ABSTRACT
Title of Dissertation: Metal hydrides as a platform for recongurable
photonic and plasmonic elements
Kevin J. Palm
Doctor of Philosophy, 2021
Dissertation directed by: Professor Jeremy N. Munday
Department of Physics
Metal hydrides often display dramatic changes in optical properties upon hy-
drogenation. These shifts make them prime candidates for many tunable optical
devices from optical hydrogen sensors and switchable mirrors to physical encryption
schemes. In order to design and fabricate optimized devices for any of these applica-
tions, we need to determine the optical and structural properties of these materials.
In this dissertation, we design and implement an apparatus that dynamically mea-
sures the gravimetric, stress, calorimetric, and optical properties of metal hydrides
as they are exposed to H2. We use this apparatus to measure the properties of 5 dif-
ferent pure metal hydrides (Pd, Mg, Ti, V, and Zr) and then use these properties to
design tunable color lters and switchable perfect absorbers, among other devices.
To widen our parameter space and to combine desirable characteristics of dierent
metal systems, we use the same apparatus to investigate the properties of dier-
ent metal alloy hydride systems including Pd-Au, Mg-Ni, Mg-Ti, and Mg-Al. We
demonstrate many improved nanophotonic designs with these materials, including
a thin lm physical encryption scheme with Pd-Au and a switchable solar absorber
with Mg-Ti.
Many of these photonic devices can be further enhanced by tailoring the sub-
strate of the device along with the metal hydride. In this dissertation, we also
investigate combining the switchable optical properties of metal hydrides with near-
zero-index substrates to further enhance the optical device changes. Near-zero-index
materials are ones where the refractive index is below 1 and can lead to a variety of
interesting optical eects, including high absorption in surrounding materials and
enhanced non-linear eects. By combining an ITO substrate with a near-zero-index
resonance at ?1250 nm with a thin Pd capped Mg lm, we demonstrate a switch-
able absorption device with >76% absorption change at 1335 nm illumination. To
further explore the possibility of large-scale fabrication of these devices, we survey
the properties of commercially available near-zero-index materials and report the
range of attainable optical properties, showing its feasibility.
METAL HYDRIDES AS A PLATFORM FOR RECONFIGURABLE
PHOTONIC AND PLASMONIC ELEMENTS
by
Kevin J. Palm
Dissertation submitted to the Faculty of the Graduate School of the
University of Maryland, College Park in partial fulllment
of the requirements for the degree of
Doctor of Philosophy
2021
Advisory Committee:
Professor Jeremy N. Munday, Chair/Advisor
Professor James Williams
Professor Daniel Lathrop
Professor Edo Waks
Professor Thomas Murphy
? Copyright by
Kevin J. Palm
2021
Dedication
I dedicate the following thesis to Rachel for her unwavering support throughout
this journey.
ii
Acknowledgments
There are many people to thank for their contributions to this thesis along
with my development as a scientist. First, I would like to thank my committee
for attending my defense: Professors James Williams, Daniel Lathrop, Edo Waks,
and Thomas Murphy. I would like to thank my advisor Professor Jeremy Mun-
day for supporting me throughout my graduate studies and always making time to
discuss either research or personal issues, editing many manuscripts and fellowship
applications, and aiding in my development as a researcher and as a person. I owe
a special thanks to Joe Murray and Tarun Narayan for mentoring me as a young
graduate student working on my rst research project. They taught me to never
stop asking questions and how to persevere through the tough times in a project
when it seems like nothing will ever work properly. I would not be the scientist I
am today without their teaching and friendship. I'd also like to thank all of my
fellow Munday Lab members for their guidance, assistance, and camaraderie along
the way: Joe Garrett, David Somers, Lisa Krayer, Sarvenaz Memarzadeh, Tristan
Deppe, and Jongbum Kim.
I would like to thank other professors I have worked with throughout my
Ph.D., including Professors Marina Leite, Yet-Ming Chang from MIT, and Curtis
Berlinguette from UBC. I also want to give a special thanks to Thomas Schenkel at
LBNL along with Matt Trevithick, David Fork, and Ross Koningstein from Google.
Working with these researchers helped to broaden my scientic horizon and hone
my skepticism when analyzing data.
iii
I'd like to thank the sta of IREAP for all of their help. Judi Gorski, Nancy
Boone, Taylor Prendergast, Dottie Brosius, and Leslie Delabar made sure that I was
fully equipped and supported throughout my studies. None of my research would
have been possible without all of the help that I received from the Fablab sta.
Tom Loughran, John Abrahams, Mark Lecates, and Jon Hummel all spent many
hours teaching me how to use all of the various tools in the cleanroom and helping
me to develop and perfect all of the fabrication and measurement methods for my
samples.
Finally, I would like to thank my family and friends. My parents and siblings
have been supporting me and pushing me in my schooling since the very beginning.
I would like to thank my Grandma and Grandad whose support has allowed me
to come this far in my schooling. I wouldn't have survived my time as a graduate
student without the fun and shenanigans with my housemates during my years
living at the 7 Seas: Antony Speranza, Chris Flower, Jon Curtis, Jaron Shrock,
Zach Eldridge, and Zack Castillo. Lastly, I want to thank Rachel who has been by
my side through all the ups and the many downs of creating this dissertation.
Parts of the work in this thesis were supported by grants from Google LLC
and the National Science Foundation. I was personally supported by a National
Defense Science and Engineering Graduate Fellowship for the nal 3 years of my
studies.
iv
Table of Contents
Dedication ii
Acknowledgements iii
Table of Contents v
List of Tables viii
List of Figures ix
List of Abbreviations xii
Publications xiii
Chapter 1: Introduction 1
1.1 Overview of metal hydride systems . . . . . . . . . . . . . . . . . . . 1
1.2 Tunable optical properties and applications . . . . . . . . . . . . . . . 4
1.3 Outline of this thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Chapter 2: Experimental Apparatus Design and Qualication 11
2.1 Introduction to QCM sensing . . . . . . . . . . . . . . . . . . . . . . 12
2.2 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.2.1 Stress measurement . . . . . . . . . . . . . . . . . . . . . . . . 14
2.2.2 Optical properties measurement . . . . . . . . . . . . . . . . . 15
2.2.3 Calorimetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.3 Instrument design and description . . . . . . . . . . . . . . . . . . . . 17
2.4 Demonstration of operation and stability . . . . . . . . . . . . . . . . 33
2.5 Example of stress and mass change measurements . . . . . . . . . . . 33
2.6 Calorimetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
2.7 Optical property measurement . . . . . . . . . . . . . . . . . . . . . . 49
2.7.1 Optical tting . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
2.8 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
Chapter 3: Dynamic Optical Properties of Pure Metal Hydrides 56
3.1 Introduction to metal hydride optical properties . . . . . . . . . . . . 57
3.2 Optical, loading, and stress characterization of pure metal hydrides . 59
3.2.1 Pd/PdHx . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
v
3.2.2 Mg/MgHx . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
3.2.3 Zr/ZrHx . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
3.2.4 Ti/TiHx . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
3.2.5 V/VHx . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
3.3 Eect of annealing Ti on TiHx hydrogenation . . . . . . . . . . . . . 67
3.4 Characterization of Pd hysteresis . . . . . . . . . . . . . . . . . . . . 68
3.5 Tunable nanophotonics with metal hydrides . . . . . . . . . . . . . . 70
3.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
3.7 Experimental methods . . . . . . . . . . . . . . . . . . . . . . . . . . 79
3.7.1 Sample fabrication . . . . . . . . . . . . . . . . . . . . . . . . 79
3.7.2 Optical measurement . . . . . . . . . . . . . . . . . . . . . . . 80
3.7.3 Loading measurement . . . . . . . . . . . . . . . . . . . . . . 81
Chapter 4: Investigation of Physical Properties of Commercial Near-Zero-
Index Materials 83
4.1 Background of NZI materials . . . . . . . . . . . . . . . . . . . . . . . 83
4.2 Measurement and optical modeling scheme . . . . . . . . . . . . . . . 85
4.3 Properties of TCO lms . . . . . . . . . . . . . . . . . . . . . . . . . 87
Chapter 5: Highly Switchable Absorption in a Metal Hydride Device Using
a Near-Zero-Index Substrate 93
5.1 Background and introduction . . . . . . . . . . . . . . . . . . . . . . 94
5.2 Concept and design . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
5.2.1 Modeling substrate as Drude material . . . . . . . . . . . . . . 104
5.2.2 Angular dependence of device . . . . . . . . . . . . . . . . . . 108
5.3 Experimental demonstration . . . . . . . . . . . . . . . . . . . . . . . 111
5.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
Chapter 6: In situ Optical and Stress Characterization of Alloyed
PdxAu1?x Hydrides 120
6.1 Background of Pd-Au alloys . . . . . . . . . . . . . . . . . . . . . . . 121
6.2 Experimental methods . . . . . . . . . . . . . . . . . . . . . . . . . . 124
6.2.1 Fabrication and characterization of PdxAu1?x thin lms . . . 124
6.2.2 Optical property measurements . . . . . . . . . . . . . . . . . 126
6.2.3 Hydrogen loading and stress measurements . . . . . . . . . . . 127
6.3 Dynamic optical property measurements of PdxAu1?x alloys . . . . . 128
6.4 Material property measurements . . . . . . . . . . . . . . . . . . . . . 136
6.5 Simulations of optical switching . . . . . . . . . . . . . . . . . . . . . 144
6.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
Chapter 7: Optical Tunability Characterization of Mg-Ni, Mg-Ti, and Mg-Al
Alloy Hydrides 152
7.1 Introduction to Mg alloys . . . . . . . . . . . . . . . . . . . . . . . . 152
7.2 Experimental methods . . . . . . . . . . . . . . . . . . . . . . . . . . 156
7.3 Optical properties of Mg alloy hydrides . . . . . . . . . . . . . . . . . 160
vi
7.3.1 Mg-Al hydrides . . . . . . . . . . . . . . . . . . . . . . . . . . 160
7.3.2 Mg-Ti hydrides . . . . . . . . . . . . . . . . . . . . . . . . . . 163
7.3.3 Mg-Ni hydrides . . . . . . . . . . . . . . . . . . . . . . . . . . 165
7.4 Stress and loading properties . . . . . . . . . . . . . . . . . . . . . . . 169
7.5 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
7.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
Chapter 8: Conclusions and Future Experiments 180
8.1 High entropy metal hydrides . . . . . . . . . . . . . . . . . . . . . . . 181
8.2 Nuclear plasmonics with metal hydrides . . . . . . . . . . . . . . . . . 184
Appendix A:Curvature to frequency derivation 187
Appendix B: Commercial NZI Materials Optical Properties 192
Appendix C: H2 Safety Protocols 200
Appendix D:H2 Sensors 202
Appendix E: CIE 1931 Color Space Calcuations 207
Appendix F: Useful Properties of Metal Hydrides and Hydrogen 210
Bibliography 213
vii
List of Tables
2.1 Apparatus specications and measurement precision . . . . . . . . . . 18
2.2 Example t parameters for a calorimetry calibration run . . . . . . . 46
3.1 Previously published work on the measured optical properties of metal
hydrides . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
6.1 Compiled RMS roughnesses of PdxAu1?x upon hydrogenation . . . . 145
B.1 Summary of commercial NZI data #1 . . . . . . . . . . . . . . . . . . 198
B.2 Summary of commercial NZI data #2 . . . . . . . . . . . . . . . . . . 199
viii
List of Figures
2.1 Overview of measurement apparatus . . . . . . . . . . . . . . . . . . 19
2.2 Schematic of environmental pressure chamber . . . . . . . . . . . . . 20
2.3 Apparatus gas ow system design . . . . . . . . . . . . . . . . . . . . 21
2.4 Gas heat exchanger design . . . . . . . . . . . . . . . . . . . . . . . . 22
2.5 Apparatus optical system design . . . . . . . . . . . . . . . . . . . . . 24
2.6 Apparatus RTD design . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.7 Sample fabrication process . . . . . . . . . . . . . . . . . . . . . . . . 27
2.8 RTD resistances as a function of measured chamber temperature . . . 28
2.9 RTD and current sensing schematic . . . . . . . . . . . . . . . . . . . 29
2.10 Schematic for the RTD current source and voltage measurement system 30
2.11 Relation of thermal conductivity to QCM resistance . . . . . . . . . . 32
2.12 Apparatus stability plots . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.13 Extracting curvature measurements from interference image . . . . . 35
2.14 Filtering raw interference pattern . . . . . . . . . . . . . . . . . . . . 36
2.15 Mass loading calculation example . . . . . . . . . . . . . . . . . . . . 39
2.16 Calorimetry model design . . . . . . . . . . . . . . . . . . . . . . . . 41
2.17 Modeled and measured calorimetry data for a thin Cr lm . . . . . . 45
2.18 Ellipsometry chamber lid . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.1 Dynamic dielectric function upon hydrogenation of Pd, Mg, Zr, Ti,
and V . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
3.2 Dynamic index of refraction upon hydrogenation of Pd, Mg, Zr, Ti,
and V . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
3.3 Optical properties of the metal hydrides as a function of hydrogen to
metal ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
3.4 Stress change upon hydrogenation for pure metals . . . . . . . . . . . 63
3.5 Eects of annealing on the optical properties of Ti and TiHx . . . . . 68
3.6 Optical and loading response of Pd during hydrogen cycling . . . . . 70
3.7 Dierential scattering cross sections for metal hydrdie nanoparticles . 72
3.8 Relative scattering cross sections (Mie eciency) of metal nanopar-
ticles and their hydrides in free space . . . . . . . . . . . . . . . . . . 72
3.9 Relative change in transmission upon hydrogenation of periodic nanorod
arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
3.10 Full transmission spectrum of periodic nanorod array . . . . . . . . . 74
3.11 Switchable perfect absorbers and tunable color lters using Mg, Pd,
and Ti . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
ix
3.12 Switchable perfect absorbers and tunable color lters using V and Zr 77
4.1 Characteristic optical properties of TCO lms . . . . . . . . . . . . . 87
4.2 Strength of NZI resonance vs the wavelength at which the resonance
occurs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
4.3 Bandwidth of NZI resonance vs the center location of the resonance . 90
4.4 Strength of NZI resonance vs the resistivity of the lm . . . . . . . . 92
5.1 Device design and simulated absorption spectra for various substrates 97
5.2 Simulated reection and transmission plots for Mg/NZI device with
350 nm NZI substrate . . . . . . . . . . . . . . . . . . . . . . . . . . 98
5.3 Thickness optimization of NZI substrate . . . . . . . . . . . . . . . . 99
5.4 Thickness optimization for Mg and Pd thin lm layers . . . . . . . . 100
5.5 Simulated absorption by layer in the Mg/NZI device stack . . . . . . 103
5.6 Eect of the substrate's optical properties on device absorption change104
5.7 Fresnel reectance R versus the imaginary part of the index of refrac-
tion of the NZI substrate . . . . . . . . . . . . . . . . . . . . . . . . . 105
5.8 Change in absorption upon hydrogenation using a Drude model for
the NZI material . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
5.9 Eect of ? parameter in Drude model on absorption change when
compared with the damping . . . . . . . . . . . . . . . . . . . . . . . 109
5.10 Eect of ? parameter in Drude model on absorption change when
compared with the plasma wavelength . . . . . . . . . . . . . . . . . 110
5.11 Angular dependence of incident light on absorption change of Mg/NZI
structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
5.12 Angular dependence of absorption of Mg/NZI device in the hydride
state for dierent illumination wavelengths . . . . . . . . . . . . . . . 112
5.13 Measured optical properties of ITO used in experiments . . . . . . . . 113
5.14 Experimental demonstration of Mg/NZI device performance . . . . . 114
5.15 Simulated transmission intensity dierence between normal and 20?
illumination for the experimental device . . . . . . . . . . . . . . . . 115
5.16 Dynamic transmission data for Mg/ITO device . . . . . . . . . . . . . 116
5.17 Reversibility of the Mg/ITO device switchable absorption . . . . . . . 117
6.1 Measured EDX spectra of fabricated PdxAu1?x alloys . . . . . . . . . 125
6.2 Measured optical properties of PdxAu1?x alloys . . . . . . . . . . . . 129
6.3 Measured optical properties of PdxAu1?x alloys: First Load . . . . . . 130
6.4 Change in dielectric functions of PdxAu1?x alloys upon hydrogenation 131
6.5 Comparison of modeled Pd and Au optical data with literature . . . . 132
6.6 Dynamic optical properties of PdxAu1?x alloys upon hydrogenation . 133
6.7 Optical response of PdxAu1?x versus hydrogen loading . . . . . . . . 134
6.8 Relationship of n and k of PdxAu1?x alloys at 1500 nm illumination . 135
6.9 Hydrogen cycling properties of alloys . . . . . . . . . . . . . . . . . . 136
6.10 Total sorption data of PdxAu1?x alloys . . . . . . . . . . . . . . . . . 138
x
6.11 Comparison of total hydrogen content per metal atom of PdxAu1?x
alloys at dierent hydrogen partial pressures . . . . . . . . . . . . . . 139
6.12 Stress characterization of PdxAu1?x alloys . . . . . . . . . . . . . . . 140
6.13 Stress dependence on atomic Pd % . . . . . . . . . . . . . . . . . . . 141
6.14 PdxAu1?x thin lm morphology characterization . . . . . . . . . . . . 143
6.15 Roughness scans upon hydrogenation for ve PdxAu1?x samples pre-
pared under dierent conditions . . . . . . . . . . . . . . . . . . . . . 144
6.16 Simulated reectance shifts upon hydrogenation for grating structures 146
6.17 NIR simulations of the reectivity of grating structures . . . . . . . . 147
6.18 PdxAu1?x alloy physical encryption scheme . . . . . . . . . . . . . . . 149
7.1 Optical properties change of Mg-Al alloys upon hydrogenation . . . . 161
7.2 Dynamic intermediate optical properties upon hydrogenation of Mg-
Al alloys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
7.3 Optical properties change of Mg-Ti alloys upon hydrogenation . . . . 163
7.4 Dynamic intermediate optical properties upon hydrogenation of Mg-
Ti alloys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
7.5 Optical properties change of Mg-Ni alloys upon hydrogenation . . . . 166
7.6 Dynamic intermediate optical properties upon hydrogenation of Mg-
Ni alloys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
7.7 Modeled absorption from backside illumination of Mg0.73Ni0.27 . . . . 169
7.8 Measured maximum loading values of thin lm Mg alloys . . . . . . . 170
7.9 Measured total stress values of thin lm Mg alloys . . . . . . . . . . . 172
7.10 Simulated switchable window performance with Mg alloys . . . . . . 174
7.11 Simulation of broadband switchable light absorption with Mg-Ti alloys176
7.12 Simulation of Mg alloy/ITO switchable absorption device . . . . . . . 177
B.1 ITO optical properties #1 . . . . . . . . . . . . . . . . . . . . . . . . 193
B.2 ITO optical properties #2 . . . . . . . . . . . . . . . . . . . . . . . . 194
B.3 ITO optical properties #3 . . . . . . . . . . . . . . . . . . . . . . . . 195
B.4 FTO optical properties . . . . . . . . . . . . . . . . . . . . . . . . . . 196
B.5 AZO optical properties . . . . . . . . . . . . . . . . . . . . . . . . . 197
F.1 Periodic table classication of metal hydrides . . . . . . . . . . . . . . 211
F.2 H-H atom spacing in dierent materials . . . . . . . . . . . . . . . . . 212
xi
List of Abbreviations
ADC Analog to Digital Converter
AFM Atomic Force Microscopy
AZO Aluminum-doped Zinc Oxide
CSID Calorimetry System Identication
EDX Energy-dispersive X-ray Spectroscopy
EELS Electron Energy Loss Spectroscopy
EMA Eective Medium Approximation
ENZ Epsilon-Near-Zero
FDTD Finite-Dierence Time-Domain
FTO Fluorine-doped Tin Oxide
HEA High Entropy Alloy
ICP-OES Inductively Coupled Plasma - Optical Emission Spectroscopy
IREAP Institute for Research in Electronics and Applied Physics
ITO Indium Tin Oxide
IR Infrared
LSPR Localized Surface Plasmon Resonance
MFC Mass Flow Controller
MSE Mean-Squared-Error
NIR Near-Infrared
NZI Near-Zero-Index
PECVD Plasma Enhanced Chemical Vapor Deposition
PWM Pulse Width Modulation
QCM Quartz Crystal Microbalance
RMS Root Mean Square
RTD Resistance Temperature Device
TCO Transparent Conducting Oxide
TE Transverse Electric
TM Transverse Magnetic
TMM Transfer-Matrix Method
xii
List of Publications
Portions of this thesis have been drawn from the publications listed below:
J. B. Murray,K. J. Palm, T. C. Narayan, D. K. Fork, S. Sadat, J. N. Munday.
Apparatus for combined nanoscale gravimetric, stress, and thermal measurements
Review of Scientic Instruments 89, 085106 (2018)
K. J. Palm, J. B. Murray, T. C. Narayan, J. N. Munday. Dynamic optical
properties of metal hydrides ACS Photonics 5, 4677-4686 (2018)
K. J. Palm, J. B. Murray, J.P. McClure, M. S. Leite, J. N. Munday. In
situ optical and stress characterization of alloyed PdxAu1-x hydrides ACS Applied
Materials and Interfaces 11, 45057-45067 (2019)
K. J. Palm, L. J. Krayer, J. N. Munday. Highly switchable absorption in a
metal hydride device using a near-zero-index substrate Submitted (2021)
Other publications co-authored during the course of the thesis include:
J. L. Garrett, L. J. Krayer, K. J. Palm, J. N. Munday. Eect of lateral tip
motion on multifrequency atomic force microscopy Applied Physics Letters 111,
043105 (2017)
D. A. T Somers, J. L. Garrett, K. J. Palm, J. N. Munday. Measurements
of the Casimir torque Nature 564, 386-389 (2018)
L. J. Krayer, K. J. Palm, C. Gong, A. Torres, C. E. P. Villegas, A. R. Rocha,
M. S. Leite, J. N. Munday. Enhanced near-infrared photoresponse from nanoscale
AgAu Alloyed lms ACS Photonics 7 (7), 1689-1698 (2020)
S. Memarzadeh, K. J. Palm, T. E. Murphy, M. S. Leite, J. N. Munday.
Control of hot-carrier relaxation time in Au-Ag thin lms through alloying Optics
Express 28 (22), 33528-33537 (2020)
T. Gong, P. Lyu, K. J. Palm, S. Memarzadeh, J. N. Munday, M. S. Leite.
Emergent opportunities with metallic alloys: From material design to optical de-
vices Advanced Optical Materials, 2001082 (2020)
J. M. Howard, K. J. Palm, Q. Wang, E. M. Tennyson, B. Roose, E. Lee, A.
Abate, J. N. Munday, M. S. Leite. Water-induced photoluminescence dynamics in
triple-cation metal halide perovskites Submitted. (2021)
xiii
Finally, I plan to submit several sections of this thesis to academic journals:
K. J. Palm, C Shelden, L. J. Krayer, J. N. Munday. Investigation of physical
properties of commercial near-zero-index materials In Preparation (2021)
K. J. Palm, M. S. Leite, J. N. Munday. Optical tunability characterization
of Mg-Ni, Mg-Ti, and Mg-Al alloys In Preparation (2021)
xiv
Chapter 1: Introduction
1.1 Overview of metal hydride systems
Metal hydrides have a long history of discovery and application. Pd, the most
commonly studied and used metal hydride, was discovered to absorb large amounts
of hydrogen in 1866 [1]. Since this discovery, there have been numerous applications
of these materials including membranes to extract or purify hydrogen [2, 3], hydrogen
gas sensors [4, 5], hydrogen storage devices [6, 7], and rechargeable batteries [8],
amongst many others. All of these applications take advantage of one or more
of the physical property changes of certain metals upon hydrogenation, including
changes in electrical resistivity, optical properties, lattice expansion, and thermal
conductivity. Many metal hydrides are non-stoichiometric compounds, where the
amount of hydrogen in the material is dependent on the driving force (pressure or
electrochemical voltage) of the hydrogen on the lattice [9]. Thus, these material
changes are on a dynamic spectrum and allow for a broad range of intermediate
properties between the fully metallic and hydride states. In this thesis, we will use
novel methods to more precisely characterize these dynamic properties to inform
the design of nanophotonic and plasmonic devices to further propel applications of
metal hydrides.
1
First, we begin with a physical picture of the hydrogenation reaction. The two
common ways of hydrogenating metals are gas-phase loading and electrochemical
loading. The hydrogen loading value is a measure of how many H atoms are located
within the metal lattice, and is measured in the units of H/M, or the number of
hydrogen atoms per metal atom in the lattice. For gas-phase loading, as certain
metals are exposed to H2 gas, the diatomic hydrogen molecule can adhere to a sur-
face site on the metal and subsequently dissociate into individual hydrogen atoms.
These hydrogen atoms can then diuse into the bulk of the metal, occupying in-
terstitial sites within the metal lattice. In many metals, this gas-phase reaction is
not spontaneous at room temperature and reasonable H2 pressures. These metals
require a catalyst to split the diatomic hydrogen molecule, which then the atomic
hydrogen atoms can diuse through the catalytic lm into the lattice sites of the
substrate metal. The most common catalyst is Pd, which also has the added ben-
et of protecting the substrate lm from oxidation when it fully encapsulates the
substrate. Electrochemical loading occurs when the metal is placed in an aqueous
(usually acidic) medium and hydrogen is driven into the lattice with an applied
voltage. This process generally allows for higher hydrogen loadings to be achieved,
because a relatively small applied voltage in an electrochemical cell is chemically
equivalent to a very high H2 gas pressure [10]. For example, a Pd loading of H/Pd
= 0.96 has been achieved with electrochemical loading, when gas phase loadings can
generally only achieve H/Pd ? 0.75 even at high pressures [10, 11]. The applications
of this process are limited to devices that can operate in an aqueous cell, thus in
this thesis, we will focus solely on the gas phase loading reaction.
2
As small amounts of hydrogen enter the interstitial sites of a metal lattice,
the hydrogen initially occupies the ? material phase describing very low hydrogen
concentrations. As the lattice lls with more hydrogen, the material enters a mixed-
phase (? + ?), which is caused by H-H interactions becoming more dominant [4].
The ? phase continues to grow in the lattice as more hydrogen is added until it be-
comes the only phase in the material. As this phase transition occurs, a signicant
amount of stress is introduced into the lattice as each hydrogen interstitial causes
displacements of the metal atoms from their usual lattice sites. This displacement
causes a distortion of the crystal lattice. A signicant volume expansion occurs
during this process, with Pd achieving a 14% expansion under 1 atm H2 [12], while
other materials have an even greater expansion, such as Mg, which has a 30% expan-
sion [13]. Taking into account these stresses, especially in constrained systems, is
very important, as they are connected with spurious phase formations. High stresses
beyond the critical yield stress of the material can also result in the formation of dis-
locations in the lattice, which further aect its optical and structural properties [12].
Ways to mitigate these high stress values include nanoscaling the active materials
to have fewer constraints on the system and alloying the active metals with other
transition metals in order to suppress the ? to ? phase transition. This mitigation
is essential for systems that desire repeatable switchable responses, as dislocations
and deformations to the lattice can signicantly degrade device response over many
H2 cycles.
3
1.2 Tunable optical properties and applications
In this thesis, we focus most of our attention on the dynamically tunable
optical properties of metal hydrides, compared with their other dynamic proper-
ties. Metal hydrides started to draw signicant attention for their tunable optical
properties in 1996 when Huiberts et al. discovered that yttrium and lanthanum
demonstrate a metal-insulator transition when exposed to H2 gas [14]. This tran-
sition caused a dramatic change in the materials' optical properties as the metallic
YH2 and LaH2 transitioned to the semiconducting YH3 and LaH3. Further work
was continued investigating the optical changes in thin-lm metal hydrides and how
these changes can be applied to dierent devices, such as for smart windows [15],
switchable solar absorbers [16], hydrogen sensors [17], and switchable mirrors [18]
using metal hydrides including Y, Mg, Pd, and Mg-Ni alloys respectively.
Beyond thin lm applications, as nanofabrication techniques improved and
nanostructuring metals became more commonplace, many groups began to study
nanoscaled metal hydrides. Nanoscaling allows for ner control of the optical re-
sponse of a device by tailoring the light-matter interaction. One common way of
doing this is to create a localized surface plasmon resonance (LSPR) on the metal
structures. The LSPR occurs when the oscillating electric eld from incident light
causes the conduction electrons in the metal to coherently oscillate. This process
allows for high absorption at the LSPR resonance wavelength. By exposing this
plasmonic metal structure to H2, you can then either shift or eliminate this reso-
nance, as has been demonstrated using Y [19] and Nb [20] nanorods and Mg [21]
4
and Pd-Au [5] nanodisks.
Recently, the primary application for these optical metal hydride devices has
been for optical hydrogen sensing. H2 gas is an ideal clean energy source candidate
because when burned, it emits no greenhouse gases and can be easily transported to
regions where high-voltage power lines cannot reach. With the increase in H2 usage,
high-quality sensors are needed to regulate the ammability risks. In particular,
optical H2 sensors are ideal due to a decreased risk of sparking in a ammable H2
environment, as opposed to electrical sensors that utilize metal hydrides' change in
resistance upon hydrogenation. They are also much less energy and space-intensive
than combustive H2 sensors that section o and combust a portion of the chamber
gas and analyze the resulting atomic spectra. Optical H2 sensors can be imple-
mented in many dierent forms. The most common current design utilizes the
LSPR resonance of nanostructured metal hydrides. By tting the peak of the LSPR
resonance, the amount of hydrogen in the atmosphere above the sensor can be de-
termined [5, 22, 23]. Another common way of implementing an optical H2 sensor
is to use a coating on an optical ber that either uses a thin metal hydride layer
at the end of the ber to detect changes in back-reected light [24] or uses a full
coating of a metal hydride that expands or contracts with the hydrogen content of
the atmosphere, changing the eective optical path length of the ber [25, 26]. Sim-
pler designs for hydrogen sensors have also been proposed that utilize resonances in
thin-lm devices by creating a cavity eect [17].
Beyond adjusting the optical properties of these materials by only changing
the H loading, alloying metals has the added benet of a wider parameter space
5
for potential optical properties. Alloying allows for the tuning of the initial optical
properties of the metal, along with its response to hydrogenation. In the case of the
Mg-Ni system, alloying has been used to tune the nal properties of the hydride to be
more clear for switchable window applications [15]. Alloying also has the potential
to solve many other issues with metal hydride devices. Many pure metal H2 sensors
suer from intracycle hysteresis, which causes a large ambiguity in the H2 pressure
reading depending on whether the sensor is loading or unloading. These sensors can
also suer from surface poisoning from various trace gases in the atmosphere, such as
CO. Surface poisoning occurs when a molecule binds to a H dissociation site on the
H2 sensor and blocks any further hydrogen splitting at this site. In particular, Pd-
based sensors suer from this, where the C atom in the CO molecule chemisorbs to H
splitting sites [27, 28]. One example of an alloying system that begins to solve these
issues is Pd-Au, which has been shown to eliminate hysteresis with high enough Au
atomic fractions and has shown a signicant reduction of surface poisoning [5, 23].
To oer an even higher resistance to poisoning, ternary systems like Pd-Au-Cu have
been explored and have shown a complete elimination of this poisoning eect [22].
Finally, some metal hydrides are too stable in their fully hydrogenated state and
must be heated in order to unload the hydrogen from the metal lattice. MgH2 in
particular suers from this eect, and much work has been done to investigate ways
to destabilize the lattice for switchable room temperature desorption for hydrogen
storage purposes [29, 30, 31]. Alloying has the potential to solve all of these issues
by retaining the large optical changes upon hydrogenation while introducing other
positive eects by alloying with other transition metals.
6
1.3 Outline of this thesis
Before we can take advantage of the wide parameter space of tunable optical
and structural properties of metal hydrides, we rst need to build and qualify an
apparatus that can characterize these systems. In Chapter 2, we describe the cus-
tom measurement system designed to measure the gravimetric, optical, thermal, and
stress properties of thin-lm and nanoscaled metal hydrides. This system is based
around a Quartz Crystal Microbalance (QCM) that is used to measure the mass of
hydrogen entering the metals under investigation. The stress of thin-lm samples
is measured using an adopted Michelson-Morely interferometer with the QCM act-
ing as one of the interferometer mirrors. By measuring the change in interference
spacing, the curvature change of the substrate can be calculated, which can then be
converted to the stress of the lm. To measure thermal signals from reactions on
the lm, we designed on-chip Resistance Temperature Devices (RTD) whose out-
puts are fed into a one-state nonlinear lumped element model. Finally, multi-angle
optical ports are available in the system to incorporate dynamic optical property
measurements with variable angle spectroscopic ellipsometry. This chapter is based
on the published manuscript J. B. Murray, K. J. Palm, T. C. Narayan, D. K. Fork,
S. Sadat, J. N. Munday. Apparatus for combined nanoscale gravimetric, stress, and
thermal measurements Review of Scientic Instruments 89, 085106 (2018).
With the demonstrated ability to measure the properties of these metal hy-
drides, we then investigate the structural and optical properties of 5 pure metals,
Pd, Mg, V, Zr, and Ti, in Chapter 3. We nd a range of dierent optical responses
7
upon hydrogenation from the dierent metals, and we use these measured dynamic
optical properties to propose dierent thin lm and nanophotonic designs including
switchable perfect absorbers and color lters utilizing the metal hydride transition.
This chapter is based on the published manuscript K. J. Palm, J. B. Murray, T.
C. Narayan, J. N. Munday. Dynamic Optical Properties of Metal Hydrides ACS
Photonics 5, 4677-4686 (2018).
In Chapter 4, we take a brief detour from metal hydride systems to explore
commercially available Near-Zero-Index (NZI) materials. NZI substrates can greatly
enhance optical eects in the wavelength range around the NZI resonance, and we
are interested in the range of potential properties of these materials for eventual
combination with a metal hydride system. We investigate the range of properties
that can be obtained with three dierent types of Transparent Conducting Oxides
(TCO): indium tin oxide (ITO), uorine-doped tin oxide (FTO), and aluminum-
doped zinc oxide (AZO). We nd positive correlations between the locations of the
NZI resonances and the strengths and bandwidths of those resonances. This chapter
is based on a manuscript currently in preparation K. J. Palm, Calum Shelden, L.
J. Krayer, J. N. Munday. Investigation of Physical Properties of Commercial Near-
Zero-Index Materials In Preparation (2021).
Utilizing the unique properties of these NZI materials, in Chapter 5 we use
an NZI substrate under a Pd/Mg stack to create a switchable thin-lm absorber.
The device can be switched from a high reectivity to a high absorption state by
exposure to H2 gas. We show that the NZI substrate is essential to obtaining the
extremely high absorption change and that the high absorption is created by a
8
destructive interference eect with the light reecting o the Mg/NZI boundary.
We experimentally demonstrate this device with a 350 nm ITO lm coated with 25
nm Mg and 3 nm Pd, which shows an absorption change >76%. This chapter is
based on a recently submitted manuscript K. J. Palm, L. J. Krayer, M. S. Leite,
J. N. Munday. Highly switchable absorption in a metal hydride device using a
near-zero-index substrate Submitted (2021).
Although pure metals have a wide range of uses in optical metal hydride de-
vices, there are a limited set of materials to choose from, and many of these metals
suer from deleterious eects such as intracycle hysteresis, degradation over multi-
ple cycles, and surface poisoning from atmospheric gases such as CO. In order to
limit these negative eects and exactly tune our optical properties to the desired
results, we can alloy our metal hydrides with other transition metals. In Chapter 6,
we investigate the optical and structural properties of the Pd-Au alloy system and
demonstrate their use as H2 gas sensors and physical encryption devices. By alloy-
ing Pd with Au, we reduce sensor hysteresis and increase chemical resistance, while
still maintaining a measurable optical signal. This chapter is based on the published
manuscript K. J. Palm, J. B. Murray, J.P. McClure, M. S. Leite, J. N. Munday.
In situ Optical and Stress Characterization of Alloyed PdxAu1-x Hydrides ACS
Applied Materials and Interfaces 11, 45057-45067 (2019).
Mg alloys are another material system of interest due to their ability to in-
crease the H absorption and desorption kinetics when compared to pure Mg and
still maintain large optical changes. This increase in kinetics is caused by the hy-
dride state being destabilized when dierent transition metals are introduced into
9
the Mg lattice. In these systems, a large range of potential optical properties can
be obtained. In Chapter 7, we investigate the optical properties of three dierent
Mg alloy systems, Mg-Ti, Mg-Ni, and Mg-Al, to determine the eects of dierent
atomic percentages on the optical and loading properties of the system. We show
that the Mg-Ti system in particular works well as both a switchable mirror and
broadband switchable light absorbers. This chapter is based on a manuscript cur-
rently in preparation K. J. Palm, M. S. Leite, J. N. Munday. Optical tunability
characterization of Mg-Ni, Mg-Ti, and Mg-Al alloys In Preparation (2021).
Finally, Chapter 8 concludes this thesis with closing remarks about the impact
of the work and future directions. In this chapter, we discuss further research areas
in dierent optical metal hydride systems, including high entropy alloys. We also go
through future applications of metal hydrides beyond optical sensors and devices,
specically for use in plasmonic beam targets for enhancing light element nuclear
reactions.
10
Chapter 2: Experimental Apparatus Design and Qualication
In this chapter, we present the design and qualication of the apparatus that
will be used to characterize various metal hydride systems throughout the rest of
this thesis. This apparatus allows for the simultaneous measurement of mass change,
heat evolution, and stress of thin-lm samples deposited on QCMs. We show device
operation at 24.85 ? 0.05 ?C under 9.31 ? 0.02 bar of H2 as a reactive gas. Using
a 335 nm palladium lm, we demonstrate that our apparatus quanties curvature
changes of 0.001 m?1. Using the QCM curvature to account for stress-induced
frequency changes, we demonstrate the measurement of mass changes of 13 ng/cm2
in material systems exhibiting large stress uctuations. We use a one-state nonlinear
lumped element model to describe our system with thermal potentials measured at
discrete positions by three RTDs lithographically printed on the QCM. By inputting
known heat amounts through lithographically dened Cr/Al wires, we demonstrate
a 150 ?W calorimetric accuracy and 20 ?W minimum detectable power. We also
show that by switching out the chamber lid and coupling our environmental chamber
with a variable angle spectroscopic ellipsometer, we can combine our dynamic mass
measurement with in-situ dynamic optical property measurement. The capabilities
of this instrument allow for a more complete characterization of reactions occurring
11
in nanoscale systems, such as the eects of hydrogenation in various metal lms and
nanostructures, as well as for direct stress compensation in QCM measurements.
2.1 Introduction to QCM sensing
Chemistry in nanoscale systems is increasingly important in a wide variety of
elds from energy and information storage to catalysis and sensing [32, 33, 34, 35, 36,
37, 38, 39]. The shift away from the macroscale allows for dimensional reduction and
dramatic changes in surface-to-volume ratios, which in turn present opportunities
to tailor the thermodynamics and kinetics. Despite the small amount of material
present in a nanoscale system, reactions can still produce signicant amounts of heat
and stress that can change the chemical and physical properties of the material. As
such, it is crucial to quantify the mechanical, optical, and thermal properties of
these systems to inform the design of devices exploiting these chemical processes.
QCMs are commonly used to observe nanoscale chemical reactions. A QCM is
a quartz wafer (typically a disc) with its crystal orientation cut to produce a shear
displacement in the presence of an electric eld normal to its face. Applying an
oscillating electric eld between two electrodes on opposite sides of the QCM excites
a shear wave in the quartz disc due to its piezoelectric response. The resonance
frequency of this oscillation is very sensitive to the material attached to the surface
of the resonator, which causes a change in the acoustic impedance of that interface.
This frequency changes (very nearly) linearly with added rigid mass, such as a
thin metal lm, and can thus be used to detect changes of mass due to chemical
12
or physical processes [40, 41]. This mass sensitivity lends itself to a number of
applications as far reaching as protein sensing and electrochemical degradation [42].
The QCM resonance frequency also has a pronounced dependence on a num-
ber of other parameters including pressure, density and viscosity of the medium
surrounding the QCM, temperature, and stress (e.g. from mounting or from stress
in an adhered thin lm) [43, 44, 45]. In order to compensate for these myriad eects
when performing a mass measurement, it is necessary to couple multiple measure-
ment techniques to independently determine the other parameters to which the QCM
is sensitive. Doing so, however, not only results in a more accurate determination
of the mass change (e.g. by accounting for stress eects) but also simultaneously
provides a greater understanding of a chemical or physical process than a QCM
measurement alone could by leveraging the knowledge of sample stresses. As such,
QCM samples can be integrated into a modular experimental apparatus to not only
correctly determine mass changes, but also a range of complementary processes in
a chemical reaction.
Our system combines the QCM platform in a pressure and temperature con-
trolled environmental chamber with optical access, which allows in situ, high-speed,
stress measurements to properly characterize mass change. It also includes optical,
calorimetric, and electrical measurements for a more complete picture of chemical
reactions on nanoscale structures. Here we capitalize on the planar nature of these
devices to be used as interferometric mirrors, for measurement of stress by means
of sample curvature, as well as substrates for photolithographically dened RTDs
that can be used for sensing or introducing known amounts of heat for calorimetry
13
modeling. All of these measurements are made in a temperature controlled, variable
pressure reaction chamber. Below, we describe the components of this system and
demonstrate the system's stability and precision. We then apply this apparatus to
a palladium hydrogenation reaction as an example of operation.
2.2 Background
Each of the disparate measurement capabilities is motivated by the need to
form a complete picture of nanoscale chemical reactions. The particular choices
of techniques are driven by constraints of integration into our system. Below we
provide background and context for these individual measurements.
2.2.1 Stress measurement
In some of the earliest work measuring lm stress on QCMs, EerNisse showed
that quartz wafers cut along dierent crystal axes have signicantly dierent re-
lationships between resonance frequency and lm stress [46, 47]. A measurement
of the frequency changes on dierently-cut quartz wafers during hydrogenation of
a palladium lm deposited on the QCM, assuming the same hydrogen absorption,
thickness, and stress level in both lms, yielded a change in both stress and mass
associated with the reaction. However, preparing two identical lms is complicated
by the fact that samples grown at slightly dierent locations within an evaporator or
with slightly dierent currents by electrochemical means could result in variations of
lm thickness and defect density. Further, mass and stress do not necessarily scale
14
linearly with the sample thickness, as thermodynamic and kinetic properties often
change at small length scales. Thus, in our system, we perform simultaneous mass,
stress, and thermal measurements on a single sample to control for sample-to-sample
variation.
Measurements of stress not only allow for corrections in QCM measurements,
but can also yield signicant insight into material systems beyond quantication
of mass changes. For example, a combined stress and optical transmittance study
on the hydrogenation of 10 nm palladium lms on glass slide substrates revealed
a gradual removal of a surface oxide layer that is often not evident in studies of
hydrogen content in palladium [48]. A curvature measurement of electrochemical
lithiation of silicon has shown that the chemical potential of lithium in silicon is
heavily governed by the stress present in the material [49]; the joint measurement
scheme also clearly delineates the extent to which the reaction can proceed before
the lm undergoes plastic deformation, which can inform further engineering of the
Li/Si system.
2.2.2 Optical properties measurement
A number of studies have been conducted that combine optical measurement
techniques with a QCM to extract unique insights into a system, surpassing what
either technique could provide individually. Many of these studies focus on nanos-
tructured samples that have a plasmonic response that depends on the chemical
reactivity within the environment. For example, a study of the corrosion of copper
15
and aluminum nanoparticles was able to distinguish between two primary oxidative
corrosion mechanisms [50, 51]. Taken separately, neither the optical nor the QCM
measurements could have distinguished between the dierent processes. By allowing
optical access to the sample, our apparatus retains the ability to distinguish these
processes.
2.2.3 Calorimetry
The most sensitive QCM-based calorimetry relies on heat conduction calorime-
try [52]. This technique detects heat using a thermopile that is thermally grounded
on one side. The generated heat ows through the thermoelectric plate creating
a voltage by the Seebeck eect. This calorimeter has been shown to accurately
measure heat from thin-lm reactions. The main drawback is that heats arising
from dierent locations on a sample are treated equally. As such, it is challeng-
ing to distinguish between local and global events. Local photochemical processes
occurring upon laser illumination can have heat conduction pathways that dier
from those that homogeneously arise from the lm, complicating the calorimetric
analysis. Thus in our apparatus, we chose to perform calorimetry with multiple
RTDs in order to allow for the localization heat eects as opposed to the global heat
conduction calorimetry. Optical calorimetry is another way to achieve this localized
measurement [53], but due to its low resolution of approximately 1 K and its depen-
dence on a multitude of environmental factors (e.g. not only the refractive index
changes but changes in size, shape, environment, etc.), we found the RTDs were a
16
superior measurement scheme in this context.
2.3 Instrument design and description
Table 2.1 outlines our system specications. Our instrument design allows
us to perform on-chip calorimetry with QCM substrates in order to simultaneously
resolve changes in curvature of 0.001 m?1 (corresponding to a stress of 0.006 MPa
per micron of lm thickness), changes in mass of 13 ng/cm2, and changes in op-
tical reectivity of 0.3%, as well as measurement of heat with 150 ?W accuracy.
This system has pressure capabilities up to 9.3 bar and a temperature range of 15
to 35 ?C with stabilities of ? 0.02 bar and ? 0.05 ?C, respectively. Our approach
compensates for sample to sample variation by performing gravimetric, stress, and
thermal measurements simultaneously on a single QCM. Calorimetry is achieved
by modeling the outputs of lithographically printed Cr/Al resistance temperature
devices on the sample substrate. We use sample curvature, measured by an interfer-
ometer integrated into our microscope, to measure in-plane stress. This non-contact
method allows for accurate mass measurements by accounting for frequency changes
due to stress eects. The optical access also allows for an external optical source
such as a laser, ellipsometer, or spectrometer to be incorporated into the system.
Figure 2.1 shows a systems overview of our apparatus. Below we describe, in turn,
each of the subsystems: environmental control, optical excitation, interferometry,
mass measurement, and calorimetry.
The reactions occur in an environmental chamber capable of achieving and
17
Parameter Characteristic Values
Typical steady state 20 ?W
minimum detectable poweri
Measured Power Accuracyii 150 ?W
Approximate minimum detectable ?1 mK
temperature change in lmi
Minimum detectable ?0.1 H/M in a 5 nm lm
concentration of H
Operation temperature 15 ?C to 35 ?C
Temperature stability 50 mK over 1 hour
150 mK over 24 hours
Dierential temperature stabilityiii ? 4 mK
Operating pressureiv 1 - 10 atm
Illumination wavelength range 250 nm to 26,000 nm
Minimum detectable ? (curvature) 0.001 m?1
Stress sensitivity 0.006 MPA per micron
of lm thickness
Optical excitation laser wavelength 660 nm
Typical absorbed laser power 1-8 mW
Optical excitation laser repetition rate 100 Hz to 500,000 Hz
iDened as the RMS noise about the mean
iiAccuracy is dened as the power equivalent to the typical 10 hour drift
of RTDs
iiiBased on typical temperature coecients. Calorimetry is performed
without reference to a temperature
ivCan reach 40 atm without optical access
Table 2.1: Apparatus Specications
maintaining pressures up to 9.3 bar. The sample substrates are 25.4 mm diame-
ter, 5 MHz polished Cr/Au QCMs (Maxtek ?). The piezoelectric resonant fre-
quency of the QCM is measured by a QCM Driver (Stanford Research Systems
Model QCM200) using a 10 MHz Rb frequency standard (Stanford Research Sys-
tems SIM940). In our system, samples consist of either a 12.7 mm-diameter lm
or a nanoparticle array of the same area deposited on the center of the QCM.
Samples are mounted on custom machined Macor stages with electrodeposited Au
contacts to make electrical connection with the QCM. A custom circular array of
18
Figure 2.1: Systems overview. The samples are deposited onto a QCM substrate.
This QCM sits in an environmental chamber that controls the pressure, temperature,
and gas composition. The QCM also has RTDs deposited on its surface that are
electrically driven, and the reective top surface of the QCM is used as a mirror in
an adapted Michelson-Morley interferometer with a bandpass ltered LED used as a
partially coherent source (see Figure 2.5 for further details). The created interference
pattern is used to calculate the curvature of the sample. This setup allows for the
introduction of outside optical sources, such as the 660 nm diode laser depicted here.
Note that the actual chamber incorporates an additional reference QCM, which we
have excluded in this image for clarity (see Figure 2.2).
clip springs (Ted Pella 16399) is used to clamp the sample to the stage and provide
in situ electrical contacts for devices such as RTDs. The springs are contacted to a
custom-designed exible printed circuit board that is fed through a tube, which is
hermetically sealed with epoxy (3M Scotch-Weld DP125 Translucent). Figure 2.2
depicts the full environmental chamber. The sample under investigation is placed
on one of the stages, while a QCM without the active lm is mounted on the sec-
ond stage. This blank QCM allows for any ambient eects in the chamber, such as
vibrations or environmental changes, to be calibrated out from the active sample
19
data.
Figure 2.2: (a) Exploded schematic of the environmental pressure chamber. The
QCM samples are centered on a Macor stage with Teon pins and held in place with
a circular spring array. A Buna-N O-Ring provides the gas seal for the chamber.
The glass window, serving as the optical port, is axed to the chamber lid with
epoxy, creating a hermetic seal. Gas, uid, and wire feedthroughs are on the sides
of the chamber. (b) Lid on and (c) lid o schematic images of the assembled sample
chamber.
Figure 2.3 outlines the gas ow system of the apparatus. The gas ows into
the system through 1/4 high pressure nylon tubing (McMaster-Carr 5173K43) with
high pressure ttings (Swagelok Ultra-TorrTM) to minimize leaks. The chamber is
kept gas-tight with a Buna-N O-Ring. For an experiment, the ow rate of the Ar gas
is regulated with a mass ow controller (MFC) (Alicat MC Series) and the reactive
gases are controlled by high pressure MFCs (Bronkhorst EF-Flow Select). The
Alicat MFC can operate at pressures up to 10 bar, so we choose to run experiments
at slightly lower pressures to avoid damaging the unit. The 3 MFCs combine into a
single gas line and are run through a heat exchanger that is temperature controlled
20
Figure 2.3: Instrument Gas Flow System. Black lines represent gas ow with dashed
blue lines representing water ow from the thermoelectric chiller.
with an Oasis Three thermoelectric chiller (Solid State Cooling Systems 10-12684-
1C). This chiller also regulates the temperature of the chamber by owing water
through copper pipes embedded in the chamber, as shown in Figure 2.2. The heat
exchanger, shown in Figure 2.4, is necessary in order to ensure that the gas is at
the same temperature as the sample chamber when it enters the system. The main
tube of the heat exchanger is a 1 hollow copper pipe. The gas is wound through
this main pipe in a 1/8" exible copper pipe with 3-5 turns per inch. Water from
the thermoelectric chiller enters and exits the exchanger through 1/4 copper pipes,
creating a thermal bath around the gas line. Both the water and gas lines that exit
the heat exchanger are run directly to the chamber through insulated tubing.
21
Figure 2.4: Gas heat exchanger design. Before entering the chamber, the gas is
owed through a copper pipe coiled inside a hollow copper tube lled with the tem-
perature controlled glycol/water solution from the thermoelectric heat exchanger.
The chamber and the exiting gas tubing are all thoroughly insulated.
The pressure of the chamber is regulated with a digital pressure controller
(Bronkhorst P-702CV-21KA-AAD-22V). The temperature is monitored with a ther-
mistor (Omega ON-402-PP) embedded in the bottom of the chamber, which is read
out with a digital panel meter (Omega DP32PT-C24). In order to provide the op-
tion of resetting the system to an inert environment, a set of valves allow purging of
all reactive gases from the chamber and gas lines. During a purge, the 3-way valves
(Swagelok SS-42GXS4) are switched to the Ar input, and Ar is own through all
MFCs at 20 sccm each. The 2-way purge valve (Swagelok SS-41GS2) is opened so
Ar can ow directly to the chamber at ?200 sccm, bypassing the MFCs, with the
ow rate regulated by a precision needle valve (McMaster-Carr 45585K85).
Figure 2.5 depicts the optical setup of the apparatus. In our system, stress is
determined using the curvature of the sample, which is monitored by measuring the
distortion of the interference pattern images produced using an adapted Michelson-
22
Morley interferometer. A 520 nm LED is passed through a bandpass lter (Thorlabs,
520 ? 10 nm), fed into the microscope (Nikon Eclipse LV 100ND), and focused with
a modied 5x interferometric objective (Nikon CF IC Epi Plan TI Interferometry
Objective) incorporating a 50:50 beamsplitter. Half of the light is reected o the
sample with the other half directed to a at, tiltable reference mirror. The tilt of this
mirror allows the user to compensate for sample tilt, with the acceptable amount of
tilt determined by the coherence length of the illumination (?2 degrees for our light
source). In this case, our ltered LED has an approximately square spectral density
that results in a fringe amplitude that is roughly a sinc function of sample height.
Thus, for ease of analysis, we typically set the tilt to ?0.1-0.25 degrees. This setting
avoids the antinodes of the fringe amplitude and results in a monotonic change in
phase of the fringe pattern which simplies the analysis by including no points of
ambiguous phase change, characterized visually by rings or crosses.
Note that the dierence in optical path length for the sample and reference
beams should be signicantly less than the coherence length of the illumination (?20
?m in our system). To that end, a compensating window cut from the same wafer
used to form the window of the chamber is inserted in the path of the reference beam.
Further, the length of the reference arm is adjusted with stainless steel spacers to
account for the change in focal length introduced by the windows, ensuring that
the image focus plane coincides with the interference focus plane. The beams are
recombined at the beam splitter to form an interference pattern that is recorded
with the microscope camera (Nikon DS-Fi2).
Our setup allows for other light sources to illuminate the QCM either for mea-
23
Figure 2.5: Instrument Optical System. Curvature measurements are obtained by
collecting the interference patterns from the adapted Michelson-Morley interferom-
eter setup with the sample acting as one of the mirrors. A 660 nm laser is fed
into the system with its input and output power values recorded with optical power
meters, allowing for the absorption within the sample to be calculated. The laser is
blocked from entering the interference arm of the setup with a spot of Bic Wite-Out
to prevent interference eects in the reected beam.
surements of reectivity, spectrometry, ellipsometry, or for other optical excitations
of the sample. In the current apparatus, a 660 nm laser diode (Vortran Stradus
660-100) is used to illuminate the sample for optical excitation or reectivity. The
incident and reected optical powers are recorded with Si power detectors (Edmund
Optics 89-309) connected to power meters (Edmund Optics 89-307) for data collec-
tion. The laser beam is reected onto the sample with a broadband polarizing plate
beamsplitter (Edmund Optics 48-545). The section of the glass window in the in-
terference arm intersecting the laser is blocked with a white scattering coating (Bic
Wite-Out) to eliminate any interference eects of the laser. A 658 nm notch reec-
tive lter (Thorlabs NF658-26) is placed before the camera to prevent the laser from
24
saturating the interference image. The output of the laser is controlled with a pulse
width modulation (PWM) signal from an arbitrary function generator (Tektronix
AFG1062).
To perform calorimetry, heating elements and temperature measurement de-
vices are integrated into the QCM device. For measuring temperature, three ?300
? Cr/Al RTDs are lithographically printed onto the QCMs, with the patterning
shown in Figure 2.6. The central and midway RTDs have intertwined heating el-
ements that are used to add known quantities of heat to the localized points (i.e.
the location of the RTDs) on the QCM by passing current through the elements.
In addition to the RTD elements and localized heaters, the system incorporates
contact pads composed of 400 nm thick Ag with a 50 nm thick Au capping layer,
which connects to the sample lm (see Figure 2.6d). These connections allow us
to pass known amounts of current through the lm to simulate distributed power
sources such as chemical reactions. In the case of discontinuous samples, such as
nanoparticle arrays, a 50 nm Cr lm is deposited below the active sample to retain
the capability of simulating distributed power. The electrically-generated localized
heat from the central heating element and the distributed heat from the lm are
used for calibration purposes in the calorimetry model, as elaborated upon in the
Calorimetry section.
To fabricate these RTDs, each QCM sample is initially rinsed with acetone,
methanol, isopropyl alcohol, and then water to clean the initial substrate. The
substrate is then further cleaned with a 100 W plasma in 1 torr of O2 for 30 min.
Immediately following this clean, 600 nm of SiO2 is deposited using plasma enhanced
25
Figure 2.6: (a) Mask design for samples' RTD pattern. The RTDs are located at
the center of the QCM, 4 mm from the center (midway RTD), and 8 mm from
the center (outer RTD). Each RTD is measured with 4-point contacts for improved
accuracy. The center and midway RTDs consist of two RTDs intertwined, with one
acting as a heater and the other as a sensor. The outlined box near the center of
the QCM is used for consistent alignment within the microscope. (b) Outer and (c)
intertwined center RTD images. (d) Sample with complete fabrication of RTDs and
Ag/Au contact tabs.
chemical vapor deposition (PECVD) (Oxford Instruments). The RTDs are then
patterned on the SiO2 using standard contact photolithography, and 50 nm Cr and
120 nm Al are deposited using electron beam evaporation (Angstrom). The excess
metal is lifted o overnight and then sonicated the following morning for 2 min to
nish the lift-o process. Another 600 nm of SiO2 is PECVD deposited on the center
of the sample using a shadow mask that excludes the RTD contact tabs. Next, 400
nm Ag and 50 nm Au are e-beam deposited using a shadow mask for the lm contact
tabs. The active lm is then e-beam deposited on the center of the sample using a
separate shadow mask. A schematic of each step of this process is shown in Figure
2.7.
26
Figure 2.7: Outline of each step of the sample fabrication process. Top row: cross-
sectional view. Bottom row: aerial view
To quantify the sensitivity of the lithographically printed RTDs to the temper-
ature of the entire system, as controlled by the thermoelectric chiller, we recorded
the measured resistances of the RTDs while sweeping the temperature of the cham-
ber from 15 to 35 ?C. The actual temperature of the chamber was monitored with
the thermistor embedded in the bottom of the chamber. The resistances from all
three RTDs on a QCM without an active sample were recorded throughout the
sweep. Normalizing to the 25 ?C resistance of each RTD, we nd that the RTDs
have an average temperature coecient of 0.0028?0.0001 (?/?C)/?. Results of this
temperature sweep are found in Figure 2.8.
The RTD sensing system driver is an Analog Devices AD7124-8 integrated
circuit, implemented here via an AD7124-8 evaluation board. This driver was cho-
sen because it met our requirements of customizability, sensitivity, noise, speed,
and integrability into our custom measurement software. The AD7124-8 uses a
multiplexed set of input/outputs, which can be internally connected to a dierential
amplier and analog to digital converter (ADC) or output to peripherals such as the
precision variable current source in use here. See Figure 2.9 for full wiring diagram.
Each of the three sensing RTDs are connected in a four-point probe conguration.
27
Figure 2.8: RTD resistances as a function of measured chamber temperature. The
average temperature coecient is 0.0028?0.0001 (?/?C)/?. The RTD resistances
shown correspond to the center (red), 4 mm from the center (blue), and outer (green)
RTDs.
The voltage drop across each RTD is compared to that across a reference resistor
in a ratiometric scheme, as seen in Figure 2.10. We use two 470 ? bias resistors
to ensure that the inputs to the ADC meet the absolute voltage requirements (0.1
V from the rails of 0 and 3.3 V). The AD7124-8 incorporates several digital lter
options, which allow the user to dene the tradeo between speed and noise. We
use a sinc4 lter with rst zero at 60 Hz (primary source of noise), which results
in a sampling time of 62 ms per channel. Our applications use 4 channels: 3 RTD
ratiometric measurements and a voltage measurement of the reference resistor (com-
pared to the on-chip 2.5 V precision voltage source). Communication with the chip
is accomplished via the SPI interface on a Teensy 3.2 development board, which
also transmits data on-demand with a USB COM port. The data exchange between
28
the controlling computer and the 4 inputs takes ?270 ms for a single measurement.
Our typical sampling period in the custom-built Windows user application is 350
ms; this leaves sucient time for the ?270 ms required for acquisition and 80 ms
for other tasks such as saving data. The currents through the center intertwined
heating element and the center disk are driven by Keithley 2450 sourcemeters in
order to model known heating eects in the sample.
Figure 2.9: RTD and current sensing schematic. The current (path shown in red)
ows through all of the RTDs and reference resistors. The voltage drops (blue
paths) across each of these resistors is probed and compared to the drop across the
reference resistor. Note that the voltage probes do draw non-zero current. However,
the current is typically < 1 nA resulting < 1 ppm oset for a 1 mA excitation
current.
Lastly, in order to properly perform the calorimetry measurements, we need
to estimate the thermal conductivity of the gas and how it changes with time.
Eq. 2.8 in our calorimetry section assumes that thermal dissipation to the gas
environment can be approximated by a quadratic equation with respect to partial gas
pressure. Below we demonstrate the accuracy of this approximation. The thermal
conductivity is related to the mole fraction of argon as shown in Figure 2.11a. This
29
Figure 2.10: Schematic for the RTD current source and voltage measurement system.
The RTD resistances are ?300 ?, the reference resistance is 500 ?, and the Rbias
resistances are 470 ?
plot suggests that the relationship is nearly linear, but has small deviations from
linearity. Another metric would be preferable to better capture the eects of heat
loss through gaseous conduction. Fortunately, the resistance of a QCM is linearly
related to the change in thermal conductivity. For a QCM in a viscous medium, the
resistance can be expressed as [54]:
( )?
?sLu 2?s?L?L
R = (2.1)
N? c?66?q
where N is the overtone number, ?s is the resonant frequency, Lu is the in-
ductance of the QCM in vacuum, ?L is the density of the uid, ?L is the viscosity of
the uid, ?q is the density of quartz, and c?66 is the piezoelectrically stiened quartz
elastic constant. N , Lu, ?q, and c?66 are all constant upon a change of atmosphere,
30
thus they do not factor into ?R. The resonant frequency is constant to four signif-
icant digits during a standard hydrogenation experiment, thus can also be ignored.
Larger frequency changes result in nonlinear behavior and are outside the scope of
this correction scheme. Thus this expression can be approximately written as:
? ??R ?? (2.2)
Where the L subscripts have been dropped for clarity. Figure 2.11b shows
that the square root of the product of density and viscosity for the Ar-H2 system,
and thus the QCM resistance, scales linearly with thermal conductivity for Ar mole
fractions greater than ?0.1 [55]. When plotted against time, with the H2 partial
pressure modeled as a decaying exponential, Figure 2.11c shows that assuming a
linear relation between mole fraction and thermal conductivity will only slightly
underestimate the thermal conductivity of the gas atmosphere. Given the small dis-
crepancy, we assume that partial pressure is proportional to thermal conductivity
which allows us to perform calorimetry while a new gas is introduced. This as-
sumption would cause for a maximum error of 1.5% to our calorimetry model (given
by the error in the conductance that this partial pressure error would create), well
below our noise level.
Fortunately, the QCM resistance is not very sensitive to changes in mass or
stress, so it serves as a good proxy for changes in thermal conductivity of the envi-
ronment [56]. Accordingly, we t the QCM resistance in the region surrounding a
change in gas content to an exponential. Using the time constant from this t and
31
Figure 2.11: (a) Plot of thermal conductivity of an Ar-H2 mixture at 9.3 bar as a
function of the mole fraction of argon. (b) Plot of thermal conductivity of an Ar-H2
mixture at 9.3 bar as a function of the product of the square roots of the density
and viscosity. This parameter serves as a proxy for the change in resistance of the
QCM. (c) Plot of both the mole fraction of argon and change in resistance proxy as
a function of time when the H2 MFC is opened at time 0 and owing at 20 sccm.
The initial atmosphere is assumed to be pure argon at 9.3 bar in a chamber with a
volume of 13 mL.
32
the initial partial pressures (as determined by the ratios of the MFC ow rates), we
generate approximate partial pressures.
2.4 Demonstration of operation and stability
To demonstrate the stability of the apparatus, a control experiment was per-
formed on a blank QCM sample with RTDs. The sample was pressurized in Ar
up to 9.3 bar, switched to H2 at 9.3 bar for 4 hours, switched back to Ar for 4
hours, and nally returned to H2. During the run, the 660 nm laser was pulsed and
incident on the sample while in H2 and Ar to test how well the calorimetry model
t the heating induced by laser absorption. The data of this run are reported in
Figure 2.12 and show the stability of the system. The partial pressure of H2 gas is
calculated using the method described in the previous section. The leak rate of the
system at 9.3 bar is measured to be 2 sccm.
2.5 Example of stress and mass change measurements
The stress in the sample is characterized by the curvature of the substrate.
The curvature is determined by converting the optical phase change measured by
the interferometric images into a sample height change, with the process depicted in
Figure 2.13. The curvature is then directly converted into a corresponding frequency
change that is linearly independent of the frequency change due to mass under the
thin-lm approximation (i.e. the lm mass is much smaller than the QCM mass).
First, the image is bandpass ltered and normalized so that spatial variations in
33
Figure 2.12: Apparatus stability for a control experiment including changes in gas
composition and laser excitation. Vertical dashed lines indicate a change in H2 ow
rate set point. (a) Calculated H2 partial pressure. (b) Total ow rate over the
course of a 10 hour experiment. (c) Pressure of the chamber for a xed pressure
set point. (d) Normalized dierential RTD resistances through a run. The spikes in
the resistances correspond to the temperature increases from the absorption of laser
light. (e) Temperature of the environmental chamber, as measured by the interior
thermistor.
image brightness and fringe sharpness are reduced. Greater delity to the original
sample surface can be achieved by removing some artifacts and by utilizing the very
narrow bandwidth of the signal. The rst artifact is simply brightness variation
in the image. This is removed by high-pass ltering the image with the cuto
frequency one-fth of the fundamental frequency. The second artifact is given by the
nite coherence length of the ltered LED light. This results in a fringe amplitude
34
variation with sinc dependence on the sample surface height. To normalize this
variation, the image is then rectied and low-pass ltered. The high-pass ltered
image is divided by this amplitude envelope. Finally, we low-pass lter this image
to reduce noise. This ltering process is graphically shown in Figure 2.14.
Figure 2.13: Image processing ow to extract curvature measurements from interfer-
ence image. The image is ltered and normalized (Figure 2.14). The instantaneous
phase is then extracted from the Hilbert transform. The phase is converted to height
and attened to give the sample topography. The curvature is extracted from the
second derivative of a 2D polynomial t to the sample height.
After the image is ltered, it is Hilbert transformed and the phase angle of
the now-complex signal is extracted. The phase angle, which varies from only 0 to
2? in a sawtooth pattern, is then unwrapped by stitching together steps in a phase
of 2? to produce a smooth phase surface. This phase surface can then be converted
to sample height by dividing by the phase change per change in height, 2?/(?/2),
where ? is the central illumination wavelength (? = 520 ? 10 nm for our LED).
Finally, a 2D polynomial is t to this phase surface and the curvature can be directly
extracted as the second spatial partial derivatives of the surface.
35
Figure 2.14: Image processing ow. The process removes variation in brightness
and interference fringe amplitude. Noise is then removed with a low-pass lter.
The hydrogen mass loading of the active sample is an important factor when
characterizing a reaction on a lm. To determine this loading, we begin with a
precise measurement of the mass of the active material deposited on the sample.
To accomplish this goal, we fabricate an additional, sacricial sample that consists
36
of a lithographically dened 1 cm x 1 cm square of the material we are character-
izing on a polished Si wafer during each sample lm deposition. We dissolve this
lm using 4 mL of either aqua regia or boiling hydrochloric acid and dilute to 100
mL as determined by lm composition. We then use inductively coupled plasma
atomic emission spectroscopy (ICP-OES) to determine the metal mass per area of
the sample deposition. We assume that the areal mass density of the metal on the
sample is the same as it is on the calibration piece (note that areal mass density is
expected to be constant even if other properties such as grain size change sample
to sample), as the deposition occurs on both pieces concurrently. Knowing the area
of the active lm on the QCM allows us to determine the exact mass of the active
material. This method is robust to multi-layer samples, where the individual layers
of each sample cannot be easily measured. If the sample is only a single layer, an
alternative method of nding the sample thickness is to use a step height measure-
ment using Atomic Force Microscopy (AFM) over one of the lithographically dened
edges of the Si wafer. Once the mass is known, we determine loading in the lm by
the QCM frequency change less the portion due to stress and environmental eects
(frequency compensation discussed below).
To calculate the contribution of stress to the frequency shift measured by the
QCM, we use a combined technique of numerical and analytical calculations. First,
we numerically calculate in COMSOL the amount of stress-induced in the QCM due
to curvature changes. In the simulation, we bound the system to be immobile at the
edges to match our conditions of the spring ring holding down the QCM. Using these
numerical stress values, we can calculate the relationship between frequency and
37
stress analytically by calculating the propagation speed of the shear wave through
the crystal and then integrating over two times the length of the crystal to get the
frequency. The nal output of our derivation gives
?f? = ??? (2.3)
where ? is the calculated curvature to frequency conversion factor equal to -777
Hz m and ?? is the change in curvature. See Appendix A for a full derivation of the
stress to frequency equations. For a typical thin metal lm (0 - 1 ?m) on a QCM,
the curvature measurement has an uncertainty of 0.001 m?1 which corresponds to a
stress uncertainty of 0.006 MPa per micron of lm thickness from this calculation.
An example of the contribution of stress to the total frequency change can be seen in
Figure 2.15. Here we use a 335 nm palladium lm, which has a well-known hydrogen
loading fraction to conrm the equations above. The gas pressure and composition of
the chamber also contribute to the frequency change and these eects are calibrated
out using the measured frequency of the QCM on the secondary control stage caused
solely by environmental eects. The frequency from the mass change of the sample
can be calculated from the total frequency change using the following equation:
?fm = ?ftot ??fgas ??f?
(2.4)
= ?ftot ??fsecondary ? ???
where ?fm is the frequency change due to a mass change, ?ftot is the total frequency
change, ?fgas is the frequency change due to environmental eects, ?f? is the
38
Figure 2.15: a) Plot of QCM frequency change during hydrogen loading of a 335 nm
thick Pd lm. The blue curve is the measured frequency of the active QCM, the
yellow curve is the measured frequency of the secondary blank QCM, and the orange
curve is the frequency contribution from the stress in the active QCM, as calculated
from the curvature. By subtracting the stress and gas composition contributions
from the total measured frequency, we obtain the frequency change due to the
added mass of the hydrogen within the Pd lattice (purple curve). b) Mass loading
fraction x upon introduction of H2.
frequency change due to stress, and?fsecondary is the measured frequency of the
blank QCM on the secondary control stage. The mass-induced frequency change
can then be converted to the hydrogen loading with the equation:
39
?fmAM
?x = (2.5)
?M tfAHCf
where ?x is the change in hydrogen atoms in the lattice per metal atom, Ai is the
atomic mass of species i (eitherM , the metal host orH, hydrogen), ?M is the density
of the unloaded metalM , tf is the lm thickness, and Cf is the Sauerbrey coecient
from the literature relating a frequency change to a corresponding mass change. The
uncertainty of the mass change measurement in our system is 13 ng/cm2, which is
dominated by the uncertainty in curvature. It should be noted that the extra mass
and varied impedance associated with RTDs, insulating layers, and the lm under
test result in a correction to the Sauerbrey coecient of < 2% (determined by full
transmission line simulation as per C. Steinem and A. Jansho [41]). However,
this correction factor lies within the uncertainty for a typical gas-phase loading
experiment and can generally be ignored. The measured loading fraction of 0.7
agrees with the well-known value for palladium at this pressure (9.3 bar) [57, 58].
2.6 Calorimetry
We use a one-state nonlinear lumped element model to describe our system
(i.e. the state is described by a single variable) with thermal potentials measured
at discrete positions by our RTDs [59]. This model takes the dierence of the
normalized resistances of the outer and center RTDs as the independent variable (i.e.
state variable of the single oating node), ? , which is a function of the dierential
temperature between the center and outside of the QCM. Thus, all that is needed
40
is a measure of relative temperature during a system calibration, which may then
be used to infer the input power (from any source) during an experiment. The
thermal power input sources are the powers from the RTDs, PRTD, the laser or
laser-induced reactions, PLaser, the heat from running a current through the sample
thin-lm or thin-lm reactions, PDisk, and the heat from running current through
the central heating element, PHeater. Figure 2.16 shows a schematic of the lumped
element model with the thermal resistor and capacitor driven by the inputs and
tied to a thermal ground. While only two RTDs are utilized in the one-state model,
we have included a third RTD on the QCM. This third (midway) RTD presents
an opportunity for a two-state model (using ? and a second state variable), which
could oer higher precision as well as additional information about heat localization,
but would require more development and an extended calibration process. Thus, a
two-state model is left as a future renement.
Figure 2.16: The four source, one-state lumped-element calorimetric model diagram.
The four contributing powers measured into the sample are the input powers from
the RTDs, PRTD, the laser or laser induced reactions, PLaser, the heat from running a
current through the sample thin lm or thin lm reactions, PDisk, and the heat from
running current through the central heating element, PHeater. The state variable,
? , is the relative thermal gradient of the system dened by the dierence between
normalized (to equilibrium resistance) center and outer RTD resistances. k and C
are the thermal conductance and thermal capacitance of the system respectively.
41
This one-state system can be modeled with a rst order dierential equation
described by a single independent state variable ? and is given by:
d?
PIn(t) = k? + C (2.6)
dt
with
RCenter(t) ROuter(t)
? ? ? (2.7)
RCenter,0 ROuter,0
Here PIn is the total eective input power (i.e. the sum of the individual input pow-
ers weighted by power distribution scaling factors, see below), Ri(t) is the measured
resistance of RTD i at time t, and Ri,0 is the measured near-equilibrium resistance
of RTD i (i.e. the resistance when the only input power to the system is the small
RTD sensing current). The non-linear thermal conductance, k, and capacitance, C,
depend on the state variable and the partial gas pressures. These values are given
by:
?N
k = k0 + k?? + k?n?n + k? ,2?2 (2.8)n n
n=0
?N
C = C0 + C?n?n + C? ,2?2 (2.9)n n
n=0
where ?n is the partial pressure of gas species n. Note that when the system is
operated at a constant total pressure and with a mix of two gases, as is often the case,
the partial pressure terms can be collapsed to a single term. With all the constants
42
above known, the total input power may be inferred. However, these constants must
rst be determined during the calibration process (i.e. system identication), when
known powers are input into the system and the constants are t using Eq. 2.6.
The eective total input power during the system identication process is given by:
( ? )N
PIn(t) = PRTDs(t) + PDisk(t) ADisk + ADisk,?n?n + PLaser(t)ALaser (2.10)
n=0
where Pi and Ai are respectively the input power and the power distribution scaling
factor of power type i, with Ai,?n being the proportion due to the change in partial
pressure n. These power distribution scaling factors are a result of the dierences
in the spatial distribution of the heat sources and are also t during the calibration
process. The terms that include partial pressures represent the dierences in eective
power due to the spatial distributions. As noted above in the case of a constant
total pressure composed of two gases, the sum can collapse to a single term.
A complete calorimetry measurement of a sample is composed of three sec-
tions: experiment where we measure the heat of a reaction, calibration where we
t a system model, and prediction where we conrm the accuracy of the model.
After each experimental section, a system identication cycle is run where various
known power inputs are applied at higher powers and frequencies than expected in
the experimental run. The results of these excitations are used to t the parameters
dened above (A's, k's, and C's). While this system identication step could theo-
retically be conducted before the experiment, it requires changing the reactive gas
43
partial pressure, which in the case of irreversible gas-based reactions would cause
the sample to be fully reacted before the experimental section. While the equations
above present a comprehensive system, generally the calibration should be tailored
to the experiment, which may allow for reduced dimensionality of the t. The tting
process uses Matlab System Identication Toolbox together with the freely avail-
able Calorimetry System Identication (CSID). This Matlab toolbox takes a set of
user-dened dierential equations with unidentied constants (A's, k's, and C's)
and uses a calibration dataset (known inputs with measured outputs) to t these
constants. The CSID adds additional functionality and gives examples specic to
the identication of calorimetric systems.
In order to fully probe the dynamics of the apparatus, it is important to excite
with a wider bandwidth and with greater amplitude than the heat pulses expected
during the course of the experiment, as well as with all possible power sources. In
our experiments, we use a set of square pulses of increasing power at a constant
total pressure of 9.3 bar and 3 dierent partial pressures of the reactive gas (see
Figure 2.17). Note that the calibration routine only calibrates for a limited range
of the system's capabilities and thus limits the scope of experiments. However, by
reducing the scope of the experiment and tailoring the calibration to the actual use
of the instrument, we can reduce the number of t parameters in our model. Table
2.2 shows the reduced set parameters actually employed in a typical experiment.
The calibration portion is followed by a prediction section. During this phase, the
system is excited more gently in an attempt to simulate run data. The data from
this section can then be used to derive error bars and help assess overtting during
44
the calibration phase.
Figure 2.17: Modeled and measured calorimetry data for a thin Cr lm during the
prediction portion of a run in which the accuracy of the calibration is assessed. (a)
Calculated H2 partial pressure of the system. (b) The system is excited with each
of the power sources. The red line is the state variable measured by the apparatus
and the blue line is the output of the model of this data. (c) Modeled and measured
excess powers due to pulses of the disk heating power during the nal portion of
the prediction section. The system is thermally excited by owing current through
the lm. This portion corresponds to the integration region used to calculate the
accuracies of the energy pulses. Here the integration is represented by the shaded
areas with dark grey for the short 65 mJ pulses and light grey for the larger 325 mJ
pulses.
With the t complete, we rst check the parameter values and the accuracy
of the model. Table 2.2 shows some typical tted calorimeter values for a standard
45
experiment using one total pressure and two gases. The parameters are scrutinized
to verify that they are physical (e.g. no negative temperature coecients for our
quartz-based system). The t is tested by using a prediction run with sinusoidal
excitation at two dierent frequencies.
Parameter Valuei
k0 2.7W?/?
k? 4.2W?
2/?2
C0 0.017 J?/?
CpH2 -0.028 J?/? Bar
CpH2,2 0.012 J?/? Bar
2
kpH2 -0.054 J?/? Bar
kpH2,2 -0.00056 J?/? Bar
2
ADisk 0.11
ALaser 0.72
ADisk,pH2 -0.0052 1/Bar
Stress sensitivity 0.006 MPA per micron
of lm thickness
iNote that the state variable, ? , is unitless but
the unit of ? , ?/?, are shown here to make
dimensional analysis clear.
Table 2.2: Example t parameters for a calorimetry calibration run
The quality of the t is then tested with another set of more gentle excita-
tions of the various power sources in a prediction run that immediately follows the
calibration cycle. Data recorded from a typical prediction run is compared to the
output of the calibrated model in Figure 2.17. This run shows a normalized (by the
mean value) root mean squared (RMS) error between measured and modeled data
of 8%. We also compare the energy calculated from the disk power in a single pulse
to that of the energy modeled. This comparison gives the accuracy of the energy
inferred from a potential chemical reaction on a similar timescale and power to that
46
of the pulse, assuming that PDisk is an analog to the power of the reaction. Here we
dene our metric as the dierence between the integrated excess calculated power
(dierence between quiescent power and pulse power) and integrated excess eective
input power (PIn) divided by the integrated modeled power over the pulse region.
In the run shown, we nd this average error to be 4% for the short 65 mJ pulses
and 2% for the larger 325 mJ pulses in 100% H2. The improved accuracy compared
to the error for the total run arises from sensor drift. Finally, to put bounds on the
instantaneous power accuracy and resolution of our system, we consider the sensor
drift over the course of a 10-hour experiment. The typical sensor drift on this time
scale is 75 ppm which corresponds to 150 ?W of apparent power drift.
There are several sources of error in our current calorimetry system. The
primary source of error in our measurements is instrument drift. The typical drift
of our state variable ? is ?75 ppm over 12 hours, which sets our accuracy at 150 ?W.
While the sources of the drift are unknown, several possibilities provide avenues for
better ts. The range of potential causes is narrowed by the fact that our drift is both
non-monotonic and does not appear to be improved by burn-in or aging. The most
likely candidate is temperature non-uniformity, which would be evidenced by drift
in the state value (dierential relative resistance). This is made more likely by the
dierences in equilibration time between the outer RTD and the center and midway
RTDs patterned above the Au electrode. Note, however, that while the drift creates
apparent inferred power osets (representing a constant additional power sink or
source), the timescale is on the order of hours which still permits the determination
of shorter time scale events by subtracting observed background power caused by
47
this drift. To potentially resolve this the chamber and gas lines could have improved
insulation and additional temperature sensors could be added to the chamber. This
would allow us to mitigate and account for non-uniformities which may be the source
of our sensor drift.
Another source of error is the use of a single state lumped element model,
which only approximates the distributed heat source produced by the disk power
or a chemical reaction of a lm on the QCM. By modeling this heat source in the
same way as the other point sources, we neglect the larger thermal capacitance
and dierences in temperature across the lm which aect the determination of the
nonlinear terms of the above equations. However, tting suggests the former, while
non-zero, is a minor eect (e.g. for the run shown in Figure 2.17 the dierence in
normalized RMS error between when the system is excited by the heating element
vs. when the system is excited by the disk power is 4% during the high-frequency
oscillations where the t is least accurate). As to the latter, the actual nonlinear
components are negligible in our experiments (e.g. a typical k0 is 2.7 W?/? with the
typical maximum product of k?? is 0.00005 W?/?, where the units originate from
the unitless state value ?). Lastly, our data suggest that stress results in an apparent
power source due to a combination of the piezoresistance and linear expansion of the
RTD, which can cause a stress induced change in the resistance. This is estimated
to result in a ?50 ppm change in resistance for a typical change in curvature but
measuring this eect is dicult and it is not currently incorporated in the model.
These errors suggest potential improvements. First, a two-state model may
improve the modeling accuracy by incorporating the additional data taken by the
48
midway RTD. This would also allow for a better simulation of the distributed power
arising from lm reactions or the disk heater. This approach has not yet been
pursued as it would require a more in-depth calibration. Second, we can begin to
incorporate the changes in input power due to stress into our system model.
2.7 Optical property measurement
By taking the system fully dened above and switching out the chamber lid
with a specically designed ellipsometry lid, we can incorporate optical property
measurement into our system. We measure the optical properties via variable angle
spectroscopic ellipsometry (Woollam M-2000). Spectroscopic ellipsometry can be
used to calculate optical properties by shining linearly polarized light on the sam-
ple of interest and measuring the change in magnitude (?) and phase (?) of the
resulting elliptically polarized beam. By modeling these values at multiple angles
across a spectrum of wavelengths, the optical properties of the measured material
can be determined. This method can also be used for measurements of thin-lm
and roughness thicknesses. In our system, in order to accommodate in situ measure-
ments, we designed a chamber lid, shown in Figure 2.18, with windows at 4 incident
angles (48?, 55?, 70?, 75?) so that light would be normally incident on the chamber
windows. It is important that the measurement beam is normal to the windows
to avoid spurious Fresnel eects in the measurement. By allowing for 4 separate
incident angles, a high-quality, multi-angle t can be achieved. One requirement
in this system is to incorporate the calorimetry scheme with the optical property
49
measurement, the sample measured must be optically thick so that the RTDs do
not interfere with the optical measurement.
Figure 2.18: a) Cut-through drawing of ellipsometry chamber lid. This cutout shows
two of the possible incident angles. The other two angles are in the plane of the
screen. This lid still allows for optical access through the top of the lid. b) Picture
of the manufactured chamber lid with the 4 incident angles (output ports are on the
other side of the lid).
This system setup allows for dynamic optical property measurement under
changing environmental conditions. For the most precise dynamic optical measure-
ment, the optical properties of the sample under consideration need to be measured
prior to being placed in the chamber. In our system, the spectral range of the ellip-
someter is 193 - 1690 nm and we take this ex situ measurement with angles varied
50
from 50? to 75? in steps of 5?. Once these initial properties are obtained, the sample
is then mounted on the QCM stage in the environmental chamber and the chamber
is closed and brought to the desired experimental pressure in Ar. The optical prop-
erties are then recorded at the 4 separate inlet angles of the environmental chamber
(48?, 55?, 70?, 75?). These results are compared to the initial ex situ measurement to
determine the change in the phase dierence between TE and TM polarizations as
light passes through the chamber windows (ellipsometric retardation eects). The
origin of this retardation is the birefringence in the glass produced by anisotropic
window stress when mounting the chamber lid. A dynamic ellipsometric measure-
ment is then taken through the 75? window (?10 data points per minute) and the
chamber ow rate is set to 20 sccm H2. This method could also be used with other
reactive gases other than H2, but in this thesis, we will only focus on the eects of H2
on metallic systems. The dynamic measurement is stopped when both the optical
properties and the measured QCM frequency have stabilized, meaning that the full
reaction has completed. While still under H2 ow, ellipsometric measurements are
taken of the metal hydride using all 4 inlet angles.
2.7.1 Optical tting
All ellipsometric data is t using the Woollam CompleteEASE tting software.
The tted optical parameters are created using a Kramers-Kronig consistent B-spline
to minimize the dierence between the modeled ellipsometric data and the measured
data. In order to decrease the number of t parameters, the surface roughness of each
51
measured sample is recorded before and after a loading run using AFM. The RMS
surface roughness from these experiments is input into the optical model for that
metal. For samples that require a Pd capping layer to catalyze the hydrogenation
reaction, the capping layer thickness in the t is constrained between 2 - 4 nm, with
the thickness value being a t parameter. The optical data used for the Pd capping
layer is from our measurements, which we will go into further detail in Chapter 3.
To determine the ellipsometric retardation phase eects (given by the added
TE/TM phase dierence) of the environmental chamber windows, the pure metal
model found above is xed and the dierence between the measured phase through
each window and the phase of the pure metal model is t using the following equa-
tion:
( )
?(f) = f C1 + C f
2
2 (2.11)
where ? is the frequency-dependent retardation eects input to the model,
f is the optical frequency of the spectroscopic beam, and C1 and C2 are the t
constants. Note that each set of windows has a dierent retardation due to dierent
stresses present in the windows, thus each set has to be calibrated separately. These
eects do not change during the course of an experiment, so they are held constant
at these values for the duration of the run. These values do vary from run to run
due to slight dierences in clamping stresses of the chamber, thus they must be
calibrated individually for each run. Because these stresses have an amplied eect
at shorter wavelengths, we generally lower bound the wavelength of our ts to 250
52
nm to eliminate extra errors introduced by these window stresses.
The nal hydride optical properties are also t using a Kramers-Kronig con-
sistent B-spline to the measured data of the 4 angles recorded after the dynamic
run (including the set delta osets found above). In some experiments, there can
be unusually high stresses in one set of windows causing a distortion in the data.
In these cases, the data from that angle is excluded from the model. The surface
roughness for the hydride is input into the model from the AFM measurement, and
the Pd thickness found in the pure metal t is held constant. We note that there
is lattice expansion from hydrogenation that would cause the PdHx thickness to
be slightly thicker ( ?12% for unbounded Pd) than the original Pd capping layer,
however this is within the error bars for the thickness determination [60]. The ex-
perimental PdHx data for the Pd sample run is used for the capping layer optical
properties.
The dynamic t uses two Bruggeman eective medium approximations (EMA):
one for the metal under investigation and one for the Pd capping layer. The two
materials input into each EMA are the pure metal model and the hydride model
found above. The three t parameters in the model are the EMA % of the metal
under investigation, the EMA % of the Pd capping layer, and the surface roughness,
which is bounded by the dened roughnesses of the metal and the hydride measured
with AFM.
One extra note with this added optical setup, with the increased height of
the ellipsometry lid, the stress compensation microscope cannot be integrated with
the optical property measurement due to the length of the interference arm. This
53
requires for two samples to be deposited at the same time, close to each other in the
deposition chamber to make as identical lms as possible. One sample is then mea-
sured in the optical setup, with the other measured in the regular apparatus dened
above. The QCM frequency changes in each of the chambers can be compared to
ensure that they are having identical responses to hydrogenation. Then, the loading
value obtained from the main apparatus can be used to scale the frequency data
from the optical sample QCM so that it can be converted to loading data. The
dynamic optical data can then be temporally aligned with the loading data from
the QCM to correlate the optical changes of the material with its loading at that
point.
2.8 Conclusions
In this manuscript, we have presented a novel apparatus with the combined
capability to measure stress, mass, and heat while pressurized under a reactive gas
and retaining optical access to the sample. The incorporation of stress compensation
allows for the accurate determination of mass loading. Prior studies on QCMs have
neglected this important factor, likely resulting in anomalous mass determinations.
We have demonstrated the ability of our chamber to hold pressure at 9.31 ? 0.02
bar of H2 at 24.85 ? 0.05 ?C with a leak rate of 2 sccm. On a 335 nm thin-lm palla-
dium sample, we demonstrated a measurement sensitivity of 0.001 m?1 of curvature,
corresponding to a stress sensitivity of 0.006 MPa per micron of lm thickness, and
a mass measurement sensitivity of 13 ng/cm2. Using a one-state lumped-parameter
54
heat transfer model, the heat creation of a reaction can be measured with 150 ?W
accuracy. The on-chip calorimetry scheme allows for a very exible system where
small heats can be detected in thin lms, which might be overwhelmed by base noise
if the temperature measuring elements were located farther from the sample.
This instrument enables the study of phenomena including the ellipsometric
determination of optical properties at varied metal hydride compositions, measure-
ment of interfacial energies between metals and various substrates, spectroscopic
measurement of nanoparticle resonance frequency for optical analysis of chemical
processes or optical calorimetry, and the direct quantication of heat and stress from
coupled plasmonic excitations. Many of these phenomena have not been thoroughly
explored because of the inability to collect all of the necessary data simultaneously
throughout a reaction. The combination of these measurements will lead to new
insights into nanoscale reactions. Further, the integration of optical access with the
other capabilities already described oers interesting possibilities for expanding the
scope of this apparatus.
Throughout the rest of this thesis, we will use this apparatus to measure the
dynamic optical and structural properties of various pure metal and metal alloy
hydride systems. The measurement of these switchable properties will allow us to
design, propose, and fabricate various nanophotonic structures for dierent applica-
tions ranging from optical hydrogen sensors to switchable high absorption devices.
55
Chapter 3: Dynamic Optical Properties of Pure Metal Hydrides
Metal hydrides often display dramatic changes in optical properties upon hy-
drogenation. These shifts make them prime candidates for many tunable optical
devices, such as optical hydrogen sensors and switchable mirrors. While some of
these metals, such as palladium, have been well studied, many other promising
materials have only been characterized over a limited optical range and lack di-
rect in situ measurements of hydrogen loading, limiting their potential applications.
Further, there have been no systematic studies that allow for a clear comparison
between these metals. In this chapter, we present such a systematic study of the
dynamically tunable optical properties of Pd, Mg, Zr, Ti, and V throughout hy-
drogenation with a wavelength range of 250 - 1690 nm. These measurements were
performed utilizing the apparatus dened in Chapter 2, which allows for us to cor-
relate these dynamic optical changes with the mass loading of the metal hydride. In
addition, we demonstrate a further tunability of the optical properties of Ti and its
hydride by altering annealing conditions, and we investigate the optical and gravi-
metric hysteresis that occurs during hydrogenation cycling of palladium. Finally,
we demonstrate several nanoscale optical and plasmonic structures based on these
dynamic properties. We show structures that, upon hydrogenation, demonstrate
56
ve orders of magnitude change in reectivity, resonance shifts of >200 nm, and
relative transmission switching of >3000%, suggesting a wide range of applications.
3.1 Introduction to metal hydride optical properties
Materials with tunable optical properties are critical to the development of
novel active nanophotonic devices ranging from plasmonic light absorbers and biosen-
sors to switchable mirrors [14, 61, 62, 63, 64, 65, 66, 67]. The ability to change the
resonances of a structure in situ allows for increased functionality and enables new
device architectures. One particular class of materials well-suited for tunable appli-
cations is metal hydrides. A number of metals have been shown to strongly absorb
hydrogen, resulting in hydrogen to metal atom ratios approaching or even exceed-
ing 1:1 [12, 57, 58, 68, 69, 70, 71]. These metals typically undergo crystalline phase
transitions, altering their crystal and electronic structures. The large changes to the
crystal structure, the additional electrons, and the additional resonances associated
with the binding energy of the hydrogen to the lattice can create dramatic shifts in
the metal's optical properties.
These metal hydrides are of great interest for switchable photonic devices, par-
ticularly for applications involving optical hydrogen sensors and switchable mirrors.
Palladium and palladium alloys have been widely used for such sensors, structured
as both thin lms and nanoparticles [72, 73, 74, 75, 76, 77, 78, 79]. Yttrium and
lanthanum have been investigated for their use as switchable mirrors due to their
metal to dielectric transition upon hydrogenation[14, 66]. Magnesium has seen re-
57
cent interest for use in reversible color lters due to its optically dramatic shift from
metal to dielectric upon hydrogenation [80]. Hafnium has been introduced as an op-
tical hydrogen sensor that can span six orders of magnitude [81]. Niobium nanorods
were recently investigated as a new material for high-temperature plasmonics with
switchable properties upon hydrogenation [20].
While work has been done on the optical properties of metals and their hy-
drides, these previous studies have a variety of limitations (see Table 3.1 for a
summary of previous data), including: narrower wavelength ranges (250 - 1690
nm in this work), lack of any dynamic or intermediate hydride data, or temperature
ranges that prevent comparison across dierent studies. On top of these limitations,
varied fabrication conditions and procedures alter the optical properties of a metal
in a given experiment, furthering the diculty of comparison. No systematic work
has compared a wide range of materials (here Pd, Mg, Zr, Ti, and V) prepared and
tested under identical broadband illumination and time-dependent hydrogenation
conditions. Further, none of the referenced studies include direct in situ loading
measurements (Azofeifa et al. indirectly infer loading in situ via resistivity and
transmission spectra) [82].
By addressing all of the above issues in this work, we present a complete,
uniform set of measurements that can be used to compare the optical properties
of ve dierent metal thin-lms (Pd, Mg, Zr, Ti, and V) before, during, and after
hydrogenation. We pair this data with simultaneously recorded measurements of the
hydrogen loading using a custom environmental chamber incorporating a QCM [95].
We also investigate the eects of annealing on the optical properties of Ti and TiHx
58
Material Wvl range (nm) Temp (K) Dynamic? Reference
MgHx ? 80-200 293 No van Setten 2007 [83]
MgHx 190-1240 293 No
i Isidorsson 2003 [84]
ZrHx 240-1040 293 No
i Azofeifa 2017 [85]
VHx 350-950 20, 140 No Azofeifa 2013 [82]
PdHx 380-1650 293 Yes Yamada 2009 [86]
PdHx 250-1200 293 No Vargas 2014 [87]
iData is oered for multiple partial pressures
Table 3.1: Previously published work on the measured optical properties of metal
hydrides. Here we show only references oering data for the pure metals and ones
which derive optical constants. There are several manuscripts on Pd alloy hydrides
[74, 76, 88, 89] and Mg alloy hydrides [89, 90, 91]. There are also some references
which oer transmission data only which are not shown above such as for MgH2 [92]
and for TiHx [93]. Also not shown here are measurements of Titanium deuteride
[94].
to quantify its strong dependence on preparation conditions, which oers another
knob for optical response tunability. We perform a cycling experiment on Pd to
study the hysteresis between hydrogenation cycles and to determine the correlation
of the loading value with cycled optical properties. Finally, we demonstrate the
applicability of the tunable optical properties of these metal hydrides in dynamically
controlled nanostructures and thin-lm cavities. We nd that the dramatic changes
in the optical response of these materials with the hydrogenation reaction present a
wealth of possibilities for practical devices.
3.2 Optical, loading, and stress characterization of pure metal hydrides
In this section, we use the optical apparatus setup dened at the end of Chapter
2 to measure the optical, loading, and stress properties of 5 dierent pure metal
hydrides. Figure 3.1 presents the dielectric functions of the ve metals (Pd, Mg,
Zr, Ti, and V) as they hydrogenate over the entire visible to near-infrared (NIR)
59
Figure 3.1: Dynamic optical properties and loading measurements of Pd, Mg, Zr,
Ti, and V metals and their hydrides. The top panel shows schematics of the thin
metal lms used for the measurements (note: the 3 nm Pd capping layer used in
the experiment is not shown). The middle panel displays the dielectric function of
each metal and its hydride, as well as the intermediate loading states. The dynamic
loading data is shown in the bottom panel where H/M is the number of hydrogen
atoms per metal atom in the metal lattice. The colored dots indicate the times
corresponding to the optical measurements in the graphs. 1 is shown as solid lines
and 2 as dashed lines. As time progresses, the shading of the lines gets lighter.
spectral range, 250 to 1690 nm. For those who prefer viewing the optical properties
in terms of the index of refraction, see Figure 3.2. The bottom panels show the
hydrogen loading (i.e. number of hydrogen atoms per metal atom in the lattice) as
a function of time, and the colored dots represent points in time where the optical
properties are recorded (bottom plots in each panel). Measurements were taken
on ?200 nm thick metal lms capped with 3 nm of Pd that were annealed for 2
hours at 350 ?C under < 1 mtorr vacuum (with the exception of Mg, which had a
thickness of 25 nm). During Mg hydrogenation, the Mg closest to the Pd capping
layer hydrogenates, turning primarily to MgH2. This MgH2 layer is a poor proton
conductor and acts as a blocking layer for more hydrogen to penetrate into the lm,
causing the time scale for total, bulk hydrogenation to be several days [96]. Because
60
Figure 3.2: Dynamic index of refraction upon hydrogenation of Pd, Mg, Zr, Ti,
and V (a) Schematic of ellipsometric measurement on a QCM substrate through
windows in the environmental chamber at 48?, 55?, 70?, and 75?. (b-f) n and k of
each metal and its hydride, as well as the intermediate loading states. The dynamic
loading data is shown on the bottom panel where H/M is dened as the number of
hydrogen atoms per metal atom in the metal lattice, with the points of the optical
measurement delineated by points of the corresponding color in the graph above. n
is shown as solid lines and k as dashed lines. As time progresses, the shading of the
lines get lighter.
of this phenomenon, we performed measurements on a 25 nm Mg thin lm (the
thickness of the formed blocking layer) in order to fully hydride the sample. The Pd
capping layer is necessary for each metal (other than Pd itself) in order to reduce
the activation energy of H2 splitting and allow for diusion into the bulk [97]. This
61
capping layer also prevents oxidation of the underlying lm, keeping it pristine. Note
that in the case of the V/VHx data, wavelengths below 300 nm are not available due
to high stress-induced ellipsometric retardation (i.e. polarization-dependent phase
change of the transmitted light) in the windows during this measurement.
Figure 3.3 shows the dependence of the optical properties on the measured
loading values of the metals. Two main points can be drawn from this data. First,
Mg's dramatic change in optical properties is clearly evident in its dependence on
the loading values. Second, Ti has the most interesting relationship because the
real part of the dielectric function has a dierent slope depending on the wavelength
of illumination. This feature could enable a resonance to shift to longer or shorter
wavelengths upon hydrogenation depending on the incident wavelength.
Figure 3.3: Optical properties of the metal hydrides as a function of hydrogen to
metal ratio. Ti is of particular interest because the slope of the real part of the
dielectric function with hydrogen loading depends on the wavelength investigated.
This may present a new scheme for optical detection of hydrogenation in Ti. Note
that the upper time axis is not linearly spaced.
Finally, during our loading measurement, we also recorded the total stress
change in these metals as they hydrogenated. We measured these values by mea-
62
suring the curvature change in the samples interferometrically and then converting
this curvature change to stress, as outlined in Chapter 2. In Figure 3.4, we show
the total amount of stress change in each of these lms compared to the total -
nal loading in the lm. In this data, we expected to see a larger stress change in
the lms with higher loading, but we do not see any strong relationship with nal
loading amount and total stress change across material systems.
Figure 3.4: Stress change upon hydrogenation for pure metals. Stress values plotted
are the total stress change of the thin lms after the hydrogneation is complete
compared with the total nal loading number of the material. We do not nd any
strong trends between the hydrogen induced stress change and the total loading
amount.The stress values here are dened to be positive.
The results of all of the optical and loading measurements for the ve dierent
metals and their hydrides are discussed in turn below.
3.2.1 Pd/PdHx
The optical properties of both the pure Pd metal and the PdHx lms agree well
with literature values across the measured range, which serves as a standard for this
measurement technique [86, 87]. There are no clear peaks or inection points in the
Pd lm's optical properties within the wavelength range under investigation. Upon
63
hydrogenation, Pd becomes distinctly less metallic at a fairly uniform rate with the
real part of the dielectric function increasing by up to 39% in the near-infrared,
and the imaginary part decreasing by similar percentages. The lm reaches a nal
loading value of 0.67 ? 0.12, where the loading value is dened as the number of
hydrogen atoms per metal atom in the lattice, in agreement with previous results
[57, 58].
3.2.2 Mg/MgHx
Of the metals investigated, Mg has the most dramatic optical changes. The
optical properties of the metal, which matches well with reported data over the
previously studied wavelength region [98, 99], is the most lossy, with the imaginary
part of the dielectric function three times greater than that of the next metal, Pd.
During hydrogenation, it can be seen that while mass loading proceeds continu-
ously, the optical properties change abruptly from a metal to an insulator, which
is consistent with previous electrical measurements [100]. When comparing the di-
electric function of MgHx with those reported in the literature, we nd that our
measured lm appears more metallic (characterized by a decrease in the real part
and increase in the imaginary part of the dielectric function with wavelength). One
likely cause for this discrepancy is that Mg and Pd readily form an alloying layer
[101, 102]. This alloy does not become entirely a dielectric, as we expect for MgH2,
which would cause our lm to appear more metallic than the pure MgH2. After
accounting for apparent oxide formation (see 3.7 Experimental Methods), the Mg
64
demonstrated loading of 1.3 ? 0.5, which is lower than the expected 2.0 loading
value. The alloying layer discussed above could account for this lower measured
value. Another potential cause could be that the bottom few nm of the lm did not
fully hydrogenate due to the blocking layer.
3.2.3 Zr/ZrHx
Zirconium has the smallest change in optical properties of the metals being
investigated. The real part of the dielectric function exhibits very small changes
during the loading process, but the imaginary part increases signicantly. The Zr
lm also exhibited the lowest amount of loading amongst the lms, only reaching
a value of 0.47 ? 0.05. The Zr and ZrHx optical data shown here dier from the
values reported by Azofeifa et al. [85] (note that those values also greatly deviate
from previous data for Zr [103]). Thus, there appears to be an important factor
in sample preparation yet to be fully described. However, the magnitude of our
measured shift in the dielectric function upon hydrogenation agrees very well with
the data presented by Azofeifa et al. over the wavelength range they explored
[85]. The dierence in initial properties could be attributed to metal preparation
conditions, as the grain size of the metal has an eect on the optical properties of
the hydrides [87]. We also observe that at 640 nm, our data shows that there is zero
change in the real dielectric function upon hydrogenation. Points like these may be
useful as reference points in a dierential measurement scheme.
65
3.2.4 Ti/TiHx
The derived optical properties of the initial, pristine Ti metal are in good
agreement with previously reported data over the same wavelength range measured
here [104]. The local maxima in the real part of the dielectric function at ?400 nm
and ?800 nm, typically ascribed to d-band transitions [105, 106], are also clearly
visible. After hydrogenation, these undulations disappear and the TiHx appears to
become more metallic, characterized by a decrease in the real part and an increase
in the imaginary parts of the dielectric function with wavelength. However, the pre-
vious studies have not observed an increase in Ti conductivity upon hydrogenation,
which would have explained this behavior [107]. This phenomenon may simply be
due to the elimination of the above mentioned d-band oscillations. Also, a small,
broad peak at ?250 nm is visible in the data, which is characteristic of the metal-
hydrogen bond [108]. Ti loads to the highest value of any metal measured in this
experiment, achieving a value of 2.04 ? 0.12.
3.2.5 V/VHx
Over the visible wavelength range previously reported, our V metal data ex-
hibits similar trends [82]. Band structure theory predicts that there should be two
absorption peaks in the data centered at 406 and 708 nm [109]. In the literature,
it has been observed that thermal broadening merges these peaks into one broad
peak. Azofeifa et al found this peak to be at 520 nm, and we nd this peak at
a similar location of 510 nm. The broadening of this peak is attributed to large
66
electron lifetimes in the excited states [110]. Upon hydrogenation, we nd that V
has a small change in the real part of the dielectric function with a much more
signicant change to the imaginary part, which increases by more than 38% in the
near-infrared region. Further, we nd that at 750 nm, there is no change in the real
part of the dielectric function even though the imaginary part changes signicantly
upon hydrogenation, showing that the imaginary part of the dielectric function can
be controlled independently from the real part through hydrogenation over this
bandwidth. This phenomenon was also observed previously, but near 500 nm. We
attribute this dierence to sample preparation and the resulting dierence in the
grain size of the metal. The V lm achieves a loading of 0.56 ? 0.03.
3.3 Eect of annealing Ti on TiHx hydrogenation
We found that the Ti and TiHx optical properties were quite sensitive to
preparation conditions. To further explore this phenomenon, we varied the anneal-
ing conditions. Ti samples were either not annealed, annealed at 200 ?C for 2 hours,
or annealed at 350 ?C for 2 hours, with each anneal occurring in under <1 mtorr
vacuum. These samples were then characterized using the same process described
above. Figure 3.5 shows the measured dielectric functions. This experiment revealed
two interesting characteristics of Ti. When not annealed, the oscillations associated
with d-band transitions mentioned above are clearly visible and become less promi-
nent with increased annealing temperature. The hydrogen loading generally reduces
the impact of annealing, as the hydrides exhibit a smaller optical property change
67
from annealing than the pure metals. There are a few possible explanations for
this eect. First, the very large stresses involved in creating the hydride may be
producing enough dislocations that it is partially undoing the impact of annealing.
Second, it is known that the eect of defects on resistivity is reduced with higher hy-
drogen concentration and perhaps this is the eect we observe; however, additional
studies are needed to understand the root cause [107]. Nevertheless, it is clear that
annealing oers an additional opportunity for modifying how hydrogenation aects
the optical properties of Ti.
Figure 3.5: Eects of annealing on the optical properties of Ti and TiHx. (a)
Dielectric function of Ti with dierent levels of annealing. (b) Dielectric function of
TiHx measured on the same samples.
3.4 Characterization of Pd hysteresis
Some devices, such as hydrogen sensors, require the ability to cycle the metal
between a hydrogenated and unhydrogenated state. Palladium is unique compared
to the other metals because it unloads at room temperature with easily realizable,
low hydrogen partial pressures, allowing cycling by simply varying the H2 gas content
68
in the chamber. We use this process to measure the optical properties and the
loading of the Pd through three consecutive loading cycles, shown in Figure 3.6.
The red shaded regions in Figure 3.6a indicate when H2 is owing into the chamber
at 20 sccm and hydrogen is loading the metal lm. The blue shaded regions indicate
unloading phases where Ar lls the chamber, displacing the H2, and the metal lm
desorbs.
We can see in Figure 3.6b that there is a clear hysteresis in the optical prop-
erties of the Pd metal between the rst loading cycle and the subsequent measure-
ments. However, after the Pd had been loaded once, there is no discernible dierence
in optical properties between the Pd after the rst and the second unload, elimi-
nating further hysteresis. Note that there is no discernible hysteresis in the optical
response of the hydride, as demonstrated in Figure 3.6c. More data would be needed
to determine performance over a longer timescale (or number of cycles), but the data
imply the optical hysteresis of Pd can be greatly reduced with an initial treatment
of the metal with H2. Note that the above refers only to the inter-cycle hysteresis.
The well-known intra-cycle hysteresis where the ? to ? phase transition happens at
a dierent hydrogen partial pressure than the ? to ? phase transition would still
be present [12]. This inter-cycle phenomenon is most likely attributed to deforma-
tions formed in the Pd lattice upon the rst hydrogen loading. We observe that the
loading value for each PdHx state remains consistent throughout all three cycles,
although we observed an increase in unloading time between the rst and the second
cycles.
69
Figure 3.6: The optical and loading response of Pd during hydrogen cycling. (a)
Dynamic loading data (black) and the real dielectric permittivity (red) at 632 nm
plotted over three loading cycles of a 200 nm Pd lm. The red shaded areas indicate
periods of loading during which H2 is owing into the chamber, while the blue shaded
sections depict times during which Ar is owing to ush out the H2. The stars on
the plot indicate the times at which the ellipsometric measurements shown in (b)
and (c) were made. (b) shows the Pd metal data and (c) shows the PdHx data. Note
for the second two Pd measurements (iii and v) and all three PdHx measurements
(ii, iv, vi) that the measured dielectric functions are indistinguishable and, thus,
overlap each other in these plots.
3.5 Tunable nanophotonics with metal hydrides
There are many potential applications presented by the tunable optical prop-
erties of the metal hydrides investigated in this work. In this section, we detail four
examples of the potential uses of these materials: nanoparticles with variable scat-
tering cross-sections, nanorod arrays with tunable transmission, thin-lm resonators
70
with adjustable color ltering, and hydrogen switchable perfect absorbers.
We begin with the simplest of these examples, showing the ability to tune
the cross-section of nanoparticles in free space by hydrogenating the sample. The
top panels in Figure 3.7 show a schematic of the conguration with the plane wave
source being scattered by the metal and its hydride, depicting whether they primarily
increase or decrease in total scattering upon hydrogenation. The bottom panels show
the dierence in the simulated nite-dierence time-domain (FDTD) scattering cross
sections between the metal and its hydride. Figure 3.8 shows the total scattering
cross sections for both the metal and the hydride. First, a clear distinction can
be made between two types of the metals: Pd and Mg particles primarily decrease
total scattering upon hydrogenation, while Zr, Ti, and V particles show mostly
increased scattering. This eect is caused by opposite responses of the metal's index
of refraction, with Pd and Mg decreasing and Zr, Ti, and V increasing. Mg clearly
has the largest dierential response to hydrogenation, having an order of magnitude
larger change than the other metals. This was expected due to its dramatic metal to
insulator transition. The Pd nanoparticles have the next largest dierential response
and behave similarly to Mg, having a large change in response in the visible to
ultraviolet range.
Ti demonstrates wavelength-dependent regions of both increased and decreased
scattering cross-section  increasing near-ultraviolet scattering while decreasing NIR
scattering. It also is the only metal under investigation that has its largest scat-
tering change in the NIR range (400 nm particle diameter), presenting applications
that the other metals could not provide. However, Ti did exhibit a fairly small
71
Figure 3.7: Dierential scattering cross sections for nanoparticles in free space com-
posed of dierent metals. The top panels show schematics illustrating the heuristic
change in the nanoparticle cross section resulting from the change in the dielectric
response. The bottom panels show the dierences in the scattering cross sections of
metals and their hydrides for multiple particle diameters ranging from 50 - 400 nm.
For Pd and Mg, hydrogenation causes decreased scattering, as opposed to Zr, Ti,
and V where the scattering increases.
Figure 3.8: Relative scattering cross sections (Mie eciency) of metal nanoparticles
and their hydrides in free space. Solid lines indicate metal scattering cross sections
and dashed lines indicate metal hydrides. Line shading indicates particle diameter.
dierential scattering response compared to the other metals. V primarily has its
largest scattering changes in the ultraviolet. The exact magnitude and location of
72
many of these peaks cannot be determined because the simulation was limited to the
experimentally measured wavelength range (> 250 nm), which cut o these peaks.
As expected, Zr has the smallest dierential response because of its small change
in optical properties upon hydrogenation. The response is on the same order as Ti,
with its maximum responses spanning a wider range of wavelengths than the other
materials, beginning in the ultraviolet and continuing to the NIR.
Next, we simulate periodic arrays of nanorods on a glass substrate, showing
large resonance shifts and changes in transmission upon hydrogenation. The relative
change in transmission upon hydrogenation for these metals is shown in Figure 3.9.
The arrays have a 500 nm period in both the parallel and perpendicular directions,
a rod width of 100 nm, and a length that varies from 150 to 400 nm (the electric
eld is parallel to the length of the rods). The total transmission spectra can be seen
in Figure 3.10. Once again, Mg and Pd behave dierently than the other metals in
terms of responses upon hydrogenation: Mg and Pd have an increase in transmission
upon hydrogenation, while Zr, Ti, and V exhibit a decrease in transmission.
Mg has the strongest relative response, exhibiting a full 3800% relative trans-
mission increase for the 400 nm rods. Pd shows the next highest response, with
a 190% relative increase in transmission at the 400 nm length. Mg and Pd have
the narrowest peaks, causing for a more spectrally localized response upon hydro-
genation. Zr and V have very similar resonant response locations in the visible and
NIR, with V exhibiting stronger changes in transmission. These materials have their
peak changes in the visible and NIR. Ti has transmission magnitude shifts similar
to Zr, but with its responses spanning further in the IR, allowing for a wider usable
73
Figure 3.9: Relative change in transmission upon hydrogenation of periodic nanorod
arrays. The top panels show schematics of a nanorod array on a glass substrate.
The rods are spaced 500 nm apart in both the parallel and perpendicular directions
and are 100 nm wide. The rod length is varied from 150 to 400 nm. The polarization
of the electric eld is parallel to the length of the rods. The bottom panels show the
relative dierence of the transmission spectra between the metals and their hydrides.
The transmission for Mg and Pd nanorods increases upon hydrogenation for most
of the spectrum, while those made of Zr, V, and Ti decrease.
Figure 3.10: Full transmission spectrum of periodic nanorod array. Solid lines indi-
cate metal scattering cross sections and dashed lines indicate metal hydrides. Line
shading indicates length of individual nanorod.
bandwidth. These transmission dierences allow for in situ tunability of optical re-
sponses for various wavelength ranges without having to physically move any piece
74
of the structure or electrically activate any part of it.
The dynamic optical properties demonstrated above suggest themselves to
two potential thin-lm applications: tunable color lters and switchable perfect
absorbers. Figure 3.11 and Figure 3.12 depict two embodiments of these devices: a
Fabry-Perot cavity with the metals acting as both the top (partial) mirror and the
bottom mirror separated by a dielectric and a thin lm Si absorber with the metals
acting as a lossy bottom mirror. For each metal, several representative thicknesses
of the Si or SiO2 layers are shown. Note that tunable color lters and switchable
perfect absorbers are not mutually exclusive and dier only in their gure of merit,
allowing a single structure to act as both. In all respects, the Fabry-Perot cavity has
better performance. However, the Si systems are of interest due to their simplicity
and because a Si-based perfect absorber is useful for several applications, such as in a
Si/metal photodetector [111]. In Figure 3.11, we present the materials that function
well as both tunable lters and switchable perfect absorbers in either structure: Mg,
Pd, and Ti.
Tunable color ltering is characterized by shifts in the wavelength of the res-
onance [80]. In each case shown, hydrogenation of the metal produces an easily
measurable and generally visible change in the resonant wavelength. The largest
shifts in the Fabry Perot modes shown are 360 nm (or 1 eV), 100 nm (or 0.4 eV),
and 50 nm (or 0.1 eV), for the Mg, Pd, and Ti, respectively. The largest shift in
the resonances of Si on metal are 40 nm (or 0.24 eV), 50 nm (or 0.17 eV), and 70
nm (0.04 eV) for the Mg, Pd, and Ti, respectively. These values compare well with
Duan et al. which demonstrated wavelength shifts of ?150 nm in a Mg nanoparticle
75
Figure 3.11: Switchable perfect absorbers and tunable color lters using thin-lm
structures containing the three metals that oer the largest optical response to
hydrogenation: Mg, Pd, and Ti. The thickness of either the Si or SiO2 layers are
labeled on the charts for each colored curve, and the bottom metal is considered
bulk. The pure metal is denoted with a single solid line and the metal hydride with
dashed lines. (a-c) Fabry-Perot cavities comprised of Mg, Pd, and Ti, respectively,
for top and bottom mirrors separated by SiO2. (d-f) Si thin-lm structures on Mg,
Pd, and Ti, respectively, undergoing hydrogenation. All of these structures perform
well as either switchable perfect absorbers (>100x changes in reectivity) or tunable
color lters (a minimum of 40 nm resonant wavelength shift).
system or wavelength shifts of ?150 nm for Pd nanorings presented by Zori? et al.
[80, 112].
On the other hand, the gure of merit for the switchable perfect absorbers is
the dierence in maximum absorption or reection [113, 114, 115]. The six structures
shown here also demonstrate excellent performance as switchable perfect absorbers.
In fact, all of the systems in Figure 3.11 have resonances that have more than
76
Figure 3.12: Switchable perfect absorbers and tunable color lters using Zr and V.
The thickness of either the Si or SiO2 layers are labeled on the subplot for each
colored curve. The pure metal is denoted with a single solid line and the hydride
with a dashed line. (a,b) Fabry-Perot cavities comprised of Zr and V respectively for
top and bottom mirrors and SiO2 centers. (c,d) Less complex Si on metal structures
for Zr and V respectively.
2 order of magnitude of switchability, including extreme cases like the Pd Fabry-
Perot structure with 121 nm of SiO2, which changes absorption by more than ve
orders of magnitude, or the Mg Fabry-Perot structure with 204 nm of SiO2, which
switches by more than four orders of magnitude. This oers signicantly improved
performance compared to Walter et al. who presented reection changes in Pd
nanodiscs or Tittl et al. who demonstrated perfect absorbers in Pd gratings, who
both reported reection ratios of <1000 [114, 115]. As above, it is interesting to
note that the structures involving Ti exhibit unique behavior amongst the materials
77
we investigated. The resonances in the Ti devices shift to shorter wavelengths after
the hydrogenation reaction due to the increase in refractive index, in contrast to
the redshifts demonstrated with the other two metals. Using this eect, it may be
possible to combine layers of, for example, Pd and Ti to exaggerate resonant shifts.
These types of simple resonant structures show the great promise for these materials
as a part of tunable optical systems.
In Figure 3.12, we show the same structures with Zr and V. In these sys-
tems, the Fabry-Perot cavity structure can still obtain large reection changes for
switchable perfect absorbers, but have much smaller resonance shift, making them
poor tunable color lters. The Si devices show poor absorption and resonance shifts
for these metals as well, thus for future device design, we will focus our attention
primarily on Mg, Pd, and Ti.
3.6 Conclusions
In summary, we have mapped the dynamic optical properties of the hydro-
genation reaction of ve dierent metal lms and have simultaneously recorded the
real-time hydrogen loading data. We have shown that the optical properties can
be tuned for the dierent metals, with the largest changes exhibited in Mg and
the smallest changes in Zr. We demonstrated further potential tunability for Ti
structures by determining the eect of annealing on optical property changes in the
metal and its hydride. Pd was shown to have a large hysteretic optical response on
the rst cycle, and then very consistent optical properties between the second and
78
third cycles, while having no hysteresis in the hydride state. Four dierent nanos-
tructured geometries were studied based on the measured optical properties, which
demonstrated several potential applications for the tunability of these metals. Mg
was shown to have the largest response for all structures due to its extreme change
in optical properties upon hydrogenation. Mg and Pd structures exhibited a redshift
in their resonances, while Zr, V, and Ti systems are characterized by blue shifts. Ti,
Mg, and Pd were all found to be excellent candidates for perfect absorbing devices,
making them widely applicable for such designs. This work acts both as a point of
comparison between these materials and to demonstrate their usefulness as tunable
optical materials for novel dynamically switchable photonic devices.
3.7 Experimental methods
3.7.1 Sample fabrication
The substrate for each of the samples is a 5 MHz Incon QCM. Before sample
deposition, each QCM is cleaned with acetone, methanol, isopropyl alcohol, and
water to remove any particles or organics on the surface of the Au electrode. The
12.7 mm diameter thin-lm disks are dened by a machined Al shadow mask. Each
metal is deposited as part of a stack using electron beam evaporation (Angstrom
NEXDEP). First, a 3 nm Cr adhesion layer is deposited, followed by 200 nm of
the metal under test (25 nm for Mg), and nished with a 3 nm Pd capping layer.
This nal layer acts as a semi-transparent and permeable surface to split the H2
molecules. For each deposition, at least two QCMs are included: the rst for the
79
optical and in situ loading measurement and the second for a more precise and com-
plete loading measurement in a separate environmental chamber which incorporates
stress compensation (see Chapter 2 for details) [95]. Having the samples deposited
in the same run ensures the similarity of the metals on each QCM for comparison
of loading. With each deposition, a lithographically dened 1 cm x 1 cm square is
included for determining the sample height via AFM (Cypher, Asylum Research).
Immediately after sample deposition, the samples are annealed in < 1 mtorr vacuum
at 350 ?C for 2 hours with the exception of the Ti samples used in the annealing
study.
3.7.2 Optical measurement
The optical properties for each of the materials are measured via in situ spec-
troscopic ellipsometry as outlined in Chapter 2. Each sample is measured imme-
diately after annealing in order to avoid contamination of the sample or excessive
oxidation. Before an optical measurement, the sample chamber is purged with Ar
at ?200 sccms for a minimum of 1 hour to remove any trace hydrogen left in the
system. Assuming complete mixing, this brings the H2 partial pressure in the 45 cm
3
chamber to <10?5 bar. Dynamic optical measurement and window compensation
are then performed as outlined in Chapter 2. During the Pd cycling measurements,
the H2 is ushed from the chamber by owing 60 sccm of Ar during the unloading
steps. The metal and nal metal hydride are t using and Kramers-Kronig con-
sistent B-spline and the dynamic data is t using two Bruggeman EMAs: one for
80
the metal under investigation and one for the Pd capping layer. The two materials
input into each EMA are the pure metal model and the hydride model. For full
modeling details see Chapter 2.
3.7.3 Loading measurement
In a separate environmental chamber, the second QCM sample from the depo-
sition is run under identical environmental conditions to the original sample. The
loading is calculated as outlined in Chapter 2, where the stress of the QCM amongst
other extraneous eects are corrected for in the loading calculation [95]. To obtain
the loading data of the optically measured sample, we rst subtract away the fre-
quency change due to the changes of gas partial pressures in the chamber. We then
oset and normalize the recorded frequency change. The baseline frequency before
H2 is introduced to the chamber is dened to be zero (oset), and the stabilized
frequency after the hydrogen loading is complete is dened to be the calculated
loading value from the duplicate sample (normalize). During the Pd hysteresis
study, a concurrent hysteresis in stress required that each cycle be normalized in-
dependently. The loading cannot be directly calculated in the optical measurement
chamber because the stress cannot be properly characterized interferometrically due
to the geometry of the setup when taking ellipsometric data.
There were two cases where we had to add extra processing to the loading
calculations of the metals: Mg and Pd. For Mg, the loading data was normalized in
a slightly dierent process due to the possibility of a thin MgO layer existing at the
81
beginning of the experimental runs. Once hydrogen was introduced to the chamber,
the QCM frequency sharply increased, indicating mass loss from the system, before
the usual reduction in frequency due to loading. We expect that this eect is due to
a MgO layer that is being reduced upon the introduction of the H2. To account for
this eect, we normalized the frequency to the top of this peak, dening it as the
baseline value. For Pd, the duplicate sample was run in a D2 environment instead
of H2. We added the known conversion factor of 0.07 to the nal loading value to
convert to the proper H2 loading [12].
82
Chapter 4: Investigation of Physical Properties of Commercial Near-
Zero-Index Materials
In this chapter, we take a brief detour from metal hydrides and investigate
commercial NZI materials to potentially be incorporated into our metal hydride
designs. TCO materials have been found to be ideal NZI materials, with strong NZI
resonances with low losses in the NIR. In this chapter, we investigate the optical
properties of 51 TCO materials and report the strength and bandwidth of their NZI
resonances. We nd positive correlations between |n| and the NZI bandwidth, as
well as between the resistivity of the lm and |n|.
4.1 Background of NZI materials
Near-Zero-Index materials are becoming essential in nanophotonic designs and
optical manipulation. As the name suggests, NZI occurs when the index of refrac-
tion of a material approaches zero over a certain wavelength range. This is achieved
when a conductive material has the real part of its permittivity cross zero com-
bined with low Drude damping in the material (low loss). These properties allow
for many extreme optical eects that stem from the phase velocity and wavelength
approaching innity as the index approaches zero, along with massive electric eld
83
enhancements within the material. Since the original seminal works on NZI materials
[116, 117, 118], there have been countless applications utilizing these novel proper-
ties. These applications range from creating various near-perfect absorption or trans-
mission devices [119, 120, 121, 122, 123, 124, 125], supercoupling light into narrow
waveguides [126, 127, 128], enhancing of optical nonlinearities [129, 130, 131, 132],
and increasing plasmonic control [133, 134], among many others.
Transparent conducting oxides have become the most common NZI materials
due to their broad NZI region located in the telecom wavelength range, low losses at
the dielectric cross-over-point (where the real part of the dielectric function crosses
zero), and tunability of the NZI wavelength based on deposition conditions. In
particular, indium-tin-oxide [129, 132, 133, 135, 136, 137, 138], aluminum-doped
zinc oxide [119, 123, 134, 136, 139], and uorine-doped tin oxide [140, 141] have
been well-studied for their NZI properties and applications. While being a benet
in some cases, the large dependence of the optical properties of these materials on
their deposition parameters can introduce diculty in experimental replication, due
to materials being deposited using dierent fabrication tools and methods across
dierent institutions. As an example, the NZI resonance of ITO varies from 1260
to 1920 nm depending on the carrier concentration in the material [142]. The dif-
ference in properties between the two ends of this region could completely change
the performance of a device.
Recently, there have been many commercial companies that sell TCO materials
on substrates that can be used for NZI applications. By being able to purchase
from one of these suppliers, the fabrication issues from batch to batch along with
84
comparisons across dierent tools are eliminated. In this chapter, we investigate
the optical properties of various commercial TCO samples and report the existence,
strength, and width of their NZI resonances. We focus on three dierent types
of commercially available TCO samples, ITO, FTO, and AZO, from a total of 12
dierent suppliers. Beyond their optical properties, we also report the thickness and
resistivity for all of the measured samples. Using these measured properties, we nd
general correlations relating the optical and electrical properties of the materials. We
nd that both the strength and the bandwidth of an NZI resonance are correlated
with the location of the resonance and that the strength of the resonance is loosely
correlated with the resistivity of the TCO lm. We hope that these results provide
a resource for future fabrications of repeatable and comparable NZI devices.
4.2 Measurement and optical modeling scheme
To measure the optical properties of these samples, we use variable-angle spec-
troscopic ellipsometry (Woollam M-2000). We take measurements of each sample
at 65?, 70?, and 75?, and we use the same tool to take transmission intensity mea-
surements. The raw optical data is t using a general oscillator model with one
Drude oscillator [143] and up to three Tauc-Lorentz oscillators [144]. The number
of Tauc-Lorentz oscillators is determined by only adding a further oscillator to the
model if this addition causes a >10% decrease in the mean-squared-error (MSE)
of the t. This method was used to avoid over-tting the data with spurious reso-
nances. For each sample where the company provided a bare substrate without the
85
TCO deposited, the optical properties of the substrate were measured separately,
and these properties were input into the model tting the TCO data. For companies
that did not provide a substrate, we used the material properties from the Woollam
database for the substrates (soda-lime glass, oat glass, or glass slide dependent on
what substrate the company stated they used) as the substrates in our ts. Our
ts also accounted for reections from the backside of the substrate. The thickness
of the TCO lms was determined from the optical tting, with the transmission
intensity data allowing us to unambiguously determine the value.
Many of the FTO samples had to have a slightly adjusted tting model. In-
stead of having a single layer of TCO, many of the FTO samples provided were
multi-layer structures. On top of the glass substrates, a thin layer of SnO2 was
deposited, followed by a thin SiO2 layer, and then nally the active FTO layer. In
this chapter, we only report the optical properties of the top active FTO layer. In
our optical model, we dene the optical properties of the SnO2 with a Cauchy model
and dene the properties of the SiO2 layer with experimental found data from sput-
tered SiO2. In the model t, we t the thicknesses of these two intermediary layers
and the thickness of the top FTO layer, along with the oscillator values dening
the properties of the FTO. These two added t parameters add extra variance to
our ts, causing a larger error in the reporting of these FTO properties. Two of the
FTO samples, sourced from Biotain Crystal and MSE Supplies, were both single-
layer samples and were t using the same process as the ITO and AZO samples
detailed above.
86
4.3 Properties of TCO lms
Typical results for the optical properties of ITO, FTO, and AZO are shown in
Figure 4.1. Although there can be a fairly large variation from sample to sample,
data from each material generally follows the same general shape and trends. ITO
normally displays a strong NZI resonance in the NIR with a higher index of refraction
in the visible and higher attenuation further into the infrared (IR). FTO usually
shows a much weaker resonance than ITO, and this NZI resonance is located further
into the IR. This same trend holds with AZO. In the short wavelength range, AZO
also exhibits a peak in attenuation. For the individual optical properties, company
names, and nominal resistances for each of the 51 samples reported in this chapter,
see Appendix B.
Figure 4.1: Characteristic optical properties of a) ITO, b) FTO, and c) AZO. These
properties are representative of the general trends in optical properties for each of
the materials. ITO samples exhibit a strong NZI resonance between 1100 -1400 nm,
while FTO and AZO have resonances beyond our experimental measurement range.
With the measured properties of this multitude of samples, we explored var-
ious correlations in the NZI resonance data. In Figure 4.2, we plot the minimum
?
magnitude of the refractive index, |n| = n2 + k2, versus the wavelength where that
minimum is found. In the plot, the grey shaded region denotes extrapolated data
87
as the range of our optical measurement ends at 1690 nm. For data in this region,
we extrapolate the optical properties using the oscillators dened from tting the
ellipsometric data. We note that if there are additional phononic resonances or ad-
ditional bound electron resonances in this region, it could further aect the optical
properties in unaccounted for ways, adding a higher error for values in this region.
The further away the resonance is from the measurement wavelength maximum of
1690 nm, the higher the uncertainty in the resonance strength and location. For
two ITO samples that we measured, we did not include the data in the plots in this
chapter because the extrapolated NZI resonances were too far in the IR to conclude
anything about their properties in that region. Their measured optical data are
reported with the other samples in Appendix B.
We see two general trends in the NZI resonances from our measurements. The
rst is that for these samples, the location of the NZI resonances is segmented by
material. The center wavelength for the ITO NZI resonance occurs in the ?1100
 1400 nm range, with one outlier. As we traverse further into the IR, we nd
the region containing the FTO resonances ?1500  1900 nm, and nally the region
where the AZO can be found ?1950  2100 nm. The second trend that we nd is
that the stronger (lower |n|) NZI resonances occur at shorter wavelengths. As the
wavelengths extend further into IR, the resonances become slightly weaker. We t
a linear trendline to this plot as a guide to the eye to show this relationship (R2 =
0.68).
The magnitude of an NZI resonance is not the only important characteristic
of the material. The bandwidth of the resonance (hereby dened as the width
88
Figure 4.2: Strength of NZI resonance vs the wavelength at which the reso-
nance?occurs. For each TCO sample, the magnitude of the index of refraction,
|n| = n2 + k2, is plotted against the wavelength where this minimum occurs. Our
experimental measurement only goes up to 1690 nm, thus the grey shaded region
beyond this wavelength shows extrapolated data of the oscillator ts in the exper-
imental region. Added Lorentz resonances in this region could aect these values.
ITO samples are shown as purple circles, AZO as green squares, and FTO as red
triangles. The dashed grey line is a linear t to the data to guide the eye.
of the spectral region where |n| < 1) is also of vital importance for broadband
devices required to perform at many wavelengths. In Figure 4.3, we plot the
bandwidth of the NZI resonance for each of our measured materials versus the
location of the center wavelength of the NZI resonance (location of the minimal |n|
value). We nd that as the location of the resonance moves further into the IR,
the bandwidth increases. The grey shaded region in this gure depicts the region
where the center wavelength of the NZI resonance is found by extrapolation. We
also note that in this gure, the two FTO samples whose center wavelengths are
located in the measured region also involve some extrapolation, as the region where
|n| < 1 extends beyond our measurement maximum of 1690 nm. For the materials
89
that do involve extrapolation, we nd a much wider range in results deviating from
our linear trendline. In this study, we cannot determine whether this high variance
eect is caused by additional error from the extrapolation, dierence in material
responses (as most of these materials are FTO and AZO), or if materials with NZI
resonances in this regime behave dierently. In all likelihood, it is some combination
of these three factors. Figure 4.3b shows the same plot with only the data with
no extrapolation, which happens to entirely consist of ITO samples. This trend is
positive with a few outliers (R2 = 0.58), with most of the data points supporting the
conclusion that materials with NZI resonances at longer wavelengths have a broader
bandwidth of this resonance.
Figure 4.3: Bandwidth of NZI resonance vs the center wavelength of the resonance.
We nd that as the NZI minimum wavelength moves further into the IR, the band-
width of the NZI region increases. Bandwidth is dened as the size of the spectral
regime where the |n| < 1. They grey shaded region depicts extrapolated data. ITO
samples are shown as purple circles, AZO as green squares, and FTO as red trian-
gles. The grey dashed line is a linear t to the data to guide the eye of the reader.
b) A plot of the samples that involved no extrapolation in the bandwidth.
This positive correlation between the location and the bandwidth of the NZI
combined with the correlation between the strength and location creates a trade-o
90
in device design. By choosing a sample with a resonance closer to the visible regime,
you can achieve a stronger resonance, but at the cost of a smaller bandwidth. Thus
for some applications, the ideal material might occur further into the IR where a
strong resonance can still be achieved and a more broadband eect is possible.
For each sample, we also took 4-point probe resistivity measurements to char-
acterize the electrical properties of the TCO lms. We used a Signatone 4-point
probe device readout with a Keithley 2400 sourcemeter. The dimensions of each
sample measured were large enough that no size compensation had to be applied to
the sample measurement. Each sample was measured in 3 dierent orientations near
the center of the sample and the results reported are the averages of these measure-
ments. In Figure 4.4, we report the resistivity of each TCO lm vs |n|min. We nd
that as the resistivity of the lms increase, |n|min also increases. This correlation is
in agreement with the literature, where it has been found that higher carrier con-
centrations in ITO samples lead to stronger NZI resonances with these resonances
located at shorter wavelengths [142]. We nd that FTO and AZO lms have higher
resistivities than the ITO lms, requiring their lms to be much thicker to obtain
similar sheet resistance values. For samples where a thin coating is required with
a high conductivity, ITO seems to be the only available commercial option of the
three materials we investigated. In Figure 4.4b, we isolate the unextrapolated ITO
samples and nd a similar correlation over a smaller scale, with all of the resistivities
of these samples located between 1.2 and 2 ?? ?m.
We summarize the complete ndings of our optical and electrical measurements
in Appendix B for reference. These include the minimum |n| achieved, the location
91
Figure 4.4: Strength of NZI resonance vs the resistivity of the lm. We nd that
generally as the resistivity of the lm increases, so does the minimum magnitude
of the index of refraction. ITO samples are shown as purple circles, AZO as green
squares, and FTO as red triangles. The grey dashed line is a linear t to the data to
guide the eye of the reader. b) A plot of the samples that require to extrapolation
in the optical properties.
of the minimum, the NZI bandwidth, the measured thickness t of the TCO, and the
measured resistance R.
In conclusion, we have surveyed the optical and electrical properties of 51 dif-
ferent TCO samples and found a wide range of potential NZI properties. We nd
that ITO is the most readily available NZI material, with many dierent samples
having a |n| < 0.6. We also nd a correlation between the location of the NZI reso-
nance and the strength and bandwidth of that resonance, with shorter wavelengths
leading to stronger resonances with a smaller bandwidth and longer wavelengths
leading to weaker resonances with a wider bandwidth. We also nd a loose corre-
lation between the strength of the NZI and the resistivity of the TCO lms, with
a lower resistivity leading to a slightly stronger NZI. We hope that these results
inform the consistent device design of future NZI materials, showing the breadth
and availability of dierent TCO materials.
92
Chapter 5: Highly Switchable Absorption in a Metal Hydride Device
Using a Near-Zero-Index Substrate
Optical switchability is an important functionality for photonic devices, which
allows them to accommodate a wide range of applications. In this chapter, we
propose a switchable absorption device consisting of a Pd-capped Mg thin lm de-
posited onto a near-zero-index substrate. By utilizing Mg's extreme optical changes
upon hydrogenation and combining it with the high optical contrast of the NZI sub-
strate, we can create a device that is fully switchable from a highly reective state
to a broadband absorbing state. When modeling the substrate as a Drude material
with a plasma wavelength of 600 nm, we calculate an absorption change of >70%
from 650  1230 nm, with a peak total absorption of 78% at 905 nm. We experi-
mentally demonstrate this eect using 25 nm of Mg with a 3 nm Pd capping layer
deposited onto an ITO-coated glass substrate. This device achieves an absorption
change of 76% at 1335 nm illumination, with a maximum absorption of 93% in the
hydride state, utilizing ITO's NZI region in the NIR. By tuning the NZI region of
the substrate, this eect can be extended from the visible through the infrared.
93
5.1 Background and introduction
Switchable absorption across the electromagnetic spectrum is a highly desir-
able functionality for applications from thermography to solar absorbers and color
displays [80, 145, 146, 147]. Dierent absorption bandwidths are necessary for these
dierent applications, with solar applications requiring high functionality in the vis-
ible, while telecom sensors require the sensitivity to be in the near-infrared. Specif-
ically, absorption in thin metal lms has found wide application in hot carrier pho-
todetection and solar cells, where hot carriers created in the metallic thin lm can be
harnessed [111, 148, 149, 150, 151, 152, 153]. By adding a switchable functionality
to this eect, devices can be turned on and o with an external stimulus, allowing a
wider range of applications, such as switchable solar windows or solar cells [16, 154].
There are multiple ways to achieve switchable absorption for these various
applications. One common method is to electrically trigger the optical change,
whether by applying a voltage across a diode structure [155, 156], by electronically
aligning a liquid crystal array [157], by actuating a micro-electro-mechanical system
[158], or by electrostatically doping an active optical material, such as graphene
[159]. Another method is to thermally activate a phase change in the device, such
as with VO2, which goes through a phase transition at 68?C [160, 161, 162, 163].
Alternatively, these optical changes can be triggered by atmospheric changes, such
as H2 gas exposure. Metal hydrides have recently been utilized for their switchable
optical properties in the visible and NIR for various purposes including H2 sensing
and physical encryption [20, 23, 80, 115, 164, 165, 166]. Mg in particular exhibits
94
a dramatic optical change upon H2 exposure, transitioning from a lossy metal to
a dielectric MgH2 [13, 165, 167]. This change upon hydrogenation can be utilized
to switch a device from a highly reective metallic state to an absorbing state.
Many other switchable absorption designs are based on metamaterials that require
complex fabrication procedures. In contrast, devices based on thin-lm Mg only
require a thin lm deposition and no lithographical processing.
To further increase the functionality of thin-lm devices, the optical properties
of the substrate can be tuned to create a Fabry-Perot-like resonance within the active
layer [111, 120, 168, 169]. By optimizing the reection at the thin lm/substrate
interface, strong destructive interference can be engineered, causing high absorption
in the lm. For many thin metal lms, this eect can be strongly enhanced in the
presence of a NZI substrate where the real part of the index of refraction approaches
zero [119, 120, 123]. Unlike traditional Fabry-Perot resonances that require the
thickness of the thin lm to be a quarter of the wavelength of the incident light,
utilizing this eect allows for interference eects in lms 10-100x thinner. Recently,
NZI materials have been utilized for a wide range of applications, from light funneling
to perfect absorbers, amongst many others [124, 128, 170].
In this chapter, we propose a thin lm device structure for switchable absorp-
tion that can be tuned across the visible into the NIR spectrum by adjusting the
NZI resonance of the substrate, achieving >90% absorption experimentally. The
structure consists of a 25 nm Mg layer capped with 3 nm of Pd on an NZI coated
glass. By utilizing the extreme optical properties change of Mg upon hydrogenation,
we can switch the device from a highly reective metallic state to a light-absorbing
95
hydride state. Simulations show this eect can create a > 70% change in absorption
over a range of 650 -1230 nm when the substrate is modeled as a Drude material
with a plasma resonance ?p = 600 nm and a damping factor ? = 10
13 Hz, with
a peak absorption change of 78% at 905 nm. We experimentally demonstrate this
structure on an ITO substrate with an NZI resonance in the NIR and nd a peak
absorption change of 76% at 1335 nm, where the absorption in the hydride state is
93%.
5.2 Concept and design
Our switchable absorption structure is depicted in Figure 5.1. It consists
of a SiO2 base coated with 350 nm of NZI (labeled substrate), 25 nm Mg, and 3
nm of Pd. The Pd capping layer is necessary for this design to catalyze the H2
dissociation reaction on the top of the structure [97, 167]. The Pd also acts as a
protective layer above the Mg, preventing oxidation into MgO [171, 172]. Exposure
to H2 gas switches the device from a reective metallic state into an absorbing
hydrogenated state. The H2 dissociates into H atoms at the Pd surface and diuses
through the thin Pd layer into the Mg to form MgH2. As Pd hydrogenates into
PdHx, it becomes less conductive, but remains in a metallic state. Meanwhile, Mg
transforms entirely into a dielectric upon hydrogenation. In this dielectric state,
the light is now able to pass through the MgH2 and interact with the substrate,
as opposed to being almost entirely reected by the Mg layer in the metallic state.
Figure 5.2 shows the simulated reection and transmission plots for each state.
96
Figure 5.1: Device design and simulated absorption spectra for various substrates.
a) Device architecture in the inert, metallic state, and b) when exposed to H2.
For each substrate (SiO2, Au, or NZI), the thicknesses of the Mg and Pd layers
were optimized between 3 - 50 nm for maximum absorption change. For the NZI
substrate, the thicknesses used were tPd = 3 nm and tMg = 25 nm. For both
the Au and SiO2 substrates, the thicknesses were tPd = 3 nm and tMg = 50 nm.
The substrate for each case was kept constant at 350 nm. c) Absorption spectra
of the metallic layers (Pd + Mg) with SiO2 (red), Au (green), and NZI (blue)
substrates. The solid (dashed) lines are the spectra in the metallic (hydride) state.
The absorption curves for SiO2 and Au substrates are almost completely overlapping
in the metallic state. The NZI substrate is dened to have an index of n? = 0.01 +
1i across the spectrum. These simulations account for the 30% volume expansion
of Mg upon hydrogenation (e.g. 25 nm metallic Mg becomes 32.5 nm MgH2).
In Figure 5.1c, we compare our proposed structure to other physical substrates
by simulating the absorption spectra using the transfer-matrix method (TMM)
found in the literature [173]. Optical properties for the Mg, Pd, and their hydrides
are taken from Palm et al. [164], Au from Johnson and Christy [174], and SiO2
from Gao et al. [175]. The index of the NZI material was dened to be n? = 0.01
+ 1i (eects of dispersion are discussed and modeled below). In the simulations,
97
Figure 5.2: Simulated a) reection and b) transmission plots for device with 350 nm
NZI substrate. Solid lines indicate response in the metallic state and the dashed
lines indicate response in the hydrogenated state. In the metallic state most of the
light is reected, as opposed to little reection in the hydride state. In neither case
is there appreciable transmission through the NZI layer.
the SiO2 base was considered to be semi-innite and each substrate was dened to
be 350 nm. For the Au substrate, 350 nm is optically thick, so that the device can
be practically thought of as an innite Au substrate, and because the SiO2 sub-
strate layer is on a SiO2 base, the dened thickness does not aect the results as
the substrate is continuous with the base layer. For the NZI material, we nd that
as the substrate thickness increases, we achieve a greater absorption change in the
material. Once the substrate becomes greater than ?300 nm, this eect begins to
level o. We chose 350 nm to achieve this high absorption change, while still having
a thin enough layer that is reasonably achieved in common NZI materials, such as
TCOs [134, 139]. Figure 5.3 shows the complete spectral dependence of the NZI
substrate thickness on the absorption change.
For each dierent substrate, we optimized the Mg and Pd thicknesses that
would create the largest absorption change upon hydrogenation. We dene the total
98
Figure 5.3: Thickness optimization of NZI substrate. Change in absorption dened
as absorption of the Mg and Pd in the metallic state subtracted from their absorption
in the hydride state. We can see that the absorption change is greater for thicker
lms of the NZI, but with diminishing returns as the lm gets larger than ?300 nm.
For this optimization the Mg layer thickness was set to 25 nm and the Pd thickness
to 3 nm. The NZI optical properties refractive index was dened to be n? = 0.01 +
1i.
absorption as the combined absorptions in the Pd and Mg layers (note: absorption in
the substrate and SiO2 are negligible, see for example Figure 5.5). In our thickness
optimization, we applied the maximum thickness of Mg as 50 nm. The full Mg lm
would not hydrogenate if it were greater than 50 nm due to the formation of a well-
documented MgH2 hydrogen blocking layer [13, 176]. We applied a lower thickness
bound of 3 nm on the Pd layer to ensure a uniform lm and complete coverage of
the Mg. Expansion of the metal hydrides also needs to be taken into account, as
the lattice expands as hydrogen enters the material. The Mg lattice is found to
expand by 30% upon hydrogenation [13] (e.g. 25 nm metallic Mg becomes 32.5 nm
MgH2), and this eect was included in all simulations. Pd has a much smaller lattice
expansion, about 12% [60], which ended up being negligible in our simulations since
99
the optimized Pd thicknesses were very small. We found the optimized thicknesses
for each layer were tPd = 3 nm and tMg = 25 nm for the NZI substrate and tPd =
3 nm and tMg = 50 nm for both the Au and SiO2 substrates. Figure 5.4 shows
the full absorption change as a function of the Mg and Pd thicknesses at 6 dierent
wavelengths.
Figure 5.4: Thickness optimization for Mg and Pd thin lm layers. For most wave-
lengths, the absorption change is maximized for thinner Pd. In order to get the
catalytic and sealing eect of the Pd in the device, at least a 3 nm coating is neces-
sary, so it was set as our lower bound for the optimization. For 400 nm illumination,
the maximum falls at 5 nm, but still has strong absorption at 3 nm. The Mg is max-
imized between 20 and 30 nm depending on the illumination wavelength. We found
that a Mg thickness of 25 nm caused for the best maximization across all of the
shown wavelengths to create the most broadband eect. The thicknesses of these
plots are the thickness of the metal layers and the simulations take into account the
30% Mg expansion upon hydrogenation.
In the metallic state, Mg is optically opaque at a thickness of 50 nm, so
the absorption curves for the SiO2 and Au substrates almost entirely overlap. In
the NZI material case, the light slightly penetrates the Mg layer at 25 nm but is
still mostly reected with little measured absorption, under 20% for most of the
100
spectrum. Upon hydrogenation, we see a dramatic increase in absorption for each
substrate. The extremely reective Mg becomes a partially transmitting MgH2. We
see that the choice of substrate greatly aects the absorption in the metal hydrides,
as the light now interacts at the MgH2/substrate interface. For the SiO2 substrate,
we see the smallest increase in absorption, mostly due to the increase in path length
of the light in the metal hydrides as the light transmits through it, with little added
benet from the reection o the substrate interface. The Au substrate shows a
much higher absorption increase by creating a cavity eect within the MgH2, with
light reecting between the Au substrate and PdHx capping layer. The largest
attainable absorption change is found for the NZI substrate. In this case, a strong
Fabry-Perot-like cavity is created within the MgH2, creating high absorption due to
the destructive interference eects. With these dened optical properties, we can
achieve an absorption of >80% from 400  1650 nm and >90% from 625  1300
nm in the hydride state. We note that the large broadband result depicted by the
NZI substrate in Figure 5.1c is not possible with a realistic substrate, because a
constant complex index of refraction over this entire wavelength region does not obey
Kramers-Kronig consistency. Instead, this result shows the potential for this high
switchable absorption response across the visible spectrum and into the infrared if
the substrate's index of refraction has a low real part with the imaginary part close
to 1 for any span of these wavelengths. We also note that there will be some alloying
between the Pd and Mg layers upon deposition [167]. To avoid this alloying in thin-
lm stacks, it is common to use a Ti interlayer between the Mg and Pd layers. The
optical properties being used for these simulations were taken on a thin lm stack
101
without this Ti interfacial layer, thus the properties being used are accurate for
these simulations and account for this interfacial layer.
The absorption of the light in the hydride layers is split between the PdHx
and MgH2. The PdHx layer is responsible for the majority of the short-wavelength
absorption up to?500 nm. At this point, the MgH2 begins to absorb more light, with
the absorption in the Pd tapering o. The reason for this delineation is that PdHx
still behaves optically like a metal with higher attenuation for shorter wavelengths.
It gets a large increase in absorption from the cavity eect, with the path length
of light traveling through it increasing dramatically from the multiple reections
within the MgH2. The MgH2 has higher attenuation beginning in the upper visible
region, which is where it begins to dominate the absorption. The combination of
the absorption in these two layers allows for the near-constant absorption change
observed in Figure 5.1c. Figure 5.5 shows the full breakdown of absorption between
each layer in the stack across the spectrum for both the metallic and hydrogenated
states.
Having found this strong absorption eect for the structure with a tailored
index of refraction n? = 0.01 + 1i, we now explore in-depth how the optical properties
of this substrate layer aect the total absorption change in the metal lms and
how robust this eect is. Figure 5.6 shows the relationship of the complex index
of refraction of the substrate, n? = n + ik, with the change in absorption of the
metal lms. ?Absorption is dened as the absorption of the metal layers (Pd and
Mg) subtracted from the absorption in the metal hydrides after hydrogen exposure
(PdHx and MgH2). The rst trend that is evident with these plots is that the
102
Figure 5.5: Simulated absorption by layer in the device stack in the a) metallic state
and b) hydride state. In the metallic state, neither Mg (red) or Pd (blue) have very
high absorption, with Mg absorbing more from the mid-visible through the IR. In
the hydride state, both Mg and Pd begin absorbing signicantly more. Pd has a
high sharp absorption peak at ?450 nm with Mg having a peak at ?600 nm that
broadly tails o into the IR. The combination of the absorption in these two layers
allows for the high broadband absorption seen in the combined structure. In neither
the metal nor the hydride state does the NZI (green) substrate have any signicant
absorption.
absorption change increases with decreasing n for each wavelength, showing the
necessity of using an NZI substrate. A more interesting trend occurs with the
imaginary part of the index. Upon rst consideration, we would expect that the
absorption in the metal hydrides would be maximized with a maximum reectivity
at the MgH2/substrate interface. We nd through our simulations that this is not
the case, because the reectivity R is maximized when k = 0 for the substrate for
all investigated wavelengths (see Figure 5.7) and that the absorption is maximized
for k ? 1 for incident wavelengths ? = 800, 1200, and 1600 nm and k ? 2 for ? =
400 nm. This analysis shows that the absorption increase in the metals is not solely
from an increased path length from a perfect reection o the NZI substrate, but
from creating an additive destructive interference condition within the metals. We
can see that the value of the imaginary part of the index has the largest eect on
103
the absorption change when the real part of the index is < 0.25. If the real part of
the index is > 0.25, the absorption change becomes comparatively fairly constant
for longer visible and NIR wavelengths. The 400 nm result tells a dierent story,
where a higher k is necessary for high absorption change regardless of the value of
the real part of the index.
Figure 5.6: Eect of the substrate's optical properties, n? = n + ik, on device ab-
sorption change. Absorption change is dened as the absorption of the device in
the metallic state subtracted from the absorption in the hydrogenated state. ? rep-
resents dierent incident wavelengths of illumination, here showing a) ? = 400 nm,
b) ? = 800 nm, c) ? = 1200 nm, d) ? = 1600 nm.
5.2.1 Modeling substrate as Drude material
So far our consideration for this structure has used specically chosen refractive
indices, but a real material has to obey causality across the spectrum. To apply a
104
Figure 5.7: Fresnel reectance R versus the imaginary part of the index of refraction
of the NZI substrate kNZI at the MgH2/NZI interface for an illumination wavelength
of a) 400 nm, b) 800 nm, c) 1200 nm, and d) 1600 nm. The real part of the index
has been set to 0.01. We can see that the reectance is maximized at kNZI = 0. In
our simulations, we nd that the absorption change in the device is maximized at
kNZI ? 1, which combined with this nding shows that the increase in absorption
is from destructive interference eects and not from pure reection intensity.
model that obeys Kramers-Kronig consistency, we modeled our substrate as a Drude
material with properties:
? ? ?2
n? = (?) = ? ? p (5.1)??2 + i??
where n? is the complex index of refraction,  is the dielectric function, ? is the
permittivity as ? ??, ?p is the plasma frequency, and ? is the damping constant.
This equation models the free-electron response of a metal without any interband
105
transitions, but provides for a realistic, yet simple, model of an NZI material.
Figure 5.8 shows the results of modeling the properties of the substrate layer
with a Drude model. For these simulations, we have set ? to 1 (for eects of ?
on the absorption change, see Figures 5.9 and 5.10). We immediately see that
with this physical model, we can still achieve very high absorption changes, >75%
for some incident wavelengths. The absorption is maximized for smaller ?, although
this eect begins to plateau for ? < 1014 Hz. This eect is what we would expect, as
a higher damping term in the substrate causes an increased real part to the index of
refraction. The increased damping eliminates the cavity eect in the MgH2 exploited
by the device structure. The ? < 1014 Hz range is easily achievable, as many TCOs
have been found to have a damping coecient between 1013 and 1014 Hz [136].
The Drude simulations also show a very strong dependence on the plasma
wavelength ?p = 2?c/?p where c is the speed of light. The resonance eect occurs
for plasma wavelengths slightly lower than the illumination wavelength. This occurs
because, as we saw in Figure 5.6, the absorption eect is maximized for k ? 1. In
the Drude model, k increases for wavelengths below the plasma frequency, and the
eect is greatest when k is large enough to create the phase matching in the MgH2
and n remains low enough to harness the NZI eect. For 400 nm illumination, we
can achieve >70% absorption change for a Drude material with ?p ? 150 nm and ?
< 3*1014 Hz. A real material that could fulll these values would be Pt with ?p =
144 nm and ? = 1.088*1013 Hz [177]. For longer wavelengths in the spectrum, we
see the continued potential for strong eects, with regions > 75% absorption change
for ? = 800 nm and ? = 1200 nm. As we continue into the infrared, we nd that
106
Figure 5.8: Change in absorption upon hydrogenation using a Drude model for the
NZI material. ?p is the plasma wavelength of the Drude substrate (?p = 2?c/?p) and
? the damping term in the Drude model. ? represents dierent incident illumination
wavelengths, here with a) ? = 400 nm, b) ? = 800 nm, c) ? = 1200 nm, d) ? =
1600 nm. e) Absorption change vs illumination wavelength for 6 dierent ?p with
? = 1013 Hz, showing how broadband the switchable eect can be with a physical
model.
107
the high absorption change eect begins to taper o.
In Figure 5.8e, we nd that even when using a Drude model for the NZI
material, we still can get a broadband result. In these simulations, we dened ? =
1013 Hz. For ?p in the long-wavelength portion of the visible regime, we can see the
bandwidth of the response is the greatest. We nd that for ?p = 600 nm and ?p
= 800 nm, we can achieve a region of > 70% absorption change for over a 580 nm
range (650  1230 nm for ?p = 600 nm and 825  1405 nm for ?p = 800 nm). If we
lower our absorption change threshold to 60%, then our bandwidth widens to 835
nm. Both of these cases reach a peak absorption change of 78%, at 905 nm for ?p
= 600 nm and at 1145 nm for ?p = 800 nm. If we further increase ?p, we still see
a broadband result, but the magnitude of the peak of the absorption change begins
to decrease. Conversely, if we lower ?p, we can still achieve a high peak absorption
switchability, but begin to lose the large broadband eect as the resonance narrows.
5.2.2 Angular dependence of device
To test the robustness of this eect to non-normal incidence illumination,
we simulated the same structures depicted in Figure 5.1 for illumination angles
varying from 0? - 85?, with the angle ? being dened from the normal. Figure
5.11 shows the dependence of the absorption change of the metals on the incident
angle. We immediately notice that the NZI substrate remains the best choice for
maximum absorption change as the angle increases when compared to the Au and
SiO2 substrates. The only exception to this is at very high incident angles in the
108
Figure 5.9: Eect of ? parameter in Drude model on absorption change when
compared with the damping. Illumination wavelength ? is varied from 400  1600
nm from a)  f). The plasma wavelength ?p was set to 700 nm. For this ?p, we
do not see any singicant absorption change until ? = 800 nm, where this change
is maximized for ? = 1. As we increase the illumination wavlength beyond this
point, the maximum abosprtion change is achieved for increasingly high ?.
long-wavelength visible and the NIR regions, where for TM polarization the Au
begins to slightly outperform the NZI. In all other cases, the NZI remains strongly
superior.
With the exception of 400 nm illumination, we nd that the absorption change
stays fairly robust (constant) for TM polarization from a 0? incident angle all the way
to 60? (for 800 nm and 1200 nm illumination the absorption change remains above
70% in this entire region). For infrared illumination, we nd that the absorption
change for TM polarization actually slightly increases for larger angles, until 35?
for 1200 nm illumination, maxing out at ?Abs = 80%, and until 39? for 1600
nm illumination, maxing at ?Abs = 71%. For TE polarization, we nd that the
absorption change drops o quicker with increasing illumination angle, where ?Abs
109
Figure 5.10: Eect of ? parameter in Drude model on absorption change when
compared with the plasma wavelength. Illumination wavelength ? is varied from
400  1600 nm from a)  f). The damping term ? was set to 1013 Hz. For this ?, we
begin to see a large absorption change for ?= 600 nm for low plasma wavelengths. As
we increase the illumination wavelength, we get broader regions of high absorption
that extend from ? = 1 to ? = 5. The high absorption change region is broadest
for lower ? and blue shifts and narrows as ? increases.
drops below 60% at 58?, 60?, and 43? incident angles for 800 nm, 1200 nm, and 1600
nm, respectively. Again, the exception is when we look at 400 nm illumination where
we actually see a dramatic increase in the absorption change, with a peak of ?Abs
= 88% at 59?. The 400 nm behaves dierently than the other wavelengths shown
here because, for this wavelength, the absorption mainly occurs in the PdHx as
opposed to the MgH2. This dierence causes dierent conditions for the destructive
interference in the device, causing very dierent responses to dierent polarizations.
We also note that with these increasing angles, the pure absorption in the
hydride state can reach almost perfect absorption. Figure 5.12 shows the absorption
for the hydrogenated state for the device structure at dierent angles of illumination.
110
For TE illumination at 400 nm, the hydride absorption reaches 99.2% at 53?. For
TM illumination at 800 nm and 1200 nm, the hydride absorption peaks at 99.3% at
56?and 98.1% at 49?respectively. These perfect absorption eects begin to trail o
as we continue further into the IR.
Figure 5.11: Angular dependence of incident light on absorption change. a) Diagram
of the device. ? dened as angle incident light makes with the normal. b-e) Plots
showing the change in absorption of the structure with increasing ? for four dierent
incident wavelengths b) 400 nm, c) 800 nm, d) 1200 nm, e) 1600 nm. Blue lines
represent NZI substrates, green lines represent Au substrates, and red lines represent
SiO2 substrates. Solid (dashed) lines are TE (TM) illumination.
5.3 Experimental demonstration
To demonstrate this eect in practice, we used a lm of ITO deposited on a
soda-lime glass substrate as our NZI substrate (MSE Supplies). The optical prop-
erties of the ITO were measured with spectroscopic ellipsometry (J.A. Woollam
111
Figure 5.12: Angular dependence of absorption of device in the hydride state for
dierent illumination wavelengths a) 400 nm, b) 800 nm, c) 1200 nm, and d) 1600
nm. Blue lines represent NZI substrates, green lines represent Au substrates, and
red lines represent SiO2 substrates. Solid (dashed) lines are TE (TM) illumination.
M-2000) and t with a Drude-Lorentz oscillator model. Figure 5.13 shows the
experimentally measured ITO optical properties. The Pd/Mg metal stack was de-
posited using electron-beam evaporation with a vacuum pressure of < 3*10?6 torr
(Denton Vacuum). 25 nm of Mg was deposited followed by a 3 nm Pd cap with-
out breaking chamber vacuum. Immediately after deposition, we measured the
transmission and reection intensities of the device. Reection measurements were
112
taken at 20? incident angle and transmission measurements were taken with normal
illumination. Using a custom-designed environmental chamber, we then exposed
the device to 1 bar of H2 gas while dynamically measuring the transmission inten-
sity. Once the sample had completely loaded, determined by the stabilization of
the transmission intensity, we removed the sample from the transmission chamber
and measured the reection intensity. After this measurement, we remeasured the
transmission intensity. This intensity spectrum exactly matched that of the in-situ
measurement, showing that the sample had not had any appreciable unloading of
hydrogen during the brief time outside of the environmental chamber. We repeated
this process with the Pd/Mg deposited on a plain glass slide as a control sample.
Figure 5.13: Measured optical properties of ITO used in experiments. Optical
properties were determined using spectroscopic ellipsometry and t using a Drude-
Lorentz model.
The experimental absorption changes are shown in Figure 5.14 and are in good
agreement with simulations. Absorption changes are calculated by using A = 1 ?
113
Figure 5.14: Experimental demonstration of device performance. Solid lines are
experimental data, while dashed lines are simulation data. Blue lines represent the
device on an ITO substrate (Top Right), while red lines are on a plain glass substrate
(Bottom Right). The experimental data shows good agreement with that expected
from simulations. The ITO device reaches a max absorption change of 76% at 1335
nm illumination, while the plain glass device never has an absorption change higher
than 21%.
R?T , where A is the absorption, R the reection intensity, and T the transmission
intensity. In this formula, scattering is disregarded. This is a reasonable assumption
in our experiments because the RMS roughness of our samples is < 3 nm, thus there
will be negligible scattering o the surface. We compare this experimental data to
transfer matrix method simulations and see strong agreement for both the ITO
and glass substrates. The simulated data uses an incident angle of 20? to match
the angle the reection intensity data was taken. The experimental transmission
intensity measurements were taken with normal illumination, but the dierence in
simulated transmission intensity of this device between normal and 20? illumination
is < 2% over the reported wavelength range, so we can assume that this dierence
114
is negligible on the scale of the eect shown. Figure 5.15 shows this simulated
transmission dierence.
Figure 5.15: Simulated transmission intensity dierence between normal and 20?
illumination for the experimental device. Dierence is dened as simulated trans-
mission intensity of 20? incidence subtracted from normal incidence. At no point is
this dierence above 2% in the investigated wavelength range.
The ITO device resonance is maximized in the NIR at 1335 nm, near ITO's NZI
resonance at 1250 nm. The device resonance is red-shifted from the NZI resonance
as expected from the simulations because of the necessity of a higher k for the full
absorption eect. The maximum experimental absorption change was found to be
76%, with a maximum total absorption in the hydrogenated state of 93%. This
slightly outperforms the simulated maxima of 74% and 87% respectively. Reasons
for this discrepancy can be attributed to the slight dierences in optical properties
of thin lms when deposited on dierent substrates, or small amounts of scattering
that were included in the absorption. The eect is relatively broadband, with an
115
absorption change > 60% from 1140 nm to 1595 nm for a bandwidth of 455 nm for
this criteria. The time of loading of the sample, as determined from the dynamic
transmission intensity data, was found to be 27 min (see Figure 5.16). This
experimental demonstration shows the promise of this device architecture with a
relative ease of fabrication.
Figure 5.16: Dynamic transmission data for Mg/ITO device. Transmission intensity
for Pd/Mg/ITO device taken through the device loading. Hydrogen gas was intro-
duced to the chamber at 8 min. Transmission intensity levels o for all wavelength
at 35 min, for a total loading time of 27 min.
As expected, when the metal stack is deposited on a plain glass substrate, there
are no resonance eects, because the glass does not have an NZI region. There is
still an absorption change in the material due to the hydrogenation of the Pd/Mg
stack, but experimentally we nd that it is fairly small and constant across the
116
near-infrared at 21%.
Figure 5.17: Reversibility of the Mg/ITO device switchable absorption. Solid lines
are the device in the metallic state and dashed lines represent the device in the
hydride state. The blue lines are the data from the original load, while the red lines
indicated the second load. The data shows that the absorption change is reversible,
but with a slight decrease in the magnitude of the eect on the second cycle.
Lastly, we demonstrate the reversibility of this device. To unload the MgH2,
we place the device on a hot plate at 85?C in the ambient atmosphere for 30 min.
During this unloading, the device visibly turns back into a metallic state. After
the unload, we retake the optical measurements and repeat the loading procedure
outlined above. Figure 5.17 shows the results of this second cycle compared to the
original loading. We can see that the eect is mostly reversible on the second cycle,
but that the magnitude of the eect is slightly diminished. The metallic state after
the unload has slightly higher absorption than the pristine metallic state, which can
be attributed to dislocations formed in the Mg lattice upon the original loading.
117
The second loaded state still reaches a high total absorption of 91%, compared to
93% on the initial run, demonstrating that the process is reversible.
5.4 Conclusions
In summary, we have proposed and experimentally demonstrated a thin lm
stack of Pd/Mg can achieve switchable high absorption changes by exposure to H2
gas. By utilizing ITO's NZI resonance in the near-infrared, our device achieves a
maximum absorption change of 76%, with a maximum total absorption of 93% at
1335 nm illumination. This device has an absorbing change > 60% over a 455 nm
bandwidth. This is a signicant improvement when compared to a control sam-
ple consisting of a plain glass slide without the ITO layer, which only achieves an
absorption change of 21% over this spectral region. We show that this process is
reversible, with only a slight deprecation in total absorption change upon a second
cycle. Our simulations varying the optical properties of the substrate show the po-
tential of expanding this device to other wavelength regimes by using materials with
NZI resonances in those frequency bands. We showed that this eect can create a
> 70% change in absorption over a range of 650 -1230 nm incident wavelengths
when being modeled as a Drude material with a plasma resonance ?p = 600 nm and
damping ? < 1014 Hz, showing that this eect has the potential to be very broad-
band. This work shows the potential for switchable high absorption devices across
dierent wavelength regions that exploit the NZI behavior of the substrate without
aecting other aspects of the device's design. Future work will involve investigating
118
new substrates in these dierent spectral regions, as well as the potential of alloying
Mg with other metals to retain this high absorption change while speeding up the
dynamics and cyclability of the loading/unloading cycles.
119
Chapter 6: In situ Optical and Stress Characterization of Alloyed
PdxAu1?x Hydrides
In the next two chapters, we move beyond pure metal hydrides and explore
improving the optical and hydrogenation properties by alloying a metal hydride with
a secondary transition metal. The rst of these that we investigate in this chapter is
the PdxAu1?x system. PdxAu1?x alloys have recently shown great promise for next-
generation optical hydrogen sensors due to their increased chemical durability while
maintaining optical sensitivity to small amounts of H2 gas. However, the correlation
between chemical composition and the dynamic optical behavior upon hydrogena-
tion/dehydrogenation is currently not well understood. A complete understanding
of this relationship is necessary to optimize future sensors and nanophotonic de-
vices. Here, we quantify the dynamic optical, chemical, and mechanical properties
of thin-lm PdxAu1?x alloys as they are exposed to H2 by combining in situ ellip-
sometry with gravimetric and stress measurements. We demonstrate the dynamic
optical property dependence of the lm upon hydrogenation and directly correlate
it with the hydrogen content up to a maximum of 7 bar H2. With this measure-
ment, we nd that the thin lms exhibit their strongest optical sensitivity to H2 in
the near-infrared. We also discover higher hydrogen loading amounts as compared
120
to previous measurements for alloys with low atomic percent Pd. Specically, a
measurable optical and gravimetric hydrogen response in alloys as low as 34% Pd
is found, when previous works have suggested a disappearance of this response near
55% Pd. This result suggests that dierences in lm stress and microstructuring
play a crucial role in the sorption behavior. We directly measure the thin lm stress
and morphology upon hydrogenation and show that the alloys have a substantially
higher relative stress change than pure Pd, with the pure Pd data point falling 0.9
GPa below the expected trend line. Finally, we use the measured optical properties
to illustrate the applicability of these alloys as grating structures and as a planar
physical encryption scheme, where we show signicant and variable changes in re-
ectivity upon hydrogenation. These results lay the foundation for the composition
and design of next-generation hydrogen sensors and tunable photonic devices.
6.1 Background of Pd-Au alloys
The alloying of dierent metals to nely tune optical and material properties
has allowed for great advancements in a wide variety of applications from plas-
monic sensors to catalysts [22, 23, 74, 113, 178, 179, 180, 181]. Alloying creates
opportunities to improve the material characteristics beyond that of the pure metal
components. Furthermore, one can combine the desirable optical and structural
properties from dierent metals into a single alloy. This process is particularly ad-
vantageous when applied to metal hydrides where the optical properties, electric
properties, and response to H2 gas are all of interest.
121
Metal hydrides are useful for a wide range of applications including color dis-
plays, switchable mirrors and tunable plasmonics [14, 20, 21, 80, 115, 182, 183, 184].
Moreover, metal hydrides possess the capability to store and detect hydrogen. The
opportunity to use hydrogen for energy storage and distribution is becoming more
attractive, and with a push for a future hydrogen economy, more high-quality sen-
sors are needed to mitigate the dangers of H2 leaks. All-optical hydrogen sensors
are the preferred method of detection due to the decreased risk of ignition. This is
in contrast to more traditional electrical sensors, which have the potential to spark
upon a device malfunction. These optical sensors are required to have fast response
times, resistance to surface poisoning, limited intracycle hysteresis between hydro-
gen absorption and desorption cycles, and a large enough signal to be reliably read
[185].
Pd has been the standard metal investigated for hydrogen sensing and storage
because it is the only pure metal that can absorb and desorb hydrogen from its
lattice at room temperature without an activation layer [186]. However, pure Pd
suers from slow response times, surface poisoning, and a large hysteresis [166,
187]. Alloying Pd with other metals has been shown as a solution to mitigate
these problems, particularly alloying with Au [23, 74, 76, 188, 189, 190, 191]. In
addition to these improved hydrogen sensing properties, PdxAu1?x alloys have been
of particular interest in improving catalysis reactions [178, 180, 192, 193, 194] as
well as being used as a hydrogen separation membrane [195, 196]. A more complete
characterization is essential to facilitate further use of these alloys.
In this chapter, we quantify the changes in the optical, mechanical, and chemi-
122
cal properties of seven dierent PdxAu1?x thin lms as a function of hydrogenation.
These alloys are fabricated by physical vapor deposition co-sputtering at room tem-
perature, a versatile process that allows for a wider variety of substrates for sensors
that are not possible with fabrication methods that require high temperature an-
nealing steps. We simultaneously investigate the dynamic optical properties and
hydrogen loading amounts upon exposure to H2 gas. We directly measure the opti-
cal responses with spectroscopic ellipsometry with wavelengths spanning from 225
to 1690 nm, and we identify changes in the complex refractive index upon hydro-
genation. We use QCM measurements to determine the hydrogen sorption in the
material and nd higher loading quantities than previously reported for gas-phase
loading experiments. For each hydrogen exposure, we also simultaneously measure
the stress change and discover that the relative change in stress of the PdxAu1?x
alloys is 0.9 GPa higher than the change of the pure Pd. Upon investigating the
correlation between this stress and the change in the surface roughness of the mate-
rial, we nd that despite the large amount of thin-lm stress present in the alloys,
there are no observed roughness changes for any of the lms after loading. Finally,
we computationally demonstrate the applicability of these alloys for enhancing light
reection in grating structures as well as demonstrate a scheme for physical encryp-
tion. Our research elucidates important material properties in PdxAu1?x alloys that
will further inform hydrogen sensor design and implementation.
123
6.2 Experimental methods
6.2.1 Fabrication and characterization of PdxAu1?x thin lms
The thin lm PdxAu1?x alloys are fabricated by room temperature physical
vapor deposition cosputtering with Pd (99.95%) and Au (99.99%) sputtering targets.
A Si chip with a lithographically dened 1 x 1 cm2 area and two separate AT-cut 5
MHz QCMs were included as substrates for each deposition run. The lm geometry
on the QCMs was dened by a 12.5 mm diameter circular shadow mask centered
on the top QCM electrode. The substrates were cleaned with acetone, methanol,
and isopropyl alcohol rinses prior to deposition (the lithographic dening of the Si
piece was performed after the solvent clean). Prior to PdxAu1?x deposition, the
base pressure in the main chamber was maintained at less than 1.8 ? 10?8 Torr.
The thin-lm alloys were deposited at room temperature, with Ar gas introduced
and adjusted to a 10 mTorr pressure. During deposition, a constant rotation of
20 rpm, a z-height of 100 mm, and a gun tilt of 7.5 mm were applied to ensure
uniform chemical composition across each sample. Direct current powers ranging
from 75 to 300 W were applied to alter the PdxAu1?x composition. Two sets of
samples were produced for these experiments. The rst set was ?100 nm thick
and was deposited directly onto the substrates. The second set was ?400 nm thick
with a 10 nm Cr adhesion layer that was sputtered onto the substrates before the
alloy deposition without breaking vacuum. The Cr was deposited with a 150 W RF
source instead of a standard DC source because the only DC sources available in
124
the system were connected to the Au and Pd targets. The composition of each alloy
was measured with energy-dispersive X-ray spectroscopy (EDX). For each sample
on a Si substrate, EDX measurements were taken on four separate points to ensure
uniformity. The raw EDX data is shown in Figure 6.1. The rst set of depositions
had the compositional fractions (PdxAu1?x) of x = 0, 0.14, 0.34, 0.42, 0.52, 0.73, 1.
For the second batch (?400 nm thick with the Cr adhesion layer), we focused on
higher Pd content samples, having the compositions x = 0.41, 0.59, 0.73, 0.77, 0.83,
0.88, 1.
Figure 6.1: Measured EDX spectra of fabricated PdxAu1?x alloys. a) Spectra for
the 100 nm samples with x = 1, 0.73, 0.52, 0.42, 0.34, 0.14, 0 and b) for the 400
nm samples with x = 1, 0.88, 0.83, 0.77, 0.73, 0.59, 0.41. Known emission lines for
Pd and Au marked with vertical dashed lines.
Film thickness and roughness measurements were taken with an AFM in an
unpressurized dry air atmosphere. The roughness measurement was taken on three
125
distinct 2 x 2 ?m2 patches on each of the alloys. The reported roughnesses in this
article are all RMS roughnesses and are calculated by taking the RMS value of each
line of the AFM image and taking the median of these values. The lm thickness
measurements were taken on the lithographically dened Si samples. 20 x 20 ?m2
scans were taken centered on the dened edge of the metal square. The step results
in a bi-modal Gaussian distribution for the histogram of the topography data. The
sample height was taken as the dierence in the mean of the two Gaussians. ICP-
OES was used to conrm both the sample thicknesses and the chemical composition
for each of the alloys. The Si squares were dissolved in boiling aqua regia and
subsequently diluted to 50 mL. ICP-OES was then used to nd the Pd and Au
concentrations of this solution. Using the known surface area of the alloy on the Si
sample, the atomic masses, and the densities of these materials, the thickness and
atomic percentage of the alloys can be calculated. These calculated values agreed
within the error of the results of the EDX and AFM measurements.
6.2.2 Optical property measurements
Optical property measurements were taken with a spectroscopic ellipsometer
in a custom environmental chamber described in Chapter 2 and demonstrated in
Chapter 3. The QCM frequency is simultaneously recorded with the optical data
to correlate the hydrogen content of the sample with the optical property changes.
Measurements for each alloy are made at four dierent angles (48?, 55?, 70?, 75?) for
the pristine metal alloy before hydrogenation. The dynamic optical properties of
126
the alloys were then taken at 75? as the environmental chamber was switched from
7 bar Ar to 7 bar of H2. The 75?orientation is chosen for the dynamic measurements
due to steeper incident angles having a higher sensitivity to optical changes. The
measurements in the hydrogenated state were then retaken at all four angles at
the end of this process. The chamber was then purged with Ar with the dynamic
data being recorded again at 75?. This process was repeated for a total of four
hydrogen absorption steps and three desorption steps for each alloy. See Chapter
2 for further details of the ellipsometric measurement, window compensation of the
environmental chamber, and the dynamic optical tting method [164].
6.2.3 Hydrogen loading and stress measurements
The hydrogen loading and stress measurements were taken in a separate envi-
ronmental chamber equipped with an interferometer using the metal alloy lms on
the QCM as one of the mirrors. The loading value for each alloy is calculated with
the method outlined in Chapter 2 [95], which compensates the total QCM frequency
change with eects from stress along with environmental eects, such as changes in
gas composition.
Because the QCM substrates used in our experiments are anisotropic, the
standard way of determining thin-lm stress using the Stoney equation is not appli-
cable [197]. The stress values reported in this article use an adapted method from
EerNisse's double resonator model [46]. Instead of using two separate cuts of QCMs,
we measure the curvature of the sample and use the curvature to frequency relation
127
detailed in Chapter 2 [95] to nd the change in QCM frequency due to stress in the
sample. The change in stress of the thin lm upon hydrogenation is then
?fs?q
?? = (6.1)
KAT ?ff0
where ?fs is the change of QCM frequency due to stress, ?q and ?f are the
thicknesses of the quartz of the QCM and the metal lm respectively, f0 is the
measured QCM frequency before hydrogenation, and KAT is a constant dened by
EerNisse to be -2.75 x 10?12 cm2/dyn for a 5 MHz AT-cut QCM [46]. Using this
method, we assume that the stress is isotropic through the thickness of the lm.
All stresses reported in this article are compressive stresses and are dened to be
positive.
6.3 Dynamic optical property measurements of PdxAu1?x alloys
Figure 6.2 shows the optical properties for the seven dierent 100 nm alloys
investigated in this study as well as their property change upon hydrogenation in 7
bar H2 for the wavelength range of 225  1690 nm. Because metal hydrides typically
have signicant dislocation formation with initial loading, the data shown here are
for the second hydrogenation of the alloys. Thus, the plots describe a typical loading
cycle. We show the data of the initial loading of the pristine alloys Figure 6.3, and
show the dielectric functions for both the rst and second loads in Figure 6.4. To
verify our optical tting model for these materials, we compare our measurements to
the Johnson and Christy values reported for Au and Pd, which can be found in Fig-
128
ure 6.5, [104, 174] and found agreement. The non-hydrogenated optical properties
of PdxAu1?x alloys have previously been inferred with modeled electron energy loss
spectroscopy (EELS) data acquired on individual nanoparticles in the range of 248
- 827 nm [198] as well as with reection and transmission measurements for a single
Pd50Au50 thin lm alloy in the range of 350 - 1050 nm [199]. However, the optical
properties have not been directly measured with spectroscopic ellipsometry and have
not been measured deeply into the NIR region. Furthermore, the optical property
dependence of these alloys upon hydrogenation has not yet been investigated, which
we present here.
Figure 6.2: Measured optical properties of seven dierent PdxAu1?x alloys. (a)
Optical properties of each alloy without any hydrogen in the lattice (unloaded).
(b) Optical properties of the hydrogenated alloys under an atmosphere of 7 bar
of H2. (c) Change in the optical properties for each alloy upon hydrogenation.
Change is dened by the optical properties of the hydride subtracted from the
optical properties of the unloaded metal.
We observe a wide range of refractive indices (n? = n+ ik) for the alloys as we
adjust the chemical composition from pure Au to pure Pd. For the imaginary part
129
Figure 6.3: Measured optical properties of seven dierent PdxAu1?x alloys for the
rst load of the material. The initial loading of the metals generally causes for
singicant dislocation formation in the metal lattice, causing for these changes to
not describe a typical loadig cycle. (a) Optical properties of each alloy without
any hydrogen in the lattice (unloaded). (b) Optical properties of the hydrogenated
alloys under an atmosphere of 7 bar of H2. (c) Change in the optical properties for
each alloy upon hydrogenation. Change is dened by the optical properties of the
hydride subtracted from the optical properties of the unloaded metal.
of the index (k), we nd that that the ve alloys with the highest atomic percent
Pd have similar values and trends, with the pure Au and Pd0.14Au0.86 exhibiting
higher values. This is consistent with visual observation, where the ve highest Pd
content alloys all appear to have the same reective grey color, with the Pd0.14Au0.86
being the only alloy with a yellow hue approaching the appearance of the Au. An
interesting trend is observed in the real part of the index (n), where one may expect
the intermediate alloys to monotonically increase in n from the lower Au values
to the higher Pd values as the Pd composition is increased. Instead, both the
Pd0.52Au0.48 and the Pd0.73Au0.27 alloys have a higher n than that of pure Pd across
the entire wavelength range investigated. The nonlinearity of the optical properties
130
Figure 6.4: Change in dielectric functions upon hydrogenation. a-c) Dielectric func-
tions and dielectric function change for the initial hydrogen exposure. d-f) Dielectric
functions and dielectric function change for the second load of the alloys. Dielectric
function change dened as the dielectric function of the hydride subtracted from the
dielectric function of the metal.
with composition is not unique to the PdxAu1?x system, but is also present in the
properties of other noble metal alloys [181]. The addition of Pd to Pd-Au results
in a nonmonotonic and dramatic increase of the valence band at the ? point [178].
The additional electronic states below the Fermi level resulting from the break in
degeneracy of the band structure are likely responsible for extra interband transitions
131
Figure 6.5: Comparison of modeled Pd and Au optical data with literature. Johnson
and Christy data [104, 174] for pure Au and pure Pd plotted in dashed lines show
good agreement with our measured values of Au and Pd recorded in our experiment,
showing the validity of our model.
in these alloys and, ultimately, to the higher values of n observed.
Upon hydrogenation, pure Pd has the largest optical change in both n and k,
as expected. The magnitude of the change decreases as the Pd content in the alloys
is decreased. The x = 0.34 sample is the smallest atomic percent Pd in an alloy that
we observe a measurable optical change, showing a slight decrease in n, but very little
change in k. For both the Au and Pd0.14Au0.86 we observe no measurable response
when H2 is introduced to the system. For the pure Pd, the addition of H2 causes
a decrease in k across all wavelengths investigated, a decrease in n for wavelengths
above 565 nm and an increase below 565 nm, which agrees with literature [86, 164].
These trends are all observed in the next three highest Pd content alloys. The zero
change intercepts in the real part of the index occur between 550  580 nm for Pd
132
and the next three highest Pd composition alloys. The results also demonstrate
that for all alloys (and the pure Pd lm), the largest response to hydrogenation
occurs in the NIR region of the electromagnetic spectrum. This higher response at
longer wavelengths has been observed with other nanostructures and metal hydrides
[23, 164, 189]. By shifting to longer detection wavelengths, PdxAu1?x alloys with
lower x become more detectable because of their increase in responsivity. This
allows for the use of higher Au concentration alloys as sensors with their increase
in anti-hysteresis and anti-poisoning benets. For the full time-dynamic changes of
the optical properties, see Figure 6.6.
Figure 6.6: Dynamic optical properties of PdxAu1?x alloys upon hydrogenation.
a) Image depicting ellipsometry apparatus setup. b-h) Dynamic optical properties
of the seven 100 nm alloys upon exposure to H2. Line color gets lighter as time
increases (and partial pressure of H2 rises in the chamber. n plotted in solid lines
and k plotted in dashed lines.
In addition to the optical properties, the amount of hydrogen in the metal
lattice was simultaneously monitored. Figure 6.7 shows the dynamic behavior of
133
the refractive index versus hydrogen loading. We specify the amount of hydrogen by
H/M, which is the number of hydrogen atoms per metal atom in the lattice (including
both Pd and Au). Figure 6.7 excludes the pure Au and Pd0.14Au0.86 alloy because
they have no measurable reaction with H2 either optically or gravimetrically. All
other alloys share similar dynamic relationships. Each alloy has a monotonic change
in optical properties with respect to loading, with the largest changes occurring
in the near-infrared. For the relationship of n and k at 1500 nm with the Pd
composition of the alloy and the nal H/M ratio, see Figure 6.8. The change in
optical response is nearly linear with loading for each of the alloys, with the Pd lm
appearing to have an inection point around H/M ? 0.25. This behavior could be
attributed to the ? to ? phase transition in the Pd that does not occur in the lower
Pd content alloys [200].
Figure 6.7: Optical response versus hydrogen loading. (a-e) Relationship of the
optical properties (n and k) with the hydrogen content for each of the ve alloys
that have an interaction with H2. Note that the top time axis has nonlinear spacing
with each top tick corresponding to the time that the alloy reached the stated H/M
amount. Time = 0 min is dened as the time that hydrogen is rst introduced into
the system.
We next consider the repeatability of the optical and gravimetric responses
134
Figure 6.8: Relationship of n and k at 1500 nm illumination with (a) the atomic
percent Pd of the alloy and (b) with the nal hydrogen to metal atom ratio in the
lattice at 7 bar H2. The grey dashed line directly connects the points and is meant
as visual guide to the reader.
upon hydrogen cycling, which is essential for any practical device. Figure 6.9a-e de-
picts the changes in n and k for the ve 100 nm-thick alloys that react with H2 for
their second through fourth H2 exposures. As stated above, the rst hydrogenation
causes a slightly dierent optical response than the subsequent loads due to lat-
tice defects and dislocations as it rst expands [12]. After this initial disturbance,
the subsequent loads cause much less degradation of the lattice, allowing for more
consistent results. For subsequent exposures (after the rst), the response is very
consistent for Pd as well as the alloys. Figure 6.9f shows the maximum hydrogen
content at 7 bar H2, where the amount of hydrogen in the lattice is also very con-
sistent from load to load after the initial exposure. As expected, these results imply
that any device based on PdxAu1?x should still be pretreated with H2 before any at-
tempted operation. Overall, the alloys provide more consistent optical changes than
the pure Pd because less hydrogen enters the lattice, causing fewer dislocations.
135
Figure 6.9: Hydrogen cycling properties of alloys. (a-e) Comparison of the change
in optical properties for each of the ve alloys that interact with hydrogen. n and k
are plotted for the second through fourth loads with the rst load excluded. n and
k are dened by the optical properties of the current hydride state subtracted from
the values of the previously unloaded state. (f) Calculated hydrogen loading values
for the second through fourth loads. Dashed lines are linear ts for each alloy. Alloy
colors match those shown in a-e with the highest (lowest) H/M ratio corresponding
to Pd (Pd0.34Au0.66).
6.4 Material property measurements
To determine the exact amount of hydrogen absorbed into the metal lattice,
we used a separate QCM sample in an environmental chamber that accounts for
extraneous eects from stress and environmental (gas composition, pressure, tem-
136
perature, etc.) changes. Figure 6.10 shows the relationship of the hydrogen content
of the alloys at 7 bar H2 with the atomic percent Pd in the alloy. We can see
from these measurements that the relationship follows a linear t, as previously de-
scribed in the literature [188, 189]. We recorded the H/M ratio versus concentration
for both the 100 nm thin lms without any adhesion layer between the alloy and
the Au QCM substrate and the 400 nm lms with a 10 nm Cr layer between the
substrate and the alloy. Both of these sets of samples follow similar linear relations
within error bars. At 7 bar, our pure Pd samples had a hydrogen content of H/M =
0.72 ? 0.03 and H/M = 0.69 ? 0.03 for the 100 nm and 400 nm samples respectively,
which are in good agreement with previous values found in literature [11, 57, 58].
Compared to previous gas-phase measurements on thin lms, we have found higher
hydrogen loading values for the alloys, especially for the lower Pd content alloys
[188]. While previous measurements suggested no hydrogen reaction below x ? 0.55
in PdxAu1?x, we record measurable amounts of hydrogen entering the lattice down
to x = 0.34 with a H/M value of 0.02 ? 0.01. Reasons for this higher measured
loading could be contributed to dierent fabrication conditions of the alloys or dif-
ferent amounts of initial thin lm stresses due to the lm thickness, substrate, or
the circular geometry of the lm. These dierent intrinsic stresses can have a large
eect on the material properties of the thin lm system and should be characterized
for each new experimental procedure.
We also performed the experiments at 0.25 bar H2 at 1 bar of total pressure
to test the eects of driving potential on the total loading amount, shown in Figure
6.11. These values are slightly lower than those taken at 7 bar, as expected due to
137
Figure 6.10: Total sorption data. The grey circles correspond to the 100 nm
PdxAu1?x alloy data. The grey dashed line is a linear t of the ve alloys that
react with hydrogen. The solid orange squares correspond to 400 nm lms that
have a 10 nm Cr adhesion layer, and the orange dashed line refers to the linear
t. The measured values are compared to prior experiments from Bannenberg et
al.[188] on 40 nm thick thin lms at 10 bar H2 partial pressure (green triangles and
t with green dashed line). The black dashed line is to guide the eye to the zero
loading point.
the smaller driving potential of hydrogen in the lattice, but the same higher than
previously observed loading trend with a shallower slope when compared to atomic
Pd percent is still clearly present at these lower pressures. In the literature, gas-
phase loading experiments have also been performed on PdxAu1?x nanoparticles,
which exhibit a loading curve with a similar slope but lower overall values than the
thin lms [188, 189]. Electrochemical loading experiments have also been performed
on these alloys in the bulk with diering results, suggesting a range of potential
mechanisms may be at play for these dierent samples [201, 202].
With these unexpectedly higher loading values, we suspect that dierences
138
Figure 6.11: Comparison of total hydrogen content per metal atom at dierent
hydrogen partial pressures. The ratio at 7 bar (orange squares) is slightly higher
than that recorded at 0.25 bar (red diamonds). Even at lower pressures we have
measurable loading down to alloys with 40 atomic percent Pd.
in lm stress and microstructuring could be the cause. When thin metal lms
are hydrogenated, they undergo a compressive strain as the metal lattice expands
to incorporate the hydrogen [12, 203, 204, 205]. This strain can aect how much
hydrogen can be absorbed into the lattice and can have a signicant impact on the
sorption kinetics [57, 167, 206, 207, 208, 209]. Because of this eect, we investigated
the amount of stress change upon hydrogenation for each of our lms.
To determine the stress in our lms, we measure the change in the curvature
of the substrate (due to lm stress) and then convert this curvature to stress using
Equation 7.1 with inputs from the simultaneous QCM data. Figure 6.12a shows
the raw curvature changes of the 100 nm and 400 nm alloy samples. The 400 nm
samples clearly have a much higher curvature change than the 100 nm samples, as
139
we would expect due to there being a larger aected lm mass to distribute the
stress to the QCM. Once we convert this change in curvature into a change in stress
(Figure 6.12b), we nd that both the 100 nm and 400 nm samples have a similar
relation between stress and hydrogen content.
Figure 6.12: Stress characterization of PdxAu1?x alloys. (a) Raw data of curvature
change and (b) calculated stress values upon hydrogenation of each alloy versus the
calculated nal hydrogen per metal value under 7 bar H2. For each plot, the grey
dashed line is a linear t of the 100 nm data (grey circles) and the orange dashed
line is a linear t of the 400 nm data (orange squares), with each linear t excluding
the pure Pd data points. The vertical error bars for the curvature in (a) are smaller
than the marker size.
An interesting observation from this stress relation is that neither the 100 nm
nor the 400 nm pure Pd samples fall upon the trend of the alloys (the Pd data points
are excluded in the linear ts of each plot). The Pd values fall a full 0.9 GPa below
the expected trend line of the stress of the alloys. An explanation for this dierence
could be attributed to the alloys having higher initial stresses than the pure Pd due
to the intrinsic strain of alloying. By starting at a higher amount of initial stress,
140
the increase of stress per hydrogen atom is enhanced. This would cause the alloys
to have a dierent response to hydrogen due to the dierent strains in the structure.
This eect is an important factor in the design of future devices using PdxAu1?x
because unaccounted for stress could have deleterious eects on device performance.
It is important to fully characterize the stress system down to the exact substrate
and deposition temperatures, otherwise there may be dierent material responses
in the system, such as the higher than expected loading observed in this work. See
Figure 6.13 for the relationship of stress and Pd alloy composition.
Figure 6.13: Stress dependence on atomic Pd %. The orange squares are 400 nm
thick samples with a 10 nm Cr adhesion layer with the orange dashed line a linear
t of the data. The grey circles are 100 nm samples with no adhesion layer with
the grey dashed line a linear t of the data. In each plot, the pure Pd data points
are excluded from the linear ts. a) Raw curvature data for each of the alloys
upon hydrogenation. b) Calculated change in stress of the alloy thin lms upon
hydrogenation.
With these high lm stresses present in the materials upon hydrogenation, we
expect to see measurable changes in the surface morphology. We performed AFM
141
scans on the alloys immediately after fabrication (i.e. before H2 introduction), after
an initial H2 exposure, and again after a second exposure. We expect an initial
increase in the surface roughness of the lms after the initial hydrogenation due
to the hydrogen forming dislocations and vacancies in the lattice followed by lesser
changes in roughness on subsequent cycles. Instead, we found that the H2 exposures
had no measurable eect on the surface roughness of any of the lms, as seen in
Figure 6.14. We measured Pd and Pd0.73Au0.27 lms for the 100 nm samples without
a Cr adhesion layer and for 400 nm samples with a 10 nm Cr layer. Both sets of
samples were measured on the Au QCM electrode (initial electrode RMS roughness
was 1.8 nm). We also evaporated, in contrast to the sputtered samples above, a
100 nm Pd lm with a 5 nm Cr adhesion layer (required to prevent delamination)
to test if the roughness change was dependent on the fabrication conditions for the
thin lms.
Both the distribution and RMS values of the roughness remain very consistent
for all sample variations (sputtering vs. e-beam evaporation, 100 nm vs. 400 nm
lms, Cr adhesion layer vs. no adhesion layer). Figure 6.14a-e shows histograms of
the topography data for all three scans of each material. The results are so consis-
tent that it is dicult to distinguish between each histogram. Figure 6.14f-j shows
the measured RMS roughnesses and Figure 6.14k-o shows the pristine metal topog-
raphy image before the lm was exposed to H2. Figure 6.15 shows each separate
histogram and the corresponding topography scan. The consistency of these rough-
ness measurements is surprising due to pure Pd having a 12% volume increase upon
hydrogenation, and Pd0.73Au0.27 having an expected 9% volume increase [12]. Note
142
Figure 6.14: PdxAu1?x thin lm morphology characterization. (a-e) Histograms for
each material before hydrogenation (yellow), after the rst hydrogen load (red), and
after the second hydrogen load (blue). Note: histograms for rst and second loadings
are barely visible due to similarities with the values obtained before loading. (f-j)
Calculated RMS roughness values for each load. Dashed lines are a linear t to the
data showing no signicant change upon each loading cycle. (k-o) Representative
AFM topography scans of the alloys before any H2 has been introduced.
that these measurements are not able to detect any vertical thickness changes, as
the height set point is reset before each measurement. We do not see any dierence
in roughness change between the materials that are adhered directly to the QCM
versus with the Cr adhesion layer, showing that in this case, the strength of the ad-
hesion does not contribute to any roughness modication in this setup. A complete
list of the measured RMS roughnesses for each of the other alloys can be found in
Table 6.1. The data also suggest that the surface roughness of the lm has no eect
on the nal loading amount in lms of these thicknesses, as is expected. The 400 nm
sputtered samples have signicantly higher roughnesses than the others, yet all three
Pd samples have the same total loading numbers within error bars. Similarly, the
143
Figure 6.15: Roughness scans upon hydrogenation for ve PdxAu1?x samples pre-
pared under dierent conditions. a-o) Topography images before H2 exposure, after
one H2 exposure, and after a second H2 exposure. a2-o2) Corresponding histograms
for each topography image.
two Pd0.73Au0.27 samples also load to similar values despite diering roughnesses.
6.5 Simulations of optical switching
To apply these measured properties to an applicable optical switching system,
we simulate (using FDTD) several grating structures to demonstrate the opportu-
nity of using these alloys as devices. The measured optical properties for both the
metals and hydrides presented in Figure 6.2 are used as inputs to the simulations.
Figure 6.16 shows the results for two separate grating structures. Both structures
consist of a grating with 100 nm width and height and a 65 nm SiO2 spacer layer
on a Au substrate. This spacer layer creates a cavity eect from the reection o
144
Composition Deposition t Pristine 1st Load 2nd Load
type (nm) (nm) (nm) (nm)
Pd Sputtered 100 2.6 2.8 2.6
Pd0.73Au0.27 Sputtered 100 2.5 2.5 2.4
Pd0.52Au0.48 Sputtered 100 2.5 2.5 2.4
Pd0.42Au0.58 Sputtered 100 - - -
Pd0.34Au0.66 Sputtered 100 2.7 2.3 2.4
Pd0.14Au0.86 Sputtered 100 2.7 2.5 2.7
Au Sputtered 100 2.4 2.2 -
Pd Sputtered 400 12.4 12.5 12.5
Pd0.88Au0.12 Sputtered 400 8.1 8.3 -
Pd0.83Au0.17 Sputtered 400 7.6 7.8 -
Pd0.77Au0.23 Sputtered 400 9.2 9.6 -
Pd0.73Au0.27 Sputtered 400 8.3 8.4 8.3
Pd0.59Au0.41 Sputtered 400 6.8 6.7 -
Pd0.41Au0.59 Sputtered 400 6.3 6.8 -
Pd Evaporated 100 2.3 2.5 2.4
Table 6.1: Compiled RMS roughnesses of PdxAu1?x upon hydrogenation. Table
entries with the symbol `-` correspond to conditions where no data were not taken.
t is thin lm layer thickness. Pristine is the RMS roughness before H2 exposure.
1st and 2nd loads are the RMS roughnesses after the 1st and 2nd full H2 exposures
respectively.
the Au substrate and enables increased sensitivity [115]. We simulated two dierent
periodicities to show the tunability of the structure. First, a 550 nm periodicity
is used to obtain large spikes in the relative reectance, where the Pd structure
shows an almost 200-fold increase. As we increase the amount of Au in the alloy,
we see the peak of the reectance ratio blue shift and decrease in magnitude, with
the Pd0.52Au0.48 alloys still having over a 2-fold relative increase in reectance. By
changing the periodicity of the structure to 300 nm, we can achieve the opposite
response to hydrogenation, where the relative reectance dips to well below 1 for
the same materials and exposures. This demonstrates the utility of these materials
for a wide range of potential photonic device designs. We note that while Pd has a
larger optical response than the alloys, the Pd0.73Au0.27 and Pd0.52Au0.48 both have
145
Figure 6.16: Simulated reectance shifts upon hydrogenation. (a) Illustration of
simulated grating structure. Gratings are 100 nm high and 100 nm wide with a
spacing d between gratings. A 65 nm SiO2 spacer layer separates the grating from
a Au substrate. Light is incident normally on the grating with its polarization
orthogonal to the grating direction. Reection spectra for each of the alloys with
grating periodicity of (b-d) 550 nm and (e-g) 300 nm plotted with the quotient of the
metal (solid line) and hydride (dashed line) reections showing the relative change
in reection with hydrogenation (dot dashed line).
signicant optical responses in these structures. The tunable responses combined
with the chemical resistance of the alloys will allow for improved sensor design. To
146
have an even higher sensitivity as an optical hydrogen sensor, the grating can be
tuned for the resonance to occur in the NIR instead of the visible, because the largest
optical changes upon hydrogenation occur at longer wavelengths. Figure 6.17 shows
simulations with resonances in the NIR.
Figure 6.17: NIR simulations of the reectivity of grating structures. (a) Schematic
of grating structure. Gratings are 100 nm high and 250 nm wide with a 750 nm
spacing between gratings. A 65 nm SiO2 spacer layer separates the grating from an
Au substrate. Light is incident normally on the grating with the polarization of the
light orthogonal to the grating direction. (b-d) Reection spectra for each of the
alloys plotted with the quotient of the metal (solid line) and hydride (dashed line)
reections showing the relative change in reection with hydrogenation (dot dashed
line).
Finally, we show that these materials can be used for physical encryption and
readout based on the gaseous H2 in the environment. Previously, Mg nanostructures
have been used for similar types of encryption schemes due to Mg's large optical
change to H2 [80, 210, 211]. In these works, e-beam lithography was used to create
147
small and intricate encryption images. In our example, we work with thin-lm
eects to make a macroscopic encryption scheme without the need for submicron
nanostructuring. We do this by taking advantage of the similarities of the optical
properties of Pd with Pd0.42Au0.58 in the non-hydrogenated state. By using the high
optical sensitivity of Fabry-Perot-like resonances, we can amplify the dierence in
the optical responses of the two metals upon hydrogenation. Figure 6.18 outlines
the encryption scheme of this setup. For the background of the image, there is an
optically thick Pd0.42Au0.58 substrate with a 94 nm SiO2 intermediate layer and a 5
nm Pd0.42Au0.58 capping layer. The lettering (encrypted message) is generated by
using the same thickness of layers as the background structure but with Pd instead of
the Pd0.42Au0.58 alloy for both the substrate and the cap. The multilayer resonance
causes the coloring of the structure. Figure 6.18b shows the true-color image of
the structure containing both the lettering and background before hydrogenation.
These colors are found by calculating the reection spectra with TMM simulations
with the optical properties presented in Figure 6.2 as inputs. We then use the CIE
XYZ color space to convert the reection spectra to a color. See Appendix E for
more details on the CIE color space. Note that all regions have indistinguishable
colorings, hiding the message under ambient atmosphere. Once the structure is
exposed to H2, the Pd lettering has a large optical change as it hydrides, causing
a shift in the resonance corresponding to an observed color shift. The Pd0.42Au0.58
background on the other hand has a limited response to the H2 gas, causing a
minimal color change of the substrate, as seen in Figure 6.18c. This allows for the
lettering to become visible, revealing the message. Because these materials are fully
148
reversible as well, all one needs to do to re-encrypt the message is to remove the
structure from the H2 atmosphere. This allows for a full encryption scheme while
only needing a small amount of H2 gas to reveal the message with no other optical
setup required. One note is that both sides of these structures would have to be
exposed to hydrogen in order to fully hydride the Pd and Pd0.42Au0.58 layers. While
the optical contrast between the two regions is relatively small, the planar nature
of the structure (without the need for nanofabrication) is advantageous.
Figure 6.18: Physical encryption scheme. a) Design for using PdxAu1?x alloys
in a physical encryption scheme. The H2" message is created using a multilayer
stack containing Pd while the background is created from a similar stack containing
Pd0.42Au0.58. b) Simulated coloring of the design before exposure to H2 and c)
coloring of design after H2 exposure. The SiO2 thickness is t = 94 nm for these
simulations with a 5 nm metal cap for both the active area and the background.
d) Chromaticity diagram showing the change in coloring of the lettering versus the
background.
149
6.6 Conclusions
In conclusion, we have directly measured the optical, sorption, and stress prop-
erties of a series of PdxAu1?x alloys with controlled chemical composition upon
hydrogenation. We implemented in situ spectroscopic ellipsometry from the mid-
ultraviolet to the near-infrared regions and showed the consistency of these proper-
ties through multiple H2 exposures after an initial hydrogenation cycle. We found
that our alloys exhibited a linear relation between Pd composition and hydrogen
loading amount, in agreement with previous reports. However, we measured a much
smaller slope with respect to Pd composition than previously observed with higher
loadings obtained in alloys with lower atomic percent Pd. We postulate that this
dierence is caused by dierent thin-lm stresses present in our system. We charac-
terized these stresses and found that the PdxAu1?x alloys have a 0.9 GPa higher rel-
ative stress than measured for pure Pd which must be accounted for in device design.
Surprisingly, even with these high stress amounts, there was no observed roughness
(morphology) change in the alloys when hydrogenated. Using our measured optical
properties, we showed the applicability of these alloys as grating devices, where the
reectivity can either be increased or decreased upon hydrogenation depending on
the periodicity of the grating. Furthermore, we demonstrate their potential for use
in a physical encryption scheme with no need for sub-micron nanostructuring.
Our results will further inform future sensor design, which must consider dier-
ences in fabrication methodology and the subsequent inherent stresses in the devices
in order to accurately describe the sorption/optical response. We have shown that
150
dierent fabrication conditions and substrate choice can alter the stress of the sys-
tem, which can be used to increase the response of new devices and allow for lower
Pd composition alloys than previously expected to be utilized in the proper sensor
design. We have also shown that moving to infrared sensing schemes could increase
the sensitivity of these alloyed sensors when compared to the visible measurements.
Future work will entail investigating new combinations of metals for potential im-
provements in sensing ability while maintaining chemical durability, potentially by
moving from binary to ternary alloys.
151
Chapter 7: Optical Tunability Characterization of Mg-Ni, Mg-Ti, and
Mg-Al Alloy Hydrides
In this chapter, we report the dynamically measured optical, loading, and
stress properties of dierent compositions of three Mg alloy systems: Mg-Al, Mg-Ti,
and Mg-Ni. By alloying Mg with secondary elements, it has been shown to increase
the hydrogenation kinetics and improve the thermodynamics of the system. We nd
that these materials all have large optical changes when exposed to H2 gas, with a
wide range of potential properties in the hydride state. The magnitude and the sign
of the optical properties change for each of the alloys is similar, but the dierences
have dramatic eects on device design. We nd that Mg-Ti alloys in particular have
applications for both switchable windows and broadband switchable light absorbers
through TMM simulations.
7.1 Introduction to Mg alloys
As has been discussed previously in this dissertation in Chapter 3, Mg ab-
sorbs a large amount of hydrogen upon H2 exposure and goes through dramatic
changes in its optical properties. One issue with pure Mg is that the hydride state
is too thermodynamically stable and must be unloaded at a high temperature [212].
152
The kinetics of both the absorption and the desorption are also slow, especially in
bulk Mg, due MgH2 being a poor proton conductor, which limits the diusion of
H through the material [213]. In order to improve upon the kinetic and thermody-
namic properties of Mg, alloying has been extensively used to destabilize the hy-
dride phase and increase H diusion. This alloying has been attempted with many
metals including Co [91, 214, 215, 216], Fe [91, 214, 217, 218], and Mn [91, 219]
among others, but the most common secondary alloying elements used in the lit-
erature are Al [220, 221, 222, 223, 224], Ti [225, 226, 227, 228, 229, 230], and Ni
[18, 183, 231, 232, 233, 234, 235, 236, 237, 238, 239]. These alloys have increased
kinetics when compared to Mg [220, 240, 241, 242], and were each found to have
interesting optical responses that are worth investigating further.
The Mg-Ni system in particular has been of interest in the optical community
due to Mg2Ni reportedly forming an intermediate optical "black state" during its
loading process [18, 235, 236]. This black state is characterized by an intermediate
loading state where the sample absorbs >75% of incident visible light when illu-
minated through a transparent substrate. The process has been investigated with
15N nuclear reaction analysis hydrogen depth proling [18] to determine the H atom
vertical distribution in the sample. It was found that the black state formation is
caused by preferential loading from the Mg2Ni/substrate interface, as opposed to
the hydride being seeded near the dissociation sites at the Mg2Ni/Pd interface. A
multi-layer interference eect occurs, where the bottom layer is fully hydrogenated
Mg2Ni and the top layer remains metallic Mg2Ni, and this orientation creates the
observed black state. In the fully hydrogenated state, some Mg-Ni alloys have also
153
been found to have good switchable window properties, switching from a reecting
state to a transparent state [183].
The Mg-Ti system has been investigated for switchable solar absorbers [228],
switchable mirrors [243], and hydrogen sensor applications [24]. These materials
have the advantage of having a reported high-absorbing state across the visible
region for thick lms (> 200 nm) in their fully hydrogenated state, as opposed to
Mg2Ni which obtains this high-absorbing state only during intermediate loading
[229]. This high-absorbing state occurs for y = 0.7 - 0.8 in MgyTi1?y. This alloying
system is also of interest due to a decrease in degradation over many hydrogenation
cycles, along with faster kinetics during these cycles, when compared to pure Mg
[229]. For very thin Mg-Ti lms (< 40 nm), the hydride state is more color-neutral
than Mg-Ni alloys (which have a yellow tint), allowing for more aesthetically pleasing
switchable windows, albeit with fairly low transmission amounts ?20% [243].
Mg-Al alloys are of interest for their potential for high hydrogen weight per-
cent (weight of hydrogen in a material divided by the total weight of the material)
for hydrogen storage along with much faster hydrogenation kinetics at room tem-
perature when compared to Mg [224]. These materials also have been suggested as
switchable window devices, as Mg-Al alloys with ? 70% Mg have been demonstrated
to have color-neutral transmission [220].
So far, all optical measurements reported in the literature of these alloys con-
sist of normal incidence reection and transmission measurements to determine the
absorption coecient of Mg-Ti alloys [229] and the absorption coecient and op-
tical bandgap of Mg-Ni alloys [234, 244] or tting normal incidence reection and
154
transmission measurements with a Drude-Lorentz model to determine the dielectric
function of Mg-Ni alloys [18, 245]. As previously discussed in this thesis, this method
of obtaining thin lm optical properties is much less sensitive than variable-angle
spectroscopic ellipsometry, especially for very thin lms < 50 nm [246]. There have
been no reports in the literature of the optical properties of Mg-Al alloys, or of any
Mg-Ti, Mg-Ni, or Mg-Al alloys in intermediate loading states. Our system allows
for in situ ellipsometry to dynamically investigate the optical properties of thin lms
(< 50 nm) of these materials with high sensitivity as they are hydrogenated. By
using similar deposition parameters to fabricate the dierent alloy systems, these
measurements will also allow us to quantitatively compare the responses of these
systems.
In this chapter, we dynamically measure the optical properties of dierent
atomic ratios of thin-lm Mg-Al, Mg-Ti, and Mg-Ni alloys as they are exposed to
H2 gas. We also dynamically record the loading and stress values of these alloys
during the loading. We nd large optical changes for all of the alloys investigated,
with Mg0.85Ti0.15 exhibiting the largest optical changes for any alloy. We conrm
the optical black state in the Mg-Ni samples through TMM simulations using our
found optical properties and observe the highest absorbing state for Mg0.73Ni0.27, as
expected with that atomic composition being the closest to Mg2Ni. We nd that
Mg-Ti alloys exhibit the best properties for both switchable windows and broad-
band switchable light absorbers. We also nd the loadings for the Mg-Ni alloys to
be slightly lower than expected from the literature and hypothesize that dierent
deposition parameters could aect the properties of the materials to cause dierent
155
loading amounts.
7.2 Experimental methods
Each of the thin-lm samples is fabricated by room temperature physical vapor
deposition cosputtering. For each deposition, two separate AT-cut 5 MHz QCMs, a
glass slide, and a Si chip are included in the deposition chamber. Prior to deposition,
each sample is cleaned with acetone, methanol, isopropanol, and water. The alloys
were deposited through a 12.5 mm diameter shadow mask centered on the QCM
substrate. The direct current powers of the sputtering tool ranged from 50 to 450
W to attain the dierent compositions of the dierent alloys. These minimum and
maximum values are determined by the minimum voltage necessary to maintain a
plasma and the voltage limit of the tool respectively. Each sample was capped with
a 3 nm Pd layer without breaking vacuum to catalyze the hydrogenation reaction
and prevent surface oxidation of the sample. The composition of each alloy was
determined with EDX taken on the Si chip included in the deposition chamber,
taking an average of 5 measurements at dierent points on the sample to ensure
uniform alloying.
The optical properties of the materials were determined with in situ spec-
troscopic ellipsometry using the system outlined in Chapter 2. To determine the
thickness of the Mg alloy lms, the glass slide sample that was included in the de-
position chamber was measured with ellipsometry and transmission measurements.
The raw ? and ? data were then t with an optical model with the properties of
156
the Mg alloy and the thickness of that alloy as t parameters. The optical proper-
ties of the glass substrate were taken before the metal deposition and were dened
in the model. The properties of the sputtered Pd were measured separately and
were dened in the model. The thickness of the Pd layer was set to be 3 nm for
all samples. With the transmission measurements, this tting procedure allows for
unambiguous determination of the thickness of the Mg lms. The thicknesses of
the lms we investigated ranged from 19 to 42 nm. As a consistency check for our
thickness measurements, we used the same model properties and thickness on the
data taken on the Au QCM electrode substrate and found good agreement without
any retting.
The dynamic optical ttings for these materials were done in a slightly dif-
ferent method than previous chapters. The volume expansion of these Mg-based
alloys becomes important in these thin-lm ttings. In previous measurements on
metal hydrides, we have measured materials that are optically thick within the mea-
surement regime (transmission across all measured wavelengths is zero). For these
measurements, thickness expansion does not aect the tting results as the materials
are modeled as bulk. However, the Mg alloy thin lms investigated in this chapter
have appreciable transmission and the optical eect of the substrate must be taken
into account. For these alloys, the volume expansion for each is found to be ?15%
over the atomic ratio region that we are investigating in this chapter [18, 247], and
we use this value for all of our samples. To include this expansion in the model,
we use the dynamic loading data to determine how much H is in the lm at each
optical time step and then scale the total thickness expansion by the same ratio as
157
the current loading to the nal loading value (i.e if a lm's nal calculated loading
value is 1 and at time step t we calculate the loading is H/M = 0.33, then we dene
the volume expansion at this time step to be 0.33*15% = 5%). We also dene the
loading amount of the Pd capping layer by this same ratio. At each time step, with
the thicknesses and Pd cap properties dened, we then use a B-spline model with
0.3 eV node spacing to t the optical properties of the Mg-alloy.
The hydrogen loading for each alloy was measured on the second QCM sample
with the method outlined in Chapter 2. To convert from the QCM frequency to
H/M loading ratio, the density of each sample must be input into the calculation.
For Mg-Ti, both Mg and Ti have hexagonal close-packed lattice structures and no
known intermetallic states form. Because of this, we believe that a linear weighting
of the densities of Mg and Ti in the same ratio as their atomic percent in the alloy
is a reasonable approximation of the density. As a check of this assumption, we
compare densities calculated from linear weighting with densities calculated from
lattice constant measurements using X-ray diraction measurements and nd good
agreement < 6% dierence. For the Mg-Al system, we also chose to use this linear
weighting scheme because there are no other alloy phases found for this material
when the atomic Mg percent > 60%. For the Mg-Ni system, we use a slightly
dierent method. Mg2Ni is a known intermetallic with a density of 3.48 g/cm
3 and
because this forms a separate phase within the Mg-Ni alloys, a linear combination of
pure metal densities is not applicable. From Ref [244], we know for MgyNi1?y if 0.67
< y < 0.89, the alloy forms a varying mixture of crystalline Mg2Ni and amorphous
Mg0.89Ni0.11 but the lattice constant remains constant in this region. If 0.89 < y <
158
0.95 the lattice constant begins to expand and the alloy is mainly nanocrystalline
Mg2Ni and crystalline Mg. Using this knowledge of the phases, we use the following
equation to calculate the densities of the Mg-Ni alloys for 0.67 < y < 0.89:
AMgy + ANi(1? y)
? = ?Mg2Ni (7.1)0.67 ? AMg + 0.33 ? ANi
where ?Mg2Ni is the density of Mg2Ni, AMg is the atomic mass of Mg, and ANi is
the atomic mass of Ni. Here we calculate the density by multiplying the mass ratio
of the alloy to Mg2Ni with the known density of Mg2Ni. This is using the fact that
the volume of the lattice is not changing from the Mg2Ni size, thus allowing us to
only account for mass. For higher Mg percent with 0.89 < y < 0.95:
( )
1 AMgy + ANi(1?
y?0.89
y) 1?y
? = y?0.89 ?Mg2Ni +1 + 0.67 ? A + 0.33 ? A 1 + y?0.89
?Mg (7.2)
1?y Mg Ni 1?y
where ?Mg is the density of pure Mg. Here we linearly weight the density of
Mg0.89Ni0.11 (calculated with Eq. 7.1) with the density of Mg. This rationale
uses the fact that up to 89% Mg, there was no expansion in the lattice and with
additional Mg added above this 89% point, we are adding crystalline Mg with the
density ?Mg. Using these calculated densities and the thicknesses found from the
optical ttings, we can then calculate the loadings and stresses in the same manner
as Chapter 2.
159
7.3 Optical properties of Mg alloy hydrides
In the following sections, we report and discuss the dynamic optical properties
of our fabricated Mg-Al, Mg-Ti, and Mg-Ni alloys as they are exposed to 1 atm H2
gas.
7.3.1 Mg-Al hydrides
Figure 7.1 shows the measured optical properties n? = n+ ik of four dierent
compositions of Mg-Al alloys and how these properties change under complete hy-
drogenation under 1 bar H2 pressure. For these materials, we observe a small spread
in the initial and nal properties of the materials, which is as expected due to all
of the alloys being fairly close together in atomic composition. The reason for this
proximity in composition is a relative lack of sensitivity of the dierent materials to
dierent deposition voltages (i.e. 200 W Mg and 200 W Al powers compared to 450
W Mg and 50 W Al powers only give a 7% dierence in atomic Mg percent). We
nd that both the real and imaginary parts of the index of refraction increase with
longer wavelengths, as we expect from most lossy metals. Interestingly, we nd that
metallic Mg0.77Ni0.23 exhibits the largest n and k across the spectrum, as well as the
largest optical change in k. We nd that for all of the Mg-Al alloys, the change in
the optical properties follow the same trend. For ?n, there is little variation from
sample to sample. Each sample exhibits an increase in n for shorter wavelengths
and a decrease for longer wavelengths, with the crossover point from positive to
negative n occurring between 1100 - 1225 nm. All samples exhibit a decrease in k
160
across the measured spectrum, with the largest decreases occurring in the NIR and
the smallest decreases in the visible. Despite having the largest ?k, fully loaded
Mg0.77Ni0.23Hx still has the largest k across the visible spectrum, although only by a
small margin, and the second highest in the NIR. In the nal hydride state, shown
in Figure 7.1c, we nd that each of the materials still has signicant attenuation
in the long-wavelength visible and into the NIR, with k > 3 after ?1000 nm for
each alloy. This is despite the fact that the NIR optical properties saw the largest
decreases in k. For the real part of the index of refraction, we nd a minimum in the
mid-visible spectrum, that increases to large values (> 3) in the NIR. The hydride
samples are still somewhat optically metallic, not exhibiting a complete transition
to a dielectric material.
Figure 7.1: Optical properties n? = n+ ik of four dierent Mg-Al alloy hydrides. a)
Optical properties in the pristine metallic state before hydrogenation. b) Change
in optical properties upon hydrogenation, dened here as the pure metal optical
properties subtracted from the hydride optical properties. c) Optical properties
in the fully hydrogenated hydride state. Each colored line on the plot represents
a dierent atomic ratio of metal hydride, with darker shades representing higher
atomic Mg percent ratios.
161
In Figure 7.2, we observe the intermediate states during the hydrogen loading.
These plots show a smooth transition of the optical properties from the metallic
(light colored curves) to the hydride states (darker colored curves). Note that the
chosen optical curves on this plot are not linearly spaced with time, but are instead
to show the range of properties of the intermediate states. Loading generally begins
slowly as small amounts of H2 are introduced to the chamber, then increases quickly
during the beginning of the ? to ? phase transition, then slows down to a long tail
for the nal ?10% of the load when most of the material is in the ? phase. Most of
our materials showed similar time dynamics, with the total time of loading ranging
from 10 - 15 min for most samples.
Figure 7.2: Optical properties of dierent Mg-Al alloys as they are exposed to H2
gas. The lightest colored line depicts the alloy in the pristine metallic state. As H2 is
introduced to the system, the material begins to hydrogenate, denoted by the lines
getting darker in the plot, with the darkest line indicating the full hydride state.
Each line is not linearly spaced in time and is instead chosen to aesthetically show
the range of possible intermediate states. The alloys shown here are a) Mg0.77Al0.23,
b) Mg0.81Al0.19, c) Mg0.83Al0.17, d) Mg0.84Al0.16
162
Figure 7.3: Optical properties n? = n + ik of ve dierent Mg-Ti alloy hydrides. a)
Optical properties in the pristine metallic state before hydrogenation. b) Change
in optical properties upon hydrogenation, dened here as the pure metal optical
properties subtracted from the hydride optical properties. c) Optical properties
in the fully hydrogenated hydride state. Each colored line on the plot represents
a dierent atomic ratio of metal hydride, with darker shades representing higher
atomic Mg percent ratios.
7.3.2 Mg-Ti hydrides
Figure 7.3a shows the metallic optical properties of 5 dierent fabricated
Mg-Ti alloys with the atomic Mg percent ranging from 82 - 91%. We nd that
the n values in the metallic state are fairly close together, with the exception of
Mg0.89Ti0.11, which exhibits a lower n for the state. The higher Mg percent alloys
exhibit higher k values, again with the exception of Mg0.89Ti0.11 (we will discuss
the discrepancies of the Mg0.89Ti0.11 sample further down in the chapter). Higher
attenuation for samples with more Mg is not unexpected, as Mg has much higher
k values than Ti in their unalloyed form. The metallic properties of these Mg-Ti
samples also align fairly closely with the metallic Mg-Al alloys investigated in the
163
previous section. As we hydrogenate these samples, we nd similar types of changes
in the optical properties when compared to the Mg-Al samples, with n increasing in
the visible and decreasing the in the NIR, and with k decreasing for all wavelengths
with larger decreases for longer wavelengths. There is a much broader range in the
changes in properties for these alloys, and two alloys, Mg0.85Ti0.15 and Mg0.87Ti0.13,
exhibit larger changes than the Mg-Al samples. Figure 7.3c shows the properties of
the hydride states. The three lowest Mg percent samples show somewhat constant
n in the visible spectrum, with peaks in the ultraviolet and then increasing with
longer wavelengths into the NIR. The two highest Mg percent samples have smaller
n values in the visible, with sharper minima that then monotonically increase with
longer wavelengths, similar to the Mg-Al hydrides. These two samples also have
the largest k by a signicant margin. The lower Mg percent samples have relatively
smaller k values, with Mg0.85Ti0.15 exhibiting almost no attenuation across the visible
and NIR.
In Figure 7.4, we observe the intermediate states during the hydrogen loading.
These plots mostly show smooth optical transitions, except for Mg0.85Ti0.15 and
Mg0.87Ti0.13, which show a resonance like dip in n between 500-600 nm for high
hydrogen content states. The Mg0.82Ti0.18 sample also slightly shows this eect. For
the Mg0.85Ti0.15 sample, we also observe a decrease in n in the NIR until the almost
fully hydrogenated state, and then a sharp increase in n when the hydrogenation
is complete. Note that again the chosen optical curves on this plot are not linearly
spaced with time but are instead shown to depict the range of properties of the
intermediate states.
164
Figure 7.4: Optical properties of dierent Mg-Ti alloys as they are exposed to H2
gas. The lightest colored line depicts the alloy in the pristine metallic state. As H2 is
introduced to the system, the material begins to hydrogenate, denoted by the lines
getting darker in the plot, with the darkest line indicating the full hydride state.
Each line is not linearly spaced in time and is instead chosen to aesthetically show
the range of possible intermediate states. The alloys shown here are a) Mg0.82Ti0.18,
b) Mg0.85Ti0.15, c) Mg0.87Ti0.13, d) Mg0.89Ti0.11, e) Mg0.91Ti0.09
7.3.3 Mg-Ni hydrides
Finally, we investigate the properties of Mg-Ni alloys. Figure 7.5a shows the
properties in the metallic state, where we see a much larger spread in initial n values
for these materials, but n and k still follow the same trend as was found with the
165
Figure 7.5: Optical properties n? = n + ik of siz dierent Mg-Ni alloy hydrides. a)
Optical properties in the pristine metallic state before hydrogenation. b) Change
in optical properties upon hydrogenation, dened here as the pure metal optical
properties subtracted from the hydride optical properties. c) Optical properties
in the fully hydrogenated hydride state. Each colored line on the plot represents
a dierent atomic ratio of metal hydride, with darker shades representing higher
atomic Mg percent ratios.
other Mg alloys (n and k generally increase with increasing wavelength). This larger
spread in initial properties is expected, as we were able to obtain a larger range of
atomic ratios for this alloy system compared to the other two, ranging from 59% -
92% Mg. In Figure 7.5b we see the same trends of optical properties change as Mg-
Al and Mg-Ti with large decreases in k across the spectrum upon hydrogenation,
and increases in n in the visible with decreases in the NIR. In the hydride state in
Figure 7.5c, we see a large range of potential nal properties. Generally, the lower
Mg percent hydrides have a higher n across the spectrum, with the Mg0.90Ni0.10
sample demonstrating the lowest n across most of the spectrum and Mg0.59Ni0.41
the highest. Most of the hydrides have low attenuation in the visible, with k < 2.
The higher Mg percent hydrides then have larger attenuation into the NIR, while the
166
lower Mg percent samples exhibit a more constant k across the measured spectrum.
Figure 7.6: Optical properties of dierent Mg-Ni alloys as they are exposed to H2
gas. The lightest colored line depicts the alloy in the pristine metallic state. As H2 is
introduced to the system, the material begins to hydrogenate, denoted by the lines
getting darker in the plot, with the darkest line indicating the full hydride state.
Each line is not linearly spaced in time and is instead chosen to aesthetically show
the range of possible intermediate states. The alloys shown here are a) Mg0.59Ni0.41,
b) Mg0.73Ni0.27, c) Mg0.84Ni0.16, d) Mg0.85Ni0.15, e) Mg0.90Ni0.10, f) Mg0.92Ni0.08
Figure 7.6 shows the dynamic transition state data for the Mg-Ni samples.
For each of the samples, the transitions are mostly monotonic without any exotic
features. For the intermediate loading states, here we are modeling the Mg-Ni
167
lm as homogeneous throughout the thickness of the lm. As was discussed in
the introduction, it has been found that Mg-Ni lms do not load homogeneously,
but instead with preferential phase formation from the substrate of the lm that
propagates to the Pd cap. Our modeling process is not contradictory to this process
and still allows for the bulk characterization of the properties of the lm in this
orientation with illumination through the Pd cap. The low MSE obtained in our
optical ts indicate that our model accurately captures the optical properties of
our samples. For our dynamic ts, the MSE for any individual t was never greater
than 10, indicating a good t. However, for a dierent illumination orientation of the
device (i.e. backside illumination through a transparent substrate), our intermediate
model ts would not account for the layering eects of the loading. To model those
bulk responses with our setup, the optical properties would have to be measured in
that same orientation. Note that this layering eect would only aect the properties
in the intermediate states, and have no eect on the modeling of the metallic or fully
hydrogenated states shown in Figure 7.5.
To determine whether any of our materials would exhibit this black state
with backside illumination, we modeled the multi-layer loading process with TMM
simulations. Figure 7.7a shows the simulation architecture. The samples consist of
illumination through a SiO2 substrate. The layers proceed from the substrate with
the fully hydrogenated Mg-Ni hydride, the fully metallic Mg-Ni alloy, and lastly
a 3 nm PdH2 capping layer. The hydrogenation of the material is simulated by
beginning with the Mg-Ni-H layer equal to 0 nm, and then increasing this layer size
while decreasing the Mg-Ni layer by the same rate, until the sample is completely
168
hydrogenated. We dened the alloy thickness to be 25 nm for these simulations.
Using this method, we did nd a high absorbing intermediate state for multiple
samples, with the largest absorption occurring for the Mg0.73Ni0.27 sample, which is
expected as this material is the closest composition to Mg2Ni for which the black
state was initially discovered. In Figure 7.7b, we show the absorption curves for
dierent loading thicknesses for this alloy, where the peak absorption occurs at 12
nm loading, which is equal to half of the thin lm being loaded.
Figure 7.7: Modeled absorption from backside illumination of Mg0.73Ni0.27. a) Simu-
lation architecture for sample loading. Sample consists of a SiO2 substrate, followed
by a fully hydrogenated Mg-Ni-H layer, then a fully metallic Mg-Ni layer, and nally
a 3 nm Pd capping layer. The total Mg alloy thickness is dened to be 25 nm. Sim-
ulated hydrogenation is dened as an increase in the thickness of the hydride layer
and a decrease of the same magnitude of the metal layer. b) Calculated absorption
for dierent loading thicknesses using Mg0.73Ni0.27 optical properties.
7.4 Stress and loading properties
On the second QCM crystal included in the deposition chamber, we dynami-
cally measured the loading and stress properties of the lms. The loading values for
169
all of the samples are shown in Figure 7.8. Figure 7.8a shows the loading values of
the Mg-Al samples. We can see that the samples have loadings near H/M = 1, with
the highest measured loading at 1.28 for Mg77Al23 and the lowest loading at 0.75 for
Mg81Al19. These values are at the low end of the range of hydrogen loading mea-
surements reported in the literature for alloys in this composition range, which nd
loading values between H/M = 0.85 - 1.5 [220, 224]. The Mg77Al23 also exhibited
the largest optical property change compared to the other Mg-Al samples. Future
work looking at these samples should investigate fabrication of higher Al atomic
percentages to determine if there is a correlation of higher loading for higher Al
percent for any range of compositions and if the optical properties change is greater
in this region.
Figure 7.8: Measured maximum loading values for dierent thin lm samples for a)
Mg-Al, b) Mg-Ti, and c) Mg-Ni alloy samples.
170
The loading values of the Mg-Ti lms are shown in Figure 7.8b. Three of
these values are in general agreement in the literature, which nds that the average
amount of loading of Mg-Ti alloys in this atomic composition range average H/M =
1.55 [229, 242]. We nd two samples that fall measurably below this average, with
Mg0.82Ti0.18 at 0.96 and Mg0.89Ti0.11 at 0.79. When we look at the optical data,
we nd that the Mg0.89Ti0.11 sample had the lowest optical change of any of the
investigated alloys as well. We suspect that there was an issue with the fabrication
of this sample that prevented a complete loading. This could have been caused by
an incomplete Pd capping layer that did not fully encapsulate the sample, allowing
for oxidation of the surface of the alloy, or potential alloying between the Pd capping
layer and Mg near the surface of the Mg-Ni alloy.
For Mg-Ni alloys in Figure 7.8c, we see generally lower calculated loadings
than for the other two alloys. We also nd a slightly negative correlation between
the loading amount and the atomic Mg percent. This does not agree with pre-
viously found data in the literature, which found that there should be a slightly
positive correlation between these values and that the values should fall between
1.2 and 1.4 H/M [234]. This loading dierence could be attributed to dierences
in sample preparation. Other thin-lm Mg-Ni alloys have been fabricated with
multi-layer metal deposition followed by a high temperature anneal, as opposed to
our cosputtering method. These dierent fabrication conditions could potentially
be forming dierent alloying phases within the metal, which would aect the total
loading amount. Further studies on the crystal structures of Mg-Ni alloys fabricated
with these two techniques should be done to determine if there is any dierence.
171
In Figure 7.9, we show the total stress change of each of the measured Mg
alloys upon full hydrogenation. These stress changes are compressive and are dened
to be positive. We nd the stresses for the samples to be fairly consistent within
a material system. Taking the averages of the stresses in each system, the Mg-Al
samples have the highest stresses with an average of 0.56 GPa, next is Mg-Ti with
an average of 0.48 GPa, and nally the Mg-Ni alloys have the lowest measured stress
with an average of 0.36 GPa.
Figure 7.9: Measured total stress change values for dierent thin lm samples for a)
Mg-Al, b) Mg-Ti, and c) Mg-Ni alloy samples. The change in stress reported here
is the stress change from the initial pristine metal mounted in the environmental
chamber to the fully hydrogenated state (does not include intrinsic stress of initial
pristine alloy).
172
7.5 Applications
In this section, we use our found optical properties of the alloys to simulate
potential applications. We use the transfer-matrix method to simulate thin-lm
responses of these materials for dierent thicknesses and on dierent substrates. The
rst application of these alloys that has been suggested is for switchable window
technologies. As a test of this application, we simulate the transmission through
the alloys in their metallic and hydride states and compare transmission amounts.
We use a metallic alloy thickness of 25 nm with a 3 nm Pd capping layer. We
nd that these thicknesses are in the ideal range for switchable window purposes
because it has just thick enough in the metallic state to create high reection,
while remaining thin enough to allow appreciable transmission in the hydride state.
These simulations also take into account the volume expansion of the alloys upon
hydrogenation of 15%. Another important factor in window technologies is to have
color-neutral transmission. Windows with non-neutral color transmission tint the
light as it transmits through the window, which is not ideal when attempting to
make a clear window. To model this color distortion, we use the CIE 1931 XYZ
color space and plot the perceived colors of the transmitted spectra. Note that only
x and y need to be plotted to fully characterize the color because x + y + z = 1. On
these plots, color-neutral is the x = y = 0.333 data point, which creates the ideal
window. We show these switchable window properties for all of the investigated
alloys in Figure 7.10.
For the Mg-Al alloys, we see poor transmission in the hydride state, with the
173
Mg0.84Al0.16 alloy having the highest transmission through the visible with values <
40% transmission for most of the spectrum. These materials generally only exhibit
?20% absolute change in transmission upon hydrogenation. This is due to the
materials still exhibiting a high attenuation in the visible, even in the hydride state.
These windows are close to color neutral, even with their low transmission, adding
a small blue-green tint to the transmitted light.
Figure 7.10: Simulated switchable window performance with Mg alloys. Simulated
transmission values of thin lm a) Mg-Al, b) Mg-Ti, and c) Mg-Ni alloys on SiO2.
The stack is dened as an SiO2 substrate, a 25 nm Mg-Al lm, and a 3 nm Pd
capping layer. Transmission in the hydride (metal) state is represented by solid
(dashed) lines. d) Chromaticity plot with the transmission color points for the
dierent alloys. Colors of the points match with the colors in part a-c) along with
the rest of the chapter. e) Zoom in of chromaticity plot. The intersection of the
blacked dashed lines represents color neutral at x = y = 0.333.
Some of the Mg-Ti alloys perform much better as switchable window technolo-
174
gies with transmissions > 60% across most of the visible spectrum for Mg0.85Ti0.15
and Mg0.87Ti0.13. For these samples, we observe a transmission change of ?40%.
Higher transmission can be achieved in the hydride state by a thinning of the sam-
ple; however, this thinning causes the transmission in the metallic state to also
become signicantly higher. We nd the transmission colors of these windows to
be similar to those of the Mg-Al samples, being mostly color neutral with a slight
blue-green tint. Mg0.85Ti0.15 is the most color neutral, in addition to having the
highest transmission across most of the visible.
The Mg-Ni alloys measured in this chapter have poor switchable window char-
acteristics, with low transmissions in the hydride state in the shorter wavelength vis-
ible region. For the longer wavelength visible, we see increased transmission but still
only achieve values of ?50%. However, we do see ?40% transmission changes in this
region. Some Mg-Ni samples exhibit very good color neutrality, with Mg0.92Ni0.08
having a transmission color value of x = 0.325 and y = 0.338. The other alloys
have a slight yellow tint, as opposed to the blue-green tint of the Mg-Al and Mg-Ti
alloys.
As mentioned in the introduction, Mg-Ti alloys have also been investigated as
broadband switchable light absorbers, with a highly reecting state in the metallic
form and a highly absorbing state when hydrogenated. The high absorption states
for these materials have been found with thicker samples > 200 nm, much thicker
than measured here, and it has been demonstrated that the total absorption of the
lm can be signicantly tuned by just varying the thickness of the lm [229, 248].
To see if our measured properties show any potential for switchable absorption, we
175
modeled a 300 nm Mg-Ti lm on and SiO2 substrate with a 10 nm Pd capping
layer. The results of these simulations are shown in Figure 7.11. We see that
for three of our measured Mg-Ti alloys, we achieve large amounts of switchable
absorption throughout the visible, with absorption tailing o into the NIR. In the
visible wavelength region, Mg0.87Ti0.13, Mg0.85Ti0.15, and Mg0.82Ti0.18 all achieve >
80% absorption in the hydride state with < 25% absorption in the metallic state
(corresponding to high reection in this state). This is a very large absorption
change upon hydrogenation for these alloy compositions and shows their potential
for broadband switchable light absorbers.
Figure 7.11: Broadband switchable light absorption with Mg-Ti alloys. a) Schematic
of switchable light absorber consisting of a 300 nm Mg-Ti alloy with a 10 nm Pd
cap on a SiO2 substrate. b) Absorption plots for this structure for dierent alloy
compositions. Solid (dashed) lines are absorption in the hydride (metallic) state.
Colors represents dierent alloy compositions
Finally, as a last application, we determine whether these Mg alloy materials
can be incorporated into the switchable absorber design demonstrated in Chapter
5. This architecture consisted of a Pd capped 25 nm Mg lm deposited on an ITO
176
Figure 7.12: Simulation of Mg alloy/ITO switchable absorption device. a) Device
architecture consisting of a SiO2 substrate coated with a 350 nm ITO lm, followed
by 25 nm Mg, and a 3 nm Pd capping layer. b-d) Absorption plots of this system in
the metallic (dashed lines) and hydrogenated (solid lines) states for Mg-Al, Mg-Ti,
and Mg-Ni systems respectively.
substrate with an NZI resonance ?1250 nm. One issue with this setup was the time
of hydrogenation and dehydrogenation were both long, around 30 min. With the
faster kinetics of these Mg alloy samples, we ran simulations to determine whether a
similar absorption change response could be obtained with any of these alloys. The
results of our simulations are shown in Figure 7.12. We see from these plots that
for any of the alloys we investigated here, we do not achieve as high of absorption
as we attained with the pure Mg sample (93%). The largest absorption that we
obtained was for Mg0.59Ni0.41 with a peak of 79% at 1524 nm. The Mg-Al samples
177
achieved nal absorptions of ?60% with two Mg-Ti samples achieving absorptions
of 70% (Mg0.87Ti0.13 and Mg0.82Ti0.18). We conclude that although the kinetics may
be improved with alloying in this system, the nal absorption is not as high as it is
for pure Mg.
7.6 Conclusions
In conclusion, we have for the rst time experimentally measured the complex
optical properties of dierent compositions of Mg-Al, Mg-Ti, and Mg-Ni alloys using
spectroscopic ellipsometry. We have found a wide range in potential optical proper-
ties for the alloys in the nal hydrogenated state, with most of the samples showing
similar properties in the metallic state. We nd mostly smooth optical transitions
during the hydrogenation process when measured through the Pd capped side of
the lm, and showed that these measurements support previous observations of the
Mg2Ni "black state". We measured the loadings and stresses of all of these samples
and found that the loadings of our materials are slightly less than those previously
reported in the literature. Future experiments with this system will explore how
dierent deposition parameters aect the properties of these alloys and how po-
tential dierences in alloy preparation could aect their nal loadings and optical
properties.
We also explored applications of these materials, showing that Mg-Ti alloys
have potential as switchable windows and broadband switchable light absorbers.
Further, we found that no alloys were able to outperform the high absorption of the
178
Mg/ITO device demonstrated in Chapter 5, indicating that more research needs to
be done to nd an alloy with increased kinetics to match the optical functionality
of pure Mg for this structure.
179
Chapter 8: Conclusions and Future Experiments
This thesis provided details of our recent work utilizing metal hydrides as tun-
able plasmonic and nanophotonic materials. In Chapter 2, we described our custom-
designed and fabricated measurement apparatus that simultaneously measures the
optical, gravimetric, thermal, and stress properties of thin lm and nanostructured
samples. This apparatus allows us to quantify curvature changes of 0.001 m?1 and
mass changes of 13 ng/cm2 in material systems exhibiting large stress uctuations.
We also showed that our calorimetry model demonstrates a 150 ?W calorimetric
accuracy and 20 ?W minimum detectable power and that we can obtain highly
accurate optical property measurements by coupling our system with a spectro-
scopic ellipsometer. In Chapter 3, we used this apparatus to measure the optical
properties of 5 dierent pure metals and demonstrated their potential use in vari-
ous nanophotonic structures, showing 5 order of magnitude reectivity changes and
over 200 nm resonance shifts upon hydrogenation. In Chapter 4, we investigated a
large variety of commercial NZI materials to identify the ranges of potential optical
and electrical properties of dierent TCO materials. In Chapter 5, we then used
one of those identied materials combined with a thin lm metal hydride structure
to experimentally demonstrate a switchable high absorption device, with a peak
180
absorption change of over 75%. Moving beyond pure metals, in Chapters 6 and 7
we introduced the benets of alloying metals for added robustness to their optical
responses. We investigated Pd-Au alloys and showed how measurable signals can be
achieved with Pd atomic percentages as low as 34% and demonstrated hydrogen sen-
sors with no surface poisoning or hysteresis, as well as a physical encryption scheme.
We then showed the wide range of potential optical properties in the Mg alloy sys-
tem, measuring and tting the optical properties of dierent atomic compositions
of Mg-Al, Mg-Ni, and Mg-Ti systems and showing potential uses of Mg-Ti alloys
as broadband switchable light absorbers and switchable windows. Ultimately, this
work has measured the properties of many types of metal hydrides and has opened
up new avenues for the simulation and design of novel nanophotonic devices using
our measured properties. In the following section, we will describe future directions
of this research focusing on two potential material directions: using high-entropy
alloys as metal hydrides to further explore the optical parameter space and by using
plasmonic resonances in metal hydrides to lower the Coulomb barrier in nuclear
reactions.
8.1 High entropy metal hydrides
In this thesis, we only investigated bi-metallic systems to improve the optical
and structural properties of pure metal hydrides. These bi-metallic alloys open a
wide parameter space, but to perfect certain designs, we can expand even further.
Work is already being done using ternary metal alloy hydrides as hydrogen sensors,
181
specically using a Pd-Au-Cu system to further improve chemical resistance and
stability [22]. In these binary and ternary systems, generally there is a principal
active element, with small additions of other secondary elements to adjust the ma-
terial properties. In the case of Pd-Au-Cu, the Pd is the primary element with the
secondary Au and Cu to enhance chemical resistance and cyclability. A new class
of alloys has recently been investigated that uses many principal elements in high
concentrations that exhibit properties that can far exceed those of traditional alloys
[249]. These materials, named high entropy alloys (HEA), have become of particular
interest in the hydrogen storage community since it has been demonstrated that a
TiVZrNbHf HEA can absorb hydrogen to a much higher loading than normal tran-
sition metal hydrides (? 2) up to a loading of 2.5 H/M [250]. This large hydrogen
storage capacity is due to increased lattice strains within the alloy, allowing hydrogen
atoms to occupy both tetrahedral and octahedral interstitial sites. Many other com-
binations and dierent ratios of transition metals are currently being investigated
to attempt to surpass this loading value [251, 252, 253].
Although HEAs have been studied for hydrogen storage applications, they
have yet to be investigated for optical devices. One reason for this absence is that
so far these materials have not been able to be fabricated at the nanoscale. The
most common way of manufacturing the alloys is by arc melting specically chosen
ratios of the pure metals multiple times under an inert atmosphere. This method
gives an ingot of the material so that the properties can be studied in bulk, but the
method cannot be easily translated to thin lms or nanostructures. As far as we
know, there have been no attempts to use these ingots to deposit thin lms through
182
a physical vapor deposition process such as electron beam evaporation or sputter-
ing, but there is concern that there will be a separation of the individual metals
during the cooling process. With this separation, the benets of the HEA could be
lost. Potentially combining a ash cooling method with this technique could yield
positive results, which needs to be investigated. Other groups have been able to fab-
ricate nanoparticles on carbon supports using ash heating and cooling, which has
been demonstrated to be successful in fabricating HEAs with up to 8 elements [254].
This process would be very interesting to investigate with metal hydride HEAs to
see if these nanoparticles can have exotic optical responses when exposed to H2 gas.
Unfortunately, this process only creates these particles on these specic supports
and is not very versatile to other systems, thus more work is necessary to attempt
to broaden the substrate choice. Lastly, co-sputtering has been shown to create
non-energetically favorable alloys under certain conditions and could potentially be
used to fabricate thin-lm HEAs. The limitation is that most sputtering systems
can only deposit 2-3 elements at a time. If a custom system was built that could
co-deposit up to 5 elements, then thin-lm and nanostructured devices could poten-
tially be implemented. Using the apparatus described in this thesis, we could then
measure the nanoscaled properties of these materials to determine whether further
improvements to hydrogen sensors or other device designs could be implemented.
183
8.2 Nuclear plasmonics with metal hydrides
Moving beyond nanophotonic devices, we believe that metal hydrides can
have future uses in increasing nuclear reaction rates. It has already been shown
at Lawrence Berkeley National Laboratory and the National Aeronautics and Space
Administration that when certain metal hydrides are loaded with a high density of
deuterium atoms, the Coulomb barrier between deuterium ions is reduced due to
the screening from the conduction electrons from the host metal [255, 256, 257].
This lowering of the Coulomb barrier allows for a greater probability of a deuterium
ion tunneling through this barrier, causing a nuclear reaction. This eect is most
prevalent at lower energies < 5 keV, as this is the energy region in which electron
screening plays the largest part in nuclear reaction probabilities (at high energies,
eects from screening become negligible). Schenkel et al. found that this screening
eect from the metal hydride lattice was 1000 ? 250 eV, causing a signicant in-
crease in the fusion rates for ion energies < 5 keV [255]. These ndings open up the
door to potentially energetically favorable nuclear fusion reactions at much lower
temperatures than previously thought.
We propose that this electron screening eect from the metal hydride lattice
can be further enhanced by plasmonically exciting the target. Plasmonic excita-
tions can increase the electric eld on the edges of nanostructures by up to 1000x
in ideal structures [258]. This large eld enhancement could dramatically decrease
the Coulomb barrier for deuterium atoms within this region. The plasmonic en-
hancements do not cause steady-state electric eld enhancements at a specic point
184
on the structures, but the time scale of the plasmonic enhancement is still much
great than that necessary for a nuclear reaction (10?9 s compared to 10?12 s [259]).
This time scale dierence allows for more than enough time to get a measurable
enhancement of D-D fusion.
Our proposed experiment uses a nanoporous Pd membrane as our plasmonic
metal hydride target. Nanoporous membranes have been well studied and can be
fabricated using the process of dealloying. In this process, rst, a bi-metallic al-
loy lm is deposited onto a substrate, where one of the metals in the alloy is
the metal of interest and the other is a sacricial metal. This alloy is then wet
etched until the sacricial metal has completely dissolved and the remaining lm is
a nanoporous sheet of the desired metal. This process has been demonstrated to
create Pd nanoporous membranes by using Pd0.2Co0.8 alloy and etching it in H2SO4
[260]. These membranes can have fairly strong plasmonic resonances, allowing for
high eld enhancements at ridges of the pores [261, 262]. We believe that these
nanoporous materials are ideal for these experiments because they are robust to
degradation from an incident ion beam. As the ion beam slowly breaks down the
top of the target, the plasmonic resonance still stays intact because of the thick,
random structure of the target. This is compared to other plasmonic targets, such
as nanogratings or bowtie structures, that would stop exhibiting a strong plasmonic
resonance after a small degradation in the structure.
To create the nuclear reactions on this target, we will use a variable energy
deuterium ion gun, where the energy can be tuned from 0.5 - 5 keV. This range is
ideal for these experiments because at the upper energy end we will be able to create
185
fusion reactions without any electron eld enhancement, thus we can compare to
literature values in this range. As we lower this ion gun energy, we then get into the
energy regime where the greatest enhancements to the fusion rate will occur due to
the reduction in the Coulomb barrier. We will then excite the nanoporous target
with a visible laser set at the experimentally determined plasmonic resonance of the
target. To nd the eect of the plasmonic excitation on the nuclear reaction rate,
we will compare the nuclear byproduct formation rate with the plasmonic excitation
laser on and o. We will measure the neutron and proton formation from the nuclear
reactions with the setup outlined in Schenkel et al. [255].
The results of these experiments will inform the possibilities to tailor nuclear
reaction rates using visible light excitation. Structuring targets in nuclear reactors
could signicantly lower the necessary energy needed to spark nuclear fusion reac-
tions. Beyond these experiments, there are many other applications of this nuclear
plasmonics concept. One in particular would be to use plasmonic excitation to
change the rates of nuclear decay of certain radioactive atoms. We know that the
rate of ? decay is dependent on the tunneling rate through the Coulomb barrier,
thus if we lower this barrier through plasmonic excitation, we could speed up this
decay process in a controlled way.
186
Appendix A: Curvature to frequency derivation
We know that there are no (DC) shear strains in this system, meaning 4
= 5 = 6 = 0. Here and in what follows we use the standard contracted, Voigt
notation and with 's being strains, ?'s being stresses, u's being displacements, and
A's being stiness elements. We can then multiply the strain by the stress tensor
to get expressions for ?1 and ?2. The relationship between these two quantities is
important in calculating wave propagation speeds in the crystal.
?? ??? ?1 ??? ???
? ?? ?
A? ? ? ? ? ?11 A12 A13 A14 A15 A16 ???? ???? ? ???? ?
1 ??
??
2
?? ?
????
?? ?
A? ? ? ? ? ? ?? ?21 A22 A23 A24 A25 A26 ????? 2? ? ? ? ?
??
? ?3 ? ? A? ?31 A32 A? A? A? A? ???? ? 33 34 35 36 ?
 ?3 ?
??? ? ???
=
????
???? ?? (A.1)
A?4 41 A
? ?
42 A43 A
? ? ? ?? ?
? ? ? 44
A45 A46 ?????? 4 ?????
?
? ??? ??? A? A? A? ? ??5 51 52 53 A54 A? A?55 56 ?????? ?5 ???
? A? A? A? ? ? ?6 61 62 63 A64 A65 A66 6
Thus, we have
?1 = A
?
11
? ?
1 + A122 + A133 (A.2)
187
? = A?2 211 + A
?
222 + A
?
233 (A.3)
Since the lm is free to expand along x3, we know that there is no stress ?3.
We can use this information to determine 3 as a function of 1 and 2.
?3 = 0 = A
?
311 + A
? ?
322 + A333 (A.4)
Solving for 3 yields
A?311 + A
?
 = 32
2
3 ? (A.5)A33
We can also determine expressions for 1 and 2 if we make an assumption of
the shape of the curved region. If the strained region has constant curvature, the
displacement in that region can be written as
?1 ?2
u = x23 1 + x
2
2 (A.6)2 2
Where ?1(?2) is the curvature in the x (y) direction. Earlier, we noted that
there are no shear strains in this system. We can use this fact again to obtain
expressions for1 and 2. We can write out the denition of strain to get
( )
1 ?u3 ?u1
5 = 0 = ? (A.7)
2 ?x1 ?x3
And thus,
188
?u1 ??u3= (A.8)
?x3 ?x1
We know that u3 is given by the expression above. Importantly, it does not
depend on x3, we can integrate with respect to x3. We can then dierentiate that
expression with respect to x1 to obtain an expression for the strain in the x1 direc-
tion.
?u ?21 u3
= 1 = ?x3 (A.9)
?x1 ?x21
Substituting in for u3 yields
1 = ??1x3 (A.10)
A similar analysis for 2 gives
2 = ??2x3 (A.11)
We can insert these expressions for the strains along with the expression for
3 as a function of 1 and 2 into an above equation to yield
? ? ? ?
? = ?A? ? A A A A1 11?1x3 ? A12?2x3 ? 13 31? ? x
13 32
1 3 ? ? ?2x3 (A.12)A33 A33
?
? ? ? ? ? A23A
? ?
31 A23A
?
? = A ? x A ? x ? x 322 21 1 3 22 2 3 ? 1 3 ? ? ?2x3 (A.13)A33 A33
189
Now, if we assume that the curvatures must be identical along both axes due
to sample clamping, we drop the ? direction distinction and nd
(
A?
)
?1 = ??x ?A? ? A? 13 ? ?3 11 12 ? ? (A31 + A32) (A.14)A33
( ? )
? = ??x ?A? ? AA? 13 ? ?2 3 11 12 ? ? (A31 + A32) (A.15)A33
As per EerNisse [46], the velocity of a shear wave through a crystal at zero
stress, W0, along a direction oriented at an angle ? relative to the x3 axis, is given
by
?0W
2
0 = C66cos
2? + C44sin
2? + 2C14sin?cos? (A.16)
Here, the stress tensor coecients are in the crystal axes, not the lab axes. In
the lab axes, the ratio of the stresses is T ?2 = ?2/?1. Along the crystal axes, we have
T2 = cos
2?T ?2
T = sin2?T ? (A.17)3 2
T ?4 = sin?cos?T2
The expression relating the angle of propagation and curvature to the wave
speed is then
190
1
W (?, x3, ?) = ? (? 2 20W0 + ?(x3, ?)(cos ?(?0.08 + 1.38T2 + 1.66T3 + 0.49T4)?0
+ sin2?(0.55 + 0.20T2 ? 2.68T3 ? 5.75T4)
+ 2sin?cos?(?1.81 + 0.28T2 ? 0.55T3 ? 0.04T ))1/24
(A.18)
Because the speed is a function of x3, it changes as the wave propagates
through the crystal. To calculate the total transit time, we need to take an in-
tegral. The transit time through the crystal and back to the starting point is given
by
? 2h/3 dx3
t(?, ?) = (A.19)
?h/3 W (?, x3, ?)
The limits of the integral are referenced to the plane of no stress in the crystal,
which is h/3 from the bottom of the crystal [263]. The change in oscillation frequency
as a function of stress is then
1 1
?f(?, ?) = ? (A.20)
2t(?, ?) 2t(?, 0)
Thus, the change in frequency depends nearly linearly on curvature with a slope of
approximately ? = -777 Hz/m?1.
191
Appendix B: Commercial NZI Materials Optical Properties
The individual optical properties of each of the TCO materials investigated in
Chapter 4 are reported here. There properties were measured with spectroscopic el-
lipsometery combined with transmission measurements and t with a Drude-Loretnz
oscillator model. Figures B.1, B.2, and B.3 show the optical properties of the ITO
samples, Figure B.4 shows the optical properties of the FTO samples, and Figure
B.5 shows the optical properties of the AZO samples. The headers in each graph
show the material, the company the sample was sourced from, and the nominal
resistance quoted by that company.
We summarize the complete ndings from these measurements in Table B.1
and Table B.2. In these tables, we report the minimum |n| achieved for each sample,
the location of the minimum, the NZI bandwidth, the measured thickness t of the
TCO, and the measured resistance R.
192
Figure B.1: Optical properties of commercially sourced ITO samples. Above each
sample plot is the name of the company the sample was sourced from and the
nominal resistance quoted for the sample.
193
Figure B.2: Optical properties of commercially sourced ITO samples. Above each
sample plot is the name of the company the sample was sourced from and the
nominal resistance quoted for the sample.
194
Figure B.3: Optical properties of commercially sourced ITO samples. Above each
sample plot is the name of the company the sample was sourced from and the
nominal resistance quoted for the sample.
195
Figure B.4: Optical properties of commercially sourced FTO samples. Above each
sample plot is the name of the company the sample was sourced from and the
nominal resistance quoted for the sample.
196
Figure B.5: Optical properties of commercially sourced AZO samples. Above each
sample plot is the name of the company the sample was sourced from and the
nominal resistance quoted for the sample.
197
Company Material Nominal R (?) t (nm) |n|min ?center BW
R (?) (nm)
Adafruit Industries ITO 10-15 8.45 181.5 0.56 1300 303
Biotain Crystal AZO 8-10 8.35 916.8 0.67 1984 492
Biotain Crystal FTO 10-15 15.06 382.3 0.72 1536 406
Biotain Crystal ITO 10-15 11.15 138 0.57 1333 314
Biotain Crystal ITO 30-40 23.15 56.6 0.56 1166 279
Biotain Crystal ITO 4-5 6.09 212.7 0.59 1205 284
Biotain Crystal ITO 6-8 8.38 180.7 0.56 1326 326
Biotain Crystal ITO 80-100 76.27 19.1 0.66 1162 237
Delta Technologies FTO 16-20 19.96 234.8 0.78 1715 364
Delta Technologies ITO 15-25 19.61 67.5 0.63 1225 273
Delta Technologies ITO 30-60 44.29 29.2 0.57 1183 257
Delta Technologies ITO 4-8 8.43 158.9 0.55 1211 288
Delta Technologies ITO 8-12 11.55 121.2 0.56 1218 284
MSE Supplies AZO 10 10.60 869 0.82 1993 390
MSE Supplies FTO 15 20.72 307 0.88 1627 296
MSE Supplies ITO 12-15 13.43 133.4 0.60 1392 344
MSE Supplies ITO 30 34.28 46.3 0.61 1251 298
MSE Supplies ITO 3 3.97 354 0.59 1241 295
MSE Supplies ITO 7-10 5.99 227.1 0.57 1229 297
MTI Corporation ITO 12-15 11.09 142.6 0.57 1325 315
MTI Corporation ITO 16-19 14.22 106.9 0.57 1277 301
MTI Corporation ITO 6-7 6.47 222.6 0.59 1254 300
MTI Corporation ITO 8-10 9.37 186.1 0.57 1297 327
NanoCS FTO 12-17 7.08 349.1 0.81 1774 393
NanoCS ITO 100 78.09 21.6 0.67 1326 272
NanoCS ITO 10 7.59 174.8 0.53 1211 281
NanoCS ITO 20 15.69 94.2 0.62 1278 284
NanoCS ITO 50 41.91 36.8 0.58 1285 282
NanoCS ITO 5 3.91 354 0.58 1282 299
Table B.1: Summary of commercial NZI data#1. In this table we report the com-
pany, the nominal resistance, the measured sheet resistance R, the measured thick-
ness t, the minimum possible magnitude of the index of refraction |n|min, the location
of that minimum ?center, and the bandwidth BW of the resonance where |n| < 1
198
Company Material Nominal R (?) t (nm) |n|min ?center BW
R (?) (nm)
Nanoshel AZO 10 10.10 891.3 0.85 2063 402
Nanoshel ITO 10 8.38 185 0.60 1246 291
Nanoshel ITO 15 25.06 61.7 0.58 1160 270
Ossila FTO 11-13 11.33 437.9 0.85 1919 347
Ossila FTO 12-14 13.22 355.5 0.75 1764 440
Ossila FTO 6-9 7.80 625 0.69 1856 524
Ossila ITO 20 17.34 111.5 0.68 1281 299
SPI ITO 30-60 34.62 126.5 0.73 1864 363
Sigma Aldrich FTO 8 8.20 603.1 0.69 1887 497
Sigma Aldrich ITO 30-60 45.04 30.2 0.59 1174 251
Sigma Aldrich ITO 70-100 85.31 18.1 0.62 1165 237
Sigma Aldrich ITO 8-12 11.64 123.4 0.55 1225 302
Techinstro AZO 10 11.03 838.6 0.90 2023 306
Techinstro FTO 15 12.89 343 0.76 1754 419
Techinstro FTO 7 7.03 574.4 0.64 1848 555
Techinstro ITO 100 83.43 18.5 0.64 1311 281
Techinstro ITO 10 8.05 183.5 0.60 1237 290
Techinstro ITO 20 10.24 148.5 0.57 1311 308
University Wafer ITO 15-20 15.42 101 0.60 1320 297
University Wafer ITO 7 6.34 213.3 0.60 1264 284
Table B.2: Summary of commercial NZI data #2. In this table we report the
company, the nominal resistance, the measured sheet resistance R, the measured
thickness t, the minimum possible magnitude of the index of refraction |n|min, the
location of that minimum ?center, and the bandwidth BW of the resonance where
|n| < 1
199
Appendix C: H2 Safety Protocols
When working with hydrogen gas, certain precautions must be taken to ensure
safety in the laboratory. H2 gas is colorless, odorless, and tasteless, so one cannot
rely on their senses alone to detect a leak in their system. H2 gas is ammable
and is dangerous when it is in concentrations between 4% and 75% at atmospheric
pressure. This is the combustion range for this gas. Below this range, there is not
enough H2 in the atmosphere for ignition, and above this range, there is not enough
O2. H2 has a very low ignition energy, so even a small spark in this partial pressure
range can set o an explosive reaction. Because of this, any room where hydrogen
is being worked with should have good ventilation that is adequate enough to deal
with the largest expected H2 leak in the laboratory. Any hydrogen that is being used
in an experiment and then discarded should be piped directly into this ventilation
shaft.
Leaks when working with H2 gas should be expected from time to time, as
hydrogen can leak from very small gaps in the tubing connectors due to its small
molecular size. It is recommended to use a hydrogen leak detector near any H2 gas
lines to alert of any leaks. Along with this, gas pressures should be periodically
checked to make sure that they are depleting by the expected value of hydrogen
200
used in an experiment. To test the leak rate of the system, pressurize the system
to the normal working pressure of an experiment, close o the input from the H2
cylinder and output at the pressure controller, and measure the time it takes for the
system to depressurize. If a leak is found, a common method to nd the leak is the
"soapy water method". Simply, soapy water is brushed onto any tubing connectors
that are suspected to contain the leak. If soap bubbles begin to form around the
connector, a leak has been identied and the tting should be replaced. The leak
rate test should then be performed again to ensure that the leak was properly xed
and that there are no more leaks in the system.
H2 gas is known to cause embrittlement in certain materials. Make sure that
all piping and gas connectors are rated to be compatible with H2 and won't degrade
over time. Only use regulators that have been specically designed for hydrogen. For
H2 cylinder storage, the cylinders must be stored upright and should be secured to
the wall. These cylinders should be kept away from any heating element, including
direct sunlight. When moving the cylinders, be sure to completely close the valve,
remove the regulator, fully screw on the safety cap, and chain the cylinder into the
cart being used.
201
Appendix D: H2 Sensors
Fast and reliable H2 are becoming essential in the newly emerging hydrogen
economy. H2 gas can be created using the electrolysis of water, which can be driven
with green electricity from sources like solar or wind power. When burned, H2 gas
gives o no greenhouse gasses, making it an ideal fuel for vehicles or as a method to
store energy. In order to use it for any of these purposes, H2 detectors are needed
to locate any leaks quickly and eciently, as H2 is very combustible. There are ve
main components that must be accounted for to make an ideal H2 gas sensor, as
laid out by the United States Department of Energy [185]:
1. Reliable: Sensors must have unambiguous readings regardless of what the
previous conditions were in the environment. This means that there can be
no intracycle or intercycle hysteresis. Each value of the sensor signal must be
mapped to a single hydrogen pressure value.
2. High Accuracy: Sensors must be able to measure H2 pressure with precision
in the range of 0.1 - 10% at atmospheric pressure.
3. Lifetime: Sensors must be able to run for many cycles of hydrogen gas,
ideally reaching 10 years of successful operation. These sensors must also be
202
able to be resistant to poisoning gases in the atmosphere such as CO and
suldes. Additionally, the sensors must be able to perform in high humidity
environments.
4. Speed: Sensor detection and recovery time should be < 1 second.
5. Cost: Sensor fabrication should be cheap enough to be feasibly made in bulk
There are many types of H2 sensors that have been demonstrated in the liter-
ature or are currently on the market that have been attempting to reach these DOE
goals. A summary of the current device types and how they operate follows:
1. Catalytic: A catalytic wire array is heated and introduced to the H2 gas,
which then reacts with O2. This reaction releases heat, which is then measured
and can be converted to H2 pressure [264].
2. Electrical Conductivity: A metal hydride wire or thin lm is exposed
to H2 gas and the change in the resistance of the metal hydride is measured
and converted to the pressure of H2 in the chamber [265, 266, 267, 268].
3. Thermal Conductivity: Heated gas is run through a pipe and changes
in the thermal conductivity of the gas are used to calculate changes in gas
composition [269].
4. Semiconductor Metal Oxides: Certain metal oxides have a large resis-
tance change when exposed to H2, such as SnO2, and this resistance is used
as a proxy for hydrogen concentration [270, 271]. Must be done at high tem-
peratures.
203
5. Electrochemical: Electrodes are submerged in solution and gas is piped in
through a porous membrane. Amount of H2 in the gas changes the measured
current [272, 273].
6. MOS/MOSFET/Schottky Diode: The top electrode is made of Pd or
another hydrogen sensitive metal. As it hydrogenates, the work function of
the structure changes and can be measured as a proxy for the H2 concentration
[274, 275, 276].
7. Surface Acoustic Wave Detector: As a material hydrogenates, the
velocity of surface acoustic waves change due to conductivity changes in the
Pd. This velocity change can be correlated with dierent H2 amounts [277,
278].
8. Microelectromechanical: Structural changes in metal hydrides upon hy-
drogenation are used to determine the H2 gas concentration. These designs
include measuring the capacitance between a metal hydride cantilever and a
plate, with the cantilevers bending at dierent angles with dierent H2 con-
centrations [279] and using metal hydride nanoparticles that expand upon
hydrogenation, completing an electrical circuit for a sharp drop in electrical
resistance [280].
Another type of hydrogen sensor that is not yet commercially available, but is
seeing signicant attention in research and development, is optical hydrogen sensors.
Optical hydrogen sensors are desirable for many reasons over some of the other
204
sensor types above. Optical sensors do not require for there to be any electrical
connections in the region of detection, which eliminates any combustion risk from
a spark from the detector. These sensors can also be designed to perform at room
temperature, have a very fast sub-second response time, and can be highly selective
to hydrogen compared with other methods. Within optical hydrogen sensors, many
dierent designs can achieve eective readings. The most common design is to
use nanoparticle arrays of metal hydrides. In these systems, a nanoparticle metal
hydride array, typically Pd or a Pd-based alloy, is deposited on a glass substrate.
The LSPR of this array is then measured with a spectrometer. By measuring the
shift of the LSPR, the atmospheric H2 percentage can be determined. By alloying
the metal hydride with other transition metals, coating the array with a hydrogen-
permeable polymer, or optimizing the nanoparticle shape and size, these sensors can
have improved speed, sensitivity, and poisoning resistance. Demonstrations of these
sensors have been widely shown in the literature [22, 23, 74, 77, 281, 282]. Other
systems utilize thin lm metal hydrides, using the reectance of light o the thin
lm as a measure of the H2 content in the atmosphere [17, 76, 283, 284]. Optical
bers are also frequently used as a base for optical H2 sensors. These bers can be
coated with metal hydrides and can measure the H2 pressure by either measuring the
change in path length due to hydride expansion upon hydrogenation [25], measuring
attenuation changes in the ber [285, 286], measuring shifts in the Surface Plasmon
Resonance within the ber [287, 288], or using a metal hydride mirror at the end of
the ber and measuring changes in reectivity [26, 247]. Finally, measuring shifts
of the Surface Plasmon Resonance without coupling into a ber have been used as
205
a sensing scheme [75, 289].
206
Appendix E: CIE 1931 Color Space Calcuations
In this thesis, we use the CIE 1931 color space to determine the colors of dif-
ferent structures under white light illumination. Our white light source is dened
by the CIE Standard Illuminant D65 [290]. This color space is used to determine
the color an average person would perceive under medium to bright illumination (at
low brightness, the human perception becomes monochromatic). Color perception
is determined by three types of cones in the average human eye that have dierent
sensitivities at dierent wavelengths. This is why most color spaces can be deter-
mined with three values, with each value representing a level of stimulus for each of
these dierent types of cones. The CIE 1931 color space is based o a "Standard
Observer" where color matching functions x?(?), y?(?), and z?(?) describe the relative
perception of red, green, and blue color for an average person. In this thesis, we use
the analytical approximation to the CIE color tables from Ref [291]:
( ( ) ) ( ( ) )
? ??
2 2
595.8 ?? 446.8
x?(?) = 1.065 exp 0.5 + 0.366 exp ?0.5 (E.1)
33.33 19.44
207
( ( ) )
? 2ln? ln556.3
y?(?) = 1.014 exp ?0.5 (E.2)
0.075
( ( ) )2
ln?? ln449.8
z?(?) = 1.839 exp ?0.5 (E.3)
0.051
With these analytical color matching functions, we can then calculate the CIE
1931 XYZ color values from a reection or transmission spectrum using the following
equations:
? ?
k = SD65(?)y?(?)d? (E.4)
0
??
SD65(?)x?(?)R(?)d?
X = 0 (E.5)
k
??
S
0 D65
(?)y?(?)R(?)d?
Y = (E.6)
k
??
SD65(?)z?(?)R(?)d?
Z = 0 (E.7)
k
Where k is a normalizing constant, X, Y, and Z are the tristimulus values,
R is the reection spectrum, and SD65 is the CIE D65 radiance spectrum. Note
that transmission can be directly substituted for the reection in these equations
when determining colors of transmission through a material (such as the switchable
windows in Chapter 7). To be able to plot these colors on a 2D diagram, we can
208
further convert these tristimulus values to the CIE xyY color space which divides
the color into two parts: chromaticity and brightness. The brightness is determined
by the Y value, while x and y are dened by:
X
x = (E.8)
X + Y + Z
Y
y = (E.9)
X + Y + Z
These two values determine the chromaticity of the spectra, with the value x
= y = 0.33 representing the color neutral point. Note that chromaticity is not the
same as a printing color, as the chromaticity is dependent on the spectrum of the
illuminating light.
209
Appendix F: Useful Properties of Metal Hydrides and Hydrogen
Metal hydrides are classied as compounds containing a metal-hydrogen bond
and can be classied into three categories: ionic, covalent, or metallic [292]. The
alkali and alkaline earth metals generally form ionic hydrides, exhibiting properties
similar to other salts formed from a combination of these elements and an element
in the halogen group [292]. Metals to the right of group VII form covalent bonds
with hydrogen, and generally can only form with complex chemical interactions,
as opposed to a direct reaction with hydrogen gas [292]. Note that MgH2 exhibits
both ionic and covalent properties. For metallic hydrides, the hydrogen enters the
interstitial sites of the lattice and forms a hydride phase within the metal. These
compounds are still conductive, have metallic properties, and can be formed with
a direct reaction with H2 gas (although some elements require a catalyst such as
Pd to aid in splitting the H2 molecule). Many of the lighter metals from Group
VI-VIII also require higher H2 pressures (>1 atm) in order to form a metal hydride
[292]. Ni is one example of this, requiring over 6000 atm for hydride formation [11].
For some of the rare earth metals, they are metal hydrides at low H contents but
become semiconductors at high H contents [293]. In Figure F.1, we depict these
classications for each element on the periodic table.
210
Figure F.1: Periodic table classication of metal hydrides. White box indicates the
element forms a metal hydride under <1 atm hydrogen pressure, purple indicates el-
ements with hydride formation >1 atm hydrogen pressure, red indicates covalent hy-
dride formation, green indicates ionic hydride formation, blue indicates non-metals
or no metal hydride formation, yellow indicates radioactive elements with no stable
isotopes, and grey indicates metal hydride formation at low H content, but at high
H contents the compound behaves as a semiconductor.
Another potentially important property of dierent metal hydrides is how close
the H atoms are to each other within the metal lattice. This spacing is particularly
important for potential fusion target designs, like the ones discussed in Chapter
8. In Figure F.2, we plot the dierent measured atomic spacings for hydrogen or
deuterium in dierent forms, including muonic D2 [294], molecular D2 [295], the
closest D spacing theoretically calculated in Pd [296], ZrV2Hx at high pressures
[297], TiH2 [298], MgH2 [298], VH [299], PdH0.7 [10], and PdD0.7 [60].
211
Figure F.2: H-H atom spacing in dierent materials. Values are the average distance
between H atoms in dierent compounds.
212
Bibliography
[1] Thomas Graham. XVIII. On the absorption and dialytic separation of gases
by colloid septa. Philosophical Transactions of the Royal Society of London,
156:399439, January 1866. Publisher: Royal Society.
[2] Yu-Ming Lin, Guo-Lin Lee, and Min-Hon Rei. An integrated purication
and production of hydrogen with a palladium membrane-catalytic reactor.
Catalysis Today, 44(1):343349, September 1998.
[3] Robert E. Buxbaum and Andrew B. Kinney. Hydrogen Transport through
Tubular Membranes of Palladium-Coated Tantalum and Niobium. Industrial
& Engineering Chemistry Research, 35(2):530537, January 1996. Publisher:
American Chemical Society.
[4] Lars J. Bannenberg, Christiaan Boelsma, Kohta Asano, Herman Schreuders,
and Bernard Dam. Metal Hydride Based Optical Hydrogen Sensors. Journal
of the Physical Society of Japan, 89(5):051003, February 2020. Publisher: The
Physical Society of Japan.
[5] Carl Wadell, Svetlana Syrenova, and Christoph Langhammer. Plasmonic Hy-
drogen Sensing with Nanostructured Metal Hydrides. ACS Nano, 8(12):11925
11940, December 2014. Publisher: American Chemical Society.
[6] Xiangyu Zhao and Liqun Ma. Recent progress in hydrogen storage alloys for
nickel/metal hydride secondary batteries. International Journal of Hydrogen
Energy, 34(11):47884796, June 2009.
[7] Billur Sakintuna, Farida Lamari-Darkrim, and Michael Hirscher. Metal hy-
dride materials for solid hydrogen storage: A review. International Journal of
Hydrogen Energy, 32(9):11211140, June 2007.
[8] S. R. Ovshinsky, M. A. Fetcenko, and J. Ross. A Nickel Metal Hydride Bat-
tery for Electric Vehicles. Science, 260(5105):176181, April 1993. Publisher:
American Association for the Advancement of Science Section: Articles.
213
[9] Adolf Sieverts. Palladium und Wassersto. II. Zeitschrift f?r Physikalische
Chemie, 88U(1):451478, June 1914. Publisher: De Gruyter Oldenbourg Sec-
tion: Zeitschrift f?r Physikalische Chemie.
[10] Jesse D. Benck, Ariel Jackson, David Young, Daniel Rettenwander, and Yet-
Ming Chiang. Producing High Concentrations of Hydrogen in Palladium via
Electrochemical Insertion from Aqueous and Solid Electrolytes. Chemistry
of Materials, 31(11):42344245, June 2019. Publisher: American Chemical
Society.
[11] G. Alefeld and J. V?lkl, editors. Hydrogen in Metals II: Application-Oriented
Properties. Topics in Applied Physics. Springer-Verlag, Berlin Heidelberg,
1978.
[12] G. Alefeld and J. V?lkl, editors. Hydrogen in Metals I: Basic Properties.
Topics in Applied Physics. Springer-Verlag, Berlin Heidelberg, 1978.
[13] Julian Karst, Florian Sterl, Heiko Linnenbank, Thomas Weiss, Mario
Hentschel, and Harald Giessen. Watching in situ the hydrogen diusion dy-
namics in magnesium on the nanoscale. Science Advances, 6(19):eaaz0566,
May 2020. Publisher: American Association for the Advancement of Science
Section: Research Article.
[14] J. N. Huiberts, R. Griessen, J. H. Rector, R. J. Wijngaarden, J. P. Dekker,
D. G. de Groot, and N. J. Koeman. Yttrium and lanthanum hydride lms
with switchable optical properties. Nature, 380(6571):231234, March 1996.
[15] K. Yoshimura, C. Langhammer, and B. Dam. Metal hydrides for smart window
and sensor applications. MRS Bulletin, 38(6):495503, June 2013.
[16] Maximilian G?tz, Maren Lengert, Norbert Osterthun, Kai Gehrke, Martin
Vehse, and Carsten Agert. Switchable Photocurrent Generation in an Ultra-
thin Resonant Cavity Solar Cell. ACS Photonics, 7(4):10221029, April 2020.
Publisher: American Chemical Society.
[17] Mohamed ElKabbash, Kandammathe V. Sreekanth, Yunus Alapan,
Myeongseop Kim, Jonathan Cole, Arwa Fraiwan, Theodore Letsou, Yandong
Li, Chunlei Guo, R. Mohan Sankaran, Umut A. Gurkan, Michael Hinczewski,
and Giuseppe Strangi. Hydrogen Sensing Using Thin-Film Perfect Light Ab-
sorber. ACS Photonics, 6(8):18891894, August 2019. Publisher: American
Chemical Society.
[18] W. Lohstroh, R. J. Westerwaal, J. L. M. van Mechelen, C. Chacon, E. Johans-
son, B. Dam, and R. Griessen. Structural and optical properties of Mg2NiHx
switchable mirrors upon hydrogen loading. Physical Review B, 70(16):165411,
October 2004. Publisher: American Physical Society.
214
[19] Nikolai Strohfeldt, Andreas Tittl, Martin Sch?ferling, Frank Neubrech, Uwe
Kreibig, Ronald Griessen, and Harald Giessen. Yttrium Hydride Nanoan-
tennas for Active Plasmonics. Nano Letters, 14(3):11401147, March 2014.
Publisher: American Chemical Society.
[20] Shahin Bagheri, Nikolai Strohfeldt, Monika Ubl, Audrey Berrier, Michael
Merker, Gunther Richter, Michael Siegel, and Harald Giessen. Niobium as
Alternative Material for Refractory and Active Plasmonics. ACS Photonics,
5(8):32983304, August 2018.
[21] Florian Sterl, Nikolai Strohfeldt, Ramon Walter, Ronald Griessen, Andreas
Tittl, and Harald Giessen. Magnesium as Novel Material for Active Plasmonics
in the Visible Wavelength Range. Nano Letters, 15(12):79497955, December
2015.
[22] Iwan Darmadi, Ferry Anggoro Ardy Nugroho, Shima Kadkhodazadeh,
Jakob B. Wagner, and Christoph Langhammer. Rationally Designed PdAuCu
Ternary Alloy Nanoparticles for Intrinsically Deactivation-Resistant Ultrafast
Plasmonic Hydrogen Sensing. ACS Sensors, page acssensors.9b00610, May
2019.
[23] Ferry A. A. Nugroho, Iwan Darmadi, Lucy Cusinato, Arturo Susarrey-Arce,
Herman Schreuders, Lars J. Bannenberg, Alice Bastos da Silva Fanta, Shima
Kadkhodazadeh, Jakob B. Wagner, Tomasz J. Antosiewicz, Anders Hell-
man, Vladimir P. Zhdanov, Bernard Dam, and Christoph Langhammer.
Metalpolymer hybrid nanomaterials for plasmonic ultrafast hydrogen detec-
tion. Nature Materials, 18(5):489495, May 2019.
[24] M. Slaman, B. Dam, H. Schreuders, and R. Griessen. Optimization of Mg-
based ber optic hydrogen detectors by alloying the catalyst. International
Journal of Hydrogen Energy, 33(3):10841089, February 2008.
[25] M. A. Butler. Optical ber hydrogen sensor. Applied Physics Letters,
45(10):10071009, November 1984. Publisher: American Institute of Physics.
[26] Michael A. Butler. Micromirror optical-ber hydrogen sensor. Sensors and
Actuators B: Chemical, 22(2):155163, November 1994.
[27] Casey P. O'Brien and Ivan C. Lee. The interaction of CO with PdCu hydrogen
separation membranes: An operando infrared spectroscopy study. Catalysis
Today, 336:216222, October 2019.
[28] Shohei Ogura, Michio Okada, and Katsuyuki Fukutani. Near-Surface Accumu-
lation of Hydrogen and CO Blocking Eects on a PdAu Alloy. The Journal
of Physical Chemistry C, 117(18):93669371, May 2013. Publisher: American
Chemical Society.
215
[29] G. Liang. Synthesis and hydrogen storage properties of Mg-based alloys. Jour-
nal of Alloys and Compounds, 370(1):123128, May 2004.
[30] M. Zhu, H. Wang, L. Z. Ouyang, and M. Q. Zeng. Composite structure
and hydrogen storage properties in Mg-base alloys. International Journal of
Hydrogen Energy, 31(2):251257, February 2006.
[31] C. Pistidda, N. Bergemann, J. Wurr, A. Rzeszutek, K. T. M?ller, B. R. S.
Hansen, S. Garroni, C. Horstmann, C. Milanese, A. Girella, O. Metz,
K. Taube, T. R. Jensen, D. Thomas, H. P. Liermann, T. Klassen, and M. Dorn-
heim. Hydrogen storage systems from waste Mg alloys. Journal of Power
Sources, 270:554563, December 2014.
[32] M. S. Wilson and S. Gottesfeld. Thin-lm catalyst layers for polymer elec-
trolyte fuel cell electrodes. Journal of Applied Electrochemistry, 22(1):17,
January 1992.
[33] Jakob Kibsgaard, Zhebo Chen, Benjamin N. Reinecke, and Thomas F.
Jaramillo. Engineering the surface structure of MoS2 to preferentially ex-
pose active edge sites for electrocatalysis. Nature Materials, 11(11):963969,
November 2012.
[34] Peter Strasser, Shirlaine Koh, Toyli Anniyev, Je Greeley, Karren More,
Chengfei Yu, Zengcai Liu, Sarp Kaya, Dennis Nordlund, Hirohito Ogasawara,
Michael F. Toney, and Anders Nilsson. Lattice-strain control of the activity in
dealloyed coreshell fuel cell catalysts. Nature Chemistry, 2(6):454460, June
2010.
[35] J. B Bates, N. J Dudney, B Neudecker, A Ueda, and C. D Evans. Thin-lm
lithium and lithium-ion batteries. Solid State Ionics, 135(1):3345, November
2000.
[36] Nian Liu, Hui Wu, Matthew T. McDowell, Yan Yao, Chongmin Wang, and
Yi Cui. A Yolk-Shell Design for Stabilized and Scalable Li-Ion Battery Alloy
Anodes. Nano Letters, 12(6):33153321, June 2012.
[37] Mark S. Gudiksen, Lincoln J. Lauhon, Jianfang Wang, David C. Smith, and
Charles M. Lieber. Growth of nanowire superlattice structures for nanoscale
photonics and electronics. Nature, 415(6872):617620, February 2002.
[38] Kenji Nomura, Hiromichi Ohta, Kazushige Ueda, Toshio Kamiya, Masahiro
Hirano, and Hideo Hosono. Thin-Film Transistor Fabricated in Single-
Crystalline Transparent Oxide Semiconductor. Science, 300(5623):12691272,
May 2003.
[39] Andrew N. Shipway, Eugenii Katz, and Itamar Willner. Nanoparticle Arrays
on Surfaces for Electronic, Optical, and Sensor Applications. ChemPhysChem,
1(1):1852, August 2000.
216
[40] G?nter Sauerbrey. Verwendung von Schwingquarzen zur W?gung d?nner
Schichten und zur Mikrow?gung. Zeitschrift f?r Physik, 155(2):206222, April
1959.
[41] Claudia Steinem and Andreas Jansho, editors. Piezoelectric Sensors.
Springer Series on Chemical Sensors and Biosensors. Springer-Verlag, Berlin
Heidelberg, 2007.
[42] Ilya Reviakine, Diethelm Johannsmann, and Ralf P. Richter. Hearing What
You Cannot See and Visualizing What You Hear: Interpreting Quartz
Crystal Microbalance Data from Solvated Interfaces. Analytical Chemistry,
83(23):88388848, December 2011.
[43] A. P. M. Glassford. Response of a quartz crystal microbalance to a liquid de-
posit. Journal of Vacuum Science and Technology, 15(6):18361843, November
1978.
[44] K. Keiji. Kanazawa and Joseph G. Gordon. Frequency of a quartz microbal-
ance in contact with liquid. Analytical Chemistry, 57(8):17701771, July 1985.
[45] K Keiji Kanazawa and Joseph G Gordon. The oscillation frequency of a quartz
resonator in contact with liquid. Analytica Chimica Acta, 175:99105, January
1985.
[46] E. P. EerNisse. Simultaneous Thin-Film Stress and Mass-Change Measure-
ments Using Quartz Resonators. Journal of Applied Physics, 43(4):13301337,
April 1972.
[47] E. P. EerNisse. Extension of the double resonator technique. Journal of
Applied Physics, 44(10):44824485, October 1973.
[48] T. P. Leervad Pedersen, C. Liesch, C. Salinga, T. Eleftheriadis, H. Weis, and
M. Wuttig. Hydrogen-induced changes of mechanical stress and optical trans-
mission in thin Pd lms. Thin Solid Films, 458(1):299303, June 2004.
[49] Vijay A. Sethuraman, Michael J. Chon, Maxwell Shimshak, Venkat Srinivasan,
and Pradeep R. Guduru. In situ measurements of stress evolution in silicon
thin lms during electrochemical lithiation and delithiation. Journal of Power
Sources, 195(15):50625066, August 2010.
[50] Markus Schwind, Saman Hosseinpour, Christoph Langhammer, Igor Zori?,
Christofer Leygraf, and Bengt Kasemo. Nanoplasmonic Sensing for Monitoring
the Initial Stages of Atmospheric Corrosion of Cu Nanodisks and Thin Films.
Journal of The Electrochemical Society, 160(10):C487C492, January 2013.
[51] Markus Schwind, Christoph Langhammer, Bengt Kasemo, and Igor Zori?.
Nanoplasmonic sensing and QCM-D as ultrasensitive complementary tech-
niques for kinetic corrosion studies of aluminum nanoparticles. Applied Surface
Science, 257(13):56795687, April 2011.
217
[52] Allan L. Smith and Hamid. M. Shirazi. Principles of quartz crystal microbal-
ance/heat conduction calorimetry: Measurement of the sorption enthalpy of
hydrogen in palladium. Thermochimica Acta, 432(2):202211, July 2005.
[53] Christoph Langhammer, Elin M. Larsson, Bengt Kasemo, and Igor Zori?. Indi-
rect Nanoplasmonic Sensing: Ultrasensitive Experimental Platform for Nano-
materials Science and Optical Nanocalorimetry. Nano Letters, 10(9):3529
3538, September 2010.
[54] Stephen J. Martin, Victoria Edwards. Gransta, and Gregory C. Frye. Charac-
terization of a quartz crystal microbalance with simultaneous mass and liquid
loading. Analytical Chemistry, 63(20):22722281, October 1991.
[55] J. Kestin, S. T. Ro, and W. A. Wakeham. Viscosity of the isotopes of hy-
drogen and their intermolecular force potentials. Journal of the Chemical
Society, Faraday Transactions 1: Physical Chemistry in Condensed Phases,
68(0):23162323, January 1972.
[56] Liang Tan, Xue'en Jia, Xiangfu Jiang, Youyu Zhang, Hao Tang, Shouzhuo
Yao, and Qingji Xie. In vitro study on the individual and synergistic cyto-
toxicity of adriamycin and selenium nanoparticles against Bel7402 cells with a
quartz crystal microbalance. Biosensors and Bioelectronics, 24(7):22682272,
March 2009.
[57] R. V. Bucur, V. Mecea, and T. B. Flanagan. The kinetics of hydrogen (deu-
terium) sorption by thin palladium layers studied with a piezoelectric quartz
crystal microbalance. Surface Science, 54(2):477488, February 1976.
[58] R. Feenstra, D. G. de Groot, J. H. Rector, E. Salomons, and R. Griessen.
Gravimetrical determination of pressure-composition isotherms of thin PdH c
lms. Journal of Physics F: Metal Physics, 16(12):1953, 1986.
[59] George Sidebotham. Heat Transfer Modeling: An Inductive Approach.
Springer International Publishing, 2015.
[60] J. E. Schirber and B. Morosin. Lat-
tice constants of $\ensuremath{\beta}\ensuremath{-
}\mathrm{P}\mathrm{d}{\mathrm{H}}_{x}$ and
$\ensuremath{\beta}\ensuremath{-}\mathrm{P}\mathrm{d}{\mathrm{D}}_{x}$
with $x$ near 1.0. Physical Review B, 12(1):117118, July 1975.
[61] Long Ju, Baisong Geng, Jason Horng, Caglar Girit, Michael Martin, Zhao
Hao, Hans A. Bechtel, Xiaogan Liang, Alex Zettl, Y. Ron Shen, and Feng
Wang. Graphene plasmonics for tunable terahertz metamaterials. Nature
Nanotechnology, 6(10):630634, October 2011.
[62] A. V. Kabashin, P. Evans, S. Pastkovsky, W. Hendren, G. A. Wurtz, R. Atkin-
son, R. Pollard, V. A. Podolskiy, and A. V. Zayats. Plasmonic nanorod meta-
materials for biosensing. Nature Materials, 8(11):867871, November 2009.
218
[63] Zhichuan Xu, Yanglong Hou, and Shouheng Sun. Magnetic Core/Shell
Fe3O4/Au and Fe3O4/Au/Ag Nanoparticles with Tunable Plasmonic Prop-
erties. Journal of the American Chemical Society, 129(28):86988699, July
2007.
[64] Andrea Tao, Prasert Sinsermsuksakul, and Peidong Yang. Tunable plasmonic
lattices of silver nanocrystals. Nature Nanotechnology, 2(7):435440, July
2007.
[65] F. J. Beck, A. Polman, and K. R. Catchpole. Tunable light trapping for
solar cells using localized surface plasmons. Journal of Applied Physics,
105(11):114310, June 2009.
[66] A. T. M. van Gogh, D. G. Nagengast, E. S. Kooij, N. J. Koeman, J. H.
Rector, R. Griessen, C. F. J. Flipse, and R. J. J. G. A. M. Smeets. Struc-
tural, electrical, and optical properties of ${\mathrm{La}}_{1\ensuremath{-
}z}{\mathrm{Y}}_{z}{\mathrm{H}}_{x}$ switchable mirrors. Physical
Review B, 63(19):195105, April 2001.
[67] Victor W. Brar, Min Seok Jang, Michelle Sherrott, Josue J. Lopez, and
Harry A. Atwater. Highly Conned Tunable Mid-Infrared Plasmonics in
Graphene Nanoresonators. Nano Letters, 13(6):25412547, June 2013.
[68] Ewald Veleckis and Russell Keith Edwards. Thermodynamic properties in the
systems vanadium-hydrogen, niobium-hydrogen, and tantalum-hydrogen. The
Journal of Physical Chemistry, 73(3):683692, March 1969.
[69] Rebecca S. Sherbo, Marta Moreno-Gonzalez, Noah J. J. Johnson, David J.
Dvorak, David K. Fork, and Curtis P. Berlinguette. Accurate Coulometric
Quantication of Hydrogen Absorption in Palladium Nanoparticles and Thin
Films. Chemistry of Materials, 30(12):39633970, June 2018.
[70] J Huot, G Liang, S Boily, A Van Neste, and R Schulz. Structural study
and hydrogen sorption kinetics of ball-milled magnesium hydride. Journal of
Alloys and Compounds, 293-295:495500, December 1999.
[71] A. San-Martin and F. D. Manchester. The H-Ti (Hydrogen-Titanium) system.
Bulletin of Alloy Phase Diagrams, 8(1):3042, February 1987.
[72] Nikolai Strohfeldt, Jun Zhao, Andreas Tittl, and Harald Giessen. Sensitivity
engineering in direct contact palladium-gold nano-sandwich hydrogen sensors
[Invited]. Optical Materials Express, 5(11):25252535, November 2015.
[73] Murat Serhatlioglu, Sencer Ayas, Necmi Biyikli, Aykutlu Dana, and
Mehmet E. Solmaz. Perfectly absorbing ultra thin interference coatings for
hydrogen sensing. Optics Letters, 41(8):17241727, April 2016.
219
[74] Carl Wadell, Ferry Anggoro Ardy Nugroho, Emil Lidstr?m, Beniamino
Iandolo, Jakob B. Wagner, and Christoph Langhammer. Hysteresis-Free
Nanoplasmonic PdAu Alloy Hydrogen Sensors. Nano Letters, 15(5):3563
3570, May 2015.
[75] Petr Tobi?ka, Olivier Hugon, Alain Trouillet, and Henri Gagnaire. An in-
tegrated optic hydrogen sensor based on SPR on palladium. Sensors and
Actuators B: Chemical, 74(1):168172, April 2001.
[76] Z. Zhao, M. A. Carpenter, H. Xia, and D. Welch. All-optical hydrogen sensor
based on a high alloy content palladium thin lm. Sensors and Actuators B:
Chemical, 113(1):532538, January 2006.
[77] Timur Shegai, Peter Johansson, Christoph Langhammer, and Mikael K?ll.
Directional Scattering and Hydrogen Sensing by Bimetallic PdAu Nanoan-
tennas. Nano Letters, 12(5):24642469, May 2012.
[78] Joel Villatoro and David Monz?n-Hern?ndez. Fast detection of hydrogen with
nano ber tapers coated with ultra thin palladium layers. Optics Express,
13(13):50875092, June 2005.
[79] David Monz?n-Hern?ndez, Donato Luna-Moreno, and Dalia Mart?nez-
Escobar. Fast response ber optic hydrogen sensor based on palladium and
gold nano-layers. Sensors and Actuators B: Chemical, 136(2):562566, March
2009.
[80] Xiaoyang Duan, Simon Kamin, and Na Liu. Dynamic plasmonic colour display.
Nature Communications, 8(1):19, February 2017.
[81] C. Boelsma, L. J. Bannenberg, M. J. van Setten, N.-J. Steinke, A. A. van
Well, and B. Dam. Hafniuman optical hydrogen sensor spanning six orders
in pressure. Nature Communications, 8(1):18, June 2017.
[82] D. E. Azofeifa, N. Clark, W. E. Vargas, H. Sol?s, G. K. P?lsson, and B. Hj?r-
varsson. Hydrogen induced changes in the optical properties of Pd capped V
thin lms. Journal of Alloys and Compounds, 580:S114S118, December 2013.
[83] M. J. van Setten, V. A. Popa, G. A. de Wijs, and G. Brocks. Electronic
structure and optical properties of lightweight metal hydrides. Physical Review
B, 75(3):035204, January 2007.
[84] J. Isidorsson, I. A. M. E. Giebels, H. Arwin, and R. Griessen. Optical proper-
ties of ${\mathrm{MgH}}_{2}$ measured in situ by ellipsometry and spec-
trophotometry. Physical Review B, 68(11):115112, September 2003.
[85] Daniel E. Azofeifa, Neville Clark, William E. Vargas, Hugo Solis, E. Aven-
dano, Michael Cambronero, and Diana Valverde-Mendez. Dielectric function
of palladium capped zirconium thin lms as a function of absorbed hydrogen.
International Journal of Hydrogen Energy, 42(35):2237322378, August 2017.
220
[86] Y. Yamada, K. Tajima, S. Bao, M. Okada, and K. Yoshimura. Hydro-
genation and dehydrogenation processes of palladium thin lms measured in
situ by spectroscopic ellipsometry. Solar Energy Materials and Solar Cells,
93(12):21432147, December 2009.
[87] William E. Vargas, Daniel E. Azofeifa, Neville Clark, Hugo Solis, Felipe Mon-
tealegre, and Michael Cambronero. Parametric formulation of the dielec-
tric function of palladium and palladium hydride thin lms. Applied Optics,
53(24):52945306, August 2014.
[88] Boyi Wang, Yong Zhu, Youping Chen, Han Song, Pengcheng Huang,
and Dzung Viet Dao. Hydrogen sensor based on palladium-yttrium alloy
nanosheet. Materials Chemistry and Physics, 194:231235, June 2017.
[89] Yogendra K. Gautam, Amit Sanger, Ashwani Kumar, and Ramesh Chandra.
A room temperature hydrogen sensor based on PdMg alloy and multilayers
prepared by magnetron sputtering. International Journal of Hydrogen Energy,
40(45):1554915555, December 2015.
[90] S. Bao, K. Tajima, Y. Yamada, M. Okada, and K. Yoshimura. Color-neutral
switchable mirrors based on magnesium-titanium thin lms. Applied Physics
A, 87(4):621624, June 2007.
[91] T. J. Richardson, J. L. Slack, B. Farangis, and M. D. Rubin. Mixed metal lms
with switchable optical properties. Applied Physics Letters, 80(8):13491351,
February 2002. Publisher: American Institute of Physics.
[92] K Yamamoto, K Higuchi, H Kajioka, H Sumida, S Orimo, and H Fujii. Op-
tical transmission of magnesium hydride thin lm with characteristic nanos-
tructure. Journal of Alloys and Compounds, 330-332:352356, January 2002.
[93] M. Singh, S. Srivastava, S. Agarwal, S. Kumar, and Y. K. Vijay. Optical prop-
erties of d.c. magneto sputtered tantalum and titanium nanostructure thin lm
metal hydrides. Bulletin of Materials Science, 33(5):569573, October 2010.
[94] Setsuo Nakao, Kazuo Saitoh, Tatsuya Hirahara, Masami Ikeyama, Masato
Tazawa, Ping Jin, Hiroaki Niwa, Seita Tanemura, Yoshiko Miyagawa, Soji
Miyagawa, and Kyoichiro Yasuda. Preparation and optical transmittance of
titanium hydride (deutende) lms by rf reactive sputtering. Thin Solid Films,
343-344:195198, April 1999.
[95] Joseph B. Murray, Kevin J. Palm, Tarun C. Narayan, David K. Fork, Seid
Sadat, and Jeremy N. Munday. Apparatus for combined nanoscale gravi-
metric, stress, and thermal measurements. Review of Scientic Instruments,
89(8):085106, August 2018.
[96] Z Luz, J Genossar, and P. S Rudman. Identication of the diusing atom in
MgH2. Journal of the Less Common Metals, 73(1):113118, September 1980.
221
[97] M. A. Pick, J. W. Davenport, Myron Strongin, and G. J. Dienes. Enhancement
of Hydrogen Uptake Rates for Nb and Ta by Thin Surface Overlayers. Physical
Review Letters, 43(4):286289, July 1979.
[98] Roberto Machorro, Jes?s M. Siqueiros, and Shu Wang. Optical properties of
Mg, from UV to IR, using ellipsometry and reectometry. Thin Solid Films,
269(1):15, November 1995.
[99] H.-J Hagemann, W Gudat, and C Kunz. Optical Constants from the Far
Infrared to the X-Ray Region: Mg, Al, Cu, Ag, Au, Bi, C, and Al2O3. Journal
of the Optical Society of America, 65(6):742744, June 1975.
[100] P. Hjort, A. Krozer, and B. Kasemo. Resistivity and hydrogen uptake mea-
surements in evaporated Mg lms at 350 K. Journal of Alloys and Compounds,
234(2):L11L15, February 1996.
[101] A. Baldi, M. Gonzalez-Silveira, V. Palmisano, B. Dam, and R. Griessen.
Destabilization of the Mg-H System through Elastic Constraints. Physical
Review Letters, 102(22):226102, June 2009.
[102] S. Singh, S. W. H. Eijt, M. W. Zandbergen, W. J. Legerstee, and V. L. Svetch-
nikov. Nanoscale structure and the hydrogenation of Pd-capped magnesium
thin lms prepared by plasma sputter and pulsed laser deposition. Journal of
Alloys and Compounds, 441(1):344351, August 2007.
[103] Marvin R. Querry. Optical Constants of Minerals and Other Materials from the
Millimeter to the Ultraviolet. Chemical Research, Development & Engineering
Center, U.S. Army Armament Munitions Chemical Command, 1987.
[104] P. B. Johnson and R. W. Christy. Optical constants of transition metals: Ti,
V, Cr, Mn, Fe, Co, Ni, and Pd. Physical Review B, 9(12):50565070, June
1974.
[105] William E. Wall, M. W. Ribarsky, and J. R. Stevenson. Optical properties of
titanium and titanium oxide surfaces. Journal of Applied Physics, 51(1):661
667, January 1980.
[106] Siham Mahmoud. Structure and optical properties of thin titanium lms
deposited on dierent substrates. Journal of Materials Science, 22(10):3693
3697, October 1987.
[107] Vitaliy Larionov, Shupeng Xu, and Maxim Syrtanov. Measurements of
hydrogenated titanium by electric methods. AIP Conference Proceedings,
1772(1):040005, October 2016.
[108] J. H. Weaver, D. J. Peterman, D. T. Peterson, and A. Franciosi. Elec-
tronic structure of metal hydrides. IV. $\mathrm{Ti}{\mathrm{H}}_{x}$,
$\mathrm{Zr}{\mathrm{H}}_{x}$, $\mathrm{Hf}{\mathrm{H}}_{x}$,
222
and the fcc-fct lattice distortion. Physical Review B, 23(4):16921698, Febru-
ary 1981.
[109] P. Romaniello, P. L. de Boeij, F. Carbone, and D. van der
Marel. Optical properties of bcc transition metals in the range
$040\phantom{\rule{0.3em}{0ex}}\mathrm{eV}$. Physical Review B,
73(7):075115, February 2006.
[110] D. G. Laurent, C. S. Wang, and J. Callaway. Energy bands, Compton pro-
le, and optical conductivity of vanadium. Physical Review B, 17(2):455461,
January 1978.
[111] Lisa J. Krayer, Elizabeth M. Tennyson, Marina S. Leite, and Jeremy N. Mun-
day. Near-IR Imaging Based on Hot Carrier Generation in Nanometer-Scale
Optical Coatings. ACS Photonics, 5(2):306311, February 2018.
[112] Igor Zoric?, Elin M. Larsson, Bengt Kasemo, and Christoph Langhammer. Lo-
calized Surface Plasmons Shed Light on Nanoscale Metal Hydrides. Advanced
Materials, 22(41):46284633, 2010.
[113] Mariama Rebello Sousa Dias, Chen Gong, Zackery A. Benson, and Marina S.
Leite. Lithography-Free, Omnidirectional, CMOS-Compatible AlCu Alloys for
Thin-Film Superabsorbers. Advanced Optical Materials, 6(2):1700830, 2018.
[114] Ramon Walter, Andreas Tittl, Audrey Berrier, Florian Sterl, Thomas Weiss,
and Harald Giessen. Large-Area Low-Cost Tunable Plasmonic Perfect Ab-
sorber in the Near Infrared by Colloidal Etching Lithography. Advanced Op-
tical Materials, 3(3):398403, 2015.
[115] Andreas Tittl, Patrick Mai, Richard Taubert, Daniel Dregely, Na Liu, and
Harald Giessen. Palladium-Based Plasmonic Perfect Absorber in the Visible
Wavelength Range and Its Application to Hydrogen Sensing. Nano Letters,
11(10):43664369, October 2011.
[116] Stefan Enoch, G?rard Tayeb, Pierre Sabouroux, Nicolas Gu?rin, and Patrick
Vincent. A Metamaterial for Directive Emission. Physical Review Letters,
89(21):213902, November 2002. Publisher: American Physical Society.
[117] N. Garcia, E. V. Ponizovskaya, and John Q. Xiao. Zero permittivity materials:
Band gaps at the visible. Applied Physics Letters, 80(7):11201122, February
2002. Publisher: American Institute of Physics.
[118] Richard W. Ziolkowski. Propagation in and scattering from a matched meta-
material having a zero index of refraction. Physical Review E, 70(4):046608,
October 2004. Publisher: American Physical Society.
[119] Lisa J. Krayer, Jongbum Kim, Joseph L. Garrett, and Jeremy N. Mun-
day. Optoelectronic Devices on Index-near-Zero Substrates. ACS Photonics,
6(9):22382244, September 2019. Publisher: American Chemical Society.
223
[120] Lisa J. Krayer, Jongbum Kim, and Jeremy N. Munday. Near-perfect absorp-
tion throughout the visible using ultra-thin metal lms on index-near-zero
substrates [Invited]. Optical Materials Express, 9(1):330338, January 2019.
Publisher: Optical Society of America.
[121] S. Vassant, A. Archambault, F. Marquier, F. Pardo, U. Gennser, A. Cavanna,
J. L. Pelouard, and J. J. Greet. Epsilon-Near-Zero Mode for Active Opto-
electronic Devices. Physical Review Letters, 109(23):237401, December 2012.
Publisher: American Physical Society.
[122] Viet Cuong Nguyen, Lang Chen, and Klaus Halterman. Total Transmission
and Total Reection by Zero Index Metamaterials with Defects. Physical Re-
view Letters, 105(23):233908, December 2010. Publisher: American Physical
Society.
[123] Jura Rensberg, You Zhou, Steen Richter, Chenghao Wan, Shuyan Zhang,
Philipp Sch?ppe, R?diger Schmidt-Grund, Shriram Ramanathan, Federico
Capasso, Mikhail A. Kats, and Carsten Ronning. Epsilon-Near-Zero Substrate
Engineering for Ultrathin-Film Perfect Absorbers. Physical Review Applied,
8(1):014009, July 2017. Publisher: American Physical Society.
[124] Simin Feng and Klaus Halterman. Coherent perfect absorption in epsilon-
near-zero metamaterials. Physical Review B, 86(16):165103, October 2012.
Publisher: American Physical Society.
[125] Klaus Halterman and Simin Feng. Resonant transmission of electromag-
netic elds through subwavelength zero-epsilon slits. Physical Review A,
78(2):021805, August 2008. Publisher: American Physical Society.
[126] Andrea Al?, M?rio G. Silveirinha, and Nader Engheta. Transmission-line
analysis of $\ensuremath{\epsilon}$-near-zerolled narrow channels. Physi-
cal Review E, 78(1):016604, July 2008. Publisher: American Physical Society.
[127] Brian Edwards, Andrea Al?, Michael E. Young, M?rio Silveirinha, and Nader
Engheta. Experimental Verication of Epsilon-Near-Zero Metamaterial Cou-
pling and Energy Squeezing Using a Microwave Waveguide. Physical Review
Letters, 100(3):033903, January 2008. Publisher: American Physical Society.
[128] D. C. Adams, S. Inampudi, T. Ribaudo, D. Slocum, S. Vangala, N. A. Kuhta,
W. D. Goodhue, V. A. Podolskiy, and D. Wasserman. Funneling Light through
a Subwavelength Aperture with Epsilon-Near-Zero Materials. Physical Review
Letters, 107(13):133901, September 2011. Publisher: American Physical Soci-
ety.
[129] Antonio Capretti, Yu Wang, Nader Engheta, and Luca Dal Negro. Enhanced
third-harmonic generation in Si-compatible epsilon-near-zero indium tin oxide
nanolayers. Optics Letters, 40(7):15001503, April 2015. Publisher: Optical
Society of America.
224
[130] Christos Argyropoulos, Pai-Yen Chen, Giuseppe D'Aguanno, Nader Engheta,
and Andrea Al?. Boosting optical nonlinearities in $\ensuremath{\epsilon}$-
near-zero plasmonic channels. Physical Review B, 85(4):045129, January 2012.
Publisher: American Physical Society.
[131] Haim Suchowski, Kevin O'Brien, Zi Jing Wong, Alessandro Salandrino, Xi-
aobo Yin, and Xiang Zhang. Phase MismatchFree Nonlinear Propagation in
Optical Zero-Index Materials. Science, 342(6163):12231226, December 2013.
Publisher: American Association for the Advancement of Science Section: Re-
port.
[132] M. Zahirul Alam, Israel De Leon, and Robert W. Boyd. Large optical
nonlinearity of indium tin oxide in its epsilon-near-zero region. Science,
352(6287):795797, May 2016. Publisher: American Association for the Ad-
vancement of Science Section: Report.
[133] Daniel Traviss, Roman Bruck, Ben Mills, Martina Abb, and Otto L. Muskens.
Ultrafast plasmonics using transparent conductive oxide hybrids in the epsilon-
near-zero regime. Applied Physics Letters, 102(12):121112, March 2013. Pub-
lisher: American Institute of Physics.
[134] Jongbum Kim, Aveek Dutta, Gururaj V. Naik, Alexander J. Giles, Francisco J.
Bezares, Chase T. Ellis, Joseph G. Tischler, Ahmed M. Mahmoud, Humeyra
Caglayan, Orest J. Glembocki, Alexander V. Kildishev, Joshua D. Caldwell,
Alexandra Boltasseva, and Nader Engheta. Role of epsilon-near-zero sub-
strates in the optical response of plasmonic antennas. Optica, 3(3):339346,
March 2016. Publisher: Optical Society of America.
[135] Zhizhen Ma, Zhuoran Li, Ke Liu, Chenran Ye, and Volker J. Sorger. Indium-
Tin-Oxide for High-performance Electro-optic Modulation. Nanophotonics,
-1(open-issue):198213, June 2015. Publisher: De Gruyter Section: Nanopho-
tonics.
[136] Gururaj V. Naik, Jongbum Kim, and Alexandra Boltasseva. Oxides and ni-
trides as alternative plasmonic materials in the optical range [Invited]. Optical
Materials Express, 1(6):10901099, October 2011. Publisher: Optical Society
of America.
[137] Salvatore Campione, Iltai Kim, Domenico de Ceglia, Gordon A. Keeler, and
Ting S. Luk. Experimental verication of epsilon-near-zero plasmon polari-
ton modes in degenerately doped semiconductor nanolayers. Optics Express,
24(16):1878218789, August 2016. Publisher: Optical Society of America.
[138] Salvatore Campione, Joel R. Wendt, Gordon A. Keeler, and Ting S. Luk.
Near-Infrared Strong Coupling between Metamaterials and Epsilon-near-Zero
Modes in Degenerately Doped Semiconductor Nanolayers. ACS Photonics,
3(2):293297, February 2016. Publisher: American Chemical Society.
225
[139] N. Kinsey, C. DeVault, J. Kim, M. Ferrera, V. M. Shalaev, and A. Boltas-
seva. Epsilon-near-zero Al-doped ZnO for ultrafast switching at telecom wave-
lengths. Optica, 2(7):616622, July 2015. Publisher: Optical Society of Amer-
ica.
[140] Francisco Javier Gonzalez, Robert E. Peale, Sepehr Benis, David Hagan, and
Eric Van Stryland. Optical Limiter using Epsilon-Near-Zero Grating. In
2019 IEEE Research and Applications of Photonics in Defense Conference
(RAPID), pages 13, August 2019.
[141] Farnood Khalilzadeh-Rezaie, Isaiah O. Oladeji, Justin W. Cleary, Nima Nader,
Janardan Nath, Imen Rezadad, and Robert E. Peale. Fluorine-doped tin
oxides for mid-infrared plasmonics. Optical Materials Express, 5(10):2184
2192, October 2015. Publisher: Optical Society of America.
[142] Hongwei Zhao, Yu Wang, Antonio Capretti, Luca Dal Negro, and Jonathan
Klamkin. Broadband Electroabsorption Modulators Design Based on Epsilon-
Near-Zero Indium Tin Oxide. IEEE Journal of Selected Topics in Quantum
Electronics, 21(4):192198, July 2015.
[143] Thomas E. Tiwald, Daniel W. Thompson, John A. Woollam, Wayne Paulson,
and Robert Hance. Application of IR variable angle spectroscopic ellipsometry
to the determination of free carrier concentration depth proles. Thin Solid
Films, 313-314:661666, February 1998.
[144] G. E. Jellison and F. A. Modine. Parameterization of the optical functions
of amorphous materials in the interband region. Applied Physics Letters,
69(3):371373, July 1996. Publisher: American Institute of Physics.
[145] A. Baldi, D. M. Borsa, H. Schreuders, J. H. Rector, T. Atmakidis, M. Bakker,
H. A. Zondag, W. G. J. van Helden, B. Dam, and R. Griessen. MgTiH thin
lms as switchable solar absorbers. International Journal of Hydrogen Energy,
33(12):31883192, June 2008.
[146] Andreas Tittl, Ann-Katrin U. Michel, Martin Sch?ferling, Xinghui
Yin, Behrad Gholipour, Long Cui, Matthias Wuttig, Thomas Taub-
ner, Frank Neubrech, and Harald Giessen. A Switchable Mid-
Infrared Plasmonic Perfect Absorber with Multispectral Thermal Imag-
ing Capability. Advanced Materials, 27(31):45974603, 2015. _eprint:
https://onlinelibrary.wiley.com/doi/pdf/10.1002/adma.201502023.
[147] Soroosh Daqiqeh Rezaei, Zhaogang Dong, John You En Chan, Jonathan
Trisno, Ray Jia Hong Ng, Qifeng Ruan, Cheng-Wei Qiu, N. Asger Mortensen,
and Joel K.W. Yang. Nanophotonic Structural Colors. ACS Photonics, July
2020. Publisher: American Chemical Society.
226
[148] Carl H?gglund, S. Peter Apell, and Bengt Kasemo. Maximized Optical
Absorption in Ultrathin Films and Its Application to Plasmon-Based Two-
Dimensional Photovoltaics. Nano Letters, 10(8):31353141, August 2010. Pub-
lisher: American Chemical Society.
[149] N. Ahmad, J. Stokes, N. A. Fox, M. Teng, and M. J. Cryan. Ultra-thin metal
lms for enhanced solar absorption. Nano Energy, 1(6):777782, November
2012.
[150] Mark L. Brongersma, Naomi J. Halas, and Peter Nordlander. Plasmon-
induced hot carrier science and technology. Nature Nanotechnology, 10(1):25
34, January 2015. Number: 1 Publisher: Nature Publishing Group.
[151] Tao Gong and Jeremy N. Munday. Materials for hot carrier plasmonics [In-
vited]. Optical Materials Express, 5(11):25012512, November 2015. Publisher:
Optical Society of America.
[152] Tao Gong and Jeremy N. Munday. Angle-Independent Hot Carrier Gener-
ation and Collection Using Transparent Conducting Oxides. Nano Letters,
15(1):147152, January 2015. Publisher: American Chemical Society.
[153] Tao Gong and Jeremy N. Munday. Aluminum-based hot carrier plasmonics.
Applied Physics Letters, 110(2):021117, January 2017. Publisher: American
Institute of Physics.
[154] Joseph Murray, Dakang Ma, and Jeremy N. Munday. Electrically Controllable
Light Trapping for Self-Powered Switchable Solar Windows. ACS Photonics,
4(1):17, January 2017. Publisher: American Chemical Society.
[155] Bo Zhu, Yijun Feng, Junming Zhao, Ci Huang, and Tian Jiang. Switchable
metamaterial reector/absorber for dierent polarized electromagnetic waves.
Applied Physics Letters, 97(5):051906, August 2010. Publisher: American
Institute of Physics.
[156] Dongju Lee, Heijun Jeong, and Sungjoon Lim. Electronically Switchable
Broadband Metamaterial Absorber. Scientic Reports, 7(1):4891, July 2017.
Number: 1 Publisher: Nature Publishing Group.
[157] David Shrekenhamer, Wen-Chen Chen, and Willie J. Padilla. Liquid Crystal
Tunable Metamaterial Absorber. Physical Review Letters, 110(17):177403,
April 2013. Publisher: American Physical Society.
[158] Prakash Pitchappa, Chong Pei Ho, Piotr Kropelnicki, Navab Singh, Dim-
Lee Kwong, and Chengkuo Lee. Micro-electro-mechanically switchable near
infrared complementary metamaterial absorber. Applied Physics Letters,
104(20):201114, May 2014. Publisher: American Institute of Physics.
227
[159] Sukosin Thongrattanasiri, Frank H. L. Koppens, and F. Javier Garc?a de
Abajo. Complete Optical Absorption in Periodically Patterned Graphene.
Physical Review Letters, 108(4):047401, January 2012. Publisher: American
Physical Society.
[160] Mikhail A. Kats, Romain Blanchard, Patrice Genevet, Zheng Yang, M. Mum-
taz Qazilbash, D. N. Basov, Shriram Ramanathan, and Federico Capasso.
Thermal tuning of mid-infrared plasmonic antenna arrays using a phase change
material. Optics Letters, 38(3):368370, February 2013. Publisher: Optical
Society of America.
[161] Hao Wang, Yue Yang, and Liping Wang. Switchable wavelength-selective and
diuse metamaterial absorber/emitter with a phase transition spacer layer.
Applied Physics Letters, 105(7):071907, August 2014. Publisher: American
Institute of Physics.
[162] Yi Zhao, Qiuping Huang, Honglei Cai, Xiaoxia Lin, and Yalin Lu. A broad-
band and switchable VO2-based perfect absorber at the THz frequency. Optics
Communications, 426:443449, November 2018.
[163] Tongling Wang, Yuping Zhang, Yuping Zhang, Huiyun Zhang, Maoyong Cao,
and Maoyong Cao. Dual-controlled switchable broadband terahertz absorber
based on a graphene-vanadium dioxide metamaterial. Optical Materials Ex-
press, 10(2):369386, February 2020. Publisher: Optical Society of America.
[164] Kevin J. Palm, Joseph B. Murray, Tarun C. Narayan, and Jeremy N. Munday.
Dynamic Optical Properties of Metal Hydrides. ACS Photonics, 5(11):4677
4686, November 2018.
[165] Kevin J. Palm, Joseph B. Murray, Joshua P. McClure, Marina S. Leite, and
Jeremy N. Munday. In Situ Optical and Stress Characterization of Alloyed
PdxAu1x Hydrides. ACS Applied Materials & Interfaces, 11(48):45057
45067, December 2019. Publisher: American Chemical Society.
[166] Reginald M. Penner. A Nose for Hydrogen Gas: Fast, Sensitive H2 Sen-
sors Using Electrodeposited Nanomaterials. Accounts of Chemical Research,
50(8):19021910, August 2017.
[167] Andrea Baldi, Lennard Mooij, Valerio Palmisano, Herman Schreuders, Gopi
Krishnan, Bart J. Kooi, Bernard Dam, and Ronald Griessen. Elastic ver-
sus Alloying Eects in Mg-Based Hydride Films. Physical Review Letters,
121(25):255503, December 2018.
[168] Mikhail A. Kats, Romain Blanchard, Patrice Genevet, and Federico Capasso.
Nanometre optical coatings based on strong interference eects in highly ab-
sorbing media. Nature Materials, 12(1):2024, January 2013. Number: 1
Publisher: Nature Publishing Group.
228
[169] C. Hilsum. Infrared Absorption of Thin Metal Films. JOSA, 44(3):188191,
March 1954. Publisher: Optical Society of America.
[170] B. Edwards, A. Al?, M. G. Silveirinha, and N. Engheta. Reectionless sharp
bends and corners in waveguides using epsilon-near-zero eects. Journal of
Applied Physics, 105(4):044905, February 2009. Publisher: American Institute
of Physics.
[171] A. M. Vredenberg, E. M. B. Heller, and D. O. Boerma. Hydriding character-
istics of FeTi/Pd lms. Journal of Alloys and Compounds, 400(1):188193,
September 2005.
[172] P. Hjort, A. Krozer, and B. Kasemo. Hydrogen sorption kinetics in partly
oxidized Mg lms. Journal of Alloys and Compounds, 237(1):7480, April
1996.
[173] Max Born and Emil Wolf. Principles of Optics. Cambridge University Press,
7th edition, 2005.
[174] P. B. Johnson and R. W. Christy. Optical Constants of the Noble Metals.
Physical Review B, 6(12):43704379, December 1972.
[175] Lihong Gao, Fabien Lemarchand, and Michel Lequime. Exploitation of mul-
tiple incidences spectrometric measurements for thin lm reverse engineering.
Optics Express, 20(14):1573415751, July 2012. Publisher: Optical Society of
America.
[176] J. Ryd?n, B. Hj?rvarsson, T. Ericsson, E. Karlsson, A. Krozer, and B. Kasemo.
Unusual kinetics of hydride formation in Mg-Pd sandwiches, studied by hy-
drogen proling and quartz crystal microbalance measurements. Journal of
the Less Common Metals, 152(2):295309, July 1989.
[177] I. R. Hooper and J. R. Sambles. Dispersion of surface plasmon polaritons
on short-pitch metal gratings. Physical Review B, 65(16):165432, April 2002.
Publisher: American Physical Society.
[178] Joshua P. McClure, Jonathan Boltersdorf, David R. Baker, Thomas G. Far-
inha, Nicholas Dzuricky, Cesar E. P. Villegas, Alexandre R. Rocha, and
Marina S. Leite. StructureProperty-Performance Relationship of Ultrathin
PdAu Alloy Catalyst Layers for Low-Temperature Ethanol Oxidation in Al-
kaline Media. ACS Applied Materials & Interfaces, May 2019.
[179] Chen Gong, Alan Kaplan, Zackery A. Benson, David R. Baker, Joshua P. Mc-
Clure, Alexandre R. Rocha, and Marina S. Leite. Band Structure Engineering
by Alloying for Photonics. Advanced Optical Materials, 6(17):1800218, 2018.
[180] Zhaoke Zheng, Takashi Tachikawa, and Tetsuro Majima. Plasmon-Enhanced
Formic Acid Dehydrogenation Using Anisotropic PdAu Nanorods Studied
229
at the Single-Particle Level. Journal of the American Chemical Society,
137(2):948957, January 2015.
[181] Mariama Rebello Sousa Dias and Marina S. Leite. Alloying: A Platform for
Metallic Materials with On-Demand Optical Response. Accounts of Chemical
Research, July 2019.
[182] Xiaoyang Duan, Simon Kamin, Florian Sterl, Harald Giessen, and Na Liu.
Hydrogen-Regulated Chiral Nanoplasmonics. Nano Letters, 16(2):14621466,
February 2016.
[183] T. J. Richardson, J. L. Slack, R. D. Armitage, R. Kostecki, B. Farangis, and
M. D. Rubin. Switchable mirrors based on nickelmagnesium lms. Applied
Physics Letters, 78(20):30473049, May 2001.
[184] D. M. Borsa, A. Baldi, M. Pasturel, H. Schreuders, B. Dam, R. Griessen,
P. Vermeulen, and P. H. L. Notten. MgTiH thin lms for smart solar col-
lectors. Applied Physics Letters, 88(24):241910, June 2006.
[185] U.S. Department of Energy, Fuel Cell Technologies Ofce., and Energy Ef-
ciency and Renewable Energy (EERE). Multi-Year Research, Development,
and Demonstration Plan, 20112020. Section 3.7 Hydrogen Safety, Codes and
Standards. 2015.
[186] Brian D. Adams and Aicheng Chen. The role of palladium in a hydrogen
economy. Materials Today, 14(6):282289, June 2011.
[187] R. B. Schwarz and A. G. Khachaturyan. Thermodynamics of open two-phase
systems with coherent interfaces: Application to metalhydrogen systems.
Acta Materialia, 54(2):313323, January 2006.
[188] Lars Johannes Bannenberg, Ferry Anggoro Ardy Nugroho, Herman Schreud-
ers, Ben Norder, Thu Trang Trinh, Nina-Juliane Steinke, Ad A. van Well,
Christoph Langhammer, and Bernard Dam. Direct Comparison of PdAu Al-
loy Thin Films and Nanoparticles upon Hydrogen Exposure. ACS Applied
Materials & Interfaces, 11(17):1548915497, May 2019.
[189] Ferry Anggoro Ardy Nugroho, Iwan Darmadi, Vladimir P. Zhdanov, and
Christoph Langhammer. Universal Scaling and Design Rules of Hydrogen-
Induced Optical Properties in Pd and Pd-Alloy Nanoparticles. ACS Nano,
12(10):99039912, October 2018.
[190] R. J. Westerwaal, J. S. A. Rooijmans, L. Leclercq, D. G. Gheorghe, T. Radeva,
L. Mooij, T. Mak, L. Polak, M. Slaman, B. Dam, and Th. Rasing. Nanostruc-
tured PdAu based ber optic sensors for probing hydrogen concentrations in
gas mixtures. International Journal of Hydrogen Energy, 38(10):42014212,
April 2013.
230
[191] Yoshiaki Nishijima, Shogo Shimizu, Keisuke Kurihara, Yoshikazu Hashimoto,
Hajime Takahashi, Armandas Bal?ytis, Gediminas Seniutinas, Shinji Okazaki,
Jurga Juodkazyte
, Takeshi Iwasa, Tetsuya Taketsugu, Yoriko Tominaga, and
Saulius Juodkazis. Optical readout of hydrogen storage in lms of Au and Pd.
Optics Express, 25(20):2408124092, October 2017.
[192] Lulu Zhang, Qiaowan Chang, Huimei Chen, and Minhua Shao. Recent ad-
vances in palladium-based electrocatalysts for fuel cell reactions and hydrogen
evolution reaction. Nano Energy, 29:198219, November 2016.
[193] Bernard Coq and Fran?ois Figueras. Bimetallic palladium catalysts: inuence
of the co-metal on the catalyst performance. Journal of Molecular Catalysis
A: Chemical, 173(1):117134, September 2001.
[194] Sarina Sarina, Huaiyong Zhu, Esa Jaatinen, Qi Xiao, Hongwei Liu, Jianfeng
Jia, Chao Chen, and Jian Zhao. Enhancing Catalytic Performance of Palla-
dium in Gold and Palladium Alloy Nanoparticles for Organic Synthesis Reac-
tions through Visible Light Irradiation at Ambient Temperatures. Journal of
the American Chemical Society, 135(15):57935801, April 2013.
[195] Kent E. Coulter, J. Douglas Way, Sabina K. Gade, Saurabh Chaudhari,
G?khan O. Alptekin, Sarah J. DeVoss, Stephen N. Paglieri, and Bill Pledger.
Sulfur tolerant PdAu and PdAuPt alloy hydrogen separation membranes.
Journal of Membrane Science, 405-406:1119, July 2012.
[196] Anthony Knapton. Palladium Alloys for Hydrogen Diusion Membranes.
1977.
[197] Stoney George Gerald and Parsons Charles Algernon. The tension of metal-
lic lms deposited by electrolysis. Proceedings of the Royal Society of Lon-
don. Series A, Containing Papers of a Mathematical and Physical Character,
82(553):172175, May 1909.
[198] Shima Kadkhodazadeh, Ferry Anggoro Ardy Nugroho, Christoph Langham-
mer, Marco Beleggia, and Jakob B. Wagner. Optical PropertyComposition
Correlation in Noble Metal Alloy Nanoparticles Studied with EELS. ACS
Photonics, 6(3):779786, March 2019.
[199] Yoshiaki Nishijima, Yoshikazu Hashimoto, Gediminas Seniutinas, Lorenzo
Rosa, and Saulius Juodkazis. Engineering gold alloys for plasmonics. Ap-
plied Physics A, 117(2):641645, November 2014.
[200] Arnulf Maeland and Ted B. Flanagan. X-Ray and Thermodynamic Studies
of the Absorption of Hydrogen by Gold-Palladium Alloys. The Journal of
Physical Chemistry, 69(10):35753581, October 1965.
[201] M. ?ukaszewski, K. Ku?mierczyk, J. Kotowski, H. Siwek, and A. Czerwi?ski.
Electrosorption of hydrogen into palladium-gold alloys. Journal of Solid State
Electrochemistry, 7(2):6976, February 2003.
231
[202] K. Hubkowska, M. ?ukaszewski, and A. Czerwi?ski. Inuence of tempera-
ture on hydrogen electrosorption into palladiumnoble metal alloys. Part 1:
Palladiumgold alloys. Electrochimica Acta, 56(1):235242, December 2010.
[203] Y. Pivak, H. Schreuders, M. Slaman, R. Griessen, and B. Dam. Thermody-
namics, stress release and hysteresis behavior in highly adhesive PdH lms.
International Journal of Hydrogen Energy, 36(6):40564067, March 2011.
[204] Stefan Wagner and Astrid Pundt. Quasi-thermodynamic model on hydride
formation in palladiumhydrogen thin lms: Impact of elastic and microstruc-
tural constraints. International Journal of Hydrogen Energy, 41(4):27272738,
January 2016.
[205] R. Gremaud, M. Gonzalez-Silveira, Y. Pivak, S. de Man, M. Slaman,
H. Schreuders, B. Dam, and R. Griessen. Hydrogenography of PdHx thin
lms: Inuence of H-induced stress relaxation processes. Acta Materialia,
57(4):12091219, February 2009.
[206] B. Ham, A. Junkaew, R. Arr?yave, J. Park, H. C. Zhou, D. Foley, S. Rios,
H. Wang, and X. Zhang. Size and stress dependent hydrogen desorption
in metastable Mg hydride lms. International Journal of Hydrogen Energy,
39(6):25972607, February 2014.
[207] C.-J. Chung, Sang-Chul Lee, James R. Groves, Edwin N. Brower, Robert
Sinclair, and Bruce M. Clemens. Interfacial Alloy Hydride Destabilization in
Mg/Pd Thin Films. Physical Review Letters, 108(10):106102, March 2012.
[208] M. Dornheim, A. Pundt, R. Kirchheim, S. J. v. d. Molen, E. S. Kooij, J. Kersse-
makers, R. Griessen, H. Harms, and U. Geyer. Stress development in thin
yttrium lms on hard substrates during hydrogen loading. Journal of Applied
Physics, 93(11):89588965, May 2003.
[209] Stefan Wagner, Philipp Klose, Vladimir Burlaka, Kai N?rthemann, Magnus
Hamm, and Astrid Pundt. Structural Phase Transitions in Niobium Hydrogen
Thin Films: Mechanical Stress, Phase Equilibria and Critical Temperatures.
ChemPhysChem, 20(14):18901904, 2019.
[210] Jianxiong Li, Simon Kamin, Guoxing Zheng, Frank Neubrech, Shuang Zhang,
and Na Liu. Addressable metasurfaces for dynamic holography and optical
information encryption. Science Advances, 4(6):eaar6768, June 2018.
[211] Thomas G. Farinha, Chen Gong, Zackery A. Benson, and Marina S. Leite.
Magnesium for Transient Photonics. ACS Photonics, 6(2):272278, February
2019.
[212] J. F. Stampfer, C. E. Holley, and J. F. Suttle. The Magnesium-Hydrogen
System1-3. Journal of the American Chemical Society, 82(14):35043508, July
1960. Publisher: American Chemical Society.
232
[213] C. M. Stander. Kinetics of decomposition of magnesium hydride. Journal of
Inorganic and Nuclear Chemistry, 39(2):221223, January 1977.
[214] A Reiser, B Bogdanovi?, and K Schlichte. The application of Mg-based metal-
hydrides as heat energy storage systems. International Journal of Hydrogen
Energy, 25(5):425430, May 2000.
[215] I. Gonz?lez Fern?ndez, F. C. Gennari, and G. O. Meyer. Inuence of sintering
parameters on formation of Mg-Co hydrides based on their thermodynamic
characterization. Journal of Alloys and Compounds, 462(1):119124, August
2008.
[216] Huaiyu Shao, Tong Liu, Yuntao Wang, Hairuo Xu, and Xingguo Li. Prepara-
tion of Mg-based hydrogen storage materials from metal nanoparticles. Jour-
nal of Alloys and Compounds, 465(1):527533, October 2008.
[217] G. F. Lima, A. M. Jorge, D. R. Leiva, C. S. Kiminami, C. Bolfarini, and W. J.
Botta. Severe plastic deformation of Mg-Fe powders to produce bulk hydrides.
Journal of Physics: Conference Series, 144:012015, January 2009. Publisher:
IOP Publishing.
[218] L. E. A. Berlouis, E. Cabrera, E. Hall-Barientos, P. J. Hall, S. B. Dodd,
S. Morris, and M. A. Imam. Thermal analysis investigation of hydriding
properties of nanocrystalline Mg-Ni- and Mg-Fe-based alloys prepared by high-
energy ball milling. Journal of Materials Research, 16(1):4557, January 2001.
[219] Yanshan Lu, Hyunjeong Kim, Kouji Sakaki, Shigenobu Hayashi, Keiko Jimura,
and Kohta Asano. Destabilizing the Dehydrogenation Thermodynamics of
Magnesium Hydride by Utilizing the Immiscibility of Mn with Mg. Inorganic
Chemistry, 58(21):1460014607, November 2019. Publisher: American Chem-
ical Society.
[220] R. Gremaud, A. Borgschulte, C. Chacon, J.L.M. van Mechelen, H. Schreuders,
A. Z?ttel, B. Hj?rvarsson, B. Dam, and R. Griessen. Structural and optical
properties of MgxAl1-xHy gradient thin lms: a combinatorial approach. Ap-
plied Physics A, 84(1):7785, July 2006.
[221] A. Zaluska, L. Zaluski, and J.O. Str?m-Olsen. Structure, catalysis and atomic
reactions on the nano-scale: a systematic approach to metal hydrides for hy-
drogen storage. Applied Physics A, 72(2):157165, February 2001.
[222] S. Bouaricha, J. P. Dodelet, D. Guay, J. Huot, S. Boily, and R. Schulz. Hydrid-
ing behavior of Mg-Al and leached Mg-Al compounds prepared by high-energy
ball-milling. Journal of Alloys and Compounds, 297(1):282293, February
2000.
[223] R. Gremaud, A. Borgschulte, W. Lohstroh, H. Schreuders, A. Z?ttel, B. Dam,
and R. Griessen. Ti-catalyzed Mg(AlH4)2: A reversible hydrogen storage
material. Journal of Alloys and Compounds, 404-406:775778, December 2005.
233
[224] H. Fritzsche, M. Saoudi, J. Haagsma, C. Ophus, E. Luber, C. T. Harrower, and
D. Mitlin. Neutron reectometry study of hydrogen desorption in destabilized
MgAl alloy thin lms. Applied Physics Letters, 92(12):121917, March 2008.
Publisher: American Institute of Physics.
[225] P. Vermeulen, P. C. J. Graat, H. J. Wondergem, and P. H. L. Notten. Crystal
structures of MgyTi100-y thin lm alloys in the as-deposited and hydrogenated
state. International Journal of Hydrogen Energy, 33(20):56465650, October
2008.
[226] Hyunjeong Kim, Herman Schreuders, Kouji Sakaki, Kohta Asano, Yu-
miko Nakamura, Naoyuki Maejima, Akihiko Machida, Tetsu Watanuki, and
Bernard Dam. Unveiling Nanoscale Compositional and Structural Hetero-
geneities of Highly Textured Mg0.7Ti0.3Hy Thin Films. Inorganic Chemistry,
59(10):68006807, May 2020. Publisher: American Chemical Society.
[227] A. Baldi, D. M. Borsa, H. Schreuders, J. H. Rector, T. Atmakidis, M. Bakker,
H. A. Zondag, W. G. J. van Helden, B. Dam, and R. Griessen. Mg-Ti-H thin
lms as switchable solar absorbers. International Journal of Hydrogen Energy,
33(12):31883192, June 2008.
[228] D. M. Borsa, A. Baldi, M. Pasturel, H. Schreuders, B. Dam, R. Griessen,
P. Vermeulen, and P. H. L. Notten. Mg-Ti-H thin lms for smart solar collec-
tors. Applied Physics Letters, 88(24):241910, June 2006. Publisher: American
Institute of PhysicsAIP.
[229] D. M. Borsa, R. Gremaud, A. Baldi, H. Schreuders, J. H. Rector, B. Kooi,
P. Vermeulen, P. H. L. Notten, B. Dam, and R. Griessen. Structural, opti-
cal, and electrical properties of MgyTi(1-y)Hx thin lms. Physical Review B,
75(20):205408, May 2007. Publisher: American Physical Society.
[230] R. a. H. Niessen and P. H. L. Notten. Electrochemical Hydrogen Storage
Characteristics of Thin Film MgX ( X = Sc , Ti , V , Cr ) Compounds.
Electrochemical and Solid-State Letters, 8(10):A534, August 2005. Publisher:
IOP Publishing.
[231] R. J. Westerwaal, A. Borgschulte, W. Lohstroh, B. Dam, B. Kooi, G. ten
Brink, M. J. P. Hopstaken, and P. H. L. Notten. The growth-induced mi-
crostructural origin of the optical black state of Mg2NiHx thin lms. Journal
of Alloys and Compounds, 416(1):210, June 2006.
[232] S. Orimo and H. Fujii. Materials science of Mg-Ni-based new hydrides. Applied
Physics A, 72(2):167186, February 2001.
[233] A. Ludwig, J. Cao, B. Dam, and R. Gremaud. Opto-mechanical characteriza-
tion of hydrogen storage properties of MgNi thin lm composition spreads.
Applied Surface Science, 254(3):682686, November 2007.
234
[234] Emil Johansson, Cyril Chacon, Claudia Zlotea, Yvonne Andersson, and
Bj?rgvin Hj?rvarsson. Hydrogen uptake and optical properties of sputtered
Mg-Ni thin lms. Journal of Physics: Condensed Matter, 16(43):76497662,
October 2004. Publisher: IOP Publishing.
[235] J. Isidorsson, I. a. M. E. Giebels, R. Griessen, and M. Di Vece. Tunable
reectance Mg-Ni-H lms. Applied Physics Letters, 80(13):23052307, March
2002. Publisher: American Institute of Physics.
[236] W. Lohstroh, R. J. Westerwaal, B. Noheda, S. Enache, I. A. M. E. Giebels,
B. Dam, and R. Griessen. Self-Organized Layered Hydrogenation in Black
Mg2NiHxSwitchable Mirrors. Physical Review Letters, 93(19):197404, Novem-
ber 2004. Publisher: American Physical Society.
[237] R. Gremaud, C. P. Broedersz, A. Borgschulte, M. J. van Setten, H. Schreuders,
M. Slaman, B. Dam, and R. Griessen. Hydrogenography of MgyNi1-yHx
gradient thin lms: Interplay between the thermodynamics and kinetics of
hydrogenation. Acta Materialia, 58(2):658668, January 2010.
[238] M. Pasturel, M. Slaman, D. M. Borsa, H. Schreuders, B. Dam, R. Griessen,
W. Lohstroh, and A. Borgschulte. Stabilized switchable black state in
Mg2NiH4-Ti-Pd thin lms for optical hydrogen sensing. Applied Physics Let-
ters, 89(2):021913, July 2006. Publisher: American Institute of Physics.
[239] Yevheniy Pivak, Valerio Palmisano, Herman Schreuders, and Bernard Dam.
The clamping eect in the complex hydride Mg2NiH4 thin lms. Journal
of Materials Chemistry A, 1(36):1097210978, August 2013. Publisher: The
Royal Society of Chemistry.
[240] E. Akiba, K. Nomura, S. Ono, and S. Suda. Kinetics of the reaction between
Mg-Ni alloys and H2. International Journal of Hydrogen Energy, 7(10):787
791, January 1982.
[241] Chengshang Zhou, Zhigang Zak Fang, Jun Lu, Xiangyi Luo, Chai Ren, Peng
Fan, Yang Ren, and Xiaoyi Zhang. Thermodynamic Destabilization of Magne-
sium Hydride Using Mg-Based Solid Solution Alloys. The Journal of Physical
Chemistry C, 118(22):1152611535, June 2014. Publisher: American Chemical
Society.
[242] P. Vermeulen, R. A. H. Niessen, and P. H. L. Notten. Hydrogen storage in
metastable MgyTi(1-y) thin lms. Electrochemistry Communications, 8(1):27
32, January 2006.
[243] S. Bao, K. Tajima, Y. Yamada, M. Okada, and K. Yoshimura. Magnesium-
titanium alloy thin-lm switchable mirrors. Solar Energy Materials and Solar
Cells, 92(2):224227, February 2008.
235
[244] R. Gremaud, J. L. M. van Mechelen, H. Schreuders, M. Slaman, B. Dam, and
R. Griessen. Structural and optical properties of MgyNi1-yHx gradient thin
lms in relation to the as-deposited metallic state. International Journal of
Hydrogen Energy, 34(21):89518957, November 2009.
[245] W. Lohstroh, R. J. Westerwaal, J. L. M. van Mechelen, H. Schreuders, B. Dam,
and R. Griessen. The dielectric function of Mgy NiHx thin lms (2<y<10).
Journal of Alloys and Compounds, 430(1):1318, March 2007.
[246] R. J. King and S. P. Talim. A Comparison of Thin Film Measurement by
Guided Waves, Ellipsometry and Reectometry. Optica Acta: International
Journal of Optics, 28(8):11071123, August 1981. Publisher: Taylor & Francis
_eprint: https://doi.org/10.1080/713820674.
[247] M. Slaman, B. Dam, M. Pasturel, D. M. Borsa, H. Schreuders, J. H. Rector,
and R. Griessen. Fiber optic hydrogen detectors containing Mg-based metal
hydrides. Sensors and Actuators B: Chemical, 123(1):538545, April 2007.
[248] B Farangis, P Nachimuthu, T. J Richardson, J. L Slack, B. K Meyer, R. C. C
Perera, and M. D Rubin. Structural and electronic properties of magnesium-
3D transition metal switchable mirrors. Solid State Ionics, 165(1):309314,
December 2003.
[249] Easo P. George, Dierk Raabe, and Robert O. Ritchie. High-entropy alloys.
Nature Reviews Materials, 4(8):515534, August 2019. Number: 8 Publisher:
Nature Publishing Group.
[250] Martin Sahlberg, Dennis Karlsson, Claudia Zlotea, and Ulf Jansson. Supe-
rior hydrogen storage in high entropy alloys. Scientic Reports, 6(1):36770,
November 2016. Number: 1 Publisher: Nature Publishing Group.
[251] Magnus Moe Nyg?rd, Gustav Ek, Dennis Karlsson, Martin Sahlberg, Mag-
nus H. S?rby, and Bj?rn C. Hauback. Hydrogen storage in high-entropy alloys
with varying degree of local lattice strain. International Journal of Hydrogen
Energy, 44(55):2914029149, November 2019.
[252] R. B. Strozi, D. R. Leiva, J. Huot, W. J. Botta, and G. Zepon. Synthesis and
hydrogen storage behavior of MgVAlCrNi high entropy alloys. Interna-
tional Journal of Hydrogen Energy, 46(2):23512361, January 2021.
[253] Cheng Zhang, Anna Song, Yuan Yuan, YuanWu, Peilong Zhang, Zhaoping Lu,
and Xiping Song. Study on the hydrogen storage properties of a TiZrNbTa
high entropy alloy. International Journal of Hydrogen Energy, 45(8):5367
5374, February 2020.
[254] Yonggang Yao, Zhennan Huang, Pengfei Xie, Steven D. Lacey, Rohit Jiji Ja-
cob, Hua Xie, Fengjuan Chen, Anmin Nie, Tiancheng Pu, Miles Rehwoldt,
Daiwei Yu, Michael R. Zachariah, Chao Wang, Reza Shahbazian-Yassar,
236
Ju Li, and Liangbing Hu. Carbothermal shock synthesis of high-entropy-alloy
nanoparticles. Science, 359(6383):14891494, March 2018. Publisher: Ameri-
can Association for the Advancement of Science Section: Research Article.
[255] T. Schenkel, A. Persaud, H. Wang, P. A. Seidl, R. MacFadyen, C. Nel-
son, W. L. Waldron, J.-L. Vay, G. Deblonde, B. Wen, Y.-M. Chiang, B. P.
MacLeod, and Q. Ji. Investigation of light ion fusion reactions with plasma
discharges. Journal of Applied Physics, 126(20):203302, November 2019. Pub-
lisher: American Institute of Physics.
[256] Vladimir Pines, Marianna Pines, Arnon Chait, Bruce M. Steinetz, Lawrence P.
Forsley, Robert C. Hendricks, Gustave C. Fralick, Theresa L. Benyo, Ba-
yarbadrakh Baramsai, Philip B. Ugorowski, Michael D. Becks, Richard E.
Martin, Nicholas Penney, and Carl E. Sandifer. Nuclear fusion reactions in
deuterated metals. Physical Review C, 101(4):044609, April 2020. Publisher:
American Physical Society.
[257] Bruce M. Steinetz, Theresa L. Benyo, Arnon Chait, Robert C. Hendricks,
Lawrence P. Forsley, Bayarbadrakh Baramsai, Philip B. Ugorowski, Michael D.
Becks, Vladimir Pines, Marianna Pines, Richard E. Martin, Nicholas Pen-
ney, Gustave C. Fralick, and Carl E. Sandifer. Novel nuclear reactions ob-
served in bremsstrahlung-irradiated deuterated metals. Physical Review C,
101(4):044610, April 2020. Publisher: American Physical Society.
[258] C. Cirac?, R. T. Hill, J. J. Mock, Y. Urzhumov, A. I. Fern?ndez-Dom?nguez,
S. A. Maier, J. B. Pendry, A. Chilkoti, and D. R. Smith. Probing the Ulti-
mate Limits of Plasmonic Enhancement. Science, 337(6098):10721074, Au-
gust 2012. Publisher: American Association for the Advancement of Science
Section: Report.
[259] David K. Fork, Jeremy N. Munday, Tarun Narayan, and Joseph B. Murray.
Target structure for enhanced electron screening, April 2019.
[260] Masataka Hakamada and Mamoru Mabuchi. Fabrication of nanoporous pal-
ladium by dealloying and its thermal coarsening. Journal of Alloys and Com-
pounds, 479(1):326329, June 2009.
[261] Xingyou Lang, Lihua Qian, Pengfei Guan, Jian Zi, and Mingwei Chen. Local-
ized surface plasmon resonance of nanoporous gold. Applied Physics Letters,
98(9):093701, February 2011. Publisher: American Institute of Physics.
[262] Juergen Biener, Gregory W. Nyce, Andrea M. Hodge, Monika M. Bi-
ener, Alex V. Hamza, and Stefan A. Maier. Nanoporous Plasmonic
Metamaterials. Advanced Materials, 20(6):12111217, 2008. _eprint:
https://onlinelibrary.wiley.com/doi/pdf/10.1002/adma.200701899.
[263] G. C. A. M. Janssen, M. M. Abdalla, F. van Keulen, B. R. Pujada, and B. van
Venrooy. Celebrating the 100th anniversary of the Stoney equation for lm
237
stress: Developments from polycrystalline steel strips to single crystal silicon
wafers. Thin Solid Films, 517(6):18581867, January 2009.
[264] Jianjun Chen, Judong Zhang, Mingming Wang, and Ye Li. High-temperature
hydrogen sensor based on platinum nanoparticle-decorated SiC nanowire de-
vice. Sensors and Actuators B: Chemical, 201:402406, October 2014.
[265] Syed Mubeen, Ting Zhang, Bongyoung Yoo, Marc A. Deshusses, and
Nosang V. Myung. Palladium Nanoparticles Decorated Single-Walled Car-
bon Nanotube Hydrogen Sensor. The Journal of Physical Chemistry C,
111(17):63216327, May 2007. Publisher: American Chemical Society.
[266] Rakesh K. Joshi, Subramanian Krishnan, Mashamichi Yoshimura, and Ashok
Kumar. Pd Nanoparticles and Thin Films for Room Temperature Hydrogen
Sensor. Nanoscale Research Letters, 4(10):1191, July 2009.
[267] D. Gupta, D. Dutta, M. Kumar, P. B. Barman, C. K. Sarkar, S. Basu, and
S. K. Hazra. A low temperature hydrogen sensor based on palladium nanopar-
ticles. Sensors and Actuators B: Chemical, 196:215222, June 2014.
[268] Juree Hong, Sanggeun Lee, Jungmok Seo, Soonjae Pyo, Jongbaeg Kim, and
Taeyoon Lee. A Highly Sensitive Hydrogen Sensor with Gas Selectivity Using
a PMMA Membrane-Coated Pd Nanoparticle/Single-Layer Graphene Hybrid.
ACS Applied Materials & Interfaces, 7(6):35543561, February 2015. Pub-
lisher: American Chemical Society.
[269] Isolde Simon and Michael Arndt. Thermal and gas-sensing properties of a
micromachined thermal conductivity sensor for the detection of hydrogen in
automotive applications. Sensors and Actuators A: Physical, 97-98:104108,
April 2002.
[270] Sukon Phanichphant. Semiconductor Metal Oxides as Hydrogen Gas Sensors.
Procedia Engineering, 87:795802, January 2014.
[271] Ananya Dey. Semiconductor metal oxide gas sensors: A review. Materials
Science and Engineering: B, 229:206217, March 2018.
[272] L. P. Martin, A. Q. Pham, and R. S. Glass. Electrochemical hydrogen sensor
for safety monitoring. Solid State Ionics, 175(1):527530, November 2004.
[273] Ghenadii Korotcenkov, Sang Do Han, and Joseph R. Stetter. Review of Elec-
trochemical Hydrogen Sensors. Chemical Reviews, 109(3):14021433, March
2009. Publisher: American Chemical Society.
[274] D. Dwivedi, R. Dwivedi, and S. K. Srivastava. Sensing properties of palladium-
gate MOS (Pd-MOS) hydrogen sensor-based on plasma grown silicon dioxide.
Sensors and Actuators B: Chemical, 71(3):161168, December 2000.
238
[275] Danick Briand, Helena Wingbrant, Hans Sundgren, Bart van der Schoot,
Lars-Gunnar Ekedahl, Ingemar Lundstr?m, and Nicolaas F de Rooij. Mod-
ulated operating temperature for MOSFET gas sensors: hydrogen recovery
time reduction and gas discrimination. Sensors and Actuators B: Chemical,
93(1):276285, August 2003.
[276] Junghui Song, Wu Lu, Jerey S. Flynn, and George R. Brandes. AlGaN/GaN
Schottky diode hydrogen sensor performance at high temperatures with dier-
ent catalytic metals. Solid-State Electronics, 49(8):13301334, August 2005.
[277] Devika Sil, Jacqueline Hines, Uduak Udeoyo, and Eric Borguet. Palladium
Nanoparticle-Based Surface Acoustic Wave Hydrogen Sensor. ACS Applied
Materials & Interfaces, 7(10):57095714, March 2015. Publisher: American
Chemical Society.
[278] A. D'Amico, A. Palma, and E. Verona. Surface acoustic wave hydrogen sensor.
Sensors and Actuators, 3:3139, January 1982.
[279] D. R. Baselt, B. Fruhberger, E. Klaassen, S. Cemalovic, C. L. Britton, S. V.
Patel, T. E. Mlsna, D. McCorkle, and B. Warmack. Design and performance of
a microcantilever-based hydrogen sensor. Sensors and Actuators B: Chemical,
88(2):120131, January 2003.
[280] J. van Lith, A. Lassesson, S. A. Brown, M. Schulze, J. G. Partridge, and
A. Ayesh. A hydrogen sensor based on tunneling between palladium clusters.
Applied Physics Letters, 91(18):181910, October 2007. Publisher: American
Institute of Physics.
[281] Ediz Herkert, Florian Sterl, Nikolai Strohfeldt, Ramon Walter, and Harald
Giessen. Low-Cost Hydrogen Sensor in the ppm Range with Purely Opti-
cal Readout. ACS Sensors, 5(4):978983, April 2020. Publisher: American
Chemical Society.
[282] Iwan Darmadi, Ferry Anggoro Ardy Nugroho, and Christoph Langham-
mer. High-Performance Nanostructured Palladium-Based Hydrogen Sen-
sorsCurrent Limitations and Strategies for Their Mitigation. ACS Sensors,
November 2020. Publisher: American Chemical Society.
[283] V. Palmisano, M. Filippi, A. Baldi, M. Slaman, H. Schreuders, and B. Dam.
An optical hydrogen sensor based on a Pd-capped Mg thin lm wedge. Inter-
national Journal of Hydrogen Energy, 35(22):1257412578, November 2010.
[284] Donato Luna-Moreno and David Monz?n-Hern?ndez. Eect of the PdAu
thin lm thickness uniformity on the performance of an optical ber hydrogen
sensor. Applied Surface Science, 253(21):86158619, August 2007.
[285] Donato Luna-Moreno, David Monz?n-Hern?ndez, Joel Villatoro, and Gon?al
Badenes. Optical ber hydrogen sensor based on core diameter mismatch and
239
annealed PdAu thin lms. Sensors and Actuators B: Chemical, 125(1):6671,
July 2007.
[286] Joel Villatoro, Donato Luna-Moreno, and David Monz?n-Hern?ndez. Opti-
cal ber hydrogen sensor for concentrations below the lower explosive limit.
Sensors and Actuators B: Chemical, 110(1):2327, September 2005.
[287] C?dric Perrotton, Ruud J. Westerwaal, Nicolas Javahiraly, Martin Slaman,
Herman Schreuders, Bernard Dam, and Patrick Meyrueis. A reliable, sensitive
and fast optical ber hydrogen sensor based on surface plasmon resonance.
Optics Express, 21(1):382390, January 2013. Publisher: Optical Society of
America.
[288] X. B?venot, A. Trouillet, C. Veillas, H. Gagnaire, and M. Cl?ment. Surface
plasmon resonance hydrogen sensor using an optical bre\ast. Measurement
Science and Technology, 13(1):118124, December 2001. Publisher: IOP Pub-
lishing.
[289] Kaiqun Lin, Yonghua Lu, Junxue Chen, Rongsheng Zheng, Pei Wang, and Hai
Ming. Surface plasmon resonance hydrogen sensor based on metallic grating
with high sensitivity. Optics Express, 16(23):1859918604, November 2008.
Publisher: Optical Society of America.
[290] J?nos Schanda. CIE Colorimetry. In Colorimetry, pages 25
78. John Wiley & Sons, Ltd, 2007. Section: 3 _eprint:
https://onlinelibrary.wiley.com/doi/pdf/10.1002/9780470175637.ch3.
[291] Chris Wyman, Peter-Pike Sloan, and Peter Shirley. Simple Analytic Approx-
imations to the CIE XYZ Color Matching Functions. Journal of Computer
Graphics Techniques, 2(2):11, 2013.
[292] G. G. Libowitz. Metallic hydrides; fundamental properties and applications.
Journal of Physics and Chemistry of Solids, 55(12):14611470, December
1994.
[293] G. G. Libowitz. Electronic properties of the rare earth hydrides. Berichte der
Bunsengesellschaft f?r physikalische Chemie, 76(8):837845, 1972. _eprint:
https://onlinelibrary.wiley.com/doi/pdf/10.1002/bbpc.19720760854.
[294] Eugene V. Sheely. Theoretical Study of the Eects of Di-Muonic Molecules
on Muon-Catalyzed Fusion. Technical report, AIR FORCE INST OF TECH
WRIGHT-PATTERSON AFB OH GRADUATE SCHOOL OF ENGINEER-
ING AND MANAGEMENT, March 2012. Section: Technical Reports.
[295] Russell D. Johnson. NIST Computational Chemistry Comparison and Bench-
mark Database, August 2020.
240
[296] Z. Sun and D. Tom?nek. Cold fusion: How close can deuterium atoms come
inside palladium? Physical Review Letters, 63(1):5961, July 1989. Publisher:
American Physical Society.
[297] Andreas Borgschulte, Jasmin Terreni, Emanuel Billeter, Luke Daemen,
Yongqiang Cheng, Anup Pandey, Zbigniew ?odziana, Russell J. Hemley, and
Anibal J. Ramirez-Cuesta. Inelastic neutron scattering evidence for anoma-
lous HH distances in metal hydrides. Proceedings of the National Academy
of Sciences, 117(8):40214026, February 2020. Publisher: National Academy
of Sciences Section: Physical Sciences.
[298] Anubhav Jain, Shyue Ping Ong, Georoy Hautier, Wei Chen, William David-
son Richards, Stephen Dacek, Shreyas Cholia, Dan Gunter, David Skin-
ner, Gerbrand Ceder, and Kristin A. Persson. Commentary: The Materials
Project: A materials genome approach to accelerating materials innovation.
APL Materials, 1(1):011002, July 2013. Publisher: American Institute of
Physics.
[299] Arnulf J. Maeland. Investigation of the VanadiumHydrogen System by
X-Ray Diraction Techniques 1,2. The Journal of Physical Chemistry,
68(8):21972200, August 1964.
241