![Characterization of nMOS-based sensors. a) Dual-gate configuration measurements without Au decoration of the sensing electrodes. Inset shows measurements in air while applying voltage through CG2 (red) and only using the single control-gate (black). b) Mean and standard deviation of multiple measurements on the same sensor, before and after Au decoration. The Au film increases the drain current shift for the same applied voltages (here, VCG1 = 0V and VCG2 = 2V). c) Consecutive measurements with the same sensor in air and with poly-d-lysine. d) ID shifts due to different liquids on Au-decorated sensor. Here, second control-gate voltages were swept (-2V to 2V) while applying constant voltage (5V) through the first control-gate. Comparison between dry (black), poly-d-lysine (blue) and KCl with different molarities (red). Culture test on the chips with hiPSC-derived cortical neurons. e-f) hiPSC-derived cortical neurons were stained for DAPI [e),blue], MAP2 [e), green] and synaptophysin [f), red] following 7 days of differentiation on chip.](https://www.researchgate.net/profile/Massimo-Mastrangeli/publication/352643692/figure/fig2/AS:1037440190013445@1624356424782/Characterization-of-nMOS-based-sensors-a-Dual-gate-configuration-measurements-without_Q320.jpg)
Content uploaded by Massimo Mastrangeli
Author content
All content in this area was uploaded by Massimo Mastrangeli on Jun 22, 2021
Content may be subject to copyright.
DUAL-GATE FET-BASED CHARGE SENSOR ENHANCED BY IN-SITU ELECTRODE
DECORATION IN A MEMS ORGANS-ON-CHIP PLATFORM
Hande Aydogmus
1
, H. Joost van Ginkel
1
, Anna-Danai Galiti
1
, Michel Hu
2,3
, Jean-Philippe Frimat
2,3
,
Arn van den Maagdenberg
2,3
, GuoQi Zhang
1
, Massimo Mastrangeli
1
, and Pasqualina M. Sarro
1
1
ECTM, Department of Microelectronics, Delft University of Technology, the Netherlands.
2
Department of Human Genetics, Leiden University Medical Centre, the Netherlands.
3
Department of Neurology, Leiden University Medical Centre, the Netherlands.
ABSTRACT
Continuous monitoring of tissue microphysiology is a key
enabling feature of the organ-on-chip (OoC) approach for
drug screening and disease modeling. Sensing charged spe-
cies in OoC tissue microenvironments is thereby essential.
However, the inherently small (i.e., cm) size of OoC de-
vices poses the challenging requirement to integrate minia-
turized and highly sensitive in situ charge sensing compo-
nents to maximize signal extraction from small volumes
(nL to µL range) of media used in these devices. Here we
meet this need by presenting a novel dual-gate field-effect
transistor-based charge sensor integrated within an opti-
cally transparent microelectromechanical (MEM) OoC de-
vice. Post-process mask-less decoration of Ti sensing elec-
trodes by spark-ablated Au nanoparticle films significantly
increases the effective electrode surface area and thus sen-
sor sensitivity while retaining the CMOS-compatibility of
the wafer-level fabrication process. We validate the bio-
compatibility of the sensor and its selective response to
poly-D-lysine and KCl, and provide a perspective on mon-
itoring cultures and differentiation of hiPSC-derived corti-
cal neurons on our OoC device.
KEYWORDS
Biosensor, charge sensing, electrodes, microfabrication,
organ-on-chip, spark ablation, neurons
INTRODUCTION
Organs-on-chip (OoCs) are dynamic tissue culture de-
vices which mimic microphysiological environments to re-
capitulate in vivo-like organ functions in vitro and thus en-
hance the efficiency of drug development and disease mod-
elling [1]. Among the key features distinguishing OoCs
from established tissue culture technologies, automated
monitoring of biological cues, such as charged species, is
essential to reduce user intervention and improve system’s
ease of use and reliability [2]. In this respect, dynamic elec-
tric sensing can meet user needs without recurring to ter-
minal optical labeling and microscopy techniques, which
deteriorate cells’ wellbeing.
To provide compact charge sensing solutions in OoCs,
the integration of a floating gate [3-5] field-effect transistor
(FG-FET)-based electrochemical sensor within a microe-
lectromechanical OoC device was previously demonstrated
[6]. Prior examples of FET-based bio-chemical sensing
include tracking metabolic activity of neurons [7-8] and
of pH [9]. Our hybrid OoC combines in a single device the
benefit of soft polymeric substrates, suitable for tissue cul-
turing under physiologically relevant conditions, with that
of extremely compact and inherently signal-amplifying
electronic sensing units, uniquely enabled by silicon-based
microfabrication [10]. Building on that, in this paper, we
show significant enhancement in sensitivity and in selec-
tive response of the FG-FET-based charge sensor. Both
breakthroughs were enabled by the introduction of a novel
dual-control-gate configuration and by in situ decoration of
the sensing electrode surface with nanoporous Au thin
films (Fig. 1).
The OoC device is fabricated by means of a mostly
CMOS-compatible co-fabrication process. The charge sen-
sor therein replaces the function of an external reference
electrode with that of a pair of control gates (CG
1
and CG
2
)
capacitively coupled to the floating gate. An extension of
the FG reaches over the suspended polymer membrane,
representing the optically accessible tissue culturing area,
and terminates with a sensing electrode. The charge stored
in the dielectric layer (silicon oxide) between CG
1
and FG
sets the working point of the transistor, whereas the elec-
trolyte bridging CG
2
and FG serves as dielectric as well as
solution-under-test (SOT). Biasing CG
2
provides electric
field towards the sensing electrode at the tip of the FG ex-
tension. This promotes the binding of charged species on
the sensing surface and thus provides higher sensitivity
(Fig. 2(a)). In turn, the charges in close proximity of the
sensing electrode can be continuously detected by monitor-
ing the drain current (I
D
) of the transistor.
Charge sensing exploits the electrical double layer
[11], as it involves (reversible) binding by charged species
in solution to sites available at the sensing surface. There-
fore, to further enhance sensitivity we aimed at expanding
the effective electrode surface and increasing the number
of available binding sites by means of surface structuring
with Au. Due to its inherent biocompatibility and inertness,
Au has been widely used for biosensing applications as a
Figure
1: Schematic of the MEM OoC device with inte-
grated FET
-based charge sensor. Right insets show sens-
ing electrodes over the suspended membrane
without (top)
and with (bottom) decoration with Au film.
selective material, including monitoring pH and neuronal
activity [8], tracking the morphology of cancer cells and
fibroblasts [12], and DNA hybridization [13]. However,
Au’s well-known incompatibility with CMOS process
flows imposes its post-processing. Here, nanostructured,
microscopically rough Au films were deposited by inertial
impaction of a pure Au nanoparticle (NP) aerosol
generated by spark ablation. The technique is mask-less
(avoiding the issues of conducting lithography over sub-
strates with high topography), spatially-selective, fast, and
is conducted at room temperature. Moreover, while it does
not presently allow batch wafer post-processing, it does not
harm the flexible and suspended polymeric membranes
supporting the electrodes, and yields surface roughness fig-
ures superior to other more cumbersome post-processing
techniques (e.g. flow coating), let alone CMOS-compatible
ones (e.g., e-beam evaporation, sputtering, atomic layer
deposition).
FABRICATION
Fabrication of the FET
A 4-inch, 525 µm-thick, single-side polished p-type Si
wafer was used for the fabrication of both nMOS- and
pMOS-based sensors using a standard BiCMOS process
(Fig. 2(b)). After defining source and drain terminals by ion
implantation, a 200 nm-thick gate oxide layer was ther-
mally grown. A 6 µm-thick PECVD oxide layer was de-
posited and patterned on the wafer backside as etch hard
mask. A 1000 nm-thick layer of AlSi was sputtered and
patterned on the front side to implement electrical intercon-
nects and floating gates (Fig. 2b(1)). A 170 nm-thick
PECVD oxide layer was used as dielectric between CG
1
and floating-gate electrodes. Polyimide (PI) was then used
to sandwich most of the extension of the floating gate onto
the OoC sensing region. This ensured electrical insulation
and mechanical robustness, while leaving the sensing elec-
trode in contact with the environment. Fig. 2b(2) sketches
the dielectric layer and first PI layer. A sputtered 300nm-
thick layer of Ti was patterned to serve as CGs and FG ex-
tensions. Then a second layer of PI was spin coated and
patterned to complete the insulation of the FG extensions
(Fig. 2b(3)). Polydimethylsiloxane (PDMS) was mixed
with its curing agent in a 10:1 ratio, degassed, spin coated
and cured to serve as 20 µm-thick membrane for the OoC
cell culturing/charge sensing area (Fig. 2b (4)). After sput-
tering a 200 nm-thick AlSi protection layer onto the PDMS
(Fig. 2a(5)), the Si substrate was etched from the backside
by deep reactive ion etching (DRIE) to land on the oxide
layer (Fig. 2b (6)). The AlSi hard mask and residual silicon
oxide were then removed by wet etching to complete the
release of the PDMS membrane (front- and backside of a
single sensor are sketched in Fig. 2b (7) and (8), respec-
tively). The wafer was finally diced into 52 equal, square
chips with footprint of 1 cm
2,
each chip containing 4 pMOS
and 4 nMOS FETs (Fig. 2(c)).
Post-processing for electrode surface structuring
The Au NP aerosols were generated by a spark dis-
charge generator (VSP G1) and deposited by means of a
Figure
2: a) Schematic of the sensing mechanism, showing a section of the sensing area with CG
2
and FG. The dual-
control-gate configuration provides measurement of charges on the floating-gate extensions. b) Wafer-
level fabrication
of the dual-gate FG-FETs. c) Laser-optical microscope image of the frontside of a single chip. Second control-gate a
nd
sensing electrodes are visible. Inset shows the full chip after dicing. d) Sketch of the nanomaterial printer setup. e) SEM
micrograph of local decoration of sensing elect
rodes with Au nanoparticle films (film morphology evidenced in the inset.)
f) Backside micrograph of the chip by laser
-optical microscopy, showing Au decoration on 3 of the 8 sensing electrodes.
nanoporous material printer (NMP) prototype developed
by VSParticle B.V. (Fig. 2(d)). The NMP operates at <1
mbar pressure to create an ultrasonic gas jet with tunable
diameter that transfers the NPs on the substrate using their
inertia [14]. The deposition process used 99.999% Au elec-
trodes and N
2
as carrier gas, resulting in pure and clean NP
film surfaces and little waste production. A 5 s deposition
resulted in < 500 µm spot size.
Figs. 2(e) and 2(f) show successful damage-less com-
pletion of the Au post-processing step on the Ti sensing
electrodes embedded in the 20 µm-thick PDMS membrane.
MEASUREMENTS
To characterize the sensors, a 4-probe measurement
system connected to a semiconductor parameter analyzer
was used to bias the FG-FETs and monitor changes in drain
current I
D
ensuing from charges at the sensing electrodes.
Measurements before electrode decoration
As baseline test, I
D
values were recorded in air with
and without applying voltage through CG
2
with the same
nMOS-based sensor (inset of Fig. 3(a)). Shift in the I
D
val-
ues demonstrate the electrical coupling between CG
2
and
FG electrodes. After characterization in air, a 2.5mM KCl
solution was introduced to the sensing area and consecutive
measurements were recorded. The sensor was flushed and
measurements with DI water were also recorded. The dual-
gate configuration shows unique I
D
characteristics for each
measurement type, as evident from Fig. 3(a). We note that
the absolute change in I
D
value is linearly proportional to
the charge concentration of the solution, whereas the nega-
tive change is attributed to the difference in dielectric con-
stants of the solutions and to the effect of coupling capaci-
tance between CG
2
and FG.
Measurements after electrode decoration
Effect of Au decoration. For the same sensor used
previously, measurements in air and with 2.5mM KCl so-
lution before and after Au post-processing were recorded
(Fig. 3(b)). The effect of electrode decoration is evident
from changes in I
D
values. In addition, for the same applied
voltage values, mean values of measurements after Au dec-
oration show an up to 2.3 times increase in sensitivity.
Effect of membrane stress. After dry measurements,
the Au decorated sensor was used to evaluate poly-d-ly-
sine, a positively-charged polypeptide widely used in
mammalian cell culturing [4] and stem cell research [15].
Fig. 3(c) shows shift in I
D
due to reversible binding of
charges in 10 µL poly-d-lysine (0.1mg/ml). Interestingly,
measurements after evaporation of the solution show simi-
lar I
D
values to those measured immediately after placing
the solution at the sensing area. This suggests that possible
membrane stress due to liquid handling did not notably af-
fect the results, and confirms that the current shifts are pri-
marily caused by charges bound on the electrodes.
SOT characterization
:
We conducted consecutive
measurements of various solutions with Au-decorated FG
extensions (Fig. 3(d)). First, a KCl solution (2.5mM) was
introduced and measurements were recorded. The sensor
was then flushed with DI water, and KCl with lower con-
centration (1.25mM) was introduced. Lastly, Poly-d-lysine
was introduced on the sensing area. The sensor was flushed
again and measurements were recorded. Changes in I
D
show differences accordingly with the different charge
concentrations in each solution. The low leakage current
from CG
2
appears due to the reversable breakthrough of the
native Ti oxide layer on top of the electrode, while applying
voltage through CG
2
.
Figure
3: Characterization of nMOS-based sensors. a) Dual-gate configuration measurements without Au decoration o
f
the sensing electrodes. Inset shows measurements in air while applying voltage through CG
2
(red) and only using the
single control-gate (black). b) Mean and standard deviation of multiple me
asurements on the same sensor, before and
after Au decoration. The Au film increases the drain current shift for the same applied voltages (here, V
CG1
= 0V and V
CG2
= 2V). c) Consecutive measurements with the same sensor in air and with poly-d-lysine. d) I
D
shifts due to different liquids
on Au-decorated sensor. Here, second control-gate voltages were swept (-
2V to 2V) while applying constant voltage (5V)
through the first control-gate. Comparison between dry (black), poly-d-lysine (blue) and KCl with diffe
rent molarities
(red). Culture test on the chips with hiPSC-derived cortical neurons. e-f) hiPSC- derived cortical neurons were
stained
for DAPI
[e),blue], MAP2 [e), green] and synaptophysin [f), red] following 7 days of differentiation on chip.
DISCUSSION
We propose the following model to explain the in-
creased sensitivity evidenced in the experimental results by
virtue of Au electrode decoration. In the saturation regime
of the transistor, the variation in the floating gate voltage
is expected to depend on the total capacitance of the
device and on the variation of charges in close prox-
imity to the sensing area [5]:
(1)
Assuming a surface charge density
forming at the sensing
electrode with surface, then:
(2)
and can be related to the floating gate voltage as:
!" # $%& (3)
where $ is the threshold voltage and ! '()*+,-.the
FET transconductance, which depends on carrier mobility
(/%, gate oxide capacitance per unit area"()*%, and on
width (W) and length (L) of the gate. According to (3) a
change in is thus in direct relation with the charge vari-
ation in close proximity to the sensing electrode. The dep-
osition of the microscopically rough Au thin film increases
the effective surface area 0 and the available sites of the
sensing electrode surface.
The proposed sensor avoids the cumbersome use of an
external reference electrode. While the control over the
charging of the surface is not direct [11], the electrode sur-
face enhancement method, coupled with the novel dual-
gate control configuration here demonstrated, enables
higher charge accumulation on the sensing electrode,
thanks to the combination of electrical field between CG2
and FG and larger effective surface area.
Cell Culture & Differentiation
On chip biocompatibility and neuronal differentiation
tests were conducted by culturing and differentiating hu-
man-induced pluripotent stem cell (hiPSC)-derived cortical
neurons. Levels of potassium in the brain higher or lower
than its threshold value might cause different disease phe-
notypes such as epilepsy [16] or chances of developing a
headache [17]. Therefore, the ability to sense potassium
from neuronal cultures in vitro is critically needed for brain
disease modelling.
Neural Progenitor Cells (NPCs) were successfully dif-
ferentiated on the chips into cortical neurons for 7 days,
with staining showing mature neuron marker
(MAP2/green) as well as the production of synaptic vesicle
(synaptophysin/red) (Fig. 3(e-f)). Long neurite extensions
are visible as well as distinct network formation, indicating
that the chip is suitable for live measurements using hiPSC-
derived cortical neurons.
CONCLUSION
A FET-based electrochemical sensor with enhanced
sensitivity and selectivity to chemical species was inte-
grated by a mostly CMOS-compatible fabrication process
within a microelectromechanical OoC device. The dual-
gate control architecture of the FETs allows higher sensi-
tivity compared to single capacitively-coupled control-gate
FETs. In addition, spark ablation was used to alter the sur-
face morphology of the sensing electrodes with thin na-
noporous Au films. This post-processing method allows
fast and spatially-selective patterning over topographical
substrates without the need for cumbersome and time-con-
suming lithography steps, and without damaging the fragile
and biocompatible polymer membranes. As a result, more
than doubled sensitivity was achieved, and the sensor could
be used to successfully identify different SOTs, including
low-concentration KCl solutions and poly-D-lysine.
Furthermore, culture tests showed that NPCs were suc-
cessfully differentiated to cortical neurons on our chip, and
visual inspection was possible due to the optical transpar-
ency of the culture area. These results suggest that our
MEM OoC device can be employed in the near future for
testing the potassium level of human brain cells (whereby
normal levels of K+ are 150 mM intracellular and 5 mM
extracellular), a crucial indicator for the wellbeing and
functioning of the brain [18].
ACKNOWLEDGEMENTS
The authors thank the staff at the Else Kooi Laboratory
of TU Delft and VSParticle B.V. for their support and for
enabling us to use the Proto-0 NMP in our lab.
This work was supported by the Netherlands Organ-
on-Chip Initiative, an NWO gravitation project funded by
the Ministry of Education, Culture and Science of the gov-
ernment of the Netherlands (024.003.001), and by the
NWO Nano Copper for Power Electronics Interconnection
Project 729.001.023.
REFERENCES
[1] U. Marx et al., ALTEX 33(3), pp. 272-321, 2016.
[2] M. Mastrangeli et al., ALTEX 36(4), p. 650-668, 2019.
[3] M. S. Thomas et al., J. Phys. Chem. Lett. 9(6), pp.
1335-1339, 2018.
[4] B. Chen et al., IEEE Sensors Journal, 11(11), pp.
2906- 2910, 2011.
[5] A Spanu et al., Scientific reports, 5, 8807, 2015.
[6] H. Aydogmus et. al., IEEE Sensors 2020.
[7] P. Fromherz et al., Science, 252, pp. 1290-1293, 1991.
[8] S. Martinoia et al., Biosensors and Bioelectronics,
16(9-12), pp. 1043-1050, 2001.
[9] A. Loi et al., Appl. Phys. Lett. 86(10), 103512, 2005.
[10] M. Kaisti. Biosensors and Bioelectronics, 98, pp. 437-
448, 2017.
[11] M. Kaisti et al., IEEE Trans. Electron Dev., 62.8, pp.
2628–2635, 2015.
[12] P. Lin et al., Adv. Mater. 22.33, pp. 3655–3660, 2010.
[13] M. H. Park et al., J. Phys. Chem. C 120.9, pp. 4854–
4859, 2016.
[14] Schmidt-Ott, A. (Ed.). Spark Ablation: Building
Blocks for Nanotechnology. CRC Press, 2019.P.
[15] P. Rocheteau et. al., Nature communications, 6(1): 1-
12, 2015.
[16] C. Simons et. al., Nature Genetics, 47, pp.73–77,
2015.
[17] Z. Guo, et al., Eneuro, 6(4), 2019.
[18] G. Florence et. al., Communications in Nonlinear Sci-
ence and Numerical Simulation, 17(12), pp. 4700-
4706, 2012.
CONTACT
*H. Aydogmus, h.aydogmus@tudelft.nl
