University of Southern California Dissertations and Theses

Parylene C bioMEMS for implantable devices with electrochemical interfaces

(USC Thesis Other)

Parylene C bioMEMS for implantable devices with electrochemical interfaces

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content i

PARYLENE C BIOMEMS FOR IMPLANTABLE DEVICES WITH ELECTROCHEMICAL
INTERFACES

by

Eugene Jisu Yoon

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
(BIOMEDICAL ENGINEERING)

August 2021

Copyright 2021 Eugene Jisu Yoon
ii

DEDICATION
To my parents and grandfather

iii

ACKNOWLEDGMENTS
When I was a small child, visits to the California Science Center on Exposition Park Drive
always excited me. My favorite exhibit was the giant mechatronic human anatomy character
named Tess
1
. I vividly remember how bright LED lights highlighted blood flow patterns in her
veins while a TV screen showed a first-person view of her playing soccer to illustrate homeostasis.
Who would have known that I would be pursuing my biomedical engineering PhD at USC just
down the block ~20 years later?
First and foremost, I want to thank and acknowledge my parents for their unconditional
love and support. I also want to apologize for not expressing my love to them more often due to
my reticent personality. It even feels awkward writing these words in English to express my
gratitude since the language barrier made deep communication challenging when I was younger
and less fluent in Korean. It is hard to fathom the magnitude of the trials and tribulations you
must have faced in this new world as first-generation immigrants… but you always supported
me and my dreams. I still remember the “chemical reaction!” science shows at the local park that
you took me to, or stories about pioneering artificial heart valves in Seoul National University.
These kinds of things stirred my subconscious. You have always made me feel like I could do
anything since I knew you would love me and cheer me on – no matter what. Thank you.
My grandfather was a healer. Formerly a pharmacist in the South Korean Navy, he took
root in Koreatown and South Central Los Angeles later on. My family would always seek his

1
https://www.roadsideamerica.com/story/11176
iv

advice whenever any of us got sick. The US healthcare system is challenging to navigate today,
but it was exacerbated back then from the language barrier again, so we always just asked
grandpa ( 할아 버지) what to do. I think that beautiful notion of leveraging knowledge to heal
sickness really moved me. I wish he could see me graduate from JHU or USC to see how that
vision breathed life into my career.
My research trajectory first took off as an undergraduate under the laboratory of Dr.
David Gracias. I was clueless and knew nothing, but he took me in. I am so grateful. Two graduate
students back then but now Dr. Changkyu Yoon and Dr. Qianru Jin mainly mentored and trained
me enough to successfully write a proposal to do research abroad at Imec in Leuven, Belgium.
There, Dr. Dries Braeken and now Dr. Jordi Cools welcomed me, trained me, and helped me grow
as a researcher. I am also deeply indebted to Dr. Alfredo Celedon, founder and CEO of Scanogen
Inc. because he was the primary catalyst for my decision to pursue a PhD. Working at his JHU-
affiliated biotech startup company as employee #3 truly sparked in me a passion for bridging
science and business with entrepreneurship as the main vehicle. He inspired me to be a future
founder myself. My strategic decision was to first stack technical and business skills first.
Searching for labs and PhD programs was challenging, but Dr. Ellis Meng was really a
beacon of light. I could really tell that she genuinely wanted to help people and patients. The
system called academia today does not necessarily incentivize that, but I could sense Ellis’s deep
passion from my PhD preview interview – especially through efforts such as Fluid Synchrony
and Senseer. Thank you so much for your guidance as well as for giving me the autonomy to
spend time on Marshall Business school classes, HTE, various pitch competitions, FDA regulatory
v

and insurance classes, etc… Throughout my time under Ellis’s wing, I met and worked with so
many bright and talented people. Dr. Kee Scholten, I really admire your encyclopedic knowledge
and straight-to-the-point approach to discussions. Dr. Lawrence Yu, thanks so much for passing
on µBPT knowledge, life tips, and all the hangouts – especially the biking group! Dr. Angelica
Cobo, I think you actually taught me the most about how believe in myself and how to express
that confidently. I never explicitly told you this because I simply observed it from your
mannerisms but thank you so much. Dr. Ahuva Weltman, thank you for your tutelage during my
PhD rotation and sorry for that one time where I was too clueless to recognize that I presented
empirically wrong research results at a lab meeting which contradicted your PhD defense
progress, oops! Dr. Alex Baldwin – you were definitely my closest friend in lab. Thanks so much
for all the tips in bouldering, life, startups, and just being so real while being fun too. I hope we
can climb again more as the pandemic settles down. Dr. Jessica Lizbeth-Ortigoza, thank you so
much for your warmth and research friendliness. James Yoo and Trevor Hudson, I am so glad
that we all started in the same cohort and could work alongside each other. Thank you for also
keeping it real, for the coffee recommendations, and being a great travel partner in Washington
D.C. Chris Larson, I really admire your patience and integrity with research. I think you definitely
walk the walk when it comes to these types of things. Xuechen Wang, I am so impressed by your
growth and intensity with research! Ping Hu, working with you on electrochemical projects has
shown me your curiosity and resilience, keep it up! Brianna Thielen, I also deeply respect and
admire how your industry experience sharpened your research prowess and technical
communication skills. It really shows during lab meetings. Garrett Soler and Nick Barrera, as the
vi

newest Mengsters (especially Nick), welcome to the group! I can already see that you are great
fits and I cannot wait to see what the future has in store for you!
Thank you also to many other collaborators and non-graduate students. Dr. Masato
Suzuki especially comes to mind. If you ever read this, thank you so much for sharing your
knowledge and being patient with me. I was such an inexperienced researcher when you came
as a visiting scholar. I hope your time here was a good one. Also, a very deep thank you goes to
Janeline Wong. Frankly, I think I was a terrible manager, especially in the beginning, but you
stuck with me throughout your entire undergraduate career and I am so grateful for your brilliant
contributions and saintly patience. I am also very grateful to other undergraduate researchers –
Ewina, Emma, Izzy, and Tanya. Also thank you so much to Beomseo Koo and Muru Zhou of the
University of Michigan for your insights and collaboration as well as NIST personnel like Dr.
Nate Orloff and Dr. Jim Booth. I am especially grateful to Dr. Angela Stelson – thank you for your
unending patience and positive attitude throughout our meetings!
Lastly, much of my research was funded through the National Science Foundation (NSF)
so I am also extremely grateful for the fiscal support.

vii

TABLE OF CONTENTS
Dedication ................................................................................................................................................. ii
Acknowledgments .................................................................................................................................. iii
List of Tables ............................................................................................................................................. ix
List of Figures ............................................................................................................................................ x
ABSTRACT .............................................................................................................................................. xx
................................................................................................................................................ 1
Introduction to Parylene-Based BioMEMS Devices with Electrochemical Interfaces ................ 1
Overview on MEMS ..................................................................................................................... 1
BioMEMS Applications ................................................................................................................ 2
Introduction to Polymers and Parylene C ................................................................................. 4
Electrochemical Interface Devices .............................................................................................. 6
Objectives ....................................................................................................................................... 8
References .................................................................................................................................... 10
.............................................................................................................................................. 16
Rat Retinal Array (RRA) to Aid Chronic Epiretinal Stimulation Research for Artificial Vision
.................................................................................................................................................................... 16
Background .................................................................................................................................. 16
Approach and System Overview .............................................................................................. 22
Design Iterations ......................................................................................................................... 28
Experimental Methods ............................................................................................................... 33
Results ........................................................................................................................................... 45
Summary and Future Directions .............................................................................................. 60
References .................................................................................................................................... 63

viii

.............................................................................................................................................. 69
Microbubble-based Pressure Transducer Development for Intracranial Pressure
Measurement ........................................................................................................................................... 69
Background .................................................................................................................................. 69
Approach and Goals ................................................................................................................... 83
Experimental Methods ............................................................................................................... 84
On Bubble Generation ................................................................................................................ 93
On Bubble Behavior .................................................................................................................. 112
On Bubble Measurement ......................................................................................................... 124
Summary and Future Directions ............................................................................................ 144
References .................................................................................................................................. 149
............................................................................................................................................ 159
Radio-Frequency Dielectric Spectroscopy on Parylene C via Microwave Microfluidics ....... 159
Introduction ............................................................................................................................... 159
Primer on Microwave Theory ................................................................................................. 166
Methods and Materials............................................................................................................. 188
Results ......................................................................................................................................... 201
Discussion and Future Directions ........................................................................................... 209
References .................................................................................................................................. 212
Chapter 5 ................................................................................................................................................. 217
Conclusion .............................................................................................................................................. 217
Appendices ............................................................................................................................................. 220
ix

LIST OF TABLES
Table 2.3-1. RRA exposed electrode diameters (µm) .................................................................. 29
Table 4.1-1. Medical conditions affecting various physiological pressure systems .................... 72
Table 3.4-1. Main results for three bubble reproducibility studies conducted from 2017 - 2019 98
Table 4.4-2. Size designations of tested counter electrodes ....................................................... 100
Table 3.5-1. Table of various bottleneck:channel ratio devices fabricated for testing .............. 113
Table 3.6-1. Color legend for universal pH indicator ................................................................ 137
Table 4.1-1. Comparison of EIS and NIST measurement techniques ....................................... 161
Table 4.1-2. Table of Parylene-based works employing high frequency signals ...................... 162
Table 4.3-1. Apportionment of chip types across wafer ............................................................ 189
Table 4.3-2. Measurement plans for test conditions of short-term wetted Parylene experiments
......................................................................................................................................... 197

x

LIST OF FIGURES
Figure 1.3-1. Poly(p-xylelene) monomer units and several Parylene variants .............................. 5
Figure 2.1-1. Anatomy of the ocular orbit and key components involved in the visual transduction
pathway ............................................................................................................................. 17
Figure 2.1-2. Normal vision compared to loss of vison from the point of view of a patient with
retinitis pigmentosa ........................................................................................................... 19
Figure 2.1-3. Illustration of the Argus II system and its main components. An electrode array is
implanted onto the retina. The visual feed from a camera fitted onto spectacles worn by the
patient is processed and then transmitted via wireless telemetry ..................................... 20
Figure 2.2-1. Schematic illustration of rat retinal array (RRA) system and main components ... 23
Figure 2.2-2. The approach for this work employed several iterations of cycling between
fabrication, implantation, and design. Each cycle yielded key insights and knowledge to
inform the next iteration.................................................................................................... 24
Figure 2.2-3. Microfabrication process flow for all rat retinal array devices .............................. 27
Figure 2.3-1. Illustration of the RRA stimulation region ............................................................. 29
Figure 2.3-2. Design schematic for RRA version 1 ..................................................................... 30
Figure 2.3-3. Design schematic for RRA version 2 ..................................................................... 31
Figure 2.3-4. Design schematic for RRA version 3 ..................................................................... 33
Figure 2.4-1. Excessive pressure applied during thermoforming created cracks at device contact
pads ................................................................................................................................... 34
Figure 2.4-2. Demonstration of first and second bend via thermoforming in RRA version 1 and 2
........................................................................................................................................... 35
xi

Figure 2.4-3. (a) Autodesk Inventor CAD illustration of PTFE mold for thermoforming RRA
version 3 devices. (b) The first bend to match retinal curvature could be obtained through
a slot and groove and (c) proper registration between device and mold was ensured through
registration pin holes ......................................................................................................... 36
Figure 2.4-4. Photograph of PTFE mold component for thermoforming RRA version 3 devices
........................................................................................................................................... 36
Figure 2.4-5. Optical micrographs of a representative device electroplated in PtIr .................... 38
Figure 2.4-6. Schematic illustration of packaging scheme between retinal array to wireless
headstage ........................................................................................................................... 39
Figure 2.4-7. (a) Bottom face and (b) top face of PCB adapter assembly ................................... 39
Figure 2.4-8. (a) Custom-built acrylic fixture CAD model and (b) photograph for crosstalk testing
........................................................................................................................................... 41
Figure 2.5-1. Thermoformed RRA version 2 device. Both version 1 and 2 employed a technique
where the first bend matched retinal curvature and the second bend ensured that the array
routes back towards the cranium ....................................................................................... 45
Figure 2.5-2. Thermoformed RRA version 3 device. The “candy cane” hook design obviated the
need for the second bend present in RRA version 1 and 2 ............................................... 46
Figure 2.5-3. Representative electrode delamination for unannealed sputtered Pt metal during
crosstalk testing ................................................................................................................. 46
Figure 2.5-4. Sample crosstalk measurements obtained at t = 0 hr of soaking ............................ 47
Figure 2.5-5. Crosstalk test results for electrodes pairs which were completely or partially
insulated ............................................................................................................................ 48
Figure 2.5-6. EIS results with (a) magnitude and (B) phase each electrode in an unannealed RRA
version 2 device (1 × phosphate buffered saline, 25 °C) .................................................. 49
xii

Figure 2.5-7. Electrochemical impedance magnitude (a) and phase (b) at different stages of
processing on RRA version 2 devices (1 × PBS, 25 °C) .................................................. 50
Figure 2.5-8. (a) Cyclic voltammetry at different stages of processing on RRA version 2 devices.
(b) Magnified and rescaled view shows features corresponding to hydrogen
adsorption/desorption and oxide redox reactions for annealed and thermoformed cases (1
× PBS, 25 °C) .................................................................................................................... 51
Figure 2.5-9. Pulse test results at different stages of processing of RRA version 2 devices. PtIr
reduced voltage transients compared to bare Pt (1 × PBS, 25 °C). .................................. 52
Figure 2.5-10. Representative photograph of enucleated rat eye undergoing sham device fitting
tests ................................................................................................................................... 52
Figure 2.5-11. Representative photograph of sham device implantation into rat cadaver ........... 53
Figure 2.5-12. Optical coherence tomagraphy image of sham array tip cross-section and retina (4
weeks post-surgery). ......................................................................................................... 54
Figure 2.5-13. (a) The eye was proptosed to enable more access to the sclera then (b) the retinal
array was sutured on. (c) Connection scheme from implanted array to potentiostat ........ 55
Figure 2.5-14. EIS before and after surgical implantation. Three out of eight electrodes were
functional prior to surgery and remained functional after implantation. .......................... 57
Figure 2.5-15. Surgical view of eye during implantation of RRA version 3 device (a) after device
suturing and (b) after scleral blanketing of the device ...................................................... 58
Figure 2.5-16. Surgical view of cranium during implantation of RRA version 3 device and
associated PCB adapter ..................................................................................................... 59
Figure 2.5-17. Live rat with PCB assembly in post-operative setting ......................................... 60
Figure 3.1-1. Illustration of several major physiological pressure systems ................................. 70
xiii

Figure 3.3-1. (a) Micrograph and (b) schematic illustrations of device where Parylene (blue) is
opaque or translucent in order to highlight hollow and/or embedded components .......... 85
Figure 3.3-2. Schematic illustration of ideal µBPT operating sequence ..................................... 86
Figure 3.3-3. Schematic illustration of fluid capillaries (a) dotted line represents cross-section of
interest (b) cross-sectional view (c) stripped view with fluid capillaries only ................. 87
Figure 3.3-4. Microfabrication process flow for the µBPT. The A-A’ cross section of interest runs
along the longitudinal length of the device ....................................................................... 88
Figure 3.3-5. Photomask illustrations and SEM images of (a) early and (b) current fluid etch port
regions ............................................................................................................................... 90
Figure 3.3-6. Custom fixture used to submerge devices in 1× PBS inside a cavity defined by the
silicone gasket and acrylic components ............................................................................ 91
Figure 3.3-7. Raw images of bubble growth were captured at 10 fps and processed with a custom
MATLAB algorithm to extract the bubble volume versus time ....................................... 93
Figure 3.4-1. Electrochemical potential diagrams to illustrate the effect of sizing on electrode
polarization. (a) At equilibrium, all points along the electrochemical circuit are at
approximately the same electrochemical potential. (b) If the electrodes in use are similarly
sized (ie. interdigitated electrodes or identical electrode pads) then polarization will occur
in equal and opposite directions. (c) If the counter electrode is significantly larger than the
working electrode, polarization will be predominantly localized to the electrode-to-
electrolyte potential difference at the working electrode interface. .................................. 94
Figure 3.4-2. Electrochemical circuit diagram to illustrate connection schemes between electrodes
of interest and equipment .................................................................................................. 99
Figure 3.4-3. Representative voltage sweeping results to determine gassing limits as a function of
counter electrode size. (a) The total electrochemical cell potential was swept at ±100
mV/second while simultaneously observing gas evolution in real-time. (b) The electrode-
xiv

to-electrode potential for both the working electrode and counter electrode is plotted for
instances where the counter electrode size was allowed to vary. Device type = G. ....... 101
Figure 3.4-4. (a) Cathodic pulses (-0.6 µA for 6 seconds) were applied to generate bubbles and
(b) the various potential differences normalized to their starting values were measured as a
function of CE size (n = 8 pulses per CE size dataset; shaded regions denote standard
deviation) (c) Multiplying the average total WE-CE potential difference per pulse by the
current amplitude and duration allowed for computation of the average energy required per
bubble. Device type = G. ................................................................................................ 102
Figure 3.4-5. Bubble volume versus time for various pulse durations (10, 15, 20, 25 s).Device type
= left-most A ................................................................................................................... 104
Figure 3.4-6. Theoretically expected and empirically measured max bubble volumes for various
pulse durations. Device type = left-most A .................................................................... 105
Figure 3.4-7. Freeze-frame images of bubbles generated via varying pulse durations at their
approximate maximum volumes. Device type = left-most A. Timepoints correspond to
those in Figure 3.4-5 ...................................................................................................... 106
Figure 3.4-8. Impedance magnitude versus time of bubbles generated via varying pulse durations.
Device type = left-most A. .............................................................................................. 106
Figure 3.4-9. Bubble relinquishes contact with Parylene wall during a detachment event at t ≈ 120
s. Device type = left-most A. .......................................................................................... 108
Figure 3.4-10. Bubble volume versus time for experiments using a large counter electrode with
positive or negative applied current pulse. Thick lines represent averaged data and thin
lines represent individual trials. Device type = G. .......................................................... 109
Figure 3.4-11. The amount of time required for generated bubbles to completely dissolve (ie.
bubble lifetime) significantly varied by the usage of positive or negative applied current
(36 ± 6.1 min versus 10.3 ± 0.7 min; respectively; mean ± standard deviation) when a large
CE was used. Device type = G. ...................................................................................... 109
xv

Figure 3.5-1. Schematic illustration of bottleneck width and channel width ............................ 113
Figure 3.5-2. Unorthodox bubble dissolution patterns adversely affected the measured impedance.
Device type = C10. ......................................................................................................... 114
Figure 3.5-3. Schematic illustration of constriction valves ....................................................... 115
Figure 3.5-4. Representative sequence of constriction valves effectively controlling bubble growth
behavior. Device type = G. ............................................................................................. 117
Figure 3.5-5. Volume versus time for 83 bubbles generated in longevity experiment. Device type
= middle A. ..................................................................................................................... 119
Figure 3.5-6. Maximum volume versus trial number in device longevity experiment. Device type
= middle A. ..................................................................................................................... 119
Figure 3.5-7. Impedance versus time for 83 bubbles generated in longevity experiment. Device
type = middle A. ............................................................................................................. 119
Figure 3.5-8. Trial 1 versus trial 83 freeze-frame images during maximum bubble volume. Device
type = middle A. ............................................................................................................. 120
Figure 3.5-9. (Left) Top-down confocal microscopy image with (right) A-A’ and B-B’ cross-
section views. Device type = I7. ..................................................................................... 121
Figure 3.6-1. Randles circuit elements to model sensing electrodes E1 and E4 ........................ 125
Figure 3.6-2. Idealized plot to illustrate theoretical relationship between impedance and bubble
length............................................................................................................................... 126
Figure 3.6-3. Expected versus measured EIS for µBPT sensing electrode pair. Device type =
middle A.......................................................................................................................... 127
Figure 3.6-4. Impedance versus time for a 60s baseline measurement with varying signal levels.
Device type = middle A. ................................................................................................. 129
xvi

Figure 3.6-5. Average current draw versus signal level to confirm device operation with linear
electrochemical regimes. Device type = middle A. ........................................................ 129
Figure 3.6-6. 3σ signal noise versus signal level in baseline measurements. Device type = middle
A. ..................................................................................................................................... 130
Figure 3.6-7. Three different measurement setups to quantify signal level noise. Device type =
middle A.......................................................................................................................... 131
Figure 3.6-8. Measured bubble length overlapped with theoretically computed impedance
magnitude ........................................................................................................................ 132
Figure 3.6-9. Measured impedance versus theoretically computed impedance. Device type =
middle A.......................................................................................................................... 132
Figure 3.6-10. Local minimum in impedance at the moment of complete bubble dissolution in a
representative trial for the residence time pressure transduction technique. Device type =
Z80. ................................................................................................................................. 134
Figure 3.6-11. Residence times of 61 bubbles illustrated against their tested pressures. Device type
= Z80. .............................................................................................................................. 135
Figure 3.6-12. Residence time versus pressure. Device type = Z80. ......................................... 136
Figure 3.6-13. Colorimetric change from pH differences caused by applying 3 V between Pt
counter wires ................................................................................................................... 137
Figure 3.6-14. Voltage between very large microelectrodes was ramped up to 3V while the
impedance between µBPT sensing electrodes was measured simultaneously. Device type
= R10. .............................................................................................................................. 138
Figure 3.6-15. pH indicator revealed growing diffusion fronts at various timepoints which affected
measured impedance at neighboring µBPT devices. Device type = R10. ...................... 139
Figure 3.6-16. One cycle of the pressure ladder waveform which was applied continuously
through the experiment. .................................................................................................. 140
xvii

Figure 3.6-17. Global view of impedance magnitude versus time for a pressure-stepped
experiment. Inset revealed that 1 mmHg pressure steps could be resolved by the measured
impedance. Device type = right-most A. ........................................................................ 141
Figure 3.6-18. Impedance versus pressure of the half cycle data in Figure 3.6-17. Device type =
Right-most A. .................................................................................................................. 141
Figure 4.1-1. Plot of highest frequencies used in several Parylene-based work ........................ 163
Figure 4.1-2. Illustration of EIS potentiostat where frequency signals may be assumed to present
uniform voltage values at all points in space per time .................................................... 164
Figure 4.2-1. Simplified workflow diagram for NIST microwave measurement techniques.... 167
Figure 4.2-2. Schematic illustration of a typical microwave measurement setup ..................... 168
Figure 4.2-3. Photograph of microwave measurement setup with microfluidic fitting laid on top
of CPW............................................................................................................................ 168
Figure 4.2-4. A signal from port 1 (a1) is sent into the measurement system ............................ 169
Figure 4.2-5. A signal from port 1 (a1) is sent into the measurement system and the transmitted
signal at port 2 (b2) is measured ...................................................................................... 169
Figure 4.2-6. A signal from port 1 (a1) is sent into the measurement system and the reflected signal
at port 1 (b1) is measured ................................................................................................ 170
Figure 4.2-7. Illustration of all input and output signals during microwave measurement operation
......................................................................................................................................... 171
Figure 4.2-8. Example screenshot of S parameter measurements from a .csv file .................... 172
Figure 4.2-9. Illustration of measurement components corresponding to measurement matrices
......................................................................................................................................... 174
xviii

Figure 4.2-10. Illustration of all input and output signals during microwave measurement
operation with additional DUT and error box components ............................................ 174
Figure 4.2-11. Illustration of the thru, reflect, and line standards in TRL measurements ......... 175
Figure 4.2-12. Illustration of the line standard with its effective length .................................... 176
Figure 4.2-13. Signal amplitude and phase change along the length of a line standard ............ 176
Figure 4.2-14. Signal amplitude and phase change along a line visualized by sample propagation
constant values ................................................................................................................ 177
Figure 4.2-15. Block diagram of a line measurement ................................................................ 177
Figure 4.2-16. Illustration of a thru standard, where the effective length defined by the reference
planes is zero ................................................................................................................... 178
Figure 4.2-17. Block diagram of a thru measurement ............................................................... 178
Figure 4.2-18. Series Resistor and Series Capacitor illustrations .............................................. 181
Figure 4.2-19. Simplified example of CPW with air and Parylene segments ........................... 183
Figure 4.2-20. A transmission line segment of differential length Δx ....................................... 184
Figure 4.2-21. Circuit model elements of previous work [192] on metal-electrolyte CPW systems
......................................................................................................................................... 187
Figure 4.3-1. Illustration of reference chip and its calibration structures .................................. 188
Figure 4.3-2. Illustration of a test chip ....................................................................................... 189
Figure 4.3-3. Chip apportionment on 100 mm diameter wafer ................................................. 190
Figure 4.3-4. Cross section of CPW to illustrate microfabrication process flow ...................... 192
Figure 4.3-5. Autodesk Inventor model of microfluidic master mold ....................................... 192
xix

Figure 4.3-6. Photograph of 3D printed microfluidic master mold ........................................... 193
Figure 4.3-7. Illustration of microfluidic structures laid on top of test chip CPWs................... 193
Figure 4.3-8. Photograph of microfluidic structures laid on top of test chip CPWs .................. 194
Figure 4.3-9. Molecular representation of Parylene before and after annealing ........................ 195
Figure 4.3-10. The polar Parylene C monomer features a chlorine atom covalently bonded to the
benzene ring .................................................................................................................... 195
Figure 4.3-11. Screenshot of the RadiCal MATLAB GUI ........................................................ 198
Figure 4.3-12. A sample MATLAB struct output from RadiCal ............................................... 199
Figure 4.4-1. High-resolution stitched micrograph of a No-Parylene reference chip ................ 202
Figure 4.4-2. The geometric model of CPW structures created in ANSYS Q3D ...................... 203
Figure 4.4-3. Physical dimensions inputted into Q3D models .................................................. 204
Figure 4.4-4. Material assignment in Q3D models .................................................................... 204
Figure 4.4-5. Q3D simulations revealed that the electric field generated during S parameter
measurement is confined predominantly to the Parylene C film .................................... 205
Figure 4.4-6. Permittivity to capacitance per unit length mapping function for varying Parylene
film thicknesses ............................................................................................................... 205
Figure 4.4-7. Resistance per unit length of simulated No Parylene and measured Yes Parylene
(unannealed) chips .......................................................................................................... 206
Figure 4.4-8. Resistance per unit length (MΩ/m) and resistivity of the CPW before and after
annealing ......................................................................................................................... 207
Figure 4.4-9. Capacitance per unit length (pF/m) and permittivity of Parylene before and after
thermal annealing ............................................................................................................ 209
xx

Whereas silicon microelectromechanical systems (MEMS) have already revolutionized the
world through various inventions, immense potential through polymer-based MEMS remains to
be fully unlocked. In particular, Parylene C is a strong candidate for biomedical innovation due to
its biocompatibility and amenability to micromachining techniques. Parylene C, herein referred to
as Parylene, may insulate thin-film metals which act as current-carrying interfaces for
electrochemical modalities in in vivo environments. This allows for the design and fabrication of
unique implantable microdevices which may serve to improve patient care or assist in basic
research.
This work presents three projects harnessing the utility of Parylene-based MEMS with
electrochemical interfaces. Chapter 1 provides an overview of fundamental considerations for
Parylene and electrochemistry to prime the reader for the ensuing chapters. Chapter 2 describes
the development of a Parylene-based novel implantable microelectrode array designed for chronic
stimulation of the retina in live rats. In Chapter 3, microbubble-based pressure transducers
constructed from Parylene microchannels are thoroughly investigated to yield improvements in
key performance specifications. Radio frequency dielectric spectroscopy techniques from the
National Institute of Standards and Technology (NIST) are used for the first time in the biomedical
context of characterizing Parylene and thin-film platinum in chapter 4. Chapter 5 provides a
conclusion and forward-looking remarks.
Although many challenges still remain, the potential for healthcare improvement continues
to drive future progress in this field. It is therefore the hope of the author that the research contained
herein may inform and aid the advancement of novel MEMS biomedical devices.
ABSTRACT
1

Overview on MEMS
The field of microelectromechanical systems (MEMS) may be traced back several decades
to the invention of the transistor in 1947 at Bell Labs [1]. Microfabrication techniques and
processes in MEMS have enabled an explosion of technology that harnessed unique advantages
afforded by its size scale, batch processing methods, versatility, and electromechanical operating
modalities [2]. Many everyday items today rely heavily on MEMS fabrication techniques –
perhaps most notably the integrated circuit (IC). The exponential growth of computing power
across the years was afforded by the ability to fit ever-smaller components into IC chips, as
famously predicted by Gordon Moore [3], and has paved the way for today’s digital era. Many
other key inventions trace their lineage to MEMS research such as the pressure sensors in cars to
detect tire deflation [4], gyroscopes in smartphones to help adjust vertical or horizontal orientation
[5], inkjet nozzles for printers [6], accelerometers in space-bound rockets such as the Falcon Heavy
[7], and many more. The very keyboard being used to type this thesis relies on MEMS-based
capacitor components to log each keystroke. MEMS has clearly demonstrated tremendous
potential and progress since its inception.

INTRODUCTION TO PARYLENE-BASED BIOMEMS DEVICES
WITH ELECTROCHEMICAL INTERFACES
2

MEMS device construction largely consists of additive processes and subtractive processes
on planar substrates called wafers [8]. Materials can be added or removed in the z-direction with
features defined in the x-y direction through subsequent layers to generate structures well-suited
for a myriad of applications. Examples of additive processes include chemical vapor deposition,
electron beam metal evaporation, sputtering, thermal evaporation, physical vapor deposition,
thermal oxidation, and more. Photolithography may be considered an additive process but warrants
special attention due to its ability pattern extremely small (µm or nm) features. These features
effectively serve as a stencil lying on the wafer or the wafer’s surface components. Subsequent
additive materials may be applied on this stencil such that some material selectively deposits in
the stencil cavities whereas others are deposited on the stencil itself. Selective remove of the stencil
material, which is termed photoresist, may effectively allow for custom-patterned deposition of
the additive material. Conversely, this process may be applied with subtractive processes to
achieve selective etching patterns. Patterning photoresist into structures known as etch masks
allows for selective protection of regions from subtractive processes while allowing others to be
eliminated. Clever design and evolution of these additive and subtractive processes in layer-by-
layer steps have enabled the invention of aforementioned ubiquitous devices.
BioMEMS Applications
1.2.1 General BioMEMS
Although MEMS is traditionally associated with the semiconductor industry, VLSI, and
computer-related systems, biologically-inspired applications have enjoyed benefits of MEMS
ingenuity in recent decades [9]. One example is the area of MEMS-based biosensors. Here,
advancements in micromachining allow for transduction of a biological event happening on a
3

MEMS-based active component. The event may be the binding of biologically specific recognition
molecules corresponding to viruses, single nucleotide polymorphisms, bacteria, insulin, and more
[10]. The transduced bio-signal can therefore assist in the diagnosis of disease or other conditions.
These systems types of systems have been termed micro total analysis (µTAS) [11] by some
researchers and have also largely experienced integration with microfluidic [12] components to
assist in the transport of small sample fluid volumes. Additionally, MEMS-based components
using soft-lithography techniques [13] have recently been employed to serve as a scaffolding test
bed for biological cells to effectively mimic organ systems [14]. These systems, coined as Lab-on-
a-chip (LOC) [15], have shown promising potential in serving as an intermediate proxy for costly
and time-consuming animal models in the context of drug development studies. A high throughput
and low-cost alternative to animal testing may be offered by MEMS-based LOC. MEMS-
components have also seen use in medical wearables [16], drug delivery [17], therapeutics [18],
and more.
1.2.2 Implantable BioMEMS
BioMEMS devices designed for chronic use inside the human body are especially
noteworthy. A unique set of design challenges exists for implantable bioMEMS due to the in vivo
environment. Biocompatibility, long-term resistance against water and ion-induced corrosion,
insulation to prevent undesired short-circuiting, and many other issues must be solved in order to
ensure that the safety and efficacy of such devices outweigh the risks of use [19]. As such, certain
MEMS-based implantable devices have the potential to save lives or dramatically improve patients’
quality of lives. The CardioMEMS device [20] is an example where cardiac pressure of a difficult-
to-reach region can be wirelessly monitored and transmitted to provide early warnings of potential
aortic rupture. Without this detection and clinical intervention, patients could suffer fatal heart
4

attacks. The small size scale of MEMS allows for the CardioMEMS device to be implanted in the
difficult-to-reach region of the heart, and electromechanical modalities allow for pressure
transduction and wireless signal transmission. Another implantable MEMS system is the Utah
Array [21]. This device is a MEMS-based multielectrode array with probe shanks designed to
penetrate brain matter. The device can record neural activity which can then be decoded to perform
tasks such as typing letters in an email or controlling robotic limbs. For patients suffering from
quadriplegia, results from the Utah array offer an uplifting avenue for interacting with the physical
world again.
Introduction to Polymers and Parylene C
1.3.1 Polymer-based BioMEMS
Although materials such as ceramics, glasses, and metals may be used in MEMS, the vast
majority of such devices are based on silicon or silicon variants [22]. CardioMEMS features two
parallel plates of fused silica and Utah Array shanks are made from silicon. Although such
materials may be permissible in limited implanted contexts, polymer-based MEMS may be much
better-suited for the in vivo environment. Unlike silicon, polymers may be biocompatible, flexible,
and mechanically softer [23]. Recent efforts have increased the amenability of polymers to
micromachining techniques, leading to developments such as the FDA-approved Argus II retinal
prosthesis from Second Sight Medical Products Inc [24]. The implantable MEMS device features
60 microelectrodes arrayed onto silicone polymer substrates to deliver controlled pulses of
electricity to retinal ganglion cells to induce artificial patients suffering from certain forms of
blindness. Neuralink Corporation, founded by Elon Musk and others, is presently developing
polyimide-based MEMS neural implants similar to those of the Utah Array. The main difference
5

is the use of implantable threads composed of polyimide polymer and gold electrodes [25] as
opposed to silicon. These threads are much closer in mechanical stiffness to brain tissue than
silicon shanks. Silicon shanks therefore may crack from cranial micromotions and also suffer from
harsher immune responses [26]. Many other polymer-based implantable bioMEMS exist,
including but not limited to peripheral nerve interfaces [27], glaucoma drainage valves [28], strain
sensors [29], drug pumps [30], artificial corneas [31], and more.
1.3.2 Poly(p-xylylene) and Parylene C
Parylene, also known as Parylene N, is the trade name for a semicrystalline class of
polymer whose chemical name is poly(p-xylelene). The monomer structure is composed of an
aromatic ring monomer with attached methylene groups. Parylene variants also exist and feature
additional atoms such as fluorine or chlorine attached to the ring as shown in Figure 1.3-1.
Although original discovered in the 1940s by Michael Mojzesz Szwarc [32], Parylene became
commercially available through the Union Carbide corporation in the 1960s after the Gorham
process was developed and enabled higher yield. The technique, developed by William Gorham
[33], employed chemical vapor deposition from a stable dimer starting material.

Figure 1.3-1. Poly(p-xylelene) monomer units and several Parylene variants

6

Parylene C is particularly noteworthy in the context of bioMEMS due to its
biocompatibility. The United States Pharmacopeia (USP) has graded Parylene C with the ISO-
10993 class VI rating – the highest biocompatibility rating for plastics. Parylene C also has low
cytotoxicity, good resistance against hydrolytic degradation, and has been used in biomedical
implants coatings for decades [34] [35] [36]. Additionally, Parylene C features the previously
described advantages of polymers such as its flexibility, mechanical softness, and is also a good
water and gas barrier [37]. Most importantly, Parylene C as well as other Parylene variants are
amenable to MEMS micromachining techniques [38]. The Gorham chemical vapor deposition
process may be applied to silicon carrier wafers to obtain pinhole-free and uniform thin films of
Parylene C. Metal features may then be patterned onto the Parylene to define wires, electrodes, or
device contact pads for packaging. Another layer of Parylene may be deposited and then selective
etching can yield various device geometries well-suited for a variety of biomedical implant
applications. These types of Parylene C-metal-Parylene C devices have been largely developed at
the Biomedical Microsystems Laboratory at the University of Southern California with examples
ranging from penetrating neural probes for the hippocampus [39], peripheral nerve cuff devices
[40], thermal flow [41] and temperature sensors [42], kirigami stretch sensors [43], and more.

Electrochemical Interface Devices
Many of these implantable Parylene C-metal-Parylene C devices feature microelectrodes
which may interact with adjacent tissue or fluids in vivo. This creates an interface between the
electrode and electrolyte. Here, the scientific branch of electrochemistry may describe the
underlying phenomena and fundamental operating mechanisms of the device [44]. In nonfaradaic
electrochemical regimes, electrons are not transferred across the interface. Instead, a capacitive
7

mechanism is involved where rapid accumulation of electrons at the metal due to an applied
potential may generate an electric field to induce movement of nearby ionic species. This
mechanism is precisely how certain neural interface stimulation devices induce voltage-gated ion
channels of neurons to depolarize and trigger action potentials [45].
Faradaic electrochemistry involves the explicit transfer of charge in redox reactions.
Whereas a simple chemical reaction may be represented by A + B → C, electrochemical reactions
can be imagined by substituting the reactant A with an electron such that e
-
+ B → C [46]. One
challenging component in understanding and harnessing electrochemistry stems from the distinct
yet overlapping time scales feasible through electronics (DC to GHz signals) compared against the
relevant kinetics, thermodynamics, and mass transfer of chemistry [47]. This treatment of time
scales allows for the distinction of reversible and nonreversible faradaic electrochemistry. Certain
neural interfaces, often via electrode-coatings, employ reversible faradaic reactions where charge
transfer may induce action potentials but the generated redox species are re-converted to their
starting forms before they have the chance to diffuse towards surrounding tissue [48]. Conversely,
irreversible faradaic electrochemical reactions allow for mass transfer to outpace kinetics such that
redox species readily diffuse away from the electrode. Although this mechanism is generally
avoided in neural interfaces, other devices may intentionally harness the phenomena for a variety
of means [49]–[51]. If AC signals are employed, various physical phenomena may be present or
absent at differing frequency ranges. An interesting and somewhat unintuitive example is double
layer charging – where electrons and ions may accumulate to effectively create a capacitor across
the electrode-to-electrolyte interface from DC to approximately 10 kHz. At higher frequencies,
fluid viscosity effectively prevents ion reorganization from catching up to the AC signal such that
the double layer effect is no longer present and only solution resistance persists [52]. At even
8

higher (GHz) frequencies, the dipolar relaxation of water molecules can be observed, as is
commonly encountered in microwave ovens [53].
Objectives
Combining expertise in electrochemistry with bioMEMS Parylene C micromachining
therefore presents an exciting area of research with many potential outcomes to improve patient
care or expand scientific inquiry. Although much work remains to be done in translating these
types of devices from the bench to the clinic, the efforts presented throughout this dissertation
contribute towards this goal through the design, fabrication, and testing of two unique devices and
the utilization of measurement techniques from the National Institute of Standards and Technology
(NIST) onto Parylene C thin films. Therefore, aim of the present work is:

To advance the state of thin-film Parylene C BioMEMS implantables featuring
electrochemical interfaces by the design, fabrication, and testing of two distinct devices as well as
usage of NIST-based measurement techniques.

Chapter 2 reviews work described by the acronym, RRA – Rat Retinal Array. Here,
Parylene C micromachining techniques were used to overcome challenges associated with
chronically implanting a microelectrode array at the inner retinal surface (epiretinal) in rats. This
is significant as it represents a step in developing a research tool to accelerate retinal research for
improving the visual acuity provided by today’s retinal prostheses to help restore sight to certain
blind patient populations.
Chapter 3 describes the microbubble pressure transducer (µBPT). Faradic reactions
involving electrolytic gas generation inside of a Parylene C microfluidic channel allowed for
9

pressure transduction due to the inverse relationship between the volume of gas and pressure.
Previous researchers in the Biomedical Microsystems Laboratory demonstrated exciting proof of
concept, and this work contributed significant strides in rigorously elucidating µBPT operating
mechanisms to yield orders or magnitude improvements across several device performance metrics.
Chapter 4 presents the first-ever use of NIST techniques in the context of biomedical thin
film Parylene C characterization. Due to its high frequency bandwidth (> 100 GHz), this
measurement technique bears several advantages over traditionally employed techniques such as
electrochemical impedance spectroscopy. A primer on the underlying microwave measurement
theory is also provided.
Lastly, a concluding summary is provided to review the contributions towards the field
provided by this work.

10

References
[1] J. W. Orton, Semiconductors and the Information Revolution: Magic Crystals That Made
IT Happen. London: Elsevier Science & Technology, 2009.
[2] M. J. Madou, Fundamentals of microfabrication and nanotechnology , 3rd ed. Boca Raton,
Fla. ; CRC, 2012.
[3] A. Thackray, Moore’s law : the life of Gordon Moore, Silicon Valley’s quiet revolutionary .
New York: Basic Books, a member of the Perseus Books Group, 2015.
[4] W. P. Eaton and J. H. Smith, “Micromachined pressure sensors: Review and recent
developments,” Smart Mater. Struct., vol. 6, no. 5, pp. 530–539, 1997.
[5] A. Allan, Basic sensors in iOS , First edit. Sebastopol, California: O’Reilly, 2011.
[6] L. G. Villanueva, J. Bausells, and J. Brugger, “Grand Challenge in N/MEMS,” Front. Mech.
Eng., vol. 1, no. January, 2016.
[7] “BPS.SPACE Falcon-Heavy,” Barnard Propulsion Systems LLV, 2021. [Online].
Available: https://bps.space/falcon-heavy. [Accessed: 30-Jan-2021].
[8] “Foundations of Microsystems (MEMS) Manufacturing.” 2015.
[9] E. Meng, Biomedical microsystems . Boca Raton: CRC Press, Taylor & Francis Group,
2011.
[10] R. D. K. Misra, “Biosensors,” Mater. Technol. (New York, N.Y.), vol. 30, no. sup7, pp.
B139–B139, 2015.
[11] D. D. Agonafer, J. Yeom, and M. A. Shannon, “Three-Dimensional CFD Model of Pressure
Drop in µTAS Devices in a Microchannel,” J. Electron. Packag., vol. 133, no. 3, 2011.
[12] T. M. Squires, “Microfluidics Fluid physics at the nanoliter.pdf,” vol. 77, no. July, 2005.

11

[13] Y. Xia and G. M. Whitesides, “Soft Lithography,” Angew. Chemie (International ed.), vol.
37, no. 5, pp. 550–575, 1998.
[14] A. P. Petersen et al., “Microenvironmental Modulation of Calcium Wave Propagation
Velocity in Engineered Cardiac Tissues,” Cell. Mol. Bioeng., vol. 11, no. 5, pp. 337–352,
2018.
[15] P. C. H. Li, Fundamentals of Microfluidics and Lab on a Chip for Biological Analysis and
Discovery. Boca Raton: CRC Press LLC, 2017.
[16] N. Nakanishi, H. Yamamoto, K. Tsuchiya, Y. Uetsuji, and E. Nakamachi, “Development of
wearable medical device for Bio-MEMS,” vol. 6036, no. 1. SPIE, pp. 60360P-60360P–9,
2005.
[17] N.-C. Tsai and C.-Y. Sue, “Review of MEMS-based drug delivery and dosing systems,”
Sensors actuators. A. Phys., vol. 134, no. 2, pp. 555–564, 2007.
[18] J. H. A. Podlevsky, “Diagnostic and Therapeutic MEMS (Micro-Electro-Mechanical
Systems) Devices for the Identification and Treatment of Human Disease.” ProQuest
Dissertations Publishing, 2018.
[19] G. Jiang, “Design challenges of implantable pressure monitoring system,” Front. Neurosci.,
vol. 4, no. FEB, pp. 2–5, 2010.
[20] C. Compton, M. Rofael, and P. McCann, “The Use of CardioMEMS Device in Detecting
Arrhythmias,” J. Am. Coll. Cardiol., vol. 75, no. 11, p. 3341, 2020.
[21] X. Xie et al., “Long-term reliability of Al2O3 and Parylene C bilayer encapsulated Utah
electrode array based neural interfaces for chronic implantation,” J. Neural Eng., vol. 11,
no. 2, p. 26016, 2014.

12

[22] S. Wolf, Silicon processing for the VLSI era , 2nd ed. Sunset Beach, Calif: Lattice Press,
2000.
[23] B. J. Kim and E. Meng, “Review of polymer MEMS micromachining,” J. Micromechanics
Microengineering, vol. 26, no. 1, 2015.
[24] H. C. Stronks and G. Dagnelie, “The functional performance of the Argus II retinal
prosthesis,” Expert Rev. Med. Devices, vol. 11, no. 1, pp. 23–30, 2014.
[25] A. N. Pisarchik, V. A. Maksimenko, and A. E. Hramov, “From Novel Technology to Novel
Applications: Comment on ‘An Integrated Brain-Machine Interface Platform With
Thousands of Channels’ by Elon Musk and Neuralink,” J. Med. Internet Res., vol. 21, no.
10, pp. e16356–e16356, 2019.
[26] H. Xu, A. W. Hirschberg, K. Scholten, T. W. Berger, D. Song, and E. Meng, “Acute in vivo
testing of a conformal polymer microelectrode array for multi-region hippocampal
recordings,” J. Neural Eng., vol. 15, no. 1, 2018.
[27] B. S. Spearman et al., “Neural Interfaces: Tissue‐Engineered Peripheral Nerve Interfaces
(Adv. Funct. Mater. 12/2018),” Adv. Funct. Mater., vol. 28, no. 12, pp. 1870076-n/a, 2018.
[28] E. Meng, P. J. Chen, D. Rodger, Y. C. Tai, and M. Humayun, “Implantable parylene MEMS
for glaucoma therapy,” 2005 3rd IEEE/EMBS Spec. Top. Conf. Microtechnology Med. Biol.,
vol. 2005, pp. 116–119, 2005.
[29] C. Lee, L. Jug, and E. Meng, “High strain biocompatible polydimethylsiloxane-based
conductive graphene and multiwalled carbon nanotube nanocomposite strain sensors,” Appl.
Phys. Lett., vol. 102, no. 18, p. 183511, 2013.
[30] R. Sheybani, H. Gensler, and E. Meng, “A MEMS electrochemical bellows actuator for
fluid metering applications,” Biomed. Microdevices, vol. 15, no. 1, pp. 37–48, 2013.
13

[31] T. V CHIRILA et al., “Artificial cornea: Polymer science and the eye,” Prog. Polym. Sci.,
vol. 23, no. 3, pp. 447–473, 1998.
[32] M. Szwarc, “Some Remarks on the CH2=CH2 Molecule,” Discuss. Faraday Soc., vol. 2,
pp. 46–49, 1947.
[33] W. F. Gorham, “A New, General Synthetic Method for the Preparation of Linear Poly-p-
xylylenes,” J. Polym. Sci. A1., vol. 4, no. 12, pp. 3027–3039, 1966.
[34] J. Lahann, “Vapor‐based polymer coatings for potential biomedical applications,” Polym.
Int., vol. 55, no. 12, pp. 1361–1370, 2006.
[35] B. A. Weisenberg and D. L. Mooradian, “Hemocompatibility of materials used in
microelectromechanical systems: Platelet adhesion and morphology in vitro,” J. Biomed.
Mater. Res., vol. 60, no. 2, pp. 283–291, 2002.
[36] “Parylene Chemical Resistance,” Parylene Engineering, 2021. [Online]. Available:
https://paryleneengineering.com/parylene_chemical_resistance.htm. [Accessed: 30-Jan-
2021].
[37] P. R. Menon, W. Li, A. Tooker, and Y. C. Tai, “Characterization of water vapor permeation
through thin film Parylene C,” SENSOR. IEEE, pp. 1892–1895, 2009.
[38] J. Ortigoza-Diaz et al., “Techniques and Considerations in the Microfabrication of Parylene
C Microelectromechanical Systems,” Micromachines, vol. 9, no. 9, p. 422, 2018.
[39] X. Wang et al., “A Parylene Neural Probe Array for Multi-Region Deep Brain Recordings,”
J. microelectromechanical Syst., vol. 29, no. 4, pp. 499–513, 2020.
[40] S. Elyahoodayan, C. Larson, A. M. Cobo, E. Meng, and D. Song, “Acute in vivo testing of
a polymer cuff electrode with integrated microfluidic channels for stimulation, recording,
and drug delivery on rat sciatic nerve,” J. Neurosci. Methods, vol. 336, p. 108634, 2020.
14

[41] T. Hudson, A. Baldwin, and E. Meng, “A Continuous, Drift-Compensated Impedimetric
Thermal Flow Sensor for in Vivo Applications,” in 2019 IEEE 32nd International
Conference on Micro Electro Mechanical Systems (MEMS), 2019, pp. 731–734.
[42] A. Baldwin, E. Yoon, T. Hudson, and E. Meng, “Fluid Temperature Measurement in
Aqueous Solution via Electrochemical Impedance,” J. Microelectromechanical Syst., vol.
PP, pp. 1–8, 2019.
[43] A. Baldwin and E. Meng, “Kirigami Strain Sensors Microfabricated From Thin-Film
Parylene C,” J. microelectromechanical Syst., vol. 27, no. 6, pp. 1082–1088, 2018.
[44] A. Bard and L. Faulker, Electrochemical Methods: Fundamentals and Applications, 2nd
Edition. John Wiley & Sons, Inc., 2000.
[45] D. R. Merrill, The Electrode - Materials and Configurations. 2011.
[46] K. J. Rountree, B. D. Mccarthy, E. S. Rountree, T. T. Eisenhart, and J. L. Dempsey, “A
Practical Beginner ’ s Guide to Cyclic Voltammetry,” 2017.
[47] S. F. Cogan, “Neural Stimulation and Recording Electrodes,” Annu. Rev. Biomed. Eng., vol.
10, no. 1, pp. 275–309, 2008.
[48] E. M. Hudak, D. W. Kumsa, H. B. Martin, and J. T. Mortimer, “Electron transfer processes
occurring on platinum neural stimulating electrodes: Calculated charge-storage capacities
are inaccessible during applied stimulation,” J. Neural Eng., vol. 14, no. 4, pp. 1–24, 2017.
[49] R. Sheybani, H. Gensler, and E. Meng, “Rapid and repeated bolus drug delivery enabled by
high efficiency electrochemical bellows actuators,” 2011 16th Int. Solid-State Sensors,
Actuators Microsystems Conf. TRANSDUCERS’11, pp. 490–493, 2011.
[50] V. B. Svetovoy, R. G. P. Sanders, K. Ma, and M. C. Elwenspoek, “New type of microengine
using internal combustion of hydrogen and oxygen.,” Sci. Rep., vol. 4, p. 4296, 2014.
15

[51] D. A. Ateya, A. A. Shah, and S. Z. Hua, “Impedance-based response of an electrolytic gas
bubble to pressure in microfluidic channels,” Sensors Actuators, A Phys., vol. 122, no. 2,
pp. 235–241, 2005.
[52] L. Yu, C. A. Gutierrez, and E. Meng, “An Electrochemical Microbubble-Based MEMS
Pressure Sensor,” J. Microelectromechanical Syst., vol. 25, no. 1, pp. 144–152, 2016.
[53] C. A. E. Little, N. D. Orloff, I. E. Hanemann, C. J. Long, V. M. Bright, and J. C. Booth,
“Modeling electrical double-layer effects for microfluidic impedance spectroscopy from
100 kHz to 110 GHz,” Lab Chip, vol. 17, no. 15, pp. 2674–2681, 2017.

16

Background
The photosensitive cells of the retina transduce light into electrical signals and enable
visual perception. Access to the retina is desired to study the neural signaling involved in vision
and also to introduce implantable retinal prostheses which have the potential to restore vision to
patients blinded by certain diseases. Both of these goals may be accomplished through use of
microelectrode array (MEA) neural interfaces. Because the retina is thin, spans a small area, and
is mechanically delicate [54], the construction of retinal interfaces is well-suited through the use
of soft materials and micromachining.
Rats are an excellent model to study retinal degeneration and the only FDA approved
retinal prosthesis employs the epiretinal approach as opposed to subretinal or suprachoroidal
approaches. However, no devices suitable for epiretinal rat studies have been reported to date. This
is largely due to surgical access challenges stemming from the small ocular orbit (~5 mm) and
extremely delicate nature of retinal tissue. Hence, biocompatible and mechanically soft Parylene-
based MEMS are well positioned to tackle these unique issues as shown through the design,
fabrication, and testing in this work. This is significant as it represents a step in developing a

RAT RETINAL ARRAY (RRA) TO AID CHRONIC EPIRETINAL
STIMULATION RESEARCH FOR ARTIFICIAL VISION
17

research tool to accelerate retinal research for improving the visual acuity provided by today’s
retinal prostheses.
2.1.1 Retinal Degeneration
In order to understand the device-related opportunities in the eye, it is first necessary to
briefly review how visible light enters the eye and is processed by the retina before information is
transmitted by the optic nerve to different downstream processing centers in the brain (Figure
2.1-1).

Figure 2.1-1. Anatomy of the ocular orbit and key components involved in the visual transduction pathway

In most mammalian eyes, visible light is first refracted by the cornea, passes through the
pupil, and is refracted once more by the lens. The lens is part of a mechanical system within the
eye; the ciliary muscle acts on the lens via zonular fibers. Contraction and relaxation of the muscle
18

causes increases in convexity or flattening which in turn allow the eye to focus on objects that are
near or far, respectively. Light exiting the lens then reaches the retina which is a membrane
spanning the back of the ocular orbit. However, light initially penetrates through many layers of
the retina as its 100-300 µm thickness renders it virtually translucent. Phototransduction occurs
when light is projected onto the photoreceptor layer. Here, rod and cone cells (~130 million)
convert the light stimulus into electrical signals via the hyperpolarization or depolarization of cell
membranes. The electrical signals are then passed from photoreceptors through a cascade of
different neuronal cell types in the opposite direction from which they entered the eye. Generally
speaking, graded potentials traverse and are processed through horizontal, bipolar, and amacrine
cells whereas the final ganglion cells output all-or-nothing action potentials [55] to the optic nerve,
visual cortex and other parts of the brain.
The malfunction of or damage to any of these ocular components may result in vision
impairment or blindness. The most common cause of blindness in the adult working population
arises from inherited retinal diseases such as age related macular degeneration (AMD) or retinitis
pigmentosa (RP) [56]. In these diseases, damage or death occurs to the photoreceptor cell layer,
thereby causing gradual blindness. For RP, the progressive loss of photoreceptors first results in
compromised peripheral vision and eventually blindness as shown in Figure 2.1-2. There is
currently no cure for RP or AMD [57]. However, a class of implantable devices known as retinal
prosthesis have recently emerged as a treatment for patients with RP or AMD.

19

Figure 2.1-2. Normal vision compared to loss of vison from the point of view of a patient with retinitis pigmentosa

20

2.1.2 Artificial Vision through Retinal Prostheses
Retinal prosthesis devices such as the FDA-approved Argus II [24] feature an implantable
multielectrode array (6 × 10 electrodes) on a polyimide substrate to provide direct electrical
stimulation to the still-functional ganglion cells of the inner retina for patients with profound visual
loss from retinal degeneration. This approach effectively bypasses the degenerated photosensitive
cells in outer retinal cell layers [58]. The visual feed from a camera fitted onto spectacles worn by
the patient is processed into output stimulation commands which are sent to the electrode array via
wireless telemetry as shown in Figure 2.1-3. The electrical stimulation patterns evoke actional
potentials on the still-functional retinal ganglion cells which mimic signals that may have been
generated from a functional visual pathway. This treatment scheme provides patients with artificial
vision which assists in improving quality of life.

Figure 2.1-3. Illustration of the Argus II system and its main components. An electrode array is implanted onto the
retina. The visual feed from a camera fitted onto spectacles worn by the patient is processed and then transmitted via
wireless telemetry

Although some semblance of vision is provided by retinal prosthetic devices, much work
still remains in order to restore visual acuity to levels typically present in society today. Rather
21

than experiencing a blurry or pixelated view of the environment, patients fitted with the Argus II
have reported seeing a phenomenon known as phosphenes [59]. A phosphene may be a ring or
spot of light produced by pressure on the eyeball or direct stimulation of the visual system other
than by light. Depending on the patient, their ensemble of phosphenes enabled varying degrees of
visual restoration. Most patients reported the ability to perceive doors, walls, or people in their
vicinity. In one study, the Argus II allowed approximately half of its 21 tested subjects to recognize
large-print letters above chance levels [60]. However, a deeper understanding of chronic retinal
stimulation on the neuroplasticity of the primary visual cortex is required to improve visual acuity
and overall clinical outcomes [61]–[63]. The best visual acuity reported in Argus II clinical trials
was 20/1,260 [64] whereas 20/200 is considered legally blind in the United States and 20/20 is
considered normal vision.
2.1.3 Improving Chronic Retinal Stimulation Studies in Animals
Animal models play an important role in vision research and may accelerate progress
towards improved retinal prostheses. Clinical devices such as the Argus II were tested in canine
models prior to human clinical trials and were placed on the inner layer of the retina (epiretinal
placement). Acute epiretinal implants in feline [65], [66] and rabbit models [67] were also reported,
yet chronic implants were tested in only a handful of cases. For instance, a 120 day long study
using earlier Argus I devices in canines included a limited number of animals (N = 6) [68].
The rat model is widely accepted and often preferred for retinal degeneration research [69],
[70]. The model is readily available at a reasonable experimental cost which facilitates meaningful
study sizes. However, the size and anatomy of the rat eye impose technically challenging
specifications on chronically indwelling retinal stimulation MEAs [71]. Development of rat MEAs
has thus been limited to the subretinal (between the outer retinal layer and the retinal pigment
22

epithelium) or suprachoroidal (between sclera and choroid) space in which the array can be held
in place between two immuno-privileged membranes [72]–[74].
However, subretinal and suprachoroidal approaches suffer the marked disadvantage of
indirect access to the ganglion cell layer [55]. Electrical attenuation as the device’s stimulation
signal traverses the inner limiting membrane must be considered [75] because increased distance
between the stimulating electrode and the retina leads to increased perceptual threshold [76].
Conversely, the main advantage of the epiretinal approach is the direct access to still-functional
retinal ganglion cells albeit accompanied by notoriously challenging surgeries. Retinal detachment,
hypotony, infection or other serious adverse events are more likely to occur in epiretinal operations
due to the retina’s extremely delicate nature [77]. Such challenges in surgical access have limited
epiretinal studies of rat retino-cortical relationships to cases where investigators resorted to manual
point-by-point epiretinal probing with needle electrodes on anaesthetized rats in acute settings [78],
[79]. Given the importance of the rat model in vision research, an indwelling epiretinal MEA was
developed to enable investigation of chronic epiretinal stimulation in rats.
Approach and System Overview
2.2.1 Design Goals and Challenges
In this work, the goal was to develop a chronically indwelling epiretinal multielectrode
array for wireless stimulation in rats. Because such a system has never been reported in the
literature, this project may contribute to the overall field of retinal prosthetics and artificial vision
by providing a tool to accelerate research on retinotopic neuroplasticity. Figure 2.2-1 illustrates
the main design concept. A Parylene-based microelectrode array is implanted into the rat eye in
the epiretinal direction. The electrodes lie on a curved Parylene substrate in order to match the
23

hemispherical geometry of the retina. The device is routed subcutaneously towards a compact
printed circuit board (PCB) which has been affixed to the cranium via a dental cement headcap. A
wireless signal provides stimulus commands such that the rat may be roaming freely to enable
awake behavioural experiments on chronic timescales unlike previous work limited to acute and
anaesthetized subjects.

Figure 2.2-1. Schematic illustration of rat retinal array (RRA) system and main components

A cyclical approach consisting of fabrication, implantation, and design was employed to
prototype multiple device iterations (Figure 2.2-2). Initial experiments were tested in rat cadavers
or enucleated eyes with sham devices that did not contain metal features in order minimize cost
while expediting lessons learned via rapid prototyping. As a deeper understanding of constraints
and trade-offs emerged, three metalized design iterations were developed in this work. Three main
factors were kept in mind throughout the design process – anatomy, electrochemistry, and surgical
feasibility.
24

Figure 2.2-2. The approach for this work employed several iterations of cycling between fabrication, implantation,
and design. Each cycle yielded key insights and knowledge to inform the next iteration

Firstly, challenging design constraints arose from the ocular and cranial anatomy of the rat.
The orbit is significantly smaller (diameter ≈ 5 mm) [80] than that of other species which have
undergone retinal array stimulation studies such as rabbit, guinea pig, sheep, human, and more.
Therefore, an alternative method of mechanical fixation of the array onto the retina was required.
Use of retinal tacks is standard in other species [23,24], but doing so in rats would create a
prohibitively large puncture and the resulting loss of intraocular pressure would induce globe
rupture. The small orbit also limits the overall number of microelectrodes which may be packed
onto the array.
To further compound geometric difficulties, the rat eye has a considerable range of motion
in a variety of axes. Rats may temporarily proptose, or shift their globes slightly out of the socket,
and they are also capable of ocular torsion (rotation around the optical axis) and cycloversion
(rotation of both eyes in the same direction) [81] . Hence, a certain degree of flexibility or cabling
slack to the stimulating region is required in order to minimize abrasion and irritation between the
device and the delicate retina. In fact, preliminary attempts at chronic stimulation via an implanted
25

microwire instead of flexible thin-film polymer-based arrays were thwarted precisely because of
this issue. Lastly, the overall length of the device’s ribbon cable must be tailored to match the
distance from the eye socket to the downstream electronics fitted on a headcap. These anatomical
challenges were overcome through expertise in Parylene micromachining which was aided by
careful observations and modifications of each design iteration.
Secondly, the fundamental electrochemical tradeoff due to the inverse relationship between
electrode size and impedance also necessitated careful design [45]. To successfully stimulate
retinal ganglion cells and elicit electrically evoked responses, electrodes must be able to deliver
sufficient charge via stimulation pulses such as the standard charge-balanced biphasic cathodic-
first waveform [82]. A larger electrode has lower electrochemical impedance and may store larger
amounts of charge. On the other hand, it is desirable to have smaller electrodes due to spatial
selectivity. Retinal ganglion soma in rats are approximately 10 to 20 µm in diameter [83] so having
electrodes of equal or smaller size may selectively simulate ganglion cells in a 1:1 manner.
However, smaller electrodes have higher impedance which may induce unwanted electrochemical
reactions such as gas evolution or generate acidic or alkaline species if their charge storage
capacity is exceeded [47]. For this reason, electrode design in this work featured diameters which
varied along one of the two rows of the multielectrode array. The other row featured constant
diameter electrodes to investigate electrical field attenuation dynamics in future in vivo work.
Additionally, the use of electroplated platinum-iridium (PtIr) coatings allowed for increase of the
electroactive surface area (ESA) while maintaining the same geometric surface area (GSA) in
order to address electrochemical constraints in this work.
Lastly, surgical feasibility was a challenge which required careful design, communication,
and optimization throughout this work. Test surgeries were initially conducted at the University of
26

Southern California and later in the University of Michigan. Although certain instances of device
mishandling were unavoidable, each attempt provided valuable lessons learned which helped to
build an intuitive understanding of reasonably obtainable design features in this work. Hard-earned
key surgical nuances are articulated in ensuing sections.
2.2.2 Microfabrication Process Flow
Devices were fabricated using established surface micromachining processes [84] as
shown in Figure 2.2-3 and also detailed in the appendix. A single 10 µm thick layer of Parylene
was deposited on a silicon carrier wafer and then AZ 5214E image reversal photoresist (Integrated
Micro Materials, Argyle, TX) was spun on and patterned to define the metal wires, electrodes, and
contact pads. An O2 plasma descum (60 s, 100 W, 100 mtorr) was performed immediately prior to
electron beam evaporation of 99.99% Pt (PraxAir Inc., Danbury, CT) by using a CHA Mark 40
system (CHA Industries, Fremont, CA).
In total, 200 nm of Pt was deposited, but special care was required to prevent metal cracking
due to the thermal expansion coefficient mismatch between Parylene and Pt. Without breaking
vacuum, the 200 nm of Pt was split into four deposition steps of 50 nm with 30-minute pauses
between each step. The deposition rate was 1.5 Å/s. The resting steps and a long chamber throw
distance (55 cm) were crucial in providing enough cooling of the substrate. Metal liftoff was
performed in sequential baths of acetone, isopropanol (IPA), and deionized water each for 10
minutes at room temperature.
After liftoff, another O2 descum was performed and then an additional 10 µm layer of
Parylene was deposited. Next, a layer of AZ P4620 photoresist (Integrated Micro Materials, Argyle,
TX) was used to define the cutout outline shape and suture holes for the arrays. The photoresist
served as the protecting mask as the wafer underwent a switched chemistry process in a deep
27

reactive ion etching (DRIE) tool that alternated between fluoropolymer deposition (C 4F8) and
oxygen plasma etching [38]. Etching with this mask proceeded until the top 10 µm of Parylene
was removed, as measured by Dektak XT Profilometer (Bruker, Millerica, MA). After photoresist
stripping, another AZ P4620 layer was spun on as a photoresist mask which defined the cutout
outline, suture holes, and also the openings for the electrodes and contact pads.
Next was another DRIE step to simultaneously expose the metal at the electrodes and
contact pads while completing the cutout and suture etch down to the silicon substrate. Deionized
water drops were applied to individual devices to facilitate release by carefully peeling off the
wafer with tweezers. To ensure that all residual photoresist was removed, devices were then
cleaned at room temperature in sequential baths of acetone, IPA, and deionized water.

Figure 2.2-3. Microfabrication process flow for all rat retinal array devices

28

Design Iterations
The rat retinal array (RRA) underwent three major design iterations. Changes throughout
device evolution were predominantly focused on overall device fit due to lessons learned from
surgical implantation trials. The stimulation region at the device’s distal tip remained constant
throughout all design iterations.
As shown in Figure 2.3-1, the array featured 2 rows of 4 electrodes, where the one row varied in electrode
diameters and bottom row featured constant diameters as a method of determining optimal electrode dimensions
[80], [85].

Table 2.3-1 lists the exposed electrode diameter values which were estimated from calculations
laid out in [86] in conjunction with the expectation of impedance reduction via PtIr electroplating.
Different stimulation region lengths were designed (2.0, 2.5, and 3.0 mm) to account for any
potential variability in rat retinal size. The stimulation region width is 780 µm in order to maximize
the number of electrodes that could be packed into the region while minimizing the scleral incision
size. 780 µm was chosen because preliminary surgical experiments revealed that intraocular
pressure will drop and cause globe rupturing if the scleral incision exceeded approximately 1 mm.
Wiring for the electrodes is 30 µm wide and separated from one another by 30 µm wide spaces.
The wires extend out from the stimulation region into the ribbon cable region where downstream
design features may vary depending on the RRA version. The three versions are recounted in the
following section, and the relevant version employed in ensuing methods and results sections will
be denoted.

29

Figure 2.3-1. Illustration of the RRA stimulation region

Table 2.3-1. RRA exposed electrode diameters (µm)

2.3.1 RRA Version 1
As shown in Figure 2.3-2, the wires tapered out into the ribbon cable region. This
intentionally created a wire-free region throughout the central length of the ribbon cable where the
Electrode Letter Exposed Diameter (µm)
A 210
B 60
C
60
D 40
E, F, G, and H 160
30

surgeon had the option grasp the device with sharp tweezers without the risk of piercing the
Parylene at the wires and consequently compromising electrical insulation. Additionally, grasping
tabs on the side of the ribbon cable region were included and designed to be snipped off with
surgical scissors when no longer needed. Suture slits were included to allow device fixation onto
the eye via suture threads onto the sclera which provided a suitable alternative to retinal tacks. The
array terminated in 8 contact pads which were sized to mate with commercial connectors attached
to downstream electronics.

Figure 2.3-2. Design schematic for RRA version 1

After implanting version 1 devices into rats, a few lessons were learned. Snipping off the
tabs after suturing the device on the eye was found to be surgically unfeasible because of the
difficult angle of approach and proximity to the wires. The suture slits allowed for too much slack
along their longitudinal axis which caused unwanted shifting and translation of the device across
time. It was suggested to create 4 holes instead of 2 slits. The width of the slits (120 µm) and their
distance away from each other in the lateral direction (760 µm) was found to be well-suited for the
suture needle (7707G: 10-0 Ethilon Black 12" TG160-4 Spatula) and suture thread (Nylon 66).
RRA version 1 devices were constructed from sputtered Pt via an external vender (LGA Thin
31

Films®, Inc, Santa Clara, CA) which suffered from poor metal liftoff whereas RRA version 2 and
3 employed the electron beam evaporated Pt described in 2.2.2.

2.3.2 RRA Version 2
In the second version (Figure 2.3-3), the suggestion regarding the suture holes was heeded
by creating 4 suture holes of 120 µm diameter spaced apart by 760 µm. The grasping tabs were
intended this time to be permanently attached to the device and were rounded to minimize potential
abrasion against ocular tissue. Features in the contact pad region were splined because the sharp
90° angles in the metal wire were postulated to create regions of concentrated electric field strength
which may accelerate metal-Parylene delamination during high charge density pulses.
Smoothening the 90° bend in the Parylene also assisted significantly when peeling devices off of
the wafer.

Figure 2.3-3. Design schematic for RRA version 2

Surgeries with this design version revealed key insights since the implantation protocol
evolved to call for routing the array subcutaneously from the ocular orbit to the top of the cranium.
This was intended to be accomplished by slotting the device through a 12 gauge needle temporarily
32

routed beneath the skin. However, the grasping tabs rendered the overall width of the device wider
than that of the needle. Forcing the array through the needle induced Parylene deformation and
rendered all electrodes nonfunctional. Careful snipping of the RRA version 2 tabs prior to suturing
sufficiently reduced device width and allowed for smooth transport through the 12 gauge needle.
Additionally, the distance from the suture holes to the electrodes proved to be a critical
parameter. The surgeon commented that this distance should be minimized due to the real limited
real estate of the rat sclera for device suturing. It was suggested to remove the distal two holes
since there would be insufficient space on the sclera to employ all four holes.
Lastly, a thermoforming technique [87] was employed to impart curvature at the
stimulation region as well as the transition between the stimulation region to the ribbon cable
region. The second curvature was found to be too severe (radius of curvature ≈ 200 µm) as it
caused cracking of approximately 50% of the 8 wires across all devices. These major lessons were
kept in mind as the third version was developed.
2.3.3 RRA Version 3
The final design iteration of this work is displayed in Figure 2.3-4. The tabs and wire-free
region have been removed as the surgeon noted that they imparted too much device width which
inhibited smooth transport through the 12 gauge shuttle needle. Such features also became
unnecessary after gaining experience through previous implantation trials. It was instead preferred
to grasp at the most proximal end of the device where the contact pad region had been potted with
biocompatible epoxy. Pulling was accomplished by threading the suture threads through the two
33

suture holes prior to implantation. The combination of grasping and pulling allowed for smooth
navigation through the 12 gauge needle since the effective width of the ribbon cable was reduced.
The thermoforming technique was still used to curve the stimulation region but the “candy
cane” hook geometry of the version 3 RRA obviated the need for bending the transition between
the stimulation region to the ribbon cable region. This allowed for 100% electrode yield post-
thermoforming. Additionally, thermoforming accuracy was improved by creating 2 alignment
holes through which pins could ensure proper registration between the Parylene device and the
thermoforming mold fixture.

Figure 2.3-4. Design schematic for RRA version 3

Experimental Methods
After the fabrication process described in section 2.2.2, devices underwent post-fabrication
processes such as thermoforming or PtIr coating. Packaging enabled electrochemical
characterization, but a newly developed reversible packaging scheme allowed for testing in
between post-fabrication steps. Testing endpoints were in vivo experiments.
34

2.4.1 Refinement of the Thermoforming Process
Devices were thermoformed (200 °C, 48 hr, 3 × nitrogen purged) by mounting them in a
custom mold to obtain the appropriate curled shape. However, several early attempts at
thermoforming caused loss of functional electrode yield primarily due to two failure modes. The
first failure mode was traced to excessive pressure applied by the mold when undergoing the
thermoforming step. Early molds were designed to sandwich the ribbon cable and contact pad
region between two polytetrafluoroethylene (PTFE) sheets surrounded by two aluminum plates.
Pressure applied by four #4-40 metal screws through the overall assembly was unintentionally
excessive (>2 ton-force) and caused cracking at the contact pads Figure 2.4-1. Reducing the force
by using binder clips or solely the weight of the aluminum plate to fix the device in place
eliminated this failure mode completely. This issue was observed during RRA version 1
development and solved during RRA version 2 development.

Figure 2.4-1. Excessive pressure applied during thermoforming created cracks at device contact pads

The second thermoforming failure mode was associated with the requirement to have a
second thermoformed bend in RRA version 1 and 2. As shown in Figure 2.4-2, the first and second
bend are realized along the same axes in order to route the device ribbon cable back towards the
35

headcap. This second bend radius (r ≈ 200 µm) was selected to be as close as possible to the
thickness of the rat sclera [88] while following guidelines established through bend tests of similar
Parylene-Pt-Parylene sandwich devices [89]. Unfortunately, microscopic cracks could be observed
in the Pt wires at the second bend region, rendering about 50% of electrodes nonfunctional.

Figure 2.4-2. Demonstration of first and second bend via thermoforming in RRA version 1 and 2

The solution was to employ the “candy cane” or hook geometry as illustrated in section
2.3.3. This enabled the ribbon cable to be routed back towards the headstage without the need for
a second bend. Autodesk Inventor CAD software was used to design the mold illustrated in Figure
2.4-3 which employed two custom-machined PTFE plates to hold the array in place while stainless
steel cylindrical dowel pins (not pictured) could be inserted intro grooves on the adjacent face to
impart solely the first curve. After further experimentation, replacing one of the custom PTFE
plates with a PTFE sheet, glass slide, and binder clips was found to be more effective and
convenient because the sheet and slide provided visibility due to their translucence. One machined
PTFE plate is shown in Figure 2.4-4.
36

Figure 2.4-3. (a) Autodesk Inventor CAD illustration of PTFE mold for thermoforming RRA version 3 devices. (b)
The first bend to match retinal curvature could be obtained through a slot and groove and (c) proper registration
between device and mold was ensured through registration pin holes

Figure 2.4-4. Photograph of PTFE mold component for thermoforming RRA version 3 devices
37

2.4.2 Platinum-Iridium Electroplating
A subset of devices was thermoformed flat (herein referred to as annealed) to allow for
high magnification imaging with compound microscopes. If nonplanar, images were unobtainable
as the specimen did not reflect light back to the objective lens. Annealed arrays were coated in
platinum iridium (PtIr) via an electrodeposition method [90] [91] to lower electrochemical
impedance and improve charge storage capacity. Collaborators at Pt Group Coatings, LLC and the
University of Michigan received annealed RRA devices and performed the coating technique.
Briefly, devices were immersed in a solution of sodium hexachloroiridate (III) hydrate,
(Na3IrCl6 ▪H2O) and sodium hexachloroplatinate (IV) hexahydrate (Na2PtCl6 ▪6H2O) in 0.1 M
nitric acid (HNO3) and the electrode potential was repeatedly cycled. Additional details on the PtIr
coating process may be found in [92].
The nanofractal surface of PtIr increases the overall surface area of the electrode, which in
turn creates additional sites for electrochemical reactions to occur. This increases the electroactive
surface area while maintaining the same geometric surface area. Additionally, the nanofractal
surface absorbs light more effectively than bare Pt. As shown Figure 2.4-5, this caused the
difference in electrode coloration from the originally reflective bare Pt to the solid black PtIr. Note
that the solid black is limited to the electrode sites and not the wires.

38

Figure 2.4-5. Optical micrographs of a representative device electroplated in PtIr

2.4.3 Packaging
2.4.3.1 Device to Flat Flexible Cable
Packaging was accomplished by zero insertion force (ZIF) connectors (Hirose Electric Co.,
Ltd, Tokyo, Japan) mated to the Pt contact pads. A polyethylene terephthalate (PET) backing
provided the necessary thickness (~200 µm) for the ZIF to engage its closing latch. The backing
was cut with a motorized cutter to match the shape of the overall contact pad region to facilitate
proper insertion into the ZIF. Unlike previous ZIF packaging techniques which employed
cyanoacrylate adhesive [93], this technique permitted reversible ZIF connections to alow
characterization between cumulative post-fabrication processes such as thermoforming, annealing
or PtIr coating.
2.4.3.2 Device to Wireless Headstage
A custom PCB assembly was required to serve as an adapter which lies between the rat
retinal array (RRA) and an off-the-shelf headstage as shown in Figure 2.4-6 and Figure 2.4-7.
The RRA was compatible with only ZIF connectors whereas the headstage required an Omnetics
connector. Since the headstage only contains 2 output stimulation channels, a dual in-line package
39

(DIP) switch was integrated to route those 2 channels to the 8 RRA electrodes. The main design
goals revolved around minimizing overall footprint and weight while maintaining reliability. This
packaging scheme was implemented for chronic implantation experiments with RRA version 3
devices.

Figure 2.4-6. Schematic illustration of packaging scheme between retinal array to wireless headstage

Figure 2.4-7. (a) Bottom face and (b) top face of PCB adapter assembly

40

2.4.4 Benchtop Testing
2.4.4.1 Crosstalk Testing with Chronic Soaking in 1 × PBS
Although undesired, finite amounts of water and ion permeation into the Parylene polymer
insulation layer were hypothesized to induce electrical crosstalk between electrodes. The creation
of such conductive pathways between adjacent traces via electrolyte penetration into the Parylene
may severely compromise device operation. In order to further elucidate the rates of such
phenomena, an experiment was conducted to quantify crosstalk.
A crosstalk measurement tool was developed by Chris Larson which employed a NI Virtual
bench unit (National Instruments, Austin, TX) and custom LabVIEW scripts in order to apply a
500 mV 1kHz sine wave across all combinations of electrode pairs possible with the 8 RRA
electrodes. For this experiment, the Parylene on top of 6 out of 8 electrodes of the RRA was
intentionally unetched during their microfabrication, rendering them fully insulated. The device
was submerged in 1 × PBS at 37 °C via convection oven in a custom-built acrylic fixture which
allowed for optical microscopy of both front and back sides of all electrodes (Figure 2.4-3).
Devices were unannealed and soaked for +1000 hours with intermittent imaging and crosstalk
percentage measurement. To obtain a crosstalk measurement, the inherent baseline crosstalk in the
overall system without the RRA (B) was first obtained by computed as the average of the ratio of
output to input voltages across all measurement terminal pairs when no device was connected to
the tool.
𝐵 = 〈
𝑉 𝑜𝑢𝑡 𝑉 𝑖𝑛
〉

Then, crosstalk between the i’th and j’th electrode pair was computed with the following:
41

𝐶𝑟𝑜𝑠𝑠𝑡𝑎𝑙𝑘 𝑖 ,𝑗 =
𝑉 𝑜𝑢𝑡 ,𝑖 − 𝐵 ∗ 𝑉 𝑖𝑛 ,𝑗 𝑉 𝑖𝑛 ,𝑗 ∗ (1 − 𝐵 )

Figure 2.4-8. (a) Custom-built acrylic fixture CAD model and (b) photograph for crosstalk testing

2.4.4.2 Electrochemical Impedance Spectroscopy
A Gamry Reference 600 Potentiostat (Gamry Instruments, Warminster, PA, USA) was
used in all other electrochemical experiments. An Ag/AgCl electrode (3M NaCl) (Basi® Inc., West
Lafayette, IN) was the reference and a 1 cm long and 0.5 mm diameter Pt wire (World Precision
Instruments, Sarasota, FL) was the counter electrode. The electrolyte was either 1 × phosphate
buffered saline (PBS) to mimic the vitreous humor [94] or 0.05 M H2SO4 (with 10 min of N2
bubbling) at room temperature. Potentiostatic electrochemical impedance spectroscopy (EIS) was
carried out from 1 Hz to 10 MHz with a 25 mV perturbation signal and no DC bias.
2.4.4.3 Cyclic Voltammetry
Cyclic voltammetry (CV) was conducted with no DC bias at 100 mV/s with -0.6 to 0.8 V
[95] endpoints with respect to the reference electrode. 30 cycles initiated and terminated at 0 V
with respect to the average open circuit potential to avoid unintentional surface conditioning.
42

2.4.4.4 Charge-Balanced Biphasic Cathodic-First Pulse Testing
Gamry Framework Custom Scripts (Galv Repeating Pulse) applied five charge balanced,
cathodic first, biphasic pulses with current densities and phase durations based on similar work
[78] (30 µC/cm
2
, 500 µs width, 100 µs gap). Typical pulse current amplitudes in rat epiretinal
studies for neuroplasticity may range from 5 to 100 µA with 1 Hz frequency, 100 µs gap, and 50%
duty cycle [78]. The recorded voltage was averaged and normalized to open circuit potential values.
2.4.5 In vivo Testing
2.4.5.1 Acute and Chronic Sham Implantations
Preliminary surgeries with Parylene-only sham devices were conducted on iterative
designs. Sham devices were sterilized by ethylene oxide gas. Acute sham studies used enucleated
eyes from rat cadavers. For chronic sham studies, the devices were then implanted in the left eye
of male Long-Evans rats (n = 2, 3-4 months, 250-350 g) to evaluate the design and surgical
technique.
All procedures for sham devices were performed in accordance to a protocol approved by
the Animal Care and Use committee at the University of Southern California. Rats were
anesthetized by an intramuscular injection of ketamine and xylazine mixture, lasting
approximately 2 hours. During the surgery, the anesthetic depth was periodically monitored via
“toe-pinch” response. If a response was elicited, the animal received another half dose of anesthetic.
In sham surgeries, the ribbon cable was sliced off such that only the stimulation region and suture
hole section remained in order to focus on implant feasibility. A 6-0 suture was used to hook and
pull away the top eyelid for better access to the eye. Surgical scissors cut 2 mm of the conjunctiva
to help create an incision on the sclera. The tip of a 30-gauge needle was used to slowly rub the
43

sclera for conjunctiva removal. Then, a 10-0 suture needle was used to pass through the suture
section, hooked on the sclera near the incision, and a loose knot was made. The tab was grasped
by fine tweezers to push the device through the incision at an acute angle to avoid contacting the
cornea. Once inside the orbit, the device was positioned at the upper temporal quadrant and the
loose knot was tightened. The animal recovered in warm bedding and returned to its holding once
able to move around normally again. Animals were monitored for pain, inflammation and infection
for up to two weeks post-surgery. From week two up to week six, animals were anesthetized once
a week to image with optical coherence tomography (OCT) and fundus imaging.

2.4.5.2 Acute Implantation for Metallized Device
Metallized device experiments were conducted at the University of Michigan in accordance
to a protocol approved by their Institutional Animal Care and Use Committee. A Long-Evans rats
was anesthetized in 3.0% isoflurane and maintained at 2.3% to eliminate toe pinch reflex. The
animal was secured on a stereotaxic frame and the head was secured with ear bars. After shaving
and alternating between 70% ethanol and betadine cleansing three times, a longitudinal incision
was made over the skull along the midline. The skin was parted, and the fascia was removed to
expose the cranial bone. A drill hole was made 2.0 mm anterior to the bregma suture with a dental
drill. A stainless-steel machine screw was bored into the drill hole for the counter/reference.
Silicone elastomer was applied on top of and around the machine screw to ensure electrical
isolation of the counter/reference electrode.
The eye was proptosed by routing a 5-0 silk suture past the temporal eyelid and pulling the
eyelid towards the midline. A peritomy was performed with iris scissors on the conjunctiva 1 mm
away from the limbus on the temporal edge of the eye. A blunt dissection of the conjunctiva
44

separated it from the underlying sclera, completing the peritomy. A 20-gauge needle was
introduced under the conjunctiva to create the initial hole for routing the array. The needle was
removed, and a 12-gauge needle was introduced to create a tunnel for the device. It was introduced
into the needle lumen and the needle was pulled back, acting as a vehicle to transport the Parylene
array into the eye. A 30-gauge needle was used to create the sclerotomy for array insertion.
Implanted devices were thermoformed with bare Pt.
Care was taken to avoid the lens during needle insertion and sclerotomy expansion to avoid
cataract formation. An 8-0 suture was threaded into the first suture hole in the device and the suture
needle was incised in the sclera 1 mm away from the sclerotomy, directing the suture parallel to
the sclerotomy. After the suture was threaded into the other suture hole, the device was lowered
towards the eye, inserting the stimulation region into the sclerotomy using a second 30-gauge
needle for support. The suture was then tied down to anchor and align the device.
Prior to surgery, the device was sterilized via ethylene oxide gas and underwent EIS testing
as described in 2.4.4.2. Immediately after implantation, EIS measurements were obtained while
the device remained in the orbit to assess device robustness and surgical feasibility of the procedure.
2.4.5.3 Chronic Implantation for Metallized Device
The surgical procedures for chronic implantation experiments of metal RRA version 3 were
nearly identical to those in previous RRA version 2 acute implants. Key differences included the
use of sterile 1 × PBS prior to implantation as a control experiment as well as additional post-
operation maintenance. Optical coherence images and in vivo EIS measurements of the
anaesthetized rat were obtained while food and water were provided ad libitum throughout the
duration of the 4-week long study.
45

Results
2.5.1 Thermoforming Results
Thermoforming was successfully achieved in both RRA version 1 or 2 as shown in Figure
2.5-1 as well as RRA version 3 (Figure 2.5-2). The curved length of the device’s stimulating
region was intentionally designed to match the arc length of the upper temporal quadrant only.
Although it was possible to create designs which spanned the entire retinal arc, preliminary
surgical experiments revealed that doing so increased the rates of retinal tearing or detachment.
The upper temporal quadrant also contains the highest density of rat retinal ganglion cells [96].
Additionally, shortening the arc length allowed the MEA and surgical procedure to better
accommodate the large lens which occupies a larger portion of the ocular orbit compared to other
species [97] which is critical because contact with the lens may leave scratches resulting in cataract
formation.
Note that RRA version 2 and 1 incorporated the first and second bend for thermoforming
as described by 2.4.1 whereas the version 3 device no longer required the second bend. This greatly
improved functional electrode yield and slightly improved surgical feasibility and electrode-to-
retina proximity since the suture point lied distal from the thermoformed region.

Figure 2.5-1. Thermoformed RRA version 2 device. Both version 1 and 2 employed a technique where the first
bend matched retinal curvature and the second bend ensured that the array routes back towards the cranium
46

Figure 2.5-2. Thermoformed RRA version 3 device. The “candy cane” hook design obviated the need for the
second bend present in RRA version 1 and 2
2.5.2 Benchtop Testing
2.5.2.1 Crosstalk Testing with Chronic Soaking in 1 × PBS
Unannealed RRA version 1 devices underwent crosstalk testing and Figure 2.5-3
illustrates results from intermittent optical microscopy for an exposed electrode. Metal
delamination was initially observed after 32 hours of soaking and progressively worsened.
However, no delamination was observed for the duration of the experiment on the 6 other
intentionally unexposed electrodes (data not shown). Devices remained inside the custom fixture
throughout the entire duration of the experiment and were never dried prior to electrical crosstalk
measurements.

Figure 2.5-3. Representative electrode delamination for unannealed sputtered Pt metal during crosstalk testing
47

Electrical measurements also corroborated the optical observations. Figure 2.5-4
illustrates an example of crosstalk measurement which was obtained immediately after soaking
began (t = 0 hours). Electrode 5 and 8 were the exposed electrodes and the remaining six electrodes
were unexposed as described in 2.4.4.1. Measurements between unexposed electrode pairs (green)
were virtually 0%, suggesting no crosstalk due to effective electrical insulation. Measurements
where one of the electrodes was an exposed electrode (blue or magenta) were approximately 5%.
This indicated that some finite current passage occurred which may be attributed to the ambient 1
× PBS electrolyte. Measurements between electrode 5 and 8 revealed extremely high crosstalk
values (approximately 90%) which was expected due to the short-circuiting that occurs by
submerging exposed electrodes in conductive electrolyte solution.

Figure 2.5-4. Sample crosstalk measurements obtained at t = 0 hr of soaking

Crosstalk measurements of each color set were averaged plotted across time (Figure
2.5-5). Standard deviations from averaging were omitted as they were consistently below 2%. Data
for E5 and E8 measurement pairs were also omitted as they represented irrelevant short circuits.
All values remained relatively stable across time, indicating that electrolyte ingress into the
Parylene polymer film was not significant within a 1024 hour time span. Chronic in vivo
E1 E2 E3 E4 E5 E6 E7 E8
E1 0.6 0.6 0.58 4.32 0.54 0.64 6.3
E2 -0.08 0.12 0.28 5 0.8 0.3 4.52
E3 0.18 0.34 0.46 3.98 0.7 0.8 5.22
E4 0.82 0.52 0.18 4.4 0.4 0.56 5.6
E5 5.56 5.38 4.6 5.38 6.18 7.36 87.26
E6 0.34 0.72 0.22 0.1 5.64 0.5 7.28
E7 0.24 0.32 0.88 0.7 5.48 0.42 4.3
E8 7 5.64 5.6 7.04 86.94 9.24 5.04
48

experiments were intended to span 1 month at a minimum. Therefore, the lack of significant
increase in crosstalk in this benchtop study indicated that devices implanted in vivo may exhibit
similar robustness.

Figure 2.5-5. Crosstalk test results for electrodes pairs which were completely or partially insulated

2.5.2.2 Electrochemical Impedance Spectroscopy
After fabrication of RRA version 2, devices were packaged with methods described in
2.4.3.1 and then underwent EIS testing. A representative plot of the impedance magnitude and
phase for one unannealed device is displayed in Figure 2.5-6. Note that distinct spectra can be
observed for each electrode size. Since electrodes E, F, G, and H have the same 160 µm diameter,
their EIS data are closely overlapped.
49

Figure 2.5-6. EIS results with (a) magnitude and (B) phase each electrode in an unannealed RRA version 2 device (1 ×
phosphate buffered saline, 25 °C)

Electrochemical properties of the electrodes after post-processing were also investigated.
EIS values from the equally sized electrodes (electrodes E, F, G, and H) were averaged and plotted
for devices that had been unannealed, annealed, thermoformed, and PtIr coated in Figure 2.5-7.
Both two-dimensional annealing and three-dimensional thermoforming caused a shift in
electrochemical impedance spectra relative to their unannealed starting point, but no significant
difference existed between annealed and thermoformed devices. This suggested that thermal
treatment yielded electrochemically similar results regardless of device planarity.

50

Figure 2.5-7. Electrochemical impedance magnitude (a) and phase (b) at different stages of processing on RRA version 2
devices (1 × PBS, 25 °C)

2.5.2.3 Cyclic Voltammetry
The same devices at these various stages of post-processing in section 2.4.4.2 were also
characterized through cyclic voltammetry and the average of all 30 cycles per device condition is
displayed in Figure 2.5-8. PtIr coated arrays exhibited the widest current range in CV plots,
indicating that they possessed the largest charge storage capacity. Between annealed and
thermoformed cases, there was no significant difference. Characteristic peaks and valleys
corresponding to hydrogen adsorption and desorption, double layer charging, and Pt-oxide redox
reactions [48] were present for bare Pt in thermally treated devices (Figure 2.5-8), but not readily
51

observable in the unannealed case. This suggests that electrochemical reactions are successfully
confined to the Pt-electrolyte interface for thermally treated devices, but such electrochemical
signatures are blurred in the unnannealed case due to leaky dielectric contribution from soaked
Parylene

Figure 2.5-8. (a) Cyclic voltammetry at different stages of processing on RRA version 2 devices. (b) Magnified and rescaled
view shows features corresponding to hydrogen adsorption/desorption and oxide redox reactions for annealed and thermoformed
cases (1 × PBS, 25 °C)

2.5.2.4 Charge-Balanced Biphasic Cathodic-First Pulse Testing
The slow sweep rates in cyclic voltammetry are known to present limitations in the context
of characterizing stimulation microelectrodes due to differing time scales present in rapid
stimulation pulses [82]. Therefore, pulse tests with parameters anticipated for in vivo rat
experiments were conducted for 160 µm diameter electrodes (Figure 2.5-9). Unannealed array
data was omitted because devices must be thermoformed prior to implantation in order to obtain
the appropriate three-dimensional structure. The voltage transients of PtIr coated devices remained
well within the water window compared to uncoated devices.

52

Figure 2.5-9. Pulse test results at different stages of processing of RRA version 2 devices. PtIr reduced voltage transients
compared to bare Pt (1 × PBS, 25 °C).
2.5.3 In Vivo Testing
2.5.3.1 Acute and Chronic Sham Implantations
Several surgeries were performed in an iterative manner on enucleated eyes (Figure
2.5-10) as well as in whole animal (Figure 2.5-11) in order to understand the interaction between
the rat anatomy and Parylene device. These experiments with pre-RRA version 1 shams were
critical in obtaining realistic estimates on the required ribbon cable length between the ocular orbit
and cranial location for the PCB adapter. They also revealed that attempting to span the entire
retinal arc with the device increased the rates of retinal tearing or detachment. This discovery
informed the key design decision to span solely the upper temporal quadrant of the retina.

Figure 2.5-10. Representative photograph of enucleated rat eye undergoing sham device fitting tests
53

Figure 2.5-11. Representative photograph of sham device implantation into rat cadaver
Optical coherence tomagraphy (OCT) images (Figure 2.5-12) from a chronic sham
experiment using the geometry of RRA version 1 devices illustrated that the thermoformed
stimulation region closely matched retinal curvature. This experiment was designed with a target
duration of 4 weeks to demonstrate chronic biocompatibility. Since the implant was well-tolerated
without inflammation or significant shifts in location well past the target duration, the experiment
was terminated at week 6.

54

Figure 2.5-12. Optical coherence tomagraphy image of sham array tip cross-section and retina (4 weeks post-surgery).

Minimizing the retina-to-device distance improves stimulation efficiency because electric
field strength attenuates with distance squared [98]. However, limitations existed which precluded
a perfectly flush fit between the retina and array. Unlike the cylindrical dowel pins of the
thermoforming mold, the rat ocular cross-section is not a perfect circle and instead is slightly
oblong [80]. The manual process of surgical attachment also introduced alignment variability.
However, because the epiretinal approach provides direct access to retinal ganglion cells,
electrodes placed up to 1000 µm away from the retina have been reported by others to elicit
electrically evoked responses in the visual cortex [76].
2.5.3.2 Acute Implantation for Metallized Device
After conducting sham device experiments, a metallized RRA version 2 device was
implanted in an acute trial (< 5 hours) with a live rat. Figure 2.5-13 illustrates the proptosing of
an eye using a silk suture hooked under the eyelid in order to provide access to scleral tissue. In
(b), the array was inserted and sutured into the ocular orbit. No notable friction against ocular
tissue and device was reported during the insertion.
55

Figure 2.5-13. (a) The eye was proptosed to enable more access to the sclera then (b) the retinal array was sutured on. (c)
Connection scheme from implanted array to potentiostat

Panel (c) shows the connection scheme between the implanted device and a potentiostat
system which applied EIS to assess electrode viability. The contact pad region of the device was
potted in marine epoxy and a flat flexible cable was fed into a breakout board. Immediately prior
to implantation, EIS was conducted in 1 × PBS and three out of eight electrodes on this retinal
array were deemed functional – meaning that their electrochemical spectra resembled those of
56

benchtop data. Nonfunctional electrodes displayed MΩ or GΩ impedance magnitudes and
approximately -90° impedance phase at higher frequencies. This was indicative of an open circuit
that may have resulted from wire breakage by damage from shipping, packaging, mishandling, or
thermoforming complications.
EIS measurements after surgery while the array was inside the orbit revealed that the three
previously functional electrodes remained functional (Figure 2.5-14). Slightly different spectra
were measured due to the change in electrolyte (1 × phosphate buffered saline before surgery and
rat vitreous humor after surgery) as well as the lack of availability of a Faraday cage in the surgery
suite. This result indicated that no significant device damage was sustained from the implantation
procedure and illustrated the robustness of the device against potential surgical trauma.
57

Figure 2.5-14. EIS before and after surgical implantation. Three out of eight electrodes were functional prior to surgery and
remained functional after implantation.

58

2.5.3.3 Chronic Implantation for Metallized Device
The “candy cane” hook geometry of RRA version 3 devices was well-suited for
implantation as shown in Figure 2.5-15. This time, the device was extruded out of the choroid
through the 12-gauge shuttle needle for improved access to the orbit. After suturing, a flap of
scleral tissue was blanketed over the device’s anchor point and sutured once more to ensure a
smooth ocular surface to minimize abrasion during saccadic eye movement post-operation.

Figure 2.5-15. Surgical view of eye during implantation of RRA version 3 device (a) after device suturing and (b)
after scleral blanketing of the device

However, sometime before sutures was tied, a surgical error was made where the PCB was
held by hand and inadvertently pulled back sharply. The hook was caught in the scleral incision
and was deformed due to the tension from the pulling motion. This caused all 8 electrodes to
become nonfunctional unfortunately. The key lesson to employ a static clamp to hold the PCB was
learned for the next surgery. Nonetheless, in order to investigate the long-term effects of
implantation, the ribbon cable was routed subcutaneously to the PCB assembly lying on the
cranium. 4 bone screws serving as recording electrodes were connected with metal wire to pin
receptacles which were mated to corresponding pin receptacles on the PCB as shown in Figure
2.5-16.
59

Figure 2.5-16. Surgical view of cranium during implantation of RRA version 3 device and associated PCB adapter

After connecting the pin receptacles, bone cement was applied on the cranium to create a
secure headcap as can be viewed in Figure 2.5-17. The Omnetics connector and DIP switch were
easily accessible on the compact packaging scheme. For the target 4 week duration, the rat was
healthy, could roam around freely, and tolerated the system implant well. Although other
researchers occasionally report that rats may scratch off headcaps, this system was secure.
Additionally, no prolonged inflammation of the retina was observed in fundus imaging and OCT
which demonstrated successful system integration. Another device and PCB shipment are en route
to surgical collaborators to re-attempt the experiment with a clamp this time.

60

Figure 2.5-17. Live rat with PCB assembly in post-operative setting
Summary and Future Directions
For the first time, an implantable epiretinal stimulation array for rats was designed,
fabricated, and tested in order to serve as a tool to accelerate research on vision and neuroscience.
This work articulated and overcame several nuances in surgical challenges associated with the rat
retinal model which soft polymer neural interfaces are uniquely positioned to overcome.
Anatomically imposed geometric constraints of the rat retina were addressed via three-
dimensional thermoforming of Parylene and iterative lessons learned through surgical trials that
spanned design conception, sham device testing, and optimized metal device implantations.
Benchtop electrochemical tests of the device revealed sufficient electrical insulation of the
Parylene and that PtIr electroplating demonstrated improved charge storage capacity. In vivo tests
demonstrated chronic biocompatibility for up to 6 weeks with non-metallized sham devices.
Functional devices containing electrodes displayed electrical integrity post-surgery.
Moving forward, additional studies may build upon the foundations laid down by this work.
Two main objectives remain. The first goal would be to achieve successful chronic in vivo
stimulation via RRA devices electroplated in PtIr. This may be supplemented by slight refinements
61

to a potential RRA version 4. For instance, slightly enlarging the suture hole diameters by 5 or 10
µm was suggested by surgical collaborators. Additionally, placing the electrodes slightly closer to
the stimulation tip may ensure electrode-to-retina contact as opposed to undesired electrode-to-
sclera contact. This work intentionally employed non-coated PtIr devices to investigate and
overcome the several challenges present in in vivo implantation in order to accelerate progress and
reduce confounding testing variables which may have hindered successful troubleshooting. In
parallel, benchtop PtIr experiments characterized electrochemical properties and long-term testing
capabilities. Results from these two parallel veins of research portend a promising combination of
retinal ganglion cell stimulation in free roaming rats through electrodes of sufficient charge storage
capacity.
Once successful stimulation with PtIr devices is obtainable, a method of measuring its
chronic effects on the visual cortex is required. Current plans with collaborators entail a
combination of imaging techniques such as magnetic resonance imaging and micro-computed
tomography as well as confirmation via histological techniques such as c-FOS protein
immunohistological detection [99]. These approaches are favorable because no other implantable
components are involved, but the second main goal of this work involves integration of a
complementary visual cortex implant along with the current system. This would enable
measurement of electrically elicited responses in the brain in near real-time, allowing for unique
experiments to be conducted to elucidate the visual pathway in the context of neural signaling in
chronic in vivo experiments. Presently, our research team may only employ and reuse commercial
visual cortex electrocortigraphy arrays (E16-500-5-200-H16, NeuroNexus, Ann Arbor, MI) in
acute settings due to the prohibitive cost of permanently implanting them across multiple animals.
62

Use of bone screws as recording electrodes with signal filtering algorithms is undergoing
development as a potential candidate for chronic recordings. Unfortunately, this has proven
challenging due to stimulation artifacts induced by the wireless headstage firmware and
exacerbated by the low impedance of the bone screws. Therefore, development of a low-cost
Parylene-based visual cortex array which is compatible with the RRA device and PCB system
would be the second future milestone of this work. Proper integration of the visual cortex device
would enable a closed-loop system between stimulation and recording in real-time system.
Experiments and data from such a system in the context of epiretinal rat studies may then drive
progress towards improvement of retinal prosthetic technology to restore visual acuity and
improving quality of life in patients suffering from retinally degenerative conditions.

63

References
[54] W. Wu, W. H. Peters, and M. E. Hammer, “Basic mechanical properties of retina in simple
elongation,” J. Biomech. Eng., vol. 109, no. 1, pp. 65–67, 1987.
[55] G. A. Goetz and P. D.V., “Electronic approaches to restoration of sight,” pp. 1–92, 2016.
[56] G. Liew, M. Michaelides, and C. Bunce, “A comparison of the causes of blindness
certifications in England and Wales in working age adults (16-64 years), 1999-2000 with
2009-2010,” BMJ Open, vol. 4, no. 2, pp. 1–6, 2014.
[57] A. C. Ho et al., “Long-Term Results from an Epiretinal Prosthesis to Restore Sight to the
Blind.,” Nonwovens Ind., vol. 44, no. 1, pp. 36–42, 2013.
[58] J. D. Weiland, S. Member, and M. S. Humayun, “Retinal Prosthesis,” IEEE Trans. Biomed.
Eng., vol. 61, no. 5, pp. 1412–1424, 2014.
[59] R. E. Marc, B. W. Jones, C. B. Watt, and E. Strettoi, “Neural remodeling in retinal
degeneration,” Prog. Retin. Eye Res., vol. 22, no. 5, pp. 607–655, 2003.
[60] L. Da Cruz et al., “The Argus II epiretinal prosthesis system allows letter and word reading
and long-term function in patients with profound vision loss,” Br. J. Ophthalmol., vol. 97,
no. 5, pp. 632–636, 2013.
[61] S. Niketeghad and N. Pouratian, “Brain Machine Interfaces for Vision Restoration: The
Current State of Cortical Visual Prosthetics,” Neurotherapeutics, vol. 16, no. 1, pp. 134–
143, 2019.
[62] C. T. Moritz, “Now is the Critical Time for Engineered Neuroplasticity,” Neurotherapeutics,
vol. 15, no. 3, pp. 628–634, 2018.

64

[63] A. M. Rosa, M. F. Silva, S. Ferreira, J. Murta, and M. Castelo-Branco, “Plasticity in the
human visual cortex: An ophthalmology-based perspective,” Biomed Res. Int., vol. 2013,
2013.
[64] S. Damle, Y. H. Lo, and W. R. Freeman, “High Visual Acuity Retinal Prosthesis:
Understanding Limitations and Advancements Toward Functional Prosthetic Vision,”
Retina, vol. 37, no. 8, pp. 1423–1427, 2017.
[65] M. Jörg-Uwe, “Retina implant—a bioMEMS challenge,” Sensors Actuators A Phys., vol.
97–98, no. 0, pp. 1–9, 2002.
[66] L. Hesse, T. Schanze, M. Wilms, and M. Eger, “Implantation of retina stimulation
electrodes and recording of electrical stimulation responses in the visual cortex of the cat,”
Graefe’s Arch. Clin. Exp. Ophthalmol., vol. 238, no. 10, pp. 840–845, 2000.
[67] H. Gerding, F. P. Benner, and S. Taneri, “Experimental implantation of epiretinal retina
implants (EPI-RET) with an IOL-type receiver unit,” J. Neural Eng., vol. 4, no. 1, 2007.
[68] D. Güven et al., “Long-term stimulation by active epiretinal implants in normal and RCD1
dogs,” J. Neural Eng., vol. 2, no. 1, 2005.
[69] A. E. H. Koh et al., “Retinal degeneration rat model: A study on the structural and functional
changes in the retina following injection of sodium iodate,” J. Photochem. Photobiol. B
Biol., vol. 196, no. February, p. 111514, 2019.
[70] E. L. Fletcher, A. I. Jobling, K. A. Vessey, C. Luu, R. H. Guymer, and P. N. Baird, Animal
models of retinal disease, vol. 100. 2011.
[71] D. R. Bertschinger et al., “A review of in vivo animal studies in retinal prosthesis research,”
Graefe’s Arch. Clin. Exp. Ophthalmol., vol. 246, no. 11, pp. 1505–1517, 2008.

65

[72] K. Mathieson et al., “Photovoltaic retinal prosthesis with high pixel density,” Nat. Photonics,
vol. 6, no. 6, pp. 391–397, 2012.
[73] V. Kansara, L. Muya, C. Wan, and T. A. Ciulla, “Suprachoroidal Delivery of Viral and
Nonviral Gene Therapy for Retinal Diseases,” J. Ocul. Pharmacol. Ther., vol. 00, no. 00,
pp. 1–9, 2020.
[74] M. T. Pardue et al., “Possible sources of neuroprotection following subretinal silicon chip
implantation in RCS rats,” J. Neural Eng., vol. 2, no. 1, 2005.
[75] V. P. Gabel, Artificial Vision. 2017.
[76] C. De Balthasar et al., “Factors affecting perceptual thresholds in epiretinal prostheses,”
Investig. Ophthalmol. Vis. Sci., vol. 49, no. 6, pp. 2303–2314, 2008.
[77] A. T. Chuang, C. E. Margo, and P. B. Greenberg, “Retinal implants : a systematic review,”
pp. 852–856, 2014.
[78] K. Nimmagadda and J. D. Weiland, “Retinotopic Responses in the Visual Cortex Elicited
by Epiretinal Electrical Stimulation in Normal and Retinal Degenerate Rats,” vol. 7, no. 5,
2018.
[79] L. L. H. Chan, E.-J. Lee, M. S. Humayun, and J. D. Weiland, “Both electrical stimulation
thresholds and SMI-32-immunoreactive retinal ganglion cell density correlate with age in
S334ter line 3 rat retina,” J. Neurophysiol., vol. 105, no. 6, pp. 2687–2697, 2011.
[80] D. C. Lozano and M. D. Twa, “Development of a rat schematic eye from in vivo biometry
and the correction of lateral magnification in SD-OCT imaging,” Investig. Ophthalmol. Vis.
Sci., vol. 54, no. 9, pp. 6446–6455, 2013.

66

[81] D. J. Wallace, D. S. Greenberg, J. Sawinski, S. Rulla, G. Notaro, and J. N. D. Kerr, “Rats
maintain an overhead binocular field at the expense of constant fusion,” Nature, vol. 498,
no. 7452, pp. 65–69, 2013.
[82] S. F. Cogan, S. Hara, and K. A. Ludwig, The Safe Delivery of Electrical Currents and
Neuromodulation, Second Edi. Elsevier Ltd, 2018.
[83] J. Danias et al., “Cytoarchitecture of the retinal ganglion cells in the rat,” Investig.
Ophthalmol. Vis. Sci., vol. 43, no. 3, pp. 587–594, 2002.
[84] and E. M. B. J. Kim, B. Chen, M. Gupta, “THREE DIMENSIONAL
TRANSFORMATION OF PARYLENE THIN FILM STRUCTURES VIA
THERMOFORMING.”
[85] P. Microelectrodes, S. F. Cogan, P. R. Troyk, J. Ehrlich, and T. D. Plante, “In vitro
comparison of the charge-injection limits of activated iridium oxide (AIROF) and platinum-
iridium microelectrodes,” In Vitro, vol. 52, no. 9, pp. 1612–1614, 2005.
[86] S. Shah, A. Hines, D. Zhou, R. J. Greenberg, M. S. Humayun, and J. D. Weiland, “Electrical
properties of retinal – electrode interface,” vol. 4, pp. 1–6, 2007.
[87] S. A. Hara, B. J. Kim, J. T. W. Kuo, and E. Meng, “An Electrochemical Investigation of the
Impact of Microfabrication Techniques on Polymer-Based Microelectrode Neural
Interfaces,” J. Microelectromechanical Syst., vol. 24, no. 4, pp. 801–809, 2015.
[88] A. Burgess, K. Shah, O. Hough, and K. Hynynen, “Rat Optic Nerve Head Anatomy within
3D Histomorphometric Reconstructions of Normal Control Eyes,” pp. 477–491, 2016.
[89] C. D. Lee and E. Meng, “Mechanical Properties of Thin-Film Parylene–Metal–Parylene
Devices,” Front. Mech. Eng., vol. 1, no. September, pp. 1–14, 2015.

67

[90] A. Petrossians, J. J. Whalen, J. D. Weiland, and F. Mansfeld, “Electrodeposition and
Characterization of Thin-Film Platinum-Iridium Alloys for Biological Interfaces [J.
Electrochem. Soc., 158, D269 (2011)],” J. Electrochem. Soc., vol. 158, no. 6, p. S15, 2011.
[91] C. D. Lee, E. M. Hudak, J. J. Whalen, A. Petrossians, and J. D. Weiland, “Low-Impedance,
High Surface Area Pt-Ir Electrodeposited on Cochlear Implant Electrodes,” J. Electrochem.
Soc., vol. 165, no. 12, pp. G3015–G3017, 2018.
[92] A. Petrossians, “Electrodeposition of Platinum-Iridium Coatings and Nanowires for
Neurostimulating Applications: Fabrication, Characterization and in-Vivo Retinal
Stimulation/Recording Eis Studies of Hexavalent and Trivalent Chromium Based Military
Coating Systems,” no. May, 2012.
[93] C. A. Gutierrez, C. Lee, B. Kim, and E. Meng, “Epoxy-less packaging methods for electrical
contact to parylene-based flat flexible cables,” 2011 16th Int. Solid-State Sensors, Actuators
Microsystems Conf. TRANSDUCERS’11, pp. 2299–2302, 2011.
[94] T. A. Silue, S. Mannikanti, and N. Peixoto, “Electrochemical Characterization of Vitreous
Humor,” Procedia Technol., vol. 27, pp. 301–303, 2017.
[95] E. M. Hudak, “Platinum for neural stimulation : voltammetry considerations,” 2010.
[96] M. Salinas-Navarro et al., “A computerized analysis of the entire retinal ganglion cell
population and its spatial distribution in adult rats,” Vision Res., vol. 49, no. 1, pp. 115–126,
2009.
[97] S. M. Thomasy, J. S. Eaton, M. J. Timberlake, P. E. Miller, S. Matsumoto, and C. J. Murphy,
“Species differences in the geometry of the anterior segment differentially affect anterior
chamber cell scoring systems in laboratory animals,” J. Ocul. Pharmacol. Ther., vol. 32, no.
1, pp. 28–37, 2016.
68

[98] A. Ray, L. L. H. Chan, A. Gonzalez, M. S. Humayun, and J. D. Weiland, “Impedance as a
method to sense proximity at the electrode-retina interface,” IEEE Trans. Neural Syst.
Rehabil. Eng., vol. 19, no. 6, pp. 696–699, 2011.
[99] A. S. Perrin-Terrin et al., “The c-FOS protein immunohistological detection: A useful tool
as a marker of central pathways involved in specific physiological responses in vivo and ex
vivo,” J. Vis. Exp., vol. 2016, no. 110, pp. 1–9, 2016.

69

Background
3.1.1 Implantable Pressure Sensing
A complex and interrelated system of organs work in concert to maintain homeostasis in
the human body on a variety of fronts. One of these critical parameters is physiological pressure.
Figure 3.1-1 graphically illustrates major pressure systems located in the human body which
include but are not limited to intraocular pressure (IOP), intracranial pressure (ICP), inner ear
pressure, joint pressure, pleural pressure, cardiac pressure, bladder pressure, and gastrointestinal
pressure.
These pressure systems are often categorized into subsystems. For example, cardiovascular
pressure distinguishes between systolic and diastolic pressure. More nuanced breakdowns exist
such as right ventricular pressure, left arterial pressure, pulmonary artery pressure, or capillary
blood pressure at the arterial or venous end. Additionally, pressure systems of different organs
may be interrelated. For instance, cerebral perfusion pressure (CPP) is the difference between
mean arterial pressure (MAP) and ICP, as shown in equation (1). CPP drives oxygen delivery to
brain tissue and is a key metric in traumatic brain injury.
𝐶𝑃𝑃 = 𝑀𝐴𝑃 − 𝐼𝐶𝑃 ()

MICROBUBBLE-BASED PRESSURE TRANSDUCER DEVELOPMENT
FOR INTRACRANIAL PRESSURE MEASUREMENT
70

Figure 3.1-1. Illustration of several major physiological pressure systems

Various physiological pressure systems maintain a dynamic balance with respect to each
other as well as other perturbations such as gravity, muscle flexion, or atmospheric pressure. They
operate in several frequency regimes with varying minimum and maximum ranges [100].
Physiological pressure may even be negative with respect to atmospheric pressure as with
71

intrapleural pressure during inspiration reaching values as low as -13 mmHg [101]. In addition,
extremely high values such as 38,000 mmHg are attainable between vertebrae in the spine
However, this value occurs during improper exercise and should be avoided.
Unfortunately, these delicate balances may be disturbed through injury or disease, leading
to adverse health effects or death if left untreated. Table 3.1-1 lists several examples of medical
conditions related to body pressure system to illustrate the wide variety of physiological pressure-
related afflictions. Although different conditions have solutions in varying degrees of satisfactory
treatment today, diseases where pressure dysregulation occurs on a chronic timescale or only
intermittently are uniquely sinister. This is because pressure measurements are difficult to interpret
due to their traditionally sparse measurement methods in the clinic. For instance, patients typically
obtain blood pressure measurements during occasional checkups. IOP measurements are also are
only recorded during visits to an ophthalmologist.

72

Unlike these discrete point measurements, continuous pressure monitoring may be
extremely helpful in the diagnoses and treatment of certain diseases. The latter provides a larger
and richer dataset which may be effectively leveraged by the recent rise of big data analytics,
evidence based healthcare, and artificial intelligence techniques [39]. Capturing trends with respect
to time can facilitate preventative care rather than resorting to reactive treatment schemes. This
may also prevent the “white coat” effect [40] in which measurements are obscured due to patient
stress associated with physical visits to a clinic.
To this end, implantable pressure sensors are especially befitting. Implantation may enable
consistent and automatic measurements in outpatient settings to obviate or heavily reduce clinic
visitations. Such devices may therefore assist in increasing patient compliance to long term
treatment schemes as well [41]. Presently, implantable devices are not widely deployed clinically
Condition System Brief Description
Glaucoma Intraocular

Chronically high IOP causes damage to the optic nerve
Traumatic Brain Injury Intracranial Sudden blow to head causes brain swelling and increased ICP
Hydrocephalus Intracranial Chronically high ICP causes damage to the brain
Stroke Cardiac High cardiac pressure may cause clot and stop brain blood flow
Coronary Artery Disease Cardiac Arteries become narrower and will increase cardiac pressure
Emphysema Intrapleural Damaged alveoli induce more negative intrapleural pressure
Irritable Bowel Syndrome Gastrointestinal Bowel dysfunction may cause bloating and higher GI pressure
Urinary Incontinence Bladder High bladder pressure may cause involuntary leakage of urine
Arthritis Joint Joint pressure may increase pain from inflammation in joints

Table 3.1-1. Medical conditions affecting various physiological pressure systems
73

because many technical issues are yet to be solved. Active and ongoing research into development
of implantable pressure sensors exist because many challenges must be tackled before the safety
and efficacy outweighs device usage risk. Such challenges include device accuracy, resolution,
resistance to drift, hermetic sealing, biocompatibility, reliability and a host of other engineering
considerations [42]. Due to their small form factor, MEMS-related technologies have great
potential to aid in treatment and diagnosis of these conditions. A review of existing technologies
is presented for major pressure systems to demonstrate the breadth of potential to augment
healthcare and to illustrate how this work contributes to the overall body of work in physiological
pressure sensing.

3.1.1.1 Cardiovascular Pressure
Cardiovascular pressure may be the most well-known physiological pressure system due
to its widespread prevalence in society. Systolic and diastolic blood pressure measurements with
sphygmomanometer “cuffs” are commonly administered in routine checkups but such point
measurements are only acquired at discrete time intervals. They may also contain “masked
hypertension” due to the white coat effect [102]. Continuous monitoring through implantable
devices may spell the difference between life and death in certain applications due to the critical
dependency of all organs on blood supply. MEMS advancements in miniaturization, sensing
capabilities, and telemetry have fostered development of a host of implantable devices today that
exist in a variety of stages of development [103] to with treatment of hypertension, coronary artery
disease, abdominal aortic aneurysms, heart failure, autonomic dysreflexia, and more [103].
Two noteworthy technologies related to abdominal aortic aneurysms (AAA) have been
developed through CardioMEMS and Remon Medical Technologies. Both have been acquired by
74

St. Jude Medical in 2014 and Boston Scientific in 2007; respectively. AAA is a dilation of aorta
near the abdomen with an 80% mortality rate if the aorta ruptures [104]. Endovascular repair of
abdominal aortic aneurysms (EVAR) is a procedure to treat AAA by placing a stent graft in the
aorta to facilitate flow through the aneurysm. EVAR’s short recovery time makes it a popular
alternative over the traditional gold standard of open heart surgery. However, follow up
complications with EVAR are difficult to detect [105]. Implantable sensors have enabled frequent
pulmonary artery pressure measurements to quickly alert patients to receive interventional
treatment if such complications do arise.
The CardioMEMS HF system received premarket approval (PMA) in 2014 and the
fundamental operating mechanism employed wireless LC resonance shifts to measure pressure
[106]. The design consists of two small flexible plates with inductor windings and distributed
capacitive structures in a hermetically sealed rigid package. The plates define a diaphragm which
will deflect under physiologically relevant pressures. Deflections cause a shift in the passive LC
circuit’s resonant frequency which may be interrogated wirelessly through an external magnetic
loop [107]. This enabled wireless and battery-less operation with performance specifications and
robustness suitable for endoleak detection throughout a patient’s lifetime.
The ImPressure from Remon Medical Technologies is also suited for AAA treatment but
employed a different mechanism. Surface acoustic waves (SAW) used ultrasound phenomena to
transmit pressure measurements acquired by a membrane-based sensor. SAW resonators are very
suitable for harsh environments and remote measurement due to their passive and wireless
operating mode [108]. The ultrasonic pressure transmission technology underwent clinical trials
with no serious adverse effects. Measurements were nearly identical to those obtained
simultaneously by the golden standard Millar catheters [109].
75

Both CardioMEMS and Remon Medical Technologies, however, use analog signal
transmission which may be considered a drawback since identifier transmissions or check-sums
cannot verify the transferred data [110]. To address this possible weakness, the ENDOCOM
project [111] is an example of the common approach of adapting application specific integrated
circuits (ASIC) with commercial pressure sensors to realize a more digitally relevant system for
cardiovascular monitoring [110]. Other clinical applications such as hypertension [112] or arterial
stenosis [53,54] have been tackled by Ken Wise’s group at the University of Michigan and have
followed this approach. However, the pressure sensing element is typically a piezoresistive
membrane [113] or monolithically integrated capacitive pressure sensor [56,57], which may be
prone to biofouling or signal quality reduction over the duration of use. Although Young [114]
investigated the effect of coating such standard implantable blood pressure sensors with Parylene
C or PDMS to improve long-term robustness, a thorough discussion on the loss of sensor
performance induced by the coatings was not provided.
Recently, several alternative blood pressure measurement methods are also undergoing
investigation on the benchtop. Hydraulic motion of embedded liquid (deionized water) was used
to transduce pressure in sensor designed to fit in a coiled cardiac stent [115]. The LC resonance
technique was adapted for a flexible Parylene C substrate to enable folding into a stent that is
suitable for a catheter-based delivery system [116]. Another technique involves the pressure-
dependent displacement of radiopaque fluid (Isovue-370) in micro-reservoir into a microfluidic
channel. The distance that the fluid traveled could be detected by noninvasive periodic x-rays and
quantified with micromachined tick marks which serve as a pressure gauge [117], [118]. Lastly, a
unique design implementing capillary action to generate a microbubble for pressure measurement
has undergone early investigation [119].
76

3.1.1.2 Gastrointestinal Pressure
Implantable pressure monitoring has also been employed in the gastrointestinal tract better to
understand gastric motility and regulation of appetite and hunger [120]. Development in this area
traces back to the 1960s when terms such as “radio pill”, “radio telemetering capsule”, or
“Endoradiosonde” were first coined [121]. The vast majority of GI implantable pressure sensors
featured an easy-to-swallow pill form factor. An elastic diaphragm served as the pressure sensing
element and analog electronics powered by a small battery amplified and transmitted signals in the
radio frequency range to external receivers [122]. Improvements in design largely focused on the
signal transmission scheme [123] although gastric juices had been observed to damage and
interfere with the pressure sensitive diaphragm [121]. Most research in implantable GI pressure
monitoring bloomed in previous decades, but new advancements do exist. For instance, a passive
LC resonant technique was recently employed for the GI system [124] and in 2017, the FDA
approved the first drug where an ingestible pressure sensor embedded in a pill assists in recording
that the medication was taken to track patient compliance to the drug regimen.
3.1.1.3 Bladder Pressure
Bladder pressure could also be measured by implantable devices to aid in urodynamic
testing to treat patients with urinary incontinence, urinary tract dysfunction, or voiding
abnormalities [125]. Presently, the gold standard for measuring bladder pressure is through an
intraurethral catheter. However, this process is painful, prone to infection, and infeasible
chronically [100]. Development of discrete wireless implantable sensors could allow for long-term
monitoring while maintaining the patient’s quality of life [126].
77

No FDA approved implantable device could be found for bladder pressure, but there are
many research efforts in this field. The “bladder pill” is an approximately 5 mm by 30 mm trans-
urethral implanted device. It used inductive coupling for wireless powering and data transmission
and is composed of an assembly of off-the-shelf parts coated in polydimethylsiloxane [127].
Recent tests in awake mini-pig demonstrated cystometric pressure measurements comparable to
that of a standard catheter measurement [128]. Another effort from the SINTEF group featured a
single crystal MEMS piezoresistive pressure sensing element designed specifically for in vivo
applications due to its small size (820 µm × 1820 µm; fitted on a 1.2 mm diameter probe tip) and
protection towards bodily fluids [129], [130]. A silicon diaphragm with 4 p-type piezo resistors
form a Wheatstone bridge circuit faces a vacuum reference cavity and hence away from the harsh
physiological environment [131]. The sensor was tested in a human patient with a percutaneous
insertion although standalone implanted formats are the eventual goal [132]. The wireless
implantable intracavity micromanometer (WIMM) was initiated in 2009 by Fletter et al. The
WIMM uses custom circuitry realized in an ASIC and is small enough to dwell beneath the bladder
mucosa and is designed to be part of a closed-loop neuromodulation system [133]. In vivo tests in
5 large animal models (female Jersey calves) have been conducted but unfortunately pressure
measurements were confounded due to the complications regarding the implantation location [134].
3.1.1.4 Intraocular Pressure
Glaucoma is a highly prevalent and incurable medical condition where increased
intraocular pressure (IOP) damages the optic nerve and eventually causes blindness [135]. Because
the disease’s slow progression makes it difficult to detect, much research has been conducted into
designing suitable implantable IOP sensors [136].
78

The Sensimed Triggerfish Contact Lens Sensor (CLS) may be the most well-known project
in this area. Having received FDA approval in 2017, the Swiss technology is composed of two
strain gauges, a microprocessor, and antennae embedded into a silicone soft contact lens designed
to be worn for 24 hours. The strain gauges detect changes in corneal shape which is highly
correlated to IOP [137]. Power and data are transmitted wirelessly to an antenna which is attached
with adhesive on the patient’s face. Relative changes in IOP can be measured and clinical trials
demonstrated that the Triggerfish CLS is safe, well-tolerated, and produces reproducible results
[138]. Chen et al. adapted the contact lens shape but employed an inductive coil instead of strain
gauges to detect changes in corneal shape. This allowed the lens to be thinner than the Triggerfish
and IOP could be correlated with shifts in LC resonance for benchtop tests conducted on a silicone
model of the eye [139].
The contact lens technique bears the advantage of being non-invasive and well-tolerated,
yet invasive methods may yield higher more accuracy. Chitnis et al. demonstrated a minimally
invasive approach where a small needle punctured the sclera to access the vitreous humor and
measured the IOP with a capacitive pressure sensor. Data was outputted wirelessly via transponder.
The technique demonstrated 1 mmHg resolution and was viable in rabbits for 1 month [140]. The
ARGOS group developed a fully implantable telemetric IOP sensor featuring 8 capacitive pressure
sensors with an ASIC and antennae encapsulated in silicone. A reader unit could power and receive
signals from the implant. Six patients were implanted with ARGOS devices in a 1 year long single-
center clinical trial. The devices were well-tolerated and yielded accurate measurements when
compared to those of classic Goldmann applanation tonometry [141].
IOP monitoring implants based off thin film Parylene substrates have also undergone
extensive research. Several design iterations exist [142] but many leveraged the highly
79

biocompatible properties of Parylene C and its amenability to surface micromachining techniques
to construct a passive LC tank circuit [28]. Pressure-induced deflection of a Parylene membrane
embedded with thin-film metal coils caused shifts in resonance frequency which were detected by
the impedance phase dip technique [143]. The sensor technology was tested in vivo for 6 months
in rabbits to demonstrate robust fixation and long-term biocompatibility. IOP measurements were
unattainable due to experimental setup limitations [140]. In addition to purely monitoring,
Parylene-based microvalves have been designed to assist in eye fluid drainage to regulate IOP
[144].
3.1.1.5 Intracranial Pressure
Intracranial pressure monitoring can be used to assist in the treatment of patients with a
variety of neurological disorders [145], but hydrocephalus is a particularly deadly chronic
condition where implantable sensors may be extremely pertinent [100]. The incurable disease is
the leading cause of brain surgery for children and imposes a $2 B/yr burden on the US healthcare
system. Increased ICP causes brain damage and eventually death in patients unless treated by
implantation of a ventricular shunt to drain cerebrospinal fluid to other regions of the body.
However, shunt failure due to clogging is very common and will cause ICP to rise lethally again.
Unfortunately, symptoms of shunt failure are difficult to distinguish from ordinary headaches or
nausea. Chronically monitoring ICP could be extremely helpful in understanding precisely when
a shunt revision surgery is required to replace the clogged shunt.
The clinical golden standard for ICP measurement is the extraventricular drain (EVD) and
manometer measurements. For EVDs, a flexible plastic catheter is placed by a neurosurgeon into
to brain to allow excess cerebrospinal fluid to drain out into a collection bag. However, such
measurements can only be performed acutely because patients have restricted mobility when
80

installed with an EVD. Risk of infection also exists. In the early 1990s, commercial groups
developed ICP microsensors for use when drainage is unnecessary or cannot be performed safely.
The Camino sensor transduced pressure by measuring the change in amount of light reflected from
a deflectable membrane. The Codman Microsensor ICP transducer operated through a strain gauge
mounted in a titanium case and the Spiegelberg used an air-filled balloon catheter to measure ICP
[146]. However, these sensors may suffer from zero drift after several days, whereas such
measurement error in an implantable device would ideally be negligible during lifetime of a patient.
Therefore, much work has gone into developing implantable ICP sensors.
For instance, materials such as polyvinylidene fluoride trifluoroethylene [147], carbon
nanotubes [148] and graphene [149] [150] have been investigated to serve as the pressure sensing
element membrane, but have yet to demonstrate chronic in vivo testing. In addition, substrate
materials such as polyimide, in the form of Kapton film, were investigated for ICP sensors
employing a polysilicon film as the pressure sensing element with aluminum interconnects. Due
to its flexibility, the sensor array was able to be rolled around a catheter [151]. Since ICP may be
dependent on orientation in space as well as body motions, a wireless multi-sensor system
comprising of pressure sensing elements and accelerometers revealed that acceleration affected
ICP more than the amplitudes of physiological ICP characteristics. The system was also able to
demonstrate viability in live porcine models for two weeks [152].
Other groups have explored methods to counter the zero drift in the context of ICP
monitoring. An air pouch method was applied where the pressure sensitive element lies away from
the CSF in order to avoid biofouling effects that cause drift. Instead, the pouch in the ventricular
space transmitted ICP to the sensor [153], [154]. Another technique involved a Parylene-oil-
encapsulation scheme to minimize drift [155]. Leung et. al employed an independent mechanical
81

switch involving a touchdown event. As a pressure sensitive diagram experienced more deflection,
only the center, and the most severely deflected region, touched down onto a micropillar. This
allowed for additional calculations and data processing techniques to reduce sensor drift [156].
3.1.2 Microbubble-based Pressure Sensing
The majority of sensing techniques in the previous section relied on a membrane or
diaphragm structure for the pressure transduction. Historically, this scheme arose from tire sensors
in the World War 2 era [110]. However, this technique may suffer biofouling, fluid ingress, short-
circuiting concerns, ion permeation, and more was because it was never intended for in vivo
environments [100]. Therefore, a new pressure transduction scheme may be more suitable in vivo.
The electrolytic bubble-based technique in this work is a promising alternative offering
unique advantages for a variety of applications. Examples include the ability to generate bubbles
on-demand and the potential to employ bubbles as sensing or actuating elements within devices
[52], [157]. Unlike traditional microfluidic bubble-based systems, the need for cumbersome tubing
for gas flow is obviated and moving parts are not required. These benefits have led to the
development of several micromachined pumps [158], sensors [159], actuators [30] and other
devices [160], [161] harnessing the utility of electrolytically generated micro- and nano-scale
bubbles. This work is one such extension of bubble-usage in the context of pressure sensing and
the project bears the research lineage described below.
Found in many elementary physics textbooks, the PV = nRT principle is a well-known
expansion of Boyle’s law, published in 1662, which relates the pressure and volume of a gas via
P1V1=P2V2. The use of bubbles as potential pressure sensing elements may be traced back to Lord
Raleigh in 1917 [162] who described the pressures developed during collapse of a bubble.
Prosperetti and Plesset also made wide contributions by building upon the theoretical foundations
82

of bubble dynamics laid out by Raleigh [163] as well as Ooi and Acosta. In the 1980s and 1990s,
Ran and Katz et al. of Johns Hopkins University built upon the same theoretical equations with
empirical measurements by using an inline holocamera and custom-built chambers to image
microbubbles subjected to various in a column of liquid [164]. Use of MEMS structures and
electrical impedance to monitor bubble size to transduce pressure was demonstrated by Ateya of
CUNY Buffalo in the early 2000s [51].
In the following decade, efforts to harness the phenomena for intracranial pressure
monitoring through a Parylene-based implantable device were pushed forward mainly through
Gutierrez and Yu in the Meng Laboratory at the University of Southern California. Initial designs
arose from impedimetric force sensing with a fluid filled chamber where mechanical chamber
deflections would alter the ionic conduction pathway [165], [166]. Later, a microbubble was
generated between the sensing leads to transduce pressure [159]. This unique principle harnessed
the liquid environment rather than being hindered by it like membrane-based systems. Hermetic
sealing may no longer be required, and bubbles may be immune to biofouling as they are generated
on demand and inherently born out of the ambient fluid. Using Parylene C as the exclusive
substrate material reduced complexity, provided mechanical flexibility, and added de-facto
biocompatibility to the system.
This microbubble pressure transducer (µBPT) system originally employed a circular
chamber for the bubble from Gutierrez et. al, but Yu et. al later adapted the rectangular geometry
from Ateya [52]. This was hypothesized to allow for thin fluid capillaries in the corners of the
channel because the gas bubble was hypothesized to maintain a longitudinally ellipsoidal geometry.
The capillaries were assumed to be modelled as simple wires to approximate the electrochemical
impedance through the bubble.
83

Several different metrics were used to transduce pressure. One example is the change in
impedance from a baseline state – where no bubble exists – to the maximum impedance when the
bubble is generated. Dissolution rate as tracked by impedance was also found to be correlated to
pressure. The bubble impedance was also found to track pressure continuously in real-time for
other design iterations.
However, standards published by the Association for the Advancement of Medical
Instrumentation (ANSI/AAMI NS28:1988 (R2015)) mandate an accuracy of ±2 mmHg in the 0 to
20 mmHg range for use in ICP sensing. Yu and Gutierrez laid the groundwork for impedimetric
microbubble-based pressure sensing, but the target device performance specifications for
hydrocephalus patient use could not be obtained. Repeatability was also difficult to achieve in
many types of device measurements. In this work, theoretical study and empirical experimentation
was carried out in numerous experiments to establish a deeper fundamental understanding of the
underlying mechanisms of various processes related to pressure transduction in the sensor. This
inquiry enabled further experiments to systematically address improvement of device
performance.
Approach and Goals
The goal of the experiments in this work was to further elucidate fundamental operating
mechanisms present in the microbubble-based pressure transducer (µBPT) in order to inform
design decision to achieve device performance that satisfied AAMI standards. Unlike the previous
chapter, the µBPT project followed a very nonlinear progression of results. Therefore, the several
experiments conducted throughout this work were thematically, not chronologically, grouped into
three main bodies – bubble generation, bubble behavior, and bubble measurement.
84

Experiments within each body serve as vignettes to illustrate key nuances that supported
the corresponding subtopic. Each body concludes with a discussion to synthesize key findings and
deeper implications. Then, a final discussion on the overall work is provided to summarize main
results and provide guidance on potential future directions.
Experimental Methods
The methods described in this section were generally used throughout the entirety of this
work. Certain experiments required modified protocols or additional materials. For those cases,
the changes are described on a per-experiment basis in ensuing sections. It may be assumed that
the following methods were employed otherwise.
3.3.1 Design, Fabrication, and Packaging
Device design was adapted from previous work [167] and can be generally described as a
hollow microfluidic structure containing embedded electrodes (Figure 3.3-1). Numerous
geometric variations were drawn via AutoCAD software and fabricated across several mask layout
iterations and designs are denoted by a letter or number ID throughout this work. Their specific
designs may be found in appendices and are stored on the laboratory server. The microfabrication
process flow remained the same for all devices. The device was constructed from thin-film
platinum (Pt) sandwiched between 2 layers of thin-film Parylene C (herein referred to as Parylene).
The Pt was patterned into four 50 µm wide traces via standard photolithography techniques. These
insulated traces emerge as exposed tips in the inner lumen of the Parylene microchannel as well as
the hollow cavity at the microchannel’s midpoint. This cavity, or nucleation core, allowed for
bubble nucleation, growth, and subsequent expansion into the microchannel. The distal ends of the
microchannel have etched openings to allow electrolyte solution to enter into the microchannel.
85

Figure 3.3-1. (a) Micrograph and (b) schematic illustrations of device where Parylene (blue) is opaque or translucent
in order to highlight hollow and/or embedded components

The device was designed to operate by following the sequence presented in Figure 3.3-2.
Initially, a bubble is generated inside the nucleation core by passing a sufficiently large current
between E2 and E3. This induces electrolysis to convert surrounding liquid H2O into H2 or O2 gas.
The bubble then grows and expands into the channel region. Once sufficiently large enough, the
current flow between E2 and E3 is terminated, allowing the bubble to dissolve. In order to
minimize surface energy, the gas in the nucleation core region dissolves first, effectively causing
the bubble to detach from the nucleation core and lie in the channel only. Dissolution continues
until the bubble is completely dissolved. Throughout the entire sequence, the electrochemical
impedance at a high frequency is measured between E1 and E4.
86

Figure 3.3-2. Schematic illustration of ideal µBPT operating sequence

Fluid capillaries on the cross-sectional corners of the channel region (Figure 3.3-3) were
conjectured to conduct electrochemical current which could be measured through high frequency
(>10 kHz) electrochemical impedance. Since bubble resonant frequencies are on the order of kHz
[164], ambient physiological pressure may be assumed to modulate the bubble volume
87

instantaneously. This would consequently modulate the fluid capillary length which served as a
variable resistor, effectively rendering electrochemical impedance as a valid proxy measurement
for pressure transduction.

Figure 3.3-3. Schematic illustration of fluid capillaries (a) dotted line represents cross-section of interest (b) cross-
sectional view (c) stripped view with fluid capillaries only

The main steps of the microfabrication process flow are illustrated in and based on
fabrication processes employed in past work [52] [167]. Briefly, 10 µm of Parylene was deposited
on a silicon carrier wafer (PDS Labcoter 2010, Specialty Coating Systems, Indianapolis, IN). AZ
5214 photoresist (Integrated Micro Materials, Argyle, TX) was then spun on to define metal
features in preparation for lift-off. An O2 plasma descum (60 s, 100 W, 100 mtorr) was performed
immediately prior to electron beam evaporation (CHA Mark 40, CHA Industries, Fremont, CA)
of 200 nm of 99.99% Pt (PraxAir Inc., Danbury, CT). Special care was taken to prevent Pt cracking
due to the thermal expansion mismatch with Parylene. Without breaking vacuum, four depositions
measuring 50 nm with 30-minute pauses in between each step were performed in a tool having a
88

long throw distance (55 cm) [38]. Metal liftoff was achieved through 10-minute long sequential
baths of acetone, isopropyl alcohol, and deionized water at room temperature.

Figure 3.3-4. Microfabrication process flow for the µBPT. The A-A’ cross section of interest runs along the
longitudinal length of the device

Next, 16 µm of AZ P4620 (Integrated Micro Materials, Argyle, TX) was spun on to define
the dimensions of the inner lumen of the microchannel and nucleation core by serving as sacrificial
photoresist. Hard baking was intentionally omitted in order to prevent photoresist reflow which
would alter the microchannel cross section. The same O2 plasma descum was applied and 10 µm
of additional Parylene was then deposited. Another layer of AZ P4620 was spun on to serve as an
etch mask in the ensuing deep reactive ion etch step. The mask selectively protected the wafer to
89

define the perimeter of each testing die, distal openings of the microchannel, and device contact
pads. At those regions, the top 10 µm of Parylene was removed by a modified Bosch process which
alternated between fluoropolymer deposition (C4F8) and inductively coupled O2 plasma etching
[168] (Oxford Plasmalab 100, Oxford Plasma Technology, UK).
With the aid of a microscope, a razor blade was used to manually separate each die using
the etched outline as a guide. Dies were then carefully peeled off the wafer with fine tweezers.
Stripping of the sacrificial photoresist inside the hollow microfluidic structures required 2-hour
long sequential baths of acetone, isopropyl alcohol and deionized water with agitation due to
diffusion-limited dissolution. After drying, devices were annealed for 48 hours at 200 °C in a
vacuum oven (TVO-2, Cascade TEK, Irving, TX) which was purged 3× with N2 gas [169]. Dies
were mated at the contact pads with zero insertion force connectors (Hirose Electric Co., Ltd,
Tokyo, Japan) which were soldered onto flat flexible cables for electrical access to downstream
testing equipment. A polyethylene terephthalate (PET) backing provided the necessary thickness
(~200 µm) for the connector to engage its closing latch.
3.3.1.1 Note on Fluidic Opening Etch
Due to vertical bias in the deep reactive ion etching (DRIE) process previous design
iterations suffered from unwanted artifacts at the fluidic ports. These shell-like structures were
unintentionally created from the passivation layer preceding the O2 etching step in the DRIE cycle.
They may have hindered liquid entry into the device when submerged in 1 × PBS due to their
hydrophobic nature and could be avoided with proper photomask design. Ensuring that the
underlying sacrificial photoresist covered a surface area that was larger than surface area of the
port defined by the etch mask yielded improved devices as shown in Figure 3.3-5.
90

Figure 3.3-5. Photomask illustrations and SEM images of (a) early and (b) current fluid etch port regions
3.3.2 Measurement Setup
Experiments were conducted with devices mounted into the custom-machined testing
fixture shown in Figure 3.3-6. Dies were sandwiched between the bottom bulk acrylic and a 0.5
mm thick silicone gasket which was cut using a motorized vinyl cutter (Graphtec CE-6000,
Graphtec America Inc., Irvine, CA). The gasket only covered a small horizontal segment across
the die to create a watertight seal while the actual test structures resided in the gasket’s opening
which was defined as the testing chamber. A laser cutter (Epilog, Golden, CO) was used to
91

fabricate an acrylic cover piece which was placed on top of the gasket. Screws were used to apply
pressure on the gasket to prevent potential leaks in the testing chamber.

Figure 3.3-6. Custom fixture used to submerge devices in 1× PBS inside a cavity defined by the silicone gasket and
acrylic components

Holes bored into the bottom acrylic block were fitted with luer connectors to allow 1× PBS
to enter the testing chamber via stopcocks and syringe tubes. The syringe tubes provided
convenient access to the testing chamber for the Ag|AgCl (3M KCl) reference electrode (RE)
(BASi®, West Lafayette, IN) and a macro scale counter electrode (CE) constructed from Pt wire
(1 cm long, 0.5 mm Ø) and potted with epoxy (Epotek 353ND-T, Epoxy Technology Inc., Billerica,
MA). A PCB breakout board was used to electrically access individual device electrodes. Wires
from the PCB were connected to an LCR meter (Agilent E4980A, Keysight, Santa Rosa, CA) or
sourcemeter unit (Keithley 2400, Keithley Instruments, Cleveland, OH). In certain experiments, a
multiplexing board (MUX) was used in lieu of the PCB. The MUX employed four 16:1 ADG1206
monolithic iCMOS multiplexers with a NI-USB6501 data acquisition unit to electronic switch
between desired channels corresponding to contact pads on the µBPT device.
92

Experiments were conducted at room temperature and atmospheric pressure. Parafilm
sheets (not shown) were used to cover syringe tube openings in between experiment trials to
account for potential changes in conductivity due to evaporation. The initial total 10 mL of 1× PBS
did not significantly decrease (<1% volume) throughout the duration of experiments.
3.3.3 Image Processing
A compound microscope (PSM-1000, Motic, Hong Kong) with 200× magnification was
used in conjunction with a mounted CMOS camera (PL-B776, Pixelink, Ottawa, Canada) for
image capture at 10 fps. The image sampling rate necessitated use of custom Windows
Management Instrumentation Command-line (WMIC) scripts to efficiently extract the file creation
time with millisecond resolution.
Custom algorithms were built with the MATLAB Image Processing Toolbox (Mathworks,
Natick, MA) to analyze bubble images and to extract the number of pixels corresponding to the
bubble per image frame. The resolution was approximately 0.2 µm per pixel which corresponded
to 0.004 pL per pixel. A baseline image containing no bubble was subtracted from subsequent
image frames containing bubbles. The resultant image was conditioned, binarized, and pixels
corresponding to the bubble were counted in each frame (Figure 3.3-7). The number of pixels was
converted into µm
2
and multiplied by the sacrificial photoresist thickness which corresponded to
the interior height of the microchannel (16 µm, as measured by profilometer) to arrive at an
estimate for bubble volume. Special care was taken in the image conditioning step to partially
count the shadow in the bubble boundary to account for the gas-liquid interface meniscus. The
same algorithm was applied to all trials.

93

Figure 3.3-7. Raw images of bubble growth were captured at 10 fps and processed with a custom MATLAB
algorithm to extract the bubble volume versus time

On Bubble Generation
3.4.1 Bubble Generation Theory
As the sensing element of the µBPT, the bubble is a crucial component of the device.
Consistent generation of gas volumes within acceptable tolerances was desired in order to recreate
ΔZ or dissolution rate results reported by Yu et. al. However, repeatable results were challenging
to obtain. Therefore, further study into bubble generation fundamental theory with empirical
validation was investigated.
94

Faraday’s law of electrolysis states that the amount of current multiplied by pulse duration
yields the total charge injected into the system which may yield stoichiometrically equivalent
amounts of gas molecules. Although this is a generally acceptable approximation for most systems,
such approximations break down at the MEMS size scale. This was because the relative sizes of
the electrodes used to generate gas was found to play a critical role in the µBPT system. The impact
of electrode size can be illustrated by using an electrochemical potential diagram (Figure 3.4-1).
Electrode polarization during application of a bubble-generating pulse can vary dramatically
depending on the relative sizes of the two electrodes in use. The underlying reasons for this size-
dependence are briefly explained here and the reader is directed to the following references for
additional background beyond the scope of this work [44], [45].

Figure 3.4-1. Electrochemical potential diagrams to illustrate the effect of sizing on electrode polarization. (a) At
equilibrium, all points along the electrochemical circuit are at approximately the same electrochemical potential. (b)
If the electrodes in use are similarly sized (ie. interdigitated electrodes or identical electrode pads) then polarization
will occur in equal and opposite directions. (c) If the counter electrode is significantly larger than the working
95

electrode, polarization will be predominantly localized to the electrode-to-electrolyte potential difference at the
working electrode interface.

An electrochemical cell contains an electrolyte solution and a minimum of two electrodes
in contact with the solution. One electrode is often designated as the ‘electrode of interest’ and is
termed the working electrode (WE). The second electrode closes the circuit and is commonly
referred to as the counter electrode (CE) although other nomenclature such as return, ground, and
auxiliary electrode exist in the literature [44].
The two electrodes may be connected to the terminals of an applied voltage source which
may drive current flow between the WE and CE and provide the driving force for redox reactions
via electrochemical potential differences throughout the cell (Figure 3.4-1). Although both voltage
and electrochemical potential are often used interchangeably, their definitions are distinct. Voltage
refers to the difference in electric potential between two points or with respect to Earth ground.
Electrochemical potential (E) specifically refers to the combination of electrical and chemical
potential (free energy). Like electric potential, electrochemical potential is not measured directly.
Electrochemical potential differences may be referenced to a dedicated reference electrode in 3-
electrode cells, as is the case in this work. In other cases, the CE may serve as both a current sink
and a pseudo-reference in 2-electrode cells. Earth ground is typically avoided because this may
render measurements susceptible to ground loops or damage.
If the driving force is sufficiently large, electron transfer between the electrode an
electrolyte may occur via faradaic reactions (ie. electrolysis). Here, we are primarily concerned
with water electrolysis with the associated half-cell reactions in (1) and (2). By convention, half-
cell reactions are written in the order of reduction.
96

If there is a sufficiently large potential difference during a pulse, reduction of the
electrolyte may occur via (2) at the cathode and (3) may occur at the anode (in the oxidative
direction). Note that the WE may serve as either the cathode or anode. Standard cell potentials in
(2) and (3) are associated for each half-cell reaction, but they are with respect to the standard
hydrogen electrode (SHE) – a theoretical construct which is physically unattainable.
2𝐻 +
(𝑎𝑞 ) + 2𝑒 −
→ 𝐻 2
(𝑔 ), 𝐸 ° (𝑉 ) = 0 ()
𝑂 2
(𝑔 ) + 4𝐻 +
(𝑎𝑞 ) + 4𝑒 −
→ 2𝐻 2
𝑂 , 𝐸 °(𝑉 ) = 1.229 ()
It is possible convert standard cell potentials into values with respect to more practical
reference electrodes such as the saturated calomel or Ag|AgCl electrode and to perform Nernst
equation calculations that determine what potential differences are required to make electrolysis
thermodynamically favorable. However, the reaction kinetics, mass transfer, double-layer
capacitance, metal cleanliness, overpotentials, temperature, buffer strength, presence/absence of
supporting electrolyte, and a myriad of other factors hinder accurate a priori prediction of
electrolytic gassing potential difference values.
In practice, such information is typically obtained by empirically cycling the cell voltage
between an upper and lower bound and measuring the resulting current via cyclic voltammetry
(CV) [170]. The peaks and troughs of i-E curves from CV effectively serve as ‘electrochemical
signatures’ to denote the presence of half-cell reactions such as those for water electrolysis.
The fundamental driving force for electrochemical reactions is the electrode-to-electrolyte
potential difference [45]. Note in Figure 3.4-1 that the total applied potential difference, EA-D
(blue), is the sum of the WE-to-electrolyte potential difference, EA-B (green), the CE-to-electrolyte
potential difference, EC-D (red), and the potential drop across the electrolyte solution, EB-C (also
97

called the iR drop). A two electrode system may only measure EA-D whereas the reference electrode
in a three electrode system may also enable approximate measurements of EA-B and EC-D.
The relative contribution of each interfacial potential drop to the total applied potential
depends strongly on the size of WE and CE. For systems with two equally sized electrodes, both
will polarize in equal and opposite directions during an applied pulse. Conversely, for systems
where the CE is much larger than the WE, polarization is predominantly restricted to the WE
interface since CE’s larger surface area imparts a larger double-layer capacitance. Electroneutrality
dictates that the same amount of charge must pass through both the WE and CE. Mostly
nonfaradaic double-layer capacitive charging therefore occurs at the CE, whereas electrode
polarization induces faradaic reactions for gas evolution at the WE. Both cases in (b) and (c) of
Figure 3.4-1 experience the same total potential difference, but the specific apportionment of
interfacial potential differences may have profound implications on electrolytic bubble generation.
Gas evolution only requires one electrode-to-electrolyte potential difference to be large
enough to drive a desired electrochemical reaction. If devices employ two electrodes of identical
size, equal and opposite electrode polarization of both the WE and CE will occur. Hence,
polarizing both electrodes can result in wasted energy whereas power savings may be critical in
applications that depend on battery life. The upcoming sections will demonstrate a variety of
experiments which were conducted in order to obtain a clearer intuition on bubble generation

98

3.4.2 Bubble Generation Experiments
3.4.2.1 Early Bubble Repeatability Tests
Several early bubble repeatability experiments were conducted where the same set of
bubble generating pulse parameters were applied across one device or identical copies of one
device. Generated bubbles were expected to demonstrate reproducibility on par with tool settings
and resolution limits, but bubble variability was unacceptably high (coefficient of variance on
estimated bubble length could be greater than 10%). Out of the dozens of attempted experiments,
table illustrates the main results from three key studies. Changing device geometries, new pulse
waveforms, and a variety of other attempts could not bring variability down to acceptable levels.
Bubble length variability was high, which precluded any attempts at pressure transduction.
Therefore, further study and experimentation into bubble generation was warranted.

Table 3.4-1. Main results for three bubble reproducibility studies conducted from 2017 - 2019
Time
Period
Device Type Test Parameters Main Result
May 2017 Z40 Type
1 µA, 2 s, one device
10 bubbles generated at 0 mmHg
10 bubbles generated at 300 mmHg
8.9% and 12.6% coefficient of
variance for estimated bubble
length of 0 and 300 mmHg tests;
respectively
December
2018
G Type
Ramp +0.1 V/s from 0 V to 3.1 V and hold at
3.1 V for 50 s
3 bubbles generated across 5 identical devices
at 0 mmHg
Approximately half of bubbles
‘overflowed’ beyond E1 and E4,
rendering coefficient of variance
calculations impossible
99

February
2019
N Type
3 µA, 4 s, one device
9 bubbles generated at 0 mmHg
15.1% coefficient of variance for
estimated bubble length

3.4.2.2 Voltage Cycling to Determine of Gassing Limits as a Function of Counter Electrode Size
Building upon the theoretical considerations laid out in 4.4.1, experiments were conducted
to investigate electrode polarization as a function of the size of the counter electrode. A specialized
electrochemical setup was employed in order to measure the potential difference between the
working electrode and reference electrode (WE-RE) and the potential difference across the counter
electrode and reference electrode (CE-R). As illustrated in Figure 3.4-2, the voltage between the
WE and CE was controlled with a sourcemeter unit (Keithley 2400) while simultaneously
measuring the WE-RE and CE-RE potential differences with a 2-channel oscilloscope (Tektronix,
Beaverton, OR). Voltage buffering was required to prevent unwanted current shunting and
accomplished by high input impedance (1 TΩ) op-amps (TLC2274ACN, Texas Instruments,
Dallas, TX) where the output was connected to its inverting input and the signal source was
connected to the non-inverting input.

Figure 3.4-2. Electrochemical circuit diagram to illustrate connection schemes between electrodes of interest and
equipment

100

An experiment similar to cyclic voltammetry (CV) was conducted to monitor the amount
of unwanted polarization of the CE during gas evolution on annealed G type devices. Using a
custom LabVIEW program, the WE-CE potential was set to start at 0 V and decrease at a sweep
rate of -100 mV/s. The potential was allowed to decrease until the onset of gas generation due to
the hydrogen evolution reaction (HER) could be observed in real-time. As soon as bubble
generation was observed, the sweep rate was changed to +100 mV/s. The WE-CE potential then
increased until the onset of the oxygen evolution reaction (OER) could be observed. At that
moment, the sweep rate was changed to -100 mV/s again and the experiment terminated when the
WE-CE potential returned to 0 V. Three trials per device per CE size were conducted. The
measured variables of interest were the WE-RE potential and CE-RE potential. This experiment
was repeatably conducted for the 3 different CE sizes shown in and representative results are
displayed in Figure 3.4-3.

Name Area (µm
2
) CE:WE Ratio
Small 2.3 × 10
2
1:1
Medium 2.5 × 10
3
10:1
Large 1.6 × 10
7
70,000:1

Table 3.4-2. Size designations of tested counter electrodes
101

Figure 3.4-3. Representative voltage sweeping results to determine gassing limits as a function of counter electrode
size. (a) The total electrochemical cell potential was swept at ±100 mV/second while simultaneously observing gas
evolution in real-time. (b) The electrode-to-electrode potential for both the working electrode and counter electrode is
plotted for instances where the counter electrode size was allowed to vary. Device type = G.

For all CE sizes, the WE-RE potential difference approached approximately -1.4 V for the
onset of H2 gas evolution and +2.0 V for O2 gas evolution. The CE-RE potential difference
displayed similar behavior, if inverted, for the small CE case. Notably, in the large CE case, the
CE-RE potential difference remained relatively unchanged from its starting value. The medium
CE case demonstrated behavior intermediate to the small and large case. Since the WE-CE
potential difference is essentially the sum of the WE-RE and CE-RE potential difference, Figure
3.4-3 illustrated that use of larger CE will confine the majority of the potential drop to the WE-to-
electrolyte interface for all applied potential difference values. Conversely, small CE usage
102

revealed unwanted sharing of the potential drop across both the WE and CE interfacial potential
differences. Additionally, data exhibited more noise for smaller CE size
3.4.2.3 Bubble Generation via Cathodic Pulses as a Function of Counter Electrode Size
As shown in Figure 3.4-4, pulses of -0.6 µA and 6 seconds were applied for different CE
sizes. These parameters consistently generated bubbles across the same set of annealed G type
devices. Potential differences were measured throughout the pulse duration and revealed that usage
of large CE effectively prevented CE polarization. If a large CE was employed, the WE-CE and
WE-RE potential differences were almost identical, indicating that the majority of the potential
drop was localized to the WE-to-electrolyte interface.

Figure 3.4-4. (a) Cathodic pulses (-0.6 µA for 6 seconds) were applied to generate bubbles and (b) the various
potential differences normalized to their starting values were measured as a function of CE size (n = 8 pulses per CE
size dataset; shaded regions denote standard deviation) (c) Multiplying the average total WE-CE potential difference
per pulse by the current amplitude and duration allowed for computation of the average energy required per bubble.
Device type = G.

103

The energy consumed to generate each bubble was computed by multiplying the average WE-CE
potential drop during a pulse, the current amplitude, and pulse duration. Results are displayed in
Figure 3.4-4 and illustrated that small CE usage required more energy than that of large CE. The
former caused both WE and CE to polarize until one of the electrode-to-electrolyte potential
differences was sufficient to drive electrolysis. A significant amount of energy is therefore wasted
in polarizing both electrodes. Conversely, a properly designed cell with a large CE effectively
prevented CE polarization and confined the potential drop to the WE interface. In this experiment,
there was a 41% difference in energy savings when comparing the small (12.0 µJ) and large (7.1
µJ) case.
Image processing tools extracted the bubble volume in each trial which allowed for
computation of the pooled standard deviation among different CE size datasets. Theoretically, the
equivalent amount of charge injected in each trial should yield equivalent bubble volumes, but the
measured average values were 0.11, 0.16, and 0.11 nL for the small, medium, and large;
respectively. Pooled standard deviations across datasets (32.7, 29.4, and 19.8 pL for small, medium,
and large cases; respectively) revealed that larger CE experiments yielded slightly more uniform
bubble volumes. In one dataset for the large CE case, the coefficient of variance in bubble volume
was as low as 3.7%.
3.4.2.4 Faraday’s Law Analysis with Increasing Pulse Durations
In order to investigate adherence to Faraday’s law as described in section 3.4.1,
experiments were conducted with one A type device where the pulse duration was varied between
10, 15, 20, and 25s while maintaining a constant +0.1 µA current amplitude and usage of the large
CE size at atmospheric pressure. The approximate volume of gas generated via the anodic reaction
104

may be computed via equation (4) which is derived from the ideal gas law and Faraday’s law of
electrolysis:
𝑉 =
𝐼𝑡𝑅𝑇 𝐹𝑣𝑃
()
where V is volume, I is current, t is the amount of time applied during the pulse, R is the ideal gas
constant, T is temperature, F is Faraday’s constant, v is the valence, and P is pressure. Note that a
linear relationship exists between volume (V) and pulse duration (t). Figure 3.4-5 illustrates the
extracted bubble volume versus time where white open circles represent the maximum volume
obtained per pulse. In order to obtain baseline impedance measurements prior to the pulse, pulses
initiated at t = 30s per trial.

Figure 3.4-5. Bubble volume versus time for various pulse durations (10, 15, 20, 25 s).Device type = left-most A).
The empirically measured volume of gas per pulse duration was compared against the theoretically expected volume
from equation (4).

Results in
105

Figure 3.4-6 show that a linear relationship between volume and time was observed, as
expected. Measured maximum bubble values were consistently about 30% of theoretically
calculated values. The reduction in yield may be attributed to some charge being spent on
charging the electric double layer as well as gas recombination.

Figure 3.4-6. Theoretically expected and empirically measured max bubble volumes for various
pulse durations. Device type = left-most A
Bubbles generated in this experiment were too small to expand into the channel region as
illustrated in Figure 3.4-7. Growth was limited such that the gas occupied only the nucleation core
as opposed to expanding into the channel region as shown illustrated in the idealized sequence of
Figure 3.3-2. However, the impedance measured from these bubbles revealed unanticipated but
valuable insights on device operation. Figure 3.4-8 demonstrates that the first 30 seconds of all
experimental trials maintained approximately the same baseline impedance value (~700 kΩ). Raw
impedance data showed that progressive trials exhibited slight decreases in impedance which
corresponded to progressive delamination of the left-most sensing electrode (E1, Figure 3.4-7)
since trials were conducted in the order of shortest to longest pulse duration. However, subtraction
of each trial’s respective baseline values (offset) or normalizing to each trial’s respective values
(% change) showed that the baseline impedance drop due to delamination was negligible.
106

Figure 3.4-7. Freeze-frame images of bubbles generated via varying pulse durations at their approximate maximum
volumes. Device type = left-most A. Timepoints correspond to those in Figure 3.4-5

Figure 3.4-8. Impedance magnitude versus time of bubbles generated via varying pulse durations. Device type =
left-most A.

107

Impedance data during the pulse was unavailable due to multiplexing, but key observations
can be made on the resulting impedance data after the pulse. For instance, slope of the impedance
recovery was progressively steeper as pulse durations were longer. This may be attributed to
changes in ambient electrolyte pH during electrolysis which temporarily depressed the resistivity
and consequently the impedance. Due to buffering effects of the 1 × PBS, the pH, resistivity, and
baseline impedance approach original values after the pulse. Since longer pulses generated more
redox species and a larger concentration gradient, the slope of impedance recovery is greatest in
the 25s case.
Interestingly, the impedance in the 20s and 25s case temporarily rose above baseline values
despite the fact that the bubble volume never entered the line of sight between sensing electrodes
E1 and E4. This phenomenon was conjectured to be a result of dissolved gas species temporarily
increasing the solution resistivity and outcompeting the resistivity decrease due to pH changes.
The 15s and 10s case may also have experienced the resistivity increase due to dissolved gas
diffusion, but in amounts insufficient to exceed baseline impedance values. All trials eventually
approached baseline values as equilibrium was reached in the span of several minutes.
Additionally, impedance discontinuities may be observed for the 20s and 25s case (t ≈
120s and t ≈ 155s; respectively). Bubbles were observed to detach and relinquish contact with a
boundary of the Parylene device wall at the precise timepoint of these impedance discontinuities
as shown in Figure 3.4-9 for the 20s case. The abrupt and discrete change in impedance from a
detachment event is conjectured to be a result of the sudden presence of an additional metal surface
such as E3 into the electrochemical measurement circuit. Prior to detachment, the Parylene walls
and the nonconductive gas bubble may have effectively served as electrical isolation for E3.
108

Figure 3.4-9. Bubble relinquishes contact with Parylene wall during a detachment event at t ≈ 120 s. Device type =
left-most A.

3.4.2.5 Bubble Generation with Positive Versus Negative Current
Original device design in section 3.3.1 called for use of 2 equally sized microelectrodes
(E2 and E3) in the nucleation core to generate the bubble. For many early experiments with this
configuration, no noticeable difference was observed if positive versus negative current was
employed to generate bubbles.
When a large CE was employed to effectively restrict polarization to the WE interface, the
use of positive versus negative current revealed clear differences. Figure 3.4-10 displays the
bubble volume versus time at the onset of the pulse (6 s duration, ±0.6 µA). Notably, the negative
applied current generated approximately double the amount of gas (0.13 nL) as the positive current
(0.07 nL). The total amount of time required for bubbles to dissolve revealed that bubble lifetime
was notably longer for bubbles generated with positive rather than negative current (36 ± 6.1 min
versus 10.3 ± 0.7 min; respectively; mean ± standard deviation; n = 8 each). Notably, these bubbles
from positive current had longer lifetimes despite their smaller starting volumes (Figure 3.4-11).
The +0.6 µA and 6 s pulse generated bubbles that only filled up approximately one third of the
total microchannel length, so a test bubble with +0.6 µA and 18 s was generated to fill the entire
109

channel. The bubble lifetime was 500 minutes which represented a 25× improvement over
previous work [171] [52].

Figure 3.4-10. Bubble volume versus time for experiments using a large counter electrode with positive or negative
applied current pulse. Thick lines represent averaged data and thin lines represent individual trials. Device type = G.

Figure 3.4-11. The amount of time required for generated bubbles to completely dissolve (ie. bubble lifetime)
significantly varied by the usage of positive or negative applied current (36 ± 6.1 min versus 10.3 ± 0.7 min;
respectively; mean ± standard deviation) when a large CE was used. Device type = G.

3.4.3 Bubble Generation Discussion
Initial dissatisfaction with bubble volume precision prompted further investigation into
fundamental bubble generation theory. This inquiry challenged previously held assumptions on
device operating mechanisms and proposed a new operating scheme which considered electrode
110

polarization effects. Usage of the large CE was able to bring bubble coefficient of variance from
~15% to 3.7%. Bubbles generated via large CE may be more uniform in terms of the total volume
of generated gas because electrode surface inhomogeneities and uncleanliness are known to affect
the repeatability of electrochemical reactions. Use of larger CE was hypothesized to reduce such
effects because the relative ratio of contaminated versus nominal surface area may be more
favorable than that of smaller CEs.
Usage of large CE in bubble generation for the µBPT also conferred addition benefits. For
instance, significant energy savings were observed per each bubble generation event (up to 41%;
from 12.0 µJ to 7.1 µJ for small versus large CE). In certain sensors designed for implantation,
such savings may significantly extend battery life because electrolytic generation consumes orders
of magnitude more energy than other operations such as sensing or data transfer [171] [52].
Another major advantage of large CE use is the control over gas chemical composition –
which in turn allowed for additional control over bubble lifetime due to each gas’s characteristic
water solubility (4.5 × 10
-5
and 2.10 × 10
-5
cm
2
/s at 25°C; H2 and O2, respectively [172])
Depending on the application, fast or slow dissolution may be desired. For instance, bubble-based
pressure sensors experienced unwanted and relatively fast bubble dissolution (< 20 minutes) [171]
due to H2 generation from equally sized electrode pairs. Figure 3.4-10 and Figure 3.4-11
illustrated adherence to stoichiometric expectations dictated by half-cell reactions. Twice the
amount of H2 gas was generated in the reduction reaction as O2 in the oxidation reaction. Although
less O2 was generated, O2 bubbles had longer lifetimes as expected by its lower solubility.
The greatest drawback in the use of large CE for the µBPT may be the device footprint.
Larger CE will increase device size which may be undesired in clinical applications. In this work,
the microfabricated medium CE size was selected to investigate a 10:1 ratio between CE:WE, but
111

the Pt wire for the large CE demonstrated most effectiveness with a 70,000:1 ratio. Although such
macro-scale and non-micromachined components may be difficult to incorporate into MEMS-
based device design, very large microfabricated pads may be designed to serve as CE to achieve
the same benefits.

112

On Bubble Behavior
3.5.1 Bubble Behavior Theory
Ideally, bubble behavior would follow the sequence described in Figure 3.3-2. The
underlying principle behind such behavior was the propensity for gas-liquid interfaces to minimize
the work (w) of surface creation by forming shapes with the smallest possible surface area (σ).
Surface tension (γ) is the constant relating these two quantities as shown in equation (5)
𝑑𝑤 = 𝛾 ∗ 𝑑𝜎 ()
However, the aspect ratio at the MEMS-scale and hysteresis at the gas-liquid phase
transition proved to pose complex issues that made bubble behavior control challenging. The
following experiments illustrate key findings which elucidate bubble behavior in the context of
this electrolytically generated nL bubbles trapped inside a Parylene C microchannel.

3.5.2 Bubble Behavior Experiments
3.5.2.1 Optimizing Bubble Detachment from Nucleation Core
Bubble detachment from the nucleation core was demonstrated by Volanschi [173] and Yu
[174] but initial attempts in this work to recreate results were highly unsuccessful. Due to the
𝑑𝑤 = 𝛾 ∗ 𝑑𝜎 principle, it was hypothesized that optimizing nucleation core and microchannel
geometries may improve rates of detachment behavior. Hence, several device geometries with
varying ratios between the channel width and bottleneck width were fabricated as shown in Figure
3.5-1. Smaller ratios were hypothesized to improve detachment because they may more effectively
“pinch off” bubbles into the microchannel. However, excessively small ratios were conjectured to
113

present fabrication issues due to tool limitations. Therefore, the experiment was design to optimize
this ratio. Table 3.5-1 displays the ratios tested in this work.

Figure 3.5-1. Schematic illustration of bottleneck width and channel width

Table 3.5-1. Table of various bottleneck:channel ratio devices fabricated for testing
Bottleneck:Channel Ratio

Channel Width (µm)
50 100 150 200
Bottleneck
Width (µm)
10 1:5 1:10 1:15 1:20
20 2:5 1:5 2:15 1:10
30 3:10 3:20
40 2:5 4:15 1:5

One major takeaway from this experiment was the discovery of consistently unorthodox
dissolution patterns in devices with channel widths of 100 µm or more. Disfigured bubble shapes
adversely affected the measured impedance as shown in Figure 3.5-2. This phenomenon was
theorized to be a consequence of the aspect ratio between the channel height (16 µm, due to
114

photoresist spin) and the channel width. Intermolecular forces from the top and bottom plane of
the channel were most likely able to sustain odd bubble shapes due to the channel height being
small relative to the large channel width. Unfortunately, such unanticipated bubble shapes
effectively altered the geometry of the fluidic capillaries proposed in Figure 3.3-3 and the caused
the measured impedance to lose their correlation to pressure. It is noteworthy that although the
impedance erratically increased and decreased with respect to bubble shape, total bubble volume
decreased monotonically (data not shown). Devices with 50 µm channel width never exhibited the
unorthodox bubble shapes.

Figure 3.5-2. Unorthodox bubble dissolution patterns adversely affected the measured impedance. Device type = C10.

Despite testing many device geometries, bubble detachment from the nucleation core
remained inconsistent. Several hundreds of bubbles were tested with various parameter
modifications including but not limited to annealing versus unannealed, alternative device cleaning
methods (cyclic voltammetry, reactive ion etching, and other solvents), varying pulse parameters,
115

and more. Regardless, an estimated 5% of bubbles demonstrated the desired detachment and
dissolution behavior outlined in Figure 3.3-2. Bubbles did tend to detach for smaller
bottleneck:channel width ratios, but the majority of these results were from devices with channel
widths greater than 100 µm and hence unusable for pressure transduction tests. Successful bubble
detachment on 50 µm channel devices appeared to occur at random with no strong correlation to
the number of previously generated bubbles per device nor other parameters.
3.5.2.2 Experiments on Controlling Bubble Location
One drawback of initial uBPT designs was the lack of control over the direction in which
the bubble grew. For instance, bubbles could expand preferentially to the right or to the left when
exiting the nucleation core and occupying the measurement channel. Often, the directional growth
would cause the bubble to encroach past one of the sensing electrodes, causing the measured
impedance to skyrocket and lose sensitivity to ambient pressure.
To counteract this unwanted phenomenon, new structures called “constriction valves” were
designed (Figure 3.5-3). Their location in space was more proximal to the nucleation core than
the sensing electrodes in order to effectively steer microbubble growth away from the sensing
electrode.

Figure 3.5-3. Schematic illustration of constriction valves

116

Constriction valve structures demonstrated excellent performance. As long as the volume
of a generated bubble did not exceed the sum of the nucleation core and channel’s capacity, the
bubble was effectively steered away with a 100% success rate and never contacted the sensing
electrodes. A representative example is illustrated in Figure 3.5-4.
In one experiment, the influence of gravity was investigated with respect to constriction
valve performance. In clinical settings, patient cranial orientation (supine, prone, etc.) may affect
the direction of the buoyant force experienced by the bubble whereas benchtop tests were
predominantly conducted on a level surface such that the longitudinal dimension of the µBPT was
parallel to the floor. In one experiment, the testing fixture was adjusted by 90° such that the µBPT
longitudinal direction laid normal to the floor. Dozens of bubbles were generated, yet 100% of
bubbles successful remained inside the µBPT due to the constriction valves. Surface tension forces
at the MEMS size scale in this work were confirmed to outweigh buoyant forces to ensure that
µBPT bubble behavior was orientation agnostic.

117

Figure 3.5-4. Representative sequence of constriction valves effectively controlling bubble growth behavior. Device
type = G.

3.5.2.3 Long-Term Bubble Considerations
An experiment was conducted to generate numerous bubbles in order to assess long-term
device and bubble behavior. Custom algorithms patched from Pulover Macro Creator
2
,
AutoClicker software, LabVIEW, Pixelink OEM capture software, and intermittent check-ins with
Teamviewer remote desktop client allowed for automatic data acquisition where 83 pulses of -0.6
µA for 6 s were applied at 2 hour intervals at atmospheric pressure. E3 was the WE and a large Pt
was the CE on an annealed A type device which had never undergone other tests. Bubble
generation pulses were applied at t = 60 s of each trial in order to obtain baseline impedance
measurements.

2
https://www.macrocreator.com/
118

As shown in Figure 3.5-5, progressively smaller amounts of gas volume are generated in
later trials. Extraction of the maximum volume per trial and plotting against trial number in Figure
3.5-6 revealed that a transition occurred at approximately the 20
th
trial. At this point, bubble
volume per trial markedly decreased from approximately 0.15 nL to 0.05 nL.
The impedance measured between E1 and E4 for these 83 trials is also plotted in Figure
3.5-7. Notably, the baseline impedance values were constant, but the impedance behavior
following bubble generation varied significantly as progressive trials were conducted. Trials
before the 20 bubble transition point experienced major increases in impedance post-bubble
generation, as expected. However, the post-bubble generation impedance for these trials
maintained an approximate plateau for 25 to 35 seconds which was at odds with the monotonically
decreasing volume versus time behavior.
The post-bubble generation impedance values for trials after the 20 bubble transition point
exhibited a temporary decrease in impedance which eventually recovered back to baseline levels.
This behavior is similar to that shown in Figure 3.4-8 which agrees with bubble behavior in cases
where the bubble volume was insufficient to expand into the nucleation core. This was confirmed
by micrographs of the freeze-frame image corresponding to the maximum bubble volume for trial
1 and trial 83 shown in Figure 3.5-8. Additionally, delamination of E2 can be observed during the
electrolytic pulse for both trials, but trial 83 demonstrated delamination severe enough to induce
Newton ring fringe interference patterns.
119

Figure 3.5-5. Volume versus time for 83 bubbles generated in longevity experiment. Device type = middle A.

Figure 3.5-6. Maximum volume versus trial number in device longevity experiment. Device type = middle A.

Figure 3.5-7. Impedance versus time for 83 bubbles generated in longevity experiment. Device type = middle A.
120

Figure 3.5-8. Trial 1 versus trial 83 freeze-frame images during maximum bubble volume. Device type = middle A.

3.5.2.4 Confocal Microscopy to View Cross-section Fluid Capillaries
Although the fluid capillaries presented in Figure 3.3-3 were hypothesized to serve as
pressure-sensitive variable resistors which conducted electrochemical current, they had never been
definitively visualized due to a lack of appropriate imaging modalities. Therefore, preliminary
experiments with confocal microscopy techniques in collaboration with the Fraser group at USC
set out to view high-resolution cross-sectional images of an annealed I5 type device. The operating
mechanism of confocal microscopy involves capture of several two-dimensional images at
different depths with a spatial pinhole technique. From those two-dimensional samples, three-
dimensional structures may be reconstructed to view the desired cross-sections of the µBPT.
Approximately 50 mL of 1 × PBS was mixed with 1 mL of fluorescein die and a bubble
was generated by applying sufficient current between E2 and E3. Figure 3.5-9 illustrates a
representative confocal image at a timepoint when the generated gas (black) displaced the fluid
(green) such that the gas occupied the nucleation core and right-half of the channel. The A-A’
121

cross section revealed that the rectangular cross-section of the left-half of the channel was
completely occupied by the green fluorescein. The B-B’ cross section of the channel was
anticipated to be predominantly black, with distinct green fluid capillaries at the corners. However,
no fluid capillaries were observed. No fluid capillaries were observable across additional trials and
devices.

Figure 3.5-9. (Left) Top-down confocal microscopy image with (right) A-A’ and B-B’ cross-section views. Device
type = I7.
3.5.3 Bubble Behavior Discussion
Experiments in this section revealed mixed results in obtaining ideal bubble behavior. For
instance, constriction valves were highly successful in preventing bubble escape from the µBPT
and bubble detachment was occasionally achieved in certain design iterations. However, deeper
inquiry revealed that many areas for improvement exist in the µBPT system. Firstly, it is
recommended to forego the use of a nucleation core structure altogether in future designs due to
the challenges associated with bubble detachment. The nucleation core’s original intent was to
retard bubble dissolution by transporting the bubble away from Pt surfaces which catalyze the gas
122

recombination reaction. However, bubble quasi-stability was achieved through an entirely
different paradigm illustrated in section 3.4.2.5 via anodic pulses for O2 generation.
In addition to failing to achieve its goal, the nucleation core presented several problems
that exacerbated poor device performance. One example was the difficulty in stripping sacrificial
photoresist during device fabrication. Photoresist removal was already challenging for the channel
region due to diffusion limited access at the fluidic ports, but the additional degree of tortuosity
presented by the nucleation core extended the amount of time required to soak devices in harsh
solvent. This was conjectured to accelerate Parylene-metal delamination and reduce device
longevity. Secondly, bubble growth was observed to unintentionally and suddenly terminate
during the applied pulse in certain nucleation core designs. Gas expansion into such a limited
amount of space was conjectured to displace all conductive liquid in the nucleation core and
effectively render an unwanted open circuit mid-pulse. A suggestion is to place only one electrode
dedicated for electrolysis as the WE in line with the E1 and E4 sensing electrodes in order to
simplify design and boost device robustness and reliability.
Another area of improvement for device behavior is the long-term reliability. µBPT
devices may ideally be present in hydrocephalus patients for several years, but current device
iterations have shown severe delamination in the span of several days. The thin-film Parylene-
metal interface has notoriously poor adhesion [38] which was clearly illustrated by Figure 3.5-8.
Repeated application of high charge density pulses only exacerbated device damage. However,
µBPT devices may employ alternative substrates such as fused silica or quartz for improved
adhesion because they do not share the same anatomically-defined geometric constraints as the
RRA project. Hydrocephalus ICP sensing may be achieved through planar but sufficiently small
µBPT dies packaged in line with catheters or sensor modules.
123

Confocal microscopy results suggested the lack of fluid capillaries. It is noteworthy that
the imaging technique did bear temporal and spatial limitations which may have precluded
capillary detection. Acquisition of 2D slices and reconstruction of the 3D images occurred on the
same timescale as bubble dissolution due to the use of E2 and E3 as bubble generation electrodes.
This created faster dissolving H2 gas but the large CE and anodic O2 gas could not be generated
due to space limitations on the confocal microscope stage. However, the plateaued impedance in
Figure 3.5-7 for early trials also corroborated the lack of fluid capillaries in the µBPT. The
impedance plateaued at timepoints where the channel was filled with gas and decreased only when
bubble dissolution progressed enough to create a fluidic path between E1 and E4. It is conjectured
that full occupation of the channel by the bubble effectively rendered an open circuit between E1
and E4 instead of the desired pressure-sensitive variable resistor via fluid capillaries. This may
explain the relatively high impedance values (approximately 2 MΩ). Transient increases and
decreases during the plateaus may be attributed to pH and dissolved gas concentration effects from
section 3.4.3. manifesting in the fluidic path that traverse around the µBPT instead of through it.
Experiments conducted by Ateya [51] modified the hydrophilicity of microchannels to obtain
similar impedance behavior measurements which may be investigated in the context of µBPT
development as well.
Lastly, impedance measurements were found to be extremely sensitive to bubble shape.
This was evidenced by the response to unorthodox bubble dissolution behavior in Figure 3.5-2.
Although undesired, a few micrometers of bubble movement in the lateral direction effected
several kΩ of impedance change. This confirmed the potential for high pressure sensitivity which
may be effectively harnessed with the desired longitudinal shape modulation of fluid capillaries.
Increasing sacrificial photoresist thickness may prevent the unorthodox bubble dissolution
124

behavior in devices with channel widths greater than 100 µm. This may also improve sacrificial
photoresist stripping by enlarging the effective cross-sectional area to improve diffusion
limitations. Additionally, raising the sacrificial photoresist thickness will alter the cross sectional
aspect ratio to approach that of a square for more amenability towards fluidic capillary creation
[175].
On Bubble Measurement
3.6.1 Bubble Measurement Theory and Modeling
Electrochemical impedance measurements were expected to serve as a valid proxy for
ambient pressure in the µBPT. Therefore, deeper investigation into measurement theory and
limitations was conducted to reveal key insights.
The measurement electrodes E1 and E4 may be approximated by using Randles circuit
elements shown in Figure 3.6-1. A charge transfer resistance Rct and double layer capacitance Cdl
exists at each electrode-electrolyte interface and the solution resistance Rs lies between the
electrodes. Rct describes charge transfer between the metal electrode and the aqueous solution due
to faradaic redox reactions and Cdl is created by the double layer of electrostatically-bound ions
that builds up at the electrode-electrolyte interface [176].

125

Figure 3.6-1. Randles circuit elements to model sensing electrodes E1 and E4

At high frequencies, ions cannot reorganize quickly enough to serve as as a capacitor. Cdl
effectively reduces into a short circuit, rendering the impedance approximately equal to solely the
solution resistance Rs. Previous work assumed that Rs may be modeled as a wire as in equation (6).
|𝑍 | ≈ 𝑅 𝑠 =
𝜌𝑙
𝐴 ()
where 𝜌 is solution resistivity, 𝑙 is the length of the fluid capillary (or bubble length), and 𝐴 is the
effective cross-sectional area off all fluid capillaries combined. If this were true, one may expect
the impedance to directly track the bubble length as shown in Figure 3.6-2 due to the linear
relationship between 𝑍 and 𝑙 . Note that 𝑍 tracks 𝑙 in a plot with separate y axes to illustrate the
expected measurement. Experiments in this section will illustrate where and why these
approximations may not hold and present potential solutions to overcome these issues.

126

Figure 3.6-2. Idealized plot to illustrate theoretical relationship between impedance and bubble length

3.6.2 Bubble Measurement Experiments
3.6.2.1 Measurement Signal Characterization
In order to confirm the validity of the Randle circuit model approximation, electrochemical
impedance spectroscopy (EIS) was conducted on the sensing pair of electrodes via a Gamry
Reference 600 potentiostat with a standard 25 mV signal. Although 3-electrode measurements may
be employed to measure the interfacial properties of the only WE as in the RRA project, this
experiment used a 2-electrode configuration to measure the properties between E1 and E4. The
WE was E1, and the RE was connected to the CE onto E4 for an annealed A type device. Figure
3.6-3 illustrates the Bode plots of an expected result (left) for a Randles circuit compared against
the empirically measured result (right).
127

Figure 3.6-3. Expected versus measured EIS for µBPT sensing electrode pair. Device type = middle A.

The low frequency spectrum impedance magnitude may be expected to plateau at values
corresponding to 2Rct + Rs whereas the high frequency impedance should reduce to Rs, as
previously discussed. In theory, the impedance phase should reach a local maximum approaching
0° when the frequency is high enough to bypass Cdl such that no imaginary component of the
impedance is measured; only the real part of the impedance should be measured through Rs. The
measured result revealed that the low frequency plateau was not observed within the tested
frequency range. A high frequency plateau exists from approximately 10 kHz to 100 kHz which
corresponded to the impedance phase maxima. However, a further downward taper of impedance
128

magnitude was observed for frequencies above 100 kHz. This result revealed that although the
Randles circuit model does not perfectly fit the data, key regions of interest (10 kHz to 100 kHz)
should effectively bypass the double layer capacitor to achieve the desire Rs measurement mode.
Impedance measurements were then acquired at a single frequency via an LCR meter. 20
kHz was used and the signal level was varied from 10 mV, 25 mV, 100 mV, 1 V, and 2 V for a
60s baseline measurement between E1 and E4 with no bubble at atmospheric pressure. No other
electrical equipment was involved in the measurement circuit. The raw impedance versus time
data is plotted in Figure 3.6-4. Figure 3.6-5 illustrates the back-calculated average current draw
per baseline impedance measurement versus signal level to confirm that all signal levels were
operating within a linear electrochemical range. Higher signal levels were hypothesized to
potentially create a large enough electrode-to-electrolyte potential difference to generate redox
species at nonreversible rates. This would cause the mass transfer of redox species to outpace
electrochemical reaction kinetics and skew measurement results due to ensuing changes in solution
resistivity. The highest signal level, 2V, demonstrated the lowest amount of 3σ signal noise
(Figure 3.6-6) while remaining in linear operation regimes so future experiments employed 2V
signal levels. The data sampling rate could be up to 20 Hz if simple virtual instruments from
LabVIEW were employed.
129

Figure 3.6-4. Impedance versus time for a 60s baseline measurement with varying signal levels. Device type =
middle A.

Figure 3.6-5. Average current draw versus signal level to confirm device operation with linear electrochemical
regimes. Device type = middle A.
130

Figure 3.6-6. 3σ signal noise versus signal level in baseline measurements. Device type = middle A.

Signal noise was also investigated as a function of additional instrumentation added to the
measurement setup. The device under test (DUT) was the same A type device in all experiments
in this section. Using the 2V signal level and 20 kHz measurement frequency, baseline impedance
measurements were obtained when the MUX board described in section 3.3.2 was included in the
measurement circuit as shown in Figure 3.6-7 in addition to another measurement with both the
SMU and LCR meter sans MUX. Despite proper grounding and shielding, the 3σ measurement
noise was 0.16 kΩ, 1152 kΩ, and 1381 kΩ, for measurement setups (a), (b), and (c); respectively.
A massive 5 orders magnitude increase in signal noise was observed when comparing setup (b)
and (c) against setup (a).

131

Figure 3.6-7. Three different measurement setups to quantify signal level noise. Device type = middle A.

3.6.2.2 Theoretical Impedance from Bubble Length Compared against Measured Impedance

To test the validity of equation (4), the theoretical impedance was calculated by using
empirically obtained bubble length (l), solution resistivity estimates (ρ) from vendor supplied
values, and the assumption that 10% of the cross sectional area (A) of the microchannel was
occupied by the fluid capillaries. Then this theoretically computed impedance was compared
against the empirically measured impedance. This process was repeated several times for several
different device geometries. Representative results for an annealed N type device are illustrated
below. Figure 3.6-8 shows the measured bubble length and the theoretical impedance which may
be expected from obeying equation (4). Note that when plotted with separate y axes, the two plots
perfectly overlap as previously discussed in section 3.6.1. However, the empirically measured
impedance did not have good agreement with the theoretical impedance as shown in Figure 3.6-9.
A variety of unanticipated features may be observed – some of which have been previously
132

attributed to phenomena such as changes in solution resistivity due to pH and dissolved gas
concentration in addition to dissolution-induced wall detachment events that abruptly introduce
new metal surfaces to the overall electrochemical circuit. Although the previous section illustrated
reasonable fit with the Randles circuit, experiments in this section revealed that equation (4) was
a poor approximation of bubble measurement modality.

Figure 3.6-8. Measured bubble length overlapped with theoretically computed impedance magnitude

Figure 3.6-9. Measured impedance versus theoretically computed impedance. Device type = middle A.
It is noteworthy to observe 3 distinct regimes of measured bubble impedance. A “mixed
phenomena” regime occurs from t = 0 to 150 s, where changes in solution resistivity are
hypothesized to occur due to changes in local pH and dissolve gas concentration. Additionally, the
133

bubble length sharply decreases in this regime. Previous ΔZ and dissolution rate techniques
unfortunately relied heavily on data in this poorly characterized regime. A “steady” regime follows
from t = 150 to 500 s. Here, the bubble length and impedance steadily decrease monotonically.
The final regime is starts at sudden impedance discontinuity at t = 520 s. Although the decrease
bubble length was steady, the abrupt impedance discontinuity was attributed to a bubble wall
detachment effect which had also been observed in Figure 3.4-9.
3.6.2.3 Pressure Measurement via Residence Time
Previous work employed various measurement modes to transduce pressure as described
in section 3.1.2. However, results in this work suggest that secondary effects such as pH and
dissolved gas concentration may have been partially responsible for loss of accuracy in modes
such as ΔZ – the change in impedance from a baseline state to the maximum impedance after
bubble generation. Results showed that the impedance maximum occurred several seconds after
the bubble length’s maximum. The dissolution rate is also adversely affected in the several seconds
after bubble generation and should not be assumed to strictly obey analytical expressions
developed by Epstein and Plesset [177]. Additionally, ΔZ and dissolution rate techniques require
one bubble per measurement, which severely limits ICP sampling rates and would quickly exhaust
the 20-bubble shelf life of current µBPT device as described in section 3.5.2.3. The real-time
impedance to pressure tracking method is preferred, but a new measurement mode has also been
developed in this work – residence time.
The residence time technique was able to correlate the ambient pressure to the amount of
time required to completely dissolve a generated bubble. At the moment of complete bubble
dissolution, the measured impedance was observed to exhibit a local minimum as shown in Figure
3.6-10 for a representative trial with an a Z80 type device.
134

Figure 3.6-10. Local minimum in impedance at the moment of complete bubble dissolution in a representative trial
for the residence time pressure transduction technique. Device type = Z80.

61 bubbles in total were generated at randomly ordered pressures and the residence times
for each trial were detected as shown in Figure 3.6-11. Residence time versus pressure in this
work is presented in Figure 3.6-12. The equation of fit (7) resembled the structure of Epstein-
Plesset solutions where:
𝑅𝑒𝑠𝑖𝑑𝑒𝑛𝑐𝑒 𝑇𝑖𝑚𝑒 = 34.15 ∗ 𝑒 −
𝑃𝑟𝑒𝑠𝑠𝑢𝑟𝑒 30.83
+ 20.89 ()

Previous work calculated pressure resolution for the ΔZ by first computing 3*SE for each
pressure value. Then, the equation of fit was solved for pressure. The difference between the
computed pressure values when the smallest 3*SE and the corresponding ΔZ value was substituted
into the equation was the pressure resolution. When this process was repeated for the residence
time technique, the pressure resolution was computed to be 0.5 mmHg – representing a 40×
improvement over the 20 mmHg resolution obtained in [52]. The 5.2 mmHg resolution obtained
135

by [159] did not employ mathematic formulae but was instead estimated by eye and this work
represented a 10× improvement.

Figure 3.6-11. Residence times of 61 bubbles illustrated against their tested pressures. Device type = Z80.

136

n
Figure 3.6-12. Residence time versus pressure. Device type = Z80.

3.6.2.4 pH experiments
Preliminary experiments with pH indicator (Laboratory-Grade Universal Indicator
Solution, Aldon Corporation, Avon, NY) were conducted in order to further investigate the relation
between measured impedance and pH. A 1:1 mixture of 1 × PBS and indicator was poured into a
plastic weigh boat and 3V was applied via Keithley SMU on two Pt counter wires. No gas bubbles
were observed, but a colorimetric change was apparent within several seconds which confirmed
that electrochemical reactions may induce local changes in pH (Figure 3.6-13). Table 3.6-1
provides a legend to distinguish alkaline versus acidic pH values against various colors.
137

Figure 3.6-13. Colorimetric change from pH differences caused by applying 3 V between Pt counter wires

Table 3.6-1. Color legend for universal pH indicator

The same 1:1 mixture was used to test several µBPT devices in the context of bubble
generation. However, colorimetric changes could not be observed. It was conjectured that although
pH changes were present, the extremely small microelectrodes in the nucleation core (E2 and E3;
230 µm
2
each) were unable to generate enough redox species to affect a perceivable color change.
138

The voltage between larger microelectrodes on the same die was ramped up to 3 V while the
impedance was simultaneously measured between neighboring µBPT sensing electrodes on an
annealed R10 type device (Figure 3.6-14). When 3 V was maintained, colorimetric changes could
be observed and decreases in impedance corresponded to timepoints when the colorimetric
diffusion front reached the vicinity of the µBPT electrodes (Figure 3.6-15).

Figure 3.6-14. Voltage between very large microelectrodes was ramped up to 3V while the impedance between µBPT
sensing electrodes was measured simultaneously. Device type = R10.

139

Figure 3.6-15. pH indicator revealed growing diffusion fronts at various timepoints which affected measured
impedance at neighboring µBPT devices. Device type = R10.

140

3.6.2.5 High Resolution Real-time Pressure Measurement

By capitalizing on the insights regarding bubble regimes and the noise reduction
experiments in section 3.6.2.1, experiments with 1 mmHg pressure steps ranging from 0 to 20
mmHg with 5 s step durations were performed (). This pressure waveform was continuously
applied throughout the microbubble’s entire lifetime.

Figure 3.6-16. One cycle of the pressure ladder waveform which was applied continuously through the experiment.

From a global view of the measured impedance data (Figure 3.6-17) characteristic peaks,
plateaus, and linearly decreasing regions confirmed that bubble regimes were indeed repeatable.
Small jagged patterns corresponded pressure cycle upswings and downswings. Closer examination
revealed that impedance closely tracked individual 1 mmHg pressure steps.
The average impedance per step was computed and plotted against the pressure to yield
sensitivity values for all pressure half cycles. Figure 3.6-18 is representative of the data previously
displayed in Figure 3.6-17. The average sensitivity of the µBPT device during regimes where R
2

values were greater than 0.995 was computed to be -87.5 Ω/mmHg whereas the 1σ standard
deviation in baseline noise signals was 37.9 Ω. Therefore, by dividing 3σ noise levels by the
sensitivity, 1.30 mmHg was calculated. This value was the highest-ever µBPT pressure
141

measurement resolution ever reported in the real-time measurement mode and was afforded by
noise reduction techniques and recognition of bubble regime considerations.

Figure 3.6-17. Global view of impedance magnitude versus time for a pressure-stepped experiment. Inset revealed
that 1 mmHg pressure steps could be resolved by the measured impedance. Device type = right-most A.

Figure 3.6-18. Impedance versus pressure of the half cycle data in Figure 3.6-17. Device type = Right-most A.

142

3.6.3 Discussion on Bubble Measurement
Experiments presented in this section (3.6) revealed that many unanticipated secondary
effects may have obscured the desired measurement of the pressure-sensitive bubble element in
the µBPT. The Randles circuit was confirmed to reasonably model the µBPT system (Figure
3.6-3) at baseline states – prior to bubble generation. The measurement frequency could be
optimized to frequencies at which the impedance phase most closely approached 0°. The
measurement tool’s maximum signal level (2 V) was confirmed to operate at linear
electrochemical regimes and yielded the smallest amounts of rms noise. Notably, the addition of
other equipment into the measurement circuit dramatically increased signal noise (5 orders of
magnitude; Figure 3.6-7).
Transient changes in resistivity which were attributed to pH and dissolved gas
concentration caused the theoretically expected impedance to significantly differ from the
empirically measured impedance (Figure 3.6-9). Bubble wall detachment events also affected the
measurement by creating abrupt impedance discontinuities. A new measurement mode, residence
time, was developed in this work and demonstrated up to 40× improvement in pressure resolution
compared to previous work.
Previous work [159] tracked pressure in real time with impedance measurements with 5.2
mmHg resolution as determined by minimum detectable pressure variation without using
mathematical formulae. This work was able to demonstrate plots where 1 mmHg steps were
visually resolvable in a similar fashion (Figure 3.6-17). Additionally, this work actually applied
calculations to arrive at a resolution of 1.30 mmHg. It is suggested to design sensing electrodes E1
and E4 as longer rectangular pads such that the additional surface area may reduce the
measurement noise as evidenced in Figure 3.4-3. Instrument noise was a key feature that thwarted
143

earlier attempts at satisfactory pressure transduction. Simple calculations from first principles
borrowing from equation (4) revealed that a 1 mmHg change in pressure would correspond to a
0.3 µm change in bubble length for a 0.6 kΩ change in impedance which were most likely drowned
out by noise levels in the MΩ range due to MUX or Keithley.
It is noteworthy that current µBPT measurement theory is predicated on measuring a
pressure-sensitive variable resistor as manifested by the bubble’s fluid capillaries. However,
mounting evidence in this work suggested that fluid capillaries do not exist. It may be assumed
that no significant inductance exists in the system due to the kHz frequency range and lack of
magnetic components. A capacitive sensing mode may actually be the dominant measurement
mode as suggested by the fact that pressure sensitivity was highest when the bubble approached a
thin membrane shape. At this timepoint, immediately before the transition into the third regime
described in section 3.6.2.2, the bubble may resemble a parallel plate capacitor where the shortest
bubble length served at the distance between plates to maximize capacitance.
In order to ultimately satisfy AAMI standards, measurement accuracy is required instead
of simply measurement resolution. This would require a priori conversion of measured impedance
values into pressure values. However, the pressure to impedance relationship demonstrated
noticeable hysteresis when pressure ladders were stepped upwards versus downwards.
Measurement drift was also present due to bubble dissolution. A potential solution to counteract
these effects may be to leverage known relationships between bubble volume versus time at
various pressure settings to create a computational model for expected impedance values.
Measured deviations from the model could then be ascribed to pressure disturbances to enable true
pressure sensor accuracy. However, more work is required to rigorously formulate a valid function
to which may map factors such as time, volume, pressure, or resistivity into an impedance
144

transform. Bubble detachment effects additionally obscure the measurement circuit and the
development of such computational models. Although additional efforts may be required, this
work successfully identified main obstacles to pressure measurement such as changes in resistivity
due to pH and dissolved gas, abrupt impedance discontinuities from bubble wall detachment,
electrochemical current leakage from shunt resistance, and measurement noise from
instrumentation and microelectrode size. Additionally, 1 mmHg pressure changes were able to be
resolved to represent a 5× improvement over previous work in the real-time pressure tracking
mode.
Summary and Future Directions
In this work, a variety of experiments were conducted to investigate µBPT operation in the
context of bubble generation, behavior, and measurement in order to continue developing the
technology to meet AAMI standards towards intracranial pressure monitoring. Significant
achievements were made, assumptions in previous work were challenged, unanticipated secondary
effects were discovered, and several recommendations for future work are presented.
For bubble generation, deeper investigation into the fundamental theory behind the
electrode to electrolyte interface led to the insight in harnessing the utility of asymmetrically sized
electrodes. Employing a large counter electrode allowed for polarization to occur at predominantly
the smaller working electrode interface which produced benefits such as improved volume
repeatability, 25 × improvement in temporal stability, power savings, and control over gas
chemical composition. Faraday’s law of electrolysis was confirmed linear growth relationships,
but at approximately 30% yields which may be attributed to charge spent in electric double layer
charging or bubble dissolution. Bubble generation experiments also hinted at the presence of
changes in solution resistivity. A major contribution of this work was the development of image
145

processing tools to correlate frame-by-frame measurements of bubble shape and volume to better
understand gas generation dynamics.
Bubble behavior was found to deviate strongly from desired results. Attempts to recreate
previously published results on detachment from the nucleation core were largely unsuccessful
despite usage of multiple device designs optimizing the channel to bottleneck width ratio. One
major contribution of this work therefore is the strong recommendation to remove the nucleation
core altogether in future designs since the intended goal of quasi-stability was achieved through
anodically generated O2 gas. The nucleation core simply added unnecessary design complexity
which also introduced fabrication struggles due to difficulties in photoresist removal. Constriction
valves are strongly recommended to remain as part of device design due to their reliable success
in shepherding bubbles inside the channel. Their MEMS-scale surface tension forces were
demonstrated to outcompete buoyancy forces which may be critical in the context of hydrocephalic
patients’ cranial orientation. Long term bubble behavior was non-ideal. 20 bubble generation
events were shown to mark a transition point between satisfactory and unsatisfactory volumes of
gas generation and gross metal-Parylene delamination. Alternate substrate materials such as fused
silica or silicon may improve the poor shelf-life of current design iterations. Parylene as a substrate
can be avoided because of the relaxed geometric requirements in the µBPT project as opposed to
the RRA project. Lastly, despite being a fundamental operating component of the µBPT, evidence
for fluid capillaries was not found in preliminary imaging experiments via confocal microscopy.
Bubble measurement experiments leveraged insights from behavior and generation
experiments to elucidate factors which confounded proper electrochemical impedance
measurements for pressure transduction. After confirming Randles circuit model validity and
optimizing the signal frequency and level, empirically obtained impedance values demonstrated
146

extremely poor fit when compared to theoretically derived values from equation (4). The main
culprits behind the deviation from theory are conjectured to be changes in resistivity due to pH
and dissolved gas, abrupt impedance discontinuities from bubble wall detachment events,
electrochemical current leakage from shunt resistance, and measurement noise from
instrumentation and microelectrode size. Identification and preliminary investigation into these
phenomena are a major contribution to this work in addition to the 40× improvement in pressure
resolution via the residence time technique and 5× improvement in real-time pressure tracking
resolution for thin membrane-like bubbles.
Several challenges and obstacles were encountered through µBPT development.
Instrumentation limitations plagued research progress for years. As an example, noise from the
addition of MUX or SMU into the measurement circuit was on the MΩ range whereas 1 mmHg
pressure steps corresponded to signals on the order of 100 Ω. The signal to noise ratio in the
inherited measurement setup therefore doomed successful pressure transduction from the start.
When pre-established protocols were discarded and current injection was achieved by manually
connecting and disconnecting the SMU by hand, instrumentation noise was reduced to acceptable
levels. Two examples of poor instrumentation reliability are also noteworthy. First, MUX digital
input/output control software suffered from low-level driver issues when incorporated into
LabVIEW virtual instruments and caused randomly unpredictable stoppages in multiplexing
control. Second, 10 fps imaging was accomplished via Pixelink capture software for high-volume
automated experiments, but next-frame image buffering firmware issues occurred at random and
unpredictable moments which prematurely curtailed experiments such as the 83 bubble experiment
in 3.5.2.3. Attempts at troubleshooting both issues with vendor support were futile. Device
fabrication at the sacrificial photoresist stripping step was also challenging and time-consuming to
147

optimize due to the diffusion-limited process exacerbated by nucleation core designs. Packaging
with ZIF connectors demonstrated nonideal reliability due to unexpected losses in connection at
key experimental steps.
Moving forward, several recommendations and potential future directions are presented.
Acute in vivo studies in porcine models had initially been planned in this work with preliminary
µBPT device designs compatible with the Integra Camino Bolt system. Obtaining acute in vivo
data will serve as a major milestone, but it should be noted that chronic studies may be challenging
due to device degradation behavior shown in 3.5.2.3. It is therefore suggested to simultaneously
invest time in efforts to untangle fundamental issues in device design and operation. For instance,
it is strongly recommended to generate a working mathematical model to describe the measured
impedance with respect to changes in resistivity due to pH and dissolved gas, abrupt impedance
discontinuities from bubble wall detachment events, electrochemical current leakage from shunt
resistance, and measurement noise from instrumentation and microelectrode size. Removal of the
nucleation core is strongly suggested as well as a rehaul of MUX design to acceptable noise levels.
Noise may also be reduced by increasing the size of the sensing electrodes. Their current 50 × 50
µm surface area is simply a vestigial parameter which may be altered in future designs. Adding
large microfabricated pads (larger than the ones in figure Figure 3.6-15) to serve as counter
electrodes is also suggested. After bubble generation, the measurement fixture in Figure 3.3-6
required manual closing of the stopcock next to the CE if pressurized experiments are desired
which may shift the fixture’s position and compromise baseline images required for image
processing. Non-Parylene substrates are highly recommended for testing due to the serious
weakness in device longevity from gross delamination. Rapid prototyping with
polydimethylsiloxane µBPT geometries plasma bonded onto metal patterned glass substrate may
148

accelerate research progress in elucidating the various confounding factors in impedance
measurement. These various suggestions may hopefully inform next-generation µBPT
development to successfully monitor ICP and provide clinical value for hydrocephalus patients.
149

References
[100] L. Yu, B. J. Kim, and E. Meng, “Chronically implanted pressure sensors: Challenges and
state of the field,” Sensors (Switzerland), vol. 14, no. 11, pp. 20620–20644, 2014.
[101] M. Refai et al., “The values of intrapleural pressure before the removal of chest tube in non-
complicated pulmonary lobectomies,” Eur. J. Cardio-thoracic Surg., vol. 41, no. 4, pp.
831–833, 2012.
[102] T. Dieterle, “Blood pressure measurement - An overview,” Swiss Med. Wkly., vol. 142, no.
1, pp. 1–9, 2012.
[103] J. A. Potkay, “Long term, implantable blood pressure monitoring systems,” Biomed.
Microdevices, vol. 10, no. 3, pp. 379–392, 2008.
[104] N. Egorova, J. K. Giacovelli, G. Greco, A. Gelijns, K. C. Kent, and J. F. Mckinsey,
“National outcomes for the treatment of ruptured abdominal aortic aneurysm,” J Vasc Surg,
vol. 48, no. 5, pp. 1092–1100, 2008.
[105] F. Springer, R. W. Günther, and T. Schmitz-Rode, “Aneurysm sac pressure measurement
with minimally invasive implantable pressure sensors: An alternative to current surveillance
regimes after EVAR?,” Cardiovasc. Intervent. Radiol., vol. 31, no. 3, pp. 460–467, 2008.
[106] M. G. Allen, “Micromachined endovascularly-implantable wireless aneurysm pressure
sensors: From concept to clinic,” Dig. Tech. Pap. - Int. Conf. Solid State Sensors Actuators
Microsystems, TRANSDUCERS ’05, vol. 1, pp. 275–278, 2005.
[107] M. A. Fonseca, M. G. Allen, J. Kroh, and J. White, “Flexible wireless passive pressure
sensors for biomedical applications,” Solid-State Sensor, Actuator, Microsystems Work., no.
1, pp. 37–42, 2006.

150

[108] B. Liang et al., “A novel implantable saw sensor for blood pressure monitoring,” 2011 16th
Int. Solid-State Sensors, Actuators Microsystems Conf. TRANSDUCERS’11, pp. 2184–2187,
2011.
[109] U. C. Hoppe et al., “Chronic monitoring of pulmonary artery pressure in patients with
severe heart failure: Multicentre experience of the monitoring Pulmonary Artery Pressure
by Implantable device Responding to Ultrasonic Signal (PAPIRUS) II study,” Heart, vol.
95, no. 13, pp. 1091–1097, 2009.
[110] R. Schlierf, M. G??rtz, T. S. Rode, W. Mokwa, U. Schnakenberg, and K. Trieu, “Pressure
sensor capsule to control the treatmnent of abdominal aorta aneurisms,” Dig. Tech. Pap. -
Int. Conf. Solid State Sensors Actuators Microsystems, TRANSDUCERS ’05, vol. 2, pp.
1656–1659, 2005.
[111] O. Romain et al., “RFID implantable pressure sensor for the follow-up of abdominal aortic
aneurysm stented,” 6th Int. Conf. Des. Technol. Integr. Syst. Nanoscale Era, DTIS’11 - Tech.
Progr., 2011.
[112] N. J. Cleven et al., “A novel fully implantable wireless sensor system for monitoring
hypertension patients,” IEEE Trans. Biomed. Eng., vol. 59, no. 12 PART2, pp. 3124–3130,
2012.
[113] J. E. Kalvesten, L. Smith, L. Tenerz, and G. Stemme, “First surface micromachined pressure
sensor for cardiovascular pressure measurements,” Proc. IEEE Micro Electro Mech. Syst.,
no. c, pp. 574–579, 1998.
[114] X. Chen and D. Young, “Robust implantable blood pressure sensor packaging for long-term
laboratory animals monitoring,” Proc. IEEE Sensors, pp. 1–3, 2017.

151

[115] A. Bulbul, A. Tandar, A. Patel, and H. Kim, “Micro hydraulic pressure sensing stent,” 2015
Transducers - 2015 18th Int. Conf. Solid-State Sensors, Actuators Microsystems,
TRANSDUCERS 2015, pp. 105–108, 2015.
[116] S.-H. Lo, M.-X. Xu, and Y.-J. Yang, “A Wireless Parylene-Based Cardiovascular Pressure
Sensor with Mxene Film,” 2019 20th Int. Conf. Solid-State Sensors, Actuators
Microsystems Eurosensors XXXIII (TRANSDUCERS EUROSENSORS XXXIII), no. June,
pp. 2231–2234, 2019.
[117] M. N. Gulari, M. Ghannad-Rezaie, P. Novelli, N. Chronis, and T. C. Marentis, “An
implantable X-ray-based blood pressure microsensor for coronary in-stent restenosis
surveillance and prevention,” J. Microelectromechanical Syst., vol. 24, no. 1, pp. 50–60,
2015.
[118] M. N. Gulari, M. Ghannad-Rezaie, P. Novelli, N. Chronis, and T. C. Marentis, “An X-ray
detectable pressure microsensor for monitoring coronary in-stent restenosis,” Proc. IEEE
Int. Conf. Micro Electro Mech. Syst., pp. 893–896, 2014.
[119] L. Tang, J. Liu, and B. Yang, “An ultraflexible microbubble blood pressure sensor for
interventional treatment,” Tech. Dig. - Int. Electron Devices Meet. IEDM, vol. 2016-Febru,
no. 800, pp. 29.7.1-29.7.4, 2015.
[120] J. T. Silverstone, G. P. Smith, and A. J. Stunkard, “Gastric pressures recorded by a
telemetering capsule,” Am. J. Dig. Dis., vol. 13, no. 7, pp. 615–618, 1968.
[121] B. Jacobson, “Endoradiosonde techniques-A survey,” Med. Electron. Biol. Eng., vol. 1, no.
2, pp. 165–180, 1963.
[122] B. Jacobson and L. Nordberg, “Endoradiosondes for pressure telemetering.,” Ire Trans.
Biomed. Electron., vol. 8, pp. 192–196, 1961.
152

[123] C. Browning, R. Valori, D. L. Wingate, and D. Maclachlan, “a New Pressure-Sensitive
Ingestible Radio-Telemetric Capsule,” Lancet, vol. 318, no. 8245, pp. 504–505, 1981.
[124] Q. Shi, J. Wang, D. Chen, J. Chen, J. Li, and K. Bao, “In Vitro and In Vivo characterization
of wireless and passive micro system enabling gastrointestinal pressure monitoring,”
Biomed. Microdevices, vol. 16, no. 6, pp. 859–868, 2014.
[125] A. Gupta, G. Defreitas, and G. E. Lemack, “The reproducibility of urodynamic findings in
healthy female volunteers: Results of repeated studies in the same setting and after short-
term follow-up,” Neurourol. Urodyn., vol. 23, no. 4, pp. 311–316, 2004.
[126] M. N. Dakurah, C. Koo, W. Choi, and Y. H. Joung, “Implantable bladder sensors: A
methodological review,” Int. Neurourol. J., vol. 19, no. 3, pp. 133–141, 2015.
[127] P. Jourand and R. Puers, “The BladderPill: An in-body system logging bladder pressure,”
Sensors Actuators, A Phys., vol. 162, no. 2, pp. 160–166, 2010.
[128] M. Bakula, A. Soebadi, D. De Ridder, and R. Puers, “The Bladder Pill: Developments
Toward Bladder Pressure Measurement in Wake Mini-pigs,” Procedia Eng., vol. 168, pp.
193–196, 2016.
[129] I. Clausen, S. T. Moe, L. G. W. Tvedt, A. Vogl, and D. T. Wang, “A miniaturized pressure
sensor with inherent biofouling protection designed for in vivo applications,” Proc. Annu.
Int. Conf. IEEE Eng. Med. Biol. Soc. EMBS, pp. 1880–1883, 2011.
[130] I. Clausen, L. G. W. Tvedt, S. Moe, and A. Vogl, “The effect of true human synovial fluid
on the functionality of an in vivo pressure sensor element,” Proc. IEEE Sensors, pp. 1–4,
2013.

153

[131] I. Clausen, L. G. W. Tvedt, A. Hellandsvik, D. K. W. Rognlien, and T. Glott, “An in vivo
MEMS sensor system for percutaneous measurement of urinary bladder,” Proc. Annu. Int.
Conf. IEEE Eng. Med. Biol. Soc. EMBS, pp. 1857–1860, 2017.
[132] I. Clausen, L. G. W. Tvedt, and T. Glott, “Measurement of urinary bladder pressure: A
comparison of methods,” Sensors (Switzerland), vol. 18, no. 7, 2018.
[133] S. J. A. Majerus, S. L. Garverick, M. A. Suster, P. C. Fletter, and M. S. Damaser, “Wireless,
ultra-low-power implantable sensor for chronic bladder pressure monitoring,” ACM J.
Emerg. Technol. Comput. Syst., vol. 8, no. 2, 2012.
[134] S. Majerus et al., “Wireless bladder pressure monitor for closed-loop bladder
neuromodulation,” Proc. IEEE Sensors, pp. 7–9, 2017.
[135] J. Jin, “Glaucoma,” JAMA, vol. 311, no. 18, p. 1934, May 2014.
[136] E. Meng, E. Yoon, and C. Gutierrez, “MEMS ENABLED TECHNOLOGIES FOR
OCULAR MONITORING AND THERAPY University of Southern California , Los
Angeles , USA and Verily Life Sciences , South San Francisco , USA,” pp. 379–382, 2017.
[137] M. Leonardi, E. M. Pitchon, A. Bertsch, P. Renaud, and A. Mermoud, “Wireless contact
lens sensor for intraocular pressure monitoring: Assessment on enucleated pig eyes,” Acta
Ophthalmol., vol. 87, no. 4, pp. 433–437, 2009.
[138] G. E. Dunbar, B. Y. Shen, and A. A. Aref, “The Sensimed Triggerfish contact lens sensor:
Efficacy, safety, and patient perspectives,” Clin. Ophthalmol., vol. 11, pp. 875–882, 2017.
[139] G. Z. Chen, I. S. Chan, L. K. K. Leung, and D. C. C. Lam, “Soft wearable contact lens
sensor for continuous intraocular pressure monitoring,” Med. Eng. Phys., vol. 36, no. 9, pp.
1134–1139, 2014.

154

[140] P. J. Chen, S. Saati, R. Varma, M. S. Humayun, and Y. C. Tai, “Wireless intraocular
pressure sensing using microfabricated minimally invasive flexible-coiled LC sensor
implant,” J. Microelectromechanical Syst., vol. 19, no. 4, pp. 721–734, 2010.
[141] A. Koutsonas, P. Walter, G. Roessler, and N. Plange, “Implantation of a novel telemetric
intraocular pressure sensor in patients with glaucoma (ARGOS study): 1-year results,”
Investig. Ophthalmol. Vis. Sci., vol. 56, no. 2, pp. 1063–1069, 2015.
[142] P.-J. Chen, D. Rodger, E. Meng, M. Humayun, and Y.-C. Tai, “Implantable Unpowered
Parylene MEMS Intraocular Pressure Sensor,” pp. 3–6, 2006.
[143] Po-Jui Chen, D. C. Rodger, S. Saati, M. S. Humayun, and Yu-Chong Tai, “Microfabricated
Implantable Parylene-Based Wireless Passive Intraocular Pressure Sensors,” J.
Microelectromechanical Syst., vol. 17, no. 6, pp. 1342–1351, 2008.
[144] P. J. Chen, D. C. Rodger, E. M. Meng, M. S. Humayun, and Y. C. Tai, “Surface-
micromachined parylene dual valves for on-chip unpowered microflow regulation,” J.
Microelectromechanical Syst., vol. 16, no. 2, pp. 223–231, 2007.
[145] R. K. Narayan et al., “Intracranial pressure: to monitor or not to monitor? A review of our
experience with severe head injury,” J. Neurosurg., vol. 56, no. 5, pp. 650–659, 1982.
[146] Y. H. Yau, I. R. Piper, R. E. Clutton, and I. R. Whittle, “Experimental evaluation of the
Spiegelberg intracranial pressure and intracranial compliance monitor. Technical note.,” J.
Neurosurg., vol. 93, no. December, pp. 1072–1077, 2000.
[147] C. Li, P. M. Wu, L. A. Shutter, and R. K. Narayan, “Dual-mode operation of flexible
piezoelectric polymer diaphragm for intracranial pressure measurement,” Appl. Phys. Lett.,
vol. 96, no. 5, pp. 2008–2011, 2010.

155

[148] S. H. A. Rahman, N. Soin, and F. Ibrahim, “Analysis of MEMS diaphragm of piezoresistive
intracranial pressure sensor,” IECBES 2014, Conf. Proc. - 2014 IEEE Conf. Biomed. Eng.
Sci. “Miri, Where Eng. Med. Biol. Humanit. Meet,” no. December, pp. 681–685, 2014.
[149] M. Mohamad, N. Soin, and F. Ibrahim, “Electromechanical Piezoresistive Sensing of
Graphene-based Intracranial Pressure Sensor,” 2018 IEEE 8th Int. Nanoelectron. Conf.
INEC 2018, no. 100, pp. 25–26, 2018.
[150] S. H. A. Rahman, N. Soin, and F. Ibrahim, “Piezoresistive Effect of Interdigitated Electrode
Spacing Graphene-based MEMS Intracranial Pressure Sensor,” 2018 IEEE 8th Int.
Nanoelectron. Conf. INEC 2018, no. 1, pp. 27–28, 2018.
[151] Z. Z. Wu, N. Bhattacharjee, C. Y. Li, J. A. Hartings, R. K. Narayan, and C. H. Ahn, “A new
intracranial pressure sensor on polyimide lab-on-a-tube using exchanged polysilicon
piezoresistors,” 2013 Transducers Eurosensors XXVII 17th Int. Conf. Solid-State Sensors,
Actuators Microsystems, TRANSDUCERS EUROSENSORS 2013, no. June, pp. 1779–1782,
2013.
[152] S. Jetzki et al., “A multisensor implant for continuous monitoring of intracranial pressure
dynamics,” IEEE Trans. Biomed. Circuits Syst., vol. 6, no. 4, pp. 356–365, 2012.
[153] Z. Wu, H. Jiang, Y. Guo, C. Zhang, W. Jia, and Z. Wang, “Live demonstration: Wireless
intracranial pressure monitoring system based on an air pressure sensor,” 2017 IEEE
Biomed. Circuits Syst. Conf. BioCAS 2017 - Proc., vol. 2018-Janua, no. 61474070, p. 1,
2018.
[154] H. Jiang, Y. Guo, Z. Wu, C. Zhang, W. Jia, and Z. Wang, “Implantable Wireless Intracranial
Pressure Monitoring Based on Air Pressure Sensing,” IEEE Trans. Biomed. Circuits Syst.,
vol. 12, no. 5, pp. 1076–1087, 2018.
156

[155] A. Shapero and Y. C. Tai, “Parylene-oil-encapsulated low-drift implantable pressure
sensors,” Proc. IEEE Int. Conf. Micro Electro Mech. Syst., vol. 2018-Janua, no. January,
pp. 47–50, 2018.
[156] D. P. Leung, D. J. McCormick, S. C. Malpas, and D. M. Budgett, “Reducing Drift in
Implantable Pressure Sensors,” IEEE Sens. J., vol. 19, no. 7, pp. 2458–2465, 2019.
[157] J. D. Suter, C. J. Hohimer, J. M. Fricke, J. Christ, H. Kim, and A. T. Evans, “Principles of
meniscus-based MEMS gas or liquid pressure sensors,” J. Microelectromechanical Syst.,
vol. 22, no. 3, pp. 670–677, 2013.
[158] I. V Uvarov, S. S. Lemekhov, A. E. Melenev, and V. B. Svetovoy, “Exploding microbubbles
driving a simple electrochemical micropump,” J. Micromechanics Microengineering, 2017.
[159] C. a Gutierrez and E. Meng, “Subnanowatt Microbubble Pressure Transducer,” Solid-State
Sensors, Actuators Microsystems Work., pp. 57–60, 2010.
[160] P. Kuo et al., “A Remotely Controlled Locomotive IC Driven by Electrolytic Bubbles and
Wireless Powering,” 2014 IEEE Int. Solid-State Circuits Conf. Dig. Tech. Pap., no. Dc, pp.
322–324, 2014.
[161] S. W. Lee, C. Jeong, and S. S. Yang, “The Fabrication of a Micro Injector Actuated by
Boiling and / or Electrolysis,” Proc. MEMS 98. IEEE. Elev. Annu. Int. Work. Micro Electro
Mech. Syst. An Investig. Micro Struct. Sensors, Actuators, Mach. Syst. (Cat. No.98CH36176,
1998.
[162] Lord Rayleigh, “VIII. On the pressure developed in a liquid during the collapse of a
spherical cavity,” London, Edinburgh, Dublin Philos. Mag. J. Sci., vol. 34, no. 200, pp. 94–
98, Aug. 1917.

157

[163] M. S. Plesset and A. Prosperetti, “Bubble Dynamics and Cavitation ~8100,” Ann. Rev. Fluid
Mech, vol. 9, pp. 145–85, 1977.
[164] B. Ran and J. Katz, “The response of microscopic bubbles to sudden changes in the ambient
pressure,” J. Fluid Mech., vol. 224, no. 4, pp. 91–115, 1991.
[165] B. J. Kim, C. A. Gutierrez, and E. Meng, “Parylene-Based Electrochemical-MEMS Force
Sensor for Studies of Intracortical Probe Insertion Mechanics,” J. Microelectromechanical
Syst., vol. 24, no. 5, pp. 1534–1544, 2015.
[166] C. A. Gutierrez, C. McCarty, B. Kim, M. Pahwa, and E. Meng, “An implantable all-parylene
liquid-impedance based MEMS force sensor,” Proc. IEEE Int. Conf. Micro Electro Mech.
Syst., pp. 600–603, 2010.
[167] L. Yu and E. Meng, “A dual mode microbubble pressure and flow sensor,” 2016 IEEE 29th
Int. Conf. Micro Electro Mech. Syst., no. January, pp. 325–328, 2016.
[168] E. Meng, P. Y. Li, and Y. C. Tai, “Plasma removal of Parylene C,” J. Micromechanics
Microengineering, vol. 18, no. 4, 2008.
[169] B. J. Kim, B. Chen, M. Gupta, and E. Meng, “Formation of three-dimensional Parylene C
structures via thermoforming,” J. Micromechanics Microengineering, vol. 24, no. 6, p.
065003, 2014.
[170] D. W. Kumsa, N. Bhadra, E. M. Hudak, S. C. Kelley, D. F. Untereker, and J. T. Mortimer,
“Electron transfer processes occurring on platinum neural stimulating electrodes: a tutorial
on the i ( V e ) profile,” J. Neural Eng., vol. 13, no. 5, p. 052001, 2016.
[171] C. a Gutierrez and E. Meng, “a Subnanowatt Microbubble Pressure Sensor Based on
Electrochemical Impedance Transduction in a Flexible All- Parylene Package,”
Measurement, pp. 549–552, 2011.
158

[172] E. L. Cussler, Diffusion : mass transfer in fluid systems, 3rd ed. Cambridge ; Cambridge
University Press.
[173] A. Volanschi, D. Oudejans, W. Olthuis, and P. Bergveld, “Gas phase nucleation core
electrodes for the electrolytical method of measuring the dynamic surface tension in
aqueous solutions,” Sensors Actuators B Chem., vol. 35, no. 1–3, pp. 73–79, 1996.
[174] L. Yu and E. Meng, “a Microbubble Pressure Transducer With Bubble Nucleation,” 27TH
Int. Conf. MEMS, no. 1, pp. 104–107, 2014.
[175] M. M. Weislogel and S. Lichter, “Capillary flow in an interior corner,” J. Fluid Mech., vol.
373, no. 1998, pp. 349–378, 1998.
[176] J. R. MaCdonald, “Double layer capacitance and relaxation in electrolytes and solids,”
Trans. Faraday Soc., vol. 66, p. 943, 1970.
[177] P. S. Epstein and M. S. Plesset, “On the Stability of Gas Bubbles in Liquid-Gas Solutions,”
vol. 1505, no. 1950, 2009.

159

Introduction
Research collaboration between the Biomedical Microsystems Laboratory at the
University of Southern California and the RF-Frequency Electronics Group at the National
Institute of Standards and Technology (NIST) was enabled through the NSF INTERN program.
NIST is a government laboratory under the U.S. Department of Commerce which has largely
developed and standardized conventions for various scientific disciplines. Travel to the campus
in Boulder, CO allowed for a fruitful exchange of research ideas and preliminary results, but the
ensuing global pandemic necessitated remote collaboration as well. For the first time, NIST
techniques have been applied in the context of biomedical thin-film Parylene C characterization
and will be presented throughout this chapter

4.1.1 Motivation for Radio Frequency Investigation
Many biomedical devices such as the rat retinal array (RRA) and microbubble pressure
transducer (µBPT) have been enabled through thin-film Parylene C micromachining. A plethora
of other devices and applications leverage similar MEMS technology including, but not limited to,
neural prostheses [178] [179], hippocampal penetrating neural probes [26], peripheral nerve

RADIO-FREQUENCY DIELECTRIC SPECTROSCOPY ON
PARYLENE C VIA MICROWAVE MICROFLUIDICS
160

interfaces [180], temperature sensors [42], implantable sensors [181], and cardiovascular implants
[182].
Many characterization techniques exist for such thin-film devices, such as Fourier
transform infrared spectroscopy with attenuated total reflectance and X-ray scattering to study
water transport [183], soak testing with leakage current measurements [184], accelerated lifetime
soak testing at elevated temperatures [185], scanning electron microscopy and energy dispersive
X-ray spectroscopy for reactive oxygen species measurement [186], and water vapor transmission
rate measurements [187]. However, electrochemical impedance spectroscopy (EIS), is by far one
of the most widely-employed and effective techniques in this area of research [38], [188]–[191].
Certain limitations exist with the EIS technique but may be overcome through advantages
associated with RF dielectric spectroscopy techniques from NIST. The main limitation stems from
the limited frequency spectrum in EIS. Whereas EIS instrumentation typically may only allow
accurate measurements up to approximately 1 MHz, the NIST technique has demonstrated use up
to 110 GHz. At higher frequencies, additional physical phenomena such as the dipolar relaxation
of water molecules can be measured. Such measurements may reveal insights on critical device
failure modes such as soaking-induced delamination rates or may be able to isolate contributions
of water versus ion permeation into the Parylene film. EIS requires samples to be submerged into
an electrolyte solution, whereas the NIST technique traditionally is performed in air but may also
incorporate microfluidic structures to soak samples. Additionally, thoroughly validated circuit
models exist in the NIST technique for electrode-electrolyte interfaces [192] and rigorous
microwave uncertainty framework (MUF) techniques exist to quantify measurement error. Table
4.1-1 summarizes comparison between EIS and NIST measurement techniques.
161

Additionally, many Parylene-based works employ radio frequency operation, especially
for wireless power and data transmission. Wireless capabilities reduce risk of infection compared
to percutaneous leads and allow increase patient mobility. Table 4.1-2 lists several Parylene-based
works and their highest frequency usage. This information is plotted in Figure 4.1-1 to graphically
illustrate the breadth of these types of devices.
As future research trends point towards data-driven wireless operation, there is a strong
need to better understand how Parylene behaves at these higher frequencies. The Parylene C base
monomer is nonpolar which will affect the real and imaginary components of the permittivity at
higher frequency ranges. Microwave frequency measurements to capture such effects are
particularly challenging due to limited availability of extremely high frequency (EHF) vector
network analyzer equipment. Because developing accurate models to describe these characteristics
may assist in future device design and improved clinical care, there is great potential and value in
investigating this vein of research.
Electrochemical Impedance Spectroscopy RF Dielectric Spectroscopy
• Typical frequency band: 0.1 Hz – 1 MHz • Typical frequency band: 10 kHz – 110 GHz
• Can model electric double layer (EDL) effects • Can model EDL, H 2O relaxation, and more
• Can only measure wetted samples • Can measure wetted and dry samples
• Electrochemical cell geometry may propagate
measurement error
• Geometric is strictly defined through coplanar
waveguide (CPW) structures
• Measurement uncertainty reporting standards vary
across research groups
• Rigorously defined and standardized microwave
uncertainty framework (MUF)

Table 4.1-1. Comparison of EIS and NIST measurement techniques
162

A [193] – B [194] – C [195] – D [196] – E [197] – F [198] – G [199] – H [200] – I [201] – J [202]
– K [203] – L [204] – M [116] – N [205] – O [206] – P [207] – Q [208]
Index Title Max Frequency
A
A Wireless, Low-Drift, Implantable Intraocular Pressure Sensor with Parylene-on-
oil Encapsulation [193]
915 MHz
B
Fabrication of a flexible wireless pressure sensor for intravascular blood pressure
monitoring [194]
211 MHz
C Implantable RF-Coiled Chip Packaging [195] 125 kHz
D
Parylene Interposer as Thin Flexible 3-D Packaging Enabler for Wireless
Applications [196]
8 GHz
E
Wireless Implantable Intraocular Pressure Sensor with Parylene-Oil-
Encapsulation and Forward-Angled RF Coil [197]
915 MHz
F
Passive, wireless transduction of electrochemical impedance across thin-film
microfabricated coils using reflected impedance [198]
5 MHz
G A Contactless Electrochemical Impedance Measurement Method [199] 2.2 MHz
H
Investigation of Parylene-C on the Performance of Millimeter-Wave Circuits
[200]
67 GHz
I
Development of a multilayered polymeric DNA biosensor using radio frequency
technology with gold and magnetic nanoparticles [201]
900 MHz
J Parylene-based integrated wireless single-channel neurostimulator [202] 500 kHz
K
Liquid pressure wireless sensor based on magnetostrictive microwires for
applications in cardiovascular localized diagnostic [203]
1.2 GHz
L Parylene-Based Fold-and-Bond Wireless Pressure Sensor [204] 320 MHz
M
A Wireless Parylene-Based Cardiovascular Pressure Sensor with Mxene Film
[116]
260 MHz
N Wireless Parylene-Based Retinal Implant [205] 160 MHz
O
Folded Dual-band (2.4/5.2GHz) Antenna Fabricated on Silicon Suspended
Parylene Membrane [206]
5.2 GHz
P
Flexible MEMS Inductors Based on Parylene-FeNi Compound Substrate for
Wireless Power Transmission System [207]
140 MHz
Q
A Compact Parylene-Coated WLAN Flexible Antenna for Implantable Electronics
[208]
5.8 GHz

Table 4.1-2. Table of Parylene-based works employing high frequency signals
163

Figure 4.1-1. Plot of highest frequencies used in several Parylene-based work

4.1.2 Challenges of RF Measurements
Although the RF techniques at NIST present key advantages over EIS, certain challenges
also exist in this technique. The cause for these challenges is rooted in the fundamental physical
relationship between electromagnetic frequency and wavelength. In a perfect vacuum, equation
(1) illustrates the inverse relationship between frequency ( λ) and wavelength ( ν), where c is the
speed of light in a vacuum (3.0×10
8
m/s). For this reason, RF measurement techniques may be
synonymously referred to as microwave measurement techniques.
𝜆 =
𝐶 𝜈 ()
This fundamental relationship causes measured RF signals to be unrepresentative of the
phenomena occur at the device under test (DUT) due to impedance mismatches occurring
164

throughout the overall measurement setup. Therefore, dedicated calibration processes are required
to correct the raw measurements into values indicative of the physics at the DUT.
To better illustrate this concept, it is helpful to revisit the operating mechanism of EIS via
potentiostatic equipment (Figure 4.1-2). Suppose the Gamry Reference 600 potentiostat was used
to send a 1 MHz signal with 25 mV. Although the signal is varying in time, for any specific point
in time, the voltage at all points along the cable, connectors, DUT, and computer can be assumed
to be equal as shown in equation (2).
𝑉 𝐴 = 𝑉 𝐵 = 𝑉 𝐶 = 𝑉 𝐷 ()

Figure 4.1-2. Illustration of EIS potentiostat where frequency signals may be assumed to present uniform voltage
values at all points in space per time

165

If the tool was hypothetically capable of sending a 1 GHz signal, then equation (2) would
no longer be valid. The electromagnetic wave traverses its peaks and troughs throughout the length
of the cabling, connectors, DUT, etc… such that the obtained measurement at the measurement
port is obscured. For any single point in time, the voltage varies across space. It is however possible
to mathematically describe and correct for these measurement inaccuracies which arise as a
consequence of such fundamental physics. Over decades, the field of high frequency electronics
has developed specialized calibration techniques and device geometries to overcome these issues.
This work will employ these techniques for the first time in the context of analyzing Parylene C
for characterization of biomedical device investigations.
4.1.3 Approach and Goals
A blend of NIST and USC techniques and expertise were used to design, fabricate, and test
samples for high-frequency interrogation of thin-film Parylene C. From such experiments, the goal
was to successfully build mathematical model elements to describe Parylene C permittivity as a
function of various test conditions such as unannealed versus annealed, soaking in DI water,
soaking in PBS, as well as soaking at various durations. From these models it may be possible to
isolate electric double layer effects from ions versus water intrusion, interpret relevant time
constants for underlying physical phenomena, or investigate capacitive versus conductive
contributions in electrochemical crosstalk. Owing to the modular nature of this research, circuit
elements produced in this work may smoothly fit into overall block diagrams developed by
previous work [192] which may be further refined and built up by future work as well.
Another main goal and contribution of this work arose from studying the underlying theory
enough to command a working knowledge of NIST techniques. An entirely distinct discipline and
body of work dedicated to microwave measurements exists and key concepts have been distilled
166

and catalogued for future students. Instead of beginning a journey towards RF measurement
techniques alone, the reader is therefore invited to review the following primer to accelerate their
learning.
Primer on Microwave Theory
A thorough description of the entire field of microwave measurements is outside the scope
of this thesis. For an in-depth text that provides firm theoretical grounding on microwave
engineering as a whole, the reader is referred to Pozar’s textbook [209]. For more granular nuances
related to microwave microfluidics and techniques similar to those in this work, Dr. Charles A. E.
Little’s doctoral thesis [210] is highly recommended.
4.2.1 Overall Workflow
Figure 4.2-1 provides a simplified roadmap of the overall workflow employed in
microwave measurement techniques. As described in section 4.1.2, raw measurements will be
obscured due to the fundamental relationship between wavelength and frequency. These raw
measurements undergo calibration techniques such as the multi-line, thru, reflect (MTRL)
calibration [211]–[213] and the series-resistor calibration [214], [215]. Once calibrated, they can
also be mathematically processed into usable forms which represent the resistance, inductance,
and capacitance (RLCG) of the DUT. Circuit elements can then be fitted to these RLCG values to
make physically meaningful interpretations of the data. This fitting bears similarities to the fitting
techniques in EIS and familiar model elements such as the double layer capacitor or constant phase
element may be applied. Other elements which cannot be employed in EIS, such as water dipole
relaxation admittances, are applicable only for microwave measurements due to the broader
measurement bandwidth. Once fitting is complete, a more complete picture of the Parylene-based
167

DUT system may be obtained. Extracted data such as water relaxation time constants can help
elucidate when or where failure modes begin. They may additionally inform device design in
Parylene-based RF applications. A video lecture series has also been recorded to facilitate learning
and may be found in the appendices.

Figure 4.2-1. Simplified workflow diagram for NIST microwave measurement techniques

4.2.2 Raw Measurements
4.2.2.1 On-Wafer Microwave Measurement Equipment
In order to understand the physical significance of raw microwave measurements, it is
beneficial to first review the equipment involved in obtaining such measurements. Unlike the
potentiostat tool used in EIS, a vector network analyzer (VNA) is the main piece of measurement
equipment. A VNA many have many measurement ports, but 2-port measurements were used
throughout this work. Measurement cables, probes and cable-to-probe connectors may be attached
with the DUT as illustrated in Figure 4.2-2.
The DUTs in this work were devices fabricated into a dedicated geometry known as
coplanar waveguides (CPW). 3 metal strips are fabricated on the same plane. The wider strips at
the top and bottom of Figure 4.2-2 were ground planes. The center strip conducted the RF signal
from the probes and was denoted as the center conductor. CPWs may be composed of purely metal
168

atop of substrate material, coated with Parylene C, or also fitted with microfluidic structures as
shown in Figure 4.2-3.

Figure 4.2-2. Schematic illustration of a typical microwave measurement setup

Figure 4.2-3. Photograph of microwave measurement setup with microfluidic fitting laid on top of CPW
169

4.2.2.2 Introduction to S Parameters
The previously described setup may then proceed to obtain raw data. More specifically,
these data are described as scattering parameters, or S parameters. S parameters represent the ratio
between the magnitude of phase of power waves at various input and output port combinations.
Suppose the left-most port, arbitrarily defined as port 1, sent a power wave signal, a1, of certain
amplitude and frequency into the measurement system which is treated with black box formalism
in Figure 4.2-4.

Figure 4.2-4. A signal from port 1 (a 1) is sent into the measurement system
At this point, two potential outcomes are possible. First, the a1 signal may be transmitted
through the system, and a measurement power wave signal, b2, may be obtained at port 2 (Figure
4.2-5). To relate the two wave quantities a1 and b2, an S parameter S21 is defined as shown in
equation (2).
𝑏 2
= 𝑆 21
𝑎 1
()

Figure 4.2-5. A signal from port 1 (a 1) is sent into the measurement system and the transmitted signal at port 2 (b 2)
is measured
170

S parameters are complex values consisting of a real and imaginary component to describe
the magnitude attenuation and frequency shift between a1 and b2 through Euler relations per
frequency. Another possible outcome is for the a1 signal to be reflected from the system such that
the measurement signal b1 is obtained at port 1. Similarly, S11 is defined via equation (3) such that:
𝑏 1
= 𝑆 11
𝑎 1
()

Figure 4.2-6. A signal from port 1 (a 1) is sent into the measurement system and the reflected signal at port 1 (b 1) is
measured

An a2 signal may similarly be sent from port 2 into the system, and the transmitted b1 signal
may be measured at port 1 or the reflected b2 signal may be measured at port 2. Although these
signals were shown in isolation to facilitate conceptual illustration, all signals are sent and
measured simultaneously during actual equipment operation as shown in Figure 4.2-7. Therefore,
equation (4) and (5) represent a linear system of equations to fully specify the relationship between
all signals.
𝑏 1
= 𝑆 11
𝑎 1
+ 𝑆 12
𝑎 2
()
𝑏 2
= 𝑆 21
𝑎 1
+ 𝑆 22
𝑎 2
()
171

Figure 4.2-7. Illustration of all input and output signals during microwave measurement operation

Equations (4) and (5) may be transformed into matrix form as shown in equation (6). A
matrix, M, may be defined by equation (7) to contain the set of these S parameters across all
frequency points. M therefore represents the raw measurements described in Figure 4.2-1. In terms
of implementation, S parameters are stored in .csv files as shown in Figure 4.2-8.

[
𝑏 1
𝑏 2
] = [
𝑆 11
𝑆 12
𝑆 21
𝑆 22
] [
𝑎 1
𝑎 2
] ()

𝑴 = [
𝑆 11
𝑆 12
𝑆 21
𝑆 22
] ()
172

Figure 4.2-8. Example screenshot of S parameter measurements from a .csv file

S parameters are obtained for every DUT. Several types of DUTs exist and their differences
will be introduced shortly. It is also noteworthy that other formalisms besides S parameters exist.
For example, T parameters, X parameters, Y parameters, Z parameters, and ABCD parameters
exist and interconversion between S parameters through mathematical formulae are possible.
However, S parameters bear the key advantage of conceptual simplicity. At this point in the
workflow, it must be noted that raw S parameter measurements are effectively meaningless
because they merely represent input and output ratios across measurement ports. Valuable
information may be hidden inside them, but they require calibration techniques.
4.2.3 Calibrations
4.2.3.1 Multi-Line Thru Reflect (MTRL) Calibration
Many calibration techniques exist for microwave calibrations, each with their own specific
set of advantages and disadvantages. For instance, the open-short-load-thru [216] and line-reflect-
173

match [217] have been used in other instances, but the thru-reflect-line (TRL) technique is favored
in this work because it is considered to be one of the most accurate techniques due to its basis in
circuit theory and amenability for on-wafer probe-tip measurements [214]. Usage of multiple lines
assists in calculating error propagation through MUF algorithms in MTRL techniques, but the
underlying theory remains the same [213].
The key to understanding MTRL lies in recognizing that the former black box treatment of
the measurement system in section 4.2.2.2 may be expanded to introduce the concept of error
boxes. The measurement matrix, M, may be designated as the matrix product of 3 distinct matrices
(8).
𝑴 = 𝑿 𝑻 𝒀 ()
Matrices X and Y refer to quantities which obscure DUT measurement due to impedance
mismatches along cables, connectors, probes, etc… Matrix T refers to quantities truly indicative
of DUT measurement. Expanding equation (8) to show each matrix element shows that the error
box X is described by elements a, b, c, and d whereas error box Y is described by e, f, g, and h (9).
These values are simply complex pairs (a + bi) which serve as a scaling factor between the obtained
measurement at the ports against the true DUT measurement. Values which truly represent the
DUT are T11, T12, T21, and T22. Graphically, the aforementioned matrices may be visualized
through the color-coded regions in Figure 4.2-9.
[
𝑆 11
𝑆 12
𝑆 21
𝑆 22
] = [
𝑎 𝑏 𝑐 𝑑 ] [
𝑇 11
𝑇 12
𝑇 21
𝑇 22
] [
𝑒 𝑓 𝑔 ℎ
] ()

174

Figure 4.2-9. Illustration of measurement components corresponding to measurement matrices

Additionally, the signal diagram presented earlier in Figure 4.2-7 may be expanded to
include the additional components, as shown in Figure 4.2-10. Note that the same signal flow
treatment described from section 4.2.2.2 may be applied to any single box or linear combination
of boxes in Figure 4.2-10.

Figure 4.2-10. Illustration of all input and output signals during microwave measurement operation with additional
DUT and error box components

175

The goal of the TRL calibration technique is to solve for the 8 unknowns (a, b, c, d, e, f,
and g). By doing so, any measurement matrix, M, may be converted into a true DUT measurement
matrix, T, by employing equation (8). To solve for the unknowns, the TRL technique obtains
measurements of known standards, and applies linear algebra to reverse-calculate a, b, c, d, e, f,
and g. Three types of standards (Figure 4.2-11) are employed – the thru, reflect, and line. The
standards are comprised of only metal atop of fused silica substrates. They do not contain Parylene
coatings or microfluidic structures on top of them.

Figure 4.2-11. Illustration of the thru, reflect, and line standards in TRL measurements

The line standard has an effective length (l), defined by designations of reference planes as
shown through the dotted red lines in Figure 4.2-12. Everything outside the reference plane is
considered to mathematically be incorporated into the error box matrices, X or Y. Everything inside
the reference planes taken to a measurement truly reflective of the line DUT.

176

Figure 4.2-12. Illustration of the line standard with its effective length

As signals traverse a line standard, amplitude and phase change occur as the wave
propagates along the length of the line (Figure 4.2-13). The propagation constant ( 𝛾 )
mathematically describes these attenuation characteristics through a set of complex numbers
in phasor notation. Equation (10) along with Figure 4.2-14 demonstrates that the real part (α)
describes amplitude attenuation and the imaginary part (β) describes phase change.

Figure 4.2-13. Signal amplitude and phase change along the length of a line standard

𝛾 = 𝛼 + 𝑖𝛽 ()
177

Figure 4.2-14. Signal amplitude and phase change along a line visualized by sample propagation constant values

Measurements of line standards may be represented through the block diagram in Figure
4.2-15 and equation (11). Euler’s relations and 𝛾 describe the transmittance of signals through the
line. Note that the top-left and bottom-right indices of the line matrix are zero, meaning that no
reflection is assumed to occur for a line. Every segment of DUT has a characteristic propagation
constant. In the present example, only a bare metal segment is measured. Future examples will
illustrate segments of Parylene-coated metal or fluid segments on top of Parylene, where each
segment has its own propagation constant as well.

Figure 4.2-15. Block diagram of a line measurement

[
𝑆 11
𝑆 12
𝑆 21
𝑆 22
] = [
𝑎 𝑏 𝑐 𝑑 ] [
0 𝑒 −𝛾𝑙
𝑒 −𝛾𝑙
0
] [
𝑒 𝑓 𝑔 ℎ
] ()
178

A thru (Figure 4.2-16) is simply a line where the effective length is equal to zero. Note
that a finite length of metal is required for probe contact, but such regions are taken to contribute
to error box matrices. Substituting zero for length in equation (11) and Figure 4.2-15 yields
equation (12) and Figure 4.2-17.

Figure 4.2-16. Illustration of a thru standard, where the effective length defined by the reference planes is zero

Figure 4.2-17. Block diagram of a thru measurement
[
𝑆 11
𝑆 12
𝑆 21
𝑆 22
] = [
𝑎 𝑏 𝑐 𝑑 ] [
0 1
1 0
] [
𝑒 𝑓 𝑔 ℎ
] ()

The reflect standard features a reflection coefficient, 𝛤 , on the top-left and bottom-right
indices of the reflect matrix. The top-right and bottom-left indices are zero, indication the lack of
179

any transmission. Equation (13) describes the relationship between reflect standard measurements
and error box terms.

[
𝑆 11
𝑆 12
𝑆 21
𝑆 22
] = [
𝑎 𝑏 𝑐 𝑑 ] [
𝛤 0
0 𝛤 ] [
𝑒 𝑓 𝑔 ℎ
] ()

After obtaining measurements of TRL standards, simple yet tedious computations may
yield values for a, b, c, d, e, f, and g and calibration is complete. The error box values are valid for
all measurements obtained in one measurement session. If measurements are made across separate
sessions, each session requires separate TRL measurements to perform calibrations. Calibration
computations are performed through MATLAB in a graphic user interface (GUI) developed by
Nate Orloff of NIST. In practice, multiple lines of varying lengths are used (MTRL) so that
measurement uncertainties may be rigorously quantified through the Microwave Uncertainty
Framework developed by Dylan Williams of NIST [218].
4.2.3.2 Series-Resistor Calibration
MTRL calibration yielded measurements which are truly representative of the DUT, but
intercomparing of those measurements to those of researchers’ systems would be impossible. This
is because MTRL measurements are referenced to the characteristic impedance of the material of
180

the thru, reflect, and line standards. In this work, the TRL CPWs were constructed from 200 nm
thick electron-beam deposited Pt. Other works may employ different geometries, thicknesses,
materials, and more. The RF scientific community therefore collectively reports measurements to
a standardized reference impedance, Z0, of 50 Ω.
The physical significance of this convention is that it is as if all components leading up to
the DUT contribute a real-only 50 Ω. In reality, there will be a real and imaginary contribution
from the various components such as the cables, connectors, probes, and the portion of CPW
contacting the probes but remaining outside the reference plane. However, reporting measurements
as if they were referenced to the 50 Ω Z0 allows for intercomparison across researchers across the
globe and throughout time. A similar analogy may be imagined as defining the kilogram as a
reference value for mass. Although mass may be initially measured in pounds, ounces, stones, or
any arbitrary unit, a conversion factor may transform the measurement into kilogram values for
standardized comparison. In this simplified mass scenario, the conversion factor may be a scalar
number, but an impedance transformer matrix real with and imaginary values accomplishes the
same principle in RF research. The series-resistor calibration technique is well-suited in this work
to further correct the MTRL calibrated measurements to Z0.
An exhaustive treatise on the mathematical proof and operating mechanism of this
calibration technique is outside the scope of this thesis. The reader is guided towards reference
[214] for a more thorough explanation. To summarize, the previously described line standard along
with two new standards – the series resistor (RS) and series capacitor (CS) are used in the Series-
Resistor calibration (Figure 4.2-18). At low frequencies, the impedance of the resistive load may
be approximated by the 50 Ω DC resistance of the strip in Rs. This allows the characteristic
impedance for transmission lines on a low-loss substrate to be simplified into equation (14):
181

𝑍 0
≈
𝛾 (𝜔 )
𝑖𝜔 𝐶 0
()
Where 𝜔 is the frequency in radians and C0 is the capacitance per unit length of transmission line
which is assumed to be constant as a function of frequency.

Figure 4.2-18. Series Resistor and Series Capacitor illustrations
4.2.4 Processed Measurements
The calibration processes described up to this point have referred to measurements of
DUTs such as the thru, reflect, line, series resistor, and series capacitor. These structures are
typically designed to consist of solely the low-loss substrate material and metal layer. This allowed
for key assumptions to be made to facilitate mathematical computations. For example,
conductivity of the air and substrate material may be assumed to be approximately zero [210].
However, measurement of these substrate-metal-only DUTs effectively yields only
calibration data. It is necessary to obtain measurements of DUTs containing thin-film Parylene
deposited on top of the metal. Additionally, regions of 1 × PBS fluid or polydimethylsiloxane
(PDMS) from microfluidic structures may also sit atop of the Parylene film during measurements.
182

This effectively creates distinct segments of CPW with differing physical characteristics.
Mathematical processes exist to handle and isolate the effect of nonhomogeneous materials
throughout CPW lengths as well as to convert such measurements into physically meaningful
elements such as resistance, inductance, capacitance, and conductance.
4.2.4.1 Reference Plane Translation
Consider the simple air-Parylene-air line in Figure 4.2-19. If calibrations have been
applied to account for the error boxes from the cables, probes, connectors, etc., then the T matrix
from equation (9) would correspond to an undesired bend of air and Parylene measurement with
no physical meaningfulness. This is because impedance mismatches are present at the air-Parylene
segment interfaces. Impedance transformer matrices are therefore required to describe the actual
DUT measurements at the segment of interest. Equation (15) describes the impedance transform
from a segment n to segment m.
𝑄 𝑍 𝑚 𝑍 𝑛 =
1
2𝑍 𝑚 |
𝑍 𝑚 𝑍 𝑛 | √
𝑅 (𝑍 𝑛 )
𝑅 (𝑍 𝑚 )
∙ [
𝑍 𝑚 + 𝑍 𝑛 𝑍 𝑚 − 𝑍 𝑛 𝑍 𝑚 − 𝑍 𝑛 𝑍 𝑚 + 𝑍 𝑛 ] ()
where Zn and Zm are the characteristic impedance of segments n and m, and R(Zn) and R(Zm) refer
to solely the real part of Zn and Zm.
183

Figure 4.2-19. Simplified example of CPW with air and Parylene segments
Therefore, the overall T matrix measured from calibrations can be written as equation (16)
and the desired T matrix from the segment of interest, such as Parylene, may be isolated.
𝑇 𝑀𝑒𝑎𝑠𝑢𝑟𝑒𝑑 = 𝑄 𝑍 𝐴𝑖𝑟 𝑍 0
𝑇 𝑙 𝐴𝑖𝑟 ,1
𝑄 𝑍 𝑃𝑎𝑟𝑦𝑙𝑒𝑛𝑒 𝑍 𝐴𝑖𝑟 𝑇 𝑙 𝑃𝑎𝑟𝑦 𝑙𝑒𝑛𝑒 ,1
𝑄 𝑍 𝐴𝑖𝑟 𝑍 𝑃𝑎𝑟𝑦𝑙𝑒𝑛𝑒 𝑇 𝑙 𝐴𝑖𝑟 ,2
𝑄 𝑍 0
𝑍 𝐴𝑖𝑟 ()

where 𝑄 𝑍 𝐴𝑖𝑟 𝑍 0
is the impedance transform obtained from the series resistor calibration to reference
all measurements to the standard 50 Ω reference impedance, Z0.

This process of using impedance transforms to isolate the segment of interest is termed
‘Reference Plane Translation’. This process may also be applied for lines with multiple segments,
as in the case with microfluidic structures with fluid and PDMS.
184

4.2.4.2 RLCG Transmission Line Quantities
Recall that raw S parameters represent the input to output ratios of various port
combinations, but do not lend well to direct interpretation of physical phenomena. In ensuing
MTRL, series-resistor, and reference plane translation techniques, the raw S parameters were
transformed into the propagation constant (𝛾 ) and characteristic impedance (Z) of desired CPW
segments. 𝛾 and Z may be transformed into transmission line quantities that represent physical
properties of the system through equations (17) and (18).
𝑍 =
√(𝑅 +𝑖𝜔𝐿 )
√ (𝐺 +𝑖𝜔𝐶 )
()
𝛾 = √(𝑅 + 𝑖𝜔𝐿 )(𝐺 + 𝑖𝜔𝐶 ) ()
The quantities R and L represent the resistance and inductance from the center conductor
ports, and CG represent the capacitance and conductance between the center conductor to the
ground planes. A differential element of transmission line is illustrated in Figure 4.2-20. Although
CPW geometries have two ground planes, they are effectively represented by the one the bottom
line of Figure 4.2-20 whereas the center conductor line is represented by the top line.

Figure 4.2-20. A transmission line segment of differential length Δx
185

Equations (17) and (18) are the solutions to the Telegrapher’s Equations which describe
the voltage and current on an electrical transmission line with respect to space and time. The
Telegrapher’s Equations are two coupled differential equations as shown in (19) and (20). The
derivation to arrive at (17) and (18) may be reviewed in [209].

𝜕 𝜕𝑥
∙ 𝑉 (𝑥 , 𝑡 ) = −𝐿 𝜕 𝜕𝑡
∙ 𝐼 (𝑥 , 𝑡 ) − 𝑅 ∙ 𝐼 (𝑥 , 𝑡 ) ()

𝜕 𝜕𝑥
∙ 𝐼 (𝑥 , 𝑡 ) = −𝐶 𝜕 𝜕𝑡
∙ 𝑉 (𝑥 , 𝑡 ) − 𝐺 ∙ 𝑉 (𝑥 , 𝑡 ) ()

4.2.5 Fitting
At this point, RLCG values representing the entire set of physical phenomena across a wide
frequency band have been obtained. However, various circuit elements may come together to make
up a holistic model of the entire set of RLCG values. Accurate fitting of these elements allows for
detailed comparison of contributing phenomena in the DUT system for a more nuanced
understanding of thin-film Parylene C behavior.
4.2.5.1 Mapping Functions
For the simple case of solely a thin-film dielectric material deposited on of the CPW,
certain assumptions may be made. It may be assumed that no inductance exists in the system due
to the lack of magnetic components (L=0). No conductance to ground planes (G = 0) may be
assumed and no resistance from the Parylene may be assumed to contribute to R.
186

Therefore, C values can be mapped directly into permittivity (ε) values. Permittivity is the
key material property useful in device design because it is an intensive property. Reporting
capacitance is possible but may be less valuable because it is strongly influenced by coating
thickness. In order to map the C values into ε, mapping functions may apply the conversion. A
classic example of a mapping function is the parallel plate capacitor in equation (21).
𝐶 =
𝜀𝐴
𝑑 =
𝑘 𝜀 0
𝐴 𝑑 ()
where A is the area between parallel plates, d is the distance between plates, εo is the permittivity
of free space, and k is a empirical scaling factor.

Although equation (21) is a valid approximation for parallel plate geometries, use of CPW
geometries renders it invalid for this work. Therefore, high performance simulation software was
used (ANSYS Quasi-3D, Canonsburg, PA) to generate mapping functions which consider relevant
CPW geometries while including Parylene thickness to compute a mathematical function which
may convert C values into ε.
4.2.5.2 Alternative RLCG Models
For more sophisticated systems such as those incorporating microfluidic structures fitted
on top of the Parylene film, more sophisticated circuit models are required. Previous work [192]
rigorously characterized and developed models describing electrical double layers (EDL) effects
with gold (Au) CPW structures and 1 × PBS fluid. The total admittance between the center
conductor and ground plane was described by the admittance of the fluid, Yf, surrounded by two
EDL admittances, YEDL, in series (Figure 4.2-21). Yf and YEDL were further broken down into
additional elements. For instance, Yf was comprised of the capacitance of the fluid far above the
187

relaxation frequency of water (𝐶 ∞
), the admittance of water (Yw), and the conductance of the fluid
(𝐺 𝜎 ). YEDL consisted of the admittance of a constant phase element (YCPE) and a Debye-type
relaxation (YD).

Figure 4.2-21. Circuit model elements of previous work [192] on metal-electrolyte CPW systems

Accurate obtainment and interpretation of these elements can yield key insights about the
measured DUT system. Because certain elements bear resemblance to circuit elements used in EIS
fittings, they may be used to corroborate biomedical device failure modes. For instance, the
constant phase element (CPE) is often used as a semi-empirical and more accurate representation
of the double layer capacitance (Cdl) as described in chapter 2 and 3. Alternatively, the solution
conductance of the fluid, 𝐺 ∞
, is an alternative expression of solution resistance, (RS). Additionally,
new elements such as the water relaxation components may assist in isolating the contribution
between water and ion permeation into Parylene films since EIS cannot yield such high frequency
insights. It is also possible to extract information such as Debye-type relaxation time constant or
solve for fitting parameters which describe water intrusion, salinity, or atomic scale
inhomogeneities to better metal-polymer delamination or other thin-film Parylene characteristics.
188

Methods and Materials
4.3.1 Fabrication
4.3.1.1 Wafer Layout
Wafer layout was strategically designed to ensure fabrication of a sufficient amount of
chips to successfully carry out desired experiments. Two types of chips were fabricated. The first
type was the reference chip which contained multiple copies of calibration standards such as the
thru, reflect, several line lengths, series resistor, and series capacitor (Figure 4.3-1). The second
type was the test chip. Here, the main structures of interest were lines which may be coated with
Parylene or fitted with microfluidics (Figure 4.3-2).

Figure 4.3-1. Illustration of reference chip and its calibration structures
189

Figure 4.3-2. Illustration of a test chip

100 mm diameter wafers were employed because of tool compatibility with USC
cleanroom facilities. This allowed for fitting of 32 chips per wafer as shown in Figure 4.3-3. The
apportionment between the number of test chips, reference chips, and chips with or without
Parylene coatings is listed in Table 4.3-1.

Table 4.3-1. Apportionment of chip types across wafer
190

Figure 4.3-3. Chip apportionment on 100 mm diameter wafer
Initially, four separate test chips were anticipated to be required for one cycle of testing.
This was because probe landing events were shown to compromise CPW contact pads if tested on
NIST’s Au CPWs. Pt CPWs in this work were found to withstand skate marks from probe landings,
allowing one test chip to be used through an entire test cycle.
Type of Chip Number of Chips per Wafer
Test chip (Parylene) 18
Reference chip (Parylene) 6
Test chip (No Parylene) 1
Reference chip (No Parylene) 7

191

4.3.1.2 Microfabrication Process Flow
1.1 cm x 1.1 cm reference chips and test chips were co-fabricated on the same 500 µm
thick and 100 mm diameter fused silica wafer to minimize processing variability. Both chip types
underwent the same microfabrication processes, but the geometric differences were obtained by
selectively patterning photoresist layers to define unique metal liftoff or etch masks.
The final fabricated device cross section is illustrated in Figure 4.3-4. Fabrication began
on the substrate by spin coating and patterning AZ 5214-IR photoresist, a thin 8 nm layer of Pt
with 1 nm of Ti for adhesion was deposited via electron beam evaporation to serve as a resistive
microstrip for series-resistor structures of 50 Ω nominal resistance. Following lift-off in acetone,
isopropanol, and deionized water, another 5214-IR layer was patterned and a thicker layer of Pt
(200 nm) with 5 nm Ti was deposited to define conductive structures of the chip with the same
lift-off technique. Then, a total of 10 µm of Parylene C was deposited via conformal room
temperature chemical vapor deposition (Labcoter, SCS). AZ P4620 photoresist was then patterned
to serve as an etch mask to protect against the following deep reactive ion etch step (Oxford
Plasmalab 100, Oxford Plasma Technology, UK) [168]. For reference chips, all the overlying
Parylene was intentionally etched off. On test chips, etching only occurred at CPW ends in order
to expose the underlying metal as contact pads. Wafers were diced (Disco DAD 3220, Disco
Corporation, Japan) and test chips were annealed at 200 °C for 48 hours in a N2 gas purged, vacuum
oven (CascadeTek TVO-2, Cascade TEK, Cornelius, OR).
192

Figure 4.3-4. Cross section of CPW to illustrate microfabrication process flow

4.3.1.3 Microfluidics Fabrication
PDMS cover layers were fabricated via soft lithography techniques to define microfluidic
channels above CPW structures on test chips. Although original plans entailed SU-8
photolithography for master mold fabrication in the NIST cleanroom, the COVID-19 pandemic
prevented on-campus access. Therefore, Autodesk Inventor was used to design 3D molds (Proto
Labs Inc., Maple Plain, MN) with special care to design channel inner lengths for proper reference
plane translation (Figure 4.3-5 and Figure 4.3-6).

Figure 4.3-5. Autodesk Inventor model of microfluidic master mold
193

Figure 4.3-6. Photograph of 3D printed microfluidic master mold

Syringe needles cored the PDMS to allow fluidic input and output. Double bubble epoxy
sealed fluidic connections to prevent leaks. PDMS covers were carefully aligned and placed atop
CPWs with the aid of a microscope. An acrylic press bar and temperature controlled stage (not
shown) ensured good contact and thermal control of the testing setup.

Figure 4.3-7. Illustration of microfluidic structures laid on top of test chip CPWs

194

Figure 4.3-8. Photograph of microfluidic structures laid on top of test chip CPWs
4.3.2 Measurement
On-chip S parameter measurements were acquired for 512 frequency points from 100 kHz
to 110 GHz on a log frequency scale via vector network analyzer (VNA). AC power was -25 dBm
(where 0 dBm is 1 milliwatt) with 50 Hz intermediate frequency (IF) bandwidth. A temperature-
controlled stage ensured thermal equilibrium at 37 °C. Measurements were calibrated by a two-
tier method similar to [192] [219] and summarized by section 4.2.3.
Although dry Parylene and wetted Parylene projects were carried out in this work, the
measurement protocol remained the same. The differences lied in the sequence and type of samples
undergoing measurement and are explained below.
4.3.2.1 Dry Parylene Measurements
The goal of this experiment was to measure and compare microwave dielectric properties
of unannealed versus annealed Parylene. Limited dielectric characterization of as-deposited
Parylene C exists and investigation as a function of different processing conditions has not been
thoroughly performed. For instance, the influence of Parylene C on passive millimeter-wave
195

circuits and a monolithic-microwave integrated circuit amplifier was studied up to 67 GHz [200],
but only for as-deposited Parylene C with no post-processing. Thermal annealing is one common
post-processing technique in which Parylene is heated above its glass transition temperature but
below its melting point in a vacuum environment. This increases polymer chain mobility to allow
semi-crystalline and amorphous regions to reorganize [169] and improve the film’s crystallinity
(Figure 4.3-9). Annealing has been reported to improve moisture barrier properties which is
beneficial for medical devices [38], but molecular rearrangement may alter dielectric properties
due to Parylene C’s polar base monomer shown in Figure 4.3-10.

Figure 4.3-9. Molecular representation of Parylene before and after annealing

Figure 4.3-10. The polar Parylene C monomer features a chlorine atom covalently bonded to the benzene ring
196

Measurements from the protocols described in section 4.3.2 were first obtained from a
reference chip that was coated with Parylene which was not annealed. The chip was then annealed
at 200 °C for 48 hours in a N2 gas purged, vacuum oven (CascadeTek TVO-2, Cascade TEK,
Cornelius, OR). A second set of measurements was then obtained from the annealed chip.

4.3.2.2 Wetted Parylene Measurements
Approximately 60 minutes are required obtain a full set of CPW measurements by
measuring across all the lines in a test chip. However, the rate of water and ion uptake into thin-
film Parylene is known to occur at comparable time scales [38]. Therefore, only one CPW line
was planned to be measured at a time, at 10 minute intervals, for 60 minutes, in the various
conditions outlined in Table 4.3-2 to better isolate such potentially confounding factors prior to
proceeding with standard measurement protocols.
After these short-term experiments of line measurements, standard full measurement sets
were planned to be obtained with a new Parylene test chip immediately after 1 × PBS was flowed,
after 1 day of soaking, 1 week of soaking, and 1 month of soaking.

197

[220]
Index Short Description Long Description and Rationale
1 Ref chips (Au + Pt)
Since Pt ref chip series-resistor may be compromised, Au ref
chip measurements must also be measured
2 Test Chip A (Air)
The baseline measurement. Do probe tips significantly skate
in the span of 1 hour and cause measurement drift?
3 Test Chip B (Methanol)
New chip. Flow methanol. Theoretically, methanol should
have negligible uptake into Parylene C [220] so that #3 may
serve as a control for #4
4 Test Chip C (DI H2O)
New chip. Flow DI H2O which should penetrate Parylene.
May quantify degree of water penetration via comparison of
data from #3 and #4
5 Test Chip D (1× PBS)
New chip. Flow 1× PBS which should penetrate Parylene.
May quantify and isolate the effect of ion versus water
penetration via comparison of data from #3, #4, and #5
6 Test Chip E (10× PBS)
New chip. Flow 10× PBS which should penetrate Parylene.
May quantify and isolate the effect concentration on the rate
of ion versus water penetration via comparison of data from
#3, #4, #5, and #6
7 Test Chip F (0.1× PBS)
New chip. Flow 0.1× PBS which should penetrate Parylene.
May quantify and isolate the effect concentration on the rate
of ion versus water penetration via comparison of data from
#3, #4, #5, #6, and #7

Table 4.3-2. Measurement plans for test conditions of short-term wetted Parylene experiments
198

4.3.3 Calibrations
4.3.3.1 RadiCal
Once raw S parameters were obtained, they were calibrated by using RadiCal, a MATLAB
graphic user interface (GUI) developed by Nate Orloff of NIST. RadiCal employs the theory
outlined in section 4.2.3 and is frequently used by members of the RF-Frequency Electronics
Group. A screenshot of the GUI front panel is provided in Figure 4.3-11.

Figure 4.3-11. Screenshot of the RadiCal MATLAB GUI

199

The .csv files containing raw S parameters are the input to the RadiCal tool. A user must
specify the directories in which S parameters for the reference chip and test chip are stored. This
information is required to perform MTRL and the series-resistor calibration. The output of RadiCal
is a .mat file (a MATLAB data struct) containing many key quantities, such as error boxes, CPW
structure names, the propagation constant, the characteristic impedance, impedance transformers,
MUF values, and more (Figure 4.3-12). Data from these .mat files may then undergo further
processing and fitting through user-specific custom MATLAB scripts.

Figure 4.3-12. A sample MATLAB struct output from RadiCal

4.3.4 Series-resistor Simulation
As mentioned in 4.2.3.2, the Series-Resistor calibration is crucial in ensuring that
measurements are referenced to a 50 Ω reference impedance, Z0. This would have been relatively
200

straightforward to accomplish with standard RadiCal processing methods and mapping functions
generated from ANSYS Q3D simulations.
However, this work required development of a new process dedicated to overcoming
unforeseen calibration challenges stemming from damaged Series-Resistor structures on the No-
Parylene (NP) reference chips. Although they were designed to have a 50 Ω DC resistance,
measurements revealed open circuit values. Interestingly, the Yes-Parylene (YP) reference chips
did indeed demonstrate approximately 50 Ω measurements. This suggested that a fabrication issue
occurred post-metal-deposition. Despite high metal-to-polymer selectivity in the deep reactive ion
etching process, over etching of the Parylene may have additionally removed the thin (8 nm) Pt
strip of the series resistor. Regardless of the specific cause of this hardware-based issue, software-
based algorithms were devised to circumvent this issue in order to proceed with this project.
To summarize, a process was developed to generate computational simulations of the NP
reference chips. Parameter sweeps in the simulation could be compared against salvageable data
of the NP reference chip until good agreement allowed for convergence to a credible model. More
specifically, the known CPW geometry was incorporated into ANSYS Q3D simulations to
generate theoretical resistance and inductance (RL) values for several swept values of thin-film Pt
resistivity. Additionally, RL values were computed from dividing the propagation constant of the
physically measured NP chip by an initial guess of C0 which had been informed by the Q3D
simulation. These two sets of RL values were compared and the sweep with greatest agreement
was selected. Because non-magnetic systems with low-loss substrates may be assumed to share
the identical RL values, the simulated and optimized RL values were compared against the
measured RL values of the YP chip to confirm the validity of this technique.
201

4.3.5 Microwave Data Analysis
Once raw S parameter measurement sets of unannealed Parylene and annealed Parylene
were obtained, they underwent analyses using custom-developed MATLAB scripts. RLCG values
were assessed with previously described methods in addition to techniques involving mathematical
fitting processes which were required to account for compromised data coming from one of the
microfabricated line standards. The permittivity before and after annealing was accomplished
through use of mapping functions and additional custom-developed MATLAB scripts.
Results
4.4.1 Fabrication
The microfabrication process flow described in section 4.3.1.2 was successfully carried out
on two wafers. Wafer dicing into chips was found to significantly compromise the Parylene coating.
However, the presence of redundant copies of test structures and multiple identical dies throughout
the wafer permitted a sufficient number of usable chips. Additionally, reference chips intentionally
containing no Parylene were spared from dicing damage (Figure 4.4-1). It is suggested to anneal
wafers prior to dicing in order to leverage improved adhesion from thermal treatment.
202

Figure 4.4-1. High-resolution stitched micrograph of a No-Parylene reference chip
203

4.4.2 ANSYS Q3D Simulations
The ANSYS Quasi-3D simulation package was utilized to create cross-sectional
computational models of CPW devices. These simulations can therefore represent per unit length
(PUL) quantities such as resistance, inductance, capacitance, and conductance. Figure 4.4-2
illustrates the geometry of the simulated CPW coated in Parylene. Note that the model dimensions
correspond to their physical counterparts. The metal backer serves as a key electrical grounding
feature and mesh refinement windows assist in processing more granular computations at regions
of interest.

Figure 4.4-2. The geometric model of CPW structures created in ANSYS Q3D
204

Physical dimensions may be numerically inputted, or they may also be programmatically
defined through other variables such as ‘substrate_width’ (Figure 4.4-3). Materials were defined
and ascribed to corresponding geometric model elements (Figure 4.4-4). For instance, the relative
permittivity of Parylene C was defined as a global variable throughout the model and swept with
respect to the thickness of Parylene to yield a mapping function between capacitance and
conductance per unit length. Since the conductance may be assumed to be zero in low-loss systems,
the capacitance may be assumed to solely map with permittivity.

Figure 4.4-3. Physical dimensions inputted into Q3D models

Figure 4.4-4. Material assignment in Q3D models
205

Simulation results revealed that the electric field is predominantly confined to the Parylene
film, as shown in (Figure 4.4-5). This confirmed that measured S parameters may be assumed to
represent physical phenomena related to the Parylene. Mapping functions between permittivity
and capacitance were then successfully computed for various Parylene thickness (Figure 4.4-6).

Figure 4.4-5. Q3D simulations revealed that the electric field generated during S parameter measurement is
confined predominantly to the Parylene C film

Figure 4.4-6. Permittivity to capacitance per unit length mapping function for varying Parylene film thicknesses
206

4.4.3 Unannealed and Annealed Parylene Dielectric Properties
4.4.3.1 Series-resistor Simulation Results
Due to unforeseen microfabrication challenges described in section 4.3.4, a dedicated
process was developed in order to simulate and optimize the resistance and inductance per unit
length of NP reference chips. Values were generated and plotted with RL values generated from
dividing the propagation constant of the unannealed YP reference chip with its capacitance per
unit length, C0. Due to the low-loss assumption with no magnetic components, the optimized
simulated NP RL values demonstrated excellent agreement with the empirical YP RL values as
shown in Figure 4.4-7. Because the theoretically derived values demonstrated no high-frequency
noise, they were critical in serving as initial guesses in subsequent cellfitRLCG fitting algorithms.

Figure 4.4-7. Resistance per unit length of simulated No Parylene and measured Yes Parylene (unannealed) chips

207

4.4.3.2 Resistance and Resistivity of Unannealed and Annealed Parylene
The resistance per unit length of the same CPW chip before and after annealing was plotted
(Figure 4.4-8). The higher frequency portions may be disregarded due to the skin effect.
Noteworthy data occur in the lower frequency plateaus corresponding to DC resistance. A 22%
decrease from 15.86 MΩ/m to 12.38 MΩ/m was measured before and after annealing. The known
geometry of the Pt layer allowed for computation of 1.586 ×10-7 Ω∙m and 1.238 × 10-7 Ω∙m
resistivities. Bulk Pt resistivity is known to be 1.06 × 10-7 Ω∙m and it is well-known that
polycrystalline metal thin film resistivities are greater than that of their bulk material counterparts
[20]. The decrease in resistivity from thermal annealing may be attributed to Pt approaching bulk
property behavior due to relaxation of metal thin film stress and grain boundary reorganization
[21].

Figure 4.4-8. Resistance per unit length (MΩ/m) and resistivity of the CPW before and after annealing

208

Although it is known that high temperature annealing (> 400 °C) of metal films can induce
similar effects, the 200 °C process in this work yielded an unanticipated result which may inform
future device design. For instance, many Parylene-Pt-Parylene devices with long aspect ratios have
reported unwanted device curvature after annealing the entire sandwich assembly. Annealing prior
to the second Parylene deposition may relax the thin film stress such that less curvature is
experienced. Additionally, lower film-to-film stress may potentially reduce rates of device
delamination.
4.4.3.3 Capacitance and Conductance of Unannealed and Annealed Parylene
Capacitance for the same CPW chip before and after annealing is was plotted (Figure
4.4-9). To within our uncertainty estimates, C was approximately constant as a function of
frequency (130 pF/m) with no significant change before and after annealing. The right y-axis
showed corresponding mapped permittivity values. Previously reported values of approximately
3.0 agreed well with results in this work up to 110 GHz [200].
This result illustrated that molecular polymer chain rearrangement did not manifest
permittivity changes, allowing future Parylene RF designers to assume that annealing does not
impart frequency-dependent electrical loss. This is significant since other commonly used
polymers such as SU-8 are known to exhibit high-frequency loss [221]. Additionally, whereas
Parylene C permittivity had previously been reported by vendors at 60 Hz, 1 kHz, and 1 MHz, this
work expanded the 3.0 permittivity value to 110 GHz to dispel concerns of potentially
unanticipated resonance or relaxation phenomena in higher frequencies

209

Figure 4.4-9. Capacitance per unit length (pF/m) and permittivity of Parylene before and after thermal annealing
Discussion and Future Directions
An entirely distinct branch of research techniques from NIST was incorporated with
existing expertise in Parylene C micromachining in order to investigate its thin-film properties at
microwave frequencies. This work successfully summarized key elements of RF theory into a
format amenable for USC Biomedical Microsystems Laboratory readership. A complementary
video lecture series has also been developed to assist in transfer of knowledge. A new fabrication
process flow was designed and successfully carried out to construct reference and test chips for
measurement with NIST facilities in Boulder, CO. After relocating from Los Angeles to Boulder,
preliminary S parameter measurements were obtained via vector network analyzer equipment.
However, the COVID-19 pandemic unfortunately restricted access to the NIST campus for the
remaining duration of the NSF INTERN program. Remote collaboration still allowed for data
acquisition from NIST staff and transfer to USC for analysis. ANSYS Q3D simulations and
various MATLAB-based scripts were developed to assist in data analysis and successful results
were obtained in differentiating dielectric properties of Parylene C before and after annealing.
210

Measurement plans and microfluidic materials for RF measurement of wetted Parylene
samples were rigorously prepared. One test chip was planned for use with each experimental
condition because CPW contact pads were assumed to be damaged from probe landing events.
Preliminary measurements revealed that the USC Pt-based CPW structures did not experience
damage from scratched skate marks from probe landing events, unlike that of Au-based CPWs at
NIST. This allowed for one test chip to be planned for longitudinal testing, allowing for additional
test conditions to investigate the effects of fluid penetration into Parylene at short (minutes) time
scales.
Future work may leverage these pre-existing plans and samples to obtain measurements
which had been delayed from the COVID-19 pandemic. Once raw S parameters are gathered from
Parylene samples which have undergone soaking with various fluids, soak times, and ionic
concentrations, they may undergo the same numerical processes of conversion into RLCG values
and fittings to generate relevant circuit model elements describing the physical phenomena at
wetted Parylene films. The circuit elements may then build on top of existing models generated by
Cully et. al as well as this work to augment the field’s understanding of electrochemical liquid
interactions with polymer thin-films. Insights acquired through these modeling endeavors may
then assist in future device design by informing engineers and scientists on water and ion
penetration rates, conductive versus capacitive contributions of signal leakage in wetted thin-film
polymer devices, and metal-polymer delamination characteristics. Additionally, moisture barrier
films may also be tested and compared against such previously described models to definitively
demonstrate improved hermetic sealing. Ultimately, such data and techniques with this vein of
research may therefore aid in basic fundamental research and improve clinical outcomes by
211

pushing the envelope towards enhanced development of implantable thin-film polymer MEMS
biomedical devices.
212

References
[178] D. D. Zhou, Implantable Neural Prostheses 2. 2010.
[179] W. M. Grill, S. E. Norman, and R. V. Bellamkonda, “Implanted Neural Interfaces:
Biochallenges and Engineered Solutions,” Annu. Rev. Biomed. Eng., vol. 11, no. 1, pp. 1–
24, 2009.
[180] A. M. Cobo et al., “Parylene-Based Cuff Electrode With Integrated Microfluidics for
Peripheral Nerve Recording, Stimulation, and Drug Delivery,” J. Microelectromechanical
Syst., vol. 28, no. 1, pp. 36–49, 2019.
[181] Y. Qin, M. M. R. Howlader, M. J. Deen, Y. M. Haddara, and P. R. Selvaganapathy,
“Polymer integration for packaging of implantable sensors,” Sensors Actuators, B Chem.,
vol. 202, pp. 758–778, 2014.
[182] S. Venkatraman, F. Boey, and L. L. Lao, “Implanted cardiovascular polymers: Natural,
synthetic and bio-inspired,” Prog. Polym. Sci., vol. 33, no. 9, pp. 853–874, 2008.
[183] E. M. Davis, N. M. Benetatos, W. F. Regnault, K. I. Winey, and Y. A. Elabd, “The influence
of thermal history on structure and water transport in Parylene C coatings,” Polymer
(Guildf)., vol. 52, no. 23, pp. 5378–5386, 2011.
[184] X. Z. Xie, L. Rieth, P. Tathireddy, and F. Solzbacher, “Long-term in-vivo Investigation of
Parylene-C as Encapsulation Material for Neural Interfaces,” Procedia Eng., vol. 25, pp.
483–486, 2011.
[185] D. W. L. Hukins, A. Mahomed, and S. N. Kukureka, “Accelerated aging for testing
polymeric biomaterials and medical devices,” Med. Eng. Phys., vol. 30, no. 10, pp. 1270–
1274, 2008.

213

[186] R. Caldwell, M. G. Street, R. Sharma, P. Takmakov, B. Baker, and L. Rieth,
“Characterization of Parylene-C degradation mechanisms: In vitro reactive accelerated
aging model compared to multiyear in vivo implantation,” Biomaterials, vol. 232, no.
January 2019, p. 119731, 2020.
[187] SPIVACK MA and FERRANTE G, “Determination of the Water Vapor Permeability and
Continuity of Ultrathin Parylene Membranes,” pp. 607–624, 1968.
[188] P. Takmakov, K. Ruda, K. Scott Phillips, I. S. Isayeva, V. Krauthamer, and C. G. Welle,
“Rapid evaluation of the durability of cortical neural implants using accelerated aging with
reactive oxygen species,” J. Neural Eng., vol. 12, no. 2, 2015.
[189] J. P. Seymour, Y. Elkasabi, H. Chen, J. Lahann, and R. Daryl, “The insulation performance
of reactive parylene films in implantable electronic devices,” vol. 30, no. 31, pp. 6158–6167,
2010.
[190] W. Chun, N. Chou, S. Cho, S. Yang, and S. Kim, “Evaluation of sub-micrometer parylene
C films as an insulation layer using electrochemical impedance spectroscopy,” Prog. Org.
Coatings, vol. 77, no. 2, pp. 537–547, 2014.
[191] A. Prasad et al., “Comprehensive characterization and failure modes of tungsten microwire
arrays in chronic neural implants,” J. Neural Eng., vol. 9, no. 5, 2012.
[192] C. A. E. Little, N. D. Orloff, I. E. Hanemann, C. J. Long, V. M. Bright, and J. C. Booth,
“Modeling electrical double-layer effects for microfluidic impedance spectroscopy from
100 kHz to 110 GHz,” Lab Chip, vol. 17, no. 15, pp. 2674–2681, 2017.
[193] A. Agarwal, A. Shapero, D. Rodger, M. Humayun, Y. C. Tai, and A. Emami, “A wireless,
low-drift, implantable intraocular pressure sensor with parylene-on-oil encapsulation,”
2018 IEEE Cust. Integr. Circuits Conf. CICC 2018, pp. 1–4, 2018.
214

[194] C. C. Yeh, S. H. Lo, M. X. Xu, and Y. J. Yang, “Fabrication of a flexible wireless pressure
sensor for intravascular blood pressure monitoring,” Microelectron. Eng., vol. 213, no.
November 2018, pp. 55–61, 2019.
[195] W. Li, D. C. Rodger, and Y. C. Tai, “Implantable RF-coiled chip packaging,” Proc. IEEE
Int. Conf. Micro Electro Mech. Syst., pp. 108–111, 2008.
[196] J. Maeng, B. Kim, D. Ha, and W. J. Chappell, “Parylene interposer as thin flexible 3-D
packaging enabler for wireless applications,” IEEE Trans. Microw. Theory Tech., vol. 59,
no. 12 PART 2, pp. 3410–3418, 2011.
[197] A. Shapero, A. Agarwal, J. C. Martinez, A. Emami, M. S. Humayun, and Y. Tai, “Wireless
Implantable Intraocular Pressure Sensor with Parylene-Oil-Encapsulation and Forward-
Angled RF Coil,” 2019 IEEE 32nd Int. Conf. Micro Electro Mech. Syst., no. January, pp.
21–24, 2019.
[198] A. Baldwin, L. Yu, M. Pratt, K. Scholten, and E. Meng, “Passive, wireless transduction of
electrochemical impedance across thin-film microfabricated coils using reflected
impedance,” Biomed. Microdevices, vol. 19, no. 4, 2017.
[199] L. Yu, E. Pun, and E. Meng, “A CONTACTLESS ELECTROCHEMICAL IMPEDANCE
MEASUREMENT METHOD,” pp. 6–9.
[200] C. Kärnfelt, C. Tegnander, J. Rudnicki, J. P. Starski, and A. Emrich, “Investigation of
parylene-C on the performance of millimeter-wave circuits,” IEEE Trans. Microw. Theory
Tech., vol. 54, no. 8, pp. 3417–3424, 2006.
[201] C. H. Yang, L. S. Kuo, P. H. Chen, C. R. Yang, and Z. M. Tsai, “Development of a
multilayered polymeric DNA biosensor using radio frequency technology with gold and
magnetic nanoparticles,” Biosens. Bioelectron., vol. 31, no. 1, pp. 349–356, 2012.
215

[202] W. Li et al., “Parylene-based integrated wireless single-channel neurostimulator,” Sensors
Actuators, A Phys., vol. 166, no. 2, pp. 193–200, 2011.
[203] A. M. Aragón, M. Hernando-Rydings, A. Hernando, and P. Marín, “Liquid pressure
wireless sensor based on magnetostrictive microwires for applications in cardiovascular
localized diagnostic,” AIP Adv., vol. 5, no. 8, 2015.
[204] B. Crum and W. Li, “Parylene-based fold-and-bond wireless pressure sensor,” 8th Annu.
IEEE Int. Conf. Nano/Micro Eng. Mol. Syst. IEEE NEMS 2013, vol. 1, pp. 1155–1158, 2013.
[205] J. H.-C. Chang, “Wireless parylene-based retinal implant,” ProQuest Diss. Theses, vol.
2014, p. 195, 2014.
[206] J. Chen, C. Huang, H. J. H. Chen, C. F. Jout, and S. R. Huang, “Folded Dual-band
(2.4/5.2GHz) Antenna Fabricated on Silicon Suspended Parylene Membrane,” vol. 1, pp.
5–8.
[207] Y. Zheng, X. Sun, Z. Li, X. Li, and H. Zhang, “Flexible MEMS inductors based on
Parylene-FeNi Compound Substrate for wireless power transmission system,” 8th Annu.
IEEE Int. Conf. Nano/Micro Eng. Mol. Syst. IEEE NEMS 2013, vol. 1, pp. 1002–1005, 2013.
[208] Y. H. Jung et al., “A Compact Parylene-Coated WLAN Flexible Antenna for Implantable
Electronics,” IEEE Antennas Wirel. Propag. Lett., vol. 15, no. 560 m, pp. 1382–1385, 2016.
[209] D. Pozar, Microwave Engineering, Fourth Edi. John Wiley & Sons, Inc., 2005.
[210] C. A. E. Little, “Microfluidic-based Broadband Measurements of Fluid Permittivity and
Permeability to 100 GHz,” University of Colorado, 2017.
[211] R. B. Marks, “A Multiline Method of Network Analyzer Calibration,” IEEE Trans. Microw.
Theory Tech., vol. 39, no. 7, pp. 1205–1215, 1991.

216

[212] D. F. Williams, C. M. Wang, and U. Arz, “An optimal multiline TRL calibration algorithm,”
IEEE MTT-S Int. Microw. Symp. Dig., vol. 3, no. July 2003, pp. 1819–1822, 2003.
[213] D. C. DeGroot, J. A. Jargon, and R. B. Marks, “Multiline TRL revealed,” 60th ARFTG Conf.
Dig. Meas. Needs Emerg. Technol. ARFTG 2002, pp. 131–155, 2002.
[214] N. D. Orloff et al., “A compact variable-temperature broadband series-resistor calibration,”
IEEE Trans. Microw. Theory Tech., vol. 59, no. 1, pp. 188–195, 2011.
[215] D. F. Williams, S. Member, and D. K. Walker, “Series-resistor calibration.”
[216] S. Padmanabhan, L. Dunleavy, J. E. Daniel, A. Rodríguez, and P. L. Kirby, “Broadband
space conservative on-wafer network analyzer calibrations with more complex load and
thru models,” IEEE Trans. Microw. Theory Tech., vol. 54, no. 9, pp. 3583–3592, 2006.
[217] D. F. Williams and R. B. Marks, “LRM Probe-Tip Calibrations using Nonideal Standards,”
IEEE Trans. Microw. Theory Tech., vol. 43, no. 2, pp. 466–469, 1995.
[218] D. Williams, “NIST Uncertainty Framework Software Package.” National Institute of
Standards and Technology, Boulder, CO, 2017.
[219] A. C. Stelson et al., “Label-free detection of conformational changes in switchable DNA
nanostructures with microwave microfluidics,” Nat. Commun., vol. 10, no. 1, 2019.
[220] H. C. Koydemir, H. Külah, and C. Özgen, “Solvent compatibility of parylene c film Layer,”
J. Microelectromechanical Syst., vol. 23, no. 2, pp. 298–307, 2014.
[221] J. C. Booth, N. D. Orloff, J. Mateu, M. Janezic, M. Rinehart, and J. A. Beall, “Quantitative
permittivity measurements of nanoliter Liquid volumes in microfluidic channels to 40 GHz,”
IEEE Trans. Instrum. Meas., vol. 59, no. 12, pp. 3279–3288, 2010.

217

BioMEMS devices continues to grow and develop in today’s scientific communities and
hold tremendous potential in bench-to-bedside clinical translation. Polymer-based bioMEMS, with
Parylene C and electrochemical interfaces in particular, are uniquely positioned to help patients
through advancements in micromachining techniques. The list of devices owing to this lineage is
ever-expanding, and the three significant bodies of work in this thesis have served as noteworthy
contributions towards the field.
A Parylene C micromachined microelectrode array was designed, fabricated, and tested for
first-ever use in rat epiretinal stimulation studies in chapter 2. Rats are an excellent model to study
retinal degeneration and the only FDA approved retinal prosthesis employs the epiretinal approach.
However, no microelectrode array devices suitable for epiretinal rat studies have been reported.
This was largely due to surgical access challenges stemming from the small ocular orbit (~5 mm)
and extremely delicate nature of retinal tissue. Hence, biocompatible and mechanically soft
Parylene-based MEMS were well positioned to tackle these unique issues. By developing this
device, retinal degeneration research may be accelerated to assist in improving the visual acuity
provided by retinal prostheses to help certain blind patient populations see again.
Similarly, in chapter 3, a Parylene C device developed by previous laboratory members
experienced redesign and rigorous fundamental experimentation to yield orders or magnitude
improvement across various performance specifications. The microbubble pressure sensor (µBPT)
CHAPTER 5
CONCLUSION
218

had demonstrated proof of concept, but efforts in this work elucidated bubble generation, behavior,
and measurement. For instance, proper electrochemical grounding allowed for anodic oxygen
bubble generation to solve quasi-stability issues with 25× improvement in bubble lifetime (500
min versus 20 min). The newly developed residence time technique in this work demonstrated 40×
improvement in measurement resolution (0.5 mmHg versus 20 mmHg) when compared against
similarly calculated dissolution rate techniques in previous work. Another previous study
employed non-mathematical techniques to claim approximately 5 mmHg in a real-time pressure-
to-impedance measurement mode. This work demonstrated 1 mmHg resolution in the same real-
time measurement mode. Additionally, impedance measurement confounding factors such as pH,
dissolved gas concentration, and bubble wall detachment events were identified and thoroughly
documented. Much potential in µBPT technology to improve the lives of hydrocephalus patients,
and this work succeeded in pushing the Parylene C device closer towards that goal.
Instead of a specific device per se, thin film Parylene C itself was the focus as chapter 4
presented first-ever collaboration between the Biomedical Microsystems Laboratory and the
National Institute of Standards and Technology (NIST). Radio frequency measurement techniques
from the government laboratory under the charter of the US Chamber of Commerce allowed for
first-ever characterization thin film Parylene C in the context of biomedical applications. Although
COVID-19 complications prevented wetted Parylene C measurements, rigorous preparations and
measurement plans have been put in place. Dry Parylene C measurements before and after thermal
annealing were able to undergo the S parameter to circuit model fitting process to demonstrate key
material property characteristics.
It is the author’s hope that the work detailed herein may serve as a useful steppingstone for
future researchers in the development of Parylene C bioMEMS to ultimately improve patient care.
219

The development of the RRA and µBPT devices in addition to the usage of NIST technique on
Parylene C represent significant contributions towards the very promising field. Although much
work remains, collective efforts in biomedical engineering have historically yielded many
consistent strides in technology to improve the health of humankind. Another leg in the relay race
of science has been completed. May the next baton-holder capitalize on previous gains to bring us
all closer to the finish line.

220

APPENDICES
Appendix A: RRA Version 3 Microfabrication Process Flow

1. Dehydration bake silicon wafers to remove moisture 110 °C, > 10 min
2. Deposit Parylene (~10 µm)
3. Descum, O2 plasma 100 W, 100 mTorr, 1 min
4. Pattern AZ 5214-IR for metal lift-off (~2 µm)
Pre spin 5 sec, 500 rpm
Spin 45 sec, 4000 rpm
Softbake 110 °C, 1 min
Exposure 42 mJ/cm
2

Image Reversal Bake 110 °C, 63 sec
Flood Exposure 280 mJ/cm
2

Development 18 sec
5. Descum, O2 plasma 100 W, 100 mTorr, 1 min
6. Platinum deposition 2000 Å (4 runs of 500 Å)
7. Lift-off in room temperature acetone, IPA, water
8. Descum, O2 plasma 100 W, 100 mTorr, 1 min
9. Deposit Parylene (~10 µm)
10. Pattern 1
st
AZ 4620 etch mask (~16 µm)
Pre spin 5 sec, 500 rpm
221

Spin 45 sec, 1100 rpm
Spin off 5 sec, 4500 rpm
Softbake 90 °C, 5 min
Rehydration > 3 min
Exposure 480 mJ/cm
2

Development ~60 sec
11. Deep Reactive Ion Etching “Meng Recipe” – 5 sets of 25 loops
12. Strip remaining etch mask with acetone, IPA, and DI water
13. Pattern 2
nd
AZ 4620 etch mask (~16 µm)
Pre spin 5 sec, 500 rpm
Spin 45 sec, 1100 rpm
Spin off 5 sec, 4500 rpm
Softbake 90 °C, 5 min
Rehydration > 3 min
Exposure 480 mJ/cm
2

Development ~60 sec
14. Deep Reactive Ion Etching “Meng Recipe” – 5 sets of 25 loops
15. Carefully peel devices off wafer by lifting corners with tweezers and allowing DI water
droplets to wick into underlying spaces to assist device release
16. Strip remaining etch mask with acetone, IPA, and DI water

222

Appendix B: RRA Version 3 Photomasks
The corresponding AutoCAD .dwg files may be found on the laboratory server at:
/Users/Eugene Yoon/RRA

223

224

225

Appendix C: µBPT Microfabrication Process Flow

1. Dehydration bake silicon wafers to remove moisture 110 °C, > 10 min
2. Deposit Parylene (~10 µm)
3. Descum, O2 plasma 100 W, 100 mTorr, 1 min
4. Pattern AZ 5214-IR for metal lift-off (~2 µm)
Pre spin 5 sec, 500 rpm
Spin 45 sec, 4000 rpm
Softbake 110 °C, 1 min
Exposure 42 mJ/cm
2

Image Reversal Bake 110 °C, 63 sec
Flood Exposure 280 mJ/cm
2

Development 18 sec
5. Descum, O2 plasma 100 W, 100 mTorr, 1 min
6. Platinum deposition 2000 Å (4 runs of 500 Å)
7. Lift-off in room temperature acetone, IPA, water
8. Descum, O2 plasma 100 W, 100 mTorr, 1 min
9. Sacrificial photoresist - AZ 4620 mask (~16 µm)
Pre spin 5 sec, 500 rpm
Spin 45 sec, 1100 rpm
Spin off 5 sec, 4500 rpm
Softbake 90 °C, 5 min
226

Rehydration > 3 min
Exposure 480 mJ/cm
2

Development ~60 sec
Do not hard bake because it will induce unwanted photoresist reflow

10. Descum, O2 plasma 100 W, 100 mTorr, 1 min
11. Deposit Parylene (~10 µm)
12. Pattern AZ 4620 etch mask (~16 µm)
Pre spin 5 sec, 500 rpm
Spin 45 sec, 1100 rpm
Spin off 5 sec, 4500 rpm
Softbake 90 °C, 5 min
Rehydration > 3 min
Exposure 480 mJ/cm
2

Development ~60 sec
13. Deep Reactive Ion Etching “Meng Recipe” – 5 sets of 25 loops
14. Slice with razor blade and carefully peel devices off wafer by lifting corners with tweezers
and allowing DI water droplets to wick into underlying spaces to assist device release
15. Strip remaining etch mask with acetone, IPA, and DI water (2 hours each)

227

Appendix D: µBPT Photomasks
The corresponding AutoCAD .dwg files may be found on the laboratory server at:
/Users/Eugene Yoon/uBPT

228

229

230

231

Appendix E: NIST Device Fabrication Process Flow

1. Dehydration bake fused silica wafers 110 °C, > 10 min
2. HMDS photoresist adhesion
Use dropper to transport 3 mL of HMDS into a small beaker. Place said beaker and
wafer into vacuum chamber and pump down for 10 minutes
3. Pattern AZ 5214-IR for metal lift-off (~2 µm)
Pre spin 5 sec, 500 rpm
Spin 45 sec, 4000 rpm
Softbake 110 °C, 1 min
Exposure 42 mJ/cm
2

Image Reversal Bake 110 °C, 63 sec
Flood Exposure 280 mJ/cm
2

Development 18 sec
4. Descum, O2 plasma 100 W, 100 mTorr, 1 min
5. Platinum deposition 2000 Å (4 runs of 500 Å)
6. Lift-off in room temperature acetone, IPA, water
7. Descum, O2 plasma 100 W, 100 mTorr, 1 min
8. Deposit Parylene (~10 µm)
9. Pattern AZ 4620 etch mask (~16 µm)
Pre spin 5 sec, 500 rpm
Spin 45 sec, 1100 rpm
232

Spin off 5 sec, 4500 rpm
Softbake 90 °C, 5 min
Rehydration > 3 min
Exposure 480 mJ/cm
2

Development ~60 sec
10. Deep Reactive Ion Etching “Meng Recipe” – 5 sets of 25 loops

233

Appendix F: NIST Microwave Data Processing Video Lectures

Video lectures and associated PowerPoint files may be found on the laboratory server at:
/Data/Projects/NIST

Abstract (if available)

Abstract Whereas silicon microelectromechanical systems (MEMS) have already revolutionized the world through various inventions, immense potential through polymer-based MEMS remains to be fully unlocked. In particular, Parylene C is a strong candidate for biomedical innovation due to its biocompatibility and amenability to micromachining techniques. Parylene C, herein referred to as Parylene, may insulate thin-film metals which act as current-carrying interfaces for electrochemical modalities in in vivo environments. This allows for the design and fabrication of unique implantable microdevices which may serve to improve patient care or assist in basic research. ? This work presents three projects harnessing the utility of Parylene-based MEMS with electrochemical interfaces. Chapter 1 provides an overview of fundamental considerations for Parylene and electrochemistry to prime the reader for the ensuing chapters. Chapter 2 describes the development of a Parylene-based novel implantable microelectrode array designed for chronic stimulation of the retina in live rats. In Chapter 3, microbubble-based pressure transducers constructed from Parylene microchannels are thoroughly investigated to yield improvements in key performance specifications. Radio frequency dielectric spectroscopy techniques from the National Institute of Standards and Technology (NIST) are used for the first time in the biomedical context of characterizing Parylene and thin-film platinum in chapter 4. Chapter 5 provides a conclusion and forward-looking remarks. ? Although many challenges still remain, the potential for healthcare improvement continues to drive future progress in this field. It is therefore the hope of the author that the research contained herein may inform and aid the advancement of novel MEMS biomedical devices.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Development of implantable Parylene-based MEMS technologies for cortical applications

PDF

Parylene-based biomems sensors for multiple physiological systems

PDF

The electrochemical evaluation of Parylene-based electrodes for neural applications

PDF

Implantable bioMEMS drug delivery systems

PDF

Parylene-based implantable interfaces for biomedical applications

PDF

Strategies for improving mechanical and biochemical interfaces between medical implants and tissue

PDF

Development of micromachined technologies for neural interfaces

PDF

Liquid flow rate sensing via solution resistance for in vivo applications

PDF

Development of fabrication technologies for robust Parylene medical implants

PDF

Fabrication and packaging of three-dimensional Parylene C neural interfaces

PDF

Modular bio microelectromechanical systems (bioMEMS): intraocular drug delivery device and microfluidic interconnects

PDF

A wireless implantable MEMS micropump system for site-specific anti-cancer drug delivery

PDF

High-resolution data acquisition with neural and dermal interfaces

PDF

Penetrating parylene neural probe array for dense, in vivo, chronic recordings

PDF

Integrated wireless piezoelectric ultrasonic transducer system for biomedical applications

PDF

Thin-film impedimetric sensors for chronic in vivo use: design and application to hydrocephalus treatment

PDF

Wireless electrochemical drug delivery micropump with fully integrated electrochemical dose tracking feedback system

PDF

Next generation neural interfaces for neural modulation and prosthesis

PDF

Multi-region recordings from the hippocampus with conformal multi-electrode arrays

PDF

Circuits for Biomedical Telemetry and Stimulation Applications

Asset Metadata

Creator Yoon, Eugene Jisu (author)

Core Title Parylene C bioMEMS for implantable devices with electrochemical interfaces

School Viterbi School of Engineering

Degree Doctor of Philosophy

Degree Program Biomedical Engineering

Degree Conferral Date 2021-08

Publication Date 07/24/2021

Defense Date 07/21/2021

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag bioMEMS,electrochemistry,implantable device,MEMS,neural interface,OAI-PMH Harvest,Parylene C

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Meng, Ellis (committee chair), Monge, Manuel (committee member), Mousavi, Maral (committee member)

Creator Email eugenejy@usc.edu,eugeneyoon562@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-oUC15621181

Unique identifier UC15621181

Legacy Identifier etd-YoonEugene-9861

Document Type Dissertation

Format application/pdf (imt)

Rights Yoon, Eugene Jisu

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 author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright. The original signature page accompanying the original submission of the work to the USC Libraries is retained by the USC Libraries and a copy of it may be obtained by authorized requesters contacting the repository e-mail address given.

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

Repository Email cisadmin@lib.usc.edu