Ishii, Gururaj V Naik, Naresh K Emani, Vladimir M Shalaev,
and Alexandra Boltasseva. Searching for better plasmonic materials. Laser & Photonics
Reviews, 4(6):795–808, 2010.
[25] Urcan Guler, Vladimir M Shalaev, and Alexandra Boltasseva. Nanoparticle plasmonics:
going practical with transition metal nitrides. Materials Today, 18(4):227–237, 2015.
[26] Martin G Blaber, Matthew D Arnold, and Michael J Ford. A review of the optical
properties of alloys and intermetallics for plasmonics. Journal of Physics: Condensed
Matter, 22(14):143201, 2010.
[27] MA Noginov, Lei Gu, J Livenere, G Zhu, AK Pradhan, R Mundle, M Bahoura, Yu A

Barnakov, and VA Podolskiy. Transparent conductive oxides: Plasmonic materials for
telecom wavelengths. Applied Physics Letters, 99(2):021101, 2011.
[28] AN Grigorenko, Marco Polini, and KS Novoselov. Graphene plasmonics. Nature photon-
ics, 6(11):749, 2012.
BIBLIOGRAPHY 154
[29] Gururaj V Naik, Vladimir M Shalaev, and Alexandra Boltasseva. Alternative plasmonic
materials: beyond gold and silver. Advanced Materials, 25(24):3264–3294, 2013.
[30] Joel Henzie, Min Hyung Lee, and Teri W Odom. Multiscale patterning of plasmonic
metamaterials. Nature nanotechnology, 2(9):549–554, 2007.
[31] Pan Wang, Mazhar E Nasir, Alexey V Krasavin, Wayne Dickson, Yunlu Jiang, and
Anatoly V Zayats. Plasmonic metamaterials for nanochemistry and sensing. Accounts of
chemical research, 52(11):3018–3028, 2019.
[32] Ioannis G Theodorou, Qianfan Jiang, Lukas Malms, Xiangyu Xie, R Charles Coombes,
Eric O Aboagye, Alexandra E Porter, Mary P Ryan, and Fang Xie. Fluorescence enhance-
ment from single gold nanostars: towards ultra-bright emission in the first and second
near-infrared biological windows. Nanoscale, 10(33):15854–15864, 2018.
[33] Jiangtian Li, Scott K Cushing, Peng Zheng, Fanke Meng, Deryn Chu, and Nianqiang
Wu. Plasmon-induced photonic and energy-transfer enhancement of solar water splitting
by a hematite nanorod array. Nature communications, 4(1):1–8, 2013.
[34] Chia-Yang Tsai, Shao-Ping Lu, Jyun-Wei Lin, and Po-Tsung Lee. High sensitivity
plasmonic index sensor using slablike gold nanoring arrays. Applied physics letters,
98(15):153108, 2011.
[35] Yang Shen, Jianhua Zhou, Tianran Liu, Yuting Tao, Ruibin Jiang, Mingxuan Liu, Guohui
Xiao, Jinhao Zhu, Zhang-Kai Zhou, Xuehua Wang, et al. Plasmonic gold mushroom
arrays with refractive index sensing figures of merit approaching the theoretical limit.
Nature communications, 4(1):1–9, 2013.
[36] Arpad Jakab, Christina Rosman, Yuriy Khalavka, Jan Becker, Andreas Trugler, Ulrich
Hohenester, and Carsten Sonnichsen. Highly sensitive plasmonic silver nanorods. ACS
nano, 5(9):6880–6885, 2011.
[37] George H Chan, Jing Zhao, Erin M Hicks, George C Schatz, and Richard P Van Duyne.
Plasmonic properties of copper nanoparticles fabricated by nanosphere lithography. Nano
letters, 7(7):1947–1952, 2007.
BIBLIOGRAPHY 155
[38] Mark W Knight, Lifei Liu, Yumin Wang, Lisa Brown, Shaunak Mukherjee, Nicholas S
King, Henry O Everitt, Peter Nordlander, and Naomi J Halas. Aluminum plasmonic
nanoantennas. Nano letters, 12(11):6000–6004, 2012.
[39] Qiang Li, Zizheng Li, Haigui Yang, Hai Liu, Xiaoyi Wang, Jinsong Gao, and Jingli Zhao.
Novel aluminum plasmonic absorber enhanced by extraordinary optical transmission.
Optics express, 24(22):25885–25893, 2016.
[40] Enyi Ye, Khin Yin Win, Hui Ru Tan, Ming Lin, Choon Peng Teng, Adnen Mlayah, and
Ming-Yong Han. Plasmonic gold nanocrosses with multidirectional excitation and strong
photothermal effect. Journal of the American Chemical Society, 133(22):8506–8509, 2011.
[41] Qingkun Liu, Yanxia Cui, Dennis Gardner, Xin Li, Sailing He, and Ivan I Smalyukh.
Self-alignment of plasmonic gold nanorods in reconfigurable anisotropic fluids for tunable
bulk metamaterial applications. Nano letters, 10(4):1347–1353, 2010.
[42] Hua Dong, Zhaoxin Wu, Fan Lu, Yucui Gao, Ahmed El-Shafei, Bo Jiao, Shuya Ning,
and Xun Hou. Optics–electrics highways: Plasmonic silver nanowires@ tio2 core–shell
nanocomposites for enhanced dye-sensitized solar cells performance. Nano Energy,
10:181–191, 2014.
[43] Qing Chang, Xiaoyu Shi, Xuan Liu, Junhua Tong, Dahe Liu, and Zhaona Wang. Broad-
band plasmonic silver nanoflowers for high-performance random lasing covering visible
region. Nanophotonics, 6(5):1151–1160, 2017.
[44] Mehmet Kahraman, Ayse Ozbay, Handan Yuksel, Ramazan Solmaz, Baran Demir, and
Humeyra Caglayan. Tunable plasmonic silver nanodomes for surface-enhanced raman
scattering. Plasmonics, 13(3):785–795, 2018.
[45] Kai Chen, Thang Duy Dao, Satoshi Ishii, Masakazu Aono, and Tadaaki Nagao. Infrared
aluminum metamaterial perfect absorbers for plasmon-enhanced infrared spectroscopy.
Advanced Functional Materials, 25(42):6637–6643, 2015.
BIBLIOGRAPHY 156
[46] Ye Zhang, Ning Zhou, Keqin Zhang, and Feng Yan. Plasmonic copper nanowire@ tio2
nanostructures for improving the performance of dye-sensitized solar cells. Journal of
Power Sources, 342:292–300, 2017.
[47] Yueying Wu, Guoliang Li, and Jon P Camden. Probing nanoparticle plasmons with
electron energy loss spectroscopy. Chemical reviews, 118(6):2994–3031, 2017.
[48] Kane Yee. Numerical solution of initial boundary value problems involving maxwell’s
equations in isotropic media. IEEE Transactions on antennas and propagation, 14(3):302–
307, 1966.
[49] Karl S Kunz and Raymond J Luebbers. The finite difference time domain method for
electromagnetics. CRC press, 1993.
[50] EA Navarro, C Wu, PY Chung, and J Litva. Application of pml superabsorbing boundary
condition to non-orthogonal fdtd method. Electronics Letters, 30(20):1654–1656, 1994.
[51] Alexei Deinega, Sergei Belousov, and Ilya Valuev. Hybrid transfer-matrix fdtd method
for layered periodic structures. Optics letters, 34(6):860–862, 2009.
[52] Yasuhiro Tachibana, Lionel Vayssieres, and James R Durrant. Artificial photosynthesis
for solar water-splitting. Nature Photonics, 6(8):511, 2012.
[53] Masaya Matsuoka, Masaaki Kitano, Masato Takeuchi, Koichiro Tsujimaru, Masakazu
Anpo, and John M Thomas. Photocatalysis for new energy production: recent advances
in photocatalytic water splitting reactions for hydrogen production. Catalysis Today,
122(1-2):51–61, 2007.
[54] Allen J Bard and Marye Anne Fox. Artificial photosynthesis: solar splitting of water to
hydrogen and oxygen. Accounts of Chemical Research, 28(3):141–145, 1995.
[55] Arthur J Esswein and Daniel G Nocera. Hydrogen production by molecular photocatal-
ysis. Chemical reviews, 107(10):4022–4047, 2007.
BIBLIOGRAPHY 157
[56] Avigail Landman, Hen Dotan, Gennady E Shter, Michael Wullenkord, Anis Houaijia,
Artjom Maljusch, Gideon S Grader, and Avner Rothschild. Photoelectrochemical water
splitting in separate oxygen and hydrogen cells. Nature materials, 16(6):646–651, 2017.
[57] Marı́a D Hernández-Alonso, Fernando Fresno, Silvia Suárez, and Juan M Coronado.
Development of alternative photocatalysts to tio 2: challenges and opportunities. Energy
& Environmental Science, 2(12):1231–1257, 2009.
[58] Jianwei Sun, Diane K Zhong, and Daniel R Gamelin. Composite photoanodes for photo-
electrochemical solar water splitting. Energy & Environmental Science, 3(9):1252–1261,
2010.
[59] Isolda Roger, Michael A Shipman, and Mark D Symes. Earth-abundant catalysts for
electrochemical and photoelectrochemical water splitting. Nature Reviews Chemistry,
1(1):1–13, 2017.
[60] Takashi Hisatomi, Jun Kubota, and Kazunari Domen. Recent advances in semiconductors
for photocatalytic and photoelectrochemical water splitting. Chemical Society Reviews,
43(22):7520–7535, 2014.
[61] Fei Cao, Gerko Oskam, Gerald J Meyer, and Peter C Searson. Electron transport in
porous nanocrystalline tio2 photoelectrochemical cells. The Journal of Physical Chem-
istry, 100(42):17021–17027, 1996.
[62] Marek Szklarczyk and John O’M Bockris. Photoelectrochemical evolution of hydrogen
on p-indium phosphide. The Journal of Physical Chemistry, 88(22):5241–5245, 1984.
[63] RL Blake, RE Hessevick, Tibor Zoltai, and Larry W Finger. Refinement of the hematite
structure. American Mineralogist: Journal of Earth and Planetary Materials, 51(1-
2):123–129, 1966.
[64] Carrick M Eggleston and Michael F Hochella. The structure of hematite {001} surfaces
by scanning tunneling microscopy: Image interpretation, surface relaxation, and step
structure. American Mineralogist, 77(9-10):911–922, 1992.
BIBLIOGRAPHY 158
[65] Mario RS Soares, Ricardo H Gonçalves, Içamira C Nogueira, Jefferson Bettini, Ade-
nilson J Chiquito, and Edson R Leite. Understanding the fundamental electrical and
photoelectrochemical behavior of a hematite photoanode. Physical Chemistry Chemical
Physics, 18(31):21780–21788, 2016.
[66] Franz Bødker, Mikkel F Hansen, Christian Bender Koch, Kim Lefmann, and Steen Mørup.
Magnetic properties of hematite nanoparticles. Physical Review B, 61(10):6826, 2000.
[67] Rongming Wang, Yaofeng Chen, Yunyi Fu, Han Zhang, and Christian Kisielowski. Bicrys-
talline hematite nanowires. The Journal of Physical Chemistry B, 109(25):12245–12249,
2005.
[68] Suoyuan Lian, Enbo Wang, Zhenhui Kang, Yunpeng Bai, Lei Gao, Min Jiang, Changwen
Hu, and Lin Xu. Synthesis of magnetite nanorods and porous hematite nanorods. Solid
State Communications, 129(8):485–490, 2004.
[69] D Muthu Gnana Theresa Nathan and S Jacob Melvin Boby. Hydrothermal preparation
of hematite nanotubes/reduced graphene oxide nanocomposites as electrode material for
high performance supercapacitors. Journal of Alloys and Compounds, 700:67–74, 2017.
[70] Jae Young Kim, Ganesan Magesh, Duck Hyun Youn, Ji-Wook Jang, Jun Kubota, Kazu-
nari Domen, and Jae Sung Lee. Single-crystalline, wormlike hematite photoanodes for
efficient solar water splitting. Scientific reports, 3:2681, 2013.
[71] Kevin Sivula, Florian Le Formal, and Michael Grätzel. Solar water splitting: progress
using hematite (α-fe2o3) photoelectrodes. ChemSusChem, 4(4):432–449, 2011.
[72] Niclas Beermann, Lionel Vayssieres, Sten-Eric Lindquist, and Anders Hagfeldt. Pho-
toelectrochemical studies of oriented nanorod thin films of hematite. Journal of the
Electrochemical Society, 147(7):2456–2461, 2000.
[73] Susanta K Mohapatra, Shiny E John, Subarna Banerjee, and Mano Misra. Water pho-
tooxidation by smooth and ultrathin α-fe2o3 nanotube arrays. Chemistry of Materials,
21(14):3048–3055, 2009.
BIBLIOGRAPHY 159
[74] Ricardo H Gonçalves and Edson R Leite. The colloidal nanocrystal deposition process: an
advanced method to prepare high performance hematite photoanodes for water splitting.
Energy & Environmental Science, 7(7):2250–2254, 2014.
[75] Dawei Ding, Bitao Dong, Jin Liang, Han Zhou, Yuanchao Pang, and Shujiang Ding.
Solvothermal-etching process induced ti-doped fe2o3 thin film with low turn-on voltage
for water splitting. ACS applied materials & interfaces, 8(37):24573–24578, 2016.
[76] Michael Grätzel. Mesoscopic solar cells for electricity and hydrogen production from
sunlight. Chemistry letters, 34(1):8–13, 2005.
[77] Alan Kleiman-Shwarsctein, Yong-Sheng Hu, Arnold J Forman, Galen D Stucky, and
Eric W McFarland. Electrodeposition of α-fe2o3 doped with mo or cr as photoanodes
for photocatalytic water splitting. The Journal of Physical Chemistry C, 112(40):15900–
15907, 2008.
[78] Kevin Sivula, Florian Le Formal, and Michael Gratzel. Wo3- fe2o3 photoanodes for water
splitting: A host scaffold, guest absorber approach. Chemistry of Materials, 21(13):2862–
2867, 2009.
[79] Ming Zhang, Yongjing Lin, Thomas J Mullen, Wei-feng Lin, Ling-Dong Sun, Chun-Hua
Yan, Timothy E Patten, Dunwei Wang, and Gang-yu Liu. Improving hematite’s solar
water splitting efficiency by incorporating rare-earth upconversion nanomaterials. The
journal of physical chemistry letters, 3(21):3188–3192, 2012.
[80] Jiujun Deng, Jun Zhong, Aiwu Pu, Duo Zhang, Ming Li, Xuhui Sun, and Shuit-Tong Lee.
Ti-doped hematite nanostructures for solar water splitting with high efficiency. Journal
of Applied Physics, 112(8):084312, 2012.
[81] Florian Le Formal, Nicolas Tetreault, Maurin Cornuz, Thomas Moehl, Michael Grätzel,
and Kevin Sivula. Passivating surface states on water splitting hematite photoanodes
with alumina overlayers. Chemical Science, 2(4):737–743, 2011.
BIBLIOGRAPHY 160
[82] RR Rangaraju, A Panday, KS Raja, and M Misra. Nanostructured anodic iron oxide film
as photoanode for water oxidation. Journal of Physics D: Applied Physics, 42(13):135303,
2009.
[83] Maurin Cornuz, Michael Grätzel, and Kevin Sivula. Preferential orientation in hematite
films for solar hydrogen production via water splitting. Chemical Vapor Deposition, 16(10-
12):291–295, 2010.
[84] Yongjing Lin, Yang Xu, Matthew T Mayer, Zachary I Simpson, Gregory McMahon,
Sa Zhou, and Dunwei Wang. Growth of p-type hematite by atomic layer deposition
and its utilization for improved solar water splitting. Journal of the American Chemical
Society, 134(12):5508–5511, 2012.
[85] Yichuan Ling, Gongming Wang, Damon A Wheeler, Jin Z Zhang, and Yat Li. Sn-
doped hematite nanostructures for photoelectrochemical water splitting. Nano letters,
11(5):2119–2125, 2011.
[86] Andreas Kay, Ilkay Cesar, and Michael Grätzel. New benchmark for water photoox-
idation by nanostructured α-fe2o3 films. Journal of the American Chemical Society,
128(49):15714–15721, 2006.
[87] Shaohua Shen, Coleman X Kronawitter, Damon A Wheeler, Penghui Guo, Sarah A Lind-
ley, Jiangang Jiang, Jin Z Zhang, Liejin Guo, and Samuel S Mao. Physical and photo-
electrochemical characterization of ti-doped hematite photoanodes prepared by solution
growth. Journal of Materials Chemistry A, 1(46):14498–14506, 2013.
[88] Sing Yang Chiam, Mulmudi Hemant Kumar, Prince Saurabh Bassi, Hwee Leng Seng,
James Barber, and Lydia Helena Wong. Improving the efficiency of hematite nanorods for
photoelectrochemical water splitting by doping with manganese. ACS applied materials
& interfaces, 6(8):5852–5859, 2014.
[89] Kenneth J McDonald and Kyoung-Shin Choi. Photodeposition of co-based oxygen evo-
lution catalysts on α-fe2o3 photoanodes. Chemistry of Materials, 23(7):1686–1693, 2011.
BIBLIOGRAPHY 161
[90] Takashi Hisatomi, Florian Le Formal, Maurin Cornuz, Jérémie Brillet, Nicolas Tétreault,
Kevin Sivula, and Michael Grätzel. Cathodic shift in onset potential of solar oxygen
evolution on hematite by 13-group oxide overlayers. Energy & Environmental Science,
4(7):2512–2515, 2011.
[91] Yong-Sheng Hu, Alan Kleiman-Shwarsctein, Galen D Stucky, and Eric W McFarland.
Improved photoelectrochemical performance of ti-doped α-fe 2 o 3 thin films by surface
modification with fluoride. Chemical Communications, (19):2652–2654, 2009.
[92] Won Jae Lee, Pravin S Shinde, Geun Ho Go, and Chil Hoon Doh. Cathodic shift and
improved photocurrent performance of cost-effective fe2o3 photoanodes. international
journal of hydrogen energy, 39(11):5575–5579, 2014.
[93] Lifei Xi, Prince Saurabh Bassi, Sing Yang Chiam, Wai Fatt Mak, Phong D Tran, James
Barber, Joachim Say Chye Loo, and Lydia Helena Wong. Surface treatment of hematite
photoanodes with zinc acetate for water oxidation. Nanoscale, 4(15):4430–4433, 2012.
[94] S David Tilley, Maurin Cornuz, Kevin Sivula, and Michael Grätzel. Light-induced water
splitting with hematite: improved nanostructure and iridium oxide catalysis. Angewandte
Chemie International Edition, 49(36):6405–6408, 2010.
[95] Lifei Xi, Phong D Tran, Sing Yang Chiam, Prince Saurabh Bassi, Wai Fatt Mak, He-
mant Kumar Mulmudi, Sudip K Batabyal, James Barber, Joachim Say Chye Loo, and
Lydia Helena Wong. Co3o4-decorated hematite nanorods as an effective photoanode for
solar water oxidation. The Journal of Physical Chemistry C, 116(26):13884–13889, 2012.
[96] Scott C Warren and Elijah Thimsen. Plasmonic solar water splitting. Energy & Envi-
ronmental Science, 5(1):5133–5146, 2012.
[97] KR Catchpole, , and Albert Polman. Plasmonic solar cells. Optics express, 16(26):21793–
21800, 2008.
[98] Ragip A Pala, Justin White, Edward Barnard, John Liu, and Mark L Brongersma. Design
of plasmonic thin-film solar cells with broadband absorption enhancements. Advanced
materials, 21(34):3504–3509, 2009.
BIBLIOGRAPHY 162
[99] Yang Tian and Tetsu Tatsuma. Plasmon-induced photoelectrochemistry at metal

nanoparticles supported on nanoporous tio 2. Chemical communications, (16):1810–1811,
2004.
[100] Brian O’regan and Michael Grätzel. A low-cost, high-efficiency solar cell based on dye-
sensitized colloidal tio 2 films. nature, 353(6346):737–740, 1991.
[101] Akihiro Furube, Luchao Du, Kohjiro Hara, Ryuzi Katoh, and Masanori Tachiya. Ultrafast
plasmon-induced electron transfer from gold nanodots into tio2 nanoparticles. Journal of
the American Chemical Society, 129(48):14852–14853, 2007.
[102] Allen Taflove and Morris E Brodwin. Numerical solution of steady-state electromagnetic
scattering problems using the time-dependent maxwell’s equations. IEEE transactions
on microwave theory and techniques, 23(8):623–630, 1975.
[103] Zhao Li, Li Shi, Daniel Franklin, Supriya Koul, Akihiro Kushima, and Yang Yang. Drastic
enhancement of photoelectrochemical water splitting performance over plasmonic al@ tio2
heterostructured nanocavity arrays. Nano Energy, 51:400–407, 2018.
[104] Xing Zhang, Yang Liu, and Zhenhui Kang. 3d branched zno nanowire arrays decorated
with plasmonic au nanoparticles for high-performance photoelectrochemical water split-
ting. ACS applied materials & interfaces, 6(6):4480–4489, 2014.
[105] Kim Hang Ng, Lorna Jeffery Minggu, Nurul Akmal Jaafar, Khuzaimah Arifin, and Mo-
hammad Bin Kassim. Enhanced plasmonic photoelectrochemical response of au sand-
wiched wo3 photoanodes. Solar Energy Materials and Solar Cells, 172:361–367, 2017.
[106] Zhonghai Zhang, Lianbin Zhang, Mohamed Nejib Hedhili, Hongnan Zhang, and Peng
Wang. Plasmonic gold nanocrystals coupled with photonic crystal seamlessly on tio2
nanotube photoelectrodes for efficient visible light photoelectrochemical water splitting.
Nano letters, 13(1):14–20, 2013.
[107] M Valenti, D Dolat, G Biskos, A Schmidt-Ott, and WA Smith. Enhancement of the pho-
toelectrochemical performance of cuwo4 thin films for solar water splitting by plasmonic
BIBLIOGRAPHY 163
nanoparticle functionalization. The Journal of Physical Chemistry C, 119(4):2096–2104,

2015.
[108] Arif Sheikh, Ashish Yengantiwar, Meenal Deo, Sarika Kelkar, and Satishchandra Ogale.
Near-field plasmonic functionalization of light harvesting oxide–oxide heterojunctions for
efficient solar photoelectrochemical water splitting: The aunp/znfe2o4/zno system. Small,
9(12):2091–2096, 2013.
[109] Zhifeng Liu, Jianyu Wu, and Jing Zhang. Quantum dots and plasmonic ag decorated wo3
nanorod photoanodes with enhanced photoelectrochemical performances. International
Journal of Hydrogen Energy, 41(45):20529–20535, 2016.
[110] Anuradha Verma, Anupam Srivastav, Anamika Banerjee, Dipika Sharma, Shailja Sharma,
Udai Bhan Singh, Vibha Rani Satsangi, Rohit Shrivastav, Devesh Kumar Avasthi, and
Sahab Dass. Plasmonic layer enhanced photoelectrochemical response of fe2o3 photoan-
odes. Journal of Power Sources, 315:152–160, 2016.
[111] Weiwei Yang, Yunjie Xiong, Liangliang Zou, Zhiqing Zou, Dongdong Li, Qixi Mi, Yanshan
Wang, and Hui Yang. plasmonic pd nanoparticle-and plasmonic pd nanorod-decorated
bivo 4 electrodes with enhanced photoelectrochemical water splitting efficiency across
visible-nir region. Nanoscale research letters, 11(1):1–8, 2016.
[112] Xiaogang Liu, Chun-Hua Yan, and John A Capobianco. Photon upconversion nanoma-
terials. Chemical Society Reviews, 44(6):1299–1301, 2015.
[113] Nicholas J Turro, V Ramamurthy, and Juan C Scaiano. Modern Molecular Photochemistry
of Organic Molecules. Viva Books University Science Books, Sausalito, 2017.
[114] Chunying Fan, Lingling Wei, Tong Niu, Ming Rao, Guo Cheng, Jason J Chruma, Wanhua
Wu, and Cheng Yang. Efficient triplet–triplet annihilation upconversion with an anti-
stokes shift of 1.08 ev achieved by chemically tuning sensitizers. Journal of the American
Chemical Society, 141(38):15070–15077, 2019.
BIBLIOGRAPHY 164
[115] Christoph Kerzig and Oliver S Wenger. Sensitized triplet–triplet annihilation upconver-
sion in water and its application to photochemical transformations. Chemical science,
9(32):6670–6678, 2018.
[116] Tanya N Singh-Rachford and Felix N Castellano. Photon upconversion based on sensi-
tized triplet–triplet annihilation. Coordination Chemistry Reviews, 254(21-22):2560–2573,
2010.
[117] Victor Gray, Damir Dzebo, Maria Abrahamsson, Bo Albinsson, and Kasper Moth-
Poulsen. Triplet–triplet annihilation photon-upconversion: towards solar energy applica-
tions. Physical Chemistry Chemical Physics, 16(22):10345–10352, 2014.
[118] Andrew Nattestad, Yuen Yap Cheng, Rowan W MacQueen, Tim F Schulze, Fletcher W
Thompson, Attila J Mozer, Burkhard Fuckel, Tony Khoury, Maxwell J Crossley, Klaus
Lips, et al. Dye-sensitized solar cell with integrated triplet–triplet annihilation upconver-
sion system. The journal of physical chemistry letters, 4(12):2073–2078, 2013.
[119] YunHui L Lin, Marius Koch, Alyssa N Brigeman, David ME Freeman, Lianfeng Zhao,
Hugo Bronstein, Noel C Giebink, Gregory D Scholes, and Barry P Rand. Enhanced sub-
bandgap efficiency of a solid-state organic intermediate band solar cell using triplet–triplet
annihilation. Energy & Environmental Science, 10(6):1465–1475, 2017.
[120] Oh Seok Kwon, Jae-Hyuk Kim, Jin Ku Cho, and Jae-Hong Kim. Triplet–triplet annihi-
lation upconversion in cds-decorated sio2 nanocapsules for sub-bandgap photocatalysis.
ACS applied materials & interfaces, 7(1):318–325, 2015.
[121] Anna L Hagstrom, Seunghyun Weon, Wonyong Choi, and Jae-Hong Kim. Triplet–triplet
annihilation upconversion in broadly absorbing layered film systems for sub-bandgap
photocatalysis. ACS applied materials & interfaces, 11(14):13304–13318, 2019.
[122] Changqing Ye, Jingjing Wang, Xiaomei Wang, Ping Ding, Zuoqin Liang, and Xutang
Tao. A new medium for triplet–triplet annihilated upconversion and photocatalytic ap-
plication. Physical Chemistry Chemical Physics, 18(5):3430–3437, 2016.
BIBLIOGRAPHY 165
[123] Wounjhang Park, Dawei Lu, and Sungmo Ahn. Plasmon enhancement of luminescence
upconversion. Chemical Society Reviews, 44(10):2940–2962, 2015.
[124] Di M Wu, Aitzol Garcı́a-Etxarri, Alberto Salleo, and Jennifer A Dionne. Plasmon-
enhanced upconversion. The journal of physical chemistry letters, 5(22):4020–4031, 2014.
[125] Wei Feng, Ling-Dong Sun, and Chun-Hua Yan. Ag nanowires enhanced upconversion
emission of nayf4: Yb, er nanocrystalsvia a direct assembly method. Chemical commu-
nications, (29):4393–4395, 2009.
[126] Hari P Paudel, Lanlan Zhong, Khadijeh Bayat, Mahdi Farrokh Baroughi, Steve Smith,
Cuikun Lin, Chaoyang Jiang, Mary T Berry, and P Stanley May. Enhancement of near-
infrared-to-visible upconversion luminescence using engineered plasmonic gold surfaces.
The Journal of Physical Chemistry C, 115(39):19028–19036, 2011.
[127] Kianoosh Poorkazem, Amelia V Hesketh, and Timothy L Kelly. Plasmon-enhanced

triplet–triplet annihilation using silver nanoplates. The Journal of Physical Chemistry C,
118(12):6398–6404, 2014.
[128] J Shen, ZQ Li, YR Chen, XH Chen, YW Chen, Z Sun, and SM Huang. Influence of sio2
layer thickness on plasmon enhanced upconversion in hybrid ag/sio2/nayf4: Yb, er, gd
structures. Applied surface science, 270:712–717, 2013.
[129] Davinson Mariano da Silva, Luciana Reyes Pires Kassab, Stefan R Lüthi, Cid B de Araújo,
Anderson SL Gomes, and Maria José Valenzuella Bell. Frequency upconversion in er 3+
doped pb o–ge o 2 glasses containing metallic nanoparticles. Applied Physics Letters,
90(8):081913, 2007.
[130] Takeho Aisaka, Minoru Fujii, and Shinji Hayashi. Enhancement of upconversion lumines-
cence of er doped al 2 o 3 films by ag island films. Applied Physics Letters, 92(13):132105,
2008.
[131] Yi Wu, Xiang Shen, Shixun Dai, Yinsheng Xu, Feifei Chen, Changgui Lin, Tiefeng
Xu, and Qiuhua Nie. Silver nanoparticles enhanced upconversion luminescence in
BIBLIOGRAPHY 166
er3+/yb3+ codoped bismuth-germanate glasses. The Journal of Physical Chemistry C,

115(50):25040–25045, 2011.
[132] Marjan Saboktakin, Xingchen Ye, Uday K Chettiar, Nader Engheta, Christopher B Mur-
ray, and Cherie R Kagan. Plasmonic enhancement of nanophosphor upconversion lumi-
nescence in au nanohole arrays. ACS nano, 7(8):7186–7192, 2013.
[133] Weihua Zhang, Fei Ding, and Stephen Y Chou. Large enhancement of upconversion lu-
minescence of nayf4: Yb3+/er3+ nanocrystal by 3d plasmonic nano-antennas. Advanced
Materials, 24(35):OP236–OP241, 2012.
[134] Xian Cao, Bo Hu, Rui Ding, and Peng Zhang. Plasmon-enhanced homogeneous and het-
erogeneous triplet–triplet annihilation by gold nanoparticles. Physical Chemistry Chem-
ical Physics, 17(22):14479–14483, 2015.
[135] Hyoung-il Kim, Seunghyun Weon, Homan Kang, Anna L Hagstrom, Oh Seok Kwon,
Yoon-Sik Lee, Wonyong Choi, and Jae-Hong Kim. Plasmon-enhanced sub-bandgap pho-
tocatalysis via triplet–triplet annihilation upconversion for volatile organic compound
degradation. Environmental science & technology, 50(20):11184–11192, 2016.
[136] Jiaojiao Fang, Cihui Zhou, Yukai Chen, Liang Fang, Wei Wang, Cheng Zhu, Yaru Ni,
and Chunhua Lu. Efficient photocatalysis of composite films based on plasmon-enhanced
triplet–triplet annihilation. ACS Applied Materials & Interfaces, 12(1):717–726, 2019.
[137] Joseph I. Goldstein, Dale E. Newbury, Joseph R. Michael, Nicholas W.M. Ritchie, John
Henry J. Scott, David C. Joy. Scanning Electron Microscopy and X-Ray Microanalysis.
Springer, 2017.
[138] David B. Williams, C. Barry Carter. Transmission Electron Microscopy: A Textbook for
Materials Science. Springer, 2009.
[139] John C. Russ. Fundamentals of Energy Dispersive X-ray Analysis. Butterworths, 1984.
[140] Robert E. Dinnebier, S. J. L. Billinge. Powder Diffraction: Theory and Practice. RSC
Publishing, 2008.
BIBLIOGRAPHY 167
[141] Tery L. Barr. Modern ESCA: The Principles and Practice of X-Ray Photoelectron Spec-
troscopy. CRC Press, 1994.
[142] Yongbin Lin, Yang Zou, Yuanyao Mo, Junpeng Guo, and Robert G Lindquist. E-beam
patterned gold nanodot arrays on optical fiber tips for localized surface plasmon resonance
biochemical sensing. Sensors, 10(10):9397–9406, 2010.
[143] Aaron R Halpern and Robert M Corn. Lithographically patterned electrodeposition

of gold, silver, and nickel nanoring arrays with widely tunable near-infrared plasmonic
resonances. ACS nano, 7(2):1755–1762, 2013.
[144] Christy L Haynes, Adam D McFarland, LinLin Zhao, Richard P Van Duyne, George C
Schatz, Linda Gunnarsson, Juris Prikulis, Bengt Kasemo, and Mikael Käll. Nanoparti-
cle optics: the importance of radiative dipole coupling in two-dimensional nanoparticle
arrays. The Journal of Physical Chemistry B, 107(30):7337–7342, 2003.
[145] Ardavan F Oskooi, David Roundy, Mihai Ibanescu, Peter Bermel, John D Joannopou-
los, and Steven G Johnson. Meep: A flexible free-software package for electromagnetic
simulations by the fdtd method. Computer Physics Communications, 181(3):687–702,
2010.
[146] Jian Zhu and Ying-juan Ren. Tuning the plasmon shift and local electric field distribution
of gold nanodumbbell: The effect of surface curvature transition from positive to negative.
Applied surface science, 285:649–656, 2013.
[147] Emilie Ringe, Jian Zhang, Mark R Langille, Kwonnam Sohn, Claire Cobley, Leslie Au,
Younan Xia, Chad A Mirkin, Jiaxing Huang, Laurence D Marks, et al. Effect of size,
shape, composition, and support film on localized surface plasmon resonance frequency:
a single particle approach applied to silver bipyramids and gold and silver nanocubes.
MRS Online Proceedings Library Archive, 1208, 2009.
[148] Peter B Catrysse and Shanhui Fan. Propagating plasmonic mode in nanoscale aper-
tures and its implications for extraordinary transmission. Journal of Nanophotonics,
2(1):021790, 2008.
BIBLIOGRAPHY 168
[149] Takazumi Ohno, Carl Wadell, Satoshi Inagaki, Ji Shi, Yoshio Nakamura, Sachiko Mat-
sushita, and Takumi Sannomiya. Hole-size tuning and sensing performance of hexagonal
plasmonic nanohole arrays. Optical Materials Express, 6(5):1594–1603, 2016.
[150] Alastair Paul Hibbins. Grating coupling of surface plasmon polaritons at visible and
microwave frequencies. PhD thesis, University of Exeter, 2000.
[151] Seung-Woo Lee, Kyeong-Seok Lee, Junhyoung Ahn, Jae-Jong Lee, Min-Gon Kim, and
Yong-Beom Shin. Highly sensitive biosensing using arrays of plasmonic au nanodisks
realized by nanoimprint lithography. ACS nano, 5(2):897–904, 2011.
[152] Andreas Tittl, Harald Giessen, and Na Liu. Plasmonic gas and chemical sensing.
Nanophotonics, 3(3):157–180, 2014.
[153] Alireza Dolatabady, Nosrat Granpayeh, and Vahid Foroughi Nezhad. A nanoscale re-
fractive index sensor in two dimensional plasmonic waveguide with nanodisk resonator.
Optics Communications, 300:265–268, 2013.
[154] Beniamino Iandolo and Michael Zäch. Enhanced water splitting on thin-film hematite
photoanodes functionalized with lithographically fabricated au nanoparticles. Australian
Journal of Chemistry, 65(6):633–637, 2012.
[155] Shouli Bai, Ke Tian, Hang Fu, Yongjun Feng, Ruixian Luo, Dianqing Li, Aifan Chen,
and Chung Chiun Liu. Novel α-fe2o3/bivo4 heterojunctions for enhancing no2 sensing
properties. Sensors and Actuators B: Chemical, 268:136–143, 2018.
[156] Shaohua Shen, Sarah A Lindley, Xiangyan Chen, and Jin Z Zhang. Hematite heterostruc-
tures for photoelectrochemical water splitting: rational materials design and charge car-
rier dynamics. Energy & Environmental Science, 9(9):2744–2775, 2016.
[157] John C Hulteen and Richard P Van Duyne. Nanosphere lithography: A materials general
fabrication process for periodic particle array surfaces. Journal of Vacuum Science &
Technology A: Vacuum, Surfaces, and Films, 13(3):1553–1558, 1995.
BIBLIOGRAPHY 169
[158] Fang Xie, Jing S Pang, Anthony Centeno, Mary P Ryan, D Jason Riley, and Neil M
Alford. Nanoscale control of ag nanostructures for plasmonic fluorescence enhancement
of near-infrared dyes. Nano Research, 6(7):496–510, 2013.
[159] Lionel Vayssieres, Niclas Beermann, Sten-Eric Lindquist, and Anders Hagfeldt. Con-
trolled aqueous chemical growth of oriented three-dimensional crystalline nanorod arrays:
application to iron (iii) oxides. Chemistry of materials, 13(2):233–235, 2001.
[160] Junghyun Choi, Taeseup Song, Jiseok Kwon, Sangkyu Lee, Hyungkyu Han, Nitish Roy,
Chiaki Terashima, Akira Fujishima, Ungyu Paik, and Sudhagar Pitchaimuthu. Wo3
nanofibrous backbone scaffolds for enhanced optical absorbance and charge transport
in metal oxide (fe2o3, bivo4) semiconductor photoanodes towards solar fuel generation.
Applied Surface Science, 447:331–337, 2018.
[161] Rachel Morrish, Mahfujur Rahman, JM Don MacElroy, and Colin A Wolden. Activation
of hematite nanorod arrays for photoelectrochemical water splitting. ChemSusChem,
4(4):474–479, 2011.
[162] Yichuan Ling and Yat Li. Review of sn-doped hematite nanostructures for photoelectro-
chemical water splitting. Particle & Particle Systems Characterization, 31(11):1113–1121,
2014.
[163] Li Zhang, Dairong Chen, and Xiuling Jiao. Monoclinic structured bivo4 nanosheets:
hydrothermal preparation, formation mechanism, and coloristic and photocatalytic prop-
erties. The Journal of Physical Chemistry B, 110(6):2668–2673, 2006.
[164] Shouli Bai, Haomiao Chu, Xu Xiang, Ruixian Luo, Jing He, and Aifan Chen. Fabricating
of fe2o3/bivo4 heterojunction based photoanode modified with nife-ldh nanosheets for
efficient solar water splitting. Chemical Engineering Journal, 350:148–156, 2018.
[165] De-Kun Ma, Mei-Li Guan, Sen-Sen Liu, Yan-Qing Zhang, Chang-Wei Zhang, Yu-Xiang
He, and Shao-Ming Huang. Controlled synthesis of olive-shaped bi 2 s 3/bivo 4 micro-
spheres through a limited chemical conversion route and enhanced visible-light-responding
photocatalytic activity. Dalton transactions, 41(18):5581–5586, 2012.
BIBLIOGRAPHY 170
[166] Jingran Xiao, Huali Huang, Qiuyang Huang, Le Zhao, Xiang Li, Xuelan Hou, Hong
Chen, and Yongdan Li. Suppressing the electron–hole recombination rate in hematite
photoanode with a rapid cooling treatment. Journal of Catalysis, 350:48–55, 2017.
[167] Xin Zhang, Yongan Niu, Yao Li, Xuemei Hou, Yibo Wang, Rui Bai, and Jiupeng Zhao.
Synthesis, optical and magnetic properties of α-fe2o3 nanoparticles with various shapes.
Materials letters, 99:111–114, 2013.
[168] Qianfan Jiang, Chengyu Ji, D Jason Riley, and Fang Xie. Boosting the efficiency of
photoelectrolysis by the addition of non-noble plasmonic metals: Al & cu. Nanomaterials,
9(1):1, 2019.
[169] Chun-Jiang Jia, Ling-Dong Sun, Zheng-Guang Yan, Li-Ping You, Feng Luo, Xiao-Dong
Han, Yu-Cheng Pang, Ze Zhang, and Chun-Hua Yan. Single-crystalline iron oxide nan-
otubes. Angewandte Chemie, 117(28):4402–4407, 2005.
[170] JYT Chan, SY Ang, EY Ye, M Sullivan, J Zhang, and M Lin. Heterogeneous photo-
fenton reaction on hematite (α-fe 2 o 3){104},{113} and {001} surface facets. Physical
Chemistry Chemical Physics, 17(38):25333–25341, 2015.
[171] Marie-Pierre Pomies, Guillaume Morin, and Colette Vignaud. Xrd study of the goethite-
hematite transformation: application to the identification of heated prehistoric pigments.
European Journal of solid state and Inorganic Chemistry, 35(1):9–25, 1998.
[172] Mario RS Soares, Ricardo H Gonçalves, Içamira C Nogueira, Jefferson Bettini, Ade-
nilson J Chiquito, and Edson R Leite. Understanding the fundamental electrical and
photoelectrochemical behavior of a hematite photoanode. Physical Chemistry Chemical
Physics, 18(31):21780–21788, 2016.
[173] Hanwei Gao, Chong Liu, Hoon Eui Jeong, and Peidong Yang. Plasmon-enhanced photo-
catalytic activity of iron oxide on gold nanopillars. ACS nano, 6(1):234–240, 2012.
[174] Qianfan Jiang, Xiangyu Xie, D Jason Riley, and Fang Xie. Harvesting the lost photon
by plasmonic enhanced hematite-upconversion nanocomposite for water splitting. The
Journal of Chemical Physics, 153(1):011102, 2020.
BIBLIOGRAPHY 171
[175] Rony S Khnayzer, Jörg Blumhoff, Jordan A Harrington, Alexandre Haefele, Fan Deng,
and Felix N Castellano. Upconversion-powered photoelectrochemistry. Chemical Com-
munications, 48(2):209–211, 2012.
[176] Tianjun Yu, Yanpeng Liu, Yi Zeng, Jinping Chen, Guoqiang Yang, and Yi Li. Triplet–
triplet annihilation upconversion for photocatalytic hydrogen evolution. Chemistry–A
European Journal, 25(71):16270–16276, 2019.
[177] Qian Liu, Wei Feng, Tianshe Yang, Tao Yi, and Fuyou Li. Upconversion luminescence
imaging of cells and small animals. Nature protocols, 8(10):2033, 2013.
[178] Junguo Xue, Chungang Wang, and Zhanfang Ma. A facile method to prepare a series of
sio2@ au core/shell structured nanoparticles. Materials Chemistry and Physics, 105(2-
3):419–425, 2007.
[179] Xinjian Shi, Kan Zhang, Kahee Shin, Jun Hyuk Moon, Tae-Woo Lee, and Jong Hyeok
Park. Constructing inverse opal structured hematite photoanodes via electrochemical
process and their application to photoelectrochemical water splitting. Physical Chemistry
Chemical Physics, 15(28):11717–11722, 2013.
[180] Benjamin Hatton, Lidiya Mishchenko, Stan Davis, Kenneth H Sandhage, and Joanna
Aizenberg. Assembly of large-area, highly ordered, crack-free inverse opal films. Proceed-
ings of the National Academy of Sciences, 107(23):10354–10359, 2010.
[181] AL González, Cecilia Noguez, J Beránek, and AS Barnard. Size, shape, stability, and color
of plasmonic silver nanoparticles. The Journal of Physical Chemistry C, 118(17):9128–
9136, 2014.

Jiang Q 2020 PHD Thesis

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Jiang Q 2020 PHD Thesis

Uploaded by

Copyright:

Available Formats

Imperial College London

Royal School of Mines

Plasmonic Enhanced Photoelectrochemical

Submitted in partial fulfilment of the requirements

I would like to express my sincere gratitude to:

2.1.1 Principles of Plasmonics . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

2.1.2 Plasmonic Metals and Nanostructures . . . . . . . . . . . . . . . . . . . . 30

2.1.3 FDTD Modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

2.2 Solar Water Splitting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

2.2.1 Photoelectrochemical Water Splitting . . . . . . . . . . . . . . . . . . . . 34

2.2.2 Hematite Based Nanocomposites as Photoanodes . . . . . . . . . . . . . 36

2.2.3 Optimization Strategies of Hematite Photoanodes . . . . . . . . . . . . . 38

2.2.4 Plasmonic Enhanced Solar Water Splitting . . . . . . . . . . . . . . . . . 40

2.3.1 Upconversion and TTA-based Upconversion . . . . . . . . . . . . . . . . 44

2.3.2 Upconversion Induced Water Splitting . . . . . . . . . . . . . . . . . . . 45

2.3.3 Plasmonic Enhanced Upconversion . . . . . . . . . . . . . . . . . . . . . 46

3 Instrumentation and Characterisation Methods 49

3.1 Scanning Electron Microscopy (SEM) . . . . . . . . . . . . . . . . . . . . . . . . 50

3.2 Transmission Electron Microscopy (TEM) . . . . . . . . . . . . . . . . . . . . . 50

3.3 Energy-dispersive X-Ray Analysis (EDX) . . . . . . . . . . . . . . . . . . . . . . 51

3.4 X-Ray Powder Diffraction (XRD) . . . . . . . . . . . . . . . . . . . . . . . . . . 51

3.5 X-Ray Photoelectron Spectroscopy (XPS) . . . . . . . . . . . . . . . . . . . . . 52

3.6 Photoluminescence (PL) Spectrometer . . . . . . . . . . . . . . . . . . . . . . . 53

3.7 Ultraviolet-visible-near-infrared (UV-Vis-NIR) Spectrometer . . . . . . . . . . . 54

3.8 Photoelectrochemical (PEC) Photocurrent Measurements . . . . . . . . . . . . . 55

3.9 Incident Photon-to-Current Efficiency (IPCE) Measurements . . . . . . . . . . . 56

4 FDTD Modelling of Plasmonic Nanoarrays 59

4.3 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

4.3.1 Nanodisk Arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

4.3.2 Nanotriangle Arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

4.3.3 Nanohole Arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75

5 Dual-layer Plasmonic Nanohole Array Enhanced BiVO4-Hematite Photoan-

5.3 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

5.3.1 Morphology and elemental analysis . . . . . . . . . . . . . . . . . . . . . 85

5.3.2 Optical properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92

5.3.3 Water splitting performance . . . . . . . . . . . . . . . . . . . . . . . . . 93

6 Aluminum Plasmonic Nanohole Array Enhanced Hematite Nanotube Pho-

6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100

6.2 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101

6.3 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103

6.3.1 Characterization of hematite nanotubes . . . . . . . . . . . . . . . . . . . 103

6.3.2 Characterization of the fabricated photoanodes . . . . . . . . . . . . . . 105

6.3.3 Characterization of water splitting performance . . . . . . . . . . . . . . 109

6.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113

7 TTA Upconversion Induced Solar Water Splitting by Hematite Photoanodes

7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116

7.2 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117

7.3 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119

7.3.1 Characterization of TTA@SiO2 upconversion nanoparticles. . . . . . . . . 119

7.3.2 Characterization of plasmonic enhanced TTA upconversion. . . . . . . . 122

7.3.3 Inverse-opal hematite photoanodes with TTA upconversion materials . . 125

7.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129

8.1 Summary of Thesis Achievements . . . . . . . . . . . . . . . . . . . . . . . . . . 131

8.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133

A Source Code of the FDTD Models 135

A.1 Nanohole Array and Nanotriangle Array . . . . . . . . . . . . . . . . . . . . . . 136

A.2 Nanohole Array with Oxidation Layer . . . . . . . . . . . . . . . . . . . . . . . . 140

A.3 Nanodisk Array . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144

A.4 Nanostars and Nanospheres . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146

B Supporting Information 149

BEM Boundary Element Method.

DDA Discrete Dipole Approximation.

EDX Energy dispersive X-Ray Analysis.