University of Southern California Dissertations and Theses

Fabrication and packaging of three-dimensional Parylene C neural interfaces

(USC Thesis Other)

Fabrication and packaging of three-dimensional Parylene C neural interfaces

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content FABRICATION AND PACKAGING OF THREE-DIMENSIONAL
PARYLENE C NEURAL INTERFACES
by
Brianna Thielen
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree of
DOCTOR OF PHILOSOPHY
(BIOMEDICAL ENGINEERING)
August 2024
Copyright 2024 Brianna Thielen

ii
To my parents and sisters

iii
Acknowledgements
There are many, many people that have supported me throughout the past six years. A PhD is
often perceived as a solo venture, but, in reality, it could not be accomplished without a community of
support both personally and professionally.
Ellis, thank you for bringing me into your research group and giving me both the freedom to
explore my own ways of attacking a problem and the guidance to ensure I went about it in the right way. I
was concerned about the adjustment going into graduate school after a few years in industry, but the
environment you have created in the lab and your trust in me taking on several new projects throughout
my PhD made that transition so much easier and made my time at USC very rewarding.
To all members of the Biomedical Microsystems Lab, I’m so glad to have been a part of the
community that we have all created. Thank you all for playing a role in getting me to where I am today.
Alberto, I’m so glad you ended up joining this lab, I’ve really valued our friendship over the past year or
so and am very grateful of your constant efforts to build community in our lab. On a professional level, I
am continuously impressed by your ability to quickly build expertise on any topic that comes your way.
Quentin, it has been amazing working so closely with you since your arrival. Your willingness to take on
any project that needs to be done has been so beneficial to our project, and your infectious positivity
always puts a positive spin on my day. After several years of very independent research, this past year and
a half was a welcome change in getting to work so closely with both of you. Chris, I’ve always admired
your ability to fully break down any problem you run into and figure out creative solutions. James, I am
impressed by your balance between work and community, always producing high quality research and
making sure no birthday or other life accomplishment goes unnoticed. Nick, your attention to detail,
especially in how you communicate your progress, is amazing, and I’ve enjoyed watching you become a
leader in the lab. Cindy and Emmanuel, I’m impressed by your growth into your research and into the lab
community. Jeannie, Max, Ruitong, and Yan, it’s been great seeing you all find your groove in the lab
and in the field and I look forward to seeing where your research takes you in the coming years. Kee, your
expertise in the field and willingness to share that knowledge has quite literally cut weeks, if not months,

iv
of time off of my workload, and I will forever be grateful for that support. Sue, Eugene, and Trevor, I am
grateful for the community and knowledge base you all contributed to before my time in the lab and
during my first few years and that will continue on in the future. To all of my collaborators from the Song
lab, the Mousavi lab, the Zhao lab, Keck, and the CARSS center, none of my work would be possible
without your efforts and support.
To all of my friends, thank you for supporting me in so many ways throughout this process.
Anshu, Pan, Rin, Natalie, Colin, and Pat, I’m so glad to have met you all and been able to muscle through
the highs and lows of our programs together. Clara, Emily, Erin, and Sophie I am so grateful to have had
your support not only for the past 6 years, but for many before that as well. Sophie, you have been a
constant support, cheerleader, confidant, and friend for more than half of my life, and I could not have
done this without you.
Finally, to my family, there are not enough words to thank you for everything you have done for
me. Mom, thank you for always being there for me, at any hour of any day, for any reason. My entire life,
I have felt supported and accepted in everything I do, and that is in large part due to you and the way you
raised us. Dad, thank you for understanding me and the way I think in a way that I’m not sure many other
people can. The way you valued and pushed for my education helped me get to where I am today. Tanya,
thank you for your constant willingness to be my friend, travel buddy, source of entertainment, and
overall a big piece of my support system. I’m so grateful to have someone so caring, funny, and smart as
my sister (and I’m sorry for now making you the second most educated person in the family). Winter,
thank you for always being in my corner for any issue, big or small. You are the most empathetic and
compassionate person I know, and I’m so lucky to have you as a sister.

v
Table of Contents
Dedication ...........................................................................................................................................ii
Acknowledgements......................................................................................................................................iii
List of Tables ..........................................................................................................................................vi
List of Figures ........................................................................................................................................viii
Abstract .......................................................................................................................................xvii
Chapter 1. Intro to Parylene-Based Neural Interfaces ........................................................................1
1.1 Neural Recording and Stimulation....................................................................................1
1.2 Immune Response to Implanted Neural Devices............................................................18
1.3 Flexible Neural Interfaces...............................................................................................20
1.4 Objectives........................................................................................................................26
1.5 References.......................................................................................................................26
Chapter 2. Thermoforming of Parylene C ........................................................................................38
2.1 Background .....................................................................................................................40
2.2 Modeling Stress and Curvature.......................................................................................43
2.3 Experimental Methods....................................................................................................52
2.4 Experimental Results ......................................................................................................60
2.5 Discussion .......................................................................................................................77
2.6 References.......................................................................................................................79
Chapter 3. Endovascular Recording Device .....................................................................................83
3.1 Background .....................................................................................................................83
3.2 Microfabricated Endovascular Electrode Array..............................................................88
3.3 Experimental Methods....................................................................................................97
3.4 Experimental Results ....................................................................................................114
3.5 Discussion .....................................................................................................................129
3.6 References.....................................................................................................................130
Chapter 4. Peripheral Nerve Stimulating Electrode........................................................................135
4.1 Background ...................................................................................................................135
4.2 Microfabricated Peripheral Nerve Cuff Electrode ........................................................137
4.3 Experimental Methods..................................................................................................144
4.4 Experimental Results ....................................................................................................159
4.5 Discussion .....................................................................................................................175
4.6 References.....................................................................................................................176
Chapter 5. Conclusion.....................................................................................................................179
Appendices .......................................................................................................................................181
A MATLAB Code for Parylene-Metal-Parylene Device Stress Model............................181
B Parylene-Metal-Parylene Device Fabrication (Double-Sided)......................................186
C Parylene-Metal-Parylene Device Fabrication (Single-Sided) .......................................198
D Annealing and Thermoforming of Parylene C..............................................................208
E Photomasks for Endovascular Electrode Arrays...........................................................216
F Peripheral Nerve Cuff Backbone Design Evaluation....................................................221
G Photomasks for Cuff Electrodes....................................................................................237
H Overmolding of Cuff Electrodes...................................................................................239

vi
List of Tables
Table 1-1: Summary of recording, stimulation, and closed-loop applications in each region of the
nervous system. Note that the cited publications are a representative sample of such
applications and not an exhaustive list............................................................................................. 2
Table 1-2: Summary of electrode types for neural interfaces. ......................................................................4
Table 1-3: Summary of insertion methods for flexible penetrating microelectrodes, including how
they impact each component of the buckling equation (increase ↑, decrease ↓, or no
impact ×), and common uses. Information sourced from [37].......................................................24
Table 2-1: Examples of microfabricated Parylene medical devices that were transformed into 3D
geometries via thermoforming. Their resulting shape and size are included.................................39
Table 2-2: Parylene C thicknesses for bare Parylene strip samples. Samples with no top layer
thickness listed were constructed of a single Parylene layer. © IOP Publishing.
Reproduced with permission from [12] (CC BY 4.0)....................................................................52
Table 2-3: Dimensions for Parylene-metal-Parylene electrode arrays........................................................54
Table 2-4: Calculated maximum Parylene thickness to prevent cracking at the given bending
diameter. ........................................................................................................................................61
Table 2-5: Thermoforming result vs. thickness, helix angle, and helix diameter for bare Parylene
strips at 200 °C thermoforming temperature and 12 hour thermoforming time. ✓ indicates
a good result, ∗ indicates minor cracking (partial-thickness), × indicates cracking (fullthickness), and ● indicates loose shape. © IOP Publishing. Reproduced with permission
from [12] (CC BY 4.0)...................................................................................................................62
Table 2-6: Thermoforming result vs. thickness and helix diameter for PMP devices at 45° helix
angle, 200 °C thermoforming temperature, and 12 hour thermoforming time. ✓ indicates
a good result, ∗ indicates minor cracking (partial-thickness), × indicates cracking (fullthickness), ● indicates loose shape, and – indicates discontinuous traces. © IOP
Publishing. Adapted with permission from [12] (CC BY 4.0). .....................................................76
Table 3-1: Summary of application, implant location, animal model, and blood vessel for
published endovascular recording studies. Common blood vessel locations in the human
brain are illustrated in Figure 3-1...................................................................................................86
Table 3-2: Comparison of size and electrode count for endovascular devices. ..........................................88
Table 3-3: Summary of fabrication changes between each design iteration of endovascular
devices. ..........................................................................................................................................98
Table 3-4: Electrode yield, listed as percentage (good electrodes/total electrodes tested) at each
fabrication step for varying processing conditions. Primary failure modes are listed for
steps which had low yield............................................................................................................121
Table 3-5: Summary of in vivo sheep studies for the endovascular device. .............................................124
Table 4-1: Microelectrode grid size and electrode surface area for each electrode type and cuff
size. ..............................................................................................................................................141
Table 4-2: Benchtop tests performed on fully fabricated devices and failure criteria resulting from
each test........................................................................................................................................151
Table 4-3: CV Parameters used for benchtop testing in each solution. ....................................................151
Table 4-4: Average ESA (± standard deviation) for cuff electrodes from batch S in the flat and
thermoformed (TF) configurations calculated using CV in ferri/ferrocyanide (FF) and
H2SO4. The right column shows the ESA in ferri/ferrocyanide divided by the ESA in
H2SO4, illustrating the differences in calculations between flat and thermoformed
configurations. .............................................................................................................................163
Table 4-5: Average ESA, CSC, 1 kHz impedance magnitude, and CIC with a pulse width of 200
and 500 µs (± standard deviation) for cuff electrodes from the most recent fabrication run
(batch U). Values are seperated by device configuration (helical vs. full, small vs. large,
prior to thermoforming (pre-TF) vs. flat thermoformed (TF flat) vs. curled thermoformed

vii
(TF curled)). Failed devices (per failure criteria in Table 4-2) were not included in
calculations. Data not listed in the table (indicated by a ‘-' mark) is unavailable due to no
testing at that stage or insufficient sample size............................................................................167

viii
List of Figures
Figure 1-1: The main types of neural recording and stimulation electrodes as implanted in the
brain, spinal cord, and peripheral nerve. Details on each electrode type are included in
sections 1.1.1 (scalp electrodes), 1.1.2 (surface electrodes), and 0 (penetrating
electrodes). Images are not to scale. ................................................................................................3
Figure 1-2: Picture of a headcap with 128 scalp EEG electrodes. Reproduced with permission
from [41] (CC BY 4.0).....................................................................................................................6
Figure 1-3: Picture of an ECoG array with 8 × 500, 750, and 1000 µm electrodes. Scale bar 5
mm. Reproduced with permission from [44] (CC BY 4.0)..............................................................7
Figure 1-4: Examples of three extraneural cuff electrodes: (A) a piano hinge split cuff, (B) a
spiral cuff, and (C) a helical cuff. Reproduced from [49] with permission from Elsevier. .............9
Figure 1-5: Picture of an AdTech macro-micro electrode with outer macroelectrodes (right arrow)
and center microwires (left arrow). © IOP Publishing. Reproduced with permission from
[108]...............................................................................................................................................10
Figure 1-6: SEM images of a Utah array containing 100 electrodes of 1.5 mm length. (A) shows
the entire 10x10 electrode array, and (B) shows a detailed image of the exposed electrode
tip and insulated probe. Reprinted from [112] with permission from Elsevier..............................11
Figure 1-7: (Left) SEM image of a microwire electrode (exposed wire tip) made of 70%/20%
platinum/iridium wire. Scale bar 10.0 µm. Reprinted from [60] with permission from
Elsevier. (Right) Picture of a microwire array with 1mm (left) and 8mm (right) wire
lengths. Reprinted from [60] with permission from Elsevier. .......................................................12
Figure 1-8: (Left) SEM image of a glass-coated carbon fiber. Reprinted from [64] with
permission from Elsevier. (Right) A 2x8 array of carbon fiber electrodes. Reproduced
with permission from [114] (CC BY 4.0)......................................................................................12
Figure 1-9: (Left) SEM image of a 64 channel silicon multisite electrode array. Reproduced with
permission from [72] (CC BY 4.0). (Right) Photograph of a Neuronexus multisite silicon
array with 16 electrodes (4 shanks with 4 electrodes at the tip of each probe). Reprinted
from [12] with permission from Elsevier.......................................................................................13
Figure 1-10: Photos of a 64 channel flexible Parylene multisite electrode array. © IOP Publishing.
Reproduced with permission from [115]. ......................................................................................13
Figure 1-11: Photographs of a prototype neural dust mote. Reprinted from [87] with permission
from Elsevier..................................................................................................................................14
Figure 1-12: Diagram (A), SEM image (B), and (C) photo of a regenerative sieve electrode.
Reproduced from [49] with permission from Elsevier. .................................................................15
Figure 1-13: Schematic of the Seeker-10 guidewire with (1) stainless steel shaft, (2) Teflon
coating, and (3) a single platinum electrode at the tip. Reprinted from [92] with
permission from Springer Nature...................................................................................................16
Figure 1-14: Schematic of a multi-electrode, SEEG-like endovascular probe used by Bower et al.
with 4 × 1 mm ring electrodes and 20 × 40 µm microelectrodes. Reprinted from [47] with
permission from Elsevier. ..............................................................................................................16
Figure 1-15: Schematic of a nanowire (0.6-1 µm wire diameter) with platinum black electrode at
the tip inside a microcatheter (90-300 µm diameter) developed by Llinas et al. Reprinted
from [13] with permission from Wiley..........................................................................................17
Figure 1-16: Picture of the Stentrode in the collapsed (top), partially-expanded (center), and
expanded (bottom) configuration with platinum disc electrodes (yellow arrow) and the
accompanying delivery catheter (green arrow). Scale bar 3 mm. Reprinted from [26] with
permission from Springer Nature...................................................................................................17
Figure 1-17: Young's Moduli of common rigid and flexible MEMS materials and human tissues
[37,141-143]. .................................................................................................................................21

ix
Figure 1-18: Insertion of a flexible probe into tissue. (A) Insertion force acts on the proximal end
of the probe; puncture force acts on the distal tip of the probe. For successful insertion,
the insertion force must be greater than the puncture force. (B) Successful insertion:
insertion force is less than the buckling force; the probe overcomes the puncture force
and inserts into the tissue. (C) Failed insertion: insertion force is greater than the buckling
force; the probe buckles before the insertion force can overcome the puncture force. ©
IOP Publishing. Reproduced with permission from [37]...............................................................23
Figure 2-1: Examples of microfabricated Parylene medical devices that were transformed into
common 3D geometries via thermoforming: (A) helices (reprinted from [5] with
permission from IEEE), (B) cylinders (reprinted from [7] with permission from IEEE),
(C) spheres (© IOP Publishing. Reproduced with permission from [8]), and (D) cones (©
IOP Publishing. Reproduced with permission from [4]). ..............................................................39
Figure 2-2: Diagram of a Parylene-metal-Parylene device configuration, with openings etched in
the top Parylene to form electrodes (section A-A’), insulated traces (section B-B’), and
openings etched in the base Parylene to form bondpads (section C-C’). © IOP Publishing.
Reproduced with permission from [12] (CC BY 4.0)....................................................................42
Figure 2-3: (A) An isometric view and (B) a cross sectional view of the simplified model
geometry, with a single metal strip sandwiched between two Parylene layers. (C) The
side view shows a small section of the device with the neutral axis of each layer and the
full device. © IOP Publishing. Reproduced with permission from [12] (CC BY 4.0). .................45
Figure 2-4: Illustration of the stresses and forces acting on a section of the device (side view, as
shown in Figure 2-3C). (A) As deposited: each film layer is deposited on a substrate with
residual stress, producing a part with stresses leading to unbalanced force and moment
(about the neutral axis). (B) Shrinkage or expansion only: When removed from the
substrate, the part shrinks or expands, adjusting the stresses in each layer to balance the
forces in the device. (C) Shrinkage/expansion and bending: After shrinking or expanding,
the device curls to balance the moment about the neutral axis, resulting in a stress
gradient in each layer. Note: the metal layer illustrated here is not to scale, so the high
stress leads to a small force, and the stress changes significantly in each step. © IOP
Publishing. Reproduced with permission from [12] (CC BY 4.0).................................................47
Figure 2-5: Modeled curvature diameter with variable base layer annealing temperature. Inset
axes are a zoomed in view of the red dotted box, showing detail for small diameter
curvature. The diameter at 200 °C is -1.2 mm. All constants were selected from literature
or from experimental PMP device dimensions (tbase = 4.4 µm, ttop = 4.7 µm, tmetal = 215 nm,
wbase = 210 µm, wtop = 210 µm, wtrace = 5 µm, ntraces = 8, σtop = -3.46 MPa, σmetal = -
100 MPa). © IOP Publishing. Reproduced with permission from [12] (CC BY 4.0)....................49
Figure 2-6: Modeled curvature diameter with variable ratio of top to bottom Parylene thickness,
with total Parylene thickness of 15 µm (blue) and 9 µm (green), and unannealed (solid)
and annealed at 200 °C (dashed) conditions. The diameter at 1:8 and 8:1 ratios is ±50.5,
±25.4, ∓3.5, and ∓1.7 mm for 15 µm unannealed, 9 µm unannealed, 15 µm annealed,
and 9 µm annealed parts, respectively. All constants were selected from literature or from
experimental PMP device dimensions (tmetal = 215 nm, wbase = 350 µm, wtop = 350 µm,
wtrace = 10 µm, ntraces = 16, σbase = -3.46 MPa (unannealed) or 25.6 MPa (annealed) σtop = -
3.46 MPa (unannealed) or 25.6 MPa (annealed), σmetal = -100 MPa). © IOP Publishing.
Reproduced with permission from [12] (CC BY 4.0)....................................................................50
Figure 2-7: Modeled curvature diameter with variable annealing temperature. The diameter at 200
°C is 4.2 mm, and the crossover from negative to positive diameter (when the device is
flat) occurs at approximately 31 °C. All constants were selected from literature or from
experimental PMP device dimensions (tbase = 3.4 µm, ttop = 11.5 µm, tmetal = 215 nm,
wbase = 350 µm, wtop = 350 µm, wtrace = 10 µm, ntraces = 16, σmetal = -100 MPa). © IOP
Publishing. Reproduced with permission from [12] (CC BY 4.0).................................................51

x
Figure 2-8: Cross-sectional view of Parylene-metal-Parylene device fabrication process flow for
(left) single-sided and (right) double-sided devices. (1) Parylene was deposited on top of
a bare silicon wafer (single-sided) or a sacrificial metal layer (15 nm titanium + 100 nm
aluminum). (2) Backside bondpads were opened via O2 etch (double-sided only). (3)
Metal (see Table 2-3 for metal stackup) electrodes, traces, and bondpads were deposited
and patterned. (4) Parylene was deposited on top of patterned metal and frontside
bondpads (single-sided only) and electrodes were opened via O2 etch. (5) The devices
were released using water (single-sided) or aluminum etchant (double-sided). © IOP
Publishing. Adapted with permission from [12] (CC BY 4.0). .....................................................53
Figure 2-9: Illustration of the helix angle – the angle between the axis of the mandrel and the long
edge of the Parylene strip or device (helix angles of 15°, 30°, and 45° were used). © IOP
Publishing. Reproduced with permission from [12] (CC BY 4.0).................................................57
Figure 2-10: Examples of bare Parylene strips (300 µm width by 20 mm length, variable
thickness) thermoformed into 0.25 mm helices, showing a good result (✓), minor
cracking (∗), cracking (×), and loose shape (●). © IOP Publishing. Reproduced with
permission from [12] (CC BY 4.0). ...............................................................................................63
Figure 2-11: Illustration of the stress in each layer of PMP devices before and after annealing.
(Ai) Thin/symmetric devices before annealing have moderate tensile stress in the base
Parylene layer due to shrinkage during the base layer anneal and a low compressive stress
in the top Parylene layer, resulting in a device curled towards the base layer. (Aii)
Thin/symmetric devices after annealing have balanced stress around the neutral plane due
to equal shrinkage in the Parylene layers and the high-stress metal layer sitting on the
neutral plane, resulting in a flat device. (Bi) Thick/asymmetric devices before annealing
have high compressive stress in the metal layer and low compressive stress in both
Parylene layers. The compressive stress in all layers results in enough expansion in the
device to produce a low tensile stress in the metal layer (below the neutral plane),
resulting in a device with mild curvature towards the base layer. (Bii) Thick/asymmetric
devices after annealing have high compressive stress in the metal layer below the neutral
plane and equal tensile stress in both Parylene layers, resulting in a device curled towards
the top (thicker) Parylene layer. © IOP Publishing. Reproduced with permission from
[12] (CC BY 4.0). ..........................................................................................................................64
Figure 2-12: Modeled and experimental curvature diameter versus base layer annealing
temperature for PMP devices. The mean value of experimental data is plotted with error
bars showing one standard deviation (SD). Modeled parameters were identical to PMP
device parameters for each device group (Table 2-3). Stress was calculated from
temperature based on fitting literature values to all data (described in section 2.2), with
σtop calculated at 20 °C (representing no anneal). © IOP Publishing. Reproduced with
permission from [12] (CC BY 4.0). ...............................................................................................65
Figure 2-13: Modeled and experimental curvature diameter versus ratio of top to base layer
thickness for PMP devices before and after 200 °C anneal. The mean value of
experimental data is plotted with error bars showing one standard deviation (SD).
Modeled parameters were identical to PMP device parameters for each device group
(Table 2-3). Stress was calculated from temperature based on fitting literature values to
all data (described in section 2.2). © IOP Publishing. Reproduced with permission from
[12] (CC BY 4.0). ..........................................................................................................................67
Figure 2-14: Illustration of the stress in each layer of annealed PMP devices in regions with
etched Parylene openings. The neutral plane shifts in etched areas, resulting in different
stress balance in each area. (A) Regions with top Parylene etched openings (electrodes)
have moderate tensile stress in the base Parylene (below the neutral plane) and high
compressive stress in the metal layer (above the neutral plane), resulting in curling
towards the base layer. (B) Regions with base Parylene etched openings (bondpads) have

xi
moderate tensile stress in the top Parylene layer (above the neutral plane) and high tensile
stress in the metal layer (due to deposition directly on silicon, not Parylene; below the
neutral plane), resulting in curling towards the base layer. © IOP Publishing. Reproduced
with permission from [12] (CC BY 4.0)........................................................................................68
Figure 2-15: Representative photos of PMP devices annealed at 100 °C for 12 hr. Column (A)
shows thin/symmetric devices (4.4 + 4.7 µm thickness) and (B) shows thick/asymmetric
devices (3.4 + 11.5 µm thickness). Row (i) shows the region of the device with no etched
openings, (ii) shows the region of the device with etched openings in the top Parylene
layer (electrodes), and (iii) shows the region of the device with etched openings in the
base Parylene layer (bondpads). Scale bar for all photos (bottom left corner) is 3 mm; the
measured average diameter (mean ± standard deviation) and number of samples
measured are shown in the bottom right corner of each image. *Indicated devices curled
towards both the base and top Parylene layers in different regions/samples, so the
minimum diameter in each direction is included (the average diameter does not capture
the physical shape because a flat device has an infinite diameter). †No standard deviation
is included due to insufficient sample size; measured values are listed. © IOP Publishing.
Reproduced with permission from [12] (CC BY 4.0)....................................................................69
Figure 2-16: Photo of the electrode region of a thin/symmetric PMP device annealed at 200 °C
for 12 hr. Local regions with etched top Parylene (marked with red arrows) are curved
towards the base layer, while regions between the etched openings (with a symmetric
Parylene-metal-Parylene stack) are noticeably flatter. © IOP Publishing. Reproduced
with permission from [12] (CC BY 4.0)........................................................................................69
Figure 2-17: Representative photos of PMP devices annealed at 200 °C for varying times.
Column (A) shows thin/symmetric devices (4.4 + 4.7 µm thickness) and (B) shows
thick/asymmetric devices (3.4 + 11.5 µm thickness). Each row represents a different
annealing time – (i) before annealing, (ii) 0.5, (iii) 6, (iv) 12, and (v) 48 hours. Scale bar
for all photos (bottom left corner) is 3 mm; the measured average diameter (mean ±
standard deviation) and number of samples measured are shown in the bottom right
corner of each image. †No standard deviation is included due to insufficient sample size;
measured values are listed. © IOP Publishing. Reproduced with permission from [12]
(CC BY 4.0)...................................................................................................................................71
Figure 2-18: Modeled and experimental curvature diameter versus full device annealing
temperature for PMP devices. The mean value of experimental data (where available) is
plotted with error bars showing one standard deviation (SD). When the mean could not
be calculated, all data points are plotted with an ‘x’. Modeled parameters were identical
to PMP device parameters for each device group (Table 2-3). Stress was calculated from
temperature based on fitting literature values to all data (described in section 2.2). © IOP
Publishing. Reproduced with permission from [12] (CC BY 4.0).................................................73
Figure 2-19: Representative photos of PMP devices annealed at varying temperatures for 12
hours. Column (A) shows thin/symmetric devices (4.4 + 4.7 µm thickness) and (B) shows
thick/asymmetric devices (3.4 + 11.5 µm thickness). Each row represents a different
annealing temperature – (i) before annealing, (ii) 100 °C, (iii) 150 °C, and (iv) 200 °C.
Scale bar for all photos (bottom left corner) is 3 mm; the measured average diameter
(mean ± standard deviation) and number of samples measured are shown in the bottom
right corner of each image. *Indicated devices curled towards both the base and top
Parylene layers in different regions/samples, so the minimum diameter in each direction
is included (the average diameter does not capture the physical shape because a flat
device has an infinite diameter). © IOP Publishing. Reproduced with permission from
[12] (CC BY 4.0). ..........................................................................................................................74
Figure 2-20: Examples of PMP devices (210-350 µm width by 20-40 mm length, variable
thickness) thermoformed into 0.25 mm helices, showing a good result (✓), minor

xii
cracking (∗), cracking (×), and loose shape (●). © IOP Publishing. Reproduced with
permission from [12] (CC BY 4.0). ...............................................................................................77
Figure 3-1: Diagrams of the cerebral arteries (left) and cerebral veins (right). Arteries and veins
commonly targeted in endovascular recording and stimulation studies are labeled. The
asterisk (*) denotes typical deep venous target for device placement. © IOP Publishing.
Reproduced with permission from [1]. ..........................................................................................86
Figure 3-2: Diagram of the novel endovascular device (straight configuration). The fully
assembled device on a guidewire (coiled for packaging) is shown at the top right, and a
detailed view of the tip of the guidewire through the beginning of the wire region is
shown at the bottom.......................................................................................................................90
Figure 3-3: Diagram of the novel endovascular device (helical configuration). The fully
assembled device on a guidewire (coiled for packaging) is shown at the top right, and a
detailed view of the tip of the guidewire through the beginning of the wire region is
shown at the bottom.......................................................................................................................90
Figure 3-4: Diagram of 4x200 µm, 3x200 µm, and 2x200 µm aggregate electrodes. Each
aggregate electrode consists of 2-4 connected 200 µm electrodes to increase the electrode
surface area without increasing the required width of the Parylene backbone. Oblong
electrodes were not used as they are at higher risk of cracking during the thermoforming
step.................................................................................................................................................91
Figure 3-5: Schematic of the first-generation endovascular electrode array. .............................................92
Figure 3-6: Schematic of the electrode region of the first-generation endovascular electrode array..........92
Figure 3-7: Schematic of the bondpad region of the first-generation endovascular electrode array. .........93
Figure 3-8: Schematic of the second-generation endovascular electrode array in 16 electrode (top)
and 8 electrode (bottom) configurations. .......................................................................................94
Figure 3-9: Schematic of the electrode region of the second-generation endovascular electrode
array in 8 electrode (top) and 16 electrode (bottom) configurations. ............................................95
Figure 3-10: Schematic of the bondpad region of the second-generation endovascular electrode
array in 8 electrode (top) and 16 electrode (bottom) configurations. ............................................96
Figure 3-11: Diagram of the wire attachment point....................................................................................97
Figure 3-12: Order of fabrication steps for each design iteration. Thermoforming (highlighted in
blue) and attachment to the connector (highlighted in green) were moved later in the
process for groups B and D. Mounting on the guidewire (highlighted in yellow) was
moved earlier in the process for group G. Thermoforming was not performed for groups
C, E, and F, and attachment to the connector was not performed for group A..............................98
Figure 3-13: Diagram of a simple Parylene-based electrode array. Center image shows a top
down view with circular electrodes at the top, rectangular bondpads at the bottom, and
traces connecting each electrode to a bondpad. Cross sections of the electrode, trace, and
bondpad regions are shown in sections A-A’, B-B’, and C-C’, respectively. Cross
sections on the left show the single-sided configuration (with all Parylene openings in the
top layer), and cross sections on the right show the double-sided configuration (with
electrode openings in the top layer and bondpad openings in the base layer). ..............................99
Figure 3-14: Parylene electrode array thin film fabrication process flow for single-sided devices,
with cross-sectional views (left) at the indicated points on the generation 1 and 2 devices
shown with the red dotted arrows (right). (A) Parylene was deposited on a silicon carrier
wafer. (B) Platinum (with titanium adhesion layer; Ti/Pt) electrodes, traces, and bondpads
(not shown) were deposited and patterned. (C) Parylene was deposited on top of patterned
metal. (D) Electrodes and bondpads (not shown) were opened via O2 etching. (E) The
electrode arrays were released. ....................................................................................................101
Figure 3-15: Photos of electrode arrays with no cracking (left) and cracked Parylene (right).
Cracking is most easily visible in high magnification images (bottom)......................................102
Figure 3-16: Photo of a thermoformed device mounted on the guidewire. ..............................................104

xiii
Figure 3-17: Photo of the electrode end of a generation 2 linear device mounted on the guidewire,
with arrows indicating locations where epoxy was applied. More epoxy was added on the
bondpad end to prevent Parylene damage due to movement of the wires. ..................................104
Figure 3-18: Photo of a flat electrode array with two wires attached. Bondpads were attached in
three rounds; two bondpads were skipped between each attached bondpad due to
interference of the Kapton tape....................................................................................................106
Figure 3-19: Photo of a thermoformed electrode array with one wire attached. Bondpads were
attached one at a time...................................................................................................................106
Figure 3-20: Examples of defects identified during visual inspection: (A) large/deep Parylene
cracks, (B) small/partial-thickness Parylene cracks, (C) broken traces, (D) shorted traces,
(E) delaminated metal which has remained attached, and (F) delaminated metal which
has detached. Most defects are visible under the stereoscope; the high magnification
microscope is required to determine the severity of Parylene cracking (as shown in B).
All scale bars are 200 µm. ...........................................................................................................110
Figure 3-21: Continuity testing setup, with a cross-sectional view of the Teflon plate holding the
electrode array. The left image shows the setup as performed before wire attach (with a
probe and small PBS droplet over one bondpad). The right image shows the setup as
performed after wire attach (with a probe connected to the attached wire).................................111
Figure 3-22: Example EIS data for a broken (green x) and functional (blue circle) electrode. ................112
Figure 3-23: Simulated surgical handling fixture, with interchangeable mandrels for variable bend
radius testing................................................................................................................................113
Figure 3-24: Generation 1 (top) and generation 2 (middle and bottom) electrode arrays
immediately after removal from the carrier wafer. Left images show the full electrode
array, right images show zoomed in views of select features on the devices at the areas
indicated by the red dotted rectangles. Scale bars shown apply to all images within the
column. ........................................................................................................................................115
Figure 3-25: Photos of the group C straight device build electrode and bondpad regions (top).
Parylene tearing failures due to torquing of the guide wire commonly occurred in the
etched electrode region (green, bottom left) and wire attach region (blue, bottom right). ..........118
Figure 3-26: Photographs of a fully fabricated novel endovascular device (straight configuration –
group E). (A) The fully assembled device alone and (B) in a container for transportation.
(C) The tip of the guidewire with the electrode array mounted on it, with dotted boxes
showing (D-green) the electrode region and (E-blue) the bondpad region of the device.
(F) The connector with 3 header pins to connect to the 3 functional electrodes..........................119
Figure 3-27: Photographs of a fully fabricated novel endovascular device (curled configuration -
group G). (A) the fully assembled device alone. (B) The tip of the guidewire with the
electrode array mounted on it, with dotted boxes showing (C-green) the electrode region
and (E-blue) the bondpad region of the device............................................................................120
Figure 3-28: EIS plots of representative electrodes of each size for a device after wire attach
(before mounting to the guidewire). ............................................................................................122
Figure 3-29: Sample EIS data for a functional electrode (blue), an electrode channel with a
cracked trace (yellow), and a fully disconnected channel (gray).................................................123
Figure 3-30: Surgical simulation results, with plots of EIS data immediately after assembly and
prior to insertion into a catheter (red), after insertion into a straight catheter (orange), and
after sequential insertions into the catheter routed around bends of decreasing size
(yellow through purple). Negligible changes are observed for all but the smallest bend
diameter, and an increased impedance magnitude and phase near -90 is observed for the
10 mm bend diameter, indicative of a broken electrode. .............................................................124
Figure 3-31: Results from the group A sham implantation demonstrating surgical feasibility. (A)
Venogram (lateral view, left to right = anterior to posterior) showing a microcatheter (tip
located at the circle) advanced into the vein of the corpus callosum (arrowhead). (B)

xiv
Post-mortem MRI (sagittal view, left to right = anterior to posterior) confirming the
location of the microcatheter (arrow). (C and D) Venograms (lateral view, left to right =
posterior to anterior) showing the sham device (arrowhead) delivered into the superior
sagittal sinus. (E) Photo of the explanted sham device advanced out of a microcatheter............125
Figure 3-32: Recordings from the novel, thin film endovascular electrode in sheep (blue; group
C) under (top) 2% and (bottom) 1% isoflurane. LFPs (amplitude spikes in the recorded
signal) increase in frequency under lighter anesthesia. This is consistent with surface
recordings in rat (black) in a similar study (adapted from [53] with permission from
Elsevier, CC BY NC ND)............................................................................................................126
Figure 3-33: Recording from the novel, thin film endovascular electrode in sheep (blue; group C)
and a conventionally-manufactured endovascular microelectrode (red) from a similar
study (adapted from [6] with permission from Elsevier). LFPs (amplitude spikes in the
recorded signal) from the thin film electrode have similar frequency and slightly higher
amplitude than the conventional microelectrode. ........................................................................127
Figure 3-34: Simultaneous recordings from a surface electrode (top) and an endovascular
electrode (bottom; group E). Although the signal from the endovascular electrode is
smaller, the spikes track with the signal on the surface electrode, suggesting it is
recording the same neural activity. ..............................................................................................127
Figure 3-35: Intraoperative angiogram from the group F device implantation (lateral view, left to
right = anterior to posterior) showing placement of endovascular electrodes (red arrow),
surface (ECoG) electrodes (black arrow) and penetrating (SEEG) electrodes (green
arrow)...........................................................................................................................................128
Figure 3-36: Simultaneous recordings from surface macro-electrodes (black), penetrating macroelectrodes (green), penetrating micro-electrodes (red) and endovascular micro-electrodes
(blue; group F) showing simultaneous bursts of electrical activity on all electrodes. .................129
Figure 4-1: Anatomy of a peripheral nerve...............................................................................................136
Figure 4-2: Diagram of the fully assembled peripheral nerve full cuff. ...................................................138
Figure 4-3: Diagram of the fully assembled peripheral nerve helical cuff. ..............................................138
Figure 4-4: Schematic of a 6x18 aggregate electrode...............................................................................140
Figure 4-5: Schematic of the bridged traces. ............................................................................................142
Figure 4-6: Schematic of the large, helical cuff electrode (design B3-). ..................................................143
Figure 4-7: Schematic of the small, helical cuff electrode (design B3-)...................................................143
Figure 4-8: Schematic of the large, full cuff electrode (design E3-).........................................................144
Figure 4-9: Schematic of the small, full cuff electrode (design E3-)........................................................144
Figure 4-10: Parylene electrode thin film fabrication process flow, with cross-sectional views
(left) at the indicated point on the B3-small device shown with the red dotted arrows
(right). (A) Parylene was deposited on a silicon carrier wafer. (B) Metal electrodes,
traces, and bondpads (not shown) were deposited and patterned. (C) Parylene was
deposited on top of patterned metal. (D) Electrodes and bondpads (not shown) were
opened via O2 etching. (E) The electrode arrays were released...................................................146
Figure 4-11: Device setup for the wire attach process..............................................................................147
Figure 4-12: (Left) device fixtured for thermoforming, with (right) zoomed in view of Parylene
electrode region wrapped around the mandrel and held in place with a Teflon film. More
detailed process images are included in appendix F. ...................................................................148
Figure 4-13: (Left) the Parylene cable reinforced with Kapton tape and hand painted PDMS, with
the thermoformed electrode region placed inside a dispensing tip to prevent loosening
during elevated temperature curing. (Center) the base and top of the mold filled with
liquid PDMS, with the device loaded into the base of the mold. (Right) the closed mold. .........150
Figure 4-14: Example CV data in H2SO4 for a functional electrode with characteristic peaks
labeled and the hydrogen desorption and cathodic charges shaded in blue and purple,
respectively. .................................................................................................................................153

xv
Figure 4-15: Example input current pulse (blue; 1 mA, 500 µs pulse width, 100 µs interphase
delay) and resulting voltage transient (red), illustrating the interphase potential (Ep).................155
Figure 4-16: Device fixtured into the glass jar filled with saline (1x PBS) for accelerated life
testing...........................................................................................................................................156
Figure 4-17: Use of the sliding surgical placement tool, in which the cuff is (left) loaded on the
tool and placed over the nerve, (center) slid off the tool and onto the nerve, and (right)
closes softly over the nerve..........................................................................................................158
Figure 4-18: Photos of the fabricated Parylene devices in (left) full and (right) helical cuff
configurations. .............................................................................................................................159
Figure 4-19: Fabricated and packaged cuff electrodes in the (left) full cuff and (right) helical cuff
configurations, with (bottom) zoomed in images of the thermoformed electrode regions,
with exposed metal electrodes on the inner surface of the cuff. ..................................................160
Figure 4-20: CV curves in H2SO4 for batch S (blue) and batch U (black) small, full cuffs prior to
thermoforming. Batch S devices have larger H+ desorption and adsorption peaks (range
of 12 to 24 µA and -47 to -28 µA, respectively) as compared to batch U devices (5 to 16
µA and -33 to -16 µA, respectively)............................................................................................161
Figure 4-21: CV curves in H2SO4 at 250 mV/s for a representative small, helical cuff before (flat
device – solid line) and after (dashed line) thermoforming. Peaks are higher magnitude
and more clearly defined before thermoforming. ........................................................................162
Figure 4-22: CV curves in ferri/ferrocyanide for a representative small, helical cuff (same device
as illustrated in Figure 4-21) before (flat – solid lines) and after (dashed lines)
thermoforming. Peak magnitudes are similar before and after thermoforming, indicating
similar ESA..................................................................................................................................163
Figure 4-23: CV curves in H2SO4 at 250 mV/s for representative flat devices of each
configuration from batch U. Note that the larger curves correspond to larger ESAs, as
listed in Table 4-5. .......................................................................................................................164
Figure 4-24: CV curves in PBS 50 mV/s for representative packaged (thermoformed and molded)
devices of each configuration from batch U. ...............................................................................165
Figure 4-25: EIS data in PBS for representative packaged (thermoformed and molded) devices of
each configuration from batch U. ................................................................................................165
Figure 4-26: VT data in PBS with 500 µs pulse width and 100 µs interphase delay for
representative packaged (thermoformed and molded) devices of each configuration from
batch U after thermoforming. ......................................................................................................166
Figure 4-27: (Left) CSC of bare platinum electrodes with (o) and without (x) stimulation over 13
days accelerated time. Each line represents a single electrode measured over several time
points. Devices show gradual decrease of CSC towards or below the failure threshold of
100 µC/cm2
. (Right) representative CV curves (200 mV/s) for stimulated (top) and
unstimulated (bottom) devices at 0, 0.5, 1, 2, 7, and 13 accelerated days, demonstrating
rapid disappearance of hydrogen and oxide peaks within the first day, followed by
gradual narrowing of the CV curve. ............................................................................................169
Figure 4-28: (Left) 1 kHz impedance magnitude of bare platinum electrodes with (o) and without
(x) stimulation over 13 days accelerated time. Each line represents a single electrode
measured over several time points. Devices show gradual increase of impedance above
the failure threshold of 1 kΩ. (Right) representative EIS curves for stimulated (top) and
unstimulated (bottom) devices at 0, 0.5, 1, 2, 7, and 13 accelerated days, demonstrating
gradual shift towards higher impedance. .....................................................................................170
Figure 4-29: (Left) CIC of bare platinum electrodes with (o) and without (x) stimulation over 13
days accelerated time. Each line represents a single electrode measured over several time
points. Devices show rapid degradation below the failure threshold of 50 µC/cm2
and
stimulation charge density of 22.1 µC/cm2
..................................................................................171

xvi
Figure 4-30: (Left) CSC of PtIr-coated electrodes with (o) and without (x) stimulation over 14
days accelerated time. Each line represents a single electrode measured over several time
points. Devices show an increase in CSC over the first few days, followed by a slight,
gradual decrease. (Right) representative CV curves (200 mV/s) for stimulated (top) and
unstimulated (bottom) devices at 0, 0.5, 1, 2, 7, and 14 accelerated days, demonstrating
rapid increase, then slow decrease in area from the starting curve (dark purple) to day 14
(red) which appears to be stabilizing. ..........................................................................................172
Figure 4-31: (Left) 1 kHz impedance magnitude of PtIr-coated electrodes with (o) and without (x)
stimulation over 14 days accelerated time. Each line represents a single electrode
measured over several time points. Devices show a rapid decrease of impedance within
the first 12 hours (in all but one sample) followed by very stable measurements. (Right)
representative EIS curves for stimulated (top) and unstimulated (bottom) devices at 0,
0.5, 1, 2, 7, and 14 accelerated days, demonstrating stability after the first 12 hours. ................173
Figure 4-32: (Left) CIC of PtIr-coated electrodes with (o) and without (x) stimulation over 14
days accelerated time. Each line represents a single electrode measured over several time
points. Devices show an increase in CIC over the first few days, followed by a slight,
gradual decrease...........................................................................................................................174
Figure 4-33: Helical cuff implanted on the sciatic nerve of a rat..............................................................175

xvii
Abstract
The field of microelectromechanical systems (MEMS) has enabled the creation of microscale
systems which have impacted a number of fields. In the field of neural interfaces, MEMS has enabled
significant miniaturization of electrodes from the millimeter scale to the micron scale. This reduction in
size allows neural interfaces to be less invasive, reducing the body’s immune response and improving
patient outcomes, and to interface with the tissue on a smaller size scale, increasing spatial resolution of
neural recording or stimulation. MEMS devices, however, are built in a planar configuration, limiting
their ability to interface with complex 3D geometry in the body. To overcome this challenge, Parylenebased planar MEMS devices can be permanently transformed into 3D shapes via post-processing,
enabling countless more applications of such devices to interface with non-planar anatomy.
This work first discusses the post-processing of Parylene-based MEMS devices to produce 3D
structures via the modulation of film stress and thermoforming, described in chapter 2. Chapters 3 and 4
apply that process to develop two novel devices to interface with complex anatomy in the body. The first
(chapter 3) is an endovascular electrode array for neural recording from within the blood vessels, aimed at
minimally invasive seizure monitoring. The second (chapter 4) is a peripheral nerve cuff electrode for
chronic stimulation of small diameter branched nerves, designed to produce more targeted
neuromodulation for a variety of different applications.

1
Chapter 1. Intro to Parylene-Based Neural Interfaces
1.1 Neural Recording and Stimulation
Neural interfaces, devices which communicate with neural tissue via electrical activity, are
integral in advancing knowledge of the nervous system and body’s functions. A variety of neural
interfaces have been developed to interface with the nervous system in different ways. As these devices
and surgical techniques have advanced, devices have been able to interface more closely with tissue to
provide greater clinical benefits and to answer more complex research questions. Neural recording is
often used clinically to diagnose abnormal nervous system function, monitor nervous system disorders,
and provide feedback to implanted or external devices. Neural stimulation is used to modulate nervous
system and end organ activity. Although rare in clinical applcations, recording and stimulation can be
used together for closed-loop monitoring and stimulation.
Electrodes can interface with the nervous system in several different locations (summarized in
Table 1-1). The most common point of recording or stimulation is the brain, which has a wide variety of
recording applications (including brain computer interfaces (BCI), monitoring of abnormal activity due to
neurological disorders, and general research [1-16]) and stimulation applications (most commonly for
treatment of neurological disorders or to produce artificial vision or hearing [1,6-9,15,17-21]) which can
be used in parallel for closed-loop systems [1,6-9,22]. A less common interface point is the spinal cord,
which is primarily accessed for the use of stimulation to treat chronic pain [17,18,23-25], however other
stimulation applications (such as limb control [2]), recording studies (for general research [4,11-14]), and
closed-loop systems [13] have been published as well. The final option is the peripheral nerves, which is
most commonly used for neural stimulation (for pain or epilepsy treatment, sensory feedback for
prosthetic limbs, or to produce artificial hearing or vision [4,9,15,26-33]) but also has several recording
applications (decoding sensory information, prosthetic limb control, and general research [4,9,11,12]).

2
Table 1-1: Summary of recording, stimulation, and closed-loop applications in each region of the nervous system. Note that the
cited publications are a representative sample of such applications and not an exhaustive list.
Region Applications
Recording Stimulation Closed-Loop
Brain
[1-22,34,35]
• BCI
• Monitoring of
neurological disorders,
such as:
o Sleep disorders
o Mood disorders
o Movement disorders
o Paralysis
o Epilepsy
• General research
• Treatment of neurological
disorders, such as
o Mood disorders
o Movement disorders
o Epilepsy
o Chronic pain
• Artificial vision
• Artificial hearing
• Epilepsy
treatment
Spinal cord
[2-4,6,15,16,
18,19,24,25]
• General research • Treatment of chronic pain
• Limb movement control
• General
research
Peripheral
nerves
[4,9,11,12,15,
26-33,36]
• Decoding sensory
information
• Prosthetic control
• General research
• Treatment of neurological
disorders, such as
oEpilepsy
oChronic pain
• Prosthetic limb sensory
feedback
• Artificial vision
• Artificial hearing
• Prosthetic
limb
feedback and
control
Many electrodes and electrode arrays have been designed to interface with the brain, spinal cord,
and peripheral nerves. In general, electrodes which are less invasive (and thus lower risk to the patient)
are capable of lower resolution recording or stimulation. As electrodes become more invasive (i.e. as they
are implanted closer to the target neural tissue), both the resolution and the implantation risk increase.
The most common types of electrodes used for neural interfacing are shown in Figure 1-1 and
Table 1-2 and described in the sections below.

3
Figure 1-1: The main types of neural recording and stimulation electrodes as implanted in the brain, spinal cord, and peripheral
nerve. Details on each electrode type are included in sections 1.1.1 (scalp electrodes), 1.1.2 (surface electrodes), and 1.1.3
(penetrating electrodes). Images are not to scale.

4
Table 1-2: Summary of electrode types for neural interfaces.
Stimulation
N
o
Yes
Yes
Yes
Yes
Yes
Yes
No
Signal Type (Recording)
Large neuronal population
(1-3 cm, <70 Hz)
Local field potentials (0.5-5
mm, <300 Hz)
Local field potentials (0.5-5
mm, <300 Hz)
Local field potentials (0.5-5
mm, <300 Hz) and single
units (200
µm, 300-5000 Hz)
Local field potentials (0.5-5
mm, <300 Hz) and single
units (microelectrodes; 200
µm, 300-5000 Hz)
Single units (200
µm, 300- 5000 Hz)
Electrode Size
4-10 mm diameter
0.5-4 mm diameter
40
µ
m-4 mm diameter
1-2.5 mm electrode
width, full nerve
circumference; 40-1500
µm edge or diameter
0.86-1.27 mm probe
diameter, 1.3-3.0 mm
electrode height
80
µm diameter tapered
to point, 50
µm exposed
length
25-80 µm wire
diameter, 0-20
µm
exposed length
3.5-40 µm fiber
diameter, 0-250 µm
exposed length
Device Configuration
10-128 circular electrodes
(over full scalp)
1-27 circular electrodes
(various sizes)
1-64 circular electrodes
(various sizes)
2-3 ring electrodes;
rectangular or circular
electrodes for special use
cases (retinal, cochlear, etc.)
4-10 ring electrodes, 5-10
mm pitch; 26-40 cm total
length; can contain
microelectrodes
1-100 insulated probes with
exposed tip; 0.5-1.5 mm
length
Single insulated wire with
exposed tip; up to 10 mm
length
Single insulated fiber with
exposed tip; up to 10 mm
length
Location
Scalp
Brain, spinal
cord (above
dura)
Brain, spinal
cord (beneath
dura)
Peripheral
nerve (above
epineurium)
Brain
Brain, spinal
cord, or
peripheral nerve
Electrode Type
Scalp electrodes
[5,37-42]
Epidural electrodes
[20,26,27,37-39,
43-46]
Subdural electrodes
[3,10,16,20,26,27,37- 39,43-45,47,48]
Extraneural
electrodes
[28,29,33,49-52]
SEEG/DBS probes
[1,7,20,37,38,48,
53-55]
Utah array
[12,35,38,56-58]
Microwires
[37,38,59-63]
Carbon fibers
[14,64-70]
Microelectrodes
Surface Penetrating

5
Table 1-2 (continued): Summary of electrode types for neural interfaces.
Stimulation
Yes
No
No
Yes
No
No
Yes
Signal Type (Recording)
Single units (200
µm, 300- 5000 Hz)
Local field potentials (0.5-5
mm, <300 Hz) and single
units (200
µm, 300-5000 Hz)
Unknown
Local field potentials (0.5-5
mm, <300 Hz) and single
units (200
µm, 300-5000 Hz)
Electrode Size
10-55 µm diameter
0.2 mm square
(theoretical:1-5
µm)
40-200
µm in diameter
0.2-0.6 mm wire
diameter, 1.5-60 mm
exposed length
1 mm probe
diameter,
1-3 mm electrode
height; 40
µm
(microelectrodes)
0.6-20 µm wire
diameter, 0-10
µm
electroplated tip length
500-750
µm diameter
Device Configuration
4-256 circular or rectangular
electrodes on rectangular
shank (10-550 width x 1-50
µm thickness) with pointed
tip
200 µ
m square electrodes on
0.8 x 3 x 1 mm dust motes
(theoretical size: 10-100
µm
cube)
Sheet or body with
electrodes in holes for nerve
growth (ring electrodes)
Single insulated wire with
exposed tip; 40-200 cm
length
4 ring electrodes, 5 mm
pitch; 20 circular
microelectrodes
Single insulated wire with
exposed tip, up to 20 mm
length; multiple wires can be
bundled
6-16 circular electrodes on
commercial stent (3-4 mm
expanded diameter, 32 mm
length)
Location
Brain, spinal
cord, or
peripheral nerve
Brain or
peripheral nerve
Peripheral
nerve
Brain, spinal
cord, or
peripheral nerve
Electrode Type
Multisite probes
[11,12,37,38,
71-86]
Neural dust
[87-89]
Regenerative electrodes
[49,50,90]
Catheter- & guidwirelike electrodes
[40,45,91-94]
Multi-electrode probes
Micro
[47,95]
Microwires
[4,13,96]
Stentrode
[20,26,43,44,97,98]
electrodes
Penetrating Endovascular

6
1.1.1 Scalp Electrodes
Scalp electrodes are large electrodes (4-10 mm in diameter [37]) placed on the surface of the
scalp (Figure 1-1A), usually in large grids (see Figure 1-2), to record general neuronal activity (scalp
electroencephalography (EEG)). This non-invasive method poses minimal risk to patients and is
commonly used for interpretations of brain states and neurological diseases such as epilepsy or for basic
BCIs [3,5]. Due to the large distance between the electrodes and the target tissue (2-3 cm of scalp and
skull), EEG recordings represent average electrical activity over a large neuronal population (1-3 cm
diameter region of tissue) [38,39] and are usually low amplitude (hundreds of microvolts) and frequency
(0-100 Hz) [37].
Figure 1-2: Picture of a headcap with 128 scalp EEG electrodes. Reproduced with permission from [41] (CC BY 4.0).
1.1.2 Surface Electrodes
Surface electrodes are any electrodes which are placed on the surface of neural tissue (or on top
of the dura or epineurium) without penetrating into it. They come in several configurations which are
tailored to their specific application. The following sections detail the main types of surface electrodes.
1.1.2.1 Epidural and Subdural Electrodes (Brain and Spinal Cord)
Epidural electrodes are placed on the surface of the dura (Figure 1-1B) and subdural electrodes
are placed under the dura directly on the neural tissue surface (Figure 1-1C). Electrodes are smaller than
those used for scalp EEG (40 µm to 4 mm in diameter [10,20,26,27,37,43,44,47]) and often configured in
a multi-electrode grid or strip to cover a region of interest on the tissue (see Figure 1-3).
In the brain, epidural and subdural electrodes are commonly used for electrocorticography
(ECoG), which aims to record local field potentials (LFPs; <300 Hz) over a smaller neuronal population
(0.5-5 mm diameter region of tissue) than is accessed with scalp EEG [10,38,39]. Signal magnitudes are

7
in the hundreds of microvolts to several millivolts. ECoG is used for BCI, seizure monitoring, or other
research purposes for which LFP recordings are useful [10,15].
Subdural electrodes are also often used for stimulation of brain regions near the cortical surface,
and epidural electrodes are used commonly in spinal cord stimulation [13,17,18,23-25]. In combination
with neurostimulators, these electrodes can be used for the treatment of neurologic disorders such as
epilepsy, Parkinson’s and other movement disorders, or chronic pain. Subdural electrodes have also been
used for closed-loop BCIs to provide user feedback during control of a robotic limb [6,25,99].
The implantation procedure of epidural and subdural electrodes is highly invasive, requiring brain
or spinal surgery with a large surgical opening (and, for subdural electrodes, opening of the dura) due to
the large size of the electrode grid, however such grids allow for more focused electrical stimulation and
provide useful information for recording applications (such as the localization of epileptic foci and the
diagnosis of other neurologic disorders) that cannot be obtained using scalp EEG.
Figure 1-3: Picture of an ECoG array with 8 × 500, 750, and 1000 µm electrodes. Scale bar 5 mm. Reproduced with permission
from [44] (CC BY 4.0).
1.1.2.2 Extraneural Electrodes (Peripheral Nerve)
Surface electrodes in the peripheral nervous system are called extraneural electrodes and record
from the surface of the epineurium (Figure 1-1D). There are a variety of extraneural electrode types
which surround the nerve fiber, such as cuff (Figure 1-4), book, and helical electrodes, with variable
electrode sizes (40-1500 µm rectangles or circles or 1-2.5 mm wide rings/helices) [28,29,49-52]. Some
devices are designed to interface with specific nerves, such as cochlear or retinal implants, which are
generally arrays of circular microelectrodes (40-250 µm in diameter [28-30,90]) that sit on the surface of
the tissue.

8
Electrical signals in the peripheral nerves are generally in the range of 1-5000 Hz. Recordings are
often noisy due to nearby external signals (such as EMG or interference from nearby nerves), however
this can be alleviated by using tri-polar electrodes for noise rejection [49]. Extraneural recording
electrodes are most commonly used to decode sensory information (for example, when nerves are intact
after spinal cord injury but cannot communicate with the brain) for prosthetic limb control [9].
Extraneural stimulation is more common than recording, with a variety of applications such as
motor function augmentation, pain treatment, epilepsy treatment, prosthetic feedback control (open- or
closed-loop), artificial vision, and artificial hearing [4,9,15,26-30]. Stimulation of larger nerve fibers is
more common, however selective stimulation of large fibers is more difficult than stimulation of
downstream, branched nerves responsible for fewer functions [9].
To implant extraneural electrodes which wrap around nerves, the tissue is opened and the nerve is
exposed, then the electrode is wrapped around the nerve and anchored in place (usually with suture). The
procedure is highly invasive, however it can pose a lower risk to the patient if the nerves do not control
functional tissue (such as nerves which control amputated limbs, which are used for prosthetic limb
sensation or control). One common complication of extraneural electrodes is over-tightening of the
electrode around the nerve, resulting in nerve constriction and damage [49]. Implantation of specialized
devices (such as retinal or cochlear electrodes) is tailored to the specific device design and use.
A variety of types of extraneural electrodes have been developed, the most common of which is
the cuff electrode. Cuff electrodes form a tube or loop around the nerve with exposed metal electrodes on
the inside of the tube. Cuffs have been designed in a variety of different configurations, including split
cuffs (Figure 1-4A), which consist of a basic tube structure which is split lengthwise so it can be placed
around the nerve, spiral cuffs (Figure 1-4B), which consist of a long, pre-curled sheet which is rolled
around the nerve, and helical cuffs (Figure 1-4C), which consist of a helical structure which wraps around
the nerve but have an open architecture which does not fully encapsulate the nerve [49,50].

9
Figure 1-4: Examples of three extraneural cuff electrodes: (A) a piano hinge split cuff, (B) a spiral cuff, and (C) a helical cuff.
Reproduced from [49] with permission from Elsevier.
To date, all clinical extraneural electrodes and most research devices target nerve diameters
approximately 1 to 5 mm in diameter [33,51,100-102]. To achieve more intimate contact with smaller
nerve fibers, some devices utilize thin film polymers or other soft materials to produce smaller diameter
devices. Most commonly, these devices are spiral cuffs (due to easier fabrication of this device
configuration at a smaller size) and can target nerves down to 150 µm in diameter [33,100].
1.1.3 Penetrating Electrodes
Penetrating electrodes are electrodes which are inserted into neural tissue (penetrating the tissue,
and often causing damage to tissue in their path). There are a variety of different penetrating electrodes
which are used for different purposes. The most commonly used penetrating electrodes are described in
the sections below.
1.1.3.1 Stereoelectroencephalography (SEEG) and Deep Brain Stimulation (DBS) Electrodes
Along with ECoG, SEEG and DBS probes are the most common clinically used invasive neural
interfaces in the brain. SEEG and DBS probes (Figure 1-1E) are long, cylindrical tubes inserted into deep
cortical tissue with an outer diameter of approximately 0.86 to 1.27 mm and 4 to 10 ring-shaped
electrodes (1.3-2.5 mm height) around the outer diameter of the probe [7,48,54,55]. Some SEEG probes
also contain microelectrodes on the tip or body of the probe to allow for more accurate spatial mapping of
electrophysiological signals (Figure 1-5) [48].
SEEG electrodes are implanted very close to the target tissue, so they are able to record LFPs
(<300 Hz) from a small neuronal population (~1 mm diameter region of tissue) with amplitudes on the
order of hundreds of microvolts (which can increase to millivolts during seizure recordings), with higher

10
resolution recordings (single units, 300-5000 Hz) on microelectrodes [13,37,38,99,103,104]. SEEG is
most commonly used for BCI, epilepsy monitoring, and movement disorder research (among other
research applications) [1,6-9].
DBS electrodes are used for stimulation in conjunction with an implantable neurostimulator to
stimulate different regions of the brain for treatment of various neurologic disorders (including epilepsy,
Parkinson’s, other movement disorders, chronic pain, or a variety of psychiatric disorders) [1,6-9,17-
19,21,48,55,105-107]. Stimulation and recording via DBS/SEEG probes can be used for closed-loop
systems, as is currently used in the Neuropace responsive neurostimulation system for epilepsy treatment
[1,6].
Insertion of SEEG or DBS leads requires a burr hole and dural incision above the implantation
site, which is a highly invasive procedure [7,8,21]. The overall size and stiffness of the probe makes it
rigid enough to directly penetrate the tissue without buckling, however the polymeric insulation does
allow the probe to flex when sufficient force is applied. To ensure straight insertion and placement at the
desired location, a rigid cannula or stylet can be used to create a track in the tissue prior to insertion of the
probe or to guide the probe along the desired trajectory [7,8,21,106]. The large size of the probe causes
the displacement of a large amount of tissue which has been found to cause irreversible damage to the
brain tissue [19]. After the probe is implanted, the body’s immune response attempts to attack and wall
off the foreign body, leading to encapsulation of the probe [9].
Figure 1-5: Picture of an AdTech macro-micro electrode with outer macroelectrodes (right arrow) and center microwires (left
arrow). © IOP Publishing. Reproduced with permission from [108].

11
1.1.3.2 Penetrating Microelectrodes
Penetrating microelectrodes are a subgroup of penetrating electrodes that have been downsized to
the micron scale to record or stimulate small regions of tissue with high resolution. They come in a
variety of configurations, the most common of which are described here.
Microelectrodes are sized to record single- or multi-units (300-5000 Hz, amplitude of tens to
hundreds of millivolts) [37,103,109,110]. They must be placed in close proximity to the target tissue
(within ~50 µm) and record signals from a very small region (200 µm diameter recording area) [13,38].
Penetrating microelectrodes are most commonly implanted in the brain for BCI and general recording
research in animal settings [2,3,9,11,12,14,34,35,56,57], however there are limited publications citing
recording in the spinal cord [11,12,14] and peripheral nerves (inter- and intrafascicular electrodes)
[11,12,49,50] and stimulation in the brain [12] and spinal cord [2].
Utah arrays (Figure 1-1F and Figure 1-6) are arrays of up to 256 needle-like silicon electrodes on
a rigid platform (Figure 1-1F). Each electrode shank is 80 µm in diameter at the base and tapers to a
point; shanks (1 to 1.5 mm in length) are fully insulated except for 35-75 µm at the tip which forms a
single electrode. The overall footprint size for a 100-electrode array is 4.2 x 4.2 mm [12,34,35,56,57,111].
Figure 1-6: SEM images of a Utah array containing 100 electrodes of 1.5 mm length. (A) shows the entire 10x10 electrode array,
and (B) shows a detailed image of the exposed electrode tip and insulated probe. Reprinted from [112] with permission from
Elsevier.
Microwires (Figure 1-1G and Figure 1-7) are long cylindrical wires with outer diameter ranging
from 25 to 80 µm [12,59-61,113]. Each microwire is a single, insulated wire with an exposed tip (0 to 20
µm exposed length). Arrays of microwires are often used for higher density recording [60].

12
Figure 1-7: (Left) SEM image of a microwire electrode (exposed wire tip) made of 70%/20% platinum/iridium wire. Scale bar
10.0 µm. Reprinted from [60] with permission from Elsevier. (Right) Picture of a microwire array with 1mm (left) and 8mm
(right) wire lengths. Reprinted from [60] with permission from Elsevier.
Carbon fibers (Figure 1-1G and Figure 1-8) are similar in shape and function to microwires; they
are small, insulated fibers with a single exposed tip ranging from 3.5 to 40 µm in diameter [64-69,114].
As with microwires, arrays of carbon fibers can be used for high density recording [68,114].
Figure 1-8: (Left) SEM image of a glass-coated carbon fiber. Reprinted from [64] with permission from Elsevier. (Right) A 2x8
array of carbon fiber electrodes. Reproduced with permission from [114] (CC BY 4.0).
Multisite arrays (Figure 1-1H) are planar probes with multiple electrodes on their face built on a
silicon (Figure 1-9) or polymer (Figure 1-10) backbone. Probe designs can vary widely, but most devices
feature multiple electrodes per shank (10 to 30 µm diameter) and an overall cross-sectional size of
approximately 120 to 200 µm wide by 15 to 50 µm thick [12,38,71-75]. Probes made of silicon are
generally stiff enough to be directly inserted into tissue without buckling but can be slightly flexible if
they are thin or long enough. Polymer-based probes are flexible and require an insertion aid (described in
section 1.3.1.2). In addition, arrays of probes are often inserted together to achieve high density recording.

13
Figure 1-9: (Left) SEM image of a 64 channel silicon multisite electrode array. Reproduced with permission from [72] (CC BY
4.0). (Right) Photograph of a Neuronexus multisite silicon array with 16 electrodes (4 shanks with 4 electrodes at the tip of each
probe). Reprinted from [12] with permission from Elsevier.
Figure 1-10: Photos of a 64 channel flexible Parylene multisite electrode array. © IOP Publishing. Reproduced with permission
from [115].
Insertion of penetrating microelectrodes is highly invasive, requiring a burr hole or craniotomy
(for brain) and incision/removal of the dura (for brain or spinal cord) or epineurium (for peripheral nerve)
above the insertion site [60,61,68]. The body’s immune response after the probe has been implanted
depends on the device configuration; devices with larger cross sections, high-density grouped shanks,
sharper edges, and stiffer materials produce a more severe immune response and will result in a thicker
encapsulation layer [9,12,75,105,116]. Penetrating microelectrodes are commonly inserted manually
(with or without a stereotactic frame) or with a motorized insertion tool mounted to a frame [12,25,56-
61,68,72,111]. To insert flexible devices (such as polymer-based multisite electrode arrays), insertion
tools or other methods are required (see section 1.3.1.2).

14
1.1.3.3 Neural Dust
Neural dust (Figure 1-1I) consists of free-floating electrodes (motes) which are implanted into
target tissue, a subdural transceiver which communicates with the motes, and an external receiver which
receives data from the transceiver. The theoretical size of the sensor motes should be approximately 10 to
100 µm cubes [88,89], however all published experimental devices are larger (0.8 x 1 x 3 mm – see
Figure 1-11), with electrodes sized for single unit recording [87].
Thus far, neural dust has only been implanted in the peripheral nerves of rodents by incising the
skin and manually placing the dust mote in contact with the nerve. The planned insertion technique for the
smaller, theoretical neural dust is by injection through a needle [87,88].
Figure 1-11: Photographs of a prototype neural dust mote. Reprinted from [87] with permission from Elsevier.
1.1.4 Regenerative Electrodes
Regenerative electrodes (Figure 1-1J and Figure 1-12) are peripheral nerve electrodes which are
implanted by transecting a nerve, placing the electrode between the two nerve ends, and allowing the
nerve to regenerate through the electrode. Devices consist of multiple electrodes, most commonly ringshaped (40-200 µm in diameter) through which the nerve will regrow [49,50].
Regenerative electrodes aim to record single- and multi-unit signals (300-5000 Hz) with
amplitudes on the scale of hundreds of microvolts [9]. Due to the highly invasive implantation procedure
and risk of unsuccessful nerve regeneration after transection, the most common application for

15
regenerative electrodes is in cases where nerve health is non-critical, such as in amputated limbs (with
nerve recordings used for prosthetic control or stimulation used for sensory feedback) [9,49].
Figure 1-12: Diagram (A), SEM image (B), and (C) photo of a regenerative sieve electrode. Reproduced from [49] with
permission from Elsevier.
1.1.5 Endovascular Electrodes
Endovascular electrodes (Figure 1-1K) are any electrodes which are inserted into the blood
vessels and routed to the target nervous tissue via a catheterization procedure. They are used to record
endovascular EEG (up to 20 Hz) and LFPs (up to 300 Hz) with signal amplitudes on the scale of 10s of
microvolts to millivolts, depending on implantation location and electrode size. Endovascular electrodes
are most commonly used to record activity in animal brain in applications such as seizure monitoring and
mapping, BCI, or general research, as a minimally invasive alternative to SEEG or ECoG electrodes [2-
4,9,13,20,26,27,43,44,47,94,96,97,117-125]. There have also been limited publications detailing
recording in spinal cord or peripheral nerve [4,13,22,126], recording in human brain (usually during
catheterization procedures for other purposes, and in one case in a human BCI trial) [40,91-
93,95,127,128], and stimulation or tissue ablation in human or animal brain [20,122,129-131]. A variety
of endovascular electrode configurations are in use, the most common of which are described here.
Early endovascular recording was performed using a standard metal guidewire with a standard
catheter used to insulate the entire length of the guidewire except for the tip (which acted as a single
electrode) [40,94,128]. Later, custom insulated guidewires with exposed metal electrodes (Figure 1-13)
[45,92,93,117-120,124,125,132] or catheters with embedded electrodes [91,122] were built specifically
for recording purposes. Guidewire- and catheter-like electrodes mimic standard guidewire and catheter

16
dimensions (0.2 to 0.6 mm diameter) with a 1.5 to 60 mm electrode length at the tip of the device
[40,45,91-94].
Figure 1-13: Schematic of the Seeker-10 guidewire with (1) stainless steel shaft, (2) Teflon coating, and (3) a single platinum
electrode at the tip. Reprinted from [92] with permission from Springer Nature.
Multi-electrode, SEEG-like probes (Figure 1-14) have multiple electrodes along their length to
achieve higher resolution recordings. These devices mimic SEEG probes, with several ring electrodes (1-
3 mm long) at the end of a catheter (~1 mm diameter), and optional microelectrodes (~40 µm diameter)
between larger contacts [47,95].
Figure 1-14: Schematic of a multi-electrode, SEEG-like endovascular probe used by Bower et al. with 4 × 1 mm ring electrodes
and 20 × 40 µm microelectrodes. Reprinted from [47] with permission from Elsevier.
Endovascular microwires (Figure 1-15) are thin, insulated wires (0.6 to 20 µm diameter) with a
conductive tip, which is often electroplated to produce a larger recording electrode area [4,13,96].
Because they are smaller than traditional guidewires or catheters, microwires can be used individually in

17
small veins or arteries (including in small animal models) or as bundles of multiple microwires (for multichannel recording).
Figure 1-15: Schematic of a nanowire (0.6-1 µm wire diameter) with platinum black electrode at the tip inside a microcatheter
(90-300 µm diameter) developed by Llinas et al. Reprinted from [13] with permission from Wiley.
The Stentrode (Figure 1-16) is a multi-electrode endovascular device built on a commercial stent
backbone. It contains 6 to 12 platinum disc electrodes (500 or 750 µm diameter, 50 µm thickness)
attached to a commercial, self-expanding stent (collapsed diameter of 1.33 mm, expanded diameter of 3 to
4 mm, 31.1 to 32 mm length), sized to target the superior sagittal sinus. A stainless steel stylet (310 to 410
µm diameter) is attached to the stent to aid insertion [20,26,43,44,97,98,133]. Of all existing endovascular
electrodes, the Stentrode is the most widely studied chronic implant, with the longest published
implantation of 190 days in sheep [97,98], a variety of studies on electrode performance over time
[20,26,27,43,44,97,98,121,134], and one human clinical trial in progress [127].
Figure 1-16: Picture of the Stentrode in the collapsed (top), partially-expanded (center), and expanded (bottom) configuration
with platinum disc electrodes (yellow arrow) and the accompanying delivery catheter (green arrow). Scale bar 3 mm. Reprinted
from [26] with permission from Springer Nature.

18
The implantation of endovascular electrodes is minimally invasive and only requires a small
incision to access a blood vessel. An access sheath is inserted into the vessel, and catheters can be inserted
through the access sheath into the endovascular space. The blood vessels act as a natural highway for
reaching deep and superficial targets within the nervous system, all the while causing minimal tissue
injury or inflammatory response. After the delivery catheter has been advanced to the targeted vessel, the
device is delivered by advancing it through the catheter and out of the distal end of the catheter [135].
A detailed review of endovascular neural interfaces is available in an external publication [135].
1.2 Immune Response to Implanted Neural Devices
As a probe is inserted into brain tissue, it causes damage to tissue along and adjacent to the
insertion path, including capillaries, extracellular matrix, glial cells, and neurons, which triggers the
body’s wound healing response. Erythrocytes, platelets, and clotting factors are released and the
complement cascade is activated. This leads to the recruitment of macrophages to remove any excess
fluid or debris and aims to rebuild the damaged tissue. At the same time, cytokines and neurotoxic free
radicals are released in an attempt to degrade the probe. This collective immune activity creates a region
of increased pressure and fluid build-up around the implant which persists for 6-8 days [136,137]. The
magnitude of this response is impacted by several aspects of the probe design, such as the size, shape, and
whether or not it is tethered to the skull. In devices with larger probes, sharper edges (as are present in
some planar silicon probes), and/or a rigid connection to the skull, the immune response is more severe
[75,138]. In addition, the immune response is generally more severe if inserting a probe into a region with
large blood vessels (>5 µm diameter) as compared to a region without large vessels [139].
The overall magnitude of this acute immune response in turn impacts the magnitude of the
chronic response [137]. As such, a smaller, less invasive insertion technique should lead to reduced acute
and chronic inflammation. Additionally, studies have shown that, when probes are inserted and quickly
removed (a stab wound), brain tissue will heal in the absence of a foreign body and the probe tracks are
not visible via histology [137]. This suggests that the continued presence of an implant within the tissue
has a significant impact on the resulting tissue response.

19
After the initial 6-8 day acute tissue response, chronic effects begin to take place. These are in
part attributed to the large difference in mechanical properties between the probe and the brain tissue
during normal brain movement (such as occurs during breathing, blood pumping, or walking) with an
increased effect when the device is tethered to the skull [74,136,137]. The Young’s modulus (which
describes the mechanical stiffness) of brain tissue and most probe materials differ by many orders of
magnitude (see Figure 1-17). The presence of a foreign body activates the body’s immune response, and
the stiffness mismatch amplifies it. Initially, activated microglia, which modulate the immune response,
and astrocytes, which later form a glial scar, travel to the probe. When macrophages cannot degrade the
probe (during the acute response), the astrocytes are recruited to wall off the implant and a glial scar
forms around the probe. This glial sheath forms around 6-10 weeks after insertion and has a typical
thickness of 50-100 µm. The formation of the glial sheath increases the distance between the probe and
the target tissue, thus increasing the impedance of the electrode-tissue interface and leading to signal
degradation [136,137]. As is the case with the acute response, the chronic response is less severe when
using a probe with a small cross-section or a small Young’s modulus (both producing a more flexible
probe), a probe that is not rigidly attached to the skull, and/or a probe with soft (non-sharp) edges [75].
In addition to these acute and chronic responses, micromotion of the probe relative to the brain
(millimeter-scale motion in the brain caused by things such as walking, breathing, and pulsing of blood in
vessels) continuously aggravates the device-tissue interface, causing repeated injuries to the tissue and
leading to a continuous immune response that may result in a thicker glial scar [105,116]. The thickness
of the glial scar resulting from this micromotion is directly related to the mass of the implant (larger mass
yields a thicker scar) because heavier implants experience greater inertial forces during motion [140]. In
addition, the glial scar is more severe when using materials that are much harder than brain tissue
(materials with a higher Young’s modulus) [105,116]. These studies indicate that a lighter, softer, and
more flexible implant which more closely matches the mechanical properties of the tissue is preferable for
mitigating the immune response and enabling long term use.

20
1.3 Flexible Neural Interfaces
Neural interfaces research is evolving, with an increasingly high need for high resolution, chronic
recording and stimulation. Among existing neural recording and stimulation electrodes, flexible,
microfabricated devices are an ideal option to fill this need, as they can contain large numbers of
microelectrodes (high resolution) and are made of soft materials which more closely match the
mechanical properties of tissue and elicit a less severe immune response. Such devices are made possible
with bioMEMS technologies.
MEMS (microelectromechanical systems) is a technology which utilizes microfabrication
techniques to produce devices with features on the nanometer to millimeter scale with electronic and
moving parts, often used to produce simple and complex sensors used in consumer products (such as
electronics or automobiles). MEMS devices are produced on a planar substrate (most commonly silicon
wafers) using a variety of additive and subtractive processes. In additive processes, materials are
deposited on top of the substrate (or on top of other features) in a conformal coating. In subtractive
processes, materials are removed from the substrate using wet or dry etching processes. Each process is
patterned using lithography (usually photolithography), allowing complex geometries to be formed.
BioMEMS borrows manufacturing processes from the MEMS field to produce sensors and
devices with biological applications. Implantable bioMEMS devices often utilize flexible polymer
materials as a device backbone (rather than silicon) to interface more closely with complex tissue
geometries and decrease the body’s immune response to the device. Devices are built on top of a carrier
substrate (usually a silicon wafer) using the same additive and subtractive processes, then released from
the substrate to produce a freed thin film bioMEMS device. These flexible devices can naturally conform
to non-planar tissue or be bent and formed into 3D geometries to seamlessly interface with more complex
anatomy or to serve a functional purpose. In the context of neural recording and stimulation, bioMEMS
devices most commonly take the form of multisite penetrating microelectrodes, however other electrode
types (such as surface electrodes, regenerative electrodes, endovascular electrodes (chapter 3), and
extraneural electrodes (chapter 4)) can also be made using bioMEMS techniques.

21
Many biocompatible polymers are compatible with bioMEMS processes. Parylene C (poly
(chloro-p-xylylene)), polyimide, and polydimethylsiloxane (PDMS), the most commonly used flexible
bioMEMS materials, have moduli below approximately 10 GPa, which more closely matches the
modulus of human biological tissue than most traditional rigid probe materials (e.g. silicon, platinum, or
carbon fiber, all of which have moduli above 50 GPa – see Figure 1-17). This mechanical compatibility is
critical to reduce the immune response and achieve more successful chronic use. Other flexible materials
such as liquid crystal polymer (LCP), SU-8 photoresist, benzocyclobutene (BCB), and Parylene HT
(poly[(2,3,5,6-tetrafluoro-1,4-phenylene) (1,1,2,2-tetrafluoro-1,2-ethanediyl)]) have also been used as
probe backbone materials but remain less common.
Figure 1-17: Young's Moduli of common rigid and flexible MEMS materials and human tissues [37,141-143].
1.3.1 Implantation of Flexible BioMEMS Electrodes
Implantation strategies for neural interfaces are critical to ensure the device can be placed
accurately with minimal damage to the device or tissue. Rigid devices are usually strong enough to be
directly placed into or on tissue using standard surgical techniques. Flexible devices, however, usually
require different insertion strategies because they are not strong enough to support themselves during
insertion and are prone to buckling. For surface electrodes in the brain and spinal cord, flexible electrodes
can be placed directly on the tissue and will conform to any topography on the tissue surface. For other
electrode types, insertion tools or strategies have been developed and electrodes have been designed to

22
facilitate the desired implantation configuration. Some of the more common designs and strategies are
described in this section.
1.3.1.1 Extraneural Electrodes
The implantation of thin film extraneural electrodes which surround the nerve fiber (such as cuff
or helical electrodes) uses a similar method as is used for traditionally manufactured electrodes (exposing
the neve and wrapping the device around it). Many thin film extraneural electrodes are flat and require a
locking mechanism on the device itself (such as serrated teeth) or a suture to hold them in place [50,51].
Other electrode designs have a curled architecture, allowing them to self-close around the nerve without
the use of a locking mechanism [90,100]. A curled, self-sizing, extraneural electrode is described in
chapter 4.
Devices which are designed to interface with specialized nerves (such as retinal or cochlear
electrodes) are often shaped to match the topography of the nerve, facilitating easier implantation, and
sometimes have tabs or other features to interface with implantation tools [28,29].
1.3.1.2 Penetrating Microelectrodes
The majority of flexible microelectrodes in literature are multisite penetrating microelectrodes
which have a rectangular cross section (1 to 50 µm thick, 10 to 550 µm wide, and 0.7 to 15 mm long)
with a pointed tip on one end (for insertion into tissue) and an interface with recording or stimulation
electronics on the other end [11,12,37,38,71-86]. Implantation of flexible penetrating microelectrodes
begins in the same way as other microelectrodes (exposing the target tissue and aligning the probe
manually or with a motorized tool). Most polymer probes cannot penetrate into the tissue without
additional mechanical reinforcement of some kind because they are too flexible and will buckle before
they penetrate the brain tissue [37]. Once the probe is inserted within the brain, it must travel to the
targeted region. Accurate insertion of longer, thinner, and narrower probes can be challenging. The
majority of flexible probes have been developed for use in small animal models and the insertion depth
rarely exceeds 5 mm. Longer flexible probes suitable for accessing deeper brain regions or for use in
larger animals can only be successful if they are accompanied by methods or tools to achieve accurate

23
implantation. The mechanics governing insertion are examined below, followed by a discussion of
different methods to achieve implantation of flexible probes.
The insertion of the probe into the tissue is governed by the buckling equation (1-1) which
captures the axial stiffness of the probe and the geometric shape to calculate the required force to insert
the probe into the tissue. For a probe to be inserted, the insertion force must be greater than the puncture
force of the tissue (Figure 1-18A). If the insertion force is less than the buckling force, the probe will
successfully insert (Figure 1-18B). Conversely, if the insertion force is greater than the buckling force, the
probe will buckle and will not insert into the tissue (Figure 1-18C).
𝐹𝑏𝑢𝑐𝑘𝑙𝑖𝑛𝑔 =
𝜋
2𝐸𝐼
(𝐾𝐿)
2
(1-1)
𝐸 = Young’s modulus, 𝐼 = second moment of area, 𝐿 = unsupported length, 𝐾 = length factor
Figure 1-18: Insertion of a flexible probe into tissue. (A) Insertion force acts on the proximal end of the probe; puncture force
acts on the distal tip of the probe. For successful insertion, the insertion force must be greater than the puncture force. (B)
Successful insertion: insertion force is less than the buckling force; the probe overcomes the puncture force and inserts into the
tissue. (C) Failed insertion: insertion force is greater than the buckling force; the probe buckles before the insertion force can
overcome the puncture force. © IOP Publishing. Reproduced with permission from [37].
Several insertion strategies and probe designs have been developed to alter components of the
buckling equation, reducing the buckling force (summarized in Table 1-3). Note that any strategies which
increase E and I are likely to produce a more severe immune response to the implanted probe due to the
larger size of tissue displacement and mechanical mismatch.

24
Table 1-3: Summary of insertion methods for flexible penetrating microelectrodes, including how they impact each component of
the buckling equation (increase ↑, decrease ↓, or no impact ×), and common uses. Information sourced from [37].
Insertion Strategy Buckling Equation
Components Common Uses
E I K L
Unaided × × × × Short, wide, and/or stiff probes
Shuttle ↑ ↑ × ×
Deep insertion (>10 mm) requiring high
targeting accuracy; arrays of probes
Dissolvable Stiffener ↑ ↑ × ×
Medium depth insertion (5-10 mm);
arrays of probes
Dissolvable Brace × × × ↓
Medium depth insertion (5-10 mm);
small arrays of probes
Stiffness Changing
Coating/Backbone ↑ ↑ × × Medium depth insertion (5-10 mm)
Surface Guide × × ↓ ↓ Rarely used
Stiff Tipped ↑ × × ↓ Rarely used
Engineered Cross
Section × ↑ × × Rarely used
Magnetically Guided n/a n/a n/a n/a Rarely used
Injection ↑ ↑ × × Insertion of neural threads or meshes
If probes are sufficiently short, wide, or stiff, they can be inserted into the tissue without an
insertion aid. Longer and thinner probes are most often inserted by laminating a rigid insertion shuttle to
the probe (which greatly increases E and slightly increases I) or by coating the probe in a dissolvable
stiffener (which slightly increases E and greatly increases I). Another strategy is to add a dissolvable
brace which supports all but the tip of the probe (decreasing the effective length, L), then slowly
dissolving the brace as it is approaches the tissue surface to expose more of the probe to be inserted.
Several less common strategies have also been used for flexible probe insertion, such as including
a stiffness-changing material in the probe construction which is stiff during insertion and flexible after
implantation (which also often requires a larger cross section backbone), engineering the cross section
with struts or curves to increase the rigidity (I) while maintaining the use of soft materials, adding a
magnetic material to the tip of the probe to steer the probe into place with a magnet (bypassing the
constraints of the buckling equation), or injecting the probe through a needle (wherein only the needle
must be stiff enough to prevent buckling).
A detailed review of each of these insertion strategies is available in an external publication [37].

25
1.3.1.3 Endovascular Electrodes
Thin film endovascular electrodes must be permanently or temporarily attached to a guidewire (or
similar structure with sufficient stiffness and steering capacity) to be routed to the target tissue in the
same way that is done for other endovascular devices (accessing the blood vessel, navigating to the target
region, and deploying the device). If the device is permanently attached to the guidewire or supporting
structure, the wire/structure is secured in place, otherwise the device is released and the guidewire
removed.
1.3.2 Three-Dimensional Neural Interfaces
Traditionally, microfabrication is used to build flat devices due to the planar nature of the
fabrication method. Flat devices can be assembled into 3D structures via stacking or linkages, however
some medical applications require more complex three-dimensional (3D) geometries to seamlessly
interface with complex anatomy. Flexible bioMEMS devices can be transformed into 3D geometries (via
post-processing) to allow for close confirmation to the target tissue and to minimize the body’s immune
response to implanted devices.
Of the most common bioMEMS polymers (Parylene C, polyimide, and PDMS), Parylene C is
particularly useful because it is a thermoplastic polymer, which allows thin films to be transformed into a
new 3D configuration by thermoforming against a template [144,145]. PDMS and polyimide are
thermoset polymers which can be bent into complex shapes but cannot be softened and re-shaped after
curing in a planar configuration, and require attachment to a supporting structure [146-148] or plastic
deformation (which can damage the device) [149] to maintain a new, non-planar shape.
Thermoformed Parylene C has been used in numerous microfabricated medical devices
[28,29,100,144,150-152] with various geometries. Of note is the helix geometry, which can be used for
strain relief, to interface with anatomical features such as nerves, muscle fibers, or blood vessels, or can
be wrapped around cylindrical supports (such as catheters, stents, and probes) to produce 3D MEMS
devices. Thermoformed Parylene C helices are explored in detail in chapter 2, and the development of
two thin film, thermoformed Parylene C devices (a helical endovascular electrode for recording in the

26
brain and an extraneural cuff electrode for stimulation in the peripheral nerves) are described in chapters 3
and 4.
1.4 Objectives
Neural recording and stimulation can be used for a wide range of applications, ranging from
general research on bodily function to the treatment of highly targeted diseases. The advancement of the
bioMEMS field and flexible devices allows for more detailed, high resolution, chronic recording and
stimulation of neural tissue. This work aims to further advance the capabilities of flexible bioMEMS
devices by developing methods to transform planar devices into complex 3D structures. These 3D devices
maintain the complexity of traditional planar, microfabricated devices but allow them to more closely
interface with the non-planar anatomy found in the body.
This work first evaluates the use of film stress to produce naturally curled thin film bioMEMS
devices and thermoforming to produce detailed 3D geometries from microfabricated Parylene thin film
devices (chapter 2). This information was used to develop two different thermoformed, thin film Parylene
devices. The first (chapter 3) is a helical electrode array for endovascular recording of neural signals in
the brain. This device is a minimally invasive alternative to current neural recording devices, such as
ECoG grids or SEEG probes, which are commonly used in clinical settings but have a highly invasive
implantation procedure with many negative side effects. The second (chapter 4) is a cuff electrode for
stimulation of small diameter branched peripheral nerves. Most clinical cuff electrodes are aimed at
stimulating larger nerves (on the scale of 1-3 mm in diameter) which control a variety of bodily functions.
This smaller (sub-mm) cuff electrode will be capable of stimulating smaller, branched nerves that lie
closer to end organs, resulting in fewer undesired side effects of stimulation.
1.5 References
[1] Connolly A T, Muralidharan A, Hendrix C, Johnson L, Gupta R, Stanslaski S, Denison T, Baker
K B, Vitek J L and Johnson M D 2015 Local field potential recordings in a non-human primate
model of Parkinsons disease using the Activa PC + S neurostimulator J. Neural Eng. 12 066012
[2] Shanechi M M, Hu R C and Williams Z M 2014 A cortical-spinal prosthesis for targeted limb
movement in paralysed primate avatars Nat. Commun. 5

27
[3] Choi J, Kim S-M, Ryu R-H, Kim S-P and Sohn J 2018 Implantable Neural Probes for BrainMachine Interfaces – Current Developments and Future Prospects Exp. Neurobiol. 27 453–71
[4] Llinas R R, Walton K D, Nakao M, Hunter I and Anquetil P A 2005 Neuro-vascular central
nervous recording/stimulating system: Using nanotechnology probes J. Nanoparticle Res. 7 111–
27
[5] Tallgren P, Vanhatalo S, Kaila K and Voipio J 2005 Evaluation of commercially available
electrodes and gels for recording of slow EEG potentials Clin. Neurophysiol. 116 799–806
[6] Lee B, Zubair M N, Marquez Y D, Lee D M, Kalayjian L A, Heck C N and Liu C Y 2015 A
Single-Center Experience with the NeuroPace RNS System: A Review of Techniques and
Potential Problems World Neurosurg. 84 719–26
[7] D’Agostino E, Kanter J, Song Y and Aronson J P 2020 Stereoencephalography Electrode
Placement Accuracy and Utility Using a Frameless Insertion Platform Without a Rigid Cannula
Oper. Neurosurg. 18 409–16
[8] Nowell M, Rodionov R, Diehl B, Wehner T, Zombori G, Kinghorn J, Ourselin S, Duncan J,
Miserocchi A and McEvoy A 2014 A Novel Method for Implementation of Frameless StereoEEG
in Epilepsy Surgery Oper. Neurosurg. 10 525–33
[9] Grill W M, Norman S E and Bellamkonda R V. 2009 Implanted neural interfaces: Biochallenges
and engineered solutions Annu. Rev. Biomed. Eng. 11 1–24
[10] Dubey A and Ray S 2019 Cortical Electrocorticogram (Ecog) Is a Local Signal J. Neurosci. 39
4299–311
[11] Na K, Sperry Z J, Lu J, Vöröslakos M, Parizi S S, Bruns T M, Yoon E and Seymour J P 2020
Novel diamond shuttle to deliver flexible neural probe with reduced tissue compression
Microsystems Nanoeng. 6 37
[12] Ward M P, Rajdev P, Ellison C and Irazoqui P P 2009 Toward a comparison of microelectrodes
for acute and chronic recordings Brain Res. 1282 183–200
[13] Watanabe H, Takahashi H, Nakao M, Walton K and Llinás R R 2009 Intravascular neural
interface with nanowire electrode Electron. Commun. Japan 92 29–37
[14] Armstrong-James M and Millar J 1979 Carbon fibre microelectrodes J. Neurosci. Methods 1 279–
87
[15] Cogan S F 2008 Neural stimulation and recording electrodes Annu. Rev. Biomed. Eng. 10 275–
309
[16] Khodagholy D, Gelinas J N, Thesen T, Doyle W, Devinsky O, Malliaras G G and Buzsáki G 2015
NeuroGrid: recording action potentials from the surface of the brain Nat. Neurosci. 18 310–5
[17] Farrell S M, Green A and Aziz T 2018 The current state of deep brain stimulation for chronic pain
and its context in other forms of neuromodulation Brain Sci. 8
[18] Gildenberg P L 2006 History of Electrical Neuromodulation for Chronic Pain Pain Med. 7 S7–13

28
[19] Richardson D E and Akil H 1977 Pain reduction by electrical brain stimulation in man. II.
Chronic self administration in the periventricular gray matter J. Neurosurg. 47 184–94
[20] Opie N L, John S E, Rind G S, Ronayne S M, Wong Y T, Gerboni G, Yoo P E, Lovell T J H,
Scordas T C M, Wilson S L, Dornom A, Vale T, O’Brien T J, Grayden D B, May C N and Oxley
T J 2018 Focal stimulation of the sheep motor cortex with a chronically implanted minimally
invasive electrode array mounted on an endovascular stent Nat. Biomed. Eng. 2 907–14
[21] Burchiel K J, McCartney S, Lee A and Raslan A M 2013 Accuracy of deep brain stimulation
electrode placement using intraoperative computed tomography without microelectrode recording
J. Neurosurg. 119 301–6
[22] Opie N L and O’Brien T J 2022 The potential of closed‐loop endovascular neurostimulation as a
viable therapeutic approach for drug‐resistant epilepsy: A critical review Artif. Organs 46 337–48
[23] North R B, Kidd D H, Zahurak M, James C S and Long D M 1993 Spinal Cord Stimulation for
Chronic, Intractable Pain: Experience over Two Decades Clinical Study Neurosurgery 32 384–95
[24] Sonntag V K H and Cooper P R 2005 Spinal cord stimulation versus repeated lumbosacral spine
surgery for chronic pain: A randomized, controlled trial - Comments Neurosurgery 56 106–7
[25] Cameron T 2004 Safety and efficacy of spinal cord stimulation for the treatment of chronic pain:
a 20-year literature review J. Neurosurg. 100 254–67
[26] Oxley T J, Opie N L, John S E, Rind G S, Ronayne S M, Wheeler T L, Judy J W, McDonald A J,
Dornom A, Lovell T J H, Steward C, Garrett D J, Moffat B A, Lui E H, Yassi N, Campbell B C V,
Wong Y T, Fox K E, Nurse E S, Bennett I E, Bauquier S H, Liyanage K A, van der Nagel N R,
Perucca P, Ahnood A, Gill K P, Yan B, Churilov L, French C R, Desmond P M, Horne M K,
Kiers L, Prawer S, Davis S M, Burkitt A N, Mitchell P J, Grayden D B, May C N and O’Brien T J
2016 Minimally invasive endovascular stent-electrode array for high-fidelity, chronic recordings
of cortical neural activity Nat. Biotechnol. 34 320–7
[27] Gerboni G, John S E, Rind G S, Ronayne S M, May C N, Oxley T J, Grayden D B, Opie N L and
Wong Y T 2018 Visual evoked potentials determine chronic signal quality in a stent-electrode
endovascular neural interface Biomed. Phys. Eng. Express 4
[28] Johnson A C and Wise K D 2012 A self-curling monolithically-backed active high-density
cochlear electrode array IEEE 25th International Conference on Micro Electro Mechanical
Systems (MEMS) (IEEE) pp 914–7
[29] Yoon E, Koo B, Wong J, Elyahoodayan S, Weiland J D, Lee C D, Petrossians A and Meng E
2020 An implantable microelectrode array for chronic in vivo epiretinal stimulation of the rat
retina J. Micromechanics Microengineering 30 124001
[30] Ordonez J S, Boehler C, Schuettler M and Stieglitz T 2012 Improved polyimide thin-film
electrodes for neural implants Proc. Annu. Int. Conf. IEEE Eng. Med. Biol. Soc. EMBS 5134–7
[31] Ben‐Menachem E, Mañon‐Espaillat R, Ristanovic R, Wilder B J, Stefan H, Mirza W, Tarver W B
and Wernicke J F 1994 Vagus Nerve Stimulation for Treatment of Partial Seizures: 1. A
Controlled Study of Effect on Seizures Epilepsia 35 616–26

29
[32] Handforth A, DeGiorgio C M, Schachter S C, Uthman B M, Naritoku D K, Tecoma E S, Henry T
R, Collins S D, Vaughn B V., Gilmartin R C, Labar D R, Morris G L, Salinsky M C, Osorio I,
Ristanovic R K, Labiner D M, Jones J C, Murphy J V., Ney G C and Wheless J W 1998 Vagus
nerve stimulation therapy for partial-onset seizures: A randomized active-control trial Neurology
51 48–55
[33] Zurita F, Freko S, Hiendlmeier L, Duca F Del, Groll T, Seelbach O, Steiger K and Wolfrum B
2023 Thermoformed Parylene-C Cuff Electrodes for Small Nerve Interfacing Adv. NanoBiomed
Res. 2300102
[34] Barrese J C, Aceros J and Donoghue J P 2016 Scanning electron microscopy of chronically
implanted intracortical microelectrode arrays in non-human primates J. Neural Eng. 13 026003
[35] Maynard E M, Nordhausen C T and Normann R A 1997 The Utah Intracortical Electrode Array:
A recording structure for potential brain-computer interfaces Electroencephalogr. Clin.
Neurophysiol. 102 228–39
[36] Saal H P and Bensmaia S J 2015 Biomimetic approaches to bionic touch through a peripheral
nerve interface Neuropsychologia 79 344–53
[37] Thielen B and Meng E 2021 A comparison of insertion methods for surgical placement of
penetrating neural interfaces J. Neural Eng. 18 041003
[38] Schwartz A B, Cui X T, Weber D J and Moran D W 2006 Brain-Controlled Interfaces: Movement
Restoration with Neural Prosthetics Neuron 52 205–20
[39] Martini M L, Oermann E K, Opie N L, Panov F, Oxley T and Yaeger K 2020 Sensor Modalities
for Brain-Computer Interface Technology: A Comprehensive Literature Review Clin. Neurosurg.
86 E108–17
[40] Hollanders T 2018 Feasability Study of Intravascular EEG-Recording (Ghent University)
[41] Shirazi S Y and Huang H J 2019 More Reliable EEG Electrode Digitizing Methods Can Reduce
Source Estimation Uncertainty, but Current Methods Already Accurately Identify Brodmann
Areas Front. Neurosci. 13
[42] Stancin I, Cifrek M, Jovic A and Igor Stancin M C and A J 2021 A Review of EEG Signal
Features and Their Application in Driver Drowsiness Detection Systems Sensors 21 3786
[43] John S E, Apollo N V., Opie N L, Rind G S, Ronayne S M, May C N, Oxley T J and Grayden D
B 2019 In Vivo Impedance Characterization of Cortical Recording Electrodes Shows Dependence
on Electrode Location and Size IEEE Trans. Biomed. Eng. 66 675–81
[44] John S E, Opie N L, Wong Y T, Rind G S, Ronayne S M, Gerboni G, Bauquier S H, O’Brien T J,
May C N, Grayden D B and Oxley T J 2018 Signal quality of simultaneously recorded
endovascular, subdural and epidural signals are comparable Sci. Rep. 8 8427
[45] Fujimoto A, Matsumaru Y, Masuda Y, Marushima A, Hosoo H, Araki K and Ishikawa E 2022
Endovascular Electroencephalogram Records Simultaneous Subdural Electrode-Detectable, Scalp
Electrode-Undetectable Interictal Epileptiform Discharges Brain Sci. 12

30
[46] Cabello C M, Bair W B, Lamore S D, Ley S, Alexandra S, Azimian S and Wondrak G T 2010 A
Parylene-Based Microelectrode Array Implant for Spinal Cord Stimulation in Rats 46 220–31
[47] Bower M R, Stead M, Van Gompel J J, Bower R S, Sulc V, Asirvatham S J and Worrell G A
2013 Intravenous recording of intracranial, broadband EEG J. Neurosci. Methods 214 21–6
[48] Ad-Tech Medical Instrument Corporation 2021 Depth Electrodes adtechmedical.com/depthelectrodes
[49] Larson C E and Meng E 2020 A review for the peripheral nerve interface designer J. Neurosci.
Methods 332 108523
[50] Russell C, Roche A D and Chakrabarty S 2019 Peripheral nerve bionic interface: a review of
electrodes Int. J. Intell. Robot. Appl. 3 11–8
[51] Cobo A M, Larson C E, Scholten K, Miranda J A, Elyahoodayan S, Song D, Pikov V and Meng E
2019 Parylene-Based Cuff Electrode With Integrated Microfluidics for Peripheral Nerve
Recording, Stimulation, and Drug Delivery J. Microelectromechanical Syst. 28 36–49
[52] Cowley A W 2013 Helical electrode for nerve stimulation
[53] Teplitzky B A, Connolly A T, Bajwa J A and Johnson M D 2014 Computational modeling of an
endovascular approach to deep brain stimulation J. Neural Eng. 11
[54] Kahan J, Papadaki A, White M, Mancini L, Yousry T, Zrinzo L, Limousin P, Hariz M, Foltynie T
and Thornton J 2015 The Safety of Using Body-Transmit MRI in Patients with Implanted Deep
Brain Stimulation Devices PLoS One 10 e0129077
[55] NeuroPace, Inc. 2021 NeuroPace RNS System neuropace.com/the-rns-system/
[56] Rousche P J and Normann R A 1992 A method for pneumatically inserting an array of penetrating
electrodes into cortical tissue Ann. Biomed. Eng. 20 413–22
[57] Blackrock Microsystems 2018 Utah Array blackrockmicro.com/electrode-types/utah-array/
[58] Blackrock Microsystems 2018 Utah Rodent Array blackrockmicro.com/utah-rodent-array/
[59] Arafat M A, Rubin L N, Jefferys J G R and Irazoqui P P 2019 A Method of Flexible Micro-Wire
Electrode Insertion in Rodent for Chronic Neural Recording and a Device for Electrode Insertion
IEEE Trans. Neural Syst. Rehabil. Eng. 27 1724–31
[60] Musallam S, Bak M J, Troyk P R and Andersen R A 2007 A floating metal microelectrode array
for chronic implantation J. Neurosci. Methods 160 122–7
[61] ChunXiang T and Jiping H 2005 Monitoring insertion force and electrode impedance during
implantation of microwire electrodes 2005 International Conference of the IEEE Engineering in
Medicine and Biology Society, Shanghai (IEEE) pp 7333–6
[62] Nevrotech 2017 Microwire Arrays - MMA nevrotech.com/products/multichannelarrays/microwire-arrays/
[63] Bamford J A, Todd K G and Mushahwar V K 2010 The Effects of Intraspinal Microstimulation

31
on Spinal Cord Tissue in the Rat Biomaterials 31 5552–63
[64] Lourenço C F, Caetano M, Ledo A and Barbosa R M 2019 Platinized carbon fiber-based glucose
microbiosensor designed for metabolic studies in brain slices Bioelectrochemistry 130 107325
[65] Koh D S and Hille B 1999 Rapid fabrication of plastic-insulated carbon-fiber electrodes for
micro-amperometry J. Neurosci. Methods 88 83–91
[66] Nicolai E N, Michelson N J, Settell M L, Hara S A, Trevathan J K, Asp A J, Stocking K C, Lujan
J L, Kozai T D Y and Ludwig K A 2018 Design Choices for Next-Generation Neurotechnology
Can Impact Motion Artifact in Electrophysiological and Fast-Scan Cyclic Voltammetry
Measurements Micromachines 9 494
[67] Patel P R, Zhang H, Robbins M T, Nofar J B, Marshall S P, Kobylarek M J, Kozai T D Y, Kotov
N A and Chestek C A 2016 Chronic in vivo stability assessment of carbon fiber microelectrode
arrays J. Neural Eng. 13 066002
[68] Patel P R, Na K, Zhang H, Kozai T D Y, Kotov N A, Yoon E and Chestek C A 2015 Insertion of
linear 8.4 μm diameter 16 channel carbon fiber electrode arrays for single unit recordings J.
Neural Eng. 12 046009
[69] Peters J L, Miner L H, Michael A C and Sesack S R 2004 Ultrastructure at carbon fiber
microelectrode implantation sites after acute voltammetric measurements in the striatum of
anesthetized rats J. Neurosci. Methods 137 9–23
[70] Malagodi M S, Horch K W and Schoenberg A A 1989 An intrafascicular electrode for recording
of action potentials in peripheral nerves Ann. Biomed. Eng. 17 397–410
[71] NeuroNexus 2021 Probe Finder neuronexus.com/products/probe-finder/
[72] Scott K M, Du J, Lester H A and Masmanidis S C 2012 Variability of acute extracellular action
potential measurements with multisite silicon probes J. Neurosci. Methods 211 22–30
[73] Barz F, Livi A, Lanzilotto M, Maranesi M, Bonini L, Paul O and Ruther P 2017 Versatile,
modular 3D microelectrode arrays for neuronal ensemble recordings: From design to fabrication,
assembly, and functional validation in non-human primates J. Neural Eng. 14 036010
[74] Biran R, Martin D C and Tresco P A 2007 The brain tissue response to implanted silicon
microelectrode arrays is increased when the device is tethered to the skull J. Biomed. Mater. Res.
Part A 82A 169–78
[75] Karumbaiah L, Saxena T, Carlson D, Patil K, Patkar R, Gaupp E A, Betancur M, Stanley G B,
Carin L and Bellamkonda R V. 2013 Relationship between intracortical electrode design and
chronic recording function Biomaterials 34 8061–74
[76] Inácio A R, Nasretdinov A, Lebedeva J and Khazipov R 2016 Sensory feedback synchronizes
motor and sensory neuronal networks in the neonatal rat spinal cord Nat. Commun. 7 13060
[77] Cheung K C 2007 Implantable microscale neural interfaces Biomed. Microdevices 9 923–38
[78] Wang X, Hirschberg A W, Xu H, Slingsby-Smith Z, Lecomte A, Scholten K, Song D and Meng E

32
2020 A Parylene Neural Probe Array for Multi-Region Deep Brain Recordings J.
Microelectromechanical Syst. 29 499–513
[79] Kato Y, Saito I, Hoshino T, Suzuki T and Mabuchi K 2006 Preliminary Study of Multichannel
Flexible Neural Probes Coated with Hybrid Biodegradable Polymer 2006 International
Conference of the IEEE Engineering in Medicine and Biology Society, New York, NY (IEEE) pp
660–3
[80] Chen C H, Chuang S C, Lee Y T, Yeh S R, Chang Y C and Yao D J 2010 Three-dimensional
flexible microprobe for recording the neural signal J. Micro/Nanolithography MEMS MOEMS 9
031007
[81] Lo M, Wang S, Singh S, Damodaran V B, Kaplan H M, Kohn J, Shreiber D I and Zahn J D 2015
Coating flexible probes with an ultra fast degrading polymer to aid in tissue insertion Biomed.
Microdevices 17 34
[82] Kil D, Carmona M B, Ceyssens F, Deprez M, Brancato L, Nuttin B, Balschun D and Puers R
2019 Dextran as a resorbable coating material for flexible neural probes Micromachines 10 61
[83] Xiang Z, Yen S-C, Xue N, Sun T, Tsang W M, Zhang S, Liao L-D, Thakor N V. and Lee C 2014
Ultra-thin flexible polyimide neural probe embedded in a dissolvable maltose-coated microneedle
J. Micromechanics Microengineering 24 065015
[84] Wu F, Im M and Yoon E 2011 A flexible fish-bone-shaped neural probe strengthened by
biodegradable silk coating for enhanced biocompatibility 2011 16th International Solid-State
Sensors, Actuators and Microsystems Conference, Beijing (IEEE) pp 966–9
[85] Luan L, Wei X, Zhao Z, Siegel J J, Potnis O, Tuppen C A, Lin S, Kazmi S, Fowler R A,
Holloway S, Dunn A K, Chitwood R A and Xie C 2017 Ultraflexible nanoelectronic probes form
reliable, glial scar–free neural integration Sci. Adv. 3 e1601966
[86] Richter A, Xie Y, Schumacher A, Löeffler S, Kirch R D, Al-Hasani J, Rapoport D H, Kruse C,
Moser A, Tronnier V, Danner S and Hofmann U G 2013 A simple implantation method for
flexible, multisite microelectrodes into rat brains Front. Neuroeng. 6 6
[87] Seo D, Neely R M, Shen K, Singhal U, Alon E, Rabaey J M, Carmena J M and Maharbiz M M
2016 Wireless Recording in the Peripheral Nervous System with Ultrasonic Neural Dust Neuron
91 529–39
[88] Seo D, Carmena J M, Rabaey J M, Maharbiz M M and Alon E 2015 Model validation of
untethered, ultrasonic neural dust motes for cortical recording J. Neurosci. Methods 244 114–22
[89] Rabaey J M 2011 Brain-machine interfaces as the new frontier in extreme miniaturization 2011
Proceedings of the European Solid-State Device Research Conference (ESSDERC), Helsinki
(IEEE) pp 19–24
[90] Stieglitz T, Beutel H, Schuettler M and Meyer J U 2000 Micromachined, polyimide-based devices
for flexible neural interfaces Biomed. Microdevices 2 283–94
[91] Penn R D, Hilal S K, Michelsen W J, Goldensohn E S and Driller J 1973 Intravascular
intracranial EEG recording J. Neurosurg. 38 239–43

33
[92] Nakase H, Ohnishi H, Touho H, Karasawa J, Yamamoto S and Shimizu K 1995 An Intra-Arterial
Electrode for Intracranial Electro-Encephalogram Recordings Acta Neurochir. (Wien). 136 103–5
[93] Stoeter P, Dieterle L, Meyer A and Prey N 1995 Intracranial Electroencephalographic and
Evoked-Potential Recording from Intravascular Guide Wires Am. J. Neuroradiol. 16 1214–7
[94] He B D, Ebrahimi M, Palafox L and Srinivasan L 2016 Signal quality of endovascular
electroencephalography J. Neural Eng. 13 16016
[95] Thomke F, Stoeter P and Stader D 1998 Endovascular electroencephalography during an
intracarotid amobarbital test with simultaneous recordings from 16 electrodes J. Neurol.
Neurosurg. Psychiatry 64 565
[96] Ruddy B P 2006 Conducting polymer wires for intravascular neural recording (Massachusetts
Institute of Technology)
[97] Opie N L, John S E, Rind G S, Ronayne S M, Grayden D B, Burkitt A N, May C N, O’Brien T J
and Oxley T J 2016 Chronic impedance spectroscopy of an endovascular stent-electrode array J.
Neural Eng. 13
[98] Opie N L, Rind G S, John S E, Ronayne S M, Grayden D B, Burkitt A N, May C N, O’Brien T J
and Oxley T J 2016 Feasibility of a chronic, minimally invasive endovascular neural interface
Proceedings of the Annual International Conference of the IEEE Engineering in Medicine and
Biology Society, EMBS (IEEE) pp 4455–8
[99] Gerboni G 2019 Stent-based Endovascular Brain Interface: Characterisation of Recording and
Stimulating Properties (University of Melbourne)
[100] Kang X, Liu J Q, Tian H, Yang B, Nuli Y and Yang C 2015 Self-closed parylene cuff electrode
for peripheral nerve recording J. Microelectromechanical Syst. 24 319–32
[101] Wheless J W, Gienapp A J and Ryvlin P 2018 Vagus nerve stimulation (VNS) therapy update
Epilepsy Behav. 88 2–10
[102] Ji B, Chen L, Wang M, Guo Z, Zhou Y, Huangfu S, Zhang K, Feng H, Chang H and Liu J 2021
Thermal Strain-Induced Self-Rolling Mesh Cuff Electrodes for Non-Linear Peripheral Nerve IEEE
34th International Conference on Micro Electro Mechanical Systems (MEMS) pp 567–70
[103] Viswam V, Obien M E J, Franke F, Frey U and Hierlemann A 2019 Optimal Electrode Size for
Multi-Scale Extracellular-Potential Recording From Neuronal Assemblies Front. Neurosci. 13 1–
23
[104] Memarian N, Madsen S K, Macey P M, Fried I, Engel J, Thompson P M and Staba R J 2015 Ictal
depth EEG and MRI structural evidence for two different epileptogenic networks in mesial
temporal lobe epilepsy PLoS One 10 1–16
[105] Groothuis J, Ramsey N F, Ramakers G M J and van der Plasse G 2014 Physiological challenges
for intracortical electrodes Brain Stimul. 7 1–6
[106] Elwassif M M, Datta A, Rahman A and Bikson M 2012 Temperature control at DBS electrodes
using a heat sink: experimentally validated FEM model of DBS lead architecture J. Neural Eng. 9

34
046009
[107] Sinnamon H M and Schwartzbaum J S 1973 Dorsal hippocampal unit and EEG responses to
rewarding and aversive brain stimulation in rats Brain Res. 56 183–202
[108] Misra A, Burke J F, Ramayya A G, Jacobs J, Sperling M R, Moxon K A, Kahana M J, Evans J J
and Sharan A D 2014 Methods for implantation of micro-wire bundles and optimization of
single/multi-unit recordings from human mesial temporal lobe J. Neural Eng. 11 026013
[109] Humphrey D R and Schmidt E M 1990 Extracellular Single-Unit Recording Methods
Neurophysiological Techniques ed A A Boulton, G B Baker and C H Vanderwolf (Humana Press)
[110] Lozano A M, Gildenberg P L and Tasker R R 2009 Textbook of Stereotactic and Functional
Neurosurgery (Springer Science & Business Media)
[111] Barrese J C, Rao N, Paroo K, Triebwasser C, Vargas-Irwin C, Franquemont L and Donoghue J P
2016 Failure mode analysis of silicon-based intracortical microelectrode arrays in non-human
primates J. Neural Eng. 10 066014
[112] Leber M, Shandhi M M H, Hogan A, Solzbacher F, Bhandari R and Negi S 2016 Different
methods to alter surface morphology of high aspect ratio structures Appl. Surf. Sci. 365 180–90
[113] Tucker-Davis Technologies 2021 Electrodes tdt.com/components/electrodes/
[114] Jiman A A, Ratze D C, Welle E J, Patel P R, Richie J M, Bottorff E C, Seymour J P, Chestek C A
and Bruns T M 2020 Multi-channel intraneural vagus nerve recordings with a novel high-density
carbon fiber microelectrode array Sci. Rep. 10 15501
[115] Xu H, Hirschberg A W, Scholten K, Berger T W, Song D and Meng E 2018 Acute in vivo testing
of a conformal polymer microelectrode array for multi-region hippocampal recordings J. Neural
Eng. 15 016017
[116] Sohal H S, Jackson A, Jackson R, Clowry G J, Vassilevski K, O’Neill A and Baker S N 2014 The
sinusoidal probe: A new approach to improve electrode longevity Front. Neuroeng. 7 10
[117] Mikuni N, Ikeda T, Murao K, Terada T, Nakahara I, Taki W, Kikuchi H and Shibasaki T 1997
“Cavernous Sinus EEG”: A New Method for the Preoperative Evaluation of Temporal Lobe
Epilepsy Epilepsia 38 472–82
[118] Boniface S J and Antoun N 1997 Endovascular electroencephalography: the technique and its
application during carotid amytal assessment J. Neurol. Neurosurg. Psychiatry 62 193–5
[119] García Asensio S, Guelbenzu S, Barrena R and Valero P 1999 Technical Aspects of Intra-arterial
Electroencephalogram Recording Interv. Neuroradiol. 5 289–300
[120] García Asensio S, Guelbenzu S, Uclés P, Barrena M R and Valero P 1999 The Role of
Intravascular EEG Riv. di Neuroradiol. 12 139–42
[121] Forsyth I A, Dunston M, Lombardi G, Rind G S, Ronayne S, Wong Y T, May C N, Grayden D B,
Oxley T, Opie N and John S E 2019 Evaluation of a minimally invasive endovascular neural
interface for decoding motor activity International IEEE/EMBS Conference on Neural

35
Engineering, NER (IEEE) pp 750–3
[122] Gaba BS P and DeSimone C V. 2016 Novel Techniques in Epilepsy Management: Venous Pacing
and Capture of Electrical Activity in the Primate Cortex J. Neurol. Neurophysiol. 7
[123] Reuter B M, Brassel F, Linke D B, Hartmann A and Kurthen M 1989 Tiefen-EEG und evozierte
Potentiale mittels intravaskulärer Katheter-Elektrode beim Pavian EEG. EMG. Z.
Elektroenzephalogr. Elektromyogr. Verwandte Geb. 20 81–3
[124] Kunieda T, Ikeda A, Ohara S, Sadato A, Taki W and Shibasaki H 2000 Use of Cavernous Sinus
EEG in the Detection of Seizure Onset and Spread in Mesial Temporal Lobe Epilepsy Posterior
Tra Anterior Jugular vein Epilepsia 41 1411–9
[125] Enomoto T, Matsumura A, Matsumaru Y, Aoki K and Nose T 2002 Multimodality examination
for detection of the epileptic foci with special reference to endovascular EEG recording Int.
Congr. Ser. 1232 655–60
[126] Liu J Y, Grayden D B, Keast J R and John S E 2021 Computational Modeling of an Endovascular
Peripheral Nerve Interface Proceedings of the Annual International Conference of the IEEE
Engineering in Medicine and Biology Society, EMBS pp 5966–9
[127] Oxley T J, Yoo P E, Rind G S, Ronayne S M, Lee C M S, Bird C, Hampshire V, Sharma R P,
Morokoff A, Williams D L, Macisaac C, Howard M E, Irving L, Vrljic I, Williams C, John S E,
Weissenborn F, Dazenko M, Balabanski A H, Friedenberg D, Burkitt A N, Wong Y T, Drummond
K J, Desmond P, Weber D, Denison T, Hochberg L R, Mathers S, Brien T J O, May C N, Mocco
J, Grayden D B, Campbell B C V, Mitchell P and Opie N L 2021 Motor neuroprosthesis implanted
with neurointerventional surgery improves capacity for activities of daily living tasks in severe
paralysis : first in- - human experience J. Neurointerv. Surg. 13 102–8
[128] Ishida S, Shimamoto H, Abe T, Mootka H, Sakurai S, Takamaya K, Yamaguchi H and Maeda H
1998 Intracranial EEG Recording from Intravascular Electrodes in Patients with Temporal Lobe
Epilepsy Epilepsia 39 1998
[129] Henz B D, Friedman P A, Bruce C J, Holmes D R, Okumura Y, Johnson S B, Packer D L and
Asirvatham S J 2008 Successful radiofrequency ablation of the cerebral cortex in pigs using the
venous system: Possible implications for treating CNS disorders Epilepsy Res. 80 213–8
[130] Henz B D, Friedman P A, Bruce C J, Holmes D R, Bower M, Madhavan M, DeSimone C V.,
Wahnschaffe D, Berhow S, Danielsen A J, Ladewig D J, Mikell S B, Johnson S B, Suddendorf S
H, Kara T, Worrell G A and Asirvatham S J 2014 Advances in radiofrequency ablation of the
cerebral cortex in primates using the venous system: Improvements for treating epilepsy with
catheter ablation technology Epilepsy Res. 108 1026–31
[131] Ammerman J, Caputy A and Potolicchio S 2005 Endovascular ablation of a temporal lobe
epileptogenic focus – a complication of Wada testing Acta Neurol. Scand. 112 189–91
[132] Nakase H, Ohnishi H, Touho H, Furuoka N, Yamada K, Takaoka M, Senoh M, Komatsu T,
Karasawa J and Yamamoto S 1993 [Spike discharge detected by intra-arterial
electroencephalography from intra-arterial guide wire in temporal lobe epilepsy] No to Shineki =
Brain Nerve 45 973–7

36
[133] Opie N L, Oxley T J, Rind G S, Ronayne S M, John S E, May C N and Grayden D B 2019
Medical Device for Sensing and or Stimulating Tissue
[134] Opie N L, Van Der Nagel N R, John S E, Vessey K, Rind G S, Ronayne S M, Fletcher E L, May
C N, Obrien T J and Oxley T J 2017 Micro-CT and Histological Evaluation of an Neural Interface
Implanted Within a Blood Vessel IEEE Trans. Biomed. Eng. 64 928–34
[135] Thielen B, Xu H, Fujii T, Rangwala S D, Jiang W and Lin M 2023 Making a case for
endovascular approaches for neural recording and stimulation J. Neural Eng. 20 011001
[136] Polikov V S, Tresco P A and Reichert W M 2005 Response of brain tissue to chronically
implanted neural electrodes J. Neurosci. Methods 148 1–18
[137] Kook G, Lee S W, Lee H C, Cho I-J and Lee H J 2016 Neural probes for chronic applications
Micromachines 7 179
[138] Kozai T D Y, Gugel Z, Li X, Gilgunn P J, Khilwani R, Ozdoganlar O B, Fedder G K, Weber D J
and Cui X T 2014 Chronic tissue response to carboxymethyl cellulose based dissolvable insertion
needle for ultra-small neural probes Biomaterials 35 9255–68
[139] Kozai T D Y, Marzullo T C, Hooi F, Langhals N B, Majewska A K, Brown E B and Kipke D R
2010 Reduction of neurovascular damage resulting from microelectrode insertion into the cerebral
cortex using in vivo two-photon mapping J. Neural Eng. 7 046011
[140] Lind G, Linsmeier C E and Schouenborg J 2013 The density difference between tissue and neural
probes is a key factor for glial scarring Sci. Rep. 3 2942
[141] Singh G and Chanda A 2021 Mechanical properties of whole-body soft human tissues: A review
Biomed. Mater. 16
[142] Green M A, Bilston L E and Sinkus R 2008 In vivo brain viscoelastic properties measured by
magnetic resonance elastography NMR Biomed. 21 755–64
[143] Paggi V, Akouissi O, Micera S and Lacour S P 2021 Compliant peripheral nerve interfaces J.
Neural Eng. 18 031001
[144] Kim B J, Chen B, Gupta M and Meng E 2014 Formation of three-dimensional Parylene C
structures via thermoforming J. Micromechanics Microengineering 24 065003
[145] Kim B J, Chen B, Gupta M and Meng E 2013 Three Dimensional Transformation of Parylene
Thin Film Structures via Thermoforming IEEE 26th International Conference on Micro Electro
Mechanical Systems (MEMS) (IEEE) pp 339–42
[146] Li C, Wu P M, Prince S J, Wu Z, Chakraborti S, Ahn C H, Hartings J A and Narayan R K 2013
Multifunctional lab-on-a-tube (LOT) probe for simultaneous neurochemical and
electrophysiological activity measurements Transducers & Eurosensors XXVII: The 17th
International Conference on Solid-State Sensors, Actuators and Microsystems (TRANSDUCERS
& EUROSENSORS XXVII) pp 880–3
[147] Li C, Wu P M, Han J and Ahn C H 2008 A flexible polymer tube lab-chip integrated with
microsensors for smart microcatheter Biomed. Microdevices 10 671–9

37
[148] Li C and Narayan R K 2012 Development of a novel catheter for early diagnosis of bacterial
meningitis caused by the ventricular drain IEEE 25th International Conference on Micro Electro
Mechanical Systems (MEMS) (IEEE) pp 120–3
[149] Guo Z, Ji B, Wang M, Wang X, Yang B and Liu J 2018 A polyimide-based 3-D ultrathin
bioelectrode with elastic sites for neural recording J. Microelectromechanical Syst. 27 1035–40
[150] Huang R and Tai Y-C 2010 Flexible parylene-based 3-D coiled cable IEEE 5th Int. Conf.
Nano/Micro Eng. Mol. Syst. 317–20
[151] Hara S A, Kim B J, Kuo J T W, Lee C D, Meng E and Pikov V 2016 Long-term stability of
intracortical recordings using perforated and arrayed Parylene sheath electrodes J. Neural Eng. 13
066020
[152] Kim B J, Kuo J T W, Hara S A, Lee C D, Yu L, Gutierrez C A, Hoang T Q, Pikov V and Meng E
2013 3D Parylene sheath neural probe for chronic recordings J. Neural Eng. 10 045002

38
Chapter 2. Thermoforming of Parylene C
Thermoforming is a process in which thermoplastic polymers can be softened and semipermanently formed into a new shape. When a thermoplastic is heated above its glass transition
temperature, the material softens but remains in a solid state, allowing the amorphous regions of the
polymer to reorganize. As the material is cooled back below the glass transition temperature, it rehardens. If the material is fixtured into a new shape prior to heating or while it is at an elevated
temperature, it will hold its fixtured shape after hardening. The process is considered semi-permanent
because the device will hold its thermoformed shape until it is heated back above the glass transition
temperature again, at which point it will re-soften and can assume a new shape.
This process is useful in the bioMEMS field because it can transform complex, microfabricated
devices into 3D shapes that more seamlessly interface with the body. For neural recording and stimulation
devices, the quality of recording and stimulation efficiency increase as the distance between the electrode
and the target tissue decreases [1-3]. Numerous thermoformed Parylene C-based devices have been
developed to achieve more intimate contact with the target tissue for a variety of applications, such as
cylindrical cuff electrodes and spherical or cylindrical retinal arrays. Thermoforming can also enable the
use of non-planar features in microfabricated devices, such as a helical cable for strain relief or a conical
probe tip to interface with an insertion stylet. Several examples of thermoformed Parylene C devices for
various applications are included in Table 2-1 and Figure 2-1. The most common (and possibly most
useful) shape in literature is the helix, which can be used for strain relief or to wrap around or insert into
cylindrical structures in the body (such as nerves, muscle fibers, or blood vessels) or external devices
(such as catheters, stents, probes, or wires).

39
Table 2-1: Examples of microfabricated Parylene medical devices that were transformed into 3D geometries via thermoforming.
Their resulting shape and size are included.
Device Curvature Shape Curvature Diameter
This work Helix 0.25-2.6 mm
Cable strain relief [4,5] Helix 1.1-5 mm
Cuff electrode [6,7] Cylinder 1-3 mm
Retinal electrode array [4,8,9] Sphere section 5 mm
Retinal electrode array [10] Cylinder section 1 mm
Penetrating cortical electrode
array [4,9,11]
Cylinder or cone (between
Parylene layers)
0.25 mm*
* 0.25 mm cylinders and cones in this study exhibited cracking in the Parylene and
metal layers, rendering the device non-functional
Figure 2-1: Examples of microfabricated Parylene medical devices that were transformed into common 3D geometries via
thermoforming: (A) helices (reprinted from [5] with permission from IEEE), (B) cylinders (reprinted from [7] with permission
from IEEE), (C) spheres (© IOP Publishing. Reproduced with permission from [8]), and (D) cones (© IOP Publishing.
Reproduced with permission from [4]).
To understand the limits of thermoforming bare Parylene C and Parylene C devices with
patterned metal layers, the capabilities of the material must be evaluated. Two important material
considerations are the natural curvature of metalized Parylene C devices due to stress mismatches
between layers in the device and the mechanical limits of bending Parylene C and patterned metal
features without damaging the Parylene or metal. These two considerations were evaluated

40
mathematically and experimentally to evaluate the design space for thermoformed Parylene and to
develop a process for producing 0.25 mm Parylene helices with functional patterned metal.
A portion of content from this chapter has been published [12] and is adapted with permission
under CC BY 4.0. © IOP Publishing. This chapter includes updates developed after publication of [12].
2.1 Background
2.1.1 Parylene C
Parylene C is semicrystalline, having both crystalline and amorphous regions. As the material is
heated above the glass transition temperature (approximately 60-90 °C [4,11,13]), the amorphous regions
will soften, allowing reorganization of the polymer chains while still in a solid state. Although Parylene is
commonly used as a coating or insulation layer in electronics, this chapter discusses Parylene-based thin
film devices microfabricated in a planar, layered formation. When used with metal layers, Parylene
served as a structural backbone and electrical insulation.
2.1.2 Film Stress
The deposition of any thin film on a substrate produces residual stresses in the film. Film stress
can either be tensile (positive), pulling the substrate into a concave “U” shape, or compressive (negative),
pushing the substrate into a convex shape and sometimes causing buckling or wrinkling in the thin film.
Metal thin films (which are commonly deposited via e-beam evaporation, sputtering, or
electroplating) generally have high stress; however, stress can be modulated by varying deposition
methods or process parameters. In most cases, sputtering produces a film with higher tensile stress than ebeam evaporation or electroplating (~185 MPa, ~95 MPa, and ~10 MPa, respectively, for 700-800 nm
gold thin films on silicon substrates [14,15]). Stress in sputtered and evaporated films can be modulated
by altering the pressure, power, and deposition rate (all of which impact deposition temperature, with
higher temperatures leading to higher (more tensile) stress) [14-19]. Electroplated metal films, while
lower stress, are usually much thicker than sputtered or evaporated films (in the range of several 100s of
nm to a few µm) and require a conductive seed layer for deposition. In polymer-based thin film devices, a
conductive surface would need to be deposited by evaporation or sputtering, negating the benefits of

41
lower-stress electroplating [14,15]. Thermal annealing of metal films after deposition has also been
shown to increase the film stress (more tensile). In gold films, this effect is seen at temperatures of 50-100
°C or higher [14,15], whereas platinum films require significantly higher temperatures to alter film stress
(~550-600 °C or higher) [17-19] which exceeds the processing temperature ranges for most thin film
polymers.
Parylene C films have lower residual stress than metal films (in the range of -6 MPa
(compressive) to 0.3 MPa (tensile) as deposited [10,13]) but are generally 10-100 times thicker (10s of
microns versus 100s of nanometers). Thermal annealing (heating above the glass transition temperature)
of Parylene films allows amorphous regions of the polymer reorganize, causing increased crystallinity and
shrinkage of the polymer, usually in the range of 1-3% [4,7,13,20-24]. These physical changes in the
Parylene result in increased stress (on the order of ~10-50 MPa), with higher annealing temperatures
resulting in higher (more tensile) stress [13,24,25].
Although it has not been widely studied, metal films deposited on top of Parylene C films often
have high compressive stress due to the thermal expansion mismatch between the materials [26]. During
deposition, the metal source is heated above its melting point and transferred to the substrate, causing
localized heating of the substrate at the surface where it is deposited. The coefficients of thermal
expansion (CTEs) of commonly used metals are higher than that of silicon wafers (14.2 ppm for gold and
9 ppm for platinum, 3-5 ppm for silicon [27]), resulting in a higher degree of shrinkage in the metal films
as compared to the silicon substrate when cooling back to room temperature, resulting in a tensile stress in
the metal film [14]. The opposite is true when depositing a metal on top of Parylene; the CTEs of
common metals are lower than that of Parylene C (35 ppm [28]), resulting in a higher degree of shrinkage
in the Parylene layer and a compressive stress in the metal film.
2.1.3 Thin Film Devices
In thin film medical devices, multiple layers of polymer and metal films are stacked to produce
devices with exposed electrodes (to interface with the tissue) and bondpads (to connect to stimulators or
recording equipment) and insulated traces (to connect the bondpads to the electrodes), such as the device

42
shown in Figure 2-2. This stacking of films, often with asymmetric thicknesses or etched openings,
results in unbalanced film stress between layers. When devices are released from the substrate, this can
lead to curvature in the device towards the layer with the highest tensile stress (as compressive layers
expand and tensile layers shrink). For dual Parylene layer devices, this curling effect is increased when
asymmetric Parylene layer thicknesses are used or in areas where Parylene has been etched away (such as
exposed metal areas for electrode sites or bondpads), as the stress between Parylene layers is not balanced
and the high-stress metal layer is no longer on or near the neutral axis (the longitudinal axis in which the
net stress is zero), resulting in curling towards the more tensile layer [7,26,29]. This curvature becomes
more severe after annealing due to Parylene shrinkage (and thus stress increase) at high temperatures (up
to ~250 °C) [10,26]. Because shrinkage of metal layers is generally observed at much higher
temperatures, metal layers can buckle or wrinkle during Parylene shrinkage, resulting in changes in the
patterned metal geometry. Curling can also be more pronounced when thin polymer layers are used, as
they cannot overcome forces imposed by the high-stress metal layer(s).
Figure 2-2: Diagram of a Parylene-metal-Parylene device configuration, with openings etched in the top Parylene to form
electrodes (section A-A’), insulated traces (section B-B’), and openings etched in the base Parylene to form bondpads (section CC’). © IOP Publishing. Reproduced with permission from [12] (CC BY 4.0).

43
Two processes utilize the stress and physical property changes of Parylene above the glass
transition temperature for functional purposes. Annealing is commonly used to increase crosslinking
between sequentially deposited, adjacent Parylene layers (preventing delamination), increase crystallinity
(decreasing water ingress through the material), and flatten parts which have a slight curvature when
taken off the wafer. Parts are clamped between two flat surfaces, heated above the glass transition
temperature while under vacuum (to prevent oxidative degradation), held for a predetermined time (in
most cases, between 100 and 300 °C for 1-48 hours), and cooled back to room temperature.
Thermoforming is a similar process that is used to transform parts into a 3D shape. Parts are fixtured into
the desired shape (rather than clamping flat), heated above the glass transition temperature, and cooled
back to room temperature, at which point the polymer re-hardens and retains the fixtured shape
[4,7,11,21,22,26,30-32].
Although thermoplastic Parylene can be annealed or thermoformed to hold a desired shape (flat
or 3D), these processes can result in unexpected curvature or shape if the film stresses in a multi-layer
(Parylene and metal) device are too high or not balanced in favor of the final shape. This work evaluates
the impact of annealing and thermoforming process parameters on the resulting shape of bare Parylene
and Parylene-metal-Parylene devices. Recommendations to manage stress by varying Parylene layer
thickness and annealing parameters to achieve the desired structure are discussed.
2.2 Modeling Stress and Curvature
To study the natural curvature of Parylene-metal-Parylene (PMP) devices, the film stress in each
layer of the device must be considered. Stress in thin films has been extensively studied and modeled,
with most studies focusing on a single material deposited on a rigid surface [33,34]. Although this
configuration differs from that of thin-film biomedical devices, which consist of multi-layer, freed thin
films, similar analysis can be used to mathematically model such devices. In addition, film stress
measurements on rigid substrates can be used to predict the curvature of freed, multi-layer films.
Although the devices used in this study (and in most thin film biomedical devices) have complex
geometries (patterned Parylene and metal layers), a simplified model structure can be used to predict the

44
behavior of thin film devices using several fundamental equations describing the stress and curvature in
composite structures. Such a model can be used to predict the curvature of fabricated devices if other
device parameters (dimensions and stress) are known.
Figure 2-3 shows a simplified thin film device consisting of a base Parylene layer, a metal layer,
and a top Parylene layer (known as a Parylene-metal-Parylene, or PMP, device), modeled as a composite
I-beam. This model simplifies the trace region of the device shown in Figure 2-2 section B-B’ by
combining the individual traces into one metal film (which should not impact the device curvature about
the Z axis if the overall metal width and thickness are identical) and ignores any effects due to contact
between the base and top Parylene layers. The model can also be used to simulate the electrode or
bondpad regions by altering the width of the top or base layers to account for the smaller amount of
Parylene in that area.

45
Figure 2-3: (A) An isometric view and (B) a cross sectional view of the simplified model geometry, with a single metal strip
sandwiched between two Parylene layers. (C) The side view shows a small section of the device with the neutral axis of each
layer and the full device. © IOP Publishing. Reproduced with permission from [12] (CC BY 4.0).
Figure 2-4A shows a small section of the device and the stresses and forces acting along one edge
of that section before any stresses have been allowed to balance (i.e., before the device has been released
from the wafer). In the case shown in the figure, the base Parylene layer has a moderate tensile stress
(𝜎𝑏𝑎𝑠𝑒), the middle metal (platinum) layer has a large compressive stress (𝜎𝑚𝑒𝑡𝑎𝑙), and the top Parylene
layer has a small compressive stress (𝜎𝑡𝑜𝑝). The stresses in each layer can be measured or estimated from
literature (common stresses in deposited Parylene and platinum films found in literature are included later
in this section). These stresses can be converted into forces using the relationship
𝐹𝑖 = 𝜎𝑖𝐴𝑖
(2-1)
where 𝐹𝑖
is the force and 𝐴𝑖
is the cross-sectional area (width, 𝑤𝑖
times thickness, 𝑡𝑖
), and the subscript 𝑖
represents each layer (i.e. base Parylene, metal, or top Parylene). When the device is released from the
wafer, the device will either compress or expand to equalize the forces in each layer (Figure 2-4B shows
the device having expanded slightly due to the net compressive force in Figure 2-4A), resulting in a net
force of zero. This is described mathematically as
∑ 𝐹𝑖
′ = 0 = 𝐹𝑏𝑎𝑠𝑒
′ + 𝐹𝑚𝑒𝑡𝑎𝑙
′ + 𝐹𝑡𝑜𝑝
′
(2-2)
= 𝜎𝑏𝑎𝑠𝑒
′ 𝐴𝑏𝑎𝑠𝑒 + 𝜎𝑚𝑒𝑡𝑎𝑙
′ 𝐴𝑚𝑒𝑡𝑎𝑙 + 𝜎𝑡𝑜𝑝
′ 𝐴𝑡𝑜𝑝
where 𝐹𝑖
′
and 𝜎𝑖
′
are the forces and stresses in each layer after shrinkage or expansion. Because each layer
is bonded to the adjacent layer, the overall strain in each layer is equal. Using this relationship, the strain
before and after expansion in each layer are described as
𝜀 =
𝜎𝑏𝑎𝑠𝑒
′ −𝜎𝑏𝑎𝑠𝑒
𝐸𝑃𝑎
=
𝜎𝑚𝑒𝑡𝑎𝑙
′ −𝜎𝑚𝑒𝑡𝑎𝑙
𝐸𝑚𝑒𝑡𝑎𝑙
=
𝜎𝑡𝑜𝑝
′ −𝜎𝑡𝑜𝑝
𝐸𝑃𝑎
(2-3)
where 𝜀 is the strain and 𝐸𝑘 is the Young’s modulus of the material for that layer (note that the
relationship in equation 2-3 assumes no delamination between films, resulting in equal strain in every
layer; Pa is Parylene). Using equations 2-2 and 2-3, 𝜎𝑖
′
for each layer can be calculated and converted into

46
𝐹𝑖
′
using equation 2-1. Although the forces in the device are now balanced, the forces are not acting at the
neutral axis of the device, resulting in a net moment. The net moment in the device is
𝑀 = ∑ 𝑀𝑖 = ∑ 𝐹𝑖
′
(𝑦𝑖 − 𝑦) (2-4)
where 𝑀 is the net moment in the device, 𝑀𝑖
is the moment in each layer, 𝑦𝑖
is the neutral axis of each
layer, and 𝑦 is the neutral axis of the device. The resulting moment (due to the expanded, balanced forces)
is what causes the device to curl. The bending moment equation describes the radius as
𝑟 =
𝐸𝐼
𝑀
(2-5)
where 𝑟 is the radius of curvature and 𝐼 is the second moment of area of the device. Note that, in order to
use this equation, the equivalent area method must be used when calculating 𝐼 to represent the full device
as a single material with Young’s modulus 𝐸. The second moment of area for each component about the
device neutral axis, 𝐼𝑖
, is
𝐼𝑖 =
𝑤𝑖𝑡𝑖
3
12
+ 𝐴𝑖
(𝑦𝑖 − 𝑦)
2
(2-6)
which can be used to calculate the device second moment of area, 𝐼, as
𝐼 = 𝐼𝑏𝑎𝑠𝑒 +
𝐸𝑝𝑡
𝐸𝑃𝑎
𝐼𝑝𝑡 + 𝐼𝑡𝑜𝑝 (2-7)
which represents the entire device as being made from Parylene using the equivalent area method.
The released, curled device has an adjusted stress gradient in each layer (due to the shrinkage or
expansion and curvature of the device), resulting in both net zero force and moment about the neutral axis
(Figure 2-4C).

47
Figure 2-4: Illustration of the stresses and forces acting on a section of the device (side view, as shown in Figure 2-3C). (A) As
deposited: each film layer is deposited on a substrate with residual stress, producing a part with stresses leading to unbalanced
force and moment (about the neutral axis). (B) Shrinkage or expansion only: When removed from the substrate, the part shrinks
or expands, adjusting the stresses in each layer to balance the forces in the device. (C) Shrinkage/expansion and bending: After
shrinking or expanding, the device curls to balance the moment about the neutral axis, resulting in a stress gradient in each layer.
Note: the metal layer illustrated here is not to scale, so the high stress leads to a small force, and the stress changes significantly
in each step. © IOP Publishing. Reproduced with permission from [12] (CC BY 4.0).
Each of the testing conditions used in this study for PMP device annealing (described in section
2.3.6) were evaluated using the model with equal dimensions to the experimental device geometries
(width and thickness of each layer). Deposited film stress and Young’s modulus for each material were
found in literature or determined by fitting experimental data to the model.
The stress in Parylene films has been shown to increase with higher annealing temperature. While
all published data referenced for this study follows this trend, exact stress values over the given
temperature range (20-200 °C) differ between publications [13,24,25]. In addition, films in these studies
were annealed in air while attached to a substrate. When exposed to oxygen at high temperatures,
Parylene will oxidize, impacting the resulting mechanical properties of the material [4,22]). In this study,
PMP device annealing was performed after devices had been released from their silicon carrier wafer and

48
clamped between two flat surfaces (as described in section 3.3) in a nitrogen-purged vacuum oven. The
difference in annealing fixturing likely resulted in different stress changes and resulting device shape
because the devices can shift slightly during the annealing process.
No data has been published on the stress in platinum films deposited via e-beam evaporation on
Parylene, however qualitative descriptions estimate it to be less compressive than sputtered platinum on
Parylene (with a value of -511 MPa) as evidenced by less wrinkling in the metal layer and less curvature
in released devices [26,29]. Modeled Parylene and platinum stress were selected by sweeping the
Parylene stress within the bounds of the published data and the platinum stress from -511 to 0 MPa and
selecting values which most accurately matched the results of the experimental results described in
section 2.4.3. This resulted in Parylene stress of -3.46 MPa at 20 °C (as deposited) linearly increasing
0.16 MPa/°C and platinum stress of -100 MPa. No published work has described the changes in Parylene
stress with varying annealing time, so the variable annealing time condition evaluated in this study
(section 2.3.6.3) could not be modeled.
Several different fabrication conditions were evaluated using the model (described in the
remainder of this section) and experimentally (described in section 2.4.3). The modeled trends in PMP
device shape under different annealing conditions and device dimensions support the experimental results
and can be used to motivate future PMP device designs. Based on the availability of stress data for
Parylene, not all experimental conditions were modeled, and some conditions which were modeled differ
slightly from experimental data (but follow a similar trend).
2.2.1 Base Layer Annealing
Annealing the base Parylene layer changes its film stress, but not that of the top Parylene or metal
layers. To model this scenario, all parameters were held constant (with dimensions matching the
thin/symmetric PMP devices) except for the stress in the base Parylene, which was varied based on the
data described previously in this section. Figure 2-5 shows the modeled device curvature diameter versus
base layer annealing temperature for a device with dimensions matching the thin/symmetric group (4.4
µm base, 4.7 µm top Parylene).

49
When the base layer is not annealed (represented as a 20 °C, or room temperature, anneal), the
magnitude of the diameter is large, representing a very flat device. This is due to equal stress and similar
thickness in the base and top Parylene layers, resulting in stress balanced about the neutral axis. As the
base annealing temperature increases, the stress in the base layer increases, resulting in a bending moment
which curls the device towards the base layer. The diameter tapers off at higher temperatures, with more
significant changes occurring at lower temperatures and more gradual changes occurring as temperature
increases above the glass transition temperature.
It is important to note that the modeled conditions do not consider increased temperatures that the
devices may be subjected to during fabrication processes, such as soft baking steps during photoresist
processing or increased temperature in a metal e-beam or sputtering chamber. Exposure to increased
temperature during processing can also result in increased Parylene stress, altering the results shown here.
Figure 2-5: Modeled curvature diameter with variable base layer annealing temperature. Inset axes are a zoomed in view of the
red dotted box, showing detail for small diameter curvature. The diameter at 200 °C is -1.2 mm. All constants were selected from
literature or from experimental PMP device dimensions (tbase = 4.4 µm, ttop = 4.7 µm, tmetal = 215 nm, wbase = 210 µm,
wtop = 210 µm, wtrace = 5 µm, ntraces = 8, σtop = -3.46 MPa, σmetal = -100 MPa). © IOP Publishing. Reproduced with permission from
[12] (CC BY 4.0).
2.2.2 Asymmetric Parylene Layers
Parylene layers of different thickness generally have the same film stress (assuming identical
processing conditions for each layer), resulting in larger forces in thicker layers due to the increased
cross-sectional area (described by equation 2-1). To model a PMP device with asymmetric Parylene

50
layers, the total device thickness was held constant and the ratio of top to bottom layer thickness was
varied.
Figure 2-6 shows the modeled device curvature diameter versus top to bottom layer thickness
ratio for total device thicknesses of 15 and 9 µm and unannealed and annealed (200 °C) conditions. All
datasets follow the same trend, with unannealed asymmetric devices (Figure 2-6, solid lines) curling
towards the thinner Parylene layer, annealed asymmetric devices (Figure 2-6, dashed lines) curling
towards the thicker Parylene layer (both to a greater extent with higher asymmetry), and symmetric
devices remaining flat (large diameter). Devices with less total thickness (Figure 2-6, green lines) are
more sensitive to asymmetry (resulting in higher curvature) due to the lower forces in the Parylene layer
(proportional to cross-sectional area per equation 2-1) and equal forces in the metal layer. Similarly,
annealed devices (Figure 2-6, dashed lines) are more sensitive to asymmetry (resulting in higher
curvature) due to the higher tensile stress in the Parylene layers contributing to the bending moment.
Figure 2-6: Modeled curvature diameter with variable ratio of top to bottom Parylene thickness, with total Parylene thickness of
15 µm (blue) and 9 µm (green), and unannealed (solid) and annealed at 200 °C (dashed) conditions. The diameter at 1:8 and 8:1
ratios is ±50.5, ±25.4, ∓3.5, and ∓1.7 mm for 15 µm unannealed, 9 µm unannealed, 15 µm annealed, and 9 µm annealed parts,
respectively. All constants were selected from literature or from experimental PMP device dimensions (tmetal = 215 nm,
wbase = 350 µm, wtop = 350 µm, wtrace = 10 µm, ntraces = 16, σbase = -3.46 MPa (unannealed) or 25.6 MPa (annealed) σtop = -3.46 MPa
(unannealed) or 25.6 MPa (annealed), σmetal = -100 MPa). © IOP Publishing. Reproduced with permission from [12] (CC BY
4.0).
2.2.3 Annealing Temperature
Annealing the full device results in film stress increase in both Parylene layers (changes in the
platinum film stress are assumed to be negligible in this temperature range), with higher annealing

51
temperatures yielding greater stress increases. A PMP device with dimensions matching thick/asymmetric
devices was modeled with base and top Parylene stress varying based on the data described previously in
this section. Figure 2-7 shows the modeled device curvature diameter versus annealing temperature. At
low temperatures, the stress in the Parylene is compressive, resulting in a slight curvature towards the
base layer. As temperature increases above ~31 °C, the Parylene stress becomes tensile, producing
curvature towards the top layer. As temperature increases further, increased tensile stress in the Parylene
layers results in increased bending moment (which is more pronounced due to the asymmetry of the
Parylene – see section 2.4.3.2), resulting in smaller curvature diameter.
Figure 2-7: Modeled curvature diameter with variable annealing temperature. The diameter at 200 °C is 4.2 mm, and the
crossover from negative to positive diameter (when the device is flat) occurs at approximately 31 °C. All constants were selected
from literature or from experimental PMP device dimensions (tbase = 3.4 µm, ttop = 11.5 µm, tmetal = 215 nm, wbase = 350 µm,
wtop = 350 µm, wtrace = 10 µm, ntraces = 16, σmetal = -100 MPa). © IOP Publishing. Reproduced with permission from [12] (CC BY
4.0).
2.2.4 Modeling Custom Geometries
Using this model and relevant datasets, custom geometries can be modeled to estimate the
curvature for any multi-layer material stackup. By programming the model into MATLAB, various
device and processing parameters can be quickly evaluated to determine their combined impact on device
shape. This model was used to select design and processing parameters for endovascular recording and
peripheral nerve stimulation electrodes (described in chapters 3 and 4, respectively). Details for
estimating curvature of devices using MATLAB are included in appendix A.

52
2.3 Experimental Methods
Although the mathematical model described in the section above provides some insight into ways
to naturally curl PMP devices by altering film stress, thermoforming is still necessary to produce most
complex geometries that are useful for bioMEMS devices. The following sections detail experiments
which evaluated the mechanical limits of thermoforming bare Parylene C and PMP devices and validated
the findings of the mathematical model.
2.3.1 Fabrication of Bare Parylene Strips
Bare Parylene strips (300 µm width, 20 mm length, varying thickness; Table 2-2) were fabricated
to evaluate thermoforming capability and cracking failure. Parylene C was deposited via a chemical vapor
deposition-like process (PDS 2010 Labcoter, Specialty Coating Systems, Indianapolis, IN) onto 4” prime
silicon wafers. Parylene thickness was varied from 5.4 to 20.9 µm by altering the amount of Parylene
dimer (Specialty Coating Systems, Indianapolis, IN) loaded into the machine. After breaking vacuum and
removing the samples, an optional second Parylene C layer was added on top of the first layer in some
samples (samples with a top layer thickness listed in Table 2-2) using the same procedure within 48 hours
of depositing the first layer. Parylene films were removed from the wafer by cutting with a scalpel and
peeling the cut region off the wafer. The film was then cut into strips by mounting on an adhesive mat and
trimming to the final shape using a cutting plotter (Graphtec CE6000-40, Irvine, CA).
Table 2-2: Parylene C thicknesses for bare Parylene strip samples. Samples with no top layer thickness listed were constructed of
a single Parylene layer. © IOP Publishing. Reproduced with permission from [12] (CC BY 4.0).
Parylene Thickness (µm)
Base Layer Top Layer Total
5.6 n/a 5.6
9.9 n/a 9.9
5.5 5.6 11.1
13.3 n/a 13.3
5.3 13.2 18.5
20.9 n/a 20.9
5.4 20.4 25.8
14.4 11.9 26.3

53
2.3.2 Fabrication of Parylene-Metal-Parylene Devices
Parylene-metal-Parylene (PMP) electrode arrays were fabricated to evaluate device curling as a
result of residual stress and test thermoforming parameters on realistic device configurations. The PMP
devices consisted of one metal layer (platinum) sandwiched between two Parylene C layers, with
openings in the Parylene C to expose electrode sites on the frontside of the device on one end and
bondpads on the backside of the device on the opposite end. A diagram of a representative PMP device is
shown in Figure 2-2, and the fabrication process is summarized in Figure 2-8.
Figure 2-8: Cross-sectional view of Parylene-metal-Parylene device fabrication process flow for (left) single-sided and (right)
double-sided devices. (1) Parylene was deposited on top of a bare silicon wafer (single-sided) or a sacrificial metal layer (15 nm
titanium + 100 nm aluminum). (2) Backside bondpads were opened via O2 etch (double-sided only). (3) Metal (see Table 2-3 for
metal stackup) electrodes, traces, and bondpads were deposited and patterned. (4) Parylene was deposited on top of patterned
metal and frontside bondpads (single-sided only) and electrodes were opened via O2 etch. (5) The devices were released using
water (single-sided) or aluminum etchant (double-sided). © IOP Publishing. Adapted with permission from [12] (CC BY 4.0).
Three groups of devices with varying dimensions were fabricated (thick/asymmetric,
thin/symmetric, and thin/asymmetric devices – dimensions for each group are listed in Table 2-3) to
evaluate different design parameters such as the total thickness, the impact of asymmetric Parylene layers,
and the impact of pre-annealing the base Parylene layer. A minimum layer thickness of ~3 µm was
chosen to prevent device damage due to handling, and a maximum total thickness of ~15 µm was chosen

54
to maintain device flexibility and to prevent excessive etching times. The thick/asymmetric and
thin/symmetric groups were fabricated using double-sided fabrication methods and were subjected to the
full battery of annealing and thermoforming tests (described in sections 2.3.6 and 2.3.7). After analyzing
test results, the thin/asymmetric group was fabricated using single-sided fabrication and optimized
parameters, after which only the thermoforming tests (section 2.3.7) were repeated.
Table 2-3: Dimensions for Parylene-metal-Parylene electrode arrays.
Group
Thick/Asymmetric Thin/Symmetric Thin/Asymmetric
Parylene
Thickness
(µm)
Base Layer 3.4 4.4 4.6
Top Layer 11.5 4.7 3.8
Total 14.9 9.1 8.4
Device Width (µm) 350 210 21
Device Length (µm) 20+501 40 40
Number of Traces 16 8 8
Trace Width (µm) 5-10 5 5
Single- or Double-Sided2 Double Double Single
Base Parylene Anneal3 No Yes Yes
Metal Stackup 15 nm Titanium +
200 nm Platinum
15 nm Titanium +
200 nm Platinum
20 nm Titanium +
25 nm Platinum +
150 nm Gold +
25 nm Platinum
Silane Treatment No No Yes
Cutout Method Etch Etch Laser
1 Thick/Asymmetric devices were “L” shaped, with one 20 mm arm and one 50 mm arm
2
Single-sided devices only had etched openings in the top Parylene layer, double-sided devices had
etched openings in the top and base Parylene layers (fabrication protocols in appendices D and E)
3
In the indicated parts, the base Parylene was annealed at 150 °C for 4 hours before adding metal or
top Parylene layers
PMP devices were fabricated using a low temperature, batch process based on prior work [35].
The fabrication process for the thick/asymmetric and thin/symmetric groups are summarized here and
described in detail in appendix B.
4” prime silicon wafers were coated with a 10 nm titanium adhesion layer followed by 100 nm of
aluminum, which was chemically roughened by etching for 8 minutes in CR-7 etchant (Transene,
Danvers, MA) to promote Parylene C adhesion to the surface. The Parylene C base layer was deposited
using a chemical vapor deposition-like process (PDS 2010 Labcoter, Specialty Coating Systems,

55
Indianapolis, IN) to the thickness defined in Table 2-3. Openings in the base Parylene for metal bondpads
on the backside of the device were etched using O2 reactive ion etching (PlasmaPro 80 RIE, Oxford
Instruments, Bristol, UK; 150 mT, 150 W, 50 sccm O2; etch rate approximately 0.2 µm/min) masked by
patterned photoresist (AZ P4620, AZ Electronic Materials, Branchburg, NJ; 12 µm thick; reflowed at 110
°C for 20 seconds to produce angled sidewalls). After etching, photoresist was removed using acetone
followed by rinsing in isopropyl alcohol and deioinized water. For the thin/symmetric device group only,
the coated wafers were baked (annealed) in an oven (TVO-2, Cascade Tek Inc., Longmont, CO) under
vacuum with nitrogen flow at 150 °C for 4 hours.
Next, lithography was performed to add a lift-off photoresist mask (AZ 5214 IR, AZ Electronic
Materials, Branchburg, NJ; 1.8 µm thick). Then, the titanium plus platinum metal stackup (see Table 2-3)
was deposited via e-beam evaporation. Electrodes, traces, and bondpads were formed following lift-off in
40 °C acetone or N-methylpyrrolidone (AZ NMP Rinse, AZ Electronic Materials, Branchburg, NJ)
followed by rinsing in isopropyl alcohol and deioinized water. After metal patterning, a top layer of
Parylene C was deposited to the thickness defined in Table 2-3. Openings in the Parylene for metal
electrodes on the frontside of the device were etched using O2 reactive ion etching (150 mT, 150 W, 50
sccm O2; etch rate approximately 0.2 µm/min) or O2 switched chemistry etching in a deep reactive ion
etcher [36] (PlasmaPro 80 or PlasmaLab 100 ICP, Oxford Instruments, Bristol, UK; switched chemistry
process parameters detailed in [36]; etch rate approximately 0.08 µm/loop) masked by patterned
photoresist (AZ P4620; 8-12 µm thick), and the outline of the device was cut out using the same etching
procedure. Photoresist was removed using acetone after each etching step followed by rinsing in
isopropyl alcohol and deioinized water. Devices were released from the wafer by dissolving the
aluminum adhesion layer in AZ MIF 726 (AZ Electronic Materials, Branchburg, NJ) at 60 °C.
The final group of PMP devices (thin/asymmetric) were fabricated using optimized parameters
and a slightly altered fabrication process. Changes to the fabrication process are summarized here and in
Table 2-3, and the full process is described in detail in appendix C.

56
For the thin/asymmetric group, the first Parylene layer was deposited directly onto a 4” prime
silicon wafer (i.e. no titanium + aluminum layer was used), and no openings were etched into the base
Parylene layer. The base Parylene anneal (4 hours at 150 °C) was performed. An altered metal stackup
containing a core gold layer (see Table 2-3) for additional flexibility was used, and wafers were treated
with an adhesion promoter prior to deposition of the top Parylene layer. The adhesion promoter solution
(1 part A-174 silane, 100 parts isopropyl alcohol, and 100 parts deionized water) was prepared 2.5 to 24
hours in advance, then wafers were submerged in the solution for 30 minutes, followed by 30 minutes of
air drying, rinsing with isopropyl alcohol, and blowing dry. After etching openings for the electrodes in
the top Parylene, bondpads were etched open in the top Parylene layer using the same procedure. The
outline of the device was cut out using a femtosecond laser (WS-Multi Head, Optec, Frameries, Belgium;
Pharos PH2-15W (Yb:KGW), Light Conversion, Vilnius, Lithuania) with 515 nm wavelength, 15 kHz
frequency, 0.75 W power, and 5 mm/s cutting speed. Devices were released from the wafer manually with
tweezers, using water to aid in the release of the device from the wafer.
2.3.3 Thermoforming and Annealing
After cutting bare Parylene strips to shape or releasing PMP devices from the wafer, the parts
were annealed flat or thermoformed into helices to evaluate the effects of heat treatment on device shape
and failure modes. The thermoforming and annealing processes are summarized here and described in
detail in appendix D.
To fixture parts for annealing, the parts were placed between two Teflon sheets (0.03 mm thick)
and clamped flat between two glass slides using clips.
To fixture parts for thermoforming, the parts were wrapped around a stainless steel mandrel of the
desired diameter into a helical shape by hand using a template to define the helix angle (see Figure 2-9)
and held in place using Teflon film (thick/asymmetric and thin/symmetric groups) or using a fixture
(thin/asymmetric group). See appendix D for details.

57
Figure 2-9: Illustration of the helix angle – the angle between the axis of the mandrel and the long edge of the Parylene strip or
device (helix angles of 15°, 30°, and 45° were used). © IOP Publishing. Reproduced with permission from [12] (CC BY 4.0).
After fixturing, the parts were placed into a programmable vacuum oven (TVO-2, Cascade Tek
Inc., Longmont, CO), placed under vacuum, then purged three times with nitrogen to minimize oxygen in
the chamber. The oven was programmed to ramp up to a defined temperature at a ramp rate of
approximately 0.7 °C/min, annealed for a defined time, then ramp down to room temperature. After
cooling, parts were removed from the fixture and inspected.
2.3.4 Annealing of Bare Parylene Strips
Bare Parylene strips (Table 2-2) were annealed flat following the procedure listed in section 2.3.3
with a thermoforming temperature of 200 °C and hold time of 12 hours to determine if bare Parylene
films (single or double layer) experienced any curling due to film stress without a metal interposing layer.
After annealing, the Parylene strips were visually inspected and photographed using a microscope
(HD60T, Caltex Scientific, Irvine, CA).
2.3.5 Thermoforming of Bare Parylene Strips
Bare Parylene strips (Table 2-2) were also thermoformed into helices (2.6, 1.6, and 0.25 mm helix
diameters; 15°, 30°, and 45° helix angles) following the procedure listed in section 2.3.3 with a
thermoforming temperature of 200 °C and hold time of 12 hours. Select samples were repeated at
thermoforming temperatures of 150 and 100 °C. Double-layer Parylene strips were tested in both winding
directions (i.e. two samples were run, one with the base layer towards the inside of the helix, and one with
the top layer towards the inside of the helix) to evaluate the effects of uneven Parylene layers.

58
After thermoforming, the Parylene helices were visually inspected using a microscope (HD60T,
Caltex Scientific, Irvine, CA and Eclipse LV100, Nikon, Tokyo, Japan) for their ability to retain the
desired shape and for any defects (cracking) in the Parylene.
2.3.6 Annealing of Parylene-Metal-Parylene Devices
PMP devices (Table 2-3) were annealed flat following the procedure listed in section 2.3.3 with
variable temperature and hold time. Due to the high number of variables (temperature, hold time,
Parylene layer thicknesses, base layer annealing), two representative groups were chosen to qualitatively
evaluate several variables experimentally while corroborating with published literature (thick/asymmetric
and thin/symmetric; thin/asymmetric devices were not included in the annealing tests).
For all parts, the curvature of the PMP devices before and after any annealing treatment was
measured by photographing the parts using a microscope (HD60T, Caltex Scientific, Irvine, CA) and
measuring the radius of curvature using edge detection in MATLAB.
2.3.6.1 Effects of Base Layer Annealing
Devices from each sample group were inspected prior to full device annealing to evaluate the
effects of base layer annealing (which was done for thin/symmetric samples, but not thick/asymmetric
samples). Although devices from each group did not have equal thickness or symmetry, results were
supported in comparison to similar processes found in literature and by the mathematical model described
in section 2.2.
2.3.6.2 Effects of Asymmetric Parylene Layers
Devices from each sample group were annealed (full annealing experiments described in sections
2.3.6.3 and 2.3.6.4) to evaluate the impact of asymmetric Parylene layers (present in the thick/asymmetric
group, but not the thin/symmetric group). In addition, the bondpad and electrode regions (with local areas
of Parylene removed in the base and top layers, respectively) were evaluated separately to determine the
effects of exposed metal on resulting device curvature. Although devices from each group did not have
equal thickness and the thin/symmetric group had a base layer annealing step, results were supported by
results from similar processes found in literature and by the mathematical model described in section 2.2.

59
2.3.6.3 Effects of Annealing Time
Devices from each sample group were annealed for 0.5, 6, 12, 24, or 48 hours at an annealing
temperature of 200 °C. The curvature diameters of devices before and after annealing were compared to
determine the effects of annealing time on Parylene shrinkage (and thus overall device curvature) and to
determine the amount of time necessary to reach the maximum shrinkage. A 200 °C annealing
temperature was chosen as it is a common annealing temperature used for Parylene in other published
devices and experiments [11,22,26,30,31,37].
2.3.6.4 Effects of Annealing Temperature
Devices from each sample group were annealed for 12 hours at an annealing temperature of 100,
150, or 200 °C. The curvature diameters of devices before and after annealing were compared to
determine the effects of annealing temperature on Parylene shrinkage (and thus overall device curvature)
and magnitude of shrinkage at each temperature. A 12 hour annealing time was selected as it was found to
be sufficient to achieve maximum shrinkage in the annealing time experiments.
2.3.7 Thermoforming of Parylene-Metal-Parylene Devices
Based on results from the bare Parylene strip thermoforming tests and PMP device annealing
tests, PMP devices were thermoformed using both the thermoforming parameters used in the bare
Parylene thermoforming experiments (200 °C for 12 hours) and optimized parameters determined from
PMP annealing experiments. For each sample group, thermoforming temperature and time were selected
which yielded devices closest to the target curvature in the PMP device annealing experiments, keeping
layer direction in mind (i.e. if devices were curled towards the base layer after annealing, the base layer
was wound to the inside of the helix; if devices were curled towards the top layer after annealing, the top
layer was wound to the inside of the helix). Several sets of parameters (low and high temperatures) were
tested for each sample group.
After thermoforming, the PMP devices were visually inspected using a microscope (HD60T,
Caltex Scientific, Irvine, CA and Eclipse LV100, Nikon, Tokyo, Japan) for their ability to retain the
desired shape and for cracking in the Parylene. PMP devices were also electrically tested for continuity

60
between the bondpads and electrodes using an LCR meter (E4980A, Agilent Technologies, Santa Clara,
CA) before and after thermoforming. The LCR meter was used to measure the impedance between each
bondpad/electrode pair using a 10 kHz, 20 mV signal. Traces were considered continuous if the
impedance magnitude was less than 100 kΩ and phase was greater than -65°.
2.4 Experimental Results
2.4.1 Annealing Tests: Bare Parylene Strips
Annealing bare Parylene strips resulted in flat parts (matching the flat, fixtured shape) for all
Parylene thicknesses tested (5.6 to 26.3 µm). The parts had relatively low stress prior to annealing due to
the low stress of deposited Parylene and absence of a high-stress metal layer. Parts were not expected to
curl as the curling is a result of mismatched stress and shrinkage in parts with multiple thin film layers.
Neither single- nor double-layer parts had mismatched stress because all layers in these parts were made
of the same material and had similar stress and shrinkage, resulting in a flat part.
2.4.2 Thermoforming Tests: Bare Parylene Strips
When fixturing bare Parylene strips into helices and thermoforming, no cracking was observed
for any Parylene thickness tested when thermoforming to a 1.6 or 2.6 mm helix diameter at any helix
angle (example of a bare Parylene helix with no cracking is shown Figure 2-10A). At 0.25 mm diameter,
helices with thinner Parylene (≤ 11.1 µm total thickness) exhibited minimal cracking (Figure 2-10B;
partial-thickness cracking) or no cracking. Thicker films (≥ 13.3 µm total thickness) exhibited cracking
(Figure 2-10C; through the full thickness of the Parylene) at the same diameter due to the increased
bending stress experienced by thicker samples. Cracking in the Parylene appeared after fixturing around
the mandrel and prior to thermoforming, indicating that the cracks were formed during the fixturing
process due to the stress induced during bending.
This result differs slightly from the theoretical maximum Parylene thickness for a given bending
diameter, which can be calculated using the definition of maximum bending stress,
𝜎𝑏𝑒𝑛𝑑,𝑚𝑎𝑥 =
𝑀(
𝑡
2
)
𝐼
(2-8)

61
and the relationship between moment and bending radius (equation 2-5). The bending strength of
Parylene is not readily available, so the yield strength was used as an estimate for this value (yield
strength and Young’s modulus are available from the manufacturer [28]). The maximum Parylene
thickness to prevent Parylene cracking at the desired bending (helix) diameter was calculated and is
shown in Table 2-4. At 2.6 and 1.6 mm diameters, Parylene up to 52 and 32 µm in thickness will not
experience sufficient stress to yield. At 0.25 mm diameter, the calculated maximum thickness is 5 µm.
Parylene strips of this thickness experienced no cracking in any tests. Parylene strips of 9.9 and 11.1 µm
experienced no or minimal cracking in all tests but are above the maximum thickness threshold,
indicating that there may be some slight stretching of the Parylene that is not visible in these samples or
that the bending strength of Parylene is greater than the yield strength (which was used for calculations).
Table 2-4: Calculated maximum Parylene thickness to prevent cracking at the given bending diameter.
Bending (Helix) Diameter Maximum Parylene Thickness
2.6 mm 52 µm
1.6 mm 32 µm
0.25 mm 5 µm
The thickness of the Parylene and the helix diameter had no apparent impact on the resulting
shape after thermoforming. All parts (except for a single outlier) held the desired shape when wrapped at
a 45° helix angle. When parts were wrapped at a 15° angle, several parts did not retain the desired shape
after thermoforming (Figure 2-10D), however there is no trend in the shape failures, indicating a fixturing
problem for these parts. Parylene strips which were tested at a 30° helix angle had identical results to
parts of equivalent thickness with a 45° angle (however not all thicknesses were tested at 30°).
Parylene strips constructed of two layers had comparable results wrapped in either direction (i.e.,
with the base layer towards the inside of the helix or the top layer towards the inside of the helix).
Although the residual stress in deposited Parylene films has been reported to vary with the thickness of
the film in some cases [10], these stress differences are minor and do not appear to impact the

62
thermoforming process when a bare Parylene sample is used. PMP parts (discussed in sections 2.3.6 and
2.3.7) are more impacted by asymmetric Parylene layers due to the high stress, interposing metal layer.
Thermoforming results for bare Parylene strips are summarized in Table 2-5, with representative
examples of a good result, cracking failure, and shape failure shown in Figure 2-10.
Table 2-5: Thermoforming result vs. thickness, helix angle, and helix diameter for bare Parylene strips at 200 °C thermoforming
temperature and 12 hour thermoforming time. ✓ indicates a good result, ∗ indicates minor cracking (partial-thickness), ×
indicates cracking (full-thickness), and ● indicates loose shape. © IOP Publishing. Reproduced with permission from [12] (CC
BY 4.0).
Parylene Thickness (µm) 30°, 45° Helix Angle 15° Helix Angle
Helix Diameter (mm) Helix Diameter (mm)
Inner
Layer
Outer
Layer Total 2.6 1.6 0.25 2.6 1.6 0.25
5.6 n/a 5.6† ✓ ✓ ✓ ✓ ✓ ✓
9.9 n/a 9.9† ✓ ✓ ∗ ● ✓ ∗
5.5 5.6 11.1 ✓ ✓ ∗ ● ● ✓
13.3 n/a 13.3‡ ✓ ✓ × ✓ ✓ ×
5.3 13.2 18.5†‡ ✓ ✓ × ● ● ×
13.2 5.3 18.5†‡ ✓ ✓ × ● ✓ × ●
20.9 n/a 20.9† ✓ ✓ × ● ✓ ×
5.4 20.4 25.8 ✓ ✓ × ● ✓ ✓ × ●
20.4 5.4 25.8 ✓ ✓ × ✓ ✓ ∗ ●
11.9 14.4 26.3† ✓ ✓ × ✓ ✓ × ●
14.4 11.9 26.3† ✓ ✓ × ● ● ×
† Only the indicated parts were tested at a 30° helix angle; all parts were tested at 15° and 45°
‡ The indicated parts were also tested at 100 and 150 °C, yielding the same results

63
Figure 2-10: Examples of bare Parylene strips (300 µm width by 20 mm length, variable thickness) thermoformed into 0.25 mm
helices, showing a good result (✓), minor cracking (∗), cracking (×), and loose shape (●). © IOP Publishing. Reproduced with
permission from [12] (CC BY 4.0).
2.4.3 Annealing Tests: Parylene-Metal-Parylene Devices
Mismatched stress between the three distinct layers in PMP devices (base Parylene, metal, and
top Parylene) caused devices to curl based on the net stress in the device (with resulting curvature towards
tensile layers and away from compressive layers). Annealing PMP devices alters the stress in each layer,
changing the magnitude and/or direction of device curling. Other parameters, such as pre-annealing the
base Parylene layer or depositing asymmetric Parylene layers (among many other design and processing
parameters not evaluated in this work), also impact the resulting curvature of the device. The impacts of
base layer annealing, asymmetric Parylene layers, annealing temperature, and annealing time were
evaluated experimentally and compared with the mathematical model. Only the thick/asymmetric and
thin/symmetric groups were evaluated using these tests.

64
2.4.3.1 Base Layer Annealing
The base Parylene layer on parts in the thin/symmetric group was annealed (150 °C, 4 hours)
prior to addition of the metal or top Parylene layers. As a result, the stress in the base Parylene layer
became more tensile (due to the shrinkage of the Parylene – see Figure 2-11Ai). When thin/symmetric
parts were taken off the silicon carrier wafer, they were curled towards the pre-annealed base layer to a
diameter of 4.2 ± 0.6 mm (mean ± standard deviation). This experimental value is slightly larger than the
value calculated by the mathematical model with equivalent conditions (2 mm towards the base layer),
likely due to the differences in processing conditions when calculating Parylene stress in literature (see
discussion in section 2.2).
Figure 2-11: Illustration of the stress in each layer of PMP devices before and after annealing. (Ai) Thin/symmetric devices
before annealing have moderate tensile stress in the base Parylene layer due to shrinkage during the base layer anneal and a low
compressive stress in the top Parylene layer, resulting in a device curled towards the base layer. (Aii) Thin/symmetric devices
after annealing have balanced stress around the neutral plane due to equal shrinkage in the Parylene layers and the high-stress
metal layer sitting on the neutral plane, resulting in a flat device. (Bi) Thick/asymmetric devices before annealing have high
compressive stress in the metal layer and low compressive stress in both Parylene layers. The compressive stress in all layers
results in enough expansion in the device to produce a low tensile stress in the metal layer (below the neutral plane), resulting in a
device with mild curvature towards the base layer. (Bii) Thick/asymmetric devices after annealing have high compressive stress
in the metal layer below the neutral plane and equal tensile stress in both Parylene layers, resulting in a device curled towards the
top (thicker) Parylene layer. © IOP Publishing. Reproduced with permission from [12] (CC BY 4.0).
Parts in the thick/asymmetric group, which did not undergo any high temperature annealing prior
to removal from the wafer, were curled towards the base layer to a much lesser extent (60 ± 22 mm
diameter); this closely matches the value produced by the model (62 mm towards the base layer). The
mild curling was due to the asymmetry of the Parylene layers in the device (described in detail in section
2.4.3.2).

65
Although the Parylene thicknesses were different in the two experimental groups so a quantitative
comparison cannot be made, the data suggest pre-annealing is largely responsible for the greater curling
in thin/symmetric group PMP device based on similarities to other devices in literature [7,38] and the
results of the mathematical model (experimental data and the mathematical model for both device groups
are plotted in Figure 2-12).
Figure 2-12: Modeled and experimental curvature diameter versus base layer annealing temperature for PMP devices. The mean
value of experimental data is plotted with error bars showing one standard deviation (SD). Modeled parameters were identical to
PMP device parameters for each device group (Table 2-3). Stress was calculated from temperature based on fitting literature
values to all data (described in section 2.2), with σtop calculated at 20 °C (representing no anneal). © IOP Publishing. Reproduced
with permission from [12] (CC BY 4.0).
It is also important to consider the impact of device processing conditions on the resulting
Parylene and metal stress. PMP device fabrication in this study included four photoresist patterning steps,
each of which required a baking step to harden the photoresist and/or for image reversal (for metal lift-off
patterning only). In total, the base Parylene layer experienced temperatures of 90 °C for 6 minutes and
110 °C for 45 seconds, and the full device (base and top Parylene layers) experienced a temperature of 90
°C for 10 minutes. The metal deposition and device release processes also exposed the base Parylene
layer and full device, respectively, to increased temperatures, however these processes did not exceed 60
°C with the equipment and parameters used in this study.
Although heating Parylene to temperatures near or above the glass transition temperature is
known to cause changes in stress [13,24,25], it is unknown how significant these changes are during such
a short heat exposure. The shortest published Parylene annealing test annealed Parylene at 50 °C for 15

66
minutes, followed by 100 °C for 15 minutes and found a stress increase from -6 to 17 MPa [13]); no
shorter duration annealing tests have been reported to our knowledge.
2.4.3.2 Asymmetric Parylene Layers
Asymmetry impacts the curvature of the device by moving the high stress metal layer away from
the neutral plane of the device. This results in unbalanced stress about the neutral axis, producing a net
moment on the device that causes it to curl. The magnitude of curling depends on the thickness of each
layer (and the resulting distance between the metal layer and the neutral plane) and the processing
conditions. Curling is generally most severe in annealed devices due to the increased tensile stress in the
Parylene layers acting in combination with the high compressive stress in the metal layer. In unannealed
devices (with no base layer anneal), every film layer has residual compressive stress (Figure 2-11Bi),
causing the device to expand and balance some forces prior to curling. This results in less severe curling
than in an annealed device, where the tensile Parylene layers and compressive metal layer are acting
against each other, resulting in a smaller degree of shrinkage or expansion and higher forces in each layer.
After annealing, thick/asymmetric devices followed the trend reported in literature and described
by the model, curling significantly towards the thicker, top Parylene layer (down to a minimum postannealing curvature diameter of 8.1 ± 2 mm when annealed for 12 hours at 200 °C) due to comparable
shrinkage in the Parylene layers and the compressive metal layer acting below the neutral plane.
Thin/symmetric devices became flatter after annealing (up to a maximum post-annealing curvature
diameter of 1300 mm when annealed for 12 hours at 100 °C). These changes are due to the stress increase
in Parylene after annealing resulting in unbalanced or balanced stress in PMP device layers. The curvature
estimated by the model with equivalent parameters for thick/asymmetric devices (with or without 200 °C
anneal) and thin/symmetric devices (with 200 °C anneal) closely matches the experimental values. For
thin/symmetric devices with only a 150 °C base anneal, the modeled curvature value is slightly smaller
than the experimental value (2 mm and 4.2 mm, respectively), likely due to the differences in processing
conditions when calculating Parylene stress in literature (see discussion in section 2.2). Experimental data

67
and modeled curvature for both device groups before and after annealing at 200 °C are plotted in Figure
2-13. Detailed post-annealing curvature analysis is included in sections 2.4.3.3 and 2.4.3.4.
Figure 2-13: Modeled and experimental curvature diameter versus ratio of top to base layer thickness for PMP devices before and
after 200 °C anneal. The mean value of experimental data is plotted with error bars showing one standard deviation (SD).
Modeled parameters were identical to PMP device parameters for each device group (Table 2-3). Stress was calculated from
temperature based on fitting literature values to all data (described in section 2.2). © IOP Publishing. Reproduced with
permission from [12] (CC BY 4.0).
In most areas on PMP devices, the top and base Parylene layers act to balance the stress in the
metal layer. In small areas where Parylene has been etched away (i.e., to produce exposed metal
electrodes or bondpads; see Figure 2-2), only two film layers are present (one Parylene, one metal),
resulting in a shift in the neutral plane of the device and unbalanced stress. When the top Parylene layer is
etched away, the neutral plane shifts down below the metal surface, allowing the high compressive stress
in the metal layer to curl the device towards the base layer (Figure 2-14A). Stronger curvature in areas
with the top Parylene etched away is commonly reported in literature [4,7,26].
When the base Parylene layer is etched away, the metal is deposited directly on the substrate
(resulting in tensile stress, instead of compressive stress when it is deposited on top of Parylene) and the
top Parylene layer deposits in a conformal layer on top of the metal. This also results in the shift of the
neutral plane towards the base, but with a tensile metal layer at the bottom surface of the device. This
produces regions with high tensile stress in the metal layer below the neutral plane (pulling the device
towards the base layer) and tensile stress in the Parylene layer above the neutral plane (pulling the device

68
towards the top layer), as illustrated in Figure 2-14B. The resulting curvature depends on the thickness of
the Parylene layer and the relative size of the local etched areas as compared to the full device area. In the
thin/symmetric group, the Parylene was not sufficiently thick to resist the tensile stress of the metal,
resulting in curvature towards the base layer. In the thick/asymmetric group, the thick top Parylene
produced a stronger force, resulting in curling towards the top layer.
Figure 2-14: Illustration of the stress in each layer of annealed PMP devices in regions with etched Parylene openings. The
neutral plane shifts in etched areas, resulting in different stress balance in each area. (A) Regions with top Parylene etched
openings (electrodes) have moderate tensile stress in the base Parylene (below the neutral plane) and high compressive stress in
the metal layer (above the neutral plane), resulting in curling towards the base layer. (B) Regions with base Parylene etched
openings (bondpads) have moderate tensile stress in the top Parylene layer (above the neutral plane) and high tensile stress in the
metal layer (due to deposition directly on silicon, not Parylene; below the neutral plane), resulting in curling towards the base
layer. © IOP Publishing. Reproduced with permission from [12] (CC BY 4.0).
These curvature changes in etched areas are most pronounced in devices which have been
annealed at low temperature (etched and non-etched regions of devices after annealing at 100 °C for 12
hours are shown in Figure 2-15; full dataset attached to the published manuscript [12]). In addition, only
the local region with etched Parylene is affected (see Figure 2-16), however these small areas can impact
the overall shape of the device.

69
Figure 2-15: Representative photos of PMP devices annealed at 100 °C for 12 hr. Column (A) shows thin/symmetric devices (4.4
+ 4.7 µm thickness) and (B) shows thick/asymmetric devices (3.4 + 11.5 µm thickness). Row (i) shows the region of the device
with no etched openings, (ii) shows the region of the device with etched openings in the top Parylene layer (electrodes), and (iii)
shows the region of the device with etched openings in the base Parylene layer (bondpads). Scale bar for all photos (bottom left
corner) is 3 mm; the measured average diameter (mean ± standard deviation) and number of samples measured are shown in the
bottom right corner of each image. *Indicated devices curled towards both the base and top Parylene layers in different
regions/samples, so the minimum diameter in each direction is included (the average diameter does not capture the physical
shape because a flat device has an infinite diameter). †No standard deviation is included due to insufficient sample size; measured
values are listed. © IOP Publishing. Reproduced with permission from [12] (CC BY 4.0).
Figure 2-16: Photo of the electrode region of a thin/symmetric PMP device annealed at 200 °C for 12 hr. Local regions with
etched top Parylene (marked with red arrows) are curved towards the base layer, while regions between the etched openings (with
a symmetric Parylene-metal-Parylene stack) are noticeably flatter. © IOP Publishing. Reproduced with permission from [12] (CC
BY 4.0).
2.4.3.3 Annealing Time
When annealing PMP devices, most stress changes in the films occur relatively quickly. Although
a 48 hour annealing time is often used to improve adhesion between Parylene layers [21,30], the resulting
curvature after annealing is similar for all tested annealing times (30 minutes to 48 hours).

70
In the thin/symmetric group, devices released off the wafer (prior to annealing) had a tight
curvature towards the base layer (-4.3 ± 0.7 mm diameter) and became significantly flatter after annealing
(21 to 76 mm diameter towards the top layer when annealed at 200 °C). Differences in annealing time did
not have a significant impact on resulting device curvature.
In the thick/asymmetric group, devices released off the wafer (prior to annealing) had a mild
curvature towards the base layer (-59 ± 22 mm diameter) and became curled towards the top layer after
annealing (4.8 to 23 mm diameter when annealed at 200 °C). Devices annealed for 30 minutes and 6
hours resulted in similar curvature diameters that were less severe (larger diameter) than devices annealed
for 12 hours and 48 hours.
Detailed curvature data and representative photos for both groups are included in Figure 2-17
(full dataset attached to the published manuscript [12]). The annealing time condition could not be
modeled due to insufficient data available in literature.

71
Figure 2-17: Representative photos of PMP devices annealed at 200 °C for varying times. Column (A) shows thin/symmetric
devices (4.4 + 4.7 µm thickness) and (B) shows thick/asymmetric devices (3.4 + 11.5 µm thickness). Each row represents a
different annealing time – (i) before annealing, (ii) 0.5, (iii) 6, (iv) 12, and (v) 48 hours. Scale bar for all photos (bottom left
corner) is 3 mm; the measured average diameter (mean ± standard deviation) and number of samples measured are shown in the
bottom right corner of each image. †No standard deviation is included due to insufficient sample size; measured values are listed.
© IOP Publishing. Reproduced with permission from [12] (CC BY 4.0).
2.4.3.4 Annealing Temperature
Annealing temperature is shown in literature to have a significant impact on film stress and can
thus be used to tailor the resulting device curvature after annealing or thermoforming, as is shown in the
modeling results in section 2.2.3. When annealing Parylene, higher temperature produces more shrinkage

72
(due to increased tensile stress in the film), as evidenced by more significant curvature changes in PMP
devices annealed at higher temperatures.
In the thin/symmetric group, devices released off the wafer (prior to annealing) had a tight
curvature towards the base layer (-4.3 ± 0.7 mm diameter – due to the 150 °C base layer anneal) and
became very flat after annealing at 100 °C, with mild curvature in both directions among various regions
on a single device. Curvature diameters ranged from -71 mm (towards the base layer) to 260 mm (towards
the top layer). As annealing temperature increased to 150 and 200 °C, the resulting parts had mild
curvature towards the top, likely due to the slightly thicker top Parylene layer.
In the thick/asymmetric group, devices released off the wafer (prior to annealing) had a mild
curvature towards the base layer (-59 ± 22 mm diameter) and similarly showed a flatter shape after a 100
°C anneal with mild curvature in both directions at different regions. Curvature diameters ranged from -
27 mm (towards the base layer) to 25 mm (towards the top layer). As annealing temperature increased to
150 and 200 °C, the resulting parts became curled towards the top due to the significantly thicker top
Parylene layer.
In both device groups, the experimental data closely matched the mathematical model in the
unannealed condition and after annealing at 200 °C, as shown in Figure 2-18. Parts annealed at 100 and
150 °C differed significantly from the modeled curvature, to a greater extent in the thin/symmetric group.
This suggests that there may be a dynamic relationship between Parylene stress and annealing
temperature around the glass transition temperature that is not captured in the linear stress models
published in literature. In addition, some of the fabrication conditions (which occur in the range of 60-110
°C on short time scales) may be impacting Parylene stress in ways that are not captured in the model.

73
Figure 2-18: Modeled and experimental curvature diameter versus full device annealing temperature for PMP devices. The mean
value of experimental data (where available) is plotted with error bars showing one standard deviation (SD). When the mean
could not be calculated, all data points are plotted with an ‘x’. Modeled parameters were identical to PMP device parameters for
each device group (Table 2-3). Stress was calculated from temperature based on fitting literature values to all data (described in
section 2.2). © IOP Publishing. Reproduced with permission from [12] (CC BY 4.0).
In both groups, the curvature of annealed devices differed significantly from unannealed devices
within the group. Detailed curvature data and representative photos for both groups are included in Figure
2-19 (full dataset attached to the published manuscript [12]).

74
Figure 2-19: Representative photos of PMP devices annealed at varying temperatures for 12 hours. Column (A) shows
thin/symmetric devices (4.4 + 4.7 µm thickness) and (B) shows thick/asymmetric devices (3.4 + 11.5 µm thickness). Each row
represents a different annealing temperature – (i) before annealing, (ii) 100 °C, (iii) 150 °C, and (iv) 200 °C. Scale bar for all
photos (bottom left corner) is 3 mm; the measured average diameter (mean ± standard deviation) and number of samples
measured are shown in the bottom right corner of each image. *Indicated devices curled towards both the base and top Parylene
layers in different regions/samples, so the minimum diameter in each direction is included (the average diameter does not capture
the physical shape because a flat device has an infinite diameter). © IOP Publishing. Reproduced with permission from [12] (CC
BY 4.0).
2.4.4 Thermoforming Tests: Parylene-Metal-Parylene Devices
As was observed with bare Parylene strips, thermoformed PMP devices did not have any
Parylene cracking when thermoforming to a 1.6 mm helix diameter at a 45° helix angle (Figure 2-20A),
with the exception of a single outlier (one sample in the thick/asymmetric group with the 11.5 µm layer
on the inside of the helix had minor cracking in some areas). All parts thermoformed to a 1.6 mm
diameter maintained electrical conductivity (i.e. continuous traces) after thermoforming.

75
At 0.25 mm diameter, thin/symmetric PMP devices with the 4.4 µm pre-annealed layer on the
inside of the helix showed minor cracking (Figure 2-20B) in some small areas and no cracking in
remaining areas. Thin/asymmetric PMP devices with the 4.6 µm pre-annealed layer on the inside of the
helix showed no cracking (Figure 2-20A) in any areas. These samples (thin/symmetric and
thin/asymmetric) were the only parts to maintain electrical conductivity after thermoforming to a 0.25
mm diameter. In thin/symmetric parts with the 4.7 µm layer formed to the inside of the helix, minor
cracking was present in all areas and metal traces did not remain continuous after thermoforming. In all
thick/asymmetric samples formed to a 0.25 mm diameter helix, Parylene cracking (Figure 2-20C) was
observed and metal traces were not continuous.
Three configurations in the thin/symmetric group did not maintain the correct shape in the
electrode region of the device after thermoforming to a 1.6 mm helix diameter (4.4 µm layer on the inside
of the helix thermoformed at 200 °C, and 4.7 µm layer on the inside of the helix thermoformed at 100 or
200 °C). In the flat annealing tests, the electrode region of PMP devices retained a different shape (with a
more significant curl away from the exposed metal surface) than the rest of the device due to high,
unbalanced film stress. In thermoformed thin/symmetric devices, this imbalance led to an unexpected
(loose) shape in the electrode region of most devices. In the optimized fabrication protocol for
thin/asymmetric devices, the thermoforming temperature was raised to 150 °C to increase Parylene
softening and prevent loose shape in thermoformed parts. In thick/asymmetric devices (all helix
diameters), thin/asymmetric devices (0.25 mm helix diameter), and thin/symmetric devices (0.25 mm
helix diameter), no loose shape was observed. Parts also maintained their shape after thermal cycling to
85 °C for 3+ hours up to 8 times over the course of 2-4 weeks.
Thermoforming results for PMP devices are summarized in Table 2-6, with representative
examples of a good result, cracking failure, and shape failure shown in Figure 2-20.

76
Table 2-6: Thermoforming result vs. thickness and helix diameter for PMP devices at 45° helix angle, 200 °C thermoforming
temperature, and 12 hour thermoforming time. ✓ indicates a good result, ∗ indicates minor cracking (partial-thickness), ×
indicates cracking (full-thickness), ● indicates loose shape, and – indicates discontinuous traces. © IOP Publishing. Adapted with
permission from [12] (CC BY 4.0).
Parylene Thick (µm) Flat
Anneal1
Thermoforming
Temperature
(°C)
Helix Diameter (mm)
Inner
Layer
Outer
Layer Total 1.6 0.25
4.62 3.8 8.4 No 150 Not
Tested
✓
4.42 4.7 9.1 No 100 ✓ ✓ ∗
4
200 ✓ ●
3 ✓ ∗
4
4.7 4.42 9.1 No 100 ✓ ●
3
∗ –
200 ✓ ●
3
∗ –
Yes 200 ✓ ∗ –
3.4 11.5 14.9 No 200 ✓ × –
11.5 3.4 14.9 No 200 ✓ × –
Yes 200 ✓ ∗
4 × –
1 The indicated parts were annealed flat at 200 °C for 48 hours prior to thermoforming.
2 The base Parylene layer on the indicated parts was deposited and annealed (150 °C, 4 hrs) before
adding metal and top Parylene.
3 Most regions of the indicated parts formed to the desired shape, however some areas in the
electrode region (with etched openings in the top, 4.7 µm Parylene) did not retain the desired
shape.
4 Most regions of the indicated parts had no cracking; some minor cracking was visible in a few
areas.

77
Figure 2-20: Examples of PMP devices (210-350 µm width by 20-40 mm length, variable thickness) thermoformed into 0.25 mm
helices, showing a good result (✓), minor cracking (∗), cracking (×), and loose shape (●). © IOP Publishing. Reproduced with
permission from [12] (CC BY 4.0).
2.5 Discussion
Annealing and thermoforming of thin film Parylene C strips and PMP devices (Parylene with
patterned thin film metal) were characterized and the results used to control the natural curvature of
microfabricated thin film polymer devices via annealing and achieve thin film Parylene-metal-Parylene
helices down to 0.25 mm in diameter via thermoforming. While linear relationships between parameters
cannot be established due to the large number of parameters and limited values tested for each parameter,
each of the individual results from sections 2.4.1 through 2.4.3 and the variety of parameter sets used in
section 2.4.4 and summarized in Table 2-6 point to an optimal set of design and process parameters for
producing 0.25 mm diameter helices. This process can also be applied to other 3D configurations, such as
cylinders (for nerve cuff electrodes), spheres (for retinal electrode arrays), or cones (for tips of depth
electrodes to interface with insertion stylets). This chapter (and the corresponding publication [12])
describes the smallest functional, thermoformed Parylene C device with patterned metal features in
thermoformed areas (0.25 mm diameter helix). Prior to this work, the smallest thermoformed Parylene

78
devices in literature had curvature diameters of 1-5 mm in functional regions [4,5,7-10] or 0.25 mm in
non-functional regions (i.e. regions with bare Parylene; with visible Parylene cracking) [4,9].
Total Parylene thickness must be less than or equal to 11.1 µm to prevent cracking, with even
thinner layers (down to 5.6 µm) performing better. This is an important consideration in medical devices,
as the Parylene usually acts as both a structural backbone and an electrical insulator and must remain
intact to prevent device damage and electrical leakage in the device. In addition, the optimization of bare
Parylene thermoforming provides insight that could inform the design of thermoformed microfluidic
channels, although additional testing would be necessary to ensure channels remained open and to
characterize any dimensional changes after thermoforming. Any cracking in Parylene microfluidic
channels would provide a pathway for fluid to escape the channel, rendering the device ineffective.
The base Parylene layer should be annealed at high temperature (150 °C used in this study, with
higher temperatures not exceeding the melting point likely having a more significant impact, as suggested
by the model) to pre-curl the device towards the base layer. Annealing the base Parylene layer increases
stress in the layer, promoting curvature towards the base after the finished device is removed from the
wafer due to shrinkage in the high-stress annealed Parylene.
The base Parylene layer should also be thicker than the top Parylene layer, keeping the total
thickness requirement (≤11.1 µm) and the minimum layer thickness necessary for electrical insulation. An
asymmetric Parylene-metal-Parylene film stack promotes curvature by moving the high-stress metal layer
away from the neutral plane of the device, with higher degrees of asymmetry producing more significant
curvature (shown experimentally and with modeled conditions). In annealed or thermoformed devices,
this results in the high stress metal layer dominating the bending moment, promoting curling towards the
thicker Parylene layer.
If a highly asymmetric device is used, a thermoforming temperature of 200 °C should be used to
take advantage of the stress imbalance in asymmetric devices. If a symmetric device is used, a
thermoforming temperature of 100 °C is preferred to minimize shrinkage of the top Parylene layer (which
negates the effects of the base Parylene anneal). High temperature annealing produces more significant

79
changes from the unannealed device shape (due to the temperature dependence of annealed Parylene
stress), while annealing time does not have as significant of an impact.
If other thicknesses, geometries, or processing parameters are required, the mathematical model
described here can estimate and optimize parameters to achieve the desired shape. The model, supported
by experimental data, provides insight into design considerations for any desired geometry, whether the
desired shape is flat or curved.
Overall, achieving a thermoformed Parylene device containing patterned thin film metal down to
0.25 mm in curvature diameter expands the design space for flexible 3D MEMS devices to allow for
miniaturized electrodes, sensors, and other patterned metal features. Evaluation of additional helix
diameters between 1.6 and 0.25 mm or below 0.25 mm would provide valuable insight into the potential
design space of thermoformed Parylene. The effects of long term and in vivo use on device geometries
have been evaluated in the testing of two novel devices (described in chapters 3 and 4). Thus far, effects
such as moisture ingress, thermal cycling, and other environmental factors have not led to changes in the
device shape or function.
2.6 References
[1] Schwartz A B, Cui X T, Weber D J and Moran D W 2006 Brain-Controlled Interfaces: Movement
Restoration with Neural Prosthetics Neuron 52 205–20
[2] Polikov V S, Tresco P A and Reichert W M 2005 Response of brain tissue to chronically
implanted neural electrodes J. Neurosci. Methods 148 1–18
[3] Kook G, Lee S W, Lee H C, Cho I-J and Lee H J 2016 Neural probes for chronic applications
Micromachines 7 179
[4] Kim B J, Chen B, Gupta M and Meng E 2014 Formation of three-dimensional Parylene C
structures via thermoforming J. Micromechanics Microengineering 24 065003
[5] Huang R and Tai Y-C 2010 Flexible parylene-based 3-D coiled cable IEEE 5th Int. Conf.
Nano/Micro Eng. Mol. Syst. 317–20
[6] Zurita F, Freko S, Hiendlmeier L, Duca F Del, Groll T, Seelbach O, Steiger K and Wolfrum B
2023 Thermoformed Parylene-C Cuff Electrodes for Small Nerve Interfacing Adv. NanoBiomed
Res. 2300102
[7] Kang X, Liu J Q, Tian H, Yang B, Nuli Y and Yang C 2015 Self-closed parylene cuff electrode
for peripheral nerve recording J. Microelectromechanical Syst. 24 319–32

80
[8] Yoon E, Koo B, Wong J, Elyahoodayan S, Weiland J D, Lee C D, Petrossians A and Meng E
2020 An implantable microelectrode array for chronic in vivo epiretinal stimulation of the rat
retina J. Micromechanics Microengineering 30 124001
[9] Hara S A, Kim B J, Kuo J T W, Lee C D, Meng E and Pikov V 2016 Long-term stability of
intracortical recordings using perforated and arrayed Parylene sheath electrodes J. Neural Eng. 13
066020
[10] Johnson A C and Wise K D 2012 A self-curling monolithically-backed active high-density
cochlear electrode array IEEE 25th International Conference on Micro Electro Mechanical
Systems (MEMS) (IEEE) pp 914–7
[11] Kim B J, Kuo J T W, Hara S A, Lee C D, Yu L, Gutierrez C A, Hoang T Q, Pikov V and Meng E
2013 3D Parylene sheath neural probe for chronic recordings J. Neural Eng. 10 045002
[12] Thielen B and Meng E 2023 Characterization of thin film Parylene C device curvature and the
formation of helices via thermoforming J. Micromechanics Microengineering 33 095007
[13] Dabral S, Van Etten J, Zhang X, Apblett C, Yang G R, Ficalora P and McDonald J F 1992 Stress
in thermally annealed parylene films J. Electron. Mater. 21 989–94
[14] Kebabi B, Malek C K and Ladan F R 1990 Stress and microstructure relationships in gold thin
films Vacuum 41 1353–5
[15] Malek C K, Kebabi B, Charai A and de la Houssaye P 1991 Effect of thermal treatment on the
mechanical and structural properties of gold thin films J. Vac. Sci. Technol. B 9 3329
[16] Parvizian M, Rahimi-Ashtari F, Goodarzi A, Sabrloui B, Sabaghzade J and Zabihi M S 2012
Residual stress improvement of platinum thin film in Au/Pt/Ti/p-GaAs ohmic contact by RF
sputtering power Appl. Surf. Sci. 260 77–9
[17] Ekkels P, Rottenberg X, Puers R and Tilmans H A C 2009 Evaluation of platinum as a structural
thin film material for RF-MEMS devices J. Micromechanics Microengineering 19
[18] Wang K, Yao K and Chua S J 2005 Titanium diffusion and residual stress of platinum thin films
on Ti/SiO 2/Si substrate J. Appl. Phys. 98
[19] Branger V, Pelosin V, Badawi K F and Goudeau P 1996 Study of the mechanical and
microstructural state of platinum thin films Thin Solid Films 275 22–4
[20] Kim B J, Chen B, Gupta M and Meng E 2013 Three Dimensional Transformation of Parylene
Thin Film Structures via Thermoforming IEEE 26th International Conference on Micro Electro
Mechanical Systems (MEMS) (IEEE) pp 339–42
[21] Kim B J, Washabaugh E P and Meng E 2014 Annealing effects on flexible multi-layered
parylene-based sensors 2014 IEEE 27th Int. Conf. Micro Electro Mech. Syst. 825–8
[22] Von Metzen R P and Stieglitz T 2013 The effects of annealing on mechanical, chemical, and
physical properties and structural stability of Parylene C Biomed. Microdevices 15 727–35
[23] Hassler C, Guy J, Nietzschmann M, Staiger J F and Stieglitz T 2011 Chronic Intracortical

81
Implantation of Saccharose-coated Flexible Shaft Electrodes into the Cortex of Rats 2011
International Conference of the IEEE Engineering in Medicine and Biology Society, Boston, MA
(IEEE) pp 644–7
[24] Hsu J M, Rieth L, Kammer S, Orthner M and Solzbacher F 2008 Effect of thermal and deposition
processes on surface morphology, crystallinity, and adhesion of Parylene-C Sensors Mater. 20 87–
102
[25] Harder T A, Yao T-J, He Q, Shih C-Y and Tai Y-C 2002 Residual stress in thin-film parylene-c
MEMS 2002 IEEE International Conference. Fifteenth IEEE International Conference on Micro
Electro Mechanical Systems pp 435–8
[26] Hirschberg A W 2018 Penetrating Parylene neural probe array for dense, in vivo, chronic
recordings (University of Southern California)
[27] The Engineering Toolbox 2003 Thermal Expansion - Linear Expansion Coefficients
engineeringtoolbox.com/linear-expansion-coefficients-d_95.html
[28] Specialty Coating Systems (SCS) 2021 SCS Parylene Properties scscoatings.com/wpcontent/uploads/2017/09/02-SCS-Parylene-Properties-1016.pdf
[29] Ortigoza-Diaz J, Scholten K, Larson C E, Cobo A, Hudson T, Yoo J, Baldwin A, Hirschberg A W
and Meng E 2018 Techniques and Considerations in the Microfabrication of Parylene C
Microelectromechanical Systems Micromachines 9
[30] Ortigoza-Diaz J, Scholten K and Meng E 2018 Characterization and Modification of Adhesion in
Dry and Wet Environments in Thin-Film Parylene Systems J. Microelectromechanical Syst. 27
874–85
[31] Hassler C, Von Metzen R P, Ruther P and Stieglitz T 2010 Characterization of parylene C as an
encapsulation material for implanted neural prostheses J. Biomed. Mater. Res. - Part B Appl.
Biomater. 93 266–74
[32] Lecomte A, Degache A, Descamps E, Dahan L and Bergaud C 2017 In vitro and in vivo
biostability assessment of chronically-implanted Parylene C neural sensors Sensors Actuators, B
Chem. 251 1001–8
[33] Stoney G G 1909 The tension of metallic films deposited by electrolysis Proc. R. Soc. London A
82 172–5
[34] Brenner A and Senderoff S 1949 Calculation of stress in electrodeposits from the curvature of a
plated strip J. Res. Natl. Bur. Stand. (1934). 42 105
[35] Scholten K, Larson C E, Xu H, Song D and Meng E 2020 A 512-Channel Multi-Layer PolymerBased Neural Probe Array J. Microelectromechanical Syst. 29 1054–8
[36] Meng E, Li P Y and Tai Y-C 2008 Plasma removal of Parylene C J. Micromechanics
Microengineering 18 045004
[37] Chang J H C, Lu B and Tai Y-C 2011 Adhesion-enhancing surface treatments for parylene
deposition 2011 16th International Solid-State Sensors, Actuators and Microsystems Conference

82
pp 390–3
[38] Ji B, Chen L, Wang M, Guo Z, Zhou Y, Huangfu S, Zhang K, Feng H, Chang H and Liu J 2021
Thermal Strain-Induced Self-Rolling Mesh Cuff Electrodes for Non-Linear Peripheral Nerve IEEE
34th International Conference on Micro Electro Mechanical Systems (MEMS) pp 567–70

83
Chapter 3. Endovascular Recording Device
3.1 Background
While conventional approaches to neural recording and stimulation (using scalp, surface, and
penetrating electrodes – see Table 1-1 and Figure 1-1) have led to significant clinical advances and
research findings, an endovascular approach is preferable in many cases. Endovascular delivery of
electrodes is a much less invasive procedure than the implantation of surface and depth electrodes and can
provide similar recording and stimulation capabilities, even with indirect access (across the vascular wall)
to the targeted tissue. Recent advances in endovascular surgical techniques and devices have made more
regions of the brain accessible via endovascular access and provided safety measures to reduce risk to the
patient. Novel endovascular devices are introduced into the market at an ever-increasing rate, making
endovascular therapy the dominant modality for many cerebrovascular conditions. Support sheaths and
distal access catheters can now provide seemingly paradoxical stiffness required for support in a co-axial
system and flexibility needed to navigate tortuous vascular anatomy to reach the targeted area. Similarly,
microcatheters are increasingly capable of safely reaching the distal ends of the cerebral vasculature.
Micro guidewires emphasize material softness to operate within delicate anatomy without compromising
torque efficiency necessary for precise steerability.
These advancements in endovascular technology allow the operator to practice with increasing
safety and efficacy. Combining these endovascular devices such as stents and guidewires with neural
recording and stimulating techniques creates the potential for minimally invasive device delivery to areas
of the nervous system that were previously challenging to reach. Subdural and epidural electrodes on the
surface of the cortex fail to reach deeper cortical areas and penetrating depth leads and microelectrodes
cause damage to any tissue in their path, including vasculature. Endovascular delivery serves as a
roadmap to accessing the cerebral venous structures that are located throughout deep and shallow brain
tissue, which were previously inaccessible or at high risk for complication with existing recording
techniques (see section 1.1 for a summary of types of electrodes used for neural interfaces).

84
There are multiple configurations of endovascular electrodes currently in use, primarily in
research settings, which are described briefly in section 1.1.5 and in detail in an external publication [1].
Endovascular electrodes have been used for both recording and stimulation with similar resolution to
surface electrodes [2-9], but with non-invasive access to deeper tissues.
3.1.1 Endovascular Recording
Endovascular electrodes generally record local field potentials (<300 Hz) or other low frequency
local signals from tissue within a few millimeters of the electrode. Higher resolution signals (such as
single or multi units) generally cannot be recorded due to filtering through the finite thickness of the
vascular wall, however single unit recording has been reported in a few cases [5,10]. This type of
recording has been termed endovascular electroencephalography (EV EEG) and is useful for a variety of
applications (see Table 3-1), the most common being seizure mapping and BCIs [11-14].
One of the most promising clinical applications of endovascular recording is for the mapping of
seizure foci (regions of the brain where seizures originate), primarily for temporal lobe epilepsy. In
clinical settings, seizures are mapped using scalp EEG followed by electrocorticography (ECoG) and/or
stereoelectroencephalography (SEEG). Scalp EEG is very low resolution and can only locate seizures
which occur in shallow brain tissue, so it is rarely sufficient for mapping regions of tissue to be resected
during epilepsy surgery. If scalp EEG does not provide an adequate seizure map, ECoG and/or SEEG
electrodes are implanted into the brain. Many electrodes are required to map the seizure foci (due to the
small recording area of each individual electrode). Hence, a large craniotomy and durotomy to is required
place large ECoG grids and/or many burr holes with significant tissue displacement for placing numerous
SEEG probes (usually on the scale of 5-15 SEEG probes per patient). Even with the placement of many
electrodes, only 30-40% of patients have seizure foci adequately localized, resulting in poor seizure
control. In addition, the tissue damage due to SEEG probe placement results in permanent neurological
damage in 12-18% of patients and death in 1-2% [15-17]. By replacing the use of ECoG grids and SEEG
probes with endovascular electrodes, the invasiveness, risk, and tissue damage associated with the
mapping process decreases significantly. In addition, if seizures are not located where electrodes are

85
placed (as often occurs with SEEG), the electrodes can be moved to a new brain region without additional
risk to the patient. Many endovascular seizure mapping studies (summarized in Table 3-1) have
demonstrated the efficacy of endovascular recording for ictal and interictal spike detection with better
performance than scalp EEG and similar performance to ECoG. The largest limitation of these studies is
the low recording resolution – most devices used contained only a single or a few electrodes, so very
limited data could be collected.
Another valuable application of endovascular electrodes is for BCI. In most cases, BCI uses scalp
EEG electrodes (due to their non-invasive nature) rather than implanted electrodes, which pose a high risk
to the patient for an elective procedure. Endovascular electrodes provide a minimally invasive option with
lower risk than surface or penetrating electrodes but significantly higher resolution than scalp electrodes.
Higher resolution BCIs allow for completion of more complex tasks (such as multiple degrees-of-freedom
prosthesis control or sensory feedback), providing more freedom to patients. Most published research on
endovascular devices focuses on device functionality and involves recording of background activity or
evoked potentials without a specific application listed (see Table 3-1). The information from these studies
is useful for BCI applications, as studies overall show better recording quality from endovascular devices
as compared to scalp electrodes. Several studies have directly tested the efficacy of the Stentrode for BCI
and have shown promising results in sheep [18] and humans [19]. The success of these simple BCIs show
promise for using the same electrodes for the control of prosthetic limbs [20].

86
Table 3-1: Summary of application, implant location, animal model, and blood vessel for published endovascular recording
studies. Common blood vessel locations in the human brain are illustrated in Figure 3-1.
Application Brain Region Animal Model Blood Vessel
Seizure Mapping Temporal lobe Human Middle cerebral artery [21-24]
Cavernous sinus [9,25]
Middle meningeal artery [26,27]
Baboon Petrosal sinus [28]
Vein of Labbe [28]
Straight sinus [28]
Subdural sinus [28]
Rat Middle cerebral artery [10]
Anterior cerebral artery [10]
Occipital lobe Baboon Petrosal sinus [28]
Vein of Labbe [28]
Vein of Galen [28]
Frontal Lobe Pig Superior sagittal sinus [6,29]
Auditory Evoked
Potentials
Temporal lobe Human Middle cerebral artery [7,30]
Basilar artery [7]
Rabbit Basilar artery [31]
Somatosensory
Evoked Potentials
Parietal lobe Human Middle meningeal artery [32]
Middle cerebral artery branches [32]
Baboon Middle cerebral artery branches [33]
Sheep Superior sagittal sinus [4,5,8,34,35]
Visual Evoked
Potentials
Occipital lobe Rabbit Basilar artery [31]
Sheep Superior sagittal sinus [36]
Not Listed Frontal lobe Human Callosomarginal artery [32]
Brain stem Human Basilar artery [32]
Spinal cord Frog (nearby vessels) [2,3]

Figure 3-1: Diagrams of the cerebral arteries (left) and cerebral veins (right). Arteries and veins commonly targeted in
endovascular recording and stimulation studies are labeled. The asterisk (*) denotes typical deep venous target for device
placement. © IOP Publishing. Reproduced with permission from [1].

87
3.1.2 Endovascular Stimulation
Neurostimulation with conventional electrodes (non-endovascular; usually surface and depth
probes) has greatly advanced the treatment of many neurologic disorders, with the largest impact on the
treatment of epilepsy. Vagus nerve stimulation (VNS) and deep brain stimulation (DBS) can interrupt
seizure activity and decrease both the frequency and severity of seizures [1]. Although these conventional
devices have proven to be clinically effective at treating epilepsy, endovascular devices offer the
advantage of a minimally invasive and more flexible implantation procedure.
Several studies have evaluated the use of endovascular electrodes for neural tissue stimulation for
various clinical and research purposes. These studies demonstrate successful endovascular stimulation by
recording from electrodes on downstream neurons [28,36] or by observing movement (after motor cortex
stimulation) [37]. There are also a few reports of successful ablation of diseased neural tissue using an
endovascular electrode [38-40].
Although DBS and peripheral nerve stimulation (PNS) have not yet been attempted using
endovascular electrodes, several computational studies have estimated the efficacy of endovascular DBS
and PNS and to identify sufficiently large blood vessels near common DBS targets. These models
estimate comparable stimulation thresholds for endovascular and stereotactic DBS in several target tissue
regions, indicating that endovascular DBS has potential to be a minimally invasive alternative to
stereotactic DBS in the treatment of conditions such as epilepsy, Alzheimer’s disease, major depressive
disorder, obsessive compulsive disorder, addiction, eating disorders, and others [41-45]. This result, along
with similar recording characteristics of endovascular and subdural/epidural electrodes, also supports the
feasibility of closed-loop DBS and PNS using endovascular electrodes [20].
3.1.3 Endovascular Electrodes
The endovascular electrodes described in section 1.1.5 have demonstrated the potential of
endovascular recording and stimulation in a variety of applications (summarized in Table 3-1). Existing
endovascular devices, however, are either too large to be inserted into small blood vessels, making many
brain regions inaccessible, or do not have a sufficient electrode count for high resolution recording and

88
stimulation, which is necessary for many clinical applications. This chapter describes the development of
a novel small diameter, high electrode count endovascular electrode array used for recording applications.
The device has a comparable size to a standard guidewire and comparable electrode count to existing
larger endovascular recording devices (see Table 3-2).
Table 3-2: Comparison of size and electrode count for endovascular devices.
Device Diameter Number of
Electrodes
✓ Small diameter
 Few electrodes
Guidewire/Catheter Electrodes [7,30,32] 0.25-0.6 mm 1-2
Microwire Electrodes [2,3,46] 0.6-20 µm 1-4
 Large diameter
✓ Many electrodes
Multi-electrode, SEEG-like Probes [6,47] 1 mm 24
Stentrode [4,5,8,34,35,37,48] 3-4 mm 6-12
✓ Small diameter
✓ Many electrodes This Work 0.4 mm 8-16
3.2 Microfabricated Endovascular Electrode Array
3.2.1 Endovascular Device Requirements
The endovascular electrode array was designed based on several requirements which were aimed
at producing a high resolution, small diameter device which could be used for a variety of applications:
1. the device must be able to be implanted into a 1.0 mm diameter vein,
2. the device must be able to record nearby neural signals (local field potentials),
3. the device must have multiple electrodes,
4. the device must be removable for acute use, and
5. the design must be flexible to allow for future changes in duration of use (acute vs chronic),
application (recording vs stimulation), and size.
With these requirements in mind, several design decisions were made. First, the goal of
implanting into a 1.0 mm diameter vein dictated the decision that the device must be able to slide freely
through a 0.76 mm inner diameter catheter (2.3 Fr). Catheters of this size are commonly used for vascular
access in 1.0 mm diameter blood vessels, and insertion of a device through a catheter (rather than on its
own) reduces the risk of damage during insertion to both the device and the blood vessel. In addition,
using standard surgical equipment reduces the difficulty of the surgical procedure and increases the odds
of a successful implantation.

89
Because the device is implanted through a catheter, it must also be strong enough to be pushed
through the full length of the catheter (often on the scale of 200 cm) and flexible enough to be able to
navigate the numerous twists and turns in the vasculature. Commercial guidewires have been advanced
over the past several decades to achieve this type of navigation (with strength, flexibility, and
torquability), however their construction (usually a solid material with micromachined grooves or coiled
wire) cannot easily be adapted into a multi-electrode system. Based on the navigating ability of
commercial guidewires and the decision to use a microfabricated electrode (described below), the device
was designed to be mounted onto a commercial guidewire with 0.25 mm outer diameter.
Only a few device configurations were feasible to achieve the small device size and high
electrode count requirements. The first option was to use a bundle of microwires (similar to an existing
endovascular device [2,3,46]). Although this would likely achieve requirements 1 through 4, it may not
allow for some advantageous or necessary features (for example, there are limitations on electrode count
and size due to the diameter of each wire and electroplating capabilities) and is not very flexible for future
changes (for example, stable chronic use is likely difficult due to electrode movement during blood flow).
Instead, a microfabricated, thin film, multi-electrode array mounted on a commercial guidewire was
chosen as it low-profile, electrode sizes can be varied within a wide range, multiple geometries are
possible, and the overall design is very flexible to allow for changes as needed.
3.2.2 Endovascular Device Design
The novel endovascular device described in this chapter (diagrams of each configuration in
Figure 3-2 and Figure 3-3) consists of several components. First, a thin film, microfabricated, Parylene C
electrode array (with embedded, patterned thin film platinum) served as the backbone for the electrodes
on the device. Insulated wires were electrically connected to exposed metal bondpads on the proximal end
of the electrode array (connected to each electrode by insulated metal traces) and insulated, leaving the
electrodes as the only recording area. The electrode array was attached to a guidewire, either by attaching
it lengthwise along the guidewire (as shown in Figure 3-2) or by wrapping in a helical configuration
around the guidewire (as shown in Figure 3-3). The wires were routed along the length of the guidewire

90
to the proximal end. Wires were electrically connected to a connector to mate with the recording
equipment.
Figure 3-2: Diagram of the novel endovascular device (straight configuration). The fully assembled device on a guidewire (coiled
for packaging) is shown at the top right, and a detailed view of the tip of the guidewire through the beginning of the wire region
is shown at the bottom.
Figure 3-3: Diagram of the novel endovascular device (helical configuration). The fully assembled device on a guidewire (coiled
for packaging) is shown at the top right, and a detailed view of the tip of the guidewire through the beginning of the wire region
is shown at the bottom.
3.2.2.1 Parylene Electrode Array
The Parylene electrode array was constructed of three layers: a base Parylene layer, a patterned
metal layer, and a top Parylene layer. The Parylene acted as a backbone material and insulation for the
patterned metal, with etched openings over the metal electrodes and bondpads. The array was configured
as a long, narrow strip to be wrapped in a helical shape around the guide wire. The helical shape allows

91
the device to be bent in any direction without stretching the Parylene backbone and causing damage to the
device, as opposed to a planar device which has limited flexibility in-plane. The overall width of the
Parylene backbone was limited to be able to wrap around the 0.25 mm diameter guidewire at a 45 degree
helix angle (see Figure 2-9) without any overlap. The distal end of each electrode array had either 8 or 16
exposed metal electrodes with 850 µm electrode pitch and sizes ranging from 100 µm in diameter to
4x200 µm diameter aggregate electrodes (see Figure 3-4). The proximal end had 8 or 16 exposed metal
bondpads (225 by 80 µm), each of which was attached to a single electrode via an insulated trace. Two
design generations were built and tested.
Figure 3-4: Diagram of 4x200 µm, 3x200 µm, and 2x200 µm aggregate electrodes. Each aggregate electrode consists of 2-4
connected 200 µm electrodes to increase the electrode surface area without increasing the required width of the Parylene
backbone. Oblong electrodes were not used as they are at higher risk of cracking during the thermoforming step.
The first-generation electrode array (Figure 3-5) had an L-shaped backbone, which allowed the
array to be wrapped in two different helix angles in the electrode and bondpad regions. The electrode
region was wrapped at a 45° angle to produce closer electrode spacing, while the cable and bondpad
regions were wrapped at a 15° angle to move the wire attachment points farther away from the recording
area (angles and helix axis are illustrated in Figure 3-5). Total Parylene thickness was 14.9 µm, with the
base layer significantly thinner than the top layer (3.4 µm vs 11.5 µm, respectively) to prevent tensile
stress in the metal (which could result in trace cracking) when forming into the helical shape. Photomasks
for the first-generation electrode array are included in appendix E.1.

92
Figure 3-5: Schematic of the first-generation endovascular electrode array.
The electrode region (Figure 3-6) of the electrode array was approximately 16 mm long, resulting
in an electrode region on the helical device spanning approximately 12 mm along the guidewire (at a 45°
helix angle). There were 16 circular or aggregate circular electrodes on the array (2 each of 4x200 µm,
3x200 µm, 2x200 µm, 200 µm, 175 µm, 150 µm, 125 µm, and 100 µm diameters) with a pitch of 850
µm. The backbone width was 350 µm to accommodate all electrodes and traces. Traces were 5 µm wide
with 5 µm spacing in the electrode region, expanding to 10 µm traces and gaps in the cable region to
decrease impedance.
Figure 3-6: Schematic of the electrode region of the first-generation endovascular electrode array.
The bondpad region (Figure 3-7) of the electrode array was approximately 15 mm long, resulting
in a bondpad region on the helical device spanning approximately 14 mm along the guidewire (at a 15°
helix angle). Each electrode was attached to a single rectangular bondpad (225 by 80 µm), with a bondpad
pitch of 1 mm. Traces were 5 µm wide with 5 µm spacing in the bondpad region.

93
Figure 3-7: Schematic of the bondpad region of the first-generation endovascular electrode array.
After benchtop testing of the first-generation electrode array (described in section 3.4), several
key issues were identified. First, the overall thickness of the Parylene (14.9 µm) was too thick to be
formed to the desired helix diameter of 0.25 mm. When wrapping the electrode array to this size, large
cracks form throughout the Parylene and metal layers, resulting in electrical failure and poor insulation
quality. Second, the asymmetry of the two Parylene layers resulted in significant stress imbalance which
promoted curvature in the opposite direction of what was intended. This device requires electrodes (which
face the top Parylene layer) to be facing the outside of the helix, whereas the thick top Parylene layer was
prone to higher stress changes during the thermoforming process, forcing the device to curl towards the
thicker top layer. (Note that the first-generation endovascular device was developed prior to the in-depth
study of Parylene-metal-Parylene device stress, curvature, and thermoforming described in chapter 2.)
Third, the “elbow” of the array (where the electrode region meets the cable region) was an area of high
stress concentration and was prone to tearing, rendering the device non-functional.
Based on these observations, a second-generation electrode array was designed (Figure 3-8). This
electrode array had a linear backbone (to remove any areas of stress concentration) and reduced overall
thickness (to reduce stress during the helical wrapping process). In addition, the base Parylene layer was
pre-annealed to promote curvature in the desired direction (toward the base layer). The electrode array
was designed to be wrapped at a 45° angle along its entire length, however any smaller helix angle was
also possible. Devices in this study were built with 45° angles and 0° angles (parallel to the guidewire), as

94
described in section 3.3. Photomasks for the second-generation electrode array are included in appendix
E.2.
A first batch of devices (generation 2A) was fabricated with total Parylene thickness of 9.1 µm,
with near-symmetric Parylene thicknesses (4.4 µm base thickness, 4.7 µm top thickness). A second batch
of devices (generation 2B) was fabricated with total Parylene thickness of 8.4 µm, with slightly
asymmetric Parylene thicknesses (4.6 µm base thickness, 3.8 µm top thickness). Two different device
sizes were fabricated, one with 8 electrodes and one with 16 electrodes.
Figure 3-8: Schematic of the second-generation endovascular electrode array in 16 electrode (top) and 8 electrode (bottom)
configurations.
The electrode region (Figure 3-9) of the electrode array was approximately 9 mm long for the 8
electrode device and 19 mm long for the 16 electrode device, resulting in electrode regions on the helical
device spanning approximately 6 and 13 mm along the guidewire (at a 45° helix angle). There were 8 and
16 circular or aggregate circular electrodes on the array (1 or 2 each, for 8 and 16 electrode devices, of
4x200 µm, 3x200 µm, 2x200 µm, 200 µm, 175 µm, 150 µm, 125 µm, and 100 µm diameters) with a pitch
of 1 mm. The backbone width was 350 µm for the 16 electrode device and 210 µm for the 8 electrode
device to accommodate all electrodes and traces. Traces were 5 µm wide with 5 µm spacing and were
patterned in a serpentine configuration (see Figure 3-9 inset) to provide some strain relief during flexing
of the device.

95
Figure 3-9: Schematic of the electrode region of the second-generation endovascular electrode array in 8 electrode (top) and 16
electrode (bottom) configurations.
The bondpad region (Figure 3-10) of the electrode array was approximately 7 mm long for the 8
electrode device and 15 mm long for the 16 electrode device, resulting in an electrode regions on the
helical device spanning approximately 5 and 11 mm along the guidewire (at a 45° helix angle). Each
electrode was attached to a single rectangular bondpad (225 by 80 µm), with a bondpad pitch of 1 mm.
Traces were 5 µm wide with 5 µm spacing with serpentine patterning in the bondpad region.

96
Figure 3-10: Schematic of the bondpad region of the second-generation endovascular electrode array in 8 electrode (top) and 16
electrode (bottom) configurations.
The second-generation electrode arrays performed significantly better than the first-generation
design, however some issues were identified in benchtop and animal testing (described in section 3.4).
Although the Parylene thicknesses is low enough to accomplish thermoforming without Parylene or metal
cracking, the devices are very delicate and prone to tearing, resulting in very low packaging yield in the
first fabrication run of second-generation devices (generation 2A). In addition, the mounting process was
identified as very low yield, prompting changes in the packaging strategy that were implemented in the
second fabrication run (generation 2B). Details on fabrication and packaging issues and how those issues
were addressed are described in section 3.4.
3.2.2.2 Insulated Wires
The insulated wires were sourced from an outside vendor (California Fine Wire, Grover Beach,
CA). The wires contained a platinum conductor (36 µm diameter) with polyimide insulation (6 µm
thickness). One wire was electrically connected to each bondpad (with conductive epoxy) and insulated
(with insulating epoxy) as shown in Figure 3-11. The bondpad region (where the wires are attached) was
the largest diameter region of the device, holding the guidewire, electrode array, insulated wires, and

97
insulation layer. The diameter and thickness of each of these components limited the total allowable
number of electrodes on the device (as this bulky area must pass through the catheter).
Figure 3-11: Diagram of the wire attachment point.
3.2.2.3 Guidewire
0.25 mm guidewires were commercially built and provided by a collaborator. The electrode array
with attached wires was mounted to the guidewire by aligning it in place (either wrapped helically around
the guidewire or lengthwise along the top surface) and securing it using epoxy. 10 mm at the tip of the
guidewire was left bare; this flexible tip is necessary for surgical navigation. The wires ran parallel to the
guidewire and were tacked in place every 3-8 cm to hold them in place.
3.2.2.4 Connector
To date, only benchtop testing and acute animal studies have been performed. As such, it was
possible to leave the catheter in place during testing (retracting it to expose the electrodes, but not fully
removing it). This allowed a connector design which was larger than the inner diameter of the catheter
and simpler to design and fabricate. The proximal end of each wire was connected to the short end of a
male header pin. These pins could either be connected manually using mini hook clips or to an adapter
which housed female header pins on one end and a male Omnetics connector on the other end (to mate
with the female Omnetics connector on the recording setup).
3.3 Experimental Methods
Devices were fabricated and tested using the steps outlined below. As fabrication issues arose,
minor changes were made to the device design, configuration, materials, and fabrication steps
(summarized in Table 3-3 and Figure 3-12). The steps outlined in sections 3.3.1 through 3.3.7 summarize

98
the most up to date fabrication process. The differences in fabrication processes for prior device builds are
described throughout these sections and summarized in section 3.3.8.
Table 3-3: Summary of fabrication changes between each design iteration of endovascular devices.
Group Electrode Array
Type
Thermoformed
Shorted
Bondpads
Insulating
Epoxy Type Testing Performed
A
Gen. 1 (L-shaped)
Sham (no metal) Yes (200 °C) No EpoTEK 301 in vivo – sheep (n = 2)
B Gen. 1 (L-shaped) Yes (200 °C) No EpoTEK 301 Benchtop
C Gen. 2a (Linear) No No EpoTEK 301 Benchtop
in vivo – sheep (n = 2)
D Gen. 2a (Linear) Yes (100 °C) No EpoTEK 301 Benchtop
E Gen. 2a (Linear) No Yes (groups
of 2-3) Loctite 4902 Benchtop
in vivo – sheep (n = 1)
F Gen. 2a (Linear) No No Loctite 4902 Benchtop
in vivo – sheep (n = 1)
G Gen. 2b (Linear) Yes (150 °C) No Loctite 4902 Benchtop
Figure 3-12: Order of fabrication steps for each design iteration. Thermoforming (highlighted in blue) and attachment to the
connector (highlighted in green) were moved later in the process for groups B and D. Mounting on the guidewire (highlighted in
yellow) was moved earlier in the process for group G. Thermoforming was not performed for groups C, E, and F, and attachment
to the connector was not performed for group A.
3.3.1 Thin Film Fabrication
The Parylene electrode arrays consisted of three thin film layers: a base Parylene C layer (3.4 µm
thick for generation 1, 4.4 µm thick for generation 2A, 4.6 µm thick for generation 2B), a metal layer (15
nm titanium adhesion layer and 200 nm platinum exposed metal layer for generations 1 and 2A, 20 nm
titanium adhesion layer, 25 nm platinum interfacing layer, 150 nm gold core layer, and 25 nm platinum
exposed metal layer for generation 2B), and a top Parylene C layer (11.5 µm thick for generation 1, 4.7

99
µm thick for generation 2A, 3.8 µm thick for generation 2B). Both single-sided (electrode and bondpad
openings in the top Parylene layer) and double-sided (electrode openings in the top Parylene layer,
bondpad openings in the bottom Parylene layer) electrode arrays were fabricated (see Figure 3-13 for an
illustration of the difference between single- and double-sided arrays). Poor performance of double-sided
thin film fabrication resulted in use of only single-sided parts for all assembled devices (see section 3.4.1
for discussion of poor thin film fabrication yield and section 3.4.2 for discussion of issues with the wire
attach process). The double-sided fabrication process was described previously in chapter 2 (section
2.3.2) and in detail in appendix B, and the single-sided fabrication process is described here and in detail
in appendix C.
Figure 3-13: Diagram of a simple Parylene-based electrode array. Center image shows a top down view with circular electrodes
at the top, rectangular bondpads at the bottom, and traces connecting each electrode to a bondpad. Cross sections of the electrode,
trace, and bondpad regions are shown in sections A-A’, B-B’, and C-C’, respectively. Cross sections on the left show the singlesided configuration (with all Parylene openings in the top layer), and cross sections on the right show the double-sided
configuration (with electrode openings in the top layer and bondpad openings in the base layer).
Single-sided Parylene electrode arrays were fabricated using a low temperature, batch process
based on prior work [49]. The fabrication process is summarized here, illustrated in Figure 3-14, and
described in detail in appendix B. 4” prime silicon wafers were used as the carrier substrate. The Parylene
C base layer was deposited (Figure 3-14A) using a chemical vapor deposition-like process (PDS 2010

100
Labcoter, Specialty Coating Systems, Indianapolis, IN) to a thickness of 3.4, 4.4, or 4.6 µm (gen. 1, 2A,
and 2B, respectively). For generation 2 devices only, the coated wafers were annealed in an oven (TVO-2,
Cascade Tek Inc., Longmont, CO) under vacuum with nitrogen flow at 150 °C for 4 hours. Next,
lithography was performed to add a lift-off photoresist mask (AZ 5214 IR, AZ Electronic Materials,
Branchburg, NJ; 1.8 µm thick). Then, a metal stackup (gen. 1 and 2A: 15 nm titanium + 200 nm
platinum; gen. 2B: 20 nm titanium + 25 nm platinum + 150 nm gold + 25 nm platinum) was deposited via
e-beam evaporation. Electrodes, traces, and bondpads were formed (Figure 3-14B) following lift-off in
40-50 °C acetone or NMP rinse followed by rinsing in isopropyl alcohol and deioinized water. For gen.
2B devices, wafers were treated with a silane adhesion promoter. The adhesion promoter solution (1 part
A-174 silane, 100 parts isopropyl alcohol, and 100 parts deionized water) was prepared 2.5 to 24 hours in
advance, then wafers were submerged in the solution for 30 minutes, followed by 30 minutes of air
drying, rinsing with isopropyl alcohol, and blowing dry. Next (for all devices), a top layer of Parylene C
was deposited (Figure 3-14C) to a thickness of 11.5, 4.7, or 3.8 µm (gen. 1, 2A, and 2B, respectively).
Openings in the Parylene for metal electrodes and bondpads on the top side of the device were etched
(Figure 3-14D) using O2 reactive ion etching (PlasmaPro 80 RIE, Oxford Instruments, Bristol, UK; 150
mT, 150 W, 50 sccm O2; etch rate approximately 0.2 µm/min) or O2 switched chemistry etching in a deep
reactive ion etcher (PlasmaPro 80 or PlasmaLab 100 ICP, Oxford Instruments, Bristol, UK; switched
chemistry process parameters detailed in [50]; etch rate approximately 0.08 µm/loop) masked by
patterned photoresist (AZ P4620, AZ Electronic Materials, Branchburg, NJ; 8-12 µm thick). The outline
of the device was cut out using the same etching procedure (gen. 1 and 2A) or using a femtosecond laser
(gen. 2B; WS-Multi Head, Optec, Frameries, Belgium; Pharos PH2-15W (Yb:KGW), Light Conversion,
Vilnius, Lithuania) with 515 nm wavelength, 15 kHz frequency, 0.75 W power, and 5 mm/s cutting
speed. Photoresist was removed using acetone after each etching step followed by rinsing in isopropyl
alcohol and deionized water. Devices were released from the wafer (Figure 3-14E) by placing a droplet of
water over one of the handling tabs, manually separating the tab from the wafer using a scalpel or sharp
tweezers, and gently pulling the device off the wafer using the tab as a handling point.

101
Figure 3-14: Parylene electrode array thin film fabrication process flow for single-sided devices, with cross-sectional views (left)
at the indicated points on the generation 1 and 2 devices shown with the red dotted arrows (right). (A) Parylene was deposited on
a silicon carrier wafer. (B) Platinum (with titanium adhesion layer; Ti/Pt) electrodes, traces, and bondpads (not shown) were
deposited and patterned. (C) Parylene was deposited on top of patterned metal. (D) Electrodes and bondpads (not shown) were
opened via O2 etching. (E) The electrode arrays were released.
After release from the wafer, parts were visually inspected (procedure in section 3.3.9.1) for
visible defects (such as torn Parylene, delaminated metal, shorted traces, or cracked traces) and
electrically tested for trace continuity (procedure in section 3.3.9.2).
3.3.2 Thermoforming
Although thermoforming is a required step to produce the final helical part design, several parts
were not thermoformed to help identify failure modes in the thermoforming and packaging steps and to
produce interim functional parts for use in animal studies. Parts that were used in the straight
configuration (groups C, E, and F) did not undergo the thermoforming step. For the remaining groups
(groups A, B, D, and G), the electrode arrays were thermoformed into 0.25 mm diameter helices. Parts
from groups B and D were thermoformed after wires had been attached and insulated (see Figure 3-12 for
process flow).
The electrode arrays were fixtured into a helical shape by wrapping around a 0.25 mm diameter
stainless steel mandrel using a template to define the helix angle (see Figure 2-9). Parts from groups A, B,
and D were manually wrapped and held in place using 0.01 mm thick Teflon film. Parts from group G

102
were mounted to a thermoforming fixture and wound around the mandrel (described in detail in appendix
D). After fixturing, the parts were placed into a programmable vacuum oven (TVO-2, Cascade Tek Inc.,
Longmont, CO), placed under vacuum, then purged three times with nitrogen to minimize oxygen in the
chamber. The oven was programmed to ramp up to the desired thermoforming temperature (200 °C for
gen. 1, 100 °C for gen. 2A, 150 °C for gen. 2B) at a ramp rate of approximately 0.7 °C/min, hold for 12
hours, then ramp down to room temperature.
After cooling, the electrode arrays were visually inspected (procedure in section 3.3.9.1) for their
ability to retain the desired shape and for cracking in the Parylene (see Figure 3-15). Arrays were also
electrically tested for continuity (procedure in section 3.3.9.2) to identify any electrical defects caused by
the thermoforming process.
Figure 3-15: Photos of electrode arrays with no cracking (left) and cracked Parylene (right). Cracking is most easily visible in
high magnification images (bottom).
3.3.3 Wire Preparation
Polyimide-insulated platinum wire (36 µm conductor diameter, 6 µm insulation thickness;
California Fine Wire, Grover Beach, CA) was cut to a length of 180 cm using a scalpel (one length per
bondpad to be attached). Both ends were stripped of insulation using a scalpel and sharp tweezers; the
bondpad attachment end was stripped for a length of 0.5-1 mm, and the connector end was stripped for a
length of 5-8 mm.

103
For sham devices (group A), bare stainless steel wire (50 µm diameter) was used instead of
insulated platinum wire.
3.3.4 Connector Attachment
Each free wire end (with 5-8 mm of stripped insulation) was electrically connected to a male
header pin to enable connection to the recording equipment. For each wire, 7-10 mm of 1.2 mm inner
diameter heat shrink was threaded onto the free wire end, then the wire was wrapped around the short side
of the header pin so the full stripped region was in contact with the header pin. Silver epoxy (EpoTEK
MED H20E) was applied to the pin/wire connection point, covering all sides of the header pin and all
exposed wire regions. The heat shrink was advanced along the wire over the silver epoxy and header pin,
then heated using a hot air rework gun to compress the wire and epoxy over the header pin. After all wires
were attached, the epoxy was cured in a convection oven (1300U, VWR, Radnor, PA) for at least 3.5
hours at 85 °C. After cooling, the heat shrink tubing was backfilled with insulating epoxy. For groups B
through D, 301 epoxy (EpoTEK, Billerica, MA) was used; for groups E through G, 4902 cyanoacrylate
(Loctite, Westlake, OH) was used. 301 epoxy was cured in a convection oven for at least 2.5 hours at 65
°C; 4902 cyanoacrylate was cured at room temperature for at least 1 hour.
For sham devices (group A), the wires were not attached to a connector because no electrical
connections were required. For devices which were thermoformed after wire attach (groups B and D), the
connector was added after thermoforming (see Figure 3-12 for process flow).
3.3.5 Mounting on Guidewire
Thermoformed devices were placed on the guidewire by wrapping the helical electrode array
around the guidewire, aligning the tip of the array 10 mm away from the tip of the guidewire (see Figure
3-16). Group G devices proceeded directly to the next step (wire attach).

104
Figure 3-16: Photo of a thermoformed device mounted on the guidewire.
For groups A through F (devices which had wires already attached), the electrode array was
aligned to the guidewire and temporarily secured using Kapton tape, aligning the tip of the array 10 mm
away from the tip of the guidewire. The device was then secured in place at the ends and every few
millimeters along the array (see Figure 3-17) using same insulative epoxy as was used for wire insulation
(groups A through D, EpoTEK 301; groups E and F, Loctite 4902), taking care to ensure no epoxy
covered any electrodes, and making sure the bondpad end (with the long, insulated wires attached) was
well secured. The epoxy was cured using the same cure schedule as was used for the insulation step
(groups A through D, at least 2.5 hours at 65 °C; groups E and F, at least 1 hour at room temperature).
Figure 3-17: Photo of the electrode end of a generation 2 linear device mounted on the guidewire, with arrows indicating
locations where epoxy was applied. More epoxy was added on the bondpad end to prevent Parylene damage due to movement of
the wires.
After securing the electrode array, the remaining length of the guidewire was carefully laid out in
a straight line and the insulated wires were laid along its length, with the header pins at the proximal end
of the guidewire. Wires were extended such that there was no slack and secured to the guidewire using
non-adhesive Teflon tape wrapped around the wires and guidewire. Kapton tape was used to secure the
guidewire and wires and prevent any movement. A small drop of the same insulative epoxy was added to
the guidewire/wire bundle every 3-10 cm to secure the bundle together. EpoTEK 301 (groups A-D) was

105
cured for at least 3 days at room temperature, and Loctite 4902 (groups E and F) was cured for at least 1
hour at room temperature. Kapton and Teflon tapes were removed after the epoxy was fully cured.
3.3.6 Wire Attach
Results of electrical testing and visual inspection were used to determine which bondpads would
have wires attached. Any channels with broken traces or delaminated bondpads or electrodes were not
attached.
Flat electrode arrays (from non-thermoformed groups C, E, and F, and from groups B and D
which were thermoformed after wire attach – see Figure 3-12 for process flow) were mounted on a Teflon
plate using Kapton tape with the bondpads facing upward (Figure 3-18). Tape was placed on the handling
tabs (which were removed at the end of the fabrication process) to prevent damage to the device.
Thermoformed devices (group A, which remained on the mandrel after thermoforming, and group G,
which were mounted on the guidewire) were secured on a Teflon plate using Kapton tape with the
bondpad to be attached facing upward (Figure 3-19). After each bondpad was attached and cured, the tape
was removed, the fixtured array was rotated, and the part was re-secured with the next bondpad facing
upward.
Wires were attached to each bondpad one at a time. The wire was held adjacent to the bondpad
attachment (0.5-1 mm stripped end) using a small piece of Kapton tape. The tip of the wire was aligned to
the bondpad by hand or using a custom vacuum tool and micrometer. The wire was secured in place using
the Kapton tape, and a second piece of tape was added for support and to ensure alignment. Silver epoxy
(MED H20E, EpoTEK, Billerica, MA) was applied over the bondpad and wire using a short piece of 50
µm wire until the entire bondpad and wire above the bondpad was covered. For flat electrode arrays, the
wire placement process was repeated for any bondpads that were not obscured by Kapton tape from
adjacent attached wires (see Figure 3-18). After all available bondpads were attached, the epoxy was
cured in a convection oven for at least 3.5 hours at 85 °C.

106
Figure 3-18: Photo of a flat electrode array with two wires attached. Bondpads were attached in three rounds; two bondpads were
skipped between each attached bondpad due to interference of the Kapton tape.
Figure 3-19: Photo of a thermoformed electrode array with one wire attached. Bondpads were attached one at a time.
After removing the electrode array from the oven, any Kapton tape which interfered with
bondpads yet to be attached was removed by placing a small droplet of isopropyl alcohol over the tape to
soften the adhesive and gently removing it with sharp tweezers. Wires were bent out of the way and
secured to the Teflon plate using Kapton tape if needed. Remaining wires were attached to remaining
bondpads using the same procedure, repeated in rounds until all desired bondpads were attached.

107
3.3.7 Wire Insulation
Insulative epoxy was prepared per manufacturer recommendations. For groups A through D, 301
epoxy (EpoTEK, Billerica, MA) was used; for groups E through G, 4902 cyanoacrylate (Loctite,
Westlake, OH) was used. The epoxy was applied over all bondpad areas, coating all silver epoxy, exposed
metal on the bondpads, and exposed metal on the wires to fully insulate the wire connections. 301 epoxy
was cured in a convection oven for at least 2.5 hours at 65 °C; 4902 cyanoacrylate was cured at room
temperature for at least 1 hour.
For group G (which was already mounted to the guidewire), the epoxy was also wicked between
the Parylene and the guidewire behind the bondpads and at the tip of the electrode region, securing the
device in place. After securing the electrode array, the remaining length of the guidewire was carefully
laid out in a straight line and the insulated wires were laid along its length, with the header pins at the
proximal end of the guidewire. Wires were extended such that there was no slack and secured to the
guidewire using non-adhesive Teflon tape wrapped around the wires and guidewire. Kapton tape was
used to secure the guidewire and wires and prevent any movement. A small drop of the same insulative
epoxy (Loctite 4902) was added to the guidewire/wire bundle every 3-10 cm to secure the bundle together
and cured for at least 1 hour at room temperature. Kapton and Teflon tapes were removed after the epoxy
was fully cured.
Visual inspection (procedure in section 3.3.9.1) and EIS (procedure in section 3.3.9.3) were
performed on fully assembled devices.
3.3.8 Fabrication Changes
Fabrication procedure changes were made iteratively with each device group to address issues
that were identified with the previous groups (changes are summarized in Table 3-3 and Figure 3-12). The
motivation for each of those changes is described here.
The first group of devices (group A) were sham devices (with no metal features) that were used to
develop and optimize the thermoforming procedure. One sham device was implanted into a sheep model
to demonstrate the surgical procedure with a realistic mechanical model device. After developing the

108
thermoforming process and demonstrating surgical feasibility, all future devices were built using
functional parts (with patterned metal features).
All generation 1 devices were thermoformed (groups A and B) because the L-shaped backbone
could not be mounted on the guidewire without thermoforming. Three of the generation 2 device groups
were mounted to the guidewire flat (without thermoforming; groups C, E, and F), and two were
thermoformed into a helical shape (groups D and G). Some flat and thermoformed groups were fabricated
concurrently (C with D, E with G) to identify issues in the thermoforming process. The thermoforming
temperature for generation 1 devices was 200 °C, 100 °C for generation 2A to reduce stress changes in
the Parylene, and increased to 150 °C for generation 2B to promote curvature in the asymmetric devices
without causing significant stress changes.
Wire attach can be performed before thermoforming (done for groups B and D) or after
thermoforming (done for groups A and G). Although attaching wires to a thermoformed device is difficult
due to the curved surface of the bondpad, fixturing for thermoforming after wire attach causes significant
mechanical failures due to the long and relatively large wires attached to the thin film device. Due to
these failures, thermoforming must be performed prior to wire attach.
Group E devices were built from electrode arrays with 8 electrodes, however electrodes were
grouped into sets of 2-3 and shorted in the bondpad region to create a device with fewer parts to improve
robustness for initial testing. In this group, a longer length of the insulated wire was stripped and a single
wire was electrically connected to 2-3 bondpads.
Early devices (groups A through D) used EpoTEK 301, a rigid insulating epoxy, while later
devices (groups E through G) used Loctite 4902, a flexible cyanoacrylate, to achieve better strain relief at
the wire attachment points and prevent damage during handling.
The generation 1 sham device (group A) and the generation 2, non-thermoformed devices (groups
C, E, and F) were tested in vivo in a sheep model, and all functional (non-sham) devices were tested on
the benchtop. Thermoformed devices from groups B and D did not survive the full fabrication process
with functional electrodes (due to poor yield in the thermoforming and mounting steps reducing the

109
functional electrode count, described in section 3.4.2), so they were not tested in vivo. Group G represents
the updated process flow based on all fabrication changes from issues in groups A through F and has only
been tested at the benchtop thus far.
3.3.9 Benchtop Testing
A series of benchtop tests were performed throughout the fabrication process (as listed in the
sections above) to identify any common failure points. A test to simulate surgical conditions was also
performed to quantitatively evaluate mechanical limitations of the device during surgery. The sections
below provide details on each test. All defects were recorded throughout the fabrication process to
evaluate fabrication yield at each step.
3.3.9.1 Visual Inspection
Visual inspection was performed using a stereoscope (HD60T, Caltex Scientific, Irvine, CA) for
larger defects and a high magnification microscope (Eclipse LV100, Nikon, Tokyo, Japan) for smaller
microscopic defects. On the electrode array, defects such as torn or cracked Parylene, delaminated metal,
shorted traces, or large cracks in the metal were visible with the stereoscope. Smaller metal cracks were
not visible with either microscope. After the electrode array was thermoformed, the stereoscope was used
to evaluate whether the array retained the desired thermoformed shape and the high magnification
microscope was used to identify the presence and severity of cracks in the Parylene. Examples of
common defects are shown in Figure 3-20.

110
Figure 3-20: Examples of defects identified during visual inspection: (A) large/deep Parylene cracks, (B) small/partial-thickness
Parylene cracks, (C) broken traces, (D) shorted traces, (E) delaminated metal which has remained attached, and (F) delaminated
metal which has detached. Most defects are visible under the stereoscope; the high magnification microscope is required to
determine the severity of Parylene cracking (as shown in B). All scale bars are 200 µm.
3.3.9.2 Continuity Testing
Continuity testing was performed using an LCR meter (E4980A, Agilent Technologies, Santa
Clara, CA) to measure the impedance between each bondpad/electrode pair using a 10 kHz, 20 mV
signal. Devices were laid flat on a Teflon plate and a large droplet of 1x phosphate buffered saline (PBS)
was applied over the electrode area to contact all electrodes. One of the LCR meter leads was attached to
a wire which was placed in contact with the PBS over the electrodes. If testing bare arrays, the second
LCR meter lead was attached to a bare wire or probe which was used to pick up a small droplet of PBS
and placed in contact with a single bondpad. If testing devices immediately after wire attach, the second
LCR meter lead was attached directly to the bare end of the wire. If testing fully packaged devices, the
second LCR meter lead was connected to the header pin (connector) attached to the wire. This setup
(shown in Figure 3-21) tested the continuity between a single bondpad or wire and the electrode region of
the device. Traces were considered continuous if the impedance magnitude was less than 150 kΩ and
phase was greater than -80°.

111
Figure 3-21: Continuity testing setup, with a cross-sectional view of the Teflon plate holding the electrode array. The left image
shows the setup as performed before wire attach (with a probe and small PBS droplet over one bondpad). The right image shows
the setup as performed after wire attach (with a probe connected to the attached wire).
3.3.9.3 Electrochemical Impedance Spectroscopy
Electrochemical impedance spectroscopy (EIS) was performed using a Reference 600 potentiostat
(Gamry Instruments, Warminster, PA) from 1 Hz to 1 MHz at 2.5 mV RMS in 1x PBS. A three-electrode
setup was used, with electrodes on the electrode array serving as the working electrode, a platinum wire
counter electrode, and a silver/silver chloride (3 M NaCl) reference electrode (BASi, West Lafayette, IN).
Functional electrodes had impedance values which leveled out at approximately 1 to 20 kΩ with phase
values from 0 to -40° at high frequency (above 10 kHz) and followed the characteristic shape shown in
the blue trace in Figure 3-22. Non-functional electrodes, such as those with broken traces or with
detached wires (which were commonly observed after the thermoforming and packaging steps – see
sections 3.4.1 and 3.4.2) had constantly rising impedance (from high to low frequency) with maximum
values in the GΩ range and phase near -90° at all frequencies (shown in the green trace in Figure 3-22).

112
Figure 3-22: Example EIS data for a broken (green x) and functional (blue circle) electrode.
3.3.9.4 Simulated Surgical Handling
Several fully assembled devices from group G underwent a simulated surgical handling test using
a custom fixture. The fixture (pictured in Figure 3-23) was designed to allow for multiple bends in a
variety of directions using interchangeable mandrels of varying size. To access deep brain areas from the
jugular vein, at least three right angle turns are expected (into the cortex, into the white matter, and from
the main trunk to branched vessels) at diameters ranging from 10 to 30 mm [51,52]. Based on these
parameters, the fixture was setup to include two 180° bends around interchangeable mandrels of 10, 15,
20, 25, and 30 mm diameter. A commercial catheter was routed around the mandrels, and the device was
inserted into the catheter until it passed through all bends. Prior to each test, the catheter was purged with
1x PBS. Tests were performed using a single bend radius at a time (in descending order), performing EIS
before and after each surgical simulation to identify any changes in electrode performance or electrode
failures.

113
Figure 3-23: Simulated surgical handling fixture, with interchangeable mandrels for variable bend radius testing.
3.3.10 In Vivo Testing
Several fully assembled devices were implanted into a sheep (Ovis aries) model targeting both
superficial and deep cerebral veins and sinuses. Each animal implantation followed the procedure
described here. The animal was anesthetized and placed in a supine position with the neck exposed.
Arterial access was established via the femoral artery and a catheter was placed into the common carotid
artery under fluoroscopic guidance. The arterial catheter was used to obtain arterial and venous
angiograms of the cerebral vasculature. Venous access was established via the internal jugular vein and a
series of catheters and guidewires of various sizes were advanced into the cerebral venous system under
fluoroscopic guidance. Fluoroscopy was used to visualize the cerebral venous anatomy and to guide a
catheter to the targeted blood vessel. For superficial cortical recordings, the target vessel was the superior
sagittal sinus. For deep cortical recordings, the target vessel was an internal cerebral vein near the
hippocampus. After the catheter was advanced to the target location, the endovascular recording device
was inserted through the catheter and advanced until the 10 mm bare tip extended outside of the catheter
(observed under fluoroscopic imaging). The catheter was then retracted approximately 20 mm, holding
the device in place, to ensure that the electrodes were exposed to the tissue but that no shearing between
the tissue and the electrodes occurred. The endovascular device and catheter were secured in place with
tape or surgical clamps, and the device was connected to a neural recording system.

114
In some cases, SEEG and/or EcoG electrodes were implanted in the same animal to provide a
comparison point for endovascular EEG recordings. To implant those electrodes, the animal was turned to
the prone position and the brain was exposed above the targeted cerebral vein. For EcoG electrodes, the
dura mater was carefully removed and the electrodes were placed onto the cortical gyrus adjacent to the
deployed endovascular device. For SEEG electrodes, a small region of the dura mater was excised and the
electrodes were inserted to a depth of approximately 18 mm from the bregma.
Neural activity was recorded from each implanted device over the course of several hours. During
recording, the anesthesia level was adjusted between “deep” and “light” levels under the guidance of a
veterinarian.
After recording was complete, a computed tomography (CT) scan was performed to verify the
implanted location of all devices. The animal was sacrificed after post-operative imaging was complete.
3.4 Experimental Results
3.4.1 Microfabrication and Thermoforming
Electrode arrays were successfully fabricated; photos of each device generation and configuration
immediately after being released from the wafer are shown in Figure 3-24. Visual inspection and
electrical testing of released thin film electrode arrays identified major yield differences between singleand double-sided parts. Bondpads on double-sided electrode arrays, which contained metal that was
exposed on the backside of the electrode array, were prone to full delamination due to poor adhesion to
the top Parylene layer (see Figure 3-20F). From visual inspection, 72% of double-sided electrode arrays
had 25% or more delaminated bondpads. No delamination was visible on single-sided electrode arrays
immediately after fabrication (partial delamination, as shown in Figure 3-20E, appeared after further
processing in some parts). A collection of generation 2 parts (14 single-sided and 6 double-sided) with 0
or 1 visible defects (after visual inspection) were continuity tested to determine the prevalence of nonvisible electrical defects. Single-sided parts exhibited better results for both electrode yield (good
electrodes per device) and device yield (devices with greater than 87% electrode yield). A wide range of
electrode yields was noted for both sets of parts (63% to 100% for single-sided, 44% to 100% for double-

115
sided), however device yield was significantly higher for single-sided parts at 64% (9/14) versus 17%
(1/6) for double-sided parts. Devices with less than 87% electrode yield were considered failed devices
and did not proceed onto further fabrication steps so that many electrodes (at least 7/8 or 14/16) would be
available for further benchtop testing (to evaluate failure modes during fabrication) and for in vivo
recordings. Devices with 100% electrode yield were chosen when available.
Figure 3-24: Generation 1 (top) and generation 2 (middle and bottom) electrode arrays immediately after removal from the carrier
wafer. Left images show the full electrode array, right images show zoomed in views of select features on the devices at the areas
indicated by the red dotted rectangles. Scale bars shown apply to all images within the column.
Thermoforming yield depended strongly on the thickness of each Parylene layer (described in
detail in chapter 2). No electrodes on generation 1 devices (14.9 µm total thickness) remained functional
after thermoforming, and many devices did not thermoform to the desired shape, instead taking on their
natural curvature as imposed by the stress in each thin film layer (curling towards the thicker, top
Parylene layer). Failure of these devices motivated the thermoforming and curvature study described in
chapter 2. Generation 2A devices (9.1 µm thickness) had some functional electrodes after thermoforming,
but with low to medium yield (0-56% of electrodes (n=17) remained functional on thermoformed
devices). Generation 2B devices (8.4 µm thickness) were not tested between thermoforming and wire
attach due to difficulty probing thermoformed bondpads, although the yield is estimated to be in the range

116
of 75-90% based on previous runs and other steps (see combined thermoforming and packaging yield and
justification for yield estimation for generation 2B (group G) in Table 3-4).
The order of fabrication steps also greatly impacted thermoforming yield. Devices which were
thermoformed after wire attach did not survive (0% yield) due to the long, rigid wires imposing
significant stress on the Parylene during fixturing, and usually resulting in detachment or tearing of the
Parylene. This result led to the final fabrication process flow used for group G, with thermoforming
performed before wire attach. Functionality of electrodes was evaluated via continuity testing and, if
wires were attached prior to thermoforming, via EIS.
3.4.2 Packaging
Yield was not assessed for mounting of thermoformed electrode arrays onto the guidewire in
group G devices due to difficulty probing thermoformed bondpads, however there are no apparent failure
modes in the process, likely producing very high yield near 100%. Combined thermoforming and
packaging yield for generation 2B (group G) is reported in Table 3-4.
When performed prior to thermoforming (as done in groups B through F), the wire attach process
was very reliable, with the only failures being due to handling issues (such as bending the attached wires
too harshly, which applied torque to the attachment area). Although the yield was near perfect (97%
success over 35 attachments in the most recent fabrication run, after optimizing technique to avoid
handling damage), the wire attach region is the most prone to handling damage during the remaining
fabrication steps due to the interface of the delicate, thin film electrode array with long, rigid wires.
Devices which had wires attached after thermoforming (group G) were not tested immediately
before wire attach, so exact yield cannot be assessed, however based on full fabrication yield (reported in
Table 3-4) and detachments during the wire attach process, yield was likely in the range of 55-75%. Of 15
attachments in the most recent fabrication run, 3 wires detached when rotating the device to expose the
next available bondpad (20% detachment), and additional failures were expected based on the final
electrode yield of 40% (n = 15 electrodes).

117
Wire insulation was very reliable, but one notable failure mode was discovered after fabricating
group D devices. The insulating epoxy used for groups A through D (EpoTEK 301) is designed as an
underfill epoxy and has very low viscosity. For bondpads with any delamination (commonly seen in
double-sided electrode arrays), the epoxy was able to wick underneath the thin film metal layer in some
parts, detaching it from the device and creating an insulative barrier between the attached wire and the
trace. In addition, EpoTEK 301 epoxy is rigid when cured, leading to higher stress concentration at the
boundary of the insulating epoxy and the Parylene electrode array. Due to these issues, EpoTEK 301 was
replaced with Loctite 4902, which has slightly higher viscosity (so it did not wick into any metal defects)
and is slightly flexible when cured (providing some stress relief at the delicate Parylene/wire interface).
Functionality of electrodes after wire insulation was evaluated via EIS to provide more detailed
information on the health of the electrodes (as compared to basic continuity testing).
Devices which were mounted to the guidewire after wire attach (groups A through F) suffered
from poor yield due to handling damage. During the mounting process, any movement of the guidewire
applied a torque on the Parylene electrode array/wire assembly, resulting in significant stress at the wire
attachment point and the electrodes (where one Parylene layer was etched away and the device was
significantly thinner). If the stress at this point was large enough, it would result in tearing of the electrode
array (see Figure 3-25) or detachment of the attached wire. When compounded with poor thermoforming
yield and insulative epoxy failures, no electrodes on helical devices from generations 1 or 2A (groups A,
B, and D) survived the full fabrication process. Straight devices (groups C, E, F, and G) also suffered
from mounting failures, but some electrodes survived on a few devices (see Figure 3-26 for example
photos of a group E device). Of the 7 straight devices which were built (4 in group C, 2 in group E), no
functional electrodes remained on 3 devices. The remaining 3 devices (which had 33 to 100% of
electrodes which were functional) were used for animal studies. Fully assembled devices were evaluated
with EIS to determine the overall performance of the electrodes.

118
Figure 3-25: Photos of the group C straight device build electrode and bondpad regions (top). Parylene tearing failures due to
torquing of the guide wire commonly occurred in the etched electrode region (green, bottom left) and wire attach region (blue,
bottom right).

119
Figure 3-26: Photographs of a fully fabricated novel endovascular device (straight configuration – group E). (A) The fully
assembled device alone and (B) in a container for transportation. (C) The tip of the guidewire with the electrode array mounted
on it, with dotted boxes showing (D-green) the electrode region and (E-blue) the bondpad region of the device. (F) The connector
with 3 header pins to connect to the 3 functional electrodes.

120
Figure 3-27: Photographs of a fully fabricated novel endovascular device (curled configuration - group G). (A) the fully
assembled device alone. (B) The tip of the guidewire with the electrode array mounted on it, with dotted boxes showing (Cgreen) the electrode region and (E-blue) the bondpad region of the device.
3.4.3 Electrical and Electrochemical Characterization
When possible, devices were tested at each fabrication step either with continuity testing or EIS.
The data from these tests provided insight into failure modes and electrode impedance changes occurring
during each fabrication step. Group C, D, E, and F devices were monitored throughout fabrication, and
the testing results (described in detail in the prior sections) are summarized in Table 3-4. Group G devices
could not be tested until they were fully fabricated because of difficulty probing bondpads after the device
was thermoformed. The impedance differences between electrodes of different sizes was minor but
visible in EIS plots (representative plots of electrodes of each size on a generation 2, non-thermoformed
device prior to mounting to the guidewire are shown in Figure 3-28).

121
Table 3-4: Electrode yield, listed as percentage (good electrodes/total electrodes tested) at each fabrication step for varying
processing conditions. Primary failure modes are listed for steps which had low yield.
Process Step Electrode Yield1 Primary Failure Modes
Thin Film Fabrication2 89% (128/144) Not evaluated
Thermoforming
Gen 2a
Before wire attach 50% (16/32) Cracking during fixturing,
excessive metal stress
Gen 2a
After wire attach 0% Long, rigid wires cause tearing at
attachment points during fixturing
Gen 2b
Before wire attach
~70-95%3
– estimated to be better than thermoforming gen 2a
before wire attach due to thinner, asymmetric Parylene layers
Wire Attach
Before
thermoforming 94% (46/49) Not evaluated
After
thermoforming
~55-75%3
– estimated to be slightly worse than wire attach
before thermoforming due to detachment of wires during
device rotation (3/15) and difficulty adhering wires to curved
surface
Wire Insulation
Double-sided
EpoTEK 301 7% (1/15) Epoxy ingress underneath poorly
adhered backside metal
Single-sided
Loctite 3902 86% (42/49) Epoxy ingress underneath poorly
adhered conductive epoxy
Mounting
Flat
After wire attach 43% (23/54) Handling damage, high torque
Thermoformed
Before wire attach ~100%3
– estimated based on no apparent failure modes
Full, Updated Process 40% (6/15) Handling damage, high torque
1 Total number of electrodes to survive a given step; calculation only considers electrodes which were
working at the beginning of each step
2 Parts with <2 visible defects were selected; yield does not consider parts with more defects which
were not tested
3 Parts in the final fabrication run were not tested at every step because testing was not possible until
wires were attached; estimated yield (based on results of prior runs) is reported instead
Estimation of unknown yields:
Group
G Yield =
Thin
Film
Yield
×
Thermoforming
Yield
×
Mounting
Yield ×
Wire Attach
Yield ×
Wire
Insulation
Yield
40%
(6/15) = 89% × ~70-95% × ~100% × ~55-75% × ~86%

122
Figure 3-28: EIS plots of representative electrodes of each size for a device after wire attach (before mounting to the guidewire).
3.4.4 Simulated Surgical Testing
One group G helical device with two functional electrodes underwent simulated surgical testing.
This device was built on a slightly larger guidewire (0.36 mm diameter) and a 2.3 Fr catheter (inner
diameter 0.77 mm) was used for the test. One of two electrodes failed after straight insertion into catheter
(no bends), likely due to the small clearance distance between the inner wall of the catheter and the epoxy
attachment points. This likely resulted in compression of the epoxy and wire at the bondpads into the
Parylene device, which can lead to cracking of the thin film metal and failure of the electrode. This
hypothesis is supported by the EIS data (shown in Figure 3-29), which shows an increase in impedance
magnitude for broken channels, but not to the same magnitude as a fully disconnected (or fully broken)
channel.

123
Figure 3-29: Sample EIS data for a functional electrode (blue), an electrode channel with a cracked trace (yellow), and a fully
disconnected channel (gray).
Testing was continued to determine the robustness of the one remaining electrode. The electrode
remained functional after simulated surgical testing through two 180° bends of 30 to 15 mm diameter and
failed at the smallest diameter of 10 mm. EIS data, shown in Figure 3-30, show minimal changes before
insertion (i.e. immediately after fabrication), after insertion through a straight catheter, and after insertion
into a catheter with bends ranging from 30 to 15 mm in diameter. After insertion through the 10 mm
diameter turns, the data suggests that a trace has been cracked, resulting in higher impedance magnitude
and impedance phase near -90°. It is unknown whether this failure is due to the device experiencing an
extreme bend diameter (10 mm) or from repeated insertions through a narrow catheter causing wear and
tear on the device.

124
Figure 3-30: Surgical simulation results, with plots of EIS data immediately after assembly and prior to insertion into a catheter
(red), after insertion into a straight catheter (orange), and after sequential insertions into the catheter routed around bends of
decreasing size (yellow through purple). Negligible changes are observed for all but the smallest bend diameter, and an increased
impedance magnitude and phase near -90 is observed for the 10 mm bend diameter, indicative of a broken electrode.
3.4.5 In Vivo Testing
Six in vivo sheep studies were performed, each with varying configurations of devices. Device
configurations used, target blood vessels, and key results of each study are described below and
summarized in Table 3-5.
Table 3-5: Summary of in vivo sheep studies for the endovascular device.
Study
Number
Number
of Sheep
Device
Configuration
Implant
Location
Functional EV
Electrodes1 Key Results
1 2 Group A
Superior
sagittal sinus
Deep vessels
n/a – sham
devices
Demonstrated surgical
feasibility
2 2 Group C Superior
sagittal sinus
1: 0/6
2: 3/8
LFP recordings under deep
and light anesthesia
3 1 Group E Superior
sagittal sinus 2/2
Simultaneous LFP
recordings with EV and
ECoG electrodes
4 1 Group F Superior
sagittal sinus 7/7
Simultaneous LFP
recordings with EV, ECoG,
and SEEG electrodes
1 Number of electrodes that remained functional after implant / total functional electrodes prior to
implant

125
In the first in vivo experiment, generation 1 sham devices (no electrodes) were successfully
implanted into the superior sagittal sinus of two sheep to demonstrate surgical feasibility. Device
placement was confirmed with a cerebral venogram (Figure 3-31C and D), and the device was explanted
(Figure 3-31E) after the animal was sacrificed. A microcatheter was also navigated to deep vessels to
demonstrate feasibility of surgical access to deeper structures (Figure 3-31A and B).
Figure 3-31: Results from the group A sham implantation demonstrating surgical feasibility. (A) Venogram (lateral view, left to
right = anterior to posterior) showing a microcatheter (tip located at the circle) advanced into the vein of the corpus callosum
(arrowhead). (B) Post-mortem MRI (sagittal view, left to right = anterior to posterior) confirming the location of the
microcatheter (arrow). (C and D) Venograms (lateral view, left to right = posterior to anterior) showing the sham device
(arrowhead) delivered into the superior sagittal sinus. (E) Photo of the explanted sham device advanced out of a microcatheter.
Generation 2 functional devices (groups C, E, and F) were successfully implanted into the
superior sagittal sinus of four sheep and the first successful electrical recordings from the microfabricated
endovascular device were collected. The first of the four animals (implanted with a group C device) did

126
not achieve recordings due to a damaged interior wall of the delivery catheter which caught the Parylene
electrode array and tore it. After replacing the damaged catheter, the implantation proceeded with the
damaged device to confirm surgical reliability.
Recordings were successfully collected in the second animal (implanted with a group C device).
The implanted device had 8 functional electrodes, 3 of which appeared to be functional after implantation
and 1 of which was able to record local field potentials (LFPs). After placement of the device, anesthesia
was lightened from 2% to 1% isoflurane, and the resulting increase in cortical activity was recorded (see
higher frequency of LFP spikes with lighter anesthesia in Figure 3-32, blue traces). This result is
consistent with findings from a similar study evaluating LFPs in rats exposed to similar isoflurane
concentrations (Figure 3-32, black traces) [53]. Recordings also showed similarities to microelectrodes
from an SEEG-like endovascular device (shown in Figure 1-14) with comparable frequencies and slightly
higher amplitudes in the thin film device (shown in Figure 3-33).
Figure 3-32: Recordings from the novel, thin film endovascular electrode in sheep (blue; group C) under (top) 2% and (bottom)
1% isoflurane. LFPs (amplitude spikes in the recorded signal) increase in frequency under lighter anesthesia. This is consistent
with surface recordings in rat (black) in a similar study (adapted from [53] with permission from Elsevier, CC BY NC ND).

127
Figure 3-33: Recording from the novel, thin film endovascular electrode in sheep (blue; group C) and a conventionallymanufactured endovascular microelectrode (red) from a similar study (adapted from [6] with permission from Elsevier). LFPs
(amplitude spikes in the recorded signal) from the thin film electrode have similar frequency and slightly higher amplitude than
the conventional microelectrode.
The third animal was simultaneously implanted with a group E device and a 2x4 subdural ECoG
grid (Ad-Tech Medical, Oak Creek, WI). The endovascular device had 2 functional electrodes, both of
which appeared to record low amplitude LFPs that matched some of the higher amplitude signals on the
surface electrodes (see Figure 3-34). Although the signal had a very small amplitude, it was dissimilar to
noise, suggesting that it was likely a low amplitude LFP recording.
Figure 3-34: Simultaneous recordings from a surface electrode (top) and an endovascular electrode (bottom; group E). Although
the signal from the endovascular electrode is smaller, the spikes track with the signal on the surface electrode, suggesting it is
recording the same neural activity.
The fourth animal was simultaneously implanted with a group F device, a 1x6 subdural ECoG
grid, and a depth probe with 6 microelectrodes and 10 macroelectrodes (ECoG and penetrating SEEG
electrodes from Ad-Tech Medical, Oak Creek, WI). Placement of each device was confirmed via
intraoperative angiogram (Figure 3-35). The endovascular device had 7 functional electrodes, all of which
recorded burst suppression (a typical pattern of low amplitude LFPs induced by administration of

128
anesthetic medication) which were highly correlated with concurrent recordings from ECoG and depth
electrodes, as quantified by high Pearson correlation coefficient values (Figure 3-36). Upon lightening
anesthesia from 3.5% isoflurane to 1.5% isoflurane (Figure 3-36 left versus right), burst activities became
more frequent due to higher brain activity during lower levels of sedation. This is consistent with findings
from previous animal studies with this device (Figure 3-32).
Figure 3-35: Intraoperative angiogram from the group F device implantation (lateral view, left to right = anterior to posterior)
showing placement of endovascular electrodes (red arrow), surface (ECoG) electrodes (black arrow) and penetrating (SEEG)
electrodes (green arrow).

129
Figure 3-36: Simultaneous recordings from surface macro-electrodes (black), penetrating macro-electrodes (green), penetrating
micro-electrodes (red) and endovascular micro-electrodes (blue; group F) showing simultaneous bursts of electrical activity on all
electrodes.
3.5 Discussion
A microfabricated endovascular recording device was designed, fabricated, and tested. This
device is the first thin film endovascular electrode array with multiple electrodes and sub-millimeter
diameter capable of accessing small blood vessels in a large animal model. The fabrication process and
device configuration were iterated and rigorously tested to identify and alleviate failure points in the
design and fabrication process. This testing identified the importance of providing sufficient strain relief
in the device, especially at the wire attachment point, where most failures occurred. The final fabrication
process flow places the high stress wire attach process last so that the Parylene electrode array is
supported by the guidewire and cannot be as easily torn by the insulated platinum wires. Although these
devices suffer from low yield (40% of electrodes from two devices remained functional after all
fabrication steps), one of two electrodes tested survived an aggressive surgical handling test, suggesting
devices will survive an implantation procedure which is expected to be more gentle. In addition, steps to
further improve yield have been identified and will be investigated. First, additional updates to the
thermoforming fixture to make it easier to handle and gentler on the device are in process. Second, new

130
wire attachment methods are being considered, including the use of a Parylene sheath over the bondpad
sites to provide more mechanical fixation of the wire onto the Parylene.
During the development of the fabrication process flow, an interim, non-thermoformed version of
the endovascular device was fabricated and implanted into a sheep model. These studies resulted in
several successful recordings, including the first published concurrent recordings from endovascular and
clinical surface and penetrating electrodes to demonstrate the recording capabilities of the novel
endovascular device. In the two most recent animal studies, all electrodes on the endovascular device
remained functional after implantation, further demonstrating the successive yield improvements made
during the design process. Future animal studies are planned to test the final configuration of the device
which utilizes the helical electrode array for additional strain relief. This more robust device will be tested
acutely in a sheep model, including awake recordings to gather more clinically relevant data. Future plans
include acute and subchronic studies while inducing seizures in the animal to gather data specific to the
target application.
3.6 References
[1] Thielen B, Xu H, Fujii T, Rangwala S D, Jiang W and Lin M 2023 Making a case for
endovascular approaches for neural recording and stimulation J. Neural Eng. 20 011001
[2] Llinas R R, Walton K D, Nakao M, Hunter I and Anquetil P A 2005 Neuro-vascular central
nervous recording/stimulating system: Using nanotechnology probes J. Nanoparticle Res. 7 111–
27
[3] Watanabe H, Takahashi H, Nakao M, Walton K and Llinás R R 2009 Intravascular neural
interface with nanowire electrode Electron. Commun. Japan 92 29–37
[4] Oxley T J, Opie N L, John S E, Rind G S, Ronayne S M, Wheeler T L, Judy J W, McDonald A J,
Dornom A, Lovell T J H, Steward C, Garrett D J, Moffat B A, Lui E H, Yassi N, Campbell B C V,
Wong Y T, Fox K E, Nurse E S, Bennett I E, Bauquier S H, Liyanage K A, van der Nagel N R,
Perucca P, Ahnood A, Gill K P, Yan B, Churilov L, French C R, Desmond P M, Horne M K,
Kiers L, Prawer S, Davis S M, Burkitt A N, Mitchell P J, Grayden D B, May C N and O’Brien T J
2016 Minimally invasive endovascular stent-electrode array for high-fidelity, chronic recordings
of cortical neural activity Nat. Biotechnol. 34 320–7
[5] John S E, Opie N L, Wong Y T, Rind G S, Ronayne S M, Gerboni G, Bauquier S H, O’Brien T J,
May C N, Grayden D B and Oxley T J 2018 Signal quality of simultaneously recorded
endovascular, subdural and epidural signals are comparable Sci. Rep. 8 8427
[6] Bower M R, Stead M, Van Gompel J J, Bower R S, Sulc V, Asirvatham S J and Worrell G A

131
2013 Intravenous recording of intracranial, broadband EEG J. Neurosci. Methods 214 21–6
[7] Nakase H, Ohnishi H, Touho H, Karasawa J, Yamamoto S and Shimizu K 1995 An Intra-Arterial
Electrode for Intracranial Electro-Encephalogram Recordings Acta Neurochir. (Wien). 136 103–5
[8] Opie N L, John S E, Rind G S, Ronayne S M, Grayden D B, Burkitt A N, May C N, O’Brien T J
and Oxley T J 2016 Chronic impedance spectroscopy of an endovascular stent-electrode array J.
Neural Eng. 13
[9] Mikuni N, Ikeda T, Murao K, Terada T, Nakahara I, Taki W, Kikuchi H and Shibasaki T 1997
“Cavernous Sinus EEG”: A New Method for the Preoperative Evaluation of Temporal Lobe
Epilepsy Epilepsia 38 472–82
[10] Zhang A, Mandeville E T, Xu L, Stary C M, Lo E H and Lieber C M 2023 Ultraflexible
endovascular probes for brain recording through micrometer-scale vasculature Science (80-. ). 381
306–12
[11] Schwartz A B, Cui X T, Weber D J and Moran D W 2006 Brain-Controlled Interfaces: Movement
Restoration with Neural Prosthetics Neuron 52 205–20
[12] Thielen B and Meng E 2021 A comparison of insertion methods for surgical placement of
penetrating neural interfaces J. Neural Eng. 18 041003
[13] Gerboni G 2019 Stent-based Endovascular Brain Interface: Characterisation of Recording and
Stimulating Properties (University of Melbourne)
[14] Viswam V, Obien M E J, Franke F, Frey U and Hierlemann A 2019 Optimal Electrode Size for
Multi-Scale Extracellular-Potential Recording From Neuronal Assemblies Front. Neurosci. 13 1–
23
[15] McGonigal A, Bartolomei F, Régis J, Guye M, Gavaret M, Fonseca A T Da, Dufour H, FigarellaBranger D, Girard N, Péragut J C and Chauvel P 2007 Stereoelectroencephalography in
presurgical assessment of MRI-negative epilepsy Brain 130 3169–83
[16] Jobst B C and Cascino G D 2015 Resective epilepsy surgery for drug-resistant focal epilepsy: A
review JAMA - J. Am. Med. Assoc. 313 285–93
[17] Klooster M A van ’t, Klink N E C van, Leijten F S S, Zelmann R, Gebbink T A, Gosselaar P H,
Braun K P J, Huiskamp G J M and Zijlmans M 2015 Residual fast ripples in the intraoperative
corticogram predict epilepsy surgery outcome Neurology 85 120–8
[18] Forsyth I A, Dunston M, Lombardi G, Rind G S, Ronayne S, Wong Y T, May C N, Grayden D B,
Oxley T, Opie N and John S E 2019 Evaluation of a minimally invasive endovascular neural
interface for decoding motor activity International IEEE/EMBS Conference on Neural
Engineering, NER (IEEE) pp 750–3
[19] Oxley T J, Yoo P E, Rind G S, Ronayne S M, Lee C M S, Bird C, Hampshire V, Sharma R P,
Morokoff A, Williams D L, Macisaac C, Howard M E, Irving L, Vrljic I, Williams C, John S E,
Weissenborn F, Dazenko M, Balabanski A H, Friedenberg D, Burkitt A N, Wong Y T, Drummond
K J, Desmond P, Weber D, Denison T, Hochberg L R, Mathers S, Brien T J O, May C N, Mocco
J, Grayden D B, Campbell B C V, Mitchell P and Opie N L 2021 Motor neuroprosthesis implanted

132
with neurointerventional surgery improves capacity for activities of daily living tasks in severe
paralysis : first in- - human experience J. Neurointerv. Surg. 13 102–8
[20] Rajah G, Saber H, Singh R and Rangel-Castilla L 2018 Endovascular delivery of leads and
stentrodes and their applications to deep brain stimulation and neuromodulation: A review
Neurosurg. Focus 45 1–6
[21] Hollanders T 2018 Feasability Study of Intravascular EEG-Recording (Ghent University)
[22] Boniface S J and Antoun N 1997 Endovascular electroencephalography: the technique and its
application during carotid amytal assessment J. Neurol. Neurosurg. Psychiatry 62 193–5
[23] Enomoto T, Matsumura A, Matsumaru Y, Aoki K and Nose T 2002 Multimodality examination
for detection of the epileptic foci with special reference to endovascular EEG recording Int.
Congr. Ser. 1232 655–60
[24] Nakase H, Ohnishi H, Touho H, Furuoka N, Yamada K, Takaoka M, Senoh M, Komatsu T,
Karasawa J and Yamamoto S 1993 [Spike discharge detected by intra-arterial
electroencephalography from intra-arterial guide wire in temporal lobe epilepsy] No to Shineki =
Brain Nerve 45 973–7
[25] Kunieda T, Ikeda A, Ohara S, Sadato A, Taki W and Shibasaki H 2000 Use of Cavernous Sinus
EEG in the Detection of Seizure Onset and Spread in Mesial Temporal Lobe Epilepsy Posterior
Tra Anterior Jugular vein Epilepsia 41 1411–9
[26] García Asensio S, Guelbenzu S, Barrena R and Valero P 1999 Technical Aspects of Intra-arterial
Electroencephalogram Recording Interv. Neuroradiol. 5 289–300
[27] García Asensio S, Guelbenzu S, Uclés P, Barrena M R and Valero P 1999 The Role of
Intravascular EEG Riv. di Neuroradiol. 12 139–42
[28] Gaba BS P and DeSimone C V. 2016 Novel Techniques in Epilepsy Management: Venous Pacing
and Capture of Electrical Activity in the Primate Cortex J. Neurol. Neurophysiol. 7
[29] Fujimoto A, Matsumaru Y, Masuda Y, Marushima A, Hosoo H, Araki K and Ishikawa E 2022
Endovascular Electroencephalogram Records Simultaneous Subdural Electrode-Detectable, Scalp
Electrode-Undetectable Interictal Epileptiform Discharges Brain Sci. 12
[30] Penn R D, Hilal S K, Michelsen W J, Goldensohn E S and Driller J 1973 Intravascular
intracranial EEG recording J. Neurosurg. 38 239–43
[31] He B D, Ebrahimi M, Palafox L and Srinivasan L 2016 Signal quality of endovascular
electroencephalography J. Neural Eng. 13 16016
[32] Stoeter P, Dieterle L, Meyer A and Prey N 1995 Intracranial Electroencephalographic and
Evoked-Potential Recording from Intravascular Guide Wires Am. J. Neuroradiol. 16 1214–7
[33] Reuter B M, Brassel F, Linke D B, Hartmann A and Kurthen M 1989 Tiefen-EEG und evozierte
Potentiale mittels intravaskulärer Katheter-Elektrode beim Pavian EEG. EMG. Z.
Elektroenzephalogr. Elektromyogr. Verwandte Geb. 20 81–3

133
[34] John S E, Apollo N V., Opie N L, Rind G S, Ronayne S M, May C N, Oxley T J and Grayden D
B 2019 In Vivo Impedance Characterization of Cortical Recording Electrodes Shows Dependence
on Electrode Location and Size IEEE Trans. Biomed. Eng. 66 675–81
[35] Opie N L, Rind G S, John S E, Ronayne S M, Grayden D B, Burkitt A N, May C N, O’Brien T J
and Oxley T J 2016 Feasibility of a chronic, minimally invasive endovascular neural interface
Proceedings of the Annual International Conference of the IEEE Engineering in Medicine and
Biology Society, EMBS (IEEE) pp 4455–8
[36] Gerboni G, John S E, Rind G S, Ronayne S M, May C N, Oxley T J, Grayden D B, Opie N L and
Wong Y T 2018 Visual evoked potentials determine chronic signal quality in a stent-electrode
endovascular neural interface Biomed. Phys. Eng. Express 4
[37] Opie N L, John S E, Rind G S, Ronayne S M, Wong Y T, Gerboni G, Yoo P E, Lovell T J H,
Scordas T C M, Wilson S L, Dornom A, Vale T, O’Brien T J, Grayden D B, May C N and Oxley
T J 2018 Focal stimulation of the sheep motor cortex with a chronically implanted minimally
invasive electrode array mounted on an endovascular stent Nat. Biomed. Eng. 2 907–14
[38] Henz B D, Friedman P A, Bruce C J, Holmes D R, Okumura Y, Johnson S B, Packer D L and
Asirvatham S J 2008 Successful radiofrequency ablation of the cerebral cortex in pigs using the
venous system: Possible implications for treating CNS disorders Epilepsy Res. 80 213–8
[39] Henz B D, Friedman P A, Bruce C J, Holmes D R, Bower M, Madhavan M, DeSimone C V.,
Wahnschaffe D, Berhow S, Danielsen A J, Ladewig D J, Mikell S B, Johnson S B, Suddendorf S
H, Kara T, Worrell G A and Asirvatham S J 2014 Advances in radiofrequency ablation of the
cerebral cortex in primates using the venous system: Improvements for treating epilepsy with
catheter ablation technology Epilepsy Res. 108 1026–31
[40] Ammerman J, Caputy A and Potolicchio S 2005 Endovascular ablation of a temporal lobe
epileptogenic focus – a complication of Wada testing Acta Neurol. Scand. 112 189–91
[41] Opie N L and O’Brien T J 2022 The potential of closed‐loop endovascular neurostimulation as a
viable therapeutic approach for drug‐resistant epilepsy: A critical review Artif. Organs 46 337–48
[42] Teplitzky B A, Connolly A T, Bajwa J A and Johnson M D 2014 Computational modeling of an
endovascular approach to deep brain stimulation J. Neural Eng. 11
[43] Liu J Y, Grayden D B, Keast J R and John S E 2021 Computational Modeling of an Endovascular
Peripheral Nerve Interface Proceedings of the Annual International Conference of the IEEE
Engineering in Medicine and Biology Society, EMBS pp 5966–9
[44] Neudorfer C, Bhatia K, Boutet A, Germann J, Elias G J, Loh A, Paff M, Krings T and Lozano A
M 2020 Endovascular deep brain stimulation: Investigating the relationship between vascular
structures and deep brain stimulation targets Brain Stimul. 13 1668–77
[45] Teplitzky B A 2016 Personalized computational models of deep brain stimulation (University of
Minnesota)
[46] Ruddy B P 2006 Conducting polymer wires for intravascular neural recording (Massachusetts
Institute of Technology)

134
[47] Thomke F, Stoeter P and Stader D 1998 Endovascular electroencephalography during an
intracarotid amobarbital test with simultaneous recordings from 16 electrodes J. Neurol.
Neurosurg. Psychiatry 64 565
[48] Opie N L, Oxley T J, Rind G S, Ronayne S M, John S E, May C N and Grayden D B 2019
Medical Device for Sensing and or Stimulating Tissue
[49] Scholten K, Larson C E, Xu H, Song D and Meng E 2020 A 512-Channel Multi-Layer PolymerBased Neural Probe Array J. Microelectromechanical Syst. 29 1054–8
[50] Meng E, Li P Y and Tai Y-C 2008 Plasma removal of Parylene C J. Micromechanics
Microengineering 18 045004
[51] Cieślicki K 2004 Experimental and Nunerical Modelling of Flow in the Human Cerebral Arteries
J. Med. Informatics Technol. 7 17–26
[52] Duvernoy H M, Delon S and Vannson J L 1981 Cortical blood vessels of the human brain Brain
Res. Bull. 7 519–79
[53] Stenroos P, Pirttimäki T, Paasonen J, Paasonen E, Salo R A, Koivisto H, Natunen T, Mäkinen P,
Kuulasmaa T, Hiltunen M, Tanila H and Gröhn O 2021 Isoflurane affects brain functional
connectivity in rats 1 month after exposure Neuroimage 234 117987

135
Chapter 4. Peripheral Nerve Stimulating Electrode
4.1 Background
Peripheral nerve stimulation (PNS) is used for the treatment of a variety of conditions, with both
systemic applications (such as chronic pain and epilepsy) and targeted applications (such as prosthetic
limb feedback and artificial hearing/vision) [1-7]. Interfacing with the peripheral nervous system provides
several advantages over the central nervous system, most notably that the surgical risk is far lower.
Placement of electrodes in the brain has high risk of complications due to infection and damage to areas
along the path of insertion for penetrating electrodes [8-10]. Stimulation of the peripheral nerves can, for
many applications, have similar effects as stimulation of brain tissue without the added risk. One such
example is the stimulation of the vagus nerve, which carries signals between the brain stem and various
organs. Vagus nerve stimulation (VNS) is has been shown to be effective in the treatment of many
different disorders in research settings and has been FDA approved for the treatment of epilepsy and
depression [11,12].
Many different types of peripheral nerve interfaces have been developed to interface with the
peripheral nerve (anatomy shown in Figure 4-1) in different ways which vary in selectivity and
invasiveness. The most common type of peripheral nerve interface used in both research and clinical
settings is the extraneural electrode, which sit outside the nerve and stimulate or record from outside the
epineurium (see section 1.1.2.2). Of the many extraneural electrode configurations, extraneural cuff
electrodes, which wrap around the nerve and stimulate it from the exterior surface of the epineurium, are
commonly used due to their simple construction and ability to contact all sides of the nerve without
causing nerve compression [11,13]. Most cuff electrodes (including those which are currently approved
for clinical use) are built out of metal foil embedded in polydimethylsiloxane (PDMS) with selectively
exposed electrode regions, resulting in a structure with a large wall thickness (≥0.4 mm); consequently,
such cuffs are suitable for interfacing with nerves in the range of 1 to 5 mm in diameter [3,4,11,13-15].

136
Figure 4-1: Anatomy of a peripheral nerve.
The clinical success of these cuff electrodes has motivated the need for a smaller diameter cuff
electrode to interface with branched nerves that lie closer to end organs. Stimulation of branched nerve
fibers allows for more targeted therapy with fewer side effects because a stimulated branched nerve will
have fewer downstream connections than common clinical nerve targets (such as the cranial nerves,
which control a wide variety of body functions). This PDMS/foil architecture that is currently used in
clinical cuff electrodes is not well suited for increasingly small parts and is likely already near its lower
size limit. It would be difficult to scale current cuff designs down to intimately interface with smaller
nerves because of the limitations of the materials and how they are assembled. In addition, these bulky
(and relatively stiff, in the case of metal foils) materials do not interface well with small nerves because
they can apply larger forces to the nerve, resulting in tissue damage [5,11].
This chapter describes the development of a smaller, more flexible cuff electrode built on a thin
film Parylene backbone. This microfabricated cuff electrode was designed to be implanted onto branched
nerves with diameters between 0.5 and 1.0 mm. An implantable pulse generator (IPG) and a variety of
other electrodes and sensors are being developed by collaborators as a part of the OpenNerve
neuromodulation platoform being developed by the Center for Autonomic Nerve Recording and
Stimulation Systems (CARSS; NIH award number U41 NS129514) with the end goal of providing an
open source, customizable neuromodulation system for use in animal models and humans.

137
4.2 Microfabricated Peripheral Nerve Cuff Electrode
4.2.1 Peripheral Nerve Cuff Requirements
The peripheral nerve cuff electrode was designed based on several requirements which were
aimed at producing a branched nerve cuff electrode for use in combination with the other OpenNerve
components:
1. the cuff must be able to fit snugly onto nerves 0.5 to 1.0 mm in diameter,
2. the cuff must be soft-closing and self-sizing,
3. the cuff must have two electrodes capable of delivering clinical levels of bipolar stimulation
(based on common values used for vagus nerve stimulation),
4. the cuff must be mechanically and electrochemically safe for chronic use, and
5. the cuff must be attached to the common OpenNerve leads (developed by a collaborator) for
compatibility with the larger OpenNerve system.
With these requirements in mind, several design decisions were made. First, a microfabricated
Parylene and thin film metal construction was selected to satisfy the size, the soft-closing requirement,
and the self-sizing requirement. Microfabricated Parylene can be thermoformed into sub-millimeter sizes
(as discussed in chapter 2) while maintaining flexibility and memory of its thermoformed shape. This
allows for the cuff to be thermoformed to a smaller diameter than the nerve, then expanded around the
nerve to allow it to gently close and fit to the exact nerve size.
To prevent nerve compression in nerves on the larger end of the desired diameter range, the cuff
was designed in two sizes: one which fits nerves 0.5 to 0.8 mm in diameter, and one which fits nerves 0.7
to 1.0 mm in diameter. The electrodes are long enough to wrap at least 1.5 times around the diameter of
the nerve to ensure they are securely held in place without adding an external anchor around the nerve.
4.2.2 Peripheral Nerve Cuff Design
The peripheral nerve cuff described in this chapter consists of two main components: the
Parylene-based cuff electrode and the lead (see Figure 4-2 and Figure 4-3). The lead, which is developed
and manufactured by a collaborator (Med-Ally, Goose Creek, SC), consists of two insulated wires with a
Bal Seal connector on one end (to be plugged into the IPG) and bare wires on the other end (to be
attached to the Parylene cuff electrode). This work focuses on the development of the Parylene cuff

138
electrode, which contains the two electrodes (for bipolar stimulation), bondpads which are used to
electrically connect to the lead, and a short “tail” to provide strain relief between the cuff and the lead.
Figure 4-2: Diagram of the fully assembled peripheral nerve full cuff.
Figure 4-3: Diagram of the fully assembled peripheral nerve helical cuff.

139
The Parylene electrode array contains three layers: a base Parylene layer, a patterned metal layer,
and a top Parylene layer. The Parylene acts as a backbone material and insulation for the patterned metal,
with etched openings over the metal electrode and bondpad sites. A multi-layer metal stack consisting of
a titanium adhesion layer, gold as a core layer, and platinum as the top, exposed metal layer was selected
to produce a flexible thin film layer (attributed to the mechanical properties of the gold) with the
favorable electrochemical properties of the exposed platinum surface. Two Parylene backbone designs
were tested to satisfy the project requirements (down-selected from a larger group of designs, as described
in appendix E). The required electrode surface area and inter-electrode distance were determined first, as
those impacted the backbone design.
4.2.2.1 Electrical Design
The distance between electrodes was set based on recommendations from experts in the field and
geometry of existing larger cuff electrodes. In general, the ideal distance between electrodes is
proportional to nerve diameter, so the electrode spacing found in the existing Liva Nova cuff (9 mm
spacing for 2 mm diameter nerve) [16] was scaled down to fit the target nerve diameters for the
microfabricated cuff (3 mm spacing for 0.5-0.8 mm diameter nerve; 4 mm spacing for 0.7-1.0 mm
diameter nerve).
The surface area of the electrodes was dictated by charge injection limits and desired stimulation
parameters. Typical stimulation parameters for vagus nerve stimulation (VNS) are at a wide range of
currents (0.5 to 4.2 mA) and pulse widths (100 to 1000 µs), with maximum charge per phase for clinical
use in published data ranging from 0.42 to 1.00 µC [16-19]. The charge injection capacity (CIC) for
platinum thin films varies in the range of 170 to 260 µC/cm2
(depending on the electrode size and pulse
width used, with smaller electrodes and longer pulse widths resulting in higher CIC) and the safe charge
injection limit for tissue varies in the range of 50 to 320 µC/cm2
(depending on electrode area) [20-24].
The CIC of electrodes can also be increased 2- to 8-fold using electrode coatings such as platinum/iridium
(PtIr), iridium oxide (IrOx), or poly(3,4-ethylenedioxythiophene):polystyrene sulfonate (PEDOT:PSS) if
bare platinum films do not produce sufficient CIC [24,25]. Based on these values from literature and

140
internal recommendations, an electrode CIC of 100 μC/cm2 was selected as a target value and used to
calculate the required electrode surface area of 1.0 mm2
(charge per phase (1.00 μC)/CIC (100 μC/cm2
)).
A safety factor of 1.5 was chosen to ensure safe stimulation of tissue and prevent damage to the
electrodes, resulting in a required electrode surface area of 1.5 mm2
.
Rather than using one large rectangular electrode, an array design consisting of closely packed,
shorted microelectrodes was selected (Figure 4-4). Large, open metal areas on thin film devices are prone
to cracking and delamination during stimulation, handling, and prolonged soaking. Using an array of
connected, circular microelectrodes (an aggregate electrode) reduces the probability of cracking or
delamination by reducing the size of the metal area and adding open areas in the thin film metal where the
top and bottom Parylene layers can adhere to each other. Circular microelectrode openings were used to
prevent stress or current concentration in any areas (as occurs in the corners of square microelectrodes).
Based on the required electrode area of 1.5 mm2
and a desired microelectrode size of 200 µm (which is
commonly used in Parylene thin film devices), at least 48 interconnected microelectrodes were used to
form each rectangular aggregate electrode.
Figure 4-4: Schematic of a 6x18 aggregate electrode.
Cuff electrodes can wrap around the nerve in either a helical or circumferential configuration. For
helical electrodes, the entire surface area of the electrode is in contact with the nerve as the cuff does not

141
overlap itself. The helix angle, electrode width, and electrode height were selected in combination to
produce sufficient surface area, to ensure the cuff is wrapped ~1.5× around the nerve, and to prevent
overlap of the electrode when implanted. For circumferential electrodes, the cuff overlaps itself, with only
a portion of the electrode in contact with the nerve. The electrode width and height were selected in
combination to produce sufficient surface area of the electrode at the smallest nerve size (where the cuff
overlaps itself). Electrode dimensions for each electrode type and cuff size are included in Table 4-1.
Table 4-1: Microelectrode grid size and electrode surface area for each electrode type and cuff size.
Electrode Type Cuff Size Microelectrode
Grid Size
Electrode Surface
Area (total)
Electrode Surface
Area (minimum
nerve contacting)
Helical
Small
(0.5-0.8 mm) 3 x 18 1.7 mm2 1.7 mm2
Large
(0.7-1.0 mm) 4 x 22 2.8 mm2 2.8 mm2
Circumferential
Small
(0.5-0.8 mm) 9 x 14 4.0 mm2 1.7 mm2
Large
(0.7-1.0 mm) 6 x 18 3.4 mm2 1.6 mm2
The minimum allowable trace width (to prevent overheating) was dictated by maximum current
requirements and calculated using a printed circuit board (PCB) trace width calculator with formulas
sourced from IPC-2221 (a set of standards for PCB design which includes calculations for required trace
width to prevent overheating of the device). A total minimum trace width of 72 µm was calculated (from
parameters: 4.2 mA, 200 nm trace thickness, 10 °C rise at 37 °C ambient temperature). Four connected,
parallel 36 µm traces with intermittent bridges between them (see Figure 4-5) were used to maintain
device flexibility and add in a safety factor of 2, still with sufficient width if two traces are broken.

142
Figure 4-5: Schematic of the bridged traces.
4.2.2.2 Mechanical Design
The Parylene backbone is the structural component of the device which supports the electrodes. It
is microfabricated in a flat configuration on a carrier wafer and then thermoformed into the final cuff
shape. There are a variety of options for backbone shape, but no clear advantages or disadvantages
surgically or electrochemically between the many options. To determine a best possible backbone shape,
several designs were fabricated as sham devices (bare Parylene with no metal layer) and evaluated for a
variety of different considerations (such as fabrication, surgery, and handling). The methods and results
for these tests are included in appendix E.
Based on the design evaluation results, two configurations of the Parylene cuff were selected: the
full cuff (backbone E) and the tilted, I-shaped, helical cuff (backbone B), both with the u-turned tail (tail
3) and no additional features. Each cuff configuration was designed in two sizes and incorporated all
electrical design features (electrode surface area, spacing, and trace width) and design requirements
(thermoformed diameter and self-sizing feature). Schematics of each cuff are shown in Figure 4-6 through
Figure 4-9.

143
Figure 4-6: Schematic of the large, helical cuff electrode (design B3-).
Figure 4-7: Schematic of the small, helical cuff electrode (design B3-).

144
Figure 4-8: Schematic of the large, full cuff electrode (design E3-).
Figure 4-9: Schematic of the small, full cuff electrode (design E3-).
4.3 Experimental Methods
Devices were fabricated and tested using the steps outlined below. The electrode arrays were built
using thin film fabrication methods, then packaged using further processing steps to allow for integration
with benchtop electrochemical testing equipment, stimulation equipment for acute in vivo studies, and the
OpenNerve IPG for chronic use. All fabrication methods and design documents are also available online
as part of the open source OpenNerve documentation set [26].

145
4.3.1 Thin Film Fabrication
The Parylene electrodes were fabricated using a low temperature, batch process based on prior
work [27], which is summarized here and in Figure 3-14 and described in detail in appendix C. The
Parylene C base layer was deposited on a 4” prime silicon wafer (used as the carrier substrate) using a
chemical vapor deposition-like process (PDS 2010 Labcoter, Specialty Coating Systems, Indianapolis,
IN) to a thickness of 5 µm. The coated wafers were annealed in an oven (TVO-2, Cascade Tek Inc.,
Longmont, CO) under vacuum with nitrogen flow at 150 °C for 4 hours. Next, lithography was performed
to add a lift-off photoresist mask (AZ 5214 IR, AZ Electronic Materials, Branchburg, NJ; 1.8 µm thick).
Then, a metal stackup (either 20 nm titanium, 25 nm platinum, 150 nm gold, and 25 nm platinum or 20
nm titanium, 155 nm gold, and 25 nm platinum) was deposited via e-beam evaporation. Electrodes,
traces, and bondpads were formed following lift-off in acetone or NMP rinse at 60 °C followed by rinsing
in isopropyl alcohol and deioinized water. After metal patterning, a silane treatment was performed
(1:100:100 A-174:DI water:IPA), and a top layer of Parylene C was deposited to a thickness of 7 µm.
Openings in the Parylene for metal electrodes and bondpads on the top side of the device were etched
using O2 switched chemistry etching in a deep reactive ion etcher (PlasmaPro 80 or PlasmaLab 100 ICP,
Oxford Instruments, Bristol, UK; switched chemistry process parameters detailed in [28]; etch rate
approximately 0.08 µm/loop) masked by patterned photoresist (AZ P4620, AZ Electronic Materials,
Branchburg, NJ; 8-12 µm thick), and the outline of the device was cut out using the same etching
procedure. Photoresist was removed using acetone after each etching step followed by rinsing in
isopropyl alcohol and deionized water. Some wafers were subsequently cleaned in O2 plasma
(CV200RFS, Yield Engineering Systems, Fremont, CA). Devices were released from the wafer by
placing a droplet of water over one of the handling tabs, manually separating the tab from the wafer using
a scalpel or sharp tweezers, and gently pulling the device off the wafer using the tab as a handling point.

146
Figure 4-10: Parylene electrode thin film fabrication process flow, with cross-sectional views (left) at the indicated point on the
B3-small device shown with the red dotted arrows (right). (A) Parylene was deposited on a silicon carrier wafer. (B) Metal
electrodes, traces, and bondpads (not shown) were deposited and patterned. (C) Parylene was deposited on top of patterned metal.
(D) Electrodes and bondpads (not shown) were opened via O2 etching. (E) The electrode arrays were released.
After release from the wafer, parts were visually inspected using a stereoscope (HD60T, Caltex
Scientific, Irvine, CA) or high magnification microscope (Eclipse LV100, Nikon, Tokyo, Japan) for
defects such as torn Parylene, delaminated metal, shorted traces, or cracked traces.
4.3.2 Lead Preparation and Attach
36 gauge stranded copper wire with silicone insulation (McMaster Carr, Santa Fe Springs, CA)
was cut to length (100 cm for lifetime testing, 30 cm for animal studies, 10 cm for other benchtop
characterization) and stripped of 2-4 mm of insulation on both ends. One end of each wire was electrically
connected to a male header pin using either solder or silver epoxy (MED-H20E, EpoTEK, Billerica, MA)
and protected using heat shrink. Two wires were bundled and threaded through a 15 cm length of 1/16”
inner diameter, 1/8” outer diameter silicone tubing (McMaster Carr, Santa Fe Springs, CA) prior to
attaching to the Parylene electrode.

147
Parylene electrodes were mounted flat on a Teflon plate using Kapton tape with the exposed
metal bondpads facing upward. Tape was placed on the handling tabs to prevent damage to the device.
The free end of each wire in the bundle was trimmed to a length of 1-2 mm using scissors. The exposed
tip of each wire was aligned to one bondpad and held in place using Kapton tape. Conductive silver epoxy
(MED H20E, EpoTEK, Billerica, MA) was painted over each wire and bondpad by hand using a pick,
needle, or rigid wire, taking care not to short the two bondpads or wires together. After both bondpads
were attached, the epoxy was cured in a convection oven for at least 3.5 hours at 85 °C. A fixtured device
with wires attached is shown in Figure 4-11.
Figure 4-11: Device setup for the wire attach process.
After removing the electrode array from the oven, the tab at the bondpad end of the device was
cut off using a scalpel or scissors and a small piece of Teflon film (25-50 µm thick) was slid underneath
the bondpad region. A thin, insulating epoxy (MED-301, EpoTEK, Billerica, MA) was prepared per
manufacturer recommendations and applied over all bondpad areas, coating all silver epoxy, exposed
metal on the bondpads, and exposed metal on the wires to fully insulate the wire connections. The Teflon
film acted as a barrier to prevent any epoxy that wicked underneath the bondpad area from adhering to the
Teflon plate. The epoxy was cured in a convection oven for at least 2.5 hours at 65 °C or at least 1.5 hours
at 85 °C, then a second coat of epoxy was added and cured using the same procedure.
After insulation was cured, devices were removed from the Teflon plate by loosening the Kapton
tape with isopropyl alcohol and removing it with tweezers. If the epoxy wicked to the back of the device
and adhered to the Teflon film, the film was slowly peeled away from the device.

148
Prior to proceeding with the remainder of the packaging procedure, a shorting test was performed
to identify if the two electrodes were electrically isolated (procedure in section 4.3.6.1), followed by
cyclic voltammetry (CV) in 0.05 M sulfuric acid (H2SO4) to clean the electrode surface and to determine
if either electrode was broken (procedure in section 4.3.6.2).
4.3.3 Thermoforming
The Parylene electrodes were fixtured into a cuff shape manually and thermoformed to hold their
shape, with the process summarized here and described in detail in appendix F. A winding fixture was
assembled by attaching a 2 cm wide Teflon film (25 µm thick) to a stainless steel dispensing tip (2”
length, 0.4 or 0.5 mm diameter for small and large cuffs) using Kapton tape. The fixture was clipped to a
glass slide, and the Parylene electrodes were placed on top of the Teflon film. The Parylene electrodes
and Teflon film were wrapped around the dispensing tip (used as a mandrel) and clamped to prevent
unwinding (shown in Figure 4-12).
Figure 4-12: (Left) device fixtured for thermoforming, with (right) zoomed in view of Parylene electrode region wrapped around
the mandrel and held in place with a Teflon film. More detailed process images are included in appendix F.
A subset of parts used for accelerated lifetime testing were clamped flat rather than curling into a
cuff shape. These parts were used for visual inspection during lifetime testing, as it was difficult to
visually inspect the inside of thermoformed cuff electrodes.

149
After fixturing, the parts were placed into a programmable vacuum oven (TVO-2, Cascade Tek
Inc., Longmont, CO), placed under vacuum, then purged three times with nitrogen to minimize oxygen in
the chamber. The oven was programmed to ramp up to 150 °C at a ramp rate of approximately 0.7
°C/min, hold for 12 hours, then ramp down to room temperature. After cooling and removing from the
fixture, the electrodes were visually inspected to ensure they held the desired cuff shape and to identify
any damage.
4.3.4 Overmolding
The lead to Parylene attach and Parylene cable regions of the device were coated in a layer of
silicone elastomer to produce a more robust device that is easier to handle surgically and to protect the
device from fluid ingress, with the process summarized here and described in detail in appendix H. To
prevent cracking in the long, cable region, small pieces of Kapton tape were applied to the front and back
side of the Parylene cable starting near the lead attachment point and extending 10-15 mm up the cable. A
thin layer of silicone elastomer was painted onto the lead attachment region and Parylene cable 2-5 mm
beyond the taped area and cured at 65 °C for at least four hours or at room temperature overnight (Figure
4-13, left). For early prototyping, benchtop testing, and acute in vivo studies, Sylgard 184 (Dow, Midland,
MI) was used; for future chronic in vivo studies, a medical grade silicone with similar properties will be
selected, such as Nusil MED-6215 (Avantor, Radnor Township, PA). After the painted layer was cured,
both sides of a custom mold were filled with mixed elastomer and degassed under vacuum. Once no
bubbles were visible, the device was loaded into the base of the mold and the thermoformed electrode
region was slid into a dispensing tip with inner diameter matching the outer diameter of the cuff (to
prevent loosening of the cuff while curing the silicone at elevated temperature; Figure 4-13, center). The
top half of the mold was placed on the base, using the pins to assure proper alignment, and slowly closed
using minimal pressure (Figure 4-13, right). The mold was placed into an oven at 65 °C for at least 4
hours to cure. After cooling, the electrode region was removed from the dispensing tip, the part was
gently removed from the mold, and all silicone flash was removed with tweezers and a scalpel. Any large

150
bubbles near the lead attach region were incised with a scalpel, filled with silicone elastomer, and cured
using the same cure schedule.
Figure 4-13: (Left) the Parylene cable reinforced with Kapton tape and hand painted PDMS, with the thermoformed electrode
region placed inside a dispensing tip to prevent loosening during elevated temperature curing. (Center) the base and top of the
mold filled with liquid PDMS, with the device loaded into the base of the mold. (Right) the closed mold.
4.3.5 Electrode Coating (Optional)
An optional electrode coating was electrodeposited to increase the charge injection capacity and
decrease the impedance of the electrodes. PtIr coatings were deposited by a collaborator (EPIC Medical,
Inc., Pasadena, CA).
4.3.6 Benchtop Testing
After the device was fully fabricated, a series of benchtop tests were performed to evaluate the
electrochemical performance of the device and identify any failed devices. The sequence of tests and
resulting failure criteria are summarized in Table 4-2 and details on each test are provided in the sections
below.

151
Table 4-2: Benchtop tests performed on fully fabricated devices and failure criteria resulting from each test.
Test Solution Failure Criteria Section
1 Shorting Test None (air) |Z| < 100 kΩ
ΦZ < -65°
4.3.6.1
2 CV H2SO4 ESA < 1 mm2
CSC < 100 µC/cm2
Uncharacteristic shape
4.3.6.2
3 (optional) CV Ferri/Ferrocyanide ESA < 1 mm2 4.3.6.2
4 CV 1× PBS Uncharacteristic shape 4.3.6.2
5 EIS 1× PBS |Z| (1 kHz) > 1 kΩ
Uncharacteristic shape
4.3.6.3
6 VT 1× PBS CIC < 50 µC/cm2 4.3.6.4
CV = cyclic voltammetry, EIS = electrochemical impedance spectroscopy, VT = voltage transient,
H2SO4 = sulfuric acid, PBS = phosphate buffered saline, Z = impedance, ESA = electroactive surface
area, CSC = charge storage capacity, CIC = charge injection capacity
4.3.6.1 Shorting Testing
A shorting test was performed to determine if the two electrodes were electrically isolated by
measuring the impedance between the two electrode wires using an LCR meter (E4980A, Agilent
Technologies, Santa Clara, CA) when the device was dry. The measurement was performed using a 10
kHz, 20 mV signal; devices with impedance magnitude less than 100 kΩ and impedance phase less than
‑65° were considered shorted.
4.3.6.2 Cyclic Voltammetry (CV)
CV was performed using a Reference 600 potentiostat (Gamry Instruments, Warminster, PA)
using a variety of parameters (see Table 4-3). To wet the device, the electrode region was submerged in
IPA for at least 30 seconds, then submerged in DI water for at least 1 minute, and finally submerged in
the test solution. A three-electrode setup was used, with each electrode on the cuff serving as a working
electrode, a platinum wire counter electrode, and a silver/silver chloride (3 M NaCl) reference electrode
(BASi, West Lafayette, IN).
Table 4-3: CV Parameters used for benchtop testing in each solution.
Solution Scan Window Scan Rate Number of Cycles
H2SO4 -0.2 to 1.2 V 250 mV/s 30
Ferri-ferrocyanide -0.4 to 1.0 V 20, 50, 80, and 100 mV/s 1
PBS -0.6 to 0.8 V 50 and 250 mV/s 3

152
When performed in H2SO4 (250 mV/s scan rate, 30 cycles), CV acts to clean the electrode surface
and provides insight on the performance of the electrode. Functional electrodes have a characteristic
shape unique to platinum (see Figure 4-14), and from that curve the electroactive surface area (ESA) and
cathodic charge storage capacity (CSC) were calculated. For all calculations, the last 5 CV cycles were
averaged. ESA was calculated by comparing the time integral of the hydrogen desorption current
(hydrogen desorption charge, 𝑄𝐻, depicted in Figure 4-14) during the average CV cycle to the
characteristic charge density of a monolayer of hydrogen atoms adsorbed to polycrystalline platinum
(𝜌𝐻 = 210 𝜇𝐶
𝑐𝑚2
) as described in equation 4-1. The CSC was calculated by dividing the time integral of
cathodic current (the cathodic charge, 𝑄𝑐𝑎𝑡ℎ𝑜𝑑𝑖𝑐, depicted in Figure 4-14) by the geometric surface area
(GSA) as described in equation 4-2. Electrodes that were not connected (identified via noise in the
nano‑amp range during CV testing) were considered non-functional and did not proceed with the
remaining fabrication steps. Electrodes that did not have the characteristic shape, had ESA less than 1
mm2
, or had CSC less than 100 µC/cm2 proceeded with fabrication but were not used for lifetime testing
as they were considered sub-standard.
𝐸𝑆𝐴 =
QH
𝜌𝐻
(4-1)
𝐶𝑆𝐶 =
Q𝑐𝑎𝑡ℎ𝑜𝑑𝑖𝑐
𝐺𝑆𝐴
(4-2)

153
Figure 4-14: Example CV data in H2SO4 for a functional electrode with characteristic peaks labeled and the hydrogen desorption
and cathodic charges shaded in blue and purple, respectively.
When performed in ferri/ferrocyanide (20, 50, 80, and 100 mV/s scan rate, 1 cycle each), CV can
be used to more accurately calculate the ESA using the Randles Sevcik relationship between peak current
and CV scan rate (equation 4-3). CV in ferri/ferrocyanide was only performed on a subset of parts to
determine the accuracy of the H2SO4 CV calculations. Any parts intended for in vivo testing were not
tested in ferri/ferrocyanide.
𝑖𝑝 = 0.4463𝑛𝐹𝐶√
𝑛𝐹𝑣𝐷
𝑅𝑇
× (𝐸𝑆𝐴) (4-3)
𝑤ℎ𝑒𝑟𝑒
{

𝑖𝑝 = 𝑝𝑒𝑎𝑘 𝑐𝑢𝑟𝑟𝑒𝑛𝑡 (𝐴)
𝑣 = 𝑠𝑐𝑎𝑛 𝑟𝑎𝑡𝑒 (
𝑚𝑉
𝑠
)
𝑛 = 𝑒𝑙𝑒𝑐𝑡𝑟𝑜𝑛𝑠 𝑡𝑟𝑎𝑛𝑠𝑓𝑒𝑟𝑟𝑒𝑑 𝑖𝑛 𝑟𝑒𝑑𝑜𝑥 𝑒𝑣𝑒𝑛𝑡 (1)
𝐹 = 𝐹𝑎𝑟𝑎𝑑𝑎𝑦 𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡 (9.6485 × 104
𝐶
𝑚𝑜𝑙)
𝐶 = 𝑠𝑜𝑙𝑢𝑡𝑖𝑜𝑛 𝑐𝑜𝑛𝑐𝑒𝑛𝑡𝑟𝑎𝑡𝑖𝑜𝑛 (5 𝑚𝑀)
𝐷 = 𝑑𝑖𝑓𝑓𝑢𝑠𝑖𝑜𝑛 𝑐𝑜𝑒𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑡 (6.56 × 10−6
𝑐𝑚2
𝑠
)
𝑅 = 𝑔𝑎𝑠 𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡 (8.3144
𝐽
𝑚𝑜𝑙 𝐾
)
𝑇 = 𝑡𝑒𝑚𝑝𝑒𝑟𝑎𝑡𝑢𝑟𝑒 (293 𝐾)

154
CV in phosphate buffered saline (PBS; 50 mV/s scan rate, 3 cycles) was used prior to and during
accelerated lifetime testing (described in section 4.3.7) to qualitatively identify any changes in electrical
properties or surface area of the electrodes (via curve shape and area inside the curve, respectively).
Functional devices had an ESA of at least 1 mm2
and a CSC of at least 100 µC/cm2
for each
electrode in the cuff and followed the characteristic CV shapes. Non-functional electrodes, such as those
with broken traces or with detached lead wires had significantly lower ESA and CSC values and CV
curves not matching the characteristic shapes.
4.3.6.3 Electrochemical Impedance Spectroscopy
Electrochemical impedance spectroscopy (EIS) was performed using a Reference 600 potentiostat
(Gamry Instruments, Warminster, PA) from 1 Hz to 1 MHz at 2.5 mV RMS in 1× PBS. A three-electrode
setup was used, with each electrode on the cuff serving as a working electrode, a platinum wire counter
electrode, and a silver/silver chloride (3 M NaCl) reference electrode (BASi, West Lafayette, IN).
Functional electrodes had 1 kHz impedance values less than 1 kΩ which leveled out at approximately 125
to 500 Ω with phase values from 0 to -20° at high frequency (above 10 kHz). Non-functional electrodes,
such as those with broken traces or with detached lead wires had constantly rising impedance (from high
to low frequency) with maximum values in the giga-ohm range and phase near -90° at all frequencies.
4.3.6.4 Voltage Transient (VT) Testing
Voltage transient (VT) testing was performed using a Reference 600 potentiostat (Gamry
Instruments, Warminster, PA) by injecting a cathodic first, balanced, biphasic square wave through the
electrodes on the device and measuring the resulting voltage transient. A three-electrode setup was used,
with the electrodes on the cuff serving as the working electrode, a platinum wire counter electrode, and a
silver/silver chloride (3 M NaCl) reference electrode (BASi, West Lafayette, IN). Two pulse widths, 200
and 500 µs, were tested with interphase gaps of 80 and 100 µs, respectively. A single pulse was injected
with a starting current of 1 mA, and the interphase potential (Ep, the difference between the potential prior
to the pulse and at the end of the interphase gap – see Figure 4-15) was calculated. Current was gradually
increased until Ep reached or exceeded -0.6 V (the edge of the water window). The precise injected

155
current to reach Ep = -0.6 V was calculated using the linear relationship between injected current and Ep.
The CIC was calculated by dividing the required injected current to reach Ep of -0.6 V by the geometric
surface area of the electrode, as described in equation 4-3. Functional devices had a CIC of at least 50
µC/cm2
.
𝐶𝐼𝐶 =
𝐼×𝑃𝑊
𝐺𝑆𝐴
(4-3)
𝑤ℎ𝑒𝑟𝑒 {
𝐼 = 𝑖𝑛𝑗𝑒𝑐𝑡𝑒𝑑 𝑐𝑢𝑟𝑟𝑒𝑛𝑡 𝑡𝑜 𝑟𝑒𝑎𝑐ℎ 𝐸𝑝 𝑙𝑖𝑚𝑖𝑡
𝑃𝑊 = 𝑝𝑢𝑙𝑠𝑒 𝑤𝑖𝑑𝑡ℎ
𝐺𝑆𝐴 = 𝑔𝑒𝑜𝑚𝑒𝑡𝑟𝑖𝑐 𝑠𝑢𝑟𝑓𝑎𝑐𝑒 𝑎𝑟𝑒𝑎
Figure 4-15: Example input current pulse (blue; 1 mA, 500 µs pulse width, 100 µs interphase delay) and resulting voltage
transient (red), illustrating the interphase potential (Ep).
4.3.7 Accelerated Lifetime Testing
Accelerated lifetime testing was performed on a subset of parts to measure the electrochemical
performance of the device over time, determine device longevity, and evaluate chronic device failure
modes. Prior to testing devices in the final configuration, flat devices with bare platinum electrodes and
with PtIr-coated electrodes were tested to evaluate performance differences and determine if an electrode
coating would be required for chronic stimulation. To evaluate these bare and coated devices and different
potential failure modes, several groups of devices were tested. Sample groups were split between bare and
coated electrodes, all in the large, full cuff configuration without silicone overmolding. Two groups of
stimulated (bipolar current stimulation between the two electrodes on the cuff) and unstimulated devices

156
were used to isolate failure modes related to stimulation from those inherent to the device. Future tests on
thermoformed and overmolded devices are planned after more coating data is available.
4.3.7.1 Device Fixturing
Five holes were drilled into the lid of a glass jar. Devices were secured into the lid through one of
the holes using marine epoxy such that the electrodes would be fully submerged when the jar was closed
and filled with PBS. The three remaining holes were sized to accommodate the counter and reference
electrodes and N2 purging tube. The electrode end of fully fabricated devices was wetted using the same
procedure used for benchtop testing (submerged in IPA for at least 30 seconds, DI water for at least 1
minute), then the jar lid was screwed onto a glass jar filled with 1× PBS such that the cuff and electrodes
were fully submerged. If needed, a kinked wire was threaded through the tab on the distal end of the
electrode region to hold it submerged. In some cases, a fifth hole was used to accommodate a second
device in a single jar (required to fit all devices in a single water bath for temperature control – see Figure
4-16).
Figure 4-16: Device fixtured into the glass jar filled with saline (1x PBS) for accelerated life testing.
4.3.7.2 Electrochemical Testing
For each device, a counter and reference electrode and the N2 purging tube were inserted into the
PBS through the holes in the jar lid and the PBS was N2 purged. CV (section 4.3.6.2), EIS (section

157
4.3.6.3), and VT (only with 500 µs pulse width, section 4.3.6.4) testing were performed for each electrode
in PBS. After testing, the counter and reference electrodes and N2 purging tube were removed from the
solution and the holes in the lid were closed using rubber stoppers. The CV and EIS curves were
evaluated for shape, and the area inside the CV curve, the 1 kHz impedance magnitude, and the CIC were
calculated.
4.3.7.3 Lifetime Testing
Following initial testing, the jars were placed into a circulating water bath maintained at 50 °C.
Per the Arrhenius equation (4-4), this equates to an acceleration factor of 2.46×. An accelerated
temperature of 50 °C was selected as it allows for a sizeable acceleration while remaining below the glass
transition temperature of Parylene. Using a temperature at or above the glass transition temperature could
introduce failure modes which would not be seen in vivo and are thus not applicable to the intended
device use.
𝑡𝑎𝑐𝑐𝑒𝑙 = 𝑡𝑟𝑒𝑎𝑙 × 2
𝑇𝑎𝑐𝑐𝑒𝑙−𝑇𝑏𝑜𝑑𝑦
10 (4-4)
𝑤ℎ𝑒𝑟𝑒
{

𝑡𝑎𝑐𝑐𝑒𝑙 = 𝑎𝑐𝑐𝑒𝑙𝑒𝑟𝑎𝑡𝑒𝑑 𝑡𝑖𝑚𝑒
𝑡𝑟𝑒𝑎𝑙 = 𝑟𝑒𝑎𝑙 𝑡𝑖𝑚𝑒
𝑇𝑎𝑐𝑐𝑒𝑙 = 𝑎𝑐𝑐𝑒𝑙𝑒𝑟𝑎𝑡𝑒𝑑 𝑡𝑒𝑚𝑝𝑒𝑟𝑎𝑡𝑢𝑟𝑒
𝑇𝑏𝑜𝑑𝑦 = 𝑏𝑜𝑑𝑦 𝑡𝑒𝑚𝑝𝑒𝑟𝑎𝑡𝑢𝑟𝑒
Parts in the stimulated group were connected to a custom stimulator board which delivered a
cathodic first, balanced biphasic square wave pulse. Stimulation parameters were selected to achieve a
charge per phase typically used for VNS (0.75 µC), resulting in a charge density of 22.1 µC/cm2
for large,
full cuffs, which was safely below the CIC of the electrodes. Due to current and frequency limitations of
the stimulator board, an amplitude of 0.5 mA was used with a pulse width of 1500 µs (resulting in 0.75
µC charge per phase) at a pulsing frequency of 50 Hz. This equates to 1.76 million pulses per accelerated
day, or 641 million pulses within one year of accelerated time.
Parts were re-tested periodically throughout the accelerated life test to monitor the
electrochemical characteristics of the electrodes. To re-test devices, the stimulator was disconnected and
the devices were removed from the heated water bath. All electrochemical tests (described in section

158
4.3.7.2) were repeated. Devices which had disconnected electrodes (identified by 1 kHz impedance in the
MΩ range and CV magnitudes in the nA range) were considered failed and did not continue in lifetime
testing. Devices with delaminated metal which was still electrically connected (identified by a decrease in
1 kHz impedance and increase in area inside the CV curve as compared to baseline values) were noted
and continued in lifetime testing until disconnection failure. Devices with degradation in CV curves not
associated with delamination or disconnection continued in lifetime testing until CV shape stabilized.
After electrochemical testing, devices were returned to the heated water bath and the stimulator was
reconnected (for stimulated parts). Parts were tested every 12 hours for 2 days, daily for two weeks,
weekly for two months, then monthly until failure (with all time points listed in accelerated time).
4.3.8 In Vivo Testing
Early acute in vivo studies were performed to demonstrate feasibility of the surgical implantation
process. The cuff electrodes were implanted onto the sciatic nerve of Sprague Dawley rats. Although this
device was designed primarily to target vagus nerve branches, rat sciatic nerve was selected as a
representative study organ to allow the use of smaller animals. The sciatic nerve in the rat is easily
accessible, appropriately sized (~1 mm in diameter), and has well characterized electrophysiology. Each
animal implantation followed the procedure described here. The animal was anesthetized and placed in a
prone position. A midsagittal incision was made at the thigh and the muscles were pulled apart at the
junction using blunt dissection. The sciatic nerve was exposed and the connective tissue surrounding the
sciatic nerve was dissected, exposing approximately 1 cm of nerve. The nerve was gently lifted away
from surrounding tissue using a retractor and the cuff was placed on the nerve using a custom surgical
implantation tool (shown in Figure 4-17). After placement of the device, the retractor was removed.
Figure 4-17: Use of the sliding surgical placement tool, in which the cuff is (left) loaded on the tool and placed over the nerve,
(center) slid off the tool and onto the nerve, and (right) closes softly over the nerve.

159
Future experiments with active devices are planned in which the sciatic nerve will be stimulated
using a variety of stimulation parameters, starting with a pulse width of 500 µs (with 100 µs interphase
gap), frequency of 30 Hz (known to obtain tetanic muscle contraction), and increasing amplitude starting
from 100 µA until full recruitment is reached. Nerve recruitment will be determined by visually observing
muscle contraction in the leg and by recording compound action potentials (CAPs) on a commercial cuff
electrode.
4.4 Experimental Results
4.4.1 Microfabrication, Thermoforming, and Packaging
Two batches of thin film cuff electrodes were successfully fabricated, with the two device
configurations (full cuff and helical cuff), each in two sizes, pictured in Figure 4-18 in their flat, postfabricated form. Thin film electrodes were thermoformed and molded, producing functional assembled
devices, with each device configuration pictured in Figure 4-19. The first batch of parts (batch S) were
fabricated with the four-layer metal stack (20 nm Ti, 25 nm Pt, 150 nm Au, and 25 nm Pt), while the
second batch of parts (batch U) used the three-layer metal stack (20 nm Ti, 155 nm Au, and 25 nm Pt).
Batch S was used for early testing and electrochemical characterization, while batch U was used to
produce parts for accelerated lifetime testing and in vivo testing.
Figure 4-18: Photos of the fabricated Parylene devices in (left) full and (right) helical cuff configurations.

160
Figure 4-19: Fabricated and packaged cuff electrodes in the (left) full cuff and (right) helical cuff configurations, with (bottom)
zoomed in images of the thermoformed electrode regions, with exposed metal electrodes on the inner surface of the cuff.
4.4.2 Electrochemical Characterization
Electrochemical results were fairly consistent between functional parts (parts which did not fail
the testing criteria listed previously), with some batch to batch variation. CV curves in H2SO4 for batch S
had relatively higher magnitudes of hydrogen desorption and adsorption peaks than batch U, as shown in
Figure 4-20, which indicates a higher ESA in batch S devices (average ESA for small, full cuffs prior to
thermoforming was 6.6 mm2
in batch S and 3.5 mm2
in batch U; GSA was 4.0 mm2
). Although the
devices from each batch were the same geometry and had the same exposed metal material (platinum),
batch S devices were likely over-etched during the first Parylene etching step (exposing the metal
electrodes using DRIE) and batch U devices had an additional oxygen plasma cleaning step (after the
final Parylene etching step). Either of these steps could cause roughening of the metal electrode surface
and a resulting increase in ESA or changes in the electrochemical properties of the metal due to plasma
exposure [29].

161
Figure 4-20: CV curves in H2SO4 for batch S (blue) and batch U (black) small, full cuffs prior to thermoforming. Batch S devices
have larger H+ desorption and adsorption peaks (range of 12 to 24 µA and -47 to -28 µA, respectively) as compared to batch U
devices (5 to 16 µA and -33 to -16 µA, respectively).
CV curves in H2SO4 also had minor differences in shape and peak magnitude between flat and
thermoformed parts in both full and helical cuffs (Figure 4-21). In full cuffs, this was expected, as the
electrodes overlapped themselves in thermoformed parts, resulting in a smaller exposed GSA. In helical
cuffs, however, there are no overlapping electrode regions and the exposed GSA is identical between flat
and thermoformed parts. Changes in the CV curve indicate changes in the electrode-electrolyte interface
during or after the thermoforming process. No material is added or removed during this step and
oxidation is not expected (or observed in the CV) because thermoforming is performed in the absence of
oxygen. One potential cause of these changes is the availability of solution around the electrode during
CV testing. In the flat configuration, solution can freely flow around the electrode. In the thermoformed
configuration, the solution may be more constrained inside the cuff, limiting the charge transfer at the
electrode-electrolyte interface.

162
Figure 4-21: CV curves in H2SO4 at 250 mV/s for a representative small, helical cuff before (flat device – solid line) and after
(dashed line) thermoforming. Peaks are higher magnitude and more clearly defined before thermoforming.
Due to inconsistencies in the CV curves in H2SO4, CV in ferri/ferrocyanide was used as a more
accurate way to characterize ESA for flat and thermoformed parts. Minor changes in peak magnitudes
before and after thermoforming indicate that the ESA of the electrode is not changing during
thermoforming, as expected. Figure 4-22 shows CV curves in ferri/ferrocyanide for one representative
device (the same device with CV curves in H2SO4 illustrated in Figure 4-21), and Table 4-4 shows
calculated ESA differences from H2SO4 and ferri/ferrocyanide calculations for all devices which were
tested in both solutions. Calculations are similar in flat devices and an average of 1.5 times higher in
thermoformed devices when measured in ferri/ferrocyanide versus H2SO4. Although ferri/ferrocyanide
was found to be a more accurate calculation, it was not used to test all devices as it is toxic and cannot be
used on in vivo parts. While ESA calculations in H2SO4 were not as accurate after thermoforming, they
were consistently lower than calculations in ferri/ferrocyanide, so parts that passed the design criteria in
H2SO4 were assumed to be able to pass in ferri/ferrocyanide. ESA calculations prior to thermoforming
were considered to be more accurate.

163
Figure 4-22: CV curves in ferri/ferrocyanide for a representative small, helical cuff (same device as illustrated in Figure 4-21)
before (flat – solid lines) and after (dashed lines) thermoforming. Peak magnitudes are similar before and after thermoforming,
indicating similar ESA.
Table 4-4: Average ESA (± standard deviation) for cuff electrodes from batch S in the flat and thermoformed (TF) configurations
calculated using CV in ferri/ferrocyanide (FF) and H2SO4. The right column shows the ESA in ferri/ferrocyanide divided by the
ESA in H2SO4, illustrating the differences in calculations between flat and thermoformed configurations.
Device Configuration GSA ESAFF ESAH2SO4 ESAFF / ESAH2SO4
Helical,
Small
Flat 1.7 3.3 ± 0.2 mm2 3.0 ± 0.3 mm2 1.1 ± 0.1
TF 1.7 2.8 ± 0.3 mm2 2.1 ± 0.4 mm2 1.4 ± 0.2
Full,
Small
Flat 4.0 7.6 ± 0.3 mm2 6.6 ± 0.5 mm2 1.2 ± 0.1
TF 1.7 5.7 ± 0.5 mm2 4.3 ± 0.8 mm2 1.4 ± 0.3
All
Devices†
Flat -- -- -- 1.1 ± 0.1
TF -- -- -- 1.5 ± 0.4
TF = Thermoformed, GSA = Geometric Surface Area, ESA = Electroactive Surface Area,
FF =Ferri/Ferrocyanide
† Average ESA not reported for all devices because GSA varies; ESAFF/ESAH2SO4 includes large
devices, which had insufficient sample size to be reported individually
After optimizing the electrochemical testing based on the results shown above, all parts in batch
U were tested using the full testing protocol listed in section 4.3.6, with results summarized here.
Representative CV curves in H2SO4 for each electrode type prior to thermoforming are shown in Figure
4-23. Representative CV, EIS, and VT data in PBS for the same devices are shown in Figure 4-24, Figure
4-25, and Figure 4-26 respectively. For all device configurations, ESA (calculated in H2SO4) was slightly

164
lower than GSA (as discussed earlier in this section). CSC exceeded the goal value of 100 µC/cm2 by
more than an order of magnitude, and CV curves showed the expected hydrogen and oxide peaks
characteristic of platinum. CIC was above the failure criteria of 50 µC/cm2
for all configurations at 200
and 500 µs pulse width and above the goal value of 100 µC/cm2
for full cuffs at both pulse widths and
small, helical cuffs at 500 µs pulse width. These CIC values equate to allowable current stimulation of 6.6
to 9.5 mA for 200 µs pulse width and 4.4 to 5.3 mA for 500 µs pulse width (current = CIC × GSA / pulse
width). Average ESA, CSC, 1 kHz impedance magnitude, and CIC for all device configurations are
summarized in Table 4-5.
Figure 4-23: CV curves in H2SO4 at 250 mV/s for representative flat devices of each configuration from batch U. Note that the
larger curves correspond to larger ESAs, as listed in Table 4-5.

165
Figure 4-24: CV curves in PBS 50 mV/s for representative packaged (thermoformed and molded) devices of each configuration
from batch U.
Figure 4-25: EIS data in PBS for representative packaged (thermoformed and molded) devices of each configuration from batch
U.

166
Figure 4-26: VT data in PBS with 500 µs pulse width and 100 µs interphase delay for representative packaged (thermoformed
and molded) devices of each configuration from batch U after thermoforming.

167
Table 4-5: Average ESA, CSC, 1 kHz impedance magnitude, and CIC with a pulse width of 200 and 500 µs (± standard
deviation) for cuff electrodes from the most recent fabrication run (batch U). Values are seperated by device configuration
(helical vs. full, small vs. large, prior to thermoforming (pre-TF) vs. flat thermoformed (TF flat) vs. curled thermoformed (TF
curled)). Failed devices (per failure criteria in Table 4-2) were not included in calculations. Data not listed in the table (indicated
by a ‘-' mark) is unavailable due to no testing at that stage or insufficient sample size.
Device Configuration GSA ESA CSC |Z|
@ 1 kHz
CIC
@ 200 µs
CIC
@ 500 µs
(mm2
) (mm2
) (µC/cm2
) (Ω) (µC/cm2
) (µC/cm2
)
Helical,
Small
Pre-TF
(n = 22)
1.7 1.3 ± 0.3 1549 ± 167 - - -
TF curled
(n = 4)
1.7 - - 300 ± 37 78 ± 11 131 ± 43
Helical,
Large
Pre-TF
(n = 16) 2.8 2.6 ± 0.9 1537 ± 136 218 ± 15 - -
TF flat
(n = 2)
2.8 - - 222 ± 9 56 ± 1 76 ± 5
TF curled
(n = 8)
2.8 2.2 ± 0.5 1443 ± 141 227 ± 51 67 ± 11 92 ± 6
Full,
Small
Pre-TF
(n = 32) 4 3.1 ± 1.1 1481 ± 209 - - -
TF curled
(n = 6) 1.7 2.3 ± 0.8 3107 ± 831 563 ± 178 101 ± 20 177 ± 48
Full,
Large
Pre-TF
(n = 14) 3.4 3.0 ± 0.7 1571 ± 176 184 ± 18 - -
TF flat
(n = 26) 3.4 1.7 ± 0.9 1317 ± 624 241 ± 53 119 ± 23 165 ± 25
TF curled
(n = 0) 1.6 - - - - -
Average
Pre-TF - - 1526 ± 178 - - -
TF flat - - 1317 ± 624 - 114 ± 28 158 ± 34
TF curled - - 2156 ± 1004 - 85 ± 22 138 ± 53
1 ESA and CSC are calculated using CV in H2SO4
2
|Z| (impedance magnitude) is calculated using EIS in 1x PBS
3 CIC is calculated using VT in 1x PBS
4 Average values for GSA, ESA, and |Z| are not given as they are area dependent
4.4.3 Device Lifetime
Early accelerated lifetime results exposed severe failures of bare platinum electrodes. Preliminary
tests showed typical behavior in all tests (CV, EIS, and VT). After lifetime testing began, CV curves
rapidly decayed, losing their characteristic hydrogen and oxide peaks within 12-24 hours accelerated time
on non-stimulated and stimulated devices (using 0.5 mA, 1500 µs pulse width, 50 Hz), resulting in
decreased CSC. Figure 4-27 illustrates the decreasing CSC and evolution of CV curves over 13 days
accelerated time on a bare platinum device without and with stimulation (indicated by lines with ‘X’ and

168
‘O’ marks in Figure 4-27, respectively). The resulting shape of the CV curve resembles that of a low- or
non-conductive surface, suggesting oxidation or fouling of the electrode surface. EDS of one device
which was removed from testing revealed expected levels of platinum, indicating no or low dissolution of
the platinum electrode surface.
Stimulated bare platinum devices (N=3 devices, 6 electrodes) slightly out-performed nonstimulated devices (N=1 device, 2 electrodes) in days 0 through 8, as evidenced by less degradation in
impedance (Figure 4-28) and CIC (Figure 4-29), then had sharp increase in impedance at days 9 and 10.
Both sets of devices were considered failed due to the severity of CV degradation and CIC dropping
below the failure threshold of 50 µC/cm2
. Tests were terminated at 13 days after CIC dropped below the
stimulation charge density of 22.1 µC/cm2
and impedance magnitude raised above the failure criteria of 1
kΩ for several days straight. Two devices were terminated early (day 1.5) for inspection.
Two additional devices (one stimulated, one non-stimulated) were placed under the same testing
conditions with a stimulation frequency of 5 Hz for 4.3 days (accelerated), resulting in similar
performance decays in a similar span of time. This result in combination with comparable results from
stimulated and non-stimulated electrodes suggest this failure mode is independent of stimulation and
either related to chronic soaking in PBS, electrochemical testing conditions, increased temperature, or a
combination of these factors. Repeated electrochemical tests on devices used for earlier benchtop testing
and in vivo studies do not show this same result, indicating that testing alone is likely not the cause.
Further testing options are being considered, including testing at room or body temperature and isolating
devices from other materials (i.e. placing only a single device per testing jar and not using a wire or other
material to hold the device submerged), to evaluate this failure mode as it has not been reported in
literature.

169
Figure 4-27: (Left) CSC of bare platinum electrodes with (o) and without (x) stimulation over 13 days accelerated time. Each line
represents a single electrode measured over several time points. Devices show gradual decrease of CSC towards or below the
failure threshold of 100 µC/cm2
. (Right) representative CV curves (200 mV/s) for stimulated (top) and unstimulated (bottom)
devices at 0, 0.5, 1, 2, 7, and 13 accelerated days, demonstrating rapid disappearance of hydrogen and oxide peaks within the first
day, followed by gradual narrowing of the CV curve.

170
Figure 4-28: (Left) 1 kHz impedance magnitude of bare platinum electrodes with (o) and without (x) stimulation over 13 days
accelerated time. Each line represents a single electrode measured over several time points. Devices show gradual increase of
impedance above the failure threshold of 1 kΩ. (Right) representative EIS curves for stimulated (top) and unstimulated (bottom)
devices at 0, 0.5, 1, 2, 7, and 13 accelerated days, demonstrating gradual shift towards higher impedance.

171
Figure 4-29: (Left) CIC of bare platinum electrodes with (o) and without (x) stimulation over 13 days accelerated time. Each line
represents a single electrode measured over several time points. Devices show rapid degradation below the failure threshold of 50
µC/cm2
and stimulation charge density of 22.1 µC/cm2
.
Electrodes which were coated in electrodeposited PtIr showed very promising results over the
first 14 days of testing, with minimal changes in CV shape (Figure 4-30), CSC (Figure 4-30), impedance
(Figure 4-31), or CIC (Figure 4-32). During the first two days of soaking, device performance improved
(increased CSC, decreased impedance, increased CIC), likely due to full hydration and conditioning of
the PtIr surface. Some minor degradation of the CV curves (and resulting CSC) and the CIC was
observed between days 2 and 14, however the electrodes appear to be stabilizing and perform
significantly above the failure criteria. Parts will be monitored during ongoing testing until failure.

172
Figure 4-30: (Left) CSC of PtIr-coated electrodes with (o) and without (x) stimulation over 14 days accelerated time. Each line
represents a single electrode measured over several time points. Devices show an increase in CSC over the first few days,
followed by a slight, gradual decrease. (Right) representative CV curves (200 mV/s) for stimulated (top) and unstimulated
(bottom) devices at 0, 0.5, 1, 2, 7, and 14 accelerated days, demonstrating rapid increase, then slow decrease in area from the
starting curve (dark purple) to day 14 (red) which appears to be stabilizing.

173
Figure 4-31: (Left) 1 kHz impedance magnitude of PtIr-coated electrodes with (o) and without (x) stimulation over 14 days
accelerated time. Each line represents a single electrode measured over several time points. Devices show a rapid decrease of
impedance within the first 12 hours (in all but one sample) followed by very stable measurements. (Right) representative EIS
curves for stimulated (top) and unstimulated (bottom) devices at 0, 0.5, 1, 2, 7, and 14 accelerated days, demonstrating stability
after the first 12 hours.

174
Figure 4-32: (Left) CIC of PtIr-coated electrodes with (o) and without (x) stimulation over 14 days accelerated time. Each line
represents a single electrode measured over several time points. Devices show an increase in CIC over the first few days,
followed by a slight, gradual decrease.
4.4.4 In Vivo Testing
Preliminary in vivo experiments (n = 3 animals) demonstrated reliable placement of the cuff
(helical and full configurations) on the sciatic nerve of a rat (shown in Figure 4-33). The custom surgical
tool was used to open the cuff and place it over the nerve, then the cuff was slid off the tool into place on
the nerve. The soft-closing and self-sizing features of the cuff allow it to close snugly over the nerve
without over constriction. The first two studies focused on surgical technique and refinement of the
surgical tool. The third study included successful stimulation of the sciatic nerve, which resulted in
contraction of the muscles in the leg. Stimulation frequencies of 1 Hz and 30 Hz (for tetanic contraction),
amplitudes of 100 to 800 µA, and a pulse width of 200 µs resulted in varying degrees of muscle
contraction (with amount of contraction proportional to stimulation amplitude). Additional studies are
planned to stimulate the nerve and record compound action potentials from a separate electrode placed
downstream and to implant the cuff for one month to study chronic use.

175
Figure 4-33: Helical cuff implanted on the sciatic nerve of a rat.
4.5 Discussion
A microfabricated peripheral nerve cuff electrode has been designed, fabricated, and extensively
tested on the benchtop and in vivo. This device provides a self-sizing, soft-closing interface for submillimeter branched nerve fibers to enable more targeted neuromodulation therapies. Its overall flexibility
and compatibility with the OpenNerve system provides easier access for clinicians and a faster path to
use.
Device design parameters were selected using a combination of literature review and benchtop
testing. Electrode size and spacing were selected based on review of typical properties of thin film
electrodes and expected clinical stimulation parameters, using a safety factor of 1.5 to account for any
discrepancies between expected and experimental values. The shape of the electrodes, cuff configuration,
and cable routing were selected based on quantitative experimental testing of a large group of design
options (summarized in Appendix F).
Two configurations of cuffs were selected and fabricated for benchtop and in vivo testing.
Electrochemical evaluation of the cuff electrodes highlighted some batch to batch variation, namely
roughening of the electrode surface due to over etching or changing of surface properties of metal during
exposure to oxygen plasma. These changes will continue to be monitored in future fabrication runs.
Minor differences in CV curves in H2SO4 between flat and curled devices were identified. These
differences did not appear when testing in ferri/ferrocyanide, suggesting that the issue is due to the limited
amount of solution inside the curled cuff which can limit charge transfer at the electrode-electrolyte

176
interface and not due to a device issue. All passing devices had ESA above the requirement of 1.0 mm2
,
low 1 kHz impedance magnitudes, and CIC above the requirement of 50 µC/cm2
, allowing stimulation
from 6.6 to 9.5 mA for 200 µs pulse width and 4.4 to 5.3 mA for 500 µs pulse width on all electrode
types.
Accelerated lifetime testing exposed severe failures of uncoated platinum electrodes, with
electrode surfaces degrading within 24 hours with and without stimulation. The mechanism of this failure
is unknown, however further tests are planned to evaluate bare platinum electrodes using lower
stimulation frequency and more frequent testing to further characterize and prevent failure if possible.
These early results, however, indicate that bare platinum electrodes are likely not an option for chronic
use in these cuff electrodes due to the rapid and severe degradation. As such, an electrode coating must be
used to make these devices clinically useful. PtIr-coated electrodes performed well in the first two weeks
of lifetime testing, with improvement over the first few days during conditioning of the electrodes
followed by a very slight and gradual decrease in performance approaching baseline levels for CSC,
impedance, and CIC. Testing for PtIr electrodes is ongoing and will continue until device failure. Testing
of other common electrode coating options (IrOx and PEDOT) is planned as well to determine the best
coating option(s) for these cuff electrodes. Once early data is available for all coatings, the best option
will be selected and fully packaged devices will be coated and tested under the same accelerated
conditions until failure.
Early in vivo testing demonstrated reliability of the surgical procedure for implanting the device
on a rat sciatic nerve and successful stimulation of the nerve resulting in leg movement using a variety of
stimulation parameters. Future studies are planned to stimulate the sciatic nerve while concurrently
recording downstream compound action potentials and to implant electrodes over the course of one month
to obtain chronic data.
4.6 References
[1] Llinas R R, Walton K D, Nakao M, Hunter I and Anquetil P A 2005 Neuro-vascular central
nervous recording/stimulating system: Using nanotechnology probes J. Nanoparticle Res. 7 111–
27

177
[2] Gerboni G, John S E, Rind G S, Ronayne S M, May C N, Oxley T J, Grayden D B, Opie N L and
Wong Y T 2018 Visual evoked potentials determine chronic signal quality in a stent-electrode
endovascular neural interface Biomed. Phys. Eng. Express 4
[3] Yoon E, Koo B, Wong J, Elyahoodayan S, Weiland J D, Lee C D, Petrossians A and Meng E
2020 An implantable microelectrode array for chronic in vivo epiretinal stimulation of the rat
retina J. Micromechanics Microengineering 30 124001
[4] Johnson A C and Wise K D 2012 A self-curling monolithically-backed active high-density
cochlear electrode array IEEE 25th International Conference on Micro Electro Mechanical
Systems (MEMS) (IEEE) pp 914–7
[5] Grill W M, Norman S E and Bellamkonda R V. 2009 Implanted neural interfaces: Biochallenges
and engineered solutions Annu. Rev. Biomed. Eng. 11 1–24
[6] Cogan S F 2008 Neural stimulation and recording electrodes Annu. Rev. Biomed. Eng. 10 275–
309
[7] Ordonez J S, Boehler C, Schuettler M and Stieglitz T 2012 Improved polyimide thin-film
electrodes for neural implants Proc. Annu. Int. Conf. IEEE Eng. Med. Biol. Soc. EMBS 5134–7
[8] McGonigal A, Bartolomei F, Régis J, Guye M, Gavaret M, Fonseca A T Da, Dufour H, FigarellaBranger D, Girard N, Péragut J C and Chauvel P 2007 Stereoelectroencephalography in
presurgical assessment of MRI-negative epilepsy Brain 130 3169–83
[9] Jobst B C and Cascino G D 2015 Resective epilepsy surgery for drug-resistant focal epilepsy: A
review JAMA - J. Am. Med. Assoc. 313 285–93
[10] Klooster M A van ’t, Klink N E C van, Leijten F S S, Zelmann R, Gebbink T A, Gosselaar P H,
Braun K P J, Huiskamp G J M and Zijlmans M 2015 Residual fast ripples in the intraoperative
corticogram predict epilepsy surgery outcome Neurology 85 120–8
[11] Larson C E and Meng E 2020 A review for the peripheral nerve interface designer J. Neurosci.
Methods 332 108523
[12] Yuan H and Silberstein S D 2016 Vagus Nerve and Vagus Nerve Stimulation, a Comprehensive
Review: Part II Headache 56 259–66
[13] Russell C, Roche A D and Chakrabarty S 2019 Peripheral nerve bionic interface: a review of
electrodes Int. J. Intell. Robot. Appl. 3 11–8
[14] Cobo A M, Larson C E, Scholten K, Miranda J A, Elyahoodayan S, Song D, Pikov V and Meng E
2019 Parylene-Based Cuff Electrode With Integrated Microfluidics for Peripheral Nerve
Recording, Stimulation, and Drug Delivery J. Microelectromechanical Syst. 28 36–49
[15] Cowley A W 2013 Helical electrode for nerve stimulation
[16] LivaNova USA, Inc. 2022 Physician’s Manual VNS Therapy Generator and Lead Manual for
Epilepsy
[17] Wheless J W, Gienapp A J and Ryvlin P 2018 Vagus nerve stimulation (VNS) therapy update

178
Epilepsy Behav. 88 2–10
[18] Koo B, Ham S D, Sood S and Tarver B 2001 Human vagus nerve electrophysiology: A guide to
vagus nerve stimulation parameters J. Clin. Neurophysiol. 18 429–33
[19] LivaNova USA, Inc. 2022 Physician’s Manual VNS Therapy Generator and Lead Manual for
Depression
[20] Ivanovskaya A N, Belle A M, Yorita A M, Qian F, Chen S, Tooker A, Lozada R G, Dahlquist D
and Tolosa V 2018 Electrochemical Roughening of Thin-Film Platinum for Neural Probe Arrays
and Biosensing Applications J. Electrochem. Soc. 165 G3125–32
[21] Shannon R V 1992 A model of safe levels for electrical stimulation IEEE Trans. Biomed. Eng. 39
424–6
[22] Cogan S F, Ludwig K A, Welle C G and Takmakov P 2016 Tissue damage thresholds during
therapeutic electrical stimulation J. Neural Eng. 13
[23] Rose T L and Robblee L S 1990 Electrical Stimulation with Pt Electrodes. VIII.
Electrochemically Safe Charge Injection Limits with 0.2 MS Pulses IEEE Trans. Biomed. Eng. 37
1118–20
[24] Ganji M, Tanaka A, Gilja V, Halgren E and Dayeh S A 2017 Scaling Effects on the
Electrochemical Stimulation Performance of Au, Pt, and PEDOT:PSS Electrocorticography
Arrays Adv. Funct. Mater. 27
[25] Shen K, Chen O, Edmunds J L, Piech D K and Maharbiz M M 2023 Translational opportunities
and challenges of invasive electrodes for neural interfaces Nat. Biomed. Eng.
[26] CARSS 2024 USC Center for Autonomic Nerve Recording and Stimulation Systems (CARSS)
sites.usc.edu/carss/
[27] Scholten K, Larson C E, Xu H, Song D and Meng E 2020 A 512-Channel Multi-Layer PolymerBased Neural Probe Array J. Microelectromechanical Syst. 29 1054–8
[28] Meng E, Li P Y and Tai Y-C 2008 Plasma removal of Parylene C J. Micromechanics
Microengineering 18 045004
[29] Li Z, Beck P, Ohlberg D A A, Stewart D R and Williams R S 2003 Surface properties of platinum
thin films as a function of plasma treatment conditions Surf. Sci. 529 410–8

179
Chapter 5. Conclusion
BioMEMS technology has played a pivotal role in enabling the development of thin film Parylene
neural interfaces, significantly expanding the scope and capabilities of the field. Although MEMS
techniques produce planar devices, the modulation of thin film polymer devices into three-dimensional
structures is made possible through the use of inherent thin film stress and thermoforming. The
characterization of film stress and thermoforming in this work has added valuable knowledge to this field,
enabling new methods to produce sub-millimeter 3D structures from planar Parylene devices, which had
not been achieved prior. These new methods enable new capabilities of neural interfaces, such as intimate
interfacing with small anatomical features such as the blood vessels and peripheral nerves. The two
devices developed in this work demonstrate these applications and expand the diverse range of bioMEMS
neural interfaces.
The main goal of this work was to produce three-dimensional Parylene C neural interfaces on a
smaller size scale than had been achieved previously. This was accomplished using a combination of two
techniques: modulation of device geometries and processing parameters to produce natural curvature via
film stress, and imposition of more complex shapes via thermoforming. Chapter 2 describes an in depth
characterization of both of these techniques separately and in combination, resulting in the transformation
of a planar thin film Parylene-metal-Parylene device into a 0.25 mm diameter helix, a 4× decrease in size
from other published work. This process was applied to two devices to demonstrate the functionality of
sub-millimeter thermoformed Parylene devices and enable new applications of thin film neural interfaces.
This process was first applied to an endovascular electrode array for the application of seizure
mapping in epilepsy patients, as described in chapter 3. The device was designed, fabricated, and tested at
the benchtop and in a sheep model, demonstrating device survival after an aggressive simulated surgical
test and the first ever simultaneous recordings from endovascular, surface, and penetrating electrodes.
Planned future tests will validate the use of the device in an awake, freely moving animal and gather more
clinically relevant data. The novel device developed in this work offers an advantage over existing

180
devices owing to its small size and multiple electrodes along its length, enabling high spatial resolution
recording from small blood vessels.
Additionally, the thermoforming process was applied to a cuff electrode for stimulation of
branched peripheral nerves, described in chapter 4. The cuff electrode was designed and fabricated and
put through a battery of benchtop testing to determine the electrochemical properties and estimated
lifetime during chronic use. Early accelerated life tests exposed the need to use electrode coatings due to
degradation of the platinum electrode surface over the first several days under test. Platinum iridium
coatings show promising results over the first two weeks of testing, and tests on PEDOT and iridium
oxide coatings are planned. Inactive devices were also implanted into a rat model to demonstrate surgical
feasibility, and ongoing studies with active devices are planned. The novel cuff electrode developed in
this work enables stimulation of small diameter branched nerves which cannot easily interface with
existing clinical cuffs. This device serves as one component of the greater OpenNerve system which will
allow easier, open source access to a full neuromodulation system by researchers and clinicians looking to
advance treatment for a wide variety of applications.
The work described here has the potential to enable future researchers to develop polymer
bioMEMS devices with more complex geometries for a wide variety of applications, with the aim of
advancing healthcare and bioelectronic medicine. The endovascular electrode array provides a safer and
more effective avenue to epilepsy care, and the cuff electrode provides more targeted therapy via
branched nerve fibers with an accompanying full, open source neuromodulation system.

181
Appendices
A MATLAB Code for Parylene-Metal-Parylene Device Stress Model
The model was coded into MATLAB as the function StressCurvature2.m. A graphical user
interface (GUI) was also created in a MATLAB live script to interactively change device parameters and
view resulting device curvature. StressCurvature2.m and screenshots of the GUI are included below.
MATLAB code for StressCurvature2 and the GUI can be found on the laboratory server at /Users/Brianna
Thielen/Curvature Model.
A.1 StressCurvature2.m
function [D, ybar] = StressCurvature2(array_params)
% StressCurvature.m calculates the diameter (in mm) of a three-layer
% device based on the following input parameters:
%
% array_params = [t_base, t_middle, t_top, ...
% w_base, w_middle, w_top, ...
% s_base, s_middle, s_top, ...
% E_base, E_middle, E_top];
%
% t_base = thickness of base layer
% t_middle = thickness of middle layer
% t_top = thickness of top layer
%
% w_base = width of base layer
% w_metal = width of middle layer
% w_top = width of top layer
% note: if a layer is patterned into multiple pieces (i.e., metal
% traces), combine the width of all portions of that layer
% (e.g., 4 metal traces of 10 um width = 40 um layer width)
%
% s_base = stress in base layer
% s_middle = stress in middle layer
% s_top = stress in top layer
%
% E_base = Young's modulus of base layer
% E_middle = Young's modulus of in middle layer
% E_top = Young's modulus of in top layer
%
% in most cases, the base and top layers are Parylene and the middle layer
% is metal
%
% all input dimensions are in microns
% all input stresses/young's moduli are in pascals
% tensile stress is positive, compressive stress is negative
%
% Output diameter is in millimeters
% Pull info from array_params
t_base = array_params(1);

182
t_middle = array_params(2);
t_top = array_params(3);
w_base = array_params(4);
w_middle = array_params(5);
w_top = array_params(6);
s_base = array_params(7);
s_middle = array_params(8);
s_top = array_params(9);
E_base = array_params(10);
E_middle = array_params(11);
E_top = array_params(12);
% Constants
E_Pa = 2.8e9; %pascals
E_pt = 172e9; %pascals
% Area calculations
A_base = t_base * w_base;
A_middle = t_middle * w_middle;
A_top = t_top * w_top;
% Neutral Axis Calculations
% ybar is distance from the bottom surface
ybar_base = t_base / 2;
ybar_middle = t_base + t_middle / 2;
ybar_top = t_base + t_middle + t_top / 2;
ybar = round((ybar_base .* A_base * E_base + ...
ybar_middle .* A_middle * E_middle + ...
ybar_top .* A_top * E_top) ...
./ (A_base * E_base + A_middle * E_middle + A_top * E_top), 4);
% Forces (not used)
% F_base = s_base .* A_base;
% F_middle = s_middle .* A_middle;
% F_top = s_top .* A_top;
% Stress Relaxation (flat)
% calculated by:
% setting strain of each layer equal during stress relaxation
% setting the sum of the new forces (F = s * A) equal to zero
%
% strain = (s_base_new - s_base) / E_base
% = (s_middle_new - s_middle) / E_middle
% = (s_top_new - s_top) / E_top
%
% sum(F) = 0 = s_base_new * A_base
% + s_middle_new * A_middle
% + s_top_new * A_top
%
s_base_new = (s_base * ...
((E_middle/E_base) * A_middle + (E_top/E_base) * A_top) - ...

183
s_middle * A_middle - s_top * A_top) / ...
(A_base + (E_middle/E_base) * A_middle + (E_top/E_base) * A_top);
s_middle_new = (s_middle * ...
((E_base/E_middle) * A_base + (E_top/E_middle) * A_top) - ...
s_base * A_base - s_top * A_top) / ...
(A_middle + (E_base/E_middle) * A_base + (E_top/E_middle) * A_top);
s_top_new = (s_top * ...
((E_base/E_top) * A_base + (E_middle/E_top) * A_middle) - ...
s_base * A_base - s_middle * A_middle) / ...
(A_top + (E_base/E_top) * A_base + (E_middle/E_top) * A_middle);
% New Forces
F_base_new = s_base_new * A_base;
F_middle_new = s_middle_new * A_middle;
F_top_new = s_top_new * A_top;
% 2nd moment of area (each component)
I_base = (w_base * t_base^3) / 12 + ...
(ybar - ybar_base)^2 * A_base;
I_middle = (w_middle * t_middle^3) / 12 + ...
(ybar - ybar_middle)^2 * A_middle;
I_top = (w_top * t_top^3) / 12 + ...
(ybar - ybar_top)^2 * A_top;
% 2nd moment of area (full device, using equivalent area method)
I = I_base + I_middle * (E_middle/E_base) + I_top * (E_top/E_base);
% Moment
M_base = -F_base_new * (ybar - ybar_base);
M_middle = -F_middle_new * (ybar - ybar_middle);
M_top = -F_top_new * (ybar - ybar_top);
M = M_base + M_top + M_middle;
% Radius
R = (E_base * I / M) / 1000; %convert to mm
D = 2 * R; %convert to diameter
end

184
A.2 Screenshots from GUI for Curvature Model
Figure A-1: Curvature model GUI - device parameters view with plot of resulting curvature.

185
Figure A-2: Curvature model GUI - device stackup render view which shows approximate device geometry as defined (real) and
as modeled (averaged for a consistent cross section).

186
B Parylene-Metal-Parylene Device Fabrication (Double-Sided)
Figure B-1: Process flow diagram for double-sided Parylene-metal-Parylene thin film devices.
1. Sacrificial Layer
1.1. Drybake wafers at 110 °C in an oven at atmosphere for at least 30 minutes (overnight
ok), or on a hot plate for >5 minutes
1.2. Deposit Aluminum
1.2.1. Deposit 10 nm titanium and 100 nm aluminum in the e-beam evaporator at 1.5-2
Å/s
1.3. Roughen Aluminum
1.3.1. Place wafer in CR-7 bath for 8 minutes with no agitation
1.3.2. Rinse thoroughly in running DI water and blow dry with N2
2. Deposit Base Parylene C
2.1. Drybake wafers at 110 °C in an oven at atmosphere for at least 30 minutes (overnight
ok), or on a hot plate for >5 minutes
2.2. Deposit Parylene C
Note: this step should be performed immediately after drybake (step 2.1)
2.2.1. Determine amount of dimer by referencing past Parylene runs
- To produce flat devices when released, base and top Parylene should have
equal thickness
- To produce curled devices when released, see Thielen B and Meng E 2023
Characterization of thin film Parylene C device curvature and the formation
of helices via thermoforming J. Micromechanics Microengineering 33
095007
1.1.2. Deposit Parylene per the Parylene tool SOP
3. Pattern Base Parylene (backside openings)

187
3.1. Inspect the backside opening etch photomask and clean (with acetone/IPA or Nanostrip)
if necessary
3.2. Drybake wafers at 60 °C in an oven under light vacuum (35-40 cmHg) and N2 flow (15-
20 sccm) for >15 minutes
3.3. Deposit Photoresist (AZ P4620)
Note: this step should be performed immediately after drybake (step 3.2)
3.3.1. Degas photoresist for >1 hour prior to spinning (open bottle and set it in the hood
with the lights off)
3.3.2. Coat 2 dummy wafers in spin coater prior to coating real wafers
Note: the remaining steps (3.3.3 through 3.3.13) are performed one wafer at a time;
repeat the following procedure once for each wafer
3.3.3. Place wafer onto the spinner chuck, center it, and engage vacuum to hold it
3.3.4. Dispense P4620 photoresist into a puddle on the center of the wafer
- ~1.5 inch diameter puddle
- Use more photoresist if surface is uneven to ensure sufficient coverage
3.3.5. Close the spinner lid and spin photoresist to desired thickness using the following
recipe:
- 5 s, 500 RPM, accl 4 (spreads out PR puddle)
- 45 s, * RPM, accl 15 (see below for speeds; defines desired thickness)
- 2 s, 4500 RPM, accl 15 (edge bead removal)
* speed is selected using past measurements; thickness should be ~1.5-2x
base Parylene thickness
Common speeds: 1100 rpm → ~15 µm; 1300 rpm → ~13.5 µm; 3200 rpm →
~8 µm
Note: steps 3.3.6 through 3.3.11 (edge bead removal – EBR) are optional and can
alternatively be performed manually (without spinning) after exposure in step 3.7
3.3.6. Open spinner lid and lower EBR shield (black plastic cylinder) over wafer
without touching the wafer and place magnet over the lid sensor to override the
interlock
3.3.7. Soak a large foam swab in EBR solvent and blot away excess solvent from the
swab
3.3.8. Place the swab on the edge of the wafer at the 3 o’clock position such that the
swab is only in contact with the edge bead (no more than 5 mm away from the
edge of the wafer)
3.3.9. Spin the wafer using the following recipe:
- 5 s, 200 RPM, accl 4 (time to position swab)
- 40 s, 750 RPM, accl 4 (EBR time)
3.3.10. Position the swab during the first 5 seconds, leave the swab in contact with the
wafer for 20 seconds, then allow the wafer to spin dry for 20 seconds
3.3.11. Repeat step 3.3.7 through 3.3.10 if necessary until the edge bead has been
removed
3.3.12. Move wafer to hot plate and soft bake at 90 °C for 5 minutes (increase time for
thicker spins)
3.3.13. Let wafer sit at room temperature for >3 minutes (rehydration)
3.4. Expose Photoresist

188
3.4.1. On the mask aligner, set the UV exposure dose to 420 mJ/cm2
3.4.2. Install etch mask 1 into the mask aligner
Note: the remaining steps (3.4.3 through 3.4.5) are performed one wafer at a time; repeat
the following procedure once for each wafer
3.4.3. Install wafer in the wafer chuck and align to the mask pattern
3.4.4. Expose wafer through etch mask 1 in soft contact mode
3.4.5. Place wafer immediately into DI water bath after exposure for at least 2 minutes
to prevent overheating
3.5. Develop Photoresist
3.5.1. Prepare developer bath (1:4 ratio of 340 developer to DI water) and DI water
rinse in separate plastic trays
- A fresh developer bath should be used for each wafer – do not re-use
developer
3.5.2. Place wafer in developer bath for desired time (typically 45-90 seconds) with
mild agitation
3.5.3. Move quickly to water bath, then flush 3x with DI water
3.5.4. Blow dry with N2
3.5.5. Inspect developed features under microscope and develop for additional time if
needed
3.6. Reflow Photoresist
Note: all steps are performed one wafer at a time; repeat the following procedure once
for each wafer
3.6.1. Place wafer with developed photoresist on a hot plate at 110 °C for 20 seconds
3.6.2. Inspect developed features under microscope for sloped sidewalls and bake for
additional time if needed
- Sidewalls should be sufficiently sloped to ensure continuity between metal
deposited on top of Parylene (traces) and directly on the wafer (etched areas)
3.7. Edge Bead Removal
Note: If edge bead removal was performed during step 3.3, skip this step and proceed to
step 3.8 or 3.9 (see note)
Note: all steps are performed one wafer at a time; repeat the following procedure once
for each wafer
3.7.1. Saturate a foam swab with acetone and dab any excess off the swab
3.7.2. Wipe all photoresist off the outer ~5 mm of the wafer using the swab, resaturating it in acetone as needed
Note: the etching process has two options that can be used. The DRIE (step 3.8) is preferred, but
the RIE (step 3.9) can be used if the DRIE is unavailable. Perform only one of the following
etching procedures (step 3.8 OR step 3.9).
3.8. Etch Parylene (DRIE)
Note: this step should etch through the thickness of the base Parylene layer (down to the
wafer)
Equipment: DRIE

189
3.8.1. Calculate necessary etch loops based on prior runs and desired etch depth (typical
rate is ~0.07-0.1 µm/loop)
3.8.2. Etch wafers in the DRIE through the patterned photoresist
- Etch in steps of less than 50 loops to prevent overheating
- DRIE recipe developed in Meng E, Li P Y and Tai Y-C 2008 Plasma
removal of Parylene C J. Micromechanics Microengineering 18 045004
3.8.3. After each step, inspect wafers for any remaining Parylene in the etched areas
and continue etching as needed
3.9. Etch Parylene (RIE)
Note: this step should etch through the thickness of the base Parylene layer (down to the
wafer)
Equipment: RIE
3.9.1. Calculate necessary etch time based on prior runs and desired etch depth (typical
rate is ~0.16-0.19 µm/minute)
3.9.2. Etch wafers in the RIE through the patterned photoresist using the following
parameters:
- 150 mT, 150 W, 50 sccm O2
- Perform in two or more steps of 15 minutes or less, rotating the wafer(s) 90-
180 degrees with each step
3.9.3. After each step, inspect wafers for any remaining Parylene in the etched areas
and continue etching as needed
3.10. Remove Remaining Photoresist
3.10.1. Strip remaining photoresist off each wafer per the following procedure:
- Soak wafer in an acetone bath for 30-60 seconds with mild agitation to
remove the majority of photoresist
- Move wafer to a second acetone bath and soak for >3 minutes with periodic
mild agitation
- Move wafer to IPA and soak for >3 minutes with periodic mild agitation
- Move mask to water and soak for >1 minutes with periodic mild agitation
- Rinse gently with water, blow dry with N2
4. Anneal Wafer (optional)
4.1. Place wafers in a vacuum oven, close, and evacuate chamber to 70 cmHg or greater
vacuum
4.2. Close vacuum valve, purge chamber with N2 to 20-30 cmHg, then re-evacuate to 70
cmHg or greater
4.3. Repeat step 4.2 twice (three total N2 purges)
4.4. Leave the vacuum valve open and open the N2 valve until 10-15 sccm of N2 are flowing
into the chamber
4.5. Bake wafers (under vacuum and N2 flow) for desired time and temperature
- Typical parameters for metal stress relaxation and adhesion: 150 °C, 4 hrs
- To produce flat devices, base Parylene and full wafer anneal should use the
same parameters
5. Deposit and Pattern Metal
5.1. Inspect the metal photomask and clean (with acetone/IPA or Nanostrip) if necessary
5.2. Drybake wafers at 60 °C in an oven under light vacuum (35-40 cmHg) and N2 flow (15-
20 sccm) for >15 minutes

190
Note: this step can be skipped if the following step (5.3 deposit photoresist) is performed
immediately after annealing (step 4)
5.3. Deposit Photoresist (AZ 5214)
Note: this step should be performed immediately after drybake (step 5.2) or annealing
(step 4)
5.3.1. Degas photoresist for 1 hour prior to spinning (open bottle and set it in the hood
with the lights off)
5.3.2. Coat 2 dummy wafers in spin coater prior to coating real wafers
Note: the remaining steps (5.3.3 through 5.3.6) are performed one wafer at a time; repeat
the following procedure once for each wafer
5.3.3. Place wafer onto the spinner chuck, center it, and engage vacuum to hold it
5.3.4. Dispense 5214 photoresist into a puddle on the center of the wafer
- ~1.5 inch diameter puddle
- Use more photoresist if surface is uneven to ensure sufficient coverage
5.3.5. Close the spinner lid and spin photoresist to ~1-2 µm thickness using the
following recipe:
- 7 s, 500 RPM, accl 8 (spreads out PR puddle)
- 45 s, 2000 RPM, accl 8 (defines desired thickness)
5.3.6. Move wafer to hot plate and soft bake at 90 °C for 70 seconds
5.4. Expose Photoresist
5.4.1. On the mask aligner, set the UV exposure dose to 36.75 mJ/cm2
5.4.2. Install metal mask into the mask aligner
Note: the remaining steps (5.4.3 through 5.4.9) are performed one wafer at a time; repeat
the following procedure once for each wafer
5.4.3. Install wafer in the wafer chuck and align to the mask pattern
5.4.4. Expose wafer through metal mask in hard contact mode
5.4.5. Bake at 110 °C for 45 seconds (image reversal bake)
5.4.6. Rest wafer for >3 minutes to cool down
5.4.7. On the mask aligner, set the UV exposure dose to 225 mJ/cm2
5.4.8. Flood expose wafer
5.4.9. Place wafer immediately into DI water bath after flood exposure for at least 2
minutes to prevent overheating
5.5. Develop Photoresist
Note: all steps are performed one wafer at a time; repeat the following procedure once
for each wafer
5.5.1. Prepare developer bath (1:4 ratio of AZ 340 developer to DI water) and DI water
rinse in separate plastic trays
- A fresh developer bath should be used for each wafer – do not re-use
developer
5.5.2. Place wafer in developer bath for the desired time (typically 16-19 seconds) with
mild agitation
5.5.3. Move quickly to water bath, then flush 3x with DI water
5.5.4. Blow dry with N2

191
5.5.5. Inspect developed features under microscope and develop for additional time if
needed
5.6. Descum Wafer
5.6.1. Descum (clean) wafers in the RIE or Asher using the following recipe:
- 100 mT, 100 W, 50 sccm O2, 1-5 minutes
5.7. Deposit Metal
Note: this step should be performed immediately after descum (step 5.6)
Note: different metals, thicknesses, and rates can be used; recommended values are
included in the steps below
5.7.1. Deposit the desired metal stackup in the e-beam evaporator at 1.5-2 Å/s
- Typical Ti/Pt stackup: 15 nm Ti (adhesion layer) + 200 nm Pt
- Typical Ti/Pt/Au/Pt stackup: 20 nm Ti (adhesion layer) + 25 nm Pt + 155 nm
Au + 25 nm Pt
5.7.2. Wait at least 30 minutes between different types of metal to allow crucible to
cool down
5.7.3. Do not deposit more than 50 nm of metal at a single time; if a layer is more than
50 nm, add a 30 minute pause between each 50 nm (e.g. for 200 nm Pt, deposit in
four 50 nm runs with 30 minute pauses between each run)
Note: the liftoff process has many options that can be used; perform only one of the following
liftoff procedures (step 5.8, step 5.9 OR step 5.10)
5.8. Pattern Metal via Liftoff (Option 1: Overnight Acetone w/ Sonication)
Note: this option is the most aggressive and should only be used if the base layer anneal
(step 4) was performed and there is good adhesion between the metal and Parylene
5.8.1. Place wafers in an acetone bath oriented vertically
5.8.2. Cover bath with aluminum foil, and leave wafers to soak overnight
Note: the remaining steps (5.8.3 through 5.8.10) are performed one wafer at a time;
repeat the following procedure once for each wafer
5.8.3. Lift wafer out of the acetone bath, rinse with acetone squeeze bottle, and move to
an NMP bath
- Use the squeeze bottle to rinse off any loose metal into the acetone bath
- Do not allow wafer to dry, as this will cause lifted-off metal to permanently
stick to the wafer surface
5.8.4. Place the NMP bath into the sonicating bath and sonicate for approximately 1-2
minutes (until all remaining metal has been lifted off)
5.8.5. Lift wafer out of the fresh NMP bath, rinse with NMP squeeze bottle, and move
to IPA bath for >5 minutes
- Do not allow wafer to dry, as this will cause lifted-off metal to permanently
stick to the wafer surface
5.8.6. Inspect the wafer for any remaining metal while submerged in IPA under the
stereoscope
- If any undesired metal remains, move wafer back to the NMP bath and repeat
process from step 5.8.4
- If any stubborn metal remains after repeating sonication, a foam swab can be
used to gently dislodge metal from the wafer surface

192
5.8.7. Lift wafer out of IPA bath, rinse with IPA squeeze bottle, and move to DI water
bath for >3 minutes
- Do not allow wafer to dry, as this will cause lifted-off metal to permanently
stick to the wafer surface
5.8.8. Re-inspect the wafer for any remaining metal while submerged in water under
the stereoscope
- If any undesired metal remains, move wafer back to the NMP bath and repeat
process from step 5.8.4
5.8.9. Rinse wafer with DI water 3 times, and blow dry with N2
5.8.10. Inspect metal features under microscope and return to step 5.8.4 if any metal or
photoresist remains
5.9. Pattern Metal via Liftoff (Option 2: Warm NMP w/ Sonication)
Note: this option is mildly aggressive and should only be used if there is sufficient
adhesion between the metal and Parylene
Note: all steps are performed one wafer at a time; repeat the following procedure once
for each wafer
5.9.1. Soak wafer in 60 °C NMP for 10-20 minutes with periodic mild agitation until
metal visibly lifts off of wafer
5.9.2. Hold the NMP bath (with the wafer inside) above the sonicating bath and
intermittently touch the sonicated water surface for intermittent sonication until
metal appears to fully lift off
5.9.3. Once liftoff appears to be complete, move NMP bath back to the hot plate
5.9.4. Lift wafer out of the warm NMP bath, rinse with NMP squeeze bottle, and move
to room temperature NMP bath for >5 minutes
- Do not allow wafer to dry, as this will cause lifted-off metal to permanently
stick to the wafer surface
5.9.5. Lift wafer out of the room temperature NMP bath, rinse with NMP squeeze
bottle, and move to IPA bath for >5 minutes
- Do not allow wafer to dry, as this will cause lifted-off metal to permanently
stick to the wafer surface
5.9.6. Inspect the wafer for any remaining metal while submerged in IPA under the
stereoscope
- If any undesired metal remains (metal that has not yet been lifted off or metal
flakes sitting on the surface), move wafer back to the warm NMP bath and
repeat process from step 5.9.2
- If any stubborn metal remains after repeating sonication, a foam swab can be
used to gently dislodge metal from the wafer surface
5.9.7. Lift wafer out of IPA bath, rinse with IPA squeeze bottle, and move to DI water
bath for >3 minutes
5.9.8. Re-inspect the wafer for any remaining metal while submerged in water under
the stereoscope
- If any undesired metal remains (metal that has not yet been lifted off or metal
flakes sitting on the surface), move wafer back to the warm NMP bath and
repeat process from step 5.9.2
5.9.9. Rinse wafer with DI water 3 times, and blow dry with N2

193
5.9.10. Inspect metal features under microscope and return to step 5.9.2 if any metal or
photoresist remains
5.10. Pattern Metal via Liftoff (Option 3: Warm Acetone w/o Sonication)
Note: this option is the least aggressive and will likely not work for liftoff of gold
Note: all steps are performed one wafer at a time; repeat the following procedure once
for each wafer
5.10.1. Soak wafer in 60 °C acetone for 10-20 minutes with periodic mild agitation until
metal visibly lifts off of wafer
- If any stubborn metal is not lifting off, a foam swab can be used to gently
dislodge metal from the wafer surface
5.10.2. Lift wafer out of the warm acetone bath, rinse with acetone squeeze bottle, and
move to room temperature acetone bath for >5 minutes
- Do not allow wafer to dry, as this will cause lifted-off metal to permanently
stick to the wafer surface
5.10.3. Lift wafer out of the room temperature acetone bath, rinse with acetone squeeze
bottle, and move to IPA bath for >5 minutes
- Do not allow wafer to dry, as this will cause lifted-off metal to permanently
stick to the wafer surface
5.10.4. Inspect the wafer for any remaining metal while submerged in IPA under the
stereoscope
- If any undesired metal remains (metal that has not yet been lifted off or metal
flakes sitting on the surface), move wafer back to the warm acetone bath and
repeat process from step 5.10.2
- If any stubborn metal remains after repeating sonication, a foam swab can be
used to gently dislodge metal from the wafer surface
5.10.5. Lift wafer out of IPA bath, rinse with IPA squeeze bottle, and move to DI water
bath for >3 minutes
- Do not allow wafer to dry, as this will cause lifted-off metal to permanently
stick to the wafer surface
5.10.6. Re-inspect the wafer for any remaining metal while submerged in water under
the stereoscope
- If any undesired metal remains (metal that has not yet been lifted off or metal
flakes sitting on the surface), move wafer back to the warm acetone bath and
repeat process from step 5.10.2
5.10.7. Rinse wafer with DI water 3 times, and blow dry with N2
5.10.8. Inspect metal features under microscope and return to step 5.10.2 if any metal or
photoresist remains
6. Deposit Top Parylene C
6.1. Descum Wafer
6.1.1. Descum (clean) wafers in the RIE or Asher using the following recipe:
- 100 mT, 100 W, 50 sccm O2, 1-5 minutes
6.2. Drybake wafers at 60 °C in an oven under light vacuum (35-40 cmHg) and N2 flow (15-
20 sccm) for >15 minutes
Note: do not perform this step if doing silanization (step 6.3)

194
6.3. Silanization
Note: this step is optional, but provides stronger adhesion between the metal and top
Parylene layers to decrease crosstalk and delamination
6.3.1. In the large A-174 beaker, prepare a mixture of 900 mL DI water, 900 mL
isopropanol, and 9 mL A-174 silane
- This volume is used for batches of 12 wafers; smaller quantities using the
same ratio can be used for smaller batches
- Stir mixture with a glass stirring rod
6.3.2. Cover the beaker with aluminum foil and let sit for at least 2.5 hours, but no more
than 24 hours
6.3.3. Decant the mixture into the crystalizing dish
6.3.4. Place wafers face up into the wafer cassette and lower the cassette into the
crystalizing dish, ensuring all wafers are fully submerged
6.3.5. Soak for 30 minutes
6.3.6. Remove the wafers and place on tex-wipes in the fume hood face up and air dry
for 30 minutes
6.3.7. Rinse the wafers thoroughly with IPA for 30 minutes
6.3.8. Blow dry with N2
6.3.9. Wafers should be coated with Parylene (step 3.4) within 12 hours after treatment
6.4. Deposit Parylene C
Note: this step should be performed immediately after drybake (step 6.2) or within 12
hours after silanization (6.3)
6.4.1. Label backside of each wafer using a permanent marker (below existing label)
with the date and which shelf it will be loaded on
6.4.2. If symmetric Parylene layers are desired, use the same amount of dimer and load
wafers onto the same shelf as the first Parylene run to increase likelihood of
symmetric layers
6.4.3. Determine amount of dimer by referencing past Parylene runs
- To produce flat devices when released, base and top Parylene should have
equal thickness
- To produce curled devices when released, see Thielen B and Meng E 2023
Characterization of thin film Parylene C device curvature and the formation
of helices via thermoforming J. Micromechanics Microengineering 33
095007
6.4.4. Deposit Parylene per the Parylene tool SOP
7. Anneal Wafer (optional)
7.1. Place wafers in a vacuum oven, close, and evacuate chamber to 70 cmHg or greater
vacuum
7.2. Close vacuum valve, purge chamber with N2 to 20-30 cmHg, then re-evacuate to 70
cmHg or greater
7.3. Repeat step 7.2 twice (three total N2 purges)
7.4. Leave the vacuum valve open and open the N2 valve until 10-15 sccm of N2 are flowing
into the chamber
7.5. Bake wafers (under vacuum and N2 flow) for desired time and temperature
- Typical parameters for adhesion: 150 °C, 4 hrs
- To produce flat devices, base Parylene and full wafer anneal should use the
same parameters

195
- To produce curled devices, see Thielen B and Meng E 2023 Characterization
of thin film Parylene C device curvature and the formation of helices via
thermoforming J. Micromechanics Microengineering 33 095007
8. Pattern Top Parylene (Step 1 - top open features and edge)
8.1. Inspect the frontside opening etch photomask and clean (with acetone/IPA or Nanostrip)
if necessary
8.2. Drybake wafers at 60 °C in an oven under light vacuum (35-40 cmHg) and N2 flow (15-
20 sccm) for >15 minutes
Note: this step can be skipped if the following step (8.3 deposit photoresist) is performed
immediately after annealing (step 7)
8.3. Deposit Photoresist (AZ P4620)
Note: this step should be performed immediately after drybake (step 8.2) or annealing
(step 7)
8.3.1. Degas photoresist for >1 hour prior to spinning (open bottle and set it in the hood
with the lights off)
8.3.2. Coat 2 dummy wafers in spin coater prior to coating real wafers
Note: the remaining steps (8.3.3 through 8.3.14) are performed one wafer at a time;
repeat the following procedure once for each wafer
8.3.3. Place wafer onto the spinner chuck, center it, and engage vacuum to hold it
8.3.4. Dispense P4620 photoresist into a puddle on the center of the wafer
- ~1.5 inch diameter puddle
- Use more photoresist if surface is uneven to ensure sufficient coverage
8.3.5. Close the spinner lid and spin photoresist to desired thickness using the following
recipe:
- 5 s, 500 RPM, accl 4 (spreads out PR puddle)
- 45 s, * RPM, accl 15 (see below for speeds; defines desired thickness)
- 2 s, 4500 RPM, accl 15 (edge bead removal)
* speed is selected using past measurements; thickness should be ~1.5-2x top
Parylene thickness
Common speeds: 1100 rpm → ~15 µm; 1300 rpm → ~13.5 µm; 3200 rpm →
~8 µm
Note: steps 8.3.6 through 8.3.12 (edge bead removal – EBR) are optional and can
alternatively be performed manually (without spinning) after exposure in step 8.6
8.3.6. Open spinner lid and lower EBR shield (black plastic cylinder) over wafer
without touching the wafer and
8.3.7. Place magnet over the lid sensor to override the interlock
8.3.8. Soak a large foam swab in EBR solvent and blot away excess solvent from the
swab
8.3.9. Place the swab on the edge of the wafer at the 3 o’clock position such that the
swab is only in contact with the edge bead (no more than 5 mm away from the
edge of the wafer)
8.3.10. Spin the wafer using the following recipe:
- 5 s, 200 RPM, accl 4 (time to position swab)
- 40 s, 750 RPM, accl 4 (EBR time)

196
8.3.11. Position the swab during the first 5 seconds, leave the swab in contact with the
wafer for 20 seconds, then allow the wafer to spin dry for 20 seconds
8.3.12. Repeat step 8.3.8 through 8.3.11 if necessary until the edge bead has been
removed
8.3.13. Move wafer to hot plate and soft bake at 90 °C for 5 minutes (increase time for
thicker spins)
8.3.14. Let wafer sit at room temperature for >3 minutes (rehydration)
8.4. Expose Photoresist
8.4.1. On the mask aligner, set the UV exposure dose to 420 mJ/cm2
8.4.2. Install etch mask 1 into the mask aligner
Note: the remaining steps (8.4.3 through 8.4.5) are performed one wafer at a time; repeat
the following procedure once for each wafer
8.4.3. Install wafer in the wafer chuck and align to the mask pattern
8.4.4. Expose wafer through etch mask 1 in soft contact mode
8.4.5. Place wafer immediately into DI water bath after exposure for at least 2 minutes
to prevent overheating
8.5. Develop Photoresist
8.5.1. Prepare developer bath (1:4 ratio of 340 developer to DI water) and DI water
rinse in separate plastic trays
- A fresh developer bath should be used for each wafer – do not re-use
developer
8.5.2. Place wafer in developer bath for desired time (typically 45-90 seconds) with
mild agitation
8.5.3. Move quickly to water bath, then flush 3x with DI water
8.5.4. Blow dry with N2
8.5.5. Inspect developed features under microscope and develop for additional time if
needed
8.6. Edge Bead Removal
Note: If edge bead removal was performed during step 8.3, skip this step and proceed to
step 8.7 or 8.8 (see note)
Note: all steps are performed one wafer at a time; repeat the following procedure once
for each wafer
8.6.1. Saturate a foam swab with acetone and dab any excess off the swab
8.6.2. Wipe all photoresist off the outer ~5 mm of the wafer using the swab, resaturating it in acetone as needed
Note: the etching process has two options that can be used. The DRIE (step 8.7) is preferred, but
the RIE (step 8.8) can be used if the DRIE is unavailable. Perform only one of the following
etching procedures (step 8.7 OR step 8.8).
8.7. Etch Parylene (DRIE)
Note: this step should etch through the thickness of the top Parylene layer (down to the
metal layer for any exposed metal features)
Equipment: DRIE

197
8.7.1. Calculate necessary etch loops based on prior runs and desired etch depth (typical
rate is ~0.07-0.1 µm/loop)
8.7.2. Etch wafers in the DRIE through the patterned photoresist
- Etch in steps of less than 50 loops to prevent overheating
- DRIE recipe developed in Meng E, Li P Y and Tai Y-C 2008 Plasma
removal of Parylene C J. Micromechanics Microengineering 18 045004
8.7.3. After each step, inspect wafers for any remaining Parylene in the etched areas
and continue etching as needed
8.8. Etch Parylene (RIE)
Note: this step should etch through the thickness of the top Parylene layer (down to the
metal layer for any exposed metal features)
Equipment: RIE
8.8.1. Calculate necessary etch time based on prior runs and desired etch depth (typical
rate is ~0.16-0.19 µm/minute)
8.8.2. Etch wafers in the RIE through the patterned photoresist using the following
parameters:
- 150 mT, 150 W, 50 sccm O2
- Perform in two or more steps of 15 minutes or less, rotating the wafer(s) 90-
180 degrees with each step
8.8.3. After each step, inspect wafers for any remaining Parylene in the etched areas
and continue etching as needed
8.9. Remove Remaining Photoresist
8.9.1. Strip remaining photoresist off each wafer per the following procedure:
- Soak wafer in an acetone bath for 30-60 seconds with mild agitation to
remove the majority of photoresist
- Move wafer to a second acetone bath and soak for >3 minutes with periodic
mild agitation
- Move wafer to IPA and soak for >3 minutes with periodic mild agitation
- Move mask to water and soak for >1 minutes with periodic mild agitation
- Rinse gently with water, blow dry with N2
9. Pattern Top Parylene (Step 2 – edge only)
9.1. Repeat step 8 using the outline etch mask and etching through any remaining Parylene
(thickness of the base Parylene)
9.2. Descum Wafer (optional)
9.2.1. Descum (clean) wafers in the RIE or Asher using the following recipe:
- 100 mT, 100 W, 50 sccm O2, 1-5 minutes
10. Release Devices
10.1. Place wafer into bath of 60 °C MIF 726 for 30-180 minutes with no agitation
- Time depends on size of devices – sacrificial aluminum dissolves from the
edges of the device, so larger areas will take longer to be released
- Released devices may get tangled when floating in the bath – take care to
remove them as they are released
10.2. As devices lift off of the wafer, move them to an IPA bath for >5 minutes
10.3. Move devices to a water bath for >3 minutes
10.4. Move devices to a second water bath for >3 minutes
10.5. Remove devices from water and store in foil-lined petri dish

198
C Parylene-Metal-Parylene Device Fabrication (Single-Sided)
Figure C-1: Process flow diagram for single-sided Parylene-metal-Parylene thin film devices.
1. Deposit Base Parylene C
1.1. Drybake wafers at 110 °C in an oven at atmosphere for at least 30 minutes (overnight
ok), or on a hot plate for >5 minutes
1.2. Deposit Parylene C
Note: this step should be performed immediately after drybake (step 1.1)
1.2.1. Determine amount of dimer by referencing past Parylene runs
- To produce flat devices when released, base and top Parylene should have
equal thickness
- To produce curled devices when released, see Thielen B and Meng E 2023
Characterization of thin film Parylene C device curvature and the formation
of helices via thermoforming J. Micromechanics Microengineering 33
095007
1.1.3. Deposit Parylene per the Parylene tool SOP
2. Anneal wafer (optional)
2.1. Place wafers in a vacuum oven, close, and evacuate chamber to 70 cmHg or greater
vacuum
2.2. Close vacuum valve, purge chamber with N2 to 20-30 cmHg, then re-evacuate to 70
cmHg or greater
2.3. Repeat step 2.2 twice (three total N2 purges)
2.4. Leave the vacuum valve open and open the N2 valve until 10-15 sccm of N2 are flowing
into the chamber
2.5. Bake wafers (under vacuum and N2 flow) for desired time and temperature
- Typical parameters for metal stress relaxation and adhesion: 150 °C, 4 hrs
- To produce flat devices, base Parylene and full wafer anneal should use the
same parameters

199
- To produce curled devices, see Thielen B and Meng E 2023 Characterization
of thin film Parylene C device curvature and the formation of helices via
thermoforming J. Micromechanics Microengineering 33 095007
3. Deposit and Pattern Metal
3.1. Inspect the metal photomask and clean (with acetone/IPA or Nanostrip) if necessary
3.2. Drybake wafers at 60 °C in an oven under light vacuum (35-40 cmHg) and N2 flow (15-
20 sccm) for >15 minutes
Note: this step can be skipped if the following step (3.3, deposit photoresist) is performed
immediately after annealing (step 2)
3.3. Deposit Photoresist (AZ 5214)
Note: this step should be performed immediately after annealing (step 2) or drybake (step
3.2)
3.3.1. Degas photoresist for 1 hour prior to spinning (open bottle and set it in the hood
with the lights off)
3.3.2. Coat 2 dummy wafers in spin coater prior to coating real wafers
Note: the remaining steps (3.3.3 through 3.3.6) are performed one wafer at a time; repeat
the following procedure once for each wafer
3.3.3. Place wafer onto the spinner chuck, center it, and engage vacuum to hold it
3.3.4. Dispense 5214 photoresist into a puddle on the center of the wafer
- ~1.5 inch diameter puddle
- Use more photoresist if surface is uneven to ensure sufficient coverage
3.3.5. Close the spinner lid and spin photoresist to ~1-2 µm thickness using the
following recipe:
- 7 s, 500 RPM, accl 8 (spreads out PR puddle)
- 45 s, 2000 RPM, accl 8 (defines desired thickness)
3.3.6. Move wafer to hot plate and soft bake at 90 °C for 70 seconds
3.4. Expose Photoresist
3.4.1. On the mask aligner, set the UV exposure dose to 36.75 mJ/cm2
3.4.2. Install metal mask into the mask aligner
Note: the remaining steps (3.4.3 through 3.4.9) are performed one wafer at a time; repeat
the following procedure once for each wafer
3.4.3. Install wafer in the wafer chuck and align to the mask pattern
3.4.4. Expose wafer through metal mask in hard contact mode
3.4.5. Bake at 110 °C for 45 seconds (image reversal bake)
3.4.6. Rest wafer for >3 minutes to cool down
3.4.7. On the mask aligner, set the UV exposure dose to 225 mJ/cm2
3.4.8. Flood expose wafer
3.4.9. Place wafer immediately into DI water bath after flood exposure for at least 2
minutes to prevent overheating
3.5. Develop Photoresist
Note: all steps are performed one wafer at a time; repeat the following procedure once
for each wafer

200
3.5.1. Prepare developer bath (1:4 ratio of AZ 340 developer to DI water) and DI water
rinse in separate plastic trays
- A fresh developer bath should be used for each wafer – do not re-use
developer
3.5.2. Place wafer in developer bath for the desired time (typically 16-19 seconds) with
mild agitation
3.5.3. Move quickly to water bath, then flush 3x with DI water
3.5.4. Blow dry with N2
3.5.5. Inspect developed features under microscope and develop for additional time if
needed
3.6. Descum Wafer
3.6.1. Descum (clean) wafers in the RIE or Asher using the following recipe:
- 100 mT, 100 W, 50 sccm O2, 1-5 minutes
3.7. Deposit Metal
Note: this step should be performed immediately after descum (step 3.6)
Note: different metals, thicknesses, and rates can be used; recommended values are
included in the steps below
3.7.1. Deposit the desired metal stackup an the e-beam evaporator at 1.5-2 Å/s
- Typical Ti/Pt stackup: 15 nm Ti (adhesion layer) + 200 nm Pt
- Typical Ti/Pt/Au/Pt stackup: 20 nm Ti (adhesion layer) + 25 nm Pt + 155 nm
Au + 25 nm Pt
3.7.2. Wait at least 30 minutes between different types of metal to allow crucible to
cool down
3.7.3. Do not deposit more than 50 nm of metal at a single time; if a layer is more than
50 nm, add a 30 minute pause between each 50 nm (e.g. for 200 nm Pt, deposit in
four 50 nm runs with 30 minute pauses between each run)
Note: the liftoff process has many options that can be used; perform only one of the following
liftoff procedures (step 3.8, step 3.9 OR step 3.10)
3.8. Pattern Metal via Liftoff (Option 1: Overnight Acetone w/ Sonication)
Note: this option is the most aggressive and should only be used if the base layer anneal
(step 2) was performed and there is good adhesion between the metal and Parylene
3.8.1. Place wafers in an acetone bath oriented vertically
3.8.2. Cover bath with aluminum foil, and leave wafers to soak overnight
Note: the remaining steps (3.8.3 through 3.8.10) are performed one wafer at a time;
repeat the following procedure once for each wafer
3.8.3. Lift wafer out of the acetone bath, rinse with acetone squeeze bottle, and move to
an NMP bath
- Use the squeeze bottle to rinse off any loose metal into the acetone bath
- Do not allow wafer to dry, as this will cause lifted-off metal to permanently
stick to the wafer surface
3.8.4. Place the NMP bath into the sonicating bath and sonicate for approximately 1-2
minutes (until all remaining metal has been lifted off)
3.8.5. Lift wafer out of the fresh NMP bath, rinse with NMP squeeze bottle, and move
to IPA bath for >5 minutes

201
- Do not allow wafer to dry, as this will cause lifted-off metal to permanently
stick to the wafer surface
3.8.6. Inspect the wafer for any remaining metal while submerged in IPA under the
stereoscope
- If any undesired metal remains, move wafer back to the NMP bath and repeat
process from step 3.8.4
- If any stubborn metal remains after repeating sonication, a foam swab can be
used to gently dislodge metal from the wafer surface
3.8.7. Lift wafer out of IPA bath, rinse with IPA squeeze bottle, and move to DI water
bath for >3 minutes
- Do not allow wafer to dry, as this will cause lifted-off metal to permanently
stick to the wafer surface
3.8.8. Re-inspect the wafer for any remaining metal while submerged in water under
the stereoscope
- If any undesired metal remains, move wafer back to the NMP bath and repeat
process from step 3.8.4
3.8.9. Rinse wafer with DI water 3 times, and blow dry with N2
3.8.10. Inspect metal features under microscope and return to step 3.8.4 if any metal or
photoresist remains
3.9. Pattern Metal via Liftoff (Option 2: Warm NMP w/ Sonication)
Note: this option is mildly aggressive and should only be used if there is sufficient
adhesion between the metal and Parylene
Note: all steps are performed one wafer at a time; repeat the following procedure once
for each wafer
3.9.1. Soak wafer in 60 °C NMP for 10-20 minutes with periodic mild agitation until
metal visibly lifts off of wafer
3.9.2. Hold the NMP bath (with the wafer inside) above the sonicating bath and
intermittently touch the sonicated water surface for intermittent sonication until
metal appears to fully lift off
3.9.3. Once liftoff appears to be complete, move NMP bath back to the hot plate
3.9.4. Lift wafer out of the warm NMP bath, rinse with NMP squeeze bottle, and move
to room temperature NMP bath for >5 minutes
- Do not allow wafer to dry, as this will cause lifted-off metal to permanently
stick to the wafer surface
3.9.5. Lift wafer out of the room temperature NMP bath, rinse with NMP squeeze
bottle, and move to IPA bath for >5 minutes
- Do not allow wafer to dry, as this will cause lifted-off metal to permanently
stick to the wafer surface
3.9.6. Inspect the wafer for any remaining metal while submerged in IPA under the
stereoscope
- If any undesired metal remains (metal that has not yet been lifted off or metal
flakes sitting on the surface), move wafer back to the warm NMP bath and
repeat process from step 3.9.2
- If any stubborn metal remains after repeating sonication, a foam swab can be
used to gently dislodge metal from the wafer surface

202
3.9.7. Lift wafer out of IPA bath, rinse with IPA squeeze bottle, and move to DI water
bath for >3 minutes
3.9.8. Re-inspect the wafer for any remaining metal while submerged in water under
the stereoscope
- If any undesired metal remains (metal that has not yet been lifted off or metal
flakes sitting on the surface), move wafer back to the warm NMP bath and
repeat process from step 3.9.2
3.9.9. Rinse wafer with DI water 3 times, and blow dry with N2
3.9.10. Inspect metal features under microscope and return to step 3.9.2 if any metal or
photoresist remains
3.10. Pattern Metal via Liftoff (Option 3: Warm Acetone w/o Sonication)
Note: this option is the least aggressive and will likely not work for liftoff of gold
Note: all steps are performed one wafer at a time; repeat the following procedure once
for each wafer
3.10.1. Soak wafer in 60 °C acetone for 10-20 minutes with periodic mild agitation until
metal visibly lifts off of wafer
- If any stubborn metal is not lifting off, a foam swab can be used to gently
dislodge metal from the wafer surface
3.10.2. Lift wafer out of the warm acetone bath, rinse with acetone squeeze bottle, and
move to room temperature acetone bath for >5 minutes
- Do not allow wafer to dry, as this will cause lifted-off metal to permanently
stick to the wafer surface
3.10.3. Lift wafer out of the room temperature acetone bath, rinse with acetone squeeze
bottle, and move to IPA bath for >5 minutes
- Do not allow wafer to dry, as this will cause lifted-off metal to permanently
stick to the wafer surface
3.10.4. Inspect the wafer for any remaining metal while submerged in IPA under the
stereoscope
- If any undesired metal remains (metal that has not yet been lifted off or metal
flakes sitting on the surface), move wafer back to the warm acetone bath and
repeat process from step 3.10.2
- If any stubborn metal remains after repeating sonication, a foam swab can be
used to gently dislodge metal from the wafer surface
3.10.5. Lift wafer out of IPA bath, rinse with IPA squeeze bottle, and move to DI water
bath for >3 minutes
- Do not allow wafer to dry, as this will cause lifted-off metal to permanently
stick to the wafer surface
3.10.6. Re-inspect the wafer for any remaining metal while submerged in water under
the stereoscope
- If any undesired metal remains (metal that has not yet been lifted off or metal
flakes sitting on the surface), move wafer back to the warm acetone bath and
repeat process from step 3.10.2
3.10.7. Rinse wafer with DI water 3 times, and blow dry with N2
3.10.8. Inspect metal features under microscope and return to step 3.10.2 if any metal or
photoresist remains
4. Deposit Top Parylene C

203
4.1. Descum Wafer
4.1.1. Descum (clean) wafers in the RIE or Asher using the following recipe:
- 100 mT, 100 W, 50 sccm O2, 1-5 minutes
4.2. Drybake wafers at 60 °C in an oven under light vacuum (35-40 cmHg) and N2 flow (15-
20 sccm) for >15 minutes
Note: do not perform this step if doing silanization (step 4.3)
4.3. Silanization (optional)
Note: this step is optional, but provides stronger adhesion between the metal and top
Parylene layers to decrease crosstalk and delamination
4.3.1. In the large A-174 beaker, prepare a mixture of 900 mL DI water, 900 mL
isopropanol, and 9 mL A-174 silane
- This volume is used for batches of 12 wafers; smaller quantities using the
same ratio can be used for smaller batches
- Stir mixture with a glass stirring rod
4.3.2. Cover the beaker with aluminum foil and let sit for at least 2.5 hours, but no more
than 24 hours
4.3.3. Decant the mixture into the crystalizing dish
4.3.4. Place wafers face up into the wafer cassette and lower the cassette into the
crystalizing dish, ensuring all wafers are fully submerged
4.3.5. Soak for 30 minutes
4.3.6. Remove the wafers and place on tex-wipes in the fume hood face up and air dry
for 30 minutes
4.3.7. Rinse the wafers thoroughly with IPA for 30 minutes
4.3.8. Blow dry with N2
4.3.9. Wafers should be coated with Parylene (step 4.4) within 12 hours after treatment
4.4. Deposit Parylene C
Note: this step should be performed immediately after drybake (step 4.2) or within 12
hours after silanization (4.3)
4.4.1. Label backside of each wafer using a permanent marker (below existing label)
with the date and which shelf it will be loaded on
4.4.2. If symmetric Parylene layers are desired, use the same amount of dimer and load
wafers onto the same shelf as the first Parylene run to increase likelihood of
symmetric layers
4.4.3. Determine amount of dimer by referencing past Parylene runs
- To produce flat devices when released, base and top Parylene should have
equal thickness
- To produce curled devices when released, see Thielen B and Meng E 2023
Characterization of thin film Parylene C device curvature and the formation
of helices via thermoforming J. Micromechanics Microengineering 33
095007
4.4.4. Deposit Parylene per the Parylene tool SOP
5. Anneal Wafer (optional)
5.1. Place wafers in a vacuum oven, close, and evacuate chamber to 70 cmHg or greater
vacuum
5.2. Close vacuum valve, purge chamber with N2 to 20-30 cmHg, then re-evacuate to 70
cmHg or greater
5.3. Repeat step 5.2 twice (three total N2 purges)

204
5.4. Leave the vacuum valve open and open the N2 valve until 10-15 sccm of N2 are flowing
into the chamber
5.5. Bake wafers (under vacuum and N2 flow) for desired time and temperature
- Typical parameters for adhesion: 150 °C, 4 hrs
- To produce flat devices, base Parylene and full wafer anneal should use the
same parameters
- To produce curled devices, see Thielen B and Meng E 2023 Characterization
of thin film Parylene C device curvature and the formation of helices via
thermoforming J. Micromechanics Microengineering 33 095007
6. Pattern Top Parylene (Step 1 - top open features and edge)
6.1. Inspect the frontside opening etch photomask and clean (with acetone/IPA or Nanostrip)
if necessary
6.2. Drybake wafers at 60 °C in an oven under light vacuum (35-40 cmHg) and N2 flow (15-
20 sccm) for >15 minutes
Note: this step can be skipped if the following step (6.3 deposit photoresist) is performed
immediately after annealing (step 5)
6.3. Deposit Photoresist (AZ P4620)
Note: this step should be performed immediately after drybake (step 6.2) or annealing
(step 5)
6.3.1. Degas photoresist for >1 hour prior to spinning (open bottle and set it in the hood
with the lights off)
6.3.2. Coat 2 dummy wafers in spin coater prior to coating real wafers
Note: the remaining steps (6.3.3 through 6.3.13) are performed one wafer at a time;
repeat the following procedure once for each wafer
6.3.3. Place wafer onto the spinner chuck, center it, and engage vacuum to hold it
6.3.4. Dispense P4620 photoresist into a puddle on the center of the wafer
- ~1.5 inch diameter puddle
- Use more photoresist if surface is uneven to ensure sufficient coverage
6.3.5. Close the spinner lid and spin photoresist to desired thickness using the following
recipe:
- 5 s, 500 RPM, accl 4 (spreads out PR puddle)
- 45 s, * RPM, accl 15 (see below for speeds; defines desired thickness)
- 2 s, 4500 RPM, accl 15 (edge bead removal)
* speed is selected using past measurements; thickness should be ~1.5-2x top
Parylene thickness
Common speeds: 1100 rpm → ~15 µm; 1300 rpm → ~13.5 µm; 3200 rpm →
~8 µm
Note: steps 6.3.6 through 6.3.11 (edge bead removal – EBR) are optional and can
alternatively be performed manually (without spinning) after exposure in step 4.6
6.3.6. Open spinner lid and lower EBR shield (black plastic cylinder) over wafer
without touching the wafer and place magnet over the lid sensor to override the
interlock
6.3.7. Soak a large foam swab in EBR solvent and blot away excess solvent from the
swab

205
6.3.8. Place the swab on the edge of the wafer at the 3 o’clock position such that the
swab is only in contact with the edge bead (no more than 5 mm away from the
edge of the wafer)
6.3.9. Spin the wafer using the following recipe:
- 5 s, 200 RPM, accl 4 (time to position swab)
- 40 s, 750 RPM, accl 4 (EBR time)
6.3.10. Position the swab during the first 5 seconds, leave the swab in contact with the
wafer for 20 seconds, then allow the wafer to spin dry for 20 seconds
6.3.11. Repeat step 6.3.7 through 6.3.10 if necessary until the edge bead has been
removed
6.3.12. Move wafer to hot plate and soft bake at 90 °C for 5 minutes (increase time for
thicker spins)
6.3.13. Let wafer sit at room temperature for >3 minutes (rehydration)
6.4. Expose Photoresist
6.4.1. On the mask aligner, set the UV exposure dose to 420 mJ/cm2
6.4.2. Install etch mask 1 into the mask aligner
Note: the remaining steps (6.4.3 through 6.4.5) are performed one wafer at a time; repeat
the following procedure once for each wafer
6.4.3. Install wafer in the wafer chuck and align to the mask pattern
6.4.4. Expose wafer through etch mask 1 in soft contact mode
6.4.5. Place wafer immediately into DI water bath after exposure for at least 2 minutes
to prevent overheating
6.5. Develop Photoresist
Note: all steps are performed one wafer at a time; repeat the following procedure once
for each wafer
6.5.1. Prepare developer bath (1:4 ratio of 340 developer to DI water) and DI water
rinse in separate plastic trays
- A fresh developer bath should be used for each wafer – do not re-use
developer
6.5.2. Place wafer in developer bath for desired time (typically 45-90 seconds) with
mild agitation
6.5.3. Move quickly to water bath, then flush 3x with DI water
6.5.4. Blow dry with N2
6.5.5. Inspect developed features under microscope and develop for additional time if
needed
6.6. Edge Bead Removal
Note: If edge bead removal was performed during step 6.3, skip this step and proceed to
step 6.7 or 6.8 (see note)
Note: all steps are performed one wafer at a time; repeat the following procedure once
for each wafer
6.6.1. Saturate a foam swab with acetone and dab any excess off the swab
6.6.2. Wipe all photoresist off the outer ~5 mm of the wafer using the swab, resaturating it in acetone as needed

206
Note: the etching process has two options that can be used. The DRIE (step 6.7) is preferred, but
the RIE (step 6.8) can be used if the DRIE is unavailable. Perform only one of the following
etching procedures (step 6.7 OR step 6.8).
6.7. Etch Parylene (DRIE)
Note: this step should etch through the thickness of the top Parylene layer (down to the
metal layer for any exposed metal features)
Equipment: DRIE
6.7.1. Calculate necessary etch loops based on prior runs and desired etch depth (typical
rate is ~0.07-0.1 µm/loop)
6.7.2. Etch wafers in the DRIE through the patterned photoresist
- Etch in steps of less than 50 loops to prevent overheating
- DRIE recipe developed in Meng E, Li P Y and Tai Y-C 2008 Plasma
removal of Parylene C J. Micromechanics Microengineering 18 045004
6.7.3. After each step, inspect wafers for any remaining Parylene in the etched areas
and continue etching as needed
6.8. Etch Parylene (RIE)
Note: this step should etch through the thickness of the top Parylene layer (down to the
metal layer for any exposed metal features)
Equipment: RIE
6.8.1. Calculate necessary etch time based on prior runs and desired etch depth (typical
rate is ~0.16-0.19 µm/minute)
6.8.2. Etch wafers in the RIE through the patterned photoresist using the following
parameters:
- 150 mT, 150 W, 50 sccm O2
- Perform in two or more steps of 15 minutes or less, rotating the wafer(s) 90-
180 degrees with each step
6.8.3. After each step, inspect wafers for any remaining Parylene in the etched areas
and continue etching as needed
6.9. Remove Remaining Photoresist
6.9.1. Strip remaining photoresist off each wafer per the following procedure:
- Soak wafer in an acetone bath for 30-60 seconds with mild agitation to
remove the majority of photoresist
- Move wafer to a second acetone bath and soak for >3 minutes with periodic
mild agitation
- Move wafer to IPA and soak for >3 minutes with periodic mild agitation
- Move mask to water and soak for >1 minutes with periodic mild agitation
- Rinse gently with water, blow dry with N2
7. Pattern Top Parylene (Step 2 – edge only)
7.1. Repeat step 4 using the outline etch mask and etching through any remaining Parylene
(thickness of the base Parylene)
7.2. Descum Wafer (optional)
7.2.1. Descum (clean) wafers in the RIE or Asher using the following recipe:
- 100 mT, 100 W, 50 sccm O2, 1-5 minutes
8. Release Devices
8.1. To remove a single device:

207
8.1.1. Place a droplet of water at the device edge (near a tab or other non-functional
feature)
8.1.2. Looking through a microscope, use a scalpel to gently lift the edge of the device,
allowing the water to wick between the Parylene and the wafer
8.1.3. Peel the device off the wafer using sharp tweezers, holding on to an open area
without critical features, allowing water to continue wicking underneath the
device as its lifted off
8.2. To remove all devices on a wafer:
8.2.1. Submerge the wafer in water
8.2.2. Devices should begin to lift off on their own, or you can peel the devices off
individually while submerged using sharp tweezers or a scalpel

208
D Annealing and Thermoforming of Parylene C
Note: fixturing for annealing and thermoforming depends on the desired final shape of the parts;
procedures for the three fixturing methods used in this document are described here (step 1 for flat
annealing (used in chapter 2), step 2 for helical thermoforming (used in chapters 2 and 3), and step 2.8
for cylindrical cuff thermoforming (used in chapter 4))
1. Fixture Parts for Annealing (Flat)
1.1. Place a piece of Teflon film on top of a glass slide
Note: the thickness of the Teflon film is not critical, but it should be thin enough to be
flexible and thick enough to prevent undesired folding; ~0.02-0.05 mm thickness
recommended
1.2. Place the Parylene device flat on top of the Teflon film and cover it with a second piece
of Teflon film
- If the Parylene device is curled, the top Teflon film can be used to hold it flat
while it is being unrolled
1.3. Place another glass slide on top of the glass-teflon-Parylene-teflon stack, taking care not
to shift the stack (which can result in bent or creased Parylene)
1.4. Clamp the full stack together using clips
Note: the glass-Teflon-device-Teflon-glass stackup does not need to be clamped (the
weight of the glass is sufficient to hold the device in place), however the clip is useful to
prevent shifting of the layers which can cause bends or creases in the Parylene
2. Fixture Parts for Thermoforming (Helical Electrodes, Manual Winding)
Note: this process requires a mandrel of the desired inner diameter of the helix and Teflon tape
(no adhesive)
2.1. Cut a piece of 0.01 mm thick Teflon tape (no adhesive) to approximately 5 cm longer
than and double the width of the Parylene device
2.2. Wrap the end of the Teflon tape (~5-10 mm) around the mandrel, securing it around itself
to hold it securely on the mandrel
2.3. Hold the mandrel horizontally with the Teflon tape underneath the mandrel and pointed
towards you, angled at the desired helix angle (relative to the mandrel)
2.4. Hold the Teflon tape tight and rotate the mandrel to wrap the tape around it for 1-2
revolutions
2.5. Place the Parylene device on top of the Teflon tape with the exposed electrodes facing
down and the tab on the electrode end of the device adjacent to the mandrel, taking care
to align it at the desired helix angle (relative to the mandrel)
2.6. While holding the Teflon tape tight, roll the mandrel towards you, rolling the Parylene
device around the mandrel, taking care to maintain the desired helix angle
2.7. Continue rolling until the entire Parylene device is rolled around the mandrel,
maintaining tension on the Teflon film and watching to ensure the Parylene does not fold
or kink
2.8. Continue rolling the mandrel until at least 1 cm of the remaining Teflon film has been
wound around the mandrel and the film is secured in place
3. Fixture Parts for Thermoforming (Helical Endovascular Electrodes, Winding with Fixture)
Note: this process requires a mandrel of the desired inner diameter of the helix, two custom
clamp assemblies, and tabs with holes on both ends of the part to be thermoformed
Note: A video showing the thermoforming process is available on the lab server at
/Data/Projects/Endovascular

209
3.1. Assemble the mandrel into the clamp as shown in Figure D-1. Both screws on the clamps
should be tight enough to prevent rotation of the clamp assembly with respect to the
mandrel.
Figure D-1: Assembly of the mandrel and clamp
3.2. Mount the Parylene device onto the first clamp by looping the holes in one Parylene tab
over the looped wire on the clamp (see Figure D-1).
3.3. Wrap the Parylene device around the mandrel a few times (if needed, e.g. if the Parylene
is pre-curled), then mount the Parylene device onto the second clamp by looping the
holes in the second Parylene tab over the looped wire on the clamp (see Figure D-2 A).
Ensure the Parylene is not twisted and will be able to smoothly lie on the mandrel when
tightened.

210
Figure D-2: The winding process, with main steps shown at left and detailed views shown at right.
3.4. Loosen one clamp slightly to allow rotation of the clamp around the mandrel (Figure D-2
A). Do not loosen it enough such that it will rotate on its own or slide along the mandrel
if the assembly is lifted off the table.
3.5. Begin rotating the clamp around the mandrel, inspecting the device as it is rotated to
ensure there are no kinks anywhere along the length of the Parylene device and that the
device is winding evenly along the mandrel (Figure D-2 B).
3.5.1. If any areas are kinked or winding unevenly, they can be adjusted gently using
tweezers or other tools.
3.6. Continue rotating and adjusting the Parylene (as needed) until the side of the device
adjacent to the semi-tight clamp is tightly wound around the mandrel.
3.7. Carefully tighten the semi-tight clamp, and loosen the tight clamp to semi-tight (Figure
D-2 C).
3.8. Rotate the semi-tight clamp, adjusting the Parylene (as needed) until the entire device is
snugly wrapped around the mandrel (Figure D-2 D). Do not over-tighten.
3.9. Carefully tighten the semi-tight clamp.
4. Fixture Parts for Thermoforming (Cuff Electrodes)
Note: this process requires 25 µm thick Teflon film, 64 µm thick (or thinner) Kapton tape, and a
mandrel 0.1-0.2 mm smaller than the desired inner diameter of the cuff; 1.5-2” long dispensing
tips of the desired diameter is recommended
Note: a video showing the thermoforming process is available on the OpenNerve website
(sites.usc.edu/carss/) and on the lab server at /Data/Projects/Endovascular
4.1. Prepare one fixture per part to be thermoformed using the following procedure
4.1.1. Cut a piece of Teflon film to 2 cm x 4 cm
4.1.2. Cut a piece of Kapton tape to 2 cm x 1 mm
4.1.3. Tape one of the 2 cm sides of the Teflon film along the length of the mandrel
using the Kapton tape

211
4.1.4. Clip the free end of the Teflon film to the short end of a glass slide so that the
film and mandrel are lying on the glass with the tape side down
4.1.5. Place the assembly underneath a microscope with the clip side towards you, prop
the free end of the glass slide up 3-5 mm, and push the slide against a heavy
block (so that you can push against it without moving the fixture in later steps);
see Figure D-3 for full fixture setup
Figure D-3: Cuff thermoforming fixture setup (cross section, shown in red dotted box, is not to scale).
4.2. Lift the mandrel away from the glass slide (while still clipped in place), lay the Parylene
cable beneath the Teflon film (with bondpad region to the left and electrodes facing up),
and lower the Teflon film on top of the cable
4.3. Slide the electrodes over the Teflon film and align them near the mandrel (see Figure D4)
Figure D-4: Initial placement of the device on the thermoforming fixture.

212
4.4. Place a small droplet of IPA underneath the electrode region of the Parylene device and
touch the Parylene down to the Teflon film, allowing it to wick between the Parylene and
Teflon and hold it in place
4.5. Align the Parylene electrode such that the center bridge is parallel to the mandrel, the top
edge of the electrodes are touching the mandrel, and the tail is extended over the edge of
the Teflon film (see Figure D-5)
Figure D-5: Alignment of the device on the thermoforming fixture, with (right) magnified view.
4.6. Hold the Teflon film tight by gently pulling the mandrel away from you and against the
heavy block, then roll the mandrel towards you, rolling the Parylene electrodes around
the mandrel (see Figure D-6)
Figure D-6: Device partially rolled into the thermoforming fixture, with (right) magnified view.
4.7. Continue rolling until the entire Parylene electrode is covered, maintaining tension on the
Teflon film and watching to ensure the Parylene does not fold or kink
4.8. Gently lift the mandrel while maintaining tension in the Teflon film and move the
Parylene cable around the back of the mandrel and on top of the Teflon film (untangling
it, see Figure D-7)

213
Figure D-7: Device fully rolled around the thermoforming fixture, with (right) magnified view.
4.9. Roll the remainder of the Teflon film around the mandrel, leaving the tail unrolled, and
hold film in place using two small pieces of Kapton tape
4.10. While maintaining tension on the film by holding the mandrel on the right side, place a
clip on the left side of the mandrel, clamping it to the glass slide
4.11. Gently rearrange the Parylene cable with wires attached so that the wires are parallel to
the long end of the glass slide and there is slack in the Parylene cable, and tape the wires
down to the glass slide using Kapton tape, taking care not to tug on the Parylene cable
and cause damage to the device (see Figure D-8)
Figure D-8: Device rolled around the thermoforming fixture and clamped on the left side of the mandrel, with wires taped to
glass slide with some slack in the Parylene cable.
4.12. Add another clip on the right side of the mandrel, taking care not to clip directly on top of
the Parylene cable
4.13. Remove the clip holding the Teflon film, and gently pull the film taut by hand to remove
any slack that was introduced during the clamping process (see Figure D-9)

214
Figure D-9: Fixtured device clamped on both sides of the mandrel while Teflon film is gently being pulled to remove any slack.
4.14. While holding the film taut, place one clip to hold the Teflon film to the glass slide as
close to the rolled device as possible without causing damage, then release the film and
place a second clip on the other side of the Teflon film (see Figure D-10)
Figure D-10: Fixtured device with two clamps holding the mandrel to the glass slide and two clamps holding the Teflon film to
the glass slide.
5. Bake Fixtured Parts
5.1. Carefully place the fixtured parts in a vacuum oven, close, and evacuate chamber to 70
cmHg or greater vacuum
5.2. Purge chamber with N2 to 20-30 cmHg, then re-evacuate to 70 cmHg or greater
5.3. Repeat step 5.2 twice (three total N2 purges)
5.4. Bake wafers or fixtured parts (under vacuum) for the specified time and temperature
5.4.1. Most on-wafer anneals are done for 4 hours at 150 °C
5.4.2. Most thermoforming steps are done for 12 hours at 100 °C or 150 °C
5.4.3. See detail in specific chapters for exact baking parameters
5.5. Remove fixtured parts from the oven
6. Remove parts from their fixture by reversing the fixturing process used

215
- Perform this step slowly, gently separating the Parylene device from the
fixture if it sticks
- If present, handle the device by the handling tabs whenever possible

216
E Photomasks for Endovascular Electrode Arrays
E.1 Generation 1 Design
Mask files can be found on the laboratory server at /Data/Projects/Endovascular/Masks/V1.

217

218
E.2 Generation 2 Design
Mask files can be found on the laboratory server at /Data/Projects/Endovascular/Masks/V2.

219

220

221
F Peripheral Nerve Cuff Backbone Design Evaluation
All design and fabrication documents are open source and included on the CARSS OpenNerve
website (sites.usc.edu/carss/).
F.1 Design Options
For the peripheral nerve cuff, five backbone shapes were considered, and four were fabricated for
testing. The first backbone (design A, a U-shaped backbone) was removed from consideration prior to
fabrication due to anticipated surgical difficulty. Designs B and C contain helical electrodes wrapped in
opposite directions (B) and the same direction (C) with a narrow bridge of Parylene between them to
define the inter-electrode distance and to hold the traces. Designs D and E contain circumferential
electrodes with a bridge of Parylene between them (D) and on a full Parylene cuff (E). Each of the
fabricated backbone designs are shown in Table F-1.
Table F-1: Summary of considered backbone shapes in flat (as fabricated) and thermoformed (as implanted) configurations.
Parylene is shown in light gray; electrode area is shown in dark gray (aggregate electrode design not shown for simplicity).
Backbone
Description Flat Configuration Thermoformed Configuration
A U-shaped Not Fabricated Not Fabricated
B
I-shaped,
tilted,
opposite
directions
C
I-shaped,
tilted, same
direction
D
I-shaped,
straight

222
E Full cuff
To connect the cuff backbone to the lead, a tail must be attached to the backbone to connect metal
traces from each electrode to bondpads at the end of the tail. The tail was designed to have a length of at
least 18 mm from the edge of the electrode to provide strain relief and prevent movement of the cuff
while the lead is being anchored during the implantation procedure.
Three tail designs were considered, and two were fabricated for testing. The first tail (design 1, a
linear tail) was removed from consideration prior to fabrication due to poor strain relief. Design 2 was a
centered tail extending from the center of the bridge between the electrodes, then turning 90 degrees to
run parallel to the nerve. Design 3 was a u-turned tail extending from the end of one electrode, then
making a U-turn (180 degrees) to run parallel to the nerve in the opposite direction.
A serpentine strain relief feature was also considered (design s) which could be added to either
tail design. The addition of an auxetic serpentine structure would allow more stretching and movement of
the tail. Each of the fabricated tail designs are shown in Table F-2.
Table F-2: Summary of considered tail designs attached to backbone B in flat (as fabricated) and thermoformed (as implanted)
configurations. Parylene is shown in light gray; electrode area is shown in dark gray (aggregate electrodes not depicted).
Tail Description Flat Configuration
(with backbone B)
Thermoformed Configuration
(with backbone B)
1 Linear Not Fabricated Not Fabricated
2 Centered

223
3 U-turn
s Serpentine
Insulation of the nerve between the electrodes helps to constrain the electric field during
stimulation. Insulation can be added with Parylene (as part of the backbone) or with an extra material,
such as PDMS, which is either attached to the cuff or added after implantation.
Two Parylene insulation designs were considered and fabricated for testing. The first option is to
use a full Parylene cuff, as is already included in backbone E. The second option is to add Parylene fins
(design f) between the electrodes (compatible with designs A-D) to maintain flexibility of the cuff but add
insulation between the electrodes. These insulation designs are shown in Table F-3.
Table F-3: Summary of considered insulation designs in flat (as fabricated) and thermoformed (as implanted) configurations.
Parylene is shown in light gray; electrode area is shown in dark gray (aggregate electrodes not depicted).
Insulation
Description
Flat Configuration
(with backbone B)
Thermoformed Configuration
(with backbone B)
Backbone E

224
f Fins
F.2 Design Evaluation Methods
The large number of design combinations from the 4 backbones, 2 tails, and 2 other features
could not be exhaustively tested, so a subset of 8 designs which spanned all design options were selected.
Designs are numbered using their design labels (summarized in Table F-4). The designs were selected to
allow direct comparison between similar designs with one component changed. For backbone shape, B2-,
C2-, D2-, and E3- were compared to each other (E2- was not a possible design – a centered tail would
interfere with the full cuff); for tails, C2- and D2f were compared to C3- and D3f; for serpentine strain
relief, B2- was compared to B2s; for fins, D2- was compared to D2f.
Table F-4: Summary of fabricated sham cuff designs.
Backbone
1
Tail
2
Other Feature
3
Description
B 2 - I-shaped/tilted/opposite with centered tail
B 2 s I-shaped/tilted/opposite with centered tail and serpentine strain relief
C 2 - I-shaped/tilted/same with centered tail
C 3 - I-shaped/tilted/same with u-turn tail
D 2 - I-shaped/straight with centered tail
D 2 f I-shaped/straight with fins and centered tail
D 3 f I-shaped/straight with fins and u-turn tail
E 3 - Full cuff with u-turn tail
1 Backbone Labels: 2 Tail Labels: 3 Other Feature Labels:
(A) U-shaped (1) End (s) Serpentine Strain Relief
(B) I-shaped (tilted/opposite) (2) Centered (f) Fins
(C) I-shaped (tilted/same) (3) U-turn (-) None
(D) I-shaped (straight)
(E) Full cuff

225
A design evaluation protocol was developed and performed to quantitatively rank each of the
design components and select the two best design options. The cuff designs listed in Table F-4 were
fabricated as sham devices with no functional metal layer to evaluate only the mechanical design. To
evaluate pre-curled parts (which are made possible via asymmetric Parylene layers and annealing steps, as
described in chapter 2), three batches of parts were evaluated with varying base layer annealing
temperature. Two samples per design per annealing temperature were fabricated (n=6 per design). Any
parts which had failures during the thermoforming step were repeated for a third sample (such that two
parts could move forward with further testing). The full fabrication and testing protocol is described in the
following sections.
F.2.1 Thin Film Fabrication
Sham Parylene cuffs were fabricated using a low temperature, batch process. 4” prime silicon
wafers were used as the carrier substrate. The Parylene C base layer was deposited using a chemical vapor
deposition-like process (PDS 2010 Labcoter, Specialty Coating Systems, Indianapolis, IN) to a thickness
of 3.3 µm. The coated wafers were annealed in an oven (TVO-2, Cascade Tek Inc., Longmont, CO) under
vacuum with nitrogen flow at 60, 100, or 150 °C (one wafer each) for 4 to 6 hours. Next, a top layer of
Parylene C was deposited to a thickness of 5.3 µm. The outline of the cuffs were etched using O2
switched chemistry etching in a deep reactive ion etcher (PlasmaPro 80 or PlasmaLab 100 ICP, Oxford
Instruments, Bristol, UK; etch rate approximately 0.08 µm/loop) masked by patterned photoresist (AZ
P4620, AZ Electronic Materials, Branchburg, NJ; 8-12 µm thick). Photoresist was removed using acetone
after etching followed by rinsing in isopropyl alcohol and deionized water. Devices were kept on the
carrier wafer until immediately prior to thermoforming.
F.2.2 Release from Wafer and Thermoforming
Immediately prior to thermoforming, the Parylene cuffs were released from the wafer by placing
a droplet of water over one of the handling tabs (which were added to the edge of each device design),
manually separating the tab from the wafer using a scalpel or sharp tweezers, and gently pulling the
device off the wafer using the tab as a handling point. The Parylene backbone was then wrapped around a

226
0.5 mm mandrel manually and held in place using 0.025 mm thick Teflon film. After fixturing, the parts
were placed into a programmable vacuum oven (TVO-2, Cascade Tek Inc., Longmont, CO), placed under
vacuum, then purged three times with nitrogen to minimize oxygen in the chamber. The oven was
programmed to ramp up to 100 °C at a ramp rate of approximately 0.7 °C/min, hold for 12 hours, then
ramp down to room temperature (see appendix D for the detailed thermoforming procedure). After
removing from the oven, parts were separated from the fixture and inspected using a stereoscope
(HD60T, Caltex Scientific, Irvine, CA) to evaluate the cuff shape and a high magnification microscope
(Eclipse LV100, Nikon, Tokyo, Japan) to observe any cracks in the Parylene.
The fixturing process (from removal of the device from the wafer through securing the Teflon
film) was timed for each sample to identify any geometries which are more difficult or time consuming to
fixture for thermoforming (test 1). Thermoformed parts were inspected for shape failures and the shape
yield for each design was calculated (test 2a). Any parts which did not conform to the desired shape (0.5
mm cuff) were considered failures and did not proceed with the remaining testing. Passing parts were
inspected for cracks in the Parylene (examples shown in chapter 2.4) at 10 or 11 inspection points (see
Figure F-1) and the total length of cracks in a 0.25 x 0.25 mm square area at each inspection point was
measured to identify designs which were more prone to handling damage (test 2b). Test 2b also served to
balance any errors introduced in test 1 (for example, if a part was handled too quickly and scored falsely
well in test 1, test 2b would identify the damage from the expedited handling and score poorly). Visual
inspection also served to identify any points on the device which are more prone to handling damage.
Figure F-1: Inspection points for cracking of thermoformed Parylene cuffs.

227
F.2.3 Simulated Implant
The implant procedure was simulated with the thermoformed cuff using an interim implantation
tool and a nerve phantom. The design for the implantation tool is in development, so an interim tool was
built by attaching 0.25 mm diameter wires to the tips of a pair of angled tweezers. The tips of the tool
were threaded into the center of the cuff and opened to expand the cuff, the cuff was placed over the
nerve, and the tool was closed and retracted off of the cuff, closing the cuff around the nerve. For the
nerve phantom, a vermicelli rice noodle was cooked and allowed to soak in warm water until it reached a
diameter of 0.6-0.9 mm (the noodle nerve phantom was recommended by several experts in the field). It
was then removed from the water and suspended between two stacks of glass slides and rehydrated as
needed using drops of water.
One sample of each design was implanted onto the nerve phantom several times in random order
over the course of several days. Due to the inconsistency of the interim implantation tool, a quantitative
comparison between designs (e.g. comparing the amount of time required to implant the cuff) was not
possible, so instead the designs were qualitatively ranked from easiest to most difficult to implant (test 3a.
During this process, the motion of the device relative to the nerve (i.e. if the implanted location of the cuff
shifted) was also observed qualitatively (test 3b).
After the simulated implant, the cuffs were re-inspected (using the same procedures as used for
tests 2a and 2b) for any changes in shape or additional damage caused by implantation (test 5a and 5b).
F.2.4 Design Scoring
Each of the tests described in the previous sections yielded quantitative or qualitative rankings of
each design which were scaled to produce a score from 0 to 1 (worst to best) for each design and each
design component (score calculations are included in Table F-8). Tests were also weighted (see Table F5), with higher weights given to more important or more significant tests, resulting in a total design score
described by equation E-1.
𝐷𝑒𝑠𝑖𝑔𝑛 𝑆𝑐𝑜𝑟𝑒 = ∑ 𝑇𝑒𝑠𝑡 𝑊𝑒𝑖𝑔ℎ𝑡 × 𝑇𝑒𝑠𝑡 𝑆𝑐𝑜𝑟𝑒 (E-1)

228
Table F-5: Weights given for each test. *Cuff movement was given a weight of 0 because no cuff movement was observed.
Test Weight
1 Fabrication Difficulty 1
2a Post-Thermoform Inspection – Shape 1
2b Post-Thermoform Inspection – Cracking 3
3a Implantation Difficulty 1
3b Cuff Movement 0*
4a Post-Implant Inspection – Shape 1
4b Post-Implant Inspection – Cracking 1
F.3 Design Evaluation Results
The design selection protocol was performed on 6 parts for each design and yielded the following
results. The tables in this section show final scoring; the full dataset is included in section E.4.
Although the ranking for tested designs (shown in Table F-6) indicated that designs E3- and D2f
performed best with scores of 0.82 and 0.83, this list of designs is not exhaustive and scores for individual
design components should be considered separately to select the best design.
Table F-6: Design scoring for the 8 tested designs. Colors scale from red (0 - worst score) to green (1 - best score). TF =
thermoforming.
Test Tested Designs
B2- B2s C2- C3- D2- D2f D3f E3-
1 (Fabrication Difficulty) 0.58 0.74 0.59 0.74 0.85 0.81 0.65 0.85
2a (Post TF Shape) 1.00 1.00 1.00 1.00 1.00 0.86 1.00 1.00
2b (Post TF Cracking) 0.86 0.81 0.53 0.74 0.57 1.00 0.75 0.71
3a (Implant Difficulty) 0.43 0.29 0.43 0.57 0.71 0.00 0.14 0.86
3b (Movement) 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00
4a (Post Implant Shape) 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00
4b (Post Implant Cracking) 0.63 0.83 0.82 0.59 0.65 0.96 0.83 0.70
Weighted Aggregate 0.77 0.78 0.68 0.77 0.74 0.83 0.74 0.82
Scores for each design component (backbone, tail, other features, and annealing temperature)
were calculated separately by comparing similar designs (as described in section E.2). These scores are
summarized in Table F-8. All backbone shapes performed similarly, with backbone E scoring highest at
0.84, closely followed by backbone B at 0.78. Tail 3 outperformed tail 2 in almost all categories except
for post-implant cracking (test 6b). This was likely an outlier given the difficulty of the implantation
process with the interim implantation tool producing inconsistent results. Due to this inconsistency, the
weights for tests 4 and 6b were reduced to 1 to have a lesser impact on the design decisions. Although the

229
serpentine structure and fins increased scores overall, they were significantly more difficult to implant,
with finned devices failing the implant process on several occasions. As such, these two features were not
included in the final design but remain in consideration for future device generations. Parts with a 100 °C
base anneal (and resulting moderate curvature when taken off the wafer) scored better than those with a
low temperature (flat) or high temperature (high curl) anneal. Due to the lower sample sizes for all tests
beyond the post-thermoforming inspection, scores for different base annealing values for tests 3a through
5b were not calculated.
In functional parts (which contain a metal layer and etched openings in the top Parylene layer),
the moderate curl will be added to the device using asymmetric Parylene layers and anneals to mimic the
curvature from the 100 °C base anneal in the sham parts. Base annealing results in curvature towards the
base layer; this project requires curvature towards the top layer such that electrodes (which are opened in
the top Parylene) can be on the inside of the cuff and facing the nerve. The curvature towards the base
layer due to the 100 °C base anneal in the test parts was calculated using the model described in chapter 2.
This curvature can be achieved towards the top layer by adjusting the ratio of the base to top layer
thickness and adding an annealing step after the top Parylene is deposited but prior to removing devices
from the wafer. A comparison of parameters used for the test parts and in the final design are shown in
Table F-7.
Table F-7: Comparison of parameters between sham test parts and the final cuff design.
Parameter Value (Test Parts) Value (Final Design)
Base Parylene
Width 1200 µm 880 µm
Thickness 3.3 µm 5 µm 1
Stress 9.4 MPa (100 °C anneal) 2 17.5 MPa (150 °C anneal)
2
Top Parylene
Width 1200 µm 580 µm 3
Thickness 5.3 µm 7 µm 1
Stress 7.8 MPa (90 °C processing)
2 17.5 MPa (150 °C anneal)
2
Metal
Width 0 (no metal layer) 568 µm 3
Thickness 0 (no metal layer) 220 nm
Stress 0 (no metal layer) -100 MPa 2
Resulting Curvature
(Modeled) -21 mm (towards base) 21 mm (towards top)
1 Total Parylene thickness is 12 µm to prevent cracking when thermoformed to 0.5 mm diameter
2 From fit data at the given processing temperature (see chapter 2)

230
3 Average of width across microelectrodes and between microelectrodes
Table F-8: Component scoring for each individual design component (backbone shape, tail, other features, and base anneal).
Colors scale from red (0 - worst score) to green (1 - best score). (Note: table is broken into two parts). TF = thermoforming.
Test Backbone Shapes Tail
B C D E 2 3
1 (Fabrication Difficulty) 0.58 0.67 0.85 0.85 0.69 0.69
2a (Post TF Shape) 1.00 1.00 1.00 1.00 0.92 1.00
2b (Post TF Cracking) 0.85 0.65 0.57 0.71 0.54 0.94
3a (Implant Difficulty) 0.43 0.50 0.71 0.86 0.21 0.36
3b (Movement) 1.00 1.00 1.00 1.00 1.00 1.00
4a (Post Implant Shape) 1.00 1.00 1.00 1.00 1.00 1.00
4b (Post Implant Cracking) 0.66 0.83 0.65 0.86 1.00 0.55
Weighted Aggregate 0.78 0.74 0.74 0.84 0.68 0.80
Table F-8 (continued): Component scoring for each individual design component (backbone shape, tail, other features, and base
anneal). Colors scale from red (0 - worst score) to green (1 - best score). (Note: table is broken into two parts). TF =
thermoforming.
Test Other Features Base Anneal
-(s) s -(f) f 60°C 100°C 150°C
1 (Fabrication Difficulty) 0.58 0.74 0.85 0.81 0.84 0.87 0.49
2a (Post TF Shape) 1.00 1.00 1.00 0.86 1.00 0.94 1.00
2b (Post TF Cracking) 0.86 0.81 0.57 1.00 0.49 0.75 0.65
3a (Implant Difficulty) 0.43 0.29 0.71 0.00 n/a n/a n/a
3b (Movement) 1.00 1.00 1.00 1.00 n/a n/a n/a
4a (Post Implant Shape) 1.00 1.00 1.00 1.00 n/a n/a n/a
4b (Post Implant Cracking) 0.63 0.83 0.65 0.96 n/a n/a n/a
Weighted Aggregate 0.77 0.78 0.74 0.83 0.66 0.81 0.69

231
F.4 Detailed Scoring of Design Selection Tests
Table E-8: Score calculations for each design selection test.
Test Score Calculation
1
Fabrication
Difficulty
1 −
𝑎𝑣𝑔 𝑓𝑖𝑥𝑡𝑢𝑟𝑖𝑛𝑔 𝑡𝑖𝑚𝑒(𝑠𝑖𝑛𝑔𝑙𝑒 𝑑𝑒𝑠𝑖𝑔𝑛) − 𝑚𝑖𝑛 𝑓𝑖𝑥𝑡𝑢𝑟𝑖𝑛𝑔 𝑡𝑖𝑚𝑒(𝑎𝑙𝑙 𝑝𝑎𝑟𝑡𝑠)
𝑚𝑎𝑥 𝑓𝑖𝑥𝑡𝑢𝑟𝑖𝑛𝑔 𝑡𝑖𝑚𝑒(𝑎𝑙𝑙 𝑝𝑎𝑟𝑡𝑠) − 𝑚𝑖𝑛 𝑓𝑖𝑥𝑡𝑢𝑟𝑖𝑛𝑔 𝑡𝑖𝑚𝑒(𝑎𝑙𝑙 𝑝𝑎𝑟𝑡𝑠)
2a
PostThermoform
Inspection –
Shape
𝑦𝑖𝑒𝑙𝑑(𝑠𝑖𝑛𝑔𝑙𝑒 𝑑𝑒𝑠𝑖𝑔𝑛)
2b
PostThermoform
Inspection –
Cracking
1
2
((1 −
𝑚𝑎𝑥 𝑐𝑟𝑎𝑐𝑘𝑠 𝑝𝑒𝑟 𝑎𝑟𝑒𝑎(𝑠𝑖𝑛𝑔𝑙𝑒 𝑑𝑒𝑠𝑖𝑔𝑛)
𝑚𝑎𝑥 𝑐𝑟𝑎𝑐𝑘𝑠 𝑝𝑒𝑟 𝑎𝑟𝑒𝑎(𝑎𝑙𝑙 𝑝𝑎𝑟𝑡𝑠)
)
+ (1 −
𝑎𝑣𝑔 𝑐𝑟𝑎𝑐𝑘𝑠 𝑝𝑒𝑟 𝑎𝑟𝑒𝑎(𝑠𝑖𝑛𝑔𝑙𝑒 𝑑𝑒𝑠𝑖𝑔𝑛)
𝑚𝑎𝑥 𝑐𝑟𝑎𝑐𝑘𝑠 𝑝𝑒𝑟 𝑎𝑟𝑒𝑎(𝑎𝑙𝑙 𝑝𝑎𝑟𝑡𝑠)
))
3a Implantation
Difficulty
1 −
𝑟𝑎𝑛𝑘𝑖𝑛𝑔(𝑠𝑖𝑛𝑔𝑙𝑒 𝑑𝑒𝑠𝑖𝑔𝑛) − ℎ𝑖𝑔ℎ𝑒𝑠𝑡 𝑟𝑎𝑛𝑘(𝑎𝑙𝑙 𝑝𝑎𝑟𝑡𝑠)
ℎ𝑖𝑔ℎ𝑒𝑠𝑡 𝑟𝑎𝑛𝑘(𝑎𝑙𝑙 𝑝𝑎𝑟𝑡𝑠)
3b Cuff
Movement
1 −
𝑎𝑣𝑔 𝑚𝑜𝑣𝑒𝑚𝑒𝑛𝑡 𝑑𝑖𝑠𝑡𝑎𝑛𝑐𝑒(𝑠𝑖𝑛𝑔𝑙𝑒 𝑑𝑒𝑠𝑖𝑔𝑛)
𝑚𝑎𝑥 𝑚𝑜𝑣𝑒𝑚𝑒𝑛𝑡 𝑑𝑖𝑠𝑡𝑎𝑛𝑐𝑒(𝑎𝑙𝑙 𝑝𝑎𝑟𝑡𝑠)
4a
Post-Implant
Inspection –
Shape
𝑦𝑖𝑒𝑙𝑑(𝑠𝑖𝑛𝑔𝑙𝑒 𝑑𝑒𝑠𝑖𝑔𝑛)
4b
Post-Implant
Inspection –
Cracking
1
2
((1 −
𝑚𝑎𝑥 𝑐𝑟𝑎𝑐𝑘𝑠 𝑝𝑒𝑟 𝑎𝑟𝑒𝑎(𝑠𝑖𝑛𝑔𝑙𝑒 𝑑𝑒𝑠𝑖𝑔𝑛)
𝑚𝑎𝑥 𝑐𝑟𝑎𝑐𝑘𝑠 𝑝𝑒𝑟 𝑎𝑟𝑒𝑎(𝑎𝑙𝑙 𝑝𝑎𝑟𝑡𝑠)
)
+ (1 −
𝑎𝑣𝑔 𝑐𝑟𝑎𝑐𝑘𝑠 𝑝𝑒𝑟 𝑎𝑟𝑒𝑎(𝑠𝑖𝑛𝑔𝑙𝑒 𝑑𝑒𝑠𝑖𝑔𝑛)
𝑚𝑎𝑥 𝑐𝑟𝑎𝑐𝑘𝑠 𝑝𝑒𝑟 𝑎𝑟𝑒𝑎(𝑎𝑙𝑙 𝑝𝑎𝑟𝑡𝑠)
))

232
Table E-9: Results from test 1 (fabrication difficulty). Numbers quoted are the average time to fixture the device for
thermoforming (in minutes). The design score is calculated per the formula in Table E-8. The sample size for this test was 6
samples per design. Parts which took longer time than 2 standard deviations (SD) above the mean were not considered in score
calculations.
Tested Designs
B2- B2s C2- C3- D2- D2f D3f E3-
Average time to fixture 3.923 3.204 3.840 3.190 2.722 2.858 3.600 2.714
DESIGN SCORE 0.575 0.737 0.594 0.740 0.845 0.815 0.648 0.847
Backbone Shapes Cable Routing
B C D E 2 3
Average time to fixture 3.923 3.515 2.722 2.714 3.397 3.395
DESIGN SCORE 0.575 0.667 0.845 0.847 0.694 0.694
Other Features Base Anneal
-(s) s -(f) f 60 C 100 C 150 C
Average time to fixture 3.923 3.204 2.722 2.858 2.750 2.590 4.321
DESIGN SCORE 0.575 0.737 0.845 0.815 0.839 0.875 0.486
Average fixturing time (all parts) 3.659 minutes
Standard deviation (all parts) 2.011 minutes
Minimum fixturing time (all parts, without outliers) 2.033 minutes
Maximum fixturing time (all parts, without
outliners) 6.483 minutes

233
Table E-10: Results from test 2a (thermoforming shape yield). Numbers quoted are the thermoforming yield. The design score is
calculated per the formula in Table E-8. The sample size for this test was 6 samples per design. Parts which had fixturing failures
not related to the backbone design were not considered in score calculations.
Tested Designs
B2- B2s C2- C3- D2- D2f D3f E3-
Thermoforming shape
yield 1.000 1.000 1.000 1.000 1.000 0.857 1.000 1.000
DESIGN SCORE 1.000 1.000 1.000 1.000 1.000 0.857 1.000 1.000
Backbone Shapes Cable Routing
B C D E 2 3
Thermoforming shape
yield 1.000 1.000 1.000 1.000 0.923 1.000
DESIGN SCORE 1.000 1.000 1.000 1.000 0.923 1.000
Other Features Base Anneal
-(s) s -(f) f 60 C 100 C 150 C
Thermoforming shape
yield 1.000 1.000 1.000 0.857 1.000 0.938 1.000
DESIGN SCORE 1.000 1.000 1.000 0.857 1.000 0.938 1.000

234
Table E-11: Results from test 2b (post-thermoforming cracking). Numbers quoted are the total length of cracks (in mm) in a 0.25
x 0.25 mm square area at the given inspection point. The design score is calculated per the formula in Table E-8. The sample size
for this test was 6 samples per design. Parts which had more cracks than 2 standard deviations (SD) above the mean were not
considered in score calculations.
Tested Designs
B2- B2s C2- C3- D2- D2f D3f E3-
Average length of cracks 0.017 0.010 0.056 0.018 0.025 0.000 0.018 0.011
Maximum length of
cracks 0.336 0.464 1.090 0.609 1.022 0.000 0.582 0.698
DESIGN SCORE 0.855 0.806 0.531 0.743 0.572 1.000 0.755 0.710
Backbone Shapes Cable Routing
B C D E 2 3
Average length of cracks 0.025 0.042 0.035 0.017 0.039 0.002
Maximum length of
cracks 0.336 0.817 1.022 0.698 1.090 0.149
DESIGN SCORE 0.852 0.648 0.568 0.708 0.538 0.938
Other Features Base Anneal
-(s) s -(f) f 60 C 100 C 150 C
Average length of cracks 0.017 0.010 0.025 0.000 0.030 0.007 0.028
Maximum length of
cracks 0.336 0.464 1.022 0.000 1.222 0.609 0.817
DESIGN SCORE 0.855 0.806 0.572 1.000 0.488 0.748 0.654
Inspection Points
1A 1B 1C 2A 2B 2C
Average length of cracks 0.017 0.024 0.020 0.000 0.067 0.021
Maximum length of
cracks 0.450 0.698 0.582 0.000 1.022 0.694
DESIGN SCORE 0.809 0.705 0.754 1.000 0.555 0.708
Inspection Points (continued)
3 4A 4B 4C 4D
Average length of cracks 0.033 0.003 0.000 0.000 0.027
Maximum length of
cracks 0.817 0.134 0.000 0.000 1.090
DESIGN SCORE 0.652 0.944 1.000 1.000 0.543
Average length of cracks (all nonzero values) 0.486 mm
Standard deviation (all nonzero values) 0.332 mm
Maximum length of cracks (all parts) 1.222 mm
Table E-12: Results from test 3a (implantation difficulty). Numbers quoted are the difficulty ranking. The design score is
calculated per the formula in Table E-8. The sample size for this test was 1 sample per design.
Tested Designs
B2- B2s C2- C3- D2- D2f D3f E3-
Average time to fixture 4 5 4 3 2 7 6 1
DESIGN SCORE 0.429 0.286 0.429 0.571 0.714 0.000 0.143 0.857
Backbone Shapes Cable Routing

235
B C D E 2 3
Average time to fixture 4 3.5 2 1 5.5 4.5
DESIGN SCORE 0.429 0.500 0.714 0.857 0.214 0.357
Other Features Base Anneal
-(s) s -(f) f 60 C 100 C 150 C
Average time to fixture 4 5 2 7
Insufficient Data DESIGN SCORE 0.429 0.286 0.714 0.000
Table E-13: Results from test 4a (post-implant shape yield). Numbers quoted are the shape yield after implantation testing. The
design score is calculated per the formula in Table E-8. The sample size for this test was 1 sample per design.
Tested Designs
B2- B2s C2- C3- D2- D2f D3f E3-
Post-implant shape yield 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000
DESIGN SCORE 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000
Backbone Shapes Cable Routing
B C D E 2 3
Post-implant shape yield 1.000 1.000 1.000 1.000 1.000 1.000
DESIGN SCORE 1.000 1.000 1.000 1.000 1.000 1.000
Other Features Base Anneal
-(s) s -(f) f 60 C 100 C 150 C
Post-implant shape yield 1.000 1.000 1.000 0.857 1.000 0.938 1.000
DESIGN SCORE 1.000 1.000 1.000 0.857 1.000 0.938 1.000
Table E-14: Results from test 4b (post-implant cracking). Numbers quoted are the total length of cracks (in mm) in a 0.25 x 0.25
mm square area at the given inspection point. The design score is calculated per the formula in Table E-8. The sample size for
this test was 1 sample per design. Parts which had more cracks than 2 standard deviations (SD) above the mean were not
considered in score calculations.
Tested Designs
B2- B2s C2- C3- D2- D2f D3f E3-
Average length of cracks 0.382 0.063 0.106 0.182 0.315 0.016 0.121 0.183
Maximum length of
cracks 1.134 0.634 0.622 1.495 1.100 0.160 0.571 1.021
DESIGN SCORE 0.626 0.828 0.821 0.587 0.651 0.957 0.829 0.703
Backbone Shapes Cable Routing
B C D E 2 3
Average length of cracks 0.275 0.076 0.311 0.094 0.000 0.334
Maximum length of
cracks 1.102 0.622 1.100 0.489 0.000 1.495
DESIGN SCORE 0.661 0.828 0.652 0.856 1.000 0.549
Other Features Base Anneal
-(s) s -(f) f 60 C 100 C 150 C
Average length of cracks 0.382 0.063 0.315 0.016
Insufficient Data Maximum length of
cracks 1.134 0.634 1.100 0.160
DESIGN SCORE 0.626 0.828 0.651 0.957

236
Inspection Points
1A 1B 1C 2A 2B 2C
Average length of cracks 0.157 0.000 0.216 0.146 0.055 0.091
Maximum length of
cracks 1.100 0.000 1.102 0.622 0.443 0.489
DESIGN SCORE 0.690 1.000 0.675 0.811 0.877 0.857
Inspection Points (continued)
3 4A 4B 4C 4D
Average length of cracks 0.000 0.297 0.400 0.593 0.064
Maximum length of
cracks 0.000 0.973 1.495 1.021 0.514
DESIGN SCORE 1.000 0.687 0.533 0.602 0.858
Average length of cracks (all nonzero values) 0.685 mm
Standard deviation (all nonzero values) 0.470 mm
Maximum length of cracks (all parts) 2.029 mm

237
G Photomasks for Cuff Electrodes
Mask files can be found on the laboratory server at /Data/Projects/HORNET NEST 4/ Designs
and Masks.

238

239
H Overmolding of Cuff Electrodes
H.1 Overmolding Procedure
1. Apply a thin layer of mold release onto all inner surfaces of the device mold and blow dry
2. Mix PDMS thoroughly and degas in a vacuum chamber until no bubbles are visible
3. Fill the base of the mold with PDMS, and degas again if necessary
4. Place device into the base of the mold using the pins to ensure correct alignment
5. Fill the top of the mold with PDMS, and degas again if necessary
6. Carefully put the top of the mold in place and screw closed, using the pins to ensure correct
alignment and taking care not to damage the device
7. Place the thermoformed electrode array inside a syringe dispensing tip with inner diameter equal
to the thermoformed diameter of the cuff, and tape the dispensing tip to the mold
7.1. Note: this step prevents loosening of the thermoformed cuff at elevated temperatures
8. Bake the molded part for at least 4 hours at 65 °C
9. Remove the mold from the oven
10. Release the electrode array from the dispensing tip
11. Unscrew the screws on the mold, and carefully separate the top and bottom pieces of the mold
12. Carefully remove the device from the mold, and remove any excess PDMS (flash) from the edges
of the molded part
H.2 Overmold Design
Mold CAD files can be found on the laboratory server at /Data/Projects/HORNET NEST 4/Mold.

Abstract (if available)

Abstract The field of microelectromechanical systems (MEMS) has enabled the creation of microscale systems which have impacted a number of fields. In the field of neural interfaces, MEMS has enabled significant miniaturization of electrodes from the millimeter scale to the micron scale. This reduction in size allows neural interfaces to be less invasive, reducing the body’s immune response and improving patient outcomes, and to interface with the tissue on a smaller size scale, increasing spatial resolution of neural recording or stimulation. MEMS devices, however, are built in a planar configuration, limiting their ability to interface with complex 3D geometry in the body. To overcome this challenge, Parylene-based planar MEMS devices can be permanently transformed into 3D shapes via post-processing, enabling countless more applications of such devices to interface with non-planar anatomy.
This work first discusses the post-processing of Parylene-based MEMS devices to produce 3D structures via the modulation of film stress and thermoforming, described in chapter 2. Chapters 3 and 4 apply that process to develop two novel devices to interface with complex anatomy in the body. The first (chapter 3) is an endovascular electrode array for neural recording from within the blood vessels, aimed at minimally invasive seizure monitoring. The second (chapter 4) is a peripheral nerve cuff electrode for chronic stimulation of small diameter branched nerves, designed to produce more targeted neuromodulation for a variety of different applications.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Development of micromachined technologies for neural interfaces

PDF

Parylene-based implantable interfaces for biomedical applications

PDF

High-resolution data acquisition with neural and dermal interfaces

PDF

Development of implantable Parylene-based MEMS technologies for cortical applications

PDF

Next generation neural interfaces for neural modulation and prosthesis

PDF

Parylene C bioMEMS for implantable devices with electrochemical interfaces

PDF

The electrochemical evaluation of Parylene-based electrodes for neural applications

PDF

Development of fabrication technologies for robust Parylene medical implants

PDF

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

PDF

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

PDF

Parylene-based biomems sensors for multiple physiological systems

PDF

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

PDF

Bidirectional neural interfaces for neuroprosthetics

PDF

Geometric and dynamical modeling of multiscale neural population activity

PDF

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

PDF

Corrosion studies of engineered interfaces

PDF

Therapeutic electrical stimulation strategies for neuroregeneration and neuroprotection of retinal neurons

PDF

Designing neural networks from the perspective of spatial reasoning

PDF

Ultrasound neuromodulation and its applications for noninvasive vision restoration

PDF

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

Asset Metadata

Creator Thielen, Brianna Leilani (author)

Core Title Fabrication and packaging of three-dimensional Parylene C neural interfaces

School Viterbi School of Engineering

Degree Doctor of Philosophy

Degree Program Biomedical Engineering

Degree Conferral Date 2024-08

Publication Date 07/31/2024

Defense Date 07/15/2024

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

Tag endovascular electrodes,epilepsy,extraneural electrodes,neural interfaces,neural recording,neuromodulation,OAI-PMH Harvest,Parylene C,thermoforming,three-dimensional

Format theses (aat)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Meng, Ellis (committee chair)

Creator Email brianna.thielen@gmail.com,bthielen@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-oUC113998LN4

Unique identifier UC113998LN4

Identifier etd-ThielenBri-13315.pdf (filename)

Legacy Identifier etd-ThielenBri-13315

Document Type Dissertation

Format theses (aat)

Rights Thielen, Brianna Leilani

Internet Media Type application/pdf

Type texts

Source 20240731-usctheses-batch-1190 (batch), University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright.

Repository Name University of Southern California Digital Library

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

Repository Email cisadmin@lib.usc.edu