of 9
Sensors & Transducers, Vol. 191,
Issue 8, August 2015, pp. 12-20
12
Sensors & Transducers
© 2015 by IFSA Publishing, S. L.
http://www.sensorsportal.com
Optimization Techniques for Miniaturized Integrated
Electrochemical Sensors
*
Muhammad Mujeeb-U-Rahman, Axel Scherer
California Institute of Technology, Pasadena, CA, 91125, USA
*
Tel.: + 1 6263956949
*
E-mail: mrahman@caltech.edu
Received: 26 June 2015 /Accepted: 30
July 2015 /Published: 31 August 2015
Abstract:
Electrochemical sensors are integral components of various integrated sensing applications. In this
work, we provide details of optimizing electrochemical se
nsors for CMOS compatible integrated designs at sub-
mm size scales. The focus is on optimization of electrode
materials and geometry. We provide design details for
both working electrode and reference electrode materials for hydrogen peroxide sensing applications which
form the basis for many metabolic sensors. We also present results on geometrical variations in designing such
sensors and demonstrate that such considerations are very relevant for optimizing the overall sensor
performance. We also present results for such optimized sensors on actual CMOS platforms. The methods
presented in this work can be adopted for countless applications of electrochemical sensing platforms.
Copyright
© 2015 IFSA Publishing, S. L.
Keywords:
Electrochemical, Sensors, Integrat
ed, Reference Electrode, Geometry.
1. Introduction
Implantable electrochemical sensors have been
attracting considerable interest owing to their
integration with electronic circuitry required for truly
wireless implants [1]. These sensors can provide
selective and sensitive results in complex
environments; in vivo environment being a very
relevant example. For example, glucose sensors
utilizing electrochemical techniques have been
shown to work for long term implanted applications
with potential for successful
human applications [2].
However, these sensors have typically been part of
large (cm scale) systems. It has been shown that
miniaturizing the overall size of the implantable
system can be very beneficial for minimizing the
foreign body response to such implants [3]. The
small size is advantageous in reducing the bio-
fouling problems associated
with such implants since
it leads to minimal disturbance in the body leading to
minimal inflammation [4]. Hence, such miniaturized
systems can provide accurate and long-term
measurements of analyte in complex environments
e.g. body fluids [5].
As the sensors are made smaller and fully
integrated; it becomes more and more challenging to
achieve high sensitivity, stability, and longevity.
However, these are the important requirements for
real-world applications. Henc
e, this work is focused
on optimizing the electroch
emical sensors for small
scale (near and sub mm) applications. The techniques
presented here can be used for many other
applications involving elect
rochemical cell designs.
The signal processing
circuitry for wireless
sensors is mostly based upon Complementary Metal
Oxide Field Effect Transi
stor (CMOS) technologies.
http://www.sensorsportal.co
m/HTML/DIGEST/P_2702.ht
m
Sensors & Transducers, Vol. 191,
Issue 8, August 2015, pp. 12-20
13
In most cases, sensors are designed on separate dies
and are then integrated
with signal processing
circuitry using bonding/assembly techniques to
realize fully functional devices [6-7]. This leads to
difficulty in miniaturization, increase in cost, and
decrease in overall system performance and
reliability [8]. CMOS technology itself provides
excellent control and processing functions with
extremely small footprint
and power requirements
[9]. Hence, a completely
integrated electrochemical
sensor as part of the CMOS platform itself can
provide extreme miniaturization and cost reduction.
Hence, the miniaturized sensors in this work are
designed to be completely integrated with the CMOS
platform designed for wireless sensing.
The top metal materials available in the CMOS
process (e.g. Al, AlCu) are not suitable for
electrochemical sensing for longevity and sensitivity
for relevant analyte [10]. Hence, material
optimization techniques have to be employed to
create sensors with long term performance.
Moreover, no rigorous design approach is available
to design these sensors optimally at small size scales
intended in this work. Typically, an empirical
approach is used with some simple sensor geometries
to provide large enough signal [10]. This approach
can be suitable for some applications but is
inefficient for optimization requirements of
implantable miniaturized sensors.
In the next sections, we present a methodological
approach towards optimizing the design of a fully
integrated electrochemical sensor on CMOS
compatible substrates. We will start with the design
criteria, and then provide methods of material
optimization, fabrication and design optimization.
Finally, we present results to demonstrate the
effectiveness of our technique.
2. Design Criteria
The underlying design criteria for our sensor are
sensitivity and longevity for metabolic sensing
applications. Glucose sensing for diabetic patients is
an important relevant example. The design
constraints include sensor size, biocompatibility and
non-toxicity of materials, and operation time. In our
typical application, the sens
or area is restricted to a
500
μ
m by 500
μ
m square on a CMOS chip, due to
other system level constraints. The sensor can be
fabricated on the top metal layer having the circuit
underneath. This allows for direct integration of the
sensor with the CMOS circuitry without the need of
any complex bonding mech
anisms. This improves
sensor reliability, minimizes system size and
allows for microfabrication of the devices in a
batch process.
The electrochemical sensor consists of three
electrodes forming a comp
lete electrical circuit
through an external r
eadout mechanism (e.g. a
current meter) [11]. The electrodes are Working
Electrode (WE), Reference Electrode (RE) and
Counter or Auxiliary Electrode (CE). For some
designs, the CE and RE can be combined to result in
a two electrode system. In this work, we focus our
attention on three electrode designs because of the
availability of a separate RE which can provide near-
nernstian stability, thus minimizing sensor response
variations. This also allows to decouple the design of
RE and CE which allows for more degrees of
optimization [11]. An integrated 3-electrode based
electrochemical sensor is
depicted in Fig. 1.
Cou
n
t
e
r
Ele
c
t
r
od
e
Re
f
e
rence
El
ectro
d
e
W
o
r
kin
g
Ele
c
t
ro
d
e
Fig. 1.
CMOS Integrated Electrochemical Sensor.
For proper electrochemica
l operation, each of the
sensor electrodes has to be designed for its particular
role in the sensor application. For example, the
sensing mechanism takes place on the WE. Hence, it
needs to be properly designed and functionalized for
sensing the analyte of interest. The CE is designed to
complete the electrical circuit by allowing some
electrochemical reaction which is equal but opposite
to the reaction at the WE. The RE should be able to
provide a near-nernstian behavior, hence always
resulting in same reference potential in the same
solution [11].
Optimization of sensor depends on material and
geometry optimization. In terms of electrode
materials, the WE should be extremely conductive
and electroactive, even in th
in films. The CE material
should allow a wide set of reactions to balance the
WE reaction current. The RE material should be inert
and stable for long time [11]. In terms of geometry,
the CE should be much larger than the WE to allow
the WE reaction current to control the current in the
readout circuit [11]. To minimize voltage drop
through the solution, the RE should also be as close
to the WE as possible [11]. RE should also be
designed to have minimal effect on diffusion of
species to the WE and CE.
All of the above mentioned considerations have
to be considered in designing integrated
electrochemical sensors. In the next sections, we will
describe the optimization procedure for both the
material and the geometry of the sensors.
3. Material Optimization
Selection of the electrode material is very
important for the sensor design. The electrode
materials are required to be very conductive and
chemically stable for sm
all-scale operation at
extended life-time. At the same time, these materials
should be electroactive and biocompatible. Noble
Sensors & Transducers, Vol. 191,
Issue 8, August 2015, pp. 12-20
14
metals such as Pt, Au, Ir and their alloys have many
favorable properties in this regard [12].
Both Au and Pt have been used for metabolic
sensing applications [13]. For our applications, a test
was conducted to verify the suitability of both metals
for our applications. Hydrogen peroxide sensing
forms the basis of many metabolic sensing
applications involving oxidoreductase enzymes [14].
Hence, we used thin film (100 nm) planar electrodes,
both of Pt and Au, for hydrogen peroxide sensing
using Cyclic Voltammetry (CV) from 0 to 1 V vs.
Ag/AgCl RE at scan rate of 0.01 V/seconds in high
peroxide concentration (50 mM). It was found that
thin film Pt (as well as bulk Pt electrode) can retain a
stable interface for much longer time than Au
(contrary to bulk Au electrode) during a continuous
CV experiment in hydrogen peroxide solutions. High
peroxide concentration (50 mM) was used for
accelerated testing and to emulate possibly large
peroxide builds up during sensor off time in real
applications. The tests were repeated for multiple
samples after different thin film fabrication methods
(sputtering, ebeam) were used to fabricate the test
samples. The results were similar for all cases.
The films after CV are shown in Fig. 2. The
interface where the electrodes were immersed in the
peroxide solution is clearly evident for Au electrode
due to corrosion. However, for Pt, the material stays
intact and the sensor response over time doesn’t
show any significant variations. This indicates that Pt
thin films can endure this process much better than
Au. For bulk electrodes,
both Pt and Au showed
sustainable performance. We believe that this is due
to the formation of a surface layer of Au based
materials (e.g. in oxidized states) which are porous
and hence are not effective in stopping the
process in thin films. This means that over long time
and with peroxide accumulation, thin layers of
Au can dissolve, hence reducing sensor lifetime
and repeatability.
(a)
(b)
Fig. 2.
Effect of Peroxide testing on WE materials
(a) Au, (b) Pt.
The CV obtained during the electrode testing
(Fig. 3) show a similar response where Au thin film
(in contrast to Au thick film) electrodes ‘corrode’ in
the solution (response dies down with time) while the
Pt electrodes (both thick and thin film) don’t corrode
an appreciable amount (response doesn’t die down).
(a) Thin Film Pt Electrode
(b) Thin Film Au Electrode
Fig. 3.
Voltamograms for Hydrogen Peroxide testing
of electrode materials (a) Pt (b) Au.
In blood or other complex fluids, both Au and Pt
can be poisoned by chloride ions and many other
species (e.g. amino acids)
due to surface adsorption.
However, nanopatterned Pt has shown to have high
stability against these blood agents [15]. Nonetheless,
since these sensors are not us
ed in direct contact with
blood, but in interstitial fluid, this is a less critical
problem [16]. For applications in blood, sensor
electrodes will have to be coated with special
materials for filtering (e.g. Nafion) [17].
Reference electrodes also play a critical role
during electrochemical sens
or operation. There are
many different reference
electrode materials which
can provide stable voltage readings in solution to
form stable and repeatable potential differences
between electrodes for proper sensor operation [18].
For solid-state integrated operation, the Ag/AgCl
electrode is a suitable optio
n, especially since the
Chloride ion concentration in body stays relatively
consistent [18]. We fabr
icated such electrodes by
depositing a thin film of Ag on the RE and using
low-power chlorine plasma to convert the top layer in
AgCl. SEM imaging of these layers confirmed the
formation of crystalline structure which is a good
indicator for AgCl formation. The resulting layers are
shown in Fig. 4.
Although, the Ag/AgCl electrode can provide
very stable performance, its formation needs extra
fabrication steps and introd
uces a different material
in the fabrication process. Also, formation of AgCl is
a special process due to the surface conditioning
requirements of the Ag electrodes.
Sensors & Transducers, Vol. 191,
Issue 8, August 2015, pp. 12-20
15
(a) Ag
(b) AgCl
Fig. 4.
SEM of Sensor Electrodes.
As Pt is already used for the other sensor
electrodes, use of Pt based reference electrode can
make the fabrication process much simpler and
reliable. For the small scale sensors with reasonably
small currents; Pt electrodes can act as quasi-
reference electrodes. For amperometric
measurements, such small variations of reference
electrode are not a signifi
cant problem. Hence, bare
noble metal (e.g. Pt) electrode can itself be used as a
RE for some applications [18].
Since Pt interface potentia
l is sensitive to pH and
peroxide interference, it can’t provide an extremely
stable reference potential required for some
applications. However, it has been shown that some
noble metals covered with their oxide also act as
pseudo-reference electrodes
(e.g. Pd/PdOx, Ir/IrOx)
[18]. Hence, coating Pt with a suitable insulating
layer can perform as a stable RE. For example, a
somewhat inert layer of surface oxide can be used to
make a Pt/PtOx RE. Such Pt based reference
electrodes can also be used in harsh environments for
longer durations compared to Ag/AgCl based
electrodes. Since Pt is a noble and very inert
material, it is not easy to oxidize it even with strong
oxidizers such as hydrogen peroxide. In this work, Pt
oxidation was attempted using strong oxygen plasma
as well as using strong acidic oxidizing agents (like
sulfuric acid) at elevated
electrochemical potentials.
Oxygen Plasma exposure
in Reactive Ion Etching
systems showed some oxidi
zing effects on the Pt
surface and the electron diff
raction studies showed
some Oxygen as part of the film. The films were
heated for long time to release any oxygen physically
adsorbed on the surface prior to such testing.
However, it is difficult to confirm the chemical
nature of such thin films reliably. Nonetheless, the
resulting RE resulted in higher electrochemical
stability suggesting that the film had become a better
reference electrode material
than bare Pt. SEM’s of
the films are shown in the Fig. 5.
4. Geometrical Optimization
To fulfill the sensor design constraints, some of
geometrical optimization pr
ocedure is required. For
large scale (macro) devices, this is not very
significant, although still relevant, since adequate
performance can be achieved using many different
types of geometries.
However for small scale devices it is very
important to study all the factors that can affect
sensor performance. In this work, the designs were
mainly based upon rectan
gular elements as the
CMOS process layouts only allow rectangular
designs. There are different possible designs that can
fit within the allowed sensor region (a 500 um by
500 um rectangle). A combination of Finite Element
Modeling (using COMSOL) and experimental
verification was used to converge to the best possible
sensor design. Some chosen designs (named by
alphabets) are shown in Fig. 6. Here, the WE and CE
material is Pt and the RE material is Ag
(before Chlorination).
(a) Pt
(b) PtOx
Fig. 5.
SEM of Pt based
RE materials.
Fig. 6.
Integrated Sensor Geometries.
The designs were tested
experimentally to
compare their performance to
choose the best design.
Deductions can be made from the best design to
formulate the criterion for
the optimized geometry as
a starting point for other applications.
5. Fabrication
Sensors were fabricated on different CMOS
compatible substrates (ins
ulated Si, glass, CMOS
dies). Sensors on insulated Si were used for
prototyping to be able to optimize the designs
N
A
K Z
Sensors & Transducers, Vol. 191,
Issue 8, August 2015, pp. 12-20
16
without the need of multiple CMOS runs. For
CMOS, the top metal pad openings were used to
define the electrode shapes for sensor design.
Photolithography was used for defining electrode
patters on the different su
bstrates. This was followed
by evaporation of 10/100 nm of Ti/Pt layer and
solvent liftoff. Another lithography step was used to
define the RE, followed by evaporation of 200 nm of
Ag followed by solvent liftoff. A low-power
(10 mW) Chlorine Plasma in a Unaxis RIE system
was used to form a thin layer of AgCl on top of the
Ag film. A typical resulting sensor is shown in
Fig. 7. In this case, the WE and CE are covered by Pt
and the RE is covered by Ag/AgCl.
Fig. 7.
Integrated
Electrochemical Sensor.
6. Cleaning
The sensors were thoroughly rinsed with DI water
before testing. Then, the sensors were connected to a
CHI 7051D Potentiostat to be cleaned
electrochemically before any further
functionalization and testing. Cyclic Voltammetry
scans were performed between -0.3 V to 1 V vs.
Ag/AgCl electrode in dilute sulfuric acid solution
(0.01M) until stable results were obtained [19]. This
ensured that the exposed surfaces are
electrochemically clean and reproducible.
7. Testing and Results
The prototype sensors were connected to a CH
Instruments 7051D Potentiostat to make
measurements. The sensing part was immersed in a
solution inside a test beaker. The solution was spiked
with different concentrations of test samples and the
electrochemical data was measured (using different
techniques including cyclic voltammetry,
chronoamperometry, impedance spectroscopy,
constant potential electrolysis and impedance-
potential curves). For some experiments, this was
done on a hot plate to get measurements at body
temperature (37
o
C). Phosphate Buffered Saline (PBS,
pH=7.4) solution was used as background solution.
Each sensor design was fabricated with both a Pt
RE and Ag/AgCl RE as shown in Fig. 8.
(a)
(b)
Fig. 8.
(a) Design M w/Ag/AgCl RE
(b) Design M w/Pt RE.
The stability of the RE was tested first by
measuring its potential relative to a commercial
stable RE, which in our
case was Ag/AgCl RE from
CH instruments. Secondly, Cyclic Voltammetric
measurements were performed using sensors
utilizing either the Ag/AgCl RE or the Pt RE. Results
indicated that Ag/AgCl RE
proved more stable than
Pt RE especially at higher reference potentials as
seen in Fig. 9. This is due to interference from
species reacting with Pt at
such potentials, most
notably hydrogen peroxide. However, it also shows
that a Pt RE can still be used for less sensitive
applications. Although, bare Ag showed a pretty
stable response, peroxide interference had some
effect on it. The formation of AgCl layer on top
makes it more independent of conditions in test fluid.
Also, since the chloride ion concentration in body
remains pretty constant, the Ag/AgCl RE can be used
without any special concen
trated coatings (e.g.
saturated KCl). After Chlorination, AgCl became
quite stable and peroxide did not have any
appreciable effect on it. Ba
re Pt showed more issues
regarding stability and peroxide interference.
Oxidation (PtOx) proved to increase stability and
decrease peroxide interferen
ce effects as expected.
The results from the RE
tests are summarized in
Table 1. These results indicate that altough the
Ag/AgCl electrode is quite stable, the ‘Pt/PtOx’
electrode can be suitable fo
r many applications where
small voltage perturbations can be tolerated.
The sensor geoemtries can be compared based
upon their perforamcne
for appropraite sensing
applications. Hence, the geometries were compared
using hydrogen peroxide testing to determine their
sensitivity as it is the underlying sensing mechanism
for most metabolic sensor
s employing oxidoredutace
enzymes [14]. For a typical peroxide sensor, results
of ameprometric experiments at 0.6 V cell potential
(vs. Ag/AgCl RE) are shown in Fig. 10.
Comaprison of different geometries for sensor
response was performed using such experiments.
Both Ag/AgCl RE and Pt RE based geometris were
compared.The comparison of maximum sensor
currents at 0.6 V potential between WE and RE
(Ag/AgCl or Pt) using amperometry is shown
in Fig. 11.
CE
WE
RE
Sensors & Transducers, Vol. 191,
Issue 8, August 2015, pp. 12-20
17
Fig. 9.
Comparison between
integrated RE Materials.
Table 1.
Comparison of RE materials.
Electrode
Material
Temporal Stability
(voltage change)
Peroxide
Interference
(voltage
change)
Ag
15 mV
120 mV
Pt
30 mV
60 mV
Ag/AgCl
5 mV
3 mV
Pt/PtOx
20 mV
10 mV
The sensors were tested using other methods as
well to confirm these indings. Spectroscopic
techniques including Electrochecmail Impedance
spectroscopy and cell potential spectroscopy are very
useful means for such confirmation [20]. These
methods were used to confirm the results obtained
using amerometric and voltammetric measurements
as noted before. For the EIS experiments using the
commerica Potentiostat, signal frequency was swept
from 100 kHz to 0.1 mHz at a bias of 50 mV.
Fig. 10.
Hydrogen Peroxide Sensing using
integrated Sensors.
Sensor Current (n
A
)
Fig. 11.
Comparison of Sensor Geometries for Hydrogen
Peroxide Sensing (a) Ag/AgCl RE (b) Pt RE.
The arrow indicates the
direction of increasing
frequency on the Nyquist plots. The results
confirmed the same trend in electrochemical
properties as shown in Fig. 12 (PBS) and Fig. 13
(10 mM Peroxide).