University of Southern California Dissertations and Theses

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

(USC Thesis Other)

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

PDF

Download

Open document

Flip pages

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

WIRELESS ELECTROCHEMICAL DRUG DELIVERY
MICROPUMP WITH FULLY INTEGRATED
ELECTROCHEMICAL DOSE TRACKING FEEDBACK SYSTEM

by
Roya Sheybani

A Dissertation Presented to
THE FACULTY OF THE USC VITERBI SCHOOL OF ENGINEERING
In Partial Fulfillment of the Requirements
for the Degree of
DOCTOR OF PHILOSOPHY
(BIOMEDICAL ENGINEERING)

August 2015
ii

ACKNOWLEDGEMENTS
“Life is not easy for any of us. But what of that? We must have perseverance and above all
confidence in ourselves. We must believe that we are gifted for something and that this thing must
be attained.” – Marie Curie
Yet perseverance is not possible without the guidance and support of others. I would not be able
to reach this point without the counsel and assistance of my family, friends, and mentors. While it
is quite impossible to name everyone that has helped me through this journey, I will try.
To my parents, who made sure that I had the best opportunities and supported me in achieving
my dreams, thank you. To my dad for being a great role model by blazing a course towards the
highest levels of education and to my mom for always being my biggest fan regardless of the
circumstances. I would also like to thank my sisters, each of them supporting me in their own
unique way; and my extended family, especially my aunt and uncle for being my home away from
home when I first started my undergraduate studies at USC and throughout the years since.
To Shahdad Irajpour, my partner, best friend, closest ally, and my most honest and supportive
critic. Thank you. I would not have reached this point without you.
As a woman in the male-dominated field of engineering, it may be difficult to find a strong
female mentor and role model. I was fortunate to meet Dr. Ellis Meng in my junior year of college.
Following the advice of one of my classmates, I met with Dr. Meng to talk to her about joining her
lab for undergraduate research. I did not have any prior research experience and did not know what
to expect. Little did I know that the meeting would put me on the course towards my dream career.
Her insight, creativity, knowledge, and approach to problem solving are undoubtedly the source
iii

of her success and amazing accomplishments. I could not have asked for a better mentor, and I
will be thriving to follow in her footsteps and emulate her in my future endeavors.
I also owe special thanks to the members of my guidance committee, Drs. Hashemi, Weiland,
Maarek, and D’Argenio; the administrative staff at the Biomedical Engineering department,
especially Mischalgrace Diasanta, Sandra Johns, Christopher Noll, Daisy Rusli, and Karen
Johnson; Dr. Donghai Zhu, our cleanroom manager; and Dr. Tuan Hoang.
To the members of the Biomedical Microsystems Laboratory, who have become closer than
family to me throughout the years, thank you. To Dr. Brian Li, my graduate mentor during my
undergraduate research, I hope to imitate his admirable patience and dedication in teaching and
training others. To Dr. Christian Gutierrez, for always challenging me to think for myself and to
consider all the aspects of a problem, and for answering questions with insightful questions. To
Dr. Heidi Tu, for being a great friend and a wonderful project partner. To Dr. Seth Hara, for not
only being my sounding board in matters of electrochemistry but also for being a true friend in all
aspects, I could not imagine a better roommate. To Angelica Cobo, in whom not only I found a
great project partner but also a kindred spirit. To Drs. Jonathan Kuo, Curtis Lee, and Brian Kim,
Lawrence Yu, Jessica Ortigoza, Ahuva Weltman, and Alex Baldwin, I am delighted to call each
of you a great friend, thank you all for your support. To Dr. Kee Scholten, who quickly became
my go to for advice while I was preparing for my defense and my career after graduate school.
I would also like to acknowledge all of the rotation, undergraduate, and highschool students I
have had the privilege of working with throughout my graduate work, they have been essential to
my success and I could not completed my work without their efforts.
Lastly, to Nan, for teaching me the magic is within me when I need it.
iv

My research was made possible through grants from the National Institute of Health (NIH),
Wallace H. Coulter Foundation, National Science Foundation (NSF), Henry M. Jackson
Foundation, Qualcomm and Telemedicine & Advanced Technology Research Center.

v

ABSTRACT
Drug delivery is essential for the treatment of chronic diseases. Implantable site-specific drug delivery
devices can deliver a potent and effective dose of drug directly to the site of therapy, improving treatment
outcomes while reducing potential side effects and the risk of infection due to catheters running through
the skin. These factors serve to increase patient comfort and decrease overall associated healthcare costs
for the treatment of chronic conditions. A MEMS approach miniaturizes infusion pumps such that they are
implantable; wirelessly powered to eliminate the use of bulky, limited lifetime batteries; and volume
efficient. Wireless data communication allows device status and performance to be monitored remotely as
well as remotely initiated changes to the drug regimen by caregivers throughout the course of treatment to
tailor the drug regimen to the individual needs of each patient. This work focuses on the design, fabrication,
and characterization of a wireless, programmable, low power electrochemical micropump capable of
delivering liquid drug formulations within a wide dynamic range of dose volumes and flow rates suitable
for the management of chronic conditions. This system rapidly responds to changes in regimen for on-
demand control and has a fully integrated wireless electrochemically based dose tracking system that is
capable of real-time tracking and confirmation of delivery, as well as assessing and recording changes in
reservoir content, delivery flow rate, blockages in the delivery catheter, and drug refills.

vi

TABLE OF CONTENTS
Acknowledgements ................................................................................................................................. ii
Abstract ................................................................................................................................................... v
Table of Contents .................................................................................................................................. vii
List of Tables ....................................................................................................................................... xiii
List of Figures ...................................................................................................................................... xiii
Chapter 1: Closed-loop On-demand Site-specific Drug Delivery ............................................................ 1
1 Background & Objectives ................................................................................................................ 1
1.1 Motivation .......................................................................................................................... 1
1.2 MEMS Pumping and Sensing Methods ............................................................................... 5
1.2.1 Micropumps ............................................................................................................. 5
1.2.2 Sensors .................................................................................................................... 7
1.3 Current State of the Art ....................................................................................................... 8
1.3.1 Commercially Available & Research Devices .......................................................... 8
Commercial Devices ................................................................................................ 8
Research Devices ................................................................................................... 10
2 System Design ............................................................................................................................... 11
3 Significance ................................................................................................................................... 12
4 Potential Applications .................................................................................................................... 13
5 References ..................................................................................................................................... 14
vii

Chapter 2: High-Efficiency MEMS Electrochemical Micropump for On-demand Site-specific Drug
Delivery .................................................................................................................................... 17
1. Preliminary Studies (Theory & Previous Work) ............................................................................. 17
1.1 Theory .............................................................................................................................. 17
1.1.1 Electrochemistry and Electrochemical Impedance Spectroscopy (EIS) ................... 17
1.1.2 Water Electrolysis .................................................................................................. 20
1.1.3 Nafion® ................................................................................................................. 23
1.1.4 Wireless Inductive Powering .................................................................................. 25
1.2 Previous Work .................................................................................................................. 27
2 Research Design & Methods .......................................................................................................... 29
2.1 Interdigitated Electrolysis Electrode .................................................................................. 29
2.1.1 Fabrication ............................................................................................................. 29
2.1.2 Electrochemical Cleaning and Characterization ...................................................... 31
Electrode Geometry ............................................................................................... 31
Electroplating and Nafion® Coating ....................................................................... 33
Electrochemical Cleaning ....................................................................................... 34
2.1.3 Flow Characterization ............................................................................................ 35
Flow Rate and Efficiency ....................................................................................... 35
Orientation ............................................................................................................. 38
Room Temperature versus Body Temperature ........................................................ 38
2.1.4 Flexible Electrodes ................................................................................................. 39
Parylene ................................................................................................................. 39
viii

PEEK ..................................................................................................................... 40
2.1.5 Recombination ....................................................................................................... 42
2.2 Electrochemical Bellows Actuator ..................................................................................... 45
2.2.1 Fabrication ............................................................................................................. 45
2.2.2 Flow Characterization ............................................................................................ 46
Continuous Operation ............................................................................................ 46
Bolus delivery ........................................................................................................ 48
2.2.3 Flow performance at body temperature ................................................................... 50
2.2.4 Effects of applied back pressure on flow performance ............................................ 51
2.2.5 Recombination ....................................................................................................... 52
Improving Recombination Efficiency ..................................................................... 53
Pt-coated bellows....................................................................................... 53
Pyrolyzing Nafion® on electrodes ............................................................. 53
Pt wire pieces ............................................................................................ 54
Pyrolyzed Nafion® coating on Pt wire pieces ............................................ 54
Recombination Orientation Dependency ................................................................. 56
Recombination Repeated Cycling ........................................................................... 58
Salt Permeation through Parylene Bellows ............................................................. 61
2.3 Electrochemically Actuated Drug Delivery Micropump..................................................... 62
2.3.1 Fabrication ............................................................................................................. 62
2.3.2 Pumping of Viscous Drug Models .......................................................................... 63
ix

2.3.3 Real-time Flow Response to Changes in Regimen .................................................. 65
2.3.4 Valved System ....................................................................................................... 66
2.3.5 Power Consumption ............................................................................................... 67
2.4 Wireless Powering & Flow Control ................................................................................... 69
2.4.1 Inductive Powering at a Single Infusion Rate .......................................................... 69
Link Efficiency Calculations .................................................................................. 72
2.4.2 Alleviating Coil Misalignment Effects on Power Transfer ...................................... 73
2.4.3 Remote Bluetooth Controllability ........................................................................... 74
2.4.4 Wireless Flow Control ........................................................................................... 75
Testing with “Dummy” Load ................................................................................. 78
Testing with Infusion Micropump .......................................................................... 79
Testing in Simulated Brain Tissue Material ............................................................ 80
3 Conclusion .................................................................................................................................... 82
4 References ..................................................................................................................................... 82
Chapter 3: Real-time Electrochemical Dose Tracking Sensors for Closed-loop Drug Delivery
Applications.............................................................................................................................. 87
Chapter 3-1: Electrochemical Dose Tracking Sensors ............................................................................ 87
1 Preliminary Studies (Theory & Previous Work) ............................................................................. 87
1.1 Theory .............................................................................................................................. 87
1.2 Previous Work .................................................................................................................. 90
2 Research Design & Methods .......................................................................................................... 91
2.1 Fabrication ........................................................................................................................ 91
x

2.2 Electrode Placement Optimization ..................................................................................... 92
2.3 Thin Film vs. Bulk Wire.................................................................................................... 93
2.4 Fluid Based Calibration ..................................................................................................... 94
2.4.1 Water Soluble Fluids .............................................................................................. 95
2.4.2 Lipid Soluble Fluids ............................................................................................... 97
2.5 Noise & Drift in Impedance Measurements ....................................................................... 97
2.5.1 Drift & Noise Surface Area Dependency ................................................................ 97
2.5.2 Drift Solution Dependency ..................................................................................... 99
2.5.3 Drift Temperature Dependency .............................................................................. 99
2.5.4 Three Electrode Configuration ............................................................................. 100
2.6 Delivery Operation & Real-time Dose Tracking .............................................................. 101
2.6.1 Smallest Detected Volume ................................................................................... 101
2.6.2 Real-time Flow Variation ..................................................................................... 102
2.6.3 Recombination Detection ..................................................................................... 104
2.6.4 Blockage & Refill Detection ................................................................................ 105
Chapter 3-2: Wireless Sensing ............................................................................................................. 107
1 Ominidirectional Wireless Data Transfer ..................................................................................... 107
2 Research Design & Methods ........................................................................................................ 109
2.1 System Architecture - 1 ................................................................................................... 109
Circuit Design and Layout .................................................................................... 110
Internal Transmitter & Receiver ............................................................... 110
xi

External Receiver .................................................................................... 111
Preliminary Experiments ...................................................................................... 111
Testing with “Dummy” Loads ................................................................. 111
Testing with Infusion Pump ..................................................................... 112
2.1.1 System Architecture - 2 ........................................................................................ 113
Circuit Design and Layout .................................................................................... 114
Internal Transmitter & Receiver ............................................................... 114
External Receiver .................................................................................... 115
2.1.2 Preliminary Testing with the Micropump ............................................................. 115
2.1.3 Calibration Testing ............................................................................................... 117
2.1.4 Wireless Sensing and Real-time Flow Variation ................................................... 120
3 Conclusion .................................................................................................................................. 122
4 References ................................................................................................................................... 123
Chapter 4: Wireless Electrochemical Drug Delivery Micropump with Fully Integrated Electrochemical
Dose Tracking Feedback System ............................................................................................. 126
1. System Design, Fabrication & Assembly ..................................................................................... 127
1.1 Micropump Fabrication ................................................................................................... 127
1.2 Wireless Power & Sensing System .................................................................................. 130
1.2.1 System Architecture ............................................................................................. 130
1.2.2 Circuit Design and Layout .................................................................................... 131
External Transceiver (Transmitter & Receiver) ..................................................... 131
Internal Transceiver (Transmitter & Receiver) ...................................................... 132
xii

2. Experimental Methods & Results ................................................................................................. 135
2.1 System Calibration .......................................................................................................... 135
2.1.1 Wireless Flow Control Calibration ....................................................................... 135
2.1.2 Wireless Dose Sensing Calibration ....................................................................... 137
2.2 Simultaneous Wireless Flow Control and Dose Sensing .................................................. 139
2.3 Recombination Detection ................................................................................................ 141
2.4 Effects of Coil Misalignment........................................................................................... 142
2.5 Delivery in simulated brain tissue .................................................................................... 143
3. Conclusion .................................................................................................................................. 145
4. References ................................................................................................................................... 145
Chapter 5: Conclusion ....................................................................................................................... 147

xiii

LIST OF TABLES
Table 2-1: Parameters for different Pt electrode layouts (modified from [14] © 2010 IEEE).
Table 2-2: Flow delivery results for uncoated and Nafion
®
coated electrodes and % increase in flow rate
after the addition of the coating (reprinted with permission from [4] © 2012 IEEE).
Table 2-3: Flow rate values for 1, 1.5, and 2 convolutions bellows actuators (reprinted with permission from
[58] © 2011 IEEE).
Table 2-4: Flow rate values for current controlled delivery at room temperature vs. body temperature
(reprinted from [1] with permission from Springer).
Table 2-5: Recombination time course expressed as % recombined for a 5.33 µL bolus was delivered with
actuators fabricated utilizing different methods to accelerate recombination [45].
Table 2-6: Average time required for complete recombination of different generated gas volumes [45].
Table 2-7: Power consumption data for uncoated and coated electrodes, actuators, and micropumps.
Table 2-8: Flow rate values for current supplied using the Class D wireless powering system (reprinted
from [1] with permission from Springer).
Table 2-9: Characteristics of the transmitting and receiving coils when tuned to 2 MHz.
Table 2-10: Percent drop in voltage output as a result of coil misalignment for one and two receivers.
LIST OF FIGURES
Figure 1-1: (a) Medtronic implantable pump [7]; (b) pump implanted in an adult male [8].
Figure 1-2: Illustrations depicting (a) continuous delivery in static therapeutic window and (b) continuous
and patient-tailored bolus delivery in dynamic therapeutic window.
xiv

Figure 1-3: Conceptual drawing of piezoelectric drug delivery device by Evans, et al. (reprinted from [42]
with permission from Elsevier).
Figure 1-4: Schematic diagram of proposed system.
Figure 2-1: Micropump operation concept (reprinted from [1] with permission from Springer).
Figure 2-2: Equivalent electrical model for electrochemical cell with two electrodes submerged in an
electrolyte solution (C dl = double layer capacitance, R p = polarization (faradaic) resistance, and
R s = electrolyte resistance; reprinted with permission from [4] © 2012 IEEE).
Figure 2-3: Interpreting an EIS Bode plot: (a) at low frequencies, the impedance value represents R s + R p.
At high frequencies, the impedance value corresponds to R s; (b) When the phase angle
approaches -90º, the slope of the impedance graph corresponds to C dl (reprinted with
permission from [4] © 2012 IEEE).
Figure 2-4: EIS interpretation examples: (a) increase in capacitance corresponds to an increase in surface
area; (b) decrease in solution resistance could be attributed to dissolving metal in the
electrolytic solution (reprinted with permission from [4] © 2012 IEEE).
Figure 2-5: An electrochemical cell; inset shows the ascent of bubbles as a result of water electrolysis [10].
Figure 2-6: The chemical structure of Nafion
®
(modified from [32] with permission John Wiley and Sons).
Figure 2-7: Equivalent electrical circuit for two Nafion
®
coated electrodes in an electrolytic solution: the
Nafion
®
is represented as a dielectric (reprinted with permission from [4] © 2012 IEEE).
Figure 2-8: Principle of inductive wireless power transmission.
Figure 2-9: Wireless powering circuit with multiple (1-3) transmitting coils (reprinted from [40] with
permission from Elsevier).
Figure 2-10: Wireless powering circuit with multiple (1-3) receiving coils (reprinted from [40] with
permission from Elsevier).
xv

Figure 2-11: Intraocular drug delivery device.
Figure 2-12: Electrolysis actuator packaged in flexible PDMS reservoir (modified from [43] with
permission from Springer).
Figure 2-13: Illustration detailing the electrode and bellows fabrication processes and bellows actuator
assembly (A 2 convolution bellows is shown; modified from [1] with permission from
Springer).
Figure 2-14: Photograph of Pt electrodes fabricated on: (a) glass, (b) Parylene (modified with permission
from [4] © 2012 IEEE), and (c) Photograph showing detail of the interdigitated electrode
design depicting the definitions for element width (100 µm) and element spacing (100 µm; the
dark regions correspond to the Pt electrodes; [45]).
Figure 2-15: EIS Bode plot for varying element width (constant overall electrode footprint) for a constant
(50μm) element spacing (reprinted with permission from [4] © 2012 IEEE).
Figure 2-16: EIS Bode plot for varying element spacing (constant overall electrode footprint) for a constant
(50μm) element width (reprinted with permission from [4] © 2012 IEEE).
Figure 2-17: Current controlled flow rate and calculated efficiency data for Pt only and electroplated Pt-Ir
electrodes (reprinted with permission from [4] © 2012 IEEE).
Figure 2-18: EIS Bode plot for uncoated Pt electrode before and after heavy anodizing (current application),
the cleaning effect is only observed after initial current application (reprinted with permission
from [4] © 2012 IEEE).
Figure 2-19: Testing setup for acquiring flow rate measurements.
Figure 2-20: Current controlled flow delivery and efficiency results for uncoated and Nafion
®
coated
electrodes (modified from [44] © 2011 IEEE).
xvi

Figure 2-21: Current controlled flow delivery results for Nafion® coated electrodes at room (25 °C) and
body temperature (37 °C).
Figure 2-22: Current controlled (a) flow rate and (b) efficiency results for Nafion® coated glass electrode
and Nafion® coated PEEK electrode (graphs modified from [44] © 2011 IEEE).
Figure 2-23: Electrodeposition on PEEK substrate (reprinted with permission from [44] © 2011 IEEE).
Figure 2-24: Recombination testing setup for experiments: (a) with one electrolysis electrode, (b) an
additional electrode is affixed to the inside lid of the test fixture as additional catalyst [45].
Figure 2-25: Comparison between recombined volume for uncoated and coated electrodes: (a) current
controlled (10 mA) delivery for different ON/OFF times (reprinted with permission from [58]
© 2011 IEEE) and (b) time controlled (2 min) flow rate delivery at different applied currents
(reprinted from [1] with permission from Springer).
Figure 2-26: Photograph of: (a) Pt electrolysis electrode, (b) 2 convolution Parylene bellows, (c) assembled
bellows actuator. Scale bars represent 4.5 mm [45]).
Figure 2-27: Current controlled flow delivery with different bellows configurations performed in a custom
test fixture (reprinted with permission from [58] © 2011 IEEE).
Figure 2-28: (a) Real-time pressure measurements of a 2 convolution Nafion
®
-coated bellows actuator
under constant current application; (b) Slope values for real-time pressure vs. time for different
current values (reprinted from [1] with permission from Springer).
Figure 2-29: Comparison between rapid-fire bolus delivery using uncoated and coated 2 convolution
bellows actuators (3 boluses at 10 mA, 15 s ON/10 s OFF, separated by 5 min OFF cycles;
reprinted with permission from [58] © 2011 IEEE).
Figure 2-30: Bolus delivery with a Nafion® coated, 2 convolution bellows actuator (at 10 mA current, 15
s/1 min ON/OFF; modified with permission from [58] © 2011 IEEE).
xvii

Figure 2-31: Bolus delivery with a Nafion® coated, 2 convolution bellows actuator (at 10 mA current, 2
s/1 min ON/OFF), inset: close-up for first10 boluses delivered (reprinted from [1] with
permission from Springer).
Figure 2-32: Flow delivery results for a range of physiologically relevant back pressures (at 5 mA constant
current). Shaded area depicts ± 5% from the flow rate value for zero backpressure (modified
from [1] with permission from Springer).
Figure 2-33: Recombination time course for a 5.33 µL bolus delivered with actuators fabricated utilizing
different methods to accelerate recombination [45].
Figure 2-34: The angle of an actuator clamped in a test fixture was varied for each volume (0°, actuator
facing up on the benchtop, 90°, actuator perpendicular to benchtop, and 180°, actuator facing
down on the benchtop) [45].
Figure 2-35: % Recombination vs. angular orientation for different delivered generated gas volumes for
unmodified actuators with Nafion®-coated electrode catalyst and actuators with additional
pyrolyzed Nafion® coating on suspended Pt wire pieces [45].
Figure 2-36: Hypothetical example: if the wait time (denoted with dashed vertical lines) between the first
and second set of generated gas boluses (10 and 50 µL, respectively) is not sufficient for
complete recombination, the bellows structure will exceed maximum inflation when the fourth
50 µL bolus is generated) [45].
Figure 2-37: (a) Repeated cycling (actuation and recombination), (b) averaged trends for actuation and
recombination of for 10 and 5 µL boluses [45].
Figure 2-38: (a) Actuation and recombination results obtained for different generated gas volumes of 0.277,
1.11, 5.33, 10, 25.33, and 50 µL for an actuator facing up on the benchtop. (b) Data for smaller
bolus volumes [45].
Figure 2-39: Drug delivery Micropump prototype with a rigid cylindrical poly propylene reservoir.
xviii

Figure 2-40: Flow delivery results for different viscosities of propylene glycol (at 8 mA constant current;
modified from [1] with permission from Springer).
Figure 2-41: Current controlled flow delivery of cocaine (in 0.9 N saline) loaded in a pump at different
concentrations (at 8 mA, 1.5 convolution bellows; modified with permission from [58] © 2011
IEEE).
Figure 2-42: Actuator response to “on the fly” changes in current: accumulated volume, and flow rate (n
=3, Mean ± SE).
Figure 2-43: Bolus delivery using an actuator with an in-line check valve (at 10 mA current, 15 sec ON/ 60
sec OFF). After 15 boluses, the actuator is turned off. Insets show photograph of valve, and
average accumulated volume for 15 boluses (modified from [1] with permission from
Springer).
Figure 2-44: Schematic diagram of Second receiver device with a low drop-out (LDO) switching regulator
(reprinted from [40] with permission from Elsevier).
Figure 2-45: The effects of (a) distance between coils and (b) foveation between the transmitter and receiver
on voltage output (reprinted from [40] with permission from Elsevier).
Figure 2-46: Link efficiency calculations using electrolysis cell resistance (Rcell) values as RL.
Figure 2-47: Wireless powering system with Bluetooth circuitry for wireless control: (a) Parallax Board of
Education, Basic Stamp 2 Module microcontroller, and an Easy Bluetooth Module; (b)
Transmitter circuit; (c) packaged micropump and receiver PCB (reprinted from [40] with
permission from Elsevier).
Figure 2-48: System architecture of ASK modulation circuit to enable wireless flow control (reprinted with
permission from [78] © 2014 IEEE).
Figure 2-49: Schematic diagram of the receiver circuit (reprinted with permission from [78] © 2014 IEEE).
xix

Figure 2-50: Wireless flow variation using the ASK modulation circuit. Output current levels are estimated
based on the results obtained for a 1 kΩ load.
Figure 2-51: Micropump with integrated receiver. Inset: Receiver flexible PCB, front and back (reprinted
with permission from [78] © 2014 IEEE).
Figure 2-52: 10 step incremental increase in receiver current output measured across a 1 kΩ resistor (at
each modification step, a specific pulse width was applied to achieve the desired change in the
wiper position; reprinted with permission from [78] © 2014 IEEE).
Figure 2-53: Current output repeatability for a specific modulation program (reprinted with permission from
[78] © 2014 IEEE).
Figure 2-54: Wireless flow variation in (a) incremental steps, and (b) random fashion (reprinted with
permission from [78] © 2014 IEEE).
Figure 2-55: Testing setup for experiments with simulated brain tissue material: (a) simulated brain tissue
placed between the pump and circuit and the transmitter (b) experiment repeated with only air
separating the transmitter and receiver at same distance.
Figure 2-56: Micropump with wireless powering coil and circuitry.
Figure 3-1: EI Dose sensing operation concept.
Figure 3-2: Electrode-electrolyte interface: excess charge on the surface of the metal and the resulting
accumulation of ions and molecules in the electrolyte create two layers of charge, named the
Helmholtz double layer (the dashed line separated the inner and outer layers).
Figure 3-3: Dose sensing is achieved by measuring the impedance across the sensing electrodes placed in
the reservoir. At a sufficiently high frequencies, the impedance represents the solution
resistance that can be correlated to the volume of fluid remaining in the reservoir.
Figure 3-4: Photograph of proof of concept prototype (modified with permission from [7] © 2011 IEEE).
xx

Figure 3-5: Photograph of (a) bulk wire electrode; (b) drug delivery system with integrated EI sensors.
Figure 3-6: High frequency inductance observed when EI electrodes are placed inside the electrolysis
chamber.
Figure 3-7: Investigating electrodes placement with respect to the bellows actuator: (a) electrodes arranged
perpendicularly, or (b) opposite one another on either side of the bellows actuator (EI sensing
electrodes are marked with X).
Figure 3-8: Cyclic voltametry for thin film electrodes and the Pt wire electrodes. Inset: electrochemically
active surface area equation.
Figure 3-9: EI spectroscopy for (a) DI water, and (b) 1× PBS. The boxed region represents the frequency
range at which the solution resistance is dominant (reprinted with permission from [11] © 2012
IEEE).
Figure 3-10: Normalized impedance response drift (1 Vpp, 1 kHz) after delivery comparing thin film
electrodes with epoxied wire and Pt wire electrodes (reprinted with permission from [11] ©
2012 IEEE).
Figure 3-11: Bolus delivery (6/10sec On/Off) impedance response (100 mVpp, 1 kHz). Two 5 µL boluses
delivered with 5 mA applied current (reprinted with permission from [11] © 2012 IEEE).
Figure 3-12: Drift solution dependency: drift in impedance magnitude for 1x PBS compared to that for
water.
Figure 3-13: Drift temperature dependency: drift in impedance magnitude for 1x PBS at room temperature
(25ºC) vs. body temperature (37ºC) over 24 hrs. Inset: first hour of measurement.
Figure 3-14: Photograph of the three electrode system.
Figure 3-15: Real-time dose tracking of actuator response to changes in the applied pump current levels
(modified from [16]).
xxi

Figure 3-16: Real-time dose tracking of lidocaine delivery [16].
Figure 3-17: Real-time blockage detection using EI sensing.
Figure 3-18: Tracking fluid refill into the reservoir using EI sensing.
Figure 3-19: Block diagram of frequency modulation circuit to enable sensing data transfer.
Figure 3-20: Block diagram of circuit using an impedance measurement integrated circuit chip to enable
sensing data transfer.
Figure 3-21: System architecture (1) of wireless sensing circuit.
Figure 3-22: Normalized received voltage values vs. frequency sweep for different resistors used in place
of the sensors (the values for each trace are normalized to the voltage measured at the resonance
frequency).
Figure 3-23: Wireless sensing with sensors integrated in an infusion pump. Each 80 µL dose leads to ~ 110
kHz shift in the resonance frequency of the internal transmitting coil.
Figure 3-24: System architecture (2) of wireless sensing circuit (reprinted with permission from [46] ©
2015 IEEE).
Figure 3-25: Schematic diagram of the internal (transmitter and receiver) circuit.
Figure 3-26: Photograph of sensors integrated in a micropump with a 1 mL reservoir. Inset photograph
shows the sensors placed vertically in the reservoir wall.
Figure 3-27: Wireless sensing with the infusion pump: each 37.5 ± 0.75 μL boluses (n =3, Mean ± SE; 3.75
% of reservoir fill volume) leads to a non-linear shift in the resonance frequency of the internal
transmitting coil (sensor response values for each trace are normalized to multiplication result
measured for the full reservoir; reprinted with permission from [46] © 2015 IEEE).
Figure 3-28: (a) pump performance and (b) sensor response to changes in applied current to the
microactuator leading to a change in the delivery flow rate (n=3, mean ± SE; sensor response
xxii

values for each trace are normalized to multiplication result measured for the full reservoir;
reprinted with permission from [46] © 2015 IEEE).
Figure 3-29: Delivery of 17.75 boluses (1.78 volume) of lidocaine HCl dissolved in saline (20mg/mL).
Shaded areas indicate 20 second on, followed by 1 min off. (Sensor response values for each
trace are normalized to multiplication result measured for the full reservoir; reprinted with
permission from [46] © 2015 IEEE).
Figure 3-30: Photograph of (a) micropump with integrated dosing sensors, (b) system setup (reprinted with
permission from [46] © 2015 IEEE).
Figure 3-31: Calibration testing for wireless sensing with three infusion pumps: seven boluses, each 14.67
µL (1.47 % of reservoir fill volume) delivered for each of the four runs (sensor response values
for each trace are normalized to multiplication result measured for the full reservoir).
Figure 3-32: Averaged results from calibration tests are used to obtain a linear relationship between the
volume delivered by the micropump and the sensor response.
Figure 3-33: Calibration curve robustness: the volume delivered by the micropump vs. the volume estimated
by the sensor response.
Figure 3-34:Real-time flow variation and simultaneous wireless sensing.
Figure 4-1: Schematic diagram of system.
Figure 4-2: Schematic diagram of system setup for small animal research.
Figure 4-3: Photographs of micropump assembly: (a) dosing sensors incorporated into the reservoir wall
across from each other, (b) vertically oriented sensors bent through the access ports to allow
connection with the wireless circuit, (c) electrolysis electrode fabricated on PEEK substrate,
(d) assembled micropump ready for the incorporation of circuit and coils (scale bars represent
4 mm).
xxiii

Figure 4-4: System architecture of wireless power and sensing system.
Figure 4-5: Schematic diagram of the external transceiver circuit.
Figure 4-6: Schematic diagram of the internal transceiver circuit.
Figure 4-7: Photograph of (a) external and (b) internal transceiver PCBs: top and bottom image show front
and back of PCB, respectively. Scale bars represent 10 mm.
Figure 4-8: Photograph of micropump with integrated sensors and circuity.
Figure 4-9: 5 step incremental increase in receiver current output measured across a 1 kΩ resistor and the
micropump (at each modification step, a specific pulse width was applied to achieve the desired
change in the wiper position).
Figure 4-10: Calibration testing for wireless sensing: 8 boluses, each 5.33 µL (1.33 % of reservoir fill
volume) delivered for each of the four runs (sensor response values for each trace are
normalized to multiplication result measured for the full reservoir).
Figure 4-11: Averaged results from calibration tests were used to obtain a linear relationship between the
volume delivered by the micropump and the natural log of the sensor response.
Figure 4-12: Linear relationship between the % accumulated volume delivered by the micropump and the
natural log of the sensor response for two different reservoir sizes. Accumulated volume is
normalized to each reservoir’s fill volume.
Figure 4-13: Simultaneous wireless flow control and dose sensing for an example dosing regimen,
comparing the expected delivered volume (based on the selected flow rate and delivery
duration), the estimated delivered volume (based on analysis of the sensor response), and the
actual delivered volume by the micropump.
xxiv

Figure 4-14: Testing setup for experiment with simulated brain tissue material: The pump and circuit were
then placed inside the simulated brain tissue so that each transmitting and receiving coil pair
was separated by 2 cm of the simulated tissue.
Figure 4-15: Photographs showing delivery of with 10x PBS mixed with blue food coloring in simulated
brain tissue material. The expected delivered volume (based on the selected flow rate and
delivery duration) and the estimated delivered volume (based on analysis of the sensor
response) are included at each time point.

1

CHAPTER 1: CLOSED-LOOP ON-DEMAND SITE-SPECIFIC DRUG DELIVERY

1 BACKGROUND & OBJECTIVES
1.1 MOTIVATION
Effective drug therapy is an essential tool in improving health outcomes in the treatment and
management of chronic conditions such as hypertension, respiratory disease, and diabetes, and therefore, a
critical healthcare need [1]. Chronic conditions affect an estimated 133 million Americans (46% of the
population in 2006). This number is estimated to reach 171 million (nearly half the population) by 2030.
Statistics gathered in 2006 showed that the management of chronic conditions contributes to an estimated
83% of total healthcare costs [2] and the average annual healthcare coverage cost for individuals suffering
from a chronic condition is $6,032, five times higher than for healthy individuals [1].
The commonly used oral, topical, or intravenous injection drug administration routes require high
systemic doses in order to achieve the intended therapeutic effect at remote sites within the body. Many
systemically administered drugs such as anti-cancer drugs, anti-fertility agents, and anti-inflammatory
steroids are associated with severe side effects [3] and therefore dramatically impact the patient’s quality
of life. Parenteral drug injection uses conventional needles and catheters and requires repeated time- and
labor-intensive procedures for chronic treatment. These are associated with frequent clinical visits, lost
productivity at the work place or in the field, low patient compliance, and elevated risk of infection. These
methods are also unsuitable for long-term treatment, have a narrow drug therapeutic window, and utilize a
complex dosing schedule with combination therapy or labile active ingredients [4]. On the other hand, many
emerging pharmaceutical agents including biologics, biosimilars, and other small molecules are not suitable
for administration through the aforementioned conventional routes [5]. For instance, the past quarter
century of outstanding progress in fundamental cancer biology has not translated into comparable advances
2

in the clinic. This discrepancy is in part due to the fact that only between 1 and 10 parts per 100,000 of
intravenously administered monoclonal antibodies reach their parenchymal targets in vivo [6].

Figure 1-1: (a) Medtronic implantable pump [7], (b) pump implanted in an adult male [8].
Shortcomings of current methods have motivated efforts to develop implantable site-specific drug
delivery systems that minimize side effects, improve safety, and have the potential to be more effective by
allowing the use of new and more potent drugs [9]. Commercially available pumps for site-specific delivery
are large and bulky, making them unsuitable for pediatrics and smaller patients (Figure 1-1) [5, 10, 11].
The decades-old mechanisms employed cannot be miniaturized and are inefficient in use of space. Their
incompatibility with new pharmaceuticals has led to recent recalls due to stalling and drug dumping
malfunctions [12]. On the other hand, while certain chronic conditions, such as drug addiction and
menopause, have associated static therapeutic windows and may be treated with constant drug
administration [13], most chronic conditions have chronobiological pattern in their pathogenesis [14-17].
More importantly, drug dosing affects the therapeutic window. Following successful drug administration,
the clinical need for additional pharmacological intervention for certain conditions diminishes transiently
while the risk of side effects dramatically increases (Figure 1-2)[5].
3

Figure 1-2: Illustrations depicting (a) continuous delivery in static therapeutic window and (b) continuous and patient-
tailored bolus delivery in dynamic therapeutic window.
Furthermore, according to reports by the US Food and Drug Administration (FDA), over a five year
period, of the 56,000 medical device reports relating to the use of infusion pumps, approximately 65% were
related to system malfunctions [18]. Unfortunately, these systems did not incorporate closed-loop feedback
sensors that could ensure that the correct dose and flow rate are delivered resulting in late diagnosis of
malfunction until physiological signs are reported by the patient. This could potentially lead to serious
health complications, injury, or even death [12, 19-21]. As a result, there are two compelling reasons for
the inclusion of sensors to track the delivered dose volume and flow rate and report to the user the state of
the pump in real-time, allowing for active control over the delivery profile. The first is to improve patient
care by confirming the correct delivery of the right drug dose to the desired therapy location for maximum
effectiveness. It has been shown that controlled administration of drugs can increase drug therapy
effectiveness, in some cases up to 60% [13]. The second, is to reduce the associated financial and human
cost of pump malfunctions or dosing errors [5, 22]. Nondestructive pharmacokinetic methods such as direct
observation, microdialysis, nuclear imaging and blood measurements have been developed to track and
4

confirm dosage level. However, these methods are indirect and do not offer real-time information, and as a
result are often subject to inherent limitations in resolution, accuracy, and detection limits [18]. The
incorporation of traditional fluid flow monitoring techniques, such as flow and pressure sensor
technologies, in implantable pumps has been quite challenging due to complicated fabrication procedures
and packaging challenges, biocompatibility, size, power consumption requirements, and the harsh nature
of the environment in which they are to function [23, 24].
Implantable drug delivery devices seek to mitigate some of challenges facing conventional drug delivery
methods by bypassing physiological barriers to enable delivery of new compounds such as biologics,
biosimilars, and other small molecules directly to the target tissue [25]; and to maximize therapeutic
efficacy of the drug and limit toxic side effects by delivering the correct amount of drug in the vicinity of
the target cells, while reducing the drug exposure to the non-target cells. Site-specific and controlled drug
administration avoids the peak and trough time course of drug concentrations in the plasma between
successive doses and the corresponding peak and trough pattern of drug action, leading to therapies that
mimic the chronobiological pattern of the condition [3, 4, 25]. Microelectromechanical systems (MEMS)
devices are scaled to allow implantation near the region of interest for localized drug therapy [4]. Moreover,
sensors can be incorporated into MEMS devices to monitor the drug release profile and provide useful
information for bioengineers and clinicians to optimize the drug therapy for the patients [4].
In order for a MEMS device to be suitable for drug delivery, certain requirements must be met. While
high volumetric flow rates are not likely to be required of implanted micropumps, precise fluid metering
and pressure generation requirements are of great importance since back pressure in the body of up to 25
kPa could be encountered depending on the delivery site [26]. The device should be capable of delivering
precise quantities of a drug at the right time and as close to the treatment site as possible. Other critical
requirements include long term reliability, accuracy, low power consumption and cost, compatibility with
an assortment of drug formulations (both existing and new), biocompatibility, and minimal patient
intervention following surgical placement [4, 26, 27]. The overall size of the device is also a major
5

consideration. The device should be as small as possible to allow for ease of implantation and removal (if
necessary) requiring only minor surgery and local anesthesia, and not cause any adverse effects after
implantation [4]. Some systems may possess many components including external patches or interfaces.
In many indications, chronic, repeated delivery is required (2-5 years), therefore, the accuracy, precision,
and reliability of the device should be maintained over the treatment duration [28]. Drug concentration
below or above the therapeutic limits might result in side-effects or even cause intoxication [29]. As a
consequence, the dose, frequency, duration, oscillatory behavior, toxicity, drug interaction and allergies
must all be considered when designing the device and possibly be customized for patients based on their
illness and history [27]. Ideally, drug can be delivered exclusively to the target tissues, cells, or cellular
components which suggests the need for developing delivery mechanisms that would equal or surpass the
selectivity of naturally occurring effectors [3]. It is also highly desirable for such drug delivery devices to
be tetherless. Eliminating the use of wires and catheters would lead to higher range of mobility for the
patient and allow for drug administration to take place outside of clinical settings. Despite advances in
battery technologies on their miniaturization, batteries are still large in comparison to MEMS devices and
can significantly increase the overall device size. Also, batteries have a limited lifetime and may pose risk
to the patient if leakage or malfunctions occur [30]. Wireless powering of drug delivery devices can
eliminate the battery and considerably reduce the size of the device. The system lifetime is also improved.
1.2 MEMS PUMPING AND SENSING METHODS
1.2.1 Micropumps
In the early 1980s, Jan Smits developed one of the first micropumps intended for use in controlled
insulin delivery systems for maintaining diabetics’ blood sugar levels without frequent needle injections
[26, 31]. This peristaltic pump consisted of three active valves actuated by piezoelectric discs. This
publication was first to demonstrate the feasibility of silicon-based micropumps and subsequently inspired
extensive research on micropumps [27].
6

A MEMS drug delivery system typically consists of micropumps, sensors, and control circuitry [29].
MEMS drug delivery devices can be divided into two categories based on the method by which the drug
release is controlled. In passive devices, the device is pre-programmed to release a certain drug dosage
profile. This can be achieved whether through passive mechanisms such as osmosis, or in response to an
environmental stimulus. Although these systems are generally less complex in terms of design and
fabrication, they have limited control of drug delivery. Active mechanisms require an external stimulus,
such as an electrical signal, radio frequency wave, or magnetic wave to alter the behavior of the drug
delivery device and cause it to dispense the drug. Active control requires power and these systems are
generally harder to design and fabricate, however, they offer precise metering of the dispensed drug [4].
For active micropumps timing and precision of delivery of liquid formulations depend on pump
microactuator performance. These micropumps can be grouped based on the actuator type as either dynamic
or displacement. In dynamic microactuators energy is continuously applied to the working fluid in a manner
that increases either its pressure directly or its momentum that is subsequently converted into pressure by
the action of an external fluid resistance [26]. Among dynamic actuation methods, magnetohydrodynamic
(MHD-DC) actuators are capable of delivering flow rates of up to 1200 µL/min [32], however, the working
fluid must be conductive and the system requires external electric and magnetic fields [29] that increase
power, size, and cost. Displacement micropumps, on the other hand, exert pressure forces on the working
fluid using one or more moving boundaries [26]. Among displacement micropumps, electrostatic [33] and
piezoelectric [34] actuators are capable of delivering high flow rates (>50 µL/min), and yet are generally
larger in size, structurally complex, and driving voltage and high power, so they are not suitable for wireless
applications [29]. Such actuator types are generally not suitable or difficult to implement in vivo. While
SMA micropumps retain linearity during deflection of the diaphragm, attain high stress (> 200 MPa), and
achieve long operation cycles; their response time to changes in regimen is slow and power consumption
may be high. Also they are usually not biocompatible.
7

Electrochemical actuation offers many advantages for displacement micropumps, including large
driving force, accurate flow control, low power consumption, and low heat generation [35]. The large
volume expansion achieved by electrolysis of water to form oxygen and hydrogen gas can be harnessed for
actuation [29]. Current control of electrolysis allows for on-demand activation of delivery and selection of
flow rate; the current magnitude and resulting flow rate are linearly correlated for flow control over a wide
range of flow rates. Dose volume is selected by pump activation time for a particular flow rate which
enables intermittent drug delivery. Electrochemical pumps have biocompatible construction and wireless
actuation [35] and are well suited for in vivo drug delivery applications. A more detailed review of MEMS
drug delivery systems can be found in [4, 29, 36-39].
1.2.2 Sensors
As previously mentioned, the incorporation of traditional fluid flow monitoring techniques, such as flow
and pressure sensor technologies, in implantable pumps has been quite challenging due to complicated
fabrication procedures and packaging challenges, biocompatibility, size, power consumption requirements,
and the harsh nature of the environment in which they are to function [23, 24]. Li, et. al., have developed
flow sensors fabricated on Parylene tape which were then rolled and fitted into catheters [40]. While, the
sensors offered high sensitivity after drift compensation (1.467 mV/ml), their operating flow range
(mL/min) is higher than flow rates typically observed in infusion pumps, their resolution is not suitable for
such applications, and they require a complicated method of fabrication and integration [41]. Evans, et. al.,
have developed a dual-chamber drug delivery microsystem, in which microvalves regulate flow from two
spring pressurized balloon reservoirs (Figure 1-3). Measured differential pressure from piezoresistive
pressure sensors embedded in the flow regulating microvalves is used to determine flow rate through the
microvalves [42, 43]. These pressure sensors offer high sensitivity, however, there operation is contingent
on the operation of the microvalves, their power consumption is not negligible, and they require complex
microfabrication techniques.
8

Figure 1-3: Conceptual drawing of piezoelectric drug delivery device by Evans, et al. (Reprinted from [42] with
permission from Elsevier).
Electrochemical impedance (EI) sensors are attractive for this application due to their simplicity and
sensitivity. This method is highly adaptable, widely compatible and can be integrated with various pumping
formats. Bohm, et. al., presented EI sensors for a closed-loop micro-dosing system used in micro total
chemical analysis systems (µTAS) and lab-on-a-chip applications. Their proposed system utilized
impedance-based measurements of the gas/liquid fraction within an electrolysis chamber and was capable
of detecting volumes as low as ±5 nL. However, the rigid silicon and glass structure was not suitable for
implantation, and the detection was limited to rather small volumes (< 1.5 µL). Another approach
demonstrated the first non-invasive, real-time monitoring of drug dissolution rate of reservoirs containing
35-45 µg of drug (in solid form) upon exposure to a liquid (1× PBS) environment for sustained release
applications [44]. However, this approach relied solely on passive diffusion processes to distribute the drug
and therefore lacked precise and active control over the delivery profile.
1.3 CURRENT STATE OF THE ART
1.3.1 Commercially Available & Research Devices
Commercial Devices - Several companies are actively commercializing miniature or MEMS-based drug
delivery devices. Durect Corporation (Cupertino, California) developed a mini osmotic pump suitable for
9

subcutaneous systematic drug delivery. The Duros® pump measures 3.8 mm in diameter and 44 mm long.
It can deliver up to 1000 mg of concentrated drug for up to a year at a constant rate (± 10%), ultimately
releasing greater than 95% of its drug content. The device was FDA approved for prostate cancer treatment
in 2000, however, clinical trials for new indications have been suspended pending redesign of the delivery
system, in order to address performance issues caused by premature shutdown of devices [45, 46]. This
system is one time use only, and depends on diffusion to deliver its drug content; therefore the flow rate
cannot be actively controlled or stopped after implantation. Also, the simple system structure does not
include any feedback mechanism to ensure correct delivery. Debiotech’s JewelPump
TM
containing a
piezoelectric actuator is currently in the final stage of development in Europe. This insulin delivery system
is worn as a patch on the skin and can be wirelessly programmed and monitored using a personal handheld
device such as a cell-phone. Each disposable reservoir is used for seven days. However, wireless technology
is only used for communication and the device still requires a battery for operation [47]. OmniPod® is an
SMA-actuated wearable mini-pump developed by Insulet Corp (Billerica, MA). The device allows for
subcutaneous delivery of insulin via a small cannula. A total insulin volume of 2000 µL is stored in the
disposable reservoir and can be delivered in 0.5 µL boluses continuously for about 72 hours. Pump
activation is achieved wirelessly using a wireless handheld device [48]. Subcutaneous delivery is not
suitable for applications were targeted site-specific delivery is required. The implantable MicroCHIPS®
micro-reservoir device utilizes thermal actuation to initiate the onset of drug release [46]. An array of 100
tiny reservoirs, each capable of holding 300 nL drug formulation per reservoir, store drug until it is ready
to be released in the body. Individual reservoirs are wirelessly activated to thermally open a membrane seal
and initiate drug release, thereby precisely control the dose delivered in single reservoir increments. Dosing
can be terminated without the need for device extraction [49, 50]. This system is one time use only, and
cannot be refilled. Also, once the membrane is opened, drug release depends on diffusion of drug out of the
reservoir. As a result, the flow cannot be precisely controlled. The second generation of the device was
approved by the FDA in 2013 (MicroCHIPS, Lexington, MA). Replenish has developed an
electrochemically driven, refillable ophthalmic micropump designed to allow variable nanoliter drug
10

delivery rates along with an external wireless programmer/charger for bi-directional communication [46,
51]. The Prometra® implantable pump system, developed by Flowonix, Medical (Mt. Olive, NJ), utilizes
a positive pressure gas expansion actuation design with battery powered valves for flow regulation [11].
The device is intended for chronic pain management and delivers morphine into intrathecal space with a
flow rate of up to 28 mL per day [52] (overall 97.5% dose accuracy). The 20 mL fixed-volume reservoir is
refillable. The system (diameter of 71 mm, 20 mm thick, weighing 150 g), is comprised of mostly immobile
parts that allow for usage for more than 10 years [53]. The system was FDA approved in 2012.
Research Devices - Evans, et al., have developed a piezoelectric micropump with low power flow
regulating check-valves, designed for intrathecal chronic pain management. Their design includes
embedded sensors for closed-loop control and error monitoring of the drug pumping, and delivers at rates
of 2.30 to 0.51 mL/h. To create a completely implantable system, electronics and 3.3 V batteries are added.
The overall system size is 5.08 cm × 9 cm × 3 cm, which is considerably smaller than commercially
available intrathecal pumps that do not employ MEMS techniques but still large for implantation especially
in smaller subjects [11, 43, 54]. Also as previously mentioned the sensor operation is contingent on proper
valve operation, and as a result they cannot be used to diagnose pump functionality independently. Another
drug delivery device developed by researchers at Cornell University consists of a small (~38 µm × 385 µm
×2.25 mm) implantable device for rapid delivery in ambulatory emergency care. Their system utilizes the
boiling of water to increase pressure in a hermetically sealed drug reservoir and deliver 20 µL of vasopressin
(a commonly used drug for cardiac resuscitation) in 45 s. This device may be implanted with and support
the function of current cardiac devices such as pacemakers and defibrillators, however, vasopressin
currently is only stable for 5 days within the reservoir and heating limits release of active drug to only 85%
of that originally loaded [55]. An electrochemical actuator specified for insulin delivery was reported by
Suzuki and Kabata in 2005. The system used hydrogen production to deflect a flat silicone membrane to
cause actuation and achieved a flow rate of 13.8 µL/min [56]. However, with this system, repeatable and
reliable delivery could not be achieved since the Ag/AgCl electrode depleted over time so that actuator
11

performance was not consistent. Also, the flat silicone membrane had limited deflection and could
potentially leak allowing toxic Ag
+
in the body.
Despite the research and commercial efforts in the past three decades, it is still clinically relevant
to investigate MEMS-based implantable drug delivery technologies that can provide controlled drug
volumes at specific times and locations within the body and include sensors that can provide feedback
for the device operation in real-time.
2 SYSTEM DESIGN
This work focuses on the design, fabrication, and characterization of a wireless electrochemical drug
delivery micropump with a fully integrated electrochemical dose tracking feedback system (Figure 1-4).
The micropump consists of an electrolysis based actuator
housed in a drug reservoir. At the heart of the actuator, a set of
interdigitated Nafion®-coated platinum (Pt) electrodes
electrolyze the water in the actuation chamber into hydrogen
and oxygen gas upon electric current application. The resulting
volume expansion is then harnessed to inflate the drug
separating bellows which in turn displaces the fluid in the drug
chamber and expels drug from the catheter to the delivery site.
The Parylene bellows separates the electrolysis reaction from
the pumped drug to prevent unwanted electrochemical or
chemical interactions. Also, electrolysis gases are trapped
within the bellows preventing their delivery with the fluid.
Once the current is removed, the gases recombine to once again
form water, allowing for repeated actuation and delivery. A pair of Pt wire segments coated in pyrolyzed
Nafion are used to facilitate and accelerate gas recombination. A normally closed check valve is required
Figure 1-4: Schematic diagram of proposed system.
12

to prevent backflow of fluids as a result of the reverse pressure gradient caused by recombination (out of
the scope of this work). A refill port is added to the drug reservoir to allow for transcutaneous refilling of
drug after implantation. Electrolysis actuation boasts low power consumption and low heat generation,
characteristics that are well suited for wireless powering and control. A class E inductive powering system
is designed, along with amplitude shift keying (ASK) modulation to allow adjustment of the infusion rate
post-implantation.
Dose tracking feedback is achieved using impedance-based sensing. A pair of Pt sensing electrodes are
incorporated in the walls of the drug reservoir. Electrochemical dose tracking is achieved through
measuring electrochemical impedance by applying a small sinusoidal excitation voltage across a these
electrodes. At sufficiently high frequencies, the measured impedance could be directly correlated to the
volume of drug remaining in the reservoir. Therefore, changes in reservoir content, delivery flow rate,
blockages in the delivery catheter, drug refills, and even damage to the electrolysis chamber could be
assessed in real-time and recorded for future analysis using A LabVIEW graphical user interface (GUI).
ASK modulation and resistive (RSK) modulation data transfer are utilized to achieve bi-directional data
telemetry between the dosing sensors and the external module.
Power application, current control, and sensing data analysis, is initiated using Bluetooth through an
off-the USB carrier board kit, microcontroller, and an Easy Bluetooth Module.
A hermetic housing could be added to encase the pump and sensors, as well as all the respective
electronics.
3 SIGNIFICANCE
The high-performance drug delivery micropump presented offers unprecedented accuracy and is capable
of delivering a diverse assortment of liquid drug formulations at the right dose, to the right tissue, and at
the right time over the entire course of treatment. In this manner, therapeutic efficacy is maximized while
13

minimizing unintended side effects. This refillable and implantable system boasts key features such as
wireless operation and on/off control, electronic dosing control, real-time dose tracking, and broad drug
compatibility. By utilizing MEMS fabrication techniques, the system components are scalable from small
animal models (including mice), up to adult human sized devices. The electrolysis-based actuation
mechanism offers large driving force, accurate flow control, generates low heat, requires low power
consumption, and enables repeated dosing. Electrolysis is controlled with electric current which allows for
on demand activation of delivery and selection of flow rate. With dose tracking, the micropump has closed-
loop feedback for monitoring pump performance that can improve device reliability and increase patient
safety. Impedance-based sensors are attractive for this application since they are simple, sensitive, widely-
compatible, independent of actuation method. Real-time monitoring of device performance will ensure the
correct dose size and flow rate are delivered and system malfunctions are reported.
4 POTENTIAL APPLICATIONS
For many chronic diseases, the timing and amount of drug dosing is critical to the effectiveness of the
therapy, and is often tied to biological rhythms. Rodent models are often used to evaluate new potential
therapeutics prior to human clinical testing, but technology for chronic drug administration in rats and mice
is quite limited, either from the point of view of lack of flow rate control or large size. The micropump
technology presented is scalable for animals as small as mice, providing researchers with flexibility in
chronic drug studies, and will lead to better understanding of and more effective treatments for chronic
disease. Ultimately the same technology can once again be scaled, this time for use in humans both
pediatrics and adults.
Device prototypes have been utilized for several applications such as intraocular drug delivery, cancer
gene therapy, drug self-administration and addiction in animal models, alcoholism, epilepsy, chronic pain
management, and intrathecal chemotherapy.
14

It is important to note that the actuation and sensing methods presented as part of a drug delivery
micropump system could also be utilized for other microfluidic fluid metering, biological, and medical
applications.
5 REFERENCES
[1] K. Menehan, "Partnership for Solutions: Better Lives for People with Chronic Conditions," Robert
Wood Johnson Foundation, 2006.
[2] The Council of State Governments, "Costs of Chronic Diseases: What Are States Facing?," 2006.
[3] D. Paolino, P. Sinha, M. Fresta, and M. Ferrari, "Drug delivery systems," Encyclopedia of Medical
Devices and Instrumentation, 2006.
[4] D. J. H. Tng, R. Hu, P. Song, I. Roy, and K. T. Yong, "Approaches and Challenges of Engineering
Implantable Microelectromechanical Systems (MEMS) Drug Delivery Systems for in Vitro and in
Vivo Applications," Micromachines, vol. 3, pp. 615-631, 2012.
[5] E. Meng and T. Hoang, "MEMS-enabled implantable drug infusion pumps for laboratory animal
research, preclinical, and clinical applications," Advanced Drug Delivery Reviews, 2012.
[6] M. Ferrari, "Cancer nanotechnology: opportunities and challenges," Nat Rev Cancer, vol. 5, pp. 161-
171, 2005.
[7] J. Barad, D. Buckland, J. Chen, J. Chang, and G. Corley. (2012), Ascenda, a Stronger Catheter for
Medtronic’s Baclofen Synchromed II Pump. MedGadget.
[8] (2011, 6/23/2013). Implanted Insulin Pump. Available:
http://lukesdday.wordpress.com/2011/04/26/implanted-insulin-pump/
[9] J. Fiering, M. Mescher, E. Leary Swan, M. Holmboe, B. Murphy, Z. Chen, et al., "Local drug
delivery with a self-contained, programmable, microfluidic system," Biomedical Microdevices, vol.
11, pp. 571-578, 2009.
[10] Ellis Meng, Tuan Hoang, and U. Sinha, "Implantable Infusion Pumps Incorporating
Nanotechnology," in The Textbook of Nanoneuroscience and Nanoneurosurgery, B. Kateb and J. D.
Heiss, Eds., ed: CRC Press, 2013, pp. 211 - 225.
[11] E. F. Lawson and M. S. Wallace, "Current developments in intraspinal agents for cancer and
noncancer pain," Current pain and headache reports, vol. 14, pp. 8-16, 2010.
[12] FDA: US Food and Drug Administration. (2012, 3/25/2012). Medtronic SynchroMed II Implantable
Drug Infusion Pump and SynchroMed EL Implantable Drug Infusion Pump. Available:
http://www.fda.gov/MedicalDevices/Safety/ListofRecalls/ucm333231.htm
[13] B. Bruguerolle and G. Labrecque, "Rhythmic pattern in pain and their chronotherapy," Advanced
Drug Delivery Reviews, vol. 59, pp. 883-895, 2007.
[14] B.-B. C. Youan, "Chronopharmaceutics: Gimmick or clinically relevant approach to drug delivery?,"
Journal of Controlled Release, vol. 98, pp. 337-353, 2004.
[15] W. Hrushesky, "Cancer chronotherapy: A drug delivery challenge," Progress in Clinical and
Biological Research A, vol. 341, pp. 1-10, 1990.
[16] F. Halberg, E. Haus, S. Cardoso, L. Scheving, J. Kühl, R. Shiotsuka, et al., "Toward a chronotherapy
of neoplasia: tolerance of treatment depends upon host rhythms," Cellular and Molecular Life
Sciences, vol. 29, pp. 909-934, 1973.
[17] C. G. Lis, J. F. Grutsch, P. Wood, M. You, I. Rich, and W. J. M. Hrushesky, "Circadian timing in
cancer treatment: the biological foundation for an integrative approach," Integrative cancer
therapies, vol. 2, pp. 105-111, 2003.
15

[18] N. Trombly. (2012) Emerging Technologies Ehance Drug Delivery Efficacy. Medical Electronics
Design. Available: http://www.medicalelectronicsdesign.com/article/emerging-technologies-
enhance-drug-delivery-efficacy
[19] FDA: US Food and Drug Administration. (2011). SIGMA Spectrum Infusion Pump Model 35700 -
Expanded Recall. Available:
http://www.fda.gov/MedicalDevices/Safety/ListofRecalls/ucm308488.htm
[20] FDA: US Food and Drug Administration. (2011). Medtronic Model 8637 SynchroMed II
Implantable Infusion Pump. Available:
http://www.fda.gov/MedicalDevices/Safety/ListofRecalls/ucm271492.htm
[21] FDA: US Food and Drug Administration. (2011, 5/10/2013). Roche Insulin Delivery Systems,
ACCU-CHEK FlexLink Plus Infusion Sets. Available:
http://www.fda.gov/MedicalDevices/Safety/ListofRecalls/ucm248772.htm
[22] R. S. Khandpur, Biomedical instrumentation: Technology and Applications: McGraw-Hill
Professional, 2004.
[23] S. Saliterman, "Fundamentals of BioMEMS and medical microdevices." vol. 153, ed: SPIE Press,
2006.
[24] C. A. Gutierrez, R. Sheybani, and E. Meng, "Electrochemically-based dose measurement for closed-
loop drug delivery applications," in 16th International Solid-State Sensors, Actuators and
Microsystems Conference, Transducers '11, Beijing, China, 2011, pp. 2839-42.
[25] J. Urquhart, J. W. Fara, and K. L. Willis, "Rate-controlled delivery systems in drug and hormone
research," Annual review of pharmacology and toxicology, vol. 24, pp. 199-236, 1984.
[26] D. Laser and J. Santiago, "A review of micropumps," Journal of Micromechanics and
Microengineering, vol. 14, p. R35, 2004.
[27] F. Amirouche, Z. Yu, and T. Johnson, "Current micropump technologies and their biomedical
applications," Microsystem Technologies, vol. 15, pp. 647-66, 2009.
[28] V. V. Ranade, "Drug delivery systems. 4. Implants in drug delivery," The Journal of Clinical
Pharmacology, vol. 30, pp. 871-889, 1990.
[29] N.-C. Tsai and C. Y. Sue, "Review of MEMS-based drug delivery and dosing systems," Sensors and
Actuators A (Physical), vol. 134, pp. 555-64, 2007.
[30] P. Si, A. P. Hu, J. W. Hsu, M. Chiang, Y. Wang, S. Malpas, et al., "Wireless power supply for
implantable biomedical device based on primary input voltage regulation," in 2007 Second IEEE
Conference on Industrial Electronics and Applications, Piscataway, NJ, USA, 2007, pp. 235-9.
[31] J. G. Smits, "Piezoelectric micropump with microvalves," in Proceedings of the Eighth Biennial
University/Government/Industry Microelectronics Symposium, Westborough, MA, USA, 1989, pp.
92-94.
[32] L. Huang, W. Wang, M. C. Murphy, K. Lian, and Z. G. Ling, "LIGA fabrication and test of a DC
type magnetohydrodynamic (MHD) micropump," Microsystem Technologies, vol. 6, pp. 235-40,
2000.
[33] J. W. Judy, T. Tamagawa, and D. L. Polla, "Surface-machined micromechanical membrane pump,"
in Proceedings. IEEE Micro Electro Mechanical Systems. An Investigation of Micro Structures,
Sensors, Actuators, Machines and Robots (Cat. No.91CH2957-9), 30 Jan.-2 Feb. 1991, New York,
NY, USA, 1991, pp. 182-6.
[34] K. Junwu, Z. Yang, T. Peng, G. Cheng, and B. Wu, "Design and test of a high-performance
piezoelectric micropump for drug delivery," Sensors and Actuators A (Physical), vol. 121, pp. 156-
61, 2005.
[35] P.-Y. Li, R. Sheybani, C. A. Gutierrez, J. T. W. Kuo, and E. Meng, "A parylene bellows
electrochemical actuator," Journal of Microelectromechanical Systems, vol. 19, pp. 215-28, 2010.
[36] A. Cobo, R. Sheybani, and E. Meng, "MEMS: Enabled Drug Delivery Systems," Advanced
Healthcare Materials, 2015.
[37] E. Meng and R. Sheybani, "Micro- and nano-fabricated implantable drug-delivery systems: current
state and future perspectives," Therapeutic delivery, vol. 5, pp. 1167-70, 2014.
16

[38] D. J. Laser and J. G. Santiago, "A review of micropumps," Journal of Micromechanics and
Microengineering, vol. 14, pp. 35-64, 2004.
[39] R. Sheybani and E. Meng, "Nano and Micro Electro Mechanical Systems (NEMS/MEMS)," in Drug
Delivery: An Integrated Clinical and Engineering Approach, Y. Rosen, Ed., ed: CRC Press: Taylor
& Francis Group, 2015.
[40] C. Li, P.-M. Wu, J. Han, and C. H. Ahn, "A flexible polymer tube lab-chip integrated with
microsensors for smart microcatheter," Biomedical Microdevices, vol. 10, pp. 671-679, 2008.
[41] J. T. Kuo, L. Yu, and E. Meng, "Micromachined Thermal Flow Sensors—A Review,"
Micromachines, vol. 3, pp. 550-573, 2012.
[42] T. Li, A. T. Evans, S. Chiravuri, R. Y. Gianchandani, and Y. B. Gianchandani, "Compact, power-
efficient architectures using microvalves and microsensors, for intrathecal, insulin, and other drug
delivery systems," Advanced Drug Delivery Reviews, 2012.
[43] A. T. Evans, J. M. Park, S. Chiravuri, and Y. B. Gianchandani, "Dual drug delivery device for
chronic pain management using micromachined elastic metal structures and silicon microvalves," in
Micro Electro Mechanical Systems, 2008. MEMS 2008. IEEE 21st International Conference on,
2008, pp. 252-255.
[44] A. M. Johnson, D. R. Sadoway, M. J. Cima, and R. Langer, "Design and testing of an impedance-
based sensor for monitoring drug delivery," Journal of the Electrochemical Society, vol. 152, pp.
H6-H11, 2005.
[45] DURECT Corporation. (2014, 7/17/2014). Products. Available:
http://www.durect.com/wt/durect/page_name/duros
[46] C. L. Stevenson, J. T. Santini, and R. Langer, "Reservoir-based drug delivery systems utilizing
microtechnology," Advanced Drug Delivery Reviews, 2012.
[47] Debiotech S.A. Switzerland. (2013). Debiotech Jewel Pump. Available:
http://www.jewelpump.com/standard.html
[48] M. Ochoa, C. Mousoulis, and B. Ziaie, "Polymeric microdevices for transdermal and subcutaneous
drug delivery," Advanced Drug Delivery Reviews, 2012.
[49] J. H. Prescott, S. Lipka, S. Baldwin, N. F. Sheppard Jr, J. M. Maloney, J. Coppeta, et al., "Chronic,
programmed polypeptide delivery from an implanted, multireservoir microchip device," Nature
Biotechnology, vol. 24, pp. 437-438, 2006.
[50] MicroCHIPS Inc. Driving Intelligent Drug Delivery. Available:
http://www.mchips.com/products/products_drugdelivery.html
[51] Replenish Inc. (2010-2013, 2/28/2013). Replenish, Inc. ophthalmic Micropumps. Available:
http://replenishinc.com/our-technology/product-description
[52] FDA: US Food and Drug Administration. (2014). Prometra® Programmable Infusion Pump System
- P080012. Available:
http://www.fda.gov/MedicalDevices/ProductsandMedicalProcedures/DeviceApprovalsandClearance
s/Recently-ApprovedDevices/ucm293502.htm
[53] S. Rosen, T. Deer, R. Rauck, J. Barsa, G. Dwarakanath, E. Dunbar, et al., "Multi-center evaluation
of drug delivery accuracy with the Prometra intrathecal infusion pump," Pain Med, vol. 10, p. 124,
2009.
[54] A. T. Evans, S. Chiravuri, and Y. B. Gianchandani, "A Multidrug Delivery System Using a
Piezoelectrically Actuated Silicon Valve Manifold With Embedded Sensors," Journal of
Microelectromechanical Systems, vol. 20, pp. 231-8, 2011.
[55] N. Elman, H. L. Ho Duc, and M. J. Cima, "An implantable MEMS drug delivery device for rapid
delivery in ambulatory emergency care," Biomedical Microdevices, vol. 11, pp. 625-631, 2009.
[56] A. Kabata, H. Suzuki, Y. Kishigami, and M. Haga, "Micro system for injection of insulin and
monitoring of glucose concentration," in Fourth IEEE Conference on Sensors, Irvine, CA, United
states, 2005, pp. 171-174.

17

CHAPTER 2: HIGH-EFFICIENCY MEMS ELECTROCHEMICAL MICROPUMP
FOR ON-DEMAND SITE-SPECIFIC DRUG DELIVERY

Electrochemical pumping is achieved by wireless inductive power
transfer resulting in constant current application to a pair of Nafion®
coated interdigitated Pt electrodes which converts water into hydrogen and
oxygen gases. This inflates a compliant Parylene bellows and expels
adjacent drug in a reservoir through a directed catheter to the delivery site
(Figure 2-1). Once the current application is ceased, the Pt electrodes
catalyze the recombination of gases into water, enabling repeatable
pumping.
1. PRELIMINARY STUDIES (THEORY & PREVIOUS WORK)
1.1 THEORY
1.1.1 Electrochemistry and Electrochemical Impedance Spectroscopy (EIS)
Electrochemistry is a branch of chemistry devoted to the study of chemical reactions involving the
transfer of electrons at the interface between an electrical conductor and an ionic solution. The beginnings
of the field can be traced back to the famous Italian physician and anatomist, Luigi Galvani, who in 1791
established a bridge between electrical current and chemical reactions in his famous frog leg twitch
experiments. Later in 1872, Oliver Heaviside created the foundation of impedance spectroscopy by
applying Laplace transforms to the transient response of electric circuits. Nernst applied the electrical bridge
invented by Wheatstone, to the measurement if dielectric constants for aqueous electrolytes. Approximately
50 years after Fick developed the laws of diffusion, Warburg developed expressions for the impedance
response using these laws, and introduced the electrical circuit analog for electrolytic systems in which the
Figure 2-1: Micropump operation concept
(reprinted from [1] with permission from
Springer).
18

capacitance and resistance are a function of frequency [2]. In 1947, a model was proposed by John E. B.
Randles [3], one of the most important theorists in the field of electrochemistry, to describe the electrode-
solution interface using the combination of basic circuit elements. The Randles’ circuit model is commonly
used as the basis for understanding the fundamental electrical response of the electrode interface.

Figure 2-2: Equivalent electrical model for electrochemical cell with two electrodes submerged in an electrolyte
solution (Cdl = double layer capacitance, Rp = polarization (faradaic) resistance, and Rs = electrolyte resistance;
reprinted with permission from [4] © 2012 IEEE).
When an electrode is placed in an electrolyte solution, two types of processes can occur at the electrode-
solution interface. In faradaic reactions, electrons are transferred across the metal-solution interface causing
oxidation and reduction reactions to take place at the metal surface. At certain applied potentials, charge
transfer does not transpire due to unfavorable thermodynamic conditions. Instead, absorption and
desorption – called nonfaradaic processes – take place. As a result, an array of charges and oriented dipoles
create an electrical double layer that acts effectively as a capacitance. The electrochemical cell can be
electrically represented using the Randles circuit model with a parallel contribution of the polarization
(faradaic) impedance and double layer capacitance, in series with the solution resistance [5]. If mass transfer
is assumed to be negligible and the processes are limited by diffusion, a two electrode system can be
modeled using the above simplified circuit (Figure 2-2) [2]. Under potentiostatic conditions, at low
frequencies, the double layer capacitance (C dl) acts as an open circuit and the impedance measured is
equivalent to the series combination of the polarization resistance (R p) and the solution resistance (R s). At
high frequencies, C dl acts as a short circuit and the impedance measured would be Rs only (Figure 2-3a).
At mid frequencies when the phase angle is closest to -90º, the slope of the impedance curve is
representative of C dl which is directly proportional to the active surface area of the electrode (Figure 2-3b).
19

In electrochemical impedance spectroscopy (EIS), a small alternating current is applied across the
electrodes at the open circuit potential (measured between one of the electrodes and a reference Ag/Cl
electrode) and through a range of frequencies to acquire impedance and phase data. The resulting Bode plot
can be fitted to the aforementioned model to extract the circuit parameters [2]. Changes in each of the
parameters (C dl, R p, and R s) offer useful information about the electrodes and electrolyte solution [5].

Figure 2-3: Interpreting an EIS Bode plot: (a) at low frequencies, the impedance value represents Rs + Rp. At high
frequencies, the impedance value corresponds to Rs; (b) When the phase angle approaches -90º, the slope of the
impedance graph corresponds to Cdl (reprinted with permission from [4] © 2012 IEEE).
For instance, an increase in the surface area, deduced based on the change in slope of the capacitive
region, may be due to surface activation as a result of electrochemical cleaning or heavy anodization,
whereas a decrease in surface area may be attributed to the addition of a dielectric coating (Figure 2-4a). A
drop in solution resistance may be attributed to dissolving metal particles as a result of delamination from
the surface of the electrode (Figure 2-4b).
(a)
(b)
20

Figure 2-4: EIS interpretation examples: (a) increase in capacitance corresponds to an increase in surface area;
(b) decrease in solution resistance could be attributed to dissolving metal in the electrolytic solution (reprinted with
permission from [4] © 2012 IEEE).
1.1.2 Water Electrolysis
Water electrolysis is possibly the oldest known example of direct conversion of electrical energy to
pressure/volume changes (Figure 2-5) [6, 7]. The first electrochemical actuator was reported by Janocha in
1988 [8]. In 1996, a patent filed in Germany described a stacked electrochemical actuator fabricated using
conventional techniques, designed to regulate process of devices such as control valves on radiators [9].

Figure 2-5: An electrochemical cell; inset shows the ascent of bubbles as a result of water electrolysis [10].
Electrolysis is initiated when a sufficient potential or current is applied to a pair of electrodes in contact
with an electrolyte (water). The reaction induces water dissociation into oxygen and hydrogen gas.
(a)
(b)
21

At the anode, water is dissociated into oxygen gas, cations, and electrons (2-1):

()⇄ 2

()+2

+0.5

() E° (25°C) = -1.23V (2-1)
where E° is the standard electrode potential. Protons are driven through the electrolyte to the cathode where
they combine with the electrons traveling through the external circuit to form hydrogen gas (2-2):
2

()+2

⇄

() E° (25°C) = 0.00V (2-2)
The net electrolysis reaction is (2-3):

()⇄

()+ 0.5

() E° (25°C) = -1.23V (2-3)
Thus, a 3:2 stoichiometric ratio conversion of gas to liquid occurs via the transfer of four equivalents of
electrons through an external circuit [11, 12]. Hydrogen is produced at twice the rate of oxygen, therefore
the rate determining step in the reaction is the production of oxygen [13]. The resulting volume of gas
produced (2-4) and water consumed (at constant pressure) can be calculated using (2-5):

=

!/# (2-4)

$%&'%(
=

×*

/+

(2-5)
where

is the moles of water transformed *

; and +

are the molecular weight and density of
water, respectively; R is the ideal gas constant (8.31 J/mol K); T is temperature; and P is pressure. Thus at
room temperature, the theoretical achievable volume expansion from liquid water to gas phase hydrogen is
(2-6) [11]:
,
-./
,
012314
=

5

67/8
.9:
;
<

=
/>
<

=
≈136,000% (2-6)
This large volume change provides an excellent mechanical power source and proceeds even in a
pressurized environment, making electrolysis an attractive actuation method [14]. There are no side
reactions during the electrolysis of water that could produce undesirable products and potentially reduce
the efficiency of gas generation [15, 16].
The efficiency (η) associated with the electrolysis reaction is expressed as (2-7):
E=
,
FGHFI1:FJ9.0
,
9KFLIF91M.0
(2-7)
22

where V experimental is the total generated gas volume (hydrogen and oxygen) and V theoretical is the theoretical
generated gas volume. For ideal conditions, the gas volume is produced proportional to the current applied
[15]. Therefore, under constant current conditions, Vtheoretical can be calculated from (2-8):

NOPQRPN%S$
=
NOPQRPN%S$
T=

U

%
V

W
T (2-8)
where qtheoretical is the theoretical gas generation rate (in m
3
/s), t is duration (in sec) over which the current
is applied, i is current (in A), F is Faraday’s constant (96.49×10
3
C/mol), and V m is the molar gas volume
at 25°C and atmospheric pressure (24.7×10
−3
m
3
/mol).
Gas generation efficiency is affected by losses due to recombination of gases into water, dissolution of
gases into the solution, Joule heating, pressure drop across the membrane the gases are acting upon, and
diffusion of gases through the membrane [15, 17]. Increasing current density by increasing active electrode
surface area and effectively increasing contact between electrodes and solution can reduce heat production
and maximize efficiency [16].
A number of molecular processes influence the rate of gas evolution and the performance of electrolysis
actuators. On a molecular level, several steps precede gas bubble evolution. Electrolytically evolved gas
first dissolves in the electrolyte adjacent to the electrodes leading to supersaturation of the adjacent
electrolyte facilitated by the low solubility and small diffusion coefficient of oxygen and hydrogen gases
in water. At low current densities, the dissolved gas remains in the electrolyte [13], and is transported to
the bulk electrolyte solely through diffusion, and so the supersaturation near the electrode becomes
constant. However, at higher current densities, the supersaturation will increase and exceed the value
required for bubble formation. Once the Archimedes’ force on the bubble exceeds the surface adhesion
forces, the bubble detaches from the electrode [18, 19]. Mass transfer of gases controls bubble growth [20]
and under constant current conditions, is proportional to the square root of time [21]. Thus, constant current
is the preferred mode for electrolysis actuation.
After the driving voltage or current is removed, generated gases begin to recombine into liquid water.
At the anode, hydrogen is oxidized, while at the cathode, oxygen is reduced. As a result, the overall reaction
23

is the oxidation of hydrogen by oxygen to form water [22]. This reaction requires the strong double bond
of molecular oxygen to be broken first, and as a result, it is limited by high overpotentials and activation
energy (~400 mV) [23]. In typical devices, electrolysis and recombination occur in the same chamber, and
therefore catalysts suitable for precipitating both forward and reverse reactions are the preferred material
of choice for electrodes [24]. Platinum (Pt) has this capability and is thus commonly selected as the catalytic
electrode material. In larger devices, due to the high cost of Pt, the electrodes are replaced with carbon-
based material. However, a significant drawback of this replacement is that the carbon-based electrodes are
corroded under oxygen evolution during electrolysis [25]. The challenge in choosing the appropriate
catalyst lies in the competing requirements for electrolysis and recombination. During electrolysis, a
hydrophilic catalytic surface is required [24, 26]. Recombination rate is subject to the slow diffusion of O 2
and H2 through the electrolyte to the surface of the Pt catalyst. The diffusion distance is affected by the ratio
of gas to liquid inside the electrolysis chamber [27] and the orientation and volume of the chamber in which
the reaction is conducted. At the surface of the catalyst, the presence of water could cause flooding,
preventing gas contact with the catalyst, which is detrimental to recombination efficiency. Thus, a
hydrophobic catalyst surface is preferred for recombination [28].
1.1.3 Nafion®
Nafion® is an ionomer developed by Dr. Walther Grot at DuPont in the late 1960’s [29]. The polymer
has a perfluorinated backbone and short pendant chains terminated by a sulfonic head group (Figure 2-6).
The sulfonic acid groups cluster to form a hydrophilic microphase surrounded by a hydrophobic
tetrafluoroethylene backbone. It has been shown that Nafion® undergoes molecular rearrangement at the
polymer-fluid/gas interface to minimize surface energy. When the ionomer is exposed to liquid water, the
hydrophilic sulfonic acid groups are drawn to the surface, and surface energy is minimized by presenting
the hydrophobic backbone to a gas interface [30].
Over time, during electrolysis, the slow diffusion of oxygen through water results in the formation of an
“O 2 film” on the electrode surface [20]. This film impedes further electrolysis and decreases pumping
24

efficiency by preventing liquid-metal contact [21]. Nafion® possesses high gas solubility (solubility of O 2
and H 2 is 1.8 times higher in Nafion® compared to water) and prevents bubble occlusion on the surface of
the Pt electrode by allowing rapid diffusion of gases away from the catalyst surface which leads to a 20%
increase in current density [31]. Faster gas diffusion can also aid in recombination by facilitating transport
of gases to the catalyst surface.

Figure 2-6: The chemical structure of Nafion
®
(modified from [32] with permission John Wiley and Sons).
The effective surface area of Nafion®-coated Pt electrodes, however, has been reported to be reduced
compared to that of the native, uncoated Pt electrode (86% of original area), suggesting that part of the Pt
surface is blocked by the electrochemically inactive hydrophobic fluorocarbon backbone of Nafion
®
[31].
Coating thickness was evaluated by EIS and selection of Nafion® concentration was guided by evidence
in research literature. The coating layer should be thick enough (>5 wt. %) to provide sufficient effective
reaction sites at the interface between Pt and the Nafion® coating. Optimal performance was reported for
coatings composed of less than 9 wt. % Nafion®. Performance degradation at >9 wt.% Nafion®, is said to
be due to the increasing resistance of mass transfer of O2 or H2 gas by diffusion to the electrode or gas
desorption from the electrode [26]. The system can be once again modeled with an electrical circuit
equivalent (Figure 2-7). Nafion® acts as a dielectric film and contributes to the double layer capacitance.

Figure 2-7: Equivalent electrical circuit for two Nafion
®
coated electrodes in an electrolytic solution: the Nafion
®
is
represented as a dielectric (reprinted with permission from [4] © 2012 IEEE).
Due to its material properties (thermal stability (up to 200ºC) and chemical and biological inertness)
25

Nafion® has been widely used as a proton exchange membrane in fuel cells [33]. Nafion® is also
biocompatible and has shown no acute or chronic foreign body response [32]. However, metallic alkali
metals (particularly sodium) can attack Nafion® directly under normal conditions of temperature and
pressure, therefore when used as a solid polymer electrolyte, it is best to hydrate it with water rather than
saline [29].
1.1.4 Wireless Inductive Powering
For transcutaneous powering, inductive transmission is most suitable compared to other forms of
wireless powering (radiative, conductive, and capacitive). Due to size restrictions, radiative powering
would require the frequency to be in the GHz range for sufficient power reception. This would lead to large
power dissipation in biological tissue and attenuation of the transferred power that are both undesirable.
Conductive and capacitive links rely on the electric field. The body acts as a bad conductor and leaky
dielectric which means both these methods would encounter large losses. Inductive powering, however
relies on magnetic field coupling and the transmission frequency can be chosen to minimize power
dissipation [34].

Figure 2-8: Principle of inductive wireless power transmission.
A typical inductive powering system consists of two separate coils, a transmitter, and a receiver placed
inside the body [35]. An alternating current (AC) applied to the transmitting coil can induce a magnetic
field that is then picked up by the receiving coil. The electromotive force that is produced as a result can be
converted to a direct current (DC) voltage or current source to drive an attached load (Figure 2-8)[34]. In
26

order to increase transmission efficiency, coils are tuned to the same resonant frequency, selected based on
the device size, placement, and the resulting attenuation of the magnetic field when traveling through air
and biological tissue. An important coil parameter for each of the coils is the quality factor, Q, which is
defined as (2-9):
X ≡
Z[
6
(2-9)
where R is the frequency dependent equivalent series resistance of the coil [36]. At the selected transmission
frequency, a high quality factor (>100) is desired to maximize the induced field on the secondary. For this
work, multi-wound Litz wire was used to fabricate the receiver coil in order to alleviate the shift in resonant
frequency caused by winding and ohmic losses [34, 37]. Sullivan’s standard analysis of proximity-effect
losses in combination with a power law modeling of insulation thickness could be used to find the optimal
stranding of Litz wire when designing coils [38]. Typically, the transmitter circuit is tuned in series to lower
load impedance and the receiver circuit is tuned with a parallel capacitor that cancels the inductive
impedance of the coil since this leads to better performance when driving a non-linear rectifier load [35].
Since DC power is required to drive the actuator, a rectifier is used to convert AC power to DC power.
Wireless powering has some limitations, such as short operating range, imperfect transfer efficiency due
to frequency tuning, and required alignment between coils [39]. When designing a wireless implant, it is
assumed that the receiver coil is placed inside a freely moving subject. Depending on the orientation of a
single transmitting and receiving coil, if the coils are oriented perpendicularly with regards to each other,
then the mutual inductance would be zero and essentially no power would be transmitted [34]. Therefore it
is important to account for the possible coil misalignment in the design. One solution would be to have
multiple transmitting or multiple receiving coils. When using multiple transmitting coils, up to three coils
could be used that are linearly independent. In this configuration, feedback from the receiver is used to
determine the transmitter coil with the best coupling and that coil is used for powering while the other two
coils are turned off (Figure 2-9). The second approach would be to use multiple receiving coils with one
transmitter. Once again up to three coils with perpendicular winding axes are sufficient. With three series
27

resonant secondary tanks, and three parallel rectifiers, the power received from all three coils can be
combined in a common filter capacitor. This approach does not require feedback (Figure 2-10) [34].

Figure 2-9: Wireless powering circuit with multiple (1-3) transmitting coils (reprinted from [40] with permission from
Elsevier).

Figure 2-10: Wireless powering circuit with multiple (1-3) receiving coils (reprinted from [40] with permission from
Elsevier).
1.2 PREVIOUS WORK
Initially, a drug delivery system integrated with an electrochemical micropump for the treatment of
intraocular diseases was reported [41]. In this pump, a silicone drug reservoir was integrated with an
interdigitated electrode to induce electrochemical phase change of liquid water into hydrogen and oxygen
gas for drug delivery (Figure 2-11). However, oxidation of the drug was observed due to the direct
electrolysis of the drug during pumping. Therefore, a separate pumping chamber is required to prevent drug
28

degradation and unwanted pH changes. A flexible Parylene bellows membrane was added to mechanically
couple the pumping chamber to the drug [14]. The design and fabrication of electrolysis-based actuators
for drug delivery were reported in [14]. Nine different layouts of interdigitated platinum (Pt/Ti) electrolysis
electrodes were fabricated and tested (with varying element width and spacing with the same overall
footprint). Flow rate and efficiency of each layout was determined; we reported operation up to 10 mA for
10 min and efficiencies of 49-90%. However, delamination of the metal traces from the glass substrate was
observed after current application at modest levels (~5 mA). Electrodes with higher surface roughness were
also produced by electrochemical deposition of Pt from aqueous ammonium hexachloroplatinate solution
onto the thin film interdigitated electrodes [42] and resulted in ~90% efficiency. However, the electroplated
metal also delaminated from the surface after a few cycles of testing. Higher efficiency with robust
construction was desired since it would lead to lower power consumption, paving the way for wireless
powering.

Figure 2-11: Intraocular drug delivery device.
While it was previously shown that electrodes having 50 μm element width and 100 μm element spacing
were associated with the highest efficiency, it was concluded that the thinner elements have higher risk of
delamination [14]. So, electrodes having 100 μm element width and 100 μm element spacing were chosen
for further studies. These electrodes were integrated with the Parylene bellows and packaged in flexible
Polydimethylsiloxane (PDMS) reservoirs (Figure 2-12). These systems were used to show preliminary
success in using gene therapy in combination with radiation to combat cancerous tumors in mice [43].
29

Figure 2-12: Electrolysis actuator packaged in flexible PDMS reservoir (modified from [43] with permission from
Springer).
2 RESEARCH DESIGN & METHODS
2.1 INTERDIGITATED ELECTROLYSIS ELECTRODE
2.1.1 Fabrication
Interdigitated Pt electrodes (100 μm wide elements separated by 100 μm gaps, 8 mm diameter footprint)
were fabricated on Borofloat® 33 glass wafers (University Wafer, Boston, MA), Parylene (Specialty
Coating Systems, Indianapolis, IN), or polyetheretherketone (PEEK) (CS Hyde, Lake Villa, IL) substrates
by liftoff (Figure 2-13 & Figure 2-14). A dual-layer photoresist process was used to create an undercut
sidewall profile to facilitate metal liftoff. First, AZ1518 photoresist (AZ Electronic Materials, Branchburg,
NJ) was spun at 4 krpm followed by global exposure (~ thickness 1.8 µm). Then, AZ4400 photoresist (AZ
Electronic Materials, Branchburg, NJ) was applied at 4 krpm and patterned (~ thickness 4 µm). Following
a short descum in oxygen plasma, a Ti/Pt film (300 Å/2000 Å) was e-beam evaporated and then liftoff was
performed. In order to separate the electrodes into individual dies, they were either diced into rectangular-
shaped dies using resin blade dicing saw (Disco DAD-2H/6, Giorgio Technologies, Mesa, AZ), or drilled
into circular-shaped dies using a circular glass hole saw (Ø24 mm). Prior to Nafion® coating, glass
electrodes were potentiostatically cleaned at ± 0.5 V (Gamry Reference 600 Potentiostat, Warminster, PA)
30

in 1× phosphate buffered saline (PBS) and rinsed in double distilled water. Nafion® (Dupont DE521
Solution, Ion Power, INC, New Castle, DE) was applied to a subset of the electrodes by dip coating 1, 2,
and 3 times which produces a 0.47, 0.97, and 1.51 µm thick coatings as measured using Dektak profilometer
(Sloan (Veeco), Plainview, NY) [44]. Conductive epoxy (Epo-tek® H20E, Epoxy Technology, Billerica,
MA) or soldering was used to affix Kynar™ silver plated copper wires (30 AWG, Jameco Electronics,
Belmont, CA) to contact pads on the electrodes. The joint was strengthened and insulated with
nonconductive marine epoxy (Loctite, Westlake, OH) [4].

Figure 2-13: Illustration detailing the electrode and bellows fabrication processes and bellows actuator assembly (A 2
convolution bellows is shown; modified from [1] with permission from Springer).

Figure 2-14: Photograph of Pt electrodes fabricated on: (a) glass, (b) Parylene (modified with permission from [4] ©
2012 IEEE), and (c) Photograph showing detail of the interdigitated electrode design depicting the definitions for
element width (100 µm) and element spacing (100 µm; the dark regions correspond to the Pt electrodes; [45]).
31

2.1.2 Electrochemical Cleaning and Characterization
In order to address delamination of metal and to improve overall actuator efficiency, the electrochemical
reactions at the surface of the Pt were investigated. Electrochemical methods such as electrochemical
impedance spectroscopy (EIS) and cyclic voltammetry (CV) have been widely used in coating evaluation,
corrosion studies, batteries, electrodeposition, semiconductor characterization, and more recently in
biotechnology applications such as characterization of biological cells, disease diagnosis, and cell culture
monitoring [46]. For this work, EIS was utilized as a tool for characterizing electrodes, cleaning,
determining surface area activity, and also understanding the effects of electrolyte coatings, electroplating,
substrate changes, and as a guide in the design and optimization of the electrodes. The results of which are
detailed in [4] and are summarized below.
Electrode Geometry: Electrode geometry, including electrode spacing, element width (Figure 2-2c), and
overall footprint, affects electrolytic gas generation and the achieved flow rate [14]. EIS is introduced as a
method to characterized electrode surface area. Using a Gamry Reference 600 Potentiostat, EIS was
performed in 1× PBS on nine different interdigitated electrode layouts (Table 2-1). The theoretical surface
area of these electrodes was calculated in [14] (Table 2-1).
Table 2-1: Parameters for different Pt electrode layouts (modified from [14] © 2010 IEEE).
Element Width Element Spacing Electrode Area
(μm) (μm) (mm
2
)
20
20 20.0
50 11.2
100 7.0
50
20 29.8
50 20.2
100 14.2
100
20 34.6
50 27.4
100 21.4
32

When the element spacing within the interdigitated electrodes is kept constant (50 μm) and the width of
the electrode elements is varied (20, 50 and 100 μm; Figure 2-15), the EIS data demonstrates that the double
layer capacitance decreases with decreasing element width, suggesting a decrease in active surface area for
electrodes having the same overall footprint. This agrees well with the theoretical surface area calculated
in [14].

Figure 2-15: EIS Bode plot for varying element width (constant overall electrode footprint) for a constant (50μm)
element spacing (reprinted with permission from [4] © 2012 IEEE).
When the element width of the interdigitated electrodes is kept constant (50 μm) and the spacing
between the elements is varied (20, 50 and 100 μm; Figure 2-16), the EIS data demonstrates that the double
layer capacitance decreases with increasing element spacing, suggesting a decrease in active surface area
for electrodes having the same overall footprint. Again, this agrees well with the theoretical surface area
calculated in [14].
33

Figure 2-16: EIS Bode plot for varying element spacing (constant overall electrode footprint) for a constant (50μm)
element width (reprinted with permission from [4] © 2012 IEEE).
Electroplating and Nafion® Coating: To increase efficiency of electrolysis, one strategy is to increase
the surface area of the Pt catalyst electrode. Since space is limited, the active surface area can be increased
through electroplating while keeping the geometric area constant. Electrochemical deposition can be carried
out with a simple process at room temperature and atmospheric pressure [47].
Previously, electroplating the surface of the electrode with Pt from aqueous ammonium
hexachloroplatinate solution [42], increased flow rate and lead to an increase in electrolysis efficiency.
However, the electroplated material delaminated from the surface of the electrode shortly after use (current
application) [14]. Electroplated Pt is subject to mechanical damage due to the softness of the deposit.
Adding iridium (Ir) increases the mechanical stiffness. Using the method developed by Petrossians et. al, a
60-40% Pt-Ir alloy was electrodeposited using the potential cycling method (E = 0.1 to -0.1 V vs. Ag/AgCl
reference, at 0.5 V/s) at a rate of 16.5 nm/min and a stable film of 500 nm thickness was produced [47].
The electrodes were then Nafion® coated.
34

Figure 2-17: Current controlled flow rate and calculated efficiency data for Pt only and electroplated Pt-Ir electrodes
(reprinted with permission from [4] © 2012 IEEE).
Significant delamination and electrode damage was not observed in the electroplated Nafion
®
coated
electrodes for currents up to 8 mA. Prolonged current application at 10 mA resulted in some loss of
electroplated Pt-Ir. While electroplating does increase flow rate and electrolysis efficiency by increasing
the active surface area of the electrode (Figure 2-17), the increase may not be significant enough to offset
the cost and time consumed to produce the electroplated electrodes when operating at higher currents. At
lower currents, a significant increase in efficiency is gained (24.24% increase at 2 mA).
Electrochemical Cleaning: Electrochemical cleaning removes impurities from the surface of the electrode,
increasing the active surface area available on the Pt to catalyze the electrolysis reaction. This can improve
electrolysis efficiency which in turns results in a higher flow rate from the actuator [48]. Electrodes were
potentiostatically cleaned at ±0.5 V (Gamry Reference 600 Potentiostat, Warminster, PA) in 1× PBS. EIS
was performed in 1× PBS on uncoated Borofloat® electrodes before and after electrochemical cleaning.
35

Figure 2-18: EIS Bode plot for uncoated Pt electrode before and after heavy anodizing (current application), the
cleaning effect is only observed after initial current application (reprinted with permission from [4] © 2012 IEEE).
Cleaning removes impurities from the electrode surface, thus increasing the effective surface area. This is
evidenced by the increase in double layer capacitance (data not shown). Indirectly, the drop in the
polarization resistance shows an increase in the rate of reaction [48]. Therefore, potentiostatic
electrochemical cleaning is an important step in pre-conditioning of electrodes before Nafion® coating in
order to maximize the electrolysis efficiency. Interestingly, “heavily anodizing” the electrode, which
occurs when current is applied, also emulates electrochemical cleaning [48]. The first time current is applied
to an uncleaned, uncoated electrode, the surface of the Pt electrode is cleaned, so that the active surface
area increases. This results in a downward shift in the EIS curve which correlates to an increase in the
double layer capacitance after the first current application. Subsequent current application, even at higher
currents, does not result in additional shifts since the electrodes are already clean (Figure 2-18). These
results suggest that either method can be used to clean electrode surfaces prior to use.
2.1.3 Flow Characterization
Flow Rate and Efficiency: The flow rate performance of uncoated and coated electrodes (one, two,
and three dip coats) was acquired and compared. The electrochemically actuated flow rate was determined
by clamping the electrode in an acrylic test fixture filled with double distilled water. The accumulated
pumped water was collected and weighed over 2 or 5 min of constant current application (Figure 2-19).
36

Figure 2-19: Testing setup for acquiring flow rate measurements.

Flow rate and efficiency were then calculated. Higher flow rates were achieved for coated in comparison
to uncoated electrodes (Table 2-2; average increase of 182.30% across conditions examined). The results
clearly indicate an increase in flow rate at all current values tested for the Nafion®-coated electrodes (Figure
2-20). The gain is more pronounced at lower currents. Also, the solid polymer electrolyte acts as a barrier
between the gas bubbles and the active electrode surface which increases the sites for gas bubble formation.
The net effect is greater linearity in the flow rate versus current relationship (uncoated R
2
= 0.971, Nafion®
coated R
2
= 0.999). Efficiency increased with current (an average of 47.58% across all current values
tested), reaching 94% for Nafion®-coated electrodes compared to 53% for uncoated at 13 mA [44]. Higher
electrolysis efficiency leads to lower power requirements [Nafion®-coated glass electrodes require 1.63–
51.31 mW power to operate under constant current conditions (0.75– 13 mA); power consumption will be
discussed more in depth later] to generate the same flow rate making this actuation method suitable for
wireless powering. Also, heat generation is inversely proportional to efficiency which is significant when
considering applications involving in vivo use [16]. However, even at low efficiency, the heat generation is
negligible in the context of in vivo operation.
37

Figure 2-20: Current controlled flow delivery and efficiency results for uncoated and Nafion
®
coated electrodes
(modified from [44] © 2011 IEEE).

Table 2-2: Flow delivery results for uncoated and Nafion
®
coated electrodes and % increase in flow rate after the
addition of the coating (reprinted with permission from [4] © 2012 IEEE).
Uncoated Coated
% increase
Current
Flow rate ± SE
(µL/min)
Flow rate ± SE
(µL/min)
2 2.6 ± 0.47 14.16 ± 0.34 444.62
5 17.65 ±0.17 46.4 ± 0.32 162.89
8 36.45 ± 1.15 82.65 ± 0.38 126.75
10 53.3 ± 1.95 106.8 ± 0.72 100.37
13 80.25 ± 1.26 141.95 ± 0.46 76.88

Coated electrodes were robust even after multiple current applications at high currents; no signs of
delamination or damage to the electrodes or the coating layer were observed. The coating layer in
combination with the Ti adhesion layer has solved the delamination issues previously observed for high
current operation (> 5 mA). EIS Bode plots for electrodes coated with a single, double, and triple layer of
Nafion
®
were not significantly different. Also, as expected based on the EIS results, the coating thicknesses
examined did not exhibit significantly different flow rates (one way analysis of variance, α = 0.05).
However, single-layer coatings were compromised after repeated operation at high current (> 8 mA) (no Pt
delamination was observed for glass electrodes even when uncoated). Sufficient robustness was achieved
38

for electrodes that were dip coated twice (same performance in over 1000 min of current application ≤ 10
mA).
Orientation: Flow rate performance tests were carried out with the electrode clamped in the test fixture
and placed facing up on the benchtop. However, for drug delivery, it is desired and sometimes inevitable
for the actuator to assume other orientations, as the device may be implanted in a certain position or the
subject may move. Flow rate dependence on actuator orientation was tested with and without Nafion®
during constant 10 mA current application by changing the angle of an actuator clamped in a test fixture
(0°, clamped electrode facing up on the benchtop, 90°, clamped electrode perpendicular to benchtop, and
180°, clamped electrode facing down on the benchtop). The Nafion® coating results in a stable flow rate
with less deviation between separate trials regardless of actuator orientation; the effect is most pronounced
at 180° orientation, at which the flow rate for the uncoated electrode decreased by 17.12% compared to the
coated electrode that showed a decrease of < 1.5%. This is likely due to the prevention of gas bubble
occlusion of the electrode surface.
Room Temperature versus Body Temperature: In implantable drug delivery applications, the actuators
will be operated at body temperature. Therefore, it is important to understand the role of temperature on
electrochemical actuation. It has been shown that Nafion® permeability to oxygen and hydrogen slightly
increases as the temperature is raised from room temperature to body temperature [49]. Increased
permeability of Nafion® could increase the rate of reaction which in turn increases flow rate. Flow rate was
evaluated at room temperature (25 °C) and body temperature (37 °C). Nafion®-coated electrodes were
clamped in a test fixture and then placed in a temperature-controlled water bath. Pumped water was
accumulated and weighed over 2 min of constant current application. As expected, the slight increase in
permeability does not have a significant effect on the actuation flow rate at any of the applied currents
(Figure 2-21). This was confirmed by performing one way analysis of variance for each current value (α =
0.05).
39

Figure 2-21: Current controlled flow delivery results for Nafion® coated electrodes at room (25 °C) and body
temperature (37 °C).
2.1.4 Flexible Electrodes
If the substrate that supports the electrodes is flexible, conforming actuators can be realized that take on
the shape of the organ to which it is applied. One application in which this feature would be beneficial is
ocular drug delivery where a pump is wrapped around the eye [41]. Also, using flexible substrates that can
be cut would enable electrodes of any desired shape and overcomes limitations set by standard dicing
methods used for the glass electrodes.
Parylene: Parylene C is a poly(p-xylylene) polymer with low Young’s modulus. It is classified as a
United States Pharmacopeia class VI polymer and highly inert which makes it suitable for biomedical
applications. Parylene is also compatible with MEMS microfabrication [50]. Thus, Parylene was selected
as a flexible substrate material. Flexible electrodes were fabricated on Parylene C films (5 µm thick)
supported on a silicon substrate. Dies containing electrodes were released after patterning. Following
removal from the supporting Si substrate, electrodes were mounted on a flexible PEEK backing to facilitate
handling. Electron beam deposition of metal is a physical process during which the metal atoms condense
on the surface of the substrate. The adhesion between the metal and the substrate is through weak Van der
Waals interactions, and the binding energy is less than a few tenths of electron volts [51]. Studies have
shown that even though Pt is the preferred material for microelectrode fabrication due to its resistance to
40

corrosion, biocompatibility, and low threshold potential [52], the electrodes have a limited lifetime under
current application. Dissolution and delamination of Pt have been shown to be dependent on the stimulation
parameters [53] and are worsened during water electrolysis [54]. Increasing the electrode surface area and
using adhesion promoters between the Pt and the electrode substrate have been suggested as ways to
enhance electrode adhesion [52]. Studies have suggested that annealing of metal electrodes on Parylene C
causes microstructure enhancement and improves the crystallinity of Parylene, which may lead to better
adhesion between the metal and the Parylene [55]. In order to combat delamination, the effects of annealing,
the addition of a Ti adhesion layer, and Nafion® coating were studied [4]. The annealing did not prevent
delamination upon current application (> 1 mA), and while the adhesion promoter delayed delamination,
damage was still observed starting with currents as low as 1 mA. Nafion® coating prevented delamination
at low currents but only delayed delamination at higher currents (> 5 mA). Therefore, it was concluded that
Parylene may not be a suitable material for applications requiring high current application without further
treatment to improve the adhesion of Pt.
PEEK: PEEK, a semi-crystalline thermoplastic, is widely used in aerospace and chemical process
industries due to high thermal stability, low elastic modulus, as well as its desirable mechanical and
chemical properties. It is also approved for medical use, one such example is its use as a substitute for
metallic implant materials. Han et al. reported that electron beam deposition of titanium on PEEK enhanced
the biological properties of implants [56]. We also investigated the use of PEEK as a flexible substrate
material. Another set of electrodes were fabricated by depositing the Pt with a Ti adhesion layer directly on
the surface of flexible PEEK sheet (0.5 mm thick, as reported in the manufacturer’s specs, CS Hyde, Lake
Villa, IL). A subset of electrodes were coated with Nafion®. Flow rate was measured by clamping the
uncoated and coated electrodes in a test fixture filled with double distilled (DD) water. The accumulated
pumped water was weighed over 5 min of constant current application and used to determine flow rate. The
efficiency was calculated using the measured accumulated volume and the calculated theoretical volume
produced. As a result of the combined effect of the rough PEEK surface (increased electrode surface area)
41

and the coating, higher efficiency of up to 97% (13 mA) was achieved compared to electrodes on
Borofloat® glass. Efficiency increased an average of 8.15% compared to Nafion®-coated glass electrodes
(Figure 2-22). Average roughness of the surface was measured by a Dektak profilometer to be 14266 Å
compared to glass which has an average roughness of 20 Å (as reported in the manufacturer’s specs, Mark
Optics, Santa Ana, CA).

Figure 2-22: Current controlled (a) flow rate and (b) efficiency results for Nafion® coated glass electrode and Nafion®
coated PEEK electrode (graphs modified from [44] © 2011 IEEE).
It is important to note that for PEEK electrodes operated at sufficiently high electrolysis currents (>5mA),
an electrodeposited material was observed on the surface (Figure 2-23). To investigate this phenomenon,
EIS was performed in 1× PBS before and after the formation of the deposits (data not shown).
The double layer capacitance increased which correlated with increased surface area from the rough
substrate and the presence of an unknown electrodeposited material following current application.
Electrodeposition was more pronounced on the cathode. This electrodeposited layer is likely the result of
the electrolytic oxidation of the PEEK polymer on Pt similar to that observed in [57]. Once coated with
Nafion®, the coating protects the surface and does not allow the formation of the electrodeposited material
even at high currents.
(b) (a)
42

Figure 2-23: Electrodeposition on PEEK substrate (reprinted with permission from [44] © 2011 IEEE).
2.1.5 Recombination
As previously mentioned, once the current is removed, gases recombine to form water, however, this
reaction is limited by high overpotentials and high activation energy [43, 49]. Recombination is an
important factor for reliable and repeatable delivery. A simple acrylic test fixture with a fixed chamber
volume (3 mL) [58] was used for all experiments. Electrodes were mounted at the bottom of the water
chamber. The generated and recombined gas volumes were calculated by measuring water front
displacement in a calibrated 100 µL micropipette attached to the test fixture outlet (Figure 2-24a).
Different volumes of gas were generated by varying the duration (1, 2, 4, 6, 8 min) of 10 mA current
application. Another set of experiments were carried out, in which the volume of gas generated was varied
by applying different constant current values (1, 2, 5, 8 mA) for 2 min. Recombination was observed was
measured based on the fluid back flow in the micropipette for an hour following each current application.

43

Figure 2-24: Recombination testing setup for experiments: (a) with one electrolysis electrode, (b) an additional
electrode is affixed to the inside lid of the test fixture as additional catalyst [45].
Figure 2-25 displays the volume of recombined gas versus the volume of electrolyzed gas for the
experiments in the 3 mL test chamber. In both tests with the Nafion®-coated electrodes, recombination
volume increased monotonically with greater amounts of available gas from electrolysis, however with
tests on uncoated electrodes this was not consistently observed. These experiments show a faster rate and
larger total volume of recombination possible with the Nafion®-coated electrodes. For uncoated electrodes,
recombination proceeds slowly as it is dominated by the slow diffusion of gases through water towards the
surface of the Pt catalyst [18]. Oxygen and hydrogen gases have a higher solubility in Nafion® compared
to water [31] (O 2 solubility is 7.2×10
−6
mol.cm
−3
in hydrated Nafion® [59] and 0.26×10
−6
mol.cm
−3
[60] in
water and H 2 solubility is 1.4×10
−6
mol.cm
−3
in hydrated Nafion® and 0.78×10
−6
mol.cm
−3
in water [31].
The higher solubility facilitates diffusion of gases to the electrode surface and may also encourage a
concentration gradient that further drives diffusion; therefore, recombination is facilitated and increased in
Nafion®-coated compared to uncoated electrodes. Recombination rate is a function of three factors:
transport of gas through the electrolyte to the surface of the catalyst, ratio of gas to liquid inside the reaction
chamber (e.g. volume of gas generated during the actuation period), and the size of the chamber. Gas
recombination occurs in two phases. The rate is diffusion limited while bubbles are diffusing through the
water and Nafion® to the surface of the catalyst but once a sufficient volume of gas bubbles are available
at the surface of the catalyst, the recombination becomes reaction limited [61].
44

For uncoated electrodes, hydrophilicity at the surface of the catalyst causes flooding of the
electrocatalyst and the gases are repelled [24, 26]. It has been shown that Nafion® undergoes molecular
rearrangement at the polymer-fluid/gas interface to minimize surface energy. As such, for Nafion®-coated
electrodes during recombination, surface energy is minimized by presenting the hydrophobic backbone to
a gas interface [30]. As a result, the rate of reaction at the surface of the catalyst is faster for Nafion®-
coated electrodes compared to uncoated electrodes.

Figure 2-25: Comparison between recombined volume for uncoated and coated electrodes: (a) current controlled (10
mA) delivery for different ON/OFF times (reprinted with permission from [58] © 2011 IEEE) and (b) time controlled (2
min) flow rate delivery at different applied currents (reprinted from [1] with permission from Springer).
In an attempt to speed up recombination, a second electrode was affixed to the inside lid of the acrylic
test fixture, directly above the electrolysis electrode (Figure 2-24b). 10 mA current was applied for 2 min
to the electrolysis electrodes and recombination was observed for an hour.
With the addition of the second Pt catalyst surface in the chamber (with the attached electrode to the
inside lid of the test fixture), of the 215 µL generated, 37.11% was recombined within 1 hour, compared to
the 11.58% for when there was no additional catalyst for recombination. This addition not only increased
the available catalyst area, it also shortened the diffusion path for gas bubbles that had traveled to the top
of the chamber during electrolysis gas generation due to their buoyancy. The addition did not affect flow
delivery during actuation (data not shown).
45

2.2 ELECTROCHEMICAL BELLOWS ACTUATOR
In order to prevent unwanted electrochemical or chemical interactions (e.g. drug degradation and
unwanted pH changes), a separate pumping chamber is required to separate the electrolysis chamber from
the drug reservoir. To this end, a flexible Parylene bellows membrane was added to mechanically couple
the pumping chamber to the drug [14]. The bellows also serves to trap electrolysis gases, preventing their
delivery with the fluid. A bellows was selected over a corrugated or flat diaphragm for its greater achievable
deflections at lower applied pressures [41]. Moreover, additional convolutions are easily added to increase
achievable deflection without significantly increasing overall device size [43]. The bellows structure was
introduced in [14]. A new high yield process developed by Gensler, et. al., produced robust bellows
consisting of arbitrary numbers of convolutions which were capable of withstanding higher pressures and
cycling encountered in high flow rate, rapid bolus delivery [62].
2.2.1 Fabrication
Parylene bellows (1, 1.5, and 2 convolutions; 9 mm outer diameter, 6 mm inner diameter) were
fabricated as described in [63] and using a new process compared to [14] (Figure 2-13). Briefly, 0.4 mm
thick silicone rubber sheets (10:1 base-to-curing agent ratio Sylgard 184, Dow Corning, Midland, MI) were
punched with 9 and 6 mm holes, aligned, and stacked in alternating layers on glass slides as to form mold
modules. These molds were filled with molten (~50°C) polyethylene glycol (PEG; 1,000 Mn, Sigma
Aldrich, St. Louis, MO), taking care to ensure that no air bubbles were introduced to the molds. Once PEG
solidified at room temperature, silicone molds were peeled away to release the solid PEG modules. The
modules were stacked and fused together by slightly moistening interfaces with water. A 13.5 µm layer of
Parylene C (Specialty Coatings Systems, Indianapolis, IN) was deposited over the PEG mold, and PEG was
dissolved by soaking in water at room temperature to complete the bellows. The number of convolutions
were chosen based on previous deflection data [14, 63] and kept low to minimize the overall actuator height.
46

Figure 2-26: Photograph of: (a) Pt electrolysis electrode, (b) 2 convolution Parylene bellows, (c) assembled bellows
actuator. Scale bars represent 4.5 mm [45]).
To assemble the actuators (Figure 2-26), bellows were cut to size using a razor blade, filled with DD
water, and carefully combined with the Nafion
®
-coated interdigitated Pt electrodes using laser cut double-
sided pressure sensitive adhesive film (3M™ Double Coated Tape 415, 3M, St. Paul, MN). The seal was
reinforced with marine epoxy (Loctite, Westlake, OH).
2.2.2 Flow Characterization
Continuous Operation: Three different bellows configurations (1, 1.5, and 2 convolutions) were
fabricated. The flow rate performance of each bellows configuration under continuous operation was
assessed by clamping the actuator within a reservoir formed by an acrylic test fixture filled with double
distilled water. Constant current was supplied to the electrolysis electrodes for 2 min (Keithley 2400,
Keithley Instruments Inc., Cleveland, OH). Flow rate was calculated using the weight of accumulated
pumped water. Flow rate is linearly dependent on applied current [4].
47

Figure 2-27: Current controlled flow delivery with different bellows configurations performed in a custom test fixture
(reprinted with permission from [58] © 2011 IEEE).
Table 2-3: Flow rate values for 1, 1.5, and 2 convolutions bellows actuators (reprinted with permission from [58] ©
2011 IEEE).
Current [mA]
Flow Rate [μL/min]
1
Convolution
1.5 Convolution 2 Convolution
1 7.33 ± 0.17 9.88 ± 0.31 7.04 ± 0.29
2 17 ± 0.29 19.54 ± 0.54 17.33 ± 0.44
5 44.33 ± 2.42 49.05 ± 3.05 48.17 ± 0.33
8 75.5 ± 1.80 72.44 ± 5.35 73.66 ± 0.44
10 N/A 100.61 ± 1.11 87.33 ± 0.83

The addition of the drug separating bellows does not affect this relationship. As expected, flow rates
were comparable for bellows with different convolutions (Table 2-3; Figure 2-27) and not significantly
different as confirmed by statistical analysis of the obtained flow rates (ANOVA, p<0.05). The 1
convolution bellows, however, was expanded beyond its elastic operation range at 10 mA current and
irreversibly damaged. Actuator operating range is therefore limited by the mechanical performance of the
bellows selected. All subsequent experiments were performed with bellows having 2 convolutions to
maximize the fluid volume accessible by the actuator for electrolysis. Real-time measurements of
48

electrolysis chamber pressure were collected at different currents (1, 2, 5, 8, and 10 mA) using a pressure
sensor (ASDX015D44R, Honeywell International Inc., Morristown, NJ) connected directly to the outlet of
the test fixture. Sensor data was sampled using a data acquisition unit (LabVIEW 9.0.1 with NI cDAQ-
9174, National Instruments Corp., Austin, TX). Each current was applied for 90 s during which the pressure
was measured in real-time. Nafion
®
coating increased electrolysis efficiency leading to faster pressure
buildup and subsequently higher flow rates.

Figure 2-28: (a) Real-time pressure measurements of a 2 convolution Nafion
®
-coated bellows actuator under
constant current application; (b) Slope values for real-time pressure vs. time for different current values (reprinted
from [1] with permission from Springer).
Figure 2-28 shows the results for real-time pressure measurement of actuators having coated electrodes
for different applied current values. As expected from the flow characterization results, the slope was
linearly dependent on the applied current (R
2
= 0.9976). Slope can be used to estimate pressure generated
at a particular instant in time for a given applied current.
Bolus delivery: One potential application for an implantable drug delivery system is self-administration
drug addiction studies in small animals. In this paradigm, doses must be administered immediately (< 5 s)
and repeatedly in response to designated activities to allow the study of correlations between behavior and
drug addiction. Bolus delivery was assessed in three studies in which flow rates, bolus volumes, and number
of boluses were varied [1].
49

Bolus delivery and recombination were compared in uncoated and Nafion®-coated bellows actuators
for rapid-fire delivery of groups of 3 boluses at 10 mA (15 s ON, 10 s OFF) separated by 5 min OFF cycles.
Forty boluses were delivered at 10 mA (15 s ON, 60 s OFF). Seventy-five boluses were delivered at 10 mA
(2 s ON, 60 s OFF) using a bellows having a larger outer diameter (10 mm).

Figure 2-29: Comparison between rapid-fire bolus delivery using uncoated and coated 2 convolution bellows
actuators (3 boluses at 10 mA, 15 s ON/10 s OFF, separated by 5 min OFF cycles; reprinted with permission from
[58] © 2011 IEEE).

Figure 2-30: Bolus delivery with a Nafion® coated, 2 convolution bellows actuator (at 10 mA current, 15 s/1 min
ON/OFF; modified with permission from [58] © 2011 IEEE).
50

When comparing bolus delivery results between coated and uncoated electrodes, differences in both
delivery rate as well as recombination are readily observed: the overall pumping efficiency under identical
operation conditions is increased and recombination rate is highly dependent on gas to liquid ratio (Figure
2-29). An average flow rate of 78.98±1.68 µL/min (mean ± SE) was calculated for the 40 boluses delivered.
As evident in Figure 2-30, the transition from diffusion limited to reaction limited regimes of recombination
occurred after ~ 175 µL of accumulated volume was delivered and limited further delivery. 75 boluses
(3.81±0.05 µL (mean ± SE) per bolus) were delivered using the Parylene bellows with the larger outer
diameter (10 mm) (Figure 2-31). The aforementioned recombination regimes are dependent on the size of
the bellows and its remaining water content. The transition between the two regimes occurs when a
particular gas to liquid ratio threshold is reached. As expected based on previous studies detailed in [63],
the bellows having larger diameter achieved higher expansion, therefore, the change in regime would occur
at larger volumes that are not reached in Figure 2-31.

Figure 2-31: Bolus delivery with a Nafion® coated, 2 convolution bellows actuator (at 10 mA current, 2 s/1 min
ON/OFF), inset: close-up for first10 boluses delivered values (reprinted from [1] with permission from Springer).
2.2.3 Flow performance at body temperature
Broka et al. [49] showed that permeability of Nafion® to oxygen and hydrogen increased slightly when
ambient temperature was raised from room (25 °C) to body temperature (37 °C). Increased permeability of
51

Nafion® may lead to increased reaction rates and more effective pumping. Temperature effects on flow
performance were first evaluated using a Nafion®-coated electrode (without bellows) in an acrylic test
fixture placed in a constant temperature water bath. Flow rate was calculated by weighing pumped water
accumulated over 2 min of current application. Then, rapid-fire delivery of 10 boluses at 8 mA (5 s ON, 1
min OFF) at room (25 ± 0.2 ºC) and body temperatures (37 ± 0.2 ºC) were performed on a Nafion®-coated
actuator with a 2 convolution bellows. Accumulated volume was calculated by measuring water
displacement in 100 µL micropipette. Flow rate calculations are summarized in the Table 2-4 below (mean
± SE, n=3) for continuous pumping with Nafion®-coated electrode only actuators at room and body
temperature (without bellows).
Table 2-4: Flow rate values for current controlled delivery at room temperature vs. body temperature (reprinted from
[1] with permission from Springer).
Current [mA]
Flow Rate @ 25 ± 0.2 ºC
[µL/min]
Flow Rate @ 37± 0.2 ºC
[µL/min]
1 8.89 ± 0.68 11.56 ± 0.47
2 21.33 ± 0.61 21.47 ± 0.50
5 54.44 ± 1.84 53.7 ± 2.96
8 89.89 ± 0.78 94.44 ± 1.15

A slight increase in flow rate was apparent at 37 °C as expected. The slight increase in flow rate was
also observed for rapid-fire boluses delivered using actuators with a 2 convolution bellows. 10 boluses with
an average volume of 15.47± 0.13 µL and 16.33±0.18 µL per bolus (mean ± SE) were delivered at room
(25 °C) and body temperature (37 °C), respectively. The flow rate differences were significant as
determined by t-test for two independent samples with unequal variances (p<0.01).
2.2.4 Effects of applied back pressure on flow performance
During in vivo pumping, the actuator must overcome any back pressure associated with the local
physiological conditions near the catheter outlet. For instance if the drug is to be delivered in the thoracic
vena cava, the backpressure would be equal to the central venous pressure (CVP). The normal range in
humans was reported to be approximately ~ 3–8 mmHg (0.4 – 1.07 kPa) [64]. A range of back pressures
52

close to the CVP were applied in order to study their impact on bolus delivery. Four boluses were delivered
at 5 mA (60 s ON, 120 s OFF). The change in flow rate across this range of relevant physiological back
pressures was less than 5.5 % (Figure 2-32). One way analysis of variance of the data demonstrated that
changes in flow rate across the different back pressure were not significantly different compared to one
another (p<0.05).

Figure 2-32: Flow delivery results for a range of physiologically relevant back pressures (at 5 mA constant current).
Shaded area depicts ± 5% from the flow rate value for zero backpressure (modified from [1] with permission from
Springer).
2.2.5 Recombination
The bellows allow for large (up to 180 µL) volume displacement without undergoing plastic deformation
compared to flat or corrugated diaphragms. Recombination of 145 µL (80% of maximum allowable volume
chosen as a safety margin to prevent irreversible plastic deformation to the bellows during electrolysis
actuation) of electrolysis generated gas was observed for one hour and compared to results obtained for a
similar volume of gas generated and recombined in the 3 mL acrylic test fixture. When the Parylene bellows
were combined with Nafion®-coated electrodes, recombination followed a predictable non-linear diffusion
dependent pattern and 99.08 % recombination was achieved within 1 hour of delivery (data not shown).
53

This is considerably larger than the fraction of gas recovered in the larger volume using the same electrodes
in the same time period. As previously mentioned, slow diffusion of gas bubbles through the electrolyte
was a determining factor in the rate of recombination. By limiting the chamber volume, the diffusion path
was limited, leading to faster recombination.
Improving Recombination Efficiency: Several methods were explored to improve the recombination
rate in a bellows actuator [45]. In each case, the prepared actuator was clamped in a 3 mL acrylic test fixture
and filled with DD water. The outlet of the test fixture was attached to a 100 µL micropipette. Generated
gas volume and recombination was indirectly measured by observing the fluid front movement in the
micropipette. 2 mA current was applied to the actuator until 4 mm of movement was observed in the
micropipette (5.33 µL of gas generated). The current was turned off and recombination was measured based
on the fluid back flow in the micropipette. The fluid meniscus position was recorded periodically for 60
minutes or until the entire bolus had recombined, whichever came first.
Pt-coated bellows: The inside surfaces of the Parylene bellows were coated with 1800 Å layer of Pt using
e-beam evaporation. Three samples were prepared. Two of these were then further coated in a thin layer of
Nafion®, one of which was cured at 60 °C (1) and the other at room temperature (2). The bellows were
each filled with DD water and then combined with a Nafion®-coated electrode to form actuators (one
device was fabricated from each of the samples).
Pyrolyzing Nafion® on electrodes: As previously mentioned, Nafion® has a perfluorinated backbone and
short pendant chains terminated by sulfonic head group. If the polymer is heated to between 280-320 °C, a
small portion of the Nafion® pyrolyzes to lose the sulfonic acid groups. These portions are hydrophobic
and repel water, however they allow the passage of gases. Zhang et. al., explored this phenomena to increase
the efficiency of their proton exchange membrane (PEM) fuel cell in [65]. The group found that at
temperatures below 300°C, pyrolysis was insufficient; however, at temperatures above 340 °C, the
hydrophobicity inversely affected electrolysis efficiency. Based on this information, the Nafion®-coated
electrolysis electrodes were heat treated at 320 °C under N 2 backflow for an hour, then slowly cooled to
54

room temperature. To further improve the adhesion between the pyrolyzed Nafion® and the Pt electrode, a
thin layer of Nafion® was applied to the surface and cured at room temperature. The electrode was then
combined with a water filled bellows (2 devices were fabricated).
Pt wire pieces: An alternate placement for the Pt catalyst is within the electrolyte solution itself. To
accomplish this, two 3 mm segments of 99.9% Pt wire (Ø 0.5 mm) (California Fine Wire, Grover Beach,
CA) were selected. The edges of the wire were smoothed using 220 grit silicon carbide sandpaper. The
pieces were suspended in the DD water filled bellows, before the bellows were affixed to electrodes (2
devices were fabricated).
Pyrolyzed Nafion® coating on Pt wire pieces: Another two 3 mm segments of 99.9% Pt wire (Ø 0.5 mm)
(California Fine Wire, Grover Beach, CA) were selected. The edges of the wire were smoothed using 220
grit silicon carbide sandpaper. The pieces were then coated with Nafion® and heat treated at 320 °C under
N 2 backflow for an hour, then slowly cooled to room temperature. The pyrolyzed- Nafion®-coated pieces
were suspended in the DD water filled bellows, before the bellows were affixed to electrodes (3 devices
were fabricated).

Figure 2-33: Recombination time course for a 5.33 µL bolus delivered with actuators fabricated utilizing different
methods to accelerate recombination [45].

55

Table 2-5: Recombination time course expressed as % recombined for a 5.33 µL bolus was delivered with actuators
fabricated utilizing different methods to accelerate recombination [45].
Time
[min]
% Recombined
Unmodified
Pyrolyzed
Nafion®
Electrode
Pt Coated
Bellows
Pt Coated
Bellows
w/Nafion®
(1)
Pt Coated
Bellows
w/Nafion®
(2)
Uncoated
Suspended
Pt
Pyrolyzed
Suspended
Pt
15 37.52 100 85.93 75.04 59.48 62.48 89.63
25 45.86 - 96.93 96.93 81.3 67.54 100
30 50.09 - 100 100 84.43 75.05 -
35 58.35 - - - 93.81 77.49 -
60 62.54 - - - - 95.12 -

The recombination performance over time for the different methods explored are presented in Figure 2-
33 and detailed in Table 2-5. The lowest rate of recombination was observed for unmodified actuators with
Nafion®-coated electrodes, while the highest rate was observed for actuators with pyrolyzed Nafion®-
coated electrodes. Recombination rate was seen to increase for bellows coated internally with thin films of
Pt and Nafion®-coated Pt. It is important to note that the recombination rate was faster for the actuator with
Nafion®-coated Pt-coated bellows (1), for which the Nafion® coating was cured at 60 °C compared to the
other cured at room temperature (2). This is likely due to the slight increase in surface hydrophobicity as a
result of heat application to the Nafion®. Suspended Pt wire pieces (coated and uncoated) produced faster
rated of recombination than unmodified actuators, but noticeably lower rates than actuators with pyrolyzed
Nafion®-coated electrodes.
For Pt-coated bellows, in all cases, the Pt delaminated from the surface of the Parylene over time likely
due to the exposure to the generated bubbles and the stress caused by the inflation of the bellows structure
during actuation. Delamination may have been facilitated by the presence of exposed metal interfaces that
arise as the deposition technique is unable to completely coat the convolutions in the bellows. This
delamination did not affect electrolysis or recombination during testing. However, given the unreliable
adhesion between the Parylene and Pt, this method cannot be used to maintain a constant catalyst surface
area. The electrodes with the pyrolyzed Nafion® coating showed the fastest recombination rate (2 devices
56

fabricated). However, despite the additional overcoat layer of Nafion®, the pyrolyzed layer delaminated
from the surface of the electrode during electrolysis and therefore repeatable results (n > 4) could not be
obtained using this method. The actuators with the pyrolyzed Nafion®-coated pieces, however, seemed to
show consistent performance (3 devices fabricated). The catalyst segments in the water did not affect
electrolysis gas generation. The modification methods were intended to accelerate recombination by either
increasing the catalyst surface area, increasing the catalyst surface hydrophobicity, or increasing the rate of
gas transport. Adding pyrolyzed-Nafion®-coated Pt segments in the actuator achieves all three functions
by providing more catalyst surface area, utilizing a hydrophobic coating attractive to gas bubbles, and
situating catalyst throughout the electrolysis chamber.
Recombination Orientation Dependency: For the above experiments, gas recombination was
characterized with the electrode clamped in the test fixture and placed facing up on the benchtop such that
the generated gases rose up away from the electrode. However, for some applications, it is desired and
sometimes inevitable for the actuator, and therefore electrodes, to assume other orientations. For instance,
the electrolysis actuator may be used in an implanted pump which may be placed at a different orientation
or change in position within the subject. Therefore, recombination dependence on actuator orientation was
also tested with unmodified actuators with Nafion®-coated electrode catalyst (henceforth referred to as
unmodified), as well as actuators with additional pyrolyzed Nafion® coating on suspended Pt wire pieces.
Several current and time combinations were applied to the actuator to achieve different generated gas
volumes of 0.277, 1.11, 5.33, 10, 25.33, and 50 µL. The angle of an actuator clamped in a test fixture was
varied for each volume (Figure 2-34; 0°, actuator facing up on the benchtop, 90°, actuator perpendicular to
benchtop, and 180°, actuator facing down on the benchtop). Recombination was monitored until the entire
delivered volume was recombined or for an hour, whichever came first.
57

Figure 2-34: The angle of an actuator clamped in a test fixture was varied for each volume (0°, actuator facing up on
the benchtop, 90°, actuator perpendicular to benchtop, and 180°, actuator facing down on the benchtop) [45].
Contrary to the results obtained for actuation [4], there is a clear distinction, between recombination
speed and the angle of the actuation chamber for a given generated gas volume for unmodified actuators
with Nafion®-coated electrode catalyst. Figure 2-35 displays fraction of recombined gas for devices with
unmodified and suspended pyrolyzed-Nafion®-coated Pt pieces at each tested orientation (0°, 90° and
180°). The highest fraction of recombined gas volume was observed for both electrode designs in the 180°
orientation. For both actuator designs greater recombination was observed for 90° than 0° orientations,
however the performance difference between the two orientations was comparatively small. The actuator
with suspended pyrolyzed-Nafion®-coated Pt pieces produced greater recombination than the unmodified
design, and the variation in performance as a function of orientation was considerably reduced. On average,
the rate of recombination was measured to be 2.3 times faster across all actuator orientations (n=3 for each
dose volume at each orientation). These observations can be explained by the behavior of the electrolyzed
gas under the influence of the buoyant force. Gas bubbles created during electrolysis tend to travel upwards
within the test fixture; after the current application is discontinued, the bubbles need to reach the surface of
the Pt catalyst to recombine; the reaction is limited by the slow diffusion of gases through water. In the 90°
and 180° orientations, the bubbles are in contact with the catalyst as soon as the current application is halted.
As a result, recombination rate is increased in these orientations. By the same reasoning, the 180°
58

orientation is faster than 90° due a larger catalyst surface area available to assist recombination. The effects
of orientation were reduced for actuators with additional pyrolyzed-Nafion® coating on Pt wire pieces.

Figure 2-35: % Recombination vs. angular orientation for different delivered generated gas volumes for unmodified
actuators with Nafion®-coated electrode catalyst and actuators with additional pyrolyzed Nafion® coating on
suspended Pt wire pieces [45].
Recombination Repeated Cycling: For a microactuator, speed of recombination determines how
quickly the generated gases can be reset to an all-liquid state. As such, repeatable and reliable microactuator
operation is dependent on repeatable complete recombination. Incomplete recombination could lead to
lower and inconsistent delivery actuation flow rates as the bellows structures approaches maximum
deflection. Further inflation as a result of electrolysis could result in permanent deformation of the bellows
via plastic deformation [62]. For instance, Figure 2-36 demonstrates an example where three 10 µL gas
boluses are generated and allowed to recombine for 30 minutes between each bolus generation. Then,
following a waiting period for further recombination to partially reset the bellows, four 50 µL gas boluses
are generated and allowed to recombine for 60 minutes between each bolus generation (data based on
electrolysis and recombination observed for an unmodified actuator). If the wait time between the first set
of boluses and the initiation of the second set is not sufficient for full recombination, the fourth 50 µL gas
59

bolus generated would cause the bellows to exceed maximum deflection (> 180 µL delivered). The wait
time is dependent on the speed of recombination, as well as repeatability of recombination for a specific
bolus volume. Therefore, it is important to study the repeatability of recombination of generated gas
volumes through repeated cycling for applications such as implantable pumps in which reliable and accurate
performance of the actuator driving drug dosing is critical.

Figure 2-36: Hypothetical example: if the wait time (denoted with dashed vertical lines) between the first and second
set of generated gas boluses (10 and 50 µL, respectively) is not sufficient for complete recombination, the bellows
structure will exceed maximum inflation when the fourth 50 µL bolus is generated) [45].
An actuator was assembled having suspended Pt wire segments with pyrolyzed Nafion® coating and
then clamped in a 3 mL acrylic test fixture. The reservoir was filled with DD water and a 100 µL
micropipette was attached to the outlet of the test fixture. Generated gas volume and recombination were
indirectly measured by observing the movement of the fluid meniscus along the micropipette. 2 mA current
was applied to the actuator until 7.5 mm of movement was observed in the micropipette (corresponding to
10 µL of gas generated). The current was turned off halting electrolysis and recombination was measured
based on the fluid back flow in the micropipette. The fluid meniscus position was recorded periodically
until the entire bolus had recombined, at which time another 10 µL of gas generated and the recombination
60

monitored. After 24 hours, three 5 µL, followed by three 10 µL boluses of gas were generated. Full
recombination of each generated bolus was monitored and recorded.
Figure 2-37 shows the results of repeated cycling (actuation and recombination) for 10 and 5 µL boluses.
Repeatable complete recombination was achieved for consecutive gas bolus volumes generated as well as
boluses delivered after 24 hours idle time. Figure 11 illustrates actuation and recombination results obtained
for different generated gas volumes of 0.277, 1.11, 5.33, 10, 25.33, and 50 µL for an actuator facing up on
the benchtop (n=3, obtained for orientation dependency testing). The average time required for complete
recombination of each of the generated gas volumes is presented in Table 2-6. An exponential trend line
was fit to the data for 25.33 and 50 µL volumes to estimate the time required for complete recombination
of the generated gas volumes. As demonstrated previously in the orientation dependency section,
recombination is slowest when the actuator is placed facing up on the benchtop. Therefore, these
recombination profile results obtained for different generated gas volumes can be used to estimate the rate
of recombination and the maximum allotted wait time required for complete recombination of generated
gases. It is important to note that changing the orientation of the actuator, as well as generation of large gas
boluses (>25 µL) leading to significant inflation of the bellows, could cause the repositioning of the
pyrolyzed-Nafion® coated Pt wire segments, which may slightly alter the recombination time profile. These
variations are represented by the error bars in Figures 2-35 and 2-38.

Figure 2-37: (a) Repeated cycling (actuation and recombination), (b) averaged trends for actuation and recombination
of for 10 and 5 µL boluses [45].
61

Figure 2-38: (a) Actuation and recombination results obtained for different generated gas volumes of 0.277, 1.11,
5.33, 10, 25.33, and 50 µL for an actuator facing up on the benchtop. (b) Data for smaller bolus volumes [45].
Table 2-6: Average time required for complete recombination of different generated gas volumes [45].
Time Required
for Complete
Recombination
[min]
Generated Gas Volume [µL]
0.277 1.11 5.33 10 25.33 50
12.75 20 15 36 300 450

Salt Permeation through Parylene Bellows: As previously mentioned, metallic alkali metals
(particularly sodium) can attack Nafion® directly under normal conditions of temperature and pressure
[29]. Damage to the Nafion coating layer of the electrolysis electrode could lead to lower electrolysis
efficiency and reduced delivery reliability and repeatability. As such, besides choosing DD water as the
electrolyte, other sources of sodium that could potentially reach the Nafion® need to be identified and
mitigated. For instance, it is important to use borosilicate glass as the electrode substrate instead of soda
lime glass. The structure of soda lime glass includes a substantial amount of sodium ions. When in contact
with aqueous solutions, the structure undergoes dealkalization and the ions leach into the surrounding
solution [66].
Sodium ions could also diffuse through the Parylene bellows into the electrolysis chamber. This is
especially important as a saline is often used as a solvent in drug therapy. EIS was used to estimate the
amount of time required for ions to diffuse from an outer reservoir filled with 1× PBS through the bellows
62

membrane (13.5 µm thick) and into the electrolysis chamber filled with DD water (data not shown). The
results showed that within several hours a small number of ions were present in the electrolysis chamber.
In order to reduce the rate of diffusion several methods of altering the bellows were investigated. 3 new
bellows were fabricated (one device each): double-Parylene, for which a second layer of Parylene was
deposited (15 µm thick); PEEK-capped, for which a circular PEEK sheet (Ø 8 mm, 66 µm thick) was placed
on top of the topmost bellows convolution prior to the deposition of a second layer of Parylene (15 µm
thick); metal-capped, for which a 2000 Å layer of Pt was evaporated onto the topmost bellows convolution
prior to the deposition of a second layer of Parylene (15 µm thick). The bellows were filled with DD water
combined with electrodes to create electrolysis actuators. The actuators were then integrated into a reservoir
filled with 1× PBS. EIS was performed daily and the Bode plot area representative of solution resistance
was monitored for the presence of ions (data not shown).
The results showed that for the double-Parylene and PEEK-capped bellows, ions were present in the
electrolysis actuator within 24 hours. However, no change in solution resistance was observed for the metal-
capped bellows actuator within 17 days. The observed difference in ion diffusion is due to the different
nature of diffusion through metals vs. polymers. In metals, the solid lattice contains periodically distributed
atoms. Interstitial positions in the lattice are positions of minimum energy. A diffusing atom requires
activation energy to pass from one position of minimum energy to another. In polymers, however,
molecules form string-types held together by van der Waals forces. An atom requires considerably less
energy diffuse through the strings [67]. These preliminary results are promising and could be further
investigated to mitigate ion diffusion from the drug reservoir into the electrolysis chamber.
2.3 ELECTROCHEMICALLY ACTUATED DRUG DELIVERY MICROPUMP
2.3.1 Fabrication
Parts for a cylindrical drug reservoir (22 mm diameter) were injection molded from polypropylene
(Chase Plastics, Clarkston, MI). A refill port was fashioned from polydimethylsiloxane (MDX-4, Factor II,
63

Lakeside, AZ) in the reservoir cap. The bellows actuator was affixed to the reservoir base. Parts were then
assembled and joints reinforced with marine epoxy (Figure 2-39).

Figure 2-39: Drug delivery Micropump prototype with a rigid cylindrical poly propylene reservoir.
2.3.2 Pumping of Viscous Drug Models
The effects of viscosity on flow performance were investigated using four different model drug solutions
(propylene glycol, diluted ISOVUE 370, cocaine, and Lidocaine). Propylene glycol is a colorless and
odorless alcohol compound that has been used as a solvent and a drug carrier in pharmaceutical products
[68-70]. ISOVUE 370 is a contrast agent containing iodine used in radiological examinations. Cocaine is a
suitable agent for studying drug addiction in a self-administration paradigm in small animals. Lidocaine is
a common local anesthetic and antiarrhythmic drug. For propylene glycol, several concentrations were
prepared by diluting stock with DD water. ISOVUE 370 (Bracco Diagnostic Inc., Princeton, NJ) was diluted
1:1 with DD water. Cocaine was dissolved in 0.9 N saline. Lidocaine HCl hydrate (Enzo Life Sciences,
Farmingdale, NY) was dissolved in 1× PBS (20 mg/mL). The viscosity of the first two solutions (e.g.
Propylene Glycol and ISOVUE 370 dye) was measured using a Cannon-Fenske routine viscometer
(Technical Glass Products, Painesville Twp., OH) immersed in a water bath at 37 °C. Model drug solutions
were loaded into a polypropylene reservoir containing a Nafion®-coated, 2 convolution bellows actuator.
Multiple boluses were delivered (propylene glycol: 10 boluses at 8 mA (10 s ON, 60 s OFF); ISOVUE 370:
3 boluses at 3 mA (90 s ON, 30 min OFF); cocaine 3 boluses at 8mA (60 s ON, 15 min OFF); Lidocaine:
64

1 mA current for 1 min, immediately followed by 10 mA for 10 s, 1 mA for 1 min, and lastly 5 mA for 30
s (n=5)) and the flow rate was calculated from measuring fluid displacement in a 100 µL calibrated
micropipette or by weighing the accumulated dispensed fluid.

Figure 2-40: Flow delivery results for different viscosities of propylene glycol (at 8 mA constant current; modified from
[1] with permission from Springer).
Despite the large range of viscosities, change in propylene glycol solution flow rate was less than 2.4 %
(Figure 2-40). One way analysis of variance of the data showed that the changes in flow rate for different
viscosities were not significantly different compared to one another (p<0.05). Similarly for the ISOVUE
370 dye, an average flow rate of 30.00 ± 0.0 µL/min was measured compared to 29.33 ± 0.0 µL/min for
water (mean ± SE). The dye solution was not viscous enough to impede pumping. However, the effects of
drug concentration (effectively viscosity) were more apparent in high concentrations of cocaine and flow
rate decreased 21 % for 25.73 mg/mL cocaine solution (viscosity could not be measured due to limited
sample, Figure 2-41). No significant difference was observed between the flow values for Lidocaine vs.
DD water or sterile PBS (measured flow rates of 11.6 ± 0.46, 135.9 ± 6.0, and 50.4 ± 1.98 µL/min, no
compensation for viscosity; t-test (p<0.05)).
65

Figure 2-41: Current controlled flow delivery of cocaine (in 0.9 N saline) loaded in a pump at different concentrations
(at 8 mA, 1.5 convolution bellows; modified with permission from [58] © 2011 IEEE).

2.3.3 Real-time Flow Response to Changes in Regimen
Pumping accuracy was evaluated under real-time adjustment of drug regimen using a series of operating
currents (10, 3, 10, and then 8 mA) in succession. Each current was supplied for 30 s using a source meter
(Keithley 2400, Keithley Instruments Inc., Cleveland, OH). Flow rate was calculated using the weight of
accumulated pumped water. The actuator responded to these “on the fly” changes in current with <1 sec
response time (Figure 2-42) [71]. Flow rate values calculated were comparable to those previously
measured for each current individually.
66

Figure 2-42: Actuator response to “on the fly” changes in current: accumulated volume, and flow rate (n =3, Mean ±
SE).
2.3.4 Valved System
While recombination is essential to reliable actuation, it also results in a reverse pressure gradient. This
pressure gradient could cause fluid to flow back through the delivery catheter to the reservoir. In order to
achieve uni-directional flow, backflow into the pump was prevented by connecting a commercial check
valve between the reservoir and the outlet catheter. A Qosina 80031 in-line check-valve with < 0.35 kPa
(0.05 psi) cracking pressure, acrylic and ethylene propylene diene monomer construction, and a 3.1–3.43
mm outer diameter (Qosina, Edgewood, NY) was selected. 15 boluses were delivered at 10 mA (15 s ON,
60 s OFF). Accumulated volume was calculated by measuring water displacement in a 100 µL calibrated
micropipette. An average bolus volume of 26.22±0.18 µL per bolus was calculated (Figure 2-43, mean ±
SE). The cracking pressure of the commercial check valve was reported by the manufacturer to be 0.345
kPa (0.05 psi, 2.6 mmHg). This cracking pressure was adequately exceeded by the actuator. The valve did
not increase response time or impede flow compared to when the actuator was operated without a valve.
These results correspond to the aforementioned real-time pressure measurements for the bellows actuator.
67

Figure 2-43: Bolus delivery using an actuator with an in-line check valve (at 10 mA current, 15 sec ON/ 60 sec OFF).
After 15 boluses, the actuator is turned off. Insets show photograph of valve, and average accumulated volume for
15 boluses (modified from [1] with permission from Springer).
However, as reported in [43], this valve is not normally-closed as advertised and requires a finite back
pressure for complete sealing. This sealing pressure (measured to be greater than 1.03 kPa) against
backward flow was not reached under this condition, leading to significant leakage after the delivery
session. It is important to note, that backflow can be prevented with a true normally closed check valve.
Designing the valve with the specifications required is beyond the scope of this work.
2.3.5 Power Consumption
Theoretically, for continuous operation at a single current, the power consumption of the electrolysis
actuator, P, depends on the voltage necessary for the electrolysis, Ecell, and on the supplied current, I (2-10)
[18]:
#= \
SP$$
×] (2-10)
The power consumption for uncoated and Nafion® coated electrodes, uncoated electrodes integrated
with 2 convolution bellows, Nafion® coated electrodes integrated with 1, 1.5, and 2 convolution bellows,
and Nafion® coated electrodes with 2 convolution bellows integrated in a cylindrical polypropylene 1 mL
reservoir was calculated (Table 2-7) using the cell resistance (R cell) measured by the source meter during
constant current application (2-11):
68

#= \
SP$$
×]=
SP$$
×]

(2-11)
Table 2-7: Power consumption data for uncoated and coated electrodes, actuators, and micropumps.
Power [mW]
Uncoated Electrode Nafion
®
Coated Electrode
Current
[mA]
Electrode
Only
+ 2
Convolution
Bellows
Electrode
Only
+ 1
Convolution
Bellows
+ 1.5
Convolution
Bellows
+ 2
Convolution
Bellows
+2 Convolution
Bellows
+ Rigid
Reservoir
0.25 - - - - - 0.6625 -
0.5 1.51 - - - - 1.25625 -
0.75 - - 1.633625 - - 2.053125 2.205
1 3.32 - 2.53 2.2 2.39 2.65 3.1975
1.5 5.32 - - - - - -
2 7.66 8.18 5.643 4.8 5.57 5.87 7.09
3 - - - - - - 10.59
5 - 79.25 16.4825 18.05 16.67 17.54 20.46
8 - 172 28.3842 58.24 31.12 34.3 34.999
10 - 180 37.205 125 41.03 54.57 44.525
13 - 286.455 51.3084 - - - 60.0795

For uncoated electrodes, the slow diffusion of oxygen through water results in the formation of an “O2
film” on the electrode surface [20]. This film impedes further electrolysis and decreases pumping efficiency
by preventing liquid-metal contact [21]. As a result higher power consumption is observed compared to
Nafion® coated electrodes, in which gases quickly diffuse away from the surface of the catalyst electrode.
As expected, adding the bellows or packaging the actuator in a rigid reservoir does not significantly affect
power consumption (ANOVA, p<0.05). Except for the 1 convolution bellows actuator under supply
currents > 8 mA; for which, the gas/water ratio contained in the bellows quickly rises, leading to a rise in

SP$$
(as a result of insufficient water supply) and a significant increase in power consumption.
Using the power data, the energy input requirements of the actuator could also be calculated using (2-
12):
^=_ ×` (2-12)
69

Where, J is energy input in Joules, W is power in Watts and s time in seconds. For instance, delivering a 10
µL bolus using a Nafion® coated actuator with a 2 convolution bellows packaged in a rigid polypropylene
reservoir with 2 mA applied current, requires 40.6 s and the energy input would be 0.288 J.
2.4 WIRELESS POWERING & FLOW CONTROL
2.4.1 Inductive Powering at a Single Infusion Rate
For the first prototype, a simple Class D amplifier wireless powering system was developed [1]. Class
D systems can provide the large current needed to create the magnetic field, however, series-tuning of the
primary coil cancels inductance leakage through the coil lowering the required driving voltage. They are
also less dependent on coupling as long as the load stays well above the switch-on resistance [72]. These
features make Class D systems suitable for simple first stage prototyping. The design was implemented
using discrete components and the effects of distance between coils and foveation between the transmitter
and receiver where studied. Flow characterization was performed with concentric placement of the
transmitter (Ø 55 mm) and receiver (Ø 23 mm) coils and in the same plane (0º foveation between coils).
The results were detailed in [1] and are summarized in Table 2-8. Twenty boluses were also delivered at
6.3 mA (5 s ON, 60 s OFF) using an actuator housed in a polypropylene reservoir with 1 mL volume. The
average flow rate calculated for the 20 boluses delivered was 68.4 ± 1.23 µL/min (mean ± SE).
Table 2-8: Flow rate values for current supplied using the Class D wireless powering system (reprinted from [1] with
permission from Springer).
Current [mA] Flow Rate with Wireless System [µL/min]
2 20.00 ± 0.14
5 59.00 ± 1.11
8 98.67± 0.41
10 125.67 ± 1.44

70

While simple in design, this Class D system using discrete components had several drawbacks. The
metal-oxide-semiconductor field-effect transistor (MOSFET) driver chip had large process variations
making it difficult to fine tune the amplifier. MOSFET switching and the open loop configuration caused
the MOSFET driver to heat up and sometimes burn out.
A second receiver device was designed with a low drop-out (LDO) switching regulator
(LTC3670EDDB#PBF, Linear Technology, Woodland Hills, CA) with a faster response time and higher
linear efficiency compared to the current regulator previously used (Figure 2-44). In this design, a half-
wave rectifier was implemented to reduce the noise level and the number of components needed. A Zener
diode (MM3Z4V3CCT-ND, Fairchild Semiconductor, San Jose, CA) was attached in parallel with the
circuit to limit the input voltage (4.3V) to the regulator.

Figure 2-44: Schematic diagram of Second receiver device with a low drop-out (LDO) switching regulator (reprinted
from [40] with permission from Elsevier).
This design was first simulated using LTSpice, before being implemented using surface mount
components on a flexible PCB on a scale that could be integrated to an implantable drug delivery device.
Benchtop testing was performed using the transmitter described in [73] originally designed for a microbolus
infusion pump used for functional neuroimaging applications in rodents. This Class E amplifier circuit
includes a handheld transmitting coil (A = 47 cm
2
). Flow rates were comparable to that achieved with
Keithley 2400 source meter (Cleveland, OH). The results of changing distance between coils and foveation
between the transmitter and receiver were once again obtained (Figure 2-45).
71

Figure 2-45: The effects of (a) distance between coils and (b) foveation between the transmitter and receiver on
voltage output (reprinted from [40] with permission from Elsevier).
The device was then packaged for in vivo testing: electronics were encapsulated using epoxy and
attached to the drug delivery device. The completely packaged system was covered in medical grade
silicone rubber (MDX-4, Factor II Inc., Lakeside, AZ) before being coated by a 9 µm thick conformal layer
of Parylene C to enhance biocompatibility and reduce moisture permeability. A 30 day trial with the
packaged device immersed in a 37 ºC water bath, showed continued accurate and reliable dosing with <
3.7% standard error.
It is important to note that due to the nature of electrolysis and the linear relationship between current
and flow rate, a receiver with a constant current output would be more suitable for this application.
Therefore, a third receiver was designed. In this design a full-wave rectifier was implemented using two
Schottky diodes (BAT54A and BAT54C, Fairchild Semiconductor, San Jose, CA). A resistor (Rset) was
used to set the output current (I set) on 3-terminal adjustable current source (LM334, Texas Instruments,
Dallas, TX) using the following equation (2-13):
a
PN
=
,
b
6
/F9
(1.059) (2-13)
Where V R is the voltage across the resistor. A second class E amplifier transmitter circuit was also
developed. This system included a rectangular transmitting coil (A = 127 cm
2
) designed to be placed
underneath a rodent cage. The results of changing distance between coils and foveation between the
72

transmitter and receiver on flow rate output of the micropump were obtained. A 25% and 33% drop in flow
rate was observed for 7 cm distance and 60° foveation respectively. Delivery results for micropumps
outfitted with this wireless powering system are extensively detailed in [74].
Link Efficiency Calculations: Link efficiency, a key factor in achieving maximum power transfer, was
calculated for the above inductive powering system using equations provided in [34]. It is important to note
that following popular methods of link efficiency calculation, only the transmitter and receiver coils and
capacitors (Table 2-9) are considered in calculations rather than including the powering system as a whole,
following the assumption that the driver, rectifier and regulator losses are small compared to the link loss
[75].
Table 2-9: Characteristics of the transmitting and receiving coils when tuned to 2 MHz.

Area
[mm
2
]
At 2 MHz
L
[µH]
C
[pF]
R
[kΩ]
Q

V (pk-pk)
[V]
Transmitting
Coil
8 turns, 20
AWG
single
strand wire
127100 31 180 50 700 112
Receiving
Coil
6 turns,
50/54 Litz
wires
1520 1300 4800 1.62 99.8 15.2

Link gain: d ≡
e,
b
f
e
e,
f
g
e
= 0.136 (2-14)
Coupling coefficient: k ~ 0.18 (2-15)
Link potential: X = h

X
i
X

=2263.46 (2-16)
Efficiency figure of merit: k=hlX
i
X

(2-17)
Optimal link efficiency: E
QmN
=
n

(i
li
n

)

=0.95 (2-18)
73

Total link efficiency: E=
o
b

b
f
(i

b

b
f

o
b

b
f
)(i

b

b
f
)
(2-19)
Where RL is the load resistance.
Link efficacy was calculated using electrolysis cell resistance (R cell) values as R L (Figure 2-46):

Figure 2-46: Link efficiency calculations using electrolysis cell resistance (Rcell) values as RL.
2.4.2 Alleviating Coil Misalignment Effects on Power Transfer
A previously mentioned, mutual inductance (and effectively power transfer) between the transmitter and
receiver coils heavily depends on their relative position, as the magnetic coupling can actually become zero
when the coils are perpendicular to each other. Multiple coils at either the transmitter or receiver side are
necessary to eliminate zero coupling [34]. The power pick-up from multiple receiver coils can be directly
combined and this method does not require feedback. Therefore it was chosen and implemented. Two
identical coils (Ø 23 mm) were fabricated using multi-wound Litz wire (50 strands of 54 g wire). The coils
were placed perpendicular to each other (on coil inside the other) and tuned. One full-wave rectifier
(BAT54A and BAT54C, Fairchild Semiconductor, San Jose, CA) was used. The voltage across the
smoothing capacitor was monitored and compared to results from a single receiver (Table 2-10).
74

Table 2-10: Percent drop in voltage output as a result of coil misalignment for one and two receivers.

100% 63% 24%

100% 97% 93%
2.4.3 Remote Bluetooth Controllability
A Parallax Board of Education USB carrier board kit, a Basic Stamp 2 Module microcontroller, and an
Easy Bluetooth Module were purchased (Parallax Inc., Rocklin, CA). The output of the Bluetooth module
was used to control a single pole OptoMos relay (LCA 717, Clare, Inc., Beverly, MA), controlling the
power to the transmitter circuit (Figure 2-47). Basic Stamp 2 Editor (v. 2.5.2) was used to program the
communication. The program created allowed for the transmitter circuit to be controlled wirelessly via
Bluetooth.

Figure 2-47: Wireless powering system with Bluetooth circuitry for wireless control: (a) Parallax Board of Education,
Basic Stamp 2 Module microcontroller, and an Easy Bluetooth Module; (b) Transmitter circuit; (c) packaged
micropump and receiver PCB (reprinted from [40] with permission from Elsevier).
75

2.4.4 Wireless Flow Control
In order to achieve true controllability and be able to accurately tailor therapy to the patient throughout
the course of treatment, the infusion rate should be adjustable post-implantation. For a single infusion rate,
the power received is rectified and applied to a current regulator to provide a single constant current to the
actuator. Variable current (and therefore variable infusion rates) could be achieved by changing the wiper
position of a current setting programmable digital potentiometer in the receiver circuit to alter the output
current. The potentiometer should be chosen to allow for the selection of the full range of desired currents
(0.6 -3.2 mA) with a resolution of 0.2 mA per step (this range was as it allows for a wide range of infusion
rates suitable for drug delivery applications). The wiper position could be changed using a data signal
wirelessly transferred through the skin to the receiver. Two methods of data transmission were considered
for this purpose: infrared and amplitude shift keying (ASK). ASK was chosen due to its simplicity, lower
number of additional components on the transmitter and receiver circuits, and lower susceptibility to
environmental noise. ASK data transmission has been previously used in biomedical applications, for
example for monitoring implanted artificial hearts [76] and for electroneurogram (ENG) signal recording
from the sciatic nerve [77]. Using this method, a low frequency data signal is carried by the power signal
and subsequently demodulated on the receiver. Pulse width of the data signal determines the number of
incremental steps of potentiometer wiper position, altering the potentiometer resistance, and changing the
output current set by the potentiometer. The transmitter output is controlled by aforementioned Easy
Bluetooth module and Basic Stamp microcontroller kit and a custom user interface program using the Basic
Stamp editor. The overall system architecture is presented in Figure 2-48.
76

Figure 2-48: System architecture of ASK modulation circuit to enable wireless flow control (reprinted with permission
from [78] © 2014 IEEE).
Similar to the third generation receiver circuit, Litz wire (6 turns, 50/54 SPN/SN Litz Wire, Wiretron,
Volcano, CA) was used for the receiving coil (Ø 22 mm). The circuit components (including the current
regulator) require a direct current (DC) power signal. Therefore, the received alternating signal was fully
rectified using two Schottky diodes (BAT54A and BAT54C, Fairchild Semiconductor, San Jose, CA). The
modulation signal was separated by half-wave rectification (BAT54A) and then filtered using a low pass
RC design. Two zener voltage regulator diodes (BZV55, NXP Semiconductors, Eindhoven, Netherlands)
regulate the output voltage of the power and demodulated signal. AD5227 (Analog Devices, Norwood,
MA) was used as the current setting potentiometer to set the output current of the current regulator. The
LM334 adjustable current source was replaced with PSSI2021SAY (NXP Semiconductors, Eindhoven,
Netherlands) to allow for the desired current range to be achieved using the potentiometer. A low frequency
oscillator (LTC 6991, Linear Technology, Milpitas, CA) set the potentiometer frequency to 500 Hz. An
LED was added in series with the infusion pump to provide visual confirmation of power supply to the
actuator (Figure 2-49). The transmitter and receiver circuits were first tested using discrete components on
breadboards. Output flow was modified four times by changing the wiper position of the current setting
potentiometer through the Bluetooth interface (Figure 2-50).
77

Figure 2-49: Schematic diagram of the receiver circuit (reprinted with permission from [78] © 2014 IEEE).

Figure 2-50: Wireless flow variation using the ASK modulation circuit. Output current levels are estimated based on
the results obtained for a 1 kΩ load.

The receiver was then implemented on a flexible printed circuit board (PCB) to allow it to be wrapped
around the round micropump. Small surface mount components were chosen to minimize the overall circuit
footprint (Figure 2-51).
78

Figure 2-51: Micropump with integrated receiver. Inset: Receiver flexible PCB, front and back (reprinted with
permission from [78] © 2014 IEEE).
Preliminary benchtop testing using a “dummy” 1 kΩ load in place of an actuator was performed to
ensure feasibility and repeatability of the approach. The circuit performance was then tested with a
micropump on benchtop. In order to study the electric and magnetic effects of tissue on the wireless
transmission, testing across simulated brain tissue material [79] was performed [78].
Testing with “Dummy” Load: The receiver current output was increased incrementally (10 steps) and
the values achieved were compared to theoretical values calculated based on equations provided in the
potentiometer (2-20) and current regulator datasheets (2-22):
Δ =q# ×
r st
uU
+70 Ω (2-20)
where CP is the number of clock pulses.

PxN
=(5 hΩ− Δ )+165 Ω ∥1.62 Ω (2-21)
a
Q'N
=
{.ui|
6
FG9
+15 }d (2-22)
The output current was calculated by measuring the voltage across a 1 kΩ load. Repeatability of the
output current for a specific modulation program was also tested by varying the current in 4 steps: 0.6, 2.6,
0.7, and 2 mA applied successively for each run (n=5).
79

Figure 2-52: 10 step incremental increase in receiver current output measured across a 1 kΩ resistor (at each
modification step, a specific pulse width was applied to achieve the desired change in the wiper position; reprinted
with permission from [78] © 2014 IEEE).
The results for the incremental increase in receiver current output are shown in Figure 2-52. The
measured value closely followed the expected value for all the modification steps. Repeatability in achieved
current output for the same modulation program was also confirmed Figure 2-53.

Figure 2-53: Current output repeatability for a specific modulation program (reprinted with permission from [78] ©
2014 IEEE).
Testing with Infusion Micropump: An infusion pump with a 1 mL reservoir was connected to the
receiver circuit Figure 2-51. Two different regimens were selected to demonstrate wireless operation at
multiple flow rates. First, the current output was increased in 0.5 mA steps from 0.6 - 2.1 mA. For the
80

second regimen, the current was increased from 0.6 to 1.2 mA, then decreased to 0.6 mA and increased to
2.2 mA. For both experiments, double distilled (DD) water was used as the drug model, and the infusion
flow rate was calculated by measuring the fluid front movement in a calibrated 100 µL micropipette
connected to the outlet of the pump. Figure 2-54 shows the variable current results with the infusion pump.
The measured flow rate closely followed the changes in current initiated using the interface program.
Measured flow rate values were comparable to those achieved for wired actuation of the infusion pump [1].

Figure 2-54: Wireless flow variation in (a) incremental steps, and (b) random fashion (reprinted with permission from
[78] © 2014 IEEE).
Testing in Simulated Brain Tissue Material: In order to mimic the effects of wireless transmission
through tissue, brain tissue was simulated according to the recipe described in [79]. Briefly, sugar, NaCl
salt, and Natrosol® (hydroxyethylcellulose, Ashland Inc., Covington, KY) were dissolved in 800 mL of
DD water (1108.9, 49.5, and 19.8 g, respectively). 1.98 g of ProClin® 950 preservative (Sigma-Aldrich,
St. Louis, MO) was added as a replacement for Dowicil 75® (1-(3-chloroallyl)-3, 5, 7-
triazalazoniaadamantanechloride). A 4 cm thick slab of gel-like material was created for testing. The pump
and circuit were then placed on top of the simulated brain tissue Figure 2-55a. Different current values were
applied using the Basic Stamp editor program and the infusion flow rate was calculated by measuring the
fluid front movement in a calibrated 100 µL micropipette. The simulated tissue material was then removed
and the experiment repeated with only air separating the transmitter and receiver Figure 2-55b.
81

Successful current modulation was achieved through the simulated brain tissue material (data not shown;
no significant difference was observed between infusion flow rates for wireless transmission through air
vs. simulated tissue; one way analysis of variance, p < 0.05). However, in both cases, the achieved flow
rate was smaller than that observed for when the pump was placed directly above the transmitter coil (15%
decrease). As previously mentioned, flow rate is linearly dependent on applied current, and therefore,
dependent on received power. As a result, decreased power transfer due to coil distance leads to decreased
flow rate. The decrease in power transfer with distance between the coils was mitigated by increasing the
power output of the transmitter, allowing for 12 cm vertical distance between the transmitter and receiver
without a decrease in flow rate.

Figure 2-55: Testing setup for experiments with simulated brain tissue material: (a) simulated brain tissue placed
between the pump and circuit and the transmitter (b) experiment repeated with only air separating the transmitter and
receiver at same distance.
82

3 CONCLUSION
A programmable, implantable, and low power (0.66-51.31 mW) electrochemical drug infusion
micropump system (Figure 2-56), capable of delivering a
diverse assortment of liquid drug formulations with high
accuracy within a wide dynamic range of dose volumes
and flow rates (0.33 - 141.9 µL/min) was demonstrated.
Viscosity independent delivery and real-time electrical
control of variable flow rate drug regimen were also
shown. A class E inductive powering system was designed,
along with ASK modulation to allow adjustment of the
infusion rate post-implantation. Successful delivery and flow rate adjustment was achieved through the
simulated brain tissue material.
4 REFERENCES
[1] R. Sheybani, H. Gensler, and E. Meng, "A MEMS electrochemical bellows actuator for fluid metering
applications," Biomedical Microdevices, pp. 1-12, 2012/07/01 2012.
[2] M. E. Orazem, Electrochemical impedance spectroscopy : selection of papers from the 6th
international symposium (EIS 2004), 16 - 21 May 2004, Cocoa Beach, FL, USA. Amsterdam Elsevier,
2006.
[3] J. Randles, "Kinetics of rapid electrode reactions," Discussions of the faraday society, vol. 1, pp. 11-
19, 1947.
[4] R. Sheybani and E. Meng, "High-Efficiency MEMS Electrochemical Actuators and Electrochemical
Impedance Spectroscopy Characterization," Microelectromechanical Systems, Journal of, vol. 21, pp.
1197-1208, 2012.
[5] L. L. Faulkner and A. J. Bard, Electrochemical methods and applications, 2nd ed. New York: Wiley,
2001.
[6] M. Faraday, "Proceedings of the Royal Society of London," Philosophical Transactions of the Royal
Society of London, vol. 124, pp. 77-122, 1834.
[7] W. Nicholson, A Journal of natural philosophy, chemistry and the arts: G. G. and J. Robinson, 1800.
[8] H. Janocha, "New Actuators," in Proc. Int. Technology-Transfer Conf., Actuator’88, Bremen,
Germany, 1988, pp. 389-427.
[9] H. Bauer, Derisavi-Fard, F. , Eckoldt, U., Gehrmann, R., Kickel, H., "Electrochemical Actuator,"
Germany Patent, 1996.
[10] "The Production of Hydrogen and Oxygen Through The Electrolysis of Water," Project Gutenberg's
Scientific American Supplement, vol. XXXII, 1891.
Figure 2-56: Micropump with wireless powering coil
and circuitry.
83

[11] C. G. Cameron and M. S. Freund, "Electrolytic actuators: Alternative, high-performance, material-
based devices," Proceedings of the National Academy of Sciences of the United States of America, vol.
99, pp. 7827-7827-7831, 2002.
[12] P. Choi, D. G. Bessarabov, and R. Datta, "A simple model for solid polymer electrolyte (SPE) water
electrolysis," in 14th International Conference on Solid State Ionics (SSI-14), 22-27 June 2003,
Monterey, CA, 2004, pp. 535-9.
[13] S. Bohm, B. Timmer, W. Olthuis, and P. Bergveld, "Closed-loop controlled electrochemically actuated
micro-dosing system," Journal of Micromechanics and Microengineering, vol. 10, pp. 498-504, 2000.
[14] P.-Y. Li, R. Sheybani, C. A. Gutierrez, J. T. W. Kuo, and E. Meng, "A parylene bellows
electrochemical actuator," Journal of Microelectromechanical Systems, vol. 19, pp. 215-28, 2010.
[15] J. Xie, Y. Miao, J. Shih, Q. He, J. Liu, Y.-C. Tai, et al., "An Electrochemical Pumping System for On-
Chip Gradient Generation," Analytical Chemistry, vol. 76, pp. 3756-3763, 2004/07/01 2004.
[16] C. Neagu, H. Jansen, H. Gardeniers, and M. Elwenspoek, "The electrolysis of water: an actuation
principle for MEMS with a big opportunity," Mechatronics, vol. 10, pp. 571-81, 2000.
[17] T. Stanczyk, B. Ilic, P. J. Hesketh, and J. G. Boyd Iv, "Microfabricated electrochemical actuator for
large displacements," Journal of Microelectromechanical Systems, vol. 9, pp. 314-320, 2000.
[18] C. R. Neagu. (1998). A medical microactuator based on an electrochemical principle. Available:
http://doc.utwente.nl/13878
[19] H. F. A. Verhaart, R. M. De Jonge, and S. J. D. Van Stralen, "Growth rate of a gas bubble during
electrolysis in supersaturated liquid," International Journal of Heat and Mass Transfer, vol. 23, pp.
293-9, 1980.
[20] S. Shibata, "Supersaturation of oxygen in acidic solution in the vicinity of an oxygen-evolving
platinum anode," Electrochimica Acta, vol. 23, pp. 619-623, 1978.
[21] C. W. M. P. Sillen, E. Barendrecht, L. J. J. Janssen, and S. J. D. van Stralen, "Gas bubble behaviour
during water electrolysis," International Journal of Hydrogen Energy, vol. 7, pp. 577-87, 1982.
[22] D. Linden, Handbook of batteries, 2nd ed.: McGraw-Hill, 1995.
[23] C. Song and J. Zhang, "Electrocatalytic Oxygen Reduction Reaction," in PEM Fuel Cell
Electrocatalysts and Catalyst Layers, J. Zhang, Ed., ed: Springer London, 2008, pp. 89-134.
[24] S. Grigoriev, P. Millet, K. Dzhus, H. Middleton, T. Saetre, and V. Fateev, "Design and characterization
of bi-functional electrocatalytic layers for application in PEM unitized regenerative fuel cells,"
International Journal of Hydrogen Energy, vol. 35, pp. 5070-5076, 2010.
[25] S. Song, H. Zhang, X. Ma, Z.-G. Shao, Y. Zhang, and B. Yi, "Bifunctional oxygen electrode with
corrosion-resistive gas diffusion layer for unitized regenerative fuel cell," Electrochemistry
Communications, vol. 8, pp. 399-405, 2006.
[26] T. Ioroi, K. Yasuda, Z. Siroma, N. Fujiwara, and Y. Miyazaki, "Thin film electrocatalyst layer for
unitized regenerative polymer electrolyte fuel cells," Journal of Power Sources, vol. 112, pp. 583-7,
2002.
[27] A. Verma and S. Basu, "Feasibility study of a simple unitized regenerative fuel cell," Journal of power
sources, vol. 135, pp. 62-65, 2004.
[28] G. Chen, H. Zhang, J. Cheng, Y. Ma, and H. Zhong, "A novel membrane electrode assembly for
improving the efficiency of the unitized regenerative fuel cell," Electrochemistry communications,
vol. 10, pp. 1373-1376, 2008.
[29] Perma Pure. (2014). Nafion: Physical and Chemical Properties. Available:
http://www.permapure.com/products/nafion-tubing/nafion-physical-and-chemical-properties/
[30] S. Goswami, S. Klaus, and J. Benziger, "Wetting and absorption of water drops on Nafion films,"
Langmuir, vol. 24, pp. 8627-8633, 2008.
[31] J. Maruyama, M. Inaba, K. Katakura, Z. Ogumi, and Z. i. Takehara, "Influence of Nafion film on the
kinetics of anodic hydrogen oxidation," Journal of Electroanalytical Chemistry, vol. 447, pp. 201-
209, 1998.
84

[32] Z. Dai and H. Mohwald, "Highly stable and biocompatible nafion-based capsules with controlled
permeability for low-molecular-weight species," Chemistry - A European Journal, vol. 8, pp. 4751-
4755, 2002.
[33] H. Kim, M.-S. Kang, D. H. Lee, J. Won, J. Kim, and Y. S. Kang, "Proton exchange membranes with
high cell performance based on Nafion/poly(p-phenylene vinylene) composite polymer electrolyte,"
Journal of Membrane Science, vol. 304, pp. 60-4, 2007.
[34] B. Lenaerts, Omnidirectional inductive powering for biomedical implants, 1st ed. New York: Springer,
2008.
[35] A. C. Albason, D. W.-Y. Chung, and S.-L. Lou, "A 2 MHz wireless CMOS transceiver for implantable
biosignal sensing systems," Journal of Signal Processing Systems, vol. 62, pp. 263-272, 2011.
[36] T. H. Glisson, Jr., Introduction to circuit analysis and design. New York: Springer Science + Business
Media, 2011.
[37] L. Liu, X. Shao, F. Wu, X. Huo, W. Li, and L. Mo, "Design and application research of implantable
wireless power transmission Micro Electro Mechanical System," in 4th International Seminar on
Modern Cutting and Measurement Engineering, Beijing, China, 2010, p. China Instrument and
Control Society; National Precision Tool Engineering Technology Research Center; Chengdu Tool
Research Institution.
[38] C. R. Sullivan, "Optimal choice for number of strands in a litz-wire transformer winding," in PESC97.
Record 28th Annual IEEE Power Electronics Specialists Conference. Formerly Power Conditioning
Specialists Conference 1970-71. Power Processing and Electronic Specialists Conference 1972, New
York, NY, USA, 1997, pp. 28-35.
[39] F. Zhang, L. Xiaoyu, S. A. Hackworth, R. J. Sclabassi, and S. Mingui, "In vitro and in vivo studies on
wireless powering of medical sensors and implantable devices," in 2009 IEEE/NIH Life Science
Systems and Applications Workshop (LiSSA 2009), Piscataway, NJ, USA, 2009, pp. 84-7.
[40] R. Sheybani, S. M. Schober, and E. Meng, "Drug Delivery Using Wireless MEMS," in Handbook of
MEMS for wireless and mobile applications, Uttamchandani, Ed., ed: Woodhead Publishing, 2013,
pp. 489-517.
[41] P. Y. Li, R. Sheybani, J. T. W. Kuo, and E. Meng, "A parylene bellows electrochemical actuator for
intraocular drug delivery," in 15th International Conference on Solid-State Sensors, Actuators and
Microsystems. Transducers 2009, Piscataway, NJ, USA, 2009, pp. 1461-4.
[42] J. J. Whalen III, J. D. Weiland, and P. C. Searson, "Electrochemical deposition of platinum from
aqueous ammonium hexachloroplatinate solution," Journal of the Electrochemical Society, vol. 152,
pp. C738-C743, 2005.
[43] H. Gensler, R. Sheybani, P.-Y. Li, R. L. Mann, and E. Meng, "An implantable MEMS micropump
system for drug delivery in small animals," Biomedical Microdevices, vol. 14, pp. 483-496, 2012.
[44] R. Sheybani and E. Meng, "High efficiency wireless electrochemical actuators: Design, fabrication
and characterization by electrochemical impedance spectroscopy," in 24th IEEE International
Conference on Micro Electro Mechanical Systems, MEMS '11, Cancun, Mexico, 2011, pp. 1233-1236.
[45] R. Sheybani and E. Meng, "Acceleration Techniques for Recombination of Gases in Electrolysis
Microactuators," Sensors & Actuators: B. Chemical, 2015 (submitted).
[46] I. O. K'Owino and O. A. Sadik, "Impedance Spectroscopy: A Powerful Tool for Rapid Biomolecular
Screening and Cell Culture Monitoring," Electroanalysis, vol. 17, pp. 2101-2113, 2005.
[47] A. Petrossians, J. J. Whalen Iii, J. D. Weiland, and F. Mansfeld, "Electrodeposition and
characterization of thin-film platinum-iridium alloys for biological interfaces," Journal of the
Electrochemical Society, vol. 158, pp. D269-D276, 2011.
[48] D. T. Sawyer and J. L. Roberts, Experimental electrochemistry for chemists. New York,: Wiley, 1974.
[49] K. Broka and P. Ekdunge, "Oxygen and hydrogen permeation properties and water uptake of Nafion
117 membrane and recast film for PEM fuel cell," Journal of Applied Electrochemistry, vol. 27, pp.
117-23, 1997.
[50] E. Meng, X. Zhang, and W. Benard, "Additive Processes for Polymeric Materials," in MEMS
Materials and Processes Handbook, R. Ghodssi and P. Li, Eds., ed: Springer, 2011, pp. 193-271.
85

[51] L. I. Maissel and R. Glang, Handbook of thin film technology. New York,: McGraw-Hill, 1970.
[52] C. de Haro, R. Mas, G. Abadal, J. Munoz, F. Perez-Murano, and C. Dominguez, "Electrochemical
platinum coatings for improving performance of implantable microelectrode arrays," Biomaterials,
vol. 23, pp. 4515-21, 2002.
[53] L. S. Robblee, J. McHardy, W. F. Agnew, and L. A. Bullara, "Electrical stimulation with Pt electrodes.
VII. Dissolution of Pt electrodes during electrical stimulation of the cat cerebral cortex," Journal of
Neuroscience Methods, vol. 9, pp. 301-8, 1983.
[54] S. B. Brummer and M. J. Turner, "Electrical stimulation with Pt electrodes. II. Estimation of maximum
surface redox (theoretical non-gassing) limits," IEEE Transactions on Biomedical Engineering, vol.
BME-24, pp. 440-3, 1977.
[55] J. H. Lee, H. Kyo Seon, Y. Ki Hyun, K. Tae Song, and A. Saeyoung, "Microstructure and adhesion of
Au deposited on parylene-c substrate with surface modification for potential immunoassay
application," IEEE Transactions on Plasma Science, vol. 32, pp. 505-9, 2004.
[56] C.-M. Han, L. Eun-Jung, K. Hyoun-Ee, K. Young-Hag, K. N. Kim, H. Yoon, et al., "The electron
beam deposition of titanium on polyetheretherketone (PEEK) and the resulting enhanced biological
properties," Biomaterials, vol. 31, pp. 3465-70, 2010.
[57] R. N. O'Brien and K. S. V. Santhanam, "Anodic oxidation and deposition of polymer films as
electrodes on metals," in Extended Abstracts, Spring Meeting - Electrochemical Society., Toronto,
Ont, Can, 1985, p. 51.
[58] R. Sheybani, H. Gensler, and E. Meng, "Rapid and repeated bolus drug delivery enabled by high
efficiency electrochemical bellows actuators," in 2011 16th International Solid-State Sensors,
Actuators and Microsystems Conference, Transducers '11 Beijing, China, 2011, pp. 490-3.
[59] Z. Ogumi, Z. Takehara, and S. Yoshizawa, "Gas Permeation in SPE Method I. Oxygen Permeation
Through Nafion and Neosepta," Journal of the Electrochemical Society, vol. 131, pp. 769-773, 1984.
[60] C. N. Murray and J. P. Riley, "The solubility of gases in distilled water and sea water—II. Oxygen,"
Deep Sea Research and Oceanographic Abstracts, vol. 16, pp. 311-320, 6// 1969.
[61] L. K. Verheij, "Kinetic modelling of the hydrogen-oxygen reaction on Pt(111) at low temperature (170
K)," Surface Science, vol. 371, pp. 100-10, 1997.
[62] H. Gensler and E. Meng, "Rapid fabrication and characterization of MEMS Parylene C bellows for
large deflection applications," Journal of Micromechanics and Microengineering, vol. 22, p. 115031,
2012.
[63] H. Gensler, R. Sheybani, and E. Meng, "Rapid non-lithography based fabrication process and
characterization of Parylene C bellows for applications in MEMS electrochemical actuators," in 16th
International Solid-State Sensors, Actuators and Microsystems Conference, Transducers '11, Beijing,
China, 2011, pp. 2347-2350.
[64] T. A. Goers, "The Washington Manual of Surgery," ed: Lippincott Williams & Wilkins, 2008.
[65] J. Zhang, X. Wang, J. Hu, B. Yi, and H. Zhang, "A novel method for preparing pemfc catalytic layers,"
Bull Chem Soc Japan, vol. 77, pp. 2289-90, 2004.
[66] D. E. Clark and B. K. Zoitos, Corrosion of Glass, Ceramics and Ceramic Superconductors: Principles,
Testing, Characterization and Applications, 1 ed.: William Andrew, 1993.
[67] R. M. Barrer, Diffusion in and through solids: CUP Archive, 1941.
[68] T. Ooya and N. Yui, "Supramolecular-structured polymers for drug delivery," ACS Symposium Series,
vol. 752, pp. 375-384, 2000.
[69] A. Kogan and N. Garti, "Microemulsions as transdermal drug delivery vehicles," Advances in Colloid
and Interface Science, vol. 123-126, pp. 369-385, 2006.
[70] S. H. Lee, J. M. Oh, J. S. Son, J. H. Lee, J. Il Kim, D. Y. Kwon, et al., "Preparation and characterization
of polymeric micelles consisting of poly(propylene glycol) and poly(caprolactone) as a drug carrier,"
in 2010 3rd International Nanoelectronics Conference, INEC 2010, Hongkong, China, 2010, pp.
1393-1394.
[71] R. Sheybani and E. Meng, "Electrochemical Drug Infusion Micropump with Wide Dynamic Range
and Viscosity Independent Pumping," in 2012 BMES Annual Meeting, Atlanta, GA, 2012.
86

[72] K. v. Schuylenbergh and R. Puers, Inductive powering : basic theory and application to biomedical
systems. Dordrecht: Springer, 2009.
[73] T. Givrad, "Induction power microbolus infusion pump used for functional neuroimaging applications
in rodents," University of Southern California Ph.D., University of Southern California, United States
-- California, 2007.
[74] A. Cobo, R. Sheybani, H. Tu, and E. Meng, "A Wireless Implantable Micropump for Localized Drug
Infusion," Biomedical Microdevices, 2015 (submitted).
[75] K. van Schuylenbergh and R. Puers, Inductive Powering: Basic Theory and Application to Biomedical
Systems: Springer Science & Business Media, 2009.
[76] E. Okamoto, Y. Kato, K. Seino, H. Miura, Y. Shiraishi, T. Yambe, et al., "A new transcutaneous
bidirectional communication for monitoring implanted artificial heart using the human body as a
conductive medium," Artificial organs, vol. 36, pp. 852-858, 2012.
[77] C.-K. Liang, J. J. J. Chen, C. Cho-Liang, C. Chen-Li, and W. Chua-Chin, "An implantable bi-
directional wireless transmission system for transcutaneous biological signal recording,"
Physiological Measurement, vol. 26, pp. 83-97, 2005.
[78] R. Sheybani and E. Meng, "On-demand wireless infusion rate control in an implantable micropump
for patient-tailored treatment of chronic conditions," in Engineering in Medicine and Biology Society
(EMBC), 2014 36th Annual International Conference of the IEEE, 2014, pp. 882-885.
[79] G. Hartsgrove, A. Kraszewski, and A. Surowiec, "Simulated biological materials for electromagnetic
radiation absorption studies," Bioelectromagnetics, vol. 8, pp. 29-36, 1987.

87

CHAPTER 3: REAL-TIME ELECTROCHEMICAL DOSE TRACKING SENSORS
FOR CLOSED-LOOP DRUG DELIVERY APPLICATIONS

Electrochemical dose tracking is achieved through
measuring electrochemical impedance (EI) by applying a
small sinusoidal excitation voltage across a set of
electrodes placed in the drug reservoir (Figure 3-1). At
sufficiently high frequencies, the measured impedance
could be directly correlated to the volume of drug
remaining in the reservoir. Therefore, changes in
reservoir content, delivery flow rate, blockages in the delivery catheter, drug refills, and even damage to
the electrolysis chamber could be assessed in real-time and recorded for future analysis using a LabVIEW
graphical user interface (GUI).
The following chapter is presented in two sections: first, the theory, fabrication, and experimental
methods and results are discussed for EI sensing using electrodes connected to a precision LCR meter. This
section is then followed by a discussion on omnidirectional wireless data transmission and the design and
characterization of the circuits with sensors, allowing for wireless dose sensing.
CHAPTER 3-1: ELECTROCHEMICAL DOSE TRACKING SENSORS

1 PRELIMINARY STUDIES (THEORY & PREVIOUS WORK)
1.1 THEORY
When a metal electrode is placed in an electrolyte, the excess charge on the surface of the metal and the
resulting accumulation of ions and molecules in the electrolyte create two layers of charge at the electrode-

Figure 3-1: EI Dose sensing operation concept.
88

electrolyte interface, named the Helmholtz double layer (Figure 3-2) [1]. When current is applied to the
electrode, two types of processes can occur at this electrode-solution interface. In Faradaic (non-reversible)
reactions, electrons are transferred across the metal solution interface leading to oxidation and reduction
reactions at the metal surface. At certain applied potentials, charge transfer does not transpire due to
unfavorable thermodynamic conditions. Instead, non-Faradaic (reversible) processes including charging
and discharging of the double-layer, oxidation and reduction of the surface monolayer, and H-atom plating
and hydride oxidation occur [2]. This results in the creation of an array of charges and oriented dipoles
called the electrical double layer that effectively acts as a capacitance. The overall electrochemical cell is
approximated by the Randles equivalent circuit model comprising the parallel contribution of the
polarization impedance and double layer capacitance in series with the solution resistance [3]. If mass
transfer is assumed to be negligible and the processes are limited by diffusion, a two electrode system can
be modeled using a simplified circuit [4].

Figure 3-2: Electrode-electrolyte interface: excess charge on the surface of the metal and the resulting accumulation
of ions and molecules in the electrolyte create two layers of charge, named the Helmholtz double layer (the dashed
line separated the inner and outer layers).
Electrochemical dose tracking is accomplished by measuring the electrochemical impedance (EI) under
a small sinusoidal excitation voltage across a set of electrodes placed in the electrolyte solution to be dosed.
At sufficiently high frequencies (1 kHz for water), the impedance response is dominated by the solution
resistance, the magnitude of which is dependent on the cross sectional area of the conductive path through
89

the electrolyte. This dependency can be used to correlate the measured impedance value with the volume
of liquid electrolyte remaining within a rigid chamber (Figure 3-3).

Figure 3-3: Dose sensing is achieved by measuring the impedance across the sensing electrodes placed in the
reservoir. At a sufficiently high frequencies, the impedance represents the solution resistance that can be correlated
to the volume of fluid remaining in the reservoir.
The measurement frequency (fm) at which the impedance response is dominated by the solution
resistance is dependent on the electrical properties of the solution; therefore, for solutions with different
conductivities, a potentiostatic sweep of frequency is required to pinpoint the appropriate measurement
frequency. These measurements can be carried out under low power conditions (< 100 µW) and the
excitation voltages required are within the “water window” (≤ 1 V pp to avoid hydrolysis of water). Under
these conditions, only non-Faradaic (reversible) reactions occur and no new chemical species are created
that can diffuse away or otherwise alter the chemical composition of the bulk solution [5].
This dose tracking method can be readily integrated with various pumping approaches; here, this
technique is demonstrated in the electrochemically actuated infusion pump presented in the previous
chapter. Briefly, actuation is achieved using a set of interdigitated Pt electrodes to electrolyze water into
hydrogen and oxygen. The resulting volume expansion inflates a Parylene bellows which in turn displaces
the drug in the adjacent reservoir through a catheter to the delivery site. Impedance measurements were
acquired in real-time using a precision impedance analyzer (Agilent E4980A, Agilent Technologies, Santa
Clara, CA) connected to the EI electrodes placed inside the drug chamber in contact with the drug and
recorded via a LabVIEW interface.
90

1.2 PREVIOUS WORK
Gutierrez, et. al., previously demonstrated a highly accurate method to realize real-time tracking of sub-
nanoliter liquid volumes, encapsulated within Parylene-based microstructures, through electrochemical
impedance measurements [6]. This work provided the basis for applying electrochemical impedance
measurements to tracking larger volumes relevant for drug delivery applications. Proof of concept was
demonstrated using simple nickel plated copper wires placed in a silicone reservoir chamber, coupled via a
silicone membrane to an electrolysis actuation chamber (Figure 3-4) [7].
To assemble the pump, first, electrolysis electrodes (described in the previous chapter) were adhered on
the pump chamber base. Then the pump diaphragm (400 µm thick) and drug reservoir were attached using
PDMS pre-polymer and cured in place. Impedance measurement electrodes were then integrated into the
drug reservoir. Two dedicated impedance measurement microelectrodes (Kynar™ silver plated copper
wires, 30 AWG, Jameco Electronics, Belmont, CA) were inserted and attached to the reservoir using PDMS
pre-polymer and cured in place. Finally, the pump chamber and reservoir were filled with double distilled
(DD) water (serving as electrolyte and model drug, respectively) using a syringe.

Figure 3-4: Photograph of proof of concept prototype (modified with permission from [7] © 2011 IEEE).
Impedance measurements were acquired in real-time via a LabVIEW-interfaced precision LCR meter
(1 V pp, 5 kHz) connected to the impedance measurement electrodes. Preliminary results demonstrated
91

detection of physiologically-relevant drug volumes (500 nL−64 µL) and on-the-fly flow rate variations
(2.78−80 µL/min) [7].
2 RESEARCH DESIGN & METHODS
2.1 FABRICATION
Studies have shown that the increased electrode surface area can increase the rate of electrochemical
reactions and decrease drift and noise in measurement [8-10]. Two types of EI measurement electrodes
with considerably different surface areas were fabricated: thin film and wire electrodes. Thin film electrodes
(2 mm × 1 mm) were fabricated on a Borofloat® 33 glass wafers (University Wafer, Boston, MA) substrate
by liftoff (Ti/Pt 300 Å/2000 Å). Kynar™ silver plated copper wires (30 AWG, Jameco Electronics,
Belmont, CA) were affixed to the electrodes using conductive epoxy (EpoTECH H20, Epoxy Technology,
Billerica, MA) to provide external electrical connections. The connection was then insulated and reinforced
with marine epoxy (Loctite, Westlake, OH). Bulk wire electrodes were fabricated from 99.9% Pt wire (Ø
0.5 mm) (California Fine Wire, Grover Beach, CA) as an alternative to the thin film impedance electrodes.
Heat shrink tubing (Zeus, Orangeburg, SC) was introduced as insulation. A 2 mm segment of the tip was
exposed and then sanded (60 and 220 grit silicon carbide sandpaper) to increase surface area (Figure 3-5a).
The electrodes were then electrochemically cleaned (±0.5 V (Gamry Reference 600 Potentiostat,
Warminster, PA) in 1× phosphate buffered saline (PBS)) and packaged into the drug reservoir (Figure 3-
5b) [11].
92

Figure 3-5: Photograph of (a) bulk wire electrode; (b) drug delivery system with integrated EI sensors.

2.2 ELECTRODE PLACEMENT OPTIMIZATION
The electrochemical dose tracking system should be able to accurately track doses in the full range
delivered by the electrolysis based drug delivery pump (from a few nL to 100’s of µL). Bohm, et al.,
reported the placement of the impedance sensing electrodes within the electrolysis chamber on the same
substrate as the electrolysis electrodes [12]. This electrochemical configuration is feasible for low delivery
volumes of < 800 nL. However, higher flow rates and delivery volumes require higher applied currents. In
this regime, the electrolysis electrodes act as a magnetic core in the magnetic field created by the EI
electrodes. This field becomes significant at high frequencies required for measurement. The model circuit
(a)
(b)
93

no longer follows the simplified Randles circuit (Figure 3-6) and therefore, dose measurements based on
tracking solution resistance are not feasible [13].

Figure 3-6: High frequency inductance observed when EI electrodes are placed inside the electrolysis chamber.
This phenomena was confirmed in EI spectroscopy measurements (0.005-100 kHz) with thin film EI
electrodes placed inside and outside the electrolysis chamber while activating electrolysis through the
actuator electrodes Therefore, for high flow operation, EI electrodes should be placed externally with
respect to the electrolysis cell enclosed by the bellows. The electrodes placement with respect to the bellows
actuator was also investigated. Electrodes were arranged opposite one another on either side of the bellows
actuator or perpendicularly. Electrode separation from the bellows actuator was evaluated (3-5 mm; Figure
3-7). The best resolution (230 nL bolus) was obtained for electrodes placed 3 mm from the bellows actuator
and directly across from one another [11].

Figure 3-7: Investigating electrodes placement with respect to the bellows actuator: (a) electrodes arranged
perpendicularly, or (b) opposite one another on either side of the bellows actuator (EI sensing electrodes are marked
with X).
2.3 THIN FILM VS. BULK WIRE
Electrochemical reactions are essentially confined to surfaces and their rate can be increased by
increasing surface area or electrocatalytic properties [8]. In order to compare the electrochemically active
surface area of the thin film and bulk wire electrodes, cyclic voltammetry (CV) was performed (Gamry
Reference 600 Potentiostat, Gamry Instruments, Warminster, PA). A platinum plate (1 cm
2
) was used as
94

the counter electrode, and an Ag/AgCl electrode as the reference (BASi, West Lafayette, IN). Twenty CV
cycles were performed under N 2 purging at a scan rate of 250 mV/s. The electrochemically active surface
area of the two electrode types was calculated by dividing the measured charge by 210 [µC/cm
2
]. Measured
charge is the integrated area under the hydrogen adsorption peaks of the CV curve (Figure 3-8) [14]. The
electrochemically active surface areas were calculated to be 3.64 mm
2
and 1.94 mm
2
for bulk wire and thin
film electrodes (compared to their geometric surface area of 3.53 mm
2
and 2 mm
2
), respectively. The bulk
wire has an estimated 73% greater electrochemically active surface area and roughening with sandpaper
further increases this by 8.6%.

Figure 3-8: Cyclic voltammetry for thin film electrodes and the Pt wire electrodes. Inset: electrochemically active
surface area equation.
2.4 FLUID BASED CALIBRATION
The electrolysis based actuator is capable of delivering a diverse assortment of liquid drug formulations
with a large range of viscosities (0.78-26.71 cSt) regardless of the fluid’s ionic properties [15]. EI
spectroscopy, however, measures impedance of at the electrode-electrolyte interface, and therefore the ionic
concentration of the solution will impact the measurement reading. As the sensors are meant to offer
feedback for the actuator operation, several water and lipid soluble fluids with different ionic concentrations
were studied as model drugs.
95

2.4.1 Water Soluble Fluids
Two solutions, DD water and 1× PBS, were chosen as water soluble model drugs. DD water has low
conductivity (5.0 × 10
-6
[S/m] at 25 ºC), whereas, by comparison, 1× PBS is highly conductive (1.9 × 10
-4

[S/m] at 25 ºC). Each solution was evaluated in separate trials, and two-electrode EI spectroscopy was
performed between 5.0 × 10
-3
– 1.0 × 10
5
Hz to determine the frequency range at which the solution
resistance dominates the impedance response (Figure 3-9).

Figure 3-9: EI spectroscopy for (a) DI water, and (b) 1× PBS. The region to the right of the dashed line represents the
frequency range at which the solution resistance is dominant (reprinted with permission from [11] © 2012 IEEE).
For thin film electrodes, 1 and 100 kHz were chosen as the appropriate measurement frequency for DD
water and PBS, respectively. 1, 3, and 5 mA currents were applied to the electrolysis electrodes, for 2, 1,
and 0.5 minutes, respectively. Each current value was applied 5 times. The volume dispensed from the
pump, as well as the change in impedance (at 1 Vpp) was recorded for each run. The results for each current
value were used to obtain averaged trends (3-1 and 3-2). The impedance values were normalized to the
baseline value for each measurement. The equation corresponding to the linear fit of the data was then used
to calculate the calibration curve for each current. The results showed that differing current values did not
affect the calibration [11].
DD Water:

= 5287 −1.0000 (3-1)
PBS:

= 4323 −0.9997 (3-2)
96

For the bulk wire electrodes, once the appropriate frequency region to bypass the double layer
capacitance, while avoiding parasitic effects encountered at higher frequencies, was chosen (1 kHz and 100
kHz were chosen for DD water and PBS, respectively), 3 mA current was applied to the electrolysis
electrodes for 30 seconds (n =10). For each run, the volume dispensed from the pump was measured by
recording fluid displacement in a 100 µL calibrated micropipette. The change in impedance was acquired
in real-time using a precision impedance analyzer (at 100 mV pp) connected to the electrodes and recorded
via a LabVIEW interface. Once again, the results were used to obtain averaged trends and the impedance
values were normalized to the baseline value for each measurement (3-3) [16]. The calibration curve was
calculated using the linear fit of the data.
A LabVIEW graphical user interface was designed to calculate the dispensed volume using the measured
impedance. Volume of drug to be delivered and fluid type (water/PBS; for correct measurement frequency
range) were inputted into the software controller. The software then used the real-time impedance
measurements and the built-in calibration curves to supply the appropriate power signal to the pump to
deliver the desired volume at the specified time intervals. All data acquired by the LabVIEW interface was
then saved for further analysis.

= 4133 −1.0005 (3-3)
As expected, differing current values applied to the actuator do not affect the calibration. Also, as
previously mentioned, the software is programmed to operate based on impedance values normalized the
baseline value for each measurement each delivery, therefore, once the appropriate frequency was selected
to match the fluid pumped, the calibration curve was no longer dependent on the fluid. Therefore the same
calibration curve could be applied to other drug model solutions once the appropriate operating frequency
for the impedance measurement is determined by EI spectroscopy.
97

2.4.2 Lipid Soluble Fluids
Purified poppy seed oil was chosen as a lipid soluble model drug. Iodized ethyl ester of poppy seed oil
is often used as a contrast and carrier for embolizing agents to treat tumors [17]. Oils typically have very
low electrical conductivity on the order of ~10
-14
[S/m] at 25 ºC [18]. The solution was loaded in the drug
reservoir, and two-electrode EI spectroscopy was performed between 5.0 × 10
-3
– 1.0 × 10
5
Hz to determine
the frequency range at which the solution resistance is dominant. The EIS results showed the conductivity
of poppy seed oil to be too low and the capacitive region was not bypassed within the frequency limits
tested. Measurement limits of our equipment do not allow measurements above 100 kHz, as parasitic effects
are encountered at higher frequencies. Therefore, with the current measurement setup, even though the
electrolysis based actuator is fully capable of delivering lipid based drugs, electrochemical dose tracking of
low conductivity lipid solutions was not possible.
2.5 NOISE & DRIFT IN IMPEDANCE MEASUREMENTS
In EI measurements, typically an Ag/AgCl reference is preferred due to its high stability. However, Ag
+

exerts toxic effects by interfering with transmembrane Ca
++
flux [19, 20]. Therefore Pt, despite its inherent
drift when used as a reference, is the preferred electrode material for in vivo measurements [21]. Drift must
be minimized to obtain accurate dose tracking in real-time, especially when actuation has ceased and the
volume in the drug reservoir remains constant. It has been documented that non-precious metals such as
silver plated copper wire and conductive epoxies may introduce a considerable amount of noise and drift
to EI measurements [9]. Increased electrode surface area and reduced excitation voltage magnitude were
reported to reduce drift [9, 10].
2.5.1 Drift & Noise Surface Area Dependency
EI measurements (1 kHz, 1 V pp) were carried out following delivery of DI water for thin film and bulk
wire electrodes. The bulk wire electrodes showed significantly less drift at 1 V pp applied excitation voltage
98

following a perturbation to the system (actuation). Also, the elimination of the silver plated copper wires
and conductive epoxy further reduced drift (Figure 3-10) [11].

Figure 3-10: Normalized impedance response drift (1 Vpp, 1 kHz) after delivery comparing thin film electrodes with
epoxied wire and Pt wire electrodes (reprinted with permission from [11] © 2012 IEEE).
On the other hand, system noise rendered measurements at an excitation voltage below 1 V pp ineffective
for thin film electrodes [11]. However, with the bulk wire electrodes, due to the increased electrochemically
active surface area, accurate measurements could be made at 100 mV pp excitation voltage (Figure 3-11).
Therefore, all subsequent testing was carried out with bulk wire electrodes. In order to further minimize the
noise in impedance measurements, an averaging filter was added to the LabVIEW interface to smooth the
recorded signals.

Figure 3-11: Bolus delivery (6/10sec On/Off) impedance response (100 mVpp, 1 kHz). Two 5 µL boluses delivered
with 5 mA applied current (reprinted with permission from [11] © 2012 IEEE).
99

2.5.2 Drift Solution Dependency
Metal ions adsorbed on the surface of the electrodes (ad-atoms) have a certain lifetime on the substrate
surface and are subject to a random walk on the surface [22]. The structure of the double layer surrounding
the electrode is dependent on the concentration of the species in solution [23]. The higher the concentration
of ions in the solution, the faster ad-atoms are replaced at the electrode surface (maintain the Nernst
equilibrium) and the lower the measurement drift. In order to observe solution dependency of measurement
drift, DD water and 1× PBS were used as a model drugs with significantly different ion concentrations. The
impedance was recorded using a 100 mV pp signal at 1 and 100 kHz (for DD water and 1× PBS respectively),
every 0.5 mins for the first 30 min, then every 30 min, and again after 45 min. The normalized impedance
magnitude for 1× PBS varied just 0.29% from baseline during the 45 min measurement, compared to 0.70%
deviation for DD water (Figure 3-12). This confirms the expectation that measurements in higher ionic
concentration 1× PBS experience less drift than in DD water.

Figure 3-12: Drift solution dependency: drift in impedance magnitude for 1x PBS compared to that for water.
2.5.3 Drift Temperature Dependency
While conductivity of metals decreases with increasing temperature, the conductivity of electrolyte
solutions increases with increasing temperature, as ion conductivity is always coupled with mass transport
[24]. As mentioned in above, less drift is observed in the impedance measurement for a solution with higher
conductivity. As the drug delivery device is meant to be used at 37 ºC, drift in impedance measurement was
100

compared for a device as room temperature 25 ºC vs. 37 ºC. 1× PBS was chosen as the model solution. The
room temperature device was operated on benchtop, whereas the 37 ºC device was immersed in a water
bath for the duration of testing. The impedance value was recorded using a 100 mVpp signal at 100 kHz,
every 5 mins for the first 30 min, then every hour for 4 hours, and again after 24 hours. As expected, the
sensors operated at 37 ºC equilibrate faster and impedance drift is lower the 24 hour period compared to
room temperature operation. It is important to note that the majority of drift is observed within the first hour
until a state of equilibrium is reached. Depending on the excitation voltage used to obtain the measurement,
a finite amount of time is required for the system to settle into equilibrium [10]. The results show that the
measurement value deviates ~2.83 - 4.07 % depending on the solution temperature for the first hour post
perturbation (Figure 3-13). This value drops to 0.18 - 0.93% in the following 24 hours. Therefore, once
equilibrium is reached the sensors are quite stable over time and could be used in systems meant for long
term use with high accuracy.

Figure 3-13: Drift temperature dependency: drift in impedance magnitude for 1x PBS at room temperature (25ºC) vs.
body temperature (37ºC) over 24 hrs. Inset: first hour of measurement.
2.5.4 Three Electrode Configuration
In EI measurements, the cell current is passed through the counter electrode, in order to create the AC
stimulation required for measurement. When using two electrodes for EI measurements, the counter
electrode also serves as the reference electrode. The addition of a third Pt electrode to separate the counter
101

and reference electrodes would improve the stability of the reference and potentially minimize drift [25]. A
1.1 mL reservoir was fabricated from acrylic. Three bulk wire electrodes were fabricated from 99.9% Pt
wire (Ø 0.5 mm) (California Fine Wire, Grover Beach, CA). A 3 mm segment of the tip was exposed and
then sanded (60 and 220 grit silicon carbide sandpaper) to increase surface area. The electrodes were placed
in the reservoir and bent to be flush with the wall. A Nafion® coated electrode with a two convolution
bellows was also incorporated. All joints were reinforced with marine epoxy (Figure 3-14).

Figure 3-14: Photograph of the three electrode system.
The reservoir was filled with 1× PBS. Two consecutive EI measurements of idle system drift were
recorded for two and three electrode configurations using a 100 mVpp signal at 100 kHz. As expected, more
drift was observed during the first test for both configurations as the electrode-electrolyte interface trends
towards an equilibrium state. Preliminary results indicate that in equilibrium, drift (deviation in
measurement impedance) in the measurement acquired using the three electrode configuration is 2.7 times
smaller than that acquired using the two electrode configuration.
2.6 DELIVERY OPERATION & REAL-TIME DOSE TRACKING
2.6.1 Smallest Detected Volume
The smallest volume detected with thin film electrodes placed 3 mm from the bellows actuator and
directly across from one another was 230 nL bolus (0.008% of reservoir) with DD water as the model drug
[11]. The bulk wire electrodes possess a higher active surface area, which should allow for the detection of
lower bolus volumes. A LabVIEW graphical user interface (GUI) was created to allow for sub second
18mm
102

control of applied current pulses. The impedance value was recorded using a 100 mV pp signal at 100 kHz
(for 1× PBS) and 1 kHz (for DD water). The smallest bolus volume detected for DD water was 83 nL
(0.003% of the reservoir; 3.25 mA applied to actuator for 1 s) and 556 nL for 1× PBS (0.019% of the
reservoir; 7 mA applied to actuator for 1 s). The conductivity of 1× PBS is two orders of magnitude higher
than that of DD water. As a result, the changes in impedance will be smaller for a given volume. Therefore,
given the same testing conditions, a smaller bolus volume could be detected for DD water.
In order to increase the sensitivity of detection, a Wheatstone bridge was devised. The three constant
bridge resistors (R 1, R 2, and R 3) were chosen based on the impedance range observed for 1× PBS at 100
kHz (R 1 = R 2 = R 3 = 3.9 kΩ). An NI 9219 24-bit universal analog input (National Instruments Corp., Austin,
TX) was used to acquire the bridge voltage signal through a LabVIEW GUI. Using Kirchhoff's first and
second rule, the impedance can be calculated using the measured bridge voltage (3-4):
#
$
=
%&''
(
)
(
*
+'.,
'.,-
(
)
(
*

(3-4)
where V S = 5 V, R x is the impedance measured using the EI electrodes, and V G the bridge voltage. The NI
9219 24-bit universal analog input has a resolution of 1 mV. Based on the previously obtained calibration
curve, this voltage corresponds to a 3.3 µL bolus. Therefore, the attainable smallest bolus volume with the
Wheatstone bridge configuration was found to be larger than that obtained with the precision LCR meter.
2.6.2 Real-time Flow Variation
The electrolysis based actuator rapidly responds to changes in regimen (<1 s). Preliminary results using
the simplified prototype with a silicone membrane with nickel coated copper wires used as EI measurement
electrodes, were demonstrated for a simple prototype demonstrated the system’s ability to discriminate on-
the-fly changes in delivery rate (and hence flow rate) [7]. Real-time flow variations for a device packaged
with bulk wire Pt electrodes filled with 1× PBS, were studied for this work. 5 , 1 , 8, and then 2 mA were
103

applied consecutively for 30 second durations to the electrolysis electrodes and real-time transitions in the
measured impedance value were recorded using a 100 mV pp signal at 100 kHz.

Figure 3-15: Real-time dose tracking of actuator response to changes in the applied pump current levels (modified
from [16]).
Monotonically increasing impedance magnitude was observed and directly corresponded to changes in
the applied pump current levels (Figure 3-15). Flow rates (35.2, 5.12, 72, and 13.34 µL/min) were measured
for pump currents of 5, 1, 8, and 2 mA, respectively, and were comparable to those previously measured
for each current individually [16].

Figure 3-16: Real-time dose tracking of lidocaine delivery [16].
104

Real-time dose tracking of the delivery of lidocaine HCl hydrate (Enzo Life Sciences, Farmingdale,
NY), a common water soluble, local anesthetic and antiarrhythmic drug [26], dissolved in PBS (20 mg/mL)
was confirmed by supplying the electrolysis actuator with 3 mA for 90 seconds (Figure 3-16) [16].
2.6.3 Recombination Detection
As mentioned in the previous chapter, once the current is removed, gases generated by the bellows
actuator recombine back to water. This reaction is limited by high overpotentials and high activation energy
[27, 28], and therefore requires Pt catalyst. The rate of reaction is limited by the rather slow diffusion of
gases through water and Nafion® to the surface of the electrolysis electrodes. In the absence of a flow
regulating check-valve in-line with the delivery catheter, dispensed drug along with bodily fluids will be
pulled back towards the inside of the catheter and potentially inside the pump during recombination.
Tracking of fluid backflow could eliminate the need for a check valve. Whenever, backflow is detected, a
small “maintenance” current can be applied to the electrolysis electrodes to maintain the fluid level. As
recombination occurs on a much slower time scale compared to actuation, the effects of drift in the EI
measurement could become dominant. Drift in tracking the recombination of various bolus sizes were
studied by closing a PEEK shut off valve immediately after delivery of the bolus and measuring changes
in impedance magnitude changes using a 100 mV pp signal at 100 kHz during the recombination period.
Drift in tracking the recombination of a 50 µL bolus of 1× PBS (5 mA applied current; n =3) is presented
as an example. The valve completely inhibits any fluid backflow during recombination (± 0.07 µL). Without
volume changes, the changes in the impedance magnitude should be negligible and any changes measured
could be attributed to drift in measurement. It is important to note that the relaxation motion of the bellows
actuator returning to its original position during recombination would change the fluidic path between the
two EI electrodes and effectively the measured impedance. However, these changes are negligible
compared to the volume change due to fluid leaving or entering the reservoir, and could not be measured
with our measurement setup. Average drift in measured impedance magnitude over 50 minutes is ~0.39%.
Using the calibration curve, the volume corresponding to the maximum change in impedance magnitude
105

was estimated to be 13.95 µL. Therefore, the dose tracking system could potentially detect a recombined
bolus of > 14 µL, when recombination occurs within 50 minutes. Smaller volumes would not be
distinguishable from the drift. Increasing the available catalyst surface area can increase the rate of
recombination (oxygen reduction) [29, 30].
As detailed in [15, 31], a limiting factor in recombination speed is the slow diffusion of gases through
water to the surface of the catalyst. Therefore, in order to increase the recombination rate, Pt film (~2000
Å) was e-beam evaporated onto the interior surface of the Parylene bellows. Using this coated bellows,
impedance magnitude changes during delivery and recombination of a 35 µL bolus of DD water (5 mA
applied current) were measured using a 100 mV pp signal at 1 kHz. In order to compare drift to the
recombined signal, recording of the impedance magnitude signal was continued for an additional 20 min
after the bolus had completely recombined. The 35 µL bolus delivered with the Pt coated Parylene bellows
actuator, completely recombined within 50 minutes and 85% of the volume recombined fast enough to be
discernible from the drift in measurement. For applications where continuous measurement of very small
volumes is desired, and therefore the system is consistently perturbed, it is possible to employ further
hardware and software calibration for the recorded EI measurement data to reduce drift [32].
It is important to note that drift does not follow a predetermined trend, and therefore repeating
experiments will not help identify and eliminate drift [10]. However, if sufficient time is allowed between
the repeat runs and when fresh solution is introduced to the system, then the system will return equilibrium
and less drift in measurement will be observed [10]. Once equilibrium is reached, it may be possible to use
real-time feedback software to minimize the effects of drift [32].
2.6.4 Blockage & Refill Detection
There have been a slew of recent recalls of commercially available implantable pumps due to issues of
stalls and malfunctions that lead to dumping of the entire drug load [33-35]. Unfortunately, without the
inclusion of sensors, such errors could go unnoticed until adverse physiological symptoms manifest and are
106

reported by the patient. The EI dose tracking system is capable of tracking a large range of doses and
therefore, any unintended delivery or lack of delivery can be tracked in real-time. However, unlike
conventional implantable pumps, the electrolysis actuator can be turned on/off and is not pre-pressurized;
therefore, risk of accidental release of the reservoir content into the body is greatly minimized.
Blockages in the catheter, on the other hand, need to be detected and reported. 5 mA current was applied
to an actuator for 25 seconds and the delivery of DD water from the adjacent reservoir was tracked using a
100 mV pp signal at 1 kHz. A blockage in the catheter was simulated by closing PEEK shut off valve
immediately after 17 seconds. In case of blockage in the catheter, fluid volume in the drug chamber does
not change even though the actuator is actively pumping. In this situation, the slope of the impedance
magnitude read-out leveled off (Figure 3-17).

Figure 3-17: Real-time blockage detection using EI sensing.
As the implantable pump is refillable through the skin, it would also be useful to be able to track the
volume of drug injected into the reservoir, to ensure a successful refill. Impedance magnitude changes
during the injection of 2 mL of DD water into an empty reservoir were recorded using a 100 mV pp signal
at 1 kHz. During the refill process, as expected, as volume was added to the reservoir, the slope of the
impedance readout became negative (Figure 3-18).
107

Figure 3-18: Tracking fluid refill into the reservoir using EI sensing.
CHAPTER 3-2: WIRELESS SENSING

1 OMINIDIRECTIONAL WIRELESS DATA TRANSFER
For wired benchtop experiments, the excitation signal application and impedance measurement were
provided by a LabVIEW-interfaced precision LCR meter directly connected to the EI sensors. For
implantable pumps, wireless power and data transmission is preferred to eliminate transcutaneous wires
and catheters which reduces surgical complexity, permits improved patient mobility, and allows drug
administration outside of clinical settings. Omnidirectional data transfer through the inductive powering
link would allow for sensor operation and extraction of sensor data to an external module. Small size and
weight, low power consumption, wireless operation and long usable lifetime are critically important
requirements for implantable drug delivery systems [36]. The sensor feedback system must also adhere to
these requirements.
For transfer through the skin towards the sensors, the current driving the power transmitter could be
manipulated through amplitude modulation (ASK) [37, 38]. For data transfer from the sensors to the
external modules two possible approaches are considered. The first method would be to use frequency
modulation data transfer. In this method, the sensor reading, in our case impedance, is converted to a
108

frequency via an oscillator. This oscillation frequency controls a modulation switch, in parallel to the
implanted inductor. As a result, the load on the inductor changes with the frequency and a modulated signal
is generated and transmitted wirelessly to the external inductor. This signal can be extracted by an envelope
detector and fed through a band-pass filter to amplify the signal and suppress the high frequency carrier.
The output signal can be read from an oscilloscope or a spectrum analyzer. Using a calibrated curve, the
sensor signal is extracted (Figure 3-19) [38]. In this system, the oscillator would have to be powered. This
power can be derived from the inductive power supplied to the electrolysis actuator. The system resolution
is a function of resolution bandwidth of the signal source analyzer. This method has been used in wireless
sensing platforms in [39-41].

Figure 3-19: Block diagram of frequency modulation circuit to enable sensing data transfer.
An alternative method is to use an impedance measurement integrated circuit chip (includes a signal
generator, high-speed analog-to-digital converter (ADC), fast Fourier transform (FFT) analyzer, high-speed
digital-to-analog converter (DAC), and low-pass filter), microprocessor for local computing, and a
telemetry protocol for data transmission (e.g. Bluetooth, ZigBit, XBee, or a wireless transmitting module
such as TX3A) (Figure 3-20) [38]. This technique has been used in several wireless sensing platforms such
as [42-44]. This method is power hungry, and often times, requires batteries on the implanted side. Extra
circuitry could be added to develop a “wake up” function to minimize power consumption. While, the
microprocessor allows additional decision making functionalities (controlling the device, managing
communications, on-chip data storage) that may be required of the implant, the required components tend
109

to be bulky; together these may increase the footprint and weight of the implant which is undesirable [37,
45]. Therefore, frequency modulation data transfer using an LC voltage controlled oscillator (LCVCO) was
chosen to minimize size, weight, and power consumption.

Figure 3-20: Block diagram of circuit using an impedance measurement integrated circuit chip to enable sensing data
transfer.
2 RESEARCH DESIGN & METHODS
2.1 SYSTEM ARCHITECTURE - 1
ASK modulation was used for data transfer from the external transmitter to the internal receiver circuit.
A lower frequency sweep signal (200-900 kHz) for sensing was carried by the 2 MHz power signal. The
power signal was rectified on the receiver and used to power implanted circuit components, as well as the
actuator of the electrolysis micropump. The frequency of the sweep signal was chosen to be higher than f m,
so that impedance response is always dominated by the solution resistance (drug in the reservoir). Changes
in the solution resistance altered the voltage across an n-channel metal-oxide-semiconductor field-effect
transistors (n-MOSFET), changing its capacitance, and therefore, the resonance frequency of the internal
transmitting coil. Changes in the resonance frequency were reflected in the external receiving coil. The
output signal was read from an oscilloscope or recorded via a LabVIEW-interfaced data acquisition system
(Figure 3-21). Shifts in frequency were correlated to the sensors’ measured impedance value and the volume
of fluid remaining in the reservoir.
110

Figure 3-21: System architecture (1) of wireless sensing circuit.
Circuit Design and Layout: External Transmitter: A 2 MHz clock oscillator (ECS -2100, ECS
international, Olathe, KS), along with a quad bilateral switch (CD4016BC, Fairchild Semiconductor, San
Jose CA) controlled by a resistor-set oscillator (LTC 6906, Linear Technologies, Milapitas, CA), were used
to generate the power and sensing signals, respectively. A 250 kΩ dual 1024-position digital potentiometer
(AD 5235, Analog Devices, Norwood, MA) controlled by LabVIEW-interfaced SPI program module (NI
USB-8451, National Instruments, Austin, TX) was used to set the various frequencies required for the
sensing frequency sweep (200 – 900 kHz). The generated signal was then amplified in two stages before
being applied to a tuned transmitting coil (8 turns of 20 AWG single strand wire, size: 310 mm × 140 mm).
Internal Transmitter & Receiver: Litz wire (6 turns, 50/54 SPN/SN Litz Wire, Wiretron, Volcano, CA)
was used for the receiving coil (Ø 22 mm). The flat internal transmitting coil (L = 110 µH at 500 kHz. Ø
50 mm) was also fabricated using Litz wire (50/54 SPN/SN Litz Wire, Wiretron, Volcano, CA). The circuit
components and the electrolysis micropump require a direct current (DC) power signal. Therefore, the
received alternating signal was fully rectified using two Schottky diodes (BAT54A and BAT54C, Fairchild
Semiconductor, San Jose, CA). The modulation signal was separated by half-wave rectification (BAT54A)
111

and applied to the sensors. An n-channel MOSFET (Si8424CDB, Vishay Siliconix, Singapore) was used
as a voltage controlled variable capacitor and placed in parallel with the sensors. A p-channel MOSFET
(Si8429DB, Vishay Siliconix, Singapore) and a BJT transistor (2N4401, Fairchild Semiconductor, San
Jose, CA), were used in a source-follower buffer configuration to maintain constant current through the
sensors (~ 100 µA).
External Receiver: The signal received with the external receiving coil (8 turns of 20 AWG single strand
wire, size: Ø 130 mm), was rectified using two Schottky diodes (BAT54A and BAT54C, Fairchild
Semiconductor, San Jose, CA). An analog input module (NI 920532, National Instruments, Austin, TX),
along with the CompactDAQ System (National Instruments, Austin, TX) and a LabVIEW interface were
used to record and display the rectified signal.
Preliminary Experiments: The external and internal transmitter and receiver circuits were first tested using
discrete components on breadboards. Preliminary benchtop testing using “dummy” loads in place of the
sensors was performed to ensure feasibility of the approach. The circuit performance was then tested with
sensors integrated in the micropump.
Testing with “Dummy” Loads: A range of voltages (V DS = 0 - 2 V) were applied across the n-MOSFET
serving as the variable capacitor in order to confirm the linear response range provided by the manufacturer’
data sheet. Capacitance varied linearly with applied voltage (2.5 – 1 nF for V DS = 0 – 2 V respectively, data
not shown). Based on these values, a constant current of 100 µA was applied to the sensors.
The external and internal transmitter and receiver circuits were implemented using discrete components
on breadboards. A 500 kΩ mechanical potentiometer was used instead of the digital potentiometer. Three
resistors (1.74, 3.24, and 12.4 kΩ) were used in place of the sensors. The transmitted sensing signal was
swept from 220 – 1000 kHz and the rectified received signal was recorded. A clear shift in resonance
frequency was observed for varying resistances used in place of the sensors (Figure 3-22). The increase in
resonance frequency corresponds to a drop in the n-MOSFET’s capacitance value, as a result of an increase
in the applied voltage.
112

Figure 3-22: Normalized received voltage values vs. frequency sweep for different resistors used in place of the
sensors (the values for each trace are normalized to the voltage measured at the resonance frequency).
Testing with Infusion Pump: Sensors integrated in an infusion pump with a 1 mL reservoir were connected
to the internal circuit. The micropump reservoir was filled with 1× PBS as a model drug. The resonance
frequency was determined before and after delivery of two 80 µL boluses using the micropump.
The results of wireless sensing with the micropump are presented in Figure 3-23. As expected, the
resonance frequency increases after each 80 µL dosing event. Delivery of each dose, leads to a rise in fluid
resistance measured by the sensors. Since the current through the sensors is kept constant, increased
resistance increases the voltage across the n-MOSFET used as a variable capacitor. The rise in voltage in
turn decreases the n-MOSFET’s capacitance value, resulting in an increase in the resonance frequency of
internal transmitting coil (~ 110 kHz shift), which is then reflected on the external receiving coil.
113

Figure 3-23: Wireless sensing with sensors integrated in an infusion pump. Each 80 µL dose leads to ~ 110 kHz shift
in the resonance frequency of the internal transmitting coil.
2.1.1 System Architecture - 2
The architecture used for the above experiments is limited by the resolution of the frequency sweep,
which in turn is limited by the step changes in the potentiometer resistance that sets the output frequency.
To alleviate this limitation, another architecture was designed and implemented to achieve wireless sensing.
Once again, ASK modulation was used for data transfer from the external transmitter to the internal
receiver circuit. A fixed 500 kHz sensing signal was carried by the power signal through skin to the implant.
Once picked up by the internal receiving coil, the signal is fully rectified, fed through a current regulator,
and applied to power the actuator. The received signal is also applied to the dosing sensors. As the current
through the sensors is held constant, changes in sensor impedance result is a change in voltage across an n-
MOSFET, causing a change in its capacitance. As a result, the resonance frequency of the implanted
transmitted coil shifts. This shift is reflected on the external receiving coil. This shift in frequency was
measured externally and correlated to changes in the reservoir fluid volume. To further improve the
sampling speed and signal quality, the externally received signal was amplified and multiplied by the
original transmitted signal (Figure 3-24).
114

Figure 3-24: System architecture (2) of wireless sensing circuit (reprinted with permission from [46] © 2015 IEEE).
Circuit Design and Layout: External Transmitter: A 2 MHz clock oscillator (ECS-2100, ECS
international, Olathe, KS), along with a quad bilateral switch (CD4016BC, Fairchild Semiconductor, San
Jose CA) controlled by a resistor-set oscillator (LTC6906, Linear Technologies, Milapitas, CA), were used
to generate the power and sensing signals (500 kHz), respectively. The generated signal was then amplified
in two stages before being applied to a tuned transmitting coil (8 turns of 20 AWG single strand wire, size:
310 mm x 140 mm).
Internal Transmitter & Receiver: Litz wire (6 turns, 50/54 SPN/SN Litz Wire, Wiretron, Volcano, CA)
was used for the receiving coil (Ø 22 mm). The flat internal transmitting coil (L = 110 µH at 500 kHz. Ø
50 mm) was also fabricated using Litz wire (50/54 SPN/SN Litz Wire, Wiretron, Volcano, CA). The circuit
components and the electrolysis micropump require a direct current (DC) power signal. Therefore, the
received alternating signal was fully rectified using two Schottky diodes (BAT54A and BAT54C, Fairchild
Semiconductor, San Jose, CA). The modulation signal was separated by half-wave rectification (BAT54A)
and applied to the sensors. An n-channel MOSFET (Si8424CDB, Vishay Siliconix, Singapore) was used
as a voltage controlled variable capacitor and placed in parallel with the sensors. A p-channel MOSFET
(Si8429DB, Vishay Siliconix, Singapore) and a BJT transistor (2N4401, Fairchild Semiconductor, San
Jose, CA), were used, in a source-follower buffer configuration, to maintain constant current through the
sensors (~ 100 µA) (Figure 3-25).
115

Figure 3-25: Schematic diagram of the internal (transmitter and receiver) circuit.
External Receiver: The signal received with the external receiving coil (8 turns of 20 AWG single strand
wire, Ø 50 mm), first filtered using a low pass elliptic filter (f cutoff = 1 MHz; LTC1560-1, Linear
Technologies, Milapitas, CA) and then amplified using an LM7171 operational amplifier (Texas
Instruments, Dallas, TX). The resulting signal was multiplied by the original 500 kHz transmitted signal
(AD835, Analog Devices, Norwood, MA). An analog input high speed data acquisition module (2
MSamples/s; NI USB-6366, National Instruments, Austin, TX), and a LabVIEW interface were used to
record, display, and analyze the resulting signal.
2.1.2 Preliminary Testing with the Micropump

Figure 3-26: Photograph of sensors integrated in a micropump with a 1 mL reservoir. Inset photograph shows the
sensors placed vertically in the reservoir wall.
116

Sensors integrated in an infusion pump with a 1 mL reservoir were connected to the internal circuit
(Figure 3-26). The micropump reservoir was filled with 1× PBS as a model drug. System performance was
evaluated during the delivery of identical and differing bolus volumes of PBS. The final multiplied signal
was recorded prior and after bolus delivery. For each run, four boluses were delivered (3 mA current applied
for 90 seconds) totaling 150 ± 3.5 µL (~83% of the actuator deliverable volume). A repeatable non-linear
response was observed between the volume remaining in the reservoir and the recorded multiplication
results (Figure 3-27).

Figure 3-27: Wireless sensing with the infusion pump: each 37.5 ± 0.75 μL boluses (n =3, Mean ± SE; 3.75 % of
reservoir fill volume) leads to a non-linear shift in the resonance frequency of the internal transmitting coil (sensor
response values for each trace are normalized to multiplication result measured for the full reservoir; reprinted with
permission from [46] © 2015 IEEE).
Different bolus volumes were delivered by changing the current applied to the pump actuator (5, 1, 5,
and 3 mA successively for 20, 180, 20, and 90 seconds, respectively; Figure 3-28). The slope of the
accumulated volume over time correlated to changes in the output flow rate; this, in turn, was reflected in
the sensor response.
117

Figure 3-28: (a) pump performance and (b) sensor response to changes in applied current to the microactuator
leading to a change in the delivery flow rate (n=3, mean ± SE; sensor response values for each trace are normalized
to multiplication result measured for the full reservoir; reprinted with permission from [46] © 2015 IEEE).
Lastly, four 17.75 ± 0.48 µL boluses of lidocaine HCl in saline (20 mg/mL) were delivered using the
micropump (20 sec on, 1 min off; Figure 3-29). A slight drift in measured impedance leads to a dip in the
sensor response during the off periods. The drift amount is not constant and does not follow a predetermined
trend [10].

Figure 3-29: Delivery of 17.75 boluses (1.78 volume) of lidocaine HCl dissolved in saline (20mg/mL). Shaded areas
indicate 20 second on, followed by 1 min off. (Sensor response values for each trace are normalized to multiplication
result measured for the full reservoir; reprinted with permission from [46] © 2015 IEEE).
2.1.3 Calibration Testing
It is important to note that for wireless sensing the signal is applied across two of the sensing electrodes
and the third electrode is grounded. During preliminary testing with the micropump, it was concluded that
the third electrode (grounded) does not act as a reference for the electrochemical cell, and as a result does
118

not offer an advantage compared to a two electrode cell. Therefore, for subsequent devices, only two
sensing electrodes were used. The electrodes were placed directly across from one another in the drug
reservoir.

Figure 3-30: Photograph of (a) micropump with integrated dosing sensors, (b) system setup (reprinted with
permission from [46] © 2015 IEEE).
Three micropumps with integrated dosing sensors and 1 mL reservoirs were fabricated and connected
to the internal circuit (Figure 3-30). The micropump reservoirs were filled with 1× PBS as a model drug.
For each run, seven boluses (~14.67 µL per bolus) were delivered (2 mA current applied for ~50 seconds).
The volume dispensed from the pump, was calculated based on fluid front movement in a calibrated 100
µL micropipette attached to the micropump catheter outlet. The final multiplied signal was recorded prior
and after each bolus delivery and normalized to the baseline value for each run. Four runs were performed
for each micropump (Figure 3-31) and the results used to obtain averaged trends.
119

Figure 3-31: Calibration testing for wireless sensing with three infusion pumps: seven boluses, each 14.67 µL (1.47 %
of reservoir fill volume) delivered for each of the four runs (sensor response values for each trace are normalized to
multiplication result measured for the full reservoir).

The following relationships exist between the parameters used for wireless sensing (3-5, 3-6, 3-7, 3-8):
∆ /
∝∆#
123145
(3-5)
∆#
123145
∝ ∆/
6789:;
(3-6)
∆/
6789:;
∝
-∆<
=>*?@A
(3-7)

B
C∆<
=>*?@A
∝ ∆D
521
∝∆EE E E (3-8)
Using these relationships, a linear relationship can be obtained between the volume delivered from the
micropump and the sensor response (3-9, Figure 3-32):
∆ /
∝ −
B
∆123145 521F4312

G
(3-9)

120

Figure 3-32: Averaged results from calibration tests are used to obtain a linear relationship between the volume
delivered by the micropump and the sensor response.
The averaged results from the calibration tests were used to obtain a linear relationship and calibration
curve between the volume delivered by the micropump and the sensor response. This curve was used to
program the LabVIEW graphical user interface to estimate the volume delivered by the micropump with
an accuracy of ±10%.
2.1.4 Wireless Sensing and Real-time Flow Variation
The robustness of the calibration curve was first tested by delivering seven consecutive boluses (~14.67
µL per bolus; 2 mA current applied for ~50 seconds). The volume dispensed from the pump, was calculated
based on fluid front movement in a calibrated 100 µL micropipette attached to the pump’s catheter outlet.
The sensor response was recorded and analyzed using the LabVIEW graphical user interface and an
estimation of the presented to the user. The results are shown in Figure 3-33.
121

Figure 3-33: Calibration curve robustness: the volume delivered by the micropump vs. the volume estimated by the
sensor response.
Real-time flow variations and simultaneous wireless dose sensing for a micropump with integrated
dosing sensors and 1 mL reservoir connected to the internal circuit was performed. The micropump
reservoir was filled with lidocaine HCl in saline (20 mg/mL). 5, 1, 3, 0.5, and then 2 mA were applied
consecutively for 20, 100, 120, 60, and 40 seconds respectively to the electrolysis electrodes. The sensor
response was recorded before and after each delivery and analyzed using the LabVIEW graphical user
interface and an estimation of the delivered volume based on the calibration curve was presented to the
user. The results are presented in Figure 3-34 comparing the programmed volume (based on the current
input and time of delivery), the sensed volume (based on analysis of the sensor response), and the actual
delivered volume as observed from fluid movement in a calibrated 100 µL micropipette connected to the
micropump catheter outlet. The smallest volume delivered during this test was 3 µL boluses which
constitutes ~1.67% of the micropump’s deliverable volume and ~0.3% of the reservoir’s fill volume.
122

Figure 3-34:Real-time flow variation and simultaneous wireless sensing.
3 CONCLUSION
A fully integrated electrochemically-based dose tracking system for closed-loop fluid drug delivery
capable of real-time volume delivery tracking and confirmation was demonstrated. Electrochemical dose
tracking is attractive for its simplicity, sensitivity, and wide compatibility and this method could easily be
adapted to other pumping methods. In addition the method requires little electrical power. Two sets of
electrodes made from thin film and bulk wire electrodes were fabricated and their performance compared.
Drift and noise in measurement were also studied. Less drift in measurement is observed for fluids with
higher ionic concentration, at higher temperatures, and also when the counter and reference electrodes were
de-coupled by using a 3 electrode configuration. Water and lipid soluble drug analogs were studied to
calibrate the system. Tracking bolus resolution of 83 nL for DD water (0.003% of the reservoir) and 556
nL for 1× PBS (0.019% of the reservoir) were shown. Real-time blockage and refill detection were also
presented. Wireless sensing was achieved through combining ASK and frequency modulation for data
transfer to and from the sensors. Simultaneous real-time flow variation and wireless dose sensing was
123

demonstrated. The smallest volume delivered during this test was 3 µL boluses (~1.67% of the micropump’s
deliverable volume and ~0.3% of the reservoir’s fill volume).
4 REFERENCES
[1] H. L. Helmholtz, "Studien uber electricshe grenzschichten," Annalen der Physic und Chemie, vol. 7,
1879.
[2] L. S. Robblee and T. L. Rose, "The electrochemistry of electrical stimulation," in Annual International
Conference of the IEEE on Engineering in Medicine and Biological Society, 1990, pp. 1479-1480.
[3] L. L. Faulkner and A. J. Bard, Electrochemical methods and applications, 2nd ed. New York: Wiley,
2001.
[4] M. E. Orazem, Electrochemical impedance spectroscopy : selection of papers from the 6th
international symposium (EIS 2004), 16 - 21 May 2004, Cocoa Beach, FL, USA. Amsterdam Elsevier,
2006.
[5] C. A. Gutierrez, "Development of flexible polymer-based MEMS technologies for integrated
mechanical sensing in neuroprosthetic systems," University of Southern California, Los Angeles,
2012.
[6] C. A. Gutierrez and E. Meng, "Improved self-sealing liquid encapsulation in parylene structures by
integrated stackable annular-plate stiction valve," in Micro Electro Mechanical Systems (MEMS),
2010 IEEE 23rd International Conference on, 2010, pp. 524-527.
[7] C. A. Gutierrez, R. Sheybani, and E. Meng, "Electrochemically-based dose measurement for closed-
loop drug delivery applications," in 16th International Solid-State Sensors, Actuators and
Microsystems Conference, Transducers '11, Beijing, China, 2011, pp. 2839-42.
[8] M. S. E. Diab, "Electrochemical applications of modified electrodes in wastewater treatment and
energy conversion systems," in Applied electrochemistry, V. Singh, Ed., ed: Nova Publishers, 2011.
[9] H. Xi, R. W. Pascal, K. Chamberlain, C. J. Banks, M. Mowlem, and H. Morgan, "A Miniature, High
Precision Conductivity and Temperature Sensor System for Ocean Monitoring," IEEE Sensors
Journal, vol. 11, pp. 3246-52, 2011.
[10] N. S. Kaisare, V. Ramani, K. Pushpavanam, and S. Ramanathan, "An analysis of drifts and
nonlinearities in electrochemical impedance spectra," Electrochimica Acta, vol. 56, pp. 7467-7475,
2011.
[11] R. Sheybani, N. E. Cabrera-Munoz, T. Sanchez, and E. Meng, "Design, fabrication, and
characterization of an electrochemically-based dose tracking system for closed-loop drug delivery,"
in Engineering in Medicine and Biology Society (EMBC), 2012 Annual International Conference of
the IEEE, 2012, pp. 519-522.
[12] S. Bohm, B. Timmer, W. Olthuis, and P. Bergveld, "Closed-loop controlled electrochemically actuated
micro-dosing system," Journal of Micromechanics and Microengineering, vol. 10, pp. 498-504, 2000.
[13] Research Solutions and Resources. (2009). Inductance and Inductive Loops. Available:
http://www.consultrsr.com/resources/eis/induct1.htm
[14] A. N. Frumkin and L. I. Krishtalik, Hydrogen Overvoltage and Adsorption Phenomena: Wiley, 1961.
[15] R. Sheybani, H. Gensler, and E. Meng, "A MEMS electrochemical bellows actuator for fluid metering
applications," Biomedical Microdevices, pp. 1-12, 2012/07/01 2012.
[16] R.Sheybani, S. Elyahoodayan, and E. Meng, "Closed-loop On-demand Drug Delivery Micropump for
Chronic Pain Management Applications," in 7th International Conference on Microtechnologies in
Medicine and Biology (MMB '13), Marina Del Rey, CA, 2013.
[17] A. Adam, A. Adam, R. F. Dondelinger, and P. R. Mueller, Interventional Radiology in Cancer:
Springer, 2004.
124

[18] T. Judendorfer, A. Pirker, and M. Muhr, "Conductivity measurements of electrical insulating oils," in
Dielectric Liquids (ICDL), 2011 IEEE International Conference on, 2011, pp. 1-4.
[19] J. Lukanov and P. Atmadjov, "Investigating the effect of silver ions on the contractile function of
smooth-muscle preparations from guinea pig stomach, in vitro," Folia medica, vol. 21, pp. 11-19,
1978.
[20] W. F. Jackson and B. R. Duling, "Toxic effects of silver-silver chloride electrodes on vascular smooth
muscle," Circulation research, vol. 53, pp. 105-108, 1983.
[21] K. R. Cote and R. C. Gill, "Development of a platinized platinum/iridium electrode for use in vitro,"
Annals of Biomedical Engineering, vol. 15, pp. 419-26, 1987.
[22] W. Plieth, Electrochemistry for Materials Science, 1 ed.: Elsevier Science, 2008.
[23] D. Pletcher and H. Saxton, A First Course in Electrode Processes, 2 ed. Cambridge, UK: The Royal
Society of Chemistry, 2009.
[24] B. Freeman, "Introduction to Membrane Science and Technology. By Heinrich Strathmann,"
Angewandte Chemie International Edition, vol. 51, pp. 9485-9485, 2012.
[25] D. R. Merrill, M. Bikson, and J. G. Jefferys, "Electrical stimulation of excitable tissue: design of
efficacious and safe protocols," Journal of Neuroscience Methods, vol. 141, pp. 171-198, 2005.
[26] J.-I. Tzeng, K.-I. Cheng, K.-L. Huang, Y.-W. Chen, K.-S. Chu, C.-C. Chu, et al., "The cutaneous
analgesic effect of class I antiarrhythmic drugs," Anesthesia & Analgesia, vol. 104, pp. 955-958, 2007.
[27] K. Broka and P. Ekdunge, "Oxygen and hydrogen permeation properties and water uptake of Nafion
117 membrane and recast film for PEM fuel cell," Journal of Applied Electrochemistry, vol. 27, pp.
117-23, 1997.
[28] H. Gensler, R. Sheybani, P.-Y. Li, R. L. Mann, and E. Meng, "An implantable MEMS micropump
system for drug delivery in small animals," Biomedical Microdevices, vol. 14, pp. 483-496, 2012.
[29] J. Fall, D. Humphreys, and S. Guo, "Design and Testing of a Unitized Regenerative Fuel Cell," Journal
of fuel cell science and technology, vol. 6, 2009.
[30] F. Nores-Pondal, I. Vilella, H. Troiani, M. Granada, S. De Miguel, O. Scelza, et al., "Catalytic activity
vs. size correlation in platinum catalysts of PEM fuel cells prepared on carbon black by different
methods," International Journal of Hydrogen Energy, vol. 34, pp. 8193-8203, 2009.
[31] C. R. Neagu. (1998). A medical microactuator based on an electrochemical principle. Available:
http://doc.utwente.nl/13878
[32] J.-F. Cheng, J.-C. Chou, T.-P. Sun, S.-K. Hsiung, and H.-L. Kao, "New calibration methods to
eliminate the non-ideal effect of drift and hysteresis in all-solid-state potassium electrode," IEEE
Sensors Journal, vol. 11, pp. 1263-1273, 2011.
[33] FDA: US Food and Drug Administration. (2011). SIGMA Spectrum Infusion Pump Model 35700 -
Expanded Recall. Available:
http://www.fda.gov/MedicalDevices/Safety/ListofRecalls/ucm308488.htm
[34] FDA: US Food and Drug Administration. (2011). Medtronic Model 8637 SynchroMed II Implantable
Infusion Pump. Available: http://www.fda.gov/MedicalDevices/Safety/ListofRecalls/ucm271492.htm
[35] FDA: US Food and Drug Administration. (2011, 5/10/2013). Roche Insulin Delivery Systems, ACCU-
CHEK FlexLink Plus Infusion Sets. Available:
http://www.fda.gov/MedicalDevices/Safety/ListofRecalls/ucm248772.htm
[36] D. J. H. Tng, R. Hu, P. Song, I. Roy, and K. T. Yong, "Approaches and Challenges of Engineering
Implantable Microelectromechanical Systems (MEMS) Drug Delivery Systems for in Vitro and in
Vivo Applications," Micromachines, vol. 3, pp. 615-631, 2012.
[37] B. Lenaerts, Omnidirectional inductive powering for biomedical implants, 1st ed. New York: Springer,
2008.
[38] R. Sheybani, S. M. Schober, and E. Meng, "Drug Delivery Using Wireless MEMS," in Handbook of
MEMS for wireless and mobile applications, Uttamchandani, Ed., ed: Woodhead Publishing, 2013,
pp. 489-517.
[39] T. Ativanichayaphong, J. Wang, W. Huang, S. Rao, and J. Chiao, "A simple wireless batteryless
sensing platform for resistive and capacitive sensors," in Sensors, 2007 IEEE, 2007, pp. 139-142.
125

[40] X. Zhang, J. Yan, B. Vermeire, F. Shadman, and J. Chae, "Passive wireless monitoring of wafer
cleanliness during rinsing of semiconductor wafers," Sensors Journal, IEEE, vol. 10, pp. 1048-1055,
2010.
[41] H. Cao, S. K. Thakar, T. Fu, M. Sheth, M. L. Oseng, V. Landge, et al., "A wireless strain sensor system
for bladder volume monitoring," in Microwave Symposium Digest (MTT), 2011 IEEE MTT-S
International, 2011, pp. 1-4.
[42] R. Tan, T. McClure, C. Lin, D. Jea, F. Dabiri, T. Massey, et al., "Development of a fully implantable
wireless pressure monitoring system," Biomedical microdevices, vol. 11, pp. 259-264, 2009.
[43] S. G. Taylor, K. M. Farinholt, G. Park, C. R. Farrar, E. B. Flynn, D. L. Mascarenas, et al., "Wireless
impedance device for electromechanical impedance sensing and low-frequency vibration data
acquisition," in SPIE Smart Structures and Materials+ Nondestructive Evaluation and Health
Monitoring, 2009, pp. 729228-729228-12.
[44] S. S. Lee, A. Kim, G. Chitnis, C. R. Powell, and B. Ziaie, "A Modular Embedded System Design for
Implantable Wireless Bladder Pressure Sensing," in International Conference on Microtechnologies
in Medicine and Biology, Marina del Rey, California, USA, 2013.
[45] D. Panescu, "Wireless communication systems for implantable medical devices," IEEE Engineering
in Medicine and Biology Magazine, vol. 27, pp. 96-101, 2008.
[46] R. Sheybani and E. Meng, "A Wireless Implantable Drug Infusion System with Integrated Dosing
Sensors," in The 18th International Conference on Solid-state Sensors, Actuators, and Microsystems
(Transducers '15), Anchorage, AK, USA, 2015 (accepted).

126

CHAPTER 4: WIRELESS ELECTROCHEMICAL DRUG DELIVERY
MICROPUMP WITH FULLY INTEGRATED ELECTROCHEMICAL DOSE
TRACKING FEEDBACK SYSTEM
The fully integrated wireless system combines the electrolysis-based micropump with the
electrochemical dose tracking system presented in chapters 2 and 3, respectively (Figure 4-1).
The micropump consists of an electrolysis based actuator
housed in a drug reservoir. Electrochemical pumping is
achieved by wireless inductive power transfer resulting in
constant current application to a pair of Nafion
®
coated
interdigitated Pt electrodes which converts water into hydrogen
and oxygen gases. The resulting volume expansion is then
harnessed to inflate the drug separating bellows which in turn
displaces the fluid in the drug chamber and expels drug from
the catheter to the delivery site. Once the current application
is ceased, the Pt electrodes catalyze the recombination of gases
into water, enabling repeatable pumping. A pair of Pt wire
segments coated in pyrolyzed Nafion may be used to facilitate
and accelerate gas recombination. A normally closed check
valve is required to prevent backflow of fluids as a result of the reverse pressure gradient caused by
recombination (not included in this work). A refill port is added to the drug reservoir to allow for
transcutaneous refilling of drug after implantation. Amplitude modulation is utilized to allow for wireless
adjustment of the infusion rate post-implantation.
Electrochemical dose tracking is achieved through measuring electrochemical impedance (EI) by
applying a small sinusoidal excitation voltage across a pair of electrodes placed in the drug reservoir. At
sufficiently high frequencies, the measured impedance could be directly correlated to the volume of drug
Figure 4-1: Schematic diagram of system.
127

remaining in the reservoir. Therefore, changes in reservoir content, delivery flow rate, blockages in the
delivery catheter, drug refills, and even damage to the electrolysis chamber could be assessed in real-time
and recorded for future analysis using A LabVIEW graphical user interface (GUI). Amplitude and resistive
modulation data transfer are utilized to achieve bi-directional data telemetry between the dosing sensors
and the external module.
Power application, current control, and sensing data analysis, is initiated using Bluetooth through an
off-the-shelf USB carrier board kit, microcontroller, and an Easy Bluetooth Module.
1. SYSTEM DESIGN, FABRICATION & ASSEMBLY
The research prototype presented in this work was scaled for small animals (rodents) which are a widely
used animal model in drug discovery and development (Figure 4-2). The design targets use of the implanted
infusion system in an animal housed inside a typical vivarium cage set that is situated inside the external
transmitting coil. The external receiving coil is located on the cage wall perpendicular to the external
transmitting coil.

Figure 4-2: Schematic diagram of system setup for small animal research.
1.1 MICROPUMP FABRICATION
An adult male mouse and rat weigh approximately 30 and 400 g, respectively [1]. As such the
implantable prototype should weigh <10% of the animal’s body weight (3 g for use in mice and 40 g for
128

use in rats). The bellows actuator can only displace the column of fluid directly above it; as such the fluid
around the bellows (between bellows outer diameter and reservoir inner wall) cannot be accessed for
pumping and is considered dead volume which should be minimized. By designing a reservoir with an inner
diameter close to that of the bellows outer diameter, the dead volume between is less than 100 µL [2, 3].
As extensively detailed in [3], new reservoir packaging was designed and fabricated from WaterShed®
XC 11122, an optically clear, water resistant material similar to the common thermoplastic acrylonitrile
butadiene styrene (ABS), using stereolithography (FineLine Prototyping, Inc., Raleigh, NC), a high-
resolution rapid prototyping technique. Two vertical indentations and access ports were incorporated in the
reservoir wall to allow for the integration of dosing sensors. The new reservoir featured two silicone-filled
refill ports (10:1 base-to-curing agent ratio Class VI MDX-4 4210; Factor II, Lakeside, AZ) instead of one
to facilitate refill in the valved configuration of the system, as well as a silicone catheter (ID. 1.016 mm,
OD 2.159 mm, VWRbrand Select Silicone Tubing, VWR International, Radnor, PA).
Other actuator and sensor components were fabricated as described in chapters 2 and 3. Briefly,
Interdigitated Pt electrodes (100 µm wide elements separated by 100 µm gaps, 8 mm diameter footprint)
were fabricated on polyetheretherketone (PEEK) sheets (thickness of 0.5 mm, CS Hyde, Lake Villa, IL) by
a liftoff method. Electrodes fabricated on PEEK were chosen due to reduce overall weight (0.112 vs. 0.260
g) and improve electrolysis efficiency (an average of 8.15%) compared to electrodes fabricated on
borosilicate glass. Individual electrodes were separated and dip coated with Nafion
®
(Dupont DE521
Solution, Ion Power, INC, New Castle, DE) twice. Kynar™ silver plated copper wires (30 AWG, Jameco
Electronics, Belmont, CA) were soldered to contact pads on the electrodes (Figure 4-3c). The joint was
strengthened and insulated with nonconductive marine epoxy (Loctite, Westlake, OH) [4]. Parylene bellows
(2 convolutions; 10 mm outer diameter, 6 mm inner diameter) were fabricated as described in [5] using a
mold formed from stacked silicone rubber sheets (10:1 base-to-curing agent ratio Sylgard 184, Dow
Corning, Midland, MI) and molten (~50°C) polyethylene glycol (PEG; 1,000 Mn, Sigma Aldrich, St. Louis,
MO). A 13.5 µm layer of Parylene C (Specialty Coatings Systems, Indianapolis, IN) was deposited over
129

the PEG mold, and PEG was dissolved by soaking in water at room temperature to complete the bellows.
Electrolysis actuators were assembled by filling the bellows with double distilled (DD) water and carefully
mounting with the Nafion
®
-coated interdigitated Pt electrodes using laser cut double-sided pressure
sensitive adhesive film (3M™ Double Coated Tape 415, 3M, St. Paul, MN). The seal was reinforced with
marine epoxy (Loctite, Westlake, OH) [6]. In order to increase the rate of recombination, a pair of 3 mm
segments of 99.9% Pt wire (Ø 0.5 mm) (California Fine Wire, Grover Beach, CA) were selected. The edges
of the wire were smoothed using 220 grit silicon carbide sandpaper. The pieces were then coated with
Nafion® and heat treated at 320 °C under N 2 backflow for an hour, then slowly cooled to room temperature.
The pyrolyzed- Nafion®-coated pieces were suspended in the DD water filled bellows, before the bellows
were affixed to electrodes [7].

Figure 4-3: Photographs of micropump assembly: (a) dosing sensors incorporated into the reservoir wall across from
each other, (b) vertically oriented sensors bent through the access ports to allow connection with the wireless circuit,
(c) electrolysis electrode fabricated on PEEK substrate, (d) assembled micropump ready for the incorporation of
circuit and coils (scale bars represent 4 mm).
Dosing sensors were fabricated from 99.9% Pt wire (Ø 0.5 mm) (California Fine Wire, Grover Beach,
CA). A 2 mm segment of the tip was exposed and then sanded (60 and 220 grit silicon carbide sandpaper)
to increase surface area. The sensors were placed inside the vertical reservoir wall indentations, bent
130

through the access ports, and soldered to Kynar™ silver plated copper wires (30 AWG, Jameco Electronics,
Belmont, CA) to establish the electrical connection with the wireless circuit. Once in place, marine epoxy
was used to create a water-resistant seal (Figure 4-3a, b).
The actuator was then incorporated into the reservoir using biocompatible epoxy (EPO-TEK® 730
unfilled, Epoxy Technologies, Billerica, MA) (Figure 4-3d). This assembly could be encapsulated with 5
µm Parylene C to further improve biocompatibility and barrier properties.
1.2 WIRELESS POWER & SENSING SYSTEM
1.2.1 System Architecture

Figure 4-4: System architecture of wireless power and sensing system.
The overall wireless system architecture is presented in Figure 4-4. A class E inductive powering scheme
was used to power the micropump as well as the internal circuit components. Once picked up by the internal
receiving coil, the signal was fully rectified, fed through a current regulator, and applied to power the
actuator. For wireless on-demand control of the infusion rate, amplitude shift keying (ASK) modulation
was used to transmit a data signal from the external transmitter to the internal circuit, changing the wiper
position of a current setting potentiometer, which in turn altered the current output of a regulator supplying
current to the actuator. Pulse width of the data signal determined the number of incremental steps of
potentiometer wiper position. ASK modulation was also used for data transfer for dose sensing. A fixed
131

500 kHz sensing signal was carried by the power signal through air (and skin when implanted) to the internal
circuit. The received signal was applied to the dosing sensors. As the current through the sensors was held
constant, changes in sensor impedance resulted in a change in voltage across an n-channel metal–oxide–
semiconductor field-effect transistor (n-MOSFET), causing a change in its capacitance. As a result, the
resonance frequency of the implanted transmitted coil shifted. This shift was reflected on the external
receiving coil. The shift in frequency was measured externally and correlated to changes in the reservoir
fluid volume. To further improve the sampling speed and signal quality, the externally received signal was
amplified and multiplied by the original transmitted signal.
A Parallax Board of Education USB carrier board kit, a Basic Stamp 2 Module microcontroller, and an
Easy Bluetooth Module (Parallax Inc., Rocklin, CA) were used to control wireless signal transmission from
the external module to the internal circuit via Bluetooth. An analog input high speed data acquisition module
(2 MSamples/s; NI USB-6366, National Instruments, Austin, TX) was used to record the externally
received sensor data. LabVIEW (v 2009, National Instruments, Austin, TX) and Basic Stamp 2 Editor (v.
2.5.2) were used to program the communication and signal processing.
1.2.2 Circuit Design and Layout
External Transceiver (Transmitter & Receiver): A 2 MHz clock oscillator (ECS -2100, ECS international,
Olathe, KS), along with a quad bilateral switch (CD4016BC, Fairchild Semiconductor, San Jose CA)
controlled by a resistor-set oscillator (LTC 6906, Linear Technologies, Milapitas, CA), were used to
generate the power and sensing signals (500 kHz), respectively. The generated signal was then amplified
in two stages before being applied to a tuned transmitting coil (8 turns of 20 AWG single strand wire, size:
310 mm x 140 mm).
The signal received with the external receiving coil (8 turns of 20 AWG single strand wire, Ø 50 mm),
was first filtered using a lowpass elliptic filter (f cutoff = 1 MHz; LTC1560-1, Linear Technologies, Milpitas,
CA) and then amplified using an LM7171 operational amplifier (Texas Instruments, Dallas, TX). The
132

resulting signal was multiplied by the original 500 kHz transmitted signal (AD835, Analog Devices,
Norwood, MA). The multiplier operates with 1 V input signals. A resistive voltage divider was used to
decrease the amplitude of the 500 kHz signal before it was supplied to the multiplier. In order to reduce
loading effects on the transmitted signal, a single inverter gate (SN74LVC1GU04DBVR, Texas
Instruments, Dallas, TX) was used as a buffer between the two stage amplifier and the voltage divider
(Figure 4-5).

Figure 4-5: Schematic diagram of the external transceiver circuit.
Internal Transceiver (Transmitter & Receiver): Litz wire (6 turns, 50/54 SPN/SN Litz Wire, Wiretron,
Volcano, CA) was used for the receiving coil (Ø 22 mm). A ferrite toroid core (OD 10 mm,
B64290L38X830, EPCOS AG, Munich, Germany) was added to the receiving coil to enhance coupling and
133

increase voltage pick up (ferrite cores have been shown to conduct the flux lines through the center of the
receiver and improve the coil coupling [8]). The flat internal transmitting coil (L = 110 µH at 500 kHz. Ø
50 mm) was also fabricated using Litz wire (50/54 SPN/SN Litz Wire, Wiretron, Volcano, CA). It is
important to note that ferrite cores can also be utilized to produce larger inductances in a smaller space [8].
Therefore, the internal transmitting coil can be significantly miniaturized with the use of a ferrite core
(ferrite toroid core, OD 6.3 mm, B64290P37X830, EPCOS AG, Munich, Germany). The circuit
components and the electrolysis micropump require a direct current (DC) power signal. Therefore, the
received alternating signal was fully rectified using two Schottky diodes (BAT54A and BAT54C, Fairchild
Semiconductor, San Jose, CA). A zener voltage regulator diode (BZV55, NXP Semiconductors, Eindhoven,
Netherlands) was used to regulate the output voltage of the power signal. AD5227 (Analog Devices,
Norwood, MA) was used as the current setting potentiometer to set the output current of the current
regulator. The PSSI2021SAY (NXP Semiconductors, Eindhoven, Netherlands) adjustable current source
was replaced with LT3092 (Linear Technology, Milpitas, CA). This current source requires less input
voltage (1.2 V compared to 5 V for PSSI2021SAY), allowing for reliable operation for lower received
power levels. The infusion rate modulation signal was separated by half-wave rectification (BAT54A) and
then filtered using a low pass RC design. A low frequency oscillator (LTC 6991, Linear Technology,
Milpitas, CA) set the potentiometer clock frequency to 213 Hz. The LED that was originally included in
series with the infusion pump to provide visual confirmation of power supply to the actuator was removed
due to its high power consumption.
The modulation sensing signal was also separated by half-wave rectification (BAT54A). A zener voltage
regulator diode (BZV55, NXP Semiconductors, Eindhoven, Netherlands) was used to limit the output
voltage of the signal to prevent damage to the MOSFETs. An n-channel MOSFET (Si8424CDB, Vishay
Siliconix, Singapore) was used as a voltage controlled variable capacitor and placed in parallel with the
sensors. A p-channel MOSFET (Si8429DB, Vishay Siliconix, Singapore) and a BJT transistor (2N4401,
Fairchild Semiconductor, San Jose, CA), were used, in a source-follower buffer configuration, to maintain
134

constant current through the sensors (~ 100 µA) (Figure 4-6). The current value is chosen based on the
expected sensor resistance range at the measurement frequency. The n-MOSFET capacitance is altered 2.5
– 1 nF for VDS = 0 – 2 V respectively. The current value should be chosen so that the voltage across the
sensors (and the MOSFET) varies between 0 – 2 V for expected sensor resistance range for the reservoir
deliverable volume. The resistance range is dependent on the ionic conductivity of the drug fluid and the
size of the reservoir.

Figure 4-6: Schematic diagram of the internal transceiver circuit.
The external (Figure 4-7a) and internal transceiver circuits were first tested using discrete components
on breadboards before being implemented on printed circuit boards (PCBs). The receiver was implemented
on a flexible PCB to allow it to be wrapped around the round micropump. Small surface mount components
were chosen to minimize the overall circuit footprint (Figure 4-7b).
135

Figure 4-7: Photograph of (a) external and (b) internal transceiver PCBs: top and bottom image show front and back
of PCB, respectively. Scale bars represent 10 mm.
2. EXPERIMENTAL METHODS & RESULTS
2.1 SYSTEM CALIBRATION
2.1.1 Wireless Flow Control Calibration
A Parallax Board of Education USB carrier board kit, a Basic Stamp 2 Module microcontroller, and an
Easy Bluetooth Module (Parallax Inc., Rocklin, CA) were used to control wireless signal transmission from
the external module to the internal circuit via Bluetooth. LabVIEW (v 2009, National Instruments, Austin,
TX) and Basic Stamp 2 Editor (v. 2.5.2) were utilized to program the communication and signal processing
functions. The communication requires the pulse width of the ASK signal to be >15 mS. Due to this timing
limitation, the frequency of the internal clock was reduced from 500 to 213 Hz. Theoretical output current
values were calculated based on equations provided in the potentiometer (4-1) and current regulator
datasheets (4-2, 4-3):
Δ = ×

+70 Ω (4-1)
where CP is the number of clock pulses.

=5 Ω− Δ ∥750 Ω (4-2)

=

!
"#$
×10 &' (4-3)
136

The program was then tested by using a 1 kΩ load in place of an actuator and calculating the output
current by measuring the voltage across the load. The current was increased in four incremental steps and
the following output current values were achieved: 0.465, 0.485, 0.530, 0.55, and 0.6 mA. The current range
is limited compared to that attained previously for the micropump without the integrated sensors. This can
be attributed to decreased power transfer efficiency (link gain, A, ~ 0.031 compared 0.136 for unmodulated
power signal transmission and coupling coefficient, k, ~ 0.0413 compared 0.136 and 0.18 for unmodulated
power signal transmission, respectively) when the power signal is modulated with the sensing signal and
could potentially be mitigated by re-designing the system to increase power amplification or increase the
operation frequency of the sensing signal.

Figure 4-8: Photograph of micropump with integrated sensors and circuity.
The circuit was then connected to the micropump (Figure 4-8) and the program was used to increase the
output current in incremental steps. 10× PBS was used as the drug model, and the infusion flow rate was
calculated by measuring the fluid front movement in a calibrated 50 µL micropipette connected to the outlet
of the pump (n = 4). Figure 4-9 shows the variable current results with the infusion pump. The measured
flow rate (0.14 – 1.04 µL/min) closely followed the changes in current initiated using the interface program.
137

Figure 4-9: 5 step incremental increase in receiver current output measured across a 1 kΩ resistor and the
micropump (at each modification step, a specific pulse width was applied to achieve the desired change in the wiper
position).
2.1.2 Wireless Dose Sensing Calibration
The micropump was filled with 10× PBS as a model drug and the integrated dosing sensors were
connected to the internal circuit. For each run, eight consecutive boluses (~5.33 µL per bolus) were
delivered (1 mA current applied using a Keithley 2400 sourcemeter, Keithley Instruments Inc., Cleveland,
OH). The volume dispensed from the pump, was calculated based on fluid front movement in a calibrated
100 µL micropipette attached to the micropump catheter outlet. The sensor signal was recorded prior and
after each bolus delivery and normalized to the baseline value for each run. Four runs were performed and
the results used to obtain averaged trends (Figure 4-10).
138

Figure 4-10: Calibration testing for wireless sensing: 8 boluses, each 5.33 µL (1.33 % of reservoir fill volume)
delivered for each of the four runs (sensor response values for each trace are normalized to multiplication result
measured for the full reservoir).
The averaged results from the calibration tests were used to obtain a linear relationship and calibration
curve between the volume delivered by the micropump and the natural log of the sensor response (Figure
4-11). This curve was used to program the LabVIEW graphical user interface to estimate the volume
delivered by the micropump with an accuracy of ±10%.

Figure 4-11: Averaged results from calibration tests were used to obtain a linear relationship between the volume
delivered by the micropump and the natural log of the sensor response.
139

Figure 4-12: Linear relationship between the % accumulated volume delivered by the micropump and the natural log
of the sensor response for two different reservoir sizes. Accumulated volume is normalized to each reservoir’s fill
volume.
It is interesting to note, that as sensor response is dependent on the cross sectional area of the fluid
contained in the reservoir, the calibration curve will be identical for different reservoir sizes if bolus delivery
is normalized to the reservoir fill volume. This relationship is evident in Figure 4-12, the accumulated
delivered volume from the 1 mL and 400 µL reservoirs during each set of calibration tests were normalized
to their total fill volume.
2.2 SIMULTANEOUS WIRELESS FLOW CONTROL AND DOSE SENSING
Real-time wireless flow rate variations and simultaneous dose sensing for a micropump was
demonstrated for an example dosing regimen (2 devices, n = 2 per device). The micropump reservoir was
filled with 10× PBS. The LabVIEW interface was used to select 1.04, 0.28, 0.52, 0.14, 0.83, and 1.04
µL/min as the expected output flow rates. Each flow rate selection lead to current application to the
electrolysis electrodes. Pumping durations were selected to be 2, 7, 1.5, 8.75, 4, and 3 min, respectively.
The sensor response was recorded before and after each delivery and then analyzed using the LabVIEW
graphical user interface according to the calibration curve obtained (Figure 4-11) and an estimation of the
delivered volume based on the calibration curve was presented to the user following each delivery. Between
140

each run, complete recombination was allowed to occur and the reservoir was refilled. The results are
presented in Figure 4-13 as a comparison between the programmed volume (based on the selected flow rate
and delivery duration), the sensed volume (based on analysis of the sensor response), and the actual
delivered volume as calculated from fluid movement in a calibrated 50 µL micropipette connected to the
micropump catheter outlet. The smallest volume delivered during this test was 0.55 µL boluses which
constitutes ~0.3% of the micropump’s deliverable volume and ~0.13% of the reservoir’s fill volume.
However, this resolution can only be attained if the accumulated delivered volume since the last refill is >
2 µL. The non-linear relationship between the sensor response and delivered volume (4-4) leads to increased
sensitivity (and bolus resolution) as accumulated volume increases:
(
)*
+,-./0
∝ 2
34
∝567589 965:8756 (4-4)

Figure 4-13: Simultaneous wireless flow control and dose sensing for an example dosing regimen, comparing the
expected delivered volume (based on the selected flow rate and delivery duration), the estimated delivered volume
(based on analysis of the sensor response), and the actual delivered volume by the micropump.
141

2.3 RECOMBINATION DETECTION
As mentioned in Chapter 2, once the current application ceases, gases generated by the bellows actuator
recombine back to water and in the absence of a flow regulating check-valve in-line with the delivery
catheter, dispensed drug along with bodily fluids will be retracted into the catheter and potentially inside
the pump during the recombination process thereby impacting dosing. Therefore, tracking this phenomena
with the dosing sensors is desirable to improve dosing accuracy. However, as recombination occurs on a
much slower time scale compared to actuation, drift may dominate the EI measurement. Results detailed in
Chapter 3 showed that for measurements attained using a precision LCR meter directly connected to the EI
sensors, depending on the speed of recombination, approximately 15-30% of the recombined volume is
indecipherable from the drift in measurement (for a ~ 50 µL bolus recombining over one hour).
In order to determine the feasibility of detection of recombination processes using wireless dosing
sensors, the outlet of the micropump filled with 10× PBS was attached to a 100 µL calibrated micropipette.
Generated gas volume and recombination were indirectly measured by observing the fluid front movement
in the micropipette. A 43 µL bolus was delivered, the current was turned off, and recombination was
measured periodically based on the fluid back flow in the micropipette for 60 minutes. At each measurement
time point wireless dose sensing was used to estimate the volume of fluid in the reservoir. This value was
subtracted from the total delivered volume to attain the recombined volume. These results were then
compared to the visually observed recombination in the micropipette (n=3). Approximately 30% of the
delivered volume recombined within 12 minutes, this volume was correctly estimated using the dosing
sensors within ±15%. Drift in measurement dominated the remainder of recombination and the volumes
could not be correctly estimated.
During EI measurements, each time the sensing signal is applied to the sensors, the electrochemical cell
is perturbed and depending on the excitation voltage used to obtain the measurement, a finite amount of
time is required for the system to settle into equilibrium [9]. Previous studies using the precision LCR meter
had shown that the majority of drift is observed within the first hour until a state of equilibrium is reached.
142

Based on these results, another experiment was performed in which an 11.52 µL bolus of 10× PBS was
delivered using the micropump attached to a 50 µL calibrated micropipette. The system was left unperturbed
for two hours before wireless dose sensing was used to estimate the volume of fluid in the reservoir. The
recombined volume was estimated to be 2.03 µL, approximately 26% less than the recombined volume
calculated from fluid backflow in the micropipette (2.78 µL). The results are promising, as combined with
techniques to speed up recombination (as presented in Chapter 2), it may be possible to estimate
recombination at certain time points following delivery.
2.4 EFFECTS OF COIL MISALIGNMENT
Wireless inductive transmission has some limitations, such as short operating range and required
alignment between coils [10]. The micropump presented in this work is meant to be implanted in a freely
moving subject. Depending on the orientation of transmitting and receiving coils, if the coils are oriented
perpendicularly with regards to each other, then the mutual inductance would be zero and essentially no
power would be transmitted [11]. Therefore it is important to characterize the effects of coil separation
distance and coil misalignment in power transmission and sensing.
Variations in the output current, as a result of distance and misalignment between the external
transmitting and internal receiving coil, were calculated by measuring the voltage across a 1 kΩ load used
in place of an actuator. The output current was unaltered when the receiving coil was placed within ± 2.5
cm of the transmitting coil. Based on these results, for animal studies, the cage (e.g. polycarbonate reusable
animal cage, 23.8 × 13.8 × 13 cm
3
, Tecniplast, Buguggiate, Italy) should be placed inside the transmitting
coil, with the coil ~2.5 cm from the bottom edge of the cage allowing for proper delivery within 5 cm of
the cage bottom. Unaltered transmission was observed for 5° misalignment between coils placed 2 cm apart.
10° misalignment between coils placed at the same distance lead to 7% drop in the received current. Current
control could not be achieved for misalignment angles >15°.
143

For wireless sensing the multiplier on the external transmitter operates with 1 V input signals. The
received signal was filtered, then amplified prior to multiplication. A potentiometer was used to set the
amplifier gain. For sensing presented here, the gain was chosen to allow for up to 5 cm of distance between
the internal transmitting and external receiving coils. The transmitter gain can be adjusted to accommodate
increased distance between the coils. However, it is important to note, that for inductive transmission, the
receiving coil should be placed within the inductive field created by the transmitting coil [8]. As such, when
implanted in a moving subject, the external receiving coil (Ø 50 mm presented here) may have to be moved
along with the moving animal. Kilinc, et. al., have developed a servo-controlled system that could track an
animal’s movements and move an external coil accordingly [12]. If needed, a similar setup could be used
to automatize moving of the receiving coil based on the movement of the internal transmitting coil.
2.5 DELIVERY IN SIMULATED BRAIN TISSUE

Figure 4-14 Testing setup for experiment with simulated brain tissue material: the pump and circuit were then placed
inside the simulated brain tissue so that each transmitting and receiving coil pair was separated by 2 cm of the
simulated tissue.
Successful power transmission through the simulated brain tissue material was previously demonstrated
(detailed in Chapter 2). The results showed no significant difference between infusion flow rates for
wireless transmission through air vs. simulated tissue; one way analysis of variance, p < 0.05). In order to
mimic the effects of wireless data transmission through tissue, brain tissue was once again simulated
according to the recipe described in [13]. Briefly, sugar, NaCl salt, and Natrosol® (hydroxyethylcellulose,
144

Ashland Inc., Covington, KY) were dissolved in 800 mL of DD water (1108.9, 49.5, and 19.8 g,
respectively). 1.98 g of ProClin® 950 preservative (Sigma-Aldrich, St. Louis, MO) was added as a
replacement for Dowicil 75® (1-(3-chloroallyl)-3, 5, 7-triazalazoniaadamantanechloride). A 4 cm thick
slab of gel-like material was created for testing. The micropump reservoir was filled with 10× PBS mixed
with blue food coloring to enhance visual contrast during delivery. Then the pump and circuit were then
placed inside the simulated brain tissue so that each transmitting and receiving coil pair was separated by 2
cm of the simulated tissue (Figure 4-14). Delivery at the catheter outlet was recorded using a Canon digital
single lens reflex (SLR) camera (EOS Rebel XSi, Canon, Tokyo, Japan). After placement, the catheter
outlet was monitored for 2 minutes, to observe potential diffusion of the reservoir contents into the
simulated tissue material. No diffusion was observed. The LabVIEW interface was then used to select 0.52
and 1.04 µL as the expected output flow rates. Each flow rate selection lead to current application to the
electrolysis electrodes. Pumping durations were selected to be 4 and 3 min, respectively. The sensor
response was recorded periodically and analyzed using the LabVIEW graphical user interface according to
the calibration curve obtained (presented in the above section) and an estimation of the delivered volume
based on the calibration curve was presented to the user at each measurement time point. Screen grabs of
the recorded video are shown in Figure 4-15. Successful wireless infusion flow control and sensing was
achieved in the simulated brain tissue material. Image analysis using ImageJ software (National Institutes
of Health) confirmed that flow rate was successfully altered.
145

Figure 4-15: Photographs showing delivery of with 10× PBS mixed with blue food coloring in simulated brain tissue
material. The programmed volume (based on the selected flow rate and delivery duration) and the sensed volume
(based on analysis of the sensor response) are included at each time point.
3. CONCLUSION
A fully integrated wireless system combining the electrolysis-based micropump with the
electrochemical dose tracking system is presented (Figure 4-8). The research prototype was scaled for small
animal (rodents) research. Smaller packaging was designed to decrease the inaccessible dead volume in the
reservoir from nearly 800 µL to less than 100 µL. The final prototype with the integrated coils and citcuitry
weighed 4.1 g (for a micropump with an empty reservoir). Wireless infusion rate control (0.14 – 1.04
µL/min) and dose sensing (bolus resolution of 0.55 – 2 µL) were each calibrated separately with the final
circuit architecture and then simultaneous wireless flow control and dose sensing was demonstrated for an
example delivery regimen. Recombination detection using the dosing system, as well as, effects of coil
distance and misalignment in wireless power and data transfer were studied. Finally, successful delivery,
infusion rate control, and dose sensing was achieved through simulated brain tissue material.
4. REFERENCES
[1] The Johns Hopkins University. (2015, 3/1/2015). Animal Care and Use Committee: The Mouse
[and] The Rat. Available: http://web.jhu.edu.libproxy.usc.edu/animalcare/procedures/
146

[2] A. M. Cobo, R. Sheybani, H. M. Tu, and E. Meng, "A Wireless Implantable Micropump for
Localized Drug Infusion," Sensors & Actuators: A. Physical, 2015 (submitted).
[3] H. M. Gensler, "A Wireless Implantable MEMS Micropump System for Site-specific Anti-cancer
Drug Delivery," University of Southern California, 2013.
[4] R. Sheybani and E. Meng, "High-Efficiency MEMS Electrochemical Actuators and
Electrochemical Impedance Spectroscopy Characterization," Microelectromechanical Systems,
Journal of, vol. 21, pp. 1197-1208, 2012.
[5] H. Gensler, R. Sheybani, and E. Meng, "Rapid non-lithography based fabrication process and
characterization of Parylene C bellows for applications in MEMS electrochemical actuators," in
16th International Solid-State Sensors, Actuators and Microsystems Conference, Transducers '11,
Beijing, China, 2011, pp. 2347-2350.
[6] R. Sheybani, H. Gensler, and E. Meng, "A MEMS electrochemical bellows actuator for fluid
metering applications," Biomedical Microdevices, pp. 1-12, 2012/07/01 2012.
[7] R. Sheybani and E. Meng, "Acceleration Techniques for Recombination of Gases in Electrolysis
Microactuators," Electrochimica Acta, 2015 (submitted).
[8] K. v. Schuylenbergh and R. Puers, Inductive powering : basic theory and application to
biomedical systems. Dordrecht: Springer, 2009.
[9] N. S. Kaisare, V. Ramani, K. Pushpavanam, and S. Ramanathan, "An analysis of drifts and
nonlinearities in electrochemical impedance spectra," Electrochimica Acta, vol. 56, pp. 7467-
7475, 2011.
[10] F. Zhang, L. Xiaoyu, S. A. Hackworth, R. J. Sclabassi, and S. Mingui, "In vitro and in vivo
studies on wireless powering of medical sensors and implantable devices," in 2009 IEEE/NIH
Life Science Systems and Applications Workshop (LiSSA 2009), Piscataway, NJ, USA, 2009, pp.
84-7.
[11] B. Lenaerts, Omnidirectional inductive powering for biomedical implants, 1st ed. New York:
Springer, 2008.
[12] E. G. Kilinc, K. Kapucu, F. Maloberti, and C. Dehollain, "Servo-controlled remote powering and
low-power data communication of implantable bio-systems for freely moving animals," in
Biomedical Circuits and Systems Conference (BioCAS), 2014 IEEE, 2014, pp. 508-511.
[13] G. Hartsgrove, A. Kraszewski, and A. Surowiec, "Simulated biological materials for
electromagnetic radiation absorption studies," Bioelectromagnetics, vol. 8, pp. 29-36, 1987.

147

CHAPTER 5: CONCLUSION
Effective drug therapy is an essential tool in improving health outcomes in the treatment and
management of chronic conditions. Despite the research and commercial efforts in the past three decades,
it is still clinically relevant to investigate MEMS-based implantable drug delivery technologies that can
provide controlled drug volumes at specific times and locations within the body and include sensors that
can provide feedback for the device operation in real-time. This work focuses on the design, fabrication,
and characterization of a wireless electrochemical drug delivery micropump with a fully integrated
electrochemical dose tracking feedback system.
A programmable, implantable, and low power (0.66-51.31 mW) electrochemical drug infusion
micropump system, capable of delivering a diverse assortment of liquid drug formulations with high
accuracy within a wide dynamic range of dose volumes and flow rates (0.33 - 141.9 µL/min) was
demonstrated. Viscosity independent delivery and real-time electrical control of variable flow rate drug
regimen were also shown. Recombination, an important factor for reliable and repeatable delivery, was
studied. Several methods were explored to improve the recombination rate in a bellows actuator. The results
of these experiments showed that actuators with floating pyrolyzed Nafion®-coated pieces added to the
electrolysis chamber, seemed to show consistent performance. The catalyst pieces in the water did not affect
electrolysis gas generation, yet the rate of recombination was measured to be 2.3 times faster across all
actuator orientations compared to unmodified Nafion®-coated electrolysis actuators. Diffusion of sodium
ions through the Parylene bellows into the electrolysis chamber was also investigated for several
modifications to the bellows fabrication. This is especially important as a saline is often used as a solvent
in drug therapy and sodium ions could irreparably damage Nafion® coating on the electrolysis electrodes,
leading to unreliable actuator performance. The preliminary results obtained showed that the rate of sodium
diffusion was significantly slower through metal-capped bellows. A class E inductive powering system was
designed, along with ASK modulation to allow adjustment of the current applied to the actuator (0.6 -3.2
148

mA, with a resolution of 0.2 mA per modulation step), resulting in a wide range of infusion rates (2.0 - 25.0
µL/min) suitable for drug delivery applications.
Next, a fully integrated electrochemically-based dose tracking system for closed-loop fluid drug delivery
capable of real-time volume delivery tracking and confirmation was demonstrated. Two sets of electrodes
made from thin film and bulk wire electrodes were fabricated and their performance compared. Drift and
noise in measurement were also studied. Less drift in measurement is observed for fluids with higher ionic
concentration, at higher temperatures, and also when the counter and reference electrodes were de-coupled
by using a 3 electrode configuration. Water and lipid soluble drug analogs were studied to calibrate the
system. The smallest bolus volume detected was 83 nL for DD water (0.003% of the reservoir) and 556 nL
for 1× PBS (0.019% of the reservoir). Real-time blockage and refill detection were also presented. Wireless
sensing was achieved through combining ASK and frequency modulation for data transfer to and from the
sensors. Simultaneous real-time flow variation and wireless dose sensing was demonstrated. The smallest
volume delivered during characterization tests was 3 µL boluses (~1.67% of the micropump’s deliverable
volume and ~0.3% of the reservoir’s fill volume).
Finally, the two systems were merged to create a fully integrated wireless system combining the
electrolysis-based micropump with the electrochemical dose tracking system. The research prototype was
scaled for small animal (rodents) research. Smaller packaging was designed to decrease the inaccessible
dead volume in the reservoir from nearly 800 µL to less than 100 µL. The final prototype with the integrated
coils and citcuitry weighed 4.1 g (for a micropump with an empty reservoir). For this system, the current
range was limited compared to that attained previously for the micropump without the integrated sensors.
This can be attributed to decreased power transfer efficiency when the power signal is modulated with the
sensing signal and could potentially be mitigated by re-designing the system to increase power
amplification or increase the operation frequency of the sensing signal. Wireless infusion rate control (0.14
– 1.04 µL/min) and dose sensing (bolus resolution of 0.55 – 2 µL) were each calibrated separately with the
final circuit architecture and then simultaneous wireless flow control and dose sensing was demonstrated
149

for an example delivery regimen. Recombination detection using the dosing system, as well as, effects of
coil distance and misalignment in wireless power and data transfer were studied. Successful delivery,
infusion rate control, and dose sensing was achieved through simulated brain tissue material.
Due to chronological nature of system design, several restrictions were placed on the performance of
the final prototype that could be mitigated through future work. The output current range (and as a result
the infusion flow rate), as well as, the allowable coil misalignment and distance, were limited compared to
that attained previously for the micropump without the integrated sensors. This can be attributed to
decreased power transfer efficiency when the power signal is modulated with the sensing signal and could
potentially be mitigated by re-designing the system to increase power amplification, including multiple
receiving coils, or increasing the operation frequency of the sensing signal. Integration of a normally closed
check valve is also required to prevent backflow of fluids as a result of the reverse pressure gradient caused
by recombination. Another potential area of improvement is decreased sodium diffusion through the
Parylene bellows. While preliminary results with metal-capped bellows were promising, further testing is
required to ensure long-term reliability. Lastly, static tissue models may not fully capture wireless power
transfer and sensing in live moving animals. While efforts were taken to simulate operation and proper
function of the system in a moving subject, a true test of functionality would require implantation in animal
models. For ultimate human use, security measures to protect the device from tampering by unauthorized
users, as well as, accidental device compromise due to interference from surrounding wireless
communication should be implemented.
Despite these shortcomings, the high-performance drug delivery micropump presented offers the
potential for unprecedented accuracy in delivering a diverse assortment of liquid drug formulations at the
right dose, to the right tissue, and at the right time over the entire course of treatment. Wireless flow
controllability and dose tracking sensors, improve patient safety and allow for active control over the
delivery profile and early warning of pump malfunctions leading to optimized and personalized patient-
tailored therapy.

Abstract (if available)

Abstract Drug delivery is essential for the treatment of chronic diseases. Implantable site‐specific drug delivery devices can deliver a potent and effective dose of drug directly to the site of therapy, improving treatment outcomes while reducing potential side effects and the risk of infection due to catheters running through the skin. These factors serve to increase patient comfort and decrease overall associated healthcare costs for the treatment of chronic conditions. A MEMS approach miniaturizes infusion pumps such that they are implantable

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

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

PDF

Implantable bioMEMS drug delivery systems

PDF

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

PDF

Parylene-based implantable interfaces for biomedical applications

PDF

Energy aware integrated circuits for communication and biomedical applications

PDF

The electrochemical evaluation of Parylene-based electrodes for neural applications

PDF

Next generation neural interfaces for neural modulation and prosthesis

PDF

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

PDF

Integrated wireless piezoelectric ultrasonic transducer system for biomedical applications

PDF

Development of implantable Parylene-based MEMS technologies for cortical applications

PDF

Parylene-based biomems sensors for multiple physiological systems

PDF

Integrated reconfigurable wireless receivers supporting carrier aggregation

PDF

Multivalent smart elastin-like polypeptide therapeutics with drug delivery and biosensing applications.

PDF

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

PDF

A percutaneously implantable wireless neurostimulator for treatment of stress urinary incontinence

PDF

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

PDF

Electronics design and in vivo evaluation of a wirelessly rechargeable fetal micropacemaker

PDF

Parylene C bioMEMS for implantable devices with electrochemical interfaces

PDF

Design and application of a C-shaped miniaturized coil for transcranial magnetic stimulation in rodents

PDF

Chitosan nanoparticle mediated oral drug delivery of kidney targeting micelles for polycystic kidney disease

Asset Metadata

Creator Sheybani, Roya (author)

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

School Viterbi School of Engineering

Degree Doctor of Philosophy

Degree Program Biomedical Engineering

Publication Date 12/19/2015

Defense Date 04/10/2015

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

Tag dose sensing,drug delivery,electrochemical sensors,electrolysis,micropump,OAI-PMH Harvest,wireless system

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Meng, Ellis (committee chair), Hashemi, Hossein (committee member), Weiland, James D. (committee member)

Creator Email rsheyban@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c3-574468

Unique identifier UC11299500

Identifier etd-SheybaniRo-3490.pdf (filename),usctheses-c3-574468 (legacy record id)

Legacy Identifier etd-SheybaniRo-3490.pdf

Dmrecord 574468

Document Type Dissertation

Format application/pdf (imt)

Rights Sheybani, Roya

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA