Thisistheacceptedversionofthearticle:
BlaschkeB.M.,LottnerM.,DrieschnerS.,CaliaA.B.,StoiberK.,
RousseauL.,LissourgesG.,GarridoJ.A..Flexiblegraphene
transistorsforrecordingcellactionpotentials.2DMaterials,
(2016).3.025007:-.10.1088/2053-1583/3/2/025007.
Availableat:
https://dx.doi.org/10.1088/2053-1583/3/2/025007
Flexible graphene transistors for recording cell
action potentials
Benno M. Blaschke
1
, Martin Lottner
1
, Simon Drieschner
1
,
Andrea Bonaccini
2
, Karolina Stoiber
1
, Lionel Rousseau
4
, Ga¨elle
Lissourges
4
and Jose A. Garrido
2,3
1
Walter Schottky Institut und Physik-Department, Technische Universit¨at M¨unchen,
Am Coulombwall 4, 85748 Garching, Germany
2
ICN2 – Catalan Institute of Nanoscience and Nanotechnology, Barcelona Institute of
Science and Technology and CSIC, Campus UAB, 08193 Bellaterra, Spain
3
ICREA, Instituci´o Catalana de Recerca i Estudis Avan¸cats, 08070 Barcelona, Spain
4
ESIEE-Paris, ESYCOM, University Paris EST, Cit´e Descartes BP99,
Noisy-Le-Grand 93160, France
E-mail: joseantonio.garrido@icn.cat
Abstract.
Graphene solution-gated field-effect transistors (SGFETs) are a promising platform
for the recording of cell action potentials due to the intrinsic high signal amplification
of graphene transistors. In addition, graphene technology fulfills important key
requirements for for in-vivo applications, such as biocompability, mechanical flexibility,
as well as ease of high density integration. In this paper we demonstrate the fabrication
of flexible arrays of graphene SGFETs on polyimide, a biocompatible polymeric
substrate. We investigate the transistor’s transconductance and intrinsic electronic
noise which are key parameters for the device sensitivity, confirming that the obtained
values are comparable to those of rigid graphene SGFETs. Furthermore, we show
that the devices do not degrade during repeated bending and the transconductance,
governed by the electronic properties of graphene, is unaffected by bending. After cell
culture, we demonstrate the recording of cell action potentials from cardiomyocyte-like
cells with a high signal-to-noise ratio that is higher or comparable to competing state
of the art technologies. Our results highlight the great capabilities of flexible graphene
SGFETs in bioelectronics, providing a solid foundation for in-vivo experiments and,
eventually, for graphene-based neuroprosthetics.
1. Introduction
In recent years, an increasing effort is being dedicated to the development of a new
generation of electronic devices that can further advance the interface to living cells
and tissue.[1, 2, 3, 4] Besides improving our understanding of the nervous system and
the brain,[5] these devices can be applied in electrically-active prostheses to restore
vision,[6] hearing,[7] or to find a solution to damaged motor or sensory functions.[8]
While some of these applications exclusively rely on the electrical stimulation of cells
Flexible graphene transistors for recording cell action potentials 2
or tissue, others also require the detection of the electrical activity of the nerve
cells. Besides microelectrode array (MEA) technologies[2, 9, 10, 11] transistor-based
concepts are receiving renewed attention for recording [12, 13, 14, 15, 16] due to the
advantages they can offer. For instance, their intrinsic signal amplification enabled by
the transistor configuration[17] and the possibility for downscaling and high density
integration in contrast to the MEA technology where the impedance is greatly affected
by the electrode size. Furthermore, the development of transistor-based designs could
enable a new generation of implants with bidirectional communication capabilities i.e.
providing both stimulation and recording, thus allowing an in-situ fine control for
electrical stimulation.[18] Therefore, there is a need to explore and identify suitable
materials for the fabrication of transistors that can be used for recording electrical
activity. In this respect the transistor material has to meet several requirements
to allow for an efficient and long-lasting interface to living systems: it has to be
biocompatible and chemically stable in harsh biological environments, and it has to
provide a broad electrochemical potential window to avoid the negative effects of
electrochemical reactions.[19] Furthermore, in order to allow for a high sensitivity in
the detection of action potentials the material of choice is expected to exhibit good
electronic performance, such as high carrier mobility and low intrinsic noise.[1] Materials
offering a high capacitance at the electrolyte/transistor interface are also of interest due
to the positive influence of the interfacial capacitance on the transistor sensitivity;[20]
additionally, a high capacitance also has a positive effect on the range of gate bias
that can be applied to these devices, which is rather limited due to the operation in
aqueous electrolytes.[14] Lastly, considering the implementation of this technology in
real applications, for instance in biomedical implants, it becomes of utmost importance
to use materials that allow the fabrication of flexible devices, a requirement needed
to lower the mechanical mismatch between the sample and the tissue, thus avoiding
the decrease in the device performance due to glial scare formation.[21] In the past,
several materials have been used for cell signal detection in a transistor configuration:
silicon,[12] gallium nitride,[22] diamond,[13] and more recently organic materials[23] and
graphene.[1] While the use of materials such as silicon, diamond and gallium nitride
introduces enormous technological challenges in terms of device flexibility, organic
materials, PEDOT:PSS for instance,[15] or novel materials such as graphene[24] can be
integrated relatively easy into flexible devices. However, many organic materials such as
P13[25] or sexithiophene only provide charge carrier mobilities below 10 cm
2
V
−1
s
−1
[26]
and have a relatively high electronic noise. Therefore, high quality chemical vapor
deposition (CVD) graphene, offering simultaneously high carrier mobility (well above
10
3
cm
2
V
−1
s
−1
), low electronic noise, high chemical stability and facile integration into
flexible devices, appears as a particularly qualified material.[14] While the first reports
of cell recordings using graphene solution-gated field-effect transistors (SGFETs) based
on rigid substrates already demonstrated the great potential of this material,[1] the
next challenge is the transfer of that rigid technology to a more suitable flexible one.
In this paper, we report on the detection of action potential of cardiomyocyte-like HL-1
Flexible graphene transistors for recording cell action potentials 3
cells[27] using flexible graphene based SGFETs. Our work confirms that flexible devices
fabricated using CVD graphene can play a significant role in the next generation of
implant technologies.
2. Results and discussion
The fabrication of the devices, described in detail in the methods section, is carried out
on an approximately 10 µm thick polyimide film spin coated on a supporting substrate.
In short, metal contacts were evaporated onto the substrate, after which CVD graphene
was transferred and the active area of the transistors was defined. Afterwards, a
second metal layer was evaporated and the metal lines were covered with an insulating
photoresist. In a last step, the device is released from the supporting substrate. The
upper panel in figure 1 a) shows a schematic of a released device. The transistors are
located in the center and connected to the bond pads via metal feed lines.
0 . 0 0 . 3 0 . 6
0
5 0
1 0 0
1 5 0
2 0 0
- 0 . 6 - 0 . 3 0 . 0 0 . 3
1 0
- 3
1 0
- 2
1 0
- 1
1 0
0
c )
I
D S
( µ A )
U
G S
( V )
b )
a )
| g
m
/ U
D S
( m S / V ) |
U
G S
- U
D i r a c
( V )
Figure 1. a) Upper panel: Schematic of a flexible graphene transistor array on a
polyimide substrate. Lower panel: Microscope image of 36 transistors of the array with
drain and source contacts and the SU8 window. Scale bar is 200 µm. b) Transistor
currents of four transistors as a function of the applied gate potential measured in
5 mm PBS buffer. c) Normalized transconductance for the same transistors (W=20 µm;
L=10 µm).
A microscope image of a 6x6 transistor array is shown in the lower panel of figure
1 a). The active area of each transistor is 10 µm (length) x 20 µm (width). Firstly,
the flexible graphene SGFETs were characterized to compare their performance to
existing technologies. The transistor measurements were performed in a 5 mm phosphate
buffered saline (PBS) solution using an Ag/AgCl reference electrode to apply the gate
voltage. Figure 1b) shows typical transistor curves in which the drain-source current,
I
DS
, was recorded as a function of the gate voltage, U
GS
, while the drain-source
voltage was fixed to U
DS
=100 mV. As expected from the graphene band structure
Flexible graphene transistors for recording cell action potentials 4
a V-shape curve is observed,[20] exhibiting the Dirac point (minimum of the curve)
around U
Dirac
=400 mV vs. Ag/AgCl. This indicates p-type doping of the device since
for an undoped device a Dirac voltage of about U
Dirac
=150 mV is expected due to
the difference of the work function of graphene (4.6 eV)[28] and the Ag/AgCl reference
electrode (4.7 eV);[29] the applied U
DS
should also be considered. Residues from PMMA
used during the transfer and interactions with the substrate have been suggested as
the origin of the p-type doping of transferred CVD graphene.[30, 31] A key figure of
merit of the device performance is the transconductance, g
m
, which is typically used
to quantify the sensitivity of the device and represents the change in the transistor
current, I
DS
, induced by a small change in the gate voltage.[17] In the particular case
of the detection of action potentials with a transistor the electrical activity of a cell in
the vicinity of the transistors active region will induce a small change of the effective
gate voltage, U
GS
, applied to the transistor. Thus, for a given U
GS
, the larger g
m
,
the larger the measured modulation of the transistor current. Figure 1 c) shows the
transconductances, normalized by U
DS
, obtained by deriving I
DS
with respect to U
GS
in figure 1 b). Values of more than 4 mS V
−1
are obtained, similar to those of rigid
graphene transistors.[1] These values are significantly higher than those reported for
transistors based on other technologies, such as silicon, diamond or AlGaN,[14] and are
comparable to other flexible technologies such as PEDOT:PSS transistors.[15] The high
transconductance of the graphene SGFETs originates from the combined effect of the
interfacial capacitance of the graphene/electrolyte interface, of several µFcm
−2
,[20] and
the high charge carrier mobilities in CVD graphene, of more than 1000 cm
2
V
−1
s
−1
.[17]
Besides the transconductance, the intrinsic electronic noise of the transistor has
to be considered in order to characterize its sensitivity: the noise figure of merit sets
the limit for the minimum modulation of the gate, and thus the minimum cell signal
that can be detected by the transistor. To assess the noise of the flexible graphene
SGFETs, the power spectral density (PSD), S
I
, of the transistor current was measured
in 5 mm PBS buffer (see methods section for details). Figure 2 a) shows the result of 200
averaged individual spectra obtained for one transistor (bias conditions: U
GS
=250 mV
and U
DS
=100 mV). A 1/f behavior of the power spectral density is observed, as reported
previously for rigid graphene SGFETs.[1, 32] To evaluate the noise performance, the
power spectral density is fitted using S
I
= A/f
b
, with A and b representing the fitting
parameters. Values of b typically range from 0.8 to 1.2. In order to understand the
origin of the noise generation mechanism and to identify the most suitable transistor bias
conditions in terms of noise, the influence of the gate bias, U
GS
, on the power spectral
density has been investigated. Figure 2 b) shows that the noise parameter A as a function
of U
GS
reaches a minimum close to U
Dirac
. For comparison, the graph also shows the
U
GS
dependence of g
m
2
(orange) and I
DS
4
(green) calculated for the same device. These
two dependences have been previously used to discuss the noise mechanisms in graphene
transistors.[32] On the one hand, a noise parameter A displaying a g
m
2
dependence has
been correlated to a noise mechanism in which charge fluctuations close to the graphene
transistor active area are coupled into the device through the interfacial capacitance.
