University of Southern California Dissertations and Theses

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

(USC Thesis Other)

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

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content I

PENETRATING PARYLENE NEURAL PROBE ARRAY FOR DENSE, IN VIVO, CHRONIC
RECORDINGS

Dissertation by

Ahuva Weltman Hirschberg

In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(BIOMEDICAL ENGINEERING)

University of Southern California

Los Angeles, California

2018

PRESENTED TO THE FACULTY OF THE USC GRADUATE SCHOOL
VITERBI SCHOOL OF ENGINEERING
AUGUST 2018

II

© 2018

Ahuva Weltman Hirschberg

All Rights Reserved

III

" ו םיקלא ארבי םדאה תא םיקלא םלצב ,ומלצב ותא ארב " )זכ:א תישארב(
“And G-d created man in His image, in the image of G-d He created him”
(Genesis 1:27)

IV

TABLE OF CONTENTS
TABLE OF CONTENTS .................................................................................................... IV
TABLE OF FIGURES ........................................................................................................ VI
INTRODUCTION: PENETRATING NEURAL PROBES FOR
HIGH-RESOLUTION, CHRONIC, ELECTRICAL RECORDINGS .................... 1
ELECTRICAL COMMUNICATION IN THE BRAIN ................................................... 1
EXTRACELLULAR ELECTROPHYSIOLOGICAL RECORDINGS ................................ 1
PENETRATING NEURAL PROBES FOR NEURAL RECORDINGS ............................... 5
LIFETIME LIMITATIONS OF RIGID NEURAL PROBES ............................................ 6
USING FLEXIBLE PROBES TO ATTENUATE IMMUNE RESPONSE AND EXTEND
RECORDING LIFETIMES ..................................................................................................... 8
OBJECTIVES .................................................................................................... 10
REFERENCES ................................................................................................... 12
MECHANICAL CHALLENGES IN THE CREATION OF FLEXIBLE,
HIPPOCAMPAL RECORDING ARRAY ....................................................... 16
FLEXIBLE ARRAY IMPLANTATION ................................................................... 17
DISSOLVABLE BRACE FOR MINIMALLY INVASIVE IMPLANTATION ................... 27
FABRICATION OF FUNCTIONAL ARRAYS .......................................................... 37
STRESS INDUCED ARRAY CURVATURE ............................................................ 45
SUMMARY ...................................................................................................... 50
REFERENCES ................................................................................................... 51
DESIGN, PACKAGING, AND ELECTRICAL TESTING OF PARYLENE-BASED
HIPPOCAMPAL RECORDING ARRAY ....................................................... 54
TARGETING THE HIPPOCAMPUS ...................................................................... 54
SPECIFIC AIMS OF THE HIPPOCAMPAL ARRAY ................................................. 55
ARRAY DESIGN AND SYSTEM OVERVIEW ........................................................ 56
ELECTROCHEMICAL CHARACTERIZATION OF HIPPOCAMPAL ARRAY ............... 64
SUMMARY ...................................................................................................... 69
REFERENCES ................................................................................................... 70
IN VIVO PERFORMANCE OF HIPPOCAMPAL RECORDING ARRAY ........... 71
METHODS OF ASSESSING PROBE PERFORMANCE OVER TIME ........................... 71
IN VIVO IMPLANTATIONS ................................................................................ 73
ACUTE RECORDINGS ....................................................................................... 77
CHRONIC RECORDINGS ................................................................................... 80
HISTOLOGY .................................................................................................... 83
DISCUSSION OF ARRAY PERFORMANCE AND IMMUNE RESPONSE ..................... 86
FUTURE DIRECTIONS ....................................................................................... 90
SUMMARY ...................................................................................................... 91
REFERENCES ................................................................................................... 91
V

LIFETIME TESTING ................................................................................ 94
IMPROVING ENCAPSULATION AND ATTENUATING CROSSTALK ........................ 94
BENCHTOP ANALYSIS OF ARRAY INTEGRITY OVER TIME ................................. 98
IN VIVO RECORDINGS FROM TREATED ARRAYS ............................................. 130
FUTURE DIRECTIONS ..................................................................................... 132
CONCLUSION ................................................................................................ 136
REFERENCES ................................................................................................. 138
APPENDIX A: FABRICATION OF SHAM ARRAYS ......................................................... 143
APPENDIX B: FABRICATION OF COMPLETE HIPPOCAMPAL ARRAYS ....................... 145

VI

TABLE OF FIGURES

Figure 1: Schematic of EEG, ECoG, and penetrating electrodes illustrating where they lay and the
characteristic amplitudes and frequencies of electrical recordings collected from each technology [1]. ..... 2
Figure 2: The presence of a rigid, penetrating neural probe in the brain instigates an immune response that
includes both glial scarring, which forms a wall of activated astrocytes and microglia around the neural
implant, and nearby neuronal death. ............................................................................................................. 7
Figure 3: Long polymer neural shanks will buckle during surgical implantation if the buckling force
threshold of the probe is less than the force required to penetrate brain tissue. .......................................... 16
Figure 4: Conceptual drawing of Parylene probe coupled to microwire for reinforcement of Parylene
during insertion (left) vs. unreinforced Parylene probe (right). .................................................................. 20
Figure 5: Layout for three probe designs for implantation. (a) A 3D pocket structure to serve as
mechanical receptacle for microwire tip for microwire-probe coupling. Blue area represents additional
Parylene layer, white represents opening into the pocket. (b) Etched out ellipse on tip of probe for
threading of microwire (c) Plain probe surface for dissolvable coating. Grid lines are 10 µm apart. ........ 20
Figure 6: First AZ4620 etch mask for pocket probes (a), ellipse probes (b), and plain probes (c). Inlet to
pocket in (a) reveals sacrificial photoresist lying beneath Parylene insulation layer. Scale bars are 20 µm.
.................................................................................................................................................................... 21
Figure 7: Fully fabricated sham arrays comprised of ellipse probes (a & b), plain probes to be reinforced
with dissolvable coatings (c & d), and pocket probe (e). Small misalignments between lithography steps
caused the inlet to the pocket structure to stretch beyond the border of the probe in (e). ........................... 22
Figure 8: (a) Macroscopic wisps across entire wafer after sacrificial photoresist lithography steps.
Streaking on various wafers occurred prior to or during deposition of the Parylene insulation layer. (b)
“Alligator skin” pattern in sacrificial photoresist after UV exposure and development of next photoresist
layer. (c) Bubbling in sacrificial photoresist after hardbake of second photoresist layer at wafer edges. .. 23
Figure 9: Insertion tool made of silicon anodically bonded to Pyrex. Trenches cut into silicon with dicing
saw served as channels for microwires. ...................................................................................................... 24
Figure 10: Threading microwires through anodically bonded insertion tool. (a) one, (b), two, (c) three, and
(d) four microwires. Scale bar is 2 mm. ...................................................................................................... 25
Figure 11: Challenges experienced in preparation of insertion tool. (a) Microwire tips were fragile and
curled upon meeting any resistance during threading procedure. (b) The width of the channel exceeded
that of the microwire precluding repeatable 250 µm spacing between each microwire. Some microwires
were closer together than others. (c & d) Parylene shearing during microwire handling pointed out by
arrows. Scale bar is 2 mm. .......................................................................................................................... 25
Figure 12: (a & b) Pocket tips puncturing after and during microwire tip insertion. (c & d) Ellipse probes
deforming or breaking during microwire threading. ................................................................................... 27
Figure 13: Difficulty in decoupling microwire from ellipse probe after insertion into agarose. (a)
Microwire threaded through a single ellipse probe. (b) Coupled microwire and ellipse probe inserted into
0.5 % agarose. Note that probes not attached to the microwire also inserted. (c) Probe remained attached
to microwire during microwire retraction. Probe is pulled out of agarose. ................................................ 27
VII

Figure 14: PDMS layers were cut out to accommodate an insertion backing and the hippocampal array
and to create a cavity surrounding the top half of the probe shanks that could be filled with PEG. The
filling of the cavity with molten PEG is shown in the image to the left, the released hippocampal array
with a PEG brace attached to the array and insertion backing is shown in the image to the right. ............ 30
Figure 15: PEG application techniques; side-view of PEG braces (created with thick PDMS sheets) and
probe profiles. (a) When molten PEG is injected into the PDMS cavity at room temperature the PEG
solidifies into large grains whose constrictions cause probes to splay in the z-direction. (b) This issue is
bypassed when the mold is injected while the mold is heated in a 50 °C oven and the probes exit the brace
completely straight. ..................................................................................................................................... 30
Figure 16: PDMS mold for application of PEG brace to hippocampal arrays, schematic (a) and
photograph (b). The first PDMS layer is placed above a PDMS base and fitted with an acrylic backing.
The hippocampal array is aligned to the acrylic backing and its probe tips overhang the cavity in PDMS
layer 1. The second PDMS layer is aligned to the prior layers and completes the mold. PDMS sheets are
0.5 mm thick and the PEG brace, once completed, is 1 mm thick. The dimensions of the PEG brace are
shown in panel a.......................................................................................................................................... 31
Figure 17: Photographs of PEG 3,350 coupon at (a) 0 min, (b) 3 min, (c) 6 min, and (d) 9 min post
addition of the PBS solution. The coupon shrinks in diameter over time and the spreading of blue dye
throughout the PBS solution indicates dissolution of the dye-laden PEG. Scale bar is 5 mm. ................... 32
Figure 18: Dissolution of PEG based on MW. PEG 1,000 dissolved the quickest at a rate of -9.5% per
minute. All linear fits had a Pearson coefficient of 0.99. Standard deviations across three coupons for each
MW are included, N = 3. ............................................................................................................................ 33
Figure 19: Typical buckling curve. The force starts of at zero. The probe tips are translated closer and
closer to the metal plate as time passes. When the probe tips contact the metal plate the force exerted on
the load cell shoots up. Buckling begins when the force begins to level off and ends in the collapse of the
probes at which point the force drops quickly back to zero. ....................................................................... 34
Figure 20: Average (and standard deviations) of buckling force thresholds measured for sham
hippocampal arrays of 8, 6, 4, 2, and 1 probes with and without PEG braces, pushed against a metal plate
at a speed of 1 mm/s. Buckling force thresholds of 1, 2, and 4 probe arrays without PEG braces fell
beneath the resolution of the load cell and could not be measured. (N= 3 or 6) ......................................... 35
Figure 21: Representative sham array insertion into 0.5% agarose brain phantom. (a) Top view of sham
array with stepped PEG brace design intended for 2-stage insertion. (b) Side view of sham probes
protruding from PEG brace. (c) Front view of entire length of sham array inserted into agarose. Some
shanks drew closer to each other during insertion. (d) Side view of sham array in agarose that inserted
without buckling, but at a slight angle. Scale bar is 1 mm. ......................................................................... 36
Figure 22: Surgical implantation of sham array into rat hippocampus. ...................................................... 37
Figure 23: Histological slices from sham array implantation into rat hippocampi. Slices stained with
hematoxylin and eosin. (a) Coronal slice and (b & c) transverse slices taken at 2.2 and 2.5 mm from the
brain surface showing probe shanks and cross-sections. ............................................................................ 37
Figure 24: Fully fabricated hippocampal array. Photograph of (a) entire array, (b) cable and probe shanks,
and (c) close-up of electrodes and traces on individual probe shanks. ....................................................... 39
Figure 25: Cracked e-beam deposited Pt imaged across (a & b) array contact pads and (c ) electrodes.
This was the mildest case of cracking seen. ................................................................................................ 40
VIII

Figure 26: Rippling in AZ5214 photoresist. 5214 after (a) sputter Pt deposition (b) heat treatment on hot
plate at 110 ° C for 3 minutes and (c) e-beam Pt deposition. Sputtering causes severe wrinkling in the
photoresist lying beneath the sputtered platinum, heating reproduces this phenomenon in photoresist, and
e-beam deposition of metal covers a smooth, undisturbed photoresist surface. Scale bar 50 µm. ............. 42
Figure 27: Overdevelopment at wafer periphery. (a) Photoresist between traces leading away from
electrode sites lift off from wafer completely forming thin filaments. (b) Thin filaments and dark erosion
visible. Scale bar is 100 µm. Increasing the length of the image reversal softbake from 70 to 75 seconds
eliminated non-uniform development across wafer and prevented overdevelopment of devices in wafer
periphery. .................................................................................................................................................... 44
Figure 28: Parylene spherules that formed during CVD of the insulation layer of Parylene. After lift-off,
the wafer surface was cleaned via a RIE O 2 descum (100 W, 100 mTorr, 1 minute). Spherules form
throughout the entire wafer but seem to collect the most densely around metal edges defined through lift-
off. ............................................................................................................................................................... 45
Figure 29: Fully fabricated array (with eight probe shanks) that curves due to compressive stress in the
thin film Pt. This curvature complicated the application of the PEG brace and was predicted to interfere
with hippocampal targeting during surgical placement. (a) High levels of compressive stress in Pt
deposited through sputtering causes intense probe curvature, whereas (b) metal deposited via e-beam
vapor deposition had less stress and probes were less curled. .................................................................... 46
Figure 30: (a) Electrode traces in hippocampal array post Pt sputtering. -511 MPa of compressive stress
were measured by the vendor and result in the wrinkled appearance of the metal surrounding the features
that will eventually undergo liftoff. This compressive stress causes devices to curl away from the exposed
side of the Parylene upon release from the silicon carrier wafer. (b) Electrode traces in hippocampal array
with metal smoothly deposited via e-beam and less compressive stress. ................................................... 46
Figure 31: Arrays thermoformed prior to release from silicon wafer cracked and were permanently
adhered to silicon. (a) Cracks in bulk Parylene due to mismatch in thermal expansion coefficients between
silicon and Parylene, (b) cracks propagating from pointed probe tips, and (c) damage to Parylene array
upon removal of device from silicon backing. ............................................................................................ 47
Figure 32: Schematic of radius of curvature calculation overlaying photograph of array profile. Black line
mimics side view of curved array. Segment L was drawn across the arc created by the curved array. θ, the
angle defined between line segment L and its perpendicular bisector, was used to calculate R. ............... 48
Figure 33: O 2 plasma etching of base layer of Parylene of arrays for 4, 8, and 12 min. Side views of arrays
compared to control 10 µm/ 10 µm array that did not undergo mounting or etching. ................................ 50
Figure 34: One-way, tri-synaptic hippocampal circuit. Neural information from the entorhinal cortex
enters the hippocampus via the DG and then passes sequentially from the DG to the CA3 and finally the
CA1. Understanding the input-output characteristics of each part of the hippocampal circuit can be the
basis for the future development of a hippocampal prosthetic device based on chronic recordings from
penetrating neural probes. ........................................................................................................................... 55
Figure 35: Overall schematic of hippocampal recording setup. Flexible neural probe array is designed to
insert into the rat brain until it reaches the hippocampal target regions. Electrical traces leading away from
the recording electrodes carry neural activities via printed circuit boards (PCBs) atop the rat’s cranium to
an external recording instrument................................................................................................................. 56
Figure 36: Diagram of Parylene neural probe array with eight probes designed to match the anatomy of
the hippocampus as the CA1 and CA3 sub-regions change in depth along the septo-temporal axis. Two
IX

groups of four Pt electrodes (30 µm exposed diameter) target the CA1 and CA3. The shape and critical
dimensions of the array are detailed above. *Thickness of insulative Parylene layer varied between 10, 14,
and 18 µm. .................................................................................................................................................. 58
Figure 37: (a) Schematic and (b) photograph of electrical packaging for in vivo neural array insertions.
Neural recordings from the hippocampal array are transmitted through a series of components soldered
onto two PCBs and are eventually recorded through Plexon recording equipment that interfaces with the
two Omnetics connectors on PCB 2. PCB 1 is permanently dental cemented to the rat’s cranium. The two
PCBs connect through a specialized PCB-PCB connector (SSB6). ........................................................... 59
Figure 38: (a) Needle “posts” used to attach PCB 2 to the micromanipulator arm during agarose and
surgical implantations. (b) Male pin connector attached to two Omnetics ground channels. Connector
mated with female pin connector on microwire serving as in vivo ground. (c) Cable clips used to stabilize
PCB 1 against a clear acrylic backing which is used for the application of a dissolvable brace that
prevents array buckling during implantation. ............................................................................................. 60
Figure 39: Damage to SSB6 plug after repeat connections. (a) prior to matings, (b) after many matings,
damaged area encircled in red. Plastic housing begins to degrade and prevents proper alignment between
plug and receptacle. (c) Marine epoxy underfilling SSB6 receptacle to fortify connection to PCB. Red “x”
indicates part of receptacle pin prone to damage during probing with multimeter. Blue circle points to
preferred probing point. .............................................................................................................................. 62
Figure 40: Cycling lifetime of SSB6 PCB to PCB connector. While these components were only rated by
their manufacturer for 30 cycles, the application of marine epoxy as an extra support between the SSB6
receptacle and PCB pads helped to extend the functional lifetime of these connectors (Trials 1-3). Marine
epoxy and improved alignment techniques allowed for mating cycles of 120 without depreciation (Trials
4-5, overlapping on graph). ......................................................................................................................... 62
Figure 41: Second generation electrical packaging design. Single PCB goes from ZIF to Omentics
connectors directly and bypasses need for PCB to PCB connectors. ......................................................... 64
Figure 42: Representative CV of electrode from a sputtered hippocampal array. The first CV curve is
relatively narrow, indicating a small electroactive area. As the cycles progress the current response
broadens at different voltages indicating increased cleanliness of the electrode surface area which allows
for more surface reactions and current flow. At cycle 30 this CV now mimics the traditional CV of Pt in
H 2SO 4. ......................................................................................................................................................... 65
Figure 43: Microscopic image of contact pad pre (a) and post (b) O 2 RIE descum at 200 W, 200 mTorr,
and 2 min. Pre (c) and post (d) images of electrode surface. Thin residual layer of Parylene scum is visible
in images prior to RIE etch, but cleared away after the descum. CV curves prior to descum revealed an
open circuit pattern, but returned to normal after RIE descum. .................................................................. 66
Figure 44: EIS taken before (a) and after (b) CV cleaning of the representative electrode shown in Figure
42. CV cleaning of this electrode dropped the 1 kHz electrode impedance from 1.3 MΩ to 360 kΩ. ....... 67
Figure 45: Impedance magnitude and phase curves for various AC stimulation voltages. Impedance
magnitude stays noise-free until 0.1 mV. The phase curve already has outliers at a stimulation voltage of
0.1 mV. ....................................................................................................................................................... 69
Figure 46: In vivo implantation of hippocampal array supported by dissolvable PEG brace into rat brain.
(a) Bare tips of hippocampal array poised over implantation site, (b) PEG incrementally dissolved in
saline and newly exposed probe length further inserted into brain with pauses for electrical recording to
X

ensure proper placement, and (c) remainder of PEG brace dissolved in preparation for dental cement cap.
Scale bar = 5 mm. ....................................................................................................................................... 75
Figure 47: Final assembly on rat’s cranium. Dental cement cap encapsulates remainder of hippocampal
array and first half of PCB 1 with care taken to leave the ground pin and SSB6 PCB-PCB connector on
PCB 1 (not visible) exposed for future reconnections to PCB 2. ................................................................ 76
Figure 48: Representative 0.1 s long spikes recorded intra-operatively with hippocampal array implanted
to 4.15 mm from the brain surface. Complex spikes (burst of 2-6 single spikes of decreasing amplitude
with ≤ 5 ms interspike intervals) recorded from the CA1 and CA3 during an experimental implantation
surgery. ....................................................................................................................................................... 78
Figure 49: Average spike amplitudes (top column stack) and noise levels (bottom column stack) with
standard deviations across all four Parylene array implantations compared to those from seven microwire
array implantations. The maximum spike amplitudes recorded by Parylene and microwire arrays (N = 7)
are represented by the circle above each bar which connects to the y-axis on the right. ............................ 79
Figure 50: Average signal to noise ratio (with standard deviations) achieved by each Parylene
hippocampal array in the acute surgical implantation setting. Dashed line represents the average signal to
noise ratio across seven microwire array implantations and dashed-dot lines represent ± 1 standard
deviation of the microwire average. The maximum signal to noise ratio achieved by each Parylene array
during implantation are noted by the circular symbols which reference the y-axis to the right of the plot.80
Figure 51: Chronic recordings obtained from the third implantation on day 24 post implantation. Complex
spikes were recorded from both the CA1 and CA3 sub-regions. ................................................................ 81
Figure 52: Averages and standard deviations of peak-valley spike amplitudes (solid symbols) and
background noise (outline symbols) across all functional electrodes, for each of the three chronic
recording experiments, with any suspected spikes due to crosstalk removed. Arrays were monitored until
recordings were no longer visible during chronic testing. N varied at each day of recording. Maximum
number of channels that recorded unique spikes during a single day was 41, minimum was 7. ................ 82
Figure 53: Average signal to noise ratio (SNR) and standard deviation across all functional electrodes for
each of the three chronic recording experiments, with any suspected spikes due to crosstalk removed.
SNR remains steady over the lifetime of recordings. N varied at each day of recording. Maximum number
of channels that recorded unique spikes during a single day was 41, minimum was 7. ............................. 83
Figure 54: Transverse, 50 µm thick hippocampal slice (at -2.7 mm) from one-month sham array
implantation. Stained with GFAP to highlight astrocytes in brown; purple corresponds to hematoxylin
counter staining; dense purple strip is the DG of hippocampus. Array was unfortunately removed prior to
tissue slicing during removal of the brain from the cranium. Scale bars are 100 µm. (a) Black arrows
indicate locations of three probes of the array visible in the DG of the hippocampus and (b) arrows mark
the location of matched controls from the same tissue slice. Color thresholding was used to measure the
astrocytic density in 25 µm rings around the central three probes and corresponding control regions,
included in table in (c). ............................................................................................................................... 85
Figure 55: Transverse, 50 µm thick hippocampal slice (at -2.75 mm) from one-month sham array
implantation. Stained with NeuN to highlight neurons in brown; purple corresponds to hematoxylin
counter staining; dense purple strip is the DG of the hippocampus. Array was removed prior to tissue
slicing. (a) Black arrows indicate locations of five probes of the array visible in the DG of the
hippocampus and (b) is the color thresholded version of the image with 50 µm thick rectangular bins
drawn around the central three probes to measure density of neurons surrounding the implantation sites.
XI

Scale bars are 100 µm. (c) A graphical representation of the concentration of neurons in implanted and
control (not shown) rectangular bins. Trace represents neuronal concentration in bins surrounding the
implant site, horizontal lines represent the average neuronal concentration ± 1 standard deviation in
matched control regions (N = 3). Neuronal concentration returned to control levels in between probes in
the array. Probe cross-sections are indicated by open filled circular markers. ........................................... 86
Figure 56: Crosstalk can occur (a) between electrodes, (b) between traces on the same probe, and (c)
between traces in different probes at the Parylene ribbon cable. ................................................................ 95
Figure 57: Crosstalk between electrodes in CA1 and CA3 groups from the first chronically implanted
animal at 24 days post implantation. (a) Pictorial representation of spikes with identical waveforms
recorded from electrodes 31-40 (excluding 32). (b) Raster plot showing most spikes occurring at identical
timestamps, providing evidence of crosstalk. ............................................................................................. 96
Figure 58: Pictorial representation of one mechanism of crosstalk. Solution permeates bulk Parylene and
collects at voids or in areas of poor adhesion between the Parylene-metal-Parylene interfaces. This forms
conductive conduits between adjacent channels which causes the transfer of voltage signals between
neighboring channels. ................................................................................................................................. 97
Figure 59: Description of crosstalk measurement in ideal circumstance, where each channel is fully
insulated from its neighbor. In this case, the only cells in the crosstalk graph that will have 100%
crosstalk occur across the diagonal, which is where the voltage signal is sent to and read from the same
channel. Crosstalk between neighboring channels should be zero, but is usually < 5% due to noise in the
system. ...................................................................................................................................................... 101
Figure 60: Packaging for initial crosstalk tests. To connect between the PCB and the crosstalk system
wires were hand-soldered to PCB pads and header pins. All metal surfaces were covered with marine
epoxy (not shown). Test vials with attached packaging were placed inside a covered water bath set to 37
°. ................................................................................................................................................................ 102
Figure 61: Crosstalk present immediately upon soaking annealed arrays in initial crosstalk tests. Crosstalk
increased in magnitude until day 8 and remained steady until the experiment ended at day 20. Channels in
set A had CV cleaning only and no EIS testing before the crosstalk experiment. Channels in set B
underwent CV cleaning and EIS testing prior to experimentation. .......................................................... 103
Figure 62: Crosstalk remaining in packaging after hippocampal cable and probes were cut off. Average
and standard deviation of crosstalk in set A decreased from 66% ± 12% to 19% ± 10% and crosstalk in set
B decreased from 95% ± 2% to 66% ± 9%. This indicates that the previous crosstalk measurements were
a combination of crosstalk inherent to the hippocampal array itself as well as crosstalk in the packaging.
.................................................................................................................................................................. 103
Figure 63: Improved packaging and soaking conditions for long-term crosstalk tests. (a) Electrical
connections from the PCB to the crosstalk system were made via isolated jumper cables rather than
header pins which could easily lead to shorts between channels if moisture penetrated any of the header
pins. (b) Complete packaging set-up: Probes (not shown) soaked in a 1x PBS filled glass vial. Parylene
cable passed through a slit drilled into the glass vial cap and contact pads insert into the ZIF of the PCB.
The vial cap and all exposed metal connections were protected with marine epoxy. Jumper cables
soldered to the back-end of the PCB end in male connectors that were inserted into the crosstalk system.
(c) Packaged vials rest in a beaker with water that is heated to 37 °C atop a hotplate. In this way
packaging is not susceptible to moisture condensation and permeation. .................................................. 104
Figure 64: Eight consecutive, physically adjacent traces, chosen to span two probes. ............................ 106
XII

Figure 65: Crosstalk in control array displayed at time points (in days) where changes occurred. Crosstalk
was not present until day 43, when catastrophic signal leakage between channels 2-6 occurred. At day 50
this crosstalk spread to include channel 1 as well. Days 57- 106 exhibit an odd trend, that of crosstalk
magnitude decreasing over time. .............................................................................................................. 106
Figure 66: Crosstalk in dehydrated sets of traces over time. Crosstalk in set A appeared at day 84 of the
soaking test, and spread from channels 3-5 to include channels 1 and 2 as well by day 90. Crosstalk
remained at these levels until day 106. Sets B and C experienced no significant crosstalk for the duration
of the experiment. ..................................................................................................................................... 107
Figure 67: Crosstalk in HF-treated set A over time. A short between channels 1 and 3 developed by day 1
of testing. Crosstalk between the shorted channels only developed at day 50 and gradually expanded over
the course of the experiment to include channels 6-8 on the same probe of the hippocampal array. ....... 108
Figure 68: Crosstalk in HF-treated set B over time. Crosstalk was not present until day 64 at which point
catastrophic crosstalk levels (>50%) became evident in channels 1-6 of the same probe. Crosstalk levels
fluctuated slightly but mostly remained steady from day 71 to day 106 of the experiment. .................... 109
Figure 69: Crosstalk in AdPro Plus® treated arrays over time. Crosstalk of magnitude < 40% became
evident in set A after only 4 hours of soaking and fluctuated over time until it seemed to disappear
completely by day 57. The other two AdPro Plus® sets of traces did not have crosstalk for the duration of
the experiment. .......................................................................................................................................... 110
Figure 70: Microscopic images of channel delamination and blister formation that correspond to the
presence of crosstalk in dehydrated sets imaged at day 113 of the crosstalk experiment. The highly
reflective Parylene surface is caused by the dimpling of array caused by the annealing process. Colored
lines indicate the eight consecutive channels tested during crosstalk experiments. Channels in dehydrated
set A corresponded to regions of poor adhesion between layers, whereas channels in dehydrated set B
appeared unaffected by this phenomenon. For dehydrated set C, delamination and blistering occurred at
locations far away from the channels tested. ............................................................................................ 111
Figure 71: Summary of channels tested (solid blue) during crosstalk experiments across arrays fabricated
with different treatment steps and which channels experienced crosstalk by day 106 of the ongoing
experiment (highlighted in purple). .......................................................................................................... 112
Figure 72: Time scale of solution evaporation from array. Repeat crosstalk tests performed in dry air after
the array had been removed from soaking in 37° C PBS solution. Each consecutive matrix was divided by
the original crosstalk matrix to yield an average percent ± standard deviation of the original crosstalk
measured in the array (N=56). .................................................................................................................. 114
Figure 73: Multimeter resistance tests in PBS over time. Channel 1 used as reference except for in AdPro
Plus devices where connections seemed to be lost by day 90. Mid-teens MΩ impedance indicated proper
connection. Impedance on scale of Ω is short circuit, and OL symbol = overload indicates that impedance
between channels is too high to measure (cell background colored in white). Resistance between channels
listed in column header, in units of MΩ. Darker blue indicates lower resistances. .................................. 117
Figure 74: Typical CV curve for Pt electrodes immersed in 0.05 M H 2SO 4 with an Ag/AgCl counter
electrode. Peaks and humps indicate the potential at which H+ adsorption and desorption occur and Pt-
oxide is formed and reduced. .................................................................................................................... 121
Figure 75: Typical impedance magnitude and phase curves from EIS data of a well-insulated
microelectrode........................................................................................................................................... 122
XIII

Figure 76: Changes to CV and EIS curve that occur with catastrophic delamination. (a) CV current range
increases by two magnitudes. Green curve taken from CVs prior to delamination, purple curve taken from
CVs post delamination. (b) EIS after delamination occurred. Impedance magnitude drops from ~ 600 to 1
kΩ. Phase fails to return to resistive regime at lower frequencies. (n=3, error bars represent standard
deviations for both panels). ....................................................................................................................... 123
Figure 77: Channels that delaminated during electrochemical testing of same array as in Figure 76,
perhaps due to the advanced age of devices. (a) “Wavy” channels at the first stage of delamination on the
Parylene ribbon cable, (b) complete dehiscence of Pt channels from Parylene base of ribbon cable, (c)
delamination present at probe tips. Double arrows point to separation between base and insulation layer of
Parylene. ................................................................................................................................................... 123
Figure 78: Crosstalk measurements taken at day 90 with corresponding CV curves, EIS impedance
magnitudes, and EIS phase curves of each of the eight channels in a single array taken at day 92 of
soaking. Data for (a) dehydrated set C, (b) dehydrated set A, (c) control set, and (d) HF-treated set A. . 125
Figure 79: Electrical circuit model and pictorial representation of current paths through a channel. Branch
1 is the traditional Randles circuit which models faradaic charge transfer as a resistor, R e and capacitive
charging across the inner and outer Helmholtz charge layers as a constant phase element with Y e and α e
and ends with a solution resistance, R s. Branch 2 represents crosstalk between two traces due to the
ingress of conductive solution through elements R delam, Y delam, and α delam. Branch 3 represents the high
frequency drop-off caused by capacitive coupling between instrument components, C wire. WE represents
the working electrode, or channel, and RE represents the Ag/AgCl reference electrode. ........................ 128
Figure 80: Model fit to EIS data from dehydrated set C, trace 7 using equivalent circuit model from
Figure 79. Goodness of fit is 3E-4 with parameters: R e = 53 MΩ, Y e = 1 nS*s^a, α e = 0.9, R s = 13 kΩ,
R delam = 3 MΩ, Y delam =1 nS*s^a, α delam = 0.9, and C wire = 13 pF. .............................................................. 129
Figure 81: Modelling EIS curves as delamination parameters change. Central parameters are R delam = 1
MΩ, Y delam =1 nS*s^a, α delam = 0.9. When α delam= 1 the constant phase element has behavior identical to a
capacitor. All other circuit values are kept at R e = 100 MΩ, Y e = 1 nS*s^a, α e = 0.9, R s = 10 kΩ, C wire = 10
pF. ............................................................................................................................................................. 129
Figure 82: Integration of screw and cranium. (a) Proper integration, yellow arrow points to cranium
hugging threads of screw. From explanted array that stayed implanted for months. (b) Poor integration
between screw and cranium. Yellow arrow points to clean screw with no cranium attached. Dental cement
and array fell off after only a week. .......................................................................................................... 130
Figure 83: In vivo recordings from array treated with AdPro Plus® at seven days post-implantation. Each
box presents neural units recorded by a single electrode (1-8) on a single probe of the Parylene array
(green, yellow, or orange). (a) High amplitude spike (pink dotted line) localized to a single electrode on
the probe, electrode 1. Neural units in yellow on electrodes 2 and 3 are independent units. (b) Electrodes
1, 2, and 4 of the same electrode group (targeting the CA3) record identical spikes, but no crosstalk
occurs to CA1 electrodes. (c) Crosstalk across both electrode groups in a probe. Neural unit centered on
electrode 2 (highest amplitude), but signal is leaked to electrodes 1, 4, 5, 7, and 8 as well. Boxes marked
with an “x” were disconnected.................................................................................................................. 131
Figure 84: Current ranges measured during CV cleaning for electrodes that did not experience crosstalk in
vivo (n= 14) compared to current ranges for electrodes that experienced crosstalk (n= 24). Although the
maximum and minimum current ranges overlap, there exists a statistically significant difference (p <
0.05) between the means of both groups. With further refinement, this could serve as a way to predict
which electrodes will last without crosstalk in vivo. ................................................................................. 132
XIV

Figure 85: Location of the hippocampus in rat, monkey, and humans. Drawn to scale according to [35].
.................................................................................................................................................................. 134
Figure 86: Anticipated fabrication scheme for hippocampal array with integrated microfluidics. A prime
silicon wafer (1) is etched with O 2 plasma to form microchannels. A sacrificial release layer is deposited
to coat the microchannels prior to deposition of Parylene via CVD (3). Another prime, silicon wafer is
coated with a release layer and then CVD coated with Parylene (4). The two wafers are laminated
together through a thermal annealing treatment (5) and the top silicon wafer is removed from the
assembly to allow for subsequent patterning and exposure of Pt electrodes and another Parylene insulation
layer (7-9). Array is released from the carrier wafer (10). ........................................................................ 136
Figure 87: Imagined top view of a single probe. Two microchannel outlets per each microchannel lie
beneath the surface patterned electrodes and channels. ............................................................................ 136

1

Electrical communication in the brain
The vast anatomical and functional complexity of the human brain makes its exploration
a daunting task yet awesome goal. Though it weighs only a small fraction of the human body, its
importance cannot be understated and the elegant manner in which it controls the actions of the
entire body is a holy grail of natural mimicry for researchers who desire to understand, imitate,
and interface with the brain.
The functional cell of the brain, the neuron, passes neural information along its axons
through high-speed electrical transmissions and messages are conveyed between nerves through
chemical synapses. Both these electrical and chemical communications control effectors that
range from other nerves to sensory organs and muscles.
The basic unit of electrical communication, the action potential (also known as the nerve
impulse or spike), consists of the physical depolarization and repolarization of an axon’s
membrane potential as ion-gated channels open and close in response to a signal from an
adjacent neuron. These voltage changes span over a hundred millivolts and pass through the
axons within milliseconds. This electrical activity is a target for neural interfaces that seek to
monitor or communicate with individual neurons in the shared language of electricity.
Extracellular electrophysiological recordings
As action potentials flow across a neuron, ionic fluctuations across cellular membranes
create extracellular electrical fields that vary in time. These potentials can be monitored and
evaluated to provide insight into the order and manner in which neurons in a neural network
interact with one another and the input and output characteristics of each leg of neural circuitry.

INTRODUCTION: PENETRATING NEURAL
PROBES FOR HIGH-RESOLUTION, CHRONIC,
ELECTRICAL RECORDINGS
2

Methods by which to record these fluctuations have developed over time, and historically
progress from less invasive to more invasive procedures as the demand for improvements in
recording resolution grew. They can appear as grid-like electrodes that lie atop the skull, as in
electroencephalography (EEG), or atop the brain directly, as in electrocorticography (ECoG),
and record summation activities from large populations of neurons. Alternatively, a penetrating
microelectrode can record from a small population of nearby neurons or even a single neuron,
enabling precision recording of individual neural activities. Penetrating neural probes will be the
focus of this thesis, however we will begin with a short survey (Figure 1) of the other electrical
recording technologies available.

Figure 1: Schematic of EEG, ECoG, and penetrating electrodes illustrating where they lay and
the characteristic amplitudes and frequencies of electrical recordings collected from each
technology, based on [1].

The original method for recording electrical activities in the form of brain waves,
developed in the 1930’s by Hans Berger, came in the form electroencephalography (EEG) [2].
This consists of an electrode placed on a scalp which measures continuous voltage fluctuations
across the entire brain tissue. EEG is often used for locating an epileptic focus non-invasively [3]
since its temporal resolution is high-enough to discover the locus site before the induced,
abnormal electrical activity spreads into a full-blown seizure, which is not the case for slower
imaging-based techniques such as magnetic resonance imaging. EEG has also been explored to
determine the severity of injury and unconsciousness in the case of traumatic brain injuries [2]
3

and to diagnose Alzheimer’s disease as well [4, 5]. However, since the electrode lies atop the
skull, the voltage signal is attenuated and distorted across the layers of soft tissue and bone that
block the electrode, resulting in a spatial resolution of only ~ 10 cm
2
. EEG signals record from
populations of 10
4
to 10
7
neurons, making it difficult to directly evaluate the relationship
between EEG recordings and individual neural firing patterns [6]. This precludes the ability to
tease apart the input and output characteristics of neural networks.
In electrocorticography (ECoG), developed in the 1950’s, electrodes are placed
underneath the cranium and the dura, and record directly from the surface of the cortex. The
placement of ECoG electrodes requires a craniotomy—it is more invasive than EEG, but less
intrusive than electrodes that are placed into the brain tissue and damage the blood-brain-barrier.
ECoG, like EEG, records electrical activity from a population of neurons, however it has an
increased spatial resolution of 4 mm
2
. This spatial resolution is two orders of magnitude superior
to that of EEG, and can used to find epileptic loci when higher spatial targeting is necessary [7].
The frequency distributions, amplitudes, and phase of these oscillations, as well as their
relationship with behavioral events, have been used to study the potentiation of functional neural
networks in memory, language, and spatial processing. However, ECoG still lacks the resolution
necessary to obtain electrical events from individual neurons.
Electrodes that penetrate the brain tissue, the most invasive type of recording, afford the
greatest spatial resolution due to the shortened distance between recording sites and neurons of
interest. Penetrating microelectrodes can record changes in the extracellular potential of a group
of neurons in the form of local field potentials (LFPs) or can measure unitary electrical activities
from individual neurons. While the presence of a penetrating electrode inside the brain tissue
itself causes an immune response against the foreign implant, which can result in nearby
neuronal death and scar formation that insulates the electrodes from nearby neurons, this remains
the sole recording solution that can measure neural activities on the order of single neurons with
high temporal resolution.
Ideally this “electrode” would have a small enough area to selectively target individual
neurons while retaining a low enough impedance to enable discrimination between spikes and
background noise. An insulated electrical trace leading away from the electrode site would then
connect to external electronics which allow for monitoring. Since an intermediate sized neuron,
for example, a pyramidal neuron in the cornu ammonis of the hippocampus, reaches ~ 30 µm in
4

diameter [8], “microelectrodes” are traditionally limited to < 10,000 µm
2
[9] in area in order to
record from specific neurons . The construction of small electrodes on this scale is challenging,
but advancements in microelectromechanical systems (MEMS) fabrication techniques that arose
from the boon in interest in the use of semiconductor technologies for electronics has enabled
microelectrode fabrication from a variety of materials in multiple designs.
For an electrode to be able to record from a neuron, the recording site should ideally lie
within 50-100 µm from the neuron and should not exceed a distance of 140 µm [10]. Surgical
placement of electrodes within this narrow window is difficult, especially due to heterogeneity in
the composition of the variegated brain structures which make some areas more difficult to
penetrate than others [11] as well as the presence of blood vessels and connective tissue
superficial to the target site which can cause electrode deflection and deviation from the desired
target. Acute damage to this vasculature and to the blood brain barrier instigates an immune
response against the foreign implant. This compels attempts to avoid disrupting surface
vasculature during implantation in attempt to avoid the initiation of these inflammatory processes
[12].
A common application of single-neuron recordings that reaches beyond basic
neuroscience research includes the development of multi-input, multi-output neural network
models that enable control of brain-machine-interfaces (BMI). In BMIs, the brain and machine
communicate via the common language of electrical activity using electrodes as the
communication medium. A BMI generally consists of at least three components: a sensor, which
monitors the electrical activity of the brain over time; a decoder, which interprets the signal
intent; and an effector, where the decoded output controls movement in a prosthetic or stimulates
neural targets further downstream. High-recording quality, sensitivity, and long recording
lifetimes are requirements for the development of a working BMI that can control movements at
various speeds with multiple degrees of freedom or even sensation. Although BMIs with
penetrating electrode technology have been shown to experimentally enable control of computer
cursors and motor prostheses in both primates and humans [13, 14], these technologies are yet
relegated to the world of “investigational devices” by the Federal Drug Administration largely
due to limitations in probe lifetimes and signal fidelity as a result of the immune response against
these foreign implants. There is a growing body of evidence that suggests that implementing
techniques to modify the material properties and size of penetrating microelectrodes in order to
5

enhance the integration of the probe into the surrounding brain tissue and limit the immune
reaction mounted against the implant may preserve the high-resolution recording advantage of
penetrating microelectrodes so they can be used for long-term experimental studies and
therapeutic interventions in cases of injury.
Penetrating neural probes for neural recordings
Penetrating neural probes have adopted various forms throughout history as technological
advancement in microfabrication enabled the creation of smaller, more densely packed
microelectrodes. Early microelectrodes that measured extracellular voltage changes consisted of
glass pipettes containing ionic solution which could respond to changes in membrane potential
and send the signal to a recording instrument through a wire housed inside the pipette. However,
these glass microelectrodes could only function for short periods of time due to leakage of ionic
solution. Next, in the 1950’s, microwires, conductive metal wires with insulation everywhere
except at the tip, were born. Microwires are not bound by the same material limitations as glass
pipettes and can record from the same neuron for many days in a row. Multiple microwires could
be wrapped together, for example, in tetrode formation, to record from multiple neurons in a
single area, or they could be arranged into microwire arrays with spaces between each
microwire. However, both these two technologies are inherently limited to a single recording site
per probe shank and require manual assembly of back-end electronics.
With the advent of microelectromechanical systems (MEMS) techniques, however, batch
fabrication of microelectrodes arrays on slender silicon shanks became possible. One model of a
multi-electrode array, called the Utah Array, is a multidirectional array that consists of
conductively doped, thin, silicon needles that have insulation everywhere except at their metal
tips. The Utah Array [15] (called the ‘Neuroport’) first received 510(k) approval for clinical
trials from the Federal Drug Administration in 2007. The second genre of multi-electrode silicon
arrays, known as the Michigan Array [16], stemmed from research performed at the University
of Michigan and consists of a one-dimensional probe array with multiple electrodes patterned
along the length of each probe. Probes in Michigan Arrays can be designed and fabricated with a
plethora of creative designs and architectures. The Michigan Array probes represent the
traditional form that current micromachined probe shanks take—a long, slender shank with a
pointed tip for tissue penetration and electrodes and traces patterned along the length of each
6

shank. As microfabrication technologies mature and develop, feature sizes that are smaller and
smaller can be patterned, yielding an increased density of electrode sites per probe shank. The
state of art technology has advanced to such a state that hundreds of individual and
independently addressable electrode sites (> 1,000) can now be patterned onto a narrow silicon
shanks (70 – 100 µm wide) through improvements in complementary metal-oxide-
semiconductor lithography and the direct integration of probes with application-specific
integrated circuitry [17-19]. For a more in-depth discussion of neural probe technologies see [20-
22].
Lifetime limitations of rigid neural probes
While there has been a tremendous amount of literature exploring the successes of silicon
probes to achieve high-quality neural recordings, one issue that continues to plague the field is
that the recording capabilities of these devices wane over time, most failing within a few months.
A retrospective evaluation of 78 intracortical, 100-site, Utah Arrays chronically implanted in
rhesus monkeys found the average recording lifetime to be 12 months, with a longest successful
recording time of 5.75 years [23]. In humans implanted with a silicon Utah array, researchers
witnessed a diminished number of recorded neural units after only 6.5 months in vivo [14].
Silicon-based probes fall short of the goal to achieve stable, long-term recordings over many
years for brain–machine interface technology [24, 25]. Implant failure can be associated with a
variety of biological (e.g., immune response) and non-biological (e.g., connector or electrode
failure) mechanisms, including mechanical damage to or chemical corrosion of electrodes and
traces, degradation of passivation layers and insulating coatings, and the foreign body response
of the brain to the implant; of these mechanisms, solutions that adequately address the biological
immune response are lacking and are critically needed to lengthen the lifetime of neural probes.
Rigid silicon and metal probes suffer inevitable signal degradation over time as chronic
tissue inflammation leads to an immune cascade that eventually may wall off of the implant. This
wall consists of glial cells (astrocytes and microglia) surrounding the implanted rigid electrodes,
which increases the distance between neurons and electrode recording sites resulting in impaired
signal-to-noise ratios over time [26]. The chronic inflammatory environment surrounding the
foreign implant leads to nearby neuronal death as well [27] (Figure 2). Whereas metals and
silicon have Young’s moduli on the order of hundreds of GPa, the stiffness of brain tissue is
7

orders of magnitudes softer, at around 10
−6
GPa. This disparity between the stiffness of brain
tissue and implantable neural probes is a likely source of tissue damage due to chronic
inflammation caused by the natural micromotion of the brain and tethering forces from anchored
electrical connections between the probes and brain exterior. It has been proposed that the use of
probes fabricated from softer materials can mitigate this damage and attenuate the adverse
immune response. In one study, soft poly(p-xylylene) (Parylene) probes were shown to induce
only a 12%–17% neuronal loss around the implantation site compared to rigid silicon probes
which incurred 40% neuronal loss at four weeks post-implantation [28].
Alternatively, instead of choosing a substrate material with a lower Young’s modulus,
some groups have explored the use of rigid substrates that when thinned to small cross sectional
dimensions can exhibit flexible properties and have lower overall stiffness factors. One such
study of carbon fibers with diameters of < 10 µm has found limited tissue disruption when
compared to larger footprint, rigid neural probes [29].

Figure 2: The presence of a rigid, penetrating neural probe in the brain instigates an immune
response that includes both glial scarring, which forms a wall of activated astrocytes and
microglia around the neural implant, and nearby neuronal death.
8

Using flexible probes to attenuate immune response and extend
recording lifetimes
The hypothesis that more compliant probes would cause less microdamage to the
surrounding brain area, and therefore limit the attendant foreign body response experienced by
the implant, has been the impetus to explore soft polymers as substrates for these devices. The
Young’s modulus (E) of rigid silicon devices is 190 GPa and 78 GPa for gold, while brain tissue
in rats can range from 0.1 to 1.2 MPa, resulting in at least five orders of magnitude difference
[30, 31]. By switching to polymers, the stiffness mismatch can be reduced. The Young’s
modulus of polydimethylsiloxane (PDMS), for example, is on the order of hundreds of kPa,
which is very closely matched to the stiffness of brain (Table 1).
One of the first penetrating probes made out of compliant materials was reported in 2001,
consisted of a polyimide-based (2.3–8.5 GPa), three probe array inserted into the rat barrel
cortex, and achieved acute recordings with a maximum signal to noise ratio of 5:1 [32]. This
proof-of-concept experiment was motivated by the desire to achieve movement of the implant in
sync with the brain during natural micromotion (e.g., arising from cardiac pulse or respiration),
thereby ameliorating damage to surrounding tissue. The idea that compliant substrates could
attenuate probe micromotion—the relative movement between implant and brain, or, for tethered
implants, the movement between the brain and skull which can then propagate to the implant-
brain interface—was first documented in reference to polyimide based peripheral electrodes and
retinal arrays [33, 34]. These early reports have been followed by a large body of research
seeking to achieve flexible penetrating neural probes suitable for basic neuroscience and
prosthetic research, including the exploration of new soft, flexible materials.
A range of studies, from in vitro and in vivo to modeling studies, have examined the
effect of implant stiffness modulation on surrounding brain tissue, providing evidence to support
the use of softer substrates in brain probes. 3D finite-element modeling work comparing silicon
(200 GPa), polyimide (3 GPa), and soft probes (with an elastic modulus of 6 MPa, made out of a
hypothetical material) showed that polyimide provided strain relief against tangential tethering
forces (probe displacing in z-direction), but that soft probes provided relief against both
tangential and radial (x-y displacement) forces [35]. Flexible probes were theorized to absorb
micromotion forces mostly in the length of probe proximal to the point of tethering, thereby
minimizing the magnitude of forces that interfered with the tip of the probe. This same study
9

indicated that soft probes could reduce interfacial strains by up to 94% during vertical
displacement, as compared to silicon counterparts. Benchtop modeling studies used gelatin (15
kPa) as a brain model and compared flexible microelectrode arrays of polyimide to rigid
microwire implants. The major findings were that bending, mechanical shock, and lateral
deflection tests, which simulated shifting of the brain within the cranium, caused tearing and
disruption of the gelatin matrix only by rigid implants, while the flexible implants imparted no
damage beyond the initial tract formed during implantation [34].
In vitro studies further support the use for soft materials, indicating that softer substrates
are more effective promoters of cell growth than their stiff counterparts. Neurons grown on
bisacrylamide gel with a stiffness of 0.23 MPa were shown to grow three times more branches
than those grown on a gel with twice the stiffness [36]. Dorsal root ganglion neurite extension
growth rate on agarose was reduced by more than half as the stiffness of agarose increased from
0.75% to 2% (wt/vol) concentrations [37], corresponding to moduli from ~10 to 200 kPa,
respectively [38].
The impact of substrate stiffness on glial cells, including microglia and astrocytes, was
examined as well. Cells plated on either a stiff (100 kPa) or soft (10–30 kPa) substrate exhibited
changes in cell morphology and protein expression that were similar to typically observed in vivo
immune responses. Glial cells on the stiffer substrate developed more expansive and denser
cellular processes than their counterparts grown on the softer substrate. This behavior is
reminiscent of the physical shape of glial cells, the main immune players, when activated. Cells
grown on stiffer substrate were also shown to upregulate inflammatory mediators toll-like
receptor 4 (TLR4) and peroxisome proliferator-activated receptor γ (PPARγ). Likewise, acute in
vivo studies involving immunohistochemical staining around implants in rat brains revealed that
cells neighboring stiffer areas of the implant exhibited elevated inflammatory marker CD11b
after just one week of implantation and increased glial fibrillary acidic protein (GFAP) levels
after three weeks of implantation. Elevated GFAP levels indicate the presence of activated
astrocytic cells, which suggests that mechanical mismatch between electrodes and nervous tissue
may enhance unfavorable tissue reactions to the implant that impair its performance [39].
Finally, there are chronic data from in vivo studies to support these claims as well. A
study of compliant poly(vinyl acetate) rat brain implants with Young’s moduli in the tens of
MPa, compared to a five-times stiffer, chemically matched silicon control showed a
10

comparatively reduced pro-inflammatory cytokine Iba1 and CD68 release around a 200-µm
radius from compliant probes at 2, 8, and 16 weeks post-implantation. By 16 weeks post-
implantation, NeuN staining for neuronal density revealed a return to normal around compliant
probes but not for silicon implants [40]. This suggests the importance of the role of scarring as
well as subsequent local neurodegeneration that adversely affect recording quality [41]. The
stability of the blood-brain barrier was assessed by IgG infiltration tests, revealing less damage
was caused by compliant probes. Neural probes made of a polydimethylsiloxane (PDMS) base
with an overall Young’s modulus of ~ 1 MPa were shown to have increased neuronal survival
and induce less microglial and astrocytic reactivity than size and surface chemistry matched
tungsten microwires at 8 weeks post implantation in rat [42] with similar beneficial results seen
for soft probes made of nanocomposites [43] and off-stoichiometry thiol-enes-epoxy (OSTE+)
polymer [44] as compared to rigid counterparts.

Table 1: Commonly used soft polymers for neural probes (silicon for reference). Values from
Scholten [30] and Hassler [45] unless otherwise noted. PDMS: Polydimethylsiloxane; Parylene:
Poly(p-xylylene); LCP: Liquid crystal polymer; BCB: Benzocyclobutene; USP: United States
Pharmacopeia Convention.
Property Silicon PDMS Polyimide
Parylene
C
SU-8 LCP BCB
Young’s Modulus
(GPa)
190
3.6 × 10
−4
–8.7
× 10
−4

2.3–8.5 2.76 2.87–4.40 [46] 10.6 3.1 [47]
Dielectric Constant
11.9
[48]
2.6–3.8 3.5 3.1 3.2 3.0 2.65 [47]
Achievable
Thicknesses
(µm)
-
10–100 (spin
coat)
1–15 1–100 1–300 25–3000 7–130
Biocompatibility - USP class VI
Yes (in vivo)
[45, 49]
USP class
VI
Mild reactivity
(in vivo)
USP
class VI
Yes (ex
vivo) [50]

Objectives
Traditionally, penetrating neural probe technology has been limited to applications whose
anatomical targets lie within the superficial cortex of the brain. This is the case with motor and
sensory brain machine interfaces which are generally implanted < 3 mm deep into brain tissue.
However, growing interest in structures that lie deep within the brain--such as the hippocampus
and other structures like the thalamus and basal ganglia [51]-- has challenged engineers to design
and develop probes whose shank lengths are designed to successfully and repeatedly reach
11

deeper brain structures without breaking, buckling, or straying from the target area of interest.
Inserting into deep-brain structures with small target areas is especially challenging. This is
compounded by the inherent difficulty in implanting thin, flexible materials into tissue without
buckling. The following work seeks to develop techniques that enable deep-brain implantation of
flexible neural probes into small targets of interest.
Only a handful of penetrating neural probes have been used to record from the
hippocampus in vivo and each have generic electrode layouts that have not been customized to
hippocampal anatomy [52-55]. The large distance between recording sites in some of these
probes preclude the ability to simultaneously record from the same neuron through neighboring
electrodes, which would provide valuable spatial positioning and localization information.
Recordings from these neural probes have been limited to those achieved during surgical
placement itself. This work seeks to fill this experimental void by designing custom built flexible
probes with electrode sites positioned specifically to target the laminar anatomy of the
hippocampus, with small enough spacing between electrodes targeting the same cell layer to
enable recording redundancy, and by achieving chronic neural recordings that enable an
evaluation of array lifetime.
Difficulties in electrically packaging probe or probe arrays, especially for probes made of
flexible substrates with many recording sites, are another barrier towards the translational use of
this technology. Many resort to the manual soldering of bulky electrical connections to probes
which is time-consuming and can introduce an added source of noise to signals as well as
recording drop out over time due to faulty connections. The eventual goal is to integrate
multiplexers directly on to individual probe shanks with built in telemetry. However, size
limitations and difficulties in bonding flexible probe materials to silicon microchips relegate this
to the realm of a future reality, a task which is currently being undertaken by the Meng lab. In
this hippocampal array study, wired electrical packaging printed circuit boards were designed to
minimize the weight and footprint atop the rat’s cranium and care was taken to automate the
assembly of the electrical packaging system as much as possible and to enhance its robustness
over time.
In short, this work focuses on advancing polymer-based neural probe technology and
recording longevity through the microfabrication of a Parylene-based hippocampal recording
system that records from multiple regions of the rat hippocampus simultaneously. This text will
12

present the design of the anatomically-conformal penetrating neural array, the microfabrication
techniques used to fabricate the array, a novel method for inserting the flexible array into deep-
brain tissue while minimizing acute insertion trauma, and a custom electrical packaging system
that enables simultaneous recordings from 64 electrodes while minimizing connector size and
weight on the rat’s cranium. In vitro array testing of the electrodes through EIS and mechanical
characterization of the array and the dissolvable insertion brace will also be discussed in the
following chapters. Acute and chronic (> 2 month) monitoring of array performance through
SNR analysis and histological staining will provide insight into the lifetime of our Parylene
neural array along with crosstalk challenges experienced with these polymer devices. The final
chapter will include a discussion of future experiments. It is the hope of the author that this
neural array furthers the development of a hippocampal circuit model that, in combination with
the array itself, may one day help restore memory to those who have suffered hippocampal
damage.
References
[1] N. V. Thakor, "Translating the brain-machine interface," Science translational medicine,
vol. 5, pp. 210ps17-210ps17, 2013.
[2] B. E. Wallace, A. K. Wagner, E. P. Wagner, and J. T. McDeavitt, "A history and review
of quantitative electroencephalography in traumatic brain injury," The Journal of head
trauma rehabilitation, vol. 16, pp. 165-190, 2001.
[3] C. M. Michel and M. M. Murray, "Towards the utilization of EEG as a brain imaging
tool," Neuroimage, vol. 61, pp. 371-385, 2012.
[4] J. Jeong, "EEG dynamics in patients with Alzheimer's disease," Clinical
neurophysiology, vol. 115, pp. 1490-1505, 2004.
[5] A. Tsolaki, D. Kazis, I. Kompatsiaris, V. Kosmidou, and M. Tsolaki,
"Electroencephalogram and Alzheimer’s disease: Clinical and research approaches,"
International Journal of Alzheimer’s Disease, vol. 2014, 2014.
[6] G. Buzsáki, C. A. Anastassiou, and C. Koch, "The origin of extracellular fields and
currents—EEG, ECoG, LFP and spikes," Nature reviews neuroscience, vol. 13, pp. 407-
420, 2012.
[7] J. Jacobs and M. J. Kahana, "Direct brain recordings fuel advances in cognitive
electrophysiology," Trends in cognitive sciences, vol. 14, pp. 162-171, 2010.
[8] P. Andersen, R. Morris, D. Amaral, T. Bliss, and J. OKeefe, The hippocampus book:
Oxford University Press, 2006.
[9] S. F. Cogan, "Neural stimulation and recording electrodes," Annu. Rev. Biomed. Eng.,
vol. 10, pp. 275-309, 2008.
[10] D. A. Henze, Z. Borhegyi, J. Csicsvari, A. Mamiya, K. D. Harris, and G. Buzsáki,
"Intracellular features predicted by extracellular recordings in the hippocampus in vivo,"
Journal of neurophysiology, vol. 84, pp. 390-400, 2000.
13

[11] J. Guo, S. Hirsch, A. Fehlner, S. Papazoglou, M. Scheel, J. Braun, et al., "Towards an
elastographic atlas of brain anatomy," PLoS One, vol. 8, p. e71807, 2013.
[12] T. Kozai, T. Marzullo, F. Hooi, N. Langhals, A. Majewska, E. Brown, et al., "Reduction
of neurovascular damage resulting from microelectrode insertion into the cerebral cortex
using in vivo two-photon mapping," Journal of neural engineering, vol. 7, p. 046011,
2010.
[13] J. M. Carmena, M. A. Lebedev, R. E. Crist, J. E. O'Doherty, D. M. Santucci, D. F.
Dimitrov, et al., "Learning to control a brain–machine interface for reaching and grasping
by primates," PLoS biol, vol. 1, p. e42, 2003.
[14] L. R. Hochberg, M. D. Serruya, G. M. Friehs, J. A. Mukand, M. Saleh, A. H. Caplan, et
al., "Neuronal ensemble control of prosthetic devices by a human with tetraplegia,"
Nature, vol. 442, pp. 164-171, 2006.
[15] E. M. Maynard, C. T. Nordhausen, and R. A. Normann, "The Utah intracortical electrode
array: a recording structure for potential brain-computer interfaces,"
Electroencephalography and clinical neurophysiology, vol. 102, pp. 228-239, 1997.
[16] A. C. Hoogerwerf and K. D. Wise, "A three-dimensional microelectrode array for chronic
neural recording," IEEE Transactions on Biomedical Engineering, vol. 41, pp. 1136-
1146, 1994.
[17] G. Rios, E. V. Lubenov, D. Chi, M. L. Roukes, and A. G. Siapas, "Nanofabricated neural
probes for dense 3-D recordings of brain activity," Nano letters, vol. 16, pp. 6857-6862,
2016.
[18] M. HajjHassan, V. Chodavarapu, and S. Musallam, "NeuroMEMS: neural probe
microtechnologies," Sensors, vol. 8, pp. 6704-6726, 2008.
[19] D. Wang, Q. Zhang, Y. Li, Y. Wang, J. Zhu, S. Zhang, et al., "Long-term decoding
stability of local field potentials from silicon arrays in primate motor cortex during a 2D
center out task," Journal of neural engineering, vol. 11, p. 036009, 2014.
[20] R. J. Vetter, J. C. Williams, J. F. Hetke, E. A. Nunamaker, and D. R. Kipke, "Chronic
neural recording using silicon-substrate microelectrode arrays implanted in cerebral
cortex," Biomedical Engineering, IEEE Transactions on, vol. 51, pp. 896-904, 2004.
[21] J. C. Williams, R. L. Rennaker, and D. R. Kipke, "Long-term neural recording
characteristics of wire microelectrode arrays implanted in cerebral cortex," Brain
Research Protocols, vol. 4, pp. 303-313, 1999.
[22] T. D. Kozai, A. S. Jaquins-Gerstl, A. L. Vazquez, A. C. Michael, and X. T. Cui, "Brain
tissue responses to neural implants impact signal sensitivity and intervention strategies,"
ACS chemical neuroscience, vol. 6, pp. 48-67, 2015.
[23] R. Biran, D. C. Martin, and P. A. Tresco, "Neuronal cell loss accompanies the brain
tissue response to chronically implanted silicon microelectrode arrays," Experimental
neurology, vol. 195, pp. 115-126, 2005.
[24] J. P. Seymour and D. R. Kipke, "Neural probe design for reduced tissue encapsulation in
CNS," Biomaterials, vol. 28, pp. 3594-607, Sep 2007.
[25] P. R. Patel, K. Na, H. Zhang, T. D. Kozai, N. A. Kotov, E. Yoon, et al., "Insertion of
linear 8.4 μm diameter 16 channel carbon fiber electrode arrays for single unit
recordings," Journal of neural engineering, vol. 12, p. 046009, 2015.
[26] K. Scholten and E. Meng, "Materials for microfabricated implantable devices: a review,"
Lab on a Chip, vol. 15, pp. 4256-4272, 2015.
14

[27] B. S. Elkin, A. Ilankovan, and B. Morrison, 3rd, "Age-dependent regional mechanical
properties of the rat hippocampus and cortex," J Biomech Eng, vol. 132, p. 011010, Jan
2010.
[28] P. J. Rousche, D. S. Pellinen, D. P. Pivin, Jr., J. C. Williams, R. J. Vetter, and D. R.
Kipke, "Flexible polyimide-based intracortical electrode arrays with bioactive
capability," IEEE Trans Biomed Eng, vol. 48, pp. 361-71, Mar 2001.
[29] T. Stieglitz, H. Beutel, and J.-U. Meyer, "A flexible, light-weight multichannel sieve
electrode with integrated cables for interfacing regenerating peripheral nerves," Sensors
and Actuators A: Physical, vol. 60, pp. 240-243, 1997.
[30] D. P. O'Brien, T. R. Nichols, and M. G. Allen, "Flexible microelectrode arrays with
integrated insertion devices," in Micro Electro Mechanical Systems, 2001. MEMS 2001.
The 14th IEEE International Conference on, 2001, pp. 216-219.
[31] 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.
[32] L. A. Flanagan, Y.-E. Ju, B. Marg, M. Osterfield, and P. A. Janmey, "Neurite branching
on deformable substrates," Neuroreport, vol. 13, p. 2411, 2002.
[33] A. Balgude, X. Yu, A. Szymanski, and R. Bellamkonda, "Agarose gel stiffness
determines rate of DRG neurite extension in 3D cultures," Biomaterials, vol. 22, pp.
1077-1084, 2001.
[34] V. Normand, D. L. Lootens, E. Amici, K. P. Plucknett, and P. Aymard, "New insight into
agarose gel mechanical properties," Biomacromolecules, vol. 1, pp. 730-738, 2000.
[35] P. Moshayedi, G. Ng, J. C. Kwok, G. S. Yeo, C. E. Bryant, J. W. Fawcett, et al., "The
relationship between glial cell mechanosensitivity and foreign body reactions in the
central nervous system," Biomaterials, vol. 35, pp. 3919-25, Apr 2014.
[36] J. K. Nguyen, D. J. Park, J. L. Skousen, A. E. Hess-Dunning, D. J. Tyler, S. J. Rowan, et
al., "Mechanically-compliant intracortical implants reduce the neuroinflammatory
response," J Neural Eng, vol. 11, p. 056014, Oct 2014.
[37] G. C. McConnell, H. D. Rees, A. I. Levey, C. A. Gutekunst, R. E. Gross, and R. V.
Bellamkonda, "Implanted neural electrodes cause chronic, local inflammation that is
correlated with local neurodegeneration," J Neural Eng, vol. 6, p. 056003, Oct 2009.
[38] Z. J. Du, C. L. Kolarcik, T. D. Kozai, S. D. Luebben, S. A. Sapp, X. S. Zheng, et al.,
"Ultrasoft microwire neural electrodes improve chronic tissue integration," Acta
Biomaterialia, 2017.
[39] J. Harris, J. Capadona, R. Miller, B. Healy, K. Shanmuganathan, S. Rowan, et al.,
"Mechanically adaptive intracortical implants improve the proximity of neuronal cell
bodies," Journal of neural engineering, vol. 8, p. 066011, 2011.
[40] H. C. Lee, F. Ejserholm, J. Gaire, S. Currlin, J. Schouenborg, L. Wallman, et al.,
"Histological evaluation of flexible neural implants; flexibility limit for reducing the
tissue response?," Journal of Neural Engineering, vol. 14, p. 036026, 2017.
[41] C. Hassler, T. Boretius, and T. Stieglitz, "Polymers for neural implants," Journal of
Polymer Science Part B: Polymer Physics, vol. 49, pp. 18-33, 2011.
[42] C. Robin, A. Vishnoi, and K. N. Jonnalagadda, "Mechanical behavior and anisotropy of
spin-coated SU-8 thin films for MEMS," Journal of Microelectromechanical Systems,
vol. 23, pp. 168-180, 2014.
15

[43] K. Lee, S. Massia, and J. He, "Biocompatible benzocyclobutene-based intracortical
neural implant with surface modification," Journal of micromechanics and
microengineering, vol. 15, p. 2149, 2005.
[44] M. J. Madou, Fundamentals of microfabrication: the science of miniaturization: CRC
press, 2002.
[45] N. Lago, D. Ceballos, F. J. Rodrı
́ guez, T. Stieglitz, and X. Navarro, "Long term
assessment of axonal regeneration through polyimide regenerative electrodes to interface
the peripheral nerve," Biomaterials, vol. 26, pp. 2021-2031, 2005.
[46] B. A. Koeneman, K.-K. Lee, A. Singh, J. He, G. B. Raupp, A. Panitch, et al., "An ex vivo
method for evaluating the biocompatibility of neural electrodes in rat brain slice
cultures," Journal of neuroscience methods, vol. 137, pp. 257-263, 2004.
[47] W. M. Grill, S. E. Norman, and R. V. Bellamkonda, "Implanted neural interfaces:
biochallenges and engineered solutions," Annual review of biomedical engineering, vol.
11, pp. 1-24, 2009.
[48] A. Bragin, J. Hetke, C. Wilson, D. Anderson, J. Engel, and G. Buzsaki, "Multiple site
silicon-based probes for chronic recordings in freely moving rats: implantation, recording
and histological verification," Journal of neuroscience methods, vol. 98, pp. 77-82, 2000.
[49] A. Berényi, Z. Somogyvári, A. J. Nagy, L. Roux, J. D. Long, S. Fujisawa, et al., "Large-
scale, high-density (up to 512 channels) recording of local circuits in behaving animals,"
Journal of neurophysiology, vol. 111, pp. 1132-1149, 2014.
[50] J. Csicsvari, D. A. Henze, B. Jamieson, K. D. Harris, A. Sirota, P. Barthó, et al.,
"Massively parallel recording of unit and local field potentials with silicon-based
electrodes," Journal of neurophysiology, vol. 90, pp. 1314-1323, 2003.
[51] V. Castagnola, E. Descamps, A. Lecestre, L. Dahan, J. Remaud, L. Nowak, et al.,
"Parylene-based flexible neural probes with PEDOT coated surface for brain stimulation
and recording," Biosensors and Bioelectronics, vol. 67, pp. 450-457, 2015.

16

The long thin beam structure of a soft polymer neural probe, especially one that targets
deep brain structures with a long shank, is prone to buckling during surgical implantation (Figure
3). To avoid buckling, the probe structure must be stiffened to tolerate the penetration force
required for surgical implantation. This chapter contains an overview of strategies used to avoid
the mechanical buckling of polymer probes during implantations as well as an account of
microwire-based shuttle strategies that were explored for hippocampal probes but were
ultimately proven unsuccessful. The final design of a novel dissolvable brace, along with data
from experimental in vitro and in vivo testing of this solution, are also included. The significance
of the bracing technique is that the bare probe can be implanted directly preserving the small
cross section of the as-fabricated polymer shank in contrast to more commonly employed
surgical stiffening techniques.

Figure 3: Long polymer neural shanks will buckle during surgical implantation if the buckling
force threshold of the probe is less than the force required to penetrate brain tissue.

Polymer probes typically consist of a thin metal layer (~100’s of nm thick) containing
electrodes and traces sandwiched between two insulating and supporting polymer layers (several
to 10’s of m thick). The individual layers can possess intrinsic stress, a result of the deposition

MECHANICAL CHALLENGES IN THE
CREATION OF FLEXIBLE, HIPPOCAMPAL
RECORDING ARRAY
17

process, and thermal stress, a result of mismatch in thermal expansion of the metal and polymer
layers, which may arise at layer interfaces during processing steps following their initial
deposition [1]. Both types of stress, whether compressive or tensile, can lead to undesirable
curvature of the probe shank once released from the carrier wafer, making precise surgical
placement virtually impossible. Strategies for stress compensation to straighten polymer neural
probes through thermal annealing and modifying the insulation thickness are included in this
chapter as well as other difficulties that were encountered when fabricating Parylene-metal-
Parylene devices.
Both techniques, which seek to control precise placement and planarity of the probes, are
generalizable to probes and probe arrays of different designs.
Flexible array implantation
Since Parylene probes lack axial stiffness, they must be temporarily stiffened to facilitate
surgical insertion and placement of electrode sites in the brain tissue. For the hippocampal
region targeted, the probes will need to be inserted to > 4 mm vertical depth. Creative
implantation solutions such as coating and insertion shuttle techniques were explored to enable
probe insertion without buckling.
Implantation techniques
A perusal of implantation strategies for both rigid and soft probes revealed two commonly
employed insertion aids for probe implantation. Many groups use either dissolvable coatings or
temporary structural shuttles to boost the buckling force threshold of the probe during insertion.
Both strategies have attendant advantages and disadvantages that are discussed below.
Common coatings used to stiffen flexible neural probes during insertion include silk [2-
4], polyethylene glycol (PEG) [4-6], tyrosine-derived polymers [7, 8], carboxy-methyl-cellulose
(CMC) [9], gelatin [10, 11], saccharose [12], table sugar [13], and maltose [14] (Table 3).
Biocompatibility of these substances is of utmost importance, as any exacerbation of the immune
response to the coated probe by surrounding tissue detracts from the quality and lifetime of
subsequent neural recordings. Coatings are biocompatible if they do not elicit an immune
response from surrounding tissue, inherent to which is the assumption that byproducts of the
18

coating can be effectively cleared from the surrounding brain tissue. It is important to note that
even if coating byproducts can be cleared, they nonetheless temporarily modify the local brain
chemistry. No matter which application technique is used, coatings have the inherent
disadvantage of increasing the cross-section of flexible probes, thereby undesirably enlarging the
acute injury of penetrating probes. Coatings differ based on methods of coating and the time
scale of dissolution of the coating. Coatings can chemically engineered together to adjust
dissolution rates or achieve coating-specific benefits simultaneously [15]. Successful coatings
must be strong enough to allow for probe penetration, coating dissolution times must be quick
enough to avoid a chronic immune response, and coating byproducts must be safe [8]. Coatings
should be uniform and smooth so as not to cause additional tissue damage during insertion [7].
Alternatively, shuttles are temporary structural supports that are another common way to
achieve placement of soft polymer probes into brain tissue. Shuttles are attached to the back-side
of a flexible probe, and for one-sided planar probes, pose no risk of covering up electrode sites
during implantation. Shuttles can be coupled to probes through biodissolvable adhesives like
PEG [16-18], by adjusting the surface chemistry of the shuttle and probe to physically attract
them to one another [19], or even through custom hold-and-release vacuum systems
[20].Considerations of shuttle design dictate minimizing the shuttle cross-section, to minimize
surgical trauma, choosing shuttle materials that are biocompatible and do not exacerbate the
foreign body response, and fabricating shuttles out of materials that are easy to manipulate.
However, if the flexible probe overhangs its stiff shuttle backing, it is likely for the probe to get
caught in superficial tissue during insertion. Therefore, shuttle bodies are made larger than the
flexible probe to eliminate this issue. This need naturally conflicts with the desire to minimize
insertion trauma mentioned above. Groups that follow this approach generally cite a study that
determined the insertion tract wounds made by silicon probes of various sizes only affected acute
tissue response, not the chronic response [21], to justify the attendant size increase caused by use
of a shuttle [19]. The greatest difficulties in implementing a structural shuttle lie in the adhesion
between the probe-shuttle couple. During implantation, it is essential to prevent sliding of the
probe against the shuttle, which could lead to premature decoupling of the two and complicate
probe placement. After the insertion is completed, the two must be decoupled without the probe
accidentally retracting from the target of interest. Probe displacement is therefore one of the
metrics of comparison between various shuttle solutions.
19

Previous efforts in our lab have explored the use of a tungsten rod as a temporarily
insertion shuttle for the implantation of a 2 × 2 array of Parylene neural probes with a 3
dimensional sheath structure that served not only as a pocket to mechanically couple the rod to
the flexible probe, but was also designed to improve integration between the implant and neural
tissue by encouraging neural ingrowth into the sheath structure based on Kennedy’s neurotrophic
cone electrode design [18]. This strategy was used as the basis for one of the probe designs
below, which involves the use of a pocket structure at the tip of the probe as a mechanical
anchoring structure for a microwire insertion shuttle.
Design of implantation strategies
Based on this background literature, three probe designs for attachment to an insertion
shuttle or for coating were fabricated and tested. The first design included a 3D pocket structure
at the tip of the probe for mechanical adhesion to a microwire insertion shuttle. The second
consisted of an elliptically shaped hole etched into the tip of the Parylene probe through which a
microwire could be threaded in order to couple the two together. See Figure 4 for a conceptual
drawing of coupling between the Parylene probe and a microwire for reinforcement during
surgery. The third consists of a plain, planar probe tip that can be coated with PEG, PNIPAM, or
silk as a supportive, dissolvable coating for probe penetration into the brain. See Figure 5 for a
pictorial representation of each design. An insertion tool, made of silicon and pyrex, was
designed to house eight microwires to serve as insertion shuttles for each of the eight probes on
an individual array and a ninth microwire to serve as a location marker for histology purposes.

20

Figure 4: Conceptual drawing of Parylene probe coupled to microwire for reinforcement of
Parylene during insertion (left) vs. unreinforced Parylene probe (right).

Figure 5: Layout for three probe designs for implantation. (a) A 3D pocket structure to serve as
mechanical receptacle for microwire tip for microwire-probe coupling. Blue area represents
additional Parylene layer, white represents opening into the pocket. (b) Etched out ellipse on tip
of probe for threading of microwire (c) Plain probe surface for dissolvable coating. Grid lines are
10 µm apart.
Fabrication of sham arrays for testing
Sham probe arrays, which lacked the metal electrodes, traces, and contact pads were
fabricated using traditional micromachining techniques applied to thin film polymers. Sham
arrays consisted of eight Parylene probes with long (5.5 mm), thin, and tapered shanks separated
from each other by a distance of 250 µm. To fabricate these arrays, a layer of 10 µm of Parylene
21

(Specialty Coating Systems, Indianapolis, IN) was chemically vapor deposited (CVD) on a
dehydrated, prime 4” silicon wafer to form the base of each probe shank. A positive photoresist
(AZ4620; Integrated Micro Materials, Argyle, TX) was patterned to define a 8 µm thick
sacrificial photoresist layer at the tip of one set of probes in the shape of pockets that were 100
µm tall with an opening width of 60 µm. A second layer of 10 µm thick Parylene was CVD
deposited as insulation. A 15 µm thick layer of AZ4620 photoresist (Integrated Micro Materials,
Argyle, TX) was spun (step 1: 5 s, 500 rpm, step 2: 45 s, 1,200 rpm) and patterned to produce an
etch mask. This first etch mask exposed the cutout of each array, an inlet to the pocket structure
for the pocket probes, and the ellipsoid cutout for the ellipse probes (see Figure 6). A switched
chemistry process in a deep reactive ion etching tool that alternated between fluoropolymer
deposition (C4F8) and oxygen plasma etching [22] was used to etch the first 10 µm of Parylene.
At the end of this etch, the pocket inlet exposed the sacrificial photoresist in the pocket probes. A
second 15 µm thick layer of AZ4620 photoresist was patterned to once again expose the cutout
of each array and the ellipses to etch the remaining 10 µm of Parylene. Any remaining
photoresist from the two resist masks were stripped through sequential rinsing of the wafer with
acetone, isopropanol and deionized water. Devices were released by gently peeling the device
away from the native oxide layer of the silicon substrate while immersed in water. Any
remaining photoresist was removed by soaking released devices for 5 minutes in sequential baths
of acetone, isopropanol, and water.

Figure 6: First AZ4620 etch mask for pocket probes (a), ellipse probes (b), and plain probes (c).
Inlet to pocket in (a) reveals sacrificial photoresist lying beneath Parylene insulation layer. Scale
bars are 20 µm.

22

Figure 7: Fully fabricated sham arrays comprised of ellipse probes (a & b), plain probes to be
reinforced with dissolvable coatings (c & d), and pocket probe (e). Small misalignments between
lithography steps caused the inlet to the pocket structure to stretch beyond the border of the
probe in (e).
Some challenges were experienced during fabrication of these sham arrays, a few of
which were subsequently attributed to the use of expired AZ4620 photoresist. The first problem
encountered was gross streaking of sacrificial photoresist across the entire wafer as shown in
Figure 8 panel (a). During lithography of this sacrificial photoresist layer, its appearance was
unremarkable. However, sometime right before or during deposition of the Parylene insulation
layer, a wispy pattern seemed to span across the entire wafer. Two processing parameters were
modified to tease apart the cause for this streaking. In order to determine if the photoresist was
insufficiently cured, softbake and hardbake temperatures were increased to a hotplate reading of
120 °C from 90 °C. However, this did not prevent wisping. In attempt to determine if the
sacrificial photoresist was undeveloped, development time was extended from 80 to 160
seconds—but streaking around the time of Parylene insulation deposition still occurred.
This same layer of photoresist continued to encounter difficulties for the remainder of
processing as well. During the second lithographic patterning step to form the first etch mask for
this wafer, the sacrificial photoresist from the prior lithography step changed to resemble
“alligator skin” after the UV exposure and development step of the second etch mask and
bubbled during a hardbake at 110 °C (Figure 8 panels (b&c)). It was posited that the UV
exposure during this second lithography step caused gas release in the finicky sacrificial
photoresist layer which led to the severe bubbling witnessed upon heating the wafer beyond its
initial soft and hard bake temperatures of 90 °C. The use of newly purchased AZ4620 and the
23

monitoring of hot plate temperatures with a temperature gun infrared thermometer to ensure that
bakes were at the desired temperatures eliminated these issues in future fabrication runs.
Etching through an entire 10 µm of Parylene at a single time in the DRIE was found to
cause widespread bubbling at the interface between the silicon wafer and the base Parylene layer
(115 loops). The Parylene inflated to such a degree that future alignments were impossible. By
splitting DRIE into groups of 25 loop this bubbling was sufficiently attenuated to enable further
processing. Perhaps smaller etching intervals attenuated the expansion of solvents trapped
beneath the Parylene layer which can occur under vacuum at elevated temperatures.

Figure 8: (a) Macroscopic wisps across entire wafer after sacrificial photoresist lithography
steps. Streaking on various wafers occurred prior to or during deposition of the Parylene
insulation layer. (b) “Alligator skin” pattern in sacrificial photoresist after UV exposure and
development of next photoresist layer. (c) Bubbling in sacrificial photoresist after hardbake of
second photoresist layer at wafer edges.
Fabrication of microwire insertion tool
Nine vertical trenches were etched onto silicon coupons that were 1 × 2 mm using a
dicing saw with a nickel blade. This created 90 µm-wide channels separated at 250 µm center-to-
center spacing from one another to match the spacing between probes on an array. Since the
microwires were purchased with a diameter of 75 µm, this allowed for a ~7.5 µm space on either
side of the microwire to enable microwire insertion into the channel. Trenches were cut 140 µm
deep. The silicon coupons were anodically bonded to Pyrex 7740 glass coupons of the same size
which served as a coverslip to create a channel for the microwires to thread through (Figure 9).
Tungsten microwires (Microprobes for Life Science, Gaithersburg, MD) were threaded through
the completed insertion tool and arranged so that a 5 mm length of microwire was left to
overhang the insertion tool for coupling with the neural array (Figure 10).
24

A few difficulties were encountered when attempting to thread microwires through the
channels of the insertion tool. Firstly, the tip of the microwire curled drastically upon meeting
the slightest bit of resistance. Threading the microwire in tip first damaged the thin tips of the
microwire. In subsequent preparations microwires were threaded in from the back end to prevent
this problem. Since the width of the channel exceeded that of the microwire (measured to be only
65 µm in diameter) some microwires crowded together and precluded equal 250 µm spacing.
Additionally, the thin insulative Parylene coating on the microwires sheared off easily during
handling with tweezers and the threading procedure. To prevent this, the microwires were passed
through a flame to melt off the Parylene insulation layer. Curled tips were sanded down with
600-grit sandpaper by gently grazing the probes across the paper. Cyanoacrylate was used to
backfill the channels and permanently fix the microwires to the insertion tool.

Figure 9: Insertion tool made of silicon anodically bonded to Pyrex. Trenches cut into silicon
with dicing saw served as channels for microwires.

25

Figure 10: Threading microwires through anodically bonded insertion tool. (a) one, (b), two, (c)
three, and (d) four microwires. Scale bar is 2 mm.

Figure 11: Challenges experienced in preparation of insertion tool. (a) Microwire tips were
fragile and curled upon meeting any resistance during threading procedure. (b) The width of the
channel exceeded that of the microwire precluding repeatable 250 µm spacing between each
microwire. Some microwires were closer together than others. (c & d) Parylene shearing during
microwire handling pointed out by arrows. Scale bar is 2 mm.
26

Testing of implantation strategies
Initial attempts to mechanically coupling microwire insertion aides to pocket and
ellipsoid sham probes were challenging. The pocket tips on probes were tightly closed due to
stiction between the two Parylene layers which prevented the microwire insertion. It was
discovered that a drop of isopropyl alcohol expanded the pockets and that further nudging with
microwires opened the pockets fully. Unfortunately, misalignment of the lithographic layers
resulted in pockets which were unsupported on the right side and repeatedly sheared off from the
probe border (Figure 12 panels a & b).
For the ellipse probes, aside from the gross difficulty in maneuvering the microwire to
thread through the ellipse, the Parylene surrounding the hole was too flimsy to withstand the
pressure of a microwire during threading. Figure 12 panels c & d show deformed and broken
probe tips caused by the pressure applied during microwire threading. A few ellipses remained
intact and were inserted into 0.5% agarose while coupled to a supportive microwire (Figure 13).
Unfortunately, it proved impossible to decouple the microwire from the probe during retraction
of the insertion aide.
During this insertion trial, it was noticed that some probe shanks were able to insert into
the agarose without any insertion aide (Figure 13 panel b). This sparked an idea that temporarily
shortening the length of the probes could perhaps enable probe penetration without a stiffener or
shuttle. The exploration and development of this strategy is discussed in the following section.

27

Figure 12: (a & b) Pocket tips puncturing after and during microwire tip insertion. (c & d) Ellipse
probes deforming or breaking during microwire threading.

Figure 13: Difficulty in decoupling microwire from ellipse probe after insertion into agarose. (a)
Microwire threaded through a single ellipse probe. (b) Coupled microwire and ellipse probe
inserted into 0.5 % agarose. Note that probes not attached to the microwire also inserted. (c)
Probe remained attached to microwire during microwire retraction. Probe was pulled out of
agarose.
Dissolvable brace for minimally invasive implantation
Once it was serendipitously discovered that a certain length of the Parylene arrays could
penetrate agarose on its own, without buckling, this idea was further explored as an alternate to
28

implantation strategies that relied on bulky coatings or shuttles that are difficult to couple to
probes and probe arrays.
Theory
If a single probe shank is modelled as a mechanical beam of Young’s modulus E, second
area moment I, column effective length factor k, and length L, Euler’s buckling formula (Eq. 1)
can be used to predict the smallest axial force applied to the beam that would initiate buckling.
For this model to hold true, the probe is assumed to have an unvarying cross-sectional area—the
tapering of the probe tip is neglected. The column effective length factor k is an expression of the
degree to which each end of the column is constrained against movement. For a probe inserting
into brain tissue or a brain phantom, the base of the probe is considered to be clamped to the
insertion tool and fixed (allowing for no translation or rotation), and the tip of the probe is pinned
in the x-y plane as soon as it contacts brain tissue (only allows for rotation, not translation).
Thus, the commonly accepted value of k is 0.7, which was validated experimentally in [23]. The
second area moment I is considered across the axis which around is more prone to buckling. For
probes with rectangular shaped cross-sections, I is directly proportional to the width of the probe
w, and follows the thickness t cubed (Eq. 2). Substituting Eq. 2 into Eq. 1, yields Eq. 3, which
explicitly shows the relationship between buckling force threshold on w, t, and L.

𝐹 𝐸𝑢𝑙𝑒𝑟 =
𝜋 2
𝐸𝐼
(𝑘𝐿 )
2

Eq. 1

𝐼 =
1
12
𝑤𝑡
3

Eq. 2

𝐹 𝐸𝑢𝑙𝑒𝑟 =
𝜋 2
𝐸𝑤 𝑡 3
5.88𝐿 2

Eq. 3

A perusal of these equations indicated that by temporarily shortening the length of the
probe during insertion, its buckling force threshold can increase according to the square of its
new, effective length (Eq. 3). This was accomplished by using a dissolvable coating to
immobilize the base of the probe shanks, while leaving the tips exposed. This decreased the
effective exposed length of the probe that is susceptible to buckling, thereby enhancing the
29

buckling force threshold of the assembly. Once the exposed length of the probe was successfully
inserted into the target tissue, and the supportive coating reaches the insertion interface, it could
be dissolved away incrementally as subsequent probe length was fed into the tissue. After
discovery of this implantation strategy, it was found that this same technique had been
previously applied to the insertion of flexible microneedles with high aspect ratios (5 µm
diameter silicon microneedles, up to 650 µm in length) which were fortified with a silk film base
scaffold for insertion into mouse brain in vivo [24]. A similar scaffolding technique, this time
using polyethylene glycol (PEG) instead of silk, was used to insert 8.4 µm carbon fiber
microwires of lengths up to 4.5 mm into agarose and rat brain [25]. For a Parylene C based
hippocampal probe of dimensions 5.5 mm shank length, 20-28 µm thicknesses, and < 200 µm
width that was not thermally treated, we would expect a single shank to buckle at just 0.8 mN,
unacceptably low compared to the commonly accepted insertion force into the brain tissue, 1 mN
[26]. By halving the length L, Fbuckling is increased four-fold to 3.3 mN (estimated).
Dissolvable brace design
Polyethylene glycol (PEG) was chosen as the dissolvable material for the probe brace due
to its tunable dissolution profile based on its molecular weight, ease of preparation, and the
favorable outcomes of biocompatibility studies which reveal its anti-immunogenic and antigenic
properties [27]. PEG is a wax-like solid at room temperature and turns to liquid at temperatures
greater than 50 °C. Initial attempts at bracing the top length of Parylene probes with PEG
involved application of the PEG with a soldering iron. This method led to non-uniform PEG
coatings over the probes, intense splaying of the probes, and was damaging to the Parylene C
which is sensitive to oxidization at temperatures greater than 125 °C in the presence of oxygen
[28]. Next, polydimethylsiloxane (PDMS) sheets (1.5 mm thick) with cut outs were layered
together to form a mold cavity around the array (Figure 14) and molten PEG was injected into
the mold cavity. When this was performed at room temperature, the PEG solidified with large
grains and this pulling once again caused splaying of the Parylene probes. By warming up the
mold in an oven heated to 50 °C and performing the PEG injection inside the warm environment,
this issue was abated (Figure 15).

30

Figure 14: PDMS layers were cut out to accommodate an insertion backing and the hippocampal
array and to create a cavity surrounding the top half of the probe shanks that could be filled with
PEG. The filling of the cavity with molten PEG is shown in the image to the left, the released
hippocampal array with a PEG brace attached to the array and insertion backing is shown in the
image to the right.

Figure 15: PEG application techniques; side-view of PEG braces (created with thick PDMS
sheets) and probe profiles. (a) When molten PEG is injected into the PDMS cavity at room
temperature the PEG solidifies into large grains whose constrictions cause probes to splay in the
z-direction. (b) This issue is bypassed when the mold is injected while the mold is heated in a 50
°C oven and the probes exit the brace completely straight.
Uniform rectangular PEG braces as well as PEG braces with a “stepped” pattern were
used in initial experiments. Braces with a stepped pattern had an abrupt transition in brace width
(Figure 21) in order to test the efficacy of an incremental dissolution and insertion paradigm.
While initially the PEG brace covered only half of the length of hippocampal probes, the final
dimensions of the mold were chosen to shorten the effective shank length by 4.55 mm such that
only 0.95 mm of unexposed probe length remained. This helped to ensure probe insertion
without buckling even in cases where variations in stiffness in the brain exceeded the 1 mN force
threshold for insertion. The final version of the PDMS mold was prepared from three layers of
0.5 mm thick polydimethylsiloxane (PDMS), the bottom of which aided in brace removal. These
sheets were thinner than the original PDMS layers used in order to create a thinner PEG brace
that could dissolve more quickly. Braces were prepared using 1,000 molecular weight PEG. The
31

first PDMS layer accommodates a 0.05 mm thick acrylic insertion backing which the array is
aligned to and laid on top of, hanging over a PDMS cavity (Figure 16). The second PDMS layer
completes the top half of the mold. The PDMS mold, hippocampal array, and PEG (Sigma-
Aldrich, Darmstadt, Germany), were heated in a 50 °C oven for 30 seconds. Molten PEG was
then injected into the mold containing the acrylic insertion backing and hippocampal array, and a
PDMS coverslip was placed on top to wipe away excess PEG. The entire assembly was then
cooled at room temperature for 5 minutes until the PEG solidified. The encapsulated device was
then removed from the mold carefully, using a pair of tweezers.

Figure 16: PDMS mold for application of PEG brace to hippocampal arrays, schematic (a) and
photograph (b). The first PDMS layer is placed above a PDMS base and fitted with an acrylic
backing. The hippocampal array is aligned to the acrylic backing and its probe tips overhang the
cavity in PDMS layer 1. The second PDMS layer is aligned to the prior layers and completes the
mold. PDMS sheets are 0.5 mm thick and the PEG brace, once completed, is 1 mm thick. The
dimensions of the PEG brace are shown in panel a.
32

PEG dissolution experiment
In initial benchtop and surgical tests, braces made out of PEG 1,000 wilted during
transport and dissolved too quickly during surgery. The dissolution rate of PEG depends directly
on the molecular weight (MW) of the PEG. In order to determine the MW of PEG that would
make the brace resilient enough during transport and would not melt during surgical handling, an
experiment was performed to determine how quickly PEG with varying MW from 1,000 –
14,000 would dissolve. PEG of each MW was melted in an oven heated to 50 °C and was mixed
with blue dye. Blue PEG of each MW was injected into three 5 mm diameter holes cut out of a 1
mm thick PDMS sheet (PDMS, Sylgard 184; Dow Corning Corp., Midland, MI). Excess PEG
was wiped away from the top of the mold and the PDMS layer was peeled away to reveal dyed
PEG coupon. Coupons were placed in hinged plastic boxes (VWR International, Brisbane, CA)
and each box was filled with 3 mL of phosphate buffered saline (PBS) which was sufficient to
fully submerge the PEG coupons. Aerial photographs of the PEG coupons were taken every 30
seconds or 1 minute and were analyzed using ImageJ to calculate the surface area of PEG
coupon remaining (Figure 17). Figure 18 shows a graphical representation of the rate of PEG
dissolution for each of the four types of PEG used in the experiment. While PEG 1,000 dissolved
by a rate of 9.5% per minute, PEG 3,350 dissolved almost half as slowly. PEG 8,000 and 14,000
were deemed to dissolve too slowly to be practical in a surgical setting. PEG 3,350 was used for
all future experiments and surgical insertions and was robust enough to withstand transport
without wilting.

Figure 17: Photographs of PEG 3,350 coupon at (a) 0 min, (b) 3 min, (c) 6 min, and (d) 9 min
post addition of the PBS solution. The coupon shrinks in diameter over time and the spreading of
blue dye throughout the PBS solution indicates dissolution of the dye-laden PEG. Scale bar is 5
mm.
33

Figure 18: Dissolution of PEG based on MW. PEG 1,000 dissolved the quickest at a rate of -
9.5% per minute. All linear fits had a Pearson coefficient of 0.99. Standard deviations across
three coupons for each MW are included, N = 3.
Testing of the dissolvable brace
In initial testing, the buckling force was directly measured for sham devices with and without
PEG encapsulation. Devices were driven into a metal plate using a computer controlled stepper
motor at 1 mm/s until they buckled, while load was measured with a 50 g load cell (Omega,
LCFA-50g, Norwalk, CT). Buckling force was measured for arrays of 8 probes which were then
reduced to arrays of 6, 4, 2, and 1 probes with and without PEG braces. Initial PEG braces that
only covered half of the probe length were used for these experiments.
A typical buckling curve is represented by Figure 19. Once the motor translates the probe
tips to the point at which they contact the metal plate the force experienced by the load cell
shoots up. Buckling begins when this force plateaus and ends in the collapse of the probes at
which point the force drops quickly back to zero.
Buckling curves were analyzed for each of the parameters described above. Buckling
force threshold averages and standard deviations across 3-6 devices are included in Figure 20.
The full array of eight probes, unbraced, buckled under a mean load of 1.26 ± 0.19 mN (N = 6).
34

The buckling threshold of a single, two, or four probe arrays could not be accurately recorded
within the resolution of our load cell. As expected, this is well below the loading anticipated
during insertion [26]. Full arrays supported with a PEG brace had an average buckling force
threshold of 4.24 ± 0.24 mN (N = 6), and for single braced-shanks buckling force threshold was
measured directly as 2.14 ± 0.25 mN (N= 3). This data supports the utility of a PEG brace for
increasing the buckling force threshold of hippocampal probes thereby enabling them to
penetrate brain phantom or brain tissue without buckling. According to Euler’s beam equation,
halving the length of probes should increase their buckling force threshold by four times. The
average increase in buckling force threshold between arrays without and with PEG braces that
reinforced half of the probe length was 3.5 times. Although this is close to the predicted value,
this difference could be attributed to the fact that our probes are not truly uniform beams even
though they are modelled as such. Because of inertia of a column’s mass, a certain amount of
time is required for columns to buckle [29] and so the buckling force threshold measured for the
above arrays may have a dependence on compression speed. Future experiments could be
improved by setting the stepper motor to the same speed at which the array is implanted during
surgery (10 µm/s).

Figure 19: Typical buckling curve. The force starts of at zero. The probe tips are translated closer
and closer to the metal plate as time passes. When the probe tips contact the metal plate the force
exerted on the load cell shoots up. Buckling begins when the force begins to level off and ends in
the collapse of the probes at which point the force drops quickly back to zero.

35

Figure 20: Average (and standard deviations) of buckling force thresholds measured for sham
hippocampal arrays of 8, 6, 4, 2, and 1 probes with and without PEG braces, pushed against a
metal plate at a speed of 1 mm/s. Buckling force thresholds of 1, 2, and 4 probe arrays without
PEG braces fell beneath the resolution of the load cell and could not be measured. (N= 3 or 6)

In preparation for in vivo implantation, benchtop testing was conducted using brain
phantoms (0.5% agarose gel solution) to simulate the consistency and density of brain tissue.
Agarose gels of concentrations near this range have been shown to represent the mechanical
stiffness of both bovine and porcine brain matter in vivo [30, 31] and through extrapolation can
be shown to represent an even higher bar of stiffness with regards to adult rat cortex and
hippocampi [32], [33]. PEG encapsulated sham arrays were inserted into gel samples using a
stereotaxic stage and the following technique: the exposed tips of the probes were manually
driven slowly (~0.1 mm/s) into the agarose up to an approximate depth of 2.75 mm, at which
point the PEG brace was dissolved with saline, and the remainder of the probe advanced to an
ultimate depth of 5.5 mm. Some sham arrays were encapsulated in a stepped PEG brace (Figure
21) and inserted incrementally into the agarose. The array was inserted to a depth of 2.75 mm,
the extent of the exposed probes. The first step of the PEG brace was dissolved and the newly
exposed shank length was inserted into the agarose. This same step was repeated with the
remaining step of the PEG brace. Probe arrays were monitored during implantation for signs of
buckling. > 9 arrays inserted without buckling with various degrees of track straightness. Some
36

probes shanks drew closer together during the insertion causing the center-to-center probe
spacing to vary across the array (Figure 21). These variables were not able to be correlated to
either style of probe brace.

Figure 21: Representative sham array insertion into 0.5% agarose brain phantom. (a) Top view of
sham array with stepped PEG brace design intended for 2-stage insertion. (b) Side view of sham
probes protruding from PEG brace. (c) Front view of entire length of sham array inserted into
agarose. Some shanks drew closer to each other during insertion. (d) Side view of sham array in
agarose that inserted without buckling, but at a slight angle. Scale bar is 1 mm.

In subsequent in vivo rat studies, sham arrays encapsulated in the PEG brace were
inserted into a male, Sprague-Dawley (SD) rat weighing between 300 to 400 g. Animal
experiments were reviewed and approved by the Institutional Animal Care and Use Committee
(IACUC) as well as the Department of Animal Resources of the University of Southern
California (USC). During surgery, the superficial cortex above the dorsal hippocampi was
exposed by removing a 2×4 mm piece of cranium from each implantation site and carefully
removing dura with forceps. Three screw holes were drilled into the cranium to serve as a port
for anchor screws (Figure 22). The animal was anesthetized throughout the procedure as
confirmed by negative toe pinch withdrawal reflexes. The neural probe array was then inserted
using the technique described above at a speed of 10 µm/s, then the subject was perfused with
paraformaldehyde and sacrificed immediately post-implantation. In one such implantation, a
sham hippocampal array was inserted into both the right and left hippocampus—one was sliced
in the coronal plane, along the length of the probe shanks, and the other was sliced in the
transverse plane, to highlight probe cross-sections (Figure 23). Successful in vivo placement of
37

the probe array in deep hippocampal layers was confirmed via hematoxylin and eosin staining of
these histological slices. The stab wounds match probe cross-sectional dimensions and minimal
damage to surrounding tissue was observed. This is a significant advantage over current insertion
techniques which rely on bulky insertion aides add dramatically to the physical footprint of the
neural probe.

Figure 22: Surgical implantation of sham array into rat hippocampus.

Figure 23: Histological slices from sham array implantation into rat hippocampi. Slices stained
with hematoxylin and eosin. (a) Coronal slice and (b & c) transverse slices taken at 2.2 and 2.5
mm from the brain surface showing probe shanks and cross-sections.
Fabrication of functional arrays
A few challenges were encountered upon fabrication of functional arrays with the thin-
film platinum conductor layer included. Some of these difficulties affected only the fabrication
38

procedure itself while others, such as probe curling due to compressive stresses in the platinum,
had wider-reaching ramifications.
The procedure for fabricating fully functional arrays follows and is similar to the above
fabrication scheme for the creation of sham arrays. To start, a base layer of Parylene C that was
10 µm thick was deposited by CVD onto a dehydrated, prime, 4” silicon wafer. A thin coat (~
1.5 µm) of AZ5214 (Integrated Micro Materials, Argyle, TX), an image-reversal photoresist, was
lithographically patterned (step 1: 8 s, 500 rpm, step 2: 45 s, 2,000 rpm) to define the metal
features of the array that would be coated with platinum, and a short O2 plasma treatment (100
W, 100 mTorr, 60 seconds) was used to clean and roughen the surface prior to metal deposition.
Details of the neural interface with electrodes designed to specifically target hippocampal layers
of interest are included in the following chapter. 2,000 Å of platinum was deposited via
sputtering (LGA Thin Films, Santa Clara, CA) or through electron-beam physical vapor
deposition (e-beam; Caltech, CA). Excess metal was lifted off in acetone heated to 50 °C with
gentle brushing used to clear the small metal traces on sputtered devices that were difficult to lift
off. A Parylene insulation layer, chosen to be 10, 14, or 18 µm thick, was deposited via CVD.
After the body of the arrays were completed, the remaining steps consisted of
lithographically masking and etching out the arrays as well as exposing the metal electrode sites
and contact pads. The first etch mask was designed to expose only the outlines of each array
while the second mask exposed both array outlines and the Parylene above electrodes and
contact pads. Exposing the electrodes on the second etch step ensured electrodes were clean after
the last processing step and obviated the need to clean a photoresist mask off an already exposed
Pt surface. Since the first etch mask need only protect an outline etch that was 10 µm deep, a 15
µm thick layer of AZ4620 as an etch mask and was followed by a subsequent DRIE etch (~ 160
loops, split into 25 loop cycles). Remaining photoresist was removed in acetone, isopropyl
alcohol (IPA), and deionized (DI) water rinses and soaks. The subsequent etching mask needed
to accommodate for etch depths that spanned 10-18 µm, so a double-spin recipe of AZ4620 (two
spins separated by a softbake, both spins consisting of step 1: 8 s, 500 rpm, step 2: 45 s, 2,000
rpm) was used as a mask for the last DRIE step which exposed electrodes and contact pads and
completed the array outline. This final resist mask was stripped through rinsing with acetone,
IPA, and DI water and devices soaked in DI water were released by using tweezers to gently peel
the device away from the native oxide layer of the silicon substrate. Released devices were
39

soaked for 10 min in baths of acetone, IPA, and DI water. Parylene contact pads were supported
with 0.002” thick polyether ether ketone (PEEK) tape with a 2.3 mil thick acrylic adhesive (CS
Hyde Co., Lake Villa, IL) which enabled insertion into the ZIF connector for electrical
packaging. Photographs of fully fabricated arrays are presented in Figure 24.

Figure 24: Fully fabricated hippocampal array. Photograph of (a) entire array, (b) cable and
probe shanks, and (c) close-up of electrodes and traces on individual probe shanks.
Fracturing of e-beam deposited Pt
One challenge that was encountered during the deposition of thin film Pt was the
presence of cracks or fractures in the exposed features not covered by the AZ5214 photoresist
mask. This was especially surprising because these cracks appeared abruptly sometime after
USC’s e-beam deposition tool had been repaired and metal deposited prior to the repair date was
smooth. This change was hypothesized to be due to excessive heating during the deposition
process as the platen that supported the wafers during deposition felt hot to the touch upon
removal of the wafers from the e-beam machine. Various techniques to aid in wafer cooling were
implemented without success. These include lowering the Pt deposition rate from 3 A/s to 0.75
A/s, lowering the power during deposition, and splitting the 2,000 Å of Pt into four runs of 500
Å instead of three runs of 666 Å with cooling periods in between each run. The thickness of the
Parylene base layer was reduced from 10 µm as well, in attempt to evaluate whether this thick
40

base coat of insulative Parylene was reducing wafer cooling, but a thinner base of Parylene did
not help either. Figure 25 contains images of the contact pads and electrodes of the hippocampal
array taken after an e-beam run performed with a new crucible, Pt, turret, and blocks of
aluminum resting upon the platen as a heat sink. There was some minor attenuation of cracking
in this run, however, fractures in the Pt are still visible.
Concurrent to the above e-beam experimentation, metal deposition was outsourced to a
company called LGA Thin Films which sputter coated the 2,000 Å of Pt in three runs of 666 Å
with 20 min cooling periods between each step. This sputtered metal film was continuous and
did not crack. Sputtering, however, is known for its more conformal metal coverage which tends
to coat the sidewalls of photoresist features complicating lift-off. Lift-off on sputtered wafers
required hours of soaking in 50 °C acetone and vigorous swabbing of the small features of the
array as will be described in the next section.
Eventually, it was discovered that e-beam evaporators with longer throw distances (22
inches at UCLA, 28 inches at Caltech, compared to 14 inches at USC) could deposit Pt at the
above parameters without metal cracking. Since e-beam evaporation only coats surfaces within
the “line of sight,” it is preferred for lift-off processes. Lift-off of metal deposited via e-beam
was significantly easier than sputtered metal and e-beam was used to deposit Pt for all future
devices.

Figure 25: Cracked e-beam deposited Pt imaged across (a & b) array contact pads and (c )
electrodes. This was the mildest case of cracking seen.
Lift-off difficulties
As briefly described above, metal deposited via sputtering severely complicated the lift-
off step. The AZ5214 lithographic photoresist mask that defined the metal features of the wafer
41

returned from sputtering with widespread rippling and wrinkling (Figure 26 (a)). Lift-off on
these wafers took ~6 hours. It required soaking in acetone heated to 50 °C and vigorous
swabbing of the metal features. Swabbing removed most of the Pt covered photoresist, yet many
traces were imperfect due to metal shorts or scrubbing damage that resulted in removal of the
actual metal features of interest.
Discussions of this difficulty with the photoresist manufacturer yielded possible sources
of these difficulties—encapsulation of the side-walls of the photoresist features with Pt and/or
excess heating of the photoresist during sputter deposition. It was hypothesized that increasing
the flood exposure step of the image reversal process could both improve the negative profile of
the resist features (thereby allowing portions of the sidewall to escape Pt encapsulation) and
increase the resilience of the photoresist mask to any heating experienced during sputtering. The
flood exposure step not only solubilizes the photoresist above the metal features so that it can be
developed away, but purportedly aids in the cross-linking of the photoresist mask as well. To
explore this, the flood exposure step was changed from 225 mJ/cm
2
to 1,000 mJ/cm
2
—the
recommended flood exposure for this image reversal resist. New wafers with this processing
change were sent out to be sputter coated by LGA. Upon return, milder wrinkling was visible
and lift-off processing time was halved.
Although the increase in flood exposure fluence seemed to aid in lift-off, a few puzzling
phenomena remained. Firstly, temperature measurements of the LGA sputtering run revealed
that the wafer only reached temperatures of 71-77 °C. It is curious that this low temperature
process could cause thermal rippling of the photoresist, especially because the softbake and
image reversal bake temperatures exceed this value (90 and 110 °C respectively) yet did not
reveal wrinkling. This seems to indicate that rippling might be caused by the physical
bombardment of photoresist with the Pt atoms. However, the placing a wafer with the metal
photoresist mask on a hotplate at 110 °C, revealed that thermal treatment replicates these
wrinkles—even with a wafer patterned with an increased flood exposure (Figure 26 (b)). It is
possible that the source of the rippling was both thermal and physical in nature. Interestingly, no
wrinkling was seen on wafers treated with e-beam Pt deposition (Figure 26 (c)).
42

Figure 26: Rippling in AZ5214 photoresist. 5214 after (a) sputter Pt deposition (b) heat treatment
on hot plate at 110 ° C for 3 minutes and (c) e-beam Pt deposition. Sputtering causes severe
wrinkling in the photoresist lying beneath the sputtered platinum, heating reproduces this
phenomenon in photoresist, and e-beam deposition of metal covers a smooth, undisturbed
photoresist surface. Scale bar 50 µm.
The results from a third sputtering run were also puzzling, this time again with
photoresist exposed to 1,000 mJ/cm
2
. The only change to the photoresist on this run was an
increase in softbake time from 70 to 75 seconds, however it is difficult to believe that this slight
modification could negatively impact the reaction of photoresist to metal sputtering. The
sputtering parameters remained the same. However, this batch returned with severe wrinkling.
Lift-off in 50 °C heated acetone failed as did attempted lift-off in heated AZ Kwik Strip, PG
remover, and developer. Soaking for such extended periods actually caused bubbles to form on
the backside of the wafers, between the Parylene and the silicon, causing the wafers to float to
the solution surface and destructive Parylene delamination. One wafer was even soaked in
Pirhana, but also failed to lift-off. At this point, sputtering was no longer considered useful for Pt
deposition of the small electrodes and traces of the hippocampal array.
A calibration run performed by LGA was performed in order to determine if magnetron
sputtering (instead of the typical radio-frequency sputtering) could further lower deposition
temperature and perhaps the stresses in the thin-film Pt layer (which were -511 MPa for the
previous run). It was hypothesized that a single deposition of 2,000 Å of metal would result in
less stress than four, thinner runs of 500 Å. This calibration run was aborted when it was
discovered the temperature reached 93-107 °C. A magnetron run split into four deposition
periods resulted in a reduced temperature of 60-65 °C, yet its stress was overwhelming (819 MPa
in the tensile direction).
43

All future metal deposition steps were performed through e-beam evaporation and
resulted in metal that lifted-off within tens of minutes, with minimal to no scrubbing required, in
acetone heated to 50 °C.
Non-uniform lithographic development
Another challenge experienced during metal lithography was the development of the
image reversal photoresist itself. Once it was uncovered that an image reversal bake temperature
of 110 °C allowed for proper development of the small electrodes and traces, it became evident
that there was a circular dependence to device clearing. If development time was set based on
devices on the perimeter of the wafer, the center devices were underdeveloped. When
development time was set based on devices at the center of the wafer, devices in the wafer
periphery grew overdeveloped—as witnessed by “lifting off” of the lines of photoresist between
the electrical trace features in thin filaments and dark erosion of the photoresist mask itself
(Figure 27).
Two experiments were performed determine whether gradations in heating of the hot
plate or intensity of the UV lamp could be responsible for this circular phenomenon. The surface
of the hotplate was measured with an IR gun and only slight temperature changes of ± 1 °C were
present. If a gradient in temperature was responsible for the results seen above, then it was
hypothesized that raising the image reversal bake temperature by a few degrees would enable
cross-linking of the photoresist mask in the periphery. Unfortunately, this experiment was
inconclusive. Increasing the image reversal bake from 110 °C to 118 °C led to appropriate
development at the periphery, but devices that appeared “overbaked” and did not develop
appropriately in the center of the wafer. We could not say with certainty that this slight gradient
in the hotplate temperature was causing this phenomenon. In a second test, increasing the fluence
of UV light beyond the 37.5 mJ/cm
2
exposure on the first exposure step did not help prevent
overdevelopment of features on the periphery of the device—even though they were now
exposed to a greater UV dose. This seemed to eliminate the UV lamp as a source of these
difficulties.
Upon Dektak profilometry measurements of the AZ5214 photoresist mask, it was
discovered that while the average thickness of the layer was ~ 1.8 µm, the edges of the wafer
44

were about 0.1 µm thinner. It is possible that this 5% difference in thickness could cause the non-
uniform development pattern witnessed.
Serendipitously, it was discovered that increasing the softbake from 70 to 75 seconds in
duration eliminated this problem entirely and enabled uniform development across the entire
wafer. Perhaps an increased softbake length creates a photoresist mask that is more resilient to
developer.

Figure 27: Overdevelopment at wafer periphery. (a) Photoresist between traces leading away
from electrode sites lift off from wafer completely forming thin filaments. (b) Thin filaments and
dark erosion visible. Scale bar is 100 µm. Increasing the length of the image reversal softbake
from 70 to 75 seconds eliminated non-uniform development across wafer and prevented
overdevelopment of devices in wafer periphery.
Parylene “spherules”
After the deposition of the Parylene insulation layer, odd “spherule” shapes inside the
insulation layer became evident. This manifested macroscopically as “milky Parylene”—
Parylene with a hazy appearance and microscopically as a bubbly scum. This phenomenon
occurred in wafers that were treated with an RIE O2 descum (100 W, 100 mTorr, 1 minute) after
lift-off and immediately prior to Parylene deposition as well as wafers that were not cleaned in
this manner. It remains uncertain if these spherules represent the presence of uncleaved Parylene
dimers or the absence of Parylene in the shape of holes. The number of spherules seemed to
increase when wafers were vigorously scrubbed during metal lift-off or in cases where fencing of
the metal features was present. It is unclear whether these “spherules” interfere with adhesion
between the Parylene and metal layers of these devices and whether they impact the longevity of
45

trace insulation in soaking conditions. However, a mention to this phenomenon in made in
literature [34] where it is posited that the “spherules” described here perhaps represent Parylene
monomers that bond to each other in the gas phase prior to deposition due to a high volume to
surface area of ratio of the deposition chamber to the samples to be coated. This same study
indicates that the coating quality suffers when this phenomenon appears—providing impetus for
further study and analysis.

Figure 28: Parylene spherules that formed during CVD of the insulation layer of Parylene. After
lift-off, the wafer surface was cleaned via a RIE O2 descum (100 W, 100 mTorr, 1 minute).
Spherules form throughout the entire wafer but seem to collect the most densely around metal
edges defined through lift-off.
Stress induced array curvature
Upon release of fully fabricated devices, that contained a metal layer, from silicon wafer
carriers it was discovered that stresses in the trilayer, particularly in the thin film Pt layer, caused
overt curvature of the arrays. Curvature was not observed in sham arrays and only present in
arrays containing metal features. Intrinsic stress in the 2,000 Å thin film sputtered film was
calculated to be -511 Mpa for sputtered metal and was visible by eye in the form of
wrinkling/ripples throughout the metal (Figure 30). The direction of stress can be compressive or
tensile. Compressive stresses have negative stress values and curve away from the exposed
electrodes and pads of the hippocampal arrays. Tensile stresses have positive stress values and
cause the array to curl toward the exposed metal surfaces. While this curvature can be harnessed
productively for some research applications, such as in the creation of a curved cochlear array
made from Parylene [35], probe curving made it difficult to apply PEG braces as it was
46

challenging to sandwich the curved array in between sheets of the PDMS mold. Array curvature
also caused concerns about the ability of a curved array to reach the small hippocampal target.

Figure 29: Fully fabricated array (with eight probe shanks) that curves due to compressive stress
in the thin film Pt. This curvature complicated the application of the PEG brace and was
predicted to interfere with hippocampal targeting during surgical placement. (a) High levels of
compressive stress in Pt deposited through sputtering causes intense probe curvature, whereas (b)
metal deposited via e-beam vapor deposition had less stress and probes were less curled.

Figure 30: (a) Electrode traces in hippocampal array post Pt sputtering. -511 MPa of compressive
stress were measured by the vendor and result in the wrinkled appearance of the metal
surrounding the features that will eventually undergo liftoff. This compressive stress causes
devices to curl away from the exposed side of the Parylene upon release from the silicon carrier
wafer. (b) Electrode traces in hippocampal array with metal smoothly deposited via e-beam and
less compressive stress.

In order to counteract this curvature and to straighten devices, stress balancing of the
trilayer was explored as well as post-processsing the device with heat and pressure [36] in a
process called annealing. This is a heating process aimed to increase Parylene chain
entanglement and reduce stress assisted in straightening the devices and in adding counter,
47

tensile stresses [37] to the system. However, this was insufficient to fully straighten arrays with
sputtered metal. In order to straighten hippocampal arrays, the Parylene insulation layer
thickness was increased, as thicker Parylene has higher magnitudes of tensile stress, and an
annealing step was added to the fabrication steps. Combinations of Parylene insulation thickness
and the presence or absence of an anneal step were used to create the straightest probes.
In initial annealing experiments, devices were thermoformed directly on the silicon wafer
carrier or released and annealed while being sandwiched between two glass slides. Arrays were
placed in an oven, vacuum purged three times with nitrogen gas, and then annealed, under
vacuum for 48 hours at 200 °C. Unfortunately, the Parylene was permanently fused to the silicon
and glass and devices could not be recovered (Figure 31). Instead, devices in the next
experimental group were sandwiched between two 0.33 mm thick Teflon sheets which were then
placed in between two glass slides and held together with four mini binder clips. This Teflon
layer prevented the Parylene arrays from irreversibly adhering to the glass slides during the
annealing treatment however, the arrays took on the surface roughness of the Teflon. Annealing
added tensile stress to counter the compressive stress of the Pt layer and was quantified as
described below.

Figure 31: Arrays thermoformed prior to release from silicon wafer cracked and were
permanently adhered to silicon. (a) Cracks in bulk Parylene due to mismatch in thermal
expansion coefficients between silicon and Parylene, (b) cracks propagating from pointed probe
tips, and (c) damage to Parylene array upon removal of device from silicon backing.
Stress balancing testing
The radius of curvature of arrays processed with various parameters (Pt deposited via
sputtering or e-beam; 10, 14, or 18 µm thick Parylene insulation layers; and annealed vs.
nonannealed devices) were measured in order to quantify the degree of array curvature. Arrays
were imaged from their sides, and ImageJ (NIH) was used to extrapolate the radius of curvature
48

(ROC) of each array. Calculations were made based on the segment, L, drawn across the curved
probes which represent the arc of interest, and θ, the angle defined between segment L and the
point where the perpendicular bisector to L intersects the arc. The ROC was calculated according
to the schematic shown in Figure 32, through Eq. 4, for arrays pre and post annealing treatment.
The higher the ROC, the straighter the array.
𝑅 =
L
2sin (2𝜃 )

Eq. 4

Figure 32: Schematic of radius of curvature calculation overlaying photograph of array profile.
Black line mimics side view of curved array. Segment L was drawn across the arc created by the
curved array. θ, the angle defined between line segment L and its perpendicular bisector, was
used to calculate R.

Table 2 presents a summary of how these processing parameters affected array curvature.
From the table, it is clear that both annealing and increasing the thickness of the Parylene
insulation layer added tensile stress to the trilayer and helped to combat the compressive stresses
intrinsic to the thin film Pt. Arrays with sputtered Pt that were annealed with 14 µm thick
Parylene insulation layer as well as annealed arrays with e-beamed Pt were found to have the
largest radius of curvatures. These arrays were deemed straightest and were used for future in
vivo recording studies.
49

Table 2: Effects of various processing parameters on array curvature (in mm).

In addition to increasing the thickness of the Parylene insulation layer, experiments were
performed to evaluate whether thinning the base layer of Parylene could potentially reduce the
overall compressive stress of the trilayer. Sputtered arrays with 10 µm thick base and insulation
layers were mounted with their base Parylene exposed onto Parylene covered glass slides and
exposed to different time lengths of reactive ion etching with oxygen plasma. AZ4400
(Integrated Micro Materials, Argyle, TX) was spun onto the slides (step 1: 5 s, 500 rpm, step 2:
10 s, 2,000 rpm) and arrays were gently placed on the photoresist with the base Parylene layer
facing upwards. Surface tension automatically drew the arrays flat atop the photoresist layer.
Arrays were baked onto the photoresist at 90 °C for four minutes in order to complete mounting
and a small window of photoresist was wiped away with an acetone soaked cotton swab to allow
monitoring of etch rates. Samples were then exposed to O2 plasma (200W, 200 mTorr) for 4, 8,
and 12 minutes. The average Parylene etch rate was 0.135 µm/min and each time period
corresponded to etching away 0.5, 1.1, 1.8 µm of Parylene thickness from the base layer. All
three etched arrays had similar reductions in curvature (Figure 33) which seemed to indicate that
the heating, not the etching, was the element responsible for array straightening probably through
50

a mechanisms similar to that proposed for annealing. Since this treatment was insufficient to
fully straighten probe shanks, it was not used as a post-fabrication treatment to straighten probes.

Figure 33: O2 plasma etching of base layer of Parylene of arrays for 4, 8, and 12 min. Side views
of arrays compared to control 10 µm/ 10 µm array that did not undergo mounting or etching.
Summary
In summary, we presented an overview of the fabrication and fabrication-induced
mechanical challenges experienced while designing a flexible neural probe array to target a
deep-brain region of interest. We developed a non-invasive strategy to reinforce the axial
stiffness of the flexible probe shanks during implantation that has met success in both in vitro
and in vivo experiments through the use of a dissolvable PEG brace. We have found the
appropriate post-fabrication treatment and probe design parameters that enable balancing of
residual stresses in the thin film Pt layer to achieve neural probes with straight shanks. Arrays
should ideally be fabricated with e-beam Pt deposition that induces lower amounts of stress and
curvature and probes can be straightened through a simple, post-processing, annealing treatment.
The fine-tuning of both the implantation and straightness of the probes is essential for the
targeting of thin, rat hippocampal layers that are only 40 to 144 µm thick [38]. Both of these
developments were necessary to pave the way for rat implantations with functional recording
arrays, as will be described in the following chapter.
51

References
[1] J. A. Thornton and D. Hoffman, "Stress-related effects in thin films," Thin solid films,
vol. 171, pp. 5-31, 1989.
[2] L. W. Tien, F. Wu, M. D. Tang-Schomer, E. Yoon, F. G. Omenetto, and D. L. Kaplan,
"Silk as a Multifunctional Biomaterial Substrate for Reduced Glial Scarring around
Brain-Penetrating Electrodes," Advanced Functional Materials, vol. 23, pp. 3185-3193,
2013.
[3] F. Wu, M. Im, and E. Yoon, "A flexible fish-bone-shaped neural probe strengthened by
biodegradable silk coating for enhanced biocompatibility," in Solid-State Sensors,
Actuators and Microsystems Conference (TRANSDUCERS), 2011 16th International,
2011, pp. 966-969.
[4] A. Lecomte, V. Castagnola, E. Descamps, L. Dahan, M. C. Blatché, T. M. Dinis, et al.,
"Silk and PEG as means to stiffen a parylene probe for insertion in the brain: toward a
double time-scale tool for local drug delivery," Journal of Micromechanics and
Microengineering, vol. 25, p. 125003, 2015.
[5] T. Suzuki, K. Mabuchi, and S. Takeuchi, "A 3D flexible parylene probe array for
multichannel neural recording," in Neural Engineering, 2003. Conference Proceedings.
First International IEEE EMBS Conference on, 2003, pp. 154-156.
[6] C.-H. Chen, "Three-dimensional flexible microprobe for recording the neural signal,"
Journal of Micro/Nanolithography, MEMS, and MOEMS, vol. 9, p. 031007, 2010.
[7] M. C. Lo, S. Wang, S. Singh, V. B. Damodaran, H. M. Kaplan, J. Kohn, et al., "Coating
flexible probes with an ultra fast degrading polymer to aid in tissue insertion," Biomed
Microdevices, vol. 17, p. 34, Apr 2015.
[8] D. Lewitus, K. L. Smith, W. Shain, and J. Kohn, "Ultrafast resorbing polymers for use as
carriers for cortical neural probes," Acta biomaterialia, vol. 7, pp. 2483-2491, 2011.
[9] P. J. Gilgunn, R. Khilwani, T. D. Y. Kozai, D. J. Weber, X. T. Cui, G. Erdos, et al., "An
ultra-compliant, scalable neural probe with molded biodissolvable delivery vehicle," pp.
56-59, 2012.
[10] J. Agorelius, F. Tsanakalis, A. Friberg, P. T. Thorbergsson, L. M. Pettersson, and J.
Schouenborg, "An array of highly flexible electrodes with a tailored configuration locked
by gelatin during implantation-initial evaluation in cortex cerebri of awake rats," Front
Neurosci, vol. 9, p. 331, 2015-September-25 2015.
[11] G. Lind, C. E. Linsmeier, J. Thelin, and J. Schouenborg, "Gelatine-embedded electrodes-
-a novel biocompatible vehicle allowing implantation of highly flexible microelectrodes,"
J Neural Eng, vol. 7, p. 046005, Aug 2010.
[12] C. Hassler, T. Boretius, and T. Stieglitz, "Polymers for neural implants," Journal of
Polymer Science Part B: Polymer Physics, vol. 49, pp. 18-33, 2011.
[13] A. Singh, H. Zhu, and J. He, "Improving mechanical stiffness of coated
benzocyclobutene (BCB) based neural implant," in Engineering in Medicine and Biology
Society, 2004. IEMBS'04. 26th Annual International Conference of the IEEE, 2004, pp.
4298-4301.
[14] Z. Xiang, S.-C. Yen, N. Xue, T. Sun, W. M. Tsang, S. Zhang, et al., "Ultra-thin flexible
polyimide neural probe embedded in a dissolvable maltose-coated microneedle," Journal
of Micromechanics and Microengineering, vol. 24, p. 065015, 2014.
52

[15] S. L. Bourke and J. Kohn, "Polymers derived from the amino acid l-tyrosine:
polycarbonates, polyarylates and copolymers with poly(ethylene glycol)," Advanced
Drug Delivery Reviews, vol. 55, pp. 447-466, 2003.
[16] S. Felix, K. Shah, D. George, V. Tolosa, A. Tooker, H. Sheth, et al., "Removable silicon
insertion stiffeners for neural probes using polyethylene glycol as a biodissolvable
adhesive," in Engineering in Medicine and Biology Society (EMBC), 2012 Annual
International Conference of the IEEE, 2012, pp. 871-874.
[17] H. S. Sohal, A. Jackson, R. Jackson, G. J. Clowry, K. Vassilevski, A. O'Neill, et al., "The
sinusoidal probe: a new approach to improve electrode longevity," Front Neuroeng, vol.
7, p. 10, 2014.
[18] J. T. 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, vol. 13, pp. 554-61, Feb 21 2013.
[19] T. D. Kozai and D. R. Kipke, "Insertion shuttle with carboxyl terminated self-assembled
monolayer coatings for implanting flexible polymer neural probes in the brain," J
Neurosci Methods, vol. 184, pp. 199-205, Nov 15 2009.
[20] D. R. Kipke, D. S. Pellinen, and R. J. Vetter, "Advanced neural implants using thin-film
polymers," in Circuits and Systems, 2002. ISCAS 2002. IEEE International Symposium
on, 2002, pp. IV-173-IV-176 vol. 4.
[21] R. Biran, D. C. Martin, and P. A. Tresco, "Neuronal cell loss accompanies the brain
tissue response to chronically implanted silicon microelectrode arrays," Experimental
neurology, vol. 195, pp. 115-126, 2005.
[22] E. Meng, P.-Y. Li, and Y.-C. Tai, "Plasma removal of Parylene C," Journal of
Micromechanics and Microengineering, vol. 18, p. 045004, 2008.
[23] S. P. Marshall, W.-C. Lin, P. R. Patel, A. J. Shih, and C. A. Chestek, "Effects of
geometry and material on the insertion oi very small neural electrode," in Engineering in
Medicine and Biology Society (EMBC), 2016 IEEE 38th Annual International
Conference of the, 2016, pp. 2784-2788.
[24] S. Yagi, S. Yamagiwa, Y. Kubota, H. Sawahata, R. Numano, T. Imashioya, et al.,
"Dissolvable base scaffolds allow tissue penetration of high-aspect-ratio flexible
microneedles," Adv Healthc Mater, vol. 4, pp. 1949-55, Sep 16 2015.
[25] P. R. Patel, K. Na, H. Zhang, T. D. Kozai, N. A. Kotov, E. Yoon, et al., "Insertion of
linear 8.4 μm diameter 16 channel carbon fiber electrode arrays for single unit
recordings," Journal of neural engineering, vol. 12, p. 046009, 2015.
[26] B. Wester, R. Lee, and M. LaPlaca, "Development and characterization of in vivo flexible
electrodes compatible with large tissue displacements," Journal of neural engineering,
vol. 6, p. 024002, 2009.
[27] N. A. Alcantar, E. S. Aydil, and J. N. Israelachvili, "Polyethylene glycol coated
biocompatible surfaces," Journal of biomedical materials research, vol. 51, pp. 343-351,
2000.
[28] M. Bera, A. Rivaton, C. Gandon, and J. Gardette, "Photooxidation of poly (para-
xylylene)," European Polymer Journal, vol. 36, pp. 1753-1764, 2000.
[29] V. A. Kuzkin and M. M. Dannert, "The maximum force in a column under constant
speed compression," arXiv preprint arXiv:1511.01358, 2015.
53

[30] F. Pervin and W. W. Chen, "Dynamic mechanical response of bovine gray matter and
white matter brain tissues under compression," J Biomech, vol. 42, pp. 731-5, Apr 16
2009.
[31] F. Pervin and W. W. Chen, "Mechanically similar gel simulants for brain tissues," in
Dynamic Behavior of Materials, Volume 1, ed: Springer, 2011, pp. 9-13.
[32] B. S. Elkin, A. Ilankovan, and B. Morrison, 3rd, "Age-dependent regional mechanical
properties of the rat hippocampus and cortex," J Biomech Eng, vol. 132, p. 011010, Jan
2010.
[33] V. Normand, D. L. Lootens, E. Amici, K. P. Plucknett, and P. Aymard, "New insight into
agarose gel mechanical properties," Biomacromolecules, vol. 1, pp. 730-738, 2000.
[34] G. Loeb, R. Peck, and J. Martyniuk, "Toward the ultimate metal microelectrode," Journal
of neuroscience methods, vol. 63, pp. 175-183, 1995.
[35] A. C. Johnson and K. D. Wise, "A self-curling monolithically-backed active high-density
cochlear electrode array," in Micro Electro Mechanical Systems (MEMS), 2012 IEEE
25th International Conference on, 2012, pp. 914-917.
[36] B. J. Kim, E. P. Washabaugh, and E. Meng, "Annealing effects on flexible multi-layered
parylene-based sensors," in 2014 IEEE 27th International Conference on Micro Electro
Mechanical Systems (MEMS), 2014, pp. 825-828.
[37] S. Dabral, J. Van Etten, X. Zhang, C. Apblett, G.-R. Yang, P. Ficalora, et al., "Stress in
thermally annealed parylene films," Journal of electronic materials, vol. 21, pp. 989-994,
1992.
[38] P. Andersen, R. Morris, D. Amaral, T. Bliss, and J. OKeefe, The hippocampus book:
Oxford University Press, 2006.

54

Targeting the hippocampus
Penetrating neural probes are the sole electrical recording technology that have a high-
enough spatial resolution to record from individual neurons. Their implantation within the brain,
at positions that lie close to the neurons of interest, make them especially essential for targeting
small structures that lie deep within the brain. One such thin, deep-brain target of interest to
researchers is the hippocampus. The hippocampus is responsible for the formation of declarative
memories-- conscious memories that can be described in words. This is in contrast to brain
structures that encode procedural memory, the unconscious recollection and performance of
learned skill sets. This essential role of the hippocampus in memory makes it a brain structure of
utmost value, without which a person no longer retains the memories and experiences that make
him the unique individual that he is. Damage to the hippocampus, in the form of traumatic brain
injury or disease processes, like Alzheimer’s, can strip a person of his identity and ability to
perform activities of daily life.
The hippocampus is a c-shaped layered structure with three subdivisions: the cornu
ammonis (CA) 1, CA3, and dentate gyrus (DG). Neural information from the entorhinal cortex
enters the hippocampus via the DG and next flows to the CA3 and finally the CA1 of the one-
way, tri-synaptic hippocampal circuit (Figure 34). Pyramidal cells in these three divisions of the
hippocampus communicate through complex spikes, a burst of 2-6 action potentials of
decreasing amplitudes separated by ≤ 5 ms. This electronic signature helps confirm placement of
a recording device in the hippocampus.
The thin, layered shape of each region of the hippocampus complicates electrode
targeting and makes it difficult to achieve simultaneous recordings from the layers of interest.
Simultaneous recordings from the different layers are necessary not only to understand how
precisely the hippocampus functions in encoding memory, but are also necessary to pave the way

DESIGN, PACKAGING, AND ELECTRICAL
TESTING OF PARYLENE-BASED
HIPPOCAMPAL RECORDING ARRAY
55

for the development of computational models which can imitate, and supplant, the proper
functioning of the hippocampus in cases of damage to the hippocampal neuronal circuitry. A
hippocampal BMI prosthetic, for example, might consist of a neural probe sensor that records
spatio-temporal spike train outputs from the CA3 and predicts the output that would naturally
stimulate the CA1 based on this modeling. Since principal cell layers in the hippocampus are
only 4-8 granule cells thick, and since each granule cell is 10 × 18 µm in size, hippocampal
layers of interest range from 40 to 144 µm thick [1]. The goal of this research project was to
design and fabricate a penetrating neural probe array with a fine enough spatial resolution to
selectively record from individual neurons in subdivisions of the rat hippocampal circuit
simultaneously.

Figure 34: One-way, tri-synaptic hippocampal circuit. Neural information from the entorhinal cortex
enters the hippocampus via the DG and then passes sequentially from the DG to the CA3 and finally the
CA1. Understanding the input-output characteristics of each part of the hippocampal circuit can be the
basis for the future development of a hippocampal prosthetic device based on chronic recordings from
penetrating neural probes.
Specific aims of the hippocampal array
The goal of this thesis is to design, fabricate, and develop a flexible, polymer-based
penetrating neural probe array that can record: (1) from multiple regions of the hippocampal
circuit simultaneously through positioning electrodes on each probe shank to match the laminar
architecture of the hippocampus, (2) from a dense number of recording sites so that many
individual neurons in each part of the network can be monitored, and (3) record complex spikes
with high signal-to-noise ratios over long periods of time.
56

Array design and system overview
The design of a complete hippocampal recording setup includes a neural interface for
collecting neural spikes and an electrical packaging system for transmitting these spikes to a
recording system that can monitor and store spikes over time. The implanted neural probe
electrode array incorporates a ribbon cable that connects to a recording system during surgery.
See Figure 35 for a schematic of the hippocampal recording setup.

Figure 35: Overall schematic of hippocampal recording setup. Flexible neural probe array is
designed to insert into the rat brain until it reaches the hippocampal target regions. Electrical
traces leading away from the recording electrodes carry neural activities via printed circuit
boards (PCBs) atop the rat’s cranium to an external recording instrument.
Neural interface
The neural interface of the hippocampal array was designed to span the 2,000 µm of the
dorsal hippocampus along the septal-temporal axis with eight probe shanks that are separated by
250 µm. Parylene C (10-18 µm thick) was chosen as the flexible substrate for the array and
platinum (2000 Å thick) was chosen as the metal layer. Parylene was chosen as a substrate due to
its compatibility with micromachining processes, conformal deposition, United States
Pharmacopeia (USP) class VI biocompatibility rating, and its flexibility. Pt was chosen to serve
as the electrode material due to its biocompatibility, high electrical conductivity, and inertness in
the body [2]. Parylene provides both structural support and electrical insulation. The sandwiched
metal layer in each probe contained two groups of platinum electrodes and their traces. Since the
57

deepest hippocampal layer of interest sits at a depth of 4-4.5 mm from the bregma (the cranial
suture intersection landmark) on the surface of the skull, probes were designed to be 5.5 mm in
length in order to reach all hippocampal regions of interest.
Principal neurons in the hippocampus pack into a cell body layer that is tens of microns
thin and curves into a double-C structure that is functionally divided into the CA1, CA3 and DG
sub-regions. In order for penetrating neural probes to record from each thin sub-region of the
hippocampus simultaneously, electrode sites must be densely patterned along the length of each
probe with some level of redundancy to overcome anatomical variations between animals.
Therefore each sub-region was targeted by a group of four linear electrodes which span the
estimated thickness of each cell layer. This ensured that at least one electrode site resides in each
targeted cell layer. The appropriate positions of electrode recording sites along the probe shanks
were determined through a combination of rat brain atlas and histological measurements
alongside in vivo recording tests which revealed changes in the depth of cell body layers across
the septo-temporal axis [3]. These locational differences were used to define the position of each
electrode group on each probe shank. The exposed electrode area was 30 µm in diameter
(electrode sites were patterned 50 µm in diameter and the edges were surrounded by insulation).
Electrode diameters were chosen to improve their selectivity to individual neurons and limit their
impact on the overall width of the shank, while balancing the need for sufficient surface area for
reducing electrode impedance (<1 MΩ at a frequency of 1 kHz) and noise [4].
The electrodes were connected to contact pads by 5 µm wide traces, with 5 µm spacing.
This small trace width was chosen to decrease the overall width of each probe and thereby
reduce the foreign body response to the implant. These design criteria enabled the realization of a
probe whose width spans from 110 µm at its most tapered part, to 150 µm at its widest part. The
shape and critical dimensions of the hippocampal array are presented in Figure 36.
58

Figure 36: Diagram of Parylene neural probe array with eight probes designed to match the
anatomy of the hippocampus as the CA1 and CA3 sub-regions change in depth along the septo-
temporal axis. Two groups of four Pt electrodes (30 µm exposed diameter) target the CA1 and
CA3. The shape and critical dimensions of the array are detailed above. *Thickness of insulative
Parylene layer varied between 10, 14, and 18 µm.
Electrical packaging
Individual probes of the hippocampal array connected into a flat, flexible cable which
terminate in Pt contact pads mated to a 71 pin zero-insertion-force (ZIF) connector (Hirose
Electric Co., Japan). The ZIF used in this work had a space-saving double row design with a 200
µm pitch between contact pads and a total length of 1.58 cm, which is less than the 2 cm
anterior-to-posterior space available on the rat’s cranium. This ZIF connector was the only
commercially available connector able to accommodate the 64 electrical connections of the
hippocampal array within the anatomical space limitations.
The electrical packaging used during in vitro electrochemical testing consisted of a single
PCB that served as an adapter between a ZIF and two 0.1” spacing header boards, whose pins
were easily connected to the working electrode of the potentiostat. Both cyclic voltammetry
cleaning and electrode impedance spectroscopy were run on each of the 64 individual electrodes
using this setup.
The electrical packaging for in vivo testing consisted of two, mated printed circuit boards
(PCB), as depicted in Figure 37. The first PCB was permanently mounted to the rat’s head with
59

dental cement, and supported the ZIF connector and an SSB6 PCB to PCB receptacle connector
(Molex Incorporated, Lisle, IL). The second PCB was detachable, and supported a male SSB6
connector and two 32 position dual row nano-miniature Omnetics connectors (Omnetics
Connector Corporation, Minneapolis, MN). The second PCB was connected during surgery and
established connection from the electrical packaging to a differential amplifier and oscilloscope.
This design was chosen to minimize the size of the permanent headmount that rested on the rat’s
cranium.

Figure 37: (a) Schematic and (b) photograph of electrical packaging for in vivo neural array
insertions. Neural recordings from the hippocampal array are transmitted through a series of
components soldered onto two PCBs and are eventually recorded through Plexon recording
equipment that interfaces with the two Omnetics connectors on PCB 2. PCB 1 is permanently
dental cemented to the rat’s cranium. The two PCBs connect through a specialized PCB-PCB
connector (SSB6).
During agarose and surgical insertions, the second PCB was attached to a
micromanipulator arm by means of a tab built into the PCB (Figure 37), or through two vertical
needle “posts” that were epoxied to each Omnetics connectors in earlier models of the PCB
packaging which lacked the tab feature (Figure 38 (a)). Either the tab or posts were threaded
through a metal ring which tightened around the head of the micromanipulator arm. A ground
wire was manually soldered onto each Omnetics connector and united to a male pin connector
(Figure 38 (b)). This reversibly mated with a female pin connector on the backend of a ground
wire which was inserted into the rat cerebellum during the implantation surgery. 1.5- 2 mm
diameter cable clips (Digi-Key Electronics, Thief River Falls, MN) (Figure 38 (c)) were used to
stabilize PCB 1 against an acrylic backing used for the application of a dissolvable brace around
60

the probe shanks to prevent array buckling during implantation (as discussed in Chapter 3).
These clips were difficult to remove prior to encapsulation of the array and PCB 1 with dental
cement at the end of surgery and replaced by double sided tape in future implantations.

Figure 38: (a) Needle “posts” used to attach PCB 2 to the micromanipulator arm during agarose
and surgical implantations. (b) Male pin connector attached to two Omnetics ground channels.
Connector mated with female pin connector on microwire serving as in vivo ground. (c) Cable
clips used to stabilize PCB 1 against a clear acrylic backing which is used for the application of a
dissolvable brace that prevents array buckling during implantation.
Assembly of electrical packaging
Custom PCBs (Gold Phoenix PCB, China) were populated with surface components
through reflow soldering performed on a hot plate. Flux (Mouser Electronics, Mansfield, TX)
was used to temporarily adhere the component to the PCB after it was aligned to pads on the
PCB under a microscope and to encourage smooth reflow of the pretinned solder pads. The PCB
and component were placed onto a hotplate which was heated to 150 °C for 90 seconds to warm
the solder and then heated to 275 °C for another 90 seconds for reflow. Custom PCBs for in vitro
electrochemical testing (OSH Park) did not come pretinned so solder was manually applied to
each pad on the PCB with a soldering iron prior to reflow. Electrical connectivity between each
channel of each component and the appropriate PCB pad by looking for electrical shorts on a
multimeter.
61

Testing of electrical packaging
Ideally, the two-PCB setup would last for 120 matings. This would accommodate 6
behavioral tests per day for five days of rat testing over a period of a month. However, the SSB6
connector was only rated for 30 cycles by its manufacturer. In initial lifetime testing of these
connectors the SSB6 receptacle component was found to degrade more rapidly than its rating, in
the form of shearing off from the PCB, while the SSB6 plug component remained functional
throughout the test.
In order to extend the functional lifetime of the SSB6 receptacle, marine epoxy was
explored as a possible means to improve adhesion between the receptacle and board which
originally relied solely on the strength of the soldering connection between the two. Marine
epoxy was carefully applied to underfill the edges of the receptacle, allowed 24 hours to fully
cure, and was mated to a plug (Figure 39 (b)). Care was taken to ensure that the height of the
marine epoxy did not exceed the height of the receptacle housing, which would interfere with the
connection between receptacle and plug. Electrical connectivity between the receptacle and
board was tested before and after the application of marine epoxy to verify that it did not result in
open circuits. To explore the effect of marine epoxy on the functional lifetime of the connectors,
the 64 connections between each component and PCB were tested with a multimeter after every
five mating cycles between the receptacle and plug until channel drop-off was deemed too
extreme to warrant continued tests. Care was taken to probe the receptacle gently, at the site
encircled in blue in Figure 39 (b) to prevent damage to pins. Trials 1-3 of Figure 40 represent the
results of these initial trials which generally supported the rated lifetime of 30 cycles.
While performing these experiments it was discovered that slight misalignments between
the SSB6 components caused physical damage to the plastic housing of the SSB6 receptacle
(Figure 39 (a)). Improvements in alignment technique, along with marine epoxy application,
resulted in two mating trials that lasted for a complete 120 mating cycles (Trials 4 and 5 of
Figure 40). It was decided to proceed with this two-PCB design and to take care to underfill
receptacles sufficiently as well as to practice alignment techniques prior to surgeries.

62

Figure 39: Damage to SSB6 plug after repeat connections. (a) prior to matings, (b) after many
matings, damaged area encircled in red. Plastic housing begins to degrade and prevents proper
alignment between plug and receptacle. (c) Marine epoxy underfilling SSB6 receptacle to fortify
connection to PCB. Red “x” indicates part of receptacle pin prone to damage during probing with
multimeter. Blue circle points to preferred probing point.

Figure 40: Cycling lifetime of SSB6 PCB to PCB connector. While these components were only
rated by their manufacturer for 30 cycles, the application of marine epoxy as an extra support
between the SSB6 receptacle and PCB pads helped to extend the functional lifetime of these
connectors (Trials 1-3). Marine epoxy and improved alignment techniques allowed for mating
cycles of 120 without depreciation (Trials 4-5, overlapping on graph).
63

Although the receptacle-plug mating lifetime was prolonged under pristine experimental
conditions, it was later found that imperfect mating in vivo caused electrical packaging damage
that led to recording fall-off over time. Rats were temporarily anesthetized with isoflurane to
enable connections between PCB 1 and 2 in vivo. However, this process was quickly performed
in order to complete it before the rat awoke from the light dose of anesthesia, and proper
alignment of the two components was made difficult by the large dental cement cap that
stabilized PCB 1 to the rat’s cranium. Accordingly, a new electrical packaging scheme was
designed to eliminate the PCB-PCB connection.
2
nd
generation electrical packaging
The goal of redesigning the electrical packaging system was to eliminate the PCB-PCB
connector system (shown to be finicky in vivo) while retaining a small packaging footprint on the
rat’s cranium. This was accomplished by creating a single PCB to interface with the
hippocampal array through a ZIF connector which directly translated signals to two Omnetics
connectors, one of which was tilted ~ 70° from the horizontal to minimize the size of the PCB
(Figure 41). This 2
nd
generation PCB is only slightly larger in dimensions than the previous
generation’s PCB 1 (1.9 × 2.1 cm vs. 1.8 × 1.7 cm, respectively). This PCB was designed to have
built in ground traces that connect the ground channels on the Omnetics with two stability pads
on either side of the ZIF connector. Microwire grounds are then directly soldered on the ZIF
stability pads and are safely embedded in dental cement at the end of the implantation surgery,
predicted to decrease external sources of noise added to the recording system by movement of
the ground wires. Another advantage of this newer packaging system is its quicker assembly
time since it contains two fewer components that need to be reflow soldered.
64

Figure 41: Second generation electrical packaging design. Single PCB goes from ZIF to
Omentics connectors directly and bypasses need for PCB to PCB connectors.
Electrochemical characterization of hippocampal array
Hippocampal arrays were inserted into the in vitro electrical packaging PCB, as described
above, and electrodes were potentiostatically cleaned using cyclic voltammetry (CV) and
analyzed using electrochemical impedance spectroscopy (EIS). Some electrodes required an
oxygen plasma etch to remove a thin residual layer of Parylene that remained, clouding the
electrode surface.
Electrode cleaning
CV (-0.2 V to 1.2 V, scan rate of 250 mVs
-1
) was run on each individual electrode trace
for 30 cycles in 0.5 M H2SO4 purged with N2 for five minutes prior to scanning in order to
systematically clean the 64 electrodes of the hippocampal array. A 1 cm
2
Pt plate was used as a
counter electrode and an Ag/AgCl electrode was used as a reference electrode. This
oxidation/reduction process was used to chemically clean off any adsorbed debris from the
surface of the electrode [5] and has been previously employed by the Meng lab as a means to
clean electrode surfaces as exhibited by an 80% decrease in 1-kHz impedance of electrodes post
CV cleaning [6]. A representative CV of an electrode from a sputtered device is shown in Figure
42. As the potential held across the electrode is cycled between negative and positive values, the
electroactive area of the electrode grows, as represented by broadened CV curves, indicating
electrode cleaning.
65

Figure 42: Representative CV of electrode from a sputtered hippocampal array. The first CV
curve is relatively narrow, indicating a small electroactive area. As the cycles progress the
current response broadens at different voltages indicating increased cleanliness of the electrode
surface area which allows for more surface reactions and current flow. At cycle 30 this CV now
mimics the traditional CV of Pt in H2SO4.
In cases where CV cleaning produced a curve with random noise, electrode traces were
checked under the microscope to see whether this resulted from an open circuit caused by defects
in the metal lithography step. In some cases, it was noted that a thin, insulative layer of Parylene
remained covering the electrodes and contact pads, occluding electrical contact between the
contact pad and ZIF connector and/or electrochemical continuity between the electrode surface
and the chemical solution. Reactive ion etching (RIE) in O2 gas at 200 W power, 200 mTorr
pressure, that lasted for two minutes was used to etch away this Parylene scum (~0.4 µm) in
order to physically clean the metal surfaces (Figure 43). CV cycles run after this cleaning step
revealed as expected CV curves, indicating proper electrical connectivity through the entire
system.
66

Figure 43: Microscopic image of contact pad pre (a) and post (b) O2 RIE descum at 200 W, 200
mTorr, and 2 min. Pre (c) and post (d) images of electrode surface. Thin residual layer of
Parylene scum is visible in images prior to RIE etch, but cleared away after the descum. CV
curves prior to descum revealed an open circuit pattern, but returned to normal after RIE descum.

CV cleaning of an additional eight electrodes on an alternate device, and EIS data taken
both pre-CV cleaning and post-CV cleaning indicate an impedance drop from 801 ± 29 kΩ to
507 ± 60 kΩ, a 37% decrease in impedance, due to the CV cleaning step.
Electrode impedance testing
EIS was performed in 1× phosphate buffered saline (PBS) (OmniPur 10× PBS, EMD
Chemicals, Darmstadt, Germany) with an excitation voltage of 25 mV (AC) over frequencies
from 1 Hz to 0.1 MHz in order to determine electrode impedance at 1 kHz—widely accepted to
be the frequency most closely matched to that of action potentials. Electrodes with impedance
measuring greater than 2 MΩ at 1kHz, or with a phase spectrum that did not transition from
resistive (close to 0°) to capacitive (close to -90°) with decreasing frequency, were considered
non-functional for purposes of implantation. See Figure 44 to see changes in EIS measurements
taken before and after the CV cleaned electrode shown in Figure 42. The average 1 kHz
impedance across 176 electrodes was 613 ± 247 kΩ. This confirms that untreated Pt electrodes
of this size exhibit a low enough electrode impedance requisite for neural recordings.

67

Figure 44: EIS taken before (a) and after (b) CV cleaning of the representative electrode shown
in Figure 42. CV cleaning of this electrode dropped the 1 kHz electrode impedance from 1.3 MΩ
to 360 kΩ.
Repeatability challenges in EIS
In some cases, repeat EIS measurements taken across the same electrode revealed a > 200
kΩ variation in 1 kHz impedance. Since electrodes in the hippocampal array have an exposed
diameter of only 30 µm, it was hypothesized that this small electrode size may necessitate a
longer open circuit potential (OCP) equilibration step than usual, in order to allow the
electrochemical cell to fully stabilize before beginning EIS. Premature termination of this
equilibration step was considered to be a possible source of 1 kHz impedance variation. To test
this theory, the OCP step was lengthened from 100, 200, 300, to 400 seconds for three separate
electrodes (n = 6 for electrodes run at OCP of 100 s) which were each tested three times at each
of the OCP values. It was expected that increasing the length of the OCP step would help reduce
variations between 1 kHz impedance measurements of the same electrodes.
Results of this experiment can be seen in Table 3. Increasing the OCP length did not
seem to increase the reproducibility of impedance measurements. There exists no trend in
relative standard deviation across the same electrode measured with various OCPs. The average
standard deviation of the 1 kHz impedance across all electrodes and OCPs was 48 kΩ with a min
and max of 12 and 190 kΩ respectively while the average 1 kHz impedance across all
measurements was 412 kΩ. Since these relatively large standard deviations do not seem to be
affected by OCP length, they could be a phenomenon of microelectrodes in general where it is
68

reasonable to imagine small changes in the concentration of chemical species at the surface of
the electrode vastly changing the surface reactions.
Table 3: Average electrode impedance at 1 kHz, stdev of average electrode impedance at 1 kHz,
and relative standard deviation across electrodes measured multiple times with OCP lengths that
varied from 100 to 400 seconds. Increasing OCP length did not impact the repeatability of 1 kHz
impedance measurements.
OCP length (s) Electrode Average 1 kHz
Impedance (kΩ)
Stdev of 1 kHz
Impedance (kΩ)
Relative Standard
Deviation (%)
100 (n=6) A 414 104 25

B 405 37 9

C 362 40 11
200 (n=3) A 407 13 3

B 362 46 13

C 371 13 4
300 (n=3) A 384 43 11

B 447 16 4

C 422 12 3
400 (n=3) A 588 190 32

B 397 51 13

C 388 13 3

Reducing current in EIS
Initial longevity tests of Parylene-Pt-Parylene electrodes performed on a peripheral nerve
interface indicated that applying a current through electrodes in a saline solution might
contribute to premature delamination or wrinkling of the metal traces from the Parylene base
layer. In attempt to attenuate this phenomenon, an exploratory experiment was performed to see
if it was possible to reduce the amount of current applied to electrodes during EIS testing without
compromising the quality of EIS measurements. The AC stimulation voltage across electrodes
was adjusted from 25 mV to 10, 5, 1, 0.5, 0.1, 0.05, 0.02, and 0.01 mV.
EIS impedance magnitude and phase graphs are presented in Figure 45 for AC voltages
that vary by a factor of ten. Graphs of impedance magnitude had little noise for 10, 1, and 0.1
mV, as evidenced by R
2
values of the linear portion of the curve that were greater than 0.99.
Visible noise was introduced to the impedance magnitude curves only when EIS was run with a
0.01 mV stimulation voltage at which point the R
2
value decreased to 0.98 and noisy outliers
were clearly visible by eye on the impedance magnitude graph. The smoothness of the phase
69

graph, however, devolved quicker and grew noisy with a stimulation voltage of 0.1 mV. In order
to get noise-free EIS measurements, 0.5 mV was chosen as the appropriate stimulation voltage
for EIS and was shown to produce smooth curves across four different electrodes and gave an
average impedance of 335 ± 15 kΩ in five trials run across the same electrode. A stimulation
voltage of 0.5 mV represents a 50 times decrease in stimulation voltage which is expected to
reduce the current running through electrodes from 42 to 0.8 nA, assuming an average
impedance of 600 kΩ. This 0.5 mV minimum was experimentally determined for the 30 µm
exposed diameter hippocampal electrodes, however, it is possible that larger microelectrodes
could reduce the voltage even further without experiencing noise in measurements.

Figure 45: Impedance magnitude and phase curves for various AC stimulation voltages.
Impedance magnitude stays noise-free until 0.1 mV. The phase curve already has outliers at a
stimulation voltage of 0.1 mV.
Summary
The development of a penetrating neural probe array that collects recordings from
individual neurons in the hippocampus has the potential to inform the development of a
computational circuit model that can be used for a future hippocampal prosthetic device. In this
chapter, we have discussed the design and assembly of such an array that contains electrodes
positioned to match the complex laminar anatomy of the hippocampus. An electrical packaging
scheme used to connect the array to a Plexon recording system was designed to minimize the
weight and size of the packaging on the rat’s cranium and modifications made to the packaging
system after testing are included here as well. Lastly, this chapter discusses using cyclic
70

voltammetry to electrochemically clean the electrode surfaces as well as the use of electrode
impedance spectroscopy to measure electrode impedance and confirm a reasonable 1 kHz
impedance that will enable the capture of neural spikes whose amplitudes are on the order of
hundreds of microvolts. This chapter is followed by an in vivo analysis of electrode performance
in both acute and chronic experiments.
References
[1] P. Andersen, R. Morris, D. Amaral, T. Bliss, and J. OKeefe, The hippocampus book:
Oxford University Press, 2006.
[2] A. Cowley and B. Woodward, "A healthy future: platinum in medical applications,"
Platinum Metals Review, vol. 55, pp. 98-107, 2011.
[3] H. Xu, A. Weltman, M.-C. Hsiao, K. Scholten, E. Meng, T. W. Berger, et al., "Design of
a flexible parylene-based multi-electrode array for multi-region recording from the rat
hippocampus," in 37th Annual International Conference of the IEEE Engineering in
Medicine and Biology Society (EMBC), 2015, pp. 7139-7142.
[4] S. F. Cogan, "Neural stimulation and recording electrodes," Annu. Rev. Biomed. Eng.,
vol. 10, pp. 275-309, 2008.
[5] A. Petrossians, J. J. Whalen, J. D. Weiland, and F. Mansfeld, "Electrodeposition and
characterization of thin-film platinum-iridium alloys for biological interfaces," Journal of
The Electrochemical Society, vol. 158, pp. D269-D276, 2011.
[6] S. A. Hara, B. J. Kim, J. T. Kuo, and E. Meng, "An Electrochemical Investigation of the
Impact of Microfabrication Techniques on Polymer-Based Microelectrode Neural
Interfaces," Journal of Microelectromechanical Systems, vol. 24, pp. 801-809, 2015.

71

Methods of assessing probe performance over time
Three methods of testing are commonly employed in order to evaluate the lifetime of a
recording array and to quantify the immune response against the neural implant. Electrode
performance can be monitored electrochemically through in vivo electrode impedance
spectroscopy (EIS) measurements, electrically through signal-to-noise ratio (SNR)
measurements and other metrics of recording quality, and through the biological response against
the foreign implant which can be monitored through various creative imaging techniques that
vary in invasiveness.
The first of these methods that monitor the chronic reliability of the neural implant is EIS,
which consists of applying a small alternating potential between a working electrode on the
neural probe and a counter electrode in a bath of conductive solution or in tissue in vivo. By
measuring the resultant current, the complex impedance of the electrode-solution or electrode-
tissue interface, split into both magnitude and phase, can be calculated. This process is repeated
across various frequencies (which often ranges from 10 Hz to 100 kHz [1]) producing two Bode
plots that represent impedance magnitude and impedance frequency. Impedance magnitude may
predict the ability of a neural probe to record small amplitude fluctuations in extracellular neural
voltages. A high impedance can signify the build-up of high resistance scar tissue at the electrode
interface which insulates the electrode from neural recordings leading to higher noise levels. A
qualitative assessment of phase changes over time may provide insight into the immune response
against the foreign implant. In one study, impedance changes were monitored up to 1 week post
implantation and increased impedances at 1 kHz (the frequency that corresponds to action
potentials) were found to correlate with extensive acute immune reactions in histologically
stained tissue surrounding microwire implantation sites [2]. In vivo EIS measurements were also
used to predict electrode functionality [3] based solely on impedance magnitude. EIS provides an
indirect way to measure electrode function and immune response.

IN VIVO PERFORMANCE OF HIPPOCAMPAL
RECORDING ARRAY
72

Traditional methods to image the biological response of brain tissue against the neural
implant use immunohistochemistry (IHC) in discrete endpoint post-mortem histology (with or
without fluorescence) to directly image the population of neurons and immune cells including
astrocytes, macrophages, and microglia, the resident neural macrophage. Traditional optical
imaging relies on staining thin tissue slices (5-50 µm) since photon scattering occurs beyond
these depths and results in image blurring [4]. Slicing the tissue block into thin sections for
imaging may also distort and damage the tissue. Unless the probe material is soft enough to be
cut with a microtome, probes must be removed prior to slicing. While these studies have been
useful to uncover phenomena such as neuronal cell loss near the implant [5], upticks in
immunoreactivity associated with the foreign body response and not just the acute stab injury
[5], the manner in which tethering increases astrocytic encapsulation of the implant [6], and the
upper limits of probe flexibility which confer protection against immune reactions [7], they are
limited in the degree to which temporal effects can be investigated and direct cause and effect
relationships can be drawn in addition to severe limitations in image depth. Because of these
limitations, and the fact that immunohistochemical analysis of tissue is terminal, some have
moved towards using in vivo two photon microscopy as a method to dynamically track tissue
changes while the animal subject is living. Imaging is performed noninvasively through a glass
window bonded over the craniotomy site. Two photon microscopy has been creatively harnessed
to monitor the movement of microglia over time [8] and to monitor neurovascular damage
incurred during the implantation surgery itself [9]. Multiphoton imaging enables optical
resolution of tissue up to a depth of hundreds of µms [4], but even cortical probes insert to mm
depths. We are currently taking part in a collaboration with an imaging lab that combines tissue
clearing techniques to remove opacity from tissue with light sheet microscopy in order to study
cellular relationships and neuronal circuitry in deep tissue systems consisting of complete, intact,
organs and bones [10, 11]. By increasing tissue transparency, photon scattering is minimized,
and optical imaging of depths > 1 mm can be explored. Although this technique is terminal, we
hope that its increased penetration depth will enable imaging of neuronal connections in the
hippocampus and the immune response surrounding hippocampal arrays in histology studies that
occur over time.
SNR measurements provide insight into a neural probe’s chronic reliability but they do
not exclusively identify the mechanism of failure which can be both biological or abiotic in
73

nature [12]. Useful SNRs are characterized as those with ratios of 5:1 or greater [13]. SNR is
most commonly calculated as the amplitude of the average waveform of a given neural unit
divided two times the standard deviation of the noise [14, 15]. Other metrics of recording quality
include average peak to trough spike amplitudes and baseline noise levels and can be monitored
to determine the lifetime of a neural probe technology. Additionally, monitoring neuronal yield
over time can help uncover the health of nearby neurons and the extent of the immune response
against the foreign implant.
In initial implantations, the chronic performance of the hippocampal arrays was chosen to
be monitored through histological staining and recording quality metrics. Recording metrics
from four acute rat implantations and chronic recordings from three out of four of the implanted
animals are included below.
In vivo implantations
Electrochemical cleaning and evaluation of electrode quality
Prior to PEG encapsulation and insertion into the ZIF, four hippocampal arrays were
potentiostatically cleaned using cyclic voltammetry (CV) [16] and analyzed using
electrochemical impedance spectroscopy (EIS). CV (-0.2 V to 1.2 V, scan rate of 250 mVs
-1
)
was run on each individual electrode trace for 30 cycles in 0.5 M H2SO4 purged with N2 for five
minutes prior to scanning. A 1 cm
2
Pt plate was used as a counter electrode and an Ag/AgCl
electrode was used as a reference electrode. EIS was performed in 1 × phosphate buffered saline
(PBS) (OmniPur 10 × PBS, EMD Chemicals, Darmstadt, Germany) with an excitation voltage of
25 V (AC) over frequencies from 1 Hz to 0.1 MHz with an Ag/AgCl reference and 1 cm
2
Pt
counter electrode. Electrodes whose phases did not change from resistive to capacitive as the
frequency decreased, and whose impedances were greater than 2 MΩ at a frequency of 1 kHz
were considered to be open circuits and were not used for impedance calculations. Average 1
kHz electrode impedances of the four hippocampal arrays implanted were 468 ± 279 kΩ (62
functioning electrodes), 734 ± 238 kΩ (53 functioning electrodes), 795 ± 364 kΩ (63 functioning
electrodes), and 718 ± 147 kΩ (62 functioning electrodes).
74

Implantation procedure
Both the Institutional Animal Care and Use Committee (IACUC) and the Department of
Animal Resources of the University of Southern California (DAR, USC) reviewed and approved
of all animal experiments. Four fully functional arrays that passed electrochemical testing were
implanted into four male Sprague-Dawley rats older than 3 months of age and weighing between
300- 450 g. Animals were pre-anesthetized with an intraperitoneal (IP) injection of a ketamine
and xylazine mixture. Anesthesia was maintained intra-operatively through the delivery of an
inhaled mixture of isoflurane and oxygen delivered to the animal. Negative toe pinch withdrawal
reflexes tested throughout surgery confirmed appropriate anesthetic level. A stereotactic frame
was used to hold the animal in place. To expose the brain surface above the implantation site, a 2
× 4 mm cranial window was drilled away above the right dorsal hippocampus and the dura was
carefully removed with forceps. Three small holes, for anchor screws, were drilled in the skull
surrounding cranial window. One anchor screw doubled as a ground, by making contact with the
cerebrospinal fluid. A small hole for the ground wire itself was drilled above the cerebellum.
Arrays were connected to an oscilloscope through the in vivo electrical packaging system during
insertions.
The exposed, bare tips (0.95 mm) of a hippocampal array were poised above the
implantation site located at ~2.5 mm posterior to bregma and ~2.45 mm left of the midline. A
micromanipulator was used to support and advance the array during implantation. The array was
angled ~30° from the midline in order to align to the septal-temporal axis of the hippocampus.
The array was slowly inserted into the brain until the bottom edge of the PEG brace reached the
surface of the brain. The brace was then gradually dissolved away with saline solution and the
newly exposed array length was advanced in increments of 0.05 mm at a speed of 10 µm/s.
Dissolving a single brace at one time took ~ 20 min. Neural signals were monitored throughout
the implantation procedure for the presence of complex spikes (a burst of 2-6 single spikes of
decreasing amplitude with ≤ 5 ms interspike intervals [17]). Complex spikes serve as an
electronic signature for pyramidal neurons of the hippocampus and helped to confirm proper
placement of the array in both the CA1 and CA3 layers of the hippocampus. Once the first group
of electrodes began recording spikes from the more superficial, CA1 layer of the hippocampus
(located at ~ 3 mm deep from the surface of the skull), the array was further advanced until the
second group of electrodes reached the CA1, at which point the first group of electrodes recorded
75

from the deeper, CA3 layer. Since neural activities are monitored at multiple depths during
insertion, the length of the full implantation surgery depends on how frequently the advancement
of the array is paused to check for neural activities. Usually the insertion is paused at each depth
for ~ 10 min and a complete implantation lasts between 2-3 hours.

Figure 46: In vivo implantation of hippocampal array supported by dissolvable PEG brace into
rat brain. (a) Bare tips of hippocampal array poised over implantation site, (b) PEG incrementally
dissolved in saline and newly exposed probe length further inserted into brain with pauses for
electrical recording to ensure proper placement, and (c) remainder of PEG brace dissolved in
preparation for dental cement cap. Scale bar = 5 mm.
After the array reached the desired target location, a thin layer of dental cement was
applied over the insertion site to hold the array in place. The ground wires were twisted onto the
screw that acted as the ground electrode for added stability, and the tips of the ground wires were
inserted into the hole above the cerebellum. The electrical packaging was lowered to the cranium
and additional dental cement was then applied to create a ‘cap’ of dental cement that
encapsulated the array and the first half of the permanent PCB, with care taken to leave the SSB6
receptacle at the top of the PCB exposed for future PCB to PCB connections during experiments.

76

Figure 47: Final assembly on rat’s cranium. Dental cement cap encapsulates remainder of
hippocampal array and first half of PCB 1 with care taken to leave the ground pin and SSB6
PCB-PCB connector on PCB 1 (not visible) exposed for future reconnections to PCB 2.
Data acquisition
During array implantation, recorded signals were bandpass filtered between 300- 3000
Hz using an alternating current differential amplifier and displayed on an oscilloscope. The
oscilloscope output was connected to a speaker to allow for auditory discrimination between
single spike and complex spike activity. Spikes from electrodes in the CA1 and CA3 electrode
groups were collected in one second long recording traces for signal to noise ratio analysis
during the surgery.
After the animal fully recovered from the surgery, 7-16 days post-implantation, neural
activities were recorded from the free-moving animal running in an open field. Neural signals
were amplified 20× by a head-stage and then were further amplified by a gain-adjustable
differential amplifier connected to a data acquisition system (Plexon Inc., Dallas, TX). The
signal was digitized at 40 kHz and bandpass filtered to remove low frequency artifacts and
enable the examination of neural spikes. Spike sorting through principle component analysis
(PCA) was conducted off-line after the data was stored.
77

Acute recordings
Neural recordings during implantation
During implantation of the array, neural signals were monitored by oscilloscope, which
allowed for observation of just a single electrode at a time. Each cluster of four closely spaced
electrodes was considered to be a ‘recording group’ and data was collected from a representative
electrode from each group throughout implantation. Recording groups containing at least two
electrodes that passed electrochemical inspection prior to implantation were considered
functional electrode groups. A total of 57 (out of 64) recording groups were considered
functional and were monitored during four implantation surgeries. Unitary activities and
complex spikes were successfully recorded from 35 recording groups (61.4% of functional
recording groups).
In each of the four implantation surgeries, signals were obtained from electrodes in both
the CA1 and CA3 sub-regions. Probes with two or more working electrodes in both the CA1 and
CA3 electrode groups, as confirmed through electrochemical testing and continuity testing of the
electrical packaging prior to implantation, were considered to be functional. A total of 28
functional probes (out of 32) were implanted, and, at the final implant location, 13 of these
shanks recorded unitary activities from both the CA1 and CA3. As the insertion procedure
progressed, the depth at which complex spikes from these two sub-regions were detectable was
recorded. The average depth at which CA1 spikes became evident was 2.84 mm from the surface
of the brain. Across the four implantations the average final depth of array placement was 4.13
mm. Figure 48 shows representative complex spikes recorded intra-operatively from a depth of
4.15 mm from the brain surface captured in 0.1 s long recording traces.
78

Figure 48: Representative 0.1 s long spikes recorded intra-operatively with hippocampal array
implanted to 4.15 mm from the brain surface. Complex spikes (burst of 2-6 single spikes of
decreasing amplitude with ≤ 5 ms interspike intervals) recorded from the CA1 and CA3 during
an experimental implantation surgery.
Quality of acute in vivo recordings and microwire comparison
Metrics for recording quality include average peak-to-valley spike amplitudes,
background noise, and signal to noise ratio (SNR). SNR was calculated using the following
formula:
SNR =
A
2 ∗ SD
noise

where A, the mean amplitude of spikes, is the average of peak-to-valley voltage of waveforms in
each 1 s trace. SDnoise is the standard deviation of the background noise. Each 1 s long recording
trace was low pass filtered to remove baseline drift. Spikes were defined as peaks with negative
amplitudes greater in magnitude than a threshold of two standard deviations of the filtered
recording trace. After removing spikes from the trace (the removed segments include both 400
µs before and 1,200 µs after threshold crossing) the standard deviation of the baseline noise was
calculated. The maximum SNR achieved by any electrode was 26.1. Figure 49 shows the mean
peak-to-valley spike amplitudes and background noise levels across all working electrodes
recorded during surgery, while Figure 50 displays the average and maximum SNR values
79

achieved by each Parylene array. Each of these two figures contains a comparison to data
recorded with conventional 25 µm diameter, stainless steel, microwire arrays, performed by a
collaborator. Average spike amplitudes (139.3 ± 75.6 µV), noise levels (37.8 ± 6.5 µV) and
SNRs (3.6 ± 1.4) of recordings collected by the Parylene hippocampal array were similar to
those of microwires deployed for hippocampal neural recordings in previous work.

Figure 49: Average spike amplitudes (top column stack) and noise levels (bottom column stack)
with standard deviations across all four Parylene array implantations compared to those from
seven microwire array implantations. The maximum spike amplitudes recorded by Parylene and
microwire arrays (N = 7) are represented by the circle above each bar which connects to the y-
axis on the right.
80

Figure 50: Average signal to noise ratio (with standard deviations) achieved by each Parylene
hippocampal array in the acute surgical implantation setting. Dashed line represents the average
signal to noise ratio across seven microwire array implantations and dashed-dot lines represent ±
1 standard deviation of the microwire average. The maximum signal to noise ratio achieved by
each Parylene array during implantation are noted by the circular symbols which reference the y-
axis to the right of the plot.
Chronic recordings
Chronic recordings from free-moving animals
Three of the subjects were given 7-16 days post implantation to recover, after which
neural activities were recorded on a 64 channel data acquisition system while the animal freely
roamed an open field. Simultaneous recording traces of all 64 electrodes were collected and
spike sorting was applied to the recorded data offline. All three arrays recorded unitary activities
from behaving animals, including complex spikes from pyramidal cells and single action
potentials locked with theta rhythms.
Crosstalk (spikes from the same neural unit, with the same waveform, that present at
identical timestamps) between traces and electrodes located in different recording groups on the
same Parylene shank was visible during the first post-surgery recording session, and remained a
confounding factor throughout all chronic recording sessions. Unfortunately, the presence of
81

crosstalk made both the identification of the signal origin and spike sorting difficult. In order to
attenuate the effect of crosstalk on the recorded data, recording traces from only the electrode
that recorded the highest amplitude signals from a particular neural unit (the primary electrode)
at specific timestamps were considered in average amplitude, noise, and signal to noise ratio
calculations.
Any electrode that recorded unitary activities, even if those activities were due to
crosstalk, were considered functional. Out of the 173 functional, implanted electrodes, 114
recorded unitary activities during recording sessions up to one month post implantation. Up to
four neural units were recorded from a single electrode. Only one implanted hippocampal array
obtained chronic unitary activities from both the CA1 and CA3 regions as seen in Figure 51.
This array recorded a total of seven neural units from the CA1 and four units from the CA3.

Figure 51: Chronic recordings obtained from the third implantation on day 24 post implantation.
Complex spikes were recorded from both the CA1 and CA3 sub-regions.
Quality of chronic in vivo recordings
Metrics for recording quality include average peak-to-valley spike amplitudes,
background noise, and SNR were calculated as above. Noise was calculated by setting the
threshold to ± 2 standard deviations from the original recording trace and then removing
waveform segments that contained spikes (the removed segments include both 250 µsec before
and 750 µsec after threshold crossing). For all data analysis purposes, spikes due to crosstalk
between channels, as well as all subsequent spikes of lower amplitude in a complex spike group,
were removed using a Matlab program. Only the highest amplitude neural unit of each electrode
82

was considered. The maximum average SNR achieved by any electrode was 13.0. Figure 52
shows the average (and standard deviation of the) peak-to-valley spike amplitudes and
background noise levels across all working electrodes for each recording day per hippocampal
array. Figure 53 displays the SNR averaged across the same period. The maximum average SNR
across an entire array was 9.4, the minimum was 4.2. SNR stayed stable throughout the lifetime of
each array. Average spike amplitudes (101.1 ± 17.6 µV), baseline noise (18.6 ± 3.0 µV), and SNR
(5.6 ± 1.3) remained stable for recording lifetimes.

Figure 52: Averages and standard deviations of peak-valley spike amplitudes (solid symbols) and
background noise (outline symbols) across all functional electrodes, for each of the three chronic
recording experiments, with any suspected spikes due to crosstalk removed. Arrays were
monitored until recordings were no longer visible during chronic testing. N varied at each day of
recording. Maximum number of channels that recorded unique spikes during a single day was
41, minimum was 7.
83

Figure 53: Average signal to noise ratio (SNR) and standard deviation across all functional
electrodes for each of the three chronic recording experiments, with any suspected spikes due to
crosstalk removed. SNR remains steady over the lifetime of recordings. N varied at each day of
recording. Maximum number of channels that recorded unique spikes during a single day was
41, minimum was 7.
Histology
Experimental procedure for one month histological study
A sham array, which lacked metal electrodes and traces, but was otherwise identical to
fully fabricated arrays, was implanted into the right hippocampus of a single rat. This animal was
sacrificed at one month post-implantation, at which time the animal was deeply anesthetized with
an IP injection of ketamine and xylazine and intracardially perfused with 10%
paraformaldehyde. The rat brain was dissected from the cranium, which forced the hippocampal
array to exit the brain. After the brain was dehydrated in 18% sucrose overnight, the tissue was
84

then embedded in 3% agarose for slicing. 50 µm thick transverse slices cut with a vibratome
(Lecia, VT1000 S). Slices taken only at depths of 2.7 mm and 2.75 mm were stained with
antibodies for glial fibrillary acidic protein (GFAP) and neuronal nuclei (NeuN) respectively for
the identification of astrocytes and neurons and counterstained with hematoxylin. Sections of
tissue on the same histological slice, on the same hemisphere, but > 500 µm away from the
implant site were chosen as control regions for comparison. Radial rings of 25 µm were drawn
around probe cross-sections and corresponding control regions. Color thresholding in ImageJ
software (National Institutes of Health, Bethesda, MD) was used to identify and count the
concentration of astrocytes in each annulus. Rectangular bins of 50 µm width, whose length
matched the thickness of the cell body layer were used to count the concentration of neurons in
implanted and control regions.
Results of one month histology study
The concentrations of astrocytes and neurons around the implantation site of a sham
array, measured one month post implantation, are presented in Figure 54 and Figure 55. Tissue
slices were taken at a depth of 2.7 mm and 2.75 mm and stained for GFAP and NeuN
respectively with hematoxylin counterstain. At 2.7 mm deep, the probes deviated slightly from
the cell body layer, whereas at 2.75 mm the probes cleanly hit the cell body layer. Control
regions were chosen to mimic the offset of each probe from the cell layer in order to control for
the heterogeneity in cellular distribution. Radial rings of 25 µm were drawn around probe cross-
sections and the corresponding control regions. Color thresholding was used to create a color
mask which allowed for the calculation of the concentration of astrocytes. Rectangular annuli fit
to the thickness of the cell body layer were used to calculate the concentration of neurons since
neurons largely lie within the cell layer. The concentration of the cell of interest was defined to
be the fraction of area in each ring occupied by the stain of interest.
Figure 54 shows histology images of the implanted and control regions with astrocytes
stained in brown with GFAP. Astrocytes, a glial support cell in the brain, are known to increase
in number around areas of injury and attempt to wall-off foreign implants through scar
formation. An increase in astrocytic concentration around the implant site serves as an immune
marker for the severity of the immune response against the individual probes in a neural array.
An examination of these images and the table included in Figure 54 reveals a small increase in
85

astrocyte concentration at rings close to the implant site compared to control, non-implanted
regions. The concentration of astrocytes within 100 µm of the implantation location was 8.5 ±
3.3% higher, on average, than their control counterparts. However, no significant increase of
astrocytes beyond those in the control region was noticed at distances greater than 100 µm from
the probe track (t test value of p=0.67).

Figure 54: Transverse, 50 µm thick hippocampal slice (at -2.7 mm) from one-month sham array
implantation. Stained with GFAP to highlight astrocytes in brown; purple corresponds to
hematoxylin counter staining; dense purple strip is the DG of hippocampus. Array was
unfortunately removed prior to tissue slicing during removal of the brain from the cranium. Scale
bars are 100 µm. (a) Black arrows indicate locations of three probes of the array visible in the
DG of the hippocampus and (b) arrows mark the location of matched controls from the same
tissue slice. Color thresholding was used to measure the astrocytic density in 25 µm rings around
the central three probes and corresponding control regions, included in table in (c).
A comparison of neuronal concentration (neurons were stained in brown with NeuN) is
presented in Figure 55 in order to evaluate whether or not excess neural death occurred near the
implant site. Rectangular bins 50 µm wide whose lengths matched the varying thickness of the
cell body layer were used to compare the concentration of neurons across the site of array
implantation to a control region. A comparison between implanted and control sites reveals that
neuronal concentrations return to normal in the ~ 100 µm space between adjacent probes in the
array.

86

Figure 55: Transverse, 50 µm thick hippocampal slice (at -2.75 mm) from one-month sham array
implantation. Stained with NeuN to highlight neurons in brown; purple corresponds to
hematoxylin counter staining; dense purple strip is the DG of the hippocampus. Array was
removed prior to tissue slicing. (a) Black arrows indicate locations of five probes of the array
visible in the DG of the hippocampus and (b) is the color thresholded version of the image with
50 µm thick rectangular bins drawn around the central three probes to measure density of
neurons surrounding the implantation sites. Scale bars are 100 µm. (c) A graphical representation
of the concentration of neurons in implanted and control (not shown) rectangular bins. Trace
represents neuronal concentration in bins surrounding the implant site, horizontal lines represent
the average neuronal concentration ± 1 standard deviation in matched control regions (N = 3).
Neuronal concentration returned to control levels in between probes in the array. Probe cross-
sections are indicated by open filled circular markers.
Discussion of array performance and immune response
The Parylene C arrays proved simple to handle and physically robust despite the
mechanical flexibility of the thin film material. This is a favorable quality in a neural interface,
considering the frustrating risk of fracture inherent to silicon probe arrays with similarly thin (20
µm) profiles. Despite the low elastic modulus of the Parylene structure, arrays were easy to
implant with the addition of the PEG brace. Use of the dissolvable PEG brace prevented
buckling or significant curving of the flexible probes during implantation and did not require
introducing the brace material to the brain tissue. The success of this approach was confirmed
through examination of histology data; minimal tissue was displaced due to insertion of the
probes (Figure 54 & Figure 55), unlike previously reported methods involving the use of large
87

insertion shuttles [18-22]. The use of a dissolvable brace was described previously for thin metal
microwire probes and carbon fiber probes [23] and our results confirm its efficacy for polymer-
based neural probe arrays.
Electrochemical characterization confirmed the Pt electrodes exhibited adequately low
impedance at 1 kHz (677 ± 297 kΩ), which corresponds well to that of commercially available
stainless steel microwires (600- 900 kΩ) with comparable electrode sizes (Parylene probe: 30
µm diameter, microwire: 25 µm diameter). These impedance values fall within the 50 kΩ – 1
MΩ range [13] required for the resolvable acquisition of neural recordings, and were largely
determined by the exposed surface area of the Pt electrode site. The 30 µm diameter Pt
electrodes were sufficiently small for the selective resolution of unitary activities and we
obtained intra-operative recordings from single neurons with spike amplitudes (139.3 ± 75.6
µV), noise levels (37.8 ± 6.5 µV) and SNRs (3.6 ± 1.4) similar to those of microwires deployed
for hippocampal neural recordings in previous work. The observed variation in spike amplitude
is attributable to variation in the distance between the recording site and adjacent neurons [24]
and neuron size. The presence of high amplitude spikes indicate that the hippocampal array was
implanted close to the cell body layer of interest. Chronic recordings from three implanted
animals had stable baseline noise levels (18.6 ± 3.0 µV) and average SNRs (5.6 ± 1.3). Noise
levels were lower for chronic recording experiments due to the increased stability and quality of
the Plexon recording system. Long-term recordings from rats 1 and 3 were terminated
prematurely because of damage to the SSB6 connectors that prevented further reconnections and
caused channel drop-off. The fact that the SNR of the last implantation remained constant for ~ 2
months seems to indicate that the severity of the immune response within this time frame is not
sufficient to negatively impact recording quality.
The quality of this recording data was more than sufficient to fully resolve spike activities
with high fidelity, and as such we believe will be suitable for acute and, ultimately, longer
chronic applications studying hippocampal neural circuits in free moving subjects. The recorded
neural data compares favorably both with commercial microwire probes and other
microfabricated neural probe arrays. Due to differences in SNR calculations between research
groups [14, 15, 25] a direct comparison between SNR measurements across literature is difficult.
However, the SNR values of this hippocampal array is peppered with quality (SNR > 4:1) and
moderate (3:1 < SNR < 4:1) units defined by the experimentally based measurements of Ludwig
88

et. al. [15], and the values reported here are similar to prior work using silicon-based neural
arrays to record from the hippocampus [24]. We also note that some noise was introduced by an
unstable SSB6 connection between the two mated PCBs in the electrical packaging. In future
work this connection will be removed entirely, and the SNR of the recording array may increase
above that reported here.
The greatest advantages of this neural array, compared to other methods of recording
hippocampal activity, are the high density of anatomically placed electrode sites (compared to
microwires), and the mitigated immune response arising from the soft-polymer construction
(compared to silicon-based arrays). If the ratio of a single probe cross-sectional area to the
number of electrode sites on probe is computed, the value for microwires (Microprobes for Life
Science, Gaithersburg, MD, 25 µm diameter, 2.5-5 µm insulation) is 707-962 µm
2
/electrode
compared to 275-525 µm
2
/electrode for the device presented here. Each Parylene probe would
occupy less than half the volume of a microwire with the same number of recording sites, which
displaces less tissue and may attenuate the immune response against the foreign implant [26].
This ratio can be further reduced with additional miniaturization.
The anatomically-matched design of the electrode layout enabled recording from both
thin layers of the hippocampus simultaneously in all four subjects. Correct placement in both
layers was confirmed through the electrophysiological detection of complex spikes,
representative of hippocampal neural activity. Simultaneous multilayer recording is a key feature
of the array, and required precise positioning of electrodes during the design of the probe and
careful placement of the array during the implantation surgeries. Anatomical variations between
rat subjects served as a potential obstacle, however recordings from both the CA1 and CA3
layers in all four animals speaks to the success of the array design and efficacy of electrode
redundancy. The similarity in the final depth of the four arrays could allow for future
implantations to the target depth directly, rather than slowly monitoring intra-operative spike
signals, reducing the complexity of surgical targeting.
An initial investigation into the foreign body response against the flexible, Parylene array
was promising and indicated limited astrocytic scarring and neuronal death as compared to
literature values from more rigid (silicon) implants. The analysis of transverse slices, stained
with immunohistochemistry, of a hippocampus implanted with a sham hippocampal array for a
single month proved to be challenging. Traditionally, when evaluating the immune response in
89

the cortex, a transverse slice is binned into rings of increasing radii, and the increase in immune
markers or decrease in neuronal population is clearly measured according to the distance away
from the implant site since the tissue is homogeneous. However, in the hippocampus, neuronal
cell bodies are restricted to the cell layer itself, and astrocytes add structural support to the
tissue—and in the histological slices seen in Figure 55, are distributed mainly in the inner area of
the cell body layer. This complicates analysis of the immune response against the hippocampal
array, since probes that have successfully penetrated through the cell layer itself will naturally be
surrounded by greater populations of neurons and fewer astrocytes, and probes that have
deviated from the cell layer will show the opposite trend. To minimize the impact of this
heterogeneity on the histological analysis of the results, control regions for GFAP were chosen
from the same slice, at a location > 500 µm away from the implant site, at comparable offsets
from the cell layer. It is important to note, that for the hippocampal slice stained with GFAP,
percent astrocyte measurements in each radial ring increase according to distance away from the
probe, as rings located farther away include more of the astrocytes that hug the interior of the cell
body layer. For the hippocampal slice stained with NeuN, rectangular bins were matched to the
width of the cell layer in order to accurately calculate the change in concentration of neurons
within the cell layer itself.
It appears that astrocytes lining the inner surface of the DG have been activated within
the cell layer itself in an attempt to physically wall off the foreign implant from natural tissue.
This type of immune reaction against an implanted neural probe is commonly documented and is
believed to encapsulate the electrodes, electrically insulating them from surrounding neurons.
This begins at 2-4 weeks post-implantation, with a chronic, glial scar of astrocytes that grows
stable at 6-8 weeks post-implantation [27]. Death of nearby neurons occurs by 2 weeks post
implantation [5]. While some astrocytic scarring and neuronal death surrounding the
hippocampal shanks of this study are present, astrocytic concentration returns to normal at
distances ≥ 100 µm from the probe track and neuronal concentration returns to control levels in
the ~ 100 µm space between the edges of adjacent shanks. Histological studies that observe the
four week histological response surrounding flexibly tethered, rigid silicon implants indicate a
far greater disruption in neuronal density that only returns to normal within 300 µm [5] from the
implant surface and an increase in astrocytic concentration that can span beyond 500 µm [5]
from the implant surface. Studies that have directly compared histological data between flexible
90

and rigid probes uncover close to a two-fold improvement in astrocytic attenuation and neuronal
density surrounding the flexible implants. A comparison of the immune response to silicon and
softer nanocomposite probes implanted for 4 weeks found that neuronal densities return to
normal at 200 µm from the rigid implant and at a distance of 100 µm for the flexible implant
[28]. The histological data from our hippocampal array perhaps exceeds the results of this study
in that neuronal density returns to normal within the 100 µm space between probe edges of
adjacent probes. Flexible probes had twice the neuronal density of rigid ones in the first 50 µm
away from the implant [29] and GFAP levels were almost half that of rigid probes up to 50 µm
away from a flexible implant [30] though in this study no statistically significant difference in
neuronal concentration was witnessed. Both of these two factors—astrocytic encapsulation of the
neural probe and death of nearby neurons—are mechanisms that directly contribute to reduced
signal quality over time. However, these immune trends continue to evolve over time as
supported by this same study which observed no neuronal loss around the compliant implant at
16 weeks post-implantation [28]. This is a positive factor in predicting the continued chronic
performance of the hippocampal arrays presented in this paper.
Future directions
Future devices will be implanted in rat models and will be monitored chronically, until
unitary activities are no longer visible, to determine the recording lifetime of our Parylene arrays
and long-term SNR stability. Future modifications to the hippocampal array are underway and
include improvements in fabrication, design of the electrical interface between the array and
recording systems, and further histology staining and analysis to quantify the immune response
against the flexible hippocampal array over time. Sputter deposited Pt was found to have a high
compressive stress that caused array curling upon lift-off and the thickness of the insulation layer
of Parylene was adjusted from 10 to 14 or 18 µm to attenuate this curvature. Future devices will
use electron beam deposited Pt, so that the compressive stress of the system can be wholly
overcome through a thermal annealing step while maintaining the Parylene insulation thickness
at 10 µm, equal to the thickness of the Parylene base layer. Changes to the electrical packaging
will eliminate the need for two PCBs, and will adapt directly from ZIF to Omnetics in a single
PCB to eliminate the noise added to the system through the SSB6 connection while only
increasing the footprint on the rat’s skull by a manageable 10%. Future immunohistochemical
91

studies will stain for Cd11b, specific to microglia, the macrophages of the brain, in addition to
the traditional GFAP and NeuN staining to expand understanding of the immune players. Tissue
from the contralateral, native hippocampus will be used as a histological control to help account
for variations in cell layer thickness. Histological analyses will be performed at multiple depths
from the surface of the brain.
Summary
Sortable complex spikes were successfully recorded from the CA1 and CA3 layers of a
rat hippocampus simultaneously using a Parylene-based neural probe array. The array featured
eight electrodes on eight individual probe shanks, designed to match the variegated anatomy of
the hippocampus for study of the trisynaptic hippocampal circuit. Average spike amplitudes,
noise baselines, and SNRs compare favorably with microwire and silicon probes implanted in
hippocampal layers in terms of signal fidelity and sensitivity. Histological staining of tissue
surrounding the implant at one month post-implantation reveals that astrocytic density increases
near the insertion site (as compared to control), but returns to normal within 100 µm from the
implant site. Neuronal density was observed to decrease in the immediate vicinity of implant
sites, but returned to control levels in the ~100 µm space between probes of the same Parylene
array. Minimal damage to the neuronal tissue was observed in tissue slices at the one month
mark. This data supports previous reports that thin-film flexible probes may mitigate immune
response and tissue damage compared to neural implants made from stiffer materials.
This report describes one of the first examples of a polymer neural probe array recording
from deep brain structures, a feat made possible from the use of a dissolvable brace to reduce
probe buckling without the use of an insertion shuttle. The results bode well for the approach and
have motivated continued work exploring the polymer array as a solution to the challenges
facing chronic recording in deeper brain structures.
References
[1] G. Kook, S. W. Lee, H. C. Lee, I.-J. Cho, and H. J. Lee, "Neural probes for chronic
applications," Micromachines, vol. 7, p. 179, 2016.
[2] J. C. Williams, J. A. Hippensteel, J. Dilgen, W. Shain, and D. R. Kipke, "Complex
impedance spectroscopy for monitoring tissue responses to inserted neural implants,"
Journal of neural engineering, vol. 4, p. 410, 2007.
92

[3] A. Prasad and J. C. Sanchez, "Quantifying long-term microelectrode array functionality
using chronic in vivo impedance testing," Journal of neural engineering, vol. 9, p.
026028, 2012.
[4] V. Ntziachristos, "Going deeper than microscopy: the optical imaging frontier in
biology," Nature methods, vol. 7, p. 603, 2010.
[5] R. Biran, D. C. Martin, and P. A. Tresco, "Neuronal cell loss accompanies the brain
tissue response to chronically implanted silicon microelectrode arrays," Experimental
neurology, vol. 195, pp. 115-126, 2005.
[6] R. Biran, D. C. Martin, and P. A. Tresco, "The brain tissue response to implanted silicon
microelectrode arrays is increased when the device is tethered to the skull," Journal of
Biomedical Materials Research Part A, vol. 82, pp. 169-178, 2007.
[7] T. D. Kozai, Z. Gugel, X. Li, P. J. Gilgunn, R. Khilwani, O. B. Ozdoganlar, et al.,
"Chronic tissue response to carboxymethyl cellulose based dissolvable insertion needle
for ultra-small neural probes," Biomaterials, vol. 35, pp. 9255-68, Nov 2014.
[8] T. D. Y. Kozai, A. L. Vazquez, C. L. Weaver, S.-G. Kim, and X. T. Cui, "In vivo two-
photon microscopy reveals immediate microglial reaction to implantation of
microelectrode through extension of processes," Journal of neural engineering, vol. 9, p.
066001, 2012.
[9] T. Kozai, T. Marzullo, F. Hooi, N. Langhals, A. Majewska, E. Brown, et al., "Reduction
of neurovascular damage resulting from microelectrode insertion into the cerebral cortex
using in vivo two-photon mapping," Journal of neural engineering, vol. 7, p. 046011,
2010.
[10] K. Chung, J. Wallace, S.-Y. Kim, S. Kalyanasundaram, A. S. Andalman, T. J. Davidson,
et al., "Structural and molecular interrogation of intact biological systems," Nature, vol.
497, pp. 332-337, 2013.
[11] A. Greenbaum, K. Y. Chan, T. Dobreva, D. Brown, D. H. Balani, R. Boyce, et al., "Bone
CLARITY: Clearing, imaging, and computational analysis of osteoprogenitors within
intact bone marrow," Science translational medicine, vol. 9, p. eaah6518, 2017.
[12] M. Jorfi, J. L. Skousen, C. Weder, and J. R. Capadona, "Progress towards biocompatible
intracortical microelectrodes for neural interfacing applications," Journal of neural
engineering, vol. 12, p. 011001, 2015.
[13] S. F. Cogan, "Neural stimulation and recording electrodes," Annu. Rev. Biomed. Eng.,
vol. 10, pp. 275-309, 2008.
[14] S. Suner, M. R. Fellows, C. Vargas-Irwin, G. K. Nakata, and J. P. Donoghue, "Reliability
of signals from a chronically implanted, silicon-based electrode array in non-human
primate primary motor cortex," IEEE Transactions on neural systems and rehabilitation
engineering, vol. 13, pp. 524-541, 2005.
[15] K. A. Ludwig, J. D. Uram, J. Yang, D. C. Martin, and D. R. Kipke, "Chronic neural
recordings using silicon microelectrode arrays electrochemically deposited with a poly
(3, 4-ethylenedioxythiophene)(PEDOT) filmThis work was supported by the Center for
Wireless Integrated Microsystems NSF EEC-9986866 and the Whitaker Foundation,"
Journal of neural engineering, vol. 3, p. 59, 2006.
[16] S. A. Hara, B. J. Kim, J. T. Kuo, and E. Meng, "An Electrochemical Investigation of the
Impact of Microfabrication Techniques on Polymer-Based Microelectrode Neural
Interfaces," Journal of Microelectromechanical Systems, vol. 24, pp. 801-809, 2015.
93

[17] J. B. Ranck, "Studies on single neurons in dorsal hippocampal formation and septum in
unrestrained rats: Part I. Behavioral correlates and firing repertoires," Experimental
neurology, vol. 41, pp. 462-531, 1973.
[18] P. J. Gilgunn, R. Khilwani, T. D. Y. Kozai, D. J. Weber, X. T. Cui, G. Erdos, et al., "An
ultra-compliant, scalable neural probe with molded biodissolvable delivery vehicle," pp.
56-59, 2012.
[19] A. Lecomte, V. Castagnola, E. Descamps, L. Dahan, M. C. Blatché, T. M. Dinis, et al.,
"Silk and PEG as means to stiffen a parylene probe for insertion in the brain: toward a
double time-scale tool for local drug delivery," Journal of Micromechanics and
Microengineering, vol. 25, p. 125003, 2015.
[20] C. Hassler, N. Ehler, V. Singh, Y. Xie, N. Martini, R. D. Kirch, et al., "Fabrication and
implantation of hydrogel coated, flexible polyimide electrodes," in Neural Engineering
(NER), 2015 7th International IEEE/EMBS Conference on, 2015, pp. 561-564.
[21] D. Lewitus, K. L. Smith, W. Shain, and J. Kohn, "Ultrafast resorbing polymers for use as
carriers for cortical neural probes," Acta biomaterialia, vol. 7, pp. 2483-2491, 2011.
[22] B. J. Kim, J. T. Kuo, S. A. Hara, C. D. Lee, L. Yu, C. A. Gutierrez, et al., "3D Parylene
sheath neural probe for chronic recordings," J Neural Eng, vol. 10, p. 045002, Aug 2013.
[23] P. R. Patel, K. Na, H. Zhang, T. D. Kozai, N. A. Kotov, E. Yoon, et al., "Insertion of
linear 8.4 μm diameter 16 channel carbon fiber electrode arrays for single unit
recordings," Journal of neural engineering, vol. 12, p. 046009, 2015.
[24] A. Bragin, J. Hetke, C. Wilson, D. Anderson, J. Engel, and G. Buzsaki, "Multiple site
silicon-based probes for chronic recordings in freely moving rats: implantation, recording
and histological verification," Journal of neuroscience methods, vol. 98, pp. 77-82, 2000.
[25] R. J. Vetter, J. C. Williams, J. F. Hetke, E. A. Nunamaker, and D. R. Kipke, "Chronic
neural recording using silicon-substrate microelectrode arrays implanted in cerebral
cortex," Biomedical Engineering, IEEE Transactions on, vol. 51, pp. 896-904, 2004.
[26] T. D. Kozai, A. S. Jaquins-Gerstl, A. L. Vazquez, A. C. Michael, and X. T. Cui, "Brain
tissue responses to neural implants impact signal sensitivity and intervention strategies,"
ACS chemical neuroscience, vol. 6, pp. 48-67, 2015.
[27] 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.
[28] J. K. Nguyen, D. J. Park, J. L. Skousen, A. E. Hess-Dunning, D. J. Tyler, S. J. Rowan, et
al., "Mechanically-compliant intracortical implants reduce the neuroinflammatory
response," J Neural Eng, vol. 11, p. 056014, Oct 2014.
[29] J. Harris, J. Capadona, R. Miller, B. Healy, K. Shanmuganathan, S. Rowan, et al.,
"Mechanically adaptive intracortical implants improve the proximity of neuronal cell
bodies," Journal of neural engineering, vol. 8, p. 066011, 2011.
[30] H. C. Lee, F. Ejserholm, J. Gaire, S. Currlin, J. Schouenborg, L. Wallman, et al.,
"Histological evaluation of flexible neural implants; flexibility limit for reducing the
tissue response?," Journal of Neural Engineering, vol. 14, p. 036026, 2017.

94

Improving encapsulation and attenuating crosstalk
Crosstalk, which is the undesirable transfer of voltage signals between neighboring
channels, was observed in vivo. Electrodes within the same electrode group can realistically
record from the same neuron, since each electrode is separated by only 70 µm center-to-center.
The amplitude of neural spikes follow the inverse square law [1] and are generally discernable
up to ~ 140 µm from hippocampal neurons [2]. However, electrodes within the same probe that
target two different layers are separated by > 1.2 mm. Thus, simultaneous spikes across groups
on the same shank suggest the presence of signal leakage from one channel to another. This
leakage can happen through three different modes: capacitive, inductive, or conductive
coupling—which can either occur between adjacent channels or through “shunting” between
channels and surrounding fluid. This shunting can occur between electrodes or between traces in
the same probe, or between different probes if occurring along the Parylene cable (Figure 56).
Capacitive and inductive coupling between channels are unlikely to be the source of crosstalk in
the hippocampal array, however, as the width of the Pt channels and the size of the spacing
between them (5 µm width, 5 µm spacing) are too large.

LIFETIME TESTING
95

Figure 56: Crosstalk can occur (a) between electrodes, (b) between traces on the same probe, and
(c) between traces in different probes at the Parylene ribbon cable.

Ideally, crosstalk should be < 1% of the recorded signal in order for it to be negligible in
comparison to the noise of the system [3]. In this Parylene neural probe array, gross (> 70%)
crosstalk was observed from the first recording day across channels within the same probe, as
well as between channels located on adjacent probes (within 7-16 days post implantation),
(Figure 57) and clouded both the identification of signal origin and spike sorting processes.
Crosstalk between electrodes and traces in polymer neural probes as has not yet been quantified
in literature to the knowledge of this author. The hippocampal array discussed here has smaller
channel widths and spacing than other polymer probes which places greater demands on the need
for robust insulation between nearby traces.

96

Figure 57: Crosstalk between electrodes in CA1 and CA3 groups from the first chronically
implanted animal at 24 days post implantation. (a) Pictorial representation of spikes with
identical waveforms recorded from electrodes 31-40 (excluding 32). (b) Raster plot showing
most spikes occurring at identical timestamps, providing evidence of crosstalk.

Studies indicate that Parylene insulation can fail, as evidenced by reduced
electrochemical impedance, within days of soaking Parylene-platinum-Parylene devices in
phosphate buffered saline. The same is true of all polymers. Some liquid penetration is expected
for all polymers immediately upon soaking. What is crucial, however, is that adhesion between
layers remains intact in order to prevent conductive paths from forming. While the failure
mechanism for Parylene insulation has been attributed to solution penetration through the bulk
97

Parylene film that disrupts insulation [4] (Figure 58), this is not the complete picture. Thermal
annealing of Parylene-Parylene films, at temperatures greater than Parylene’s glass transition
point (60-90 °C) but below its melting temperature (290 °C), can reduce water diffusivity
through the Parylene by increasing its crystallinity [5, 6]. However, the decrease in impedance
cannot be solely attributed to a decline in the barrier characteristics of Parylene. Poor adhesion
between Parylene-Parylene and Parylene-metal layers, especially during long-term soaking
conditions, can compromise device functionality by causing conduits of conductive solution to
form between the layers and short adjacent channels together. Increasing the adhesion between
the Parylene-metal interface as well as increasing the hydrophobicity of the Parylene-solute
interface are two methods that have been shown to halt the decline of impedance over time (18
days, nonterminal experiment) [7] and may be effective in reducing crosstalk and maximizing
the lifetime of Parylene insulation.

Figure 58: Pictorial representation of one mechanism of crosstalk. Solution permeates bulk
Parylene and collects at voids or in areas of poor adhesion between the Parylene-metal-Parylene
interfaces. This forms conductive conduits between adjacent channels which causes the transfer
of voltage signals between neighboring channels.
Many studies have performed a targeted evaluation of the effect of adhesion promoters
on prolonging the lifetime of Parylene insulation through a mechanical analysis of adhesion
between dry and soaked Parylene layers and Parylene-metal interfaces [8-11]. A comprehensive
summary of these literature sources as well as original data exploring the use of AdPro Plus® (a
proprietary adhesion promoter from Specialty Coating Systems designed to increase Parylene
adhesion to metals), ethylene glycol diacrylate (EGDA) to improve Parylene-Parylene adhesion,
diamond-like carbon, and thermal annealing to extend the mechanical integrity between layers
over long term, room temperature wet soaking studies can be found in a current publication from
the Meng lab which was recently accepted for publication (“Methods for improving adhesion of
98

Parylene C films for dry & wet environments”). In this study annealed Parylene-metal-Parylene
coupons experienced interface failure at 2 weeks of soaking, but devices with an AdPro Plus
applied before the insulating Parylene layer were found to last up to 24 weeks. These are
promising results. It is logical to imagine that an increase in mechanical adhesion under soaking
conditions would translate into improved electrochemical isolation between neighboring
channels of metal sandwiched between two layers of Parylene. However, a direct evaluation of
electrical device integrity is necessary to evaluate whether or not these mechanical tests predict
long-term, electrical channel isolation.
To this end, a number of different methods designed to improve adhesion between the
Parylene-metal-Parylene interfaces were studied for their ability to promote electrical isolation
between metal channels of the hippocampal recording array. These strategies include the use of
AdPro Plus® (Specialty Coating Systems, Indianapolis, Indiana), as described above, to be
deposited on top of the base Parylene and metal device features prior to insulation with Parylene.
A treatment of dilute hydrofluoric acid was performed on another set of devices prior to Parylene
encapsulation in attempt to clean the surface and prevent the formation of voids during the
deposition of the insulative Parylene layer [12] as well a vacuum-based dehydration step of the
base layer of Parylene prior to Parylene encapuslation. Additionally, plasma treatment with gases
other than oxygen can be explored as a way to improve inter-layer adhesion [13]. Plasma
treatments are posited to benefit adhesion through roughening the surface and increasing the
rugosity of the base Parylene, which can promote mechanical interlocking between the base and
insulative Parylene layers or by creating free radicals which can then bind to future Parylene
monomers. The use of a metal adhesion layer between the Parylene and platinum surfaces holds
promise towards attenuating crosstalk as well [14]. A variety of these techniques will be
explored to determine which process may be best suited for the long-term, electrical integrity of
the Parylene hippocampal array.
Benchtop analysis of array integrity over time
Long term crosstalk tests
For a material to provide lasting physical and electrical insulation to metal traces, it must
have properties that include: robustness to the saline-based physiological environment, low
99

moisture intake, low dielectric constant, pinhole free deposition, low film stress so it does not
delaminate or crack, and adequate adhesion to the metal and substrate. Four methods to monitor
encapsulation integrity are mentioned in literature. These include impedance spectroscopy,
measuring DC leakage current, measuring shunt and coupling capacitance, and, lastly, applying
either a current or voltage signal to a single channel and measuring the resultant current or
voltage signal that leaks across to an adjacent channel. These techniques were applied to study
the properties of polymers used to encapsulate rigid substrates, but have yet to be studied
rigorously on free-film polymer devices with metal traces.
Impedance spectroscopy reveals electrode properties over a range of frequencies, in
contrast to the other three methods above, and is commonly used to monitor encapsulation
properties [15-21]. This is an advantage as some electrode changes occur with greater sensitivity
at particular frequencies, and may be missed entirely if monitoring solely at one frequency [15].
By creating an electrical circuit model to fit the impedance data, insight can be gained into the
physical mechanism of delamination or crosstalk. When channels are fully insulated and
electrodes are not exposed, only capacitive signals can escape and EIS measurements should
always have a phase of -90°. Deviation of this phase towards resistive values (closer to 0°)
indicates possible compromise of the insulation layer and suggests the emergence of a new,
conductive pathway. Impedance magnitude will also decrease in value as insulation failure
occurs and resistance to current flow dwindles. In one such study, any deviations in impedance
modulus below 0.1 GΩ, or a phase > -80 °, at frequencies greater than 1 Hz were automatically
considered to represent failed insulation in a study with interdigitated electrodes on a silicon
substrate [15]. While impedance results are clear for encapsulated channels, they may be more
challenging to interpret for channels with exposed electrodes, like in the Parylene-based
hippocampal array discussed here.
The second technique used to evaluate insulation integrity entails measuring the direct
current (DC) current leakage between encapsulated interdigitated electrodes, or coils, as current
flows between them [5, 15, 22-24]. When insulation is intact, leakage current is small, but
increases by many orders of magnitude when insulation fails. One group adopted this technique
and modified it for interdigitated comb electrodes that connect to an inductively coupled circuit.
Changes in the resonant frequency and the Q-factor of the circuit were then directly mapped to
moisture intrusion in this study [25]. Challenges in this technique include difficulties in setting
100

the threshold for current values that represent failure, as this must be experimentally determined
for each device, and this technique inherently requires the traces tested to be physically large
enough to have resolvable leakage currents.
Shunt capacitance, between an embedded trace and solution surrounding the device, and
coupling capacitance, between adjacent traces in the same device, can sometimes be directly
measured to calculate the amount of crosstalk and signal sharing expected with that particular
device configuration and insulation materials [26, 27].
Many groups choose apply an electrical signal to a single channel in a device and to
monitor how much of that signal, if any, is unintentionally transferred to adjacent channels
through leakage paths [28-34]. This test can be performed on any device of any configuration. It
can be performed directly in solution if the electrode is fully encapsulated, and can be performed
outside of solution under dry conditions if the electrode is exposed. Performing this test with a
voltage signal has the benefit of insuring that the potential applied across that channel remains
within the safety of the water window. This has been adopted as the primary method of
monitoring device integrity in this work, but variations of the other three techniques have also
been implemented as will be described below.
Experimental Methods
To evaluate how long adjacent channels in hippocampal arrays remain electrically
insulated and isolated from one other in soaking conditions, an automated testing system was
developed by the Meng lab to quantify the magnitude of voltage leakage between channels. A 1
kHz voltage sinusoidal signal with an amplitude of 0.5 V (against ground) was applied to the
contact pad of a single electrode channel of a dry array through a waveform generator
(VirtualBench tool, National Instruments, Austin, TX). Simultaneously, the voltage of adjacent
channels were measured with respect to ground. This process repeated itself until a voltage
signal had been applied to all channels once. Crosstalk measurements were calibrated by
measuring crosstalk inherent to the measurement system itself prior to each experiment and
subtracting these values from the crosstalk measured in each array. In the case of intact Parylene
insulation between channels, no voltage signals should be measurable from neighboring
channels. In this ideal case, crosstalk would be represented by Figure 59, where crosstalk levels
are 100% when voltage signals are sent to and read from the same channel, but are zero
101

everywhere else. Compromised Parylene insulation, which may allow solution to permeate the
array and collect between channels, is likely to result in signal leakage. In this case, neighboring
channels will carry voltage signals as well. Percent crosstalk was calculated to be the ratio of
Vread,n /Vsend multiplied by 100 for each of the neighboring electrodes surrounding the channel
through which the voltage signal was applied. Crosstalk measurements were performed on arrays
that had been temporarily removed from solution and dried off so as not to include voltage
transferred from one channel to another through solution via the exposed electrodes.

Figure 59: Description of crosstalk measurement in ideal circumstance, where each channel is
fully insulated from its neighbor. In this case, the only cells in the crosstalk graph that will have
100% crosstalk occur across the diagonal, which is where the voltage signal is sent to and read
from the same channel. Crosstalk between neighboring channels should be zero, but is usually <
5% due to noise in the system.
In order to replicate long-term, in vivo, soaking conditions, arrays were immersed in
phosphate buffered saline (1 × PBS) contained in glass vials and maintained at body temperature
(37 °C). Initial crosstalk tests on two groups of eight channels in a traditional, annealed
hippocampal array revealed the immediate presence of crosstalk that gradually increased as
soaking was extended from 1 to 8 days and remained steady at high values until day 20 (Figure
61). The first set of channels underwent only CV cleaning and no EIS testing before the crosstalk
experiment. The second set underwent CV cleaning and EIS testing prior to experimentation.
These arrays were packaged as demonstrated in Figure 60. Arrays were connected to a PCB that
had a ZIF connector on one end and pads for two Omnetics connectors on the other end. Wires
were hand soldered to two groups of eight, consecutive pads of the Omnetics pads on the PCB on
both the right and left sides of the PCB, sets A and B. Wires were soldered to header pins. All
exposed metal surfaces were covered with marine epoxy to attenuate water penetration through
the packaging. Arrays in glass vials were warmed in a covered water bath and header pins were
covered with Parafilm (Bemis NA, Neenah, WI) in attempt to prevent moisture from
102

accumulating in the header pins. During intermittent inspections, however, it was evident that
moisture did indeed collect on the header pins due to condensation in the covered water bath. In
order to determine if crosstalk was solely caused by moisture penetration into the hippocampal
array itself or whether crosstalk values were being confounded due to imperfections in the
packaging, the Parylene cable that extended through the top of the glass vial into the PBS
solution was cut at day 20. The crosstalk test was repeated and revealed intermediate and high
levels of crosstalk remaining in the packaging of sets A and B of the hippocampal array (Figure
62). This indicated that the packaging was susceptible to water permeation and that crosstalk
measurements were a mixture of crosstalk inherent to the array itself as well as crosstalk
occurring through the packaging.

Figure 60: Packaging for initial crosstalk tests. To connect between the PCB and the crosstalk
system wires were hand-soldered to PCB pads and header pins. All metal surfaces were covered
with marine epoxy (not shown). Test vials with attached packaging were placed inside a covered
water bath set to 37 °.
103

Figure 61: Crosstalk present immediately upon soaking annealed arrays in initial crosstalk tests.
Crosstalk increased in magnitude until day 8 and remained steady until the experiment ended at
day 20. Channels in set A had CV cleaning only and no EIS testing before the crosstalk
experiment. Channels in set B underwent CV cleaning and EIS testing prior to experimentation.

Figure 62: Crosstalk remaining in packaging after hippocampal cable and probes were cut off.
Average and standard deviation of crosstalk in set A decreased from 66% ± 12% to 19% ± 10%
and crosstalk in set B decreased from 95% ± 2% to 66% ± 9%. This indicates that the previous
crosstalk measurements were a combination of crosstalk inherent to the hippocampal array itself
as well as crosstalk in the packaging.
104

In order to isolate crosstalk to the array itself, improvements in packaging and
modifications to soaking conditions were performed. During testing it was noted that water
condensed in between adjacent header pins while inside the covered water bath. To address this,
future tests used jumper cables, which functioned as header pins that were isolated from one
another, to attach to the crosstalk setup. Additionally, devices were moved from a covered water
bath and were instead heated to 37 °C by resting in a water bath (confirmed through a
temperature probe) on a hotplate (Figure 63). This allowed for the jumper cables to protrude
from the wet bath environment and limited their risk of exposure to moisture penetration. In
crosstalk tests performed on six sets of packaging—three with no array inserted into the ZIF, and
three others with PEEK-backed Parylene inserted into the teeth of the ZIF, but lacking a cable
and probes, all sitting above PBS in glass vials—there was no crosstalk present beyond noise
levels at 114 days of soaking (test ongoing).

Figure 63: Improved packaging and soaking conditions for long-term crosstalk tests. (a)
Electrical connections from the PCB to the crosstalk system were made via isolated jumper
cables rather than header pins which could easily lead to shorts between channels if moisture
penetrated any of the header pins. (b) Complete packaging set-up: Probes (not shown) soaked in
a 1x PBS filled glass vial. Parylene cable passed through a slit drilled into the glass vial cap and
contact pads insert into the ZIF of the PCB. The vial cap and all exposed metal connections were
protected with marine epoxy. Jumper cables soldered to the back-end of the PCB end in male
connectors that were inserted into the crosstalk system. (c) Packaged vials rest in a beaker with
water that is heated to 37 °C atop a hotplate. In this way packaging is not susceptible to moisture
condensation and permeation.
105

Once it was confirmed that improvements in packaging and soaking conditions isolated
crosstalk to within the array itself, a new set of long-term crosstalk tests were performed. A
control array and three sets of eight channels from two arrays fabricated with the experimental
conditions described below were packaged and soaked in vials of 1x PBS placed in a beaker of
water heated to 37° C by a hot plate (Table 4). Two arrays were treated with AdPro Plus®

between the metal layer and the Parylene insulation layer. Two arrays were dehydrated overnight
in vacuum prior to the deposition of a Parylene insulation layer, with no N2 backfill. Two others
were cleaned by soaking in diluted 4.9% hydrofluoric acid (HF) for two minutes prior to the
deposition of the Parylene insulation layer. All arrays except for those treated with AdPro Plus®
were annealed at 200 °C for 48 hours in an O2 free environment purged with N2 as this
temperature driven process had previously been shown to impair the function of AdPro Plus®.
Two sets of eight channels from the left and right side of the array were monitored from a single
array of each type; a single set of eight channels was monitored on the remaining array. The term
“channel” refers to the electrode site, trace, and electrical packaging that isolates each single,
independent, recording unit from its peers. Arrays were CV cleaned and EIS tested prior to
packaging in order to find eight consecutive, functioning channels that spanned two probes
(Figure 64). This channel arrangement was chosen in order to monitor both intra- and inter-probe
crosstalk. Arrays were packaged as described above. Crosstalk measurements were performed in
dry conditions before soaking, after four hours of soaking, and then daily for the first week,
triweekly until week three, and then weekly for the remainder of the study. PBS solution was
replaced monthly. Prior to testing, probes were removed from the PBS, cleaned with Millipore
water, and dried—a process that took less than five minutes, too short a time for solution that
ingressed between layers to evaporate from the array. By monitoring the time point at which
crosstalk becomes evident between neighboring channels we can compare which methods of
adhesion promotion were more effective at attenuating crosstalk and use this to help guide future
array fabrication.
Table 4: Fabrication conditions of arrays used in long-term soaking tests.
Fabrication Conditions
1 Control Annealed
2 AdPro Plus® under insulation layer Non-annealed
106

3 Dehydrated overnight prior to deposition of insulation layer Annealed
4 Soaked in 4.9% HF for two minutes prior to deposition of insulation layer Annealed

Figure 64: Eight consecutive, physically adjacent traces, chosen to span two probes.
Results
The control array failed most quickly; > 70% crosstalk was evident suddenly on day 43
of testing in channels 2- 6. By day 50, > 80% crosstalk was present in channels 1-6. Crosstalk in
channels 7 and 8, which lie in the adjacent probe, grew up to 40% by day 57 and while hovering
around those values, seemed to decrease slightly until day 106 (Figure 65, testing ongoing).

Figure 65: Crosstalk in control array displayed at time points (in days) where changes occurred.
Crosstalk was not present until day 43, when catastrophic signal leakage between channels 2-6
occurred. At day 50 this crosstalk spread to include channel 1 as well. Days 57- 106 exhibit an
odd trend, that of crosstalk magnitude decreasing over time.
107

One set of traces that underwent a dehydration step prior to the deposition of an
insulative layer of Parylene developed > 60% crosstalk in channels 3-5 and spread to channels 1-
5 (> 50% crosstalk) by days 84 and 90 respectively. It remained stable for the remainder of the
experiment. The other two dehydrated sets of traces reached day 106 of testing without any
signal leakage developing between channels (Figure 66).

Figure 66: Crosstalk in dehydrated sets of traces over time. Crosstalk in set A appeared at day 84
of the soaking test, and spread from channels 3-5 to include channels 1 and 2 as well by day 90.
Crosstalk remained at these levels until day 106. Sets B and C experienced no significant
crosstalk for the duration of the experiment.
One HF-treated developed an electrical short between channels 1 and 3 at day 1 of
testing. This was indicated by 100% crosstalk between channels 1 and 3, whereas crosstalk
between channels 1 and 2 and channels 2 and 3 were within the noise of measurement. This set
of shorts remained electrically isolated from neighboring channels until day 50, at which point a
108

nidus of crosstalk connected channels 1-3 and eventually stretched to span channels 1-6 at day
90. At day 106 channels 7 and 8 have little (<15% crosstalk) with their neighbors. The second
HF-treated array remained crosstalk free until day 64, at which point >50% crosstalk developed
in channels 1-6, increasing only slightly in magnitude on the testing days since. The last HF-
treated set of traces remained crosstalk free at day 106 of the study.

Figure 67: Crosstalk in HF-treated set A over time. A short between channels 1 and 3 developed
by day 1 of testing. Crosstalk between the shorted channels only developed at day 50 and
gradually expanded over the course of the experiment to include channels 6-8 on the same probe
of the hippocampal array.
109

Figure 68: Crosstalk in HF-treated set B over time. Crosstalk was not present until day 64 at
which point catastrophic crosstalk levels (>50%) became evident in channels 1-6 of the same
probe. Crosstalk levels fluctuated slightly but mostly remained steady from day 71 to day 106 of
the experiment.
Two of the three sets of traces treated with AdPro Plus® never developed crosstalk, while
a single set developed low levels of crosstalk (<40%), focused in channels 3-6, that became
evident at four hours of soaking, fluctuated from week to week, and seemed to disappear
completely by day 57 of testing.
110

Figure 69: Crosstalk in AdPro Plus® treated arrays over time. Crosstalk of magnitude < 40%
became evident in set A after only 4 hours of soaking and fluctuated over time until it seemed to
disappear completely by day 57. The other two AdPro Plus® sets of traces did not have crosstalk
for the duration of the experiment.
Signs of delamination or water penetration were challenging to visualize due to small
feature sizes and the highly-reflective, dimpled pattern of Parylene surfaces caused by transfer of
texture from Teflon sheets; the sheets were use to apply pressure while preventing bonding to the
fixture during annealing. Additionally, the Parylene cable of each array lies was obscured by the
opaque glass vial caps, preventing visual inspection. Two cases where channel delamination and
blister formation were readily visible were in dehydrated sets of channels A, B, and C at day 113
of the soaking study, as shown in Figure 70. The presence of delaminated, wavy channels as well
as blisters are indicated in these photographs. This delamination was not visible during
111

microscopic inspection of the channels last performed on day 71 of soaking. Their positional
correlation to the channels tested in sets A, B, and C are noted as well.

Figure 70: Microscopic images of channel delamination and blister formation that correspond to
the presence of crosstalk in dehydrated sets imaged at day 113 of the crosstalk experiment. The
highly reflective Parylene surface is caused by the dimpling of array caused by the annealing
process. Colored lines indicate the eight consecutive channels tested during crosstalk
experiments. Channels in dehydrated set A corresponded to regions of poor adhesion between
layers, whereas channels in dehydrated set B appeared unaffected by this phenomenon. For
dehydrated set C, delamination and blistering occurred at locations far away from the channels
tested.
Discussion
Another graphical summary of channels tested as well as which channels experienced
crosstalk is presented in Figure 71. The first array to experience crosstalk levels of > 50% was
the control array at day 43 of soaking. To put this in perspective, crosstalk in vivo was already
present within 7-14 days post implantation for the control, annealed device with e-beam
deposited metal. This could indicate that some of the crosstalk witnessed in vivo may have been
due to failures in packaging rather than inherent limitations in device adhesion as devices tested
at the benchtop did not fail this quickly. Two HF-treated set of traces began to experience
crosstalk of this magnitude at days 50 and 64 of testing. A single dehydrated set showed
112

crosstalk at day 84 of testing. Interestingly, crosstalk appeared suddenly in each of these devices,
with a number of channels experiencing > 50% crosstalk simultaneously. No other sets of traces
experienced crosstalk at this time, with the exception of a single AdPro Plus® treated set whose
crosstalk waned over time. All ten of these sets of traces tested lasted longer than a month
without experiencing crosstalk and five of the sets (two HF treated arrays, and three AdPro
Plus® arrays) did not experience significant crosstalk over the 106 day duration of this
experiment. This indicates that Parylene encapsulated devices have the potential to remain
undisturbed in simulated in vivo conditions for greater than three months and indicates variability
across treatment conditions.

Figure 71: Summary of channels tested (solid blue) during crosstalk experiments across arrays
fabricated with different treatment steps and which channels experienced crosstalk by day 106 of
the ongoing experiment (highlighted in purple).
Limitations in drawing conclusions from crosstalk results will be discussed in the next
section. Since the sample size in this study was limited, it may be too early to draw conclusions
as to the preferred performance of arrays that had been dehydrated prior to the deposition of the
second layer of Parylene. However, a simple perusal of the data seems to indicate that arrays
fabricated with AdPro Plus® had the longest lifetimes in this in vitro, benchtop soak.
Since arrays treated with the same experimental conditions differed in performance from
one another, it is important to consider that quality control issues may play a role in the lifetime
113

of these arrays. Therefore, it is valuable to explore methods that will assist in determining the
quality of a particular array before its use, perhaps through electrochemical studies, that may be
able to predict which arrays are more prone or resilient to the development of crosstalk, as will
be discussed in section 5.2.3. Some quality controls that may be useful in improving array
adhesion include assuring cleanliness and proper photoresist removal in between processing
steps, monitoring the age of Parylene layers, and tracking whether or not the Parylene deposits
perfectly conformally or if Parylene spherules could possibly represent voids in Parylene layers.
While crosstalk generally increased in magnitude over time, both the control array and
array A from the AdPro Plus® devices, and, to a lesser degree, array B that was treated with HF,
experienced a fluctuation in crosstalk magnitude over time. Sometimes drying off arrays and
connecting the jumper cables to the crosstalk setup can take longer than other times. It is feasible
that longer waits prior to crosstalk measurements to allow for the evaporation of solution that
gathered between channels leading to smaller amounts of crosstalk. In order to determine
whether this fluctuation in crosstalk could be due to this form of experimental error, the control
array with high levels of crosstalk was allowed to dry for longer than usual and crosstalk was
tested at 5, 8, and 10 minutes post the initial crosstalk test, and then at 5 minute intervals.
Crosstalk values remained plateaued for the first half of testing and then began to drop at greater
and greater speeds. At 40 minutes, crosstalk had only dropped to 81% of its initial value (Figure
72). Since unintended pauses between array testing were not longer than 10 minutes, evaporation
of solution that penetrated the array is unlikely to occur in that time scale and is not likely to be
114

responsible for the weekly fluctuations in crosstalk that were sometimes observed above. At this
point the source of these fluctuations is unknown.

Figure 72: Time scale of solution evaporation from array. Repeat crosstalk tests performed in dry
air after the array had been removed from soaking in 37° C PBS solution. Each consecutive
matrix was divided by the original crosstalk matrix to yield an average percent ± standard
deviation of the original crosstalk measured in the array (N=56).
While physical signs of delamination in the form of blistering and wavy Pt became
evident in some arrays at day 113 of soaking, they were challenging to visualize and appeared
much later than crosstalk became evident electrically. Optical microscopy is insufficient to
monitor array adhesion and status of the encapsulation. The photographs presented in Figure 70
serve as a reminder that monitoring only 8 or 16 out of a total of 64 traces does not yield a
complete picture of whether or not delamination is present anywhere across the array. In fact, set
C of the dehydrated traces revealed no crosstalk, but photographs of the array reveal
delamination and blistering on that array at locations far away from the channels tested.
Reproduction of these experiments on a more comprehensive set of channels (perhaps one
channel from each probe) and more arrays treated with each experimental condition may reveal
which treatment method works best.
115

Multimeter resistance tests
During long-term crosstalk experiments it was noted that channels that were not
connected to the crosstalk system revealed crosstalk levels that were within the noise of the
system, indicating “zero” crosstalk. Therefore, it was necessary to institute a control to confirm
that crosstalk results were meaningful, and that null crosstalk did not result from disconnections
between the crosstalk system and the contact pad of a channel.
Experimental Method
To check connectivity, resistance tests between two jumper cables of an array soaking in
the same glass vial used for the above crosstalk experiments were performed using a handheld
digital multimeter (Fluke Corporation, Everett, WA) connected to a jumper cable of each
channel using alligator clip cables. The average DC resistance across two, intact adjacent
channels in PBS heated to 37 °C was 12.7 ± 0.2 MΩ. If an electrical short was present between
two channels the resistance would drop to the order of magnitude of Ωs. In a case where one or
more of the channels was disconnected, the resistance between the two channels would be too
high to measure and the multimeter would indicate “OL” or overload. When crosstalk between
two channels was present, resistance dropped to the single digit MΩs.
DC resistance between channel 1 of arrays and each fellow channel were recorded from
some arrays before the beginning of long-term soaking for crosstalk tests and for all arrays at
various time increments (days 8, 44, 67, and 90) during crosstalk testing. Channel 1 was
consistently used as the reference unless the multimeter indicated all seven pairs of
measurements as “OL” in which case each of the other seven channels were tried as possible
references. Thus, if channel 1 was no longer connected, resistance measurements would still be
achievable between the remaining channels measured with respect to a new reference.
Results
At day 8 post-soaking it was evident that channel 6 of AdPro Plus® Array C was not
electrically connected as it measured “OL” with a working channel 1. It also was clear that there
was an electrical short between channels 1 and 3 of Array A of the HF treated devices as it gave
a resistance measurement of 4 Ω with respect to channel 1 (Figure 73). This short was likely
caused by contact between the jumper cables at the point at which they were soldered to the PCB
116

prior to marine epoxy encapsulation as if this short was due to failed lift-off between two
features in the microfabricated array, there would be no way to short between channel 1 and 3
while skipping channel 2. As expected, resistance measurements of these shorted and open
channels remained consistent throughout the duration of the trial.

117

Figure 73: Multimeter resistance tests in PBS over time. Channel 1 used as reference except for in AdPro Plus devices where
connections seemed to be lost by day 90. Mid-teens MΩ impedance indicated proper connection. Impedance on scale of Ω is
short circuit, and OL symbol = overload indicates that impedance between channels is too high to measure (cell background
colored in white). Resistance between channels listed in column header, in units of MΩ. Darker blue indicates lower
resistances.

118

Resistance measurements of the control array indicated intact metal channels and intact
insulation on day 8 post-soaking. However, resistance values dropped to single digit MΩs by day
44 of testing. Resistance measurements from the control array increased slightly on day 90, but
did not return to original levels.
DC resistances across all channels in dehydrated sets A, B, and C remained steady across
67 days of this experiment. On day 90, only set A exhibited decreased resistances (≤ 10 MΩ)
between channels 1-5.
As mentioned above, HF treated array A did indeed consistently measure a short between
channels 1 and 3 over all four testing days. Resistance levels between channels 1 and 2, 4 and 1
and 1-6 were smaller than usual at days 67 and 90 of soaking respectively. For array B,
resistance measurements did not drop until day 67. The first measurement, between channels 1
and 2 taken on array C on the first day of multimeter resistance testing revealed a value of 5 MΩ,
but in subsequent days of testing the resistance varied from 11 to 12 MΩ which more closely
match control resistance levels.
AdPro Plus® treated arrays had consistent, control resistance values until day 90.
Surprisingly, on day 90, all channels indicated “OL” when tested in reference to channel 1 and in
reference to all other channels on the array. This same anomaly occurred on channels 2, 3, and 4
of array B, and were a new development for channel 5 of array C as well.
Discussion
The decrease in resistance values in the control array that became evident at day 44 of
testing corresponds well with the development of crosstalk that was noted at day 43. The fact
that each pair of channels provided a resistance measurement on the order of MΩs confirms that
the crosstalk results are an accurate representation of the status of channel isolation, and were not
errantly recording “zero crosstalk” in days prior. The same held true for dehydrated arrays and
arrays that were cleaned with HF prior to the deposition of the Parylene insulation layer:
resistance measurements dropped in magnitude at time points that correlate to when crosstalk
began to appear in the above set of experiments. For dehydrated set A this occurred by day 90,
which fits well with the appearance of crosstalk noted on day 84. For HF-treated sets A and B
low resistance values become evident at day 67 of testing, and crosstalk appeared at days 50 and
119

64 respectively. This general trend of resistance measurements falling as crosstalk develops is to
be expected. Crosstalk likely occurs through paths of solution that penetrate the Parylene and
electrically connects the two channels together. This additional path from one channel to the next
would result in lower resistances between the two channels as well. The 5 MΩ resistance
measure between channels 1 and 2 of HF-treated array C seems likely to be an experimental
error, perhaps due to the alligator clips not attaching fully to the jumper cables, which are often
seen to cause resistance values to fluctuate wildly.
Resistance measurements from the control array drop on day 67 and increase, but still not
to control levels, at day 90. If array delamination is a progression, it is hard to imagine why
resistances would increase over time instead of continuously decreasing. Perhaps this can be
explained by array cleanliness. The arrays were not removed from PBS-filled glass vials and
cleaned before resistance testing. It is possible for salt to have deposited on the electrode surface
over time, thereby impeding the path of current between one channel and the next.
Perhaps the most curious result of this experiment, as well as the most impactful result,
may be the “OL” measurements from the AdPro Plus® treated arrays. It is curious why many
channel connections would suddenly be disrupted after three months of soaking and why this
phenomenon only occurred in arrays treated with AdPro Plus®. These multimeter resistance
tests, combined with null crosstalk data across all three of these arrays at day 90 puts into
question whether any of the crosstalk results after day 90 on channels with “OL” readings can be
trusted since channels that are not connected will still give “zero” crosstalk readings. One
difference between the AdPro Plus® devices and the other arrays is that the former lacked an
annealing treatment. Perhaps annealing, by increasing the crystallinity of Parylene, increases the
robustness of the Parylene supporting the thin Pt contact pad that connects to the ZIF pins and
makes this connection more resilient and less prone to ZIF pins puncturing through the thin
metal-polymer interface.
Overall, resistance tests are an essential method of confirming the accuracy of crosstalk
results. They serve as a way to confirm that meaningfulness of null crosstalk results and should
be performed alongside crosstalk experiments.
120

Correlation between crosstalk and electrochemical data
Originally electrochemical evaluation of channels that were part of the crosstalk tests
were not performed out of concern that the application of currents and voltages involved in EIS
and CV testing might promote delamination and hasten the arrival of crosstalk. Once gross
crosstalk was present in certain arrays, however, this concern was abated. Therefore, EIS and CV
curves of the control array, two dehydrated arrays, and a single array treated with HF at day 97
of soaking. Current ranges and impedance and phase trends of CV and EIS curves respectively
were compared between each set of channels and correlated to crosstalk measurements taken
close to that time period.
Experimental Methods
CV (-0.2 V to 1.2 V, scan rate of 250 mVs
-1
) was run on each individual electrode channel
for 5 cycles in 0.5 M H2SO4 purged with N2 for five minutes prior to scanning. A 1 cm
2
Pt plate
was used as a counter electrode and an Ag/AgCl electrode was used as a reference electrode. EIS
was performed in 1 × phosphate buffered saline (PBS) (OmniPur 10 × PBS, EMD Chemicals,
Darmstadt, Germany) with an excitation voltage of 25 V (AC) over frequencies from 1 Hz to 0.1
MHz with an Ag/AgCl reference and 1 cm
2
Pt counter electrode.
The typical shape of a CV curve of a Pt electrode in H2SO4 is presented in Figure 74. The
peaks in this curve are attenuated slightly due to the small diameter of the hippocampal electrodes
(30 µm exposed).
121

Figure 74: Typical CV curve for Pt electrodes immersed in 0.05 M H2SO4 with an Ag/AgCl
counter electrode. Peaks and humps indicate the potential at which H+ adsorption and desorption
occur and Pt-oxide is formed and reduced.
The expected shape of a healthy EIS curve of an electrode immersed in solution is
presented in Figure 75. According to a simplified Randles equivalent circuit model of an
electrode in solution [35], the electrode surface can be modelled as a capacitor, Cdl, which
represents the Helmholtz double-layer of ionic charges, in parallel with a resistor, Rct, that
represents charge-transfer reactions. These two circuit elements lie in series with another resistor
element that represents the solution resistance, Rs. At the low frequency regime, the capacitor
acts as an open circuit. Therefore, the Randles model simplifies to the charge transfer resistance
in series with the solution resistance. In the high frequency range, (> 100 kHz for Pt electrodes of
the size used here), the capacitive element is shorted, and the Randles model can be simplified to
a single element circuit composed of a resistor modeling solution resistance. This model matches
the shape of EIS impedance magnitude and phase curves well. As the stimulus begins in the low
frequency regime, the impedance magnitude is plateaued at a maximum value equal to Rs + Rct
while the phase curve is close to 0°, since the system is purely resistive. As the frequency
increases, the impedance drops linearly as capacitive reactions begin to occur, which is matched
by a capacitive phase shift of almost -90°. In the third region of the curve, the impedance
magnitude again plateaus at a value representative of Rs and the phase returns to close to 0°.

122

Figure 75: Typical impedance magnitude and phase curves from EIS data of a well-insulated
microelectrode.
Once delamination of channel insulation occurs, however, CV and EIS curves can change
quite dramatically, as shown in Figure 76. These data were collected from an array of advanced
age, of uncertain processing history, that appeared to delaminate within the ~30 minutes of
electrochemical testing performed on it. As CV testing proceeded, each subsequent electrode
tested appeared to have a larger CV current range than the prior electrode. Upon inspection,
wavy channels and delamination was apparent (Figure 77). CV curves taken prior to
delamination had an average current range of -123 ± 9 nA to 110 ± 23 nA, curves taken after
delamination had occurred had an average current range of -1,350 ± 106 nA to 1,400 ± 82 nA.
As the array delaminates, it seems that adjacent channels act as one single, larger electrode
which allows for more current flow through it during voltage application in CV tests.
Additionally, this newer, “larger” electrode has a higher capacitance than a single, small, isolated
electrode. As C increases in an RC circuit, the time taken to charge the capacitor increases as
well. This means that lower frequency signals can now act to “short” the capacitor earlier than
before, as confirmed by the transition to the resistive regime of intact electrodes occurring ~ 10
5

Hz (Figure 75) but occurring as early as 10
2
– 10
3
Hz for delaminated channels (Figure 76).
123

Figure 76: Changes to CV and EIS curve that occur with catastrophic delamination. (a) CV
current range increases by two magnitudes. Green curve taken from CVs prior to delamination,
purple curve taken from CVs post delamination. (b) EIS after delamination occurred. Impedance
magnitude drops from ~ 600 to 1 kΩ. Phase fails to return to resistive regime at lower
frequencies. (n=3, error bars represent standard deviations for both panels).

Figure 77: Channels that delaminated during electrochemical testing of same array as in Figure
76, perhaps due to the advanced age of devices. (a) “Wavy” channels at the first stage of
delamination on the Parylene ribbon cable, (b) complete dehiscence of Pt channels from Parylene
base of ribbon cable, (c) delamination present at probe tips. Double arrows point to separation
between base and insulation layer of Parylene.
Results
Figure 78 includes a summary of crosstalk data from day 90 and CV and EIS curves
taken at day 92 for the control array, two dehydrated arrays, and a single HF-treated array used
for the crosstalk and multimeter resistance tests described above.
124

Column A of Figure 78 presents data from dehydrated set C at which no crosstalk was
measured for the duration of the soaking study. The CV curves for channels 2-8 are typical for
those recorded from hippocampal array electrodes. Oxidation and reduction peaks and hydrogen
adsorption and desorption peaks are readily visible on this cyclic voltammogram. The current
ranges from ~ 20 to -15 nA. The EIS impedance magnitude and phase curves are also typical of
an electrode whose channels are fully insulated from neighboring channels. The average 1 kHz
impedance for these electrodes is 437 ± 43 kΩ (stdev) and the phase curves are capacitive at
middle frequencies and resistive on either end of the frequency spectrum. Channel 1, however,
had a CV curve with a very small range of current, an impedance magnitude greater than the
usual, and its phase did not return to the resistive regime at high current stimulation frequencies.
Data in column B of Figure 78 was taken from dehydrated set A that began to show
crosstalk on day 84 of testing. Channels 1-5 which measured high levels of crosstalk had CV
curves with larger ranges than channels 6-8 which had minimal crosstalk levels. With regards to
EIS curves, channels 1-5 had impedance curves shifted down from their neighboring channels.
The 1 kHz impedance of channels 1-4 was 125 ± 9 kΩ, channel 5 had an intermediate impedance
of 229 kΩ, and channels 5-8 had the highest impedances of 353 ± 34 kΩ. Channels 1-3 where
crosstalk levels were highest had EIS phase curves that appeared similar to those measured from
completely delaminated channels (Figure 76) in that the phase offset did not return to the
resistive regime at lower frequencies and that transitioned from capacitive to resistive at ~ 10
3
Hz
on the high frequency end of the spectrum. Interestingly, channels 4-8 had two humps in the
capacitive region, but shared the overall shape and positioning of phase curves from delaminated
channels. This “double time constant” in the phase curve suggests delamination beginning to
occur, as will be further explained in the discussion.
125

Figure 78: Crosstalk measurements taken at day 90 with corresponding CV curves, EIS impedance magnitudes, and EIS phase
curves of each of the eight channels in a single array taken at day 92 of soaking. Data for (a) dehydrated set C, (b) dehydrated set
A, (c) control set, and (d) HF-treated set A.
126

Column C shows CV and EIS curves of the control array which began exhibiting
significant crosstalk levels on channels 2-6 on soaking day 43. By day 90 of soaking crosstalk >
20% was present in all eight channels of the array. Impedance magnitude curves were shifted
down relative to controls and phase curves also suggested a double time constant. Channel 1
oddly had an impedance magnitude that was higher than its neighbors and a phase curve that
went from capacitive to resistive and back to capacitive—which was an anomaly.
Column D contains CV and EIS measurements from HF-treated set A which had
channels 1 and 3 shorted together and began exhibiting crosstalk in other channels in the array by
day 50. These two channels exhibit almost identical CV curves with highest current ranges of all
8 channels. The phase curves of these two channels appear similar to the completely delaminated
channels shown in Figure 76. Channels 7 and 8 which had crosstalk levels of < 15% had CV
curves with the narrowest ranges and EIS phase curves that shot back to resistive phase regions
at lower frequencies. The remaining channels of this array exhibited intermediate results with
EIS phase curves that were somewhere in between those with a double time constant and the
fully delaminated case.
Discussion
Overall, crosstalk correlates to individual CV curves that have increased current ranges.
As conductive solution permeates the insulation separating two adjacent channels, it can act to
electrically “short” channels together. This increases the effective surface area of conductive
material that voltage is being applied to, thereby decreasing its resistance, and enabling the flow
of higher levels of charge through the new, combined, channel group. The dehydrated set with no
crosstalk presented in column A of Figure 78 had CV curves that ranged from -19 ± 1 nA to 13 ±
1 nA (n=8, excluding anomalous channel 1). This current range is representative of that recorded
from intact channels. The second dehydrated set studied had channels with poor insulation
(channels 1-5, crosstalk > 50%) and channels whose insulation appeared intact (channels 6-8,
crosstalk < 15%). Intact channels of this partially delaminated array had CV current ranges
similar to dehydrated set C: -18 ± 0 nA to 12 ± 0 nA (n=3). However, channels that measured
crosstalk had larger current ranges of -30 ± 3 nA to 19 ± 2 nA (n=4, excluding anomalous
channel 1). This same trend holds true for channels with and without crosstalk in HF-treated set
A. Channels without crosstalk had a smaller current range of -22 ± 0 nA to 12 ± 0 nA and
127

channels with crosstalk had a larger current range of -25 ± 2 nA to 16 ± 1 nA (channels 1 and 3
were excluded from this analysis since they are assumed to be physically shorted together
through the packaging). Although the degree to which the current ranges of delaminated
channels in this experiment have increased does not match the 10-fold magnitude seen in cases
of catastrophic delamination, this milder trend may be valuable to monitor and characterize
during future tests of array lifetime. Initial results indicate that higher current ranges on CV
curves of electrodes during the cleaning step correlate to channels that develop crosstalk in vivo,
as will be discussed in section 5.3.
EIS curves also changed as crosstalk developed between channels, and was particularly
evident in two general trends. First, the magnitude curve shifted downward, to lower
impedances. Dehydrated set C that had no measurable crosstalk had an average 1 kHz impedance
of 436 ± 43 kΩ (n=7, excluding trace 1 as an outlier). All other traces had significantly smaller 1
kHz impedances (p < 0.0001), averaging 229 ± 77 kΩ (n=23, excluding trace 1 on control set as
an outlier). Also, not only did the shape of the phase curves for channels that exhibited crosstalk
not return back to the resistive regime at lower frequencies, many of these channels exhibited
“double humps” which represent a second time constant in the EIS phase curve. Physically this
could represent an intermediate stage of delamination, where current applied to an individual
channel can now take one of two paths: it can either travel down the trace through the exposed
electrode to the solution or it can travel through a trace but leak through solution that permeated
the Parylene insulation to nearby traces.
In fact, this dual path can be modelled as a Randles circuit (branch 1) in parallel with
another circuit leg (branch 2) that contains a resistor and a constant phase element (Rdelam and
Ydelam, αdelam) as shown in [36].These elements represent crosstalk, the physical transfer of
current through traces connected to each other by the ingress of conductive solution. The
constant phase element represents an interface with intermediate capacitive properties and the
resistance may be some combination of solution resistance and faradaic reactions across the trace
interfaces. This circuit also contains a third path (branch 3) between the working and reference
electrodes to mimic the high-frequency drop-off (> 10
5
MHz) that occurs in the electrical leads of
the Gamry system due to capacitive coupling from Cwire. This equivalent circuit models the data
extremely well, as shown in a representative example from trace 7 of dehydrated set C (Figure
128

80). Figure 81 shows how modifying the crosstalk parameters, Rdelam, Ydelam, and αdelam, would
change the shape of the EIS curves.

Figure 79: Electrical circuit model and pictorial representation of current paths through a
channel. Branch 1 is the traditional Randles circuit which models faradaic charge transfer as a
resistor, Re and capacitive charging across the inner and outer Helmholtz charge layers as a
constant phase element with Ye and αe and ends with a solution resistance, Rs. Branch 2
represents crosstalk between two traces due to the ingress of conductive solution through
elements Rdelam, Ydelam, and αdelam. Branch 3 represents the high frequency drop-off caused by
capacitive coupling between instrument components, Cwire. WE represents the working electrode,
or channel, and RE represents the Ag/AgCl reference electrode.
129

Figure 80: Model fit to EIS data from dehydrated set C, trace 7 using equivalent circuit model
from Figure 79. Goodness of fit is 3E-4 with parameters: Re = 53 MΩ, Ye = 1 nS*s^ αe, αe = 0.9,
Rs = 13 kΩ, Rdelam = 3 MΩ, Ydelam =1 nS*s^ αdelam, αdelam = 0.9, and Cwire = 13 pF.

Figure 81: Modelling EIS curves as delamination parameters change. Central parameters are
Rdelam = 1 MΩ, Ydelam =1 nS*s^a, αdelam = 0.9. When αdelam= 1 the constant phase element has
behavior identical to a capacitor. All other circuit values are kept at Re = 100 MΩ, Ye = 1
nS*s^a, αe = 0.9, Rs = 10 kΩ, Cwire = 10 pF.

130

In vivo recordings from treated arrays
Control arrays and arrays treated with AdPro Plus®, HF, and a dehydration step were
implanted into six Sprague-Dawley rats according to the implantation procedure described in
Chapter 4. Prior to implantation, probe shanks were soaked in IPA for 5 minutes to clean and
sterilize the implant. All materials used to prepare the PEG brace were sterilized in 70% ethanol
for five minutes; PEG was autoclaved by heating it to > 120 °C for 15 minutes. Ground wires
that lead away from electrical packaging were also sterilized in 70% ethanol for five minutes.
The second generation electrical packaging system design was used in attempt to eliminate any
in vivo crosstalk or channel fall-off that was due to damage in the SSB6 PCB-PCB connectors
that accumulated over many mating cycles.
Three of the implanted arrays could not provide chronic recordings due to debonding of
the dental cement cap and cranium which caused the Parylene array to pull out of the brain.
During previous insertion surgeries the screws, integrated well with the cranium and stuck tightly
to the dental cement cap. An examination of these explanted arrays, however, revealed no
integration between the screws and cranium (Figure 82). This complication has since been
avoided by drilling a slightly smaller hole for the self-tapping screws, reinforcing the screws
with cyanoacrylate, and adding a fourth screw to the cranium to help support the packaging.
Other implantations failed due to a surgical complication from the delivery of anesthesia, user
accident, and failure to properly insert the back-end of the array into the ZIF connector.

Figure 82: Integration of screw and cranium. (a) Proper integration, yellow arrow points to
cranium hugging threads of screw. From explanted array that stayed implanted for months. (b)
Poor integration between screw and cranium. Yellow arrow points to clean screw with no
cranium attached. Dental cement and array fell off after only a week.
131

Though this prevented an accumulation of chronic data that would enable the evaluation
of which treatment type experienced the least crosstalk in vivo, one of the implantations provided
promising initial results. In neural recordings taken at day 7 post-implantation from an array
treated with AdPro Plus®, a mixture of results was achieved (Figure 83). Some probes were
resilient to crosstalk and experienced no signal crosstalk between a particular electrode site and
its neighbors (green boxes). Others experienced intermediate levels of crosstalk, where
recordings spread across electrodes in the same, CA3, electrode group but did not leak to the
CA1 group of electrodes (yellow boxes). Yet others experienced widespread crosstalk, with
signals being shared across electrode sites throughout the entire probe (orange boxes).

Figure 83: In vivo recordings from array treated with AdPro Plus® at seven days post-
implantation. Each box presents neural units recorded by a single electrode (1-8) on a single
probe of the Parylene array (green, yellow, or orange). (a) High amplitude spike (pink dotted
line) localized to a single electrode on the probe, electrode 1. Neural units in yellow on
electrodes 2 and 3 are independent units. (b) Electrodes 1, 2, and 4 of the same electrode group
(targeting the CA3) record identical spikes, but no crosstalk occurs to CA1 electrodes. (c)
Crosstalk across both electrode groups in a probe. Neural unit centered on electrode 2 (highest
amplitude), but signal is leaked to electrodes 1, 4, 5, 7, and 8 as well. Boxes marked with an “x”
were disconnected.
An evaluation of the maximum and minimum current ranges during CV cleaning
revealed differences in these values according to the electrodes tested. Electrodes with larger
current ranges during CV cleaning corresponded well to electrodes that experienced crosstalk in
132

vivo (Figure 84). Perhaps, in future, experiments, CV current ranges can be used as a predictor of
electrode quality and as an indicator of whether or not crosstalk will occur.

Figure 84: Current ranges measured during CV cleaning for electrodes that did not experience
crosstalk in vivo (n= 14) compared to current ranges for electrodes that experienced crosstalk (n=
24). Although the maximum and minimum current ranges overlap, there exists a statistically
significant difference (p < 0.05) between the means of both groups. With further refinement, this
could serve as a way to predict which electrodes will last without crosstalk in vivo.
Future directions
Chronic, in vivo implantations with optimized arrays
Hippocampal arrays of treatment types described above will be prepared for chronic
implantation. Recordings will be taken chronically to re-evaluate device performance and
lifetime and special attention will be paid to whether the appearance of crosstalk can be delayed
in time as compared to the prior implantations. If unsuccessful, manually filtering out crosstalk
will be explored as a way to overcome this confounding issue. During this time our collaborators
will perform functional connectivity studies of the hippocampus using the Parylene recording
arrays and will monitor and evaluate the activity of place cells, specialized hippocampal cells
133

that fire when an animal revisits a particular environmental location [37] in brain plasticity
studies.
Denser probes and deeper targets
With an eye towards further increasing the reach of this device, arrays with higher
number of electrode sites will be fabricated and longer shanks will be explored to enable access
of deeper brain targets. Efforts towards achieving both of these goals are currently underway.
Double-sided and multi-layer fabrication techniques, as presented in [38, 39], are being explored
to create dense arrays with 64 electrodes on each probe, for a total of 512 electrodes on an array
of eight probes. In order to route arrays with this many channels to a packaging system that is
small enough to fit on a rat’s cranium, application-specific integrated circuits (ASICs) will be
integrated into each probe to multiplex and amplify recorded signals to a smaller number of
external wire connections. Since longer probes are necessary for even deeper brain targets and
for future experiments with larger animal models (Figure 85), buckling mechanics of longer
shanks will be performed. With an eye towards future neural probe implantations in humans,
creating a penetrating neural probe that can be scaled to tens of mms in length and implanted
without buckling is crucial.
134

Figure 85: Location of the hippocampus in rat, monkey, and humans. Drawn to scale according
to [40].
Expand probe functionality
Adding other sensing or delivery functionalities to the current version of the hippocampal
array has the potential to expand its usefulness in basic neuroscience research and as a future
multifunctional, neuroprosthetic tool. One such expansion, naturally, would be the addition of
stimulation capabilities to the current electrode design. Since the charge injection limit of Pt ~ 50
- 350 µC/cm
2
[41], electrodes with 30 µm exposed diameters can theoretically deliver 0.4 – 2.5
nC of charge. However, in unpublished research from collaborators, 10 nC of charge is required
to stimulate ex vivo hippocampal tissue. Applying a PtIr coating through electrodeposition, is one
potential way to increase the charge storage capacity of electrodes [42]. Initial experiments
involving electrodepositing PtIr atop the thin-film, Pt, hippocampal showed a 50-fold drop in the
1 kHz impedance of electrodes after plating. Further research must be performed to explore
adhesion between Pt and the PtIr coating and to determine the bounds of current application that
do not cause delamination of the embedded Pt traces from the Pt substrate.
Although the more common avenue of communication with neurons is through the
sensing or transmission of electrical impulses, neural probes can communicate chemically with
135

brain tissue as well. While recording electrodes focuses on interfacing with the electrical activity
of the brain, communication between adjacent neurons relies on the release of neurotransmitters
at the neuronal synapse. The delivery of pharmacological agents that bind to the synapse of
select neurons or populations of neurons and modulate their activity can help understand the
functional connectivity between neurons through selective silencing or enhancement of action
potentials. This type of control can also provide insight into the underlying physiological
processes that occur across various types of neural cells, synapses, and receptors.
Traditionally, chemical infusions into the brain rely on microdialysis methods that inject
the pharmacological agent via pressure injection or iontophoresis through large diameter
cannulas that are an order of magnitude larger than that of neural cells. This limits focal delivery
and the catheter tubing attached to such systems is often bulky, contains large amounts of dead
space, and poses a risk for infection, entanglement, and tethers the animal preventing freedom of
movement. The development of a microfluidic system with microfluidic channels whose outlets
are similar in size to neurons and the development of a head mounted pump could help address
these issues.
While microfluidic channels have been integrated into neural probes previously in both
silicon [43] and polymer [44, 45] shanks, drug delivery and its effect on recordings in vivo has
only been studied rarely and has not yet been monitored beyond the acute time period during
surgery. As such the goals of this project would be two-fold: (1) to design and fabricate
microchannels with multiple outlets along the probe shank for targeted, uniform delivery of
various drugs to small populations of neurons within reach of the recording electrodes, (2) to
design and develop a pumping mechanism that enables the delivery of pharmacological agents at
nL/min while ensuring that both the pump and drug reservoir fit directly atop the cranial
packaging of the animal and that obviates the dead space inherent to cannular connections to
fluid pumps. The anticipated fabrication scheme of such neural probes is expressed in Figure 86.
In brief, an attempt will be made to use thermal bonding to form microfluidic channels on
Parylene probes [45] without the use of a sacrificial photoresist layer which is time-consuming
and difficult to remove from the interior of microchannels with small cross-sections.

136

Figure 86: Anticipated fabrication scheme for hippocampal array with integrated microfluidics.
A prime silicon wafer (1) is etched with O2 plasma to form microchannels. A sacrificial release
layer is deposited to coat the microchannels prior to deposition of Parylene via CVD (3). Another
prime, silicon wafer is coated with a release layer and then CVD coated with Parylene (4). The
two wafers are laminated together through a thermal annealing treatment (5) and the top silicon
wafer is removed from the assembly to allow for subsequent patterning and exposure of Pt
electrodes and another Parylene insulation layer (7-9). Array is released from the carrier wafer
(10).

Figure 87: Imagined top view of a single probe. Two microchannel outlets per each
microchannel lie beneath the surface patterned electrodes and channels.
Conclusion
Recent advances in the creation of central neural implants for neural stimulation and
recording have led to a burgeoning hope for relevant therapeutic treatments. Such devices serve
as the critical component of brain-machine-interfaces, and are set to play a transformative role in
the future of medicine and in enabling the restoration of function to a damaged part of the human
body.
Towards accomplishing these goals, this dissertation details the design, fabrication, and
use of a novel neural probe array composed of ‘soft’ polymeric materials, boasting greater
137

flexibility and mechanical compliance than existing probes of silicon, glass and metal. Such rigid
probes suffer inevitable signal degradation over time as chronic tissue irritation drives an
immune cascade that may wall-off the implant. This mechanical disparity between the stiffness
of brain tissue and implantable neural probes can lead to tissue damage as micro-motion of the
brain causes chronic tissue irritation. It is speculated that the use of polymer probes might
mitigate this damage and attenuate the immune response and subsequent glial cell sheath that
degrade signal-to-noise ratios, thereby enabling the production of an effective, life-long brain
machine interface, as described in Chapter 1.
The reduced stiffness of probes, however, presents a technical challenge for surgical
insertion into brain tissue. Polymer probes must be temporarily stiffened in order to penetrate
brain tissue and for accurate surgical placement, typically via bulky biodegradable overcoats or
insertion shuttles which can increase probe cross-section many-fold times significantly adding to
acute tissue injury. The need for temporary stiffening of probes during implantation has limited
development of polymer probes to designs with short shanks (typically 1-2 mm), as shorter
probes have larger buckling force thresholds, and has constrained recording sites to superficial
cortical structures. Our strategy overcomes these issues, and opens the door for the large scale
acquisition of neural recordings from deep brain structures such as the cornu ammonis and
dentate gyrus in the hippocampus, as described in Chapter 2.
Chapter 3 discusses the design and packaging of a functional recording system that
overcomes the challenges inherent in connecting to a neural interface with small, dense, traces
embedded in a polymer. It also details an effective cleaning and testing scheme to ensure
electrode functionality prior to implantation. Chapter 4 reveals in vivo successes in achieving
high-quality recordings from probes whose explanted histology reveals an attenuated immune
response against these flexible, foreign implants.
While our flexible, polymer-based neural probes for brain recordings are designed with
principles of deep-brain targets and longevity in mind, challenges in the material properties of
these devices have required further research into failure analysis and reliability testing, as
described in Chapter 5. Various treatments aimed at improving channel insulation and Parylene
encapsulation properties have been explored with promising initial results. Further work towards
the continued exploration of the efficacy of these treatments is necessary to determine the ideal
138

fabrication scheme for these devices. Only after chronic, in vivo successes have been achieved
can this work be considered complete.
It is inspiring to note the incredible speed at which the field of penetrating polymer
probes is currently expanding. At the start of this study, this work represented one of the first
polymer arrays to record unitary activities from the deep-brain. It boasted a total of eight
electrodes per probe shank—the highest number of electrodes patterned onto a Parylene shank
[46, 47]. Recent improvements in multi-layer fabrication (up to four metal layers thick [48]), on
more heat resilient, flexible substrates such as polyimide, have enabled the creation of thin probe
shanks containing up to 16 electrodes and traces [39, 49]. In addition to expanding the electrode
density, improvements are also being made to better integrate soft polymer substrates with rigid
packaging outputs in a small-footprint design that can fit well onto the cranium of an animal
subject. One recent study created a modular headstage design that could be stacked together to
support a total of 1024 channels on the cranium of a single rat using multiplexing [50]. This
same study expanded the traditional, acute scope of polymer probe recording experiments by
monitoring neural recordings for up to 5-8 months post-implantation.
It is the hope of this author that this work will be a cog in a well-oiled research machine
that paves the way for the successful creation and realization of a hippocampal prosthetic to
restore brain function in individuals who have traumatic brain injuries or disease-driven memory
loss.
References
[1] W. Rall, "Electrophysiology of a dendritic neuron model," Biophysical journal, vol. 2,
pp. 145-167, 1962.
[2] D. A. Henze, Z. Borhegyi, J. Csicsvari, A. Mamiya, K. D. Harris, and G. Buzsáki,
"Intracellular features predicted by extracellular recordings in the hippocampus in vivo,"
Journal of neurophysiology, vol. 84, pp. 390-400, 2000.
[3] K. Najafi, J. Ji, and K. Wise, "Scaling limitations of silicon multichannel recording
probes," IEEE Transactions on Biomedical Engineering, vol. 37, pp. 1-11, 1990.
[4] S. A. Hara, "The Electrochemical Evaluation of Parylene-based Electrodes for Neural
Applications," 2015.
[5] J.-M. Hsu, S. Kammer, E. Jung, L. Rieth, R. Normann, and F. Solzbacher,
"Characterization of Parylene-C film as an encapsulation material for neural interface
devices," in Conference on Multi-Material Micro Manufacture, 2007.
[6] 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.
139

[7] H. Yasuda, Q. Yu, and M. Chen, "Interfacial factors in corrosion protection: an EIS study
of model systems," Progress in organic Coatings, vol. 41, pp. 273-279, 2001.
[8] V. Radun, R. von Metzen, T. Stieglitz, V. Bucher, and A. Stett, "Evaluation of adhesion
promoters for Parylene C on gold metallization," Current Directions in Biomedical
Engineering, vol. 1, pp. 493-497, 2015.
[9] D. Zeniieh, A. Bajwa, L. Ledernez, and G. Urban, "Effect of Plasma Treatments and
Plasma‐Polymerized Films on the Adhesion of Parylene‐C to Substrates," Plasma
Processes and Polymers, vol. 10, pp. 1081-1089, 2013.
[10] J. H.-C. Chang, B. Lu, and Y.-C. Tai, "Adhesion-enhancing surface treatments for
parylene deposition," in Solid-State Sensors, Actuators and Microsystems Conference
(TRANSDUCERS), 2011 16th International, 2011, pp. 390-393.
[11] R. Kumar, "Advances in Adhesion Solutions for Medical Applications," in Proceedings
of the SMTA Medical Electronics Symposium, 2008.
[12] J. H. Chang, L. Bo, and T. Yu-Chong, "Adhesion-enhancing surface treatments for
parylene deposition," in Solid-State Sensors, Actuators and Microsystems Conference
(TRANSDUCERS), 2011 16th International, 2011, pp. 390-393.
[13] A. K. Sharma and H. Yasuda, "Effect of glow discharge treatment of substrates on
parylene‐substrate adhesion," Journal of Vacuum Science and Technology, vol. 21, pp.
994-998, 1982.
[14] K. P. Gibbons, L. C. Lau, and F. E. Woodard, "Metal on plastic films with adhesion-
promoting layer," ed: Google Patents, 1996.
[15] S. Minnikanti, G. Diao, J. J. Pancrazio, X. Xie, L. Rieth, F. Solzbacher, et al., "Lifetime
assessment of atomic-layer-deposited Al 2 O 3–Parylene C bilayer coating for neural
interfaces using accelerated age testing and electrochemical characterization," Acta
biomaterialia, vol. 10, pp. 960-967, 2014.
[16] X. Xie, L. Rieth, L. Williams, S. Negi, R. Bhandari, R. Caldwell, et al., "Long-term
reliability of Al2O3 and Parylene C bilayer encapsulated Utah electrode array based
neural interfaces for chronic implantation," Journal of neural engineering, vol. 11, p.
026016, 2014.
[17] J. P. Seymour, "Advanced polymer-based microfabricated neural probes using
biologically driven designs," University of Michigan, 2009.
[18] C. Hassler, R. P. von Metzen, P. Ruther, and T. Stieglitz, "Characterization of parylene C
as an encapsulation material for implanted neural prostheses," Journal of Biomedical
Materials Research Part B: Applied Biomaterials, vol. 93, pp. 266-274, 2010.
[19] J. C. Williams, J. A. Hippensteel, J. Dilgen, W. Shain, and D. R. Kipke, "Complex
impedance spectroscopy for monitoring tissue responses to inserted neural implants,"
Journal of neural engineering, vol. 4, p. 410, 2007.
[20] X. Xie, L. Rieth, P. Tathireddy, and F. Solzbacher, "Long-term in-vivo investigation of
parylene-C as encapsulation material for neural interfaces," Procedia Engineering, vol.
25, pp. 483-486, 2011.
[21] J. P. Seymour, Y. M. Elkasabi, H.-Y. Chen, J. Lahann, and D. R. Kipke, "The insulation
performance of reactive parylene films in implantable electronic devices," Biomaterials,
vol. 30, pp. 6158-6167, 2009.
[22] S. W. Lee, K. S. Min, J. Jeong, J. Kim, and S. J. Kim, "Monolithic encapsulation of
implantable neuroprosthetic devices using liquid crystal polymers," IEEE Transactions
on Biomedical Engineering, vol. 58, pp. 2255-2263, 2011.
140

[23] J.-M. Hsu, L. Rieth, R. A. Normann, P. Tathireddy, and F. Solzbacher, "Encapsulation of
an integrated neural interface device with Parylene C," IEEE Transactions on Biomedical
Engineering, vol. 56, pp. 23-29, 2009.
[24] W. Li, D. C. Rodger, E. Meng, J. D. Weiland, M. S. Humayun, and Y.-C. Tai, "Flexible
parylene packaged intraocular coil for retinal prostheses," in Microtechnologies in
Medicine and Biology, 2006 International Conference on, 2006, pp. 105-108.
[25] X. Huang, P. M. Denprasert, L. Zhou, A. N. Vest, S. Kohan, and G. E. Loeb,
"Accelerated life-test methods and results for implantable electronic devices with
adhesive encapsulation," Biomedical Microdevices, vol. 19, p. 46, 2017.
[26] R. Huang, C. Pang, Y. Tai, J. Emken, C. Ustun, R. Andersen, et al., "Integrated parylene-
cabled silicon probes for neural prosthetics," in Micro Electro Mechanical Systems, 2008.
MEMS 2008. IEEE 21st International Conference on, 2008, pp. 240-243.
[27] J. Du, I. H. Riedel-Kruse, J. C. Nawroth, M. L. Roukes, G. Laurent, and S. C.
Masmanidis, "High-resolution three-dimensional extracellular recording of neuronal
activity with microfabricated electrode arrays," Journal of neurophysiology, vol. 101, pp.
1671-1678, 2009.
[28] J. J. Pancrazio, P. P. Bey Jr, A. Loloee, S. Manne, H.-C. Chao, L. L. Howard, et al.,
"Description and demonstration of a CMOS amplifier-based-system with measurement
and stimulation capability for bioelectrical signal transduction," Biosensors and
Bioelectronics, vol. 13, pp. 971-979, 1998.
[29] S. A. Hara, B. J. Kim, J. T. Kuo, C. Lee, C. A. Gutierrez, T. Hoang, et al., "Pre-
implantation electrochemical characterization of a Parylene C sheath microelectrode
array probe," in Engineering in Medicine and Biology Society (EMBC), 2012 Annual
International Conference of the IEEE, 2012, pp. 5126-5129.
[30] T. Akin, B. Ziaie, S. A. Nikles, and K. Najafi, "A modular micromachined high-density
connector system for biomedical applications," IEEE transactions on biomedical
engineering, vol. 46, pp. 471-480, 1999.
[31] J. J. Pancrazio, S. A. Gray, Y. S. Shubin, N. Kulagina, D. S. Cuttino, K. M. Shaffer, et
al., "A portable microelectrode array recording system incorporating cultured neuronal
networks for neurotoxin detection," Biosensors and bioelectronics, vol. 18, pp. 1339-
1347, 2003.
[32] T. J. Blanche, M. A. Spacek, J. F. Hetke, and N. V. Swindale, "Polytrodes: high-density
silicon electrode arrays for large-scale multiunit recording," Journal of neurophysiology,
vol. 93, pp. 2987-3000, 2005.
[33] M. K. Zachek, J. Park, P. Takmakov, R. M. Wightman, and G. S. McCarty,
"Microfabricated FSCV-compatible microelectrode array for real-time monitoring of
heterogeneous dopamine release," Analyst, vol. 135, pp. 1556-1563, 2010.
[34] M. K. Zachek, P. Takmakov, B. Moody, R. M. Wightman, and G. S. McCarty,
"Simultaneous decoupled detection of dopamine and oxygen using pyrolyzed carbon
microarrays and fast-scan cyclic voltammetry," Analytical chemistry, vol. 81, pp. 6258-
6265, 2009.
[35] J. E. B. Randles, "Kinetics of rapid electrode reactions," Discussions of the faraday
society, vol. 1, pp. 11-19, 1947.
[36] G. Brug, A. Van Den Eeden, M. Sluyters-Rehbach, and J. Sluyters, "The analysis of
electrode impedances complicated by the presence of a constant phase element," Journal
141

of electroanalytical chemistry and interfacial electrochemistry, vol. 176, pp. 275-295,
1984.
[37] P. J. Best, A. M. White, and A. Minai, "Spatial processing in the brain: the activity of
hippocampal place cells," Annual review of neuroscience, vol. 24, pp. 459-486, 2001.
[38] J. P. Seymour, N. B. Langhals, D. J. Anderson, and D. R. Kipke, "Novel multi-sided,
microelectrode arrays for implantable neural applications," Biomedical microdevices, vol.
13, pp. 441-451, 2011.
[39] A. Tooker, V. Tolosa, K. G. Shah, H. Sheth, S. Felix, T. Delima, et al., "Polymer neural
interface with dual-sided electrodes for neural stimulation and recording," in Engineering
in Medicine and Biology Society (EMBC), 2012 Annual International Conference of the
IEEE, 2012, pp. 5999-6002.
[40] R. Bakker, P. Tiesinga, and R. Kötter, "The scalable brain atlas: instant web-based access
to public brain atlases and related content," Neuroinformatics, vol. 13, pp. 353-366, 2015.
[41] D. R. Merrill, M. Bikson, and J. G. Jefferys, "Electrical stimulation of excitable tissue:
design of efficacious and safe protocols," Journal of neuroscience methods, vol. 141, pp.
171-198, 2005.
[42] A. Petrossians, J. J. Whalen, J. D. Weiland, and F. Mansfeld, "Electrodeposition and
characterization of thin-film platinum-iridium alloys for biological interfaces," Journal of
The Electrochemical Society, vol. 158, pp. D269-D276, 2011.
[43] K. Neeves, C. Lo, C. Foley, W. Saltzman, and W. Olbricht, "Fabrication and
characterization of microfluidic probes for convection enhanced drug delivery," Journal
of controlled release, vol. 111, pp. 252-262, 2006.
[44] S. Metz, A. Bertsch, D. Bertrand, and P. Renaud, "Flexible polyimide probes with
microelectrodes and embedded microfluidic channels for simultaneous drug delivery and
multi-channel monitoring of bioelectric activity," Biosensors and bioelectronics, vol. 19,
pp. 1309-1318, 2004.
[45] D. Ziegler, T. Suzuki, and S. Takeuchi, "Fabrication of flexible neural probes with built-
in microfluidic channels by thermal bonding of Parylene," Journal of
Microelectromechanical Systems, vol. 15, pp. 1477-1482, Dec 2006.
[46] F. Wu, L. W. Tien, F. Chen, J. D. Berke, D. L. Kaplan, and E. Yoon, "Silk-backed
structural optimization of high-density flexible intracortical neural probes," Journal of
Microelectromechanical Systems, vol. 24, pp. 62-69, 2015.
[47] A. W. Hirschberg, H. Xu, K. Scholien, T. W. Berger, D. Song, and E. Meng,
"Development of an anatomically conformal parylene neural probe array for multi-region
hippocampal recordings," in Micro Electro Mechanical Systems (MEMS), 2017 IEEE
30th International Conference on, 2017, pp. 129-132.
[48] A. Tooker, V. Tolosa, K. G. Shah, H. Sheth, S. Felix, T. Delima, et al., "Optimization of
multi-layer metal neural probe design," in Engineering in Medicine and Biology Society
(EMBC), 2012 Annual International Conference of the IEEE, 2012, pp. 5995-5998.
[49] A. Mercanzini, K. Cheung, D. L. Buhl, M. Boers, A. Maillard, P. Colin, et al.,
"Demonstration of cortical recording using novel flexible polymer neural probes,"
Sensors and Actuators A: Physical, vol. 143, pp. 90-96, 2008.
[50] J. E. Chung, H. R. Joo, J. L. Fan, D. F. Liu, A. H. Barnett, S. Chen, et al., "High-density,
long-lasting, and multi-region electrophysiological recordings using polymer electrode
arrays," bioRxiv, p. 242693, 2018.

142

143

Appendix A: Fabrication of Sham Arrays

Figure A- 1: Hippocampal sham array photomask transparencies.

1. Bake clean 4” silicon wafer to remove moisture 110 °C, > 10 mins

2. Deposit Parylene (10 µm)

3. Pattern AZ 4620 etch mask (~ 8 µm thick) (Mask 1 – Sacrificial photoresist pockets)
Pre spin 5 sec, 500 rpm
Spin 45 sec, 3500 rpm
Softbake 90 °C, 5 minutes
Hydration 45 minutes
Exposure 400 mJ/cm
2
(25 mW/cm
2
, 16 sec)
Development 80 seconds/ 20 seconds (two baths)
Hard bake 1 90 °C, 5 minutes, hotplate
Hard bake 2 110 °C, 15 minutes, under vacuum

4. Deposit Parylene ( 10 µm)

5. Pattern AZ 4620 etch mask (> 10 µm thick) (Mask 2 –Pocket inlet; cutout and ellipse)
Pre spin 5 sec, 500 rpm
Spin 40 sec, 1200 rpm
Softbake 90 °C, 5 minutes
Hydration 30 minutes
Align and expose 550 mJ/cm
2
(25 mW/cm
2
, 22 sec)
Development 90 seconds
Hard bake 1 90 °C, 5 minutes, hotplate
Hard bake 2 90 °C, 15 minutes, under vacuum

144

6. Deep Reactive Ion Etching (Oxygen plasma) 125 loops, rotate wafer 90° every 25 loops

Table A- 1: DRIE parameters for deposition and etch steps
Parameter Deposition Etch
ICP Power (W) 700 700
RF Power (W) 80 80
O2 (ccm) 1 60
C4F8 (ccm) 35 1
Ar (ccm) 40 40
SF6 (ccm) 0 0
Pressure (mTorr) 23 23
Time (s) 3 10

7. Strip remaining photoresist mask with Acetone, IPA, and DI water

8. Pattern AZ 4620 etch mask (> 10 µm thick) (Mask 3 –Cutout and ellipse completion)
Pre spin 5 sec, 500 rpm
Spin 40 sec, 1200 rpm
Softbake 90 °C, 5 minutes
Hydration 30 minutes
Align and expose 550 mJ/cm
2
(25 mW/cm
2
, 22 sec)
Development 90 seconds
Hard bake 1 90 °C, 5 minutes, hotplate
Hard bake 2 90 °C, 15 minutes, under vacuum

9. Deep Reactive Ion Etching (Oxygen plasma) 125 loops, rotate wafer 90° every 25 loops

10. Release Clean surface with acetone and IPA
Peel carefully while immersed in water

11. Strip any remaining photoresist mask with 5 min soaks in acetone, IPA, and DI water

145

Appendix B: Fabrication of Complete Hippocampal Arrays

Figure A- 2: Hippocampal array photomask transparencies.
1. Bake clean 4” silicon wafer to remove moisture 110 °C, > 10 mins

2. Deposit Parylene (10 µm)

3. Pattern AZ 5214-IR for lift-off (2 µm thick) (Mask 1 - Metal)
Pre spin 8 sec, 500 rpm
Spin 45 sec, 1800 rpm
Softbake 90 °C, 70 seconds
Exposure 37.5 mJ/cm
2
(25 mW/cm
2
, 1.5 sec)
IR bake 110 °C , 55 sec
Hydration 3 minutes
Global exposure 1000 mJ/cm
2
(25 mW/cm
2
, 40 seconds)
Development (AZ 351 1:4 dilution) 18 seconds

4. Descum, O2 plasma 100 W, 100 mTorr, 1 min

5. Metal deposition (Pt) 2000 Å (in 4 runs of 500 Å )

6. Lift-off in acetone (gentle scrub if necessary) In warm acetone 50 °C

7. *Descum, O2 plasma. Sometimes this step was eliminated without obvious changes in array
performance. 100 W, 100 mTorr, 1 min

8. *Deposit Parylene (10 µm). Treat with HF dip, dehydration, or AdPro Plus® prior to
deposition for array integrity experiments.

9. Pattern AZ 4620 etch mask (12.8 – 13.5 µm thick) (Mask 2 – Insulation Cutout)
Pre spin 5 sec, 500 rpm
Spin 45 sec, 1000 rpm
146

Softbake 90 °C, 5 minutes
Hydration 45 minutes
Exposure 550 mJ/cm
2
(20 mW/cm
2
, 27.5 sec)
Development 1.5 minutes
Hard bake 1 90 °C, 15 minutes, hotplate
Hard bake 2 90 °C, 15 minutes, under vacuum

10. Deep Reactive Ion Etching (Oxygen plasma) 100 loops, rotate wafer 90° every 25
loops
Table 5: DRIE etch and deposition parameters for each loop.
Parameter Deposition Etch
ICP Power (W) 700 700
RF Power (W) 80 80
O2 (ccm) 1 60
C4F8 (ccm) 35 1
Ar (ccm) 40 40
SF6 (ccm) 0 0
Pressure (mTorr) 23 23
Time (s) 3 10

11. Strip remaining photoresist mask with Acetone, IPA, and DI water

12. Pattern AZ 4620 etch mask (12.8 – 13.5 µm thick) (Mask 3 – Insulation and electrode etch)
Pre spin 5 sec, 500 rpm
Spin 45 sec, 1000 rpm
Softbake 90 °C, 5 minutes
Hydration 45 minutes
Exposure 550 mJ/cm
2
(20 mW/cm
2
, 27.5 sec)
Development 1.5 minutes
Hard bake 1 90 °C, 15 minutes, hotplate
Hard bake 2 90 °C, 15 minutes, under vacuum

13. Deep Reactive Ion Etching (Oxygen plasma) 100 loops, rotate wafer 90° every 25
loops
147

Table 6: DRIE etch and deposition parameters for each loop.
Parameter Deposition Etch
ICP Power (W) 700 700
RF Power (W) 80 80
O2 (ccm) 1 60
C4F8 (ccm) 35 1
Ar (ccm) 40 40
SF6 (ccm) 0 0
Pressure (mTorr) 23 23
Time (s) 3 10

14. Strip remaining photoresist mask with Acetone, IPA, and DI water

15. Release Clean surface with acetone and IPA.
Peel carefully while immersed in water

Abstract (if available)

Abstract Recent advances in the creation and development of central, penetrating neural implants for the stimulation and recording of precise neural circuitry have unlocked growing interest in the powerful, potential applications of such technology. These “neural probes” not only serve to advance basic neuroscience studies of the mechanism of neural encoding, but also serve as crucial components to brain-machine-interfaces which can enable the restoration of sensory, motor, or cognitive functionalities to a damaged part of the human body. The small size and detail required for such implants have been realized through strides in MEMS fabrication capabilities. Traditionally, probes are made of rigid materials, like silicon, which can irritate the adjacent tissue and lead to signal degradation over time. This dissertation focuses on the use of a flexible polymer, Parylene C, as a substrate and insulator for the creation of a penetrating neural probe array that targets a particularly thin and deep target in the rat brain, the hippocampus, with the goal of recording longevity in mind. ❧ In this dissertation, Chapter 1 presents the detailed design and fabrication of a novel Parylene neural array with improved reach, number of recording sites, and distribution of electrodes to target recordings from individual neurons of two particular layers of interest in the hippocampus, deep within the brain. Chapter 2 introduces technical challenges that exist when implanting flexible probes into deep brain structures

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

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

PDF

The electrochemical evaluation of Parylene-based electrodes for neural applications

PDF

Development of implantable Parylene-based MEMS technologies for cortical applications

PDF

Parylene-based biomems sensors for multiple physiological systems

PDF

Parylene-based implantable interfaces for biomedical applications

PDF

High-resolution data acquisition with neural and dermal interfaces

PDF

Fabrication and packaging of three-dimensional Parylene C neural interfaces

PDF

Development of fabrication technologies for robust Parylene medical implants

PDF

Parylene C bioMEMS for implantable devices with electrochemical interfaces

PDF

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

PDF

Simulating electrical stimulation and recording in a multi-scale model of the hippocampus

PDF

Liquid flow rate sensing via solution resistance for in vivo applications

PDF

Next generation neural interfaces for neural modulation and prosthesis

PDF

Miniature phased-array transducer for colorectal tissue characterization during TEM robotic surgery; and, Forward-looking phased-array transducer for intravascular imaging

PDF

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

PDF

Bidirectional neural interfaces for neuroprosthetics

PDF

Scale-up of vapor-phase deposition of polymers: towards large-scale processing

PDF

Design and development of ultrasonic array transducers for specialized applications

PDF

Biosensing and biomimetic electronics

PDF

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

Asset Metadata

Creator Hirschberg, Ahuva Weltman (author)

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

School Viterbi School of Engineering

Degree Doctor of Philosophy

Degree Program Biomedical Engineering

Publication Date 08/12/2020

Defense Date 08/12/2018

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

Tag flexible,hippocampus,in vivo recordings,multi-region recordings,neural array,neural probe,OAI-PMH Harvest,Parylene C,recording lifetime,subcortical

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Meng, Ellis (committee chair)

Creator Email aweltman1@gmail.com,weltman@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-66954

Unique identifier UC11672623

Identifier etd-Hirschberg-6734.pdf (filename),usctheses-c89-66954 (legacy record id)

Legacy Identifier etd-Hirschberg-6734.pdf

Dmrecord 66954

Document Type Dissertation

Format application/pdf (imt)

Rights Hirschberg, Ahuva Weltman

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