Towards a Microfabricated Flexible Graphene-Based
Active Implant for Tissue Monitoring During
Optogenetic Spinal Cord Stimulation
Andrada Iulia Velea
, Sten Vollebregt
, Vasiliki Giagka
Bioelectronics Section and
Electronic Components, Technology and Materials Section,
Department of Microelectronics, Delft University of Technology, Mekelweg 4, 2628 CD, Delft, The Netherlands
Technologies for Bioelectronics Group, Department of System Integration and Interconnection Technologies, Fraunhofer Institute for
Reliability and Microintegration IZM, Berlin, Germany
A.I.Velea@student.tudelft.nl, S.Vollebregt@tudelft.nl, V.Giagka@tudelft.nl
Abstract – Our aim is to develop a smart neural interface
with transparent electrodes to allow for electrical
monitoring of the site of interest during optogenetic
stimulation of the spinal cord. In this work, we present
the microfabrication process for the wafer-level
development of such a compact, active, transparent and
flexible implant. The transparent, passive array of
electrodes and tracks have been developed using
graphene, on top of which chips have been bonded using
flip-chip bonding techniques. To provide high flexibility,
soft encapsulation, using polydimethylsiloxane (PDMS)
has been used. Preliminary measurements after the
bonding process have shown resistance values in the
range of kΩ for the combined tracks and ball-bonds.
Keywords – Neural interface, optogenetic stimulation,
active implant, graphene, PDMS.
Epidural spinal cord stimulation (ESCS) has been
proven to promote locomotion recovery in patients affected
by spinal cord injuries (SCIs) . However, optimization of
the specifications for such therapies is still under research
and identifying the mechanism of action could greatly
benefit from parallel monitoring of the response of the
biological tissue during stimulation. Usually, in ESCS,
energy is injected into the tissue in the form of electrical
pulses, leading to activation. Alternatively, energy in the
form of light can also be used to activate the tissue, in a
more spatially specific manner, using optogenetics.
Already available electrode arrays for ESCS feature
opaque electrodes which limit electrical monitoring of the
tissue response during optogenetic excitation . Therefore
there is the need to develop optically transparent,
conductive, and flexible electrodes. This will allow for
capturing the electrical activity of the neurons below the
activation site at the time of stimulation. One potential
material for these electrodes is graphene, as the material is
optically transparent, bendable, potentially biocompatible,
and has excellent electrical properties [3, 4]. Graphene
microelectrode arrays have been previously reported as
passive implants with Au tracks to interface the electrodes
with the outside active system . However, for a smart
implant, ultimately, active components, e.g. integrated
circuits (ICs), have to be embedded with the electrodes to
allow for signal acquisition, in-situ amplification and
The aim of the current work is the development, by
means of microfabrication, of a compact, flexible, graphene-
based, active spinal cord monitoring implant for optogenetic
A microfabrication process has been used to ensure
repeatability and maintain the small size of the implant
while achieving high resolution. Fig. 1 depicts the process
steps of the current work.
Chemical vapor deposited (CVD) graphene tracks and
electrodes have been microfabricated on a wafer level, using
a pre-patterned 50 nm Mo layer as a catalyst, deposited on
as described in detail in . On top, 100 nm of Ti and
675 nm of Al have been deposited and patterned in order to
create a bonding interface between graphene and the Au
Fig. 1. Process steps of the proposed method. A molybdenum catalyst layer
is used for graphene growth, on top of which chips are bonded and later, the
complete structure is encapsulated in polymer.
a) b) c)
Fig. 4. Implant structure with Mo tracks after DRIE. In a)
, the complete
suspended implant can be observed. In b),
a detailed perspective of the
PDMS membrane and tracks is presented. In c),
the high flexibility of the
Fig. 3. Raman spectroscopy measurement, taken with a 633 nm laser.
peaks indicate the presence of graphene.
a) b) c) d)
Fig. 5. Preliminary results after bonding. CT scans and 2-
measurements have been performed to ensure that the bonding process was
stud bumps existent on the pads of the chips. Later, using a
thermocompression flip-chip bonding technique, chips were
bonded to the substrate. Next, 50 µ m of Sylgard 184
polydimethylsiloxane (PDMS), 1:10 ratio, have been spin
coated on top of the structure and cured at 90 °C for 1 hour.
At this point, the complete structure had to be transferred or
released from the original wafer in order to spin coat the
final PDMS encapsulation layer. To do so, two approaches
have been investigated, as illustrated in fig. 2.
The approach in a) consists of the creation of through-
silicon vias (TSV), before graphene growth, to increase the
number of access points for the etchant, while the approach
in b) consists of a deep reactive ion etching (DRIE) process
for cm-size areas (that can later be coated with PDMS as
final encapsulation), performed after having the complete
structure on the wafer, an approach known as F2R .
First, Raman spectroscopy was performed to ensure the
presence of graphene (fig. 3). The ratio I
suggests that a
multilayer graphene has been grown on the substrate, while
the ratio I
estimates the amount of defects present in the
graphene layer (the greater the ratio, the more defects can be
found). The defects originate from the growth process .
A. “Wet” transfer of structure
BHF etching of the oxide layer from beneath the
structure, has been tested. The expected etch rate was 150
nm/min and the total calculated etching time was 40 min.
Yet, after 7 hours, no etching around the TSV has been
observed. Possibly, DRIE of the TSV resulted in the
deposition of a polymer layer which could not be removed
-plasma treatment. To circumvent this, potassium
hydroxide (KOH) etching was performed to widen the
pathways for the BHF. However, after these long wet
etching steps, the structures were highly damaged or even
removed. Since later in the process, at this step, chips
containing active components will also be present, wet
transfer of the structure is to be avoided.
B. “Flex-to-rigid” approach
Fig. 4 shows the results obtained after the DRIE process,
for silicon removal, in combination with wet etching of the
remaining oxide layer. The complete area of the implant
was successfully suspended and the membranes did not
contain any significant wrinkles or damages.
C. Flip-chip bonding
So far, no flip-chip bonding processes on graphene
substrates have been reported in the literature and since our
initial attempt showed that the adhesion between graphene
and Au stud bumps is poor, the creation of a metal interface
in between has been chosen as an alternative.
Fig. 5 a) depicts a visual representation of the structures
before and after flip-chip bonding. In b), a computer
tomography (CT) scan, after the bonding process is
illustrated, while c) and d) show a preliminary 2-point
measurement result. The resistance, ~7.6 kΩ, is the sum of
the ball-bond resistance and 2 graphene tracks resistances
(one graphene track indicated a resistance value of ~3.7
kΩ). Currently new devices are being fabricated which will
allow the measurement of the flip-chip ball-bond only.
This work presents the process for developing flexible,
active, graphene-based epidural spinal cord monitoring
implant, by means of microfabrication only. It has been
shown that F2R approach can be used to suspend large
areas, thus avoiding as much as possible wet process steps
that can damage the structures. Moreover, it has been
demonstrated that flip-chip bonding of chips on a graphene
substrate, using metal interfaces, is possible and initial
measurements have shown that there is electrical
conductivity after the bonding process. To the authors’ best
knowledge, this is the first reported graphene-based active
van den Brand et al., Science, vol. 336, pp. 1182-1185, 2012.
 V. Giagka et al., Biomed. Microdev., vol. 17, pp. 106 – 118, 2015.
D. W. Park et al., Nat. Commun., vol. 5, 2014.
 J. Y. Hong et al., ACS Nano, vol. 10, 2016.
S. Vollebregt et al., Proc. of IEEE-MEMS, pp.17-20, 2016.
 B. Mimoun et al., IEEE Sensors, vol. 13, pp. 3873-3882, 2013.
Fig. 2. Approaches used to transfer/ release the structure. In a), a wet
transfer approach, using buffered hydrofluoric acid (BHF) 7:1 for oxide
etching. In b), a “flex-to-rigid” (F2R) approach .