Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Nano-fabricated devices in electrochemistry and cancer therapy
(USC Thesis Other)
Nano-fabricated devices in electrochemistry and cancer therapy
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
NANO-FABRICATED DEVICES IN ELECTROCHEMISTRY
AND CANCER THERAPY
By
Yifei Wang
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(ELECTRICAL ENGINEERING)
May 2018
Copyright 2018 Yifei Wang
Dedication
Dedicated to my family.
Acknowledgements
I will always remember the past several years studying at USC as a PhD student. The
exposure to the University environment, to great people I have met, to excellent research
atmosphere, have provided me invaluable life experience. There are numerous people
who have helped, encouraged, and inspired me, giving me valuable advices on both
research and various aspects of my life. I would like to thank them all for making this
dissertation possible.
First of all, I wish to thank my advisor, Prof. Wei Wu, for offering me the valuable
experience in experimental nanotechnology research. His sincere interests in
nanotechnology, and knowledge in interdisciplinary areas have been a great inspiration to
me. I am also very grateful for his valuable suggestions and help for my academic career.
Without his insight, guidance and support, this work would have never been achieved.
I would also like to express my great thanks to my dissertation committee members Prof.
Han Wang and Prof. S.R. Narayanan for their very helpful insights, comments and
suggestions as well as Prof. E.S. Kim and Prof. Mahta Moghaddam for serving on my
qualifying exam committee.
My PhD work in the lab carries a lot of support and help from my lab colleagues,
including Dr. He Liu, Dr. Yuhan Yao, Boxiang Song, Yuanrui Li, Shujin Huang, Hao
Yang, Deming Meng, Buyun Chen, Pan Hu, Yunxiang Wang, Zhexian Ou, Meng Yu,
Chi-Chan Wu, Xuan Qin. All of you are not only excellent researchers but also fabulous
people, and I really enjoyed working with you in such a friendly, encouraging, and active
group.
Additionally, I would like to acknowledge all of the collaborators, including Prof. S.R.
Narayanan, Prof. Mahta Moghaddam, Prof. John Stang, Prof. Eugene Chung, Prof. Pin
Wang, Prof. Qifa Zhou, Prof. Stephen Cronin, Prof. Chongwu Zhou, Prof. Rehan
Kapadia, Prof. Mark Thompson from USC, Dr. Adam Schwartzberg, Dr. Deirdre Olynick,
Dr. Stefano Cabrini from Lawrence Berkeley National Laboratory, Prof. Essam Heggy
from Jet Propulsion Laboratory, and Prof. Fanxin Liu from Zhejiang University of
Technology (China), who have provided invaluable expertise in different research areas
has greatly helped me diversify my knowledge and broaden my vision.
Lastly, and most importantly, I greatly appreciate the endless love, understand, and
encouragement from my parents, my family. I would not have all these achievements
today without you.
Contents
Abstract ............................................................................................................................................ i
List of Figures and Tables .............................................................................................................. iv
Project 1: Nanogap Electrochemical Cells for Hydrogen Generation ............................................. 1
Chapter 1. Introduction for Hydrogen Production....................................................................... 1
Chapter 2. Theoretical Analysis .................................................................................................. 5
2.1. Background Analysis ........................................................................................................ 5
2.1.1. Debye-length ......................................................................................................... 5
2.1.2. Electric Field Distribution ..................................................................................... 8
2.2. Water Electrolysis in Three Systems .............................................................................. 14
2.1.1. Water Electrolysis in Macrosystem ..................................................................... 14
2.2.2. Pure Water Electrolysis in Nanogap Cell ............................................................ 17
2.3. Virtual Breakdown Mechanism ...................................................................................... 18
2.3.1. Virtual Breakdown Mechanism ........................................................................... 18
2.3.2. RC-circuit Model for Nanogap Cell .................................................................... 21
2.4. Summary ........................................................................................................................ 22
Chapter 3. Device Fabrication ................................................................................................... 23
3.1. Device Design ................................................................................................................ 23
3.1.1. Material Selection ............................................................................................... 24
3.1.2. Electrode System Design .................................................................................... 25
3.2. Fabrication Process ......................................................................................................... 27
3.2.1. Process Details .................................................................................................... 27
3.2.2. Efforts to Avoid Short-circuit .............................................................................. 30
3.3. Low DC-bias Silicon Nitride Etching ............................................................................ 34
3.3.1. Traditional High DC-bias Etching ....................................................................... 34
3.3.2. DC-bias vs. Etching Parameters .......................................................................... 36
3.3.3. Optimal Etching Recipe ...................................................................................... 39
3.4. Summary ........................................................................................................................ 41
Chapter 4. Water Splitting Experiment Results......................................................................... 42
4.1. Pure Water Experiments ................................................................................................. 42
4.1.1. IV-curve Measurement ........................................................................................ 42
4.1.2. Electron-transfer Limited Reactions and Phase Diagram ................................... 46
4.1.3. Electrolysis Efficiency Analysis .......................................................................... 50
4.2. Anode Oxidation Damage .............................................................................................. 53
4.3. Sodium Hydroxide Solution Experiments ...................................................................... 57
4.3.1. Pure Water vs. Sodium Hydroxide Solution ........................................................ 57
4.3.2. Bubble Effect ...................................................................................................... 62
4.3.3. log I vs. V curves ................................................................................................. 63
4.4. Summary ........................................................................................................................ 66
Chapter 5. Methanol Solution Splitting Experiment Results .................................................... 67
5.1. Background Analysis ...................................................................................................... 67
5.2. Device Fabrication .......................................................................................................... 71
5.3. Experiment Results ......................................................................................................... 77
5.4. Summary ........................................................................................................................ 83
Conclusion and Future Work..................................................................................................... 84
Project 2: Lithographically Defined Particles for Cancer Therapy ............................................... 86
Chapter 6. Introduction for Cancer Therapy ............................................................................. 86
Chapter 7. Micro-Resonators for Microwave Cancer Therapy ................................................. 91
7.1. Resonance Microparticle Design .................................................................................... 91
7.2. Resonance Frequency and Q-value ................................................................................ 94
7.3. Q-value Optimization ..................................................................................................... 96
7.4. Summary ...................................................................................................................... 103
Chapter 8. Non-Resonance Microparticles: Design and Fabrication ...................................... 105
8.1. Magnetic Dipoles and Electric Dipoles ........................................................................ 105
8.2. Magnetic Dipoles Optimization ................................................................................... 110
8.3. Fabrication and Releasing Process ............................................................................... 115
8.4. Summary ...................................................................................................................... 118
Chapter 9. Particle Collection and Heating Enhancement Characterization ........................... 119
9.1. Polymer-Film Nanocomposite Setup............................................................................ 120
9.2. Particle-Suspended Hydrogel Setup ............................................................................. 124
9.3. Summary ...................................................................................................................... 130
Chapter 10. Particle Optimization and Fundamental Characterization ................................... 131
10.1. Pure Gold Particles without Parylene ......................................................................... 132
10.2. Permittivity Characterization ...................................................................................... 136
10.3. Summary .................................................................................................................... 141
Conclusion and Future Work................................................................................................... 142
Reference ..................................................................................................................................... 144
Publication ................................................................................................................................... 156
i
Abstract
Top-down nanofabrication is a powerful tool to achieve desired nano-devices that can
contribute to various types of applications. Besides the traditional semiconductor
nanofabrication, an increasing attention has been focused on many other fields, e.g.,
nano-optics, nano-energy, nano-biomedicine, etc. Sometimes the performance of such
nano-devices can be quite different from those macro-devices, and new knowledge and
new functions can be generated by studying these nano-devices. In this dissertation, two
top-down nanofabrication projects have been discussed.
In the first project (Chapter 1 - 5), we utilized deep-sub-Debye-length nanogap
electrochemical cells to achieve efficient electrolysis of pure water and pure methanol
solution (without any added electrolyte) at room temperature. Here we have
fundamentally broken through the common knowledge that only conductive solution (or
solution with strong electrolyte) can be efficiently electrolyzed. A field-assisted effect
resulted from overlapped electrical double layers can greatly enhance molecules
ionization and mass transport, leading to electron-transfer limited reactions. We have
named this process “virtual breakdown mechanism” (which is completely different from
traditional mechanisms) that couples the two half-reactions together, greatly reducing the
energy losses arising from ion transport. This fundamental discovery has been
theoretically discussed in this dissertation and experimentally demonstrated in a group of
electrochemical cells with nanogaps between two electrodes down to 37 nm. Many
ii
efforts were put onto the improvement of fabrication yield of such
deep-sub-Debye-length nanogap electrochemical cells. Based on our nanogap
electrochemical cells, the electrolysis current density from pure water can be significantly
larger than that from 1 mol/L sodium hydroxide solution, indicating the much better
performance of pure water splitting as a potential for on-demand clean hydrogen
production.
In the second project (Chapter 6 - 10), lithographically defined nanoparticles with high
efficiency of microwave absorption have been used for cancer hyperthermia therapy. Two
types of particles have been studied: resonance particles and non-resonance particles. For
the LC-resonance particles, finite element method has been used for design and
optimization. It was found that the limitation of the micro-resonator is not the
requirement of the low resonance frequency but the low Q-value due to insufficient
conductivity of natural materials. For the non-resonance particles, disk-shaped magnetic
dipoles have been chosen as the high-efficiency microwave-absorbers. The fabrication
and collection processes of the particles have been developed. Our measurements, based
on both polymer-film nanocomposite setup and particle-suspended hydrogel setup, have
demonstrated that the micro/nano-particles can absorb the external electromagnetic field
efficiently and greatly enhance the localized heat generation compared to control groups
without particles. Further characterization has shown that both the real and imaginary
part of the permittivity of our nanocomposite have been enhanced, which is probably the
fundamental reason why the microwave absorption efficiency can be improved.
iii
In the first project, Chapter 1 discusses the state-of-art methods of industrial hydrogen
generation. Chapter 2 discusses the fundamental difference between macro electrode
system and nanogap electrochemical cells (taking water electrolysis as an example).
Chapter 3 discusses the fabrication process of our vertical-designed nanogap
electrochemical cells. Chapter 4 and 5 discusses the experiment results of pure water
electrolysis and pure methanol solution electrolysis, respectively, both based on our
nanogap electrochemical cells.
In the second project, Chapter 6 discusses the state-of-art methods of cancer therapy.
Chapter 7 discusses the method of LC-circuit resonance particles. Chapter 8 discusses the
design and fabrication of the non-resonance particles. Chapter 9 discuss the method to
collect / transfer such lithographically defined particles, and uses our disk-shaped
magnetic dipoles as an example to show the heating enhancement characterization.
Chapter 10 further discusses the fundamental characterization of such particle/matrix
nanocomposite mixtures.
iv
List of Figures and Tables
Figure 2.1 Schematic diagram of the concept “Debye-length”……………………. 6
Figure 2.2 Schematic diagram of potential distribution comparison between
macrosystem and our sandwiched-like nanogap cells. High electric
field uniformly distributed in the entire gap between anode and cathode
in NECs………………………………………………………………… 9
Figure 2.3 Simulation results to show the electric field distribution (1-D plot and
2-D plot) between two electrodes with gap distance of 37 nm (the
smallest gap distance in our experiments), 0.1 μm (0.1λ d,
sandwiched-like NEC), 5.0 μm (5λ d, sandwiched-like NEC) and 100
μm (macrosystem, plate electrodes)……………………………………. 10
Figure 2.4 Geometry and boundary conditions setting in finite element
calculations……………………………………………………………... 12
Figure 2.5 Schematic diagram of water splitting reactions in macrosystems. (a)
Pure water in macrosystem cannot be split efficiently due to the lack of
rapid ions transport inside bulk solution. (b) In sodium hydroxide
solution, water splitting reaction can keep occurring but is limited by
mass transport (mainly diffusion)………………………………………. 15
Figure 2.6 In nanogap cell, high electric field in the entire gap can enhance water
ionization and mass transport (mainly migration), leading to pure water
splitting efficiently limited by electron-transfer………………………... 17
Figure 2.7 The field-assisted effect is equivalent to the scenario that water
molecules are split into H 3O
+
and OH
-
ions in the middle of the
nanogap…………………………………………………………………. 19
v
Figure 2.8 RC-circuit model of half-reaction of water splitting in nanogap cells…. 21
Figure 3.1 Pourbaix diagrams of gold and nickel………………………………….. 24
Figure 3.2 Vertical nanogap system and horizontal nanogap system……………… 25
Figure 3.3 Fabrication procedures and results of our metal-dielectric-metal
sandwiched-like NECs. This fabrication method can be simply
fabricated on large area with high yield. Dimensions: the key
parameter of the gap distance between the two electrodes, or thickness
of silicon nitride, varied from 37 nm to 1.4 μm; thermal silicon
dioxide, 100 nm thick; Pt, 100 nm thick; Ti, 2 nm thick; gold, 40 nm
thick; Cr, 10 nm thick; the contact pads were 3.5 mm by 3.5 mm; the
grating regions were 1 cm by 1 mm, with different pitches from 10 μm
to 80 μm………………………………………………………………… 28
Figure 3.4 “Rabbit ear” structure formed after lift-off process due to the
non-vertical photoresist sidewall. The figure shown here was after the
silicon nitride etching, which made the “rabbit ear” clearer…………… 31
Figure 3.5 Top metal layer with one single line under optical microscope. Top
metal branches can form before silicon nitride etching, leading to
short-circuit between top and bottom metal layer……………………… 32
Figure 3.6 Sputtering yield of different materials [41]…………………………….. 33
Figure 3.7 Metal atoms sputtered out and re-deposited at high DC-bias etching,
leading to short-circuit between two metal layers……………………… 35
Figure 3.8 The relationship between DC-bias / etching profile and (a) capacitively
coupled RF power, (b) inductively coupled plasma (ICP) power, (c) the
pressure in the etching chamber, and (d) the combination of etching
vi
gases. The scale bar is 400 nm…………………………………………. 36
Figure 3.9 The recipe parameters and the etching profile of our low DC-bias
silicon nitride etching…………………………………………………... 40
Figure 4.1 I-V curves measurement based on our NECs with pure DI water. (a)
Linear I-V curves showed larger current generated from smaller gap
distances. (b) A voltage plateau around 0.9 V shown on the log I vs. V
curves…………………………………………………………………… 42
Figure 4.2 A typical I-V curve comparison between dry sweep and pure water test.
The device measured here was with 72 nm silicon nitride thickness and
20 μm grating pitch…………………………………………………….. 45
Figure 4.3 The plot of electrolysis current vs. gap distance
-1
at different voltages
demonstrated that the pure water splitting was limited by
electron-transfer step…………………………………………………… 47
Figure 4.4 Phase diagram of electrochemical performance vs. gap distance….…… 49
Figure 4.5 Even gold anode can become roughed during water electrolysis [48].
Here * stands for an adsorption site on the electrode…………………... 54
Figure 4.6 Anode damage formed in pure water splitting when voltage above 5 V.
The device shown here is with 72 nm gap and 40 μm pitch…………… 55
Figure 4.7 Pourbaix diagrams of indium and tin………………………….……….. 56
Figure 4.8 Electrolysis current at 1.8 V vs. the number of edges from pure water
splitting and water splitting in 1 mol/L sodium hydroxide solution,
both based on our NECs………………………………………….…….. 57
Figure 4.9 Schematic diagram of the mechanisms showing the different reaction
locations in pure water splitting and water splitting in sodium
vii
hydroxide solution……………………………………………………… 58
Figure 4.10 Evidence of the entire surface involved into the reactions in sodium
hydroxide solutions. (a) Larger droplet provided larger current. (b)
Bubbles formed at non-grating region. The devices were with 72 nm
gap distance…………………………………………………………….. 59
Figure 4.11 A PDMS coating to confine the reaction area………………………….. 60
Figure 4.12 (a) Bubble generation around 2 V from pure water electrolysis. (b)
Bubble effects on plateaus (or peaks) around 2 V in I-V curves based
on devices with 72 nm gap and 10 μm pitch in pure water, and (c) in 1
mol/L sodium hydroxide solution………………………………………. 62
Figure 4.13 Plateaus in log I vs. V curves from (a) pure water tests and (b) sodium
hydroxide solution tests. The devices were with 72 nm gap distance….. 64
Figure 5.1 Comparison of the required voltage between water electrolysis and
methanol electrolysis…………………………………………………… 69
Figure 5.2 Parameter comparison between water and methanol…………………... 70
Figure 5.3 High-magnification SEM image of the fabrication result of the device
for pure methanol solution electrolysis………………………………… 74
Figure 5.4 Two compatibility issues of Ru-involved fabrication process. (a)
Bubbles generated when the sample with Pt/Ru was immersed into the
Cr-etchant, indicating Ru had been wet-etched. (b) Pt/Ru layer became
non-uniform after silicon nitride etching, indicating Ru had been
dry-etched………………………………………………………………. 75
Figure 5.5 IV-curve measurements of pure methanol solution based on the
nanogap electrochemical cells, (a) linear I vs. V, (b) log I vs. V where
viii
the background data from pure water had been removed……………… 78
Figure 5.6 Comparison between the electrolysis of methanol solution and ethanol
solution, both based on our nanogap cells……………………………… 80
Figure 5.7 The experiment results of IV-curve measurement when the anode was
changed to pure Pt, (a) methanol solution, (b) ethanol solution……….. 82
Figure 7.1 Design of multi-layer LC-structures, (a) schematic diagram, (b)
equivalent circuit diagram, and (c) to-scaled structure model with
five-layer inductor………………………………………………………
93
Figure 7.2 (a) A typical LC-structure and the heating loss curves vs. frequency,
with conductivity (b) 10
10
S/m, (c) 4.1×10
7
S/m (gold), and (d) 10
12
S/m…………………………………………………………………........ 94
Figure 7.3 Effect of the inner diameter and outer diameter of the coil. Square line:
keep inner diameter zero; circle line: keep outer diameter 8 μm………. 97
Figure 7.4 (a) The total heating loss from different linewidth structures, including
basic heating loss. (b) Modified Q-to-f 0 ratio vs. linewidth…………….. 98
Figure 7.5 Effects of the number of the inductor layers (circle line) and the
thickness of the inductor coil (square line). The bottom x-axis is also
corresponding to the total thickness of the coil layers (circle line)…….. 102
Figure 8.1 Simulation results of comparison between magnetic dipoles and
electric dipoles. Volume-averaged loss density vs. the feature size D
and L (when parameter a was 100 nm). The direction of the EM field is
also shown on the figure………………………………………………... 106
Figure 8.2 Simulation results of comparison between magnetic dipoles and
electric dipoles. Volume-averaged loss density vs. the parameter a
ix
(when D and L were 8 μm)……………………………………………... 108
Figure 8.3 Simulation result of the size effect of the electric dipoles. If the aspect
ratio did not change, then the volume-averaged loss density would not
change…………………………………………………………………... 109
Figure 8.4 Simulation results of comparison between the simple gold ring
structures and the nickel/gold-core/ring structures. The outer diameter
was fixed at 8 μm and the height was fixed at 1 μm, and the diameter
of nickel core was equal to the inner diameter of the gold ring………... 111
Figure 8.5 Simulation results of (a) comparison of the total loss from particle
arrays (1×1, 3×3, 10×10, and 30×30) with the same total volume (the
height of the particles was 100 nm), and (b) the localized distribution
of the volume loss density from the four arrays (the scale bar is in
log-scale)………………………………………………………………..
114
Figure 8.6 The fabrication and releasing processes of the disk-shaped LDPs. The
results shown here were observed by naked eye, by low magnification
under SEM (scale bar: 100 μm), and by high magnification under SEM
(scale bar: 5 μm), respectively…………………………………………. 116
Figure 9.1 (a-d) A parylene film with gold LDPs peeled off from a quarter of
3-inch wafer gradually in the LOL stripper solution. (e) An entire
3-inch parylene film peeled off. (f) A blank parylene film and a
parylene film with LDPs after drying……………………….………….. 120
Figure 9.2 (a) The experimental measurement set-up. (b) The microwave aperture
near the end of the microwave probe (scale bar: 1 cm), with a thermal
sensor close to it. (c) During experiments, the microwave aperture was
completely covered by the parylene film inside DI water……………… 122
x
Figure 9.3 The microwave heating characterization results of temperature vs. time
based on the parylene-film nanocomposite setup. The microwave input
power in (b) was 20 W…………………………………………………. 123
Figure 9.4 Transferring the LDPs from wafers to DI water by LOL stripper
flushing and multiple times centrifugation……………………………... 125
Figure 9.5 Combining the agarose hydrogel mash (as the matrix) and centrifuged
LDP / water solution by ultrasound mixing to obtain the agarose/LDP
composite……………………………………………………………….. 127
Figure 9.6 Heating enhancement characterization of temperature vs. time based on
different LDP concentrations. The microwave input power was 20 W at
1.9 GHz………………………………………………………………….
129
Figure 10.1 The ideal fabrication process to fabrication pure gold particles without
parylene as the underneath support…………………………………….. 133
Figure 10.2 The parameters of Mg e-beam evaporation…………………………….. 134
Figure 10.3 Pure gold micro-disks without parylene support. The gold surface was
not uniform due to the high deposition rate of Mg evaporation………... 135
Figure 10.4 The LDP/PDMS mixture was obtained by mixing the liquid PDMS (A)
and LDP/acetone solution by ultrasound agitation………….………….. 137
Figure 10.5 The measurement set-up for permittivity characterization. A
ring-holder was used to fix the liquid composite between the two
parallel-planar electrodes……………………………………………….. 138
Figure 10.6 The characterization results of permittivity of LDP/PDMS(A)
composites. (a) The composites with four different LDP
concentrations, pure PDMS (A), 0.75×10
7
/mL, 1.5×10
7
/mL, 3×10
7
/mL.
xi
(b) Real part of permittivity measurement. (c) Imaginary part of
permittivity measurement……………………………………………….
140
Table I Parameters setting in simulations for electric filed distribution in
nanogap cells…………………………………………………………… 12
1
Project 1: Nanogap Electrochemical Cells for
Hydrogen Generation
Chapter 1. Introduction for Hydrogen Production
Since the first oil embargo in the 1970s there has been interest in developing alternative
fuels to power our society. As a clean and renewable resource to substitute fossil fuels in
future, efficient hydrogen production has become increasingly important. Moreover,
hydrogen has a huge demand in variety of industrial fields such as metal refining,
ammonia synthesis, petroleum refining and energy storage. The annual production of
hydrogen is estimated to be about 55 million tons with its consumption increasing by
approximately 6% per year [1]. From the perspective of element abundance, hydrogen is
one of the most abundant elements on earth. Generally hydrogen combines with other
chemical elements, and it can be found as part of other substances, such as water,
hydrocarbon, or alcohol. How to efficiently separate hydrogen from such compounds is
drawing attentions from world-widely researchers.
Today, more than 90% of the industrially-produced hydrogen comes from steam
reforming of natural gas and gasification of coal and petroleum coke [2]. Fuel processing
technologies convert a hydrogen containing material such as gasoline, ammonia, or
methanol into a hydrogen rich stream. Close to 50% of the global demand for hydrogen is
currently generated via steam reforming of natural gas, about 30% from oil/naphtha
2
reforming from refinery/chemical industrial off-gases, 18% from coal gasification [1].
However, such approaches have two serious issues. Firstly, such hydrocarbon resources
themselves are non-renewable energy supplier, which are relatively scarce resources at
present and in the future. Secondly, massive emissions of greenhouse gases are produced
during such hydrogen production.
Photolysis of water is another promising technology for future hydrogen production [3-5].
Currently, it is the least expensive and the most effective method of hydrogen production
from renewable resources. The element source (e.g., sea water) and the energy source
(e.g., sunlight) are unlimited. The semiconductor photoelectrode can generate
electron-hole pairs to create the necessary voltage for the direct decomposition of water
molecule into oxygen and hydrogen. However, this method is still under development
because of low overall efficiencies.
Water electrolysis, especially when connected to other renewable energy supplies such as
wind turbines, solar photovoltaic, and hydroelectric generation, can provide high-purity
hydrogen and could be a solution for a sustainable energy supply. However, at present
only 4% industrial hydrogen production comes from water electrolysis [6-8], basically
due to the low conversion efficiency resulting from the high cell voltage, which arises
from the large overpotential at the electrodes and ohmic loss in the solution, especially at
large operating current density [9]. Water electrolysis has been known for more than 200
years [10] and applied on industrial hydrogen production for over 100 years [6, 11].
However, efforts to increase the energy efficiency and reduce the cost of water
3
electrolysis continue even today. Research has focused on electrocatalytic materials
[12-14], temperature and pressure effects [9, 15, 16], and optimization of electrolyzer
design [17, 18].
Different from these foregoing approaches, we have demonstrated a new approach to
improve the electrochemical reaction efficiency, by using electrochemical cells with
distance between anode and cathode in nanometer scale, even much smaller than
Debye-length. With these nanogap electrochemical cells (NECs), even pure water or pure
methanol solution (without any added electrolyte) can be electrochemically electrolyzed
much more efficiently than traditional theory predicted.
In this project, we mainly achieved two breakthroughs. The first one is fundamental
breakthrough. Different from common knowledge that we learned from high-school
Chemistry, we successfully achieve pure water electrolysis, and proposed a totally novel
theory, what we called “virtual breakdown mechanism”, to explain our experiment results.
The second breakthrough is fabrication breakthrough. we successfully fabricated nanogap
electrochemical cells with gap distance down to 37 nm, with much higher fabrication
yield than other reported. Our experiments have demonstrated that the performance of
NECs with pure water can be comparable to or even better than 1 mol/L sodium
hydroxide solution, which results from completely different microscopic mechanism of
field-driven ions transport to enhance water ionization and even virtual breakdown. We
have pushed the limitation of such electrochemical reactions to electron-transfer step.
Moreover, we even attempted non-conductive pure methanol solution and successfully
4
achieve the electrolysis of pure methanol solution with much lower required voltage. Our
experiments have demonstrated that by using the NECs, such electrolysis of pure water or
pure methanol solution could provide a great potential for high energy-efficiency
on-demand hydrogen production for both mass manufacturing and portable devices.
5
Chapter 2. Theoretical Analysis
In this chapter, we will discuss the fundamental difference when we shrink the gap
distance between anode and cathode to much smaller than Debye-length. We will take
water electrolysis as an example, to analyze the electrochemical reactions in both
macrosystem and NECs. Starting from the concept of Debye-length, we point out that the
electric field distribution is quite different between traditional macrosystem and NECs.
Then the different reaction mechanisms in these two systems are discussed. Due to the
large electric field uniformly distributed in the entire nanogap, mass transport inside
NECs can be enhanced significantly, and even “virtual breakdown mechanism” can be
achieved.
2.1. Background Analysis
2.1.1. Debye-length
The key concept of this project is “Debye-length”, named after the Dutch physicist and
physical chemist Peter Debye, is a measure of a charge carrier's net electrostatic effect in
solution and how far its electrostatic effect persists.
Consider putting a negative electrode into a solution with positive and negative ions in it.
Due to the electrostatic force, such positive ions can be attracted to the electrode, and
negative ions can be repelled away from the electrode. Such ion movement can lead to
the change of ion concentration distribution, which can further lead to the ion diffusion
6
due to concentration gradient. Under steady-state, the ion drift and the ion diffusion can
keep balance and the entire ion concentration distribution achieves equilibrium.
In solution, the Poisson Equation can be applied,
arg
0
()
( ) ( )
ch e
r
r
rr
•
(2.1)
Where εr is the dielectric constant of the solution, ε0 is the permittivity of free space,
ρcharge is the local charge density (net charge per volume), φr is the local electrostatic
potential. The local charge density ρcharge can be presented as the sum of all ions i,
arg
1
( ) ( )
n
ch e i i
i
r n r ez
(2.2)
Where ni is the number of ions of type i, e is the electronic charge, and zi is the valence of
ion i. Assume that the concentration of ions of type i is governed by the Boltzmann
distribution of statistical mechanics,
Figure 2.1. Schematic diagram of the concept “Debye-length”.
7
,
()
( ) exp[ ]
i
i i bulk
B
ze r
n r n
kT
(2.3)
Where kB is the Boltzmann constant and T is the temperature. We combine the equation
(2.1) - (2.3) and obtain the Poisson-Boltzmann Equation,
,
1
0
() 1
( ) ( ) [ exp( )]
n
i
r i i bulk
i
B
ze r
r r ezn
kT
•
(2.4)
When considering the requirement of Gouy-Chapman theory
()
1
i
B
ze r
kT
(2.5)
By Taylor expanding we can ignore the second and higher orders. In this way,
( ) ( )
exp( ) 1
ii
BB
ze r ze r
k T k T
(2.6)
Substitute this into equation (2.4), we can obtain
22
,
1
0
1
( ) ( ) [ ] ( )
n
i
r i bulk
i
B
ze
r r n r
kT
•
(2.7)
When dielectric constant does not depend on location, then the equation can be rewritten
as
22
( ) ( ) rr
(2.8)
Where
2
22
,
1
0
n
i bulk i
i
rB
e
nz
kT
(2.9)
When we just consider 1-D situation, the solution of the equation (2.8) is
0
1
( ) exp( )
x
x
(2.10)
8
Notice that the electrostatic potential decreases along with the distance far away from the
electrode, meaning that the counter ions inside the solution has screened the electric field
from the electrode (the “screening effect”, or the “double layer effect”). We define the
“Debye-length” is just the distance where the potential decrease to 1/e of the surface
potential. Here, κ
-1
is just the Debye-length. Theoretically,
2
1 2 1/2
,
1
0
()
n
i bulk i
i
rB
e
nz
kT
(2.11)
Notice that, the Debye-length is significantly dependent on the concentration of the
solution. For pure water, the Debye-length is 970 nm; if we consider CO2 from air
dissolved into pure water, the Debye-length has been reduced to around 220 nm
(saturated equilibrium); for 1 mol/L sodium hydroxide solution, the Debye-length has
been down to only 0.3 nm.
2.1.2. Electric Field Distribution
Based on the concept of Debye-length, let us discuss the difference between the
traditional electrochemical cells and nanogap electrochemical cells. As shown in Figure
2.2, the fundamental difference between traditional cells and nanogap cells are their
electric potential distribution. For water electrolysis with strong electrolyte in
macrosystem (i.e., high-concentration solution), the Debye-length is quite small. Due to
the screening effect, almost all the potential drop is confined within such small
Debye-length region (or double layer region). The potential in bulk solution (far from the
9
electrodes) would not change too much, meaning that there is nearly zero electric field
inside the bulk solution. However, when we put the counter electrode within the
Debye-length region, two double layers from anode and cathode can be overlapped with
each other. The electrostatic potential inside the entire gap has to change continuously
and dramatically, meaning that huge electric field can be uniformly distributed in the
entire gap. This is the fundamental difference between traditional macrosystem and our
NECs. In our project, we utilized metal-dielectric-metal sandwiched-like nanostructures
to achieve NECs. The gap distance is tuned by adjusting the thickness of the silicon
Figure 2.2. Schematic diagram of potential distribution comparison between macrosystem and our
sandwiched-like nanogap cells. High electric field uniformly distributed in the entire gap between
anode and cathode in NECs.
10
nitride between the two electrodes and can be easily achieved to deep-sub-Debye-length
in pure water.
Figure 2.3 shows the simulation results of electric field distribution between two
electrodes with different gap distances, from 37 nm (the smallest gap distance we
Figure 2.3. Simulation results to show the electric field distribution (1-D plot and 2-D plot)
between two electrodes with gap distance of 37 nm (the smallest gap distance in our experiments),
0.1 μm (0.1λ d, sandwiched-like NEC), 5.0 μm (5λ d, sandwiched-like NEC) and 100 μm
(macrosystem, plate electrodes).
11
achieved in our experiments) to 100 μm (considered as macrosystem). Close to the
electrode regions both the nanogap cell and the macrosystem present a high electric field
due to the screening effect; however, in bulk solution the electric field in 100 μm
macrosystem is only 10 V/m while in 0.1 μm gap the field can obtain 10
7
V/m. The
rainbow-scale 2-D plot shows more clearly that huge electric field can form in the entire
gap between the two electrodes in nanogap cells. Such high electric field is the key factor
to fundamentally change the mechanism of electrochemical reactions in NECs: it can
result in significant enhancement of ion enrichment and ion migration [19, 20], and even
further water ionization and virtual breakdown.
The simulation results shown in Figure 2.3 were achieved by commercial software
Comsol Multiphysics® 5.2. The 2-D geometry and boundary conditions setting are
shown in Figure 2.4 (take gap distance of 5 μm as an example, only one boundary edge of
our sandwiched-like nanogap cells was simulated). The parameters setting is shown in
Table I. The equations that governed the ions movement and distribution were the
steady-state Nernst-Planck equation and the Poisson equation,
( / )
i i i i i i
J D C zF RT DC
(2.12)
2
0
/
r
(2.13)
where Ji, Di, Ci, and zi are the current density, diffusion coefficient, concentration and
charge of species i, φ is the local electric potential, ρ is the local net charge density in the
solution, εr is the static dielectric constant, ε0, F, R, and T are the permittivity of vacuum,
Faraday constant, gas constant and temperature. To simplify the problem, the εr of pure
12
water was set constant 80 in the entire solution even though near the electrode surface the
εr can be reduced to less than 10 [21].
Table I. Parameters setting in simulations for electric filed distribution in nanogap cells.
Name Value Unit Description
T0 25 degC Temperature
c_H_bulk 0.0001 mol/m^3 Bulk cation concentration
c_OH_bulk c_H_bulk mol/m^3 Bulk anion concentration
z_H 1 Cation charge
z_OH -1 Anion charge
Figure 2.4. Geometry and boundary conditions setting in finite element calculations.
13
D_H 9.31E-09 m^2/s
Diffusion coefficient,
cation
D_OH 5.26E-09 m^2/s
Diffusion coefficient,
anion
eps_H2O 80
Relative permittivity of
water
xS 0.2 nm Stern layer thickness
phi_anode 0.5 V Anode potential
rho_space F_const*(z_H*c_H+z_OH*c_OH) C/m³ Space charge density
deltaphi phiM-phi V
Electrode-OHP potential
difference
rho_surf epsilon0_const*eps_H2O*deltaphi/xS C/m² Surface charge density
phiM (at anode) phi_anode/2 V Anode potential
phiM (at cathode) -phi_anode/2 V Cathode potential
thk_nitride Manually setting μm
Thickness of silicon nitride
layer
The calculation of the Debye-length of pure water, around 1 μm, from Gouy-Chapman
theory requires infinite electrode plane and potential much smaller than 26 mV at room
temperature. Simulation results showed that, even though our modeling could not satisfy
the two requirements of Gouy-Chapman theory, the approximation value of 1 μm could
still be valid since little difference showed up between the theoretical value (from the
Gouy-Chapman theory) and simulated value (from the software simulation). Besides, the
smallest nanogap between the two electrodes we achieved was 37 nm, which was much
smaller than both theoretical value and simulated value. Thus, the claim of
“deep-sub-Debye-length” is still valid.
Stern layer had been considered in the initial setting; however, the final results had little
dependence on with or without Stern layer setting. This is probably because the
14
simulation mesh was not fine enough near the electrode surface. Mesh quality is a key
factor of the simulation results. We discovered that finer mesh near the surface greatly
enhanced the surface concentration (more obvious when large potential added). However,
further finer meshing was not possible due to limited computational resources. Here,
more accurate results might not be necessary. Quantitatively, we have demonstrated the
double layer overlapping effect, and high electric field (just voltage divided by gap
distance) uniformly distributed in the entire gap. For our current research, we determined
that these simulation results are sufficient.
2.2. Water Electrolysis in Three Systems
2.1.1. Water Electrolysis in Macrosystem
To fully explain the fundamental difference between pure water splitting and
conventional electrolyte-added water splitting, we first have to discuss why pure water
cannot be split efficiently in traditional macrosystem. As shown in Figure 2.5(a), we take
cathode and H3O
+
ions as an example. Initially near the cathode surface water molecules
can be dissociated into H3O
+
and OH
-
ions. H3O
+
ions obtain electrons from cathode
leading to hydrogen evolution; while newly-generated OH
-
ions can only transport very
slowly through the bulk solution by slow diffusion or hopping process facilitated by a
weak electric field in bulk solution. Moreover, the intrinsic concentration of H 3O
+
ions in
15
bulk solution of pure water is too low to neutralize the OH
-
ions produced near the
cathode. These lead to local OH
-
ions accumulation (so that the solution near cathode
turns alkaline) especially at the cathode surface, causing the potential of the cathode
Helmholtz plane to decrease (because of negatively-charged OH
-
ions). Such a potential
decrease reduces the potential difference between the cathode and the Helmholtz plane,
further reducing the reaction rate of hydrogen evolution and thus water splitting. In other
words, the reaction becomes very slow or even self-limited, showing a large equivalent
resistance between the cathode and the anode. These phenomena also explain the rise in
cathode overpotential, since a more negative cathode potential is necessary to allow the
reaction to continue. That is why pure water in macrosystem cannot be split efficiently.
The fundamental reason is the lack of rapid ions transport inside bulk solution.
Figure 2.5. Schematic diagram of water splitting reactions in macrosystems. (a) Pure water in
macrosystem cannot be split efficiently due to the lack of rapid ions transport inside bulk solution.
(b) In sodium hydroxide solution, water splitting reaction can keep occurring but is limited by mass
transport (mainly diffusion).
16
When a high-concentration of sodium hydroxide is present in the electrolyte (Figure
2.5(b)), there are plenty of Na
+
ions from the bulk solution that can move to partially
compensate for the charge from the newly-generated OH
-
ions near the cathode, restoring
the potential difference between the cathode and the Helmholtz plane, to reduce the
overpotential requirement and thus sustain the reaction current. A similar process occurs
at the anode, where OH
-
ions from the sodium hydroxide in the bulk solution can
compensate the H3O
+
accumulation at the cathode. In this way, water electrolysis with
strong electrolyte shows a small resistance between the two electrodes and the whole
reaction of water splitting can continue.
However, we should notice that even though the ions transport inside bulk electrolyte
solution is large enough to keep the reactions continuing, at cathode the sodium ions
transport is still limited mainly by diffusion (because of weak electric field in bulk
solution) [19, 22], which is often slower than OH
-
ions generation (i.e., reaction
electron-transfer) especially when the current density is large enough. Under steady-state
conditions, the generation rate is equal to the transport rate, but a net OH
-
ions
accumulation still occurs at the cathode. In this case, the potential on the cathode
Helmholtz plane is still affected by the OH
-
ions accumulation so that cathode
overpotential requirement still rises. Similar effect also occurs on anode, indicating low
electrolysis efficiency.
17
2.2.2. Pure Water Electrolysis in Nanogap Cell
In pure water, when the counter-electrode is placed within the Debye-length (Figure 2.6),
double layer regions of the cathode and the anode are overlapping with each other so that
high electric field exists in the entire gap. Still at cathode, newly-generated OH
-
ions can
be migrated rapidly from cathode towards anode due to large electric field in the entire
gap. When the gap distance is small enough, initially the transport rate can be even higher
than the electron-transfer rate. Once OH
-
ions are generated, they are immediately drawn
from cathode to anode, leading to such OH
-
ions waiting for electron-transfer at the anode,
rather than accumulated at the cathode. In this way, the whole reactions would continue
even in pure water, but now are limited by electron-transfer. This implies that, net OH
-
ions accumulate near the anode and net H3O
+
ions accumulate near the cathode, leading
Figure 2.6. In nanogap cell, high electric field in the entire gap can enhance water ionization and
mass transport (mainly migration), leading to pure water splitting efficiently limited by
electron-transfer.
18
to completely opposite pH-value distribution compared to macrosystem (which maybe be
good for protecting the anode against corrosion). Moreover, such net OH
-
ion enrichment
near the anode not only enhances the local concentration of the reactant ions, but also
increases the potential difference between anode and anode Helmholtz-plane (which in
fact decreases the overpotential requirement, as in the Frumkin effect [23]). According to
Butler–V olmer equation [24],
0' 0'
0 ( )/ (1 ) ( )/ F E E RT F E E RT
OR
j Fk C e C e
(2.14)
such ions accumulation can significantly increase the electrolysis current, namely water
splitting throughput.
2.3. Virtual Breakdown Mechanism
2.3.1. Virtual Breakdown Mechanism
The two half-reactions become coupled together in pure water splitting in NECs. Take the
anode as an example. At the anode OH
-
ions (the reaction ions) come from two parts: one
is from water ionization near the anode; the other part comes from the OH
-
ions migrated
from the cathode to the anode. Under steady-state condition, such two parts of OH
-
ions
are balanced with the amount of electrons from the external circuit. Notice that, although
water molecule dissociation still occurs only near the electrode (due to local ions
continuous consumption), it appears like that the water molecules are split into H 3O
+
and
19
OH
-
ions in the middle of the gap (as shown in Figure 2.7), allowing H3O
+
ions to drift
towards the cathode and OH
-
ions to drift towards the anode, respectively. In other words,
such huge electric field not only increases the transport rate, but also enhances the water
molecules ionization (i.e., ions concentration). From a microscopic perspective, the
conductivity of water has been enhanced “equivalently”. From the equation of
conductivity,
nq
(2.15)
where q is the ion charge, μ is the ion mobility and n is the ion concentration. Here the
ion charges have not changed. The ion concentration is increased but only partially
contributes to the conductivity. The fundamental change is that two half reactions are
coupled together in contrary to macrosystem, and the huge electric field within the NEC
gap leads to a significantly enhanced “apparent mobility” (for macrosystem, even though
the intrinsic mobility is high, but since there is no electric field in the bulk solution, the
mobility cannot serve to the conductivity). The total effect looks like breakdown of pure
Figure 2.7. The field-assisted effect is equivalent to the scenario that water molecules are split into
H 3O
+
and OH
-
ions in the middle of the nanogap.
20
water. However, we should point out that this effect is not traditional breakdown of pure
water, which actually requires the electric field around 1 V/Å [25], about two magnitude
orders larger than what we achieved in our experiments and what we have discussed here.
The high electric field in our NECs could not split water molecules directly. However, it
can take advantage of the self-ionization of the water molecule to H3O
+
and OH
-
ions that
are continuously consumed at the electrodes, facilitating the following equilibrium
reaction to shift in the ionization direction,
23
2H O H O OH
(2.16)
Such field-assisted ionization, plus the strong ion transport, performs in a manner similar
to the breakdown of pure water. That is why we called this field-assisted effect, “virtual
breakdown mechanism”. Consider the equivalent resistance of the solution between the
anode and the cathode for water splitting, as given by
l
R
S
(2.17)
where ρ is the resistivity, l is the length and S is the cross-section area of the resistor.
When we shrink the gap distance between the two electrodes, not only the value of L
decreases, but also the value of resistivity decreases, which in fact contributes more to the
decrease of the total resistance.
The traditional view should be revised that even pure water can be electrolyzed, when the
electrode gap is small enough. This “virtual breakdown mechanism” can be applied on
21
almost all types of weakly-ionized materials: such weak ionization actually helps to
achieve the virtual breakdown effect. These findings may inspire interesting possibilities
in electrochemical applications.
2.3.2. RC-circuit Model for Nanogap Cell
In electrochemical reactions, two processes are in serial: electron-transfer step at the
electrode, and mass transport step in bulk solution. The electron-transfer rate mainly
depends on the voltage or potential on the electrode; while the mass transport rate is the
sum of diffusion rate (depending on the concentration gradient) and migration rate
(depending on the electric field) in bulk solution. In nanogap cells, the migration rate is
much larger than the diffusion rate due to high electric field inside the entire gap. The
mass transport is mainly migration. Figure 2.8 just shows the RC-circuit model of
half-reaction of water splitting in nanogap cells.
The capacitor represents the double layer adsorption. R1 represents the reaction rate of
electron-transfer. R2 represents the mass transport rate, which is related to both voltage
Figure 2.8. RC-circuit model of half-reaction of water splitting in nanogap cells.
22
and gap distance (i.e., electric field in the gap). R3 represents the water ionization rate.
When gap distance is smaller, R2 becomes smaller; while R1 can also be smaller slightly
since localized ion concentrations become higher (new H
+
and OH
-
transported), meaning
that R1 is coupled to R2. R3 may depend on R1 because ions being consumed continuously
can enhance the water ionization. When gap distance is around Debye-length, R2 is the
largest one (bottleneck) and determines the whole reaction rate; however, when gap is
much smaller than Debye-length, R2 can be even smaller than R1, indicating
electron-transfer limited reaction. That is to say, when gap distance further decreases, the
current can obtain a saturated value that only depends on voltage. More discussion will
be in Chapter 4.
2.4. Summary
In this chapter, we theoretically analyzed the fundamental difference of the
electrochemical reactions in traditional macrosystem and in nanogap cells. Due to the
double layer regions from the two electrodes overlapping with each other, huge electric
field can be uniformly distributed inside the entire nanogap, which can enhance the ion
mass transport during the electrochemical reactions, and even “virtual breakdown
mechanism” can be achieved. In this way, even pure water can be efficiently electrolyzed
based on the nanogap electrochemical cells.
23
Chapter 3. Device Fabrication
In this chapter, we will discuss how we designed and fabricated our nanogap
electrochemical cells. Metal-dielectric-metal sandwiches-like nanostructures were
selected as our nanogap cells. We will also discuss our techniques to prevent short-circuit
of our nanogap cells, including the low DC-bias silicon nitride anisotropic etching
technology, in order to enhance the fabrication yield.
3.1. Device Design
Many efforts [26, 27] have been attempted to fabricate nanogap electrodes.
Electron-beam and ion-beam lithography-defined nanogap electrodes may not be scaled
to large-area fabrication. Chemical-synthesized electrodes [28, 29] and
mechanically-fabricated electrodes [30, 31] may suffer from the lack of controllability.
Sacrificial-layer based nanogaps [32-34] require complicated processes and thus give a
poor yield [22] especially for nanogaps less than 100 nm. Bohn et al [19, 35, 36] and
White et al [20, 22] have done excellent work on nanogap based reversible redox cycling
analysis at low ionic strength; however, their structures may not be suitable for
irreversible reactions, especially with gas evolution.
The requirements of our nanogap electrochemical cells are: first, the gap distance should
be easily achieved down to deep-sub-Debye-length region; second, it should be robust
enough during water electrolysis; third, it should be suitable for gas evolution.
24
3.1.1. Material Selection
In terms of material selection for cathode and anode for water electrolysis, we selected Pt
as the cathode and gold as the anode. Pt is famous for its ability to catalyze hydrogen
evolution [37] (due to the low overpotential requirement). Normally the anode material in
industrial water electrolysis is nickel [37] due to its low cost and good ability to catalyze
oxygen evolution; however, during our experiments we discovered that nickel would get
electrochemically eroded during water electrolysis. Figure 3.1 shows the Pourbaix
diagrams of gold and nickel [38]: gold is more stable (meaning that gold is harder to
become ion-state) under higher voltage, although even gold can become ion-state when
the voltage is high enough. The ion-state of the anode material can lead to anode damages
due to anode oxidation, meaning that the metal atoms can be dissolved and then
Figure 3.1. Pourbaix diagrams of gold and nickel.
25
redeposited [39] during water electrolysis, which would cause short-circuit between two
electrodes especially when the distance of the two electrodes is too small. That is why,
even though gold may not be the best catalytic material for oxygen evolution, it is still
selected as our anode materials, basically due to its stability towards anodic oxidation, in
order to avoid short-circuit between the two electrodes during electrolysis and extend the
device lifetime.
3.1.2. Electrode System Design
In our project, we did not use the sacrificial layer technique to achieve the nanogap cells,
but use open cells which is good for gas evolution. Initially we had two different designs
of the nanogap electrochemical cells, vertical nanogap design and horizontal nanogap
Figure 3.2. Vertical nanogap system and horizontal nanogap system.
26
design, as shown in Figure 3.2. The vertical design is the triple-layer
metal-dielectric-metal sandwiched-like nanostructure, where two electrodes are in stack
and separated by the middle layer of dielectric. The gap distance is just the thickness of
the dielectric layer, while the lateral size of the electrodes could be in micron-scale. That
is to say, no nano-dimensional fabrication techniques have to be involved, meaning that
the processes are relatively easy to handle. However, it requires the good quality of the
dielectric otherwise leakage current cannot be avoided. Meanwhile, notice that the
deep-sub-Debye-length nanogap is only at the edge of the electrodes (where the two
electrodes close enough to each other), meaning that the reactions can only occur at the
edge region, which would reduce the effective reaction area. Dense lateral electrode
pattern can enhance the effective reaction area but it will require nano-dimensional
fabrication (e.g., nanoimprint). At our current stage, that may not be necessary.
Another approach to achieve deep-sub-Debye-length nanogap cells is the comb-like
horizontal design. Different from the vertical nanogap electrodes system, both anode and
cathode are in the same plane in the horizontal system. In another word, both anode and
cathode are contacting to the substrate directly, and the gap distance is just the lateral
comb-gap. In this way, the effective reaction area is just about the entire electrode
covered area, which can be easily calculated. However, notice that such horizontal design
requires nano-dimensional fabrication since now the critical dimension is down to several
tens of nanometer: electron-beam lithography and nanoimprint lithography should be
used. Electron-beam lithography is used firstly to make initial mold, and then
27
nanoimprint lithography can be used to transfer the pattern from the mold to the substrate.
However, in this case, two nanoimprint runs should be done for anode and cathode,
respectively, since they require two different metals. In this way, the alignment in
nanometer scale is necessary. With the capability of our facility, that would become quite
challenge.
Therefore, we selected vertical design as our nanogap electrochemical cells. Silicon
nitride has been selected as the dielectric layer because of its stability in terms of both
electrochemistry and mechanics (other commonly-used dielectric materials, e.g., silicon
dioxide, aluminum oxide, would get chemically eroded during water electrolysis). The
details of the fabrication process of the vertical nanogap cells are discussed in the next
section.
3.2. Fabrication Process
3.2.1. Process Details
The fabrication procedure of our open-cell sandwiched-like NECs is shown in Figure 3.3.
First, a layer of 100 nm silicon oxide film was grown onto a 3-inch silicon wafer (<100>)
by thermal oxidation in a furnace (Thermco Products Corporation, MB-71) for 50 min.
Then, 100 nm Pt on top of 3 nm Ti (adhesive layer) was deposited onto the substrate by
e-beam evaporation (Temescal BJD-1800 E-Beam Evaporator) with deposition rate of 0.3
28
Å/s. A layer of silicon nitride (thickness varied from 37 nm to 1.4 μm) was then deposited
onto the substrate by plasma-enhanced chemical vapor deposition at 350 °C (PECVD,
Oxford PlasmaPro System 100). Afterwards, a lift-off layer (Shipley Microposit LOL
Figure 3.3. Fabrication procedures and results of our metal-dielectric-metal sandwiched-like NECs.
This fabrication method can be simply fabricated on large area with high yield. Dimensions: the key
parameter of the gap distance between the two electrodes, or thickness of silicon nitride, varied from
37 nm to 1.4 μm; thermal silicon dioxide, 100 nm thick; Pt, 100 nm thick; Ti, 2 nm thick; gold, 40
nm thick; Cr, 10 nm thick; the contact pads were 3.5 mm by 3.5 mm; the grating regions were 1 cm
by 1 mm, with different pitches from 10 μm to 80 μm.
29
2000, 3000 rpm spin coating for 60 sec and baking at 170 °C for 10 min) and photoresist
layer (AZ MiR
TM
701, 3000 rpm spin coating for 40 sec and baking at 90 °C for 1 min)
was deposited onto the substrate. UV light exposure was performed at 54.3 mJ/cm
2
with
custom-designed photomask. Afterwards, post-bake was performed at 110 °C for 1 min
and the substrate was immersed into the developer solution (AZ 300 MIF Developer) for
1 min. Then Ti (2 nm, adhesion layer), gold (40 nm, top anode metal) and Cr (10 nm, dry
etching mask) were directionally deposited onto the substrate by e-beam evaporation
(Temescal BJD-1800 E-Beam Evaporator) with deposition rate of 0.3 Å/s. Next, a lift-off
process was performed by immersing the substrate into acetone solution with ultrasound
vibration and spraying the substrate with acetone solution. The substrate then was
immersed into the developer solution (AZ 300 MIF Developer) for 90 sec to remove the
residual lift-off layer. The patterns after the lift-off process consist of contact pads and
1-D gratings with different values of pitch (the contact pads were 3.5 mm by 3.5 mm; the
grating regions were 1 cm by 1 mm, with different grating pitches from 10 μm to 80 μm,
and 50% duty cycle). Here only the top gold anode was patterned and the bottom Pt
cathode was just a blank film. Next, the silicon nitride was etched with Cr as mask by
low DC-bias anisotropic reactive-ion etching (RIE, Oxford PlasmaPro System 100) that
was developed by us [40], to avoid short-circuit between the top and bottom electrodes
(more details in Chapter 3.3). Then, the Cr mask was removed by Cr wet etching (Cr-7,
MicroChemicals GMDH, 5 min immerse), which can also increase the hydrophilicity of
the entire surface. Finally, the sample was completely cleaned by acetone, isopropyl
30
alcohol, and deionized water (DI water) to remove all residual ions for further
measurement. The whole process is yield-controlled and can be scalable to mass
manufacturing.
The experimental set-up is also schematically shown in Figure 3.3, with two electrode
tips connected to the anode and cathode, and pure DI water was dropped to cover the
grating region. The hydrophilicity of the entire surface guaranteed that the water
completely wetted the whole electrode structure and gaps. Notice that the field-driven
pure water splitting only occurs at the boundary (edges) of each grating line, meaning
that denser gratings can provide larger total electrolysis current. Figure 3.3 also shows the
fabrication results (40 μm grating pitch and 72 nm gap distance as an example) observed
by unaided eyes (top view), by optical microscopy (top view) and by scanning electron
microscopy (SEM) (cross-section view). Compared to the nanogap cells reported in other
researches, we have pushed the controllable electrode gap distance to 37 nm (mainly
limited by the capacity of our PECVD equipment, thinner nitride films would have more
defects that will lead to leakage current), with high fabrication yield.
3.2.2. Efforts to Avoid Short-circuit
Much effort has been made to guarantee the high fabrication yield of our nanogap
electrochemical cells. Except the most important step, low DC-bias silicon nitride etching,
that will be discussed in the next section, three other processes have been developed to
avoid short-circuit of our NECs. The first process is to add one layer of LOL layer under
31
the photoresist layer before we do photolithography. Because we use lift-off process to
pattern the top metal layer, the sidewall of photoresist is required to be vertical enough
for good lift-off results. However, we discovered that sometimes if the result of
photolithography was not good enough, “rabbit ear” structure would form after lift-off
process due to the non-vertical photoresist sidewall, as shown in Figure 3.4. In this case,
the metal atoms on the sharp tips are more easily to be sputtered out during the silicon
nitride etching, which would lead to top and bottom electrodes short-circuit. The solution
to solve this issue is to add one layer of LOL 2000 under the photoresist layer before we
do the photolithography, and make sure that the thickness of LOL layer is larger than that
of metal layer. During photoresist development, the developer can also wet-etch LOL
layer, creating lateral undercut below the photoresist layer. In this way, during the top
metal evaporation, the metal layer cannot contact the sidewall of either the photoresist
Figure 3.4. “Rabbit ear” structure formed after lift-off process due to the non-vertical photoresist
sidewall. The figure shown here was after the silicon nitride etching, which made the “rabbit ear”
clearer.
32
layer or the LOL layer, and thus the “rabbit ear” structure can be avoided.
Another process to avoid short-circuit is to use spray gun to clean the device during
lift-off process. During the normal lift-off process where the sample was immersed into
acetone with ultrasound vibration, sometimes the metal particles cannot be removed
completely. Even worse, sometimes top metal branches can form if the e-beam
evaporation is not vertical enough, as shown in Figure 3.5. During the nominal lift-off
process, the ultrasound vibration can break the metal branches, forming a bridge from top
metal to bottom metal after nitride etching (meaning leading to short-circuit). The
fundamental reason is just ultrasound vibration is not strong enough to completely
remove all such metal particles or branches. That is why we have to use spray gun with
acetone to fiercely clean the device after ultrasound vibration.
The third effort to avoid short-circuit is to use Cr as the etching mask during the silicon
nitride etching. We compared two methods: to directly use gold as the etching mask and
Figure 3.5. Top metal layer with one single line under optical microscope. Top metal branches can
form before silicon nitride etching, leading to short-circuit between top and bottom metal layer.
33
to add Cr above gold as the etching mask. The results showed that with gold as the
etching mask directly, almost all the devices got short-circuit after etching; however, Cr
as the etching mask can prevent the short-circuit significantly. This is because gold has
much larger sputtering yield than Cr [41], as shown in Figure 3.6, indicating that with the
same etching bombardment effect, Cr atoms are much harder to be sputtered out than
gold atoms. That is why Cr as the etching mask can avoid short-circuit between the top
and bottom electrodes significantly.
Other details have also been noticed in order to prevent short-circuit. We discovered that
by using the multimeter to measure the resistance between the top and bottom electrodes,
the device was probably got short-circuit. We found that it was because the multimeter
Figure 3.6. Sputtering yield of different materials [41].
34
probe could punch through such thin silicon nitride layer. In this case, we did not use
multimeter to measure the device resistance any more, but used probe station to directly
measure the IV-curves.
3.3. Low DC-bias Silicon Nitride Etching
3.3.1. Traditional High DC-bias Etching
Low DC-bias silicon nitride etching is the key breakthrough in our project to achieve
high-yield nanogap electrochemical cells. In experiments, we discovered that traditional
nitride etching with high bombardment effect could lead to very low yield of device
fabrication: most of the devices got short-circuit after nitride etching. Commonly-used
RIE methods usually take advantages of high DC-bias (self-bias) voltage to achieve
perfect anisotropic etching profile; however, as shown in Figure 3.7, high-energy ions
will bombard the top and bottom metal layers during etching process when DC-bias is
high, resulting in metal atoms sputtered out everywhere and then re-deposited back onto
the substrate [42-44]. In some cases, especially when the silicon nitride layer is very thin,
those metal atoms will be unfortunately re-deposited onto the sidewall of silicon nitride
and form an electrical current path, in another word, short-circuit between two metal
layers.
In another word, low bombardment, or low DC-bias, is necessary in order to avoid
35
short-circuit between nanogap electrodes. However, there is a tradeoff between low
DC-bias and anisotropic etching profile. In general, both physical etching and chemical
etching are exploited during dry etching processes, to keep balance between vertical
profile and high selectivity. In experiments, we only used SF6 and C4F8 (and sometimes
O2) as etching gases. SF6 was mainly used to generate F and SFx free radicals to etch
SiNx, while C4F8 was the source of carbon-based passivation layer to protect the sidewall
from etching. By carefully controlling the parameters of etching recipes, passivation and
etching will be in dynamic equilibrium at the sidewalls and therefore vertical sidewalls
can be achieved even when DC-bias is very low.
Figure 3.7. Metal atoms sputtered out and re-deposited at high DC-bias etching, leading to
short-circuit between two metal layers.
36
3.3.2. DC-bias v s . Etching Parameters
Four factors were studied: capacitively coupled RF power, inductively coupled plasma
(ICP) power, the pressure in the etching chamber, and the combination of etching gases.
Here we tried different etching recipes, and observed the DC-bias variation and the final
etching profiles. The RIE etcher was Oxford Plasmalab system 100.
Figure 3.8. The relationship between DC-bias / etching profile and (a) capacitively coupled RF
power, (b) inductively coupled plasma (ICP) power, (c) the pressure in the etching chamber, and (d)
the combination of etching gases. The scale bar is 400 nm.
37
Figure 3.8(a) shows the relationship between DC-bias and capacitively coupled RF power
(or “RF power”). When RF power increases, DC-bias increases monotonically. However,
there is a striking RF power below which the DC-bias keeps at zero. At this striking
threshold, DC-bias is not stable, and could be either very high or zero. However, once the
DC-bias has been activated, it is maintained easily. When the pressure is low enough, it is
easier for striking because the mean-free-path gets larger so electrons can accumulate
enough energy to excite the gas molecules. And, when ICP power is higher, the striking
RF power needed is smaller because ICP power increases the plasma density. Moreover,
we discovered that it is difficult to find low DC-bias recipe at high pressure. Experiments
showed that the minimum DC-bias with struck plasma at high pressure is very large.
Figure 3.8(a) shows two SEM images at relatively low DC-bias (15 V and 25 V
respectively). The one with lower RF power had lower DC-bias, and correspondingly
achieved lower etching rate and more isotropic etching profile due to lower ion energy to
remove the passivation layer and enhance the chemical etching rate at the bottom.
The relationship between DC-bias and ICP power is shown in Figure 3.8(b). The red line
shows that DC-bias first turns on suddenly (the plasma strikes) and then decreases slowly
along with ICP power increasing. The turn-on points are when the ionization rate is
sufficient to compensate the electron loss (the blue line only shows the data after its
turn-on point and the black line only shows the data before its turn-on point). The turn-on
points increase if the pressure increases, because mean-free-path gets shorter so more
power are necessary to generate sufficient ionization rate. After the turn-on points,
38
increasing ICP power results in larger plasma density, so the ion flow current towards the
substrate increases. At fixed RF power, the product of ion flow current and bias voltage
between two capacitive plates is conserved, therefore larger ICP power will lead to lower
DC-bias. Two SEM images from the red line and the black line in Figure 3.8(b) show
non-etching results when the values of the ICP power are smaller than the turning points.
The etching rate is almost zero even at high DC-bias because the plasma density is
insufficient to generate enough radicals and to etch efficiently. Comparison within three
SEM images from the blue line indicates that ICP power affects the etching rate
significantly: larger ICP power leads to faster etching. However, larger ICP power may
not be a good choice even it will decrease the DC-bias, because lateral etching will also
become faster so that the profile will become even more isotropic.
The relationship between DC-bias and the pressure in the chamber is shown in Figure
3.8(c). At low pressure, DC-bias increases as the pressure increases, which may result
from increasing amount of charged particles (ions and electrons). However, we also
discovered that at high pressure, DC-bias decreases as the pressure increases due to
shorter mean-free-path. We also found that at very high pressure, the etching rate is much
slower, which results from insufficient kinetic energy of the ions at high pressure with the
fixed RF power and ICP power. When the pressure is too high (e.g., 50 mtorr), in fact the
plasma is difficult to be struck and the DC-bias keeps zero when RF power is relatively
low. People [45] have discovered that, low pressure can avoid etching product substrate
39
redeposition and also may reduce the negative taper angle for trench etching. Therefore,
here low pressure is our primary selection for low DC-bias etching.
When the total flow rate is constant and the pressure is low, DC-bias decreases as C-to-F
ratio increases, as shown in Figure 3.8(d). That may be because carbon atoms are easier
to be ionized than fluorine atoms [46]; therefore, the gas in the chamber is more
conductive with increasing C-to-F ratio. Another reason may be the electronic attachment
of fluorine plasma, in which F ions tend to catch electrons to form F
-
ions. This leads to
less free electron density between the capacitive plates then higher DC-bias voltage.
While at high pressure, the DC-bias can either increase or decrease along with the
increasing C-to-F ratio, depending on different recipes. The C-to-F ratio only plays a big
role on the etching profile, coupled with the effect by DC-bias.
3.3.3. Optimal Etching Recipe
By meticulously controlling the parameters of etching recipes, we eventually found an
optimal recipe with DC-bias even below 20 V and nearly 90° sidewall, as shown in
Figure 3.9, and the etching rate can be larger than 80 nm/min, which was fast enough for
etching thin silicon nitride film in nanometer scale. The parameters were also shown in
Figure 3.9. In fact, exactly vertical sidewalls were not required because in fact a little bit
isotropic etching was desired since anode tips at the boundary could form higher electric
field, which may benefit for the electrochemical reactions. Moreover, further study shows
that the selectivity between silicon nitride and Cr based on our optimal recipe was larger
40
than 100:1.
More significantly, our experiments showed that this recipe can be used to fabricate
sandwiched-like nanostructures without causing short circuit between top and bottom
metal layers. The samples were tested under a common silicon nitride dry etching recipe
with DC-bias of around 300 V and our optimal recipe with DC-bias of only 19 V for
comparison. By measuring the resistances, we demonstrated that our recipe can avoid
short-circuit nearly 100% between two metal layers (resistances out of the range of our
multimeter) while the common recipe always leads to short-circuit (smaller than 100 ohm)
between the top and the bottom metal layers. This means that our low DC-bias etching
technology can significantly enhance the fabrication yield of our nanogap
electrochemical cells.
In experiments, wet etching (without Cr mask) of silicon nitride using buffered oxide etch
(7:1) was attempted as well because it naturally avoids the ion bombardment effect.
However, the two electrodes always got short-circuit which may be attributed to capillary
Figure 3.9. The recipe parameters and the etching profile of our low DC-bias silicon nitride
etching.
41
contact of the two electrodes because of lateral etching undercut. Therefore, this method
had been given up.
3.4. Summary
In this chapter, we systematically discussed how we designed and fabricated our nanogap
electrochemical cells. The vertical-design metal-dielectric-metal sandwiched-like
nanostructures were selected as our nanogap cells. The efforts to avoid short-circuit
between the two electrodes, especially the low DC-bias etching technology to lower the
bombardment effect, have been discussed. We have pushed the gap distance between
anode and cathode to 37 nm, with high fabrication yield and good controllability, which
is a big breakthrough compared to other reported nanogap cells.
42
Chapter 4. Water Splitting Experiment Results
In this chapter, we will discuss and compare the experiment results of pure water
electrolysis and high-concentration sodium hydroxide solution electrolysis, both based on
our nanogap electrochemical cells. Fundamental analysis of the effect from the electrode
gap distance has been discussed and the phase diagram has been proposed. This is the
first time that people achieve pure water electrolysis, and the first time that people realize
that by just adjusting the gap distance, the fundamental performance of electrochemical
reactions has been changed significantly. Efficiency analysis, anode damages analysis
and bubble effects have been also included.
4.1. Pure Water Experiments
4.1.1. IV-curve Measurement
Figure 4.1. I-V curves measurement based on our NECs with pure DI water. (a) Linear I-V curves
showed larger current generated from smaller gap distances. (b) A voltage plateau around 0.9 V
shown on the log I vs. V curves.
43
When exposed in air, CO2 can dissolve into pure water, forming HCO3
-
or CO3
2-
. Such
ions can increase the ion concentration inside water, leading to Debye-length decreasing.
When at CO2 saturated equilibrium, the pH value of water is around 5.7 [47], and the
Debye-length of “pure water” decreases from 1 μm to around 220 nm. Even in this case,
we can still claim that our nanogap cells are within deep-sub-Debye-length range, since
our smallest gap distance was 37 nm, the double layer at each electrode has been at least
compressed into 1/10 of the original Debye-length. In this case, still the huge electric
field can uniformly distribute inside the entire gap, which is inversely proportional to the
gap distance at a given voltage.
Figure 4.1 shows the I-V curves from pure water measurements based on different gap
distances from 37 nm to 1.4 μm. The experiments conditions were 22 °C, 1 atm, humidity:
45%, scanning step: 50 mV , hold time: 1.5 s, delay time: 1.5 s to guarantee steady state.
The devices were with 40 μm pitch (50% duty cycle). Quite different from the common
knowledge we learned from high-school Chemistry, this is the first time that even pure
water can be electrolyzed at room temperature! The fundamental mechanism has been
discussed in Chapter 2. We can also see that, when gap distance shrank, the electrolysis
current became larger. This makes sense because higher electric field can form between
two electrodes when the gap distance is smaller, indicating larger migration rate.
Theoretically, the energy requirement for water electrolysis can be calculated from the
difference of entropy or enthalpy between reactant “water” and product “hydrogen” and
“oxygen”, as shown in the equation,
44
2 2 2
22 H O H O
(4.1)
We can calculate the minimum required voltage for the reaction in water from the change
of Gibbs free energy. For pure water reaction:
2 2 2 _
2 ( ) ( ) 2 ( )
0 0 2*( 237.14 / )
474.28 /
liquid
G G H G O G H O
kJ mol
kJ mol
(4.2)
In this reaction, 4 mole electrons are transferred. Therefore, the voltage required is
/ (4* * ) 1.229
GA
V G e N V
(4.3)
However, it is necessary to bring external energy corresponding to enthalpy to operate the
electrolysis cell since all these reactions are absorbing heat. Gibbs free energy represents
the minimum work necessary for the reaction to proceed, and the reaction enthalpy is the
amount of energy (both work and heat) that has to be provided so the reaction products
are at the same temperature as the reactant. Therefore, if enthalpy considered, the voltage
for water splitting can be calculated from the Standard enthalpy of formation,
2 2 2 _
2 ( ) ( ) 2 ( )
0 0 2*( 285.8 / )
571.6 /
/ (4* * ) 1.48
liquid
HA
H H H H O H H O
kJ mol
kJ mol
V H e N V
(4.4)
Therefore, 1.48 V is the theoretically minimum required enthalpic potential to achieve
water electrolysis [9] if there is no heat that absorbed by the reactant water. In the
experiments, we did not intentionally input heating into our measurement platform.
Moreover, we also attempted nickel (famous for its lower overpotential for oxygen
evolution) as the anode. However, the threshold voltage was also above 1.5 V . (Nickel
45
was more easily to get damaged by anodic oxidation than gold (more details in Chapter
4.2), therefore we gave up nickel at that time.) Therefore, we believe that in our
experiments the minimum required voltage for water electrolysis is just 1.48 V , and the
difference between the threshold voltage 1.5 V (shown in Figure 4.1(a)) and
thermodynamic voltage 1.23 V mainly comes from the enthalpy requirement, rather than
the overpotential. That is to say, the threshold voltage around 1.5 V indicated a small
overpotential on the gold anode. The hypothesis reason has been discussed in Chapter 2:
it is mainly because of the ion distribution that is different from that in macrosystem.
Figure 4.1(b) shows the log I vs. V curves, still based on the same data shown in Figure
4.1(a). A voltage plateau around 0.9 V was observed in this figure, which was
independent of the gap distance. This may result from the dissolved oxygen reduction
since the DI water was not saturated with inert gas, e.g., nitrogen; another reason may be
surface oxide formation on gold during water electrolysis [48-50]. The entire surface
Figure 4.2. A typical I-V curve comparison between dry sweep and pure water test. The device
measured here was with 72 nm silicon nitride thickness and 20 μm grating pitch.
46
became more hydrophilic after the first test, which was consistent with surface oxidation
or hydroxide formation. More detailed discussion can be found in Chapter 4.2, where the
log I vs. V curves are compared from pure water and from high-concentration sodium
hydroxide solution.
Before each test of IV-curve measurement with pure water or high-concentration sodium
hydroxide solution, dry voltage-sweep without any liquid was always executed first, in
order to characterize how large the leakage current is (or, how good our fabrication result
is). If the leakage current is very large, which is mainly due to some defects or pin-holes
in silicon nitride layer, the device will be given up. Normally, the quality of the silicon
nitride layer in most of the devices were good enough within our measurement scope. As
shown in Figure 4.2, typically the dry sweep current was more than three orders of
magnitude smaller than the current from pure water tests. Moreover, such background
current was with little hysteresis, indicating there were few shunts or ion migrations
inside our silicon nitride layer.
4.1.2. Electron-transfer Limited Reactions and Phase Diagram
As discussed in Chapter 2, in electrochemical reactions there are two processes in serial:
electron-transfer step at the electrode, and mass transport step in bulk solution. The
electron-transfer rate exponentially depends on the potential added onto the electrode.
Normally the mass transport rate is the sum of diffusion rate (depending on the
concentration gradient) and migration rate (depending on the electric field) in bulk
47
solution. In nanogap cells, the migration rate is much larger than the diffusion rate due to
the huge electric field inside the entire gap, meaning that the diffusion rate can be ignored
when we consider the entire mass transport rate. In nanogap cells, the electric field, as
well as the migration rate, is of course dependent on the voltage on the two electrodes
and gap distance between them. In another word, in nanogap cells the mass transport rate
depends on the gap distance while the electron-transfer rate does not.
In chemical kinetics, the overall rate of a reaction is often approximately determined by
the slowest step, known as the rate determining step. In a plot of electrolysis current vs.
gap distance
-1
(a scale of electric field) at each voltage, if the reaction is limited by
Figure 4.3. The plot of electrolysis current vs. gap distance
-1
at different voltages demonstrated that
the pure water splitting was limited by electron-transfer step.
48
electron-transfer, the current should be independent of the gap distance; however, if the
reaction is limited by mass transport, the current should be sensitive to the gap distance
(showing a large slope). Figure 4.3 clearly demonstrated such effects. For large gaps (still
comparable to Debye-length), a large slope appeared on the figure, indicating that the
reaction was mass-transport limited. That is because the gap distance was not small
enough to enhance the migration rate to the same level of electron-transfer rate: the
migration is still the bottleneck. However, when the gap was small enough
(deep-sub-Debye-length), the enhanced migration rate can be even faster than the original
electron-transfer rate. In this case, as shown in Figure 4.3, the current reached saturation
value that only depended on the voltage, indicating an electron-transfer limited reaction.
This is the first time that people have successfully pushed the electrochemical reactions
to electron-transfer step limitation, over the entire range of current densities.
The critical gap distance (or “turning point”) between such two states became smaller
(moved to the right on the figure) with increasing voltage. This is due to the different
effects from the voltage to electron-transfer rate and mass transport rate: electron-transfer
rate increases exponentially along with voltage increasing; while mass transport (mainly
by migration) rate increases linearly along with voltage increasing. It means that the
electron-transfer rate increases faster than the mass transport when voltage increases.
Therefore, in order to achieve saturation current (electron-transfer limited) at higher
voltages, smaller gaps are necessary as a compensation to the slow increasing of the mass
transport compared to the electron-transfer step.
49
Such electron-transfer limited reactions also indicate that, completely opposite
distribution of pH-value (compared to the macrosystem) occurs in pure water splitting in
nanogap cells: H3O
+
ions accumulate near the cathode and OH
-
ions accumulate neat the
anode. Otherwise, the electrolysis current cannot achieve saturation value. This also
demonstrates the fundamental analysis in Chapter 2.
We can even deduce this analysis to a more general picture, as shown in Figure 4.4: the
phase diagram of electrochemical performance vs. gap distance. Three phase regions are
considered. For traditional macrosystem where the electrode gap distance is much larger
than the Debye-length, two half-reactions are decoupled and cannot feel each other.
Normally the electrochemical current is limited by slow diffusion step. When gap
Figure 4.4. Phase diagram of electrochemical performance vs. gap distance.
50
distance has been reduced to around Debye-length, large electric field can form between
the two electrodes to enhance the mass transport rate. In this region, the current is very
sensitive to the gap distance and the reactions are migration limited. When the gap
distance is further reduced to deep-sub-Debye-length region, the mass transport can be
enhanced further to the level even faster than electron-transfer step. In this region, even
we shrink the gap distance further the current cannot be enlarged any more, meaning that
the current has reached saturation. Here the two half-reactions are coupled together and
the reactions are electron-transfer limited. This is the first time that people have proposed
such phase diagram and realized that by just adjusting the gap distance the fundamental
performance of the electrochemical reactions can be significantly changed.
4.1.3. Electrolysis Efficiency Analysis
There are two ways to calculate the electrolysis efficiency. When the electrolysis occurs
at voltage Velectrolysis and current Ielectrolysis, the total power consumption is
Velectrolysis*Ielectrolysis. Then the efficiency can be calculated as,
2
/
_
*
*
*
total
electrolysis electrolysis
mol s H
electrolysis electrolysis
Hydrogen Energy
Efficiency
VI
Hydeogen Heat
VI
(4.5)
Where the Hydrogenmol/s is the hydrogen generate rate (mol/s), and Heat H2 is the total
combustion heat of each mole hydrogen, which is 286 kJ/mol. Here we use the value
known as “higher heating value”, or HHV (compared to the other term, “lower heating
51
value”), which is determined by bringing all the products of combustion back to the
original pre-combustion temperature, and in particular condensing any vapor produced
[51]. This is the same as the thermodynamic heat of combustion since the enthalpy
change for the reaction assumes a common temperature of the compounds before and
after combustion, in which case the water produced by combustion is condensed to a
liquid, hence yielding its latent heat of vaporization. In fact,
2
2* * *
HH
Heat e Na V
(4.6)
Where e is the one electron charge, and Na is the Avogadro constant, and VH is 1.48 V ,
the theoretically minimum required enthalpic potential to achieve water electrolysis, as
we discussed. It means that both higher heating value and enthalpic voltage are the
energy terms that are related to enthalpy change in the reaction. In fact, this enthalpic
voltage is also called “higher heating value” voltage, which corresponds to the heat
content of the dry product gases with respect to liquid water at 25°C. It is thus the correct
parameter for calculation of electrical energy efficiency [9].
If we consider the ideal case that all the electrons from Ielectrolysis make contributions to the
eventual hydrogen generation (meaning that, for each electron, it is corresponding to one
hydrogen ion and such hydrogen atom becomes part of the hydrogen gas that are
evolving out), then the value of Ielectrolysis can be calculated as:
/
**
electrolysis mol s
I electron Na e
(4.7)
Where electronmol/s is the number of electrons that are combined with hydrogen ions per
second. Notice that, each hydrogen molecule consumes two electrons. Combine equation
52
(4.5) - (4.7), we can obtain that,
1.48
H
electrolysis electrolysis
V
Efficiency
VV
(4.8)
This result means that the final efficiency is only related to the electrolysis operating
voltage, and lower operating voltage can provide higher efficiency! However, very low
voltage cannot turn on the electrolysis. The operating voltage should be at least larger
than Vthreshold, i.e., 1.48 V plus the overpotential in our measurement. In another word, the
smaller the overpotential requirement, the higher the efficiency. Our measured IV-curves
showing the Vthreshold around 1.5 V demonstrated that the efficiency in our pure water
electrolysis based NECs could be almost 100%, which is a huge improvement compared
to the industrial water electrolysis with efficiency around 60% - 70%.
However, we should notice that, if the operation voltage is low, the hydrogen throughput
would be exponentially slow although the efficiency is high. There is a trad-off between
efficiency and hydrogen throughput. That is why in industrial water electrolysis, a little
bit higher voltage (1.85 V – 2.2 V) is usually used [52]. Another point we should notice is
that the calculation and analysis above are based on the assumption that every electron
from the electrolysis current are contributing to the final hydrogen evolution. However,
sometimes this might not be true: several factors ensure that somewhat lower amounts of
gas are actually found. Some electrons (and product) are used up in side reactions (e.g.,
oxygen reduction), some of the products are catalytically reconverted to water at the
electrodes particularly if there is no membrane dividing the electrolysis compartments,
53
some hydrogen may absorb into the cathode. Finally, some gas remains held up in the
nanobubbles for a considerable time and some gas may escape measurement. In another
word, if we want to calculate the electrolysis efficiency accurately, the overall hydrogen
generation rate should be measured and the equation (4.5) should be used. At our current
stage, we did not measure such hydrogen generation rate because it is very difficult to
collect all the hydrogen gas based on our current set-up platform, mainly due to the quite
small quantity of the generated hydrogen from the nanogap electrochemical cells. In this
case, the hydrogen generation rate cannot be measured accurately. New devices should be
designed and new set-up method should be used if we want to enhance the reliability of
the measurement of hydrogen generation rate, in order to calculate the electrolysis
efficiency more accurately.
4.2. Anode Oxidation Damage
In the experiments, we found that sometimes the gold anode got damaged after multiple
IV-curve measurements or at higher measuring voltage. This is due to anode oxidation
during water electrolysis. In fact, anode can be roughed during redox cycling [39, 48].
The roughness comes from the electrochemically oxidation (forward sweeping) and
re-deposition (backward sweeping) of the anode metal, even for gold [48, 53]. With gold
as the anode for water electrolysis, people found Au
3+
in electrochemically grown gold
oxide layers, and concluded that electrochemical Au2O3 is highly disordered. The
54
Pourbaix diagram of gold shown in Chapter 3 also indicates that at higher voltage the
gold anode would be oxidized.
In our experiments, we also discovered such anode oxidation. The entire surface became
more hydrophilic after the first test, which was also consistent with surface oxidation or
hydroxide formation. The anode damage sometimes occurred when the applied voltage
values were above 5 V , as shown in Figure 4.6. Thinner-gap samples were more likely to
suffer damage. Moreover, damage always showed up near the grating boundary where the
electric field was the highest. Such damage, especially the re-deposition of gold atoms,
Figure 4.5. Even gold anode can become roughed during water electrolysis [48]. Here * stands for
an adsorption site on the electrode.
55
can form conductive gold (or oxide) clusters sometimes with the size comparable to the
gap distance in our nanogap electrochemical cells. This would probably lead to
short-circuit between the anode and the cathode especially for smaller-gap devices, and
thus reducing the lifetime of the devices. To avoid such short-circuit, the backward
sweeping was removed and only forward sweeping was accomplished in order to
minimize the re-deposition process, and the maximum external voltage for forward
sweeping was set to be 2.5 V (to reduce the current density in fact). In this way, the
devices can be measured repetitively without obvious damage or short-circuit.
Two possible approaches are proposed here to avoid such anode damage. Since such
anode damage came from the anode oxidation, if one material cannot be oxidized further
it would not have such damage. The conductive indium tin oxide (ITO) may be used to
replace gold as the anode material. Figure 4.7 shows the Pourbaix diagrams of indium
and tin. The highest oxidation state of both indium and tin can be quite stable over a large
Figure 4.6. Anode damage formed in pure water splitting when voltage above 5 V . The device
shown here is with 72 nm gap and 40 μm pitch.
56
range of voltage. In fact, ITO just occupies its highest oxidation state which cannot be
oxidized further [54, 55]. This may be a good material for preventing the anode damage.
However, the fabrication process of ITO should be developed to make it compatible with
the current fabrication process of our nanogap cells. Another approach to avoid anode
damage is to use an ultrathin layer of dielectric material coated onto the gold anode as the
energy-band offset material [56, 57]. The first requirement of this dielectric materials is
that the energy-band should match the anode material (e.g., gold); the second requirement
is that the thickness of this dielectric material should be small enough to be conductive to
gold while thick enough for preventing gold contacting water directly. At our current
stage, we did not attempt such methods because the anode damage was not a significant
Figure 4.7. Pourbaix diagrams of indium and tin.
57
issue. However, those methods should be developed if we want to optimize the NEC
devices further, especially when operating at large current or voltage.
4.3. Sodium Hydroxide Solution Experiments
4.3.1. Pure Water vs . Sodium Hydroxide Solution
The electrolysis of pure water and 1 mol/L sodium hydroxide solution were compared in
Figure 4.8, both based on our nanogap cells with the same gap distance 72 nm and
different grating pitches from 10 μm to 80 μm. The experiment conditions were the same
as in Figure 4.1. The number of edges are corresponding to the value of 1-D grating
pitches: since we fixed the total width of the 1-D grating region, smaller pitch can
Figure 4.8. Electrolysis current at 1.8 V vs. the number of edges from pure water splitting and water
splitting in 1 mol/L sodium hydroxide solution, both based on our NECs.
58
provide more line edges, i.e., larger number of edges. From the figure, we can conclude
that for pure water the electrolysis current at 1.8 V linearly increased with the number of
grating edges; while for sodium hydroxide solution, the current was less dependent on the
number of edges and the data dispersion was significantly larger than that of pure water.
The mechanism is shown in Figure 4.9. For pure water splitting, the reaction only occurs
at the edges where the two electrodes close to each other to couple the two half-reactions
together; at the “non-edge” region (i.e., top face of the grating line), the scenario is just
like pure water splitting in macrosystem where the reactions are self-limited. On the
contrary, in sodium hydroxide solution the entire surface could be involved in supporting
the reaction. That is because the Debye-length in 1 mol/L sodium hydroxide solution is
less than 1 nm (as discussed in Chapter 2.1), still significantly smaller than the electrode
Figure 4.9. Schematic diagram of the mechanisms showing the different reaction locations in pure
water splitting and water splitting in sodium hydroxide solution.
59
distance shown 72 nm here. In this way, the two half reactions are still decoupled and
limited by diffusion, just like the scenario in macrosystem where the reaction occurs on
all accessible parts of the electrodes. In another word, the electrolysis current from
sodium hydroxide solution significantly depends on the entire effective reaction area.
Two evidences can demonstrate the analysis above. In our present experiments, the
effective reaction area for sodium hydroxide solution was just the solution droplet
covered region. As shown in Figure 4.10, larger droplet of sodium hydroxide solution
provided larger current, indicating more surface area involved into the reactions, though
the total number of the edges was independent of the droplet size. The second evidence is,
bubbles could be even generated far away from the counter-electrode (i.e., non-grating
region), indicating that the reactions can occur even very far from the grating edges. That
is to say, the reactions in sodium hydroxide solutions occur not only at the grating edges,
Figure 4.10. Evidence of the entire surface involved into the reactions in sodium hydroxide
solutions. (a) Larger droplet provided larger current. (b) Bubbles formed at non-grating region. The
devices were with 72 nm gap distance.
60
but also over the entire region covered by the droplet. In comparison, we did not observe
such two phenomena in pure water electrolysis: the current from pure water was not
sensitive to the area of the droplet region, and bubbles could only form within the grating
region.
In our present experiments, we did not accurately control the volume or covered region of
the droplet for each test. That is why the electrolysis current from sodium hydroxide
solution presented significant variability. However, this will be improved by fixing the
reaction area, as shown in Figure 4.11. A piece of PDMS film can be covered on the
device with an open window left above the grating region. The liquid solution can be just
dropped in this open window. Moreover, the droplet volume will also be accurately
controlled. In this way, we can make sure that the effective reaction area and the amount
of the solution are the same for every measurement.
Notice that, the effective reaction area in sodium hydroxide solution was much larger
than that in pure water (the entire surface area vs. only edge region). Even under such
unfavorable conditions, the electrolysis current from pure water is comparable to that
Figure 4.11. A PDMS coating to confine the reaction area.
61
from 1 mol/L sodium hydroxide solution, indicating that the electrolysis current density
from pure water was much larger than that from 1 mol/L sodium hydroxide solution. This
is quite different from our common knowledge: the current density from non-conductive
materials could be even larger than that from conductive materials! First, it was because
the reactions in pure water electrolysis were electron-transfer limited (as shown in Figure
4.3), much faster than the diffusion-limited reactions in sodium hydroxide solution.
Second, in fact the apparent conductivity of pure water had been significantly enhanced,
as discussed in Chapter 2.3. (The conductivity of 1 mol/L sodium hydroxide solution and
CO2-equilibrated pure water are 2×10
5
μS/cm and 1 μS/cm [47], respectively, indicating
more than 10
5
-fold enhancement of the apparent conductivity of pure water.) In Figure
4.8, the slope of the linear fitting (increased electrolysis current per edge) from the pure
water curve was significantly larger than that from the sodium hydroxide solution curve,
indicating much larger contribution to electrolysis current from the field-assisted effect
than from the diffusion effect. The extrapolated intercept value 0.32 mA of sodium
hydroxide solution also indicated the nature of entire surface being involved into the
reaction. For pure water, the background current 0.031 mA was much smaller, probably
resulting from capacitive current or ionic impurities. From the linear fitting, we can also
conclude that when the grating pitches are smaller than 2 μm, even the electrolysis
current from pure water can be higher than that from 1 mol/L sodium hydroxide solution
(2 μm pitch was beyond our photolithography capability, therefore we did not attempt it
at present stage). These results demonstrate a great potential of splitting of pure water
62
with better performance compared to conventional electrolyte-added water splitting for
hydrogen production.
4.3.2. Bubble Effect
Figure 4.12(a) shows part of our experimental set-up and bubbles generation around 2 V
during the pure water splitting. Such obvious bubbles could definitely demonstrate that
pure water had been electrolyzed with even very high electrolysis current. Sometimes
those generated bubbles were very few, which may result from nanobubbles dissolved
Figure 4.12. (a) Bubble generation around 2 V from pure water electrolysis. (b) Bubble effects on
plateaus (or peaks) around 2 V in I-V curves based on devices with 72 nm gap and 10 μm pitch in
pure water, and (c) in 1 mol/L sodium hydroxide solution.
63
into water [58, 59].
Figure 4.12 also shows plateaus (or peaks) around 2 V in I-V curves, both in pure water
measurement and 1 mol/L sodium hydroxide solution measurement. We believe that it
was due to bubble effects. Around 2 V bubble generation started to be vigorous enough so
that it could be observed by the naked eye. Moreover, devices with smaller gap distance
or smaller grating pitches had more obvious plateaus around 2 V , indicating that such
plateaus were determined by the geometry of the structures, rather than electrode
electrochemical reactions. This observation is reasonable since bubbles are more likely to
be trapped within the smaller gap or smaller pitch structures before releasing, excluding
the water involved in the reaction. Therefore, larger voltage leads to larger excluding
effect, reducing the current and showing negative resistance which performs like a
plateau or peak in I-V curves.
Notice that Figure 4.12 also shows the result consistency among several tests. The data
from different devices were almost exactly the same, especially below 2 V . Above 2 V,
the data had a relatively larger error range which we think was due to the bubble effects
on the current performance. Therefore, we always selected the current data below 2V for
analysis and comparison to be free from bubble related artifacts.
4.3.3. log I vs . V curves
Figure 4.13 shows the log I vs. V curves from tests of pure water and sodium hydroxide
solution on the same device, respectively. For pure water, one plateau appeared around
64
0.9 V (also shown in Figure 4.1(b)), which became flatter after the first test on the same
device. For sodium hydroxide solution, two plateaus, around 0.4 V and 1.2 V respectively,
were shown on the log I vs. V curves. After the first test, the 0.4 V plateau still existed but
the 1.2 V plateau disappeared, and the current became much larger (the 2V plateau can be
ignored since it was due to the bubble effects). This plateau phenomenon was quite
repeatable, no matter what the gap distance or pitch was, indicating that it was more
likely related to the intrinsic electrochemical reactions, rather than geometry factors.
However, the fundamental mechanism is still not clear.
Only a few literature reviews discussed about such plateaus in log I vs. V curves. Our
hypothesis is the following. The 0.9 V plateau from pure water tests may be attributed to
dissolved oxygen reduction or anode gold oxidation (and these two effects might be
coupled). For sodium hydroxide solution, the 0.4 V plateau most likely came from the
reduction of dissolved oxygen; while the 1.2 V plateau was related to anode gold
Figure 4.13. Plateaus in log I vs. V curves from (a) pure water tests and (b) sodium hydroxide
solution tests. The devices were with 72 nm gap distance.
65
oxidation. The different values of the oxidation plateaus in pure water and in sodium
hydroxide solution was most likely due to the difference in pH values. The 0.4 V plateau
would not disappear since for every test fresh sodium hydroxide solution (without inert
gas saturation) was used. For the gold anode, non-conductive oxide state I and conductive
oxide state II can form during water splitting [60]. During the first test in sodium
hydroxide solution, OH
-
ions concentration was so large that all surface gold could be
oxidized to state II, therefore during the second or third tests no surface gold could be
oxidized further (thus the 1.2 V plateau disappeared). Also, because oxide state II was
porous and conductive, the distance between anode and cathode had been shortened due
to gold oxide islands, and the current after the first test could become larger (the larger
current could be also attributed to roughness of the surface so that effective reaction area
became larger [39]). However, for pure water, OH
-
ions concentration was small so that
only oxide state I might form, therefore gold could still be oxidized further into the
formation of state I during the second or even third tests (until two or three monolayers of
the oxide state I coverage reached [48, 49]), with almost the same electrolysis current or
smaller since oxide state I was non-conductive. However, such plateaus may also result
from the formation of oxygen coverage [50], inhibition layer [61] or inert sites [62].
Detailed experiments are necessary to get a clearer fundamental understanding of the
mechanism underlying such plateaus. First, inert gas saturated pure water and sodium
hydroxide solution should be utilized; second, anode current and cathode current should
be studied separately by adding reference electrode into the system; third, crystal plane of
66
original gold and final anode oxidation should be further analyzed by spectroscopy
measurement. However, since this problem is beyond the scope of our present study, we
have not included such experiments in this thesis.
4.4. Summary
In this chapter, we systematically analyze and compare the experiment results from pure
water electrolysis and sodium hydroxide solution electrolysis, both based on our nanogap
electrochemical cells. This is the first time that people achieved high-efficiency pure
water electrolysis at room temperature. We also demonstrated that the reactions from pure
water electrolysis was limited by electron-transfer step, and the current density could be
even much larger than that from sodium hydroxide solution. Moreover, we successfully
proposed the phase diagram to show that by just adjusting the gap distance between two
electrodes, the fundamental performance of electrochemical reactions can be significantly
changed!
67
Chapter 5. Methanol Solution Splitting Experiment Results
In this chapter, we will discuss the electrolysis of pure methanol solution based on our
nanogap electrochemical cells. Methanol is usually regarded as a non-conductive organic
solvent. We want to utilize methanol solution to further demonstrate our concept of
virtual breakdown mechanism which can enhance the equivalent conductivity of the
solution, in order to eliminate the electrolyte requirement for electrolysis. Moreover, the
minimum required voltage for the electrolysis of methanol solution is much smaller than
that of water electrolysis, indicating its potential as a more promising technology for
hydrogen generation. In this chapter, the performance of ethanol solution was also
characterized, and compared to methanol solution, both based on different electrode
materials.
5.1. Background Analysis
Although hydrogen generated from water electrolysis is clean and pure, water electrolysis
for large-scale hydrogen production is unattractive because of its large voltage
requirement, leading to high electricity consumption. Therefore, at present more than 90%
of the industrial-produced hydrogen comes from steam reforming of natural gas and
gasification of coal and petroleum coke, even though large amount of greenhouse gas is
generated as by-product. The fundamental reason for the high-power consumption of
water electrolysis is that large energy is required at anode where O
2-
is oxidized to O2.
68
However, this half-reaction at anode is not necessary as we do not need to generate
oxygen by water electrolysis. To minimize the energy requirement, methanol electrolysis
is desired, in which the anode half-reaction is to oxidize C
2-
to C
2+
or C
4+
, with much
lower power consumption. Theoretically the minimum required voltage can be calculated
as following, the same way we calculated the energy requirement for water electrolysis as
discussed in Chapter 4.1.
For methanol electrolysis, people usually use methanol-water mixture as the reactant.
When electrolyzing, hydrogen is generated at cathode, and CO2 is generated at anode, as
the following reaction,
3 2 2 2
3 CH OH H O H CO
(5.1)
In this case, the change of Gibbs free energy in the system and the required voltage are,
2 2 2 _ 3
3 ( ) ( ) ( ) ( )
0 ( 394.39 / ) ( 237.14 / ) ( 166.4 / )
9.15 /
/ (6* * ) 0.0158
liquid
A
G G H G CO G H O G CH OH
kJ mol kJ mol kJ mol
kJ mol
V G e N V
(5.2)
However, we can also use pure methanol for electrolysis. Theoretically, the reaction is,
32
2 CH OH H CO
(5.3)
And the change of Gibbs free energy in the system and the required voltage are,
23
2 ( ) ( ) ( )
0 ( 137.16 / ) ( 166.4 / )
29.24 /
/ (4* * ) 0.076
A
G G H G CO G CH OH
kJ mol kJ mol
kJ mol
V G e N V
(5.4)
However, as we discussed in Chapter 4.1, it is necessary to bring external energy
69
corresponding to enthalpy to operate the electrolysis cell since all these reactions are
absorbing heat. Therefore, if enthalpy considered, the voltage for the electrolysis of
water-methanol mixture can be calculated from the Standard enthalpy of formation,
2 2 2 _ 3
3 ( ) ( ) ( ) ( )
0 ( 393.5 / ) ( 285.8 / ) ( 238.4 / )
130.7 /
/ (6* * ) 0.227
liquid
A
H H H H CO H H O H CH OH
kJ mol kJ mol kJ mol
kJ mol
V H e N V
(5.5)
And the enthalpic voltage for pure methanol electrolysis is
23
2 ( ) ( ) ( )
0 ( 110.5 / ) ( 238.4 / )
127.9 /
/ (4* * ) 0.331
A
H H H H CO H CH OH
kJ mol kJ mol
kJ mol
V H e N V
(5.6)
The calculations above are summarized in Figure 5.1, which also includes the
information of water electrolysis. We can see from Figure 5.1 that, the required voltage
for water electrolysis is 1.48 V , much larger than that in methanol electrolysis if water
involved into the reaction (0.227 V) or if pure methanol is considered (0.331 V). In
another word, when pure methanol is used, the required voltage can be only 22% of that
Figure 5.1. Comparison of the required voltage between water electrolysis and methanol
electrolysis.
70
in water electrolysis; when methanol-water mixture is used, the voltage can be down to
only 15% of that in water electrolysis. This certainly indicates the low-power
consumption of methanol electrolysis. Another advantage of such low electrolysis voltage
is that, the serious issue of oxidation corrosion at anode in water electrolysis can be
avoided. By using this method, Prof. Shen from Sun Yat-Sen University utilized only 1/3
voltage of water electrolysis to achieve the electrolysis of methanol-water mixture [63].
However, there is still large space to improve.
Notice that, for the electrolysis of methanol-water mixture, hydrogen can be generated
not only from methanol, but also from water. Still from Figure 5.1, we can see that carbon
dioxide or carbon monoxide is generated during the methanol electrolysis. Similar
by-products are also produced in current chemical reforming and gasification process for
industrial hydrogen production. That is to say, the similar approaches in industry to
reduce the by-products emission can be also applied on methanol electrolysis.
Figure 5.2 shows some parameters of water and methanol as a comparison. The ion
product of pure methanol is only 10
-16.7
(S.I.), indicating even fewer ions in methanol
than in pure water. However, the conductivity of pure methanol is higher than that of pure
Figure 5.2. Parameter comparison between water and methanol.
71
water. This indicates that the mobility of the ions in pure methanol is much larger than
that in pure water, which in fact can benefit the virtual breakdown mechanism where the
ion mobility is quite important in order to enhance the mass transport. The Debye-length
in pure methanol is around 3 μm, 3 times larger than that in pure water. In this way, the
deep-sub-Debye-length nanogap cells can be easier to achieve in pure methanol than in
pure water, in terms of fabrication complexity. We should point out that, in experiments
the measured conductivity of pure methanol was 1.37 μS/cm, still smaller than the
measured conductivity of “pure water” (maybe CO2 from air dissolved into it) 1.88
μS/cm. It is more challenging to electrolyze pure methanol solution than “pure water”
because methanol is even more non-conductive. However, this is a good chance to
demonstrate our concept of virtual breakdown mechanism, where strong electrolyte can
be eliminated in electrolysis.
5.2. Device Fabrication
The first question for device fabrication is the material selection for the cathode and
anode for methanol electrolysis. People discovered that the overpotential at cathode and
anode in methanol electrolysis is higher than that in water electrolysis, even for the best
catalyst for now. The mechanism is quite unknown. However, at present researchers are
usually using low-overpotential platinum-based materials for both cathode and anode.
The cathode materials are usually Pt-C, and the anode materials are usually Pt/Ru-C.
72
People also attempted Pt/WO3, Pt/Sn, and Pt/Sn-C as anode materials.
The significant reason why people use such compound materials is due to the carbon
monoxide molecules adsorption on Pt. During methanol electrolysis (or methanol
oxidation, more specifically), CO molecules, *COH, *CHOH, *CH2OH would be
massively generated at anode as intermediate by-products [64]. Those molecules or active
chemical groups could possess strong specific adsorption on Pt. Such adsorption is so
strong that can occupy the adsorption sites of hydrogen on Pt. In this case, the reaction
becomes self-limited and the current becomes much smaller. This problem is called
“carbon monoxide poisoning”. When Ru or Sn participates into Pt as alloy, hydroxy
groups can form on the anode surface to further oxidize such carbon monoxide molecules
or active chemical groups to carbon dioxide molecules or other final products (e.g.,
HCOOH, HCHO). Because those final products would not possess strong specific
adsorption on Pt, the occupied sites for hydrogen adsorption can be released. In this way,
the electrolysis reaction can keep continuing with relatively large current.
Due to the capability of our equipment, we selected Pt as the cathode (for hydrogen
evolution), and Pt/Ru alloy as the anode (for methanol oxidation). They are not the best
catalyst for the reactions; however, they are good enough at our current stage to
demonstrate the electrolysis of pure methanol solution, based on our nanogap
electrochemical cells.
As discussed in Chapter 3.1, there are two options for the design of our nanogap
electrochemical cells: vertical nanogap design and horizontal nanogap design, as shown
73
in Figure 3.2. For methanol electrolysis, we selected the vertical nanogap design as well
due to the fabrication complexity. The first reason is that we selected two different
materials for cathode and anode respectively, the same as the case in pure water
electrolysis. In this case, if horizontal nanogap design was selected, the nano-dimensional
alignment would be still necessary. With the capability of our facility, that would become
quite challenge. The second reason is that our anode material, Pt/Ru alloy, can be only
deposited by sputtering rather than metal evaporation (due to different vapor pressures).
Such sputtered film is quite conformal and cannot be patterned by lift-off process.
However, if we use etching technology to pattern the metal alloy, there is no mature
recipe for Pt/Ru alloy etching (no matter wet etching or dry etching), either. Therefore,
we had to select the metal-dielectric-metal sandwiched-like vertical nanogap design for
our nanogap electrochemical cells. Moreover, the Pt/Ru alloy (anode material) should be
the bottom metal layer, since it was difficult to pattern it. It was different from the devices
used in pure water electrolysis, where the bottom Pt layer was cathode.
The fabrication process was quite similar with the process of the devices used in pure
water electrolysis, as discussed in Figure 3.3. First, a layer of silicon dioxide was grown
by thermal oxidation. Then the bottom metal layer was deposited by physical vapor
deposition (PVD) and silicon nitride layer was deposited by PECVD on to the wafer.
Afterwards, the top metal layer was patterned on the nitride layer by photolithography,
metal deposition, and lift-off process. Finally, the silicon nitride was etched by our low
DC-bias silicon nitride etching technology. All the other parameters are the same as
74
shown in Figure 3.3, and the mainly changes here were replacing the anode material with
Pt/Ru alloy (metal sputtering, 60 nm, 0.3 Å/s, and with 3 nm Ti as adhesion layer), and
replacing the cathode material with Pt (e-beam evaporation, 60 nm, 0.3 Å/s, and with 3
nm Ti as adhesion layer). For the anode, the atom ratio of Pt and Ru was set 1:1, which
was regarded as the optimal composite ratio for catalysis [65]. High-magnification SEM
image of the fabrication result was shown in Figure 5.3.
There were two compatibility issues of Ru-involved fabrication process that we
discovered during our experiments. Notice that we did not use Cr above the top Pt layer
as the etching mask when doing the silicon nitride etching. That was because we
discovered that the Cr wet-etchant (Cr-7, MicroChemicals GMDH) could also attack
Pt/Ru alloy (mainly attacked Ru) when we removed the Cr layer by immersing the
Figure 5.3. High-magnification SEM image of the fabrication result of the device for pure
methanol solution electrolysis.
75
sample into the etchant solution, as shown in Figure 5.4(a) where some bubbles generated
(meaning that Ru had also been etched). Some other solutions, e.g., hydrochloric acid,
had also been attempted. However, the Cr layer could not be efficiently removed by
hydrochloric acid, probably due to the change of the surface properties during nitride
etching. In another word, there was no good method to remove the Cr layer if it was used
as the etching mask. However, we found that Pt itself was already a good etching mask
for nitride etching, even though the sputtering yield of Pt is a little larger than Cr (as
shown in Figure 3.6). Our low DC-bias silicon nitride etching would not attack Pt layer
observably. In this way, we did not use Cr, but used the top cathode Pt layer directly, as
the etching mask.
Another important change, compared to the devices used in pure water electrolysis, was
that the thickness of the silicon nitride layer was fixed at 72 nm. We did not attempt
Figure 5.4. Two compatibility issues of Ru-involved fabrication process. (a) Bubbles generated
when the sample with Pt/Ru was immersed into the Cr-etchant, indicating Ru had been wet-etched.
(b) Pt/Ru layer became non-uniform after silicon nitride etching, indicating Ru had been
dry-etched.
76
thinner thickness because we wanted to guarantee the high fabrication yield of our
devices. Morevoer, 72 nm was already small enough to achieve the
deep-sub-Debye-length distance (notice that in methanol the Debye-length is 3 μm). We
did not attempt thicker thickness, either, because we discovered that the etching recipe for
silicon nitride etching could also etch Ru. Figure 5.4(b) shows the silicon nitride etching
result of our device with 144 nm thick silicon nitride. The non-uniform surface indicated
that the bottom Pt/Ru layer had been etched. We did not discover such non-uniformity
when we etched the devices for pure water electrolysis where the bottom layer was Pt. In
another word, the mainly etched part should be Ru, consistent with some literature
reports [66]. Due to the unstable etching rate, thicker silicon nitride samples always had
to be overetched for longer time in order to guarantee that all nitride had been etched
away. In this way, the surface Ru layer would be etched more, sometimes even
completely etched away, leading to pure Pt left as the anode surface (which would lose
the catalysis capability as we discussed). However, for 72 nm-thick nitride samples, we
did not discover the observable Ru-etching phenomenon because the overetching time
was relatively short. Therefore, we did not use thicker silicon nitride samples, but just
fixed the nitride layer to 72 nm thick, for our experiments of methanol solution
electrolysis.
77
5.3. Experiment Results
In experiments, pure methanol solution was characterized by IV-curve measurements
based on different concentrations from 0.1 mol/L – 10 mol/L (pure methanol / pure water
mixture). The experiments conditions were 22 °C, 1 atm, humidity: 45%, scanning rage:
0 – 1.1 V to prevent water electrolysis and anode oxidation, scanning step: 15 mV , hold
time: 1 s, delay time: 1 s to guarantee steady state. The devices were with 20 μm pitch
(50% duty cycle). During the experiments, we did not use pure methanol because pure
methanol was so “hydrophilic” (or wetting) when dropped onto the device, leading to
very fast evaporation and large data variability. Moreover, pure methanol has more
serious issue of “carbon monoxide poisoning” intrinsically. However, notice that the
Debye-length from those methanol-water mixtures were in between the Debye-length
form pure methanol and pure water; even if CO2 from air considered, Debye-length from
those methanol-water mixtures should still be larger than 220 nm, meaning that our 72
nm gap distance was still within deep-sub-Debye-length region.
Figure 5.5(a) shows the IV-curve measurement results of such pure methanol solutions
based on different concentrations. The data from pure water was also included here, in
order to prove that the increasing current above 0.6 V came from the reactions from
methanol electrolysis. This definitely demonstrates that even non-conductive pure
methanol solution can be successfully electrolyzed (the conductivity of all such solutions
were measured between 0.7 – 2.0 μS/cm), which further demonstrates the virtual
78
breakdown mechanism that we discussed in Chapter 2: by using nanogap electrochemical
cells, huge electric field can be uniformly distributed inside the entire gap due to the
Figure 5.5. IV-curve measurements of pure methanol solution based on the nanogap
electrochemical cells, (a) linear I vs. V , (b) log I vs. V where the background data from pure water
had been removed.
79
double layers from the two electrodes overlapping with each other, which can
significantly enhance the mass transport inside bulk solution and thus enhance the
apparent conductivity. The mechanism of such pure methanol solution electrolysis was
quite similar with the case of pure water electrolysis, as we discussed in Chapter 2.2.
Notice that, the threshold voltage shown here, around 0.6 V , was much smaller than the
threshold voltage of pure water electrolysis (1.5 V, as shown in Figure 4.1(a)), indicating
the potential of low-power consumption from methanol electrolysis as we discussed in
Chapter 5.1. However, such threshold voltage was larger than the theoretically minimum
required voltage (as shown in Figure 5.1), meaning that the overpotentials from the
electrodes were large. This was expected since we mentioned that the materials we
selected were not the best catalyst. In terms of the concentration effect, the results show
an optimal range, from 1 mol/L – 4 mol/L, where the electrochemical current went
maximum. Within this range, the current was almost the same. Below this range (e.g., 0.1
mol/L), the current went down because there was no enough methanol that could be
involved into the reaction; above this range (e.g., 10 mol/L), the concentration of
methanol (or, those active poisoning groups, specifically) was so large that the poisoning
effect would be dominant, which could further reduce the electrochemical current. This
result was also consistent with other’s discovery [63]. During the experiments, we did not
observe any obvious bubbles generation by naked eyes, probably because the
electrochemical current was so small due to the low applied voltage, compared to the
case in pure water electrolysis. Therefore, the bubble generation rate was very small.
80
Figure 5.5(b) shows the log I vs. V plots where the background data from pure water was
subtracted from each curve, meaning that those data presented only the methanol
reactions at the anode. An interesting trend was shown in this figure that there was a pit
around 0.5 V , and the pit becomes larger when increasing the concentration. As we
mentioned before, during methanol electrolysis, two groups of chemicals are competing
the available sites of Pt if using Pt/Ru alloy as the anode material. The *CH2OH,
*CHOH, *COH, and *CO during methanol electrolysis will poison the Pt; the *OH,
however, is used to recover the sites of Pt for Pt-Hadsorption. We believe that, the current
going down between 0.2 V and 0.5 V was due to the anode adsorption of such active
poisoning chemicals: for larger-concentration methanol solutions, this adsorption effect
Figure 5.6. Comparison between the electrolysis of methanol solution and ethanol solution, both
based on our nanogap cells.
81
was more obvious. When the voltage above the threshold (around 0.5 V shown here), the
electrolysis started, and thus the local adsorbed chemicals started to be consumed. In this
case, the current became increasing again. Larger threshold voltage was necessary for the
methanol solution with larger concentration in order to overcome the poisoning effect, as
shown in this figure.
We also characterized the IV-curve measurements of ethanol solution based on the same
device. The result is shown in Figure 5.6. To make it clearer, the data from methanol
solution was also included, and the data background from pure water had been removed
for all the curves shown here. First, the results demonstrate that even more
non-conductive ethanol solution can also be electrolyzed based on our nanogap cells,
with threshold voltage around 0.7 V , which further demonstrates our virtual breakdown
mechanism. For ethanol solution, the data from both 2 mol/L solution and 10 ml/L
solution were much smaller than the data from 2 mol/L methanol solution. That is
because, in fact, the poisoning effect can also occur in ethanol electrolysis. However, Ru
might not be the best choice to recover the adsorption sites. Moreover, in ethanol there
are more poisoning chemicals (e.g., *CH2OH, *CHOH) than in methanol, leading to
more obvious poisoning effect. That is also why the data from 2 mol/L ethanol solution
and 10 ml/L ethanol solution had no big difference.
We further changed our electrode system to Pt for both anode and cathode to further
demonstrate the benefit of Ru as part of the anode to prevent the poisoning effect. Both
methanol solution and ethanol solution were characterized, and the results are shown in
82
Figure 5.7. As we can see, when anode was changed to Pt, the current from both
methanol solution and ethanol solution had been reduced: the decrease in methanol
solution was large and the decrease in ethanol solution was quite small. Moreover, for
both methanol solution and ethanol solution, the data from 2 mol/L and 10 mol/L were
almost the same. As we mentioned that, pure Pt could not prevent the poisoning effect,
Figure 5.7. The experiment results of IV-curve measurement when the anode was changed to pure
Pt, (a) methanol solution, (b) ethanol solution.
83
especially for methanol solution. That is why the current in methanol solution decreased
significantly. For ethanol solution, since it had been already poisoned seriously even with
Pt/Ru anode, there was no big difference when we changed the anode to Pt. That is also
why the concentration would not affect the electrochemical performance any more when
we changed the anode to Pt, no matter in methanol solution or ethanol solution, since all
the adsorption sites on Pt had been occupied by those poisoning chemicals. To compare
methanol solution and ethanol solution, the scale of y-axis of these two figures were set
the same. In fact, except the data from 2 mol/L methanol solution with Pt/Ru anode, all
the other data were almost the same, indicating that the reactions from those systems
were all self-limited due to serious poisoning effect.
5.4. Summary
In this chapter, we have successfully achieved the electrolysis of non-conductive pure
organic solvent based on our nanogap electrochemical cells. Pt/Ru alloy was utilized as
the anode to prevent the poisoning effect. Both methanol solution and ethanol solution
had been characterized, and further demonstrated the concept of virtual breakdown
mechanism in which strong electrolyte can be eliminated for the electrolysis of
non-conductive solvent. In our experiments, the electrolysis of methanol solution showed
the great potential of low-power consumption for hydrogen generation.
84
Conclusion and Future Work
Fundamentally different from conventional and well-established electrolysis that relies on
high concentrations of added electrolyte, field-assisted splitting of pure water and pure
alcohol solution (methanol solution and ethanol solution) at room temperature has been
successfully achieved in this project based on our metal-dielectric-metal sandwiched-like
nanogap electrochemical cells. Both pure water electrolysis and pure methanol solution
electrolysis showed great potentials for hydrogen generation. In this project, the gap
distance between anode and cathode down to 37 nm has been demonstrated. In such
deep-sub-Debye-length region where the double layers overlapped, high electric field in
the entire gap couples the two half-reactions together, and significantly enhances
molecules ionization and mass transport, leading to an electron-transfer limited reaction.
This virtual breakdown mechanism can reduce the energy losses arising from ion
transport, and greatly enhance the equivalent conductivity of the electrolyzed solution. In
our project, the electrolysis current density from pure water could be even much larger
than that from 1 mol/L sodium hydroxide solution. We propose to investigate this virtual
breakdown mechanism further. For example, reference electrode can be added to study
cathode current and anode current separately; characterizations of capacitance-voltage
curves will also provide important information for theoretical analysis. Besides,
compared to other NECs, our open cells can be simply fabricated on large area with high
fabrication yield, and have a great potential to enhance the redox cycling rate for
85
ultra-sensitivity/selectivity. At last, such virtual breakdown mechanism can be applied on
almost all weakly-ionized materials, and may provide inspiring directions in many
electrochemical fields, including ultrafast charging, alcohol electrolysis, carbon-dioxide
reduction and fuel cells.
86
Project 2: Lithographically Defined Particles
for Cancer Therapy
Chapter 6. Introduction for Cancer Therapy
Cancer is one of the leading causes of death all over the world [67]. In 2012 about 14.1
million new cases of cancer occurred globally, and about 8.2 million deaths (14.6% of all
deaths) were caused by cancer [68]. The most common methods of cancer treatment
include surgical resection, radio/chemotherapy, and the various combinations thereof.
However, surgery often fails to remove all of the cancerous cells, resulting in the
regrowth of malignancies. Furthermore, many tumors are inoperable because of their
positions too close to critical tissues or the conditions of the patients, and
radio/chemotherapy attacks not just malignancies but also healthy cells, leading to many
acute and chronic side effects. In recent decades, a promising approach called
hyperthermia therapy, which utilizes heating effect for tumor therapy, has been drawing
significant attention. Researchers have found that moderate heating, in the range of 42 to
47 °C, can destroy the tumor while leaving the normal tissue unaffected [69]. This is due
to the reduced heat tolerance of tumors compared to normal tissues: hyperthermia can
cause apoptosis of the cells through lysis and rupture of membranes and release of
digestive enzymes, leading to protein denaturation and irreversible cell damages [70, 71].
Compared to surgery or radio/chemotherapy, hyperthermia therapy is noninvasive and
nontoxic, and has the capability of treating deep embedded tumors inside the human body
87
[72]. Patients can also tolerate this therapy without significant discomfort or serious side
effects [73]. Conventional sources of thermal treatments include radiofrequency [74-76]
and microwaves [77-79], ultrasound waves [80, 81], as well as laser therapy [82, 83].
However, in clinical practice, these methods currently lack the capability to
noninvasively discriminate between malignancies and surrounding healthy tissues,
resulting in elevating the temperature of intervening tissue between the source and the
target. It is difficult to selectively heat tumor tissues by those techniques, which require
directing externally applied energy toward only tumor regions [84].
Alternatively, the energy absorption of the specific target tumors can be enhanced by the
introduction of contrast agents, and the emerging field of micro/nano-technology offers
the potential for developing novel agents. Micro/nano-particles with specifically
engineered energy absorption properties can accumulate around or inside malignant
regions, to facilitate localized heating only affecting tumors. Approaches to
micro/nano-particle-mediated hyperthermia include plasmonic photothermal therapy
(PPTT), magnetically-induced heating, and microwave irradiation. For PPTT methods,
noble metal nanoparticles, e.g., gold nanospheres [71, 85], gold nanorods [86, 87], and
gold nanoshells [88, 89], can possess strongly enhanced visible and near-infrared light
absorption due to the phenomenon of surface plasmon resonance [90]. Unfortunately, the
penetration depth of near infrared light in human body is only a few millimeters [91] due
to significant scattering and attenuation of light by biological tissues [92]. Moreover,
toxicity of a surfactant used in gold nanoparticle synthesis can degrade membranes and
88
peptides [93]. Magnetically-induced heating [94-97] takes advantages of magnetic
nanoparticles excited by the external electromagnetic (EM) field with much longer
wavelength, which can penetrate deeper through human body [93]. Under the EM field,
magnetic nanoparticles can perform magnetic hysteresis [98], Néel or Brown relaxation
[99], and magnetic resonance [100], coupling the external EM field power into
nanoparticles to generate heat. However, the absorption efficiency of these magnetic
nanoparticles highly depends on their morphology and size [73, 92, 101], which are
difficult to control due to their complex chemical synthesis processes: the intrinsic
complex synthesis limits their optimal structures. Moreover, besides the problem of
particle aggregation [102, 103], biocompatibility [104] remains another significant hurdle
for magnetic nanoparticle hyperthermia. The third type of nanoparticle-based
hyperthermia therapy, called microwave irradiation, is based on microwave absorbers
with large conductivity, e.g., gold nanoparticles [105, 106] and single-walled carbon
nanotubes [107, 108]. The mechanism behind microwave irradiation is mainly due to
space charge polarization [109, 110] with a total effect as damped vibrations of electric
dipoles. The property of equivalent conductivity and dielectric constant can also be
adjusted by adding those nanoparticles [108]. However, those microwave-absorber
nanoparticles also highly rely on their chemical synthesis processes [110], which limits
their applications for hyperthermia. Other microwave-absorber nanoparticles [111-113]
have also been investigated by researchers, but they have not been shown to be suitable
for medical applications.
89
We propose to solve this issue by using lithographically defined micro/nano-particles
(LDPs) to achieve selectively localized heating under microwave radiation. By using
top-down nanofabrication processes, particles with complex structures, which could have
never been achieved using chemical synthesis, can be fabricated. Comparing to the
methods mentioned above, the key advantage of our new approach is that optimal
geometry can be easily obtained and controlled without synthesis limitations. In our
project, microwave was selected as the external EM filed energy source since the
penetration depth of microwaves in the range of 0.9-3 GHz into human body can be
several centimeters, which is suitable for medical applications. On the other hand, in
order to make the particles injectable into the entire human body, the size of our particles
should be limited to smaller than Red Blood Cells, which are usually 6-8 μm [114] in
diameter and about three orders of magnitude smaller than the wavelength of the
microwave. The challenge of this project is just to design those extremely
sub-wavelength particles with high efficiency of microwave absorption. Two schemes
have been considered in this project: resonance particles and non-resonance particles. For
resonance particles, LC-circuit-like particles as micro-resonators at microwave
frequencies are presented. When resonating at a microwave frequency with a high
Q-value, the energy losses resulting from currents flowing in the circuit can be enhanced
and can produce significant heating. However, we discovered that the limitation of the
micro-resonator is not the low resonance frequency trade-off with the small size but the
low Q-value due to insufficient conductivity of materials. For non-resonance particles,
90
disk-shaped gold microparticles as magnetic dipoles have been chosen to be our LDPs.
The fabrication and collection/transfer processes of the particles have been developed.
Significant heating enhancement of these LDPs has been demonstrated both in our
polymer-film nanocomposite setup and particle-suspended hydrogel setup. It is shown
that the LDPs with microwave radiation are a promising candidate to selectively deliver
heat for cancer hyperthermia therapy, which can potentially offer non-invasive,
cost-effective and low side-effect cancer treatment.
91
Chapter 7. Micro-Resonators for Microwave Cancer Therapy
In this chapter, we present a numerical analysis on metallic micro/nano-particles as
LC-circuit resonators at microwave frequencies for cancer hyperthermia therapy. We
pushed the LC-structure design to the limit to provide an optimized structure within
desired size limits, which can achieve resonance at the frequencies commonly used in
interstitial microwave thermal therapies. Moreover, we discovered that the limitation of
the micro-resonator is not the low resonance frequency trade-off with the small size, but
the low Q-value due to insufficient conductivity of materials. This issue can be solved by
using multiple-layer-inductor design of LC-structures, which could increase the Q-value
linearly with the number of inductor layers, but at the cost of fabrication process
complexity.
7.1. Resonance Microparticle Design
Here we want to take advantage of resonance effect to enhance the total heating loss
generated from our lithographically defined particles. In the range of microwave
frequency, oscillating energy of carriers is low and the plasmonic resonance effect on
materials is minimal. We cannot take advantages of resonant waveguide / chamber to be
designed in order to enhance the local resonance at microwave frequency, either, since
the size should be comparable to wavelength, much larger than our size limitation of 6-8
microns. Based on our application and limitation, the most straightforward method to
92
achieve resonance effect in microwave range is the LC-circuit resonance.
Top-down nanofabrication technology can be used to obtain such micron-scale particles
as LC-circuits. In a resonant circuit with a coil inductor and a plate capacitor, the coil
inductor couples the magnetic energy most efficiently if the incident wave frequency is
the same as the resonance frequency of the circuit, with the plane of the coil
perpendicular to the direction of the magnetic flux. Then large currents will be generated
in the coil due to the high Q-value. Because the skin depth of the metal in the microwave
region is much larger than the thickness of the metal layers in our designs, we could
consider the metal as bulk resistors very well. According to Joule's law, significant heat
from the resistance will be generated. Those micro-resonator particles can be injected into
the human body, allowing heat to be transferred to local tissues. This could provide for a
localized cancer hyperthermia therapy.
The first challenge was how to achieve resonance frequency with a structure whose
dimensions are 3 orders of magnitude smaller than the wavelength. Figure 7.1 shows our
design of LC-circuit structures within the size limit. For clarity, Figure 7.1(a) only shows
one turn of each coil and the dimensions in z-direction are not to scale. In Figure 7.1(a),
the bottom three plates constitute two capacitors in parallel (comb-like structures). The
first and second plates are connected to the two ends of the inductor, respectively, which
actually consists of three coils in series in three different layers. Figure 7.1(b) shows the
schematic of the full LC-structure. The side length of the plates is fixed to 8 μm. The
whole structure is made of metal, while the material between the capacitor plates consists
93
of TiO2 thin films, with a relative permittivity around 81 at 2.45 GHz range (a
commonly-used operating frequency for cancer therapy). This design is achievable by
conventional nanofabrication processes with multiple lithography steps. Moreover, in this
way both inductor and capacitor structures could be extended to more layers, in order to
achieve larger inductance and capacitance, meaning that the resonant frequency can be
even lower than 1 GHz if desired. Figure 7.1(c) shows a to-scale LC-structure design
with five-layer inductor (dense coils) and only one capacitor. In order to avoid magnetic
field leakage, the separation between the coils must be small. In practice, sometimes the
magnetic field will not be aligned with the normal of the coil planes of all particles.
Assuming a random distribution of particles, this can be compensated for by an averaging,
reducing the total heating loss by a factor of 1/3.
Figure 7.1. Design of multi-layer LC-structures, (a) schematic diagram, (b) equivalent circuit
diagram, and (c) to-scaled structure model with five-layer inductor.
94
7.2. Resonance Frequency and Q-value
There are two requirements of the micro-LC-circuit: the resonance frequency should be
low enough and the Q-value should be as large as possible. In this project, the
commercial software ANSYS HFSS was used to simulate the characteristics of the
LC-circuit design shown in Figure 7.2(a). The electric field intensity of the input
microwave was set at 1 V/m (a.k.a., power density: 1.19 μW/cm
2
), and the microwave
source was put far away from the LC-structure in order to achieve the far-field effect. In
the first example, relatively typical parameters were picked in terms of the
Figure 7.2. (a) A typical LC-structure and the heating loss curves vs. frequency, with conductivity
(b) 10
10
S/m, (c) 4.1×10
7
S/m (gold), and (d) 10
12
S/m.
95
nanofabrication difficulty. The width of the capacitor was 8 μm, the line width of the
coils was 200 nm with 200nm space, the thickness of the metal was 50 nm, and the gap
distance between two plates was 20 nm, with only a 3-layer inductor and one capacitor
(the inductor coving the entire capacitor plate area, as shown in Figure 7.2(a)), and
magnetic field perpendicular to the inductor plane. Figure 7.2(b) shows the simulation
result of the heating loss curve generated from this structure (one single
micro/nano-particle) vs. frequency from 1.5 – 2.3 GHz. The structure material was set
with conductivity of 10
10
S/m (a fictitious material). The resonance peak of the heating
loss appeared at 1.94 GHz. Notice that the parameters were not pushed to the limits;
however, the resonance frequency was already even lower than the 2.45 GHz requirement.
This definitely demonstrates that our design could work for low-frequency
micro-resonators even though the size of the LC-circuit was three orders of magnitude
smaller than the wavelength of the applied EM field.
However, when we only replaced the material with gold (conductivity 4.1×10
7
S/m) or
with a fictitiously higher-conductivity material (conductivity 10
12
S/m), the peak was
eliminated or became sharper, as shown in Figure 7.2(c) and Figure 7.2(d), respectively.
This is because Q-value is proportional to the conductivity of the material: low
conductivity can only provide small Q-value. This indicates that for gold model the
Q-value will be too low to have an observable resonance due to very high resistance from
the coils. The Q-value is the key parameters in this project, which can enhance the
heating loss from the LC-resonance effect for localized cancer hyperthermia therapy. If
96
the Q-value is so small, it will reduce the heating effect significantly for realistic
applications.
Considering the standard Q-value equations below,
0
2
L
Qf
R
(7.1)
where f0 is the resonance frequency (2.45 GHz commonly used, or below), we need a
large L-to-R ratio if a high Q-value is necessary. Values of R and L are predominantly
determined by the properties of the inductor coils. Therefore, we studied how to optimize
the structure, and in particular, the coil structure. Here we only consider the Q-to-f0-ratio
(or L-to-R-ratio) and fix f0 because we can always extend or reduce the number of the
capacitor layers to achieve the desired resonance frequency. The key point is to find a
large Q-to-f0-ratio inductor structure. For single-layer inductors, three geometry factors
are important, the inner diameter and the outer diameter of the coil, and the linewidth of
the coil.
7.3. Q-value Optimization
The Q-value can be calculated from the simulation result of heating loss vs. frequency,
shown in the following equation,
0
21
f
Q
ff
(7.2)
Here the f0 is the resonance frequency, and f2 and f1 are the high and low half-loss
97
frequency, respectively. It describes how “broad” the resonance peak shown in the loss
vs. frequency plot.
First, the effect of the inner diameter and the outer diameter of the coil was characterized.
We fixed the linewidth of the coil at 100 nm with 100 nm space, the thickness of the
metal was still 50 nm, and only one capacitor was presented here. The inner diameter and
the outer diameter of the coil were changed from 2 – 8 μm. The simulations were based
on a fictitious material with conductivity 10
10
S/m. The result is shown in Figure 7.3.
Q-to-f0 ratio will decrease if we increase the inner diameter, and will increase if we
increase the outer diameter. This indicates that the inductance will increase faster along
with the total coil length than the resistance, and we discovered that the outer part
Figure 7.3. Effect of the inner diameter and outer diameter of the coil. Square line: keep inner
diameter zero; circle line: keep outer diameter 8 μm.
98
contributed more to the improve of Q-to-f0 ratio than the inner part, which was because
more magnetic flux passed through the outer part than the inner part. In another word, the
maximum value of Q-to-f0 ratio will occur when the coil covers the area as much as
possible; in our case, it was just the entire area of the capacitor plate.
Figure 7.4. (a) The total heating loss from different linewidth structures, including basic heating
loss. (b) Modified Q-to-f0 ratio vs. linewidth.
99
The effect of the linewidth of the coil is shown in Figure 7.4. The conductivity here was
still set as 10
10
S/m, the gap distance between the two plates was fixed at 10 nm, and the
coils covered the entire plates in order to achieve as large Q-to-f0 ratio as possible. The
linewidth of the coil changed from 40 nm to 800 nm, with the space the same with the
linewidth. Figure 7.4(a) shows the total heating loss generated from the structure vs. the
linewidth. In fact, the total heating loss generated from our LC structure comes from two
contributing sources: the first part is basic heating loss, which mainly comes from the
eddy current inside the capacitor plates, and is only linearly proportional to frequency;
the second part is the Ohmic heating from the current in the RLC circuit. When resonance
is reached with a large Q-value, this part is greatly enhanced due to the enhancement of
current. In most of our cases, the first part can be ignored compared to the second part.
However, when Q-value is small, in fact the first part can be comparable to the second
part and cannot be ignored. We discovered that, for small-linewidth structures, the
Q-value is relatively small: the basic loss is a significant portion of the total loss so it
should be subtracted from the total heating loss when studying the loss from the
resonance (in order to calculate the “truly” Q-value). The basic heating loss can be
calculated by cutting the current path (simply removing the connecter between the
inductor and the capacitor, therefore there is no resonance at all), but keeping the rest of
the structure unchanged. This value is also shown here.
After subtracting the basic loss from the total loss, the “truly” Q-value can be calculated
from the modified plot of only resonance heating vs. frequency. In this way, the modified
100
Q-to-f0 ratio based on the “truly” Q-value can be obtained, as shown in Figure 7.4(b). It
shows the clear trends that, the Q-to-f0 ratio increases if the linewidth decreases from 1
μm to 100 nm. In fact, when the linewidth is larger than 100 nm, it is hard to form a
typical “coil” with many turns because the linewidth is too wide. In another word, the
coil cannot perform as a normal inductor. When the linewidth decreases from 1 μm to
100 nm, the inductance from the coil can be enhanced; that is why the Q-to-f0 ratio can be
increased. When the linewidth is smaller than 100 nm (when the coil is dense enough),
the Q-to-f0 ratio keep around 1.75-1.80 GHz
-1
. This is consistent with the theoretical
predictions [115] that the Q-to-f0 ratio should be a constant if the parasitic capacitance is
neglected. In fact, both inductance and resistance are inversely proportional to the square
of linewidth (the fluctuation from the results is due to numerical errors). When the coil is
dense enough, for the resistance,
2
/
*
1
total
cross
linewidth
linewidth thickness
linewidth
L
R
s
Sl
lh
l
(7.3)
Here Ltotal is the total length of the coil, σ is the conductivity of the material, scross is the
cross-section area of the coil wire, S is the total area of the plate (the coil covering the
entire plate area), and hthickness is the thickness of the coil wire. For the inductance,
101
2
2
1
linewidth
Ln
l
(7.4)
Here n is the number of the turn from the coil. Therefore,
0
2.
QL
const
fR
(7.5)
For even smaller linewidth (e.g., 40 nm), the Q-to-f0 ratio decreases. This is probably
because the parasitic capacitances from the sidewall of the coil starts to affect the total
performance; another reason may be the numerical errors from the software, because 40
nm is quite small compared to the entire simulated structure and the local meshing may
not be insufficient. In summary, the optimal linewidth should be around 50 nm to 100 nm,
with Q-to-f0 ratio around 1.75-1.80 GHz
-1
. This linewidth requirement can be easily
achieved by the state-of-art nanofabrication technology. However, this result is based on
the fictitious material with conductivity of 10
10
S/m.
To further increase the Q-to-f0 ratio, multi-layer-inductor designs have to be used. Two
ways were attempted here: one is to increase the number of the inductor layers directly, to
increase both inductance and resistance together; the second way is to only increase the
thickness of the coil, to decrease resistance but keep the inductance. For the first method,
we have to make sure that the inductor layers should be very close to each other, in order
to prevent any magnetic field leakage. The result is shown as the circle line in Figure 7.5:
the Q-to-f0 ratio is linearly proportional to the number of the layers. Further study proves
that the inductance, including both self- and mutual-inductance, is proportional to the
102
square of the number of the layers, whereas the resistance is simply linearly increasing
with that. For the second method, the result is shown as the square line in Figure 7.5.
With the same total thickness of the metal layers as the first method, the Q-to-f0 ratios are
much lower than that in the first method, mainly due to the increasing parasitic
capacitance from the increasing sidewalls inside the coil. Therefore, the first method is
better and higher Q-to-f0 ratio can be achieved by simply increasing the total number of
the inductor layers.
Combine all the analysis above, if we want the f0 to be around 2 GHz and Q-value to be
Figure 7.5. Effects of the number of the inductor layers (circle line) and the thickness of the
inductor coil (square line). The bottom x-axis is also corresponding to the total thickness of the coil
layers (circle line).
103
around 10, then at least three layers of the inductor are necessary if the optimal design of
the LC-circuit is exploited. However, this is based on the fictitious material with
conductivity of 10
10
S/m. Lower conductivity has to require more layers of the inductor
as a compensation of the resistance increasing, in order to achieve the same Q-to-f0 ratio
(or the L-to-R ratio). Unfortunately, this is linear relationship between the conductivity
and the number of the inductor layers. If gold is used, with conductivity only 4.1×10
7
S/m, the number of layers should be at least two hundred to achieve the same Q-values at
the same resonance frequency. This is quite challenging in terms of the nanofabrication
process, not mentioned that the total thickness of the particle has been above the size
limitation of 8 μm. If we compromise with such complexity of fabrication, Q-value
would be decreased to the level can be neglected. In this case, the “resonance” effect
would become the non-resonance scenario, which we will discuss in the next chapter.
7.4. Summary
By using the finite element method with commercial software (Ansoft HFSS), this article
gives a numerical analysis on fictitious micro/nano-particles as micro-resonators for
enhancing microwave thermal absorption for applications such as cancer thermal therapy.
Within the size limits needed for such applications, we present the optimal design of an
LC-circuit approach to maximize absorption of energy from a microwave source. With
the structure we provide, the desired low resonance frequencies well below 2 GHz can be
104
achieved. We also discovered that due to insufficient conductivity of available materials,
a multi-layer inductor design is desired to achieve reasonable Q-values and therefore
meaningful heat absorption enhancement.
105
Chapter 8. Non-Resonance Microparticles: Design and
Fabrication
In this chapter, we will focus on the non-resonance particles: the design optimization and
fabrication process will be discussed. Different from the micro-resonators, the
performance of non-resonance particles is not dependent on particular frequencies, but
responds to wide-band range of microwave frequency. Two schemes have been
considered: magnetic dipoles and electric dipoles. The particle structures, including both
geometry and material, have been designed, tested and optimized by the finite element
method using Ansoft HFSS to obtain high microwave absorption efficiency. After
numerical analysis and comparison, simply disk-shaped magnetic dipoles have been
selected as our lithographically defined particles (LDPs) as the microwave absorbers to
achieve localized heating for hyperthermia therapy. The fabrication and releasing
processes of the particles have been discussed as well.
8.1. Magnetic Dipoles and Electric Dipoles
As we discussed in the previous chapter, in order to make our lithographically defined
particles (LDPs) injectable into the entire human body, the size of the LDPs should be
limited to smaller than Red Blood Cells, which are usually 6-8 μm [114] in diameter. We
set 8 μm to be the upper bound of the LDP sizes. However, the microwave wavelength is
at the centimeter scale, which is 3 orders of magnitude larger than the size limitation of
106
our LDPs. In such an extreme sub-wavelength size range, two non-resonance schemes
can be used to couple the EM energy into the particle, namely through the use of
magnetic dipoles and electric dipoles. Here we use the commercial software Ansys HFSS
to do the numerical analysis to compare the volume-averaged heating loss density
generated from these two types of dipoles. Both magnetic dipoles and electric dipoles
require high-conductivity materials. We chose gold for this purpose because not only it is
highly conductive, but also it is a human-tissue inert material [71, 72, 116, 117].
Magnetic dipoles can be realized by a gold disk, and electric dipoles can be realized by a
gold rod. Figure 8.1 shows the effect of the feature size of such two dipoles on the
volume-averaged heating loss density. The structures were excited by a 1.9 GHz far-field
Figure 8.1. Simulation results of comparison between magnetic dipoles and electric dipoles.
V olume-averaged loss density vs. the feature size D and L (when parameter a was 100 nm). The
direction of the EM field is also shown on the figure.
107
harmonic plane wave with electric field intensity of 1 V/m (a.k.a., power density: 1.19
μW/cm
2
), and deionized (DI) water as surroundings. The magnetic field direction is
perpendicular to the disk, while the electric field direction is parallel to the rod. This
results in the maximum absorption efficiency for magnetic dipoles and electric dipoles,
respectively. The value a, or the thickness of the disk and the width of the cross section of
the rod, was fixed at 100 nm in Figure 8.1. When the feature sizes (the diameter of the
disk D or the length of the rod L) increase, the volume-averaged loss density will increase.
For magnetic dipoles, it is due to more magnetic flux going through the effective
projection area, leading to larger eddy currents inside the disk; for electric dipoles,
heating loss is generated by electron movement under external electric field, hence longer
rods will result in faster movement of electrons and larger dissipation. Therefore, in order
to achieve higher loss density, larger feature sizes for both magnetic dipoles and electric
dipoles are required. Notice that, here when we increased the feature sizes for both the
two dipoles, not only the total heating loss but also the loss density increased, which is
more important to us because this is the more efficient way to generate heating at the
same material cost. However, the feature sizes of both magnetic dipoles and electric
dipoles have to be limited by the 8 μm upper bound.
In order to see the effect of parameter a, we fixed the feature size D and L to be 8 μm for
both magnetic dipoles and electric dipoles. The result is shown in Figure 8.2. For
magnetic dipoles, the loss density will not be altered significantly with a change in the
thickness of the disk. However, when the thickness became comparable to the diameter of
108
the disk, the loss density decreased. This is due to the magnetic field being screened by
the sidewall of the disk (or cylinder, precisely), leading to more reflection and less
penetration. For electric dipoles, when a is very small compared to L, loss density
decreases greatly when a increases. It is because larger a values will screen the EM field
significantly thus leading to lower loss density for electric dipoles, while for magnetic
dipoles the screening effect is much weaker. However, when a becomes even larger, the
loss density will increase. That is because larger a values (when it can be comparable to L)
provide larger projection areas from the perspective of magnetic field, therefore actually
magnetic dipoles start to contribute more. In fact, there is a phase boundary of a value,
smaller than which the electric dipole dominates, and larger than which the shape
Figure 8.2. Simulation results of comparison between magnetic dipoles and electric dipoles.
Volume-averaged loss density vs. the parameter a (when D and L were 8 μm).
109
becomes more isotropic and the magnetic dipole dominates.
In order to obtain higher absorption efficiency, magnetic dipoles require larger diameter;
while electric dipoles require longer rod length and smaller cross-section. However, in
order to offer the same loss density as the 8 μm diameter magnetic dipole, the aspect ratio
of the electric dipole should be at least 130:1 when rod length is 8 μm. Moreover, we
discovered (as shown in Figure 8.3) that even if we change the size of the rod, the 130:1
aspect ratio is still necessary to achieve the similar loss density. This stringent
requirement comes from the fact that the intrinsic impedance of gold, which can be
expressed as,
E
j
Hj
(8.1)
Figure 8.3. Simulation result of the size effect of the electric dipoles. If the aspect ratio did not
change, then the volume-averaged loss density would not change.
110
The intrinsic impedance quantitatively couples the magnitude of electric field intensity
and magnetic field intensity together. For gold, it is a very small value due to the very
large equivalent permittivity of gold (mainly because the conductivity, or the imaginary
part of the permittivity is quite large). Compared to the magnetic field intensity, the
electric field intensity is too small inside gold. This can also explain why the screening
effect on magnetic dipoles is much weaker than that on electric dipoles. Therefore, very
high aspect ratio rod is necessary for electric dipoles in order to generate comparable loss
density to magnetic dipoles. Carbon nanotubes (CNTs) can achieve such high aspect ratio;
however, in reality, the geometry of CNTs with such high aspect ratio would probably
bend and could not keep rod-like shapes, leading to electric field redistribution and much
lower loss density. In summary, magnetic dipoles more easily achieve high efficiency of
microwave absorption than electric dipoles. Therefore, disk-shaped magnetic dipoles
were chosen as our LDPs, which cannot be synthesized by chemical reactions due to
entropy limitations.
8.2. Magnetic Dipoles Optimization
For magnetic dipoles, a significant approach to increase the heating loss density is to
enhance the central magnetic flux density (thus eddy currents). We further optimized the
magnetic dipoles by adding a nickel core inside the gold disk, as shown in Figure 8.4.
The outer diameter was fixed at 8 μm here to satisfy the size limitation, and the thickness
111
was fixed at 1 μm. The diameter of the nickel core was just set to be the inner diameter of
the gold ring. The input microwave was still set at 1.9 GHz with electric field intensity of
1 V/m (a.k.a., power density: 1.19 μW/cm
2
). As a comparison, we also simulated hollow
gold ring structures. In this simulation, we compared the total heating loss generated from
those structures, since the entire “apparent” volume are actually the same. Comparing the
red circle curve (only the ring part of the gold-nickel structure) and the black square
curve (hollow gold ring structure), the total heating loss from the ring can be enhanced
when a nickel core is put into the central area of the ring, due to the enhanced central
permeability and therefore magnetic flux density caused by nickel. The blue triangle
Figure 8.4. Simulation results of comparison between the simple gold ring structures and the
nickel/gold-core/ring structures. The outer diameter was fixed at 8 μm and the height was fixed at 1
μm, and the diameter of nickel core was equal to the inner diameter of the gold ring.
112
curve shows the heating loss from the entire gold-nickel structure. When compared to the
red circle curve, it indicates that most of the heating loss of the gold-nickel structure is
from the gold ring part. This means that the nickel core only increases the central
magnetic flux density but does not contribute to the Ohmic loss directly. The heating loss
curve shows an optimal inner diameter value to offer maximum heating loss: when the
inner diameter is small, the larger nickel core will provide larger magnetic flux density;
but when it is too large, the lower conductivity of nickel reduces the heating current, and
thus reduces the entire heating loss. There is a tradeoff between the high permeability and
the low conductivity of the nickel core. From the black square curve, when the inner
diameter is increasing, the total heating loss decreases slowly at first and then decreases
rapidly. The turning point of the inner diameter was around 4.5 μm, which was consistent
with the skin depth of 1.7 μm in gold at 1.9 GHz since the total outer diameter is 8 μm
(4.5 + 2*1.7 ≈ 8): when the inner diameter was smaller than 4.5 μm, the inner diameter
had no obvious effect on the heating loss generated from the gold ring structure because
most of the heating came from the skin-depth region; however, when the inner diameter
was very large, the effective region became smaller and the total heating loss started to
decrease. This also indicates that the inner part will not contribute much, while the outer
part contributes significantly to the total heating loss.
Although the gold-nickel structure can provide about 20% higher absorption efficiency
(290 compared to 240, shown in Figure 8.4), the simple gold disk structure is still
preferred because of the much simpler fabrication process and almost the same
113
absorption efficiency. In our first stage, the simple gold disks can provide a convenient
way to demonstrate the feasibility and efficiency of our concept of the lithographically
defined particles.
We also studied the sizing effects by comparing the heating loss of one big microparticle
and the amount of heating loss from thousands of nanoparticles (such that researchers are
focusing on currently) with the same total volume (or material cost). The results are
shown in Figure 8.5. Here we studied square particles because it is more convenient to
keep the total volume the same. One-layer particle arrays (1×1, 3×3, 10×10, and 30×30)
with the same total side length 4.5 μm were simulated. Only one-layer arrays were
simulated because the effect from multi-layer arrays were just the linear increasing of the
one-layer arrays. Here we set all the particles with height of 100 nm. For the array 30×30,
each small particle was 150 nm×150 nm×100 nm, similar to the size of the chemically
synthesized gold nanoparticles. As shown in Figure 8.5(a), the total heating loss from the
array 1×1 (just one sheet-like microparticle) is more than two orders of magnitude larger
than the total heating loss from the array 30×30 (900 small nanoparticles), while the total
volumes are the same. Figure 8.5(b) shows the localized volume-averaged heating loss
density distribution of the four arrays (the scale bar is in log-scale). This obviously shows
that the heating loss density decreases rapidly when the sizes of the particles become
smaller. For these nanoparticles, the contribution from electric dipoles can be ignored due
to the low aspect ratio; however, the eddy current generated is also multiple orders of
magnitude smaller than that from one microparticle. For magnetic dipoles, higher
114
absorption efficiency requires larger and more isotropic (circular) projection areas that
perpendicular to the magnetic field direction. The big microparticle possessed the largest
circular current path, while for nanoparticles the current paths had to be confined in such
Figure 8.5. Simulation results of (a) comparison of the total loss from particle arrays (1×1, 3×3,
10×10, and 30×30) with the same total volume (the height of the particles was 100 nm), and (b) the
localized distribution of the volume loss density from the four arrays (the scale bar is in log-scale).
115
small regions, and hence one big microparticle can generate more heating loss than
thousands of small nanoparticles under the same conditions even though with the same
material cost. However, we still have the size limitation of 8 μm. Therefore, 8 μm
diameter disk-shaped magnetic dipoles are our choice as LDPs for further fabrications
and measurements.
8.3. Fabrication and Releasing Process
It is relatively easy to pattern the gold micro-disks on silicon wafers, but it is a little bit
challenging to release them from the silicon wafer. Here we utilized sacrificial-layer
technology to achieve the releasing process. Lift-off Layer (SHIPLEY MICROPOSIT®
LOL 2000) is a commonly-used material as the sacrificial-layer. However, it is very
easily to get wet-etched by such chemical solvent used during the photolithography
process, and thus should be protected until the releasing step. Figure 8.6 describes the
fabrication and releasing processes of the 8 μm diameter disk-shaped LDPs. Firstly, one
Lift-off Layer around 150 nm thick was spin-coated on a silicon wafer as a sacrificial
layer (3000 rpm spin coating for 60 sec and baking at 170 °C for 10 min). Then 1 μm
thick parylene-D was chemical vapor deposited (SCS Labcoter PDS 2010) to cover the
whole wafer conformally. The parylene film is an chemically-inert layer covering the
LOL 2000 layer in order to protect against all of the following chemicals until the
releasing step. After that, photoresist (AZ MiR
TM
701, 3000 rpm spin coating for 40 sec
116
and baking at 90 °C for 1 min) was spin-coated above the parylene film and was
patterned by conventional photolithography process to define the LDPs (54.3 mJ/cm
2
UV-light exposure with custom-designed photomask, post-bake performed at 110 °C for
1 min and then the substrate immersed into the developer solution, AZ 300 MIF
Developer, for 1 min). O2 plasma treatment (Tegal Plasmaline 411 Asher, 20 W for 30 sec)
was then utilized to enhance the adhesion between the parylene film and the metal layer,
by creating a number of activated chemical sites on the parylene film. Afterwards, a 300
Figure 8.6. The fabrication and releasing processes of the disk-shaped LDPs. The results shown
here were observed by naked eye, by low magnification under SEM (scale bar: 100 μm), and by
high magnification under SEM (scale bar: 5 μm), respectively.
117
nm thick layer of gold with 3 nm thick layer of Ti as the adhesion layer were evaporated
directionally onto the substrate (Temescal BJD-1800 E-Beam Evaporator, evaporation
rate 1 Å/s), following a lift-off step by removing the photoresist with the metal above in
acetone with ultrasound vibration for 15 min to achieve the patterned gold micro-disks on
the parylene film. Then both the parylene film and the underneath LOL layer were
dry-etched by O2 plasma with the gold disks as the etching mask (ICP etching, Oxford
PlasmaPro System 100. The etching rate was around 700 nm/min with relatively
anisotropic profile), to expose the bottom LOL layer. Finally, the wafer was immersed
into the LOL stripper solution (AZ 400K Developer, 1 min) to dissolve the LOL layer and
to release all the gold micro-disks. The whole processes can be easily applied on mass
manufacturing by roll-to-roll nanoimprint technology.
The gold thickness here was only 300 nm, limited by the lift-off step with the photoresist
we used which was around 1 μm thick. 500 nm thick gold could be also list-off but not
that easily sometimes. However, from the simulation results we know that, the thicker the
magnetic dipoles, the larger the total heating loss generated. In another word, thicker gold
disks perform better for microwave heating. Generally, they can be easily obtained by
using thicker photoresists. However, the numerical studies also show that overly thick
magnetic dipoles will decrease the heating loss density because when the sizes are closer
to the wavelength of the input microwave, the screening effect will become more
dominant. Here, we focused on demonstrating the feasibility of LDPs, and hence 300 nm
gold was used. Also notice that there was a 1 μm thick parylene support under the gold
118
disk, which had little effect on the microwave heating, since it is non-conductive (lossless)
and thin enough compared to the microwave wavelength. On the other hand, this
parylene support can improve the mechanical strength of the entire particle. The LDP
fabrication and releasing results are shown in Figure 8.6, observed by naked eye, by low
magnification under scanning electron microscope (SEM), and by high magnification
under SEM, respectively. The lateral etching of parylene by the O2 plasma etching was
obvious from the view of high magnification, but it had no harm in terms of microwave
absorption of the entire particle.
8.4. Summary
This chapter systematically compare microwave-absorption performance of the two
schemes in this extremely sub-wavelength range based on numerical analysis. In order to
achieve higher heating loss density, larger feature sizes for both magnetic dipoles and
electric dipoles are required. However, very high aspect ratio rod is necessary for electric
dipoles in order to generate comparable loss density to magnetic dipoles. After comparing
with the nickel-core structure, simple gold disk structure is still preferred because of the
much simpler fabrication process and almost the same absorption efficiency. We also
discussed the fabrication and releasing process of such gold micro-disks by using LOL
2000 as the sacrificial layer and parylene film as the protection layer.
119
Chapter 9. Particle Collection and Heating Enhancement
Characterization
Compared to the chemically-synthesized particles, lithographically-defined particles
(LDPs) are more easily to obtain the desired structure; however, it is challenging to
collect them and transfer them to the testing tubes for the heating measurement. In this
chapter, we will discuss our methods for particle collection and heating enhancement
characterization. Notice that, those methods can be applied not only on our disk-shaped
gold particles, but also on other types of LDPs (e.g., micro-resonators) if applicable. Here
we just used our disk-shaped gold particles as an example to discuss such methods.
The microwave heating enhancement measurements were accomplished in two different
setups: polymer-film nanocomposite setup and particle-suspended hydrogel setup, both of
which were in fact LDP/matrix nanocomposites. The two setups used in the
measurements had two different kinds of matrix. For polymer-film nanocomposite setup,
the entire-wafer parylene film during the LDP fabrication processes can intrinsically
provide a polymer matrix for the LDP-based micro/nano-composite. It was very easy to
collect 100% LDPs and to transfer them from wafers into DI water for heating tests.
Despite its convenience to characterize the heating enhancement of the LDPs, this
entire-polymer-film method had no completely released particles, and thus cannot imitate
the real clinical situations. We therefore developed the second setup by using
centrifugation techniques to transfer and concentrate the LDPs from wafers to DI water,
120
and using agarose hydrogel as the matrix to support the LDPs, in order to mimic the
clinical application. Both of these two setups have demonstrated the microwave
absorption enhancement by our LDPs, and have indicated the potential of our LDPs for
selectively localized heating hyperthermia in medical applications.
9.1. Polymer-Film Nanocomposite Setup
The parylene film we deposited during the LDP fabrication processes can be peeled off
entirely from the wafer, along with tens of millions of gold LDPs on it. This is naturally a
good matrix to support those LDPs.
Figure 9.1. (a-d) A parylene film with gold LDPs peeled off from a quarter of 3-inch wafer
gradually in the LOL stripper solution. (e) An entire 3-inch parylene film peeled off. (f) A blank
parylene film and a parylene film with LDPs after drying.
121
Instead of O2 plasma ICP etching after the lift-off step shown in Figure 8.6, the wafer was
immersed into LOL stripper solution directly without O2 plasma etching. The underneath
LOL layer can be dissolved by the stripper, and in this way the entire parylene film can
be separated from the wafer. Figure 9.1 shows a parylene film with gold LDPs peeled off
from a quarter of 3-inch wafer gradually (Figure 9.1(a) - (d)). An entire 3-inch film can
also be peeled off very easily as shown in Figure 9.1(e), with parylene thickness only 1
μm. Figure 9.1(f) shows two dry parylene films: one is a blank and transparent parylene
film; the other is yellowish in color due to the gold LDPs. By using this method, we
could collect almost 100% LDPs without any loss, and transfer them very easily into DI
water for heating tests.
The heating enhancement effects of the LDP/parylene-film composite were characterized
under microwave radiation. The experimental set-up is shown in Figure 9.2(a). A
microwave probe, connected to a microwave source and amplifier, was immersed
centrally into the testing sample. The microwave aperture was around 1cm long, located
near the end of the probe as shown in Figure 9.2(b). A thermal sensor, connected to an
external digital thermometer, was fixed very close to the microwave aperture to measure
the localized temperature, which was recorded by Labview software. The testing sample
was fixed inside a Faraday cage to prevent microwave leakage. Both the microwave
probe and the testing sample were fixed to achieve the reliable heating effect
characterization. During experiments, the microwave aperture was completely covered by
the parylene film, with 3 mL DI water as the surrounding media (as shown in Figure
122
9.2(c)). DI water was required since the imaginary part of the permittivity (i.e.,
Figure 9.2. (a) The experimental measurement set-up. (b) The microwave aperture near the end of
the microwave probe (scale bar: 1 cm), with a thermal sensor close to it. (c) During experiments,
the microwave aperture was completely covered by the parylene film inside DI water.
123
conductivity) of DI water can be neglected, in order to minimize the surrounding’s effect
on the microwave heating generation. The frequency of the microwave was 1.9 GHz.
Figure 9.3(a) shows the heating effects compared between a blank 3-inch diameter
parylene film and a 3-inch diameter parylene film with 3×10
7
gold LDPs. At both 10 W
and 20 W microwave input, the temperature increase of LDP-based films were at least
2.1 times of that of blank films (for 10 W, in one minute the temperature increased from
20.6 °C to 40.7 °C for LDP-based films versus 21.5 °C to 30.6 °C for blank films; for 20
W, 23.0 °C to 46.7 °C versus 22.3 °C to 33.8 °C, respectively). The temperature
enhancement of LDP-based films at 10 W was even 1.75 times of that of blank films at
20 W (20.6 °C to 40.7 °C versus 22.3 °C to 33.8 °C). This definitely demonstrates that
the gold LDPs can significantly improve the heating effect, while the filling volume ratio
was only around 1/3500. The heating effects of the films were also compared to the pure
DI water case (Figure 9.3(b)). The microwave input power was fixed at 20 W. The case
Figure 9.3. The microwave heating characterization results of temperature vs. time based on the
parylene-film nanocomposite setup. The microwave input power in (b) was 20 W.
124
with blank parylene film showed lower temperature enhancement than the pure water
case, indicating that parylene film itself was not a good microwave absorber. However,
the gold LDPs can considerably improve the microwave absorption efficiency of the
parylene film matrix. Experiments also showed that heating occurred in two steps: an
initial rapid heating, followed by a gradual heating. Such a significant degree of heating
within a brief period is due to intense localized heating before thermal relaxation into the
surroundings [91], and indicates the potential for localized heating hyperthermia
treatments.
In summary, this parylene-film nanocomposite setup could easily collect the LDPs and
transfer them into testing samples, and had demonstrated the heating enhancement from
the LDPs’ contribution. However, this nanocomposite was not suitable for injection in
medical therapy. Therefore, experiments based on completely released LDPs were
desired.
9.2. Particle-Suspended Hydrogel Setup
In real thermal therapy, completely released particles are necessary. However, in
experiments, the characterization of the released particles is much more challenging than
the characterization of the LDPs on parylene films. The first challenge was how to
transfer the released LDPs from wafers into DI water, and the second challenge was to
select an appropriate matrix to mimic the human tissue environments and to support the
125
LDPs randomly distributed inside the testing samples during the heating tests. Our
approach for the first challenge was shown in Figure 9.4. The LOL stripper was used to
flush the wafer. Due to the very fast dissolution rate of the underneath LOL layer in the
stripper, the LDPs could be flushed out of the wafer into a testing tube. In contrary to
collecting the LDPs by soaking the entire wafer into the stripper, which resulted in a large
amount of LDPs stuck onto the sidewall of the striper container (even very hydrophobic
Teflon dish), almost all the LDPs could be collected and transferred into the testing tube
by using this flushing method.
Then we utilized centrifuge technology to clean and concentrate those particles. First, the
tube was put inside a centrifugation machine with 3000 rpm (1740 ×g) for 5 min, and all
the LDPs could be pushed to the bottom of the tube, as shown in Figure 9.4. The upper
Figure 9.4. Transferring the LDPs from wafers to DI water by LOL stripper flushing and multiple
times centrifugation.
126
layer of the stripper (alkaline solution) was removed from the tube and new DI water was
added into the tube to dilute the solution, then the centrifugation process was repeated
again. Sometimes, some particles might stick onto the sidewall of the testing tube. We
could use ultrasound agitation to separate them from the sidewall and to make them
uniformly distributed inside the testing tube again. After 8-10 times dilution, the pH value
of the solution was almost 7.0, and this solution, where the LDPs were in, was effectively
pure DI water. We can always spread the LDPs by ultrasound agitation and gather them
by centrifugation. It is also very easy to combine or divide the LDPs quantity within a
desired water volume by using this method. Moreover, SEM images show that the LDPs
had no morphological damages after such multiple centrifugations.
Since the surface-volume ratio of our micro-LDPs was relatively small, there was no
serious aggregation problem which usually occurs to nanoparticles [91, 95, 104, 110].
However, just because of the large size, the LDPs always precipitated rapidly inside the
water within one minute, which was a problem for heating effect characterization since
random distribution of the LDPs inside the testing samples were desired to mimic the real
clinical situations of LDPs around/inside tumor regions. Agarose, a jelly-like hydrogel,
was utilized as the matrix to support the LDPs and to mimic the human tissue
environments. The melting temperature of agarose hydrogel is above 85 °C – enough for
heating effect characterization. The process of our particle-suspended hydrogel setup is
shown in Figure 9.5. First, 1 gram agarose powder was dissolved into 100 mL hot DI
water, and cooled down for network cross-linked. This jelly-like hydrogel was then
127
ground completely by a regular blender into agarose mash, as shown in Figure 9.5(a) with
4 mL agarose mash in a testing tube after the first centrifugation (3000 rpm, 5min) was
executed to compress the agarose mash tightly inside the testing tube. The prepared DI
water-based LDPs solution was then poured into this testing tube (this may need several
times to completely transfer all the particles into the testing tube), and another
centrifugation (3000 rpm, 5min) was executed to gather the LDPs to the interface of the
agarose mash and water (Figure 9.5(b)). The upper layer of DI water was removed from
the tube to keep the total volume at 6 mL (volume ratio of agarose mash and water was
2:1). Afterwards an ultrasound probe was used to mix the bottom agarose mash and the
DI water with LDPs, leading to LDPs uniformly distributed inside the whole mixture.
The last centrifugation (1000 rpm (193 ×g), 2min) was executed to remove all the
bubbles inside the mixture induced by the ultrasound probe. Under such low speed
centrifugation, the LDPs distribution would not change. The final mixture was shown in
Figure 9.5. Combining the agarose hydrogel mash (as the matrix) and centrifuged LDP / water
solution by ultrasound mixing to obtain the agarose/LDP composite.
128
Figure 9.5(c), with total volume of 6 mL. The uniformly suspended LDPs inside this
agarose mash were very stable and have not precipitated yet even after several months,
probably because the LDPs had bonded onto the pieces of the agarose mash. Further
studies showed that the pure agarose mash and DI water have nearly the same
permittivity within the microwave range. Moreover, the final agarose mash mixture was
very viscous with much less convection compared to water. This indicated that this
agarose mash matrix was a good imitation of human tissue environments, and this
LDP-suspended hydrogel setup could mimic the real medical situations of LDPs
around/inside tumor regions.
The heating effect characterization based on different LDP concentrations are shown in
Figure 9.6. The measurement set up was the same as described in Figure 9.3. The
microwave input was 20 W at 1.9 GHz. High concentration (LDPs: 1×10
7
/mL), low
concentration (LDPs: 5×10
6
/mL), and pure agarose mash (combination of 4 mL pure
agarose mash plus 2 mL DI water) were shown as inserted in Figure 9.6. The microwave
reflection loss were measured during the experiments, which were -15.0 dB, -14.0 dB,
and -12.9 dB for high concentration, low concentration, and pure agarose mash,
respectively, meaning that more than 95% input power was delivered out of the
microwave probe aperture. It is obviously shown that higher concentration of the LDPs
provided a significantly larger enhancement of heating effect: at 60 seconds, with the
same input power of 20 W, the temperature enhancement of the high concentration case
was 1.64 times of that of the pure agarose case (21.2 °C to 56.6 °C versus 20.6 °C to
129
42.1 °C). The mechanism has been discussed as above: induced eddy current inside the
magnetic dipole could generate huge Ohmic loss. Moreover, this particle-suspended
hydrogel setup shows the two-step heating as well as described in the polymer-film
nanocomposite setup. Higher concentration showed larger differences between those two
steps, exhibiting greatly localized rapid heating enhancement, which was resulted from
small convection efficiency inside this agarose mash mixture. This also indicated that
alternative dosage regimens of the hyperthermia (higher dosage, less exposure time) may
Figure 9.6. Heating enhancement characterization of temperature vs. time based on different LDP
concentrations. The microwave input power was 20 W at 1.9 GHz.
130
provide improved local heating while minimizing heating in surrounding tissues [91].
9.3. Summary
In this chapter, we present our methods to collect / transfer the lithographically-defined
particles from wafers to testing tubes for the characterization of heating enhancement
effects. Two methods are shown here. For the polymer-film nanocomposite setup, the
entire parylene film can be easily peeled off from the wafer as a good matrix for our
LDPs. This method can easily collect all the particles and transfer them into DI water for
heating test; however, completely released particles are necessary for the medical
applications. In our second method, the particle-suspended hydrogel setup, centrifuge
technology was used to wash and concentrate the LDPs, and agarose mash (as the matrix)
was used to support the LDPs and mimic the human tissue environment. Notice that,
those methods can be applied not only on our disk-shaped gold particles, but also on
other types of LDPs (e.g., micro-resonators) if applicable. Moreover, the heating
enhancement effect was characterized based on these two setups. Both methods showed
great potentials of our LDPs as high-efficiency microwave absorbers to enhance the
localized selective heating for cancer hyperthermia therapy.
131
Chapter 10. Particle Optimization and Fundamental
Characterization
In this chapter, we will discuss the fabrication optimization of our particles (to remove
the parylene support), and the measurement results of the fundamental characterization.
In fact, our LDP/matrix were one type of nanocomposite for microwave absorption.
Microwave-absorbing nanocomposite attracts a great amount of interest nowadays,
because nanocomposite absorbers can combine advantages of both dielectric matrix and
nanoparticle absorbers and have significantly better characteristics [95]. Adding
nanoparticles into a bulk matrix can markedly adjust the microwave properties of the
composite [107-109] since the fillers are too small for microwave so that an averaged EM
effect will occur. From the perspective of dielectric loss, the enhancement of both real
and imaginary part of permittivity or permeability can improve the microwave absorption.
Larger real part of the permittivity or permeability can help to confine the same input
power into a smaller matrix volume due to smaller wavelength inside the matrix; while
larger imaginary part can directly increase the heating loss (i.e., microwave absorption).
In our previous chapter, we only measured the overall effect of the temperature vs.
particle concentration along with time. However, we did not know how the particles
affect the fundamental properties of the nanocomposite. In this chapter, the fundamental
parameters, real and imaginary part of permittivity, will be further characterized. This can
help us to better understand how the gold particles affect the composite properties
132
fundamentally.
10.1. Pure Gold Particles without Parylene
As shown in Figure 8.6, our disk-shaped lithographically defined particles were with
parylene support under the gold disk. Parylene is a lossless material for microwave,
which would reduce the volume loss density of the entire particle. Also, we only focus on
the effect from the gold micro-disks; any other factors in terms of structure or material
should be removed. In another word, in order to characterize the effect of the LDPs on
the permittivity of the nanocomposite, we first need to remove the parylene support in
our disk-shaped particles. In this way, new fabrication process should be proposed to
make pure gold particles without parylene.
In our previous fabrication process, parylene not only acted as the support for gold
micro-disks, but also protected the underneath sacrificial layer LOL 2000 layer during the
process. Generally, there are two ways to remove the parylene layer: the first way is to
remove the parylene after we fabricated the gold-parylene particles; the second way is to
avoid parylene being involved into the whole fabrication process. We did not attempt the
first method because parylene is quite chemically inert: most of the commonly-used and
safe etchant cannot wet-etch it. For the second way, if parylene is avoided from the whole
fabrication process, new sacrificial layer and releasing solvent should be put forward.
The ideal fabrication approach is shown in Figure 10.1: first, one sacrificial layer is
133
deposited onto silicon wafer, and then photolithography, metal evaporation and lift-off
process are executed to pattern the gold micro-disks directly on this sacrificial layer.
Finally, the sacrificial layer is dissolved by a particular releasing solvent and gold
particles can be released out of the wafer. The key point is to find a new sacrificial layer
and relevant dissolution solvent.
Some requirements should be limited to the material of the sacrificial layer: first, it
should be easily for deposition with good film quality; second, it should be
chemically-inert enough and would not be attacked during the following fabrication steps,
especially the photolithography step, until the releasing step; third, it should have fast
dissolution rate inside the releasing solvent, in order to guarantee that all the particles can
be flushed out of the wafer by the releasing solvent.
We proposed to use active metal as the sacrificial layer and hydrochloride acid as the
Figure 10.1. The ideal fabrication process to fabrication pure gold particles without parylene as the
underneath support.
134
releasing solvent. The metal layer can be easily deposited onto the silicon wafer by
e-beam evaporation. When releasing the particles by hydrochloride acid, the generated H2
gas can help to detach the particles out of the wafer. For active metals, potassium,
calcium, and sodium are so active that even can react with water. Aluminum is an
amphoteric metal, which would be attacked by the photolithography developer. Titanium
is an active metal; however, its intrinsic oxidation layer can protect it from both acidic
and alkaline solution. Zinc presents desired chemical properties, but the film quality by
e-beam evaporation is not good enough. Chromium may not have fast reaction rate with
hydrochloride acid, meaning that the particles cannot be flushed out of the wafer rapidly.
In the contrary, magnesium (Mg) satisfies all our requirements for the sacrificial layer.
Therefore, Mg is selected as our active-metal sacrificial layer.
The recipe of Mg e-beam evaporation had been developed. The parameters are shown in
Figure 10.2. The parameters of all other fabrication processes (photolithography, metal
evaporation and lift-off process) were the same as described in Chapter 8.3. Notice that
by using Mg as the sacrificial layer, no adhesion layer (e.g., 3 nm Ti) was necessary since
Mg was adhesive enough to gold. After such gold micro-disks were patterned onto the
Figure 10.2. The parameters of Mg e-beam evaporation.
135
Mg layer, 10% - 15% hydrochloride acid was used to rapidly dissolve the Mg layer and
flush the particles out of the wafer and into the testing tube. The fabrication results, such
pure gold micro-disks without parylene support, are shown in Figure 10.3.
Due to its very low melting temperature and high vapor pressure, the deposition rate of
Mg evaporation was very high and could not be controlled very well. In our experiments,
the deposition rate was very unstable and could be varied from 6 Å/s to 15 Å/s. Thus, the
surface was not uniform and had a great amount of defects. After gold deposition, this
non-uniformity was transferred to the surface of the gold micro-disks, as shown in Figure
10.3. However, such small defects had no harm to the overall microwave absorption of
the entire particles. Another issue of this process was that acid solution could lead to
aggregation of gold particles, even though our particles were in micron scale. To solve
this issue, acetone could be added first inside the particle-releasing tube as the separator,
Figure 10.3. Pure gold micro-disks without parylene support. The gold surface was not uniform due
to the high deposition rate of Mg evaporation.
136
to neutralize the charges accumulation. Further study shows that, these 8 μm diameter
and 300 nm thick gold particles were mechanically robust enough during the
centrifugation process and ultrasound vibration process, meaning that in fact there is no
need for the parylene support to enhance the mechanical strength of the entire particles.
10.2. Permittivity Characterization
In experiments, the permittivity of our LDP/matrix nanocomposite had been
characterized by commercialized equipment. Both the real part and the imaginary part of
the permittivity are the function of frequency. In our experiments, they are also related to
the concentration of the gold particles.
The liquid-state PDMS (A) (poly(dimethyl-methylvinylsiloxane)) was selected as the
matrix. The first reason is that PDMS (A) is viscous enough, meaning that there was no
obvious precipitation phenomenon of our heavy gold particles inside such liquid.
Experiments found that those particles could be relatively uniformly distributed inside
PDMS (A) for several days, long enough for measurement. The second reason is that
PDMS (A) is lossless and does not absorb microwave. In another word, the permittivity
of PDMS (A) are relatively small. In this case, significant enhancement of the value of
the permittivity can be easily observed if gold particles added into the composite. That is
to say, lossless PDMS is a good background contrast to demonstrate the obvious effects
on the increasing permittivity of our LDP/matrix nanocomposite.
137
The method to transfer such pure gold micro-disks to PDMS (A) was quite similar with
the method to make the hydrogel-based nanocomposite, as discussed in Chapter 9.2. First,
the gold micro-disks were flushed out of the wafer by 10% - 15% hydrochloride acid and
were collected into the testing tube with hydrochloride acid as the surroundings.
Centrifugation was executed with 3000 rpm (1740 ×g) for 5 min to push all the LDPs to
the bottom of the tube. Then the upper layer liquid inside the testing tube was replaced by
pure acetone directly, in order to dilute the hydrochloride acid surroundings. After 8 – 10
times dilution, the surroundings could be considered as pure acetone. Here we utilized
pure acetone directly because it is a good separator for such gold particles in acidic
environment. Then such LDP/acetone solution was poured into another testing tube with
volume-fixed liquid-state PDMS (A) in it, as shown in Figure 10.4. Those gold
microparticles would precipitate down to the interface of acetone and PDMS (A) rapidly
Figure 10.4. The LDP/PDMS mixture was obtained by mixing the liquid PDMS (A) and
LDP/acetone solution by ultrasound agitation.
138
due to the large difference of the density between gold microparticles and acetone
solution. When all the particles precipitated down to the interface (normally within 20
min), the upper layer acetone was removed and only 2 mL acetone was left inside the
tube, in order to fix the total volume of acetone as well. Afterwards, ultrasound agitation
was used to completely mix the liquid PDMS (A) and LDP/acetone solution. In this way,
all the particles could be uniformly distributed inside the liquid PDMS (A). Finally, the
air bubbles induced by the ultrasound agitation and the residue acetone were removed by
mechanical pumping for 15 min. The final LDP/PDMS mixture is shown in Figure 10.4.
The permittivity was measured by the impedance analyzer Agilent 4991A with the
adapter Keysight 16453A (as shown in Figure 10.5). The material to be measured was
clamped between two cylinder-like parallel-planar electrodes and thus formed a capacitor.
Because our nanocomposite was liquid-based, a homemade ring-holder was necessary to
Figure 10.5. The measurement set-up for permittivity characterization. A ring-holder was used to
fix the liquid composite between the two parallel-planar electrodes.
139
fix the liquid composite between the two parallel-planar electrodes, also as shown in
Figure 10.5. The ring-holder was made of acrylic, which is also a lossless material with
relatively small real and imaginary permittivity. During the experiments, such set-up can
only measure the overall effective permittivity of the entire structure, including both our
liquid nanocomposite and the ring-holder. However, in fact the two capacitors from the
liquid composite and the ring-holder were in parallel between the two electrodes,
meaning that theoretically the effective permittivity of the entire structure can be
calculated as,
12 eff
AB
(10.1)
Here εeff is the measured overall effective permittivity, ε1 and ε2 are the permittivity of the
liquid composite and the ring-holder, respectively, A and B are two geometric factors,
which would not change if we did not change the ring-holder. During the experiments, we
first obtained those two geometric factors A and B by measuring the overall effective
permittivity based on ε1-known materials (e.g., air and acrylic, still using the ring-holder),
then used such two known geometric factors and the measured overall effective
permittivity to calculate the permittivity of our liquid composite. Notice that this method
could be applied on both real and imaginary part of the permittivity. In this way, the
permittivity of our liquid LDP/PDMS mixture could be achieved. Due to the equipment
capability, the sweeping frequency range for permittivity characterization was from 300
MHz to 1 GHz, which were also commonly-used frequencies for cancer hyperthermia
therapy.
140
In our experiments, the original PDMS (A) was 4 mL and the LDP/acetone solution was
2 mL. After ultrasound mixing and mechanical pumping, the total volume of the
LDP/PDMS composite was still around 4 mL, as shown in Figure 10.6(a). The
characterized composites were based on four different concentrations of the gold
micro-disks: pure PDMS (A), 0.75×10
7
/mL, 1.5×10
7
/mL, 3×10
7
/mL. The calculated
permittivity of those four samples (both real part and imaginary part) were shown in
Figure 10.6(b)-(c). Both real part and imaginary part of the permittivity value were
increased after adding the gold particles into PDMS (A), and it is very obvious that
Figure 10.6. The characterization results of permittivity of LDP/PDMS(A) composites. (a) The
composites with four different LDP concentrations, pure PDMS (A), 0.75×10
7
/mL, 1.5×10
7
/mL,
3×10
7
/mL. (b) Real part of permittivity measurement. (c) Imaginary part of permittivity
measurement.
141
larger-concentration samples provided larger enhancement of the permittivity value. For
the largest-concentration sample in our experiment, the real permittivity increased from
3.4 to around 6.0, and imaginary permittivity increased from around 0.01 to 0.1 (10 times
enhancement); however, the volume ratio of our gold particles to PDMS matrix was just
around 1 : 2500, indicating very promising potentials of our disk-shaped lithographically
defined particles as high-efficiency microwave absorbers for cancer hyperthermia
therapy.
10.3. Summary
In this chapter, we discussed our method to use Mg as the sacrificial layer and
hydrochloride acid as the releasing solvent to fabricate pure gold particles without
parylene support. Those particles were mixed with viscous and lossless PDMS (A) as the
LDP/matrix composite for permittivity measurement. The permittivity characterization
based on our homemade ring-shaped holders showed that both real part and imaginary
part of the permittivity value were enhanced after adding the gold particles into the
matrix, indicating great potentials of our disk-shaped lithographically defined particles as
high-efficiency microwave absorbers for cancer hyperthermia therapy.
142
Conclusion and Future Work
This project presents a novel approach to utilize lithographically defined
micro/nano-particles to significantly enhance the heating effects under microwave
radiation, with a great potential to achieve selectively localized heating for cancer
hyperthermia therapy. LC-circuit-like resonance particles were numerically studied using
finite element methods, and was discovered that the limitation of the micro-resonators
was not the low resonance frequency trade-off with the small size, but the low Q-value
due to insufficient conductivity of materials. For non-resonance particles, the structures
and materials were optimized by numerical studies, and disk-shaped gold magnetic
dipoles with induced eddy current were selected as the particles to improve the
microwave absorption efficiency. The LDPs fabrication, releasing and collection
processes were developed. The microwave heating enhancement effects have been
successfully demonstrated by both polymer-film nanocomposite setup with parylene film
as matrix and particle-suspended hydrogel setup with completely released LDPs
randomly distributed inside agarose mash. Moreover, fundamental characterization
showed that both real part and imaginary part of the permittivity value of the composite
were enhanced after adding the gold particles into the matrix, which was probably the
reason why the microwave absorption efficiency can be improved. Due to the capability
of our equipment, here we only measured the permittivity value. In order to further
understand the effect from our gold magnetic dipoles, permeability should also be
143
measured and analyzed. Moreover, here we only demonstrated the disk-shaped LDPs,
which in fact can be further optimized by new structures or materials, e.g., gold-nickel
stacking layers, carbon nanotubes conjugation, or matching-enhanced coating.
Meanwhile, our disk-shaped gold LDPs are also an ideal platform for chemical group
conjugation [90, 92, 93], (e.g., poly (ethylene glycol) (PEG), or targeting agents) to
extend the circulation time inside human bodies [72, 84], and to enhance the specific
accumulation of LDPs near tumor regions [71]. Besides that, the modified LDPs can also
be used for microwave imaging since it can significantly increase the dielectric contrast
between tumors and benign regions [108].
At last, in vivo experiments based on LDPs to
measure the hyperthermia effect and thermal biotoxicity should be investigated in the
future work.
144
Reference
1. Kalamaras, C. M., & Efstathiou, A. M. (2013, June). Hydrogen production technologies:
current state and future developments. In Conference Papers in Science (Vol. 2013). Hindawi
Publishing Corporation.
2. Holladay, J. D., Hu, J., King, D. L., & Wang, Y . (2009). An overview of hydrogen production
technologies. Catalysis today, 139(4), 244-260.
3. Fujishima, A., & Honda, K. (1972). Electrochemical photolysis of water at a semiconductor
electrode. Nature, 238(5358), 37-38.
4. Liu, C., Tang, J., Chen, H. M., Liu, B., & Yang, P. (2013). A fully integrated nanosystem of
semiconductor nanowires for direct solar water splitting. Nano letters, 13(6), 2989-2992.
5. Luo, J., Im, J. H., Mayer, M. T., Schreier, M., Nazeeruddin, M. K., Park, N. G., ... & Grätzel,
M. (2014). Water photolysis at 12.3% efficiency via perovskite photovoltaics and
Earth-abundant catalysts. Science, 345(6204), 1593-1596.
6. Zoulias, E., Varkaraki, E., Lymberopoulos, N., Christodoulou, C. N., & Karagiorgis, G. N.
(2004). A review on water electrolysis. TCJST, 4(2), 41-71.
7. De Souza, R. F., Padilha, J. C., Gonçalves, R. S., De Souza, M. O., & Rault-Berthelot, J.
(2007). Electrochemical hydrogen production from water electrolysis using ionic liquid as
electrolytes: towards the best device. Journal of Power Sources, 164(2), 792-798.
8. Ursua, A., Gandia, L. M., & Sanchis, P. (2012). Hydrogen production from water electrolysis:
current status and future trends. Proceedings of the IEEE, 100(2), 410-426.
9. LeRoy, R. L. (1983). Industrial water electrolysis: present and future. International Journal of
Hydrogen Energy, 8(6), 401-417.
10. De Levie, R. (1999). The electrolysis of water. Journal of Electroanalytical
Chemistry, 476(1), 92-93.
11. Santos, D. M., Sequeira, C. A., & Figueiredo, J. L. (2013). Hydrogen production by alkaline
water electrolysis. Química Nova, 36(8), 1176-1193.
145
12. Gong, M., Zhou, W., Tsai, M. C., Zhou, J., Guan, M., Lin, M. C., ... & Pennycook, S. J.
(2014). Nanoscale nickel oxide/nickel heterostructures for active hydrogen evolution
electrocatalysis. Nature communications, 5, 4695.
13. Wang, H., Lee, H. W., Deng, Y ., Lu, Z., Hsu, P. C., Liu, Y ., ... & Cui, Y . (2015). Bifunctional
non-noble metal oxide nanoparticle electrocatalysts through lithium-induced conversion for
overall water splitting. Nature communications, 6.
14. Cobo, S., Heidkamp, J., Jacques, P. A., Fize, J., Fourmond, V., Guetaz, L., ... & Fontecave, M.
(2012). A Janus cobalt-based catalytic material for electro-splitting of water. Nature
materials, 11(9).
15. Doenitz, W., Schmidberger, R., Steinheil, E., & Streicher, R. (1980). Hydrogen production by
high temperature electrolysis of water vapour. International Journal of Hydrogen
Energy, 5(1), 55-63.
16. LeRoy, R. L., Bowen, C. T., & LeRoy, D. J. (1980). The thermodynamics of aqueous water
electrolysis. Journal of The Electrochemical Society, 127(9), 1954-1962.
17. Carmo, M., Fritz, D. L., Mergel, J., & Stolten, D. (2013). A comprehensive review on PEM
water electrolysis. International journal of hydrogen energy, 38(12), 4901-4934.
18. Rashid, M. M., Al Mesfer, M. K., Naseem, H., & Danish, M. (2015). Hydrogen production by
water electrolysis: a review of alkaline water electrolysis, PEM water electrolysis and high
temperature water electrolysis. Int J Eng Adv Technol, 4(3), 80-93.
19. Ma, C., Contento, N. M., & Bohn, P. W. (2014). Redox cycling on recessed ring-disk
nanoelectrode arrays in the absence of supporting electrolyte. Journal of the American
Chemical Society, 136(20), 7225-7228.
20. Chen, Q., McKelvey, K., Edwards, M. A., & White, H. S. (2016). Redox Cycling in Nanogap
Electrochemical Cells. The Role of Electrostatics in Determining the Cell Response. The
Journal of Physical Chemistry C, 120(31), 17251-17260.
21. Joshi, R. P., Qian, J., Schoenbach, K. H., & Schamiloglu, E. (2004). Microscopic analysis for
water stressed by high electric fields in the prebreakdown regime. Journal of applied
146
physics, 96(7), 3617-3625.
22. Xiong, J., Chen, Q., Edwards, M. A., & White, H. S. (2015). Ion Transport within High
Electric Fields in Nanogap Electrochemical Cells.
23. De Kreuk, C. W., Sluyters-Rehbach, M., & Sluyters, J. H. (1970). Electrode kinetics and
double-layer structure. Journal of Electroanalytical Chemistry, 28(2), 391-407.
24. Bard, A. J., Faulkner, L. R., Leddy, J., & Zoski, C. G. (1980). Electrochemical methods:
fundamentals and applications (Vol. 2). New York: wiley.
25. Stuve, E. M. (2012). Ionization of water in interfacial electric fields: An electrochemical
view. Chemical Physics Letters, 519, 1-17.
26. Li, T., Hu, W., & Zhu, D. (2010). Nanogap electrodes. Advanced Materials, 22(2), 286-300.
27. Rassaei, L., Singh, P. S., & Lemay, S. G. (2011). Lithography-based nanoelectrochemistry.
28. Chen, F., Qing, Q., Ren, L., Wu, Z., & Liu, Z. (2005). Electrochemical approach for
fabricating nanogap electrodes with well controllable separation. Applied Physics
Letters, 86(12), 123105.
29. Liu, S., Tok, J. B. H., & Bao, Z. (2005). Nanowire lithography: fabricating controllable
electrode gaps using Au− Ag− Au nanowires. Nano letters, 5(6), 1071-1076.
30. Zhao, Q., Wang, W., Shao, J., Li, X., Tian, H., Liu, L., ... & Lu, B. (2016). Nanoscale
Electrodes for Flexible Electronics by Swelling Controlled Cracking. Advanced
Materials, 28(30), 6337-6344.
31. Beesley, D. J., Semple, J., Jagadamma, L. K., Amassian, A., McLachlan, M. A., &
Anthopoulos, T. D. (2014). Sub-15-nm patterning of asymmetric metal electrodes and devices
by adhesion lithography. Nature communications, 5.
32. Wolfrum, B., Zevenbergen, M., & Lemay, S. (2008). Nanofluidic redox cycling amplification
for the selective detection of catechol. Analytical chemistry, 80(4), 972-977.
33. Ka ̈ telho ̈ n, E., Hofmann, B., Lemay, S. G., Zevenbergen, M. A., Offenha ̈ usser, A., & Wolfrum,
B. (2010). Nanocavity redox cycling sensors for the detection of dopamine fluctuations in
microfluidic gradients. Analytical chemistry, 82(20), 8502-8509.
147
34. Zevenbergen, M. A., Wolfrum, B. L., Goluch, E. D., Singh, P. S., & Lemay, S. G. (2009). Fast
electron-transfer kinetics probed in nanofluidic channels. Journal of the American Chemical
Society, 131(32), 11471-11477.
35. Ma, C., Xu, W., Wichert, W. R., & Bohn, P. W. (2016). Ion accumulation and migration
effects on redox cycling in nanopore electrode arrays at low ionic strength. ACS nano, 10(3),
3658-3664.
36. Fu, K., Han, D., Ma, C., & Bohn, P. W. (2016). Electrochemistry at single molecule
occupancy in nanopore-confined recessed ring-disk electrode arrays. Faraday
discussions, 193, 51-64.
37. Miles, M. H. (1975). Evaluation of electrocatalysts for water electrolysis in alkaline
solutions. Journal of Electroanalytical Chemistry and Interfacial Electrochemistry, 60(1),
89-96.
38. Takeno, N. (2005). Atlas of Eh-pH diagrams. Geological survey of Japan open file
report, 419, 102.
39. Gao, P., Gosztola, D., Leung, L. W. H., & Weaver, M. J. (1987). Surface-enhanced Raman
scattering at gold electrodes: dependence on electrochemical pretreatment conditions and
comparisons with silver. Journal of electroanalytical chemistry and interfacial
electrochemistry, 233(1-2), 211-222.
40. Wang, Y ., Liu, H., Li, Y ., & Wu, W. (2015). Low DC-bias silicon nitride anisotropic
etching. Journal of Vacuum Science & Technology B, Nanotechnology and Microelectronics:
Materials, Processing, Measurement, and Phenomena, 33(6), 06FA01.
41. Sigmund, P. (2016). Six decades of atomic collisions in solids. Nuclear Instruments and
Methods in Physics Research Section B: Beam Interactions with Materials and Atoms.
42. Gao, D., Wijesundara, M. B., Carraro, C., Howe, R. T., & Maboudian, R. (2004). Recent
progress toward a manufacturable polycrystalline SiC surface micromachining
technology. IEEE Sensors Journal, 4(4), 441-448.
43. Abramova, V., Slesarev, A. S., & Tour, J. M. (2015). Meniscus-Mask Lithography for
148
fabrication of narrow nanowires. Nano letters, 15(5), 2933-2937.
44. Hübner, H. (1992). Calculations on deposition and redeposition in plasma etch
processes. Journal of The Electrochemical Society, 139(11), 3302-3309.
45. Chen, W., Sugita, K., Morikawa, Y ., Yasunami, S., Hayashi, T., & Uchida, T. (2001).
Application of magnetic neutral loop discharge plasma in deep silica etching. Journal of
Vacuum Science & Technology A: Vacuum, Surfaces, and Films, 19(6), 2936-2940.
46. Lide, D. R., Baysinger, G., Berger, L. I., Goldberg, R. N., Kehiaian, H. V., Kuchitsu, K., ... &
Zwillinger, D. CRC Handbook of Chemistry and Physics.
47. Pashley, R. M., Rzechowicz, M., Pashley, L. R., & Francis, M. J. (2005). De-gassed water is a
better cleaning agent. The Journal of Physical Chemistry B, 109(3), 1231-1238.
48. Diaz-Morales, O., Calle-Vallejo, F., de Munck, C., & Koper, M. T. (2013). Electrochemical
water splitting by gold: evidence for an oxide decomposition mechanism. Chemical
Science, 4(6), 2334-2343.
49. Oesch, U., & Janata, J. (1983). Electrochemical study of gold electrodes with anodic oxide
films—I. Formation and reduction behaviour of anodic oxides on gold. Electrochimica
Acta, 28(9), 1237-1246.
50. Rossmeisl, J., Logadottir, A., & Nørskov, J. K. (2005). Electrolysis of water on (oxidized)
metal surfaces. Chemical physics, 319(1), 178-184.
51. “Heat of combustion”. In Wikipedia. Retrieved October 20, 2017, from
https://en.wikipedia.org/wiki/Heat_of_combustion
52. Wendt, H., & Imarisio, G. (1988). Nine years of research and development on advanced water
electrolysis. A review of the research programme of the Commission of the European
Communities. Journal of Applied Electrochemistry, 18(1), 1-14.
53. Yeo, B. S., Klaus, S. L., Ross, P. N., Mathies, R. A., & Bell, A. T. (2010). Identification of
hydroperoxy species as reaction intermediates in the electrochemical evolution of oxygen on
gold. ChemPhysChem, 11(9), 1854-1857.
54. WANG, H., ZHONG, C., JIANG, C. J., GU, X., LI, J., & JIANG, Y . M. (2009).
149
Electrochemical Behavior of ITO Films during Anodic and Cathodic Polarization in Sodium
Hydroxide Solutions. Acta Physico-Chimica Sinica, 25(5), 835-839.
55. Matsumoto, Y ., & Sato, E. (1986). Electrocatalytic properties of transition metal oxides for
oxygen evolution reaction. Materials chemistry and physics, 14(5), 397-426.
56. Chen, Y . W., Prange, J. D., Dühnen, S., Park, Y ., Gunji, M., Chidsey, C. E., & McIntyre, P. C.
(2011). Atomic layer-deposited tunnel oxide stabilizes silicon photoanodes for water
oxidation. Nature materials, 10(7), 539-544.
57. Bao, J. (2015). Photoelectrochemical water splitting: A new use for bandgap
engineering. Nature nanotechnology, 10(1), 19-20.
58. Kikuchi, K., Nagata, S., Tanaka, Y ., Saihara, Y ., & Ogumi, Z. (2007). Characteristics of
hydrogen nanobubbles in solutions obtained with water electrolysis. Journal of
Electroanalytical Chemistry, 600(2), 303-310.
59. Kikuchi, K., Tanaka, Y ., Saihara, Y ., Maeda, M., Kawamura, M., & Ogumi, Z. (2006).
Concentration of hydrogen nanobubbles in electrolyzed water. Journal of colloid and
interface science, 298(2), 914-919.
60. Lohrengel, M. M., & Schultze, J. W. (1976). Electrochemical properties of anodic gold oxide
layers—I: Potentiostatic oxide growth and double layer capacity. Electrochimica Acta, 21(11),
957-965.
61. Conway, B. E., Sattar, M. A., & Gilroy, D. (1969). Electrochemistry of the nickel-oxide
electrode—V. Self-passivation effects in oxygen-evolution kinetics. Electrochimica
Acta, 14(8), 677-694.
62. Lu, P. W. T., & Srinivasan, S. (1978). Electrochemical ‐Ellipsometric Studies of Oxide Film
Formed on Nickel during Oxygen Evolution. Journal of the Electrochemical Society, 125(9),
1416-1422.
63. Hu, Z., Wu, M., Wei, Z., Song, S., & Shen, P. K. (2007). Pt-WC/C as a cathode
electrocatalyst for hydrogen production by methanol electrolysis. Journal of Power
Sources, 166(2), 458-461.
150
64. Yu, E. H., Krewer, U., & Scott, K. (2010). Principles and materials aspects of direct alkaline
alcohol fuel cells. Energies, 3(8), 1499-1528.
65. Dinh, H. N., Ren, X., Garzon, F. H., Zelenay, P., & Gottesfeld, S. (2000). Electrocatalysis in
direct methanol fuel cells: in-situ probing of PtRu anode catalyst surfaces. Journal of
Electroanalytical Chemistry, 491(1), 222-233.
66. Lim, K. T., Kim, D. P., Kim, K. T., & Kim, C. I. (2003). The effect of CF 4 addition on Ru
etching with inductively coupled plasma. Journal of the Korean Physical Society, 42(Suppl.),
829-832.
67. Heron, M. P. (2015). Deaths: leading causes for 2012.
68. McGuire, S. (2016). World cancer report 2014. Geneva, Switzerland: World Health
Organization, international agency for research on cancer, WHO Press, 2015. Advances in
Nutrition: An International Review Journal, 7(2), 418-419.
69. Svaasand, L. O., Gomer, C. J., & Morinelli, E. (1990). On the physical rationale of laser
induced hyperthermia. Lasers in Medical Science, 5(2), 121-128.
70. Milleron, R. S., & Bratton, S. B. (2007). ‘Heated’debates in apoptosis. Cellular and
Molecular Life Sciences, 64(18), 2329-2333.
71. El-Sayed, I. H., Huang, X., & El-Sayed, M. A. (2006). Selective laser photo-thermal therapy
of epithelial carcinoma using anti-EGFR antibody conjugated gold nanoparticles. Cancer
letters, 239(1), 129-135.
72. O'Neal, D. P., Hirsch, L. R., Halas, N. J., Payne, J. D., & West, J. L. (2004). Photo-thermal
tumor ablation in mice using near infrared-absorbing nanoparticles. Cancer letters, 209(2),
171-176.
73. Dutz, S., & Hergt, R. (2014). Magnetic particle hyperthermia—a promising tumour
therapy?. Nanotechnology, 25(45), 452001.
74. Gazelle, G. S., Goldberg, S. N., Solbiati, L., & Livraghi, T. (2000). Tumor ablation with
radio-frequency energy. Radiology, 217(3), 633-646.
75. Ikeda, N., Hayashida, O., Kameda, H., Ito, H., & Matsuda, T. (1994). Experimental study on
151
thermal damage to dog normal brain. International journal of hyperthermia, 10(4), 553-561.
76. Mirza, A. N., Fornage, B. D., Sneige, N., Kuerer, H. M., Newman, L. A., Ames, F. C., &
Singletary, S. E. (2001). Radiofrequency ablation of solid tumors. Cancer journal (Sudbury,
Mass.), 7(2), 95-102.
77. Luyen, H., Gao, F., Hagness, S. C., & Behdad, N. (2014). Microwave ablation at 10.0 GHz
achieves comparable ablation zones to 1.9 GHz in ex vivo bovine liver. IEEE Transactions on
Biomedical Engineering, 61(6), 1702-1710.
78. Stang, J., Haynes, M., Carson, P., & Moghaddam, M. (2012). A preclinical system prototype
for focused microwave thermal therapy of the breast. IEEE Transactions on Biomedical
Engineering, 59(9), 2431-2438.
79. Seki, T., Wakabayashi, M., Nakagawa, T., Imamura, M., Tamai, T., Nishimura, A., ... & Inoue,
K. (1999). Percutaneous microwave coagulation therapy for patients with small
hepatocellular carcinoma. Cancer, 85(8), 1694-1702.
80. Huber, P., Debus, J., Jenne, J., Jöchle, K., van Kaick, G., Lorenz, W. J., & Wannenmacher, M.
(1996). Therapeutic ultrasound in tumor therapy. Principles, applications and new
developments. Der Radiologe, 36(1), 64-71.
81. Wu, F., Chen, W. Z., Bai, J., Zou, J. Z., Wang, Z. L., Zhu, H., & Wang, Z. B. (2001).
Pathological changes in human malignant carcinoma treated with high-intensity focused
ultrasound. Ultrasound in medicine & biology, 27(8), 1099-1106.
82. Philipp, C. M., Rohde, E., & Berlien, H. (1995, July). Nd: YAG laser procedures in tumor
treatment. In Seminars in surgical oncology (V ol. 11, No. 4, pp. 290-298). John Wiley & Sons,
Inc..
83. Prudhomme, M., Tang, J., Rouy, S., Delacrétaz, G., Salathé, R. P., & Godlewskl, G. (1996).
Interstitial diode laser hyperthermia in the treatment of subcutaneous tumor. Lasers in surgery
and medicine, 19(4), 445-450.
84. von Maltzahn, G., Park, J. H., Agrawal, A., Bandaru, N. K., Das, S. K., Sailor, M. J., &
Bhatia, S. N. (2009). Computationally guided photothermal tumor therapy using
152
long-circulating gold nanorod antennas. Cancer research, 69(9), 3892-3900.
85. Zharov, V. P., Galitovsky, V., & Viegas, M. (2003). Photothermal detection of local thermal
effects during selective nanophotothermolysis. Applied Physics Letters, 83(24), 4897-4899.
86. Huang, X., El-Sayed, I. H., Qian, W., & El-Sayed, M. A. (2006). Cancer cell imaging and
photothermal therapy in the near-infrared region by using gold nanorods. Journal of the
American Chemical Society, 128(6), 2115-2120.
87. Takahashi, H., Niidome, T., Nariai, A., Niidome, Y ., & Yamada, S. (2006). Photothermal
reshaping of gold nanorods prevents further cell death. Nanotechnology, 17(17), 4431.
88. Loo, C., Lin, A., Hirsch, L., Lee, M. H., Barton, J., Halas, N., ... & Drezek, R. (2004).
Nanoshell-enabled photonics-based imaging and therapy of cancer. Technology in cancer
research & treatment, 3(1), 33-40.
89. Loo, C., Lowery, A., Halas, N., West, J., & Drezek, R. (2005). Immunotargeted nanoshells for
integrated cancer imaging and therapy. Nano letters, 5(4), 709-711.
90. Huang, X., Jain, P. K., El-Sayed, I. H., & El-Sayed, M. A. (2008). Plasmonic photothermal
therapy (PPTT) using gold nanoparticles. Lasers in medical science, 23(3), 217.
91. Hirsch, L. R., Stafford, R. J., Bankson, J. A., Sershen, S. R., Rivera, B., Price, R. E., ... &
West, J. L. (2003). Nanoshell-mediated near-infrared thermal therapy of tumors under
magnetic resonance guidance. Proceedings of the National Academy of Sciences, 100(23),
13549-13554.
92. Cherukuri, P., Glazer, E. S., & Curley, S. A. (2010). Targeted hyperthermia using metal
nanoparticles. Advanced drug delivery reviews, 62(3), 339-345.
93. Day, E. S., Morton, J. G., & West, J. L. (2009). Nanoparticles for thermal cancer
therapy. Journal of biomechanical engineering, 131(7), 074001.
94. Hilger, I., Hergt, R., & Kaiser, W. A. (2005). Towards breast cancer treatment by magnetic
heating. Journal of Magnetism and Magnetic Materials, 293(1), 314-319.
95. Bregar, V. B. (2004). Advantages of ferromagnetic nanoparticle composites in microwave
absorbers. IEEE Transactions on Magnetics, 40(3), 1679-1684.
153
96. Hergt, R., Hiergeist, R., Zeisberger, M., Glöckl, G., Weitschies, W., Ramirez, L. P., ... &
Kaiser, A. W. (2004). Enhancement of AC-losses of magnetic nanoparticles for heating
applications. Journal of Magnetism and Magnetic Materials, 280(2), 358-368.
97. Johannsen, M., Gneveckow, U., Eckelt, L., Feussner, A., Waldöfner, N., Scholz, R., ... &
Jordan, A. (2005). Clinical hyperthermia of prostate cancer using magnetic nanoparticles:
presentation of a new interstitial technique. International journal of hyperthermia, 21(7),
637-647.
98. Ito, A., Shinkai, M., Honda, H., & Kobayashi, T. (2005). Medical application of
functionalized magnetic nanoparticles. Journal of bioscience and bioengineering, 100(1),
1-11.
99. Hergt, R., Andra, W., d'Ambly, C. G., Hilger, I., Kaiser, W. A., Richter, U., & Schmidt, H. G.
(1998). Physical limits of hyperthermia using magnetite fine particles. IEEE Transactions on
Magnetics, 34(5), 3745-3754.
100. Lu, B., Dong, X. L., Huang, H., Zhang, X. F., Zhu, X. G., Lei, J. P., & Sun, J. P. (2008).
Microwave absorption properties of the core/shell-type iron and nickel nanoparticles. Journal
of magnetism and magnetic materials, 320(6), 1106-1111.
101. Toneguzzo, P., Viau, G., Acher, O., Fiévet ‐Vincent, F., & Fiévet, F. (1998). Monodisperse
ferromagnetic particles for microwave applications. Advanced Materials, 10(13), 1032-1035.
102. Deatsch, A. E., & Evans, B. A. (2014). Heating efficiency in magnetic nanoparticle
hyperthermia. Journal of Magnetism and Magnetic Materials, 354, 163-172.
103. Eberbeck, D., Wiekhorst, F., Steinhoff, U., & Trahms, L. (2006). Aggregation behaviour of
magnetic nanoparticle suspensions investigated by magnetorelaxometry. Journal of Physics:
Condensed Matter, 18(38), S2829.
104. Hergt, R., Dutz, S., & Röder, M. (2008). Effects of size distribution on hysteresis losses of
magnetic nanoparticles for hyperthermia. Journal of Physics: Condensed Matter, 20(38),
385214.
105. Cardinal, J., Klune, J. R., Chory, E., Jeyabalan, G., Kanzius, J. S., Nalesnik, M., & Geller, D.
154
A. (2008). Noninvasive radiofrequency ablation of cancer targeted by gold
nanoparticles. Surgery, 144(2), 125-132.
106. Gannon, C. J., Patra, C. R., Bhattacharya, R., Mukherjee, P., & Curley, S. A. (2008).
Intracellular gold nanoparticles enhance non-invasive radiofrequency thermal destruction of
human gastrointestinal cancer cells. Journal of nanobiotechnology, 6(1), 2.
107. Gannon, C. J., Cherukuri, P., Yakobson, B. I., Cognet, L., Kanzius, J. S., Kittrell, C., ... &
Curley, S. A. (2007). Carbon nanotube ‐enhanced thermal destruction of cancer cells in a
noninvasive radiofrequency field. Cancer, 110(12), 2654-2665.
108. Mashal, A., Sitharaman, B., Li, X., Avti, P. K., Sahakian, A. V., Booske, J. H., & Hagness, S.
C. (2010). Toward carbon-nanotube-based theranostic agents for microwave detection and
treatment of breast cancer: Enhanced dielectric and heating response of tissue-mimicking
materials. IEEE Transactions on Biomedical Engineering, 57(8), 1831-1834.
109. Yan, D., Cheng, S., Zhuo, R. F., Chen, J. T., Feng, J. J., Feng, H. T., ... & Yan, P. X. (2009).
Nanoparticles and 3D sponge-like porous networks of manganese oxides and their microwave
absorption properties. Nanotechnology, 20(10), 105706.
110. Phang, S. W., Tadokoro, M., Watanabe, J., & Kuramoto, N. (2008). Synthesis,
characterization and microwave absorption property of doped polyaniline nanocomposites
containing TiO 2 nanoparticles and carbon nanotubes. Synthetic Metals, 158(6), 251-258.
111. Phang, S. W., Tadokoro, M., Watanabe, J., & Kuramoto, N. (2008). Microwave absorption
behaviors of polyaniline nanocomposites containing TiO 2 nanoparticles. Current Applied
Physics, 8(3), 391-394.
112. Basavaraja, C., Kim, W. J., & Do Kim, Y . (2011). Synthesis of polyaniline-gold/graphene
oxide composite and microwave absorption characteristics of the composite films. Materials
Letters, 65(19), 3120-3123.
113. Ohlan, A., Singh, K., Chandra, A., & Dhawan, S. K. (2008). Microwave absorption
properties of conducting polymer composite with barium ferrite nanoparticles in 12.4–18
GHz. Applied Physics Letters, 93(5), 053114.
155
114. Turgeon, M. L. (2005). Clinical hematology: theory and procedures. Lippincott Williams &
Wilkins.
115. Mohan, S. S., del Mar Hershenson, M., Boyd, S. P., & Lee, T. H. (1999). Simple accurate
expressions for planar spiral inductances. IEEE Journal of solid-state circuits, 34(10),
1419-1424.
116. Sherman, A. I., & Ter ‐Pogossian, M. (1953). Lymph ‐node concentration of radioactive
colloidal gold following interstitial injection. Cancer, 6(6), 1238-1240.
117. El-Sayed, I. H., Singer, M. I., & Civantos, F. (2005). Sentinel lymph node biopsy in head and
neck cancer. Otolaryngologic Clinics of North America, 38(1), 145-160.
156
Publication
1. Wang, Y., Narayanan, S.R. and Wu, W. Field-Assisted Splitting of Pure Water Based
on Deep-Sub-Debye-Length Nanogap Electrochemical Cells. ACS Nano, 2017. (reported
by Chinese leading internet media)
2. Wang, Y., Stang, J., Yu, M., Tsvetkov, M., Wu, C. C., Qin, X., Chung, E.,
Moghaddam, M. and Wu, W. Microwave Selective Heating Enhancement for Cancer
Hyperthermia Therapy based on Lithographically Defined Micro/Nanoparticles.
Advanced Materials Technologies, 2016. (reported by MaterialsViewsChina)
3. Wang, Y., Yu, M., Stang, J., Chung, E., Moghaddam, M. and Wu, W. Micro-resonator
for Microwave Cancer Therapy. IEEE Journal on Multiscale and Multiphysics
Computational Techniques, 2016.
4. Wang, Y., Liu, H., Li, Y . and Wu, W. Low DC-bias silicon nitride anisotropic etching.
Journal of Vacuum Science & Technology B, 33(6), 06FA01, 2015.
5. Wang, Y., Ma, P., Feng, K., Chen, Z. and Wu, W. Hierarchical
superhydrophobic/hydrophilic substrates based on nanospheres self-assembly onto
micro-pillars. Materials Research Express, 1(4), 045010, 2014.
6. Song, B., Yao, Y ., Groenewald, R.E., Wang, Y ., Liu, H., Wang, Y., Li, Y ., Liu, F.,
Cronin, S.B., Schwartzberg, A.M., Cabrini, S., Haas, S. and Wu, W. Probing Gap
Plasmons Down to Sub-Nanometer Scales Using Collapsible Nano-Fingers. ACS Nano,
2017.
157
7. Yao, Y ., Liu, H., Wang, Y., Li, Y ., Song, B., Wang, R. P., Povinelli, M. L. and Wu, W.
Nanoimprint-defined, large-area meta-surfaces for unidirectional optical transmission
with superior extinction in the visible-to-infrared range. Optics Express, 24(14),
15362-15372, 2016.
8. Liu, H., Yao, Y ., Wang, Y. and Wu, W. Full-color reflective display system based on
high contrast gratings. Journal of Vacuum Science & Technology B, 32(6), 06FE04, 2014.
9. Yao, Y ., Liu, H., Wang, Y., Li, Y ., Song, B., Bratkovsk, A., Wang, S. and Wu, W.
Nanoimprint lithography: an enabling technology for nanophotonics. Applied Physics A,
1-7, 2015. (invited paper)
10. Yao, Y ., Wang, Y., Liu, H., Li, Y ., Song, B. and Wu, W. Line width tuning and
smoothening for periodical grating fabrication in nanoimprint lithography. Applied
Physics A, 1-5, 2015. (invited paper)
11. Qian, C., Ni, C., Yu, W., Wu, W., Mao, H., Wang, Y. and Xu, J. Highly-Ordered, 3D
Petal-Like Array for Surface-Enhanced Raman Scattering. Small, 7(13), 1801-1806, 2011.
12. Li, Y ., Mao, H., Liu, H., Yao, Y ., Wang, Y., Song, B., Chen, Y . and Wu, W.
Stereolithography with variable resolutions using optical filter with high-contrast gratings.
Journal of Vacuum Science & Technology B, 33(6), 06F604, 2015.
13. Sun, M., Qian, C., Wu, W., Yu, W., Wang, Y. and Mao, H. Self-assembly nanoparticle
based tripetaloid structure arrays as surface-enhanced Raman scattering substrates.
Nanotechnology, 23(38), 385303, 2012.
158
Conference Presentation (only first-author presentation listed)
1. Y.F. Wang, S.R. Narayanan and W. Wu, “Field-Driven Splitting of Pure Water based
on Deep-sub-Debye-length Nanogap Cells”, the 61th international conference on electron,
ion, and photon beam technology and nanofabrication (EIPBN 2017), May 30 - June 2,
2017, Orlando, FL.
2. Y.F. Wang, Y .Q. Su, J. Stang, M. Moghaddam and W. Wu, “Microwave
Characterization of Nanocomposite based on Lithographically Defined Nanoparticles”,
the 60th international conference on electron, ion, and photon beam technology and
nanofabrication (EIPBN 2016), May 31 - June 3, 2016, Pittsburgh, PA.
3. Y.F. Wang and W. Wu, “High-efficiency Water Electrolysis based on
Nanoelectrodes”, the 59th international conference on electron, ion, and photon beam
technology and nanofabrication (EIPBN 2015), May 26-29, 2015, San Diego, CA.
4. Y.F. Wang, H. Liu and W. Wu, “Low DC-bias Silicon Nitride Anisotropic Etching”,
the 59th international conference on electron, ion, and photon beam technology and
nanofabrication (EIPBN 2015), May 26-29, 2015, San Diego, CA.
5. Y.F. Wang, M. Yu, J. Stang, M. Moghaddam, Y .R. Li and W. Wu, “Focused
Microwave Cancer Therapy Using Lithographically Defined Nanoparticles”, the 59th
international conference on electron, ion, and photon beam technology and
nanofabrication (EIPBN 2015), May 26-29, 2015, San Diego, CA.
6. Y.F. Wang, H. Liu, Y .H. Yao, S.J. Barcelo, M. Hu, Z.Y . Li and W. Wu, “Nanoimprint
Lithography of 3-D Structures for SERS Sensor”, SERS workshop: Flexible SERS
159
substrates: Challenges and Opportunities, June 25-26, 2014, Washington University, St.
Louis. (Invited)
7. Y.F. Wang, Y . Tian, K.J. Feng, C. Li, D.D. She and W.G. Wu, “Vertical Deposition of
Nanospheres on the Open Sidewalls of Silicon Pillars”, 7th IEEE International
Conference on Nano/Micro Engineered and Molecular Systems (IEEE NEMS’12), Mar.
5-8, 2012, Kyoto, Japan.
8. Y.F. Wang, D.D. She, Y . Tian, C. Li, K.J. Feng and W.G. Wu, “Lotus-like
Hierarchical Structures based on Self-assembly of Nanospheres on Surfaces of Silicon
Cylinders”, 4th International Symposium on Microchemistry and Microsystems
(ISMM’12), June 10-13, 2012, Zhubei, Taiwan.
Patent
1. Yifei Wang, Wei Wu, Mahta Moghaddam, John Stang and Eugene Chung,
Lithographically Defined Nanoparticles for Microwave Absorption, U.S. Patent
15/224,562, 2016. (Filed)
2. Yifei Wang and Wei Wu, Nanoelectrodes for Water Splitting, U.S. Patent 15/605,805,
2017. (Filed)
Abstract (if available)
Abstract
Top-down nanofabrication is a powerful tool to achieve desired nano-devices that can contribute to various types of applications. Besides the traditional semiconductor nanofabrication, an increasing attention has been focused on many other fields, e.g., nano-optics, nano-energy, nano-biomedicine, etc. Sometimes the performance of such nano-devices can be quite different from those macro-devices, and new knowledge and new functions can be generated by studying these nano-devices. In this dissertation, two top-down nanofabrication projects have been discussed. ❧ In the first project (Chapter 1 - 5), we utilized deep-sub-Debye-length nanogap electrochemical cells to achieve efficient electrolysis of pure water and pure methanol solution (without any added electrolyte) at room temperature. Here we have fundamentally broken through the common knowledge that only conductive solution (or solution with strong electrolyte) can be efficiently electrolyzed. A field-assisted effect resulted from overlapped electrical double layers can greatly enhance molecules ionization and mass transport, leading to electron-transfer limited reactions. We have named this process “virtual breakdown mechanism” (which is completely different from traditional mechanisms) that couples the two half-reactions together, greatly reducing the energy losses arising from ion transport. This fundamental discovery has been theoretically discussed in this dissertation and experimentally demonstrated in a group of electrochemical cells with nanogaps between two electrodes down to 37 nm. Many efforts were put onto the improvement of fabrication yield of such deep-sub-Debye-length nanogap electrochemical cells. Based on our nanogap electrochemical cells, the electrolysis current density from pure water can be significantly larger than that from 1 mol/L sodium hydroxide solution, indicating the much better performance of pure water splitting as a potential for on-demand clean hydrogen production. ❧ In the second project (Chapter 6 - 10), lithographically defined nanoparticles with high efficiency of microwave absorption have been used for cancer hyperthermia therapy. Two types of particles have been studied: resonance particles and non-resonance particles. For the LC-resonance particles, finite element method has been used for design and optimization. It was found that the limitation of the micro-resonator is not the requirement of the low resonance frequency but the low Q-value due to insufficient conductivity of natural materials. For the non-resonance particles, disk-shaped magnetic dipoles have been chosen as the high-efficiency microwave-absorbers. The fabrication and collection processes of the particles have been developed. Our measurements, based on both polymer-film nanocomposite setup and particle-suspended hydrogel setup, have demonstrated that the micro/nano-particles can absorb the external electromagnetic field efficiently and greatly enhance the localized heat generation compared to control groups without particles. Further characterization has shown that both the real and imaginary part of the permittivity of our nanocomposite have been enhanced, which is probably the fundamental reason why the microwave absorption efficiency can be improved. ❧ In the first project, Chapter 1 discusses the state-of-art methods of industrial hydrogen generation. Chapter 2 discusses the fundamental difference between macro electrode system and nanogap electrochemical cells (taking water electrolysis as an example). Chapter 3 discusses the fabrication process of our vertical-designed nanogap electrochemical cells. Chapter 4 and 5 discusses the experiment results of pure water electrolysis and pure methanol solution electrolysis, respectively, both based on our nanogap electrochemical cells. ❧ In the second project, Chapter 6 discusses the state-of-art methods of cancer therapy. Chapter 7 discusses the method of LC-circuit resonance particles. Chapter 8 discusses the design and fabrication of the non-resonance particles. Chapter 9 discuss the method to collect / transfer such lithographically defined particles, and uses our disk-shaped magnetic dipoles as an example to show the heating enhancement characterization. Chapter 10 further discusses the fundamental characterization of such particle/matrix nanocomposite mixtures.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Fabrication and application of plasmonic nanostructures: a story of nano-fingers
PDF
Light management in nanostructures: nanowire solar cells and optical epitaxial growth
PDF
Flame synthesis of TiO₂ nanoparticles and its application in dye sensitized solar cells
PDF
Improving antitumor efficacy of chimeric antigen receptor-engineered immune cell therapy with synthetic biology and combination therapy approaches
PDF
High-throughput nanoparticle fabrication and nano-biomembrane interactions
PDF
Semiconductor devices for vacuum electronics, electrochemical reactions, and ultra-low power in-sensor computing
PDF
Plasma emission across metal nanoparticle surfaces and semiconductor -liquid interface characterizations
PDF
Light-matter interactions in engineered microstructures: hybrid opto-thermal devices and infrared thermal emission control
PDF
Nano-engineered devices for display and analog computing
PDF
Hot carriers in bare metals and photocatalytically active defect sites in dielectric/metal structures
PDF
Nanophotonic light management in thin film silicon photovoltaics
PDF
Engineering immunotoxin and viral vectors for cancer therapy
PDF
Metasurfaces in 3D applications: multiscale stereolithography and inverse design of diffractive optical elements for structured light
PDF
Active integrated photonic devices in single crystal LiNbO₃ micro-platelets and a hybrid Si-LiNbO₃ platform
PDF
Optoelectronic, thermoelectric, and photocatalytic properties of low dimensional materials
PDF
GaN power devices with innovative structures and great performance
PDF
Transient pulsed plasma for pollution remediation and energy conversion applications
PDF
Electromagnetic modeling of plasmonic nanostructures
PDF
Printed electronics based on carbon nanotubes and two-dimensional transition metal dichalcogenides
PDF
Optical simulation and development of novel whispering gallery mode microresonators
Asset Metadata
Creator
Wang, Yifei
(author)
Core Title
Nano-fabricated devices in electrochemistry and cancer therapy
School
Viterbi School of Engineering
Degree
Doctor of Philosophy
Degree Program
Electrical Engineering
Publication Date
08/01/2018
Defense Date
12/11/2017
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
cancer hyperthermia therapy,lithographically defined nanoparticles,nanofabrication,nanogap electrochemical cells,OAI-PMH Harvest,pure water electrolysis,virtual breakdown mechanism
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Wu, Wei (
committee chair
), Narayanan, S. R. (
committee member
), Wang, Han (
committee member
)
Creator Email
wangyf.jeffery@gmail.com,yifeiwan@usc.edu
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c40-467556
Unique identifier
UC11268026
Identifier
etd-WangYifei-5987.pdf (filename),usctheses-c40-467556 (legacy record id)
Legacy Identifier
etd-WangYifei-5987.pdf
Dmrecord
467556
Document Type
Dissertation
Rights
Wang, Yifei
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA
Tags
cancer hyperthermia therapy
lithographically defined nanoparticles
nanofabrication
nanogap electrochemical cells
pure water electrolysis
virtual breakdown mechanism