University of Southern California Dissertations and Theses

Development of micromachined technologies for neural interfaces

(USC Thesis Other)

Development of micromachined technologies for neural interfaces

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content
DEVELOPMENT OF MICROMACHINED TECHNOLOGIES FOR NEURAL
INTERFACES

by

Jonathan T.W. Kuo

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

December 2013

Copyright 2013 Jonathan T.W. Kuo
i

To my family
ii
ACKNOWLEDGEMENTS

Give me wisdom and knowledge…
King Solomon to God
2 Chronicles 1:10

As I put the finishing touches on my thesis I find myself reflecting often upon my
journey here. Many people have been integral to propelling me to this point. The first to
come to mind is my family, without whose support I would be lost. My grandmother,
who looked after me full-time in the first couple years of my life as my parents worked.
My parents’ tolerance and encouragement of my explorations into the world during my
childhood were integral; without which I would not be who I am today. My brother,
Jeffrey has been a source of companionship and brotherhood since I was two and half
years of age. To my family, I cannot fully convey the gratitude of which you are owed;
thank you.
My time at USC has been one of the best in my life. The education I have
received here has been more than I could have expected going into grad school and no
doubt will serve me well for life. I chose Dr. Ellis Meng for my advisor because she is a
brilliant innovator. Thankfully, she did not reject me as a student. I have not been
disappointed in the knowledge I have gained from her. Her ability to take entrenched
problems and engineer novel solutions that truly work well is inspiring every time I
witness the process. The first class education I received in research and communication
skills (literature searching, reading between the lines, thinking a problem through, not
iii
comparing apples and oranges, writing, poster making, presenting a complete story etc.)
cannot be matched. I am grateful for the skills and knowledge she has imparted.
The camaraderie from the other students in lab has been integral to making the
grad school experience something I looked forward to everyday. The warm welcome and
words of wisdom I received from Dr. Po-Ying (Brian) Li alleviated much of my anxieties
surrounding graduate research. Dr. Ronalee Lo’s advice was also helpful. I am grateful to
both Drs Brian Li and Christian Gutierrez for training me in microfabrication processes. I
am especially indebted to Dr. Christian Gutierrez’s keen observation that cleanroom air
currents and SU-8 photoresist do not go well together. The DARPA team; Brian Kim,
Seth Hara, and Curtis Lee were the best partners to have in working on a challenging but
fulfilling project. Many thanks for the late nights and weekends spent helping me with
fabrication. Heidi Gensler and Roya Sheybani have been constant pillars of support and
humor over the years; it has been fun sitting in-between these two every day for all these
years. It is encouraging to see bright new students emerge, Lawrence Yu (not so new-one
of my first undergrads) and Angelica Cobo, their enthusiasm and energy are
heartwarming. Of course, I must mention the lab traditions: the daily lunch and coffee
communions, Togo’s Days, happy hours, and the newer Fried Chicken Fridays; may
these survive and be a source of joy to others as they have been to me.
iv
TABLE OF CONTENTS

ACKNOWLEDGEMENTS .................................................................................................... II
TABLE OF CONTENTS ..................................................................................................... IV
LIST OF TABLES .............................................................................................................. VI
LIST OF FIGURES ........................................................................................................... VII
ABSTRACT …… ............................................................................................................. XIV
CHAPTER 1 INTRODUCTION ............................................................................................ 1
1.1 THE STUDY OF NEURONS ........................................................................................... 1
1.2 NEURAL INTERFACING METHODS .............................................................................. 3
1.3 THE CASE FOR MEMS NEURAL INTERFACES ........................................................... 11
1.4 MEMS FABRICATION TECHNOLOGY ....................................................................... 14
1.5 MEMS NEURAL INTERFACES .................................................................................. 23
CHAPTER 2 PARYLENE SHEATH PROBES FOR LONG-TERM IN VITRO SIGNAL
RECORDING ........................................................................................... 30
2.1 BACKGROUND .......................................................................................................... 30
2.2 MATERIAL SELECTION ............................................................................................. 33
2.3 TOP ELECTRODE SHEATH PROBES ........................................................................... 33
2.4 WING ELECTRODE PROBES ...................................................................................... 39
2.5 ARRAYED SHEATH PROBES ...................................................................................... 49
2.6 MINIMALLY SIZED PROBES ...................................................................................... 56
2.7 DISCUSSION & CONCLUSION .................................................................................... 60
CHAPTER 3 MICROMACHINED CHEMICAL NEURAL INTERFACES .............................. 64
3.1 BACKGROUND .......................................................................................................... 64
3.2 AIMS.. ...................................................................................................................... 65
3.3 MICROFLUIDIC PLATFORM FOR CHEMICAL INTERFACING ........................................ 65
3.4 MICROFLUIDIC PLATFORM DISCUSSION ................................................................... 75
3.5 MICROFLUIDIC PLATFORM CONCLUSION ................................................................. 80
3.6 MODULES FOR CHEMICAL INTERFACING .................................................................. 81
3.7  CCM ..................................................................................................................... 83
3.8  BAM ..................................................................................................................... 88
3.9 CHEMICAL INTERFACING MODULES CONCLUSION ................................................. 101
CHAPTER 4 INCLINED SU-8 MIRRORS FOR OPTICAL STIMULATION ........................ 108
4.1 BACKGROUND ........................................................................................................ 108
v
4.2 AIM… .................................................................................................................... 111
4.3 THEORY ................................................................................................................. 112
4.4 DESIGN .................................................................................................................. 114
4.5 FABRICATION ......................................................................................................... 114
4.6 RESULTS ................................................................................................................ 118
4.7 CONCLUSION .......................................................................................................... 126
CHAPTER 5 CONCLUSION ............................................................................................ 133
APPENDIX A: SILANE A-174 PARYLENE ADHESION TREATMENT ............................. 137
APPENDIX B: PARYLENE SHEATH PROBE VERSION 1 FAB RECIPE ........................... 138
APPENDIX C: PARYLENE SHEATH PROBE VERSION 1 MASK SET ............................. 140
APPENDIX E: PARYLENE SHEATH PROBE VERSION 2 MASK SET ............................. 145
APPENDIX F: PARYLENE SHEATH PROBE VERSION 3 FAB RECIPE ........................... 149
APPENDIX G: PARYLENE SHEATH PROBE VERSION 3 MASK SET ............................. 151
APPENDIX H: MINIMALLY SIZED PROBES FAB RECIPE ............................................ 154
APPENDIX I: MINIMALLY SIZED PROBES MASK SET ................................................. 155
APPENDIX J: ARRAYED PARYLENE SHEATH PROBES WITH PERFORATIONS FAB
RECIPE ................................................................................................. 161
APPENDIX K: ARRAYED PARYLENE SHEATH PROBES WITH PERFORATIONS MASK
SET ....................................................................................................... 163
APPENDIX L: INCLINED SU-8 MIRRORS FAB RECIPE ................................................ 169
APPENDIX M: INCLINED SU-8 MIRRORS MASK ........................................................ 170
APPENDIX N: µCCM SU-8 MOLD FAB RECIPE .......................................................... 171
APPENDIX O: µCCM SU-8 MOLD MASK SET ............................................................ 172
APPENDIX P: µBAM FAB RECIPE ............................................................................... 173
APPENDIX Q: µBAM MASK SET ................................................................................. 174
vi
LIST OF TABLES
Table 2-1: Dimensions of the different fabricated sheath structures for top electrode
designs................................................................................................................... 38

vii
LIST OF FIGURES
Figure 1-1: Basic neural circuitry. Neurons communicate with one another across the
synapse by the release of neurotransmitters in response to an electrical signal from
the presynaptic neuron's axon. ................................................................................ 2
Figure 1-2: Experimental setup of Galvani where an electric machine is attached to a
nerve-muscle preparation of frog legs with a metal wire [2]. Electrical stimulation
caused muscle contraction. With kind permission from Springer Science and
Business Media. ...................................................................................................... 3
Figure 1-3: (a) A representative microwire array adapted from [6] © 1968 IEEE. (b) Utah
array from [9]. Reprinted by permission from Macmillan Publishers Ltd: Nature
Neuroscience, copyright 2002. (c) Michigan probe from [10] © 2008 IEEE. ....... 5
Figure 1-4: ChR2 mechanism. The Na
+
ion channel is opened when exposed to blue light
(λ = 470 nm) of sufficient optical power. The subsequent flow of Na
+
ions across
many opened channels lead to action potential generation. .................................... 8
Figure 1-5: Photostimulation of an optogenetic neuron causes generation of action
potentials due to the actions of channerhdopsin ion channels. ............................... 9
Figure 1-6: Neurotransmitter uncaging leads to an increase in localized concentration of
biologically active neurotransmitter that acts upon the neuron. ........................... 10
Figure 1-7: A neural probe integrated with optical waveguides for optogenetic
applications from [38] © 2011 IEEE. ................................................................... 11
Figure 1-8: Illustration of photolithography. The Chinese character for Kuo in the
photomask is transferred to the substrate. A light source, commonly in the form of
a Hg lamp, has its light focused and collimated through optics. The light passes
through a photomask, exposing the photopattern to the photoresist covered wafer
substrate which is then developed. Mechanical stages can be used to precisely
move the substrate for alignment purposes prior to exposure. ............................. 16
Figure 1-9: Chemical structure of SU-8 polymer (modified from [45]). .......................... 18
Figure 1-10: Chemical structure of Parylene C polymer. The characteristic single chlorine
(Cl) attached to the benzene ring is what differentiates Parylene C from other
Parylene types. ...................................................................................................... 20
Figure 1-11: Metal patterning through the liftoff process. (a) Photoresist is first patterned
and developed with a negative step profile. (b) Metal is then deposited over the
entire surface. (c) Next, an acetone soak attacks and dissolves the photoresist
viii
mask lifting off the undesired metal. (d) Finally, the desired metal pattern is
obtained. ................................................................................................................ 21
Figure 2-1: (Left) Neurotrophic cone electrode concept showing ingrowth of neural
processes toward recording microwires from [11]. (Right) Explanted device with
neural tissue ingrowth from monkey after 6 months of implantation; Scale bar:
100 µm [7]. ........................................................................................................... 32
Figure 2-2: Conceptual drawing of Parylene sheath probes for long-term intracortical
recordings. The 3D sheath structure allows for ingrowth of neural processes
toward recording electrodes inside the sheath. ..................................................... 34
Figure 2-3: Fabrication process for sheath probe having “top” electrodes on sheath
surface. Only the final outline of the device is shown. ......................................... 35
Figure 2-4: Thermoforming process steps. (a) Released probes were (b) shaped around a
microwire mold and thermally treated. (c) Subsequently, wires were removed to
reveal the final structure. ....................................................................................... 36
Figure 2-5: (a) Sheath probe with integrated Parylene cable attached to a ZIF connector
for external electrical connections. (b) Released sheath probe. The 3D Parylene
sheath structure holds its shape post-thermoforming. ........................................... 37
Figure 2-6: Post-thermoformed probes of the three different sheath shapes. ................... 38
Figure 2-7: (a) Top electrodes were prone to fracture (arrows) due to the strain induced
by opening the Parylene microchannel for thermoforming. (b) An SEM image of
a top electrode showing multiple cracks. .............................................................. 39
Figure 2-8: Comparison of top electrode (a) and wing electrode (b) probes. Scale bar is
150 m. Moving outer electrodes from the top to the wings of the sheath avoids
straining these recording sites when opening the Parylene structure with a
microwire for thermoforming. .............................................................................. 40
Figure 2-9: Fabrication process for sheath probe having peripheral “wing” electrodes.
Only the final outline of the device is shown. Note that the time-consuming steps
of an additional metal patterning and deposition as well as Parylene insulation
deposition and patterning are eliminated as compared to the previous top
electrode sheath probe fabrication. ....................................................................... 41
Figure 2-10: Comparison of partially etched Parylene insulation (a) to completely open
electrodes (b). A partial etch can be rectified with more plasma etching until the
eletrodes look “shiny.” .......................................................................................... 42
Figure 2-11: Scum deposited on metals after a standard clean of the PR etch mask. ...... 42
ix
Figure 2-12: Scum is still present after an aggressive descum in plasma. ........................ 43
Figure 2-13: (a) Evidence of partial PR reflow can be seen as PR starts to obscure the
recording site (arrow). (b) The previously open pattern above the recording site is
now completely covered due to PR reflow (arrow); thereby preventing plasma
etching of the underlying insulation. .................................................................... 44
Figure 2-14: Bubbles are generated underneath the Parylene when exposing and
developing the PR mask for plasma etching on wafer. Note that bubbles originate
in exposed areas. ................................................................................................... 45
Figure 2-15: Bubbles are generated underneath the Parylene when exposing and
developing the PR mask for plasma etching on wafer. Note that bubbles originate
in exposed areas. ................................................................................................... 46
Figure 2-16: Probes possessing wing electrodes were fabricated with three different
sheath shapes. ........................................................................................................ 47
Figure 2-17: Representative (left) CV in H
2
SO
4
and (right) EIS in 1XPBS of a single
recording site. ........................................................................................................ 48
Figure 2-18: Representative rat neuronal activity from a sheath probe recording site at
different days post-implantation. .......................................................................... 49
Figure 2-19: Comparison between (a) sheath probe shape A from before and (b)
perforated sheath probe. Perforations were added throughout the sheath for better
neural integration and a sleeker outline was obtained by etching away the tip and
introducing a taper to the wings. A sleeker outline may also minimize neural
damage during implantation. ................................................................................ 51
Figure 2-20: A left and right handed sheath probe array. ................................................. 52
Figure 2-21: Fabrication process for arrayed sheath probe possessing perforations. Only
the final outline of the device is shown. ............................................................... 53
Figure 2-22: An electrically packaged array. Two devices were packaged together
through ZIF connectors onto a PCB. .................................................................... 54
Figure 2-23: SEM of perforated sheath probe after themoforming. ................................. 55
Figure 2-24: Representative (left) CV in H
2
SO
4
and (right) EIS in 1XPBS of Pt recording
sites (n = 32) of arrayed perforated probes electrically packaged into a 2×2 array
for implantation. .................................................................................................... 56
Figure 2-25: Representative in vivo signal recording from rat implanted perforated sheath
probe at 21 days. ................................................................................................... 56
x
Figure 2-26: Evaluation of tissue response to implanted probes at 30 days. 10 µm cortical
sections sliced perpendicular to the probe were stained with NeuN (red) to
visualize neurons and GFAP (blue) highlighting reactive astrocytes. The presence
of astrocytes indicates tissue damage. A tissue void matching the shape of the
microwire used to insert the three different sheath probe shapes is evident (a)
Sheath shape A, sharpest taper. (b) Sheath shape B, wider taper (c) Blunt
cylindrical sheath probe, from [27]. ...................................................................... 57
Figure 2-27: (a) Optical micrograph of the smallest sheath probe fabricated with 10 µm
width wings. Even with such a narrow wing width, the multi-layered Parylene
structure held. The three Pt recording electrodes are visible. (b) SEM image of
another minimal sheath probe with 30 µm width wings. 15 µm diameter
perforations were added to the non-thermoformed probe. .................................... 58
Figure 2-28: A minimal sheath probe with a wing width of 40 µm opened with a
microwire. Structural integrity of the probe was compromised as evidenced by
slight Parylene delamination and tearing (arrow) with microwire insertion. ....... 59
Figure 3-1: Top view of a microfluidic platform with three 100 µm wide Parylene
microchannels. Each channel is centrally perforated with a pore having a size
indicated by the label on the upper right corner of each channel (5, 10, 20 µm
diameter). Eight Pt thermal flow sensors line each channel and flank the central
pore. ...................................................................................................................... 66
Figure 3-2: Microfluidic platform design (not to scale). .................................................. 67
Figure 3-3: Illustration showing concept of hot film thermal flow sensor. The resistor
serves as a heater and sensing element. Resistance value is dependent on
temperature. .......................................................................................................... 68
Figure 3-4: Fabrication process for creating the microfluidic platform. .......................... 71
Figure 3-5: Packaged microfluidic platform in a custom made acrylic jig with gold wire-
bonds connecting the embedded Pt thermal flow sensors to the PCB. Wires were
soldered to the PCB and convered with epoxy to make robust external electrical
connections. Glass capillaries aligned to and clamped on top of the SU-8
inlet/outlets provided a fluidic connection to the microchannel. .......................... 73
Figure 3-6: Hot film response for 4 mA constant current biasing for two flow sensors, one
upstream and one downstream of the pore in the channel (mean ± S.E. with n =
58). Both sensors were monitored simultaneously. Data points marked with an
asterisk have non-significant changes due to change in flow rate based on
statistical analysis. This is attributed to low signal-to-noise ratio. ....................... 75
xi
Figure 3-7: Time-lapse fluorescent images of PC12 cells focally stimulated by bradykinin
on top of the biomimetic chemical interface. PC12 cells increasingly fluoresce in
a radial nature as more bradykinin is delivered through the pore (arrow) and
diffuses outward. ................................................................................................... 76
Figure 3-8: The microfluidic platform requires many external connections (fluidic and
electrical) in order to operate; leading to a bulky setup and limiting this
technology’s scalability due to costs associated with the syringe pump and
electronics. ............................................................................................................ 78
Figure 3-9: Illustration of the two complementary modules for chemical interfacing.
Neurons are cultured within µCCM which isolates the axons and somata with its
microchannels. Chemicals are delivered to the axons with the placement of
µBAM within the axon compartment. Scaling is possible with the use of multiple
µBAMs and the setup is compatible with standard microscopy. .......................... 82
Figure 3-10: Illustration of µCCM working principle (not to scale) with two axon
compartments. Neurons are loaded into the soma compartment and cultured
within the device. Fluidic pressure from the soma to the axon compartments
moves the neurons into the central channel toward the axon compartments. The
microchannels physically impede the neuronal soma from entering the axon
compartment while allowing it to extend axons. .................................................. 85
Figure 3-11: Neurons growing in vitro within µCCM. Axons can be seen growing into
the axonal compartment (arrows) with the somata in the other compartment. ..... 88
Figure 3-12: Time-lapse mockup of passive pumping within µBAM. Flow within the
microchannel is self-driven from the smaller fluid droplet into the larger. Outlines
of the original droplets are show to better illustrate the change in droplet size. ... 89
Figure 3-13: (a) An initially sized pore is etched into Parylene with a cross-sectional view
on the left and the top view on the right. (b) Subsequent Parylene coating over this
pore creates a smaller diameter pore that is tunable by adjusting the deposited
Parylene thickness. The nature of conformal coating is exaggerated in this
illustration. ............................................................................................................ 92
Figure 3-14: Illustration of µBAM. Passive pumping is exploited to generate flow in the
microchannel. Chemicals are released through a pore etched in the Parylene layer.
The microchannel is constructed of SU-8 that is capped with Parylene. .............. 93
Figure 3-15: Illustration of how sacrificial PR can be used to construct the embedded SU-
8 microchannel required for µBAM. .................................................................... 94
Figure 3-16: Thermocompression between two polymer surfaces leads to polymer
entanglement and a subsequent bond. ................................................................... 96
xii
Figure 3-17: Effects of temperature on Parylene/Parylene thermocompressive bonds.
Generally, higher bond strength is achieved with elevated bonding temperature.
From [41] © 2005 IEEE. ...................................................................................... 98
Figure 3-18: Effects of temperature on SU-8/SU-8 thermocompressive bond strengths.
Generally, higher bond strength is achieved with elevated bonding temperature.
Adapted from [37] © IOP Publishing. Reproduced by permission of IOP
Publishing. All rights reserved. ............................................................................. 98
Figure 3-19: Jig for thermocompressive bonding. The samples to be bonded, a patterned
SU-8 die and Parylene-coated glass slide are first placed between two Teflon
sheets (white). The two aluminum plates are then tightened together with the use
of screws to apply compression to the samples. ................................................... 99
Figure 3-20: A SU-8 die bonded to a Parylene coated glass slide. The SU-8/Parylene
bond keeps the die suspended. ............................................................................ 100
Figure 4-1: In vivo study of neural circuitry using optogenetic techniques. An optical
fiber is inserted into the rodent's brain and the light activates the underlying neural
tissue (adapted from [1]). .................................................................................... 108
Figure 4-2: Illustration of setup in which photostimulus is delivered via a microscope
objective. Only one light source is available at a fixed location on the stage. ... 109
Figure 4-3: Cross section of the proposed device (not to scale). The metal layer over the
inclined SU-8 structure acts as a mirror to guide the incoming blue light upwards
in order to locally activate the optogenetic cell. ................................................. 112
Figure 4-4: Inclined exposure of SU-8 in a glycerol bath. An anti-reflection layer between
Si substrate and SU-8 photoresist is required to prevent undesired exposure from
reflections off the polished wafer surface. .......................................................... 116
Figure 4-5: Improved fabrication process with the soda lime substrate serving as the mask
(left). Inclined SU-8 exposure takes place within a glycerol bath placed under the
UV lamp (right)................................................................................................... 117
Figure 4-6: Fabrication results (a) without and (b) with a Parylene layer between the soda
lime glass substrate and SU-8. Delamination of SU-8 across the wafer cannot be
prevented without a Parylene stress relief layer. 100% yield was achieved over a
3” soda lime wafer with the addition of a Parylene layer. No delamination was
observed. (c) Consistent, well-defined angled structures over the entire 3" wafer
were obtained, even at the edges where the photoresist is thinner. .................... 119
Figure 4-7: (a) Ambient air flow during the softbake process causes undesirable patterns
to form in the SU-8. (b) A funnel inverted over the wafer during softbaking
xiii
protects the SU-8 from external perturbations and leads to successful softbake
with a level SU-8 surface. ................................................................................... 121
Figure 4-8: Development of SU-8 with the use of an orbital shaker. ............................. 122
Figure 4-9: SEM images of the resulting SU-8 structures showing that the desired angle
of 45º was obtained. ............................................................................................ 123
Figure 4-10: Representative surface profile of the resulting SU-8 angled structure. The
variation is less than 8 nm; thus the structure is suitable for use in optical
applications. ........................................................................................................ 124
Figure 4-11: Light output from an optical fiber butt-coupled to a halogen light source.
Light is seen exiting the core of the fiber. Light is also seen escaping through the
cladding which is indicative of butt-coupling’s natural inefficiency causing
optical loss. ......................................................................................................... 125
Figure 4-12: Light is delivered to the mirror by an optical fiber butt-coupled to a halogen
light source. ......................................................................................................... 125
Figure 4-13: Representative intensity profile of the reflected light from the SU-8 mirror
structure. The intensity is in a Gaussian distribution which is as expected for non-
collimated light source. ....................................................................................... 125
Figure 4-14: (a) Brightfield image of cultured optogenetic PC12 cells. (b) Transfected
PC12 cells fluoresce due to the expression of a fluorescent protein tag indicating
successful expression of hChR2 channels. 10× magnification. .......................... 127
Figure 4-15: Differentiated PC12 cells cultured on a MEA. Some electrodes are damaged
in the form of metal delamination. ...................................................................... 127
xiv
ABSTRACT

BioMEMS is uniquely positioned to impact the field of neuroscience by
advancing neural interface state of the art. Traditional neural interfaces rely on macro-
world technologies that can be limited when attempting to interface with neurons on a
cellular level. Micromachining enables construction of microengineered devices
possessing form factors that are suitable for interfacing with neurons or complex neural
networks. Improved tools for neural interfacing can improve neuroscience knowledge by
enabling more sophisticated experimental studies or advancing the use of
neuroprosthetics.
In this work, three micromachined neural interfaces are developed and described;
each addressing a neural interfacing mode; electrical, chemical, and optical. Novel
micromachining technologies were developed in order to construct these neural
interfaces.
A neural probe in the novel form of a three-dimensional Parylene sheath is first
described in chapter 2. This sheath probe was constructed with biocompatible materials
and designed to obtain long-term in vivo neural signal recording by promoting neural
growth through the sheath structure. Multiple improving sheath probe designs and their
fabrication details are described. A Parylene thermoforming process was utilized to shape
the sheath. The microfabricated recording electrodes were characterized in vitro and
found to possess desirable electrochemical properties for neural recording. Sheath probes
were implanted into rat and in vivo neural signals were successfully recorded.
xv
Next, chemical neural interfacing with microdevices is described in chapter 3. A
microfluidic platform was first designed and developed to deliver chemicals focally to
neurons. This was achieved by micromachining Parylene microchannels that possessed
plasma etched pores through which focal delivery could occur. Thermal flow sensors
were also embedded in the microchannels which enable real-time sensing of the focal
delivery. Many challenges were encountered with the microfluidic platform and so
microfluidic modules were designed and constructed to simplify ease of use. Separate
microdevice modules separated the challenges of in vitro neural cell culturing and
chemical delivery so that each module’s packaging requirement was simplified by
addressing only one challenge. Soft lithography was utilized to construct CCM
(compartmented culture module), the module for guiding and separating axons and
somata in in vitro cultured neurons. Thermocompressive bonding was developed to
construct embedded SU-8 microchannels with a Parylene cap for BAM (biochemical
administration module), the module for chemical delivery. BAM was designed to
exploit the principle of passive pumping to self-generate flow and chemical delivery;
simplifying packaging requirements.
Finally, optical neural interfacing is explored through the creation of a
micromachined mirror array in chapter 4. The mirror array was designed to overcome
throughput issues that limit the use of traditional optogenetic or photocaging tools. A
novel fabrication process was developed and utilized to construct inclined structures that
were then coated with metal to become mirrors. The fabricated optical device was
xvi
characterized and found to possess acceptable optical properties for optical neural
interrogations.
1

1.1 The Study of Neurons
The mammalian nervous system is a complex structure comprised of networks of
neurons that work together to achieve incredibly complex tasks such as pattern
recognition, coordination of multiple sensory inputs, and initiating and controlling the
appropriate physiological responses to stimuli. The prevalence of neurological diseases
such as Parkinson’s and Alzheimer’s as well as the potential to heal paralysis has spurred
much research in the basic studies of neurons. Understanding the nervous system will
provide new treatments for neurobiological diseases.
The nervous system is highly complex with an estimated density of 10
5
neurons
per mm
3
in the human cortex with each neuron connecting to 10
3
or more other neurons
[1]. The basic structure of all the neurons within the nervous system is similar (Figure 1-
1). The diameter is measured in the 10’s of µm and the axons range in length from less
than 1 µm to more than 1 m.
Chapter 1 INTRODUCTION
2

Figure 1-1: Basic neural circuitry. Neurons communicate with one another across the
synapse by the release of neurotransmitters in response to an electrical signal from the
presynaptic neuron's axon.

Neurons are the primary unit of the nervous system which gives rise to
neurophysiological control over bodily functions, environmental sensing including vision,
as well as cognition. The nervous system achieves these complex tasks through networks
of neurons that process and transmit information. Information is transmitted in two ways;
through the generation of an electrical signal, an action potential (AP) which is
transmitted through a neuron’s cell body and axon, and through the transmission of
chemical signals, neurotransmitters which are released in the synaptic cleft between
neurons and signal information between one neuron to the next.
3
1.2 Neural Interfacing Methods
1.2.1 Electrical
Electrophysiological methods to communicate and probe the nervous system have
been explored extensively since the discovery in the 18th century that electrical signals
modulate muscle movements [2]. The most common method involves implanting metal
electrodes to record and stimulate neurons. Initial efforts used bulk metals that interacted
with entire neuron populations. Patch clamping was introduced in 1978 in which an
individual neuron ion channel’s electrical activity could be monitored through the use of
a pulled glass pipette [3]. Later technologies focused on interfacing with the brain
through brain machine interfaces (BMI).

Figure 1-2: Experimental setup of Galvani where an electric machine is attached to a
nerve-muscle preparation of frog legs with a metal wire [2]. Electrical stimulation caused
muscle contraction. With kind permission from Springer Science and Business Media.

4
Reliable and robust BMI that enable direct patient control of motor prosthetics are
an ongoing challenge. Implantable probes with single or arrays of neural recording and
stimulation electrodes, simply referred to sometimes as “neural probes,” that reliably
interact and communicate with neurons have served as standard neuroscience research
tools for studies of brain function for decades but have yet to achieve their potential in
clinical use. Individuals with neurodegenerative disorders or other neurological deficits
may benefit from neural probe technologies to restore brain function lost by disease
(stimulating electrodes) or reestablish previously severed brain-limb connections as in
limb control (recording electrodes).
Earlier BMI technology consisted of individual metal microwires (>75 µm
diameter [4]) with de-insulated tips for recording neural signals [5]. Later, arrays of
microwires held together with adhesive led to improved spatiotemporal neural recordings
[6]. With the emergence of microfabrication technologies, silicon (Si) shanks supporting
patterned microelectrodes were developed for use as neural probes; today, two probe
types are prevalent. The Michigan probes, introduced in 1986, consists of Si
micromachined probes with multiple electrode recording sites patterned along the shank
(later commercialized by NeuroNexus) [7]. The Utah probe array, introduced shortly
thereafter, consists of arrays of Si micromachined shanks each with an individual metal
recording site located at the shank tip (later commercialized by Blackrock Microsystems)
[8]. Dense arrays of recording sites are possible with either technology platform.
5

Figure 1-3: (a) A representative microwire array adapted from [6] © 1968 IEEE. (b) Utah
array from [9]. Reprinted by permission from Macmillan Publishers Ltd: Nature
Neuroscience, copyright 2002. (c) Michigan probe from [10] © 2008 IEEE.
1.2.2 Chemical
There has been effort to investigate the microenvironments of neurons since
microgradients can influence cellular behavior [11]. For example, axonal growth
direction has been demonstrated to be controlled by gradients of neurotrophic factors [12].
The cellular microenvironment contains a number of chemical gradients that
fundamentally influence cell functions such as cell signaling, nutrient uptake, waste
disposal, and gas exchange. Thus, control of these microenvironments is sought to enable
precise studies of cells and tissues both in vitro and in vivo. The conventional method to
control microenvironments is by releasing soluble factors to cells through bulk fluid flow
delivered by hand-pipetting or other forms of perfusion and targets only large cell
populations. However, this crude method does not allow for precise modulation of the
cellular microenvironment for a targeted cell or small cell group within a population [13,
14]. A focused chemical stimulus in the targeted microenvironment of a cell within a
population can evoke a specific response that elucidates its relationship within the
cellular network or tissue organization.
6
Devices that allow precise and repeatable modulation of the cellular
microenvironment would benefit biological applications in which precise control over the
microenvironment is required such as stem cell niches, selective differentiation,
metaplasia (reversible transformation of cell from normal to abnormal state),
synaptogenesis, and chemotaxis [13, 15-20]. Microfluidic dosing devices that enable
selective stimulation of cells have been introduced [13, 14, 21-24]. Devices that produce
controlled laminar flows defined by microchannels have demonstrated cellular and sub-
cellular resolution of cell inactivation by changing the localized microenvironment of
cells cultured in the microchannels [14, 22, 23]. For example, Tourovskaia et al.
demonstrated localized stimulation of muscle cells cultured within a main
poly(dimethylsiloxane) channel 2 cm long, 1500 µm wide, and 250 µm high with a
hydrodynamically focused agrin stream (~100-150 µm wide) delivered by a central inlet
channel connecting to the main channel [14]. However, these devices are suitable for cell
culture studies but are difficult to use with tissue samples.
Another method of controlling the microenvironment is to release soluble factors
from a small aperture (2-8 µm in diameter) [13, 14, 21]. This technique can produce
repeatable localized chemical gradients at multiple locations on a device and is applicable
for tissue applications. Agrin was focally applied via a microaperture array to myotubes
cultured on top of the array; apertures (ranging 2-8 m diameter) were etched into a low-
stress silicon nitride membrane and useful in studies of neuromuscular synaptogenesis
[13]. Technology for the localized delivery of soluble agents is also beneficial not only
for the study of in vitro or ex vivo systems but also for in vivo therapeutic applications
7
such as drug delivery and neurotransmitter-based biomimetic neural prostheses [13, 14,
21, 24-26]. A neural prostheses intended for restoring vision was devised in which a flow
control channel and aperture were used to control neurotransmitter release from each
aperture and thus achieve localized neurotransmitter stimulation of neurons in the visual
pathway [24].
1.2.3 Optical
The discovery and application of channelrhodopsins to neurons through
optogenetics significantly advanced the study of neurobiology earlier this decade [27].
Channelrhodopsins are specialized ion channels that are activated by light.
Channelrhodopsin-2 (ChR2) was first integrated into neurons by genetic transfection in
2005 [28]. ChR2 is a light activated cation channel that opens when activated by blue
light. This opening of ion channels within a neuron causes ion flows that ultimately lead
to the generation of an action potential if a threshold membrane voltage is crossed. Thus,
neurons that are transfected and express ChR2 are controlled by researchers wishing to
study neural circuits. Researchers are able to generate action potentials in neurons at will
and so probe the neural circuitry of the neural tissue without invasively attaching
electrodes.
8

Figure 1-4: ChR2 mechanism. The Na
+
ion channel is opened when exposed to blue light
(λ = 470 nm) of sufficient optical power. The subsequent flow of Na
+
ions across many
opened channels lead to action potential generation.

Additionally, an archael light-driven chloride pump (NpHR) from Natronomonas
pharaonis has been developed for neuronal studies which acts in concert with ChR2 [29].
NpHR acts similarly to ChR2 in that it is an ion channel that is activated by light.
However, the activation of NpHR serves the opposite function of ChR2; instead of
inducing an excitation potential in a neuron, NpHR inhibits the generation of an action
potential.
The use of these channelrhodopsins has allowed researchers to advance the study
of the nervous system beyond what was achievable with decades old tools such as
microelectrodes. With genetically transfected neurons that express channelrhodopsins,
researchers are able to optically probe neural circuitry [30].

9

Figure 1-5: Photostimulation of an optogenetic neuron causes generation of action
potentials due to the actions of channerhdopsin ion channels.

Neurotransmitters that are photocaged have also been developed [31]. These
neurotransmitters are inactive until exposure to optical energy induces a photolytic
chemical reaction thereby releasing the neurotrasnmitter to act on the neuron. Caged
neurotransmitters enable complete diffusion across a cell population without loss of
neurotransmitter neuroactivity as opposed to natural neurotransmitters which possess
short half-lives [31]. A finely focused laser is used to uncage and activate a select
concentration of neurotransmitters at a specific synaptic location (confined to the laser
beam diameter [32]) and time which presents the cell with a physiologically relevant
neurotransmitter challenge. Subsequent laser bursts can be exploited to present complex
synaptic neurotransmitter inputs in a precisely controlled manner without disturbing the
surrounding media since no fluid exchange is required. In principle, multiple caged
neurotransmitters can be activated simultaneously or in sequence with different
wavelengths providing a high degree of spatiotemporal control over synaptic input. Thus,
neurotransmitter activity can be tightly controlled both spatially and temporally by
optical methods with no physical perturbation.
10

Figure 1-6: Neurotransmitter uncaging leads to an increase in localized concentration of
biologically active neurotransmitter that acts upon the neuron.

Optical interrogation of neurons utilizing caged neurotranmitters or optogenetic
neurons expressing channelrhodopsins requires directing a light source to the target
location. This is commonly achieved through delivering light through a microscope
objective to the stage where the target is placed [33]. A drawback of this method is that
only one light source is available and the stage needs to be moved for optically targeting
other points. A fast-scanning laser system is used to overcome this limitation and
optically target multiple locations in a precise and short amount of time (< 4
microseconds) with custom hardware (multiple mirror actuators, lenses, controllers, and
programmable shutter) mounted to a microscope [34]. Another less infrastructure
intensive method utilizes optical fibers to guide light to target locations on a neuron [31].
Fibers are positioned with micromanipulators and multiple fibers can be used per stage
enabling multiple targets. Optical fiber setups are not limited to in vitro use, fibers have
been inserted into frogs for uncaging of neurotransmitters [35] as well as brains of
optogenetic mice [36].
Microdevices have been constructed to take advantage of optical neural
interfacing techniques. Optical fibers can be gold plated forming “optrodes” such that
11
they can deliver light and electrically record simultaneously [37]. Multiple electrical
recordings per light source has been achieved by fabicating microelectrode arrays
(MEA’s) adjacent to an integrated waveguide on a neural probe (Figure 1-7) [38].
Photocaging experiments have been achieved with the use of a fused silica (glass) chip
with etched waveguides that direct light to a chamber that houses the photocaged solution
and cells enabling multiple light sources without the use of optical fibers or
micromanipulators [39].

Figure 1-7: A neural probe integrated with optical waveguides for optogenetic
applications from [38] © 2011 IEEE.

1.3 The Case for MEMS Neural Interfaces
Conventional methods of neural interfacing rely on macro-scale technologies and
as a consequence, are unable to interact with neurons on a cellular level without great
difficulties or limitations. The conventional methods of electrical, chemical, and optical
interfacing are patch clamping, puffer pipetting, and microscopy respectively. These
technologies are capable of interacting with neurons on a subcellular scale albeit with
great technical effort and limited throughput.
12
The gold standard of neural electrophysiology is the patch clamp. Single ion
channels can be recorded from with a pulled glass pipette and subcellular ion flows or
current giving rise to the generation of an action potential can be observed with this
technique. Micromanipulators are first utilized to manually position a pipette to a target
cell membrane; suction is then applied in order to achieve a seal for electrical recording.
This seemingly straightforward procedure is actually riddled with technical difficulties; a
successful patch clamp experiment is highly dependent upon operator skill. Cells possess
limited viability under the microscope stage and so need to be patched as quickly as
possible. The suctioning procedure requires subjective judgment as to how much force to
apply; too little and the required “gigaseal” is not obtained. Too much and the cell lyses.
This laborious process is time-consuming and not scalable limiting the experimental
throughput. Achieving just one patch clamp per cell is difficult enough with a success
rate of ~30% for skilled operators [40], thus multiple patch clamps per cell or neural
network are rare. Moreover, there is a spatial limit imposed by the micromanipulators; a
setup can only hold a few at once. A robotic patch clamper has been developed in an
effort to automate the difficult process however it was only successful around half the
time [40] and did not address the spatial limit preventing multiple simultaneous
recordings.
Puffer pipetting is similar to patch clamping in that it utilizes a pulled glass
pipette positioned close to a target location with the use of a micromanipulator. Instead of
recording, the pipette is a channel through which a fluid stream of chemicals can be
delivered quickly with the use of a solenoid valve [41] or slowly through passive
13
diffusion. Nanoliter volumes can be delivered in a spatiotemporal manner. However,
convective flow may distort the localized chemical concentration gradient negatively
impacting the experiment and impeding repeatability. As with patch clamping, there is a
limit to how many micromanipulators and thus pipettes can be positioned and utilized at
once, limiting throughput with this mode of chemical interfacing.
For a long time conventional optical neural interfacing was limited to passive
interaction through microscopy. Microscopy exposes the sample cell to condensed light
which is then collected through an objective to obtain a magnified image of the cell.
Since neurons, with the exception of retina, are not normally affected by light,
microscopy only passively records neural behavior with regards to growth or movement.
The advent of photocaged neurotransmitters and optogenetics has made it possible to
actively interface with neural networks with light. The microscope light source can be
switched to a wavelength such that it uncages neurotransmitters or generates action
potentials [33]. In principle, targeted subcellular regions can be illuminated by moving
the microscope stage to the desired position and condensing the light to a subcellular area.
As with the other conventional methods, there is limited throughput since only one spot
can be illuminated at a time.
Ideally, a neural interface will be able to interact with neurons on a cellular level
as one neuron and its signals make up the basic unit of the nervous system. Neural cell
body diameters are in the range of microns (10
-6
meters) and thus neural interfaces in the
size of microns are best suited for interfacing with neural networks. The spacing between
the neural interface and neuron(s) should be minimized as this will produce better signal
14
through minimal noise in electrical interfacing, minimal diffusion due to Brownian
motion in chemical interfacing, and minimal light attenuation and diffusion in optical
interfacing.
Neural interfaces that can be precisely fabricated in the micron range with the
complexity of neural networks require the use of microfabrication. Microfabrication
technologies allow for the integration of optics, mechanical structures, sensors, actuators,
and electronics on one device [42]. Structures with fine features can be constructed in
micron ( m) and submicron scales with a high degree of control, repeatability, and
complexity. Microdevices with multiple elements can be fabricated in various dimensions
and form factors that match the micron size and geometrical dependencies of neurons
while minimizing the cell-interface distance through MEMS fabrication technologies.
1.4 MEMS Fabrication Technology
MEMS fabrication technologies share many characteristics with the processes
involved in making IC (integrated circuits) reflecting their common history. The advent
of mass produced IC involved the development of single crystal Si bulk material that was
then thinly sliced (commonly 500 m thick) with a circular form factor-commonly
known as a wafer. This pure Si material could then be doped with various impurities to
produce and modulate different electrical properties that are desirable for producing
transistors, fundamental components of IC. The initial distinction between MEMS and IC
could be seen in the production of a transistor that utilized a physical gate whose position
was controlled by electrostatic forces by Nathanson and Wickstrom [43]. In addition to
15
circuitry, MEMS can integrate a variety of different elements that are not necessarily
electrical such as moving mechanical parts or optical components together to create a
functional micromachine or microdevice.
The intellectual birth of MEMS can be traced to the seminal talk given by the
famous physicist Richard Feynman titled “There’s Plenty of Room at the Bottom” to the
American Physical Society in 1959 [44]. Here, the Nobel Laureate imagined a new field
of science and engineering that delved into the miniature and how that might affect
society. Notably, he pointed to biological cellular systems which functioned as complex
micromachines with characteristics of computers as an inspiration. He then predicted
bioMEMS in the form of microdevices which can perform microsurgery on a patient.
Feynman then ended his talk with a challenge to create the technology for realizing such
microdevices; inspiring generations of scientists and engineers to delve into micro and
nano-scale fabrication.
Feynman imagined writing an entire encyclopedia onto a pin head by use of a
focused electron beam through a photo process, describing the fundamental process on
which all if not most other micromachining techniques rely; lithography. Lithography
was originally invented for producing artwork using a smooth limestone as a master for
transferring illustrated ink patterns onto paper by creating hydrophobic and hydrophilic
regions through a chemical process. The IC industry made use of photoresists, chemicals
which are affected by light to transfer patterns onto a Si wafer to define circuit elements
through photolithography. Photolithography transfers light through a photomask to the
photoresist (Figure 1-8). The photoresist is then developed in the pattern of the
16
photomask due to the light exposing certain photoresist regions. Lenses can be utilized to
shrink the light pattern to expose a small region of photoresist resulting in a smaller
pattern then the original or precise plotters can be used to print photomasks with small
features. Realization of MEMS devices also utilizes photolithography to define and
transfer small patterns onto a substrate which then become the elements of a microdevice
with or through subsequent processes such as metal deposition, etching, chemical vapor
deposition, oxidation, etc.

Figure 1-8: Illustration of photolithography. The Chinese character for Kuo in the
photomask is transferred to the substrate. A light source, commonly in the form of a Hg
lamp, has its light focused and collimated through optics. The light passes through a
photomask, exposing the photopattern to the photoresist covered wafer substrate which is
then developed. Mechanical stages can be used to precisely move the substrate for
alignment purposes prior to exposure.
17
1.4.1 Fabrication Technologies Used in this Work
Numerous different microfabrication technologies are available to create just
about any type of device; their descriptions are beyond the scope of this dissertation and
can be found elsewhere [42, 43]. What follows are descriptions of the microfabrication
techniques used to realize the devices described in this work.
1.4.1.1 SU-8 Photoresist
SU-8 is an epoxy-based negative photoresist introduced in 1995 by IBM created
from commercially available EPON SU-8 resin by adding photoinitiators so that it
became photopatternable [45]. The patent for SU-8 was later licensed to Microchem and
Gersteltec who have become the major suppliers for this resist. SU-8 is used to cheaply
realize high-aspect ratio structures and is chemically inert as well as mechanically robust
through a large thermal range due to its high degree of cross-linking, leading to its
popularity in many MEMS applications. Optically, SU-8 is relatively transparent post-
development however it autofluoresces. SU-8’s viscosity can be tuned leading to many
different formulations of SU-8 that allow for variations in achievable thickness (up to
mm range), adhesion, and aspect ratios (over 14.5) [46].
18

Figure 1-9: Chemical structure of SU-8 polymer (modified from [45]).

SU-8 is most commonly utilized in bioMEMS applications for constructing molds
for soft lithography. Microchannel SU-8 master molds are defined by lithography; upon
development, silicone such as polydimethylsiloxane (PDMS) is cast and cured over the
mold. Upon release, a silicone device with the desired microchannel features is obtained.
Recently, SU-8 has been integrated into neural probes in the form of optical waveguides
for optogenetic studies [47]. SU-8 has also been used as insulating material for MEA’s or
utilized for microneedle construction. There are claims that SU-8 is biocompatible
however the veracity of this claim is questionable as the photoresist has been shown to
leech cytotoxic chemicals over time into cell culture [48].
The processing of SU-8 begins with its spin coating onto a substrate. A softbake
is then performed to evaporate the solvents and leave a solid film of SU-8 that can be
photopatterned. A hot plate is preferably utilized as the use of an oven can create a “skin”
that traps solvent within the resist. A leveled hot plate is critical at this point as the
viscous nature of SU-8 can create an uneven coating if baked over an uneven surface.
19
The resist is exposed under a photomask with an appropriate exposure dose that
corresponds to the resist thickness. The UV light initiates a chemical reaction in the
exposed SU-8 that leads to cross-linking. Next, a post-exposure bake (PEB) is performed
for SU-8 cross-linking as an elevated temperature (~90 °C) is required to complete this
step. Patterns can be visually seen at this point. SU-8 is then developed in a solvent
solution to remove the uncross-linked regions. A rinse with isopropyl alcohol (IPA) is
commonly utilized to test for complete development; the presence of white streaks due to
IPA is an indicator that there is underdevelopment. Finally, a hardbake can be performed
on the developed SU-8 structure.
1.4.1.2 Parylene Microfabrication
Parylene, the trade name for poly(para-xylylene), is a polymer that was
commercialized by Union Carbide Corporation in the 1960’s. Many different types of
Parylene were developed but the most popular and well known is Parylene C. Parylene C
is a United States Pharmacopeia (USP) class VI polymer that is conformally deposited
through a chemical vapor deposition (CVD) process. The deposited polymer film has a
relatively low Young’s modulus of ~4 GPa and is optically transparent [49]. Very thin
layers (~1 µm) can be deposited that are pinhole free. Parylene C is commonly used in
electronics as an insulating material due to its chemical inertness. This property is useful
in biological applications as Parylene C is hydrophobic and thus resistant to biofouling.

20

Figure 1-10: Chemical structure of Parylene C polymer. The characteristic single chlorine
(Cl) attached to the benzene ring is what differentiates Parylene C from other Parylene
types.

The most useful property of Parylene C is that it can be selectively etched with O
2
plasma with high resolution (µm range) making it micromachinable [50]. RIE (reactive
ion etching) of Parylene C with a patterned PR mask transfers the pattern into the
Parylene layer in an isotropic fashion. Empirically, the etch selectivity between the
masking photoresist (AZ) and exposed Parylene has been found to be 1:1. Anisotropic
etching is also possible with a DRIE switched chemistry process with achievable aspect
ratios of up to two (2) [51, 52] This allows for a variety of Parylene based devices in the
form of pumps, valves, and neurocages amongst others.
Parylene can be shaped with heat due to its thermoplastic semicrystalline nature.
This process is referred to as thermoforming [53]. A Parylene film is attached to a mold
and heated past its glass transition temperature (~60-90 °C) but below its melting
temperature (~290 °C) for a set time. Upon cooling, the Parylene recrystallinizes and
hardens conformally to the mold and can be released; holding its new shape. This
Parylene process enables many different form factors for Parylene-based devices; which
is useful for creating devices that conform to biological shapes as is often the case for
implantable bioMEMS.
21
1.4.1.1 Metal deposition and liftoff
Metals can be patterned for MEMS with the use of photoresist. A photoresist is
patterned and developed exposing certain regions of the substrate while masking the rest.
Metals are then deposited over the entire area through use of a deposition method such as
sputtering or e-beam evaporation. Finally, the masked region is removed by a liftoff
process in which an acetone soak dissolves the resist and the unwanted metal on top is
lifted off; leaving the patterned metal on the substrate. Acetone access to the photoresist
through the photoresist step is necessary for completion of the liftoff process; negative
step profiles are less likely to be coated with metal and thus produce better results in
general compared to positive profiles (Figure 1-11b).

Figure 1-11: Metal patterning through the liftoff process. (a) Photoresist is first patterned
and developed with a negative step profile. (b) Metal is then deposited over the entire
surface. (c) Next, an acetone soak attacks and dissolves the photoresist mask lifting off
the undesired metal. (d) Finally, the desired metal pattern is obtained.

E-beam evaporation is used for depositing metals in this work. A metal source
within a crucible is heated to its evaporation point within a vacuum chamber with the
22
substrate under high vacuum. A high-temperature filament is induced to generate an
electron beam (e-beam) by passing a high current through it. This e-beam is directed to
the metal source by use of a magnet which then melts the metal; evaporating it. The
evaporated metal in gaseous form then coats everything within the chamber upon
precipitation. A high vacuum (preferably <10
-7
Torr pressure) is necessary for uniform
metal deposition as the presence of atmospheric gas molecules impedes the mean free
path of the gaseous metal. Also, a higher vacuum prevents metal oxidation and thus
contamination of the deposited metal film as in the case with aluminum. The high
temperatures generated through evaporation of metal raises the temperature of the
chamber and the substrate; possibly leading to substrate temperatures of >60 °C.
The heat experienced by the substrate can affect the process depending on the
materials utilized. In this work, Parylene C is used often and is known to experience
thermal degradation as evidenced by delamination from the wafer. Thus it is necessary to
minimize the heat Parylene C is exposed to. This can be achieved by making use of
multiple e-beam depositions with adequate cool-downs in-between so that the chamber is
not exposed to excessive temperatures for thicker coatings (e.g. >600 Å for platinum (Pt)).
Photoresist may also reflow at high temperatures destroying the desired pattern; thus it
may be advantageous to reduce the temperature at which deposition occurs.
A metal that deserves highlighting is platinum (Pt) due to its usefulness for
bioMEMS, especially within neural interfaces. Pt is a common material used for neural
electrodes and preferred over other metals since it is known to be inert under biological
environments and possesses a high charge injection limit [54]. Pt has been utilized as
23
neural recording electrode material in the fabrication of microwires, Michigan probes,
and Utah arrays with much success [55]. A retinal prosthesis in the form of a Pt
multielectrode array (MEA) patterned on a Parylene substrate was successful in
stimulating and recording from retinal tissue [56]. Pt is also used for electrochemical
(EC) sensor and flow sensor fabrication [43, 57]. No adhesion layer is required when
depositing Pt onto Parylene substrate [58] whereas a titanium (Ti) adhesion layer is
needed when depositing onto soda lime.
1.5 MEMS Neural Interfaces
This work investigates each of the neural interfacing modalities: electrical,
chemical, and optical through the development of microdevices that improve neural
studies over that of traditional technologies by reaping the benefits of microfabrication
and miniaturization. An improved BMI technology in the form of a Parylene sheath probe
is first introduced for long-term neural recordings. Controlled chemical interaction with
cells is then investigated through the development of biomimetic chemical interfaces.
Finally, a micromachined mirror array is described for optically stimulating optogenetic
or light-sensitive cells. Novel microfabrication technologies were developed to realize
these devices and are described in detail. The development and use of these
micromachined neural interfaces will improve understanding of the nervous system and
potentially lead to better treatments for neurophysiological disorders as well as catalyze
improvements in neural engineering and neuroprosthetics.
24
REFERENCES
[1] E. R. Kandel, J.H. Schwartz, and T.M. Jessell, Principles of neural science, 4th ed.
New York: McGraw-Hill, Health Professions Division., 2000.
[2] L. Galvani, De viribus electricitatis in motu musculari commentarius vol. 7: Ex
Typographia Instituti Scientiarum Bologna, 1791.
[3] E. Neher, B. Sakmann, and J. H. Steinbach, "The extracellular patch clamp: A
method for resolving currents through individual open channels in biological
membranes," Pflügers Archiv, vol. 375, pp. 219-228, 1978/07/01 1978.
[4] P. Gilgunn, R. Khilwani, T. Kozai, D. Weber, X. Cui, G. Erdos, et al., "An ultra-
compliant, scalable neural probe with molded biodissolvable delivery vehicle,"
presented at the The 25th IEEE International Conference on Micro Electro
Mechanical Systems, Paris, France, 2012.
[5] D. A. Robinson, "The electrical properties of metal microelectrodes," Proceedings
of the IEEE, vol. 56, pp. 1065-1071, 1968.
[6] M. A. L. Nicolelis, D. Dimitrov, J. M. Carmena, R. Crist, G. Lehew, J. D. Kralik,
et al., "Chronic, multisite, multielectrode recordings in macaque monkeys,"
Proceedings of the National Academy of Sciences, vol. 100, p. 11041, 2003.
[7] K. Najafi and K. D. Wise, "An implantable multielectrode array with on-chip
signal processing," Solid-State Circuits, IEEE Journal of, vol. 21, pp. 1035-1044,
1986.
[8] P. K. Campbell, K. E. Jones, R. J. Huber, K. W. Horch, and R. A. Normann, "A
silicon-based, three-dimensional neural interface: manufacturing processes for an
intracortical electrode array," IEEE Transactions on Biomedical Engineering, vol.
38, pp. 758-768, 1991.
[9] J. P. Donoghue, "Connecting cortex to machines: recent advances in brain
interfaces," nature neuroscience, vol. 5, pp. 1085-1088, 2002.
25
[10] K. D. Wise, A. M. Sodagar, Y. Yao, M. N. Gulari, G. E. Perlin, and K. Najafi,
"Microelectrodes, microelectronics, and implantable neural microsystems,"
Proceedings of the IEEE, vol. 96, pp. 1184-1202, 2008.
[11] T. M. Keenan and A. Folch, "Biomolecular gradients in cell culture systems," Lab
on a Chip, vol. 8, pp. 34-57, 2008.
[12] S. J. Butler and J. Dodd, "A Role for BMP Heterodimers in Roof Plate-Mediated
Repulsion of Commissural Axons," Neuron, vol. 38, pp. 389-401, 2003.
[13] T. F. Kosar, A. Tourovskaia, X. Figueroa-Masot, M. E. Adams, and A. Folch, "A
nanofabricated planar aperture as a mimic of the nerve-muscle contact during
synaptogenesis," Lab on a Chip, vol. 6, pp. 632-638, 2006.
[14] A. Tourovskaia, T. F. Kosar, and A. Folch, "Local induction of acetylcholine
receptor clustering in myotube cultures using microfluidic application of agrin,"
Biophysical Journal, vol. 90, pp. 2192-2198, Mar 2006.
[15] H. R. Bourne and O. Weiner, "Cell polarity: A chemical compass," Nature, vol.
419, pp. 21-21, 2002.
[16] H. Lin, "The stem-cell niche theory: lessons from flies," Nat Rev Genet, vol. 3, pp.
931-940, 2002.
[17] H. J. Song and M. M. Poo, "The cell biology of neuronal navigation," Nature Cell
Biology, vol. 3, pp. E81-E88, Mar 2001.
[18] A. Spradling, D. Drummond-Barbosa, and T. Kai, "Stem cells find their niche,"
Nature, vol. 414, pp. 98-104, 2001.
[19] D. Tosh and J. M. W. Slack, "How cells change their phenotype," Nature Reviews
Molecular Cell Biology, vol. 3, pp. 187-194, Mar 2002.
[20] M. Ueda, Y. Sako, T. Tanaka, P. Devreotes, and T. Yanagida, "Single-Molecule
Analysis of Chemotactic Signaling in Dictyostelium Cells," Science, vol. 294, pp.
864-867, 2001.
26
[21] M. C. Peterman, J. Noolandi, M. S. Blumenkranz, and H. A. Fishman, "Localized
chemical release from an artificial synapse chip," Proceedings of the National
Academy of Sciences of the United States of America, vol. 101, pp. 9951-9954, Jul
2004.
[22] H. Kaji, M. Nishizawa, and T. Matsue, "Localized chemical stimulation to
micropatterned cells using multiple laminar fluid flows," Lab on a Chip, vol. 3, pp.
208-211, 2003.
[23] S. Takayama, E. Ostuni, P. LeDuc, K. Naruse, D. E. Ingber, and G. M.
Whitesides, "Selective chemical treatment of cellular microdomains using
multiple laminar streams," Chemistry & Biology, vol. 10, pp. 123-130, Feb 2003.
[24] N. Z. Mehenti, H. A. Fishman, and S. F. Bent, "A model neural interface based on
functional chemical stimulation," Biomedical Microdevices, vol. 9, pp. 579-586,
Aug 2007.
[25] J. T. Santini, M. J. Cima, and R. Langer, "A controlled-release microchip,"
Nature, vol. 397, pp. 335-338, 1999.
[26] B. Vastag, "Future Eye Implants Focus on Neurotransmitters," JAMA: The
Journal of the American Medical Association, vol. 288, pp. 1833-1834, 2002.
[27] G. Nagel, D. Ollig, M. Fuhrmann, S. Kateriya, A. M. Musti, E. Bamberg, et al.,
"Channelrhodopsin-1: a light-gated proton channel in green algae," Science, vol.
296, pp. 2395-8, Jun 28 2002.
[28] E. S. Boyden, F. Zhang, E. Bamberg, G. Nagel, and K. Deisseroth, "Millisecond-
timescale, genetically targeted optical control of neural activity," Nature
Neuroscience, vol. 8, pp. 1263-1268, Sep 2005.
[29] F. Zhang, L. P. Wang, M. Brauner, J. F. Liewald, K. Kay, N. Watzke, et al.,
"Multimodal fast optical interrogation of neural circuitry," Nature, vol. 446, pp.
633-9, Apr 5 2007.
[30] G. Miller, "OPTOGENETICS: Shining New Light on Neural Circuits," Science,
vol. 314, pp. 1674-1676, December 15, 2006 2006.
27
[31] D. W. Godwin, D. P. Che, D. M. OMalley, and Q. Zhou, "Photostimulation with
caged neurotransmitters using fiber optic lightguides," Journal of Neuroscience
Methods, vol. 73, pp. 91-106, Apr 25 1997.
[32] H.-U. Dodt, M. Eder, A. Frick, and W. Zieglgänsberger, "Precisely localized LTD
in the neocortex revealed by infrared-guided laser stimulation," Science, vol. 286,
pp. 110-113, 1999.
[33] E. M. Callaway and L. C. Katz, "Photostimulation using caged glutamate reveals
functional circuitry in living brain slices," Proceedings of the National Academy
of Sciences, vol. 90, pp. 7661-7665, 1993.
[34] C. Boucsein, M. Nawrot, S. Rotter, A. Aertsen, and D. Heck, "Controlling
synaptic input patterns in vitro by dynamic photo stimulation," J Neurophysiol,
vol. 94, pp. 2948-58, Oct 2005.
[35] S. Giszter, J. Scabich, G. Ellis-Davies, K. Simansky, and M. Lemay, "Photolytic
uncaging of neurotransmitters as a control and stimulation device for neural
tissues," in Engineering in Medicine and Biology, 2002. 24th Annual Conference
and the Annual Fall Meeting of the Biomedical Engineering Society EMBS/BMES
Conference, 2002. Proceedings of the Second Joint, 2002, pp. 2072-2073.
[36] A. M. Aravanis, L.-P. Wang, F. Zhang, L. A. Meltzer, M. Z. Mogri, M. B.
Schneider, et al., "An optical neural interface: in vivo control of rodent motor
cortex with integrated fiberoptic and optogenetic technology," Journal of Neural
Engineering, vol. 4, pp. S143-S156, 2007.
[37] J. Zhang, F. Laiwalla, J. A. Kim, H. Urabe, R. Van Wagenen, Y.-K. Song, et al.,
"Integrated device for optical stimulation and spatiotemporal electrical recording
of neural activity in light-sensitized brain tissue," Journal of Neural Engineering,
vol. 6, 2009.
[38] M. Im, I.-J. Cho, F. Wu, K. D. Wise, and E. Yoon, "Neural probes integrated with
optical mixer/splitter waveguides and multiple stimulation sites," in Micro Electro
Mechanical Systems (MEMS), 2011 IEEE 24th International Conference on, 2011,
pp. 1051-1054.
28
[39] D. Ghezzi, R. Vazquez, R. Osellame, F. Valtorta, A. Pedrocchi, G. Valle, et al.,
"Femtosecond Laser Microfabrication of an Integrated Device for Optical Release
and Sensing of Bioactive Compounds," Sensors, vol. 8, pp. 6595-6604, 2008.
[40] S. B. Kodandaramaiah, G. T. Franzesi, B. Y. Chow, E. S. Boyden, and C. R.
Forest, "Automated whole-cell patch-clamp electrophysiology of neurons in
vivo," Nature methods, vol. 9, pp. 585-587, 2012.
[41] K. Spitzer and J. Bridge, "A simple device for rapidly exchanging solution
surrounding a single cardiac cell," American Journal of Physiology-Cell
Physiology, vol. 256, pp. C441-C447, 1989.
[42] M. J. Madou, Fundamentals of Microfabrication, 2nd ed. Boca Raton: CRC Press,
2002.
[43] E. Meng, Biomedical Microsystems, 1 ed.: CRC Press, 2010.
[44] R. P. Feynman, "There's plenty of room at the bottom," Journal of
Microelectromechanical Systems, vol. 1, pp. 60-66, 1992.
[45] N. C. LaBianca and J. D. Gelorme, "High-aspect-ratio resist for thick-film
applications," in SPIE's 1995 Symposium on Microlithography, 1995, pp. 846-852.
[46] H. Lorenz, M. Despont, N. Fahrni, N. LaBianca, P. Renaud, and P. Vettiger, "SU-
8: a low-cost negative resist for MEMS," Journal of Micromechanics and
Microengineering, vol. 7, p. 121, 1997.
[47] M. Im, I.-J. Cho, F. Wu, K. D. Wise, and E. Yoon, "Neural probes integrated with
optical mixer/splitter waveguides and multiple stimulation sites," in 24th IEEE
International Conference on Micro Electro Mechanical Systems, MEMS 2011,
January 23, 2011 - January 27, 2011, Cancun, Mexico, 2011, pp. 1051-1054.
[48] V. N. Vernekar, D. K. Cullen, N. Fogleman, Y. Choi, A. J. García, M. G. Allen, et
al., "SU-8 2000 rendered cytocompatible for neuronal bioMEMS applications,"
Journal of Biomedical Materials Research Part A, vol. 89A, pp. 138-151, 2009.
29
[49] C. Shih, T. A. Harder, and Y. C. Tai, "Yield strength of thin-film parylene-C,"
Microsystem technologies, vol. 10, pp. 407-411, 2004.
[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.
[51] E. Meng and Y.-C. Tai, "Parylene Etching Techniques for Microfluidics and
BioMEMS," presented at the MEMS 2005, Miami, Florida, 2005.
[52] E. Meng, P.-Y. Li, and Y.-C. Tai, "Plasma removal of Parylene C," Journal of
Micromechanics and Microengineering, vol. 18, p. 045004, 2008.
[53] B. J. Kim, B. Chen, M. Gupta, and E. Meng, "Three dimensional transformation
of Parylene Thin Film Structures Via Thermoforming," in MEMS 2013, Taipei,
Taiwan, 2013, pp. 339-342.
[54] Y. Nam, "Material considerations for in vitro neural interface technology," MRS
bulletin, vol. 37, pp. 566-572, 2012.
[55] M. HajjHassan, V. Chodavarapu, and S. Musallam, "NeuroMEMS: neural probe
microtechnologies," Sensors, vol. 8, pp. 6704-6726, 2008.
[56] D. C. Rodger, A. J. Fong, W. Li, H. Ameri, A. K. Ahuja, C. Gutierrez, et al.,
"Flexible parylene-based multielectrode array technology for high-density neural
stimulation and recording," Sensors and Actuators B: Chemical, vol. 132, pp.
449-460, 2008.
[57] J. T. W. Kuo, L. Yu, and E. Meng, "Micromachined Thermal Flow Sensors—A
Review," Micromachines, vol. 3, pp. 550-573, 2012.
[58] A. V. Vasenkov, "Atomistic modeling of parylene-metal interactions for surface
micro-structuring," Journal of Molecular Modeling, vol. 17, pp. 3219-3228, 2011.

30

2.1 Background
Despite many decades of development current implanted neural probes possess
short recording lifetimes; reliable neural recordings are typically obtained in durations
measured in months. This short lifetime is attributed to the physiological response of
surrounding neural tissue to the implanted neural probe [1]. Upon implantation, brain
micromotion due to blood flow, respiration, and head movement aggravates the tissue
surrounding the device and a physiological response begins that involves inflammation,
glial scar encapsulation of implant, and retraction of neurons from recording sites leading
to a loss of recordable neural signal. Tissue aggravation is attributed to the mechanical
mismatch in elastic moduli between the stiff probe material and surrounding soft neural
tissue. Microwires made of noble metals or micromachined Si shanks (Young's modulus,
E~100-400 GPa) possess Young's moduli many orders higher than neural tissue (E~0.1-6
kPa) [2]. A chronic strain is induced on the probe-tissue interface upon implantation with
resulting shear forces scaling with the moduli difference between probe and tissue [3].
Besides the material mechanical mismatch, the common practice of anchoring the
implanted neural probe to a stationery object such as the skull also increases tissue
aggravation since the skull-anchored probe opposes natural brain micromotion.
Chapter 2 PARYLENE SHEATH PROBES FOR LONG-TERM IN
VITRO SIGNAL RECORDING
31
A neural probe technology which has achieved long recording lifetime (in both
animals and humans) measured in years is the neurotrophic cone electrode [4-10]. This
non-micromachined technology is fundamentally different from other microwire or
silicon-based neural probes. The long term success of the neurotrophic cone electrode is
attributed to its use of growth factors and its hollow structure. The overall structure is
based on a glass cone formed by pulling a heated pipette and then cutting to achieve the
desired length (1-2 mm). Microwires are then manually positioned within the cone and
glued into place. Next, growth factors that promote neural tissue growth such as Matrigel
and nerve growth factor (NGF) are coated within the cone prior to implantation. The
release of factors in vivo is thought to attract neural tissue growth into the cone towards
the microwire recording sites. The ingrowth of neural tissue into the cone mechanically
integrates and anchors the probe to the tissue; thus the detrimental effects of brain
micromotion are minimized by reducing chronic strain at probe-tissue interface. The
ingrowth of neural processes into the cone toward recording sites leads to better neural
signal recording since the distance between neuron and recording site is minimized.

32

Figure 2-1: (Left) Neurotrophic cone electrode concept showing ingrowth of neural
processes toward recording microwires from [11]. (Right) Explanted device with neural
tissue ingrowth from monkey after 6 months of implantation; Scale bar: 100 µm [7]
1
.

Unlike silicon probes, neurotrophic cone electrodes are individually assembled
using a manual labor intensive process that is prone to low yield, throughput, and probe-
probe repeatability. Glass (E~100 GPa) still poses a mechanical mismatch issue with
neural tissue. Finally, microwire spatial requirements limit the number of recording sites
possible per cone; a maximum of four recording sites per cone electrode has been
demonstrated.
To overcome manufacturing and electrode density limitations of cone electrodes
that prevent its widespread use, we developed a batch fabricated Parylene flexible neural
probe using a novel process that enables a three dimensional (3D) sheath structure [12,
13]. Microfabrication technologies allow for precise patterning of multiple recording
electrodes per sheath (eight electrodes for this device). Electrodes can be placed on both
inner and outer surfaces of the sheath. Batch fabrication ensures high throughput with

1
Reprinted from P. R. Kennedy, et al., "The cone electrode: ultrastructural studies following long-term recording in rat
and monkey cortex," Neuroscience Letters, vol. 142, pp. 89-94, 1992., with permission from Elsevier.
33
repeatability of dimensions and high fidelity of characteristics and behaviors throughout
multiple probes. A post-fabrication thermoforming process is used to obtain the final 3D
structure and open the lumen of the sheath. The lumen can accommodate neurotrophic
factor coatings and allows for tissue ingrowth upon implantation thereby anchoring the
neural probe to tissue and improving its recording by attracting neural processes to
interior recording sites. Coatings of eluting anti-inflammatory factors may also be applied
the sheath to mitigate inflammatory and encapsulation processes in the surrounding
tissue.
2.2 Material Selection
Parylene C, a micromachinable and USP class VI biocompatible polymer [14,
15], was selected as the structural material for the 3D sheath structure and cable substrate
as it improves upon the mechanical mismatch between neural probe and neural tissue
since (E~4 GPa is an order of magnitude lower than glass and silicon [16]). Platinum
(Pt), a highly inert and biocompatible metal [14], was selected as the recording electrode
material. Pt is a common material used for neural electrodes and preferred over other
metals since it is known to be inert under biological environments and possesses a high
charge injection limit [17]. Also no adhesion layer is required when depositing Pt onto
Parylene substrate [18].
2.3 Top Electrode Sheath Probes
Probes consist of sheath structures with four Pt electrodes (45 µm diameter) on
each of the inner and outer surfaces (eight total electrodes). Sheath dimensions and
34
electrode placement were selected to match the anatomy of rat barrel motor cortex to
target recordings of vibrissal (whisker)-related neuronal activity.
Probes consist of sheath structures with four Pt electrodes (45 µm diameter) on
each of the inner and outer surfaces (eight total electrodes). Two probe types were
fabricated differing only in the location of the outer electrodes – either directly on the
sheath or the flaps (“wings”) peripheral to the sheath. Sheath dimensions and electrode
placement were selected to match the anatomy of rat barrel motor cortex to target
recordings of vibrissal (whisker)-related neuronal activity.

Figure 2-2: Conceptual drawing of Parylene sheath probes for long-term intracortical
recordings. The 3D sheath structure allows for ingrowth of neural processes toward
recording electrodes inside the sheath.

2.3.1 Fabrication
A bare Si wafer with native oxide was used as a carrier substrate during the
microfabrication process and aided in the subsequent release of Parylene probes from the
wafer.
Sheath probes having sheath-top electrodes were fabricated by first depositing 5
µm Parylene (Figure 2-3). A liftoff process using negative photoresist (AZ 5214 E-IR)
35
was utilized to pattern inner sheath electrodes created with e-beam deposited Pt (2000 Å).
A 1 µm Parylene insulation layer was then deposited and selectively plasma etched to
expose electrodes and contact pads. Sheath outlines were constructed by patterning
sacrificial photoresist (AZ 4620, 9.6 µm) and overcoating with 5 µm Parylene. A dual
layer liftoff scheme (AZ 1518/AZ 4620) with negative sidewall profile was utilized to
pattern outer electrodes on top of the sheath structure and was necessary to ensure that
resulting wire traces were continuous from the top of the microchannel structure to the
base [19]. Pt was then e-beam deposited (2000 Å) to form the outer electrodes. A final 1
µm Parylene insulation layer was deposited and plasma etched to create openings for
outer electrodes and contact pads. A final plasma etch was performed to create sheath
openings and cut out the individual probes. Probes were released from the substrate and
sacrificial photoresist removed with an acetone soak.

Figure 2-3: Fabrication process for sheath probe having “top” electrodes on sheath
surface. Only the final outline of the device is shown.
36
2.3.2 Parylene Thermoforming
A custom fabricated tungsten (MicroProbes for Life Science, Gaithersburg, MD)
or stainless steel microwire (Cooner Wire Co., Chatsworth, CA) matching the desired
probe shape was carefully inserted into the microchannel opening using a microscope
(Figure 2-4). The apparatus was then held in a custom made aluminum jig and placed into
a vacuum oven (V0914A, Lindberg/Blue, Asheville, NC) for thermoforming to produce
the final 3D sheath shape. Although this process was performed on individual sheaths,
this technique is easily extended to linear sheath arrays with corresponding microwire
arrays.

Figure 2-4: Thermoforming process steps. (a) Released probes were (b) shaped around a
microwire mold and thermally treated. (c) Subsequently, wires were removed to reveal
the final structure.

Parylene is a thermoplastic that can be thermoformed with a mold to transform its
as-fabricated shape [20, 21]. The thermoforming process takes place at 200°C (well
above Parylene's glass transition temperature of ~90°C [22]), results in increased
crystallinity of the polymer, and allows the thermoplastic film to take on the shape of the
mold even after the mold is removed. Parylene sheaths can thus be obtained by
37
thermoforming following the surface micromachining process. Thermoforming was
performed under vacuum to prevent Parylene oxidative degradation. Nitrogen purging
was used to minimize oven oxygen content. The probes were soaked at 200°C for 48
hours. After cooling to room temperature, the microwire was easily removed without the
assistance of any release coatings.
2.3.3 Electrical Packaging
A linear contact pad array located at the end of the Parylene ribbon cable was
attached using adhesive to a thin polyetheretherketone (PEEK) sheet and inserted directly
into a zero insertion force (ZIF) connector. This method allowed simple, reversible
electrical connections with contact pads (Figure 2-5a).

Figure 2-5: (a) Sheath probe with integrated Parylene cable attached to a ZIF connector
for external electrical connections. (b) Released sheath probe. The 3D Parylene sheath
structure holds its shape post-thermoforming.

2.3.4 Results
Sheath probes with different sheath shapes (A, B, and C) were successfully
fabricated and thermoformed (total of three different designs, Figure 2-6, Table 2-1). The
different shapes were for testing which design would be optimal for implantation into rat.
Thermoformed Parylene sheaths were mechanically robust and could withstand repeated
indentation by a probe.
38

Table 2-1: Dimensions of the different fabricated sheath structures for top electrode
designs.
Sheath
Shape
Base
Opening
Diameter
Tip
Opening
Diameter
Sheath
Length
A, tapered 300 m 50 m 800 m
B, tapered 450 m 50 m 800 m
C, cylindrical 300 m 300 m 800 m

Figure 2-6: Post-thermoformed probes of the three different sheath shapes.

The dual photoresist layer liftoff method successfully patterned the outer surface
electrodes and traces despite the step from the periphery to the top of the sheath. Lead
resistances were measured by directly probing the top electrodes and corresponding
contact pads (typically 325-389 Ω). No lead discontinuities were observed. However,
39
some sheath-top electrodes exhibited tensile fracture at locations with high strain (Figure
2-7). Interestingly, inner sheath electrodes did not crack.

Figure 2-7: (a) Top electrodes were prone to fracture (arrows) due to the strain induced
by opening the Parylene microchannel for thermoforming. (b) An SEM image of a top
electrode showing multiple cracks.

The presence of cracks in the top electrodes made the probe unsuitable for
implantation since these electrodes were destroyed and unable to record signals. While
not all thermoformed probes experienced electrode cracking the fabrication yield of
useful sheath probes was adversely impacted enough to warrant redesign.
2.4 Wing Electrode Probes
Outer sheath electrodes were moved to the flaps (“wings”) peripheral to the
sheath to improve fabrication yield (Figure 2-8). Moving electrodes to the periphery and
away from potential strain induced by thermoforming would mitigate electrode cracking
and thus device loss. These wing electrodes may experience improved recordings by
providing greater access to surrounding neurons due to their peripheral locations [23].
40

Figure 2-8: Comparison of top electrode (a) and wing electrode (b) probes. Scale bar is
150 m. Moving outer electrodes from the top to the wings of the sheath avoids straining
these recording sites when opening the Parylene structure with a microwire for
thermoforming.

2.4.1 Fabrication
The “wing” sheath probes simplified the fabrication process by moving the outer
electrodes to the periphery; this reduced the number of steps required and also prevented
occasional cracking of the top electrodes encountered during the sheath forming process
(Figure 2-9). As before, a bare Si wafer with a native oxide layer was coated with 5 µm
Parylene. Electrodes, traces, and contact pads were liftoff patterned using AZ 5214 E-IR
and e-beam deposited Pt (2000 Å). A 1 µm Parylene insulation layer was then deposited,
patterned, and plasma etched, exposing recording sites and contact pads. Next, AZ 4620
(9.6 µm) was patterned and covered with 5 µm Parylene to create sheath structures.
Selective plasma etching of Parylene exposed peripheral electrodes, contact pads, and
sheath openings while simultaneously defining each individual probe. Finally, the probe
was lifted from the wafer surface and an acetone soak released the sheath structure.

41

Figure 2-9: Fabrication process for sheath probe having peripheral “wing” electrodes.
Only the final outline of the device is shown. Note that the time-consuming steps of an
additional metal patterning and deposition as well as Parylene insulation deposition and
patterning are eliminated as compared to the previous top electrode sheath probe
fabrication.

2.4.1.1 Plasma Etching of Parylene Insulation
Exposing probe recording sites through a complete plasma etch of an insulation
layer is a critical step. Any Parylene that covers the electrode reduces its electroactive
area and increases its impedance; making it useless for recording. The general rule of
thumb to ensure complete Parylene removal is to overetch. However, there is danger in
overetching as the isotropic nature of plasma etching increases the diameter of the etch
pattern and the PR mask becomes thinner. Parylene is also heated in the process;
unnecessary Parylene heating should be avoided to minimize Parylene thermal
degradation. Lastly, Parylene plasma etching can be time-consuming and one should not
waste time performing unnecessary processes. In practice, optical observation through
42
microscopy can confirm that the insulation layer is completely etched when the
underlying Pt metal is reflective. In contrast, an incomplete etch will leave evidence of
partially etched Parylene residue in the form of a rough layer that looks dark over the
electrodes (Figure 2-10).

Figure 2-10: Comparison of partially etched Parylene insulation (a) to completely open
electrodes (b). A partial etch can be rectified with more plasma etching until the eletrodes
look “shiny.”

2.4.1.2 Scum
Scum was present on metals, both recording sites and contact pads after washing
off the PR mask with standard clean (acetone, IPA, and DI water rinse) when plasma
etching was complete (e.g. after an insulation etch) (Figure 2-11). Metals were observed
to be clean prior to mask removal.

Figure 2-11: Scum deposited on metals after a standard clean of the PR etch mask.
43
This scum proved to be resistant to subsequent cleanings. Soaking the wafers in
acetone and rubbing with cotton swabs did not remove any scum. An aggressive descum
in plasma (100 W: 100 mT: >10 min did not have any effect either (Figure 2-12).

Figure 2-12: Scum is still present after an aggressive descum in plasma.

Emperically, the scum was preferentially deposited on the exposed metal rather
than surrounding Parylene. More scum was present after a particularly thick Parylene
etch as compared to thinner etches; thus the amount of PR mask exposure to plasma
seemed to correspond to the amount of scum. The hypothesis was that a top layer of PR is
hardened during plasma etching as it is heated and perhaps dried out. Credence to this
theory came from the observation that wisps of PR were present when soaking the wafer
into an acetone bath. These PR wisps did not dissolve in acetone as PR regularly does.
The solution of avoiding scum was found to be careful soaking and washing of
the wafer in multiple solvent baths. After plasma etching, the wafer was placed carefully
into a bath of acetone. After the bulk of the PR mask was attacked by acetone, cotton
swabs were used to move the insoluble PR wisps to the bath edge. Clean cotton swabs
were then used to scrub the wafer. The wafer was then removed from the bath while
44
avoiding the wisps, as these PR clouds may redeposit onto the wafer and present as
irremovible scum. The wafer was placed next placed into a second acetone bath and
scrubbed vigourously with cotton swabs. The wafer was then placed into an IPA bath,
followed by rinsing in DI water. Finally, the wafer was dried with inert N
2
gas.
2.4.1.3 PR Mask Reflow
Etching through thicker Parylene layers entailed a longer etch time such as the
case when cutting out the probe outline. The prolonged exposure to high energy plasma
heated up the Parylene as well as the PR mask. In many cases, this heat exposure led to
undesirable thermal degradation of Parylene. Moreover, PR could reflow at elevated
temperatures and obscure the Parylene etch (Figure 2-13).

Figure 2-13: (a) Evidence of partial PR reflow can be seen as PR starts to obscure the
recording site (arrow). (b) The previously open pattern above the recording site is now
completely covered due to PR reflow (arrow); thereby preventing plasma etching of the
underlying insulation.

One simple method to avoid PR reflow and Parylene thermal degradation was to
reduce the etch time. While total etch time remained unchanged, multiple etches of
45
shorter durations with adequate cooldowns in-between were performed to minimize
heating and its associated problems.
2.4.1.4 Parylene bubbles
Bubbles under the Parylene layers were encountered during the fabrication
process as is often the case when patterning Parylene deposited with no adhesion
promoter (A-174) [20]. This time, it was observed that the bubbles formed over time
under the exposed PR (Figure 2-14). Experimentally, it was found that a quick transition
from exposure to the development bath mitigated the undesired creation of bubbles. This
is possibly due to the bath cooling the localized temperature rise due to an exothermic
reaction undergone by the exposed PR [24].

Figure 2-14: Bubbles are generated underneath the Parylene when exposing and
developing the PR mask for plasma etching on wafer. Note that bubbles originate in
exposed areas.
46
2.4.2 Thermoforming and Packaging
Wing electrode sheath probes were thermoformed around a microwire just as
before with top electrode probes under vacuum (Figure 2-15). PEEK backing was
attached to the bottom of the contact pad area and the same ZIF connector from before
was used for electrically packaging.

Figure 2-15: Bubbles are generated underneath the Parylene when exposing and
developing the PR mask for plasma etching on wafer. Note that bubbles originate in
exposed areas.

2.4.3 Results
Three different wing electrode sheath shapes were constructed with the same
dimensions as before with the top electrode probes (Figure 2-16). The probes were
successfully thermoformed and held their shape permanently after the process.
Fabrication yield was drastically improved to > 90 % since no electrodes were destroyed
through strain induced fracturing.
47

Figure 2-16: Probes possessing wing electrodes were fabricated with three different
sheath shapes.

2.4.3.1 Electrochemical Characterization
Electrodes housed within a Faraday cage were assessed using a Gamry Reference
600 potentiostat for cyclic voltammetry (CV) and electrochemical impedance
spectroscopy (EIS) characterization. Probes were immersed in 0.05 M sulfuric acid
(H
2
SO
4
) and a separate Ag/AgCl reference electrode was used for CV measurements.
One probe electrode was set as the working electrode while an adjacent Pt electrode was
used as the counter. Voltage was cycled between -0.2 to 1.2 V with a scan rate of 250
mV/sec. 50 scans were taken to ensure a stable voltammogram was recorded.
48
The classic voltammogram of Pt immersed in sulfuric acid was obtained for all
thin film Pt electrodes. Peaks corresponding to hydrogen adsorption and desorption as
well as Pt oxide formation and reduction were observed (Figure 2-17).

Figure 2-17: Representative (left) CV in H
2
SO
4
and (right) EIS in 1XPBS of a single
recording site.

The sheath probe was immersed in 1X phosphate buffer solution (PBS) for EIS
with a separate Ag/AgCl reference electrode. Impedances for all electrodes were recorded
over 1-10
5
Hz (Figure 2-17). Impedances ranging from 50-250 kΩ at 1 kHz were
obtained indicating acceptable electrode properties for neural recording. Impedances of
less than 1 MΩ portend neural recording success with high SNR upon probe implantation
[25].
2.4.3.2 In Vivo Intracortical Recordings
Probes with wing electrodes were implanted into the rat barrel motor cortex
following a protocol approved by the Huntington Medical Research Institutes
Institutional Animal Care and Use Committee (IACUC). Neural recordings were
49
successfully obtained at weekly intervals post-implantation and neural signal was found
to improve over time with increasing SNR and spike rate (Figure 2-18).

Figure 2-18: Representative rat neuronal activity from a sheath probe recording site at
different days post-implantation.

2.5 Arrayed Sheath Probes
Initial results were promising however it was uneconomical and inefficient to
only implant one probe per rat. Packaged arrays of single sheath probes were constructed
by machining an inserter tool made of multiple wires which could temporarily hold and
implant probes in a 2×1 or 2×2 pattern. In theory, this allowed for more electrodes, more
signals and thus data from each implanted rat and led to increased statistical robustness
for experiments studying coating efficacy and signal modulation due to whisker
movement. In practice, this was difficult to achieve and required considerable skill on
part of the experimenters throughout all the numerous steps required prior to implantation.
With each step, some loss of device could occur necessesating a brute force approach to
counteract the loss leading to overall inefficiency.
50
2.5.1 Arrayed Probe Design
The dimensions of sheath probe shape A were chosen for the probe dimensions of
the new device which consists of two probes arrayed in a 2×1 pattern with all electrode
leads converging to a single contact pad (16 pads total). In previous rat implantations of
the different probe shapes, shape A was found to record the clearest signals after one
month; this is attributed to its narrower and sleeker outline compared to shapes B and C
which may cause less damage to the surrounding neural tissue during implantation.
The probe dimension was further streamlined by introducing a taper to the wings
at the tip. The fabrication was also changed slightly to etch the base tip away so the final
device did not possess a blunt tip; creating a sleeker device profile (Figure 2-19). The two
probes were patterned 1 mm apart center-to-center to more easily fix onto the introducer
tool for implantation which consisted of two microwires spaced 1 mm apart.
Perforations were added to the probe design. Literature suggests that an open
architecture in the form of perforations would improve neural tissue integration of the
probe and lead to stable recordings [23]. Tissue integration is hypothesized to improve
due to increased anchoring of the probe with growth through the perforations. This would
counteract the undesirable effects of brain micromotion which can pulsate by as much as
30 µm in rat [26]. Maintenance of functional neurons may depend on communication
between neurons and astrocytes which is dependent upon cell-cell interactions.
Perforations would increase the area over which these interactions can take place. An
improvement in local diffusion of biochemicals due to the addition of perforations may
also stabilize recordings by maintaining the health of local neurons and minimizing scar
51
tissue formation. Thus 15 µm diameter perforations throughout the probe were plasma
etched (Figure 2-19).

Figure 2-19: Comparison between (a) sheath probe shape A from before and (b)
perforated sheath probe. Perforations were added throughout the sheath for better neural
integration and a sleeker outline was obtained by etching away the tip and introducing a
taper to the wings. A sleeker outline may also minimize neural damage during
implantation.

Contact pad spacing and size were increased by ~2-3% to compensate for
Parylene shrinkage encountered during the thermoforming process. Previously with
individual probes that possessed a total of only 8 electrodes and contact pads, the
shrinkage was found to be within tolerance and ZIF connectors successfully made
electrical contacts to each contact pad. However, with a doubling of contact pads, the
spatial error introduced through thermoforming propagated to such a degree that
electrical connections through ZIF connectors failed. Thus, a characterization of the
shrinkage was conducted and the contact pad patterns enlarged to compensate.
Left and right handed versions of the arrays were constructed so that they could
be packaged on top of one another, allowing for implantation of four probes (2×2 pattern)
or two arrayed devices into a single rat (Figure 2-20). However, it is important to note
that these are not simply mirror images of each other. The leads were mirror-imaged
52
while the probe electrode pattern was not. This was done to preserve the relative
electrode locations between left and and right handed versions. The conservation of
heightwise electrode placement within the cortical layers is paramount for experimentally
comparing neural signals between the subtly different arrays.

Figure 2-20: A left and right handed sheath probe array.

2.5.2 Arrayed Probe Fabrication
The fabrication of the arrayed probes follows in a similar fashion to previous
designs (Figure 2-21). Extensive use of Parylene patterning through plasma etching
resulted in successful construction of the device on a wafer level. One change to be noted
is that the process was moved from a 3” to a 4” wafer for these devices. Tighter device
placement and larger patternable wafer surface area resulted in increased device yield per
wafer (from 27 to 51 per wafer).
As before, a bare Si wafer with a native oxide layer was coated with 5 µm
Parylene. Electrodes, traces, and contact pads were liftoff patterned using AZ 5214 E-IR
and e-beam deposited Pt (2000 Å). Perforations were next plasma etched into the base
Parylene layer. A 2 µm Parylene insulation layer was then deposited, patterned, and
plasma etched, exposing perforation and recording sites in addition to contact pads. Next,
53
AZ 4620 (9.6 µm) was patterned and overcoated with 5 µm Parylene to create
microchannel structures. Selective plasma etching of Parylene exposed peripheral
electrodes, contact pads, and channel openings while simultaneously defining each
individual probe cutout and creating perforations. Finally, the probe was lifted from the
wafer surface and an acetone soak released the channel structure.

Figure 2-21: Fabrication process for arrayed sheath probe possessing perforations. Only
the final outline of the device is shown.

The Parylene insulation layer thickness was doubled from 1 to 2 µm for these
arrayed devices. This was done to guarantee electrical insulation. While 1 µm insulation
thickness was found to be adequate before, some variations in EIS readings were
encountered and so the insulation thickness was doubled to preemptively eliminate
fabrication process vagaries as a source for experimental errors.
54
2.5.3 Themoforming and Electrical Packaging
Themoforming of the probes was performed just as before. Microwires were
carefully inserted into the probe openings and the apparatus was placed into a vacuum
oven and soaked for 48 hours at 200 °C. Nitrogen purging was found to be unnecessary if
the vacuum seal on the oven was great enough to prevent Parylene oxidative degradation
due to atmospheric leaks. ZIF connectors were used to connect to PEEK backed contact
pads as before for electrical connections. A right and left handed array were packaged
together back-to-back to create a 2×2 array for implantation (Figure 2-22).

Figure 2-22: An electrically packaged array. Two devices were packaged together
through ZIF connectors onto a PCB.

55
2.5.4 Arrayed Probe Results

Figure 2-23: SEM of perforated sheath probe after themoforming.

The arrayed probes were fabricated on wafer, released, and thermoformed (Figure
2-23). No difference in mechanical behavior was observed between previous designs that
lacked perforations and these arrayed perforated sheaths. The addition of perforations did
not compromise the mechanics of the probe due to the minimal amount of Parylene lost
to perforation etching.
After packaging, the Pt electrodes were electrochemically characterized just as
before with CV and EIS. Arrayed probes were immersed in 0.05 M sulfuric acid and a
separate Ag/AgCl reference electrode was used for CV measurements. One probe
electrode was set as the working electrode while an adjacent Pt electrode was used as the
counter. Voltage was cycled between -0.2 to 1.2 V with a scan rate of 250 mV/sec.
Multiple (30) scans were taken to ensure a stable voltammogram was recorded. The
classic voltammogram of Pt immersed in sulfuric acid was obtained for all thin film Pt
electrodes with observable peaks corresponding to hydrogen adsorption and desorption as
well as Pt oxide formation and reduction. EIS of all recording sites in 1X PBS with a
56
separate Ag/AgCl reference electrode demonstrated low impedances suitable for neural
recording at 1 kHz (Figure 2-24).

Figure 2-24: Representative (left) CV in H
2
SO
4
and (right) EIS in 1XPBS of Pt recording
sites (n = 32) of arrayed perforated probes electrically packaged into a 2×2 array for
implantation.

Arrayed perforated probes were then implanted into the rat cortex following a
protocol approved by the Huntington Medical Research Institutes Institutional Animal
Care and Use Committee (IACUC). Recordings were taken just as before and neural
signals were successfully recorded with the probes (Figure 2-25).

Figure 2-25: Representative in vivo signal recording from rat implanted perforated sheath
probe at 21 days.

2.6 Minimally Sized Probes
Minimally sized sheath probes were designed and constructed in response to
preliminary tissue histology data that showed some local neural damage from rat-
57
implanted devices. The damaged tissue surrounded the implanted sheath and a tissue void
was created due to the peircing action of the microwire that was used to temporarily
stiffen and implant the probe without buckling (Figure 2-26) [27]. Much of the tissue
damage was due to the size of the microwire, which can be reduced greatly through
electrochemical etching. Correspondingly, minimizing the size of the sheath probe to fit
smaller needles would lead to less local tissue damage.

Figure 2-26: Evaluation of tissue response to implanted probes at 30 days. 10 µm cortical
sections sliced perpendicular to the probe were stained with NeuN (red) to visualize
neurons and GFAP (blue) highlighting reactive astrocytes. The presence of astrocytes
indicates tissue damage. A tissue void matching the shape of the microwire used to insert
the three different sheath probe shapes is evident (a) Sheath shape A, sharpest taper. (b)
Sheath shape B, wider taper (c) Blunt cylindrical sheath probe, from [27].

Sheath probes 3-4 Χ smaller than previously implanted ones were designed. A
variety of devices with permutations in wing width (10-50 µm), sheath taper (50-90 µm),
and cylinder diameter (50-90 µm) were designed and fabricated. Such permutations were
necessary to determine the smallest achievable sheath probe since there are dimensional
limits due to the fabrication process; lithography, alignment, anisotropy of Parylene
etching, packaging etc. Critically, the wing width of the device is what holds the two
58
halves of the sheath together and so there is a limit to how small this surface area can be
before sheath structural integrity is compromised (Figure 2-27).
Fabrication of these small sheath devices proceeded similar to the process
described before. A Parylene base layer was patterned with Pt which was then insulated
with another Parylene layer. Parylene microchannels were then constructed by
overcoating patterned sacrificial PR. Finally, openings for electrical connections and
microchannels as well as the device outlines were plasma etched and the devices were
released from the wafer followed by an acetone soak to release the microchannels.

Figure 2-27: (a) Optical micrograph of the smallest sheath probe fabricated with 10 µm
width wings. Even with such a narrow wing width, the multi-layered Parylene structure
held. The three Pt recording electrodes are visible. (b) SEM image of another minimal
sheath probe with 30 µm width wings. 15 µm diameter perforations were added to the
non-thermoformed probe.

Microwires were then used to open the microchannels in these devices. Sheath
probes with smaller wing widths were observed to be destroyed through delamination of
the Parylene layers. The small surface areas at the wings required for holding the
structure together could not withstand the mechanical strains associated with creating the
59
sheath through the insertion of the microwire mold (Figure 2-28). Wing widths of at least
50 µm were required to maintain sheath integrity, 64% smaller than previously implanted
devices. It should be noted that this is the minimum width required for creating sheaths
through the thermoforming process; even smaller devices with 10 µm wing widths could
be used as probes without thermoforming. The Parylene microchannel possesses enough
clearance (~10 µm) to permit neural tissue ingrowth. However, electrode surface area (as
small as 150 µm
2
compared to 1589 µm
2
originally) and number (<3/probe vs 8/probe)
were severely constrained with the smaller Parylene surface available for metal
patterning.

Figure 2-28: A minimal sheath probe with a wing width of 40 µm opened with a
microwire. Structural integrity of the probe was compromised as evidenced by slight
Parylene delamination and tearing (arrow) with microwire insertion.

While such a minimal device could mitigate localized neural tissue damage
during implantation, a severe tradeoff in recording capability is encountered. A lower
number of electrodes limits how much neural signal can be recorded. Also, lower
electrode surface area can possess undesirable neural recording properties such as high
impedance and poor frequency response [25] with issues of noise and pickup in the
electronic packaging due to high input impedances [28]. On the other hand, a minimal
60
probe form factor may be more desirable when a large number of neural recordings is not
necessary and the focus is on minimizing localized damage.
2.7 Discussion & Conclusion
A Parylene neural probe with a 3D sheath structure for improved neural
recordings and integration was designed and fabricated. A novel fabrication scheme was
introduced in which a 3D hollow sheath structure was produced with multiple Pt
recording electrodes on the interior and exterior using a combination of surface
micromachining and thermoforming processes. Different sheath shapes were constructed
and experimentally tested for neural probe suitability. An optimal geometrical
dependency was found for sheath probes; sleeker profile (sheath shape A) with the tip
removed minimizes tissue damage compared to larger tapers or cylinderical shapes.
Smaller sheath probes could also be fabricated with the tradeoff of less capable neural
recording electrodes. Arrays of sheath probes were constructed and improved recording
efficiency in number of recording sites possible per implant. Convenient and reusable
electrical connections with flexible Parylene cables were achieved with ZIF connectors.
Electrochemical characterization of the electrodes demonstrated desirable properties for
neural recordings. Successful in vivo neural recordings were demonstrated in rat barrel
motor cortex. The novel fabrication method presented in which flat surface
micromachined structures can be modified post-fabrication to form 3D structures is not
only an enabling technology for next generation neural probes but also has implications
on flexible and 3D microfluidic platforms for fluid delivery or tissue ingrowth.
61
REFERENCES

[1] V. S. Polikov, P. A. Tresco, and W. M. Reichert, "Response of brain tissue to
chronically implanted neural electrodes," Journal of Neuroscience Methods, vol.
148, pp. 1-18, 2005.
[2] J. R. Capadona, D. J. Tyler, C. A. Zorman, S. J. Rowan, and C. Weder,
"Mechanically adaptive nanocomposites for neural interfacing," MRS Bulletin, vol.
37, pp. 581-589, 2012.
[3] J. Subbaroyan, D. C. Martin, and D. R. Kipke, "A finite-element model of the
mechanical effects of implantable microelectrodes in the cerebral cortex," Journal
of neural engineering, vol. 2, p. 103, 2005.
[4] P. R. Kennedy, "The cone electrode: a long-term electrode that records from
neurites grown onto its recording surface," Journal of Neuroscience Methods, vol.
29, pp. 181-193, 1989.
[5] P. R. Kennedy and R. A. Bakay, "Restoration of neural output from a paralyzed
patient by a direct brain connection," Neuroreport, vol. 9, pp. 1707-11, Jun 1
1998.
[6] P. R. Kennedy, R. A. Bakay, and S. M. Sharpe, "Behavioral correlates of action
potentials recorded chronically inside the Cone Electrode," Neuroreport: An
International Journal for the Rapid Communication of Research in Neuroscience,
1992.
[7] P. R. Kennedy, S. S. Mirra, and R. A. E. Bakay, "The cone electrode:
ultrastructural studies following long-term recording in rat and monkey cortex,"
Neuroscience letters, vol. 142, pp. 89-94, 1992.
[8] P. Kennedy, D. Andreasen, P. Ehirim, B. King, T. Kirby, H. Mao, et al., "Using
human extra-cortical local field potentials to control a switch," Journal of Neural
Engineering, vol. 1, p. 72, 2004.
62
[9] J. Bartels, D. Andreasen, P. Ehirim, H. Mao, S. Seibert, E. J. Wright, et al.,
"Neurotrophic electrode: Method of assembly and implantation into human motor
speech cortex," Journal of Neuroscience Methods, vol. 174, pp. 168-176, Sep 30
2008.
[10] F. H. Guenther, J. S. Brumberg, E. J. Wright, A. Nieto-Castanon, J. A. Tourville,
M. Panko, et al., "A Wireless Brain-Machine Interface for Real-Time Speech
Synthesis," Plos One, vol. 4, Dec 9 2009.
[11] P. Kennedy, "Changes in emotional state modulate neuronal firing rates of human
speech motor cortex: A case study in long-term recording," Neurocase, vol. 17,
pp. 381-393, 2011.
[12] J. T. W. Kuo, B. J. Kim, S. A. Hara, C. Lee, C. A. Gutierrez, T. Hoang, et al.,
"Fabrication of 3D Parylene Sheath Probes for Reliable Neuroprosthetic
Recordings," presented at the Hilton Head 2012: Solid State Sensors, Actuators
and Microsystems Workshop, Hilton Head Island, South Carolina, USA, 2012.
[13] J. T. W. Kuo, B. J. Kim, S. A. Hara, C. D. Lee, C. A. Gutierrez, T. Q. Hoang, et
al., "Novel flexible Parylene neural probe with 3D sheath structure for enhancing
tissue integration," Lab Chip, 2013.
[14] E. Meng, Biomedical Microsystems, 1 ed.: CRC Press, 2010.
[15] 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.
[16] C. Shih, T. A. Harder, and Y. C. Tai, "Yield strength of thin-film parylene-C,"
Microsystem technologies, vol. 10, pp. 407-411, 2004.
[17] Y. Nam, "Material considerations for in vitro neural interface technology," MRS
bulletin, vol. 37, pp. 566-572, 2012.
[18] A. V. Vasenkov, "Atomistic modeling of parylene-metal interactions for surface
micro-structuring," Journal of Molecular Modeling, vol. 17, pp. 3219-3228, 2011.
63
[19] C. A. Gutierrez and E. Meng, "A dual function Parylene-based biomimetic tactile
sensor and actuator for next generation mechanically responsive microelectrode
arrays," presented at the The 15th International Conference on Solid-State Sensors,
Actuators and Microsystems (TRANSDUCERS), Denver, Colorado, USA, , 2009.
[20] C. A. Gutierrez, "DEVELOPMENT OF FLEXIBLE POLYMER-BASED MEMS
TECHNOLOGIES FOR INTEGRATED MECHANICAL SENSING IN
NEUROPROSTHETIC SYSTEMS," Doctor of Philosophy Doctoral, Biomedical
Engineering, University of Southern California, Los Angeles, 2011.
[21] D. C. Rodger, A. J. Fong, W. Li, H. Ameri, A. K. Ahuja, C. Gutierrez, et al.,
"Flexible parylene-based multielectrode array technology for high-density neural
stimulation and recording," Sensors and Actuators B: Chemical, vol. 132, pp.
449-460, 2008.
[22] H.-S. Noh, Y. Huang, and P. J. Hesketh, "Parylene micromolding, a rapid and
low-cost fabrication method for parylene microchannel," Sensors and Actuators
B: Chemical, vol. 102, pp. 78-85, 2004.
[23] J. P. Seymour and D. R. Kipke, "Neural probe design for reduced tissue
encapsulation in CNS," Biomaterials, vol. 28, pp. 3594-3607, 2007.
[24] A. Reiser, J. Huang, X. He, T. Yeh, S. Jha, H. Shih, et al., "The molecular
mechanism of novolak–diazonaphthoquinone resists," European polymer journal,
vol. 38, pp. 619-629, 2002.
[25] V. Pikov, "Personal communication," ed. Los Angeles, 2012.
[26] A. Gilletti and J. Muthuswamy, "Brain micromotion around implants in the rodent
somatosensory cortex," Journal of neural engineering, vol. 3, p. 189, 2006.
[27] B. J. Kim, J. T. W. Kuo, S. A. Hara, C. D. Lee, L. Yu, C. A. Gutierrez, et al., "3D
Parylene sheath neural probe for chronic recordings," Journal of Neural
Engineering, vol. 10, p. 045002, 2013.
[28] D. J. Ives, G. J. Janz, and C. King, Eds., Reference electrodes: theory and
practice. New York City: Academic Press, 1961, p.^pp. Pages.
64

3.1 Background
Studying neurons with electrical methods is problematic since the
neurotransmitter mechanism of signal propagation is ignored or disrupted. Also, studying
neurons with patch clamping is not biomimetic and details of how neurons behave
naturally may be lost with traditional neuroscience techniques since the natural
neurotransmitter to action potential mechanism is disrupted. Patch clamping physically
deforms the cell membrane which may affect the membrane embedded ion channels.
Driving an entire cell membrane to a certain voltage level affects all ion channels in the
same method and so different voltage-gated ion channels’ behaviors are circumvented, a
problem encountered with the use of any microelectrode.
As an example, retinal ganglion cells (RGC) are known to possess ON and OFF
types in which cells respond in opposite manners to the same neurotransmitter, glutamate
[1]. ON cells will activate and generate an action potential to glutamate while OFF cells
will become inhibited. This property forms the basis for object contrast in vision.
Subjecting RGC to a microelectrode or patch clamp will elicit the same response from
both ON and OFF cells since both will respond to the same electrical input. In order to
properly study such cells and achieve selective response to stimulation, neural interaction
Chapter 3 MICROMACHINED CHEMICAL NEURAL INTERFACES
65
through the use of neurotransmitters through a biomimetic chemical neural interface is
required.
3.2 Aims
Such a biomimetic chemical neural interface can be achieved through the use of
microfabrication technologies. Microfluidics can be used to deliver precise amounts of
neurotransmitter to neurons with high spatiotemporal resolution. Integrated flow sensors
can monitor delivery amount in the nL scale. Electrochemical (EC) sensors can be used
to monitor neurotransmitter release in situ leading to a bi-directional chemical interface in
which neurotransmitters can be delivered and received from neurons.
3.3 Microfluidic Platform for Chemical Interfacing
A microfluidic platform for precise biochemical control of the extracellular
microenvironment was developed (Figure 3-1). Controlled fluid delivery from a “pore” to
a cellular microenvironment was achieved with a dedicated microchannel. Thermal flow
sensors were integrated into each microchannel for flow rate monitoring of individual
microchannels; thus feedback control of focal pore delivery was achievable.

66

Figure 3-1: Top view of a microfluidic platform with three 100 µm wide Parylene
microchannels. Each channel is centrally perforated with a pore having a size indicated
by the label on the upper right corner of each channel (5, 10, 20 µm diameter). Eight Pt
thermal flow sensors line each channel and flank the central pore.
3.3.1 Microfluidic Platform Design
The microfluidic platform consists of an array of three microchannels constructed
of Parylene C through micromachining technologies on a soda lime substrate. The
microchannels measure 6 mm × 100 µm × 4 µm (L × W × H) with 2 µm thick walls.
Each microchannel is centrally perforated with a pore for focal delivery to extracellular
microenvironments. A linear array of Pt thermal flow sensors flanks each pore within the
microchannel. Pore sizes of 5, 10 and 20 µm in diameter were fabricated. SU-8 structures
(1 mm in diameter) were patterned to construct fluidic inlet and outlet access ports to the
microchannels and eliminate the need for through-wafer etching. A Parylene C funnel
structure connects access ports to the microchannels and is supported with patterned
Parylene C posts to prevent collapse of the wide, free-standing structure.

67

Figure 3-2: Microfluidic platform design (not to scale).

Use of a soda lime substrate and Parylene C give the microfluidic platform optical
transparency. Real-time visual inspection of microchannel flow, focal pore delivery, and
biological response of cells/tissue to the resulting chemical gradient is possible through
optical microscopy. Pt thermal flow sensors give real-time flow measurements and allow
precise measurement of focal delivery rate through differential flow sensing before and
after the pore.
3.3.2 Thermal flow sensing theory
Thermal flow sensors transduce flow to varying electrical signals by taking
advantage of thermal phenomenon when fluid flow affects heat transfer. Sensor response
to flow change is captured through the heat transfer between sensor and fluid that varies
with flow. A constant fluid temperature is required for proper flow sensor operation and
68
readout; temperature compensation is required for instances when fluid temperature
varies during operation.
Hot-film sensors operate by heat transfer from a heated element to surrounding
cooler fluid. The thin film resistive element is heated and subjected to fluid flow.
Convective heat loss increases from the heated element as fluid flow increases. Sensor
electrical resistance varies based upon change in element temperature; thus heat transfer
rate can be transduced into an electrical signal that varies with respect to fluid flow.

Figure 3-3: Illustration showing concept of hot film thermal flow sensor. The resistor
serves as a heater and sensing element. Resistance value is dependent on temperature.

For typical thermal flow sensor materials, the resistance relationship to
temperature is given by:

 
00
( ) ( ) 1 R T R T T T     

(1)

where R(T) is the resistance at temperature T and α is the temperature coefficient
of resistivity (TCR). TCR can be determined experimentally by:

0
0
0
( ) ( )
( ) or
()
R
R T R T R
a T T T
R T R

 
     (2)

69
in which a
R
is the resistance overheat ratio and determined by measuring the
change in resistance of the material at two different temperatures.
Hot-film sensors are first characterized by imposing a known fluid flow and
measuring the resulting resistance change. Fluids used in sensor characterization need to
possess the same or similar thermal conductivity properties as the measured fluid since
thermal convection is integral to the transduction mechanism.
Hot-film sensors can be operated in six operational modes by controlling either
the heater power or temperature and observing the heater temperature, power, or
temperature difference resulting from fluid flow [2, 3]. Constant heater power mode
involves imposing a constant current bias on the heating element and monitoring the
change in resistance or voltage due to flow. Constant temperature mode requires
feedback circuitry that monitors and holds constant the sensor temperature; the increase
in power required to maintain temperature under higher flow rates is monitored. While
more complex in implementation, constant temperature mode can deliver better sensor
resolution and frequency response [4].
Pt in the form of thin film resistors was chosen for thermal flow sensor
construction due to ease of fabrication and possessing desirable flow sensing properties:
high corrosion and oxidation resistance, large temperature range, linear resistance versus
temperature relationship, and high TCR (39.2×10
-4
/K) [5].
70
3.3.3 Microfluidic Platform Fabrication
Fabrication began with piranha (H
2
SO
4
:H
2
O
2
4:1) cleaned 3” soda lime substrates.
Thermal flow sensors were patterned with AZ4400 photoresist (AZ Electronic Materials,
Branchburg, NJ), O
2
plasma descummed (60 W: 100 mT: 1 min) to clean the surface, and
e-beam deposited Ti adhesion layer (200 Å) and Pt (2000 Å) followed by liftoff with
gentle mechanical agitation in acetone, isopropyl alcohol (IPA) and deionized (DI) water
immersions. A short descum (60 W: 100 mT: 1 min) was performed to clean the surface
after the liftoff process and a Parylene adhesion promoter (A-174, Specialty Coating
Systems, Indianapolis, IN) was applied. A 2 µm insulation layer of Parylene (Parylene C,
Specialty Coating Systems, Indianapolis, IN) was then vapor deposited to prevent direct
contact between flow sensors and conductive fluid flow.
Parylene microchannels were fabricated using surface micromachining techniques.
AZ 4400 photoresist was spun on (4 µm) and patterned to form the sacrificial
microchannel structure as well as small pits for eventual formation of Parylene posts
along the channel length to prevent collapse. The sacrificial photoresist was then
overcoated with Parylene (4 µm). This Parylene layer was then selectively O
2
plasma
etched (150 W: 200 mT) to obtain fluidic access to the channel, form the central fluid
delivery pore, and create contact pad openings to the thermal sensor elements.
Finally, fluidic access ports to channels were constructed with SU-8 to prevent
mechanical microchannel damage during packaging. SU-8 2100 (Microchem Corp.,
Newton, MA) was spun on (75 µm), patterned, and hardbaked (30 min at 100 °C with a
3 °C/min ramp). Individual dies were diced out and immersed in an IPA bath over several
71
days followed by a DI water rinse. IPA immersion was utilized to release the
microchannels instead of acetone despite slower photoresist dissolution. Acetone was
avoided since exposure to acetone results in swelling, deformation, and delamination of
SU-8 structures.

Figure 3-4: Fabrication process for creating the microfluidic platform.
3.3.4 Microfluidic Platform Packaging
The device was first bonded to a PCB with epoxy. Electrical connections between
thermal sensor contact pads and PCB were created through wire bonding (Au). Wires
were then soldered to the PCB to create external electrical connections.
72
An acrylic jig was designed and fabricated with a laser drill (Mini/Helix 8000,
Epilog, Golden, CO) for simultaneous fluidic and electrical connections to the
microfluidic platform. Silicone (Sylgard 184, Dow Corning, Midland, MI) was used to
construct gaskets to create a seal around the inlet and outlet ports of the microchannel.
Glass capillaries (5 µL, Alltech, Deerfield, IL) were inserted through the top of the jig
into the PDMS gaskets to create fluidic connections into and out of the channel.
The PCB bonded microfluidic platform was placed between the bottom and top
acrylic jig pieces and carefully aligned under microscopy to position the glass capillaries
to the SU-8 fluidic ports. The device was secured by tightening the two jig pieces with
screws.
The glass capillaries were connected with a short piece of silicone tubing acting
as a transition tube that ended with Polyetheretherketone (PEEK) tubing. The other end
of this PEEK tubing connected to a precision glass syringe (Gastight Syringes, Hamilton
Company, Reno, NV) driven by a syringe pump (PHD 2000, Harvard Apparatus Inc.,
Holliston, MA).
73

Figure 3-5: Packaged microfluidic platform in a custom made acrylic jig with gold wire-
bonds connecting the embedded Pt thermal flow sensors to the PCB. Wires were soldered
to the PCB and convered with epoxy to make robust external electrical connections.
Glass capillaries aligned to and clamped on top of the SU-8 inlet/outlets provided a
fluidic connection to the microchannel.
3.3.5 Thermal Flow Sensor Use
The integrated Pt flow sensors were characterized and their performance
validated. Fluids could be detected in the nL/min flow regime. The description of how the
thermal sensors were characterized and the different possible modalities available for use
are beyond the scope and can be found in [5]. Here, ultimate use of the flow sensors
utilizing differential sensing to detect how much chemical was delivered through the pore
is described.
A microchannel was investigated for electrical sensing of fluid delivery through a
pore with a diameter of 20 µm. Two flow sensors, one before the pore and one after, were
biased at 4 mA in hot-film constant current (CC) mode using two separate 2400
74
SourceMeters as a CC source for each (sensors equidistant from pore, distance between
sensors = 4.5 mm). Both sensors were placed in two separate and identical Wheatstone
quarter-bridges. The bridge outputs were simultaneously recorded (n=58) for each flow
rate setting (filtered DI water) using two separate channels of the LabView controlled
2700 MultiMeter (Keithley Instruments Inc., Cleveland, OH). The data recorded at each
flow rate were then averaged, and the resulting curves were compared to one another.
Again, known flow rates from 0-1 µL/min were used and the responses of two sensors
were measured and recorded in CC mode (Figure 3-6). Fluid delivery through the pore
was confirmed by optical inspection under a microscope. For higher flow rates, the
upstream sensor exhibited a greater voltage output that the downstream sensor indicating
that fluid delivery through the pore can be measured by using this differential
measurement technique. However, increased resolution is required for nL/min flows and
may be achieved by incorporating thermal isolation structures underneath the sensor
regions to prevent undesirable heat transfer to the glass substrate. Statistical analysis
(Kruskal-Wallis non-parametric test) was performed on the data and statistically
significant differences (p < 0.01) were found for all but three points (400 and 1000
nL/min upstream of the pore and 400 nL/min downstream of the pore). One contributing
factor is likely the low signal-to-noise ratio.
75

Figure 3-6: Hot film response for 4 mA constant current biasing for two flow sensors, one
upstream and one downstream of the pore in the channel (mean ± S.E. with n = 58). Both
sensors were monitored simultaneously. Data points marked with an asterisk have non-
significant changes due to change in flow rate based on statistical analysis. This is
attributed to low signal-to-noise ratio.
3.4 Microfluidic Platform Discussion
A microfluidic platform for precise focal delivery of soluble factors to cells and
tissues was realized. An array of three Parylene microchannels each centrally perforated
with a uniquely sized small diameter pore was investigated. This arrangement allows a
one-to-one correlation between the flow delivery channel and each pore. Each channel
possesses Pt thermal flow sensors for integrated flow sensing. Focal fluid delivery
capabilities of the system were characterized by calibration and operation of the thermal
flow sensors and preliminary demonstration of focal fluid delivery. The thermal flow
sensors operated under steady state conditions were able to detect nL/min flow rates.
76
In work done by a colleague, focal chemical stimulation of a neuronal cell line
was demonstrated with the microfluidic platform. PC12 cells were cultured on the device
and loaded with fluo-4, a Ca
2+
indicator. Bradykinin was flowed through the Parylene
microchannel and delivered through the pore to the cells while they were monitored
under fluorescence microscopy. PC12 cells release intracellular Ca
2+
upon contact with
bradykinin. PC12 cells closest to the pore were found to fluoresce followed by others in a
radial manner indicating successful chemical stimulation with bradykinin (Figure 3-7).
However, the fragility of the microchannels prevented additional cell stimulation
experiments.

Figure 3-7: Time-lapse fluorescent images of PC12 cells focally stimulated by bradykinin
on top of the biomimetic chemical interface. PC12 cells increasingly fluoresce in a radial
nature as more bradykinin is delivered through the pore (arrow) and diffuses outward.
3.4.1.1 Microfluidic Platform Shortcomings
The packaging required for the microfluidic platform was cumbersome and
resulted in low yield. The first step was electrical packaging which required the use of Au
wire bonding in order to make electrical contact with the micromachined Pt sensors. This
was a time consuming process that resulted in bonds that were prone to breakage. The
77
cause of this weak bond was due to using Au wires instead of Al as is appropriate for
wire bonding to thin-film Pt on a glass substrate [6]. The second packaging step involved
creating macro-to-micro fluidic connections to the Parylene microchannels. This was
achieved with machined acrylic pieces with the fluidic connectors precisely aligned to the
microchannel inlet/outlets. Alignment was done by hand under a microscope however the
design of the packaging obscured views of the inlet/outlets making the process difficult.
Misalignment was often encountered since movement due to tightening of the acrylic
pieces often shifted the previously aligned microfluidic platform. In most cases, the
device would be released from the acrylic packaging and realigned. Many attempts were
required before a successful fluidic connection was achieved. However, sometimes the
microfluidic platform would be critically damaged due to misalignment; the Parylene
microchannel or funnel would be crushed by the fluidic connector and so a new device
was needed - lowering yield.
Even after packaging, Parylene microchannels were found to be delicate and
prone to collapse despite the addition of Parylene support posts within the channels.
Collapsed channels drastically reduced yield due to the effort and time required to
fabricate replacement devices. The stiction forces keeping the Parylene channels
collapsed could not be overcome with increased fluid pressure to reopen the channels
without permanently deforming or bursting the channels. In practice, a collapsed channel
meant a failed device with no hope of repair; a new microfluidic platform was required.
Another cause of device failure was that the micromachined pores were prone to clogging
due to ambient dust particles or cellular debris.
78
Another drawback of the microfluidic platform was its reliance on an external
syringe pump for controlling microchannel flow and delivery through the pore. While an
array of three microchannels were fabricated, in practice only one could be used at once.
This limitation was twofold. First, the number of syringe pumps available was limited
due to financial constraints (~$3k each). Second, each syringe pump required a
significant amount of space. Thus, reliance on syringe pumps for flowing fluids through
the microfluidic platform limited the scalability of the platform. Even though a MEMS
device was utilized, much of the benefits that were supposed to be realized through
miniaturization were lost due to reliance on a bulky external technology for operation
(Figure 3-8).

Figure 3-8: The microfluidic platform requires many external connections (fluidic and
electrical) in order to operate; leading to a bulky setup and limiting this technology’s
scalability due to costs associated with the syringe pump and electronics.

79
Focal delivery to neurons was achieved with the microfluidic platform however
there was no method to target neurons. In other words, neurons had to be positioned to
where the pore was in order to be exposed to delivered chemicals. For tissue studies, the
region of interest was manually aligned to the pores prior to focal delivery. For cell
culturing, no method was immediately available for culturing neurons to grow in a
controlled manner around the pore. Of course, there are examples within literature for
achieving patterned cell cultures utilizing stamping [7, 8] however this method is not
compatible with the delicate microchannels, which readily collapse with any application
of downward force.
With the microfluidic platform, a random population of neurons was exposed to
chemicals with a concentration gradient relative to their distance from the pore. Neurons
closest to the pore were completely flooded with chemicals while neurons further away
were exposed to diluted chemicals. It was impossible to chemically interface with
neurons at specific spatial points in a biomimetic fashion. Specifically, neural networks
are formed by the growth of axons linking to other neurons through synapses at specific
geometrical locations. Study of such networks relies upon the ability to influence how
networks are formed through the generation of chemical gradients [9]. Axons have been
demonstrated to preferentially grow in response to chemical cues [10]. Such chemical
guidance is often dynamic and spatially precise, to even subcellular regions, thus a true
chemical neural interface would possess the ability to chemically interact with neurons on
a subcellular level in a targeted fashion. Such ability would allow for a chemical interface
80
that influences and studies neural network formation as well as subsequent neural
processing within this network.
3.5 Microfluidic Platform Conclusion
A device for chemically interacting with neurons with flow through
micromachined pores was constructed with integrated flow sensors. Proof of concept of
chemical interfacing was demonstrated through focal delivery to cultured cells [11] or
neural tissue [5, 12]. However, this design was found lacking in several key areas. First,
the reliance on extensive packaging reduced yield and was time-consuming. Next,
channels and pores were delicate and easily failed. The need for syringe pumps for flow
actuation made the setup bulky and limited throughput. Finally, cells were randomly
exposed to delivered chemicals in a decidedly non-biomimetic fashion.
An improved device for chemical interfacing can be designed that overcomes
these limitations. Specific sub-cellular regions of neurons can be targeted by constructing
a device that allows for on-chip patterning of cell cultures whose axons or somata are
fluidically separated and accessed. Microchannel mechanical strength can be improved
by utilizing sturdier materials for fabrication. Packaging requirements can be simplified if
not done away with through careful design. In short, a better micromachined chemical
interface needs to be realized for biomimetic neural interfacing and for adoption as a
routine experimental tool by the neuroscience field.
81
3.6 Modules for Chemical Interfacing
The major challenges presented for chemical interfacing can be delineated as
follows. One is patterning cells in a controlled manner and isolating the soma and axon.
The second is a method for delivering chemicals to these neural components. Any device
that meets these challenges must also be easy to use and reliable. In practice, this requires
a design that is free of undue packaging requirements.
As more features and requirements are expected from a device, the design
becomes more complicated and ease of use is compromised. One method to avoid this
problem is to create different devices for every requirement which work together in a
modular fashion to create the desired operation - a usable chemical interface that is able
to target neurons both spatially and temporally for studying neural networks. By
addressing challenges separately, each modular device can be specialized and focused on
overcoming a specific problem without becoming unduly complicated. Complex
operations are then possible by combining simple modules.
Patterning neural cell cultures such that somata or axons are isolated and precisely
targeted without affecting the other can be achieved with a microdevice with
compartments that guide and separate the cell soma and axon as they grow. This can most
easily be achieved with soft lithography to construct soma and axon compartments with
microchannels that act as guides for axonal growth from the cell body: a CCM
(compartmented culture module) for in vitro on-chip cell culturing.
82
Complementing the CCM is a device for delivering chemicals to the neurons,
BAM (biochemical administration module). This device can be constructed from
mechanically rigid materials so that microchannels are not collapsible. BAM can also be
easy to use by utilizing passive pumping and doing away with electrical components for
vast simplification in packaging. The placement of BAM within CCM determines the
location of chemical delivery enabling targeting of specific locations for chemical
delivery.

Figure 3-9: Illustration of the two complementary modules for chemical interfacing.
Neurons are cultured within µCCM which isolates the axons and somata with its
microchannels. Chemicals are delivered to the axons with the placement of µBAM within
the axon compartment. Scaling is possible with the use of multiple µBAMs and the setup
is compatible with standard microscopy.

Improvements in chemical interfacing are enabled utilizing these complementary
modular components. Primarily, device simplification is achieved since packaging is
minimized and thus can be expected to lead to higher yields in module fabrication.
Improved designs prevent device fragility minimizing the chance of module failure and
83
maximizing ease of use post-fabrication. The combined modules are compatible with
microscopy as well as incubators and due to their small form factors are scalable with
multiple simultaneous experiments possible with limited space.
3.7 CCM
3.7.1 CCM Design
The CCM’s purpose is to enable patterned neural cell culturing with separation
of the soma and axon. Therefore, it must be compatible with standard cell culturing
technologies such as ethanol sterilization, pipetting, and incubation. The device should
also be compatible with standard optical microscopy as is normally employed with cell
culturing for visual observations of the growing neurons.
The simplest method to meet these requirements is to utilize PDMS for
constructing the CCM. PDMS-based devices have commonly been used for in vitro cell
culturing in various lab-on-a-chip applications and are mechanically robust enough to
withstand the demands of cell culturing [13]. PDMS can survive within an incubator and
if necessary can be sterilized through autoclaving (~121 ºC) [14]. PDMS is optically
transparent and thus amenable to optical microscopy. A caveat is that PDMS surfaces
have been found to release possible toxins in the form of uncured PDMS oligomers into
the surrounding culture media [15] though this effect may be media and thus cell type
specific. In any case, should neural cell death or injury be observed within CCM, a
84
simple solution would be to overcoat with Parylene C to serve as a surface barrier and
prevent undesirable interactions [16].
A widely adapted PDMS device creates separate compartments with a mechanical
filter made of microchannel arrays between compartments for fluidic isolation; enabling
the generation of chemical gradients across the compartments for chemotaxic studies
involving neurons. Neurons are then exposed to these gradients with some studies
culturing cells such that the somata are exposed to one gradient while the axons to
another [17, 18]. A drawback with this approach is that the two compartments are not
directly accessible. A bulk fluid flow into the device exposes the entire cell population;
generation of different geometrical gradients is not possible.
In a similar fashion, the CCM can be constructed from PDMS with both soma
and axonal compartments separated with an array of microchannels with the axonal
compartment accessible to the outside. The dimensions of the microchannel are smaller
than the somata such that it serves as a barrier preventing the whole neuron migrating
through; only the axons can grow pass (Figure 3-10). Cells seeded into the soma
compartment and then induced towards the microchannels through manipulation of the
fluid pressure. A positive pressure from the soma to axonal compartment can be created
by increasing the fluid height in the soma compartment relative to the axonal. A higher
height corresponds to higher pressure than the other compartment leading to a positive
pressure towards the axonal compartment. The axonal compartment can be directly
accessed with placement of BAM to generate chemical gradients.
85

Figure 3-10: Illustration of µCCM working principle (not to scale) with two axon
compartments. Neurons are loaded into the soma compartment and cultured within the
device. Fluidic pressure from the soma to the axon compartments moves the neurons into
the central channel toward the axon compartments. The microchannels physically impede
the neuronal soma from entering the axon compartment while allowing it to extend
axons.
3.7.2 CCM Fabrication
The CCM was defined through soft lithography utilizing SU-8 molds. A prime
Si wafer was first obtained and treated with A-174, an adhesion promoter for Parylene. A
base layer of Parylene (~2-5 m) was then deposited. This base Parylene layer was
optional but utilized to promote the adhesion of subsequent SU-8 layers. While SU-8
adhesion to Si is adequate, SU-8 PR is still susceptible to thermally induced stress and
strain as well as swelling during development. All of these effects can lower the adhesion
between the final SU-8 mold to the substrate. Anecdotally, many PDMS devices created
through soft lithography break their molds upon release due to SU-8 delamination,
negating the benefits of constructing only one mold for fabricating multiple devices. The
addition of a soft Parylene layer adhered to the substrate can absorb the stress and strain
86
SU-8 undergoes during processing [19, 20] and present a better surface for SU-8
adhesion than bare Si.
A thin layer of SU-8 2 (Microchem Corp., Newton, MA) was spun on to a height
of 3 m over the base Parylene. This SU-8 layer was then patterned and developed to
create molds for CCM microchannels. After developing and drying, SU-8 2035
(Microchem Corp., Newton, MA) was spun on to a height of ~100 m. A ramped
softbake from room temperature to 95 ºC followed by a controlled temperature ramp
down was performed to minimize SU-8 thermal shock due to its thicker nature. The same
temperature ramp up and down was utilized for post-exposure baking (PEB) for the same
reason. This thicker SU-8 layer was patterned to create molds for the soma and axonal
compartments of CCM as well as pipetting access ports.
After SU-8 processing, the wafer was diced using a scribe into the separate device
molds. SU-8 device outlines had been patterned which helped guide the scribe during
dicing. After dicing, the molds were standard cleaned and N
2
dried. Acrylic blocks
matching the dimensions of the axonal compartment and soma compartment access ports
were laser machined and epoxied to the SU-8 mold. This was done to add height to the
compartment so that the resulting chambers were ~1/8 inch (3.175 mm) in height to
contain adequate volume as opposed to ~100 m. Care was taken to ensure excess epoxy
did not encapsulate and destroy the microchannel molds. Upon epoxy curing, the molds
were cleaned with DI water and N
2
dried and placed into a clean environment (sealed
petri dish) in preparation for soft lithography.
87
PDMS (Sylgard 184, Dow Corning, Midland, MI) was prepared by mixing the
prepolymer and curing agent in a 1:10 ratio. A mold was placed into a petri dish and
PDMS was poured until the PDMS just covered the mold. A transparency sheet cut to
size was placed over the mold and made contact with the acrylic blocks due to the surface
tension pulling effect of PDMS prepolymer [21]. A weight was then placed over the
transparency sheet, nominally pushing out PDMS prepolymer between the transparency
sheet and acrylic block surfaces creating an intimate contact. This was done to ensure the
fabricated module possessed the desired thickness while keeping the acrylic molded
access ports open and free of PDMS. PDMS was cured at ~60 ºC for 1 hour in an oven.
Upon curing and cooldown to room temperature, the CCM was released from the mold
and placed in a sealed petri dish in preparation for cell culturing.
3.7.3 CCM Results
CCM was sterilized with ethanol and dried prior to cell culturing. The module
was then placed onto a clean, sterile glass slide and pressed gently but firmly to seal.
Chick embryonic neurons were seeded into the soma compartment and a pressure
towards the axonal compartment was induced by increasing the soma compartment fluid
height relative to the axonal. Cells were allowed to grow for several days.
Optical microscopy of the cells showed separation of somata and axons with
axons extending into the axonal compartment (Figure 3-11).
88

Figure 3-11: Neurons growing in vitro within µCCM. Axons can be seen growing into
the axonal compartment (arrows) with the somata in the other compartment.
3.8 BAM
3.8.1 BAM Design
The module for delivering chemicals to the neurons needs to be easy to use; thus
the device should possess minimal packaging and be mechanically robust as to not easily
fail. Minimizing packaging reduces the size of the complete chemical interfacing setup
enabling system scalability. Finally, compatibility with cell culture technologies is also
required.
Choosing which material to construct BAM from requires careful consideration
of the requirements for mechanical robustness and compatibility with cell culturing.
Surfaces which are immersed in cell media need to be resistant to biofouling; thus
Parylene C is a logical choice. Parylene microchannels are fragile as experienced before
89
so another micromachinable material to give support and mechanical rigidity is needed;
SU-8. Both Parylene and SU-8 are used extensively in lab on a chip applications and are
compatible with neuronal cell culturing; especially if Parylene overcoats the SU-8 [22].
The size of the system can be drastically reduced by eliminating an external pump
for flowing chemicals through the device. Passive pumping can be exploited to allow
fluids to self-pump [23]. Elimination of a syringe pump dramatically simplifies
packaging since macro-to-micro fluidic connections are not required in addition to the
various plumbing in the form of tubes. Self-pumping increases BAM’s ease of use since
simple pipetting, a staple of cell culturing technique, is utilized to generate droplets of
various sizes (Figure 3-12). Exploiting passive pumping allows for massive scaling of
BAM use since the system is not dependent on the availability of expensive pumps,
therefore multiple modules can be used at once even within an incubator.

Figure 3-12: Time-lapse mockup of passive pumping within µBAM. Flow within the
microchannel is self-driven from the smaller fluid droplet into the larger. Outlines of the
original droplets are show to better illustrate the change in droplet size.
3.8.2 Passive Pumping Theory
Passive pumping takes advantage of the Laplace Law which states that a droplet
of fluid possesses an internal pressure that is a function of the droplet radius [24]. The
90
internal pressure is inversely related to the radius; therefore smaller droplets possess
higher pressures than larger ones. When two droplets of differing sizes are fluidically
connected, there exists a pressure difference between the two due to the different internal
pressures. This pressure difference generates flow from the smaller droplet to the larger
one until the smaller droplet collapses into the larger.
The flow rate that results from passive pumping can be modeled from the Young-
Laplace equation as such. The internal pressure difference arising within an input
(smaller) droplet is:






  
2 1
1 1
R R
P  (3)
where P is pressure, γ is surface tension and R
1
and R
2
are the principal radii of the
curvature. Assuming that the input droplet is a perfect sphere for simplification:
2 1 R R  and
R
P
 2
  (4)
As the input droplet flows into the larger, its radius changes as a function of time
thus internal pressure driving the flow also changes with time. Passive pumping flow rate
is related to the change in internal pressure [24]:
Q(t) K
t R
W 
) (
2 
(5)
where K
W
is a function of channel dimensions and liquid viscosity and Q(t) is flow rate.
While flow rate changes over time with passive pumping, a steady state is possible with
91
refilling the input droplet before it becomes too small, maintaining its internal pressure
[24].
As seen in the Young-Laplace equation as well as intuitively, the greatest pressure
is generated when the drop is spherical. As the droplet spreads out, its internal pressure
decreases since its radius increases and so achievable flow rate is lowered. Therefore,
efficiency of passive pumping is greatest when droplets are placed on a hydrophobic
surface in order to increase their internal pressures by “beading” up.
PDMS is used for passive pumping since it is highly hydrophobic (contact angle
of 105.9 ± 4.5 º [25]) and is easy to construct microchannels with through soft
lithography [23]. Other hydrophobic micromachinable materials can be used such as
Parylene (contact angle of 97.2 ± 4.2 º [25]) which will maintain high internal pressure of
the input droplet.
3.8.3 BAM Fabrication
Parylene is used to create a hydrophobic surface upon which droplets can be
placed for passive pumping through a microchannel. The microchannel is constructed
from SU-8 so that it is rigid and not collapsible, contributing to BAM’s ease of use. A
pore is plasma etched into the Parylene covering the microchannel providing a route
through which chemicals are delivered.
This pore can be tuned to smaller diameters not achievable with other materials or
even Parylene plasma etching [26]. This is achieved by taking advantage of Parylene’s
92
conformal coating nature; i.e. a pore with a diameter of 3 m deposited with an additional
1 m of Parylene will result in a pore diameter of 1 m because the additional Parylene
will coat around the pore circumference shrinking its diameter (Figure 3-13). While this
concept can work with other base materials such as PDMS or SU-8, the achievable initial
pore diameters would not be as small as that for Parylene.

Figure 3-13: (a) An initially sized pore is etched into Parylene with a cross-sectional view
on the left and the top view on the right. (b) Subsequent Parylene coating over this pore
creates a smaller diameter pore that is tunable by adjusting the deposited Parylene
thickness. The nature of conformal coating is exaggerated in this illustration.

BAM fabrication requires a SU-8 microchannel that is capped with a Parylene
roof (Figure 3-14). Creating embedded SU-8 microchannels is not a trivial task [27];
especially with another material such as Parylene. The usual method of overcoating PR
with Parylene [28] and then removing the sacrificial material will not work with SU-8.
SU-8 is difficult to use as a sacrificial material due to its low dissolution rate and the need
to protect this unexposed SU-8 from subsequent exposure and crosslinking which make it
insoluble [29, 30].
93

Figure 3-14: Illustration of µBAM. Passive pumping is exploited to generate flow in the
microchannel. Chemicals are released through a pore etched in the Parylene layer. The
microchannel is constructed of SU-8 that is capped with Parylene.

AZ PR can be utilized as a sacrificial material without the difficulties involved
with SU-8. These PR’s have been used in the construction of Parylene microchannels as
described before for the microfluidic platform [5] and for the construction of Parylene
sheath probes [31, 32]. A sacrificial PR is patterned which is then overcoated with
Parylene. Dissolution of the sacrificial material in acetone or IPA releases the Parylene
microchannel.
AZ PR’s use as a sacrificial material could be utilized in the construction of
BAM such that its microchannels are first patterned in SU-8 and then filled with the
sacrificial material. An overcoat of Parylene caps the microchannel while the sacrificial
PR prevents undesirable conformal coating within the microchannel (Figure 3-15). Pores
and access holes can be plasma etched into the Parylene through which the sacrificial PR
can be removed through solvent dissolution. The pore diameter can then be tuned to a
desired size.
94

Figure 3-15: Illustration of how sacrificial PR can be used to construct the embedded SU-
8 microchannel required for µBAM.

There are a number of difficulties involved with this fabrication method chiefly
related to the use of sacrificial PR. The application of sacrificial PR to the developed SU-
8 microchannels needs to be precise such that the microchannels are filled exactly. Any
overflow of sacrificial material over the SU-8 will result in Parylene not sealing the
microchannel. Conversely, underflow of sacrificial PR will result in a sagging Parylene
roof; which is undesirable since a non-planar surface can impede proper patterning and
etching of the Parylene layer. The actual act of sacrificial material filling is can be
problematic as well. Spin coating of AZ PR is limited in achievable thickness (~20-40
m) and is not suited for thicker SU-8 applications of < 100 m as would be the case for
BAM. Multiple spins to achieve thicker sacrificial PR height may be possible however
the multiple repetitive steps involved is time-consuming and subsequent softbakes may
thermally degrade the Parylene base layer which is required for releasing the device from
95
the wafer. It is unlikely that the non-planar nature of the patterned SU-8 microchannels
will be filled uniformly through spin coating. Another method to filling in the patterned
microchannels with sacrificial PR is to manually pipette PR into the microchannels. This
method can potentially fill the microchannels in one step and avoid the issues with
multiple spinning. However, difficulties may arise since SU-8 possesses a relatively high
contact angle (80 º [33]) and thus the patterned narrow channels may be resistant to
filling. The manual process can be time-consuming since each microchannel on a wafer
must be individually filled. Finally, there is no guarantee that a perfect fill will be
achieved, surface tension effects may result in uneven sacrificial material filling with
both overflow and underflow present within the same microchannel leading to an
undesirable Parylene cap.
A simpler method would be to utilize polymer bonding to cap the SU-8
microchannel for BAM construction.
3.8.4 Thermocompressive Polymer Bonding
Polymers can be mechanically bonded to each other through a process of
thermocompression in which a combination of pressure and thermal energy leads to
polymer entanglement at the polymer interface (Figure 3-16).
96

Figure 3-16: Thermocompression between two polymer surfaces leads to polymer
entanglement and a subsequent bond.

This technique (originally referred to as Parylene micromolding [34]) has been
utilized to construct Parylene microchannels for use in neural probes [35, 36]. An etched
Si wafer was overcoated with Parylene to create open microchannels. The microchannels
were then capped through thermocompressive wafer bonding with another Parylene
coated wafer. Upon release from both wafers, a complete Parylene microchannel was
obtained.
In addition to Parylene/Parylene bonding, thermocompression has also been
utilized in SU-8/SU-8 bonding. Many applications have been proposed such as wafer
level packaging with SU-8 [37-39] and 3D embedded SU-8 microchannels [40]. The
process is similar to Parylene/Parylene bonding in that a patterned SU-8 layer is
interfaced to another SU-8 layer within a jig. Pressure and heat are applied and the
polymers entangle at the SU-8/SU-8 interface creating a bond.
The resultant polymer bonds can be characterized through measurements of shear
strength via pull tests [40, 41]. Four variables affect the quality and strength of the bond;
temperature, pressure, time, and the presence of vacuum during the process. Vacuum is
97
required for Parylene/Parylene bonding due to the thermal oxidative degradation of the
polymer at elevated temperatures. Within the literature for SU-8/SU-8 bonding, vacuum
has always been applied as part of the bonding process utilizing a commercial wafer
bonder; usually to achieve vacuum packaging of devices.
Unfortunately, no examples exist within literature of SU-8/Parylene bonding so
far. From literature reporting on SU-8/SU-8 and Parylene/Parylene bonding; variables
compatible with both SU-8 and Parylene can be extrapolated and utilized for SU-
8/Parylene bonding. Parylene/Parylene thermocompressive bonding was found to occur
at bonding temperatures of 130-290 °C with an optimal temperature of 230 °C (Figure 3-
17) [41]. Similarly, SU-8/SU-8 bonding was tested across a temperature range of 50-150
°C and a maximum bond strength was obtained at 90 °C (Figure 3-18) [37]. In another
report, SU-8/SU-8 bonding was achieved at 200 °C [42]. Temperatures much greater than
150 °C are compatible with SU-8 since the glass transition temperature of SU-8 is ~200
°C with thermal degradation not occurring until ~380 °C [43]. Therefore, a
thermocompressive bonding temperature range of ~130-200 °C can be assumed to be
compatible for both SU-8 and Parylene polymers with an expected general trend of
stronger resultant bond strengths from higher bond temperatures.
98

Figure 3-17: Effects of temperature on Parylene/Parylene thermocompressive bonds.
Generally, higher bond strength is achieved with elevated bonding temperature. From
[41] © 2005 IEEE.

Figure 3-18: Effects of temperature on SU-8/SU-8 thermocompressive bond strengths.
Generally, higher bond strength is achieved with elevated bonding temperature. Adapted
from [37] © IOP Publishing. Reproduced by permission of IOP Publishing. All rights
reserved.
99
3.8.5 SU-8/Parylene Bonding
A jig was constructed for applying compressive forces (Figure 3-19) for polymer
bonding. Screws are used to tighten the two aluminum plates together and apply
compressive stress. The samples to be bonded are sandwiched between two Teflon sheets
to ensure an even distribution of stress.

Figure 3-19: Jig for thermocompressive bonding. The samples to be bonded, a patterned
SU-8 die and Parylene-coated glass slide are first placed between two Teflon sheets
(white). The two aluminum plates are then tightened together with the use of screws to
apply compression to the samples.

Thermocompression commenced by first placing a micromachined SU-8
microchannel die under a Parylene coated (~10 µm) glass slide within the jig. The screws
of the jig were tightened to apply compression and placed into a vacuum oven
(Laboratory Vacuum Oven Model V0914A, Lindberg/Blue, Asheville, NC). Vacuum was
applied and the oven heated to 200 °C for ~18 hours. The oven was then allowed to cool
down to room temperature still under vacuum in order to prevent Parylene oxidative
100
degradation. After cool down, the oven chamber was vented back to atmospheric pressure
and the jig removed. The bonded sample was then removed.
A successful SU-8/Parylene bond was obtained (Figure 3-20) providing proof of
principle for SU-8/Parylene bonding. Empirically, the bond was strong as attempting to
pull the SU-8 die away from the Parylene coated slide sometimes resulted in
delamination of the SU-8 from the Si substrate.

Figure 3-20: A SU-8 die bonded to a Parylene coated glass slide. The SU-8/Parylene
bond keeps the die suspended.
3.8.6 SU-8/Parylene Bond Discussion
A release layer may be necessary to release the bonded Parylene film from its
original substrate to the SU-8 layer. Anecdotally, soda lime substrates may be used for
transferring Parylene films though these substrates require a Parylene release layer
application, Micro-90 (International Products Corporation, Burlington, NJ), prior to
Parylene deposition as done in the construction of Parylene neural probes through
Parylene/Parylene bonding [35]. Sacrificial PR may also be used for the release of
Parylene [44]. Other more complicated release steps can involve metal release layers
such as Al which is then etched away through a wet chemical process. Any release layer
process chosen must be compatible with the bonded SU-8.
101
SU-8/Parylene thermocompressive bonding was demonstrated on a die level
however this technique is scalable to whole wafer level bonding wherein many devices
are bonded at once. This scalability makes this bonding method attractive as compared to
other manually intensive methods of capping SU-8 microchannels; such as individually
filling with sacrificial PR and then overcoating with Parylene which requires a final PR
dissolution step. The uniformity of the resultant SU-8/Parylene bond over many devices
through thermocompressive bonding may be superior to that achieved with the use of
sacrificial PR microchannel filling.
The ability to cap SU-8 microchannels with Parylene enables a key requirement
of BAM; mechanically robust microchannels possessing a small pore for chemical
delivery. Rigid SU-8 walls make the channel highly resistant to collapse while the
Parylene cap presents a biocompatible surface for interfacing with cells and media.
Moreover, the Parylene can be plasma etched to create small pores for chemical delivery.
SU-8/Parylene bonding technology can also enable the construction of other devices
which require capped SU-8 channels such as 3D networked microchannels. Embedded
SU-8 structures with high aspect ratios can be constructed with multiple PR layers
utilizing this technology since a planar surface is presented after bonding for subsequent
lithographic processing as non-planar surfaces can inhibit the spin coating of additional
PR.
3.9 Chemical Interfacing Modules Conclusion
New devices in the form of modules were designed and constructed to improve
micromachined chemical interfacing. First, the challenge of targeted chemical delivery to
102
a specific location was delegated to CCM by patterning neural cell culture growth in
vitro. Neural axons and somata were separated and could be targeted for chemical
delivery for axonal studies. The challenge of delivering chemicals in an easy to use
manner was delegated to another module, BAM. Passive pumping eliminated the
reliance on external, bulky pumps and enables experimental scaling. A novel fabrication
technology, SU-8/Parylene thermocompressive bonding was developed for capping SU-8
microchannels with a Parylene layer that could be subsequently processed upon through
plasma etching. This enables a chemical interface with non-collapsible SU-8
microchannels that possesses biocompatibility properties with the Parylene layer.
Coupled with high ease of use since only the application of droplets are required to drive
passive pumping; a much improved neural chemical interface is realized.
103
REFERENCES
[1] E. R. Kandel, J. H. Schwartz, and T. M. Jessell, Principles of neural science, 4th
ed. New York: McGraw-Hill, Health Professions Division, 2000.
[2] N. T. Nguyen, "Micromachined flow sensors - a review," Flow Measurement and
Instrumentation, vol. 8, pp. 7-16, Mar 1997.
[3] J. T. W. Kuo, L. Yu, and E. Meng, "Micromachined Thermal Flow Sensors—A
Review," Micromachines, vol. 3, pp. 550-573, 2012.
[4] E. F.-C. Meng, "MEMS technology and devices for a microfluid dosing system,"
Ph.D. 3093554, Electrical Engineering, California Institute of Technology,
Pasadena, California, USA, 2003.
[5] J. T. W. Kuo, L.-Y. Chang, P.-Y. Li, T. Hoang, and E. Meng, "A microfluidic
platform with integrated flow sensing for focal chemical stimulation of cells and
tissue," Sensors and Actuators B: Chemical, vol. 152, pp. 267-276, 2011.
[6] G. G. Harman, Wire Bonding in Microelectronics: Materials, Processes,
Reliability, and Yield, 2nd ed. New York: McGraw-Hill, 1997.
[7] H. Kaji, M. Nishizawa, and T. Matsue, "Localized chemical stimulation to
micropatterned cells using multiple laminar fluid flows," Lab Chip, vol. 3, pp.
208-211, 2003.
[8] R. Pal, K. E. Sung, and M. A. Burns, "Microstencils for the patterning of
nontraditional materials," Langmuir, vol. 22, pp. 5392-5397, 2006.
[9] T. M. Keenan and A. Folch, "Biomolecular gradients in cell culture systems," Lab
on a Chip, vol. 8, pp. 34-57, 2008.
[10] S. J. Butler and J. Dodd, "A Role for BMP Heterodimers in Roof Plate-Mediated
Repulsion of Commissural Axons," Neuron, vol. 38, pp. 389-401, 2003.
104
[11] G. Mallén-Ornelas, L. Y. Chang, P. Y. Li, T. Hoang, L. J. Ho, K. Swertfager, et
al., "A Microfluidic Platform for Focal Chemical Stimulation of Cells," in Micro
Total Analysis Systems 2008, San Diego, CA, 2008, pp. 613-615.
[12] L. Y. Chang, P. Y. Li, L. Zhao, T. Hoang, and E. Meng, "Integrated Flow Sensing
for Focal Biochemical Stimulation," in NEMS 2008, Sanya, China, 2008, pp. 921-
6.
[13] S. K. Sia and G. M. Whitesides, "Microfluidic devices fabricated in poly
(dimethylsiloxane) for biological studies," Electrophoresis, vol. 24, pp. 3563-
3576, 2003.
[14] A. Mata, A. Fleischman, and S. Roy, "Characterization of Polydimethylsiloxane
(PDMS) Properties for Biomedical Micro/Nanosystems," Biomedical
Microdevices, vol. 7, pp. 281-293, 2005/12/01 2005.
[15] K. J. Regehr, M. Domenech, J. T. Koepsel, K. C. Carver, S. J. Ellison-Zelski, W.
L. Murphy, et al., "Biological implications of polydimethylsiloxane-based
microfluidic cell culture," Lab on a Chip, vol. 9, pp. 2132-2139, 2009.
[16] Y. Lei, Y. Liu, W. Wang, W. Wu, and Z. Li, "Studies on Parylene C-caulked
PDMS (pcPDMS) for low permeability required microfluidics applications," Lab
Chip, vol. 11, pp. 1385-1388, 2011.
[17] A. M. Taylor, M. Blurton-Jones, S. W. Rhee, D. H. Cribbs, C. W. Cotman, and N.
L. Jeon, "A microfluidic culture platform for CNS axonal injury, regeneration and
transport," Nature methods, vol. 2, pp. 599-605, 2005.
[18] A. M. Taylor, S. W. Rhee, C. H. Tu, D. H. Cribbs, C. W. Cotman, and N. L. Jeon,
"Microfluidic multicompartment device for neuroscience research," Langmuir,
vol. 19, pp. 1551-1556, 2003.
[19] R. Lo and E. Meng, "Integrated and reusable in-plane microfluidic interconnects,"
Sensors and Actuators B: Chemical, vol. 132, pp. 531-539, 2008.
[20] J. T. W. Kuo and E. Meng, "Improved process for high yield 3D inclined SU-8
structures on soda lime substrate towards applications in optogenetic studies,"
105
presented at the 2012 IEEE 25th International Conference on Micro Electro
Mechanical Systems (MEMS), 2012.
[21] B.-H. Jo, L. M. Van Lerberghe, K. M. Motsegood, and D. J. Beebe, "Three-
Dimensional Micro-Channel Fabrication in Polydimethylsiloxane (PDMS)
Elastomer," Journal of Microelectromechanical Systems, vol. 9, pp. 76-81, 2000.
[22] V. N. Vernekar, D. K. Cullen, N. Fogleman, Y. Choi, A. J. García, M. G. Allen, et
al., "SU-8 2000 rendered cytocompatible for neuronal bioMEMS applications,"
Journal of Biomedical Materials Research Part A, vol. 89A, pp. 138-151, 2009.
[23] G. M. Walker and D. J. Beebe, "A passive pumping method for microfluidic
devices," Lab on a Chip, vol. 2, pp. 131-134, 2002.
[24] E. Berthier and D. J. Beebe, "Flow rate analysis of a surface tension driven
passive micropump," Lab on a Chip, vol. 7, pp. 1475-1478, 2007.
[25] T. Y. Chang, V. G. Yadav, S. De Leo, A. Mohedas, B. Rajalingam, C.-L. Chen, et
al., "Cell and Protein Compatibility of Parylene-C Surfaces," Langmuir, vol. 23,
pp. 11718-11725, 2007/11/01 2007.
[26] R. Kawano, T. Osaki, H. Sasaki, and S. Takeuchi, "A Polymer-Based Nanopore-
Integrated Microfluidic Device for Generating Stable Bilayer Lipid Membranes,"
Small, vol. 6, pp. 2100-2104, 2010.
[27] Y. J. Chuang, F. G. Tseng, J. H. Cheng, and W. K. Lin, "A novel fabrication
method of embedded micro-channels by using SU-8 thick-film photoresists,"
Sensors and Actuators A: Physical, vol. 103, pp. 64-69, 2003.
[28] K. Walsh, J. Norville, and Y.-C. Tai, "Photoresist as a sacrificial layer by
dissolution in acetone," in Micro Electro Mechanical Systems, 2001. MEMS 2001.
The 14th IEEE International Conference on, 2001, pp. 114-117.
[29] L. J. Guerin, M. Bossel, M. Demierre, S. Calmes, and P. Renaud, "Simple and
low cost fabrication of embedded micro-channels by using a new thick-film
photoplastic," in Solid State Sensors and Actuators, 1997. TRANSDUCERS '97
Chicago., 1997 International Conference on, 1997, pp. 1419-1422 vol.2.
106
[30] C. Chung and M. Allen, "Uncrosslinked SU-8 as a sacrificial material," Journal of
Micromechanics and Microengineering, vol. 15, pp. N1-N5, Jan 2005.
[31] J. T. W. Kuo, B. J. Kim, S. A. Hara, C. Lee, C. A. Gutierrez, T. Hoang, et al.,
"Fabrication of 3D Parylene Sheath Probes for Reliable Neuroprosthetic
Recordings," presented at the Hilton Head 2012: Solid State Sensors, Actuators
and Microsystems Workshop, Hilton Head Island, South Carolina, USA, 2012.
[32] J. T. W. Kuo, B. J. Kim, S. A. Hara, C. D. Lee, C. A. Gutierrez, T. Q. Hoang, et
al., "Novel flexible Parylene neural probe with 3D sheath structure for enhancing
tissue integration," Lab on a Chip, vol. 13, pp. 554-561, 2013.
[33] N. J. Shirtcliffe, S. Aqil, C. Evans, G. McHale, M. I. Newton, C. C. Perry, et al.,
"The use of high aspect ratio photoresist (SU-8) for super-hydrophobic pattern
prototyping," Journal of Micromechanics and Microengineering, vol. 14, p. 1384,
2004.
[34] H.-S. Noh, Y. Huang, and P. J. Hesketh, "Parylene micromolding, a rapid and
low-cost fabrication method for parylene microchannel," Sensors and Actuators
B: Chemical, vol. 102, pp. 78-85, 2004.
[35] Y. Yu-Chang, L. Yu-Tao, C. Yen-Chung, and F. Weileun, "A pneumatic neural
probe structure fabricated by parylene thermal bonding technique," in Micro
Electro Mechanical Systems (MEMS), 2012 IEEE 25th International Conference
on, 2012, pp. 381-384.
[36] D. Ziegler, T. Suzuki, and S. Takeuchi, "Fabrication of Flexible Neural Probes
With Built-In Microfluidic Channels by Thermal Bonding of Parylene,"
Microelectromechanical Systems, Journal of, vol. 15, pp. 1477-1482, 2006.
[37] C. Pan, H. Yang, S. Shen, M. Chou, and H. Chou, "A low-temperature wafer
bonding technique using patternable materials," Journal of Micromechanics and
Microengineering, vol. 12, p. 611, 2002.
[38] Y. Kim, S. Yi, S. Kim, and B. Ju, "A novel low-temperature microcap packaging
using SU-8 bonding," in Solid-State Sensors, Actuators and Microsystems
Conference, 2007. TRANSDUCERS 2007. International, 2007, pp. 2107-2110.
107
[39] G. Murillo, Z. J. Davis, S. Keller, G. Abadal, J. Agusti, A. Cagliani, et al., "Novel
SU-8 based vacuum wafer-level packaging for MEMS devices," Microelectronic
Engineering, vol. 87, pp. 1173-1176, 2010.
[40] F. J. Blanco, M. Agirregabiria, J. Garcia, J. Berganzo, M. Tijero, M. T. Arroyo, et
al., "Novel three-dimensional embedded SU-8 microchannels fabricated using a
low temperature full wafer adhesive bonding," Journal of Micromechanics and
Microengineering, vol. 14, pp. 1047-1056, Jul 2004.
[41] H. Kim and K. Najafi, "Characterization of low-temperature wafer bonding using
thin-film parylene," Microelectromechanical Systems, Journal of, vol. 14, pp.
1347-1355, 2005.
[42] S. Rajaraman, H.-s. M. Noh, P. J. Hesketh, and D. S. Gottfried, "Rapid, low cost
microfabrication technologies toward realization of devices for dielectrophoretic
manipulation of particles and nanowires," Sensors and Actuators B: Chemical,
vol. 114, pp. 392-401, 2006.
[43] A. Del Campo and C. Greiner, "SU-8: a photoresist for high-aspect-ratio and 3D
submicron lithography," Journal of Micromechanics and Microengineering, vol.
17, p. R81, 2007.
[44] D. C. Rodger, A. J. Fong, W. Li, H. Ameri, A. K. Ahuja, C. Gutierrez, et al.,
"Flexible parylene-based multielectrode array technology for high-density neural
stimulation and recording," Sensors and Actuators B: Chemical, vol. 132, pp.
449-460, 2008.

108

4.1 Background
Optogenetics, the term used to describe the use of channelrhodopsins (ChR2)
within neural tissue has proven to be very promising in its in vivo applications of
studying neural circuits [1]. Animal models (typically rodents) are transfected so that
their neurons express channelrhodopsins and researchers are then able to control and
probe the animals’ neural circuitry optically and see how the animals’ behavior is
affected. This is typically done by inserting an optical fiber and transmitting the
appropriate light wavelength to activate or deactivate the neural tissue (Figure 4-1).
While exciting, this technique is currently in its infancy and the current state of the art is
limited to one optical fiber coupled to an MEA [2].

Figure 4-1: In vivo study of neural circuitry using optogenetic techniques. An optical
fiber is inserted into the rodent's brain and the light activates the underlying neural tissue
(adapted from [1]).

Chapter 4 INCLINED SU-8 MIRRORS FOR OPTICAL
STIMULATION
109
Surprisingly, in vitro use of optogenetics or photocaged neurotransmitters has
been limited by the abilities of the tools currently available. Typically, a glass coverslip
with photostimulatable neural tissue and/or photocaged neurotransmitter solution is
placed under a microscope. The light from the microscope activates the neurotransmitter
or transgenic neural tissue directly (Figure 4-2). Such a setup is limited since there is no
precise control over how much light is activating the sample since the light source is from
the microscope. Also, because the light source is from the microscope, there is no control
over the location of the light activation; one can only move the stage of the microscope if
a different point of activation is desired. Finally, because there is only one light source,
there is no method of optically probing multiple points at once.

Figure 4-2: Illustration of setup in which photostimulus is delivered via a microscope
objective. Only one light source is available at a fixed location on the stage.

The same techniques have been applied to optically probe cell cultures
photosensitized through transfection. The field of optogenetics was born with the first
110
insertion of ChR2 ions into cultured hippocampal neurons [3]. Later work from the same
group developed and inserted NpHR, a light sensitive chloride pump that complements
ChR2 by inhibiting action potential generation, again into hippocampal neurons [4].
Likewise, photosensitized PC12 cells were created to study ChR2 kinetics [5]. In these
works, whole-cell patch clamping was utilized to monitor electrical activity while
transfected neurons were optically probed to verify depolarization/hyperpolarization with
ChR2 and NpHR respectively. After the establishment and adoption of ChR2 and NpHR
throughout the neurosciences, transfected optogenetic cells and neural tissue slices have
been utilized to study visual processing pathways [6], long-range callosal projections in
the somatosensory cortex [7], the spatial distribution of synaptic circuits within the
cereberal cortex [8], feed-forward processing within the thalamus and neocortex [9], and
the formation of glutamatergic synaptic networks [10]. As before, experimental
throughput is severely limited due to a finite light source of one out of the microscope
objective and spatial requirements associated with patch clamping. Voltage-sensitive
dyes have been utilized in conjunction with optogenetics to avoid patch clamping
however the source of light stimulation is still limited to output from the microscope
objective and the photostimulus interferes with the dyes’ function [11].
Some have proposed a solution to these limitations by using optical fibers. Optical
fibers allow for targeted light sources that can be as small as a few µm’s [12]. The setup
typically involves a light source coupled to optical fibers that then direct the light to the
sample for localized light activation [13, 14]. Such setups are limited in their spatial and
temporal abilities since micromanipulators are typically used to position the optical fibers.
111
The spacing requirements of the micromanipulators as well as the optical fibers limit how
many fibers can be simultaneously positioned over the sample; thus most studies utilize
only one optical fiber. The temporal resolution is also limited since there is a severe time
limitation on how fast one can position the sample to another point of interest due to the
experimental setup.
Development of an integrated device with an optical fiber coupled with an MEA
has been initiated. Such a device allows for local photostimulation of the neural tissue or
neurotransmitters and localized recordings from the MEA [2, 15]. While this is an
improvement over the cumbersome method of using micromanipulators, this technique
suffers from the limitation of having only one optical fiber and so can only activate one
area at a time.
The use of a fast-scanning laser as a light source has been used to address the lack
of spatial and temporal control [16]. While this addresses the issue of probing multiple
points quickly, it still suffers from not being able to simultaneously probe many points
since there is only one light source.
4.2 Aim
A microfabricated mirror array chip possessing embedded optical fibers that
direct light towards target tissue areas for photostimulation will solve issues with
previous technologies. Mirrors are created from metal coated inclined SU-8 structures.
The device also possesses optical fiber guides for optical fiber integration.
112
The optical fibers guide light into the chip and light is directed upwards toward
the sample for photostimulation (Figure 4-2). Such a device allows for temporal and
spatial control of light activation since multiple points can be photostimulated at once and
quickly since there would be no need to move the sample. Using microfabrication
techniques, an array of optical fibers can be densely arranged and so match the
complexity of neural circuitry.

Figure 4-3: Cross section of the proposed device (not to scale). The metal layer over the
inclined SU-8 structure acts as a mirror to guide the incoming blue light upwards in order
to locally activate the optogenetic cell.

4.3 Theory
ChR2 channels require a blue wavelength (λ = 470 nm) in order to activate with a
minimum optical power of 1 mW/mm
2
[1]. Therefore optical power loss from light
transmission through media (air, glass, tissue, etc.) needs to be minimized.
SU-8 is an epoxy based negative photoresist (PR) introduced by IBM. It has been
used for thick photoresist applications and is an alternative to LIGA processing. SU-8 has
also found much use as a mold for soft lithography processing since large aspect ratios
are possible. The obtainable thickness of SU-8 makes it amenable to inclined processing
113
in which light is exposed at a controlled angle to the PR and inclined structures are
obtained.
Backside inclined processing of SU-8 has been pursued by various groups since
its introduction by Peterman for microneedle fabrication [17]. Backside inclined
processing is an unconventional method in which the soda lime substrate is used as the
mask for the PR. As indicated by the nomenclature, SU-8 is deposited onto a UV
transparent substrate patterned with a mask. After softbaking, exposure light is
introduced through the backside of the substrate. This method ensures a perfect gapless
contact between mask and SU-8 throughout the entire substrate and high aspect ratios are
achivable with high yield. An angle can be introduced to create inclined structures.
Inclined exposure has inherent limitations in the available exposure angle of SU-8
(~35º) due to Snell’s law. Light must pass through different mediums and the respective
refraction indexes, the incident angle changes through each medium as described by:

There is a limit of degrees achievable with only air and SU-8 medium due to the different
indices of refraction.
One method of extending the achievable angle is by use of an index matching
medium. Glyerol has been used in which the wafer, PR, mask are immersed and UV light
is exposed through the glycerol bath [18]. The exposure light path passes through air,
glycerol, mask medium (soda lime), and finally through SU-8 PR. Glycerol’s refraction
114
index corrects the light path thereby extending the achievable angled exposure as
described:

Water has also been used for index matching however the achievable angle is less than
glycerol [19].
4.4 Design
An array of mirrors with fiber guides was fabricated using SU-8 processing for
photostimulation studies. The fiber guides are 250 µm high in order to accommodate the
inserted fibers, 245 µm diameter with coating (F-SV, Newport Corporation, Irvine, CA).
The mirror array is fabricated by creating inclined SU-8 structures through
backside inclined exposure. 45º angled SU-8 structures are then coated with a reflective
metal, Pt or Ti to create mirrors.
4.5 Fabrication
Processing was first performed on a Si substrate since SU-8 has been reported to
have adhesion issues with soda lime [20]. An anti-reflection coating CLK600 was first
spun on and softbaked. This anti-reflection layer is required to prevent unwanted
115
exposure from reflectance of UV light off of wafer surface. A 250 µm layer of SU-8 was
then spun on and softbaked. A 10 minute wait before softbaking was performed for SU-8
reflow. Softbaking was done on a level hotplate with an inverted funnel covering the
wafer to protect against ambient air currents. Wafer was softbaked in a controlled manner
with a temperature ramp of 5ºC/min from room temperature until 95ºC at which a 2 hour
bake occurred. A controlled cool down to room temperature then occurred. The
controlled temperature ramping is required to avoid SU-8 thermal shock. The wafer was
attached to the mask and set at an incline on a mask holder. The assembly was placed in a
glycerol bath placed under exposure lamp and a 10 minute wait ensued for glycerol
settling. Exposure dose was determined by this formula:
Exposure dose = 20 mJ/cm
2
*SU-8 thickness (µm)
After exposing, the wafer was removed from the bath and wafer holder and placed
on hot plate for post-exposure baking (PEB). As with softbaking, a controlled
temperature ramp from room temperature occurred with baking at 95ºC for 2 hours. A
controlled cool down also occurred. After PEB, the wafer was developed in SU-8
developer. After development, the wafer was rinsed with IPA and dried with N
2
gas.
116

Figure 4-4: Inclined exposure of SU-8 in a glycerol bath. An anti-reflection layer between
Si substrate and SU-8 photoresist is required to prevent undesired exposure from
reflections off the polished wafer surface.

The fabrication process was then modified to address issues with SU-8 processing
[Appendix L]. A soda lime substrate was used for backside inclined exposure. A soda
lime wafer was first patterned with a Ti mask using lithography and lift-off technologies.
A silane, A-174, treatment then occurred to adhere a subsequent Parylene layer to soda
lime substrate. A 1-2 µm Parylene layer was then deposited. This Parylene layer serves as
an adhesion layer between SU-8 and soda lime since the TCE of SU-8 and soda lime are
mismatched [20]. SU-8 was then spun on at 250 µm thickness and softbaked as before.
The soda lime wafer was then placed on the mask holder positioned so that exposure
occurred through the backside through the mask pattern. The assembly was placed in
glycerol bath as before. Exposure dose, PEB, and development proceeded as before.
117

Figure 4-5: Improved fabrication process with the soda lime substrate serving as the mask
(left). Inclined SU-8 exposure takes place within a glycerol bath placed under the UV
lamp (right).

The SU-8 structure was then coated with reflective metal (Pt or Ti) through
ebeam deposition to create a mirror surface for light reflection. Finally, the devices were
diced out.
118
4.6 Results
4.6.1 Fabrication
Processing on Si substrate with inclined exposure produced adequate SU-8
inclined structures albeit with low yield. SU-8 edge bead introduced gaps between mask
and PR that produced undesirable structures. Even with edge bead removal, slight
unevenness in SU-8 surface produced gaps in random wafer areas. The use of anti-
reflection coating also increased the processing time required.
Processing on soda lime substrate was first attempted without a Parylene adhesion
layer. Resulting SU-8 structures were also very poor in yield due to large portions of SU-
8 delaminating from substrate during processing. Evidence of thermal residual stress was
observed on SU-8 structures that delaminated (Figure 4-5a).
The addition of a Parylene adhesion layer solved all delamination issues. Gapless
contact between mask and SU-8 during exposure led to high yield in obtained structures
(~100%) (Figure 4-5b). High aspect ratios were achieved with this fabrication method.

119

Figure 4-6: Fabrication results (a) without and (b) with a Parylene layer between the soda
lime glass substrate and SU-8. Delamination of SU-8 across the wafer cannot be
prevented without a Parylene stress relief layer. 100% yield was achieved over a 3” soda
lime wafer with the addition of a Parylene layer. No delamination was observed. (c)
Consistent, well-defined angled structures over the entire 3" wafer were obtained, even at
the edges where the photoresist is thinner.

While the major impediments to efficiently constructing inclined SU-8 structures
were solved with backside exposure and a Parylene adhesion layer, there were other
processing issues unique to SU-8 that had to be overcome.
The first of these was ensuring an even coating of SU-8 during the spin on
process. SU-8 is a very viscous fluid prior to softbaking. As such, application of SU-8 to
the center of the wafer substrate is more critical than processing with other less viscous
photoresists such as the AZ line. An off-center pour can lead to complete substrate
coverage but due to the viscous nature of SU-8, an uneven thickness gradient over the
wafer is obtained. Spin speed settings are also critical as a prominent edge bead can be
created with faster spin speeds and longer spin times leading to uneven thickness coating.
SU-8 edge beads can be dealt with in two ways: the first is to physically scrape off the
edge bead using a squeegee sacrificing the edges of the wafer but resulting in consistent
SU-8 thickness in the center; the other simpler method is to carefully calibrate the spin
120
speed with a lower spin time such that minimal to no edge bead results. A minimal edge
bead can be eliminated during the reflow process that occurs during softbaking.
Air bubbles within the photoresist may be encountered during the SU-8 spin on
process which are undesirable since they become part of the SU-8 structure and can
negatively impact the desired pattern or device. Bubbles can be introduced during the
SU-8 pour on process which should be avoided with pouring in a circular, outward
fashion from the substrate center without introducing air into the photoresist (Appendix
L). Despite best efforts, bubbles may be encountered and seen after photoresist spinning.
The use of a small pipette tip (10 L) or transparency sheet corner can pop the bubbles; it
is critical to use a fresh new one for each bubble. Popping the bubble in this fashion
works by introducing a hole; the bubble does not collapse immediately due to the viscous
nature of SU-8. Instead, bubble collapse occurs during photoresist reflow in the
softbaking step.
The SU-8 softbaking step possesses some critical elements which are not found
with other photoresists. SU-8 reflows during softbacking and is vulnerable to negative
effects by the environment due to the long reflow time. The first critical element is to
ensure the hotplate is level during softbaking. An unlevel plate surface will result in an
uneven SU-8 layer due to SU-8 reflow. The second element is to eliminate exposure to
ambient air flow which can induce uneven SU-8 reflow during the softbake process
(Figure 4-7a). A inverted funnel was used to protect the SU-8 during reflow (Figure 4-7b).
Anecdotally, an oven may be used for softbaking which protects the photoresist from
ambient air flow however a SU-8 “skin” can form as solvents on the photoresist surface
121
are first driven out trapping solvents in the center impeding a complete softbake;
therefore softbaking atop a hotplate in which solvents are driven out from the bottom-up
is usually recommended.

Figure 4-7: (a) Ambient air flow during the softbake process causes undesirable patterns
to form in the SU-8. (b) A funnel inverted over the wafer during softbaking protects the
SU-8 from external perturbations and leads to successful softbake with a level SU-8
surface.

Development of thick SU-8 can take a significant amount of time (~hours) in
order to fully develop the structures. The use of an orbital shaker (Incubating Microplate
Shaker, VWR International, Radnor, PA) can automate the process (Figure 4-8) however
the circular motion is not ideal and can result in a longer development time as compared
to development with a side-to-side motion which presents better convection within the
developer bath. The most ideal equipment would be a reciprocating shaker for
automating SU-8 development. SU-8 can absorb the solvents present within SU-8
developer and swell, destroying the desired structure, therefore timely checks throughout
the development process are required.
122

Figure 4-8: Development of SU-8 with the use of an orbital shaker.

4.6.2 Device characterization
4.6.2.1 Fabrication Validation
The fabricated structures were coated with gold and imaged with scanning
electron microscopy (SEM) to verify that the desired SU-8 thickness and angle were
fabricated. Consistent angled structures were obtained over the entire wafer,
demonstrating the effectiveness of the fabrication process.
123

Figure 4-9: SEM images of the resulting SU-8 structures showing that the desired angle
of 45º was obtained.

The bond strength between SU-8 structure to Parylene to underlying soda lime
substrate was first investigated with scotch tape test. SU-8 to soda lime and SU-8 to Si
bonds were subjected to this test as well for comparison. SU-8/Si bond survived the
scotch tape test which is expected due to reported SU-8 adhesion to Si. SU-8/soda lime
bond did not survive the scotch tape test. SU-8/Parylene/soda lime structures did survive
the scotch tape test demonstrating that Parylene plays an integral part in adhering SU-8
photoresist to soda lime substrate.
A pull test was then performed on the bond. A string was attached with epoxy to
the SU-8 structure and the other end was fixed to an Accu-Weigh scale. Force was slowly
increased until bond failure or maximum measurable force (3.4 N). SU-8/Si bond
124
survived the maximum force indicating a bond strength greater than 0.85 N/cm
2
. In
contrast, SU-8/soda lime bond failed at 0.1 N/cm
2
. SU-8/Parylene/soda lime bond
obtained with this fabrication process possessed bond strength greater than 0.85 N/cm
2
.

4.6.2.2 Optical Quality
The SU-8 structures are to be used as optical mirrors for delivering light in
optogenetic applications. Therefore, the mirror quality is critical and determined by the
smoothness of the inclined structure. A rough surface will diffract and scatter light which
is undesirable. A DekTek profilometer was used to measure the surface roughness of the
inclined structure. The SU-8 surface was found to not vary by more than 8 nanometers
which is smooth enough for use as a mirror.

Figure 4-10: Representative surface profile of the resulting SU-8 angled structure. The
variation is less than 8 nm; thus the structure is suitable for use in optical applications.

Next, the reflected light intensity profile from SU-8 mirror structure was imaged.
The mirror array was placed under a microscope and optical fibers placed into the
embedded fiber grooves. Optical fiber was butt-coupled to a high powered halogen light
source (Model 6202, 20W, Electrix Illumination, New Haven, CT). The fiber emitted
125
light reflected off the inclined mirror and toward the microscope camera which captured
the image. The intensity profile was analyzed by ImageJ and a Gaussian distribution was
seen since the light was uncollimated.

Figure 4-11: Light output from an optical fiber butt-coupled to a halogen light source.
Light is seen exiting the core of the fiber. Light is also seen escaping through the
cladding which is indicative of butt-coupling’s natural inefficiency causing optical loss.

Figure 4-12: Light is delivered to the mirror by an optical fiber butt-coupled to a halogen
light source.

Figure 4-13: Representative intensity profile of the reflected light from the SU-8 mirror
structure. The intensity is in a Gaussian distribution which is as expected for non-
collimated light source.
126
4.7 Conclusion
The technique utilized here improves backside inclined lithography of SU-8
structures through use of a Parylene adhesion layer. SU-8 adhesion to the soda lime
substrate is greatly improved and structures with large contact areas with the substrate
can be fabricated without delamination. Consistent inclined angled structures throughout
the entire wafer can be achieved due to the gapless contact between the masked substrate
and the SU-8 photoresist. An optical device for use in neural photostimulation studies
has been fabricated utilizing this technique and the reflected light characterized.
The optical device was designed to be used with a created optogenetic neural cell
line, transfected PC12 cells expressing hChR2 (human Channelrhodopsin-2) ion channels.
PC12 cells were transfected with a lentiviral vector [3] and allowed to grow (gift from Dr.
Pin Wang’s lab, USC, Los Angeles, CA) (Figure 4-14). Cells were then cultured on a
MEA with an attached culture well (Figure 4-15) and allowed to differentiate and grow
axons with the application of NGF (nerve growth factor). The MEA was to be placed
over the optical device and cell electrical recordings would be taken in response to light
stimulation from the reflected light.
127

Figure 4-14: (a) Brightfield image of cultured optogenetic PC12 cells. (b) Transfected
PC12 cells fluoresce due to the expression of a fluorescent protein tag indicating
successful expression of hChR2 channels. 10× magnification.

Figure 4-15: Differentiated PC12 cells cultured on a MEA. Some electrodes are damaged
in the form of metal delamination.

A number of challenges relating to packaging remain to be addressed before this
optical device can be used for photostimulation studies. First, a suitable light source
128
needs to be obtained that generates the relatively high optical power with the correct
wavelength required for optogenetics. The method of butt-coupling a fiber to a lamp or
other light source is inefficient and unsuitable for this application. A better method would
be to utilize a laser with a condenser that focuses the laser output into the optical fiber
core so that minimal optical power loss occurs [21]. Next, a fast shutter or controller is
required to turn the light on and off for temporal control over the photostimulation.
Attaching cells to the optical device requires a cover such as a glass coverslip over the
mirror which presents another medium for light to travel across causing optical power
attenuation. Minimizing the light path length would be optimal; this can be achieved by
utilizing thin-film Parylene which can be several orders of magnitude thinner than glass
and amenable to cell culture. Parylene can be used as a substrate for patterning of
transparent indium tin oxide (ITO) electrodes [22] which enable recording cell
electrophysiological response to photostimulation without impeding the reflected light
path. That being said, a suitable recording system would be required to fully capture the
expected neural signals (action potentials) in response to the transmembrane ion flows
due to hChR-2 photoactivation.
Future generations of the inclined mirror device can explore constructing
embedded optics in addition to the mirrors to focus the light reflecting off the chip. A
modular design may also add versatility. Currently, a 1D mirror array is constructed for
each device. The combination of several optical devices post-fabrication in a modular
fashion can enable various light stimulation patterns. Finally, applications other than
129
photostimulation with the optical device are possible such as for cell imaging purposes
when side view microscopy is desired [23-25].
130
REFERENCES
[1] A. M. Aravanis, L.-P. Wang, F. Zhang, L. A. Meltzer, M. Z. Mogri, M. B.
Schneider, et al., "An optical neural interface: in vivo control of rodent motor
cortex with integrated fiberoptic and optogenetic technology," Journal of Neural
Engineering, vol. 4, pp. S143-S156, 2007.
[2] J. Zhang, F. Laiwalla, J. A. Kim, H. Urabe, R. Van Wagenen, Y.-K. Song, et al.,
"Integrated device for optical stimulation and spatiotemporal electrical recording
of neural activity in light-sensitized brain tissue," Journal of Neural Engineering,
vol. 6, 2009.
[3] E. S. Boyden, F. Zhang, E. Bamberg, G. Nagel, and K. Deisseroth, "Millisecond-
timescale, genetically targeted optical control of neural activity," Nature
Neuroscience, vol. 8, pp. 1263-1268, Sep 2005.
[4] F. Zhang, L. P. Wang, M. Brauner, J. F. Liewald, K. Kay, N. Watzke, et al.,
"Multimodal fast optical interrogation of neural circuitry," Nature, vol. 446, pp.
633-9, Apr 5 2007.
[5] T. Ishizuka, M. Kakuda, R. Araki, and H. Yawo, "Kinetic evaluation of
photosensitivity in genetically engineered neurons expressing green algae light-
gated channels," Neuroscience research, vol. 54, pp. 85-94, 2006.
[6] K. P. Greenberg, A. Pham, and F. S. Werblin, "Differential targeting of optical
neuromodulators to ganglion cell soma and dendrites allows dynamic control of
center-surround antagonism," Neuron, vol. 69, pp. 713-720, 2011.
[7] L. Petreanu, D. Huber, A. Sobczyk, and K. Svoboda, "Channelrhodopsin-2–
assisted circuit mapping of long-range callosal projections," Nature Neuroscience,
vol. 10, pp. 663-668, 2007.
[8] H. Wang, J. Peca, M. Matsuzaki, K. Matsuzaki, J. Noguchi, L. Qiu, et al., "High-
speed mapping of synaptic connectivity using photostimulation in
Channelrhodopsin-2 transgenic mice," Proceedings of the National Academy of
Sciences, vol. 104, pp. 8143-8148, 2007.
131
[9] S. J. Cruikshank, H. Urabe, A. V. Nurmikko, and B. W. Connors, "Pathway-
Specific Feedforward Circuits between Thalamus and Neocortex Revealed by
Selective Optical Stimulation of Axons," Neuron, vol. 65, pp. 230-245, 2010.
[10] V. Jokinen, P. Sakha, P. Suvanto, C. Rivera, S. Franssila, S. E. Lauri, et al., "A
microfluidic chip for axonal isolation and electrophysiological measurements,"
Journal of neuroscience methods, 2012.
[11] S. Tsuda, M. Z. Kee, C. Cunha, J. Kim, P. Yan, L. M. Loew, et al., "Probing the
function of neuronal populations: combining micromirror-based optogenetic
photostimulation with voltage-sensitive dye imaging," Neuroscience research,
2012.
[12] C. Eberius and D. Schild, "Local photolysis using tapered quartz fibres," Pflugers
Archiv-European Journal of Physiology, vol. 443, pp. 323-330, Nov 2001.
[13] L. Campagnola, H. Wang, and M. J. Zylka, "Fiber-coupled light-emitting diode
for localized photostimulation of neurons expressing channelrhodopsin-2," J
Neurosci Methods, vol. 169, pp. 27-33, Mar 30 2008.
[14] D. W. Godwin, D. P. Che, D. M. OMalley, and Q. Zhou, "Photostimulation with
caged neurotransmitters using fiber optic lightguides," Journal of Neuroscience
Methods, vol. 73, pp. 91-106, Apr 25 1997.
[15] D. Ghezzi, A. Menegon, A. Pedrocchi, F. Valtorta, and G. Ferrigno, "A Micro-
Electrode Array device coupled to a laser-based system for the local stimulation
of neurons by optical release of glutamate," Journal of Neuroscience Methods, vol.
175, pp. 70-78, Oct 30 2008.
[16] C. Boucsein, M. Nawrot, S. Rotter, A. Aertsen, and D. Heck, "Controlling
synaptic input patterns in vitro by dynamic photo stimulation," J Neurophysiol,
vol. 94, pp. 2948-58, Oct 2005.
[17] M. C. Peterman and et al., "Building thick photoresist structures from the bottom
up," Journal of Micromechanics and Microengineering, vol. 13, p. 380, 2003.
132
[18] K. Y. Hung, H. T. Hu, and F. G. Tseng, "Application of 3D glycerol-compensated
inclined-exposure technology to an integrated optical pick-up head," Journal of
Micromechanics and Microengineering, vol. 14, pp. 975-983, Jul 2004.
[19] Z. Ling and K. Lian, "SU-8 3D microoptic components fabricated by inclined UV
lithography in water," Microsystem technologies, vol. 13, pp. 245-251, 2007.
[20] R. Lo and E. Meng, "Integrated and reusable in-plane microfluidic interconnects,"
Sensors and Actuators B: Chemical, vol. 132, pp. 531-539, 2008.
[21] M. S. Y. Lung, P. Pilowky, and E. M. Goldys, "Activation of the Mammalian
Cells by Using Light-Sensitive Ion Channels," in Spectroscopic Methods of
Analysis, W. M. Bujalowski, Ed., ed: Humana Press, 2012, pp. 241-251.
[22] P. Ledochowitsch, E. Olivero, T. Blanche, and M. M. Maharbiz, "A transparent
μECoG array for simultaneous recording and optogenetic stimulation," presented
at the 33rd Annual International Conference of the IEEE Engineering in Medicine
& Biology Society, Boston, Masachusetts, USA, 2011.
[23] C. Boocock, A. Brown, and G. Dunn, "A simple chamber for observing
microscopic specimens in both top and side views," Journal of microscopy, vol.
137, pp. 29-34, 1985.
[24] J. Cao, S. Usami, and C. Dong, "Development of a side-view chamber for
studying cell-surface adhesion under flow conditions," Annals of biomedical
engineering, vol. 25, pp. 573-580, 1997.
[25] K. Seale, R. Reiserer, D. Markov, I. Ges, C. Wright, C. Janetopoulos, et al.,
"Mirrored pyramidal wells for simultaneous multiple vantage point microscopy,"
Journal of microscopy, vol. 232, pp. 1-6, 2008.
133

Micromachining technologies are uniquely suited for constructing neural
interfaces. Neural networks are complex and are made up of neurons which have
dimensions in the micron range. The fabrication technologies available for
micromachining provide a perfect solution for matching the micron dimensions and
geometrical complexities that are demanded of advanced neural interfaces as well as high
throughput, reliability and repeatability; much more so than is possible with traditional
neural interfacing technologies such as electrical wires, pipettes, or lasers.
In this work, the Parylene sheath probe concept for increasing the long term
reliability of electrophysiological recordings from the motor cortex was first introduced
for use in improved BMIs. Parylene patterning through plasma etching was exploited to
produce desired probe shapes possessing multiple Pt electrodes patterned throughout the
sheath for neural recordings. A 3D structure was obtained through thermoforming the
Parylene device around a microwire mold. The purpose of this 3D structure was to
encourage neural ingrowth and therefore neural integration of the probe post-implantation.
Multiple improving sheath probe generations were designed and fabricated and minimum
achievable feature dimensions were determined. Pt electrodes were electrochemically
characterized and found to possess desirable neural recording properties. Probes were
then implanted into rats and in vivo neural signals were successfully recorded. While
Chapter 5 CONCLUSION
134
initial results are promising, this sheath probe technology still needs to be validated
through histology and comparison with other neural probe types with regard to neural
ingrowth and how the unique sheath probe properties (3D structure, Parylene material,
shape, etc.) influence such neural integration. A full understanding of these properties
and their effects on promoting neural integration will lead to a true long term neural
probe for lifetime electrophysiological recordings.
Next, micromachined chemical interfaces were described. A microfluidic
platform possessing an array of Parylene microchannels with flow sensors was
constructed and used for chemical delivery to neurons through an etched pore atop each
microchannel. Modular devices were then created to address the challenges experienced
by the microfluidic platform. CCM was constructed to enable in vitro neural cell
patterning with isolation of neural axons and soma for subcellular chemical targeting.
Chemical delivery was to be achieved with BAM, a device possessing mechanically
rigid SU-8 microchannels capped with Parylene and operated with passive pumping. A
novel fabrication technique for constructing BAM was developed: SU-8/Parylene
thermocompressive bonding, which can enable construction of many other microdevices
such as multilayered 3D microfluidic devices. Simple to use robust modules for
delivering chemicals to neurons with little to no infrastructure requirements lend this
technology to wide adoptability. Before that can happen, a complete characterization of
the achievable chemical flow and gradient patterns is necessary. BAM dimensions can
be defined to produce optimal chemical delivery with knowledge of how channel and
135
pore dimensions impact passive pumping operation, enabling experimental studies that
are not possible with existing methodologies.
Finally, a device was micromachined for optically interfacing with neurons.
Inclined mirrors were constructed to direct light to cultured neurons for optical
stimulation. These mirrors were fabricated through inclined exposure which produced
SU-8 structures at a 45 º angle. A fabrication process was developed that built upon
previous inclined exposure techniques to achieve improved yields by utilizing the
substrate as a mask and exposing within a glycerol medium. These inclined structures
were then coated with metal to create a reflective surface. Light delivered through optical
fibers into the device was directed by the mirrors upwards toward cultured cells or tissue.
Thus, an optical device compatible with standard microscopy and possessing multiple
light stimulation sites was achieved for neural interfacing. Currently, reflected light is
diffracted thereby diffusing the optical power and increasing the light diameter which can
be undesirable for precision optical stimulation. Embedded lenses for light focusing can
be constructed to enhance the optical device. Inclined mirrors can also be utilized for
other purposes such as cell motility studies when imaging angles other than a
perpendicular one are desired. Inclined exposure enables different angles other than 45 º
which can be exploited to construct different mirror types. Different sizes are also
possible with this fabrication technology. The addition of a MEA constructed of
transparent indium tin oxide (ITO) electrodes can enable multimodal neural interfacing
through electrophysiological recording while optical stimulation takes place
simultaneously.
136
Different micromachined neural interfaces have been developed in three
modalities for electrical, chemical, and optical interrogations of neural networks. Novel
fabrication methods were developed and utilized in the construction of these
microdevices which improve the state of the art in neural interfacing technologies.
Hopefully, use of these developed neural interfacing tools will advance neuroscience
knowledge and thus pave the way for improved clinical interventions for neurological
disorders or prosthetics use.

137
APPENDIX A: SILANE A-174 PARYLENE ADHESION TREATMENT

CAUTION: To be performed under the solvents Fume Hood only.

Mixing steps:
1. Always check the shelf-life tag on the A-174 bottle. If > 6 months, solution needs to be
replaced.

2. Make 0.5% of A-174 in IPA and DI water.

3. Mixing recipe:
a. IPA: DI H20: A-174 = 500:500:5 ml. Stir for 30 seconds. Make sure it is mixed well
b. Let it stand for 2 hours prior to use.
c. Life time of mixed solution is 24 hours.

4. After 2 hours, submerge the cleaned wafer(s) in the solution for 15-30 min.

5. Air dry for 15-30 min or carefully blow dry or spin dry in spin dryer (wafer must be dried
before proceeding to next step).
a. Do not let surfaces to be coated to come into contact with dirty surfaces (prop up
wafers at an angle).

6. Clean with IPA for 15-30 sec. Agitate gently up and down.

7. Air dry for 30 – 60 seconds or spin dry in spin dryer.

8. Bake dry to remove moisture in an oven for 15 min @ 100°C (wafer must be dried before
Parylene coating).

138
APPENDIX B: PARYLENE SHEATH PROBE VERSION 1 FAB RECIPE

The first version only had inner electrodes (4) patterned.

Released sheath probe with only inner electrodes patterned.

Obtain prime 3" Si wafer out of box
Deposit Parylene (4-5 µm)
AZ 5214 E-IR Mask 1b
5 sec @ 500 rpm; 45 sec @ 2krpm (~2 µm)
Bake 1:10 min @ 90 °C
Expose 50 mJ = 2.5 mW/cm
2
* 20 sec
IR bake 45 sec @ 120 °C
Turn aligner to Soft Contact mode
Global exposure 300 mJ = 2.5 mW * 120 sec
Develop: 20-22 sec
Descum 100 W:100 mT: 1 min
Ebeam Pt (3 runs of 666 Å for total of ~2000 Å)
Liftoff
Deposit Parylene (1 µm)
Etch mask for insulation layer
AZ 4400 Mask 2
5 sec @ 500 rpm; 45 sec @ 4 krpm (~4 µm)
Bake 2 min @ 90 °C
Mask 2
Expose 150 mJ = 3 mW/cm
2
* 50 sec
Develop ~ 45-50 sec
RIE insulation layer etch 100 W: 100 mT: 5 min
Clean off mask with standard clean
Sacrificial PR for cone
AZ4620 Mask 3
5 sec @ 500 rpm; 45 sec @ 3.5 krpm (~8 µm)
Bake 5 min @ 90 °C
Expose 400 mJ = 4 mW/cm
2
* 100 sec
Develop 50-60 sec
Hardbake - not done originally but to prevent bubbles in PR later on
139
Bake 3 min @ 90 C
Descum 100 W : 100 mT : 1 min
Deposit Parylene (5 m)
AZ 4620 cutout mask Mask 6b
5 sec @ 500 rpm; 45 sec @ 3.5 krpm (~8 µm)
Bake 5 min @ 90 °C
Mask 6b Channel Release, Partial etch of cutout
Expose 400 mJ = 4 mW/cm
2
* 100 sec
Develop 50-60 sec
RIE 100W: 100mT: 5 min
4 times rotating wafers each time
AZ 4620 cutout mask Mask 7
5 sec @ 500 rpm; 45 sec @ 3.5 krpm (~8 µm)
Bake 5 min @ 90 °C
Mask 7 Device Release
Expose 400 mJ = 4 mW/cm
2
* 100 sec
Develop 50-60 sec
RIE 100 W: 100 mT: 5 min
4 times rotating wafers each time

140
APPENDIX C: PARYLENE SHEATH PROBE VERSION 1 MASK SET

141

142

143
APPENDIX D: PARYLENE SHEATH PROBE VERSION 2 FAB
RECIPE

The second version had eight Pt electrodes, four inside sheath and four on sheath surface
and took ~2 weeks to process.

Eight-electrode sheath probes; four on sheath interior and four on the exterior. Three
different sheath designs are presented with their dimensions.

Obtain prime 3" Si wafer out of box
Deposit Parylene (4-5 µm)
AZ 5214 E-IR Mask 1b
5 sec @ 500 rpm; 45 sec @ 2krpm (~2 µm)
Bake 1:10 min @ 90 °C
Expose 50 mJ = 2.5 mW/cm
2
* 20 sec
IR bake 45 sec @ 120 °C
Turn aligner to Soft Contact mode
global exposure 300 mJ = 2.5 mW/cm
2
* 120 sec
Develop: 20-22 sec
Descum 100W:100mT:1 min
Ebeam Pt (3 runs of 666Å for total of ~2000Å)
Liftoff
Deposit Parylene (1 µm)
Etch mask for insulation layer
144
AZ 4400 Mask 2
5 sec @ 500 rpm; 45 sec @ 4 krpm (~4 µm)
Bake 2 min @ 90 °C
Expose 150 mJ = 3 mW/cm
2
* 50 sec
Develop ~ 45-50 sec
RIE insulation layer etch 100 W:100 mT:5 min
Clean off mask with standard clean
Sacrificial PR for cone
AZ4620 Mask 3
5 sec @ 500 rpm; 45 sec @ 3.5 krpm (~8 µm)
Bake 5 min @ 90 °C
Expose 400 mJ = 4 mW/cm
2
* 100 sec
Develop 50-60 sec
Hardbake - not done originally but to prevent bubbles in PR later on
Bake 3 min @ 90 °C
Descum 100W:100mT:1 min
Deposit Parylene (5 µm)
Dual layer liftoff for top electrodes
AZ 1518
5 sec @ 1 krpm; 45 sec @ 4 krpm (~2 µm)
Bake 45 sec @ 90 °C
Turn aligner to Soft Contact mode
Global Exposure 90 mJ = 3 mW/cm
2
* 30 sec
AZ 4620 Mask 4
5 sec @ 500 rpm; 45 sec @ 2 krpm (~9 µm)
Bake 5 min @ 90 °C
Expose 200 mJ = 3 mW/cm
2
* 66.6 sec
Develop 52 sec with high agitation ****very sensitive;
Descum
Ebeam Pt (3 runs of 666 Å for total of ~2000 Å)
Liftoff
Deposit Parylene (1 µm)
AZ 4400 Mask 5b
5 sec @ 500 rpm; 45 sec @ 4 krpm (~4 µm)
Bake 2 min @ 90 °C
Expose 150 mJ = 3 mW/cm
2
* 50 sec
Develop ~ 45-50 sec
RIE etch (~0.3 µm/min Parylene etch)
100 W : 100 mT : 3.5 min
Turn wafers 180 degrees
100 W : 100 mT : 3.5 min
Clean off mask with standard clean
AZ 4620 cutout mask Mask 6b
5 sec @ 500 rpm; 45 sec @ 3.5 krpm for (~8 µm)
Bake 5 min @ 90 °C
Expose 400 mJ = 4 mW/cm
2
* 100 s
Develop 50-60 sec
RIE 100W: 100mT: 5 min
4 times rotating wafers each time
AZ 4620 cutout mask Mask 7
5 sec @ 500 rpm; 45 sec @ 3.5 krpm for (~8 µm)
Bake 5 min @ 90 °C
Expose 400 mJ = 4 mW/cm
2
* 100 s
Develop 50-60 sec
RIE 100W: 100mT: 5 min
4 times rotating wafers each time

145
APPENDIX E: PARYLENE SHEATH PROBE VERSION 2 MASK SET

146

147

148

149
APPENDIX F: PARYLENE SHEATH PROBE VERSION 3 FAB RECIPE

The four Pt electrodes on sheath surface were prone to cracking during the
thermoforming process and so were moved to positions adjacent to the sheath. These
became the so-called wing electrodes and were used in the initial implantation studies
(acute):

Three wing electrode sheath probe designs shown as-fabricated with photoresist
sacrificial layer: a) 300 m to 50 m taper, b) 450 m to 50 m taper, and c) 300 m
diameter cylinder

With electrodes located on wings, all metal could be processed in one step with a
reduction in Parylene coating and plasma etching thus total processing time was
significantly reduced (~1 week).

Obtain prime 3" Si wafer out of box
Deposit Parylene (4-5 µm)
AZ 5214 E-IR Mask 1b v3
5 sec @ 500 rpm; 45 sec @ 2krpm (~2 µm)
Bake 1:10 min @ 90 °C
Expose 50 mJ = 2.5 mW/cm
2
* 20 s
IR bake 45 sec @ 120 °C
Turn aligner to Soft Contact mode
Global exposure 300 mJ = 2.5 mW /cm
2
* 120 s
Develop: 20-22 sec
Descum 100 W : 100 mT : 30 sec
Ebeam Pt (3 runs of 666 Å for total of ~2000 Å)
Liftoff
Deposit Parylene (1 µm)
Etch mask for insulation layer
150
AZ 4400 Mask 2 v3
5 sec @ 500 rpm; 45 sec @ 4 krpm (~4 µm)
Bake 2 min @ 90 °C
Expose 150 mJ = 3 mW/cm
2
* 50 sec
Develop ~ 45-50 sec
RIE insulation layer etch 100 W : 100 mT : 5 min
Clean off mask with standard clean
Sacrificial PR for cone
AZ4620 Mask 3
5 sec @ 500 rpm; 45 sec @ 2 krpm (~9.6 µm)
Bake 5 min @ 90 °C
Expose 400 mJ = 4 mW/cm
2
* 100 s
Develop 50-60 sec
Pseudo-hardbake: Bake 3 min @ 90 °C
Descum 100 W : 100 mT : 1 min
Deposit Parylene (5 µm)
AZ 4620 cutout mask Mask 5b-6b v3
5 sec @ 500 rpm; 45 sec @ 2 krpm for (~9.6 µm)
Bake 5 min @ 90 °C
Expose 400 mJ = 4 mW/cm
2
* 100 sec
Develop 50-60 sec
RIE 100W : 100mT : 11.5 min
2 times rotating wafers each time
DO NOT CLEAN OFF MASK
AZ 4620 cutout mask Mask 7
5 sec @ 500 rpm; 45 sec @ 2 krpm for (~9.6 µm)
Bake 5 min @ 90 °C
Expose 400 mJ = 4 mW/cm
2
* 100 sec
Develop 50-60 sec
RIE 100 W : 100 mT : 11.5 min
2 times rotating wafers each time

151
APPENDIX G: PARYLENE SHEATH PROBE VERSION 3 MASK SET

152

153

154
APPENDIX H: MINIMALLY SIZED PROBES FAB RECIPE

Obtain prime 4" Si wafer out of box
Deposit Parylene (5 µm)
AZ 5214 E-IR Mask 1
5 sec @ 500 rpm; 45 sec @ 2 krpm (~2 µm)
Bake 1:10 min @ 90 °C
Expose 50 mJ = 10 mW/cm
2
* 5 sec
IR bake 45 sec @ 120 °C
Turn aligner to Soft Contact mode
Global exposure 300 mJ = 10 mW /cm
2
* 30 sec
Develop: 20-22 sec
Descum 100 W : 100 mT : 30 sec
Ebeam Pt (3 runs of 666 Å for total of ~2000 Å)
Liftoff
Etch mask for base perforation
AZ 4620 Mask 2
5 sec @ 500 rpm; 45 sec @ 2 krpm (~9.6 µm)
Bake 5 min @ 90 °C
Expose 400 mJ = 10 mW/cm
2
* 40 sec
Develop 50-60 sec
RIE 100W : 100 mT : 11.5 min
2 times rotating wafers each time
Deposit Parylene (1 µm)
Etch mask for insulation layer
AZ 4400 Mask 3
5 sec @ 500 rpm; 45 sec @ 4 krpm (~4 µm)
Bake 2 min @ 90 °C
Expose 150 mJ = 10 mW/cm
2
* 15 sec
Develop ~ 45-50 sec
RIE insulation layer etch 100 W : 100 mT : 5 min
Clean off mask with standard clean
Sacrificial PR for cone
AZ4620 Mask 4
5 sec @ 500 rpm; 45 sec @ 2 krpm (~9.6 µm)
Bake 5 min @ 90 °C
Expose 400 mJ = 10 mW/cm
2
* 40 sec
Develop 50-60 sec
Pseudo-hardbake: Bake 3 min @ 90 °C
Descum 100 W : 100 mT : 1 min
Deposit Parylene (5 µm)
AZ 4620 cutout Mask 5
5 sec @ 500 rpm; 45 sec @ 2 krpm for (~9.6 µm)
Bake 5 min @ 90 °C
Expose 400 mJ = 10 mW/cm
2
* 40 sec
Develop 50-60 sec
RIE 100W : 100 mT : 11.5 min
2 times rotating wafers each time
DO NOT CLEAN OFF MASK
AZ 4620 cutout Mask 6
5 sec @ 500 rpm; 45 sec @ 2 krpm for (~9.6 µm)
Bake 5 min @ 90 °C
Expose 400 mJ = 10 mW/cm
2
* 40 sec
Develop 50-60 sec
RIE 100 W : 100 mT : 11.5 min
2 times rotating wafers each time

155
APPENDIX I: MINIMALLY SIZED PROBES MASK SET

156

157

158

159

160

161
APPENDIX J: ARRAYED PARYLENE SHEATH PROBES WITH
PERFORATIONS FAB RECIPE

Obtain prime 4" Si wafer out of box
Deposit Parylene (5 µm)
AZ 5214 E-IR Mask 1
5 sec @ 500 rpm; 45 sec @ 2 krpm (~2 µm)
Bake 1:10 min @ 90 °C
Expose 50 mJ = 10 mW/cm
2
* 5 sec
IR bake 45 sec @ 120 °C
Global exposure 300 mJ = 10 mW /cm
2
* 30 sec
Develop: 20-22 sec
Descum 100 W : 100 mT : 30 sec
Ebeam Pt (3 runs of 666 Å for total of ~2000 Å)
Liftoff
Etch mask for base perforation
AZ4620 Mask 2
5 sec @ 500 rpm; 45 sec @ 2 krpm (~9.6 µm)
Bake 5 min @ 90 °C
Expose 400 mJ = 10 mW/cm
2
* 40 sec
Develop 50-60 sec
RIE of base 100 W : 100 mT : 5 min (5 times rotating each time)
Wash off PR mask with triple IPA bath with cotton swabbing
Descum 100 W : 100 mT : 1 min
Deposit Parylene (2 µm)
Etch mask for insulation layer
AZ 4620 Mask 3
5 sec @ 500 rpm; 45 sec @ 2 krpm (~9.6 µm)
Bake 5 min @ 90 °C
Expose 400 mJ = 10 mW/cm
2
* 40 sec
Develop ~ 50-60 sec
RIE insulation layer etch 100 W : 100 mT : 5 min (3 times rotating each time)
Wash off PR mask with triple acetone & IPA bath with cotton swabbing
Sacrificial PR for cone
AZ4620 Mask 4
5 sec @ 500 rpm; 45 sec @ 2 krpm (~9.6 µm)
Bake 5 min @ 90 °C
Expose 400 mJ = 10 mW/cm
2
* 40 sec
Develop 50-60 sec
Pseudo-hardbake: Bake 3 min @ 90 °C
Descum 100 W : 100 mT : 1 min
Deposit Parylene (5 µm)
AZ 4620 channel release/top perforation/partial cutout Mask 5
5 sec @ 500 rpm; 45 sec @ 2 krpm for (~9.6 µm)
Bake 5 min @ 90 °C
Expose 400 mJ = 10 mW/cm
2
* 40 sec
Develop 50-60 sec
RIE 100W : 100 mT : 5 min (5 times rotating each time)
DO NOT CLEAN OFF MASK
AZ 4620 cutout Mask 6 (double spun PR)
5 sec @ 500 rpm; 45 sec @ 2 krpm for (~9.6 µm)
Bake 5 min @ 90 °C
5 sec @ 500 rpm; 45 sec @ 2 krpm
Bake 6 min @ 90 °C
Expose 540 mJ = 10 mW/cm
2
* 54 sec
162
Develop 50-60 sec
RIE 100 W : 100 mT : 9 min (4 times rotating each time)

163
APPENDIX K: ARRAYED PARYLENE SHEATH PROBES WITH
PERFORATIONS MASK SET

164

165

166

167

168

169
APPENDIX L: INCLINED SU-8 MIRRORS FAB RECIPE

Obtain 3” soda lime wafer
AZ 4400 Mask
5 sec @ 500 rpm; 45 sec @ 4 krpm (~4 µm)
Bake 2 min @ 90 °C
Expose 150 mJ = 10 mW/cm2 * 15 sec
Develop ~ 45-50 sec
Descum 100 W : 100 mT : 30 sec
Ebeam Ti (1 run of 500 Å)
Liftoff
A-174
Deposit Parylene (2-10 µm)
Spin on SU-8 2035
Cast off 1st drop of SU-8 from applicator
Apply a large amount in a circular motion
Avoid bubbles
15 sec @ 600 rpm for ~200 µm of thickness
10 minute wait for reflow
Use thin pipette tip to poke out bubbles if any
Use plastic (transparency) strips to scrape off edge bead
Softbake of SU-8
Ramp 5°C/3min from room temperature
Bake 2 hrs @ 95°C
Turn off hotplate and wait 2 hours for controlled cooldown to room temperature
Prepare inclined hoder (bookend)
Bend to desired exposure angle of PR based upon Snell’s Law
Place wafer PR side down on inclined holder
Attach with magic tape
Immerse into glycerol bath
Wait 10 minutes for bath to clear
Expose 20 mJ/µm thickness of SU-8 = 20 mJ/µm * 200 µm = 4000 mJ
Remove inclined holder from bath
Let excess glycerol drip back into bath
Remove wafer
PEB
Ramp 5°C/3min from room temperature
Bake 2 hrs @ 95°C
Turn off hotplate and wait 2 hours for controlled cooldown to room temperature
Develop
Rinse with IPA
Rinse with DI water
Blow dry

170
APPENDIX M: INCLINED SU-8 MIRRORS MASK

171
APPENDIX N: µCCM SU-8 MOLD FAB RECIPE

Obtain 3” Si wafer
AZ 4400 Mask 1
5 sec @ 500 rpm; 45 sec @ 4 krpm (~4 µm)
Bake 2 min @ 90 °C
Expose 150 mJ = 10 mW/cm
2
* 15 sec
Develop ~ 45-50 sec
Descum 100 W : 100 mT : 30 sec
Ebeam Ti (1 run of 200 Å) – for alignment marks to be seen in latter SU-8 processing
Liftoff
A-174
Deposit Parylene (2 µm)
Base mold for 3 µm high channels
SU-8 2 Mask 1
5 sec @ 500 rpm; 45 sec @ 1425 rpm (~3 µm)
Bake 1 min @ 65 °C
Bake 3 min @ 95 °C
Expose 100 mJ = 10 mW/cm
2
* 10 sec
PEB 1 min @ 65 °C; 1 min @ 95 °C
Develop 2 min
Second layer of SU-8 for cell compartments
SU-8 2035 Mask 2
15 sec @ 925 rpm (~140 µm) or 15 sec @ 2 krpm (~100 µm)
Relax 10 min
Softbake programmed ramp up to 95 °C for 2 hrs and controlled ramp down (turn off hotplate)
Expose 500 mJ = 10 mW/cm
2
* 50 sec
PEB programmed ramp up to 95 °C for 2 hrs and controlled ramp down (turn off hotplate)
Develop
Optional hardbake
172
APPENDIX O: µCCM SU-8 MOLD MASK SET

173
APPENDIX P: µBAM FAB RECIPE

Obtain 3” Si wafer
Deposit Parylene (2-5 µm)

Optional AZ 5214 E-IR Mask 1
5 sec @ 500 rpm; 45 sec @ 2 krpm (~2 µm)
Bake 1:10 min @ 90 °C
Expose 50 mJ = 10 mW/cm
2
* 5 sec
IR bake 45 sec @ 120 °C
Global exposure 300 mJ = 10 mW /cm
2
* 30 sec
Develop: 20-22 sec
Descum 100 W : 100 mT : 30 sec
Ebeam Ti (1 run of 200 Å) – for alignment marks to be seen in latter SU-8 processing
Liftoff

SU-8 Base
SU-8 2035 Mask 2
45 sec @ 4 krpm (~30 µm)
Relax 10 min
Softbake programmed ramp up to 95 °C for 2 hrs and controlled ramp down (turn off hotplate)
Expose 500 mJ = 10 mW/cm
2
* 50 sec
PEB programmed ramp up to 95 °C for 2 hrs and controlled ramp down (turn off hotplate)
Develop
2
nd
SU-8
SU-8 2035 Mask 3
45 sec @ 2 krpm (~55 µm)
Relax 10 min
Softbake programmed ramp up to 95 °C for 2 hrs and controlled ramp down (turn off hotplate)
Expose 500 mJ = 10 mW/cm
2
* 50 sec
PEB programmed ramp up to 95 °C for 2 hrs and controlled ramp down (turn off hotplate)
Develop
Parylene base etch
AZ 4620 Mask 4
5 sec @ 500 rpm; 45 sec @ 2 krpm (~9.6 µm)
Bake 5 min @ 90 °C
Expose 400 mJ = 10 mW/cm
2
* 40 sec
Develop 50-60 sec
RIE 100W : 100 mT : 11.5 min
2 times rotating wafers each time
Parylene wafer bonding

Parylene pore etch
AZ 4620 Mask 5
5 sec @ 500 rpm; 45 sec @ 2 krpm (~9.6 µm)
Bake 5 min @ 90 °C
Expose 400 mJ = 10 mW/cm
2
* 40 sec
Develop 50-60 sec
RIE 100W : 100 mT : 11.5 min
2 times rotating wafers each time

174
APPENDIX Q: µBAM MASK SET

175

176

Abstract (if available)

Abstract BioMEMS is uniquely positioned to impact the field of neuroscience by advancing neural interface state of the art. Traditional neural interfaces rely on macro-world technologies that can be limited when attempting to interface with neurons on a cellular level. Micromachining enables construction of microengineered devices possessing form factors that are suitable for interfacing with neurons or complex neural networks. Improved tools for neural interfacing can improve neuroscience knowledge by enabling more sophisticated experimental studies or advancing the use of neuroprosthetics. ❧ In this work, three micromachined neural interfaces are developed and described

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

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

PDF

Development of implantable Parylene-based MEMS technologies for cortical applications

PDF

Parylene C bioMEMS for implantable devices with electrochemical interfaces

PDF

The electrochemical evaluation of Parylene-based electrodes for neural applications

PDF

Development of fabrication technologies for robust Parylene medical implants

PDF

Fabrication and packaging of three-dimensional Parylene C neural interfaces

PDF

Parylene-based implantable interfaces for biomedical applications

PDF

Parylene-based biomems sensors for multiple physiological systems

PDF

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

PDF

High-resolution data acquisition with neural and dermal interfaces

PDF

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

PDF

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

PDF

Bidirectional neural interfaces for neuroprosthetics

PDF

Next generation neural interfaces for neural modulation and prosthesis

PDF

Implantable bioMEMS drug delivery systems

PDF

Integrated wireless piezoelectric ultrasonic transducer system for biomedical applications

PDF

Detection and decoding of cognitive states from neural activity to enable a performance-improving brain-computer interface

PDF

Novel beamforming techniques for robust contrast enhancement in ultrasound imaging

PDF

Piezoelectric ultrasonic and acoustic microelectromechanical systems (MEMS) for biomedical, manipulation, and actuation applications

PDF

Investigation of the electrode-tissue interface of retinal prostheses

Asset Metadata

Creator Kuo, Jonathan T. W. (author)

Core Title Development of micromachined technologies for neural interfaces

School Viterbi School of Engineering

Degree Doctor of Philosophy

Degree Program Biomedical Engineering

Publication Date 10/29/2013

Defense Date 10/14/2013

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

Tag bioMEMS,microfabrication,micromachining,neural interfaces,OAI-PMH Harvest,Parylene C,SU-8

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Meng, Ellis (committee chair), Gupta, Malancha (committee member), Zhou, Qifa (committee member)

Creator Email jonathan.k3@gmail.com,jonathtk@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c3-340448

Unique identifier UC11296464

Identifier etd-KuoJonatha-2117.pdf (filename),usctheses-c3-340448 (legacy record id)

Legacy Identifier etd-KuoJonatha-2117.pdf

Dmrecord 340448

Document Type Dissertation

Format application/pdf (imt)

Rights Kuo, Jonathan T. W.

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