WEARABLE GAS SENSOR MICROSYSTEM FOR PERSONAL
HEALTHCARE AND ENVIRONMENTAL MONITORING
By
Xiaoyi Mu

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Electrical Engineering – Doctor of Philosophy
2013

ABSTRACT
WEARABLE GAS SENSOR MICROSYSTEM FOR PERSONAL HEALTHCARE AND
ENVIRONMENTAL MONITORING
By
Xiaoyi Mu
Exposure to air pollutants is an ever present concern for human health in modern society.
Most existing tools for monitoring air pollution are stationary and incapable of accurately
measuring individual exposures that vary with personal lifestyles and environments. To study the
health impacts of personal exposure, researchers need new monitoring tools that are suitable for
long term personal use. This dissertation seeks to overcome the challenges of implementing an
inexpensive, highly miniaturized, and rapidly responding gas sensor array system to measure air
pollutants for personal health. By thoroughly studying the requirements of wearable gas sensing
and the characteristics of prevailing gas sensing technologies, room-temperature-ionic-liquid
(RTIL) electrochemical gas sensing technology is identified for the wearable microsystem
development. To overcome the RTIL-based sensor’s inherent slow response, a planar
porous-electrode-on-permeable-membrane (PEoPM) structure is introduced, demonstrating a
180% faster response to SO2 compared to a traditional structure. Microfabrication processes are
utilized to miniaturize PEoPM sensors and construct a physically flexible sensing platform. To
achieve a low-cost low-power sensory system, a new electrochemical instrumentation circuit
topology is introduced. The new circuit uniquely utilizes the inherent nature of electrochemical

sensor properties to significantly reduce circuit complexity compared to traditional architectures.
Finally, this thesis presents the first ever single-chip CMOS electrochemical gas sensor that
further miniaturizes the sensory microsystem. The monolithic CMOS-RTIL gas sensor occupies
2

0.48mm per sensing channel and demonstrates a limit of detection that is seven times better
than a hybrid (multi-chip) solution. The results of this research lay a solid foundation for a
personally wearable environmental monitor that could greatly improve human healthcare.

ACKNOWLEDGEMENTS
Though only my name appears on the cover of this dissertation, a great many people have
contributed to its production. I owe my gratitude to all those people who have made this
dissertation possible and because of whom my graduate experience has been one that I will cherish
forever.
First of all, I would like to thank my advisor Professor Andrew J. Mason for his continued
invaluable mentorship, encouragement and inspiration through my PhD research. I would also like
to acknowledge my dissertation committee including Professor Wen Li, Professor Prem Chahal
and Professor Scott Calabrese Barton for their valuable feedback that made this dissertation
possible.
I must also thank my colleagues at AMSaC lab who helped me put my ideas to practice.
Special thanks to Yue’s direction, idea sharing and answering my questions. I am grateful to
Xiaowen, Haitao, Lin and Yuning for their help. It is my great pleasure to know and work with
them.
I would like to thank our collaborators Professor Zeng’s group. Special thanks to Dr. Wang’s
valuable discussions and idea sharing. I also appreciate Professor Zeng and Min Guo for their
cooperation, valuable knowledge share and discussions.
Finally and most importantly, I want to thank my parents and brother for believing in me, and
my wife for supporting me every step of the way.

iii

TABLE OF CONTENTS

LIST OF TABLES ....................................................................................................................... viii
LIST OF FIGURES ....................................................................................................................... ix
1

Introduction ....................................................................................................................... 1
1.1 Motivation .......................................................................................................................... 1
1.1.1 Air pollutants effect on human health and research method ........................................ 1
1.1.2 Air pollutant monitoring and data acquisition platform............................................... 2
1.2 Challenge ............................................................................................................................ 3
1.3 Goal .................................................................................................................................... 6
1.4 Thesis outline ..................................................................................................................... 7

2

Background ....................................................................................................................... 8
2.1 Gas sensing technology for wearable air pollutant monitoring device ............................. 8
2.1.1 Gas sensor requirements for wearable system implementation ................................... 8
2.1.2 Analysis of gas sensing technologies for wearable system.......................................... 9
2.1.3 Electrochemical sensing technology .......................................................................... 12
2.1.4 Summary of motivation for RTIL-based electrochemical gas sensor ........................ 15
2.2 Electrochemical interface IC ............................................................................................ 15
2.2.1 Potentiometric technique ............................................................................................ 16
2.2.2 Amperometric technique ............................................................................................ 16
2.2.3 EIS technique .............................................................................................................. 19
2.2.4 Summary of electrochemical IC circuit design .......................................................... 20
2.3 Integration approaches for gas sensor microsystem ........................................................ 21
2.3.1 Comparison between hybrid and monolithic approach.............................................. 21
2.3.2 Analysis of monolithic gas sensor implementation approach .................................... 22
2.3.3 Summary of integration approach for electrochemical gas sensor microsystem ...... 28

3

Development of RTIL-based Sensor for Microsystem Platform ............................... 30
3.1 Overview .......................................................................................................................... 30
3.2 Development of fast-response RTILs-based electrochemical sensor .............................. 30
3.2.1 Sensor’s structure design for fast response ................................................................ 30
3.2.2 Fabrication of PEoPM sensors ................................................................................... 33
3.2.3 Test of PEoEM sensors ............................................................................................... 36
3.3 Miniaturization approach of PEoPM sensor arrays ......................................................... 41
3.3.1 Fabrication of miniaturized PEoPM sensors .............................................................. 41
3.3.2 Package of miniaturized PEoPM sensor arrays ......................................................... 46

iv

3.3.3 Test of miniaturized PEoPM sensor arrays ................................................................ 48
3.3.4 Integration of PEoPM sensor array within a wearable monitoring system ............... 53
3.4 Development of flexible PEoPM sensors ........................................................................ 53
3.4.1 Fabrication of flexible PEoPM sensors ...................................................................... 54
3.4.2 Test of flexible PEoPM sensors .................................................................................. 58
3.5 Conclusion ........................................................................................................................ 63
4

Development of a Compact Low-Power Amperometric Instrumentation with
Current-to-Digital Readout ........................................................................................... 64
4.1 Overview .......................................................................................................................... 64
4.2 Background ...................................................................................................................... 65
4.2.1 Electrochemical sensors and equivalent circuit models ............................................. 65
4.2.2 Traditional amperometric instrumentation ................................................................. 67
4.3 Compact amperometric instrumentation design .............................................................. 70
4.4 Performance analysis ....................................................................................................... 75
4.4.1 Performance difference from traditional instrumentation .......................................... 75
4.4.2 Second-order effects of the sensor equivalent circuit model ..................................... 77
4.5 Results .............................................................................................................................. 79
4.5.1 CCDAI implementation .............................................................................................. 79
4.5.2 Experimental results ................................................................................................... 81
4.5.3 Performance comparison ............................................................................................ 84
4.6 Conclusion ........................................................................................................................ 86

5

Development of a CMOS Monolithic Electrochemical Gas Sensor Microsystem ... 87
5.1 Overview .......................................................................................................................... 87
5.2 System architecture .......................................................................................................... 87
5.2.1 Electrochemical sensor system ................................................................................... 87
5.2.2 CMOS monolithic sensor microsystem concept ........................................................ 89
5.3 CMOS design ................................................................................................................... 91
5.3.1 Potentiostat ................................................................................................................. 91
5.3.2 Amperometric readout circuit..................................................................................... 92
5.4 Monolithic sensor fabrication .......................................................................................... 93
5.5 Results .............................................................................................................................. 96
5.5.1 Test setup .................................................................................................................... 96
5.5.2 Electrical experimental results ................................................................................... 97
5.5.3 Chemical experimental results ................................................................................... 98
5.6 Conclusion ...................................................................................................................... 105

6

Summary, Contributions and Future Work .............................................................. 106
Summary ........................................................................................................................ 106

6.1

v

6.2
6.3

Contributions .................................................................................................................. 107
Future work .................................................................................................................... 108

APPENDICES ............................................................................................................................. 111
Appendix A: EIS equivalent circuit modeling ......................................................................... 112
Appendix B: Fabrication procedure for PEoPM sensors......................................................... 118
BIBLIOGRAPHY ...................................................................................................................... 125

vi

LIST OF TABLES

Table 2.1. Summary of gas sensing technologies. ····················································· 12
Table 2.2. Summary of monolithic gas sensors. ······················································· 28
Table 3.1. Average sheet resistance and standard deviation with different width configurations. ·
···································································································· 44
Table 4.1. Area occupation of IC modules in ON 0.5µm CMOS fabrication process and the
comparison between the traditional amperometric instrumentation circuit and the CCDAI. ····· 85
Table 4.2. Power consumption of IC modules in ON 0.5µm CMOS fabrication process and the
comparison between the traditional amperometric instrumentation circuit and the CCDAI. ····· 85
Table 5.1. Performance summary of the CMOS interface circuit. ·································· 99
Table 5.2. Off-CMOS and on-CMOS electrochemical RTIL sensor measured parameters and
calculated volume and height. ············································································ 104
Table 5.3. Off-CMOS and on-CMOS electrochemical RTIL sensor noise level comparison.
······························································································· 104

vii

LIST OF FIGURES

Figure 1.1. Concept of the wearable platform for air pollutant monitoring and data collection. 3
Figure 1.2. Architecture of a portable air pollutant monitoring device. ····························· 5
Figure 2.1. Simplified block diagram of circuits for potentiometric measurements. ············· 16
Figure 2.2. Simplified block diagram of circuits for amperometric measurements. ·············· 17
Figure 2.3. The structure of basic potentiostat. ························································ 18
Figure 3.1. (a) Conventional sensor structure: response time is slow due to slow gas diffusion
through RTIL; (b) RTIL-in-permeable-membrane structure: response time is improved by a
controllable thin electrolyte layer; (c) Porous-electrodes-on-RTIL structure: response time is
improved by bypassing gas diffusion in RTIL. (d) Electrodes-on-permeable-membrane structure:
response time is improved due to fast gas diffusion in the permeable membrane. ················· 32
Figure 3.2. (a) Image of RTIL [C4mpy][NTf2] droplet on porous PTFE. The contact angle is 95˚.
(b) Image of RTIL [C4mpy][NTf2] droplet on gold-coated porous PTFE. The contact angle is 18˚.
·································································································· 34
Figure 3.3. PEoPM sensor fabrication process. A stainless steel mask was mounted against the
porous PTFE and then a gold film was deposited. After removing the mask, patterned planar gold
electrodes were formed. Finally, RTIL was coated on the electrodes to form the sensor. ········· 35
Figure 3.4. Photograph of fabricated PEoPM device. The WE and CE are interdigitated with
finger widths of and the gaps of 200 µm. The O-ring isolates the sensor area from the contact pads.
·································································································· 36
Figure 3.5. Cross-sectional view of the gas testing chamber. Separate inlets and outlets on the top
and bottom of the chamber allow gas to be introduced along two separate paths to effectively
characterize two different sensor structures, where path1 corresponds to a conventional sensor
structure shown in Figure 3.1(a) and path 2 corresponds to the PEoPM sensor structure shown in
Figure 3.1(d).·································································································· 37
Figure 3.6. Impedance amplitude |Z| spectrum from 0.1 Hz to 10 kHz recorded from an
RTIL-based PEoPM sensor presenting under a SO2 environment. SO2 concentration varies from 0

viii

to 25ppm. Obvious impedance amplitude change to different SO2 concentration was observed,
with a stronger response below 100Hz. ··································································· 38
Figure 3.7. SO2 calibration curve for RTIL-based PEoPM sensor. Blue dots represent the
recorded -Zim values at 10Hz. Red curve represents the fitting curve follows Equation 3.1. The
2

fitting curve achieves high least-squares correlation coefficients (R =0.9957). ···················· 39

Figure 3.8. Transient response of RTIL-PEoPM sensor to 25 ppm SO2, with gas flow from two
different paths (see Figure 3.5.) representing a conventional sensor structure with 56 s full scale
response and a new rapid-response structure that reduces response time to 20 s. ·················· 40
Figure 3.9. Profile scan by Dektak3 Surface Profiler. The photoresist AZ4620 covered the rough
surface of porous PTFE and a good step was formed. ·················································· 43
Figure 3.10. Miniaturized RTIL-based PEoPM sensor fabrication process. ························ 45
Figure 3.11. Photograph of fabricated miniaturized PEoPM device. The WE and CE are
interdigitated with finger gaps of 100µm. The O-ring isolates the RTIL area from the contact pads.
·································································································· 46
Figure 3.12. (a) Illustration of the parts assembly for PEoPM sensor array. First, four
electrodes-patterned PEoPM sheets were placed on wire-routed PCB substrate, with pads
connected to wire contacts on PCB by conductive epoxy. Then O-rings were attached on a
windows-opened PCB lib and mounted on the wire-routed PCB substrate by tighten bolts, to create
RTIL reservoirs. Finally, [C4mpy][NTf2] were dropped in the reservoirs to create RTIL sensing
layer. (b) Photographs of assembled PEoPM sensor array from front, back and side views. The
sensor array occupies 1.55” x 1.25” including the connector to be electrically connected with
sensor interface circuit. ······················································································ 48
Figure 3.13. CH4, SO2, NO2, O2 calibration curves for RTIL sensors from EIS test data at 10Hz.
·································································································· 51
Figure 3.14. Selectivity experiment. 20.4% O2, 3% CH4, 10 ppm SO2, and 10 ppm NO2 were
purged into the test chamber sequentially in four steps and the sensor array’s impedance responses

ix

were recorded in each step. Taking the impedance response to O2 as 100% signal S, the sensor
array responses to all four gases are normalized as ΔS/S (%) and plotted. ·························· 52
Figure 3.15. 2.6” × 2” × 1” prototype wearable air pollutant monitoring system with miniaturized
sensor array, multi-electrochemical instrumentation board (MEIB), and a µ-Controller (µC) board.
·································································································· 53
Figure 3.16. Coating experiments illustration. Case1 illustrates small volume pure RTIL coating
(Vol.=0.2 μL) result. RTIL droplets were coated only electrodes surface. Case2 illustrates large
volume pure RTIL coating (vol.=1 μL) result. RTIL droplets were coated on electrode surfaces as
well as the gap between electrodes. Case3 illustrates ethanol/RTIL mixture coating
(Rethanol/RTIL=2, RTIL’s vol.=0.2 μL) result. The initial mixture droplet was coated the electrodes
area as well as the gap between electrodes. After ethanol removal, RTIL still occupied the initial
droplet region. Case4 illustrates ethanol/RTIL mixture coating (Rethanol/RTIL=4, RTIL’s vol.=0.2
μL) result. The initial mixture droplet filled the pores inside the porous PTFE as well as covered
electrodes surface. After ethanol removal, RTIL still stayed in the porous PTFE.·················· 55
Figure 3.17. (a) Photograph of microfabricated flexible PEoPM RTIL-based sensor. (b)
Photograph of convex bend of the PEoPM RTIL-based sensor. (c) Photograph of concave bend of
the PEoPM RTIL-based sensor. ············································································ 57
Figure 3.18. Gas test setup and position of the sensor within the flow cell.························· 58
Figure 3.19. Cyclic voltammograms (scan rate = 500 mV/s) for the reduction of O2 at WE. O2
concentration varies from 0-21%. O2 reduction current peak was observed at -1.4V vs Au. The
relationship between peak current and O2 concentration from 0% to 21% is linear. ··············· 59
Figure 3.20. Current vs. time curve at various O2 concentrations when the potential is held at –1.4
V vs Au. N2 is the background gas. O2 concentration steps up from 0% to 21% and steps down
from 21% to 0%. ····························································································· 61
Figure 3.21. O2 calibration curve for RTIL sensor using data extracted from Figure 3.19.······· 61

x

Figure 3.22. Constant-potential current response measured over three cycles of alternate exposure
to 1% O2 and pure N2 flow at an applied potential of -1.4 V. ········································· 62
Figure 4.1. Equivalent circuit model of electrochemical sensor cell. (a) Randles model (b)
Complete model considering both AC and DC stimulus. (c) Simplified model for circuit analysis.
·································································································· 66
Figure 4.2. Schematic of a traditional amperometric instrumentation circuit including
potentiostat and current-mode ADC. ······································································ 68
Figure 4.3. Waveforms of the current on the integrator input Iint, the voltage on the integrator
output Vint, and the digital output of the comparator Dn. ·············································· 69
Figure 4.4. Schematic of the electrochemical sensor system. It consists of the traditional
amperometric instrumentation circuit and the simplified electrochemical sensor equivalent circuit
model.
·································································································· 71
Figure 4.5. Derivation of the instrumentation topology. The input current source is folded to
parallel connect with integrator capacitor. ································································ 72
Figure 4.6. Schematic of the modified amperometric instrumentation circuit with sensor
equivalent circuit model. ···················································································· 73
Figure 4.7. Schematic of the optimized amperometric instrumentation circuit with sensor
equivalent circuit model. ···················································································· 74
Figure 4.8. Schematic of CCDAI with the simplified sensor equivalent circuit model. ········· 74
Figure 4.9. Vint waveform illustration when considering Rs in the equivalent circuit model. ·· 77
Figure 4.10. Illustration of Isens in time domain. ························································ 79
Figure 4.11. Test setup for electrical and chemical experiment. ······································ 80
Figure 4.12. A hysteresis comparator realization with adjustable upper/lower bound. ············ 80
Figure 4.13. DNL and INL of the CCDAI. Both DNL and INL in the current range are better than
-49dB, implying an 8 bit of effective resolution. ························································ 81

xi

Figure 4.14. The Faradaic current generated by 6 mM of potassium ferricyanide as function of
time when VWE-RE=190mV. Red line represents data recorded by CHI760C and blue line
represents data recorded by CCDAI. ······································································ 83
Figure 4.15. The Faradaic current recorded by the CCDAI at VWE-RE=190mV as function of time
for 0- 6 mM of potassium ferricyanide. ··································································· 83
Figure 4.16. Calibration curve of Faradaic current vs potassium ferricyanide concentration. The
current values were the average values from 200s to 300s. Fitting curve was presented as a straight
2

line. R values of the fitting line are 0.991 and 0.996 for the data acquired by CCDAI and
CHI760C, respectively. ······················································································ 84
Figure 5.1. Block diagram of the electrochemical sensor system. ·································· 88
Figure 5.2. Conceptual illustration of the CMOS monolithic electrochemical gas sensor
microsystem. ·································································································· 91
Figure 5.3. Schematic of the monolithic microsystem, consisting of the amperometric
instrumentation circuit and the RTIL sensor. The green rectangle block illustrates the clock
waveform. ·································································································· 92
Figure 5.4. Post-CMOS pocess flow for on-chip RTIL sensor fabrication: Shipley 1813
photoresist is spin-coated (a) and developed (b). Titanium and gold in deposited by PVD (c).
Photoresist is then rinsed off to leave the electrode (d). CMOS chip with on-chip electrode is wire
bonded to package (e). A droplet of RTIL is casted on the electrodes (f). ··························· 95
Figure 5.5. Die photographs of CMOS chip fabricated by foundry and on-CMOS RTIL sensor
formed by post-CMOS fabrication. ········································································ 95
Figure 5.6. Chemical test setup. ·········································································· 97
Figure 5.7. The rail-to-rail opamp output voltage range and linearity error as a function of input
voltage. The absolute error of the output voltage is less than 0.05% over the range from 0.03 V to
4.90 V.
·································································································· 99
Figure 5.8. Amperometric readout output range and linearity errors as a function of input current
when fs = 100 kHz, Vref2 = 1.3 V. In the range of -1 µA to 1 µA, the absolute linearity errors are
less than 4 nA. ································································································ 99
Figure 5.9. Output of amperometric readout circuit for on-CMOS RTIL sensor response to
oxygen. Oxygen concentration varies from 0 - 21% with a step of 2.1%. Oxygen exposures were

xii

10 s in duration, preceded and followed by exposure to pure nitrogen for 30 s. The amperometric
circuit was set to Vref1=2.1 V and Vref2=1.3 V (the sensor was applied at -0.8V respectively), and
fs=100 kHz. ································································································ 101
Figure 5.10. Oxygen calibration curve for monolithic RTIL gas sensor using data extracted from
2

Figure 5.8. A high level of linearity (R = 0.995) was achieved. ···································· 101
Figure 5.11. Repeatability test. Constant-potential current response measured over five cycles of
alternate exposure to 4.2% oxygen and pure nitrogen flow at an applied bias of -0.8 V. The
standard deviation of the five voltage data is 89 µV and corresponding repeatable LOD of oxygen
is 0.06%.
································································································ 102
Figure 5.12. Off-CMOS and on-CMOS sensor response time (10% - 90%) characterized by
amperometric readout circuit. ············································································ 104
Figure A.1. Randle equivalent circuit model. ························································· 112
Figure A.2. The equivalent circuit model when bias is much lower than over-potential. ······ 115

xiii

1 Introduction

1.1 Motivation
1.1.1

Air pollutants effect on human health and research method
Airborne pollutants are known to threaten human health and safety, causing discomfort,

illness, and even death, particularly among susceptible individuals such as those with pre-existing
cardiovascular or respiratory problems as well as infants and the elderly [1, 2]. Exposure to air
pollutants consistently ranks among the leading causes of illness and mortality. US Environmental
Protection Agency has released National Ambient Air Quality Standards (NAAQS) [3] for the
evaluation of the air pollution effects to public human health, where air quality is quantitatively
identified by six principle pollutants including NOx, CO, O3 and SO2 from short-term and
long-term aspects.
The most pervasive methodology used to study the effects of air pollution on human health is
epidemiology [1-3]. Epidemiological studies model the impact of air pollutants by illuminating
statistical associations between levels of air pollutants and impacts on public human health both in
short and long term. Such studies heavily rely on air pollutant data (species and concentration),
which are acquired from stationary air monitoring sites.

1

1.1.2

Air pollutant monitoring and data acquisition platform

Although epidemiology reveals statistical correlation between air pollutant exposure and
human health, current research cannot produce precise models on the impacts of exposure to air
pollutants on personal health. There are several limitations of existing data acquisition platforms
that hinder the research [1]. Firstly, because of the high cost and bulky size of multi-gas analyzers
using for air pollutant data recording, only a limited number of stationary sites are distributed in
certain regions. Given the density of the distributed stations, the spatial resolution needed to track
variations of pollutant concentrations within the region of interest is not usually provided.
Secondly, a limited number of stationary sites are not able to provide pollutant data covering
human-scale microenvironments. In addition, different daily activity patterns of varied sub-group
of populations—for example adults and children—lead to different levels of exposure. Fixed
monitoring sites fail to capture such differences. Given all these limitations, ambient levels of air
pollutants were shown to be poor predictors of personal exposure, in particular, of acute exposure.
The lack of an effective platform to acquire sufficient air pollutant data, in terms of both temporal
and spatial resolution, remains the key obstacle to fully understand the health effects of pollutant.
The size, cost and immobility of the platform limits its distribution, thus is the bottleneck of the
research on health effects of air pollutants.
To overcome the limitations of existing data acquisition platforms, it is desirable to have an
alternative platform, that can be worn by many people, collect data continuously and automatically

2

Figure 1.1. Concept of the wearable platform for air pollutant monitoring and data collection.
(For interpretation of the references to color in this and all other figures, the reader is referred to
the electronic version of this dissertation.)

upload data to data clouds, thus providing data at high resolution both temporally and spatially as
shown in Figure 1.1. In addition to providing information for epidemiology, this platform would
also provide preventative alerts and managed care services to individuals susceptible to the effects
of exposure.

1.2 Challenge
To realize a wearable platform for air pollutant monitoring, individual monitoring device in
the platform should have features as follows. To be able to monitor air pollutants, the device
should be sensitive to the air pollutants in the range of interest, be reliable, and respond in real time.
To provide preventative alerts and managed care services, the device should have signal
processing and data storage. To upload air pollutant data to data clouds, the device should have

3

network function, like WiFi, Bluetooth or GPRS. To be able to be distributed to many people, the
device should be inexpensive. To be easy to be worn daily, the device should be small and light like
a cellphone or even a wrist watch. To be powered up to days or even months by battery, the device
should be highly power efficient.
Recently, a few portable air pollutant monitoring device prototypes [4-8] have been
developed to realize the wearable platform concept. Their typical architecture is shown in Figure
1.2. It consists of four parts: (1) Multiple-gas sensor array transduces air pollutant information
(species/concentration) to electrical signal; (2) Sensor interface circuit drives the sensor array as
well as readouts the electrical signal; (3) µController/ARM module digitizes and processes the
sensor signal; and (4) Bluetooth/WiFi/GPRS module uploads the data to the data clouds. Those
device prototypes meet some criteria of the wearable platform. They are capable of monitoring
single or multiple air pollutants and uploading data to the Internet. However, they are impractical
to distribute to many people and continuously collect data due to three common drawbacks: high
product cost, high power consumption, and relative large size. Each device costs several hundred
dollars. Their power consumption is around 100mW to 1W, limiting their battery-powered lifetime
only to a few hours. Although the devices are portable, they are too large to be wearable, like the
size of a smart phone or a wrist watch.
The power/size/cost drawbacks of existing gas monitors can be analyzed in terms of their
general architecture. Thanks to modern digital and RF IC technologies accelerated by smart
phones/tablets, the µController/ARM module and Bluetooth/WiFi/GPRS module are typically

4

Figure 1.2. Architecture of a portable air pollutant monitoring device.
implemented using commercial IC chips with low size, low cost and low power. In contrast,
currently commercial sensor arrays and sensor interface circuits are of high cost and large size.
Therefore, due to the underdeveloped sensor array and sensor interface circuit technologies,
existing portable air pollutant monitoring device prototypes cannot directly be used for the
wearable platform. Mirroring the success of embedded camera sensor units in smart phones, if the
price of the sensor unit (including sensor interface circuit) falls below $1 and the size smaller than
3

1cm , it is possible to develop a wearable microsystem with similar size to Google Glass or smart
watches by incorporating sensors, sensor interface circuits and digital/RF modules together for air
pollutant monitoring.
To overcome the bottlenecks of the wearable platform, research efforts are needed that focus
on the miniaturization of the sensor array and power-efficient interface circuits utilizing
cost-effective fabrication processes. From the system point of view, several technical challenges
remain in order to achieve a cost-effective gas sensor microsystem for long-term use.
The first critical challenge lies in developing a miniaturized high-performance sensor device,
because available commercial sensors cannot satisfy the wearable system’s requirement in terms

5

of power and size. The miniaturized sensor device should be able to sense the air pollutants in the
range of interest in real time, with long lifetime, small feature, and low power.
The second challenge lies in developing sensor interface electronic circuits. Traditionally,
high-performance, high-cost, bulky instrumentation is used to characterize gas sensors in sensing
technology development. When applied to field sensing, a low-power, low-cost IC solution is
required.
The third critical challenge lies in the integration approach for sensor devices and electronics
circuits. The monolithic microsystem approach, that implements both sensor and electronics
circuits in a single CMOS chip, is considered to be a cost-effective and size-compact solution.
However it rouses technique challenges like process compatibility during implementation.

1.3 Goal
The goal of this dissertation is to experimentally develop a wearable gas sensor microsystem
utilizing a sensing technology and integration approach that satisfies all the requirements,
including performance, cost, and size. Scientific contributions are expected in the areas of
miniaturization of high performance gas sensor arrays, compact sensor interface circuits, and
sensory microsystem integration.

6

1.4 Thesis outline
Chapter 2 describes the background of this research. Options of potential gas sensing
technologies suitable for wearable gas sensor microsystem are discussed. With the best sensing
technology selected, options of potential electronics are discussed. The possible integration
approaches are discussed with a review of current research progress. Chapter 3 reports the sensor
device development utilizing the selected sensing technique. Its miniaturization approach is also
presented. Chapter 4 introduces an optimized compact sensor interface circuit implementation
tailored to the chosen sensing methodology. Chapter 5 introduces the development of a monolithic
sensor microsystem platform that enables highly-miniaturized wearable air pollutant monitoring
device. Conclusions and contributions are given in Chapter 6.

7

2

Background

2.1 Gas sensing technology for wearable air pollutant monitoring device
2.1.1

Gas sensor requirements for wearable system implementation
To provide information for epidemiology to build accurate personal exposure model, and

provide preventative alerts and managed care services to individuals susceptible to the effects of
exposure, a wearable air pollutant monitoring platform is demanded. To realize a wearable
platform for air pollutant monitoring, individual monitoring devices should meet following
criteria.
(1) Sensitive to the air pollutants (NO2, SO2, O3, CO, volatile organic compounds (VOC),
etc.) in the range of interests, be reliable and respond in real time.
(2) Signal processing and data storage capability.
(3) Networking capability.
(4) Cost effective to be widely distributed.
(5) Miniaturized size and light weight equivalent to a cellphone or even a wrist watch.
(6) Power efficient that are can be powered up to days or even months by battery. Power
range from mW to µW is desirable.
To satisfy the device criteria, the gas sensor in the device should meet following
requirements.

8

(1) Sensitive to the air pollutants (NO2, SO2, O3, CO, VOCs, etc) in the range of interest
with good selectivity.
(2) Good reliability. It should function for months to years without extract maintenance.
(3) Rapid response time to achieve high temporal resolution, sampling rate up to 10Hz.
3

(4) Miniaturized size to be fit into cellphone or even wrist watch. Size around cm or even
3

mm is desirable.
(5) Low power consumption around µW is desirable.

2.1.2

Analysis of gas sensing technologies for wearable system
To develop a miniaturized gas sensor that meets all the requirement, prevalent gas sensing

technologies should be analyzed, and the most promising sensing technology would be chosen for
the sensor development. The five types of prevalent gas sensing technologies are: gas
chromatography (GC)-based [9], non-dispersive infrared (NDIR) [10], metal oxide semiconductor
[11], conductive polymer [12], and electrochemical sensing technology [13].
GC-based

sensing technology [9] refers to the gas sensing technology that utilizes a

semi-selective gas sensor and adopts GC to enhance selectivity. GC plus mass spectrometry (MS)
[14] and flame ionization detector (FID) [15] are standard gas analyzers for stationary air pollutant
monitoring sites [16]. GC-based gas analyzers achieve high sensitivity, selectivity, and reliability.
However, traditional GC-based gas analyzers are bulky and expensive, require complex operations
and regular maintenance. Although development of micro GC [17, 18] , micro MS [19, 20] and

9

micro FID [21] have been reported during last decade, the GC’s high degree of complexity and
sophistication still results high production cost and maintenance effects. In addition, the operation
of GC-based gas analyzers involves a gas sampling process and a pre-concentration process,
which brings significant drawbacks. The gas sampling process is time consuming, usually taking
minutes to output one dataset. Thus the sampling rate is pretty low. The gas pre-concentration
process needs a high-power-consumption heater to release absorbed gas analytes. Thus the power
consumption of GC-based gas analyzers is inefficient.
NDIR sensing technology [10] detects gas by determining the absorption of an emitted
infrared light source through a certain air sample. Commercial NDIR sensors can detect air
pollutants – CO, NO, SO2, O3 and VOCs [22]. NDIR sensors exhibit good sensitivity and good
stability. However, they necessitate a relatively complicated optical system, making it expensive,
bulky and high power consumption (hundred mW).
Metal oxide semiconductor gas sensing technology [11] has been developed since 1962 and
demonstrated for a variety of air pollutants—CO, NO, NO2, O3 and VOCs [22]. The sensing
principle is: the film resistance changes when gases are absorbed or combust with catalyst material.
By choosing different metal oxide semiconductor materials and doping/coating different catalyst
materials, metal oxide semiconductor sensors exhibit good selectivity to gas analytes. Other
advantages include light weight, miniaturized size, quick response time, and low cost. However,
metal oxide semiconductor gas sensor requires operating at very high temperature
(usually >250°C), and thus consume a lot of power. Even a miniaturized sensor could consume

10

power around 200 mW [23]. In addition, although they have long operation lifetime and
reasonable parameter stability, they degrade as they are used.
Conductive polymer sensing technology [12] uses different types of conductive polymer to
absorb or react with a specific gas, causing a change of the polymer film resistance. It has been
reported as capable of sensing CO, NO2 and a few VOCs in the air pollutant category. Conductive
polymer based sensors consume low power given their room-temperature operations. They exhibit
high sensitivity and short response time to analytes. They have low fabrication cost because of a
simple fabrication process and are easy to miniaturize. However, they suffer low selectivity,
long-term instability and irreversibility.
Electrochemical sensing technology [13] utilizes electrochemical reaction to generate an
electrical signal (current or potential) to sense gases. Electrochemical sensors have been proven to
be effective for detecting environmental gases such as CH4, CO, CO2, NO, NO2, SO2, H2, O3, O2
and VOCs [22, 24-27]. They exhibit good selectivity, low power consumption (in liquid
electrolyte), wide dynamic range and low production cost. Their response time and reliability vary
with the electrolyte. They have the potential to be miniaturized with the newly-developed
electrolytes.
In all, the five gas sensing technologies are qualitatively summarized in Table 2.1 in terms of
gas sensor criteria for microsystem implementation. Although GC-based sensing technology is
capable of sensing all air pollutants of interest, and exhibits excellent selectivity and reliability, its
high cost, slow response, and high power consumption exclude itself from being an ideal candidate

11

for wearable air pollutant monitoring microsystem. Apart from GC-based sensors, only
electrochemical sensors are sensitive to all the air pollutants of interest and they exhibit the best
comprehensive strength to satisfy all requirements for air pollutant monitoring microsystem.
Therefore, the electrochemical sensing technique is chosen to develop miniaturized sensor arrays
for gas sensor microsystems.
2.1.3

Electrochemical sensing technology
To develop an electrochemical gas sensor that satisfies all the requirements for the air

pollutant monitoring microsystem, electrochemical sensing technology is reviewed in detail.
Technique challenges of the wearable electrochemical sensor development are thereby revealed.
Electrochemical gas sensing [13] is based on the measurement of an electrical signal caused
by chemical reaction in which the gas analyte is involved. The reaction takes place when the gas
diffuses to the interface of an electrode and a supporting electrolyte and a certain electrical
Table 2.1.
Summary of gas sensing technologies.
Sensing technology
Detectable air
pollutants

GC/MS or
GC/FID

Metal oxide Conductive
Electrochemical
semiconductor polymer

NDIR

CO, NO, NO2, CO, NO,

SO2, O3, and
VOCs
Selectivity
Excellent
Reliability
Excellent
Low cost
Bad
Fast response
Bad
Miniaturized size
Good
Power consumption Bad

CO, NO,

CO, NO, NO2,
CO, NO2,
SO2, O3 NO2, O3 and
and VOCs SO2, O3, and
and VOCs VOCs
VOCs
Good
Good
Bad
Good
Good
Moderate
Bad
Moderate
Bad
Good
Good
Good
Good
Good
Good
Moderate
Bad
Good
Good
Good
Bad
Bad
Good
Good

12

condition is satisfied. This type of sensor has been utilized in industry over the last 50 years.
Electrochemical gas sensors provide one of the lowest-power approaches combined with
outstanding performance that includes high sensitivity and good selectivity. They are sensitive to a
wide range of air pollutants such as CH4, CO, CO2, NO, NO2, SO2, H2, O3 and O2. Furthermore,
the instrumentation needed for such sensors can readily be implemented as a single
microelectronics chip suitable for wearable/portable devices [28-30].
Different planar electrodes configuration are adopted for different working modes.
Concentric electrodes are used for amperometric mode to achieve a uniform electrical field.
Interdigitated electrodes are used for electrochemical impedance spectroscopy (EIS) mode to
minimize the solution resistance.
In terms of electrolyte type, there are two types of electrochemical gas sensors:
solid-electrolyte sensors and liquid-electrolyte sensors. Solid-electrolyte electrochemical sensors
are utilized in the automotive industry to detect exhaust gases. Because they can only function at
high temperature, to achieve sufficient ion mobility to sense gases, heaters are required if the
solid-electrolyte electrochemical sensors are applied to air pollutant monitoring. This requirement
brings extra power consumption and is not favorable. In contrast, liquid-electrolyte
electrochemical sensors are one of the lowest-power gas sensors because of room temperature
operation. They can detect a variety of gases (e.g. O2, CO, SO2, H2S, NO2) by employing
conventional solvents (e.g. H2SO4/H2O mixtures, or organic solvents such as acetonitrile or
propylene carbonate). However, traditional liquid-electrolyte electrochemical sensors cannot

13

survive in drastic temperature changes and cannot function in extremely dry or humid conditions;
such conditions exist in normal circumstances. In addition, the lifetime of such liquid-state
electrolyte sensors is often determined by the volume of the electrolyte in the sensor and how
quickly the electrolyte dries up. Such features make traditional liquid-electrolyte electrochemical
sensors unfavorable for miniaturization.
To overcome the drawbacks of the traditional liquid-electrolyte electrochemical sensors,
room temperature ionic liquids (RTILs) have been adopted as a new type of liquid electrolyte
recently [31-33]. RTILs are nonvolatile and conductive compounds consisting entirely of ions [31].
Given the special properties, RTILs are a promising electrolyte for robust electrochemical gas
sensors that can operate in extreme conditions. On the one hand, RTIL-based electrochemical gas
sensors inherit the advantages of liquid-electrolyte electrochemical gas sensor. RTILs can function
at room temperature. Thus, power demands are greatly reduced by eliminating the heater required
by many other sensors. On the other hand, RTILs overcome the drawbacks of traditional
liquid-electrolyte electrochemical gas sensor. RTIL has negligible vapor pressure and high thermal
stability. It features longer lifetime, structure that does not require a permeable membrane to seal
liquid, and capability to function in a very small volume. The selectivity of an electrochemical gas
sensor is introduced by the electrochemical potential applied to the sensor. Therefore, an
electrolyte that can be sustained at a wide range of electrochemical potential has a possibility to
sense many gases. Thanks to their wide electrochemical potential windows, RTILs are capable of
sensing a variety of gases, including oxygen [34-39], ambient toxic gases (e.g. NO2 [40], NO [41],

14

NH3 [42], H2S [43]) and VOCs [44, 45]. However, as the viscosity of RTILs is typically larger
than traditional electrolytes by 1-2 orders of magnitude, the diffusion coefficients of gaseous
analytes are significantly lower in RTILs. Because the reaction only takes place when the gas
reaches the interface of the electrode and electrolyte by diffusion, the response of RTIL-based
sensor is slow in a traditional electrochemical cell structure where a thick layer of electrolyte is
usually deployed between the electrodes and atmosphere.
2.1.4

Summary of motivation for RTIL-based electrochemical gas sensor
In all, the electrochemical sensing technique demonstrates overall good performance, and

thus has the potential to meet all requirements for air pollutant gas sensor microsystems. Choosing
RTIL as electrolyte, the electrochemical sensors overcome drawbacks of traditional
liquid-electrolyte electrochemical sensors. The challenge of slow response requires an innovation
in sensor structure. In addition, an approach to sensor array miniaturization is demanded.

2.2 Electrochemical interface IC
Because an RTIL-based electrochemical gas sensor array has been selected for microsystem
implementation, a corresponding sensor interface circuit should be developed to meet system
criteria as well. Traditionally, high-performance, high-cost, bulky instrumentation is used to
characterize gas sensors in sensing technology development. When applied to wearable
microsystems, a low-power, low-cost IC solution is required. Although no IC solution for

15

electrochemical gas sensors has been reported, ICs for other electrochemical applications [28-30]
can be referenced for developing a gas sensor interface IC, given the same working principle.
Based on implementation methods, electrochemical instrumentation circuits can be classified
into three types: potentiometric, amperometric, and EIS technique. In this section, a brief review
on existing electrochemical instrumentation circuits implemented in CMOS IC is presented.
2.2.1

Potentiometric technique
The potentiometric technique measures potential between a working electrode (WE) and a

reference electrode (RE) when a controlled current is passed between the WE and a counter
electrode (CE) [46] (Figure 2.1). Due to relatively large error caused by double-layer charging
effects, this method is seldom used for sensors. No IC instrumentation circuit implementing
potentiometric sensing has been reported.
2.2.2

Amperometric technique
The amperometric technique measures current between a CE to a WE with a current readout

circuit when a controlled potential (sustained by potentiostat) is applied between RE and WE as

Figure 2.1. Simplified block diagram of circuits for potentiometric measurements.

16

shown in Figure 2.2 [46]. In the amperometric sensor, the current is associated with an oxidation
and/or reduction reaction involving the analyte. The controlled potential typically is a constant or
ramp potential. Applying constant or ramp potential stimulus respectively, chronoamperometry
and cyclic voltammetry technologies are widely used in enzymes, DNA and protein detections. In
order to meet different demands for sensor applications, numerous CMOS integrated potentiostats
and current readout circuits have been developed with various functions and performances in the
past decades.
A basic potentiostat is implemented by an amplifier as shown in Figure 2.3. By incorporating
amplifier-based adder, multiple stimulus signals can be added to the electrochemical sensors. By
adjusting stimulus to the WE side, a hardware-efficient bi-potentiostat scheme was implemented
in CMOS IC for redox-enzyme-based biosensors [47]. To improve the output swing and dynamic
range, a fully-differential potentiostat was developed to be capable of being fabricated in a
low-voltage CMOS process [48]. A current-mirror-based potentiostat was presented in [49] to
achieve lower power consumption and occupy smaller area.

Figure 2.2. Simplified block diagram of circuits for amperometric measurements.

17

Figure 2.3. The structure of basic potentiostat.

A basic current readout circuit is composed of a current-voltage convertor and a
voltage-mode analog-to-digital converter (ADC). The former part typically is implemented by a
resistive feedback amplifier or an integrator; the later part typically is implemented by a dual-slope
ADC [50, 51]. An advanced readout circuit approach adapts current-mode ADC to directly digitize
the current [52]. In the current readout circuits design, the key point is to improve the detection
resolution. The main barriers are 1/f noise and thermal noise in the circuits. 1/f noise is the
dominant noise source for DC (or low frequency for the low scan rate case in CV) current
detection circuits. Several techniques, such as chopping and correlated double sampling (CDS)
can eliminate 1/f noise. In [53], the entire current readout chain utilized CDS to reduce 1/f noise as
well as amplifier offset. To reduce thermal noise, one solution is to use a longer integration time
[54, 55], in which a switched-capacitor first-order single-bit Σ-Δ modulator was implemented in
the current-to-digital converter. However, the single Σ-Δ technique reduces the thermal noise by

18

sacrificing the speed of the converter. In [52], a semi-synchronous Σ-Δ analog-to-digital
conversion algorithm, combines continuous time Σ-Δ with time-encoding machines (TEM) to
improve the readout current resolution.
For high throughput array microsystems, power consumption and area are important circuit
design parameters to consider besides noise performance. A half-amplifier structure was
developed to measure small currents without increasing the power consumption and layout area by
eliminating amplifying circuitry [56].

2.2.3

EIS technique
In the EIS technique, a small-amplitude AC stimulus with known frequency content is

applied to a sensor interface via a potentiostat. Current is then recorded for signal conditioning to
extract a phase or amplitude shift, algebraically related to sensor interface’s impedance (real and
imaginary components). When applying an analyte to the sensor, the impedance of the sensor
changes respectively. This technique has been implemented in CMOS chips for applications such
as protein sensors [30] and DNA sensors [57].
Regarding the extraction method, fast Fourier transform (FFT) algorithm and frequency
response analyzer (FRA) based methods are typically used [58]. The FFT algorithm utilizes a
broadband stimulus (e.g. a pulse) and computes the results at all frequency points simultaneously,
producing an impedance spectrum. The composite signal source and the computationally intensive
FFT require a digital signal processor (DSP) with extensive computational resources [59].

19

Although a few studies have been conducted to implement FFT with analog circuits [60], it is
impractical for microsystem implementations due to poor multiplier linearity and accuracy, and
demands for preprocessing/pre-storing of the input data set. Alternatively, in FRA, a small
amplitude (10mV) sinusoid voltage stimulus is applied between the two biosensor electrodes, and
the current response is measured as the frequency of the stimulus voltage is swept. The FRA
method can be realized with compact analog circuits [28-30].
Recently many FRA-based instrumentation circuits have been developed. The lock-in
technique, which is a two phase reference coherent demodulation method, has been employed to
extract the value of the magnitude and phase of impedance [28]. A circuit combined lock-in
multiplier with Σ-Δ modulation reports ultra-low power consumption and high resolution [61]. A
dual-slope ADC with lock-in function as well as an on-chip sinusoid signal generator was reported
in [62]. To address the impediments of solution resistance, a differential FRA approach was
applied in [63].
2.2.4

Summary of electrochemical IC circuit design
The literature review reveals that electrochemical IC circuit design is tailored to these specific

applications. To maximize the sensor’s performance and meet system requirements, various kinds
of measurement techniques, stimulus schemes, and readout topologies have been developed
respectively. Therefore, customized electrochemical interface IC circuits should be optimized
according to the gas sensor microsystem’s compact, low power, and low cost requirement.

20

2.3 Integration approaches for gas sensor microsystem
To achieve highly miniaturized gas sensor microsystem, the gas sensor and sensor interface
circuit should be integrated in a cost-effective and highly-compact way. Therefore, it is necessary
to analyze existing integration approaches and select a path for the gas sensor microsystem
integration, tailoring to electrochemical sensing technology.
2.3.1

Comparison between hybrid and monolithic approach
Reviewing the past 20 years of academic and industrial efforts toward gas detector

miniaturization, two main development pathways have emerged for the integration of gas sensor
elements and sensor interface circuits. They are referred to as hybrid integration and monolithic
integration [64].
In the case of hybrid integration, the sensor element and readout circuitry are formed by two
separate elements that are assembled into a single module. Typically this involves packaged
components combined on a printed circuit board (PCB). This approach permits sensor and
circuitry to be fabricated using specific processes, materials, and thermal budgets that are tailored
to the individual needs of each element. However, this approach has larger parasitic
capacitance/resistance due to interconnection between two elements. It also has the problem of
higher packaging cost and larger size than monolithic integration.
In the case of monolithic integration, the sensor element and readout circuitry are formed
within a single chip. Compared to hybrid integration, monolithic integration is superior in the

21

following perspectives. It is of smaller size. It has lower noise by eliminating interference coupled
through wire bonds. It is also of higher sensitivity by minimizing parasitic capacitance/resistance,
and of lower cost by reducing chip packaging. In conclusion, following the trend of
miniaturization in electronic products, monolithic integration is a promising approach for wearable
gas sensor microsystems.
2.3.2

Analysis of monolithic gas sensor implementation approach
The monolithic integration approach has been applied to several sensor technologies

including micro cantilever sensors, thermoelectric sensors, semiconductor sensors, optical sensors,
chemiresistor sensors and chemical field effect transistor (ChemFET) sensors. Although no
monolithic electrochemical gas sensor has been reported, it is valuable to understand all
monolithic gas sensors’ working principles, fabrication process, pros, cons, etc.

2.3.2.1

Micro cantilever sensors

Micro cantilever sensors are a type of gravimetric sensors which respond to the mass
accumulated in a sensing layer. In the case of gas sensing, the sensing layer must be capable of
selectively absorbing a target gas molecule. Micro cantilevers are fabricated on silicon substrate
by using micro/nano-electromechanical systems (MEMS/NEMS) technology and then coated by a
selective gas sensing polymer. The total mass of the cantilever changes when the sensing polymer
absorbs gas molecules, causing alteration of resonant frequency [65, 66].

22

The cantilever resonance is actuated by a heater through thermal-mechanical transduction,
and is sensed by piezoresistor through mechanical-electrical transduction. A feedback-loop circuit
controls the hotplate’s periodically fluctuating temperature to drive the cantilever vibration at its
resonant frequency. Bulk micromachining is adopted to implement cantilevers by using wet
etchants to isotropic etch the silicon substrate, leaving a p-doped (p+) region as cantilever. Coated
with a specific gas-sensitive polymer, the micro cantilever is capable of detecting VOCs such as
toluene, octane and ethanol. However, this type of cantilever gas sensor has two drawbacks. First,
it has limited detection resolution. Second, power consumption is high. Due to the hotplate, power
consumption can reach up to 10mW.

2.3.2.2

Thermoelectric sensors

Thermoelectric sensors convert the temperature difference on junctions of two conductive
materials to a voltage difference. After coating a gas sensitive polymer with conductive materials,
the enthalpy change produced by ambient gas absorption or desorption can induce temperature
change, thus create voltage change. In this way, the gas concentration is transduced to an electrical
signal.
Such sensors could be implemented by the CMOS process, utilizing polysilicon and
aluminum as thermopiles. To fully implement these sensors, several extra components should be
included. First, a heater is needed to generate a temperature difference, which can be easily
realized by CMOS process. Second, a thermally isolated structure—such as a membrane and a

23

bridge—is required to enhance the temperature gradient between hot and cold junctions in a
limited area and protect the circuit, which operates at room temperature. This requires an extra
post-CMOS back-side etch. Third, a gas sensitive polymer needs to be coated on the thermopile to
enable gas sensing. A monolithic thermoelectric sensor integrated with a pre-amplifier was
reported to detect different VOCs [67]. An improved sensor structure with an N-well island
membrane and a more accurate calibration was presented in a follow up paper [68]. However,
this improvement does not solve the problem of extra power consumption introduced by the heater.
Similar to micro cantilevers, extra power consumption remains one of the major drawbacks of
thermoelectric sensors.

2.3.2.3

Semiconductor sensors

Semiconductor sensors utilize metal oxide and noble metal additives as catalytic materials to
detect combustible gases [69]. The combustible gases can be oxidized below the ignition
temperature when catalytic materials are presented. The enthalpy change during the reaction
induces a change in temperature and eventually in the resistance of the catalytic materials.
Similar to the thermoelectric sensors, the implementation of monolithic semiconductor
sensors also requires a heater, a thermally isolated island, sensing material coating, and readout
circuit, either by CMOS or post CMOS process fabrication. Furthermore, to achieve a controllable
high temperature (usually >400 °C), a hotplate with homogeneous temperature distribution, a
temperature sensor and temperature control circuits are required. Such a monolithic sensor was

24

first reported in 2004 [2]. Later, an improved monolithic metal oxide catalytic sensor with
post-CMOS fabricated temperature sensor that can operate at 500 °C was reported [23]. This
compact array demonstrated improved analyte discrimination with a detection limit of 1 ppm CO
and 100 ppm CH4. However, those monolithic semiconductor sensors suffer similar disadvantages
as thermoelectric sensors, given the similarly high working temperature and complicated post
CMOS process.

2.3.2.4

Optical sensors

Compared to other sensing methods, optical techniques provide great selectivity by reading
different properties of the electromagnetic waves such as frequency, phase, amplitude, and state of
polarization. Photodiodes as optical sensors are readily fabricated by modern CMOS technique.
However, since optical stimulus is difficult to integrate into monolithic chips, there is very little
research on monolithic optical gas sensors. A technique utilizing bioluminescence was developed
to implement a monolithic gas sensor called bioluminescent bioreporter integrated circuits (BBIC)
[70-72]. This technique employs engineered bioluminescent bacteria, which can metabolize target
compounds such as toluene and convert the energy directly into light.
A conceptual BBIC microsystem is composed of porous light-tight closure, bioluminescent
bacteria in a microenvironment, and IC with photodiodes on a chip. The IC chip detects the
intensity of the luminescent light signal proportional to the target concentration. 10 ppb detection
limit of toluene was reported. However, the disadvantages of the BBIC limit its application. First,

25

the target analyte is limited by the type of bacteria. Second, the BBIC is not applicable to real-time
monitoring due to long response time ranging from minutes to hours. Third, the sensor needs a
cultivated microenvironment, and the bacteria amount varies over time, causing signal drift.

2.3.2.5

Chemiresistor sensors

Chemiresistor sensors rely on changes in the conductivity of a polymeric or organic sensitive
film upon interaction with a gas analyte (absorption and desorption). Their operation temperature
ranges from room temperature to 60 °C.
Depending on the target analytes, the sensing materials vary from conductive polymers [73],
carbon black polymers [74-76], thiolate-monolayer-protected gold nanoparticles (MPN) [77] to
single wall carbon nanotubes (SWCNT) [78]. Because most sensing materials work at room
temperature, island structure formed by back-side etch are not necessary and only on-chip surface
modifications are involved. Generally, those materials are dissolved / dispensed into volatile
organic solutions and sprayed onto the sensing area of the chips. On-chip electrodes were
fabricated either by CMOS process or by post-CMOS metal electroplating, sputtering and thermal
deposition. Special post-CMOS treatments includes annealing [74], dielectrophoretic assembly,
parylene-C encapsulation [78].

26

2.3.2.6

ChemFET sensors

ChemFET sensors function similarly to MOSFET. The FET’s source-drain current changes
with channel work function change, which is directly related to ambient gas concentration. Three
types of ChemFET have been developed with CMOS-compactable process. The first type is called
adsorption-porous-silicon-FET(APSFET) where the transistor channel is directly modified to a
gas sensitive layer [79, 80].The second type is called Lundstrom-FET where the sensitive layer is
deposited on top of the transistor channel [81]. The third type is called suspended-gate-FET
(SGFET) where the gas sensitive layer is suspended at a certain distance above the channel [81].
All ChemFET fabrication processes include standard CMOS and post-CMOS processes.
Regarding post-CMOS process, APSFET requires a BHF etching to create a porous silicon region
for sensing; Lundstrom-FET requires palladium gate and SGFET need extra electrode deposition
and suspended gate attaching.
The gases sensed by ChemFET in air monitoring applications include NO2 and H2. The gas
variety is mainly constrained by the sensing material. To date, limited metal (palladium in
Lundstrom-FET and Pt in SGFET) could be deposited on silicon substrate and limited materials
(boron in APSFET) could be doped in silicon substrate.

2.3.2.7

Summary of monolithic gas sensor technologies

The monolithic gas sensors discussed above are summarized in Table 2.2. The summary is
organized in terms of sensor technology, sensing material, detectable gases, and disadvantages.

27

Regarding performance, each type of sensor can only detect limited types of gases confined by the
compactable sensing material it adopts. Besides the popular sensing materials like metal oxide and
conductive polymer, others have limited potential to extend the monolithic platform to detect
multiple gases. In addition, all sensors inherit disadvantages from the sensing materials they
adopted, such as slow response to gases, sensor baseline drift, and extra power needed to heat up
the sensor to the desired operation temperate. Lastly, our literature review leads us to conclude
that no monolithic integration of electrochemical gas sensors has been reported.
Table 2.2.
Summary of monolithic gas sensors
Sensing technology

Detectable Cons
gases

Ref

Micro cantilever Polymer (photoresist)

VOC

Power consuming

[65, 66]

Thermoelectric

Polymer

VOC

Power consuming

[67, 68]

Semiconductor

Metal oxide

Power consuming

[2, 23, 69]

Optical

Bacteria

VOC

Not reliable,
response

Chemiresistor

Conductive polymer, etc.

VOC

Sensor drift

[73-78]

ChemFET

Porous silicon/
metal oxide

NO2, H2

Slow response

[79-81]

Sensor type

2.3.3

Sensing material

CO, CH4

slow [70-72]

Summary of integration approach for electrochemical gas sensor microsystem
The comparison between hybrid integration and monolithic integration indicates that the

monolithic approach leads to a cost-effective and highly-miniaturized microsystem. The

28

monolithic gas sensor literature review reveals to date no electrochemical gas sensor microsystem
has been implemented by monolithic integration yet. Considering the dual benefits that the
electrochemical sensing technique and monolithic approach bring, it is promising to develop an
electrochemical gas sensor microsystem by monolithic integration.

29

3 Development of RTIL-based Sensor for Microsystem Platform

3.1 Overview
RTIL-based electrochemical sensing technology was identified and chosen to develop sensor
devices for wearable air pollutant monitoring microsystems due to its comprehensive strengths
that are superior to other sensing technologies. However, traditional electrochemical sensor
structure suffers slow response to gases. In this chapter, a new type of RTIL-based gas sensor
structure is introduced to achieve rapid response time. Then, approaches for sensor miniaturization
and sensor array packaging are reported. After that, the development of a flexible version of the
RTIL-based gas sensor is presented. The achievements pave a path to realize a wearable gas sensor
microsystem for air pollutant monitoring.

3.2 Development of fast-response RTILs-based electrochemical sensor
3.2.1

Sensor’s structure design for fast response
RTIL-based electrochemical sensing technology inherits the outstanding performance of

liquid-electrolyte electrochemical gas sensors, but also suffers slow response. The main obstacle
to response time in existing RTIL-based electrochemical gas sensors is the slow diffusion of target
gases from the RTIL surface to the electrodes. As illustrated in Figure 3.1(a), in the conventional
sensor structure, the electrodes are fabricated on a substrate and RTIL is then coated on top of the

30

electrodes. Because the electrochemical reaction happens at the interface of the RTIL and the
electrodes, a response cannot be measured until the gas analyte dissolves into the RTIL and reaches
the electrodes. A full-scale response to gas analyte is achieved only after an equilibrium forms on
the interface. The time interval between the gas reaching the electrode surface and establishing
equilibrium is primarily dependent upon gas diffusivity in the RTIL. Due to RTIL’s high viscosity,
RTIL-based electrochemical gas sensors suffer slow response caused by the low gas diffusivity in
the RTIL.
Different sensor structures have been developed to solve the issue of slow response. On the
one hand, as response time is reciprocal to square of the thickness of electrolyte [82], it can be
significantly reduced by thinning the RTIL layer [38, 45]. However, formation of thin RTIL layer
is unstable and irreproducible [39]. By wicking RTIL in a permeable membrane, a controllable
uniform electrolyte layer can be produced [35] (Figure 3.1(b)). However, the thickness of RTIL is
limited by the thickness of membrane. On the other hand, the response time can be shortened by
eliminating the gas diffusion process in RTIL. By placing electrodes between gaseous analyte and
RTIL as shown in Figure 3.1(c) [37, 44], the response time can be significantly reduced. The
drawback of this structure is that a very precise volume control is required to maximize the
sensitivity.
In

contrast

to

the

traditional

structure,

we

have

designed

a

planar-electrode-on-permeable-membrane (PEoPM) structure that bypasses the slow diffusion of
gas across the RTIL. As illustrated in Figure 3.1(d), the electrodes are fabricated directly on a

31

Figure 3.1. (a) Conventional sensor structure: response time is slow due to slow gas diffusion
through RTIL; (b) RTIL-in-permeable-membrane structure: response time is improved by a
controllable thin electrolyte layer; (c) Porous-electrodes-on-RTIL structure: response time is
improved by bypassing gas diffusion in RTIL. (d) Electrodes-on-permeable-membrane
structure: response time is improved due to fast gas diffusion in the permeable membrane.

gas-permeable membrane, allowing gas to reach the electrodes/RTIL interface through the
permeable membrane, where diffusion is much faster than in the RTIL. If the porous membrane
has hydrophobic response to RTILs while porous electrode has hydrophilic response to RTILs,
RTIL can fill the porous electrodes but cannot soak into the porous membrane. In such case, the
effective sensing area is only determined by the surface geometry of the porous electrodes rather
than RTIL volume. Therefore, this solution can solve the volume control issue in Figure 3.1(c).

32

Choosing inert material to form the porous membrane, the sensor structure can be used to detect
toxic gases. To simplify the sensor structure complexity and lower the fabrication cost, WE, CE
and RE should be all fabricated on the permeable membrane.

3.2.2

Fabrication of PEoPM sensors
Proper materials were selected for the PEoPM structure. POREX® porous PTFE with 35%

porosity and 4μm pore size (Zitex TM, Chemplast, Incorporated, Wayne, New Jersey) was chosen
as the permeable membrane due to its excellent inertness and hydrophobic response to RTIL. Gold
was chosen as the electrode material because it is highly inert and can readily be deposited in thin
films and patterned in a planar process. High-purity RTIL 1-butyl-1-methylpyrrolidinium
bis(trifluoromethylsulfonyl)imide ([C4mpy][NTf2]) was chosen as the electrolyte due to its low
viscosity and high chemical stability. It was prepared and refined to gravimetric purities exceeding
99.9% using reported method [37, 44].
To verify the hydrophobic/hydrophilic property of blank porous PTFE membrane and
gold-coated porous PTFE membrane, contact angles of an RTIL droplet on both membranes were
measured using Video Control Angle System 2000. As shown in Figure 3.2, the contact angles are
95˚ and 18˚, respectively. The results demonstrate that the porous PTFE membrane has a
hydrophobic response to RTILs while porous gold electrode has a hydrophilic response to RTILs.
Therefore, the selected materials are capable of forming the proposed sensor structure.

33

Figure 3.2 (a) Image of RTIL [C4mpy][NTf2] droplet on porous PTFE. The contact angle is 95˚.
(b) Image of RTIL [C4mpy][NTf2] droplet on gold-coated porous PTFE. The contact angle is
18˚.

To realize the PEoPM structure, a fabrication process was developed. A gold film was
deposited on a porous PTFE sheet using physical vapor deposition (PVD, Edward 360 thermal
evaporator). Thicker films help to ensure continuity of the thin film electrodes applied to a porous
substrate. Thinner films reduce blockage of PTFE pores, thus improving diffusion of gas to the
WE/IL interface. This tradeoff was experimentally studied, and a gold film thickness of 300nm
was chosen balance consistency and performance. Metal deposited on porous PTFE results in a
very uneven surface that complicates patterning by traditional photoresist processes. To solve this
problem, a stainless steel hard mask was prepared by electric discharge machining. The mask was
tightly mounted against the PTFE substrate on the PVD sample stage, and after deposition it was
removed to leave the desired electrode pattern. Finally, to add the RTIL layer, the electrodes were
then coated with 200 µm-thick [C4mpy][NTf2] with a droplet process. The fabrication process is
illustrated in Figure 3.3.

34

Figure 3.3. PEoPM sensor fabrication process. A stainless steel mask was mounted against the
porous PTFE and then a gold film was deposited. After removing the mask, patterned planar
gold electrodes were formed. Finally, RTIL was coated on the electrodes to form the sensor.

Because RTIL is a highly resistive electrolyte (the conductivity of [C4mpy][NTf2] is only
0.22 S/m at room temperature [83] compared to 62 S/m for 50% H2SO4 [84]), the sensor
electrodes were designed to minimize the electrolyte resistance and maximize the response current
by interdigitating the WE and CE electrodes. An RE was included alongside the WE. The
electrode structure shown in Figure 3.4 occupies 0.75” × 0.5” including large pads to facilitate
2

connections to instrumentation. The effective WE area is 10 mm . The finger width of and the gap
between interdigitated CE and WE are both 200 µm. An O-ring was used to separate the sensing
electrode area from the connection pad area during testing.

35

Figure 3.4. Photograph of fabricated PEoPM device. The WE and CE are interdigitated with
finger widths of and the gaps of 200 µm. The O-ring isolates the sensor area from the contact
pads.

3.2.3

Test of PEoEM sensors
The fabricated sensor was placed in a gas chamber and sealed with an O-ring, as illustrated in

Figure 3.5. To compare the response time of the different sensor structures shown in Figure 3.1(a)
and Figure 3.1(d), both the upper chamber and lower chamber have a gas inlet and outlet for
testing with gas flow. Path 1 over the top of the sensor corresponds to a conventional sensor
structure while path 2 at bottom of the sensor corresponds to the PEoPM sensor structure. The total
gas flow was maintained at 200 ± 0.05 sccm (standard cubic centimeters per minute) by digital
mass-flow controllers (MKS Instruments Inc).
EIS testing was carried out with a VersaStat MC potentiostat (Princeton Applied Research,
Oak ridge, TN, U.S.A.). During EIS measurement, a sinusoidal AC signal with a 10mV
peak-to-peak amplitude was applied. The DC bias was set to the electrochemical potential that

36

Figure 3.5. Cross-sectional view of the gas testing chamber. Separate inlets and outlets on the top
and bottom of the chamber allow gas to be introduced along two separate paths to effectively
characterize two different sensor structures, where path1 corresponds to a conventional sensor
structure shown in Figure 3.1(a) and path 2 corresponds to the PEoPM sensor structure shown in
Figure 3.1(d).

generated the largest response for the chosen RTIL and the target gas. For the example pollutant
gas SO2 reported below, the DC bias was set to -0.5V.
To characterize the sensing response for the PEoPM device, SO2 was selected as an example
pollutant. The sensor was exposed to SO2 and the sensor’s impedance spectrum was recorded. The
AC stimulus signal was swept from 0.1 Hz to 10 kHz, and SO2 concentration was varied from 0 to
25 ppm. The recorded amplitude of the complex impedance |Z| is plotted in Figure 3.6., where each
curve corresponds to the impedance spectrum under a fixed-concentration SO2. The RTIL sensor
shows an obvious impedance amplitude change to different SO2 concentration, with a stronger
response below 100 Hz.

37

To further illustrate the relation between the sensor impedance and SO2 concentration, the
imaginary portion of the complex impedance Zim at 10 Hz were extracted to form an SO2
calibration curve. Figure 3.7 plots -Zim verses SO2 concentration. The experimental data was fit to
an equation given by (see explanation from EIS model of RTIL-based gas sensor in Appendix A)

 Z im  0.377 

1.239CSO2
C SO2  7.588

(3.1)

Figure 3.6. Impedance amplitude |Z| spectrum from 0.1 Hz to 10 kHz recorded from an
RTIL-based PEoPM sensor presenting under a SO2 environment. SO2 concentration varies
from 0 to 25ppm. Obvious impedance amplitude change to different SO2 concentration was
observed, with a stronger response below 100Hz.

38

-Zim(kohm)

1.5

1

0.5

0

0

5

10
15
SO2 (ppm)

20

25

Figure 3.7. SO2 calibration curve for RTIL-based PEoPM sensor. Blue dots represent the
recorded -Zim values at 10Hz. Red curve represents the fitting curve follows Equation 3.1. The
2

fitting curve achieves high least-squares correlation coefficients (R =0.9957).

where CSO2 is the concentration of SO2 in ppm units. This fitting curve matches test data very well,
2

with least-squares correlation coefficients R =0.9957. Although the calibration curve is nonlinear,
the sensitivity at any given concentration can be obtained by calculating the slope of the fitting
curve (3.1) at the desired concentration.
To compare the response times of the two different sensor structures shown in Figure 3.1(a)
and Figure 3.1(d), 25 ppm SO2 was purged through path 1 and path 2 (Figure 3.5.) separately, and
the transient response of impedance amplitude was recorded. Defining the baseline as 0% and the
steady state as 100%, the data were normalized, and the transient response of the RTIL sensor is
plot in Figure 3.8. For the conventional structure (corresponding to path 1), the RTIL sensor

39

reaches 90% of full-scale response (T90) in 56 s. However, using the PEoPM sensor structure
(corresponding to path 2) the response time is reduced to only T90 = 20 s. The PEoPM sensor
structure thus is 180% faster than the conventional structure in the case of SO2 sensing.
These preliminary results demonstrate that an RTIL sensor on a PEoPM electrode structure
using EIS methods provides a rapid and sensitive response to SO2. Experiments using other gas
analytes such as CH4 show that similar rapid responses can be achieved. This new sensor structure
thus appears well suited to measure multiple gases in a real-time air pollution monitoring system.

Figure 3.8. Transient response of RTIL-PEoPM sensor to 25 ppm SO2, with gas flow from two
different paths (see Figure 3.5.) representing a conventional sensor structure with 56 s full scale
response and a new rapid-response structure that reduces response time to 20 s.

40

3.3 Miniaturization approach of PEoPM sensor arrays
Preliminary results demonstrate that an RTIL-based sensor on a PEoPM structure provides a
rapid and highly sensitive response to example air pollutants. To utilize the PEoPM structure in a
wearable air pollutant monitoring microsystem, miniaturization of the sensor arrays is required. In
this section, a miniaturization approach of the PEoPM sensor array is presented.
3.3.1

Fabrication of miniaturized PEoPM sensors
The process described in the previous section successfully resulted a macro sensor device.

However, the process is not applicable for miniaturized sensor fabrication. The stainless steel hard
mask used in the macro sensor device fabrication was prepared by electric discharge machining
only provides a resolution of around 200 µm. Thus the WE-CE gap could not be reduced below
200 µm, which required the electrodes to occupy a large area to keep the gap resistance reasonable.
To overcome this significant limitation, a new fabrication process utilizing photolithography was
developed.
Microfabrication processes have been well developed and widely used in the semiconductor
industry. Using photolithography followed by thermal metal deposition and liftoff process,
electrode patterns on smooth substrates can easily be formed. Traditional photolithography has a
resolution on the order of 1 µm and should be suitable for fabricating miniaturized gas sensors.
However, directly applying traditional photolithography on porous PTFE substrates, as desired

41

here, introduces several problems that require development of a new photolithography process for
porous substrates.
First, note that photoresist is generally spin-coated onto a substrate that is vacuum-attached to
the spinner. However, porous PTFE is soft and cannot be reliably held by vacuum because it is
porous. A preliminary experiment demonstrated that, when a PTFE sheet was put directly on the
spinner, significant photoresist would be sucked into the PTFE and could not be removed during
liftoff. To resolve this problem, the porous PTFE was clamped to a glass substrate, and the glass
was placed on the spinner and held by vacuum.
Second, because porous PTFE has a rough surface, standard photoresist coatings of PTFE are
not reliably smooth enough for patterning. For example, with porous PTFE having a 4 μm-pore
size, around 4 μm of surface roughness can be expected. To resolve this problem, a thick film
photoresist was selected. Hoechst AZ4620 was spin-coated at 2100 rpm to create 10 μm-thick
layer. Lithography by 60 s UV exposure and 300 s developing in AZ300 MIF developer was found
to reliably remove the thick resist. The profile scan by a Dektak3 Surface Profiler in Figure 3.9
demonstrates that the photoresist AZ4620 covered the rough surfaces of porous PTFE and that a
good edge step, suitable for liftoff, was formed.
Third, note that preliminary experiments found metal liftoff to be difficult, even after a
gold-deposited sample was soaked in acetone for several days. One possible reason is that the
photoresist became polymerized during the half hour PVD, where the temperature on the sample is
over 100°C. Because polymerized photoresist cannot easily be dissolved in acetone, this would

42

Figure 3.9. Profile scan by Dektak3 Surface Profiler. The photoresist AZ4620 covered the
cause the liftoff to fail. To solve this problem, note that exposure to UV should prevent photoresist
from polymerizing. Thus, a 60 s flood exposure was performed after developing the photoresist
and before metal deposition.
Finally, because of the uneven distribution of pores on PTFE surface, the conductivity of
metal traces were found to be unreliable and in the worst case could result in entire electrode
fingers being disconnected, resulting in significant device-to-device variation. To resolve this
problem, one possible solution is to increase the width of the electrodes thereby reducing the
chance of a discontinuity. However, this approach conflicts with the goal to miniaturize the
device. To analyze the relation between device variation and electrode width, 4 groups of
10mm-long gold traces with different widths were fabricated. Each group had eight identical
elements, and resistance values were measured to compare their average sheet resistance and

43

standard deviation as shown in Table 3.1. As expected, the results show that traces with larger
widths have better consistency. 200 µm width was chosen as a compromise between reliability and
the miniaturization goal.
By addressing each of the challenges described above, the RTIL-based gas sensor fabrication
process was optimized to achieve a miniaturized PEoPM structure followed by the process
addressed as shown in Figure 3.10.
The porous PTFE sheet was clamped on glass wafer and spin-coated by AZ4620 photoresist.
5 min soft bake at 95°C was performed, followed by a 60 s UV exposure through the electrode
pattern mask. After 300 s developing in AZ300 MIF developer, the sample was flood exposed for
60 s. Then a 300 nm-thick gold film was deposited on the porous PTFE sheet using physical vapor
deposition (PVD, Edward 360 thermal evaporator). After deposition, the sample was soaked in
acetone overnight and the gold electrodes pattern was formed by liftoff. Finally, to add the RTIL

Table 3.1.
Average sheet resistance and standard deviation with different width configurations.

standard
deviation
(%)

0.288

standard
deviation
(Ω)
0.044

200

0.323

0.041

12.7

400

0.327

0.034

10.3

800

0.306

0.023

7.5

width
(µm)

average
(Ω)

100

44

15.2

Figure 3.10. Miniaturized RTIL-based PEoPM sensor fabrication process.

interface, the electrodes were then coated with 200 µm-thick [C4mpy][NTf2] with a droplet
process.

45

Following the microfabrication process above, the miniaturized PEoPM device shown in
Figure 3.11 was fabricated. The electrode structure occupies a 2 mm × 2 mm sensing area, only 8%
of the area in the macro-scale device as shown in Figure 3.4. The WE and the CE were
interdigitated for impedance measurement with a 200 µm width and a 100 µm gap.
3.3.2

Package of miniaturized PEoPM sensor arrays
A miniaturized PEoPM device fabrication process was describe in previous section. To

integrate the sensor within the wearable air pollutant monitoring system, the microfabricated
PEoPM sensors should be packaged in an array form and electrically connected with sensor
interface circuit.
An individual RTIL-based PEoPM sensor element is a standard electrochemical cell that
consists of three electrodes (WE, CE, and RE), and electrolyte. The three electrodes are on the
PEoPM sheet and RTIL is the electrolyte. To form PEoPM sensor arrays, there are two schemes:

Figure 3.11. Photograph of fabricated miniaturized PEoPM device. The WE and CE are
interdigitated with finger gaps of 100µm. The O-ring isolates the RTIL area from the contact
pads.

46

electrodes array that all sensor elements share the same RTIL or electrochemical cell array that
each group of three-electrodes pairs with separate RTIL. The former scheme would be to directly
pattern multiple electrodes on one piece of porous PTFE sheet with coating a continuing RTIL
layer. However, this approach fails the sensors’ function according to the discussion below. The
gas sensing selectivity is introduced by electrochemical biasing potential. Each sensor tailored to a
specific gas target is biased at a specific potential. If the sensor elements in the array share the same
electrolyte environment, the different electrical bias conditions would introduce significant
interference among sensors. Therefore, electrochemical cell array scheme was chosen to develop
PEoPM sensor arrays.
To properly separate the electrochemical cells, O-rings were used to confine the RTIL in
individual sensor element region. To avoid RTIL leaking from the edge between O-rings and
porous PTFE, O-rings and PEoPM sheet were tightly clamped between two rigid boards. Both the
up-board (lid) and lower-board (substrate) were designed containing a hole for gas diffusing to
porous PTFE and for introducing RTIL to the electrochemical cell, respectively. In our design,
Viton O-rings were chosen for its good seal and chemical resist performance; printing circuit board
(PCB) was chosen to construct the lid and the substrate because it is mechanically rigid and easy
for wire routing. The PEoPM sensor array were assembled as illustrated in Figure 3.12(a). First,
four electrodes-patterned PEoPM sheets were placed on wire-routed PCB substrate, with pads
connected to wire contacts on PCB by conductive epoxy. Then O-rings were attached on a
windows-opened PCB lib and mounted on the wire-routed PCB substrate by tighten bolts, to create

47

RTIL reservoirs. Finally, [C4mpy][NTf2] were dropped in the reservoirs to create 200µm-thick
RTIL sensing layer. Photos of the PEoPM sensor array are shown in Figure 3.12 (b). The sensor
array occupies 1.55” x 1.25” including the connector to be electrically connected with sensor
interface circuit. This array contains only four sensors because only four gas analytes were
available for preliminary test. Nevertheless, using the package process described above, arrays
contain eight sensor elements or more can be readily implemented.
3.3.3

Test of miniaturized PEoPM sensor arrays
To demonstrate the functionality of the microfabricated PEoPM sensor array, the packaged

Figure 3.12. (a) Illustration of the parts assembly for PEoPM sensor array. First, four
electrodes-patterned PEoPM sheets were placed on wire-routed PCB substrate, with pads
connected to wire contacts on PCB by conductive epoxy. Then O-rings were attached on a
windows-opened PCB lib and mounted on the wire-routed PCB substrate by tighten bolts, to
create RTIL reservoirs. Finally, [C4mpy][NTf2] were dropped in the reservoirs to create RTIL
sensing layer. (b) Photographs of assembled PEoPM sensor array from front, back and side
views. The sensor array occupies 1.55” x 1.25” including the connector to be electrically
connected with sensor interface circuit.

48

sensor array was placed in a dessicant chamber and gas analytes were purged with gas flow
maintained at 200 ± 0.05 sccm. CH4, SO2, NO2, and O2 were selected as example pollutants. CH4
concentration was varied from 0 to 5%; SO2 concentration was varied from 0 to 25ppm; NO2
concentration was varied from 0 to 35ppm; O2 concentration was varied from 10% to 20%.
EIS testing was carried out with VersaStat MC potentiostat. The DC bias was set to the
electrochemical potential that generated the largest response for the chosen RTIL and the target gas.
For the example gas CH4, SO2, NO2, and O2, the DC bias of each channel was set at -0.3 V, -0.5 V,
1.2 V and -1.2 V to acquire the maximum response to target each gas analyte, respectively. The
impedance spectra were recorded.
Depending on the gas analyte being measured, the sensor’s impedance will change either due
to a physical absorption of the analyte [85] or by chemical interaction with the analyte [46].
Physical absorption mechanisms will result in a change of the complex impedance (Z = Zre+jZim)
at low frequency that will be most evident within the imaginary portion (Zim) and have little
impact on the real portion (Zre). Chemical interaction mechanisms will results in a change of
recorded current at low frequency that will be most evident within the amplitude of impedance
(|Z|).
For CH4 and SO2 sensing, given their physical absorption mechanisms, Zim at 10 Hz was
extracted to plot calibration curves. For NO2 and O2 sensing, given their chemical reaction
mechanisms, |Z| at 10 Hz was extracted to plot calibration curves. Figure 3.13 plots all the four gas
analytes’ calibration curves for the RTIL sensor array. The red lines represent the fitting curves

49

with equation forms that are derived from the theory (see details in Appendix A). The impedance
–Zim or |Z| of each gas sensor are expressed as follows

 Z im ,CH  26.89 
4

7.815CCH

4

CCH  1.622
4

2

(R =0.994)

(3.2)

 Z im, SO  22.54 
2

10.95CSO

2

CSO  6.28
2

2

(R =0.986)

(3.3)

| Z |NO  2.05  101  [1.49  103  C NO  (4.72  102 +7.85  104  C NO ) j ]1
2

2

2

2

(R =0.967)

(3.4)

| Z |O  2.05  10 1  [8.51  10 3  CO  (2.92  10 2 +6.12  10 4  CO ) j ]1
2

2

2

(R =0.999)

2

(3.5)

50

35

32

-Zim(kohm)

-Zim(kohm)

34

30
28
26

0

2
4
CH4 (%)

0

10
20
SO2 (ppm)

30

10
O2(%)

20

40

20

|Z|(kohm)

|Z|(kohm)

25
20

6

25

15
10

30

0

10

20
30
NO2(ppm)

30
20
10
0

40

0

5

15

Figure 3.13. CH4, SO2, NO2, O2 calibration curves for RTIL sensors from EIS test data at 10Hz.

where CCH4 is the concentration of CH4 in % units, CSO2 is the concentration of SO2 in ppm units,
CNO2 is the concentration of NO2 in ppm units, and CO2 is the concentration of O2 in % units.
2

All least-squares correlation coefficients R are higher than 0.96, demonstrating the test
results have good agreement with the theory. The results also demonstrate good sensitivity over the
range of interest for the gas targets.
To demonstrate the selectivity of the miniaturized sensor array, 20.4% O2, 3% CH4, 10 ppm
SO2, and 10 ppm NO2 were purged into the test chamber sequentially in four steps and the sensor

51

array’s impedance responses were recorded in each step. Taking the impedance response to O2 as
100% signal S, the sensor array responses to all four gases are normalized as ΔS/S (%) and plotted
in Figure 3.14. The appended gases have negligible interference to other gas sensors. This
miniaturized sensor array is thus well suited to measure multiple gases in a wearable real-time gas
monitoring microsystem.

Figure 3.14. Selectivity experiment. 20.4% O2, 3% CH4, 10 ppm SO2, and 10 ppm NO2 were
purged into the test chamber sequentially in four steps and the sensor array’s impedance
responses were recorded in each step. Taking the impedance response to O2 as 100% signal S,
the sensor array responses to all four gases are normalized as ΔS/S (%) and plotted.

52

3.3.4

Integration of PEoPM sensor array within a wearable monitoring system

The miniaturized sensor array was successfully installed in a prototype of a wearable air
pollutant monitoring system as shown in Figure 3.15. It consists of a packaged miniaturized
sensor array, multi-electrochemical instrumentation board (MEIB), and a µ-Controller (µC)
board. The size of the system is 2.6” × 2” × 1”. Details of the wearable system is described in
[86].Development of the microsystem is still ongoing.

3.4 Development of flexible PEoPM sensors
Recently, research on wearable sensors that features flexibility has drawn a lot of attention
because physical flexibility of sensors could open a lot of new applications such as artificial skin,
conformal monitoring, etc. Likewise, by featuring flexibility on gas sensors, the wearable air

Figure 3.15. 2.6” × 2” × 1” prototype wearable air pollutant monitoring system with
miniaturized sensor array, multi-electrochemical instrumentation board (MEIB), and a
µ-Controller (µC) board.

53

pollutants monitoring microsystem could even be integrated into clothing. The previous section
introduced a PEoPM structure that provides a fast response. However, the sensor’s package is
mechanically stiff. The following section discusses how to design planar PEoPM sensors with
mechanical flexibility.
3.4.1

Fabrication of flexible PEoPM sensors
A PEoPM sensor array is mechanically stiff because two rigid boards are required to clamp

the O-ring and PEoPM sheet to create reservoirs for RTILs. Therefore, flexible PEoPM sensor is
possible only when the O-ring reservoir can be removed in the sensor design. However, without
constrained by O-ring reservoir, thick RTIL (200µm in PEoPM sensor array) can easily flow out
with small mechanical disturbance. Only thin RTIL layer can be formed on PEoPM sheet. Because
RTIL has high viscosity, it would typically stick to the porous surface regardless mechanical
disturbance. However, a thin layer cannot be formed by directly dropping pure RTIL on the
PEoPM sheet. That’s because porous PTFE is hydrophobic to RTIL (Figure 3.2(a)), while gold on
porous PTFE is hydrophilic to RTIL (Figure 3.2(b)). The surface energy difference for different
substrates results in RTIL holding on individual electrodes as “islands” when applying a small
volume (0.2 μL in the experiment) of pure RTIL on a PEoPM sheet, as shown in Figure 3.16 Case
1. If the RTIL islands become disconnected, the electrolyte fails to form an electrochemical cell

54

Figure 3.16. Coating experiments illustration. Case1 illustrates small volume pure RTIL coating
(Vol.=0.2 μL) result. RTIL droplets were coated only electrodes surface. Case2 illustrates large
volume pure RTIL coating (vol.=1 μL) result. RTIL droplets were coated on electrode surfaces
as well as the gap between electrodes. Case3 illustrates ethanol/RTIL mixture coating
(Rethanol/RTIL=2, RTIL’s vol.=0.2 μL) result. The initial mixture droplet was coated the
electrodes area as well as the gap between electrodes. After ethanol removal, RTIL still
occupied the initial droplet region. Case4 illustrates ethanol/RTIL mixture coating
(Rethanol/RTIL=4, RTIL’s vol.=0.2 μL) result. The initial mixture droplet filled the pores inside
the porous PTFE as well as covered electrodes surface. After ethanol removal, RTIL still stayed
in the porous PTFE.

over the electrodes. Although islands could merge into a bulky droplet when applying a large
volume (1μL in the experiment) of pure RTIL, as shown in Figure 3.16 Case 2, this large droplet
would easily flow away with small mechanical disturbances, ruining the sensor.

55

One approach is to mix RTIL with solvents that have low surface tension on porous PTFE.
Ethanol was chosen for low surface tension on both porous PTFE and gold. It can be mixed with
RTIL in any ratio. In addition, ethanol can easily be removed through evaporation. By fixing the
RTIL volume to 0.2 μL, droplets with different ratios of ethanol to RTIL (denoted by Rethanol/RTIL)
were dropped on the PEoPM surface, and the PEoPM sheet was then vacuumed for two hours to
remove ethanol. These experiments are illustrated in Figure 3.16 Case 3 and Case 4. In Case 3, the
initial mixture droplet coated not only the electrode area but also the gap between electrodes in the
same electrochemical cell. After ethanol removal, RTIL still occupied the initial droplet region,
and a conformal thin layer of RTIL was formed on both PTFE and electrodes surface. AC
2

impedance measurement between electrodes shows a resistance on the order of 10 Ω, implying
good conductivity of electrolyte within the electrochemical cell. In Case 4 with higher ethanol
concentration, the mixture’s surface tension was too small and the droplet filled the pores inside
the porous PTFE as well as covering electrode surface. After ethanol removal, RTIL still remained
within the porous PTFE. AC impedance measurement between electrodes shows an open circuit.
6

Even if the volume of the droplet was increased, a large resistance (on the order of 10 Ω) was
measured for this ethanol/RTIL ratio due to uneven pore distribution and limited pore connection
inside the porous PTFE. Following this procedure, optimized mixing ratio (Rethanol/RTIL=2) and
droplet volume (VRTIL=0.2 μL) were experimental obtained to form a thin and conformal layer on
the PEoPM surface that achieved a good, conductive electrochemical cell.

56

After the microfabrication (Figure 3.10) and droplet processes above, the PEoPM
RTIL-based sensor shown in Figure 3.17(a) was produced. Concentric ring electrode geometry
was adopted for amperometric-type planar sensors. A disk-shaped WE with radius of 1 mm was
patterned in the center of porous PTFE sheet. An annular CE with inner radius of 1.125 mm and
outer radius of 1.75 mm and an RE was patterned around the WE. The gap between WE and CE is
2

125µm. The total sensing region occupies 9.6 mm . A 21 µm-thick [C4mpy][NTf2] layer was
coated on the sensing region, covering all electrodes and the gaps between them. The resulting
sensor structure can be bent either convex or concave as shown in Figure 3.17, demonstrating its
physical flexibility.

Figure 3.17. (a) Photograph of microfabricated flexible PEoPM RTIL-based sensor. (b)
Photograph of convex bend of the PEoPM RTIL-based sensor. (c) Photograph of concave bend
of the PEoPM RTIL-based sensor.

57

3.4.2

Test of flexible PEoPM sensors
Chemical experiments were carried out using the test setup as shown in Figure 3.15. O2 was

selected as an example analyte. O2 and N2 (as background gas) concentrations were controlled
using digital mass-flow controllers (MKS Instruments, Inc.) to vary O2 concentration from 0% to
21%. The total flow rate was fixed to 100 sccm to circumvent the influence of differential gas flow
rate. The PEoPM RTIL-based sensor was placed in a flow cell within a desiccant gas chamber
parallel to the gas flow as shown in Figure 3.18. The gas flows across both side of the sensor and
reaches the electrodes both by permeating through porous PTFE and by dissolving into RTIL.
Based on previously reported experiments [87], the response time is determined by the faster path
permeating through the porous PTFE. Through wires that were attached to the sensor pads with
conductive epoxy, the sensor was connected to a VersaStatMC potentiostat for electrochemical
test.

Figure 3.18. Gas test setup and position of the sensor within the flow cell.

58

To verify the functionality of the PEoPM RTIL-based sensor, cyclic voltammetry (CV) was
performed between 0 - -1.8 V with 500 mV/s scan rate for the reduction of O2 at the WE, as shown
in Figure 3.19 where the O2 concentration ranges from 0% to 21%. A broad current peak was
observed at about -1.4V vs Au, which is in agreement with the CVs obtained for O2 reduction in
RTIL reported in [88]. The relationship between peak current Ip(µA) and O2 concentration CO2(%)
was found to fit the following linear calibration condition

Figure 3.19. Cyclic voltammograms (scan rate = 500 mV/s) for the reduction of O2 at WE. O2
concentration varies from 0-21%. O2 reduction current peak was observed at -1.4V vs Au.
The relationship between peak current and O2 concentration from 0% to 21% is linear.

59

2

Ip=-1.25 CO2 -3.35 (R =0.989)

(3.6)

This relationship demonstrates the functionality and linearity of the PEoPM RTIL-based
sensor.
For real-time monitoring, the constant-potential amperometry method is usually used due to
its simple hardware implementation. To characterize the performance of the PEoPM RTIL-based
sensor using constant-potential amperometry, the sensor was biased at the O2 reduction potential
of -1.4 V and the current was measured. The sensor’s sensitivity, linearity, repeatability and limit
of detection (LOD) were characterized as follows.
The sensor’s response linearity was characterized by a step change of O2 concentration from
0% to 21% and then back to 0%. The response current vs. time curve shown in Figure 3.20 was
recorded. The stable values at each step were collected to form the O2 calibration curve shown in
Figure 3.21. The relationship between amperometric current Ia (µA) and O2 concentration CO2 (%)
can be expressed by
2

Ia=-0.48 CO2 -0.64 (R =0.997)

60

(3.7)

Figure 3.20. Current vs. time curve at various O2 concentrations when the potential is held at
–1.4 V vs Au. N2 is the background gas. O2 concentration steps up from 0% to 21% and steps
down from 21% to 0%.

Figure 3.21. O2 calibration curve for RTIL sensor using data extracted from Figure 3.19.

61

2

The sensitivity of the amperometric response was measured as 0.48μA/%. The R value of 0.997
implies a good linearity of the sensor response to O2.
Repeatability test was performed by alternatingly purging 1% O2 and pure N2. The resulting
amperometric current is shown in Figure 3.22. Although the sensor shows a slight drift due to
decay of the double layer capacitor charging current, by subtracting the baseline value the response
current ΔI shows a good repeatability with an average value of 0.73 µA and a standard deviation of
only 1.5 nA. The corresponding LOD for O2, defined as three times the baseline noise or 0.038 µA,
was found to be 0.08 vol %.

Figure 3.22. Constant-potential current response measured over three cycles of alternate exposure
to 1% O2 and pure N2 flow at an applied potential of -1.4 V.

62

3.5 Conclusion
This chapter introduced the implementation of robust RTIL-based electrochemical gas
sensors featuring a planar-electrode-on-permeable-membrane structure that provides flexibility,
high sensitivity and fast response. Fabrication processes have been developed for PEoPM sensor,
miniaturized PEoPM sensor array, and flexible PEoPM sensor, respectively. Using SO2, CH4,
NO2, and O2 as example analytes, the sensors’ functionality and performance were
demonstrated.

63

4 Development of a Compact Low-Power Amperometric
Instrumentation with Current-to-Digital Readout

4.1 Overview
To develop wearable air pollutant monitoring microsystem, the sensor interface electronic
circuits need optimized to minimize its power and complexity. For electrochemical sensors,
amperometric technique is widely used. Miniaturization and optimization of amperometric
instrumentation is essential for wearable applications such as gas sensor microsystems. The
amperometric instrumentation consists of two parts: a potentiostat and a current readout circuit.
The potentiostat provides current required for the reaction while maintaining the
electrode/electrolyte interface at the correct potential. The current readout circuit conditions and
digitizes the reaction current. Existing research focuses on optimizing individual parts (either
potentiostat or readout circuit) given power/size/resolution requirements [47-49, 52, 56]. In
contrast, an opportunity exists to optimize instrumentation circuit topology in terms of complexity
from the perspective of circuit system level, or even the perspective of sensor system level. This
chapter introduces a novel amperometric instrumentation circuit topology optimization from the
perspective of sensor system level, resulting in a compact amperometric instrumentation circuit
with current-to-digital readout.

64

4.2 Background
4.2.1

Electrochemical sensors and equivalent circuit models
Electrochemical sensors in amperometric mode work under the following sensing principle:

the reaction current is proportional to the analyte concentration when reacted electrode/electrolyte
interface is biased at a constant voltage. To accurately control the reaction taking place at the
interface, a three-electrode cell configuration has been applied to amperometric electrochemical
sensors. In such three-electrode cell, the reaction takes place at the interface between the WE and
electrolyte. A constant potential is maintained between the RE and the WE. The CE provides a
current path to the WE.
To analyze the electrochemical sensor’s electrical response, equivalent circuit models have
been proposed in EIS theory. Randles circuit model [46] shown in Figure 4.1 (a) is a classic
equivalent circuit model widely used to describe a three-electrode sensor. The impedance between
RE and WE consists of an uncompensated solution resistor Rs (relatively small), in series with the
parallel combination of the double layer capacitor Cdl at the WE interface (charging current ic
follows through this path), and an impedance of a Faradaic reaction caused by AC stimulus (AC
Faradaic current if follows through this path). The Faradaic reaction consists of a charge transfer
resistor Rct and Warburg element Zw

ZW 

65

AW
j

(4.1)

Figure 4.1. Equivalent circuit model of electrochemical sensor cell. (a) Randles model (b)
Complete model considering both AC and DC stimulus. (c) Simplified model for circuit
analysis.

where Aw is Warburg coefficient and ω is the angular frequency. Since our only interest is at the
WE interface, the impedance between the CE and the RE is denominated as simple impedance Z.
Notice that this model only represents sensor’s response to small AC stimulus. To represent both
AC and DC response, a complete equivalent circuit model is shown in Figure 4.1(b) [89, 90]. A

66

current source is added to represent DC Faradaic current If. Here, If is the constant reaction current
proportional to the analyte concentration in amperometric electrochemical sensors, which is the
main interest in sensor current measurements. In general, if << If, and Rs is relative small. They
can be considered as second-order effects in sensors response. For analysis simplicity, Rs and if are
initially omitted in the following circuit topology derivation, and the negligible impact of this
omission will be discussed in Section 4.4. The simplified model is shown in Figure 4.1(c).
4.2.2

Traditional amperometric instrumentation
An amperometric instrumentation circuit consists of two parts: a potentiostat and a current

readout circuit. The potentiostat provides current from the CE to the WE while maintaining the
voltage between RE and WE. A typical potentiostat can be implemented by a single opamp with
appropriate connections [89, 90]: the positive input node is connected with bias for RE VRE, the
negative input node is connected with RE, and the output is connected with CE to provide current.
The current readout circuit collects If either at WE or CE, then conditions and digitizes it. Two
topologies are used to implement the current readout circuit: a current-to-voltage convertor
followed by a voltage-mode analog-to-digital convertor (ADC) [49, 91-93], or a single
current-mode ADC [52, 94]. Given its compact structure, the current-mode ADC is usually chosen
for readout circuit design.

67

Given the requirement of sensor applications for low power and low complexity, a traditional
amperometric instrumentation circuit utilizes the single opamp for potentiostat design and the
current-mode ADC [94] for current readout circuit design. The schematic is shown in Figure 4.2.
The circuit working principle is described as follows. Two reference current sources Iref with
opposite directions alternatively connect with the integrator through switches, which are
controlled by the digital output of the hysteresis comparator Dn. Thus the input current of the
integrator Iint is

I int  I f  (1) Dn  I ref

(4.2)

As the waveforms in Figure 4.3 illustrate, the integrator’s capacitor is charged/discharged
according to the direction of Iint. Consequently, the output of the integrator Vint rises/falls
correspondingly to Iint direction. While Vint reaches the hysteresis comparator upper/lower bound

Figure 4.2. Schematic of a traditional amperometric instrumentation circuit including potentiostat
and current-mode ADC.

68

Figure 4.3. Waveforms of the current on the integrator input Iint, the voltage on the integrator
output Vint, and the digital output of the comparator Dn.

(Vref+/-ΔV/2) (where ΔV is the hysteresis window width and Vref is the reference voltage), Dn
flips, changing Iint following (4.2). The square waveform at the output of the hysteresis
comparator is then digitized by a counter with the reference clock at a much higher frequency. The
time interval of the digital “high” for Dn is

T1 

Cint  V
I ref  I f

and the time interval of the digital “low” for Dn is

69

(4.3)

Cint  V
I ref  I f

T0 

(4.4)

From (4.3) and (4.4), If can be expressed as a function of Iref, T1, T0

If 

T0  T1
I ref
T0  T1

(4.5)

Defining the duty cycle α as



T1
T1  T0

(4.6)

Substituting T1 and T0 by α in (4.5), If can be expressed as a function of α and Iref

I f  (1  2 )  I ref

(4.7)

Therefore, given a known Iref, If is obtained by measuring duty cycle of Dn. Notice that If is
independent of both the integrator capacitor Cint and the hysteresis comparator parameters (ΔV
and Vref).

4.3 Compact amperometric instrumentation design
The previous section introduced the electrochemical sensor equivalent circuit model and the
typical amperometric instrumentation circuit. Substituting the sensor symbol in Figure 4.2 by the
simplified sensor equivalent circuit model (Figure 4.1(c)), the schematic of the electrochemical

70

Figure 4.4. Schematic of the electrochemical sensor system. It consists of the traditional
amperometric instrumentation circuit and the simplified electrochemical sensor equivalent
circuit model.

sensor system is given in Figure 4.4. Notice that the sensor operates at the steady state when no
current flows through Cdl, only If is collected in the readout circuit. From systematic point of view,
the sensor system contains two capacitors: Cdl and Cint. Cint is part of the readout circuit and used
for charging/discharging; Cdl is the inherent interface capacitor, but doesn’t play a functional role
in the system. Since capacitors usually occupy large area in a single IC chip, if Cdl can play the role
of Cint, Cint can be eliminated to save area. How to modify the circuit to incorporate Cdl into the
amperometric instrumentation? An answer is given in this section by deriving a compact
amperometric instrumentation topology from the traditional structure.
Part of the sensor model and the current-readout circuit are taken as starting point for the
derivation. Ignoring sensor’s bias requirement at this step, a current source If is used to substitute
the sensor model. As shown in Figure 4.5, given node B is a low-impedance node, folding the
current source to the output of the integrator is equivalent to the typical topology of the current

71

Figure 4.5. Derivation of the instrumentation topology. The input current source is folded to
parallel connect with integrator capacitor.

readout circuit. (4.7) holds given a satisfied sensor’s bias condition. Notice that the parallel
connection of If and Cint is the same as the equivalent circuit between RE and WE as shown in
Figure 4.1(c) and Cint can be arbitrary, (4.7) is still valid when Cint is substituted by Cdl.
To satisfy sensor’s bias condition, a potentiostat function is incorporated into the
current-mode ADC by the following modification steps. First, by flipping the direction of If and
substituting Vref with VWE and VWE with VRE, the voltage between the RE and the WE can be
held by feedback loops of the integrator (loop1) and of the ADC (loop2). Although the WE
potential is not strictly held constant due to a nonzero value of ΔV in the loop2, the perturbation on
the WE does not affect the sensor’s steady state as long as ΔV is set small enough (less than 10 mV)
[95]. In addition, because current can only flow from the CE to the WE, node A should be
connected to the CE rather than the RE.

72

Following the modification steps described above, a modified amperometric instrumentation
circuit with the sensor model is illustrated in Figure 4.6. Because the direction of the current source
If is opposite from the direction of If in Figure 4.5, If in Figure 4.6 is written as

I f  (2  1)  I ref

(4.8)

This topology successfully realizes the functions of both current-mode ADC and potentiostat.
Compared to a traditional topology in Figure 4.2, it utilizes Cdl for integrator, and eliminates one
opamp required by the potentiostat and Cint required by the integrator.
Additional optimization is conducted as follows. Notice that voltages at the RE and the WE
are both held by the feedback loops and no constrains are required for VRE and VWE from the
circuit perspective. Therefore nodes RE and WE are interchangeable. By swapping WE with RE,
an optimized structure is shown in Figure 4.7. This topology is equivalent to the topology in Figure
4.6. Thus If follows (4.8). Notice that the WE is connected with a unit-gain buffer. This buffer can

Figure 4.6. Schematic of the modified amperometric instrumentation circuit with sensor
equivalent circuit model.

73

be discarded for further optimization. By connecting the WE with ground and substituting VRE
with VRE-WE, a compact current-to-digital amperometric instrumentation (CCDAI) topology is
given in Figure 4.8 as equivalent to the optimized structure in Figure 4.7. Here, we assume that
sensor’s bias requires VRE>VWE, concluding VRE-WE>0. If sensor’s bias requires VRE<VWE,
alternatively, the WE can be connected to the power supply.
Following the derivation from the schematic in Figure 4.4 to the one in Figure 4.8, the
CCDAI topology was designed functionally equivalent to the traditional amperometric

Figure 4.7. Schematic of the optimized amperometric instrumentation circuit with sensor
equivalent circuit model.

Figure 4.8. Schematic of CCDAI with the simplified sensor equivalent circuit model.

74

instrumentation, when the parameters of the hysteresis comparator in the CCDAI meet the
following constraints: Vref is set to VRE-WE, and ΔV is set to 10mV.

4.4 Performance analysis
Although the function of the CCDAI is equivalent to the traditional amperometric
instrumentation, structure differences and additional constraints will cause performance
differences. In addition, the equivalent circuit used to derive the circuit topology was the
simplified model in Figure 4.1 (c). The second-order effects of the sensor equivalent circuit should
be fully considered in terms of performance. This section evaluates the performance of the CCDAI
in two aspects: performance difference from the traditional amperometric instrumentation and
performance effects of second-order elements in equivalent circuit model.
4.4.1

Performance difference from traditional instrumentation
Compared to the traditional potentiostat, the CCDAI drives the electrochemical cell by a

constant current source rather than an opamp. Therefore, it would take longer time to stabilize the
electrochemical cell potential. Nevertheless, differences in the potential stabilization time would
not affect steady state operation of the electrochemical cell.
Compared to the traditional current mode ADC, the main differences of the CCDAI include:
(1) the integrator’s capacitor Cint is replaced by sensor’s double layer capacitor Cdl; (2) the
hysteresis comparator voltage window is limited to 10mV. These two differences affect the
resolution of the calculated If. From (4.7), If is obtained by calculating measured α with a known

75

Iref value. From (4.6), the resolution of α is determined by how short T0 and T1 are given a fixed
counter reference clock frequency. From (4.3) and (4.4), T0 and T1 are proportional to Cdl and ΔV.
Therefore, Cdl and ΔV do affect the resolution of If. Assuming |If|<Imax, the max time interval is

Tmax 

Cdl  V
(4.9)

I ref  I max

For a fixed counter reference clock frequency f0, the maximum relative quantization error |δq|max
[94] is

q

max

I ref  I max
1


f 0  Tmax f 0  Cdl  V

(4.10)

The ADC’s effective resolution, N, in units of number of bits, is determined by [94]

N   log 2  q

max

 log 2 ( f 0  Cdl  V )  log 2 ( I ref  I max ) (4.11)

Therefore, larger ΔV and Cdl would improve the effective resolution N. In the traditional
current-mode ADC, ΔV can be up to the power supply voltage, Vdd, which can be 5 V in a
portable device. In the CCDAI, ΔV is restricted to maximum 10 mV. ΔV in the CCDAI is 500
times smaller than in the traditional current-mode ADC, resulting in 9 bits of effective resolution
loss for the CCDAI. However, in the meantime, electrochemical double layer capacitor Cdl has
much larger capacitance density than a capacitor that can be fabricated by CMOS process in a
2

single IC chip. For instance, double layer formed on 1 mm electrode can generate μF level
capacitance; while a capacitor in a single IC chip is up to ten pF. The 10000 times larger
capacitance in the CCDAI would result in 13 bits of effective resolution improvement for the

76

CCDAI. Therefore, the total effect of Cdl and ΔV provides an improvement of around 4 bits of the
effective resolution. As a tradeoff, the sampling rate drops as the effective resolution increases.
Fortunately, electrochemical systems typically have a slow response and do not need a fast
sampling rate.
4.4.2

Second-order effects of the sensor equivalent circuit model
The derivation in Section 4.2 was based on the simplified equivalent model in Figure 4.1(c).

Given a complete model in Figure 4.1(b), an evaluation is needed to determine whether the
solution resistor Rs and AC Faradaic components in the complete equivalent circuit model would
introduce significant errors.
First, Rs is added in the simplified model and the corresponding waveform of Vint is
illustrated as shown in Figure 4.9. From the observation, the abrupt jump caused by Rs and does
not change T1 and T0. Thus (4.8) is still valid. However, the abrupt jump decreases effective
charging/discharging window from ΔV to ΔV - Iref·Rs. Within a standard electrochemical cell

Figure 4.9. Vint waveform illustration when considering Rs in the equivalent circuit model.

77

2

configuration, the RE is placed close to the WE. Thus Rs value is in the order of 10 ~ 10 Ω
experimentally. With μA level of Iref, this only gives 10 ~ 100μV error, which is less than 1% of 10
mV. Therefore, Rs has negligible impact on the resolution.
Next, AC Faradaic components are evaluated. The AC Faradaic components are parallel
connected with double layer capacitor Cdl and DC Faradaic current source If. Because Warburg
element blocks DC current same as Cdl, only AC current if can pass through those AC Faradaic
components. The sensor current Isens is the sum of the DC current If and the AC current ic + if.
Observe that the sensor current Isens should be equal to the current provided by the current source
at any time,

I sens (t )  I f  ic  i f ,ac

 I ref during T1


 I ref during T0


(4.12)

where T1 is the time interval when Dn = 1 in the CCDAI, T0 is the time interval when Dn = 0 in the
CCDAI. Here T1 and T0 do not follow (4.3) and (4.4). The waveform of Isens is illustrated in
Figure 4.10. Given the waveform in the time domain, Isens can also be expressed by Fourier series


sin(k )
cos(k  2 f c  t ) I ref
k
k 1

I sens  (2  1) I ref  4

(4.13)

where fc = 1/(T1+T0). The first term in (4.13) represents DC part of Isens, and the second term in
(4.13) represents AC part of Isens. Because AC components (Cdl and Warburg elements) block DC
current, and DC current source blocks AC current, If is equal to the DC part of Isens.Thus If is

I f  (2  1)  I ref

78

(4.14)

Figure 4.10. Illustration of Isens in time domain.

(4.14) is identical to (4.8), which leads to the conclusion that the readout value of the CCDAI is the
same as the result obtained in Section 4.2, even when considering the complete sensor equivalent
circuit model.

4.5 Results
4.5.1

CCDAI implementation
To verify the functionality and performance of the CCDAI, a test setup shown in Figure 4.11

was built. The CCDAI was implemented on a printed circuit board using commercial IC chips. The
circuit power supply was set to 5 V and current bias Iref was set to 1 μA, which are suitable values
for a portable application. To implement a hysteresis comparator with upper and lower bounds that
can be adjusted independently during testing, the circuit shown in Fig. 12 was implemented using
two comparators, an AND gate and an OR gate. A USB-6259 data acquisition card (National
Instrumentations Inc.) was used to set the voltage on the reference electrode, VRE-WE, and the
comparator’s upper/lower bound voltages, Vh and Vl. It was also used to measure the time
intervals T1 and T0 of comparator output Dn using an internal 10MHz clock. A Labview user
79

interface was built for communication between a PC and the data acquisition card. The current If
was calculated using (8) with the measured T1 and T0 values. The results were stored in files.

Figure 4.11. Test setup for electrical and chemical experiment.

Figure 4.12. A hysteresis comparator realization with adjustable upper/lower bound.

80

4.5.2

Experimental results
To evaluate the ADC performance of the CCDAI, an electrical test model was connected with

the CCDAI board. To implement the simplified model in Figure 4.1(c), the electrical test model
consisted of a 1 μF capacitor and a Keithley 6430 Source Meter connected in parallel. CE and RE
were shorted in the test. The current readout accuracy of the CCDAI was tested by sweeping If
from -800 nA to 800 nA with 2 nA step. Differential non-linearity (DNL) and integral non-linearity
(INL) of the readout current are plotted in Figure 4.13. The worst DNL equals to -56dB and the

DNL(dB)

-54
-56
-58
-60
-0.8

-0.4
0
0.4
Input current(A)

0.8

-0.4
0
0.4
Input current(A)

0.8

INL(dB)

0
-50
-100
-150
-0.8

Figure 4.13. DNL and INL of the CCDAI. Both DNL and INL in the current range are better
than -49dB, implying an 8 bit of effective resolution.

81

worst INL equals to -49dB, meaning that CCDAI achieves better than 8 bits of effective resolution,
Such resolution is sufficient for most sensor applications.
The Faradaic current generated by potassium ferricyanide was recorded both by the CCDAI
and CHI760C as a function of time when VWE-RE was constantly held to 190 mV. As an example,
data in the case of concentration of 6 mM was plotted in Figure 4.14. Both currents recorded by
CCDAI and CHI760C converged to the same level with negligible differences after the chemical
system reached the steady state. The transit pattern differences are caused by the different stimulus
provided by the two instrumentations. CHI760C applies large current to set the initial VWE-RE to
the desired voltage in the very short time; while CCDAI applies a constant current to raise
VWE-RE to the desired voltage in a gentle way. In addition, initial current recorded by CHI760C
includes charging current caused by step stimulus, while current recorded by the CCDAI does not
contain the charging current. Due to unavoidable convection in the solution [96], the currents at the
steady state fluctuate in small amplitude. This phenomenon was observed from the data recorded
by both instrumentations. Results obtained by the CCDAI at different potassium ferricyanide
concentrations are plotted in Figure 4.15. The average current values from 200 s to 300 s, which
were recorded by CCDAI and CHI760C, were taken to plot the calibration curve as shown in
2

Figure 4.16, respectively. The least-squares correlation coefficients (R ) of the fitting curve are
0.991 and 0.996 for the data acquired by CCDAI and CHI760C respectively. The chemical
experiment results demonstrate the functionality and the accuracy of the CCDAI.

82

1.5

Current(uA)

CHI760C
CCDAI
1

0.5

0

0

100

200
Time(s)

300

400

Figure 4.14. The Faradaic current generated by 6 mM of potassium ferricyanide as function of
time when VWE-RE=190mV. Red line represents data recorded by CHI760C and blue line
represents data recorded by CCDAI.

0.8
0mM
1mM
2mM
4mM
6mM

Current(A)

0.6

0.4
0.2

0

0

50

100

150
200
Time(s)

250

300

Figure 4.15. The Faradaic current recorded by the CCDAI at VWE-RE=190mV as function of
time for 0- 6 mM of potassium ferricyanide.

83

0.80
Current(μA)

CHI760C
0.60

CCDAI

0.40
0.20
0.00
0

2
4
Concentration(mM)

6

Figure 4.16. Calibration curve of Faradaic current vs potassium ferricyanide concentration. The
current values were the average values from 200s to 300s. Fitting curve was presented as a
2

straight line. R values of the fitting line are 0.991 and 0.996 for the data acquired by CCDAI
and CHI760C, respectively.

4.5.3

Performance comparison
The CCDAI realize a compact topology while functions as the same as the traditional

amperometric instrumentation circuit. Comparison between Figure 4.4 and Figure 4.8 shows that
the CCDAI eliminates two opamps and one integrator capacitor compared to the traditional one. In
microelectronic circuits, both of these components usually occupy larger area than comparators
and current sources. In addition, opamps are the main power source in ICs. To provide a qualitative
and fair comparison, Table 4.1 and Table 4.2 list each component’s area and consumption in ON
0.5µm CMOS fabrication process based on our previous IC chip design [97]. The area and power
of the CCDAI and the traditional amperometric instrumentation are calculated as a summary of the

84

Table 4.1.
Area occupation of IC modules in ON 0.5µm CMOS fabrication process and the comparison
between the traditional amperometric instrumentation circuit and the CCDAI.

Opamp
Comparator
Current source pair
(with switch)
8-bit counter
@100kHz
Capacitor (1pF)
Total area (µm2)

Area (µm2) Traditional CCDAI
1200
2
0
1000
1
1
600

1

1

12000
2200

1
1

1
0

18200

13600

Table 4.2.
Power consumption of IC modules in ON 0.5µm CMOS fabrication process and the
comparison between the traditional amperometric instrumentation circuit and the CCDAI.

Opamp
Comparator
Current source
pair (with switch)
8-bit counter
@100kHz
Capacitor (1pF)

Power@5V (µW) Traditional CCDAI
7.5
2
0
5
1
1
0.5

1

11
N/A

Total power(µW)

1
1
1

1
0
16.5

31.5

85

required components’ area and power consumption. As results, the CCDAI reduces area by 25.3%
and power consumption by 47.6% compared to the traditional amperometric instrumentation
circuit. Notice that the 8-bit counter occupies the major area because the feature width of the
process is 0.5µm. With advanced process, the area reduction by the CCDAI would be more
significant.

4.6 Conclusion
A novel compact amperometric instrumentation design with current-to-digital readout for
electrochemical sensor was presented. This design utilizes the EIS concept and the inherent
interface capacitor to reduce the circuit complexity. Compared to a traditional amperometric
instrumentation structure, this design dramatically saves area, cost and power. A board-level
CCDAI was implemented and tested, demonstrating an 8-bit of effective resolution in the range of
-800 nA to 800 nA. Functionality of the instrumentation was verified by potassium ferricyanide
electrochemical experiment. High linearity of current-concentration relation was acquired with
2

0.991 of R . T CMOS implementation of the CCDAI could save 25.3% of area, and 47.6% of
power compared to a traditional amperometric instrumentation structure. This design suits for
wearable electrochemical sensor device applications.

86

5 Development of a CMOS Monolithic Electrochemical Gas Sensor
Microsystem

5.1 Overview
Chapter 3 introduced a miniaturized RTIL-based electrochemical sensor array for hybrid
integration of a wearable air pollutant monitoring device. To implement an ultra-low cost and
highly-miniaturized microsystem, monolithic integration needs to be explored utilizing
RTIL-based electrochemical sensing technology. This chapter presents a development of the first
CMOS monolithic electrochemical gas sensor microsystem. The monolithic gas sensor integrates
an RTIL-based electrochemical sensor and CMOS sensor interface circuit in a single chip.

5.2 System architecture
5.2.1

Electrochemical sensor system
The two techniques most commonly used to acquire qualitative information in

electrochemical sensors are amperometry and impedance spectroscopy. The amperometry
technique is adopted for the electrochemical sensor system design. An electrochemical sensor
system based on amperometry consists of two parts: an electrochemical sensor and an
amperometric sensor interface circuit as shown in Figure 5.1. The sensor is the interface that
converts gas information (species and concentrations) to electrical signal (current in amperometric

87

Figure 5.1. Block diagram of the electrochemical sensor system.

mode). The sensor interface circuit provides the essential sensor’s electrical conditions and records
the current.
An electrochemical gas sensor is composed of electrodes and an electrolyte. The sensing
principle is described as follows [13]. A specific reduction/oxidation (redox) reaction involving a
gas analyte that has dissolved within the electrolyte takes place at the electrode/electrolyte
interface and thus generates a redox current when the electrode/electrolyte interface is biased at a
specific voltage. The redox current is proportional to the gas analyte concentration in the
electrolyte. To realize a sensor using this principle, the three-electrode configuration is typically
used. The WE is where the reaction takes place. The CE provides an electrical current path to the
WE. To eliminate the potential errors, an RE is added to maintain a constant reaction potential at
the WE.
A sensor interface circuit consists of potentiostat and readout circuit, connecting sensors
through electrodes. The potentiostat provides the required voltage and current for a three-electrode
configuration, and the amperometric readout circuit is used to measure the redox current.

88

5.2.2

CMOS monolithic sensor microsystem concept
As introduced in Chapter 2, monolithic approach for sensor systems can significantly lower

production cost and power consumption, minimize the system size, and improve the detection
limit. To explore a monolithic electrochemical sensor system, the physical implementation of
electrochemical sensors and interface circuits are reviewed and the structure compatibility is
evaluated.
RTIL-based electrochemical sensors can be implemented in a variety of structures. Because
RTIL doe not evaporate, it is not necessary to use traditional electrochemical structures that need
containers and gas permeable membranes to seal the volatile electrolytes. Microfabrication
technology enables planar electrochemical cell structures that only need three layers: a substrate
for physical support, planar electrodes and RTILs as a supporting electrolyte. Thus, an RTIL-based
electrochemical sensor can be implemented by two simple steps [45]: planar electrodes are
patterned on a chemical-inert, non-conductive substrate (such as silicon nitride); consequently,
RTILs are coated on the electrodes to form the sensor. Although porous PTFE is ideal substrate for
fast response, it is not easy to combine it with CMOS.
In modern electrical instrumentation, the sensor interface circuit can be implemented as a
CMOS IC chip using a standard CMOS process in a commercial foundry. Such IC chip is a rigid
silicon die with a passivation layer on top. The passivation layer is made of silicon nitride layer to
protect circuits from any chemical corrosion.

89

Consider implementing the monolithic microsystem by stacking an IC chip and a planar
RTIL-base sensor together, structure compatibility is discussed as follows. Through overlapping,
the top passivation layer of the IC chip can be the substrate of the RTIL-based electrochemical
sensor. The passivation layer will not only provide physical support as a substrate for an
RTIL-based electrochemical sensor, but also protect the circuit from any potential electrochemical
corrosion introduced by the electrochemical reaction on the WE. In addition, due to silicon
nitride’s hydrophilic response to RTILs, it is possible to form a thin layer of RTIL on chip’s surface,
enabling the sensor a fast response to gas analytes. Because of the high surface tension,
small-volume RTIL droplets could firmly stick to the chip surface regardless of mechanical
disturbance. The CMOS process also forms on-chip contacts, enabling a direct connection
between circuits and electrochemical sensors if electrodes are on the surface of the CMOS chip.
Based on the discussion of the sensors and circuits’ structure compatibility, a CMOS
monolithic electrochemical microsystem concept utilizing RTIL is proposed in Figure 5.2. By
stacking electrochemical sensor layer and sensor interface circuit layer, and using IC chip’s surface
passivation layer as sensor’s substrate, a monolithic microsystem is achieved. To realize such
RTIL-based sensor on CMOS IC chip, the fabrication process only involves electrodes pattern and
RTIL coating. The post-CMOS process is much simpler compared to other types of monolithic gas
sensor, providing a competitive low-cost solution for monolithic sensor.

90

Figure 5.2. Conceptual illustration of the CMOS monolithic electrochemical gas sensor
microsystem.

5.3 CMOS design
To realize the monolithic microsystem concept, the first step is to implement a CMOS sensor
interface circuit IC chip. As shown in Figure 5.1, the sensor interface circuit consists of a
potentiostat and an amperometric readout circuit. This section describes the CMOS circuit design
of each block.

5.3.1

Potentiostat
As shown in Figure 5.3, a typical potentiostat topology was adopted using only one

operational amplifier (opamp) to save power consumption [89, 90]. The WE is connected to the
negative input of the opamp and the CE is connected to the output of the opamp. Input voltage
Vref1 is applied to the positive input of the opamp. Through the feedback loop from the CE to the
RE, the voltage of the RE VRE can be precisely equal to Vref1. A rail-to-rail opamp (R2rop) [98]

91

Figure 5.3. Schematic of the monolithic microsystem, consisting of the amperometric
instrumentation circuit and the RTIL sensor. The green rectangle block illustrates the clock
waveform.

was implemented in the opamp design to maximize the potential window for electrochemical
sensor, in particular, in low-power application using battery power source. The design
specification of the power supply is 5 V and R2rop is expected to drive the RE and the CE in the
range of 0.1V to 4.9 V.
5.3.2

Amperometric readout circuit
To maximize the sensor sensitivity, the low-noise amperometric readout circuit topology

shown in Figure 5.3 was adopted from our previous work [93]. Since capacitive feedback readout
achieves lower noise compared to resistive feedback and current conveyer [99], an integrator is

92

used to convert the current to voltage. In the integrator, a bias Vref2 is applied on the positive input
of OP1 to hold the voltage of the WE equal to Vref2. Thus the voltage between the WE and the RE
can be held constantly equal to V ref2-V ref1 during sensor operation. The output voltage of the
integrator is sent to a programmable gain amplifier (PGA) with a gain of CC1/CC2. To output a
stable voltage, sample and hold (S/H) circuit is used afterward.
To reduce 1/f noise and amplifier DC offset, correlated double sampling (CDS) technique
[100] was implemented. As illustrated in Figure 5.3, CDS requires several switches and
non-overlapping switch clock signals φ1, φ2, and φ3 that were produced by an on-chip clock
generator block (not shown). The output of amperometric readout is given by

VO 

IWE
C
 C1
f s  Cint CC 2

(5.1)

where IWE is the current collected at WE, fs is the frequency of the switch clocks, Cint is the
capacitor value of the integrator.

5.4 Monolithic sensor fabrication
As described in Section 5.2.2, monolithic sensor fabrication involves standard CMOS IC
fabrication process and post-CMOS sensor fabrication process. The post-CMOS fabrication
includes electrodes patterning and RTIL coating.
Two concerns need be considered in the circuit layout placement and routing. First, to connect
circuits directly with on-chip electrodes, open glass contacts were designed in the layout. In

93

addition, considering the reliability of electrode pattern and surface routing, top metal layer were
avoid to use for circuit routing to create a flat chip surface. With those considerations, the sensor
interface circuit described in Section 5.3 was fabricated in ON 0.5 µm CMOS process and the
fabricated die is 2.2 mm × 2.2 mm.
Because the cut die was received from the foundry, the electrodes pattern process was
followed by die-level processing [101]. The die was glued at radius of 30 mm at a 3” wafer.
Shipley 1813 photoresist was then dispensed on the die and the supporting wafer was spun at 3000
◦

rpm followed by a 10 min soft bake at 95 C in the oven. 10 s of UV exposure through the electrode
pattern mask was performed followed by a 45 s of development by AZ 352 developer. Then
5nm-thick titanium film and 100nm-thick gold film was deposited on the die using physical vapor
deposition by Edward 360 thermal evaporator. After deposition, the photoresist was rinsed off by
acetone, methanol and DI water, thus the gold electrodes pattern was formed by liftoff.
After electrodes patterned, the CMOS chip was wire bonded to a DIP-40 package. A droplet
of RTIL (around 16nL) was casted on the electrodes by a syringe tip. Here, high-purity
1-butyl-1-methylpyrrolidinium bis(trifluoro-methylsulfonyl)imide ([C4mpy][NTf2]) RTIL was
chosen due to its low viscosity, high chemical stability, and its success in sensing oxygen, sulfur
dioxide and methane [36, 85, 102, 103]. The whole process flow is illustrated in Figure 5.4.
At the end of all this process, a disk-shaped WE with radius of 100 μm was patterned on the
surface of the CMOS chip. An annular CE with inner radius of 120 µm and outer radius of 200 µm
and a RE were patterned around the WE. The gap between WE and CE is 20 µm. The RTIL covers

94

the whole electrode area. The average thickness of RTIL is around 26 µm based on the calculation
of the RTIL volume used and the droplet coverage area on the die. The photographs shown in
Figure 5.5 illustrates the CMOS circuit fabricated in foundry, CMOS chip with on-chip electrodes

Figure 5.4. Post-CMOS pocess flow for on-chip RTIL sensor fabrication: Shipley 1813 photoresist
is spin-coated (a) and developed (b). Titanium and gold in deposited by PVD (c). Photoresist is
then rinsed off to leave the electrode (d). CMOS chip with on-chip electrode is wire bonded to
package (e). A droplet of RTIL is casted on the electrodes (f).

Figure 5.5. Die photographs of CMOS chip fabricated by foundry and on-CMOS RTIL sensor
formed by post-CMOS fabrication.

95

and monolithic CMOS gas sensor including both CMOS circuit and on-CMOS RTIL sensor,
respectively.

5.5 Results
5.5.1

Test setup
To characterize operation and performance of the amperometric instrumentation circuit chip,

a custom evaluation board was designed to interface the chip with a data acquisition (DAQ) card
(National Instruments, DAQ USB-6259) and low noise battery supply. The DAQ generates digital
signals to configure and control the chip, provides input voltage, acquires and digitizes analog
outputs from the chip, and displays and stores data on a PC running LabVIEW. Keithley 6430
source meter was used as an accurate input current source to evaluate the current readout circuit.
To test the sensing performance of the CMOS monolithic RTIL sensor, the chemical
experiment station shown in Figure 5.6 was built. Oxygen was selected as an example analyte.
Oxygen and nitrogen (as background gas) concentrations were controlled and mixed using gas
blender 100 (MCI Instruments, Inc.) to vary oxygen concentration from 0% to 21%. The total flow
rate was fixed to 100 standard cubic centimeters per minute (sccm) to circumvent the influence of
differential gas flow rate. A 3D-printed gas chamber (by Connex 350 3D printer) was capped on
the DIP40 package carrier, sealed by Viton o-ring. Including the cavity of the carrier, the total
3

chamber space is 0.48 cm , allowing a total chamber response time of 0.3 s.

96

5.5.2

Electrical experimental results
The amperometric instrumentation circuit occupies 800 µm × 600 µm as shown in Figure 5.5.

The circuit consumes 280 µA with a 5 V supply in the static operation mode.
To characterize the effective voltage working range of the potentiostat, the rail-to-rail opamp
was connected as a unit-gain buffer and the positive input node was applied a voltage from 0 to 5 V.
The output voltage and the linearity error are plotted as a function of input voltage in Figure 5.7.
The absolute error of the output voltage is less than 0.05 % over the range from 0.03 V to 4.90 V.
The result is in good agreement with the design expectation (from 0.1 V to 4.9 V), maximizing the
applicable potential window required for electrochemical sensors in a limited power supply

Figure 5.6. Chemical test setup.

97

voltage.
To characterize the performance of amperometric readout circuit, input current was swept
from -1.5 µA to 1.5 µA and output voltage Vo was recorded by DAQ USB6259. The amperometric
readout clock frequency fs was set to 100 kHz and Vref2 was set to 1.3 V. The output voltage and
linearity error are plotted as a function of the input current in Figure 5.8. The input current from -1
µA to 1 µA shows a good linearity, with the output voltage in the range of 1.74 V to 0.78 V where
the absolute linearity errors are less than 4 nA. By adjusting fs from 10 Hz to 1 MHz, the dynamic
readout range varies from 100 pA to 10 µA and the least linearity error is 2 pA at fs = 10 Hz. In the
oxygen test of the on-CMOS RTIL sensor, current range from -1 µA to 1 µA and linearity error of
4 nA is proper, thus fs was set to 100 kHz. Performance summary of the CMOS interface circuit is
given in Table 5.1.
5.5.3

Chemical experimental results
To verify the functionality of the monolithic RTIL CMOS sensor, the monolithic sensor was

exposed to oxygen. Oxygen concentration was varied from 0 - 21% with a step of 2.1%. Oxygen

98

Figure 5.7. The rail-to-rail opamp output voltage range and linearity error as a function of input
voltage. The absolute error of the output voltage is less than 0.05% over the range from 0.03 V to
4.90 V.

Figure 5.8. Amperometric readout output range and linearity errors as a function of input current
when fs = 100 kHz, Vref2 = 1.3 V. In the range of -1 µA to 1 µA, the absolute linearity errors are
less than 4 nA.
Table 5.1.
Performance summary of the CMOS interface circuit.
Power supply
Steady current
Area
Input/output range of potentiostat
Current readout range(bi-direction)
Current linearity error

99

5V
280 μA
600 μm x 800 μm
0.03 - 4.90V
100 pA - 10 μA
< 2 pA

exposures were 10 s in duration, preceded and followed by exposure to pure nitrogen for 30 s. The
amperometric instrumentation circuit was set Vref1 = 2.1 V, Vref2 = 1.3 V (the sensor was applied
at -0.8 V respectively), and fs = 100 kHz. The output voltage of the monolithic sensor is plotted as
a function of time in Figure 5.9. The step response values at each rising step were collected to form
the oxygen calibration curve shown in Figure 5.10. The relationship between response voltage
Vr(mV) and oxygen concentration CO2(%) can be expressed by
2

Vr = 4.23·CO2 - 3.995 (R = 0.995)

(5.2)
2

The sensitivity of the monolithic sensor response was measured as 4.23 mV/%. The R value of
0.995 implies a high level of linearity of the monolithic sensor response to oxygen.
Repeatability test was performed by alternatingly purging 4.2% oxygen and pure nitrogen.
The resulting output voltage is shown in Figure 5.11. Although the sensor shows a slight drift due
to the latency of oxygen desorption, by subtracting the baseline value the response voltage ΔV
shows a good repeatability with an average value of 10.36 mV and a standard deviation of only
89µV. The corresponding repeatable limit of detection (LOD) for oxygen, defined as three times
the standard deviation, was found to be 0.11%.

100

Figure 5.9. Output of amperometric readout circuit for on-CMOS RTIL sensor response to oxygen.
Oxygen concentration varies from 0 - 21% with a step of 2.1%. Oxygen exposures were 10 s in
duration, preceded and followed by exposure to pure nitrogen for 30 s. The amperometric circuit
was set to Vref1=2.1 V and Vref2=1.3 V (the sensor was applied at -0.8V respectively), and fs=100
kHz.

Response voltage (mV)

100
75
50
25
0
0

3

6

9
12
O2(%)

15

18

21

Figure 5.10. Oxygen calibration curve for monolithic RTIL gas sensor using data extracted from
2

Figure 5.8. A high level of linearity (R = 0.995) was achieved.

101

Figure 5.11. Repeatability test. Constant-potential current response measured over five cycles of
alternate exposure to 4.2% oxygen and pure nitrogen flow at an applied bias of -0.8 V. The
standard deviation of the five voltage data is 89 µV and corresponding repeatable LOD of oxygen
is 0.06%.
To evaluate the benefits brought by the monolithic integration, another RTIL off-CMOS
sensor was tested to compare with the on-CMOS sensor. The off-CMOS sensor was fabricated on
a thermal oxide silicon substrate with the same-size electrodes as the on-CMOS sensor. Assuming
the droplets forming on the electrodes are both a spherical crown shape, based on the measured
surface radius and the contact angle, the volumes and average height of the droplet on on-CMOS
and off-CMOS sensors can be calculated as shown in Table 5.2. Due to the different surface
tension on silicon nitride (CMOS top layer) and silicon dioxide, the RTIL droplets have different
volumes and heights in the on-CMOS sensor and the off-CMOS sensor.
The off-CMOS sensor was connected with CMOS interface circuit and tested in oxygen.
Response time of both off-CMOS sensor and on-CMOS sensor were extracted from sensor step
response test and plotted as a function of oxygen concentration in Figure 5.12. In the test of 21%

102

O2, the response times of on-CMOS and off-CMOS sensor are 1.1 s and 1.5 s, respectively. The
on-CMOS sensor demonstrates a faster response, accounting for thinner RTIL layer formed on
silicon nitride. Based on response time equation on a one-dimension model [82],
2

t = h /D

(5.3)
2

where h is the thickness of the RTIL and D is the oxygen diffusion coefficient in RTIL (876μm /s
in [C4mpy][NTf2] [31]), the theoretical response time in on-CMOS and off-CMOS sensors are
0.75 s and 1.62 s, respectively. The test results have a good agreement with the theoretical values.
Signal noise level and full response range were measured from both off-CMOS sensor and
on-CMOS and listed in Table 5.3. The LOD of oxygen, defined as three times of noise standard
deviation, were calculated based on the sensitivity and measured voltage noise level. The
monolithic sensor demonstrates seven times of LOD improvement by eliminating environment
noise.

103

Table 5.2.
Off-CMOS and on-CMOS electrochemical RTIL sensor measured parameters and calculated
volume and height.

on-CMOS
Contact angle
(degree)
Radius(μm)
Volume(μm 3)
Avg height(μm)

off-CMOS

13.25
440

18.70
454

1.56E+07
26

2.44E+07
38

Figure 5.12. Off-CMOS and on-CMOS sensor response time (10% - 90%) characterized by
amperometric readout circuit.
Table 5.3.
Off-CMOS and on-CMOS electrochemical RTIL sensor noise level comparison.

Noise level

off-CMOS on-CMOS
608µV
55µV

Response at 21%O2 136mV

86mV

LOD of O2 (%)

0.04%

0.28%

104

5.6 Conclusion
This chapter presents the first monolithic electrochemical CMOS gas sensor microsystem.
The monolithic microsystem includes a compact amperometric instrumentation circuit. The
circuit was implemented in 0.5 μm CMOS with a 5 V supply and consumes 280 µA. The chip
provides both potentiostat and readout functions, achieving rail-to-rail voltage support and range
of bi-direction current readout from 100 pA to 10 µA with linearity error better than 2 pA.
Utilizing post-CMOS fabrication process, RTIL sensor was fabricated on the surface of the
CMOS instrumentation chip. Using oxygen as demonstrated analyte, the monolithic gas sensor
2

exhibits a high level of linear response (R = 0.995) in the range of 0 - 21% oxygen, with LOD
of 0.04%. Because RTIL has demonstrated sensing many gases especially including air
pollutants, this monolithic gas sensor microsystem provides a platform suitable for
portable/wearable air-pollutants monitoring applications.

105

6 Summary, Contributions and Future Work

6.1 Summary
Exposure to air toxic pollutants is an ever present concern for human health in modern society.
This research developed technologies that overcome the challenges of implementing an
inexpensive, highly miniaturized, and rapidly responding sensor array microsystem to measure
several toxic gases and VOCs for wearable air pollutant monitoring. The requirements for
wearable microsystems were studied and prevailing gas sensing technologies were evaluated,
followed by the identification of the challenges for the wearable sensor development. A planar
porous-electrode-on-permeable-membrane (PEoPM) structure was designed for RTIL-based
electrochemical sensor to overcome sensor-inherent slow response shortcoming, followed by the
development of miniaturized, fast-response, and flexible sensor structures. Utilizing the unique
inherent nature of electrochemical sensor interface, a compact amperometric instrumentation
circuit with current-to-digital output was developed as a low-cost low-power solution for
electrochemical gas sensor drive/readout. Finally, the first ever CMOS monolithic electrochemical
gas sensor using RTIL was implemented, enabling a highly-miniaturized wearable air pollutant
monitoring system. The research addressed some of the most important challenges of the air
pollutant monitoring microsystem by developing new sensor structures, a new compact
electrochemical sensor interface circuit and the first CMOS monolithic electrochemical gas sensor
platform.

106

6.2 Contributions
a.

Developed an electrochemical gas sensor array structure that responds 180% faster than
the traditional liquid-electrolyte structure and achieves a unique profile of performance that
is well suited for wearable air pollutant monitoring.

A planar porous-electrode-on-permeable-membrane structure was designed for RTIL-based
electrochemical sensor to overcome sensor’s inherent shortcoming of slow response. This
structure achieves 180% faster response compared to a traditional structure for sensing SO2. With
a unique microfabrication process that enables miniaturization of electrochemical sensors, a
miniaturized RTIL-based electrochemical gas sensor with array package and a planar flexible
sensor were implemented. Sensor response to CH4, SO2, NO2 and O2 was characterized,
demonstrating multiple air pollutants sensing capability.
b. Developed a compact power-efficient amperometric instrumentation circuit that
uniquely utilizes the inherent nature of electrochemical sensor properties to significantly
reduce circuit complexity compared to traditional architectures.

By fully understanding the electrochemical equivalent circuit model, incorporating the
sensor’s inherent interface capacitor into the instrumentation, a compact power-efficient
amperometric instrumentation circuit with digital output was developed. Compared to a traditional
amperometric instrumentation structure, this instrumentation design eliminates operational
amplifiers and a capacitor, thus dramatically saves area, cost and power. The compact

107

amperometric instrumentation was implemented and tested, demonstrating an 8-bit of effective
resolution in the range of -800 nA to 800 nA. Functionality of the instrumentation was verified by
potassium ferricyanide electrochemical experiment. High linearity of current-concentration
2

relation was acquired with 0.991 of R . This design suits for wearable electrochemical sensor
applications including electrochemical gas sensor microsystem.
c. Implemented the first ever CMOS monolithic electrochemical gas sensor; it achieves
detection limit seven times better than a hybrid solution and enables a highly-miniaturized
wearable air pollutant monitoring system

The first ever CMOS monolithic electrochemical gas sensor microsystem was implemented
2

using RTIL, occupying only 0.48mm per channel. Using oxygen as analyte example, the sensor
2

microsystem demonstrates a high-linearity response (R = 0.995) in the range of 0 - 21% oxygen,
limit of detection of 0.04%, and 1 second response time. Using monolithic integration, the sensor
eliminates environmental noise and achieves seven times resolution better than a hybrid solution.
Utilizing RTIL electrochemical sensing technique, this monolithic gas sensor approach provides
an ideal highly-miniaturized platform that enable gas sensor for wearable air pollutant monitoring
application.

6.3 Future work
This research has established the foundation for a wearable air pollutant monitoring. Future
full integration of RTIL-based sensor array on a compact sensor interface IC would further

108

demonstrate the dual benefit of electrochemical sensor and monolithic integration. To achieve this
goal, the following works are crucial.
a. Analyze miniaturization effect of RTIL-based sensor and develop a reliable miniaturized
RTIL-based sensor

Traditional electrochemical research focuses on one-electrode reaction based on the
assumption that the reaction occurring on the counter electrode would not affect the reaction
occurring on the working electrode. However, the assumption is invalid in the miniaturized
RTIL-based electrochemical sensor. Experiment discovered, within a minute, the products
generated on the CE diffuse to the WE and interfere the amperometric current. Such
miniaturization effect needs to be thoroughly analyzed and consequently the geometry and/or
structure need to be optimized for a reliable electrochemical sensor.
b. Implement compact electrochemical sensor interface IC with on-CMOS RTIL sensor

IC realization of the multi-channel compact amperometric instrumentation circuit is essential
for a compact low-power low-cost array microsystem. Electrical specifications of the on-CMOS
RTIL sensor array will be identified and multi-channel compact amperometric instrumentation
circuit will be customized to meet the specifications. In addition, impedance-mode
electrochemical instrumentation circuit will be incorporated into this interface IC to fully extract
the sensor response.
c. Develop an array fabrication process for CMOS monolithic electrochemical sensor array

109

On-CMOS sensor fabrication process has been developed for monolithic RTIL-based
electrochemical sensor. However, this process lacks capability to implement monolithic arrays. To
restrict RTIL in individual electrochemical cells, chemical-resist micro reservoir implementation
will be studied to develop an on-CMOS array for CMOS monolithic electrochemical sensor array.

110

APPENDICES

111

Appendix A: EIS equivalent circuit modeling
A.1 EIS equivalent circuit modeling Faradaic reaction
Randle model is a typical EIS equivalent circuit model to describe electrode interface [46].
The impedance spectrum of the electrode interface in RTIL in the case of NO2 and O2 has a good
agreement with Randle model. Thus the relation between the interface impedance and the
concentration of gas analytes can be revealed by Randle model.
As shown in Figure A.1, a Randle model consists of solution resistance Rs, double layer
capacitor Cdl, charge transfer resistance Rct, and Warburg impedance ZW. The impedance of
Randle model is expressed as

Z  Rs  ( jCdl  ( Rct  ZW )1)1

(A.1)

In them, Rct and ZW are functions of the analyte’s concentration.

Figure A.1. Randle equivalent circuit model.
Rct is expressed as

Rct 

RT
Fi0

112

(A.2)

where R is the Molar gas constant, equal to 8.31 J/mol·K; T is the temperature; F is the Faraday
4

constant, equal to 9.65×10 C/mol; and i0 is the exchange current that describes balanced faradaic
activity, which is defined as

*(1
*
i0  FAk 0CO  )CR

(A.3)

0

*

where A is area; k is standard heterogeneous rate constant; α is the transfer coefficient; CO is the
bulk concentration of O (the concentration in the supporting electrolyte for the gas analyte); and
*

CR is the bulk concentration of R which is the product in the one-electron one-step reaction

Oe R.
ZW is expressed as

ZW   (

1





j



)

(A.4)

where ω is the angular frequency and σ is Warburg coefficient, defined as



RT

1
1
(

)
*
2 F 2 A D C*
DR CR
O O

(A.5)

where DO is the diffusion coefficient of O and DR is the diffusion coefficient of R.
Nernst equation is used to provide link between electrode potential E and the concentrations
of participants in the electrode process described above

*
RT CO
E  E0 
ln
*
F
CR

113

(A.6)

0

where E is the standard potential. Therefore, given a steady state at a fixed potential, the
*

*

concentration ratio kC = CO / CR , is constant. Subsequently, Rct and ZW can be described as
*

functions of CO

k1
*
CO

(A.7)

k2 1
j
(

)
*
CO 


(A.8)

Rct 
ZW 
where k1 is defined as

k1 

RT

F 2 Ak 0 kC

(A.9)

and k2 is defined as

k2 

RT
1
1
(

)
2
2 F A DO
DR kC

*

(A.10)
*

k1 and k2 are independent of CO . Since gas concentration in the supporting analyte CO is
proportional to gas concentration in ambient air Cg with Henry law, kH was defined as the Henry
constant:
*
CO
kH 
Cg

(A.11)

Thus for a fixing frequency, the impedance of Randle model is expressed as a function of
analyte’s concentration Cg

| Z | Rs  (k3  Cg  j (k4  k5  C g )) 1

114

(A.12)

where k1, k2, and k3 are defined as follows

k3  k H

k1  k2 / 
k12  2k1k2 /   2k2 2 / 

k4  Cdl
k5  k H

(A.13)

(A.14)

k2 / 
k12  2k1k2 /   2k2 2 / 

(A.15)

2

Using Equation (A.12) to fit the test data, high R values for calibration curves were achieved
in the cases of NO2 and O2.

A.2 EIS equivalent circuit modeling absorption process
When the bias is much lower than over-potential, no faradaic reaction occurs (in other words,
faradaic current is much smaller than charging current). In such bias condition, the equivalent
circuit model between RE and WE is a series connection of solution resistance Rs and double layer
capacitor Cdl as shown in Figure A.2. The interface impedance is

Z  Rs 

1
jCdl

(A.16)

Figure A.2. The equivalent circuit model when bias is much lower than over-potential.

115

The imaginary part of impedance is

Zim  

1
Cdl

(A.17)

At a certain bias, dissolved gas molecules are absorbed on the electrode, thus changing Cdl.
The relation between Cdl and gas concentration Cg can be revealed by Langmuir absorption
mechanism.
The WE surface coverage of gas molecules is determined by Langmuir equation



K1Cg ,IL
(A.18)

1  K1Cg ,IL

where Cg,IL is the concentration of gas analyte in IL,

K1 is the ratio of the direct and inverse

rate constants in absorption equivalent equation. Since Cg,IL is proportional to Cg in ambient air
with Henry law, kH was defined as the Henry constant:

kH 

C g , IL
Cg

(A.19)

If K = kH ×K1 , Equation (A.17) can be rewritten as



KC g
1  KC g

(A.20)

Define Cdl0 as double-layer capacitance value when θ=0 and Cdl1 as double-layer capacitance
value when θ=100%. The double-layer capacitance Cdl can be expressed in terms of θ as follows

Cdl  Cdl 0  (1   )  Cdl1 

116

(A.21)

Derived from Equation (A.20) and (A.21), Cdl is rewritten as

KC g

Cdl  Cdl 0  (Cdl1  Cdl 0 ) 
1  KC

g

(A.22)

Replacing Cdl in Equation (A.17) by Equation (A.22), Zim is a function of Cg

k Cg
-Zim  k6  7
C k
g

(A.23)

8

where k6, k7, and k8 are defined as

k6 

k7 

1
Cdl 0

(A.24)

Cdl 0  Cdl1

Cdl1Cdl 0

k8 

(A.25)

Cdl 0
(A.26)

KCdl1
2

Using Equation (A.23) to fit the test data, high R values for calibration curves were achieved
in the cases of CH4 and SO2.

117

Appendix B: Fabrication procedure for PEoPM sensors
B.1 Fabrication process for macro PEoPM sensors
Material requirement
1. Equipment and tools: Edward 360 thermal evaporator, tweezers and scissors.
2. Chemicals: acetone, IPA.
3. Materials: POREX® porous PTFE sheet with 35% porosity and 4μm pore size (PM23J,
from Interstate Specialty Products, Inc.), gold pellet.
Process
1. Prepare a stainless steel hard mask by electric discharge machining (EDM).
2. Cut PTFE sheet to the size of 0.75” x 0.5”.
3. Clean the PTFE sheet by spraying acetone to sheet, followed by IPA cleaning and nitrogen
blowing.
4. Gold deposition
4.1 Tool: Edward 360 thermal evaporator
4.2 Mount the hard mask against the PTFE substrate on the PVD sample stage.
4.3. Deposit 300nm thick gold by following the SOP of Edward 360 thermal evaporator.

118

B.2. Fabrication process for miniaturized PEoPM sensors
Material requirement
1. Equipment and tools: Spin processor, oven, Edward 360 thermal evaporator, aluminum foil,
cleanroom wipes, tweezers, glass slides, binder clips, and scissors.
2. Chemicals: AZ4620 photoresist, AZ300 MIF developer, acetone, IPA, DI water.
3. Materials: POREX® porous PTFE sheet with 35% porosity and 4μm pore size (PM23J,
from Interstate Specialty Products, Inc.), gold pellet.
Process
1. Cut PTFE sheet to the size of 1.6” x 1.6”.
2. Clean PTFE sheet, glass slides, and binder clips.
Clean the sheet, glass slides and binder clips by spraying acetone to sheet, followed by IPA
cleaning and nitrogen blowing.
3. Spin-on AZ4620
3.1 Tool: Spin Coater
3.2 Fix the PTFE sheet on the glass slide by binder clips.
3.3. Blow-clean the surface of PTFE briefly with nitrogen.
3.3 Center the glass slide on chuck and perform a test spin to check the PTFE sheet is
firmly held on the slide. Ramp the spin to 2100 rpm in 5 seconds and hold at this speed for 40
seconds.

119

3.4 Pour AZ4620 from the bottle directly onto the PTFE sheet. Try not to introduce
bubbles while pouring. Let the SU-8 flow and relax for a few seconds. Start the spin cycle
identical to the test spin as above.
4. Pre-bake SU-8
4.1 Tool: Oven (preheat to 95C)
4.2 Release the PTFE sheet from the glass slide. Place the PTFE in the oven to bake for
5 minute. Remove the sheet and allow it to cool down for 5 minutes before exposure.
5. Exposure
5.1 Tool: Mask aligner
5.2 Design and print the transparency mask
5.3 Mount the mask on the aligner. Note the side of the emulsion print should be facing
the wafer and resist.
5.4 Set the exposure time for 60 seconds.
NOTE: If needed, clean the transparency with a fab wipe using IPA and blow dry with
nitrogen.
6. Development
6.1 Tool: Chemical hood
6.2 Under hood, pour AZ300 MIF developer into a beaker. Place the sheet into the
beaker for ~5 minutes. Make sure the pattern has clearly appeared on the sheet. Carefully
pick up the sheet with a tweezer, rinse with water, and blow dry with nitrogen.

120

7. Gold deposition
7.1 Tool: Edward 360 thermal evaporator
7.2 Deposit 300 nm thick gold by following the SOP of Edward 360 thermal evaporator.
8. Liftoff
8.1 Tool: Chemical hood
8.2 Under hood, pour acetone in the beaker and soak gold coated PTFE sheet in acetone.
Seal the beaker by aluminum foil and place under hood overnight.
8.3. Rinse PTFE sheet by acetone to liftoff, followed by IPA cleaning and nitrogen
blowing.

B.3. Assembly process for miniaturized PEoPM sensor arrays
Material requirement
1. Equipment and tools: Cleanroom wipes, tweezers, screw drivers, felting needles, pipettes
(Rainin Classic PR-2) and scissors.
2. Materials: Fabricated PEoPM sensor device (followed by B.2), silicone rubber (TSE397),
conductive epoxy (MG Chemicals 8331), fabricated PCB (by 4PCB) for lower-frame and
upper-frame, Viton O-rings, and RTIL-[C4mpy][NTf2].
.
Process

121

1. Place four pieces of PEoPM sensor device on the lower-frame PCB, glue them by applying
the silica on the four corners, wait 12 h to dry out.
2. Place four pieces of Viton O-rings on the upper-frame PCB, glue it by applying the silica on
the edge, wait 12 h to dry out.
3. Mix conductive epoxy tube A and B together, stir them thoroughly by felting needles, and
apply the mixed conductive epoxy to connect the PEoPM sensors’ pads with via on the
lower-frame PCB.
4. Align the upper-frame PCB (with O-rings) and the lower-frame PCB (with PEoPM
sensors), tighten four parts by screws.
5. Drop 1 µL RTILs by the pipettes into each O-ring.

B.4. Fabrication process for flexible PEoPM sensors
Material requirement
1. Equipment and tools: Cleanroom wipes, tweezers, pipettes, and vacuum oven.
2. Chemicals: acetone.
3. Materials: Fabricated PEoPM sensor device (followed by B.2), and RTIL-[C4mpy][NTf2].
Process
1. Mix acetone and RTIL with volume ratio of 2:1.
2. Pick 0.6µL of the mixture by the pipettes and drop on the PEoPM sensor device.

122

3. Place the sensor in the vacuum and vacuum for overnight.

123

BIBLIOGRAPHY

124

BIBLIOGRAPHY

[1]

J. A. Bernstein, N. Alexis, C. Barnes, I. L. Bernstein, A. Nel, D. Peden, et al., "Health effects
of air pollution," Journal of Allergy and Clinical Immunology, vol. 114, pp. 1116-1123, 2004.

[2]

B. Brunekreef and S. T. Holgate, "Air pollution and health," The Lancet, vol. 360, pp.
1233-1242, 2002.

[3]

X. Han and L. P. Naeher, "A review of traffic-related air pollution exposure assessment
studies in the developing world," Environment International, vol. 32, pp. 106-120, 2006.

[4]

N. Nikzad, N. Verma, C. Ziftci, E. Bales, N. Quick, P. Zappi, et al., "CitiSense: improving
geospatial environmental assessment of air quality using a wireless personal exposure
monitoring system," presented at the Proceedings of the Conference on Wireless Health, San
Diego, CA, 2012.

[5]

Common sense Available: http://www.communitysensing.org/technology.php

[6]

Aircasting Available: http://aircasting.org/

[7]

S. Choi, N. Kim, H. Cha, and R. Ha, "Micro Sensor Node for Air Pollutant Monitoring:
Hardware and Software Issues," Sensors, vol. 9, pp. 7970-7987, 2009.

[8]

L. Jen-Hao, C. Yu-Fan, L. Tzu-Shiang, L. Da-Wei, W. Tzai-Hung, S. Chih-Hong, et al.,
"Developed urban air quality monitoring system based on wireless sensor networks," in
International Conference on Sensing Technology, 2011, pp. 549-554.

[9]

Pavia, D. L., G. M. Lampman, G. S. Kritz, and R. G. Engel, 4th ed.: Thomson Brooks/Cole.,
2006.

[10] Available: http://en.wikipedia.org/wiki/Infrared_gas_analyzer
[11] G. F. Fine, L. M. Cavanagh, A. Afonja, and R. Binions, "Metal Oxide Semi-Conductor Gas
Sensors in Environmental Monitoring," Sensors, vol. 10, pp. 5469-5502, 2010.
[12] H. Bai and G. Shi, "Gas Sensors Based on Conducting Polymers," Sensors, vol. 7, pp.
267-307, 2007.

125

[13] J. R. Stetter, G. Korotcenkov, X. Zeng, Y. Tang, and Y. Liu, "Electrochemical Gas Sensors:
Fundamentals, Fabrication, and Parameters," in Chemical Sensors Comprehensive Sensor
Technologies. vol. 5, G. Korotcenkov, Ed., ed New York: Momentum Press, 2011, pp. 1-89.
[14] C. Dass, Fundamentals of Contemporary Mass Spectrometry: Wiley, 2007.
[15] I. G. McWilliam and R. A. Dewar, "Flame Ionization Detector for Gas Chromatography,"
Nature, vol. 181, pp. 760-760, 1958.
[16] Compendium of Methods for the Determination of Toxic Organic Compounds in Ambient
Air. Available: http://www.epa.gov/ttnamti1/files/ambient/airtox/tocomp99.pdf
[17] C.-J. Lu, W. H. Steinecker, W.-C. Tian, M. C. Oborny, J. M. Nichols, M. Agah, et al.,
"First-generation hybrid MEMS gas chromatograph," Lab on a Chip, vol. 5, pp. 1123-1131,
2005.
[18] S. K. Kim, D. R. Burris, H. Chang, J. Bryant-Genevier, and E. T. Zellers, "Microfabricated
Gas Chromatograph for On-Site Determination of Trichloroethylene in Indoor Air Arising
from Vapor Intrusion. 1. Field Evaluation," Environmental Science & Technology, vol. 46, pp.
6065-6072, 2012.
[19] M. Geear, R. R. A. Syms, S. Wright, and A. S. Holmes, "Monolithic MEMS quadrupole mass
spectrometers by deep silicon etching," Journal of Microelectromechanical Systems, vol. 14,
pp. 1156-1166, 2005.
[20] Z. Ouyang and R. G. Cooks, "Miniature Mass Spectrometers," Annual Review of Analytical
Chemistry, vol. 2, pp. 187-214, 2009.
[21] J. Kim, B. Bae, J. Hammonds, T. Kang, and M. A. Shannon, "Development of a micro-flame
ionization detector using a diffusion flame," Sensors and Actuators B: Chemical, vol. 168,
pp. 111-117, 2012.
[22] M. Aleixandre and M. Gerboles, "Review of Small Commercial Sensors for Indicative
Monitoring of Ambient Gas," Chemical Engineering, vol. 30, pp. 169-174, 2012.
[23] M. Graf, U. Frey, S. Taschini, and A. Hierlemann, "Micro Hot Plate-Based Sensor Array
System for the Detection of Environmentally Relevant Gases," Analytical Chemistry, vol. 78,
pp. 6801-6808, 2006.
[24] R. Knake, P. Jacquinot, and P. C. Hauser, "Amperometric Detection of Gaseous
Formaldehydein the ppb Range," Electroanalysis, vol. 13, pp. 631-634, 2001.

126

[25] G. Schiavon, G. Zotti, G. Bontempelli, and G. Farnia, "Amperometric Monitoring of Ozone in
Gaseous Media by Gold Electrodes Supported on Ion Exchange Membranes (Solid Polymer
Electrolytes)," Anal. Chem., vol. 62, pp. 293-298, 1990.
[26] A. W. E. Hodgson, P. Jacquinot, and P. C. Hauser, "Electrochemical Sensor for the Detection
of SO2 in the Low-ppb Range," Anal. Chem., vol. 71, pp. 2831-2837, 1999.
[27] R. Knake, P. Jacquinot, and P. C. Hauser, "A Direct Comparison of Amperometric Gas
Sensors with Gas-Diffusion and Ion-Exchange Membrane Based Electrodes," Analyst, vol.
127, pp. 114-118, 2002.
[28] A. Manickam, A. Chevalier, M. McDermott, A. D. Ellington, and A. Hassibi, "A CMOS
Electrochemical Impedance Spectroscopy (EIS) Biosensor Array," IEEE Transactions on
Biomedical Circuits and Systems, vol. 4, pp. 379-390, 2010.
[29] P. M. Levine, G. Ping, R. Levicky, and K. L. Shepard, "Active CMOS Sensor Array for
Electrochemical Biomolecular Detection," IEEE Journal of Solid-State Circuits, vol. 43, pp.
1859-1871, 2008.
[30] Y. Chao, S. R. Jadhav, R. M. Worden, and A. J. Mason, "Compact Low-Power
Impedance-to-Digital Converter for Sensor Array Microsystems," IEEE Journal of
Solid-State Circuits, vol. 44, pp. 2844-2855, 2009.
[31] E. I. Rogers, A. M. O'Mahony, L. Aldous, and R. G. Compton, "Amperometric Gas Detection
Using Room Temperature Ionic Liquid Solvents," ECS Transactions, vol. 33, pp. 473-502,
October 1, 2010 2010.
[32] M. J. A. Shiddiky and A. A. J. Torriero, "Application of ionic liquids in electrochemical
sensing systems," Biosensors and Bioelectronics, vol. 26, pp. 1775-1787, 2011.
[33] D. S. Silvester, "Recent advances in the use of ionic liquids for electrochemical sensing,"
Analyst, vol. 136, pp. 4871-4882, 2011.
[34] R. Toniolo, N. Dossi, A. Pizzariello, A. P. Doherty, S. Susmel, and G. Bontempelli, "An
oxygen amperometric gas sensor based on its electrocatalytic reduction in room temperature
ionic liquids," Journal of Electroanalytical Chemistry, vol. 670, pp. 23-29, 2012.
[35] R. Wang, T. Okajima, F. Kitamura, and T. Ohsaka, "A Novel Amperometric O2 Gas Sensor
Based on Supported Room-Temperature Ionic Liquid Porous Polyethylene
Membrane-Coated Electrodes," Electroanalysis, vol. 16, pp. 66-72, 2004.

127

[36] Z. Wang, P. Lin, G. A. Baker, J. Stetter, and X. Zeng, "Ionic Liquids as Electrolytes for the
Development of a Robust Amperometric Oxygen Sensor," Analytical Chemistry, vol. 83, pp.
7066-7073, 2011/09/15 2011.
[37] C. Hu, X. Bai, Y. Wang, W. Jin, X. Zhang, and S. Hu, "Inkjet Printing of Nanoporous Gold
Electrode Arrays on Cellulose Membranes for High-Sensitive Paper-Like Electrochemical
Oxygen Sensors Using Ionic Liquid Electrolytes," Analytical Chemistry, vol. 84, pp.
3745-3750, 2012/04/17 2012.
[38] X.-J. Huang, L. Aldous, A. M. O’Mahony, F. J. del Campo, and R. G. Compton, "Toward
Membrane-Free Amperometric Gas Sensors: A Microelectrode Array Approach," Analytical
Chemistry, vol. 82, pp. 5238-5245, 2010.
[39] S.-Q. Xiong, Y. Wei, Z. Guo, X. Chen, J. Wang, J.-H. Liu, et al., "Toward Membrane-Free
Amperometric Gas Sensors: An Ionic Liquid–Nanoparticle Composite Approach," The
Journal of Physical Chemistry C, vol. 115, pp. 17471-17478, 2011/09/08 2011.
[40] M. Nádherná, F. Opekar, and J. Reiter, "Ionic liquid–polymer electrolyte for amperometric
solid-state NO2 sensor," Electrochimica Acta, vol. 56, pp. 5650-5655, 2011.
[41] S. R. Ng, C. X. Guo, and C. M. Li, "Highly Sensitive Nitric Oxide Sensing Using
Three-Dimensional Graphene/Ionic Liquid Nanocomposite," Electroanalysis, vol. 23, pp.
442-448, 2011.
[42] X. Ji, C. E. Banks, D. S. Silvester, L. Aldous, C. Hardacre, and R. G. Compton,
"Electrochemical Ammonia Gas Sensing in Nonaqueous Systems: A Comparison of
Propylene Carbonate with Room Temperature Ionic Liquids," Electroanalysis, vol. 19, pp.
2194-2201, 2007.
[43] A. M. O’Mahony, D. S. Silvester, L. Aldous, C. Hardacre, and R. G. Compton, "The
Electrochemical Reduction of Hydrogen Sulfide on Platinum in Several Room Temperature
Ionic Liquids," The Journal of Physical Chemistry C, vol. 112, pp. 7725-7730, 2008/05/01
2008.
[44] N. Dossi, R. Toniolo, A. Pizzariello, E. Carrilho, E. Piccin, S. Battiston, et al., "An
electrochemical gas sensor based on paper supported room temperature ionic liquids," Lab
on a Chip, vol. 12, pp. 153-158, 2012.
[45] M. A. G. Zevenbergen, D. Wouters, V.-A. T. Dam, S. H. Brongersma, and M. Crego-Calama,
"Electrochemical Sensing of Ethylene Employing a Thin Ionic-Liquid Layer," Analytical
Chemistry, vol. 83, pp. 6300-6307, 2011.

128

[46] A. J. Bard and L. R. Faulkner, Electrochemical
Applications, 2 ed.: Wiley, 2000.

Methods:

Fundamentals

and

[47] H. Yue and A. J. Mason, "A redox-enzyme-based electrochemical biosensor with a CMOS
integrated bipotentiostat," in IEEE Biomedical Circuits and Systems Conf., 2009, pp. 29-32.
[48] S. M. Martin, F. H. Gebara, T. D. Strong, and R. B. Brown, "A low-voltage, chemical sensor
interface for systems-on-chip: the fully-differential potentiostat," in IEEE Int. Symp. on
Circuits and Systems, Vancouver, Canada, 2004, pp. 892-895.
[49] M. M. Ahmadi and G. A. Jullien, "Current-Mirror-Based Potentiostats for Three-Electrode
Amperometric Electrochemical Sensors," IEEE Transactions on Circuits and Systems I, vol.
56, pp. 1339-1348, 2009.
[50] R. J. Reay, S. P. Kounaves, and G. T. A. Kovacs, "An Integrated CMOS Potentiostat for
Miniaturized Electroanalytical Instrumentation," presented at the Proc. Inter. Solid-State
Circuits Conf., 1994.
[51] P. M. Levine, G. Ping, R. Levicky, and K. L. Shepard, "Active CMOS Sensor Array for
Electrochemical Biomolecular Detection," IEEE J. of Solid-State Circuits, vol. 43, pp.
1859-1871, 2008.
[52] A. Gore, S. Chakrabartty, S. Pal, and E. C. Alocilja, "A Multichannel
Femtoampere-Sensitivity Potentiostat Array for Biosensing Applications," IEEE
Transactions on Circuits and Systems I, vol. 53, pp. 2357-2363, 2006.
[53] L. Li, X. Liu, W. A. Qureshi, and A. J. Mason, "CMOS Amperometric Instrumentation and
Packaging for Biosensor Array Applications," IEEE Transactions on Biomedical Circuits
and Systems, vol. 5, pp. 439-448, 2011.
[54] M. Stanacevic, K. Murari, A. Rege, G. Cauwenberghs, and N. V. Thakor, "VLSI Potentiostat
Array With Oversampling Gain Modulation for Wide-Range Neurotransmitter Sensing,"
IEEE Transactions on Biomedical Circuits and Systems, vol. 1, pp. 63-72, 2007.
[55] K. Murari, M. Stanacevic, G. Cauwenberghs, and N. V. Thakor, "Integrated Potentiostat for
Neurotransmitter Sensing," IEEE Eng. Medicine Biol. Mag., vol. 24, pp. 23-29, 2005.
[56] S. Ayers, K. D. Gillis, M. Lindau, and B. A. Minch, "Design of a CMOS Potentiostat Circuit
for Electrochemical Detector Arrays," IEEE Transactions on Circuits and Systems I, vol. 54,
pp. 736-744, 2007.

129

[57] A. Frey, M. Schienle, C. Paulus, Z. Jun, F. Hofmann, P. Schindler-Bauer, et al., "A digital
CMOS DNA chip," in IEEE International Symposium on Circuits and Systems, 2005, pp.
2915-2918
[58] E. Barsoukov and J. Macdonald, Impedance Spectroscopy: Theory, Experiment, and
Applications, 2 ed.: John Wiley and Sons, 2005.
[59] D. Rairigh, C. Yang, and A. Mason, "Analysis of On-Chip Impedance Spectroscopy
Methodologies for Sensor Arrays," Sensor Letters, vol. 4, pp. 398-402, 2006.
[60] G. Ramos, "Analog Computation of the Fast Fourier Transform," presented at the IEEE
Proceedings, 1970.
[61] Y. Chao, S. R. Jadhav, R. M. Worden, and A. J. Mason, "Compact Low-Power
Impedance-to-Digital Converter for Sensor Array Microsystems," IEEE J. of Solid-State
Circuits, vol. 44, pp. 2844-2855, 2009.
[62] H. M. Jafari and R. Genov, "CMOS impedance spectrum analyzer with dual-slope
multiplying ADC," in IEEE Biomedical Circuits and Systems Conf., 2011, pp. 361-364.
[63] A. Hassibi and T. H. Lee, "A Programmable 0.18um CMOS Electrochemical Sensor
Microarray for Biomolecular Detection," IEEE Sensors Journal, vol. 6, pp. 1380-1388,
2006.
[64] J. W. Gardner, P. K. Guha, F. Udrea, and J. A. Covington, "CMOS Interfacing for Integrated
Gas Sensors: A Review," IEEE Sensors Journal, vol. 10, pp. 1833-1848, 2010.
[65] D. Lange, C. Hagleitner, A. Hierlemann, O. Brand, and H. Baltes, "Complementary Metal
Oxide Semiconductor Cantilever Arrays on a Single Chip:  Mass-Sensitive Detection of
Volatile Organic Compounds," Analytical Chemistry, vol. 74, pp. 3084-3095, 2002.
[66] M. Zimmermann, T. Volden, K. U. Kirstein, S. Hafizovic, J. Lichtenberg, and A. Hierlemann,
"A CMOS-based sensor array system for chemical and biochemical applications," in
Proceedings of European Solid-State Circuits Conference, 2005, pp. 343-346.
[67] A. Koll, A. Schaufelbuhl, N. Scheeberger, U. Munch, O. Brand, H. Baltes, et al.,
"Micromachined CMOS calorimetric chemical sensor with on-chip low noise amplifier," in
IEEE International Conference on Micro Electro Mechanical Systems, 1999, pp. 547-551.
[68] N. Kerness, A. Koll, A. Schaufelbuhl, C. Hagleitner, A. Hierlemann, O. Brand, et al.,
"N-well based CMOS calorimetric chemical sensors," in International Conference on Micro
Electro Mechanical Systems, 2000, pp. 96-101.

130

[69] J. Chou, Hazardous gas monitors : a practical guide to selection, operation and applications.
New York [etc.]: McGraw-Hill, 2000.
[70] M. L. Simpson, G. S. Sayler, B. M. Applegate, S. Ripp, D. E. Nivens, M. J. Paulus, et al.,
"Bioluminescent-bioreporter integrated circuits form novel whole-cell biosensors," Trends in
Biotechnology, vol. 16, pp. 332-338, 1998.
[71] D. E. Nivens, T. E. McKnight, S. A. Moser, S. J. Osbourn, M. L. Simpson, and G. S. Sayler,
"Bioluminescent bioreporter integrated circuits: potentially small, rugged and inexpensive
whole-cell biosensors for remote environmental monitoring," Journal of Applied
Microbiology, vol. 96, pp. 33-46, 2004.
[72] K. I. Syed, V. Rajagopal, Z. Mo, R. Steven, D. C. Sam, W. Brandon, et al., "Integrated Circuit
Biosensors Using Living Whole-Cell Bioreporters," IEEE Transactions on Circuits and
Systems I, vol. 54, pp. 89-98, 2007.
[73] P. Murali, R. K, N. Bhat, G. Banerjee, B. Amrutur, K. N. Bhat, et al., "A CMOS Gas Sensor
Array Platform With Fourier Transform Based Impedance Spectroscopy," IEEE
Transactions on Circuits and Systems I, vol. PP, pp. 1-11, 2012.
[74] K. Thomas Jacob, H. Alister, T. Su Lim, A. C. James, W. G. Julian, and C. P. Tim, "Analog
VLSI Circuit Implementation of an Adaptive Neuromorphic Olfaction Chip," IEEE
Transactions on Circuits and Systems I, vol. 54, pp. 60-73, 2007.
[75] J. A. Dickson and R. M. Goodman, "Integrated chemical sensors based on carbon black and
polymer films using a standard CMOS process and post-processing," in IEEE International
Symposium on Circuits and Systems, 2000, pp. 341-344.
[76] S. M. Cho, K. Sang Choon, H. Seung-Chul, K. Yong Shin, K. Young Jun, Y. Yoonseok, et al.,
"Monolithic electronic nose system fabricated by post CMOS micromachining," in IEEE
Sensors, 2005, pp. 420-423.
[77] X. Mu, C. E., R. D., C. Kurdak, E. Zellers, and A. J. Mason, "CMOS Monolithic
Nanoparticle-Coated Chemiresistor Array for Micro-Scale Gas Chromatography," IEEE
Sensors, vol. 12, pp. 2444-2452, 2012.
[78] C. Chia Ling, V. Agarwal, S. Sonkusale, and M. R. Dokmeci, "Integration of Single-Walled
Carbon Nanotubes on to CMOS Circuitry with Parylene-C Encapsulation," in IEEE
Conference on Nanotechnology, 2008, pp. 480-483.
[79] G. Barillaro and L. M. Strambini, "An integrated CMOS sensing chip for NO2 detection,"
Sensors and Actuators B, vol. 134, pp. 585-590, 2008.

131

[80] G. Barillaro, P. Bruschi, F. Pieri, and L. M. Strambini, "CMOS-compatible fabrication of
porous silicon gas sensors and their readout electronics on the same chip," physica status
solidi (a), vol. 204, pp. 1423-1428, 2007.
[81] C. Wilbertz, H. P. Frerichs, I. Freund, and M. Lehmann, "Suspended-Gate- and
Lundstrom-FET integrated on a CMOS-chip," Sensors and Actuators A, vol. 123–124, pp.
2-6, 2005.
[82] M. C. Buzzeo, C. Hardacre, and R. G. Compton, "Use of Room Temperature Ionic Liquids in
Gas Sensor Design," Analytical Chemistry, vol. 76, pp. 4583-4588, 2004/08/01 2004.
[83] H. Tokuda, S. Tsuzuki, M. A. B. H. Susan, K. Hayamizu, and M. Watanabe, "How Ionic Are
Room-Temperature Ionic Liquids? An Indicator of the Physicochemical Properties," The
Journal of Physical Chemistry B, vol. 110, pp. 19593-19600, 2006.
[84] Conductivity
ordering
guide.
Available:
http://myweb.wit.edu/sandinic/Research/conductivity%20v%20concentration.pdf
[85] Z. Wang, X. Mu, M. Guo, Y. Huang, A. J. Mason, and X. Zeng, "Methane Recognition and
Quantification by Differential Capacitance at the Hydrophobic Ionic Liquid-Electrified
Metal Electrode Interface," Journal of The Electrochemical Society, vol. 160, pp. B83-B89,
2013.
[86] H. Li, X. Mu, Z. Wang, M. Guo, X. Zeng, and A. J. Mason, "Room Temperature Ionic-Liquid
Electrochemical Gas Sensor Array System for Real-Time Mine Safety Monitoring," in IEEE
Sensors Conf., 2013.
[87] X. Mu, Z. Wang, M. Guo, X. Zeng, and A. J. Mason, "Fabrication of a miniaturized room
temperature ionic liquid gas sensor for human health and safety monitoring," in IEEE
Biomedical Circuits and Systems Conf., Hsinchu, Taiwan, 2012, pp. 140-143.
[88] Z. Wang, P. Lin, G. A. Baker, J. Stetter, and X. Zeng, "Ionic Liquids as Electrolytes for the
Development of a Robust Amperometric Oxygen Sensor," Anal. Chem., vol. 83, pp.
7066-7073, 2011.
[89] C. M. Pettit, P. C. Goonetilleke, C. M. Sulyma, and D. Roy, "Combining Impedance
Spectroscopy with Cyclic Voltammetry:  Measurement and Analysis of Kinetic Parameters
for Faradaic and Nonfaradaic Reactions on Thin-Film Gold," Analytical Chemistry, vol. 78,
pp. 3723-3729, 2006.

132

[90] S. M. Martin, F. H. Gebara, B. J. Larivee, and R. B. Brown, "A CMOS-integrated
microinstrument for trace detection of heavy metals," IEEE Journal of Solid-State Circuits,
vol. 40, pp. 2777-2786, 2005.
[91] P. M. Levine, P. Gong, R. Levicky, and K. L. Shepard, "Active CMOS Sensor Array for
Electrochemical Biomolecular Detection," IEEE Journal of Solid-State Circuits, vol. 43, pp.
1859-1871, 2008.
[92] E. Lauwers, J. Suls, W. Gumbrecht, D. Maes, G. Gielen, and W. Sansen, "A CMOS
multiparameter biochemical microsensor with temperature control and signal interfacing,"
IEEE Journal of Solid-State Circuits, vol. 36, pp. 2030-2038, 2001.
[93] L. Li, L. Xiaowen, W. A. Qureshi, and A. J. Mason, "CMOS Amperometric Instrumentation
and Packaging for Biosensor Array Applications," IEEE Transactions on Biomedical
Circuits and Systems, vol. 5, pp. 439-448, 2011.
[94] D. Kościelnik and M. Miśkowicz, "Asynchronous Sigma-Delta analog-to digital converter
based on the charge pump integrator," Analog Integrated Circuits and Signal Processing, vol.
55, pp. 223-238, 2008.
[95] M. E. Orazem and B. Tribollet, Electrochemical Impedance Spectroscopy: Wiley, 2008.
[96] J. A. Plambeck, Electroanalytical Chemistry: Basic Principles and Applications. New York:
Wiley, 1982.
[97] Y. Chao, S. R. Jadhav, R. M. Worden, and A. J. Mason, "Compact Low-Power
Impedance-to-Digital Converter for Sensor Array Microsystems," IEEE Journal of
Solid-State Circuits, vol. 44, pp. 2844-2855, 2009.
[98] R. Hogervorst, J. P. Tero, R. G. H. Eschauzier, and J. H. Huijsing, "A compact power-efficient
3 V CMOS rail-to-rail input/output operational amplifier for VLSI cell libraries," IEEE
Journal of Solid-State Circuits, vol. 29, pp. 1505-1513, 1994.
[99] K. Dongsoo, B. Goldstein, T. Wei, F. J. Sigworth, and E. Culurciello, "Noise Analysis and
Performance Comparison of Low Current Measurement Systems for Biomedical
Applications," IEEE Transactions on Biomedical Circuits and Systems, vol. 7, pp. 52-62,
2013.
[100] C. C. Enz and G. C. Temes, "Circuit techniques for reducing the effects of op-amp
imperfections: autozeroing, correlated double sampling, and chopper stabilization," IEEE
Proceedings, vol. 84, pp. 1584-1614, 1996.

133

[101] L. Lin, L. Xiaowen, and A. J. Mason, "Die-level photolithography and etchless parylene
packaging processes for on-CMOS electrochemical biosensors," in IEEE International
Symposium on Circuits and Systems, 2012, pp. 2401-2404.
[102] X. Mu, Z. Wang, X. Zeng, and A. J. Mason, "A Robust Flexible Electrochemical Gas Sensor
Using Room Temperature Ionic Liquid," IEEE Sensors Journal, vol. PP, pp. 1-1, 2013.
[103] X. Mu, Z. Wang, M. Guo, X. Zeng, and A. J. Mason, "Fabrication of a miniaturized room
temperature ionic liquid gas sensor for human health and safety monitoring," in IEEE
Biomedical Circuits and Systems Conference, 2012, pp. 140-143.

134