microtechnology and mems
microtechnology and mems Series Editor: H. Baltes
H. Fujita
D. Liepmann
The series Microtechnology and MEMS comprises text books, monographs, and state-of-the-art reports in the very active field of microsystems and microtechnology. Written by leading physicists and engineers, the books describe the basic science, device design, and applications. They will appeal to researchers, engineers, and advanced students. Mechanical Microsensors By M. Elwenspoek and R. Wiegerink CMOS Cantilever Sensor Systems Atomic Force Microscopy and Gas Sensing Applications By D. Lange, O. Brand, and H. Baltes Micromachines as Tools for Nanotechnology Editor: H. Fujita Modelling of Microfabrication Systems By R. Nassar and W. Dai Laser Diode Microsystems By H. Zappe Silicon Microchannel Heat Sinks Theories and Phenomena By L. Zhang, K.E. Goodson, and T.W. Kenny Shape Memory Microactuators By M. Kohl Force Sensors for Microelectronic Packaging Applications By J. Schwizer, M. Mayer and O. Brand Integrated Chemical Microsensor Systems in CMOS Technology By A. Hierlemann
A. Hierlemann
Integrated Chemical Microsensor Systems in CMOS Technology With 125 Figures
123
Professor Dr. Andreas Hierlemann Physical Electronics Laboratory ETH Hoenggerberg, HPT-H 4.2, IQE 8093 Zurich Switzerland Email:
[email protected]
Series Editors: Professor Dr. H. Baltes ETH Zürich, Physical Electronics Laboratory ETH Hoenggerberg, HPT-H6, 8093 Zürich, Switzerland
Professor Dr. Hiroyuki Fujita University of Tokyo, Institute of Industrial Science 4-6-1 Komaba, Meguro-ku, Tokyo 153-8505, Japan
Professor Dr. Dorian Liepmann University of California, Department of Bioengineering 466 Evans Hall, #1762, Berkeley, CA 94720-1762, USA
ISSN 1439-6599 ISBN 3-540-23782-8 Springer Berlin Heidelberg New York Library of Congress Control Number: 2004114045 This work is subject to copyright. All rights are reserved, whether the whole or part of the material is concerned, specif ically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microf ilm or in any other way, and storage in data banks. Duplication of this publication or parts thereof is permitted only under the provisions of the German Copyright Law of September 9, 1965, in its current version, and permission for use must always be obtained from Springer. Violations are liable to prosecution under the German Copyright Law. Springer is a part of Springer Science+Business Media springeronline.com © Springer-Verlag Berlin Heidelberg 2005 Printed in Germany The use of general descriptive names, registered names, trademarks, etc. in this publication does not imply, even in the absence of a specif ic statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. Typesetting: by the authors and TechBooks using a Springer LATEX macro package Cover concept: eStudio Calamar Steinen Cover production: design & production GmbH, Heidelberg Printed on acid-free paper
57/3141/jl - 5 4 3 2 1 0
Preface
This book provides a comprehensive treatment of the very interdisciplinary field of CMOS technology-based chemical microsensor systems. It is, on the one hand, targeted at scientists and engineers interested in getting first insights in the field of chemical sensing since all necessary fundamental knowledge is included. On the other hand, it also addresses experts in the field since it provides detailed information on all important issues related to realizing chemical microsensors and, specifically, chemical microsensors in CMOS technology. A large number of sensor and integrated-sensor-system implementations illustrate the current state of the art and help to identify the possibilities for future developments. Since microsensors produce “microsignals”, sensor miniaturization without sensor integration is in many cases prone to failure. This book will help to reveal the benefits of using integrated electronics and CMOS-technology for developing chemical microsensor systems and, in particular, the advantages that result from realizing monolithically integrated sensor systems comprising transducers and associated circuitry on a single chip. After a brief introduction, the fundamentals of chemical sensing are laid out, including a short excursion into the related thermodynamics and kinetics. Fabrication and processing steps that are commonly used in semiconductor industry are then abstracted. These more fundamental sections are followed by a short description of microfabrication techniques and the CMOS substrate and materials. Thereafter, a comprehensive overview of semiconductorbased and CMOS-based transducer structures for chemical sensors is given. The corresponding chemically sensitive materials and the related applications are mentioned in the context of each transducer structure. CMOS-technology is then introduced as platform technology, which allows the fabrication of microtransducers and, moreover, enables the integration of these microtransducers with the necessary driving and signal conditioning circuitry on the same chip. Several examples such as microcapacitors, microcalorimeters, microcantilevers, and microhotplates are described in great detail. In a next step, the development of monolithic multisensor arrays and fully developed microsystems with on-chip sensor control and standard interfaces is depicted. A short section on packaging shows that techniques from the semiconductor industry can also be applied to chemical microsensor packaging. The book
VI
Preface
concludes with a short outlook to future developments such as developing more complex integrated microsensor systems and interfacing biological materials such as cells with CMOS microelectronics. As with all interdisciplinary efforts, teamwork plays a central role in being successful. Therefore I am particularly grateful to many colleagues and former students, who contributed much to the work that is the topic of this book. I would like to thank Prof. Henry Baltes for giving me the opportunity and the support to enter in the field of CMOS-based sensors in his laboratory. I very much appreciated his continual interest in discovering new things and exploring new fields of science. I am also very grateful to Prof. Oliver Brand, who was always a valuable source of information on microtechnology and microfabrication. I am very much obliged to several highly motivated and excellent coworkers, whose work is amply cited in this book: Christoph Hagleitner and Kay-Uwe Kirstein, the chief circuit designers, the microhotplate group: Markus Graf, Diego Barrettino, Stefano Taschini, Urs Frey, and Martin Zimmermann, the guys working on cantilevers: Dirk Lange, Cyril Vancura, Yue Li, Jan Lichtenberg, the capacitor freaks: Andreas Koll, Adrian Kummer, the microcalorimeter people: Nicole Kerness and Petra Kurzawski, and, finally, Wan Ho Song, who did the microsensor packaging. In the outlook some first results on the combination of microelectronics and cells are mentioned. These rely on the work of Flavio Heer, Wendy Franks, Sadik Hafizovic, Robert Sunier, and Frauke Greve. I am very grateful for all their efforts, and I am looking forward to exciting new results in this research area. I am also indebted to European collaboration partners, Udo Weimar and Nicolae Barsan, University of T¨ ubingen, and to AppliedSensor GmbH, Reutlingen, who provided many of the chemically sensitive materials such as the metal oxides. The fruitful collaboration with Sensirion AG, Z¨ urich, namely Felix Mayer and Mark Hornung, is also gratefully acknowledged. Financial support for the CMOS chemical-sensor projects came from the European Union (FP5, FP6, IST-program), the Swiss Bundesamt f¨ ur Bildung und Wissenschaft (BBW), the Swiss Commission for Technology and Innovation (CTI), and the K¨ orber Foundation, Hamburg, Germany. Zurich, September 2004
Andreas Hierlemann
Contents
1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1
2
Fundamentals of Chemical Sensing . . . . . . . . . . . . . . . . . . . . . . .
9
3
Microtechnology for Chemical Sensors . . . . . . . . . . . . . . . . . . . . 3.1 Microtechnology Substrate Materials . . . . . . . . . . . . . . . . . . . . . . 3.2 Fundamental Semiconductor Processing Steps . . . . . . . . . . . . . . 3.2.1 Deposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.2 Patterning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.3 Etching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.4 Doping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3 CMOS Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.4 Microfabrication for Chemical Sensors . . . . . . . . . . . . . . . . . . . . . 3.4.1 Micromachining for Chemical Microsensors . . . . . . . . . . 3.4.1.1 Bulk Micromachining . . . . . . . . . . . . . . . . . . . . . 3.4.1.2 Surface Micromachining . . . . . . . . . . . . . . . . . . . 3.4.2 Wafer Bonding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.4.3 Sensitive-Layer Deposition . . . . . . . . . . . . . . . . . . . . . . . . .
15 16 16 17 18 19 20 20 22 22 23 25 26 26
4
Microfabricated Chemical Sensors . . . . . . . . . . . . . . . . . . . . . . . . 4.1 Chemomechanical Sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1.1 Rayleigh SAW Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1.2 Flexural-Plate-Wave or Lamb-Wave Devices . . . . . . . . . 4.1.3 Resonating Cantilevers . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2 Thermal Sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.1 Catalytic Thermal Sensors (Pellistors) . . . . . . . . . . . . . . 4.2.2 Thermoelectric or Seebeck-Effect Sensors . . . . . . . . . . . 4.3 Optical Sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.1 Integrated Optics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.2 Microspectrometers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.2.1 Fabry-Perot-Type Structures . . . . . . . . . . . . . . . 4.3.2.2 Grating-Type Structures . . . . . . . . . . . . . . . . . . . 4.3.3 Bioluminescent Bioreporter Integrated Circuits (BBIC) 4.3.4 Surface Plasmon Resonance (SPR) Devices . . . . . . . . . . 4.4 Electrochemical Sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
29 29 32 35 37 39 40 43 45 49 53 53 54 55 57 59
VIII
Contents
4.4.1 Voltammetric Sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4.2 Potentiometric Sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4.2.1 Electrochemical Cell . . . . . . . . . . . . . . . . . . . . . . 4.4.2.2 Field-Effect-Based Devices . . . . . . . . . . . . . . . . 4.4.2.2.1 MOS Field-Effect Transistors, MOSFETs, and Ion-Selective FieldEffect Transistors, ISFETs (Chemotransistors) . . . . . . . . . . . . . . 4.4.2.2.2 MOS Diode and Ion-Controlled Diode, ICD (Chemodiodes) . . . . . . . 4.4.2.2.3 MOS Capacitor and Ion-Selective Capacitor (Chemocapacitors) . . . . . . 4.4.2.2.4 Measuring Work Functions: Kelvin Probe and Suspended-Gate Field-Effect Transistor, SGFET . . . . . . . . . . . . . . 4.4.2.2.5 Light-Addressable Potentiometric Sensor, LAPS . . . . . . . . . . . . . . . . . . . 4.4.2.2.6 Field-Effect Device Fabrication . . . . 4.4.3 Conductometric Sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4.3.1 Chemoresistors . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4.3.1.1 Low-Temperature Chemoresistors . . 4.4.3.1.2 High-Temperature Chemoresistors (Hotplate Sensors) . . . . . . . . . . . . . . . 4.4.3.2 Chemocapacitors . . . . . . . . . . . . . . . . . . . . . . . . . 5
CMOS Platform Technology for Chemical Sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1 CMOS Capacitive Microsystems . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1.1 CMOS Capacitive Transducer . . . . . . . . . . . . . . . . . . . . . . 5.1.2 On-Chip Circuitry of the Capacitive Microsystem . . . . 5.1.3 Capacitive Gas Sensing . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1.3.1 Selectivity Through Sensitive Layer Thickness 5.1.3.2 Insensitivity to Low-ε Analytes . . . . . . . . . . . . . 5.1.3.3 Humidity Interference . . . . . . . . . . . . . . . . . . . . . 5.2 CMOS Calorimetric Device . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2.1 CMOS Calorimetric Transducer . . . . . . . . . . . . . . . . . . . . 5.2.2 Calorimeter Circuitry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2.3 Calorimetric Gas Sensing . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3 CMOS Integrated Resonant Cantilever . . . . . . . . . . . . . . . . . . . . 5.3.1 Resonant Cantilever Transducers . . . . . . . . . . . . . . . . . . . 5.3.1.1 Thermal Actuation . . . . . . . . . . . . . . . . . . . . . . . 5.3.1.2 Magnetic Actuation . . . . . . . . . . . . . . . . . . . . . . . 5.3.1.3 Vibration Detection . . . . . . . . . . . . . . . . . . . . . . .
60 64 64 66
67 71 72
73 75 76 76 77 78 80 83
85 88 88 90 92 94 97 98 100 100 104 105 109 110 110 113 114
Contents
5.3.1.4 Cantilever Temperature . . . . . . . . . . . . . . . . . . . 5.3.2 Microcantilever Circuitry . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3.2.1 Thermal Actuation . . . . . . . . . . . . . . . . . . . . . . . 5.3.2.2 Magnetic Actuation . . . . . . . . . . . . . . . . . . . . . . . 5.3.3 Microcantilevers as Chemical Sensors . . . . . . . . . . . . . . . 5.3.3.1 Polymer Coating . . . . . . . . . . . . . . . . . . . . . . . . . 5.3.3.2 Analyte Absorption . . . . . . . . . . . . . . . . . . . . . . . 5.3.4 Comparison of Cantilevers to Other Mass-Sensitive Devices . . . . . . . . . . . . . . . . . . . . 5.4 CMOS Microhotplate System Development . . . . . . . . . . . . . . . . 5.4.1 CMOS Microhotplates . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.4.1.1 Temperature Sensor Calibration . . . . . . . . . . . . 5.4.1.2 Thermal Microhotplate Modeling and Characterization . . . . . . . . . . . . . . . . . . . . . . 5.4.1.3 Microhotplate Heaters: Resistor and Transistor . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.4.1.4 Microhotplate Sensor Fabrication . . . . . . . . . . . 5.4.2 Hotplate-Based CMOS Monolithic Microsystems . . . . . 5.4.2.1 Analog Hotplate Microsystem . . . . . . . . . . . . . . 5.4.2.2 Analog/Digital Hotplate Microsystem . . . . . . . 5.4.2.3 Digital Hotplate Array Microsystem . . . . . . . . . 5.5 CMOS Chemical Multisensor Systems . . . . . . . . . . . . . . . . . . . . . 5.5.1 CMOS Multiparameter Biochemical Microsystem . . . . . 5.5.2 CMOS Gas-Phase Multisensor System . . . . . . . . . . . . . . 5.5.2.1 Multisystem Architecture . . . . . . . . . . . . . . . . . . 5.5.2.2 Multisystem Circuitry Components, Design and Fabrication . . . . . . . . . . . . . . . . . . . . . . . . . . 5.5.2.3 Multisystem Gas Sensor Measurements . . . . . . 5.5.2.4 Multisystem Applications and Operation Modes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.6 CMOS Chemical Microsensor and System Packaging . . . . . . . . 5.6.1 Simple Epoxy-Based Package . . . . . . . . . . . . . . . . . . . . . . 5.6.2 Chip-on-Board Package . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.6.3 Flip-Chip Package . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
IX
115 116 116 118 122 122 124 130 133 133 139 139 142 145 148 148 153 159 163 164 165 166 168 174 177 181 181 182 184
Outlook and Future Developments . . . . . . . . . . . . . . . . . . . . . . . . 187
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193 Abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219 Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
1 Introduction
The detection of molecules or chemical compounds is a general analytical task in the efforts of chemists to obtain qualitative and/or quantitative time- and spatially resolved information on specific chemical components [1]. Examples of qualitative information include the presence or absence of certain odorant, toxic, carcinogenic or hazardous compounds. Examples of quantitative information include concentrations, activities, or partial pressures of such specific compounds exceeding, e.g., a certain threshold-limited value (TLV), or the lower explosive limits (LEL) of combustible gases. All this information can, in principle, be obtained from either a chemical analysis system or alternatively by using chemical sensors. In both cases sampling, sample pretreatment, separation of the components and data treatment are the tasks to be fulfilled. The main components of a state-of-the-art chemical analysis or sensor system are depicted schematically in Fig. 1.1.
cA yi
xi
x1 x2 x3 . . .
comparison with calibration data
cA
result: chemical composition (quantitative or qualitative)
pattern recognition multicomponent analysis
feature extraction
data pretreatment (sensor electronics)
detection (chemical sensors )
sample conditioning (catalyst, enzyme...)
filtering (membrane)
sample uptake
sample (gas, liquid...)
Fig. 1.1. Components of a chemical analysis or sensor system. Adapted from [15]
2
1 Introduction
It is not easy to clearly distinguish between a chemical sensor and a complex analytical system. Integrated or miniaturized chromatographs or spectrometers may be denoted chemical sensors as well. However, a typical chemical sensor is, in most cases, a cheaper, smaller, and less complex device as compared to miniaturized analytical systems. A draft of the IUPAC (International Union of Pure and Applied Chemistry) provides a definition of a chemical sensor [2]: “A chemical sensor is a device that transforms chemical information, ranging from the concentration of a specific sample component to total composition analysis, into an analytically useful signal”. This rather wide definition does not require that the sensor is continuously operating and that the sensing process is reversible. But intermittently operating devices exhibiting irreversible characteristics are usually referred to as dosimeters [3]. In this context it is useful to introduce some important keywords used extensively throughout the chemical sensor literature [1, 4–11]. Reversibility Thermodynamic reversibility, strictly speaking, requires that the sensor measurand is related to a thermodynamic state function. This implies that, e.g., a certain sensor response unequivocally corresponds to a certain analyte concentration (analyte here denotes the chemical compound to be monitored). The sensor signal may not depend on the history of previous exposures or how a certain analyte concentration is reached (no memory effects or hysteresis). More details on fundamental thermodynamics of the chemical sensing process will be given in Chap. 2. Sensitivity and Cross-Sensitivity Sensitivity usually is defined as the slope of the analytical calibration curve, i.e., how largely the change in the sensor signal depends upon a certain change in the analyte concentration. Cross-sensitivity hence refers to the contributions of compounds other than the desired compound to the overall sensor response. Selectivity/Specificity Selectivity or specificity can be defined according to Janata [4] as the ability of a sensor to respond primarily to only one species in the presence of other species (usually denoted as interferants). Limit of Detection and Limit of Determination The limit of detection (LOD) corresponds to a signal equal to k-times the standard deviation of the background noise (i.e., k represents the signal-tonoise ratio) with a typical value of k = 3. Values above the LOD indicate the presence of an analyte, whereas values below LOD indicate that no analyte is detectable. The limit of determination implies qualitative information, i.e., that the signal can be attributed to a specific analyte. This in turn requires more information and, therefore, the limit of determination is always higher than the limit of detection.
1 Introduction
3
Transducer Transducer is derived from Latin “transducere”, which means to “transfer or translate”. Therefore, a device that translates energy from one kind of system (e.g., chemical) to another (e.g., physical) is termed a transducer. Biosensor Biosensors are usually considered a subset of chemical sensors that make use of biological or living material for their sensing function [10, 11]. Since this book covers mostly chemical sensors, there will not be any further diversification into chemo- and biosensors within this work. Using the above definitions, chemical sensors usually consist of a sensitive layer or coating and a transducer. Upon interaction with a chemical species (absorption, chemical reaction, charge transfer etc.), the physicochemical properties of the coating, such as its mass, volume, optical properties or resistance, reversibly change (Fig. 1.2).
analyte molecules
oscillator circuit
frequency signal, ∆f
physical measurand
data recording and processing
microbalance
driving circuitry
(physico-) chemical interaction
polymer
transducer
mass change ∆m
sensitive layer
example:
chemical sensor
Fig. 1.2. Components of a chemical sensor exemplified for the mass-sensitive principle
These changes in the sensitive layer are detected by the respective transducer and are translated into an electrical signal such as a frequency, current, or voltage, which is then read out and subjected to further data treatment and processing. In Fig. 1.2, this is exemplified for the mass-sensitive principle. Analyte molecules are absorbed into a coating material (polymer) to an extent governed by intermolecular forces. The change in mass of the polymeric coating in turn causes a shift in the resonance frequency of the transducer, e.g., a quartz microbalance. This frequency shift constitutes the electrical signal that is used in subsequent data processing.
4
1 Introduction
To supply the different needs in chemical sensing, a variety of transducers based on different physical principles has been devised. Following the suggestion of Janata [4, 5], chemical sensors can be classified into four principal categories according to their transduction principles: 1. 2. 3. 4.
Chemomechanical sensors (e.g., mass changes due to bulk absorption) Thermal sensors (e.g., temperature changes through chemical interaction) Optical sensors (e.g., changes of light intensity by absorption) Electrochemical sensors (e.g., changes of potential or resistance through charge transfer)
Each of those four categories of chemical sensors will be treated in great detail in Chap. 4. An overview of more recent literature on chemical sensors with regard to different transduction principles is given in [5]. Various inorganic and organic materials serve as chemically sensitive layers that can be coated onto the different transducers. Typical inorganic materials include metal oxides like tin dioxide (SnO2 ) for monitoring reducing gases such as hydrogen or carbon monoxide, or zirconium dioxide (ZrO2 ) to detect oxygen, nitrogen oxide, and ammonia. Organic layers mostly consisting of polymers such as polysiloxanes or polyurethanes are used to monitor hydrocarbons, halogenated compounds and different toxic volatile organics. A survey of typical chemically sensitive materials and their applications is given in Table 1.1. Further information on the coating materials will be provided, e.g., in the context of the different transducers in Chap. 4. Current research and development work in chemical sensors and sensitive materials evolves in three main directions: 1. Miniaturization and monolithic integration of transducers with electronics and, possibly, auxiliary sensors. 2. Search for highly selective (bio)chemical layer materials (molecular recognition, key-lock-type interactions). 3. Using arrays of sensors exhibiting different partial selectivity (polymers, metal oxides) and developing pattern recognition (odors, aromas) and multicomponent analysis methods (mixtures of gases and liquids). The latter strategy has grown very popular [12–17], especially since compact sensor arrays can presently be fabricated at low costs, and interferants, which are present in almost any practical application, can be handled. Chemical sensors meanwhile have also reached the stage of exploratory use in a variety of industrial and environmental applications, some examples being quality control or on-line process monitoring in the food-industry as well as preliminary tests in the areas of medical practice and personal (workplace) safety [18]. In particular in environmental monitoring, there is an urgent need for low-cost sensor systems detecting various pollutants at trace level.
1 Introduction
5
Table 1.1. Typical sensitive materials and applications Materials metals
Examples
Applications
Pt, Pd, Ni, Ag, Sb, Rh, . . .
inorganic gases like CH4 ,H2 , . . .
ionic compounds
electronic conductors (SnO2 , TiO2 , Ta2 O5 , In2 O3 , AlVO4 , . . . ) mixed conductors (SrTiO3 ,Ga2 O3 , perowskites, . . . ) ionic conductors (ZrO2 , LaF3 , CeO2 , nasicon, . . . )
inorganic gases (CO, NOx , CH4 . . . ) exhaust gases, oxygen, ions in water, . . .
molecular crystals
phthalocyanines (Pcs): PbPc, LuPc2 , . . .
nitrogen dioxide, volatile organics
Langmuir-Blodgett films cage compounds
lipid bilayers, polydiacetylene . . . zeolites, calixarenes, cyclodextrins, crown ethers, cyclophanes, . . .
organic molecules in medical applications, biosensing, . . . water analysis (ions), volatile organics, . . .
polymers
components of biological entities
nonconducting polymers detection of volatile organics, polyurethanes, polysiloxanes, . . . food industry (odor conducting polymers and aroma), environmental polypyrroles, polythiophenes, monitoring in gas nafion, . . . and liquid phase, . . . synthetic entities phospholipids, lipids, HIVepitopes, . . . natural entities enzymes, receptors, proteins, cells, membranes, . . .
medical applications, biosensing, water and blood analysis, pharmascreening, . . .
Key requirements for a successful chemical sensor include: • • • • • • • • • • • •
High sensitivity and low limit of detection (LOD) High selectivity to target analyte and low cross-sensitivity to interferants Short recovery and response times Large dynamic range Reversibility Accuracy, precision and reproducibility of the signal Long-term stability and reliability (self-calibration) Low drift Low temperature dependence or temperature compensation mechanisms Ruggedness Low costs (batch fabrication) and low maintenance Ease of use
6
1 Introduction
Semiconductor technology provides excellent means to effectively realize device miniaturization and to meet some of the chemical-sensor key criteria listed above (low cost, batch fabrication). The rapid development of the integrated-circuit (IC) technology during the past decades has initiated many initiatives to fabricate chemical sensors consisting of a chemically sensitive layer on a signal-transducing silicon chip [19,20]. The earliest types of chemical sensors realized in silicon technology were based on field-effect transistors (FETs) [21, 22]. Reviews of silicon-based sensors (not only chemical sensors) are given in [23–25]. In this context two more keywords have to be introduced here. Integrated Sensor A sensor is denoted an integrated sensor if the chemical sensing operation is based on a direct influence on an electric component (resistor, transistor, capacitor) integrated in silicon or another semiconductor material [11]. Smart or Intelligent Sensor The combination of interface electronics and an integrated sensor on a single chip results in a so-called “smart sensor”. At least some basic signal conditioning is usually carried out on chip. One major advantage of smart sensors is the improved signal-to-noise and electromagnetic interference characteristics [11]. In addition the connectivity problem, which occurs especially in multisensor arrays, can be eased by using on-chip multiplexers and by using bus interfaces. For more details on sensor system integration, see Chap. 5. The largely planar integrated-circuit (IC) and chemical-sensor structures processed by combining lithographic, thin film, etching, diffusive and oxidative steps have been recently extended into the third dimension using microfabrication technologies (see Chap. 3 in this book). A variety of micromechanical structures including cantilever beams, suspended membranes, freestanding bridges, gears, rotors, and valves have been produced using micromachining technology (MicroElectroMechanicalSystems MEMS ) [26–29]. MEMS technology thus provides a number of key features, which can serve to enhance the functionality of chemical sensor systems [9, 11, 26, 29–34]. Micromechanical structures (MEMS-structures) and microelectronics can be realized on a single chip allowing for on-chip control and monitoring of the mechanical functions as well as for data preprocessing such as signal amplification, signal conditioning, and data reduction [29–34]. ComplementaryMetal-Oxide-Semiconductor or CMOS-technology is the dominant semiconductor IC technology for microprocessors and Application-Specific Integrated Circuits (ASICs) and has also been used to fabricate integrated chemical microsensors. The use of CMOS technology entails a limited selection of device materials (see Sect. 3.3) and a predefined fabrication process for the CMOS part. Sensor-specific or transducer-specific materials and fabrication steps have to be introduced in most cases as post-processing after the CMOS fabrication.
1 Introduction
7
In the next chapters the fundamentals of the chemical sensing process itself will be laid out (Chap. 2) followed by a short description of microfabrication techniques and the CMOS substrate (Chap. 3). In Chap. 4, there will be an extensive treatment of the different microtransducers that are commonly used for chemical sensors. This transducer overview will be restricted to semiconductor-based and CMOS-based devices and will, for the sake of completeness, also include short abstracts on devices, which are described in much more detail in the subsequent Chap. 5 on the CMOS technology platform for chemical sensors. Chapter 5 will show the evolution from single transducers, which are integrated with the necessary driving and signal conditioning circuitry to monolithic multisensor arrays and fully developed systems with on-chip sensor control and standard interfaces. The concluding Chap. 6 will include a short glance at future developments such as combining cells and CMOS devices to develop biosensors or bioelectric interfaces.
2 Fundamentals of Chemical Sensing
The interaction of a chemical species with a chemical sensor can either be confined to the surface of the sensing layer, or it can take place in the whole volume of the sensitive coating. Surface interaction implies that the species of interest is adsorbed at the surface or interface (gas/solid or liquid/solid) only, whereas volume interaction requires the absorption of the species and a partitioning between sample phase and the bulk of the sensitive material. The different types of chemical interactions involved in a sensing process range from very weak physisorption through rather strong chemisorption to charge transfer and chemical reactions. Physisorption in this context implies that the compound is only physically ab/adsorbed (London or Van-der-Waals dispersion forces) with an interaction energy of 0–30 kJ/mol, whereas in the case of the much stronger chemisorption (interaction energy >120 kJ/mol), the particles stick to the surface by forming a chemical (usually covalent) bond. Charge transfer and chemical reactions involve, in most cases, interaction energies comparable to those of chemisorption and higher. Some of the most common interaction mechanisms and associated energies are listed in Table 2.1, for further details, see [35]. Table 2.1. Typical intermolecular interactions and energies Interaction type covalent bond ion-ion coordination, complexation, charge-transfer bonding
Typical energy [kJ/mol]
Comment
120–800
chemical reaction
250
only between ions
8–200
weak “chemical” interaction
ion-dipole
15
hydrogen bond
20
hydrogen bond: A–H· · ·B
dipole-dipole
0.3–2
between polar molecules
London dispersion (induced dipole-induced dipole)
0.1–2
physical interaction between any molecules
10
2 Fundamentals of Chemical Sensing
High chemical selectivity and rapid reversibility place contradictory constraints on desired interactions between chemical sensor coating materials and analytes. Low-energy, perfectly reversible (physisorptive) interactions generally lack high selectivity, while chemisorptive processes, the strongest of which result in the formation of new chemical bonds, offer selectivity, but are inherently less reversible. A practicable compromise has to be achieved with due regard to the specific application. In this context it should be noted that the commonly accepted limit of reversibility up to 20 kJ/mol refers to room temperature and will not apply to the case of, e.g., tin-dioxide–coated semiconductor sensors operated at 300◦ C to 400◦ C. On the other hand, spontaneous chemical reactions occurring at room temperature often require a tedious regeneration of, e.g., biological recognition units (enzymes). Any interaction between a coating material and an analyte is governed by chemical thermodynamics and kinetics. Thus, a fundamental thermodynamic function, the Gibbs free energy, G [J], is the most important descriptor in all chemical sensing processes: The direction of spontaneous reactions is always towards lower values of G (minimization of the Gibbs energy). The Gibbs free energy is a state function in the thermodynamic sense, i.e., its value depends only on the current state of the system and is independent of how that state has been prepared. This implies that any chemical (sensing) process described by a Gibbs energy function moves towards a dynamic equilibrium (∆G = 0, G minimal), in which both reactants and products are present but have no tendency to undergo net change. This equilibrium is reversible, i.e., an infinitesimal change in the conditions in opposite directions results in opposite changes in its state. The interaction equilibrium of an analyte, A, with a sensor coating, S, can thus be represented by: →
k
A···S . A+S⇔ ←
(2.1)
k
→
←
Here, k and k denote the rate constants of the forward reaction and the reverse reaction, which will be detailed below. Such equilibrium can be described by an equilibrium constant, K, which relates the activity, a, of reaction products (A· · ·S) to those of the reactants (A and S). This constant is thus a characteristic value for the progression of the reaction (K ≤ 1 : no reaction takes place), its numerical value depends on the system temperature. aA···S and in general : K = ani i . (2.2) K= aA · aS i The index i denotes the chemical substance, ni are the corresponding stoichiometric numbers in the chemical equation. This expression signifies that each activity (or fugacity) is raised to the power equal to its stoichiometric number, and, then, all such terms are multiplied together. Stoichiometric numbers of the products are positive and those of the reactants are negative, i.e., reactants appear as the denominator and reaction products as the
2 Fundamentals of Chemical Sensing
11
numerator. The activity1 , ai (fugacity, fi , for gases), which denotes the effective quantity of compound i participating in, e.g., a chemical reaction, is related to the mole fraction, xi , (partial pressure, p, for gases) of a species via: aA = γA · xA with γI ≤ 1. The activity coefficient, γ i , measures the degree of departure of a components behavior from ideal or ideally dilute behavior. The equilibrium constant, K, is also related to kinetics. For the simple → reaction in (2.1), two kinetic constants can be defined: k for the reaction ←
leading to the product A· · ·S, and k for the reaction in the opposite direction. → ← daA = − k aA aS + k aS···A . dt
(2.3)
K then represents the ratio of those two kinetic constants in equilibrium state. → k (2.4) K= ←. k Both, thermodynamics and kinetics hence affect the progress of any chemical process or reaction. Thermodynamics, namely the Gibbs free energy (minimum) or the equilibrium constant, can tell us the direction of spontaneous change and the composition at the equilibrium state, whereas kinetics tell us, whether a kinetically viable pathway exists for that change to occur, and how fast an equilibrium state will be achieved. Kinetics are important in the context of chemical sensors, since there exist chemical processes, the activation barrier of which is too high to get a reaction going, although the Gibbs free energy of the products would be below that of the reactants. Such effects can be used to advantage in tuning the selectivity of, e.g., catalytic chemical sensors (see, e.g., Sect. 4.4.3.1.2). A chemical potential has been introduced in thermodynamics. The chemical potential shows how the Gibbs energy of a system changes when a portion of a specific chemical compound is added to it or removed from it. The chemical potential of the i-th component, µi , is defined as: ∂G and ∆G = µi dni . (2.5) µi = ∂ni p,T i 1
The activities and activity coefficients used throughout this book are related to mole fractions for simplicity reasons. The standard states include (a) a pure compound or (b) infinite dilution. There exist also activities and activity coefficients that are related to molalities (mol/kg) or concentrations (mol/m3 ). The standard state of molality is 1 mol/kg, that of concentration is 1 mol/liter. The values of the activity coefficients related to molalities or concentrations are significantly different from those for molar fractions. A detailed discussion of this issue can be found, e.g., in Levine, I., Physical Chemistry, 2nd edition, McGraw-Hill 1983, New York, pp. 249–258.
12
2 Fundamentals of Chemical Sensing
Here ni denotes the stoichiometric number or the amount of substance in moles. Again, the stoichiometric numbers of the products are positive and those of the reactants negative. The pressure, p, and the temperature, T , are kept constant. The chemical potential can be expressed in terms of mole fractions, xi , or activities, ai , in liquids, and partial pressures, pi , or fugacities, fi , in the gas phase [35]: µi = µ0i (p, T ) + RT ln ai
or
µi = µ0i (p, T ) + RT ln fi .
(2.6)
µ0i (p, T ) here denotes the chemical potential of an appropriately defined standard state such as, e.g., “infinite dilution” or a “pure compound”; R is the molar gas constant (8.314 J/Kmol) and T denotes the temperature in [K]. So there are two terms in (2.6), a reference term and an activity-dependent term. Plugging the terms of (2.6) into (2.5), the reference terms (µ0i ) can be subsumed into ∆G0 as shown in the following equation: ni µ0i (p, T ) + RT ni ln ai = ∆G0 (p, T ) + RT ln ani i . (2.7) ∆G = i
i
i
Both, ∆G and K are characteristic descriptors for the direction of a chemical reaction. In comparing (2.2) with (2.7), it is evident that in a thermodynamic equilibrium state (∆G = 0) ∆G0 and K are interrelated via the following equation (for details, see [35]): ln K = −
∆G0 . RT
(2.8)
The more negative ∆G0 , the larger is K, or in other words, the higher the chemical potential of the reactants with regard to the products, the larger is the reaction extent, and the more spontaneous will the reaction occur in case that (already discussed) kinetic factors will not upset such predictions. According to the Gibbs fundamental equation, ∆G0 is composed of an enthalpy term, ∆H 0 , representing the reaction heat at constant pressure, and an entropy term, ∆S 0 , representing the degree of “disorder” or, thermodynamically more precise, the number of different ways in which the energy of a system can be achieved by rearranging the atoms or molecules among the states available to them (for details, see [35]): ∆G0 = ∆H 0 − T ∆S 0 .
(2.9)
For spontaneous reactions (∆G0 negative), the entropy increases and/or the enthalpy term is negative, i.e., heat is released during the chemical reaction. In the following, the thermodynamics of three prototype reactions of chemical sensors will be briefly discussed.
2 Fundamentals of Chemical Sensing
13
Simple Adsorption/Absorption At thermodynamic equilibrium state, the free species and the ad/absorbed species are in dynamic equilibrium, i.e., the chemical potentials of a certain compound A in gaseous and polymeric phase are identical: µgas A (p, T ) = polymer 0 (p, T ) (2.6). Absorbing all the constant terms (µi , R, T ) into a sorpµA tion constant, Ksorption , or so-called partition coefficient, the equilibrium state can be described by: asorbed (2.10) Ksorption = Afree . aA The partition coefficient is a dimensionless “enrichment factor” relating, e.g., ) to that in the probed the activity of a compound in the sensing layer (asorbed A ) and also represents a thermodynamic equilibrium gas or liquid phase (afree A constant, which is related to ∆G0 via (2.8). For surface adsorption, it is more common to relate the fractional coverage of the surface, θ, to the concentration of the analyte in the probed phase and to use different types of adsorption isotherm like Langmuir-, Freundlich-, or BET-(Brunauer-Emett-Teller) isotherms [35]. Chemical Reaction In this case, (2.2) can be applied in principle. It has to be modified with regard to the respective reaction mechanism occurring. For a simple reaction like nA A + nB B ↔ nC C + nD D, the equilibrium constant is given in analogy to (2.2): anC · anDD ani i and in particular : K = C . (2.11) K= anAA · anBB i The chemical potentials as defined in (2.6) can be used, and (2.8) holds. As already mentioned, the interaction leading to a true chemical reaction may be too strong to be reversible. Charge Transfer and Electrochemical Reaction For a reaction of type A+ + e− ↔ A, an electrochemical potential has to be introduced. The contribution of an electrical potential to the chemical potential is calculated by noting that the electrical work, We , of adding a charge, z . e (z denotes the number of elementary charges, e), to a region where the potential is φ (φ denotes the Galvani potential, which represents the bulk-to-bulk inner contact potential of two materials and is defined as the difference of the Fermi levels of these two materials), is: We = z · e · φ ; hence, the work per mole is : We = z · F · φ
(2.12)
F here denotes the Faraday constant, 96485 C/mol, which is equivalent to one mole of elementary charges. Consequently, the electrochemical potential is (compare 2.6): (2.13) µi = µ0i (p, T ) + RT ln ai + zF φ .
14
2 Fundamentals of Chemical Sensing
When z = 0 (neutral species), the electrochemical potential is equal to the chemical potential (2.6). Rewriting (2.7) for the electrochemical potentials leads to: ani i + zF · ∆φ . (2.14) ∆G = ∆G0 (p, T ) + RT ln i
In the equilibrium state (∆G = 0), (2.13) can be expressed in terms of K (2.2). By replacing E, the “electromotive force”, for ∆φ and by replacing E 0 , the standard cell potential, for −∆G0 /zF (a positive voltage per convention always corresponds to a negative ∆G: spontaneous reaction), the so-called “Nernst-equation” results: E = E0 −
RT ln K . zF
(2.15)
The Nernst equation now can be used to derive an expression for the potential of any electrochemical cell or, in our case, electrochemical sensor. Electrochemical reactions can be triggered by applying currents or voltages via electrodes to a sensing layer. After this short excursion into thermodynamics and kinetics, the different microfabrication techniques and the fundamentals of CMOS-devices will be detailed in Chap. 3.
3 Microtechnology for Chemical Sensors
Microtechnology and microfabrication processes are used to produce devices with dimensions in the micrometer to millimeter range. Microfabrication processes can be effectively applied to yield a single device or thousands of devices. The so-called “batch processing”, i.e., the fabrication of many devices in parallel, does not only lead to a tremendous cost reduction, but also enables the production of array structures or large device series with minute fabrication tolerances. Microfabrication processes hence significantly differ from conventional machining processes, such as drilling or milling with mechanical tools. Integrated circuit (IC) fabrication processes are the most important microfabrication processes [36–38]. The success of CMOS-technology, which is one of the enabling technologies of the information age, clearly demonstrates the efficiency of microfabrication technologies. Standard processing steps originating from semiconductor technology can be used in combination with dedicated micromachining steps to fabricate three-dimensional mechanical structures, which form the basis for the chemical microsensors detailed in Chap. 4. Key advantages of microfabricated chemical sensors include small device size and sampling volume, the possibility of batch processing, and the reproducibility of transducer/sensor characteristics due to the precise geometric control in the fabrication steps. Microfabrication techniques also can be used to either significantly improve sensor characteristics in comparison to conventionally fabricated devices, or to develop devices with new functionality that cannot be realized in conventional fabrication technology. Microsensor success stories, such as micromachined pressure sensors and accelerometers, show that microfabrication techniques are especially suitable for high-volume applications in, e.g., automotive industry. In high-volume production, the advantage of batch processing is paramount, and the high development and setup costs amortize. This chapter is organized in the following way: Microsystem substrate materials and standard processing steps originating from semiconductor technology are detailed in the first sections, followed by a short introduction to CMOS technology and a description of micromachining and layer-deposition processes that are specific to chemical microsensors and -systems.
16
3 Microtechnology for Chemical Sensors
3.1 Microtechnology Substrate Materials Silicon is the standard substrate material for IC fabrication and, thus, the most common substrate material in microfabrication. It is supplied as single crystal wafers with diameters from 100–300 mm. The use of silicon substrate material enables the co-integration of transducers and circuitry, which is used to advantage, e.g., in realizing CMOS-based microsystems [39]. Besides its favorable electrical properties, single crystal silicon also has excellent physical properties (mechanical strength, thermal conductivity) [40], which enable the design of micromechanical structures. Therefore, silicon is also the most common substrate material for microfabricated chemical and biosensors. A large number of micromachining techniques have been developed to structure silicon substrates (see also Sect. 3.4) [26–29, 41–43]. Glass exhibits attractive dielectric and optical properties. Glass is also supplied in wafer form in different compositions (e.g., quartz, fused silica, and borosilicate glass) and diameters. Since glass is transparent for visible light, it is particularly suited for devices with optical detection principles. Single-crystal quartz with its hexagonal lattice structure is piezoelectric and is, therefore, used, e.g., as substrate material for acoustic-wave devices (see Sect. 4.1). Last but not least, glasses are chemically inert and suitable for high-temperature applications. A number of micromachining techniques, such as isotropic wet etching or anisotropic dry etching, have been developed to structure glass. Ceramics have been extensively used as substrate material for hybrid microelectronics and in microelectronics packaging [44] . The standard material is alumina (Al2 O3 ), other materials include beryllium oxide (BeO) and aluminum nitride (AlN). Their chemical inertness, biocompatibility, and mechanical stability render ceramics a very interesting material for microsystems. Most microfabrication techniques for ceramic materials have been adapted from microelectronics packaging processes. Polymers have been more and more explored over the last years as an inexpensive substrate material. Due to the cost advantage, disposable devices, such as microfluidic arrays or microstructured biosensor assays, are often based on polymers [45]. Special processes, such as hot embossing, injection molding, laser machining, or stereolithography, have been developed to structure polymer materials even in the micrometer to nanometer range [46].
3.2 Fundamental Semiconductor Processing Steps The four basic microfabrication techniques for chemical/biosensors are identical with those used in integrated-circuit fabrication [36–38]: Deposition, patterning, doping and etching. The sequential application of these techniques to build up a device layer by layer is illustrated in Fig. 3.1.
3.2 Fundamental Semiconductor Processing Steps
wafer
17
wafer in
film deposition photolithography
doping
mask set etching
wafer out Fig. 3.1. Flow diagram of an integrated-circuit fabrication process using the four basic microfabrication techniques: Deposition, photolithography, etching, and doping. Adapted from [36]
A thin layer, such as an insulating silicon dioxide film, is deposited on a substrate. A light-sensitive photoresist layer is then deposited on top and is patterned using photolithography. The pattern is then transferred from the photoresist layer to the silicon dioxide layer by an etching process. After removing the remaining photoresist, the next layer is deposited and structured, and so on. Doping of a semiconductor material by ion implantation can be done directly after photolithography or after patterning an implantation mask (e.g., a patterned (sacrificial) silicon dioxide layer). In the following, a brief overview on the four fundamental microfabrication steps will be given. More details can be found in books on semiconductor processing [36–38]. 3.2.1 Deposition The two most common thin-film deposition methods in microfabrication are chemical vapor deposition (CVD), performed at low pressure (LPCVD), atmospheric pressure (APCVD) or plasma-enhanced (PECVD), and physical vapor deposition (PVD), such as sputtering and thermal evaporation. Typical CVD and PVD film thicknesses are a few micrometers. Other techniques include electroplating of metal films and spin- or spray-coating of polymeric films such as photoresist. Both processes can yield film thicknesses from less than 1 µm up to several hundred micrometers.
18
3 Microtechnology for Chemical Sensors
Dielectric layers, predominantly silicon dioxide and silicon nitride, are used as insulating material, as mask material, and for passivation. Silicon dioxide is either thermally grown on top of a silicon surface (thermal oxide) at high temperatures in an oxidation furnace (900◦ –1200◦ C), or it is deposited in a CVD system like silicon nitride. CVD oxides are deposited at temperatures between 300◦ and 900◦ C. Metal layers are used for electrical connections, as leads or electrode material, for resistive temperature sensors (thermistors) or as mirror surfaces. Metals, which are widely used in the microelectronics industry, such as aluminum, titanium, and tungsten, are routinely deposited by sputtering or by electron-beam evaporation. Depending on the application, a large number of other metals, including gold, palladium, platinum, silver or alloys, can be deposited with PVD methods. Whereas aluminum has been the standard metallization in IC-fabrication for many years, the state-of-the-art, sub-0.25 -µm CMOS technologies often feature copper metallizations due to their lower resistivity and higher electromigration resistance as compared to aluminum. Highly-doped polycrystalline silicon (polysilicon) is used as gate material for metal-oxide-semiconductor field-effect transistors (MOSFET), for electrodes and resistors, and as thermoelectric or piezoresistive material. Polysilicon is usually deposited in a LPCVD process using silane (SiH4 ) as gaseous precursor. Polymers such as photoresist are commonly deposited by spin- or spraycoating. Similar techniques are also used to coat chemical sensors with sensitive polymer films [1, 4]. 3.2.2 Patterning Photolithography is the standard process to transfer a pattern, which has been designed with a computer-assisted design (CAD) program, onto a certain material. The process sequence is illustrated in Fig. 3.2 [47]. A mask with the desired pattern is created. The mask is a glass plate with a patterned opaque layer (typically chromium) on the surface. Electron-beam lithography is then used to write the mask pattern from the CAD data. In the photolithographic process, a photoresist layer (photostructurable polymer) is spin-coated onto the material to be patterned. Next, the photoresist layer is exposed to ultraviolet (UV) light through the mask. This step is done in a mask aligner, in which mask and wafer are aligned before the subsequent exposure step is performed. Depending on whether positive or negative photoresist was used, the exposed or the unexposed photoresist areas are removed during the resist development process. The remaining photoresist acts as a protective mask during the etching process, which transfers the pattern onto the underlying material. Patterned photoresist can also be used as mask for a subsequent ion-implantation step. After the etching or ion-implantation step, the remaining photoresist is removed, and the next layer can be deposited and patterned.
3.2 Fundamental Semiconductor Processing Steps
19
thin film substrate apply photoresist
photoresist
expose photoresist
UV mask light
develop photoresist
transfer pattern
Fig. 3.2. Schematic of a photolithographic process sequence for structuring a thinfilm layer [47]
The so-called lift-off technique is a way to pattern a thin film material, which would be difficult to etch. Here, the thin film material is deposited on top of the patterned photoresist layer. In order to avoid a continuous film, the thickness of the deposited film must be less than the resist thickness. By removing the underneath photoresist, the thin film material on top is also removed by “lifting it off”, leaving a structured thin film on the substrate [28, 41, 42]. 3.2.3 Etching The two different categories of etching processes include wet etching using liquid chemicals and dry etching using gas-phase chemistry. Both methods can be either isotropic, i.e., provide the same etch rate in all directions, or anisotropic, i.e., provide different etch rates in different directions (see also Sect. 3.4 Micromachining) [37,38]. The important criteria for selecting a particular etching process encompass the material etch rate, the selectivity to the material to be etched, and the isotropy/anisotropy of the etching process. An overview on various etching chemistries used in microfabrication can be found in [48]. Wet etching is usually isotropic with the important exception of anisotropic silicon wet etching in, e.g., alkaline solution, such as potassium hydroxide (see Fig. 3.4, Sect. 3.4.1.1). Moreover, wet etching typically provides a better etch selectivity for the material to be etched in comparison to accompanying
20
3 Microtechnology for Chemical Sensors
other materials. An example includes wet etching of silicon dioxide using hydrofluoric-acid-based chemistry. SiO2 is isotropically etched in diluted hydrofluoric acid (HF : H2 O) or buffered oxide etch, BOE (HF : NH4 F). Typical etch rates for high-quality (thermally grown) silicon dioxide films are 0.1 µm/min in BOE. Dry etching, however, is often anisotropic, resulting in a better pattern transfer, as mask underetching is avoided (see Fig. 3.4, Sect. 3.4.1.1). Therefore, anisotropic dry etching processes, such as reactive-ion etching (RIE), of thin film materials are very common in the microelectronics industry. In an RIE system, reactive ions are generated using a plasma and are accelerated towards the surface to be etched, thus providing directional etching characteristics. Higher ion energies typically result in more anisotropic etching characteristics, but also lead to reduced etching selectivity. Though a large number of dry etching chemicals and recipes exist, mainly fluorine- or chlorine-based etching chemistry is commonly used [37, 38]. 3.2.4 Doping Doping is used to modify the electrical conductivity of semiconducting materials such as silicon or gallium arsenide [36–38]. It is hence the key process step to fabricate semiconductor devices such as diodes and transistors. In the case of silicon, doping with phosphorus or arsenic yields n-type silicon, whereas ptype silicon results from boron doping. By varying the dopant concentration of n-type silicon from 1014 to 1020 cm−3 , the resistivity at room temperature can be tuned from approximately 40 Ω cm to 7·10−4 Ω cm. Dopant atoms are introduced by either ion implantation or diffusion from a gaseous, liquid, or solid source. Ion implantation allows for introducing precisely defined quantities of dopants into the semiconductor material and is, hence, a key process of microelectronics fabrication. The substrate material, e.g., a silicon wafer, is bombarded with accelerated ionized dopant atoms in an ion implanter. The result is an approximately Gaussian distribution of the dopant atoms in the substrate wafer with a mean penetration depth controlled by the acceleration voltage. A high-temperature process is then used for annealing and activating the dopants in the case of ion implantation or for “driving-in” the dopant atoms until a desired doping profile has been achieved in the case of utilizing diffusion processes [36–38].
3.3 CMOS Technology CMOS is the dominant semiconductor technology for microprocessors, memories and application-specific integrated circuits (ASICs) [36–38]. CMOS-chips generally consist of a substrate, the transistor components, the metal layers and a passivation layer on top. The substrate is a silicon wafer, the thickness of which depends on the wafer size: 525 µm for a four-inch wafer and
3.3 CMOS Technology
21
850 µm for an eight-inch wafer. Implanted in the silicon wafer are doped regions, which form together with two polysilicon layers (e.g., transistor gate regions) and silicon oxide layers the transistor structures as defined during the CMOS process. Up to 8 metal layers consisting of aluminum (down to a feature size of 0.18 µm) or copper (0.13 µm and 0.09 µm CMOS) are used to wire the electronic components and to establish connections to the outside world (bondpads). Intermetal oxide (Si-oxide) layers are used as electrical insulator between the different metal layers. Finally, silicon nitride, silicon oxinitride, or silicon oxide layers passivate the device and protect the electronics (Fig. 3.3). The overall CMOS device is fabricated in a defined sequence of material deposition, doping, lithography and etching steps [36–38]. intermetal passivation oxide nitride metal 2 metal 1
S
D
D
n-well
n+ p+ PMOS
polysilicon
S
gate contact oxide oxide
field oxide
NMOS
p-substrate Fig. 3.3. Cross-section of a n-well, double-metal CMOS-chip. A p-doped silicon wafer substrate that includes n-well implantations hosts the NMOS and PMOS transistors. Two metal layers, in this case aluminum, are used for wiring the electronic components (electrical isolation by intermetal oxide). On top is a silicon nitride passivation as protection layer
Undoped semiconductor materials conduct electricity but not enthusiastically. Semiconductor areas that are doped become conductors of either extra electrons with a negative charge (n-doping, e.g., by phosphorus) or of positive charge carriers (p-doping, by e.g., boron). The denomination “complementary” metal oxide semiconductor is owing to the fact that both, n-channel and p-channel transistors are realized on the same substrate. The substrate is, for example, a lightly p-doped wafer material that exhibits n-doped areas (n-wells), see Fig. 3.3. More recently twin-well processes are used, in which n-wells and p-wells are implanted in the wafer substrate material. Both transistor types, n-channel (NMOS) and p-channel (PMOS) transistors are used to realize logic functions. An exemplary CMOS-chip cross-section (0.8 µm nwell, double-metal CMOS process as used for most of the devices in Chap. 5)
22
3 Microtechnology for Chemical Sensors
is shown in Fig. 3.3. The p-substrate exhibits n-well areas created by implantation. Heavily p-doped structures (p+) in the n-well and heavily n-doped structures (n+) in the p-substrate form transistor source and drain. The transistor gate is made of polysilicon on top of the gate oxide (SiO2 ). By applying an appropriate voltage (positive for NMOS, negative for PMOS) to the gate via the polysilicon, the substrate majority charge carriers (electrons in n-well, holes in p-substrate or p-well) are depleted in the surface area below the gate owing to the field effect, and a conducting channel is formed between source and drain (n-channel in p-substrate or p-well, p-channel in n-well). Variation of the gate voltage modulates the source-drain current. The modulation is continuous within a certain range in analog circuits, and produces only two states, “on” or “off”, in digital circuits. The metal layers (Fig. 3.3: metal 1 and metal 2, aluminum) are used to wire the transistors. Dielectric layers such as gate oxide, field oxide, contact oxide, and intermetal oxide (SiO2 ) serve as electrical insulation between conducting or semiconducting layers. The silicon nitride or other materials (oxinitride, oxide) on top serve as passivation and provide electrical and mechanical/chemical protection of the circuitry. Silicon nitride and, to a lesser extent, silicon oxide (or silicon with a native oxide layer) are durable in case of liquid exposure and are biocompatible, which principally allows for using CMOS chips also with cells or living material. CMOS aluminum, however, is neither stable in liquids or in air at higher temperatures nor is it biocompatible, and, consequently, has to be covered with noble-metal coatings such as gold or platinum for many microsystem applications. For more details on semiconductor technology see, e.g., the standard textbooks of Sze [36–38].
3.4 Microfabrication for Chemical Sensors 3.4.1 Micromachining for Chemical Microsensors Micromachined structures such as membranes and cantilevers are widely used in bio(chemical) sensors. Membranes provide, e.g., the thermal isolation required for thermal chemical sensors, whereas cantilevers can be used as resonant structures for mass-sensitive chemical sensors (see Chap. 4.1.3). In the following, the fundamental micromachining techniques are briefly reviewed. More details on micromachining techniques can be found in dedicated books on microsystem technology [25–29, 41, 42]. A recent review on microfabrication in biology and medicine can be found in [43]. The micromachining techniques are categorized into bulk micromachining [49] and surface micromachining processes [50] (see Fig. 3.4). In the case of bulk micromachining, the microstructure is formed by machining the
3.4 Microfabrication for Chemical Sensors
23
relatively thick bulk substrate material, whereas in the case of surface micromachining, the microstructure consists of thin-film layers, which are deposited on top of a substrate, and which are selectively removed in a defined sequence to yield the MEMS structure. 3.4.1.1 Bulk Micromachining One approach to enhance the functionality of IC-based devices includes micromachining the bulk substrate, which, in most cases, consists of silicon. Bulk micromachining techniques can be classified into isotropic and anisotropic etching techniques (structure geometry), or into wet and dry etching techniques (reactant phase: liquid or gaseous) [25–29, 41, 42, 49, 50]. In the case of isotropic etching, the same etch rate applies to all directions (Fig. 3.4a), whereas in the case of anisotropic etching, the substrate is preferentially etched away along certain crystal planes while it is preserved in other directions (Fig. 3.4a).
(a) bulk micromachining isotropic etching
(b) surface micromachining substrate (Si)
sacrificial layer
structural layer
microstructure
substrate (Si) anisotropic etching
substrate (Si) Fig. 3.4. Micromachining techniques: (a) Bulk micromachining, anisotropic and isotropic etching, (b) surface micromachining with sacrificial layer, structural layer and a subsequent etch step
The most common isotropic wet silicon etchant is HNA, a mixture of hydrofluoric acid (HF), nitric acid (HNO3 ), and acetic acid (CH3 COOH): Nitric acid oxidizes the silicon surface, and hydrofluoric acid etches the grown silicon dioxide layer. The acetic acid controls the dissociation of HNO3 , which provides the oxidation of the silicon. The etch rates and the resulting surface quality strongly depend on the chemical composition [41, 42, 49]. Anisotropic wet etching of silicon is the most common micromachining technique and is used to release, e.g., membranes and cantilevers for chemical and biosensors. Anisotropic silicon etchants etch single-crystal silicon at
24
3 Microtechnology for Chemical Sensors
different etch rates in distinct crystal directions. The etch grooves are limited by crystal planes, along which etching proceeds at slowest speed, i.e., the 111 planes of silicon. In case of 100 silicon wafers, the 111 planes are intersecting the wafer surface at an angle of 54.7◦ , so that the typical pyramid-shape etch grooves as shown in Fig. 3.4 are formed. Mask materials for anisotropic silicon etchants are silicon dioxide and silicon nitride. It is important to note, that “convex” corners of the etch mask are underetched in case of 100 silicon substrates, leading to, e.g., completely underetched cantilever structures. The etch rates in preferentially etched crystal directions such as the 100 and the 110 direction, and the ratio of the etching rates in different crystal directions strongly depend on the exact chemical composition of the etching solution and the process temperature [51]. The most common anisotropic silicon etching solution is potassium hydroxide, KOH. As an example, a 6-molar KOH solution at 95◦ C provides a 100 etch rate of 150 µm/hour and an anisotropy between 100 and 111 direction etching of 30–100:1 [52]. Since the etch rate of silicon dioxide in KOH solution is rather high (for thermal oxide approx. 1 µm/hour in 6-molar KOH solution [42]), silicon nitride films are often used as etching mask. KOH solution is very stable, yields reproducible etching results, is relatively inexpensive, and is, therefore, the most common anisotropic wet etching chemical in industrial manufacturing. The disadvantages of KOH include the relatively high SiO2 - and Al-etch rates, which require a protection of, e.g., integratedcircuit structures during etching. Etching with KOH is typically performed from the back side of the wafer, with the front side protected by a mechanical cover and/or a protective film [52]. Alternative silicon wet etchants are ammonium hydroxide compounds, such as tetramethyl ammonium hydroxide (TMAH), and ethylene diamine/ pyrochatechol (EDP) solutions. Some EDP formulations, such as EDP type S, exhibit relatively low Al- and SiO2 -etch rates, which render them suitable for releasing microstructures on the front side of CMOS-wafers [53]. More detailed discussions of wet etching of silicon can be found, e.g., in [26] and [42]. Reliable etch stop techniques are very important for achieving reproducible etching results. As already mentioned, wet anisotropic silicon etchants “stop” etching, i.e., the etch rate is reduced by at least 1–2 orders of magnitude, as soon as a 111 silicon plane or a silicon dioxide/nitride layer is reached. In addition, the etch rate is greatly reduced in highly boron-doped regions (doping concentration ≥1020 cm−3 ). The etching can also be stopped at a p-n-junction using a so-called electrochemical etch stop technique (ECE) [41]. This method has been extensively used to release silicon membranes and n-well structures (see, e.g., Chap. 5.4). ECE relies on the passivation of silicon surfaces through application of a sufficiently high anodic potential with respect to the potential of the etching solution. Isotropic dry etching of silicon is done with xenon difluoride, XeF2 . This vapor-phase etching method exhibits excellent etch selectivity with respect
3.4 Microfabrication for Chemical Sensors
25
to aluminum, silicon dioxide, silicon nitride, and photoresist, all of which can be used as etch masks. The XeF2 silicon etch rates depend on the loading (size of the overall silicon surface exposed to the etchant) with typical values of approx. 1µm/min [54]. Anisotropic dry etching of silicon is usually carried out as reactiveion-etching (RIE) with plasma-assisted etching systems. By controlling the process parameters, such as process gases and process pressure, the etching can be rendered either isotropic or anisotropic. The dry-etching anisotropy originates from experimental parameters such as the direction of the ion bombardment, and is, therefore, independent of the crystal orientation of the substrate material. Most bulk etching of silicon is accomplished using fluorine free radicals with SF6 as a typical process gas. Adding chlorofluorocarbons results in polymer deposition in parallel with etching, which leads to enhanced anisotropy. Very-high-aspect-ratio microstructures can be achieved with deep (D)RIE , a method, which has gained importance during the last years. Deep-RIE systems rely on high-density plasma sources and an alternation of etching and polymer-assisted sidewall protection steps. In a process known as the “Bosch” process [55], a mixture of trifluoromethane and argon is used for polymer deposition. Due to the ion bombardment, the polymer deposition on the horizontal surfaces can almost be prevented, while the sidewalls are passivated with a TeflonTM -like polymer. In the second process step, an SF6 -based etching chemistry provides silicon etching in the non-passivated regions, i.e., the horizontal surfaces. Both process steps are alternated, resulting in typical silicon etch rates of 2–3 µm/min with an anisotropy on the order of 30:1 [49]. Silicon dioxide and photoresist layers can be used a etch masks. Even though an exceptional anisotropy can be achieved, which is independent of the crystal orientation, one should keep in mind that deep RIE systems are by far more expensive than a simple wet-etching setup, and that only one wafer is processed at a time [56, 57]. 3.4.1.2 Surface Micromachining Surface micromachining comprises a number of techniques to produce microstructures from thin films previously deposited onto a substrate and is based on a sacrificial-layer method (Fig. 3.4b). In contrast to bulk micromachining, surface micromachining leaves the substrate intact [41,50]. A sacrificial layer is deposited and patterned on a substrate. After that, a structural thin film, in most cases polysilicon, is deposited and patterned, which will perform the mechanical or electrical functions in the final device. A selective etchant then removes exclusively the sacrificial-layer material. The thickness of the sacrificial layer determines the distance of the structural parts from the substrate surface. Common sacrificial-layer materials include silicon oxide etched by hydrogen fluoride and aluminum etched by a mixture of phosphoric, nitric and acetic acid.
26
3 Microtechnology for Chemical Sensors
Clamped beams, microbridges, or microchannels can be fabricated this way, microrotors and even microgears can be realized by repeated layer deposition and etching [25–29, 41]. 3.4.2 Wafer Bonding While the majority of microstructures are fabricated from a single substrate or wafer, several wafers can be joined by wafer bonding [58]. The substrates/wafers are bonded onto each other either directly or via an intermediate layer. Direct bonding techniques include silicon fusion bonding and anodic bonding. Silicon fusion bonding involves two silicon wafers, which are bonded to each other at high temperatures (T≥ 1000◦ C). Low-temperature (T < 400◦ C) fusion bonding has been demonstrated after using special cleaning procedures, but exhibits reduced bonding strength. Anodic bonding of a sodiumrich glass wafer onto a silicon wafer is accomplished by applying an electric field across the bonding interface at moderate bonding temperatures (T ≈ 300–400◦ C). Glass materials with a thermal expansion coefficient similar to that of silicon (e.g., Pyrex glass 7740) are used for anodic bonding in order to minimize thermomechanical stress. Anodic bonding and silicon fusion bonding require very clean and smooth wafer surfaces to achieve void-free bonding. The surface quality and roughness is less important if an intermediate layer is used for wafer bonding. Possible intermediate bonding layers include adhesives, low-melting-temperature glass, solder films, or metallic films such as gold. Examples will be given in Sects. 4.3.2.2 and 4.3.4. 3.4.3 Sensitive-Layer Deposition The set of microfabrication processes used for chemical/biosensors is completed by various deposition techniques for chemically or biologically sensitive layers. Chemically sensitive polymer layers or organic molecules that are used for the detection of volatile organics in air can be deposited by, e.g., dispensing, spray coating or by using self-assembled monolayers (SAMs). Metal-oxide films for the detection of, e.g., carbon monoxide and nitrogen oxides can be deposited either by a sol/gel process, by drop coating, or by sputtering [59]. Recently, microcontact printing or soft lithography [60] has been introduced as an additional method for pattern transfer. A soft polymeric stamp is used to reproduce a desired pattern directly on a substrate. Feature sizes on the order of 1 µm can be routinely achieved with this technique. The polymer stamp, often made from poly(dimethylsiloxane) (PDMS), is formed by a molding process using, e.g., a silicon master fabricated with conventional microfabrication techniques. After “inking” the stamp with the material to be printed, the stamp is brought in contact with the substrate material, and the
3.4 Microfabrication for Chemical Sensors
27
pattern of the stamp is reproduced. Surface properties of the substrate thus can be modified to, e.g., locally promote or prevent molecule adhesion. Soft lithography has been specifically developed for biological applications such as patterning cells or proteins with the help of, e.g., self-assembled monolayers (SAM) [60].
4 Microfabricated Chemical Sensors1
Chemical sensors usually consist of a sensitive layer or coating and a transducer (see Chap. 1, Fig. 1.2) [1,4–9]. Upon interaction with a chemical species (absorption, chemical reaction, charge transfer etc.), the physicochemical properties of the coating, such as its mass, volume, optical properties or resistance reversibly change. These changes in the sensitive layer properties can be detected by a variety of transducers and can be translated into electrical signals such as frequency, current, or voltage changes, which are then subjected to further data treatment and processing. As already mentioned in the introduction to this book, chemical sensors can be classified into four principal categories according to their transduction principles [4, 5]: (a) chemomechanical sensors (b) thermal sensors (c) optical sensors, and (d) electrochemical sensors. Each of those four sensor categories will be briefly introduced, and, then, specific exemplary microfabricated transducers and devices will be abstracted. The transducer overview will be restricted to semiconductor-based and CMOS-based devices and will, for the sake of completeness, also include short abstracts on devices, which are described in much more detail in the subsequent Chap. 5 on the CMOS technology platform for chemical sensors. Typical chemically sensitive materials and sensor applications will be brought up in the context of the respective transducer structures.
4.1 Chemomechanical Sensors The change in mechanical properties (e.g., mass) of a sensitive layer upon interaction with an analyte can be conveniently recorded by using micromechanical structures. Any species that can be immobilized on the sensor can, in principle, be sensed. As with most of the chemical sensors (excluding thermal sensors), the measurements are performed at a thermodynamic equilibrium state (2.1–2.4, 2.10), which is defined by the Gibbs energy minimum 1
Large parts of the material of Chap. 4 were originally published in the book: MEMS: A Practical Guide to Design, Analysis, and Applications, edited by Jan Korvink and Oliver Paul, William Andrew Publishing, Norwich, NY, 2005. Reprinted here with permission.
30
4 Microfabricated Chemical Sensors
of the system. In the simplest case such chemomechanical sensors are gravimetric sensors responding to the mass of species accumulated in a sensing layer [61–63]. Some of the sensor devices additionally respond to changes in a variety of other mechanical properties of solid or fluid media in contact with their surface such as polymer moduli, liquid density and viscosity [61–63], which will not be discussed here. The high sensitivity of gravimetric sensors provides good chemical sensitivity: mass changes in the picogram range can be detected, and ppm (parts per million) to ppb (parts per billion) detection levels have been reported for, e.g., gas and vapor sensors [61–63]. The large number of chemical species that can be present in the environment, and the difficulty in selectively and, at the same time, reversibly sorbing these species on the sensor, however, makes specific detection difficult. Most of the gravimetric sensors rely on piezoelectric materials such as quartz, lithium tantalate or niobate, aluminum nitride, zinc oxide and others. Piezoelectricity results in general from coupling of electrical and mechanical effects. The prerequisite is an anisotropic, noncentrosymmetric crystal lattice. Upon mechanical stress, charged particles are displaced and thus generate a measurable electric charge in the crystal. In turn, mechanical deformations can be achieved by applying a voltage to such a crystal (for details, see [4]). Using an alternating current (AC), the crystals can be electrically excited into a fundamental mechanical resonance mode. The resonance frequency, which is the recorded sensor output in most cases, changes in proportion to the mass loading on the crystal or device. The more mass (analyte molecules) is absorbed, e.g., in a polymer coated onto a piezoelectric substrate or transducer, the lower is the resonance frequency of the device: ∆f = −C f02 ∆m/A .
(4.1)
This equation was published by Sauerbrey in 1959 [64]. ∆f here denotes the frequency shift due to the added mass in [Hz], C is a constant, f0 is the fundamental frequency of the quartz crystal in [Hz] and ∆ m/A is the surface mass loading in [g cm−2 ]. The following equation describes the relationship between analyte gas phase concentration change, ∆c A , and the responses of mass-sensitive sensors: ∆fA = Γ · MA · K · ∆cA .
(4.2)
Here, ∆fA [Hz] denotes the frequency shift (sensor response) measured upon exposure to analyte at a concentration cA [mol/L]. MA [kg/mol] is the molar mass of the analyte vapor, K is the partition coefficient (2.10), and Γ is a gravimetric constant [L/kg·s] including, e.g., the frequency shift measured upon initial deposition of the sensitive layer, the coating density, transducer dimensions, etc. A typical signal of a gravimetric sensor is displayed in Fig. 4.1 showing the frequency shifts of a resonant cantilever coated with poly(etherurethane),
4.1 Chemomechanical Sensors
31
frequency shift [Hz]
250
500
750
1000
1250
1500
1500
1250
750
1000
500
0
250
ppm
20
-20 -40 -60
analyte: n-octane polymer: PEUT
-80 -100
0
50
100 time [min]
150
200
Fig. 4.1. Typical responses of mass-sensitive sensors. Frequency shifts of a polymer-coated (poly(etherurethane), PEUT) cantilever upon exposure to different concentrations2 (250–1500 ppm) of an organic volatile: n-Octane
(PEUT), upon exposure to various concentrations of n-octane. At low analyte concentrations (trace level), a linear correlation between the frequency shift due to analyte absorption and the corresponding analyte concentration in the gas phase is usually observed (Fig. 4.1), provided that the sensing film on the transducer moves synchronously with the oscillating crystal surface. Significant deformations across the film thickness result in a more complex relationship between mass changes and resonant frequency due to, e.g., viscoelastic effects (concept of “acoustically thin and thick” films as detailed in [65]). The most common devices are the thickness-shear-mode resonator (TSMR) or quartz microbalance (QMB), a bulk resonator, and the Rayleigh surfaceacoustic-wave (SAW) device, both based on quartz substrates. The QMB was demonstrated to function as an organic vapor sensor by King in 1964 [66], the SAW device became popular after introducing interdigital transducers to 2
Concentration are usually given in mole-per-volume (mol/m3 ) or mass-pervolume (kg/m3 ) units. In gas sensorics ppm-units (parts per million volume or parts per million pressure) are widely used, which are strictly speaking no valid concentration units, in particular since ppm-units are dimensionless. The ppm-units are nevertheless used here owing to their popularity. Assuming the validity of the ideal-gas law, which holds true for low analyte concentrations, ppm-units can be easily converted into mass-per-volume units, e.g., (µg/L) by division through the molar volume of an ideal gas at 25◦ C (24.45 l) and multiplication with the molar mass of the analyte compound. Moreover, under normal pressure conditions 10 ppm analyte correspond to 1 Pa partial pressure of the respective analyte.
32
4 Microfabricated Chemical Sensors
acoustic sensors in 1970 [67]. Since the TSMR is not semiconductor-based and not compatible with IC technology, it will not be treated here any further. Shear-horizontal-acoustic-plate-mode (SH-APM) devices devices, sheartransverse-wave device (STW) and Love-wave devices devices require quartz, lithium niobate or lithium tantalate substrates [61–63] and, hence, will not be dealt with here as well. For details and further information it is referred to a wealth of literature [61–63, 68–70]. Silicon is not a piezoelectric material. The realization of silicon-based piezoelectric transducers hence requires an additional piezoelectric layer to be patterned on the silicon. Different materials have been used such as cadmium sulfide [71], aluminum nitride [72, 73], and in particular zinc oxide (ZnO) [74–76], which will be subject to further discussion in this chapter. In the following, three semiconductor-technology-compatible types of mass-sensitive devices will be described in more detail: (1) SAW-devices on Sisubstrates with piezoelectric overlay, (2) flexural-plate-wave devices (FPWs), and (3) micromachined cantilevers. Operability in gas or liquid media, typical coating materials, target analytes, and applications will be discussed in the context of each transducer. An overview of micromachined resonant sensors is given in [70, 77]. 4.1.1 Rayleigh SAW Devices Transduction Principle and Sensing Characteristics Interdigital transducers can be used to launch and detect a surface-acoustic wave on a piezoelectric substrate [67] as schematically shown in Fig. 4.2. By applying an AC voltage to a set of interdigital transducers patterned on a piezoelectric substrate with appropriate orientation of the crystal axes, top view
Rayleigh SAW
interdigitated electrodes
excitation propagation metal electrodes
side view detection ZnO
wave propagation
particle displacement
substrate: silicon
Fig. 4.2. Launching, propagation and detection of a Rayleigh-type surface acoustic wave by interdigitated transducers on a zinc-oxide-covered silicon substrate. The top view shows the electrode configuration and the wave propagation. The side view shows the elliptical particle displacement
4.1 Chemomechanical Sensors
33
one set of the fingers moves downwards, the other upwards, thereby creating an oscillating mechanical surface deformation. This surface deformation generates an acoustic wave, which propagates along the surface and is converted back into an electrical signal by deforming the surface in the region of the receiving transducer. The electrical signal of the receiving transducer is recorded and represents the sensor signal. For a given piezoelectric substrate, the acoustic wavelength and, thus, the operating frequency of the SAW is determined by the transducer periodicity, which is equal to the acoustic wavelength at the transducer center frequency. Typical frequencies range between 100 and 500 MHz [61–63]. Such frequencies require a sophisticated high-frequency circuit design. Therefore, a bare reference oscillator is operated together with the sensor in many cases, and the outputs are mixed to produce a difference frequency with values in the kHz-range that is recorded [74, 76]. The acoustic wave is confined to a surface region of approximately one acoustic wavelength thickness. The velocity and damping characteristics of the acoustic wave hence are extremely sensitive to changes at the transducer surface. When used in an oscillator circuit, relative changes in the wave velocity are reflected as equivalent changes in fractional oscillation frequency. A change in mass due to, e.g., absorption of a gaseous analyte in a polymeric sensing layer thus changes the device frequency according to (4.1). The acoustic (Rayleigh) wave causes an elliptical particle movement at the transducer surface (Fig. 4.2), i.e., the sensitive films deposited on top of the transducers and the piezoelectric substrate are severely deformed. Thus, additional effects such as changes in viscoelastic properties of the sensing layer can affect the sensor response [65]. Fabrication Since there exists a variety of custom-designed semiconductor and silicon processes in the literature, only the transducer-related additional fabrication steps after semiconductor processing and the sensitive-layer-processing steps will be listed in this chapter. Industrial standard IC processes like CMOS fabrication will be explicitely identified, especially since CMOS-based devices are extensively treated in Chap. 5. For more details on semiconductor and MEMS process steps, see Chap. 3. Fabrication Steps: • Optional back etching using potassium hydroxide, KOH, or ethylenediamine-pyrocatechol, EDP, to achieve a membrane structure [74, 78]). • Zinc oxide processing: Deposition mainly by sputtering techniques at 150◦ – 450◦ C. Highly oriented layers of 5–50 µm with a high degree of surface flatness [74]. • Electrode processing: Vacuum evaporation of aluminum or gold, layer thickness >200 nm.
34
4 Microfabricated Chemical Sensors
• Sensitive layer: Spin or spray coating of polymers, organic layers, or biological entities. Applications Since surface-normal particle displacements occur (Fig. 4.2), and the acoustic wave velocity is larger than the compressional velocity of sound in water, the device radiates compressional waves into the liquid phase, which causes severe attenuation. Rayleigh SAW devices hence cannot be used in liquids [61–63]. Typical applications areas are environmental monitoring or personal safety devices. This includes the detection of different kinds of organic volatiles (hydrocarbons, chlorinated hydrocarbons, alcohols, etc.) by using polymeric layers [74] or porphyrins [79], and the detection of nitrogen dioxide using phthalocyanines [76]. The interaction mechanisms involve, in most cases, fully reversible physisorption and bulk/gas phase partitioning (see 2.10, 4.2). Integrated Gallium Arsenide (GaAs) SAW Sensor GaAs is a well-developed semiconductor device material for fabricating highfrequency integrated circuits, and GaAs is piezoelectric. The piezoelectric properties of GaAs and, hence, the device characteristics are similar to those of quartz except for the strong temperature dependence. An integrated GaAsSAW sensor is shown in Fig. 4.3 [80]. It consists of a 470 MHz SAW device along with a multistage amplifier (4 gain stages and impedance matching
4-gain-stage amplifier
output stage
470 MHz delay line
Fig. 4.3. Micrograph of a monolithically integrated GaAs surface-acoustic-wave device showing the delay line, the amplifier and the output stage. Reprinted from [80] with permission
4.1 Chemomechanical Sensors
35
output stage) forming a monolithic oscillator circuit thus eliminating the need for high frequency interconnections [80]. 4.1.2 Flexural-Plate-Wave or Lamb-Wave Devices Transduction Principle and Sensing Characteristics Flexural-plate-wave devices have been introduced in 1988 [81]. Their chief advantage is their high sensitivity to added mass at a low operating frequency (typically 3–10 MHz) [82]. FPW devices feature plates that are only a few percent of an acoustic wavelength thick (typically 2–3 µm). The plates are composite structures (Fig. 4.4) consisting of a silicon nitride layer, an aluminum ground plane, a sputtered zinc oxide piezoelectric layer, all supported by a silicon substrate [61, 63, 81, 83]. FPW top view
side view silicon frame
silicon nitride membrane bottom
aluminum
ZnO
composite membrane
metal electrodes
composite membrane wave propagation
particle displacement
Fig. 4.4. Schematic of a flexural-plate-wave device. The side view shows the different layers and the membrane movement. Interdigitated electrodes are used for actuation
The interdigital transducers (IDTs) on these devices generate flexural waves (Lamb waves, Fig. 4.4) with retrograde elliptical particle motions as in the SAW devices. However, the velocity in the membrane is much less than in a solid substrate, and the operating frequency for a given transducer periodicity is, hence, considerably lower [61–63, 70, 81]. The Lamb waves give rise to a series of plate modes, one of which has a frequency that is much lower than those of the other possible modes. The velocity of this unique wave decreases with decreasing plate thickness. The entire thickness of the plate is set in motion like the ripples in a flag [61–63, 70, 81, 83]. The confinement of acoustic energy in the thin membrane results in a very high mass sensitivity. The sensor response (frequency shift) is proportional to the mass loading (4.1). Since the Lamb wave causes an elliptical particle movement at the transducer surface (Fig. 4.4), the sensitive films are deformed as it is the case with
36
4 Microfabricated Chemical Sensors
the SAW. The frequency, however, is much lower, and, therefore, changes in viscoelastic properties of the sensing layer do not severely affect the sensor response. The sensitive layer can be deposited on either side of the membrane. Deposition on the backside (non-processed side of the wafer) has the advantage, that on-chip circuitry will not be exposed to chemicals [81–84]. Transducer Modification: Magnetically Excited FPW Magnetic excitation requires an externally applied magnetic field, but eliminates the need for a piezoelectric layer, which frequently contains elements (Zn, etc.) that pose contamination problems in IC fabrication. The device consists of a silicon nitride membrane suspended in a silicon frame. A metal meander-line transducer is patterned on the membrane surface (Fig. 4.5). Alternating current flowing in the transducer interacts with a static in-plane magnetic field to generate time-varying Lorentz forces (Fig. 4.5). These deform the membrane, exciting it into a resonant mode [70, 85]. To efficiently excite the mode, the current lines of the transducer must be positioned along lines of maximum mode displacement (Fig. 4.5). This requires a critical alignment between the top metallization pattern and the backside etch mask [70, 85]. silicon nitride membrane
I meander-line transducer
FPW
λ silicon substrate B (static in-plane magnetic field)
Fig. 4.5. Schematic representation of the magnetically excited flexural-plate-wave device. Lorentz forces are generated between an impressed alternating current in a serpentine conductor and a static in-plane magnetic field. Reprinted from [85] with permission
Fabrication • Evaporation of Al and Si-nitride (LPCVD) [81]. • Back etching (KOH, or EDP) to achieve a membrane structure [81, 83]. • Zinc oxide or lead zirconate titanate (PZT) [86] processing if necessary (see SAW).
4.1 Chemomechanical Sensors
37
• IDT processing: Vacuum evaporation of aluminum or gold, see SAW. • Sensitive layer: spin or spray coating of polymers, deposition of biological entities. IC process-compatible fabrication sequences for monolithic integration of the Lamb device with electronics are detailed in [83, 87]. Applications Surface-normal particle displacements occur (Fig. 4.4), but the acoustic-wave velocity is much less than the compressional velocity of sound in water. FPW devices thus can be used in the liquid phase [61, 63, 83, 88]. Typical application areas are environmental monitoring (gas and liquid phase) or biosensing in liquids. These include the detection of different organic volatiles in the gas phase (hydrocarbons, chlorinated hydrocarbons, alcohols, etc.) by using polymeric layers [81, 84, 89–91], the detection of the weight percentage of alcohol/water [83] or glycol/water [88] mixtures and the use of an FPW-based immunoassay for the detection of breast cancer antigens [92]. The interaction mechanisms involve reversible physisorption and bulk/gas phase partitioning (see 2.10, 4.2) as well as antigen/antibody binding [92]. 4.1.3 Resonating Cantilevers Transduction Principle and Sensing Characteristics Micromachined cantilevers commonly employed in atomic force microscopy (AFM) constitute a promising type of mass-sensitive transducer for chemical sensors [93–107]. The sensing principle is quite simple. The cantilever is a layered structure composed of, e.g., the dielectric layers of a standard CMOS process, silicon, metallizations, and eventually, add-on piezoelectric zinc oxide. The cantilever base is firmly attached to the silicon support. The freestanding cantilever end is coated with a sensitive layer (Fig. 4.6). The excitation of a cantilever in the resonant mode is usually performed by applying piezoelectric materials (ZnO) [106] or by making use of the bimorph effect, i.e., the different thermal expansion coefficients of the various polymeric coating
silicon frame
cantilever
Fig. 4.6. Schematic representation of a resonating cantilever
38
4 Microfabricated Chemical Sensors
layer materials forming the cantilever [93–105]. This difference in material properties gives rise to a cantilever deflection upon heating. Periodic heating pulses in the cantilever base thus can be used to thermally excite the cantilever in its resonance mode at 10–500 kHz [93–95, 106]. There are two fundamentally different operation methods: (a) static mode: measurement of the cantilever deflection upon analyte-sorption-induced stress changes by means of, e.g., laser-light reflection [100, 103–105], (b) dynamic mode: excitation of the cantilever in its fundamental mode and measurement of the change in resonance frequency upon mass loading [93–95, 101, 102] in analogy to other mass-sensitive devices (4.1, 4.2). These two methods impose completely different constraints on the cantilever design for maximum sensitivity. Method (a) requires long and deformable cantilevers to achieve large deflections, whereas method (b) requires short and stiff cantilevers to achieve high operation frequencies. Method (b) is preferable with regard to integration of electronics and simplicity of the setup (feedback loop) [93–95, 101, 102, 106, 107]. Method (a) can be applied in liquids as well [101, 104], which is rather difficult using the dynamic mode. The detection of the frequency changes can be done by embedding piezoresistors in the cantilever base [93–95, 101, 102], by measuring motional capacitance changes [107], or by using optical detection by means of laser light reflection on the cantilever [97–100, 103–105]. The mass resolution of the cantilevers is in the range of a few picograms [93–96, 103–105]. This high mass-sensitivity does not necessarily imply an exceptionally high sensitivity to analytes since the area coated with the sensitive layer usually is very small (on the order of 100 × 150 µm2 ) [93]. The sensing layer is deformed upon motion of the cantilever; therefore, modulus effects are expected to contribute to the overall signal, especially since the coating thickness may exceed the thickness of the cantilever. Fabrication • • • • •
Eventually additional Al and Si-nitride (LPCVD) [93–105]. Back etching (KOH, or EDP) to achieve a membrane structure [93–107]. Zinc oxide processing if necessary [106]. Release of the cantilevers by front-side reactive-ion etching [93–103]. Sensitive layer: Spray or drop coating of polymers, deposition of biological entities.
IC process- and CMOS compatible fabrication sequences for monolithic integration of the cantilevers with electronics are detailed in [93–95,101,102,107]. Applications The application of the dynamic mode is mostly restricted to the gas phase, whereas the static mode can be used to detect analytes in liquid phase as well. Due to the bimorph effect (cantilever deformation upon heat generation), cantilevers have also been used in microcalorimetric applications [103, 108].
4.2 Thermal Sensors
39
Typical applications are environmental monitoring (gas and liquid phase) or biosensing in liquids. These include the detection of different kinds of organic volatiles (hydrocarbons, chlorinated hydrocarbons, alcohols, etc., see Fig. 4.1) or humidity in the gas phase by using polymeric layers [93–100, 102, 105], the detection of alcohol in water [101], and the hybridization and detection of complementary strands of oligonucleotides [104]. The interaction mechanisms involve reversible physisorption and bulk/gas phase partitioning (see 2.10, 4.2), as well as receptor-ligand binding [104].
4.2 Thermal Sensors Calorimetric or thermal sensors rely on determining the presence or concentration of a chemical species by measurement of an enthalpy change produced by the chemical to be detected [1,4,64,109]. Any chemical reaction (2.1, 2.11) or physisorption process (2.1, 2.10) releases or absorbs from its surroundings a certain quantity of heat (enthalpy term, ∆H 0 , in 2.9). Reactions liberating heat are termed exothermic, reactions abstracting heat are termed endothermic. This thermal effect shows a transient behavior: Continuous heat liberation/abstraction occurs only as long as the reaction proceeds. This implies that only a steady-state situation can be achieved: A chemical reaction is proceeding at a constant rate and is thus releasing/abstracting permanently a constant amount of heat. There will be, however, no heat production, and, hence, no measurable signal at thermodynamic equilibrium (∆G = 0) in contrast to mass-sensitive, optical, or electrochemical sensors. Conflicting constraints are imposed on the design of a thermal sensor: The sensor has to interact with the chemical species (exchange of matter) and thus constitutes a thermodynamically open system, but, at the same time, the sensing area should be thermally as isolated as possible. The liberation or abstraction of heat is conveniently measured as a change in temperature, which can be easily transduced into an electrical signal. All sensors aimed at thermal infrared radiation detection can, in principle, be used as chemical sensors as well. The various types of calorimetric sensors differ in the way that the evolved heat is transduced. The catalytic sensor (often denoted “pellistor”) employs platinum resistance thermometry [110–122], the thermistor employs composite oxide resistance thermometry [123–129], whereas the pyroelectric [130, 131] and Seebeck-effect [132–144] sensors utilize the respective effects to measure the temperature change. In addition, there are thermal (flow) sensors based on the different thermal conductivity of gaseous analytes [145, 146]. The micromachined cantilever enabling microthermal analysis due to the bimorph effect [103, 108] has already been mentioned in Sect. 4.1.3. Catalytic sensors, Seebeck-effect or thermoelectric sensors, pyroelectric sensors, and thermal conductivity sensors are semiconductor-technologycompatible. Since the latter is essentially a flow sensor responding to thermo-
40
4 Microfabricated Chemical Sensors
physical properties of a gas (thermal conductivity, heat capacity) [145, 146], it will not be subject to further discussion here. Sensors based on the pyroelectric effect (anisotropic, noncentrosymmetric crystal lattice, permanent polarization, creation of macroscopic charges due to thermal stress in the crystal) require the deposition of pyroelectric material (lithium tantalate, zinc oxide [130], polycyclic organic compounds [131]) on the silicon chip. There are very few chemical-sensing applications reported in literature [34, 130, 131]. Therefore, this class of sensor will not be discussed any further. Thermistors are temperature-sensitive bead resistors composed of either oxide semiconductors with a negative (NTC) or a positive (PTC) temperature coefficient; that is, their resistance decreases or increases nonlinearly with temperature. PTC resistors are made, e.g., from barium or lead titanate, while NTC resistors are made from sintered transition metal oxides (titanium oxide) doped with aliovalent ions. The beads are contacted via two metallic (platinum) leads and coated with glass for chemical inertness [4, 123]. The thermistors themselves have not been fabricated in planar semiconductor technology yet, but have been integrated into silicon-based biosensors due to their small size [124, 125]. They have been used as thermal biosensors in strongly exothermic enzymatic reactions to detect urea [125] glucose [126– 128], uric acid [129], and other compounds of relevance in blood analysis. In the following, thermoelectric and catalytic calorimetric sensors will be detailed. 4.2.1 Catalytic Thermal Sensors (Pellistors) Transduction Principle and Sensing Characteristics The development of the catalytic sensor is derived from the need for a handheld detector for methane to replace the flame safety lamp in coalmines. The catalytic device measures the heat evolved during the controlled combustion of flammable gaseous compounds in ambient air on the surface of a hot catalyst by means of a resistance thermometer in proximity with the catalyst. This method is therefore calorimetric. A catalyst is a chemical compound (often a noble metal like platinum (Pt)) enabling or accelerating a chemical reaction by provision of alternative reaction paths involving intermediates with lower activation energies than the uncatalyzed mechanism (Fig. 4.7). The catalyst itself is not permanently altered by the reaction. The heated catalyst here permits oxidation of the gas at reduced temperature and at concentrations below the lower explosive limit (LEL). Three elements are necessary for this method: A catalyst, a method to heat it, and a means to measure the heat of catalytic oxidation. The term “pellistor” originally refers to a device consisting of a small platinum coil embedded in a ceramic bead impregnated with a noble metal catalyst [110]. A ceramic
4.2 Thermal Sensors
41
A*
A*cat educts A intermediates
products B
reaction: A ↔ A* ↔ B Fig. 4.7. Working principle of a catalyst: Provision of an alternative reaction path with less activation energy for the reaction A↔ A∗ ↔ B via the intermediate state A∗cat
(a) catalyst (Pd, Pt)
platinum resistors
Si-nitride passivation
(b)
silicon frame
100
m
supporting Si-oxynitride membrane
Fig. 4.8. (a) Cross-section of side-by-side microhotplates composed of a dielectric membrane on an etched silicon wafer, and platinum resistors/heaters. The device on the left has a deposited catalyst making it the active element. Redrawn from [111]. (b) Micrograph (SEM) of two meandered polysilicon microbridges. The lower meandered bridge is coated with a thin (approx. 0.1 µm) layer of platinum (CVD). In a differential gas-sensing mode, the upper uncoated filament acts to compensate changes in the ambient temperature, thermal conductivity and flow rate, while the lower filament is used to calorimetrically detect combustible gases. Reprinted from [113] with permission
bead is used since the rate of reaction (and thus the sensor signal) is directly proportional to the active surface area. Figure 4.8 shows two different micromachined designs to realize a catalytic calorimetric sensor: A meander structure on a micromachined membrane [111] and a freestanding, Pt-coated polysilicon microfilament (10 µm wide, 2 µm thick) separated from the substrate by a 2-µm air gap [112,113]. Heat losses to the silicon frame are minimized in these designs. By passing an electric current through the meander, the membrane/microbridge is heated to a temperature sufficient for the Pt surface to catalytically oxidize the combustible mixture;
42
4 Microfabricated Chemical Sensors
the heat of oxidation is then measured as a resistance variation in the Pt. The combustion of methane, e.g., generates 800 kJ/mol heat, which translates into a corresponding temperature change. The structures described here are very similar to hotplate structures discussed in the electrochemical section. (Sect. 4.4.3.1.2) The temperature change of the sensor element is proportional to the combustible concentration when the device is operated in excess oxygen and in the mass-transfer-limited regime [114, 115]. The combustion of hydrogen in dry air is exemplified in Fig. 4.9 [112, 113]. The circuit maintains a constant sensor temperature by adjusting the supplied current to keep the filament resistance at a constant value. Note that the sensor response is measured at steady state, i.e., continuous combustion. In most realizations, the measuring resistor forms part of a Wheatstone bridge configuration [111, 114]. Temperature–modulated operation has been reported in [116, 117].
sensor response [V]
0.6 1.6% 1.0% 0.5% 0.1% 0.55
0.5 hydrogen on Pt-filament 0.45 200
400
600
800 1000 1200
time [s] Fig. 4.9. Sensor response of a Pt-coated filament exposed to various concentrations of hydrogen in synthetic air. Reprinted from [112] with permission
Fabrication • Back etching (KOH, or EDP) for membranes [111, 116, 118]. • Surface micromachining: Sacrificial-layer etching (HF) for the bridges [112, 113]. • Pt or catalyst processing: Sputtering [116, 117], evaporation [93–95], LPCVD [112, 113]. A processing sequence for microbridges is given in [121, 122]. Applications Main applications include monitoring and detection of flammable gas hazards in industrial, commercial and domestic environments. The lower explosive
4.2 Thermal Sensors
43
limit (LEL) is the concentration of gas in air, below which it cannot be ignited. Target gases include methane [114,118,122], hydrogen [111–115,118], propane [111], carbon monoxide [111, 119], and organic volatiles [116, 117, 120]. The detectable gas concentrations usually range between the lower-few-percent (1–5%) and the some-hundreds-of-ppm region. The interaction process is an irreversible chemical combustion reaction at high temperature liberating the respective reaction enthalpy (2.9, 2.11). 4.2.2 Thermoelectric or Seebeck-Effect Sensors Transduction Principle and Sensing Characteristics This type of sensor relies on the thermoelectric or Seebeck-effect: When two different semiconductors or metals are connected at a hot junction, and a temperature difference is maintained between this hot junction and a colder point, then an open-circuit voltage is developed between the different leads at the cold point. This thermovoltage is proportional to the difference of the Galvani potentials (inner contact potential, difference of the Fermi levels of the two materials) at the two temperatures and, thus, proportional to the temperature difference itself [132]. This effect can be used to develop a thermal sensor by placing the hot junction on a thermally isolated structure like a membrane, bridge, etc., and the cold part on the bulk chip with the thermally well-conducting silicon underneath [95, 133–135]. To achieve a higher thermoelectric voltage, several thermocouples are connected in series to form a thermopile. The membrane structure (hot junctions) is covered with a sensitive or chemically active layer liberating or abstracting heat upon interaction with an analyte. The resulting temperature gradient between hot and cold junctions then generates a thermovoltage, which can be measured. Figure 4.10 displays the schematic of a CMOS thermopile. The sensor system relies on polysilicon/aluminum thermocouples exhibiting a Seebeck coefficient of 111 µV/K. The hot junctions are in the center of the membrane, the cold junctions on the bulk wafer material. The center part (hot junctions) of the membrane is coated with a gas-sensitive layer such as a polymer. The detection process includes four principal steps: (I) absorption and partitioning or chemical reaction, (II) generation of heat, which causes (III) temperature changes to be transformed in (IV) thermovoltage changes (see, e.g., [95, 134]). Each of the four steps contributes to the overall sensor signal. The calorimetric sensor only detects changes in the heat budget at nonequilibrium state (transients) as a consequence of changes in the analyte concentration (for details, see Sect. 5.2.3). Therefore, the sensor provides a signal upon absorption (condensation heat) and desorption (vaporization heat) of gaseous analytes into the polymer [134–138], or during chemical reaction of an analyte with the sensing material [139–141]. The recorded thermovoltage change, ∆U [V], is, therefore, proportional to the derivative of the analyte concentration as a function of time, dcA /dt [mol/m3 s]:
44
4 Microfabricated Chemical Sensors cold junctions
CMOS n-well
polymer
hot junctions
analyte
dielectric membrane
p-substrate
Fig. 4.10. Schematic of a thermoelectric sensor. Polysilicon/aluminum thermopiles are used (hot junctions on the membrane, cold junctions on the bulk chip) to record temperature variations caused by analyte sorption in the polymer
dcA . (4.3) dt Here A [K·s/J] and B [V/K] are device- and coating-specific constants describing the translation of a generated molar absorption/reaction enthalpy, ∆H [J/mol], via a temperature change into a thermovoltage change. Vsens denotes the sensitive-layer volume, and K is the partition coefficient (2.10) or reaction equilibrium constant (2.2, 2.11). ∆U = A · B · Vsens · ∆H · K ·
Fabrication • Back etching (KOH, or EDP) for membranes [95, 133, 135, 142]. • Processing of the sensitive layer: Airbrush [95, 135], dispensing, spin coating, or enzyme immobilization methods [139–141]. Processing sequences for the integration of thermoelectric sensors with circuitry in a CMOS standard process are detailed in [95, 135, 142]. Sensors are commercially available from Xensor [143]. Applications Typical applications areas are environmental monitoring (gas and liquid phase) or biosensing in liquids. These include the detection of different organic volatiles in the gas phase (hydrocarbons, chlorinated hydrocarbons, alcohols etc.) by using polymeric layers [95, 135–138] or metal oxides [144], the monitoring of acid/base neutralization [139, 141], and the biosensing of glucose, urea and penicillin in the liquid phase by using suitable enzymes [134, 139–141]. The interaction mechanisms involve reversible physisorption and bulk/gas phase partitioning (see 2.10) as well as enzymatic chemical reactions (2.11) [134, 139–141].
4.3 Optical Sensors
45
4.3 Optical Sensors Light can be considered consisting of either particles (photons) or electromagnetic waves according to the principle of duality. The characteristic properties of the electromagnetic waves such as amplitude, frequency, phase, and/or state of polarization can be used to devise optical sensors [1,4,6,10,147–149]. The energy, E, of an electromagnetic wave is quantized, a quantum being termed a photon (h is Planck’s constant, 6.626 · 1034 Js, ν denotes the frequency): E =h·ν . (4.4) When light interacts with matter, several processes can take place, sometimes simultaneously. Absorption If a sample is irradiated with visible light or electromagnetic waves, the radiation can be absorbed, which results in a decrease of the intensity in the detected radiation as compared to the primary beam (Fig. 4.11). Alternatively, the radiation can be transmitted without attenuation. A prerequisite for absorption is that the absorbing matter (atom, molecule, etc.) exhibits unoccupied energy states with an energetic difference exactly equal or less than the energy of the incoming radiation quanta. The matter then absorbs the radiation energy by transition into a so-called excited state with higher internal energy. The absorption of radiation forms the base for most traditional spectroscopic methods, which are usually distinguished according to the different radiation wavelengths or frequency ranges as given in Table 4.1 [35]. Table 4.1. Different spectroscopy methods and radiation energies Radiation
Energy [J/mol]
Wavelength [m]
Transition
γ-radiation X-rays Ultraviolet (UV) Visible (VIS) Infrared (IR) Microwaves Radio waves
109 –1011 107 –109 106 –107 105 –106 102 –105 10−2 –102 <10−2
10−13 –10−11 10−11 –10−9 10−9 –10−7 400–800 nm 10−6 –10−4 10−4 –1 >1m
Nucleus excitation Core electron excitation Shell electron excitation Shell electron excitation Vibrational states Rotation states Electron spin, Nuclear spin
The absorption of monochromatic radiation (only one selected wavelength) can be quantitatively determined using the well-known Lambert-Beer relation: (4.5) I = I0 · e−ελ cA l . Here I denotes the transmitted radiation intensity at the detector, I0 the intensity of the incident radiation, ελ is the molar absorptivity at the measured
46
4 Microfabricated Chemical Sensors
wavelength, cA the analyte concentration, and l the optical path length in the probed volume. Scattering Changes of the direction and/or the frequency of light are commonly denoted as scattering (Fig. 4.11). Scattering of light does not necessarily involve a transition between quantized energy levels in atoms or molecules. A randomization in the direction of light radiation occurs. Particles with sizes that are small compared to the wavelength of radiation give rise to Rayleigh scattering, while particles that are large compared to the wavelength give rise to Mie scattering [147, 148]. In both processes the particle polarization is unaltered. However, the incident radiation can promote vibrational changes (energy quantum absorption), which can alter the polarization of the irradiated particle/molecule. The frequency of the light scattered by these molecules will be different from that of the incident light and the light intensity will be much lower. Such a phenomenon is known as Raman scattering [6, 35, 150]. Fluorescence and Phosphorescence The mechanism of those two phenomena is an absorption-emission process. The wavelength or energy of the incident radiation is absorbed and promotes changes in the molecular energy states. The resulting excited state is unstable, and the molecule dissipates some of its energy to rotational and/or vibrational energy states. The molecule then can return into the ground state by emitting light at a lower frequency than the incident radiation, this process being termed fluorescence. If a more complex and slower intersystem crossing process into a triplet state, and then, a radiative transition from there to the ground state occurs, the process is called phosphorescence. For details and the respective Jablonski diagrams, which represent simplified portrayals of the relative positions of the electronic energy levels of a molecule, see for example [4, 35, 147]. Both processes, fluorescence and phosphorescence are sometimes subsumed under luminescence processes, but the term “luminescence” here will be used exclusively for chemoluminescence processes as detailed below. Fluorescence processes are extensively used in gene analysis techniques, where a defined array of single-stranded deoxyribonucleic acid (DNA) fragments is hybridized with the respective complementary strands labeled with a fluorescent marker. By illuminating the array with a laser, the sites, where the labeled DNA fragments are bound by interaction between the two complementary strands, can be detected by their positive fluorescence response [149]. This technique has been commercialized by several companies [151]. Chemoluminescence The excited state of a molecule (C∗ ) is created by a chemical reaction [4, 10, 152]; the molecule emits light during transition to the ground state according to: (4.6) A + B ⇒ C∗ ⇒ C + h · ν .
4.3 Optical Sensors
47
Chemical energy is thus directly converted into light energy in most cases without additional heat generation (cold luminescence). In the biological domain, this process is denoted bioluminescence and, e.g., occurs in glowworms. Reflection and Refraction Reflection and refraction take place when light infringes on a boundary surface between two media of distinct optical properties (refraction index). The light can either be reflected back into the original medium or be refracted (transmitted) into the adjacent medium (Fig. 4.11). Several distinct types of reflection are possible. The first is a “mirror type” or specular external reflection (Fig. 4.11) occurring at, e.g., a metal surface or generally at interfaces of media with no transmission through (evanescent waves will be treated in the context of refraction below). Another type is diffuse reflection, where the light penetrates the medium and subsequently reappears at the surface after partial absorption and multiple scattering within the medium. The optical characteristics of diffusely reflected radiation provide information on the composition of the reflecting medium [6]. Thin films (10 µm and less) on a surface can strongly affect the propagation of incident light due to reflection at each of the thin film interfaces causing a multitude of reflected, coherent beams with small phase shifts. Sensor techniques to interrogate such thin film structures include ellipsometry [153, 154] and thin-film reflectometric interference spectroscopy (RIFS), which is based upon spectral modulation of the reflectance of a thin film under white-light illumination without using the polarization information. The spectral characteristics are a function of the film thickness, and, therefore, any ad/absorption of organic matter leads to changes in the interferograms of the reflected beams [155, 156]. A variety of transducers respond to changes in the refractive index in immediate vicinity to the device surface. The propagation behavior of a wave guided by nonmetallic total internal reflection (Fig. 4.11) in a medium of high refractivity depends on the dielectric characteristics of the surrounding medium. This effect is mediated by the evanescent field, which penetrates from the optically denser guiding medium a few hundred nanometers into the optically rarer environment [149, 157, 158] (more details in Sect. 4.3.1 on integrated optics, see also Fig. 4.12). If the environment absorbs, energy will be transferred from the evanescent wave to the environment and attenuation of the traveling wave will occur (attenuated total reflection, ATR) [147, 148]. The energy of the evanescent wave can also be used to initiate fluorescence (total internal reflectance fluorescence, TIRF) [157]. Light propagating in waveguide structures without absorbing cover layers is not attenuated by environmental influences (frustrated total reflection, FTR) [147, 148]. Its propagation velocity, however, changes depending on the refractive index in the vicinity of the waveguide. Several
48
4 Microfabricated Chemical Sensors
scattering
J0 light source
sample
detector
absorption
J
L
reflection and refraction
external reflection
reflected n1 n2 n 1> n2
φ1 φr
φ 1 = φr
φ2 refracted
λ
internal reflection
air
n 1> n2
standing wave
n1
metal
n2 evanescent wave
TIRF
Fig. 4.11. Schematic representation of the different processes taking place upon interaction of light with matter: Absorption, scattering, reflection/refraction, external and internal reflection. n denotes the refraction index, J the light intensity, λ the wavelength and φ the angle. For details, see text and [147–149]
setups have been proposed and will be discussed in Sect. 4.3.1 on integrated optics [147, 149, 158]. Surface plasmon resonance (SPR) is based on collective fluctuations in electron density at the surface of thin films, typically gold or silver, on a waveguide. Surface plasmon waves show the maximum of the electrical field distribution located at the waveguide/metal interface, which is exponentially decaying into the metal and the adjacent medium. SPR is detected as a strong attenuation of the reflected light beam sensitive to the medium adjacent to the metal film [159–161]. SPR will be described in more detail in Sect. 4.3.4. In comparison to other chemical sensing methods, optical techniques offer a great deal of selectivity already inherent in the various transduction mechanisms. Characteristic properties of electromagnetic waves, such as amplitude, frequency, phase, and/or state of polarization can be used to advantage. The wavelength of the radiation, e.g., can be tuned to specifically match the energy of a desired resonance or absorption process. Geometric effects (scattering) can provide additional information. Moreover, optical sensors, like any other chemical sensor, can capitalize on all the selectivity effects originating from the use of a sensitive layer. Optical sensors and classical spectroscopy methods are often very similar in methodology but differ in the arrangement of the experiment and equipment. In particular, the introduction of fiber-optic techniques has promoted the development of comparatively inexpensive optical sensor setups. In the following the focus will be exclusively on semiconductor- and MOEMS(micro-opto-electro-mechanical [162]) based transducers and their respective
4.3 Optical Sensors
49
mechanisms. The wide field of glass-based and fiber-optical techniques such as optodes [163] or micro optodes [164, 165] will not be covered. For interested readers reviews and articles by Wolfbeis and others are recommended [147,148,166–170]. The light-addressable potentiometric sensor (LAPS) [171] will be discussed in the electrochemical sensor section with the field-effect devices (Sect. 4.4.2.2.5). 4.3.1 Integrated Optics Transduction Principle and Sensing Characteristics The generation of light in silicon devices is very difficult since there is no first-order transition from the valence band to the conduction band without the involvement of a phonon (lattice vibrations) [36, 172]. Only directbandgap semiconductors like gallium arsenide (GaAs) or indium phosphide (InP) show first–order radiative electron-hole recombinations with high quantum efficiency (see section on GaAs devices below). The detection of light is possible with either silicon-based devices (photodiodes) or other semiconductor materials. Integrated optical (IO) sensors make use of guided waves or modes in planar optical waveguides. The waveguide materials usually include highrefractivity silicon dioxide or titanium dioxide and silicon nitride films on oxidized silicon wafer substrates. The guided waves or modes in planar optical waveguides include the TE (transverse electric or s-polarized, surfacenormal) and the TM (transverse magnetic or p-polarized, surface-parallel) modes. Changes in the effective refractive index of a guided mode are induced by changes of the refractive index distribution in the immediate vicinity of the waveguide surface, i.e., within the penetration depth (some hundred nanometers) of the evanescent field in the sample (Fig. 4.12a) [149, 158]. The evanescent field decays exponentially with increasing distance from the waveguide surface. Changes in the effective refractive index can be induced by absorption of an adlayer onto the surface of the waveguide from gas or liquid phase [158, 173, 174], by interaction of an analyte molecule with a recognition structure immobilized on the waveguide surface [158,175–179], or by changes of the refractive index of the medium adjacent to the waveguide in a flow-through configuration [158, 175, 176]. In the case of microporous waveguides, analyte molecule absorption or desorption directly into the pores of the wave-guiding film itself can change the waveguide refractive index [158]. A number of different IO sensors have been developed to transform the changes of the effective refractive index into readily measurable physical quantities. Grating Couplers Among the first integrated optical devices were grating coupler structures embossed with a monomode film waveguide as proposed by Lukosz [158,174– 176, 180]. A periodic grating on the surface of the waveguide can be used
50
4 Microfabricated Chemical Sensors
(a) evanescent wave
(b) grating coupler
fluorophores (TIRF) evanescent sensitive field layer nl waveguide nw substrate ns < nl < nw
flow cell liquid sample
sensitive layer protection
grating waveguide substrate
ns φ
Fig. 4.12. (a) Schematic of an evanescent wave in an optical waveguide. When fluorophores are within the reach of the evanescent wave, they can be excited, and the fluorescence can be detected (TIRF). (b) Schematic of a grating coupler. A periodic grating on the surface of the waveguide is used for in- or out-coupling of radiation to/from the waveguide. The deflection angle depends on the light wavelength and the grating period and is altered by binding of an analyte on the grating. Redrawn from [158]
for in- or out-coupling of radiation (TE and TM mode) from the waveguide. In- and out-coupling are governed by the same physical laws, as the reciprocity theorem permits the reversal of the propagation direction of all light waves. The deflection angles (coupling angles) of the TE and TM modes depend on the light wavelength and the grating period (Fig. 4.12b). Binding of an analyte on the grating alters the coupling angle, which can be detected using position-sensitive detectors. The sensitivity of the grating coupler is related to the lateral dimensions of the grating region interacting with the sample. Prism couplers [180] will not be treated here. Difference Interferometer In a planar waveguide, the TE and TM mode are coherently excited by a laser. Both propagate along a common path down the same waveguide and interact with the sample within a certain length of the waveguide. The polarization-dependent interaction induces a phase difference between the two modes, which can be measured using a dedicated interferometer setup [158, 173, 174, 181, 182]. A variant of this method involves a Zeeman laser to generate two orthogonally polarized modes in a silicon nitride waveguide [183, 184].
4.3 Optical Sensors
(a) Mach-Zehnder interferometer sensitive layer
51
(b) integrated GaAs interferometer
Si-oxide Si-nitride Si-oxide
sensor area
Si substrate electric field
phase modulator
reference branch input
output
DBR laser
sensor pad
detector
L
L sensor branch
Fig. 4.13. Schematic of a conventional Mach-Zehnder interferometer (a), and an integrated Mach Zehnder interferometer (light source and detector on chip) in GaAstechnology (b). The cross section shows the separate sensor (left side, open) and reference (right side, covered) branches. Redrawn from [191]
Two-Beam Interferometers Mach-Zehnder IO-devices (Fig. 4.13a) are monomode channel waveguides (TE or TM mode) and allow for a straightforward implementation of an interferometer structure [158,177–179,185]. A waveguide is split into an open measurement path and a protected reference path and recombined after some distance. The phase difference, introduced by analyte interaction (refractive index change) in the sensing path, is detected by interference effects. Detection limits are in the range of some picograms/mm2 [186]. Integrated Waveguide Absorbance Optodes (IWAO) A membrane inserted between two micromachined waveguides acts simultaneously as the light-guiding medium and sensing element and hence changes its spectral properties while interacting with an analyte. First results with potassium-selective optode membranes are reported in [187]. Fabrication • Patterning of silicon nitride as waveguide (LPCVD, RIE, lithography) [158, 177–179] • Deposition of an silicon oxide cladding layer (PECVD) [158, 177–179] • Deposition of the chemically sensitive layer (immobilization of biological entities) [175–179] Electro-optical modulation techniques of the sensor signal, when using zinc oxide as the optical waveguide on a chip fabricated in silicon oxinitride technology, have been reported in [188, 189].
52
4 Microfabricated Chemical Sensors
∆Φ[2π]
-10
PBS
10 µg/ml avidin 10 µg/ml biotinylated protein A 0 rabbit anti-atrazine 1 serum 1:100 2
PBS
-20 PBS
3 25 µg/ml atrazine-HRP
-30 PBS
4
-40
surface mass density [ng/mm2]
0
5 0
100
200 300 time [min]
400
Fig. 4.14. Sensor response (phase shift, ∆Φ) of a difference interferometer used for biosensing. Adsorption of avidin at the surface, affinity binding of biotinylated protein A to the avidin layer, binding of the rabbit anti-atrazine serum to the protein A layer, immunoreaction of the immobilized anti-atrazine antibodies with atrazine (atrazine-horseradish peroxidase, atrazine-HRP). PBS denotes phosphate buffer solution washing steps. Reprinted from [173] with permission
Applications Typical applications are humidity sensors [158,173,174], gas sensors [184,185] (adsorption on the device or absorption in a microporous waveguide), and environmental monitoring or biosensing. Examples include the detection of different organic solvents in the liquid phase (hydrocarbons, alcohols etc.), the monitoring of sucrose and buffer solutions [158, 173, 175, 176], and biotin/streptavidin-mediated immunosensing (Fig. 4.14) involving antibody/antigen binding experiments [158, 173–179]. The interaction mechanisms include reversible physisorption (2.10) as well as biochemical affinity reactions (2.11) [158, 175–179]. Gallium-Arsenide-Based Devices Due to their direct band gap, III-V-semiconductors offer the opportunity of fabricating and integration of lasers, waveguides, phase modulators and waveguide detectors on the same chip. GaAs/AlGaAs-based Mach-Zehnder devices with integrated light sources and detectors have been developed as shown in Fig. 4.13b [190–192]. The light source is a distributed Bragg reflector (DBR) laser, which was fabricated with a simplified grating recess technology [193, 194] and is operating on a single mode. A dielectric waveguide pad (silicon oxide, tantalum oxide) is integrated in the measurement arm of the interferometer [190–192]. The detector is a long-absorbing-length photodiode with high quantum efficiency [193]. A more recent development for optical gas sensing is the vertical-cavity surface-emitting laser (VCSEL). The cavity is formed vertically on the wafer
4.3 Optical Sensors
53
surface. Epitaxially grown Bragg mirrors serve as distributed reflectors above and below the laser’s very short active region. GaAs/AlGaAs-based VCSELs emit in the near infrared region. Oxygen sensing at 762 nm (absorption due to magnetic dipole transitions in the gas molecule without interference from other gases) was demonstrated in first experiments [195, 196]. 4.3.2 Microspectrometers 4.3.2.1 Fabry-Perot-Type Structures Transduction Principle and Sensing Characteristics A Fabry-Perot interferometer (FPI) is an optical element consisting of two partially reflecting, low-loss, parallel mirrors separated by a gap. The optical transmission characteristics through the mirrors consist of a series of sharp resonant transmission peaks occurring when the gap is equal to multiples of a half wavelength of the incident light. These transmission peaks are caused by multiple reflections of the light in the cavity. By using highly reflective mirrors, small changes in the gap (width, absorptivity) can produce large changes in the transmission response. Even though two reflective mirrors are used, transmission through the element at the peak wavelengths approaches unity. The transmission is a function of both, the gap spacing and the radiation wavelength. The devices can be used as wavelength selector or monochromator by adjusting the gap width to achieve the desired wavelength. Tunable devices with a gap width variable by electrostatic actuation using electrodes on movable micromachined parts have been reported (Fig. 4.15a) [197–200]. Such devices operate preferably in the near infrared region at wavelengths larger than 1 µm, where silicon substrates become transparent [197]. (a) tunable Fabry-Perot
(b) CMOS Fabry-Perot etalon
radiation antireflection coating
silver wafer
optical control coating (Al) electrodes
PECVD oxide
gap
fusionbonded layer corrugated support
movable mesa
n-well p+ implanted layer (SP)
Fabry-Perot cavity
aluminum
p -epilayer
}
}pnp-phototransistor
p+ substrate
Fig. 4.15. (a) Schematic of a tunable Fabry-Perot interferometer. The wavelengthdefining gap width can be changed by applying DC to the control electrodes. Aluminum is used as optical coating material. Redrawn from [197]. (b) Cross section of an integrated Fabry-Perot etalon. The gap width is determined by the thickness of the PECVD oxide, silver and aluminum are used as optical coatings. A pnp-phototransistor (p+ implanted layer/n-well/p-epilayer) is located directly underneath the etalon. Redrawn from [201]
54
4 Microfabricated Chemical Sensors
A single-chip CMOS optical microspectrometer based on FPI and operating in the UV/VIS-region is reported in [201, 202]. It contains an array of 16 addressable Fabry–Perot etalons (500 × 500 µm2 ) each with a different resonant cavity length realized as a PECVD silicon oxide layer sandwiched in between an aluminum and a silver layer and placed on top an array of vertical pnp phototransistors (Fig. 4.15b) [203]. It additionally includes circuits for readout, multiplexing and driving a serial bus interface. Fabrication • • • •
Surface micromachining techniques to achieve an air gap (HF) [200, 202] Deposition of the lower mirror (evaporation and lift-off of aluminum) [201] Deposition of a silicon oxide layer of defined thickness (PECVD) [201] Deposition of the upper mirror layer (silver) [201]
Some devices are derived from air gap pressure sensors [198–200, 204, 205], the fabrication sequences of which can then be applied. The same holds for wafer-stacking techniques to achieve the FPI cavities [198,199]. The complete fabrication sequence of a monolithic CMOS VIS spectrometer is given in [201]. Applications Typical applications include gas sensors [198–200, 203–207]. The characteristic absorption wavelengths of carbon monoxide are 4.7 µm, that of carbon dioxide 4.2 µm, and that of methane or hydrocarbons 3.3 µm (IR region, molecular vibrations). Polymeric coatings in the FPI cavity have been used to detect iodine [204, 205]. Carbon dioxide sensors based on tunable FPIs (dual wavelength measurements) are commercially available from [207]. The radiation source in most cases is a light bulb or light-emitting diode (LED). 4.3.2.2 Grating-Type Structures Transduction Principle and Sensing Characteristics Micromachined diffraction gratings have been used in combination with imaging devices to set up microspectrometer arrangements [208–210], see Fig. 4.16a for a schematic. Two basic types of gratings can be easily micromachined: (1) amplitude gratings by blocking out light with an array of opaque and transparent sections, or (2) phase gratings where the light phase is modulated by variations in the grating shape [208, 209]. In the example displayed in Fig. 4.16a [210], a phase grating with a fine grating pitch creating high dispersion angles was etched into a quartz wafer. The transmission grating was then mounted directly over a charge-coupled-device (CCD) imager [208], or a CMOS imaging chip consisting of an array of custom-designed photodiodes [210]. A glass spacer is placed in the optical path between the grating and the detector. In another approach, two wafers have been micromachined so that an optical path of about 4 mm length was obtained. Light dispersed by a 32-slit diffraction grating travels along the optical path and is directed to an array of photodiodes (Fig. 4.16b) [211]. The interior of one of
4.3 Optical Sensors
(a) CMOS microspectrometer diffraction grating
photodetector array
blue
grating
glass spacer
CMOS imager chip
(b) integrated spectrometer
red
incident light
55
bevel p-wafer mirror
increasing
wavelength
n - epilayer
Fig. 4.16. (a) CMOS microspectrometer consisting of a glass and a CMOS imager chip: Different wavelengths are diffracted at different angles from the surface normal. Redrawn from [210]. (b) Schematic of an integrated spectrometer using two fusion-bonded chips, one of them carrying a grating and a photodetector array. Redrawn from [211]
the wafers is coated with a reflective film. The grating and the diode array are integrated in the second wafer, which remains uncoated. The wafers are bonded together by silicon/silicon fusion bonding. The performance of such micromachined spectrometers is comparable to that of low-end bench-top spectrometers. Diffractive optical elements, which produce analyte infrared spectra of compounds such as hydrogen fluoride (HF) for chemical sensor systems based on correlation spectroscopy, are reported in [212]. Fabrication • • • •
Patterning of gratings, quartz micromachining [208, 211] Back etching using electrochemical etch stop [211] Deposition of reflective layers [211] Wafer fusion bonding [211]
Applications Typical applications include biochemical and chemical analysis. Emission gas spectra have been recorded for carbon dioxide and helium. Fluorescence of a dye (fluorescein) was induced by a laser and detected with the microspectrometer [208]. 4.3.3 Bioluminescent Bioreporter Integrated Circuits (BBIC) Transduction Principle and Sensing Characteristics This technique employs bioluminescent bacteria placed on an applicationspecific optical integrated circuit (standard CMOS) [213–215]. The bacteria have been engineered to luminesce (4.5) when a target compound such
56
4 Microfabricated Chemical Sensors
signal processing circuit
photodetectors
Fig. 4.17. Micrograph of a bioluminescent bioreporter integrated circuit (BBIC). For details, see text. Reprinted from [214] with permission
as toluene is metabolized. The integrated circuit detects, processes, and reports the magnitude of the optical signal. The microluminometer uses the p-diffusion (source and drain diffusions of p-channel MOSFETs) as the photodiode. The shallow p-diffusion has a strong response to the 490-nm bioluminescent signal. The entire sensor including all signal processing and communication functions can be realized on a single chip. The integrated circuitry contains the subunits needed to detect the optical signal, to perform analog or digital signal processing, to communicate the results, and to perform auxiliary functions (temperature, position measurement) (Fig. 4.17). Many types of bioluminescent transcriptional gene fusions have been used to develop light-emitting bioreporter bacterial strains to sense the presence, bioavailability, and biodegradation of different kinds of pollutants. The cells here were entrapped on the chip by encapsulation in natural or synthetic polymers providing a nutrient-rich hydrated environment [213–215]. Fabrication • Silicon nitride protective coating using a jet vapor-deposition technique [216] • Deposition of the cell-containing polymer (drop coating, spraying, beads) [213–215] Applications Typical applications include chemical analysis in gas or liquid phase. Depending on the integration time of the device, trace amounts of toluene and naphthalene were detected in the gas phase using suitable cell colonies (Pseudomonas putida) [213–215]. The interaction mechanism is a chemical reaction (2.11, 4.5).
4.3 Optical Sensors
57
4.3.4 Surface Plasmon Resonance (SPR) Devices Transduction Principle and Sensing Characteristics The quantum optical-electronic basis of SPR is due to the fact that the energy carried by photons of light can be “coupled” or transferred to electrons in a metal [159]. The wavelength of light, at which coupling (i.e., energy transfer) occurs, is characteristic of the particular metal and the environment, in which the metal surface is illuminated; gold being the preferred metal. The coupling can be observed by measuring the amount of light reflected by the metal surface. All the light is reflected except at the resonant wavelength, where almost all the light is absorbed (Fig. 4.18). The coupling of light into a metal surface results in the creation of a plasmon, a group of excited electrons, which behave like a single electrical entity. The plasmon, in turn, generates an electrical field, which extends about 100 nm above and below the metal surface [159]. The characteristic of this phenomenon, which makes SPR an analytical tool, is that any change in the chemical composition of the environment within the range of the plasmon field causes a change in the wavelength of light that resonates with the plasmon. That is, a chemical change results in a shift in the wavelength of light, which is absorbed rather than reflected, and the magnitude of the shift is quantitatively related to the magnitude of the chemical change [161, 217]. The phenomenon of SPR is non-specific. Different chemical changes cannot be distinguished.
light source
φ polarized light
prism
1 2
optical detection unit
reflected light
intensity
SPR monochromatic, angle variation
1
2
sensor chip metal film
angle
flow channel
Y
1 : receptor only
2 : receptor and target analyte
Fig. 4.18. Surface plasmon resonance (SPR) principle and a typical sensor response diagram: Intensity versus angle. Monochromatic light is used to excite the surface plasmon, which leads to a drastic intensity decrease at a defined reflection angle. Adsorption of an analyte changes this angle
A hybrid SPR system consisting of two micromachined silicon layers, one micromachined glass layer and an alumina substrate, is shown in Fig. 4.19 [218, 219]. Bonding was achieved using a low-temperature-curing polyimide and a solder sealing. The micromachined silicon layers contain a torsional,
58
4 Microfabricated Chemical Sensors
Fig. 4.19. Cross-sectional diagram of the SPR microsystem, labeling individual components and showing the light path. Color image available at http://www. ece.ucdavis.edu/misl/web/pages/projects/plasmon.html. Reprinted from [218] with permission
all-silicon micromirror, V-grooves for optical fiber and lens, and a positionsensing photodiode (PSD). The silicon micromirror was electrostatically deflected through 9–10 degrees to direct the light beam emitted from the end of a fiber through a range of angles incident onto the metal film, setting up a surface plasmon. The position and intensity of the reflected beam was recorded with the position-sensing photodiode. The microsystem measures approximately 1 × 2 × 0.2 cm3 [218, 219]. Fabrication • Alumina substrate: Laser drilling, screen printing of thick-film conductors [218] • Anisotropic etching of the silicon micromirror and the fiber-to-lens alignment groove [218] • Filling the interior of the assembly with index-matching fluid [218, 219] • Glass slide: Deposition of the sensitive layer (drop coating, spraying, beads) [160, 161] Small SPR instruments are commercially available from Texas Instruments [220]. Applications Any pair of molecules that exhibit specific binding can be adapted to SPR measurements. These may be an antigen and antibody, a DNA probe and complementary DNA strand, an enzyme and its substrate, or a chelating agent and a metal ion. Typical applications areas are environmental monitoring (gas sensing [160,218]) or bio- and immunosensing in liquids [160,161].
4.4 Electrochemical Sensors
59
With the micromachined instrument the surface adsorption of bovine serum albumin (BSA) was tested [218, 219]. The interaction mechanisms involve specific biochemical reactions (2.11) [160, 161].
4.4 Electrochemical Sensors Electrochemical sensors constitute the largest and oldest group of chemical sensors [4, 6, 8–11, 23, 26, 34]. They rely on electrochemical or charge-transfer reactions: A+ + e− ↔ A (see 2.12–2.15 and related text in Chap. 2). Electrochemistry includes charge transfer from an electrode to a solid or liquid sample phase or vice versa. Chemical changes take place at the electrodes or in the probed sample volume, and the resulting charge or current is measured. Both, electrode reactions and charge transport in the sample are subject to changes by chemical processes, and, hence, are at the base of electrochemical sensing [4]. A key requirement for electrochemical sensors is a closed electrical circuit, though there may be no current flow (see potentiometry below). An electrochemical cell is always composed of, at least, two electrodes with two electrical connections: One through the probed sample, the other via transducer and measuring equipment. The charge transport in the sample can be ionic, electronic or mixed, while that in the transducer branch is always electronic. Electrochemical sensors are usually classified according to their electroanalytical principles [4, 8, 11]. Voltammetry Voltammetric sensors are based on the measurement of the current-voltage relationship in an electrochemical cell comprising electrodes in a sample phase. A potential is applied to the electrodes, and a current is measured, which is proportional to the concentration of the electro-active species of interest. Amperometry is a special case of voltammetry, where the potential is kept constant as a function of time. Potentiometry Potentiometric sensors are based on the measurement of the potential at an electrode, which is, in most cases, immersed in a solution. The potential is measured at equilibrium state, i.e., no current is allowed to flow during the measurement. According to the Nernst equation (2.15), the potential is proportional to the logarithm of the concentration of the electro-active species (Work function sensors will be discussed in the context of field effect devices in Sect. 4.4.2.2.4). Conductometry Conductometric sensors are based on the measurement of a conductance between two electrodes in a sample phase. The conductance is usually measured
60
4 Microfabricated Chemical Sensors
by applying an AC potential with a small amplitude to the electrodes in order to prevent polarization. The presence of charge carriers determines the sample conductance. In contrast to conductometry, AC-impedance measurements are not really used for analytical applications. Impedance measurements are of special interest for membrane, electrode and electrolyte characterization. The goal of impedance measurements is to find an equivalent electronic circuit model and to correlate that model with electrochemical phenomena. Another categorization method relies on discerning the electronic components [9, 34]. There are chemoresistors, chemodiodes, chemocapacitors and chemotransistors. Within this book the electroanalytical principles will be used as the superordinated classification scheme, and the component notation will be used within this scheme. Again, the focus will be on semiconductor-based systems, and a wealth of literature on, e.g., other designs of ion-sensitive electrodes [221], or siliconcarbide-based devices operating at extremely high temperature [222–225] will be omitted. Review articles are recommended to explore further details (see, e.g., [226–231]). 4.4.1 Voltammetric Sensors Transduction Principle and Sensing Characteristics Voltammetry, in general, is the measurement of the current that flows at an electrode as a function of the potential applied to the electrode. The result of a voltammetric experiment is a current/potential curve. Amperometry is more frequently applied in chemical sensors and provides a linear current/analyte concentration relationship at a constant potential, which is predefined with regard to the target analyte. Two different electrode configurations are normally used. The two-electrode configuration [4, 8, 11] consists of a reference electrode (RE) and a working electrode (WE) (Fig. 4.20a) [11]. The disadvantage of this method is, that the RE carries current and may become polarized if it is less than one hundred times the size of the WE. Material consumption due to the current in the RE is another problem. A better approach is, therefore, the use of a three-electrode-system [4, 8, 11] in a potentiostatic configuration. An additional auxiliary electrode (AE, sometimes denoted counter electrode, CE) is introduced for current injection in the analyte (Fig. 4.20b) [11]. The reference electrode is now a true RE with a well-defined potential since no current is flowing through the RE. The potentiostat controls the current at the auxiliary electrode as a function of the applied potential. This is realized in practice with an operational amplifier (opamp) [11]. The potential is applied to the positive input of the opamp. The RE is connected to the negative input and measures the potential in the solution. The AE is connected to the output. The opamp injects a current into the solution through the AE. Due to the feedback mechanism, the current is controlled in such a way that the potential at the negative input equals
4.4 Electrochemical Sensors
2 - electrode system (a)
61
3 - electrode system
U
+ _
(b)
U
WE RE
WE RE AE
Fig. 4.20. Schematic of a two-electrode (a) and a three-electrode configuration (b) used for voltammetric measurements. For details, see text
the potential at the positive input. The potential difference between WE and RE hence equals the applied potential. No current is flowing through the RE since the opamp has a very high input impedance. The sensor signal current is measured at the working electrode. The measured current at any given potential difference depends on the material properties, the composition and geometry of the electrodes, the concentration of the electro-active species (presumably the target analyte) and the mass transport mechanisms in the analyte phase [4, 8, 11, 34]. Among those are migration, the movement of charged particles in an electric field, convection, the movement of material by forced means like stirring or as a consequence of density or temperature gradients, and diffusion, the movement of material from high-concentration regions to low-concentration regions. The electrochemical reactions at the electrodes are normally fast in comparison to the transport and supply mechanisms. Since convection in the electrode vicinity is avoided, and migration is suppressed by, e.g., a large excess of electro-inactive salts (i.e., electro-inactive at the respective applied potential), diffusion is normally regarded to be the dominant mechanism. There are two components to the measured current, a capacitive component resulting from redistribution of charged and polar particles in the electrode vicinity, and a component resulting from the electron exchange between the electrode and the redox species (analyte) termed faradaic current [4,8,11,34]. The faradaic component is the important measurand and is, for the case of diffusion-limited conditions, directly and linearly proportional to the target analyte concentration. The limiting current (all analyte ions are immediately
62
4 Microfabricated Chemical Sensors
charged or discharged upon arrival at the electrode) is then given by the Cottrell equation [4, 8, 11, 26, 34]: I∞ = ne · F · A · Ddiff
cA . Ldiff
(4.7)
Here, ne denotes the number of electrons, F is the Faraday constant, A the effective electrode area, Ddiff the diffusion coefficient, cA the target analyte concentration and Ldiff the diffusion length (for more mechanistic details see [4,8,11,26,34]). Correction terms to this equation for small electrodes have to be introduced to take into account the respective electrode geometry [4,8,26]. Cyclic Voltammetry A stationary WE is used, and a cyclic ramp potential versus some RE is applied. The potential versus time is triangular: It increases at a rate linear with time, then reverses and decreases at the same rate. The current flows as a consequence of the applied potential between the WE and an AE. This technique is used to study the electrode/sample interface [4, 8, 11, 232–235]. Stripping Voltammetry This method is used to detect heavy-metal ions at trace level by means of a mercury electrode [236–239]. The method involves an initial preconcentration phase, in which the array is held at a cathodic potential such that the metal ions from solution are reduced and amalgamated into the mercury. Then, the electrode potential is reversed to anodic, and the metals in the mercury are re-oxidized and stripped from the mercury into the solution. The charge required to strip a given metal completely from the mercury is proportional to its initial concentration in the test solution [236–239]. Fabrication • Additional silicon nitride as protective coating • Optional back etching, membrane formation and perforation for liquid electrolyte access to membrane-covered, sensitive electrodes or gas permeation [240, 241] • Deposition/patterning of metal electrodes (lift-off, thermal evaporation, sputtering) [8, 11, 240–256] • Deposition of electro-active polymers, membrane materials, hydrogels (spincasting, spraying, screen printing) [240–257] Sensor processing sequences are given in [11, 240, 241, 251, 253]. The fabrication of electrochemical sensors that are integrated with CMOS circuitry components is described in [11, 251]. Microsensor arrays with up to 1024 individually addressable elements have been reported on [254, 255]. A picture and a schematic of a CMOS-based 3-electrode amperometric sensor are shown in Fig. 4.21 [251]. The monolithic device includes the electrochemical sensor, a temperature sensor, and interface circuitry. The circuitry contains an operational amplifier as potentiostat, a switched-capacitor
4.4 Electrochemical Sensors
63
(a)
(b)
Fig. 4.21. Micrograph (a) and layout (b) of a CMOS-based 3-electrode amperometric sensor. Reprinted from [251] with permission
current-to-voltage converter and a clock generator. Interface circuitry and the temperature sensor are realized in 3 -µm CMOS-technology. The circuitry needs a supply voltage of ±2.5 V, can apply voltages from +1 V to −1 V to the sensor and handles currents from 30 nA full scale to 1 µA full scale. The output voltage of the temperature sensor is proportional to the absolute temperature and has a sensitivity of 125 µV/K. The total sensor dimensions are 0.75 mm by 5 mm. Applications Typical applications include chemical analysis in the gas or liquid phase. If the target analyte is not an electro-active species like, e.g., glucose, oxygen, or carbon dioxide, then polymer electrolytes or enzymes (glucose oxidase) producing analyte-related ionic species are used as components of the sensitive electrode coatings. Typical target analytes in the gas phase are nitrogen oxides, hydrogen sulfide, using Nafion polymer electrolyte [241–243], as well as oxygen [248, 253, 255], and carbon dioxide using liquid electrolytes [240]. Target analytes in the liquid phase comprise dissolved oxygen [235, 251], glucose [245, 246, 251], hydrogen peroxide [240, 247], and chlorine in drinking water [244, 249] Fig. 4.22. One of the best-known voltammetric cells is the Clark cell, which is based on a two-step-reduction of oxygen via hydrogen peroxide to hydroxyl ions in aqueous solution. The Clark cell is used to measure dissolved oxygen in blood and tissue [258]. The reference electrodes in the liquid phase are in most cases silver/silver-chloride elements. The interaction mechanism in all cases is an electrochemical redox reaction (2.12–2.15).
64
4 Microfabricated Chemical Sensors
100
(b)
PDMS
WE
75 CE
RE
current [pA]
(a)
50 25 0 hypochlorous acid -25 0
5
10 15 time [min]
20
25
Fig. 4.22. (a) Micrograph of an amperometric 3-electrode free-chlorine sensor with a central membrane-covered working electrode (WE) surrounded by the counter electrode (CE, ring) and the reference electrode (RE, ring segment). A polysiloxane (PDMS) encapsulation ring guides the liquid sample phase. (b) Sensor response upon exposure to chlorine from hypochlorous acid near the limiting threshold (ppb range). Reprinted from [249] with permission
4.4.2 Potentiometric Sensors Potentiometry is the direct application of the Nernst equation (2.15) through measurement of the potential between nonpolarized electrodes (WE and RE) under conditions of zero current. The measurement is carried out at thermodynamic equilibrium. In the following, it will be distinguished between two different types of potentiometric devices: • Electrochemical cell with metal electrodes. • Field-effect semiconductor devices. 4.4.2.1 Electrochemical Cell Transduction Principle and Sensing Characteristics The electrochemical cell used for potentiometric microsensors consists of two metal electrodes, a WE covered with a ion-selective membrane or gel, which preferably hosts a specific target ion, and an RE, which, in most cases, is a silver electrode covered by a thin silver chloride film [259, 260]. The WE is termed an ion-selective electrode (ISE). Both electrodes are on the same chip and are simultaneously exposed to the analyte phase (Fig. 4.23a). The ISE is, in principle, the oldest solid-state chemical sensor [261]. The design of modern micromachined nonsymmetrical ISEs, however, is completely different from that of conventional symmetrical ISEs such as a pH-glass-electrode or a
4.4 Electrochemical Sensors
(a) electrochemical cell
(b) concentration cell electrode 1
ion-selective membrane
Ag/AgCl reference
65
electrode 2
perm-selective membrane
metal electrodes (Pt) silicon substrate solution 1
solution 2
a1 (A+)
a2 (A+)
Fig. 4.23. (a) Schematic of a potentiometric nonsymmetrical electrochemical cell with metal electrodes. Ion-selective electrode (ISE) and RE are on the same chip exposed to the analyte. The RE is protected by a membrane. (b) Schematic of a classical symmetrical potentiometric concentration cell. The potential is measured between two half-cells containing different activities/concentrations of the same analyte (A+ )
lanthanium-fluoride membrane electrode. The traditional symmetric arrangements exhibit a liquid phase reservoir separated by a perm-selective membrane from the analyte phase and, thus, essentially constitute electrochemical concentration cells (Fig. 4.23b). The charge transfer processes at all interfaces (solution/solid electrolyte, solid electrolyte/metal, ionic conductor/electronic conductor) of a modern, nonsymmetrical ISE must be carefully designed. If the exchange current den2 sity of the charged species of interest is sufficiently high (i.e., >10−3 A/cm ), such interfaces are well defined, and the devices are stable. The exchange current is due to significant and continuous movement of charge carriers in both directions through the interphase region at an electrode at equilibrium (dynamic equilibrium). The magnitude of these mutually compensating currents (no net current) flowing at any zero-current potential is called exchange current. The Nernst equation (2.15) can be applied to calculate the electrochemical potential evoked by a certain analyte concentration. The design of metal-electrode potentiometric sensors is very similar to the voltammetric sensors (two-electrode configuration) described in Sect. 4.4.1. In comparison to voltammetric techniques it should be noted that the concentration dependence of the measured potential is logarithmic (2.15). Using amperometric techniques, the target ions can be selected by careful tuning of the appropriate redox potential. In potentiometry, all ions in the sample exhibiting comparable exchange current density contribute to the measured overall potential. In order to achieve selectivity to one specific ion, this target ion must provide a significantly higher exchange current density than the
66
4 Microfabricated Chemical Sensors
other interfering ions. This condition can be achieved by incorporating selective binding sites or ionophores in a membrane or gel material. The sensitive coating hence has to provide selectivity. Fabrication The fabrication is very similar to that of voltammetric devices and is, therefore, not further specified here. Sensor processing sequences are given in [259, 262, 263]. The fabrication of potentiometric sensors integrated with CMOS circuitry components is described in [263, 264]. Applications Prototype applications include chemical analysis in the gas or liquid phase. Typical target analytes in the gas phase are nitrogen oxide, sulfur dioxide and carbon dioxide using sintered ceramic electrolytes (sodium, barium and silver sulfate or Nasicon) [262, 265]. Target analytes in the liquid phase comprise all kinds of ionic species like hydrogen (pH: “potentia hydrogenii”, negative decadic logarithm of the hydrogen ion concentration), potassium, ammonium, calcium, chloride, cyanide, or nitrate using ionophores in polymeric membranes [259, 260, 263, 264, 266] or chalcogenide glasses [267]. The interaction mechanism in all cases is an electrochemical chargetransfer reaction (2.12–2.15). 4.4.2.2 Field-Effect-Based Devices Transduction Principle and Sensing Characteristics The field-effect-based microfabricated potentiometric sensors, like metaloxide semiconductor (MOS) devices in electronics, rely on variations in the charge distribution within the semiconductor surface space-charge region. The three field-effect device structures commonly used include the MOS capacitor (MOSCAP), the MOS diode (MOS-diode), and the MOS fieldeffect transistor (MOSFET) [4, 8, 11, 34]. These structures are displayed in Fig. 4.24. It is important to point out, “that one of the most critical thrusts of the electronics industry’s efforts to develop commercial field-effect devices has been predicated by the need to isolate these devices from any variation in their chemical environment” [34], since the field-effect device characteristics are very sensitive to such variation. Those efforts of the electronics industry are in diametrical opposition to developing, e.g., FET-based chemical sensors. By replacing, the MOSFET metal gate with an ionic solution and a reference electrode immersed into this solution, ion-sensitive device structures have been developed, the basic structure of which is analogous to the respective MOS devices: Ion-sensitive capacitor (ISCAP), ion-controlled diode (ICD) and ion-sensitive field effect transistor (ISFET) (Figs. 4.24, 4.25). The gate region is exposed to any ion present in the solution [4, 8, 11, 34]. The device applications will be discussed in the individual device-related sections, whereas the fabrication of the family of field-effect devices will be detailed in summary at the end in Sect. 4.4.2.2.6.
4.4 Electrochemical Sensors
(a) MOSCAP
(b) MOS-Diode
gas phase
p-silicon
gas phase metal gate (Pd)
metal (Pd) U
Si-oxide
(c) MOSFET
gas phase
metal (Pd) U
67
Ug Si-oxide
gate oxide n n source drain
p-silicon
p-silicon Ud
capacitance
current
drain voltage constant
∆Ugas
∆Ugas
∆Ugas
hydrogen no hydrogen
hydrogen no hydrogen voltage
drain current
hydrogen no hydrogen voltage
gate voltage
Fig. 4.24. Schematic representation of the different MOS field-effect devices: (a) capacitor, (b) diode, (c) transistor. Ug denotes the gate voltage, Ud the source-drain voltage. Characteristic sensor responses (voltage shift upon exposure to hydrogen) are given at the bottom
4.4.2.2.1 MOS Field-Effect Transistors, MOSFETs, and Ion-Selective Field-Effect Transistors, ISFETs (Chemotransistors) MOSFET Transduction Principle and Sensing Characteristics A MOSFET in electronics is a transistor, the source drain current (conductance) of which is modulated through an electric field perpendicular to the device surface. This electric field is generated by an isolated gate electrode and influences the charge carrier density in the conductance path between source and drain (semiconductor field effect). The MOSFET as used for chemical sensing has, e.g., a p-type silicon substrate (bulk) with two n-type diffusion regions (source and drain). The structure is covered with a silicon-dioxide insulating layer, on top of which a metal gate electrode (originally palladium [22], later other platinide metals [268, 269, 271, 272]) is deposited. When a positive voltage (with respect to the silicon) is applied to the gate electrode, holes, which are the majority carriers in the p-substrate, are depleted near the semiconductor surface. Upon applying a voltage between drain and source (Ud ), the electrons from the n-doped source can pass through this depleted surface region to the drain so that a conducting n-channel between source and drain is generated in the p-substrate near the silicon/silicon dioxide interface. The conductivity of this n-channel, i.e., the magnitude of
68
4 Microfabricated Chemical Sensors
the source-drain current (Id ) can be modulated by adjusting the strength of the electrical field perpendicular to the substrate surface between gate electrode and the silicon substrate [22, 268, 269]. Applications Palladium (Pd)-gate FET structures were demonstrated to function as hydrogen sensors by Lundstr¨ om [22] and others [270,273,274]. Hydrogen molecules readily absorb on the gate metal (platinum, iridium, palladium) and dissociate into hydrogen atoms. These H-atoms can diffuse rapidly through the Pd and absorb at the metal/silicon oxide interface partly on the metal, partly on the oxide side of the interface [268,269]. Due to the absorbed species and the resulting polarization phenomena at the interface, the drain current (Id ) is altered, and the threshold voltage (Ud ) is shifted. The voltage shift (∆Ud ) is proportional to the concentration or coverage of hydrogen at the oxide/metal interface. The presence of oxygen promotes the formation of water at the gas phase/metal interface due to the catalytic reaction of atomic hydrogen with atomic oxygen. With thin catalytic metal gates (containing holes and cracks) ammonia [269,276], amines, and any kind of molecule that gives rise to polarization in a thin metal film (hydrogen sulfide, ethene, etc.) or causes charges/dipoles on the insulator surface, can be detected [278,279]. Detailed models for those processes, however, do not yet exist. Sensitivity and selectivity patterns of gas-sensitive FET devices depend on the type and thickness of the catalytic metal used, the chemical reactions at the metal surface, and the device operation temperature. For extremely high temperatures (600◦ –800◦ C), silicon carbide devices have been developed [222–225]. A paramount problem with MOSFET sensors has been long-term drift, which seems to be mitigated to some extent by the deposition of a thin alumina layer between the Pd gate and the silicon oxide [277]. The temperature should be kept constant. Alternative gate materials include polyaniline for the detection of water and ammonia [278, 279], and high-temperaturesuperconducting cuprate to monitor ammonia and nitrogen oxides [280]. ISFET Transduction Principle and Sensing Characteristics For the case of the ISFET, the gate metal electrode of the MOSFET is replaced by an electrolyte solution, which is in contact with the reference electrode, i.e., the silicon gate oxide is directly exposed to aqueous electrolyte solution Fig. 4.25 [21]. An external reference electrode is required for a stable operation of an ISFET [4, 8, 11, 34, 281–284]. Unfortunately, including such a reference is nontrivial and subject to dedicated research [34, 285–289]. An electric current (Id ) flows from the source to the drain via the channel, and, like in MOSFETs, the channel resistance depends on the electric field perpendicular to the direction of the current. It additionally depends on the potential difference across the gate oxide. Therefore, the source-drain current, Id , is influenced by the interface potential at the oxide/aqueous solution boundary. Though the electric resistance of the channel provides a measure
4.4 Electrochemical Sensors
(a) MOSFET
(b) ISFET
gas phase metal gate (Pd) Ug
69
U
reference electrode liquid phase
gate oxide n n source drain
gate oxide n n source drain
p-silicon Ud
p-silicon Ud
Fig. 4.25. Schematic representation of a MOSFET (a) and an ISFET structure (b). Ug denotes the gate voltage, Ud the source-drain voltage. By replacing the metal gate of the MOSFET with an ionic solution and a reference electrode immersed into this solution, the ISFET has been developed
for the gate oxide potential, the direct measurement of this resistance gives no indication of the absolute value of this potential. However, at a defined source-drain potential (Ud ), changes in the gate potential can be compensated by a modulation of Ug . This adjustment can be carried out in a way that the changes in Ug applied to the reference electrode exactly compensate for the changes in the gate oxide potential. This is automatically performed by ISFET amplifiers with feedback, which allow for obtaining a constant source-drain current. In this particular case, the gate-source potential is determined by the surface potential at the insulator/electrolyte interface. Mechanistic studies of the processes occurring at the solution/gate oxide interface (site binding model [290]) and the oxide/semiconductor interface can be found in the literature [4, 8, 11, 34, 283, 284, 290–293]. The insulator/solution interface is assumed to represent in most cases a polarizable interface, i.e., there will be charge accumulation across the structure but no net charge passing through.Interfaces with net charge passing through are termed “faradaic”, see, e.g., [4, 8, 11, 34, 229]. Applications The gate oxide surface contains reactive Si-OH groups, which, besides providing pH sensitivity, can be used for covalent attachment of a variety of organic molecules and polymers. pH-FET (pH: “potentia hydrogenii”, negative decadic logarithm of the H+ -ion concentration).The basic ISFET is an exposed-gate-oxide FET and functions as a pH sensor [21]. The surface of the gate oxide contains OHfunctionalities, which are in electrochemical equilibrium with ions in the sample solutions (H+ and OH− ). The hydroxyl groups at the gate oxide surface can be protonated and deprotonated and, thus, when the gate oxide contacts an aqueous solution, a change of pH will change the silicon oxide surface potential. A site-binding model describes the signal transduction as a function
70
4 Microfabricated Chemical Sensors
sensor signal [mV]
-1450
-1470
-1490
-1510 0
20
40
60 80 time [sec]
100
120
140
Fig. 4.26. Dynamic response of a pH-ISFET in a flow-through configuration upon repeated pH-changes by one unit (4 to 5). The signals exhibit a stable baseline and a good reproducibility. According to the Nernst equation (2.15), a pH-change of one unit causes a voltage change of 59 mV. Reprinted from [299] with permission
of the state of ionization of the amphoteric surface Si-OH groups [294, 295]. The change in the charge state of these sites leads to a variation of the dipole layer and consequently to a change in the semiconductor space charge region [4, 34, 276, 293]. Typical pH-sensitivities measured with silicon-oxide ISFETs are 37-40 mV per pH unit [295]. Sensor signals achieved with a silicon nitride membrane are given in Fig. 4.26 [299]. Inorganic gate materials for pH sensors like silicon nitride (CMOS process material) [296–302], oxynitride [303], alumina [304], tantalum oxide [304,305], and iridium oxide [306] have better properties than silicon oxide with regard to pH response, hysteresis and drift (see Fig. 4.26). In practice, these layers are deposited on top of the first layer of silicon oxide by means of chemical vapor deposition (CVD). CHEMFET The CHEMFET or chemically sensitive FET [307,308], is a modification of an ISFET with the original inorganic gate material covered by organic ionselective membranes like polyurethane, silicone rubber, polystyrene, polyamide and polyacrylates containing ionophores (Fig. 4.27). A critical point of the CHEMFET is the attachment of the sensitive membrane, which can be improved by mechanical [309] or chemical [308, 310, 311] anchoring to the surface of the gate oxide. CHEMFETs selective to K+ [308, 312–315], Na+ [316–318], Ag+ [319], transition metal cations (Pb2+ , Cd2+ ) [320–322] and some anions (NO− 3 ) [323–325] have been developed. Highly specific organic or biological compounds can be incorporated in the membrane as well: Enzymes (ENFET) [326, 327] like glucose oxidase for the detection of glucose [326, 328–330], and penicillinase for the detection
4.4 Electrochemical Sensors
(b)
Ug
reference electrode
membrane hydrogel gate oxide n n source drain p-silicon Ud
resin
sensor response [mV]
(a) chemFET
liquid phase
71
320
a
b
a c
b
a
c
160
0 0
3
6
9
time [min] a: 10-5 M lead nitrate in acetate buffer b: 10-4 M lead nitrate in acetate buffer c: 10-2 M acetate buffer, pH: 5.5
Fig. 4.27. (a) Schematic representation of a CHEMFET, a modification of an ISFET with the original inorganic gate material covered by an organic ion-selective membrane and a hydrogel. (b) Sensor response of a CHEMFET with a lead-selective membrane upon different target analyte concentrations in an acetate buffer solution: 10−5 molar lead nitrate solution (a), 10−4 molar lead nitrate solution (b), and pure buffer solution (c). Redrawn, adapted from [322]
of penicillin [326]. Other enzymes can be used to detect pesticides and organophosphorous compounds [331–334] or aldehydes [335]. Antibodies can be immobilized for immunoreactions (IMFET), and biological entities like whole cells [336–338] or insect antennae [339] have been used (BIOFET). Differential CHEMFET Configuration CHEMFETS can be applied in a differential measurement configuration [340– 347] (Fig. 4.28a). This device configuration offers the advantage that a variety of experimental parameters and external disturbances, which affect both FET structures (e.g., light and temperature), cancel out. The inorganic gate of one of the CHEMFETS has to be rendered insensitive to pH or other target analytes by chemical surface treatment. This “insensitive” FET (sometimes denoted reference FET, REFET) should ideally show no response to the target species present in the sample phase (Fig. 4.28b) [348]. pH-sensitivity, e.g., can be suppressed by plasma-deposition of a hydrophobic polymer on the gate [349–353], or by realizing a reservoir of buffered solution with constant pH separated by a membrane from the sample phase Fig. (4.28) [348]. The applications comprise similar target analytes as for the case of CHEMFETS. 4.4.2.2.2 MOS Diode and Ion-Controlled Diode, ICD (Chemodiodes) Transduction Principle and Sensing Characteristics The chemical sensing mechanism is identical with that of the MOSFET, only the transduction method is slightly different. When the gas molecules (hydrogen) diffuse to the metal/oxide interface to form a polarization layer
72
4 Microfabricated Chemical Sensors
(b)
(a) differential FET configuration gas phase nonbuffered hydrogel layer
gas-permeable buffered membrane hydrogel layer
gate oxide
Ug
n
source drain p-silicon
gate oxide
n
a
n Ud
n
drain source
b
Ug
EMF, voltage [mV]
Ag/AgCl reference electrodes
2
500
10-4 M 1
10-3 M
3
400
10-2 M 3•10-2 M
1: CO2-FET 2: insensitive FET 3: differential signal
300 0
10
20 time [min]
30
Fig. 4.28. (a) Schematic of a differential FET configuration for sensing carbon dioxide. The carbon-dioxide FET (formation of “carbonic acid”, pH-FET) exhibits a nonbuffered hydrogel layer, whereas the insensitive FET has a buffered one. The pH of the latter will not change upon carbon dioxide absorption as shown in the sensor responses (b). The differential configuration thus eliminates effects common to both devices (temperature fluctuations, flow disturbances etc.). Redrawn/adapted from [348]
at the interface, the height of the energy barrier of the diode is altered. This leads to a change in either the forward voltage or the reverse current of the diode. The diode characteristics are hence shifted along the voltage axis (see Fig. 4.24), as it is the case with the MOSFET [22, 268, 269, 354–357]. The response time and sensitivity of the sensor can be improved by operation at elevated temperature. Therefore, sensing structures have been placed on thermally isolated membranes [358, 359]. The ion-controlled diode (ICD) was first described by Zemel [360]. Its operation is not too different from that of an ISFET and is analogous to that of the MOS-diode [8,34,293,360,361]. Polarization resulting from ion adsorption on the insulator (silicon oxide) causes changes in the effective forward voltage or reverse current, which can be measured. A more sophisticated structure with a through-the-chip p-n-junction, which enables the contacts on the back side of the chip to be isolated from the aqueous electrochemistry occurring at the front (“gate”) side, is described in [34,360,361]. The applications are the same as for the FETs. 4.4.2.2.3 MOS Capacitor and Ion-Selective Capacitor (Chemocapacitors) Transduction Principle and Sensing Characteristics The sensor element is a standard MOS or electrolyte insulator semiconductor (EIS) structure as shown in Fig. 4.24. Again, the target analyte species change the polarization at the metal/oxide or electrolyte/oxide interface thus affecting the flat-band voltage of the capacitor. The capacitance-voltage curve of the capacitor is shifted by a certain amount, which is proportional to the target analyte concentration in the gas or liquid phase. One can either record
4.4 Electrochemical Sensors
73
capacitance [nF]
1000
porous silicon
800 600 400 200 0 -0.8
voltage [mV]
the capacitance as a measure of the analyte concentration or use some circuitry to keep the capacitance constant by varying the necessary bias voltage. Capacitor-type structures are straightforward to realize. The applications (MOSCAP [4,268,269,273,362,363], ISCAP/EIS [364– 367]) are similar to those of the FETs. A set of capacitance/voltage curves is shown in Fig. 4.29 [366].
-100
-300 4
5
6
7
8
pH
pH 8 pH 7 pH 6 pH 5 pH 4
-0.4
40 mV / pH
-200
0 voltage [V]
0.4
0.8
Fig. 4.29. Set of capacitance/voltage curves for a porous electrolyte-insulatorsemiconductor (EIS) structure exposed to solutions of different pH. The insert shows the calibration curve (pH versus voltage) Adapted from [366] with permission
4.4.2.2.4 Measuring Work Functions: Kelvin Probe and Suspended-Gate Field-Effect Transistor, SGFET Transduction Principle and Sensing Characteristics The work function is defined as the minimum work required to extract an electron in vacuum from the Fermi level of a conducting phase through a surface and place it outside the reach of electrostatic forces at the so-called vacuum level [368, 369]. When two different electronic conductors are in contact, electrons flow from the material with the lower electron affinity to that with the higher one according to the difference in the (electro-)chemical potential (2.5, 2.6, 2.13) of the electrons until an equilibrium is reached. A contact or Galvani potential arises, which represents the bulk-to-bulk inner contact potential of the two materials or the difference of the Fermi levels of the two materials. If surfaces of two different materials are parallel and separated by a very thin insulator or air gap, a potential across the gap is formed, the Volta potential or outer potential, which represents the difference of the work functions of both materials. A palladium plate separated by a thin air gap (few
74
4 Microfabricated Chemical Sensors
µm) from a copper plate will become positively charged, and correspondingly, the copper plate will be negatively charged. The work function of a certain material, which includes the chemical potential of the electrons (changes in electron affinity, eventual band bending due to electron transfer) and the surface dipole field (which exists even at absolutely clean surfaces) is changed upon formation of surface adsorbates, e.g., from the gas phase. Therefore, work function measurements can be used to advantage in chemical or gas sensing. The Kelvin probe relies on the displacement of one of the surfaces in a periodic oscillation. This oscillation induces charges across the surfaces and generates a sinusoidal current in the sensing plate, which is proportional to the work function difference between the sensing plate and the reference plate. This current thus directly depends on the surface chemistry of the plates. Micromachined Kelvin probes with a metal sensing film supported by a dielectric membrane and a 2.5-µm-thick silicon reference plate that is electrostatically deflected by a drive electrode (schematically depicted in Fig. 4.30a) have been fabricated and have been used to detect the surface adsorption of oxygen [370].
(a) integrated Kelvin probe heater Si-wafer
3 µm glass substrate drive electrode
temperature sensor
(b) suspended-gate FET metal gate (Pd) dielectric membrane metal sensing film electroplated contacts
movable Si-reference plane
Ug
gas phase
chemically sensitive layer
gate oxide
n
n
source
drain
p-silicon
Ud
Fig. 4.30. Schematic representation of a micromachined Kelvin probe (a) and a suspended-gate FET (b). For details, see text. Redrawn from [370] (a) and [368] (b)
The SGFET (or even the standard MOSFET) is very similar to the Kelvin probe [368, 371]. The notion of work functions was not used in the context of FETs yet, though FET transducers are sometimes denoted “work function devices”. A metal plate (suspended gate) is separated by an air gap (or in the case of the MOSFET silicon oxide) from a silicon plate (Fig. 4.30b) [368]. A variable voltage source (Ug ) is located between the back of the silicon and the gate metal. The plate distance, however, cannot be varied. Therefore, the drain-source current is used to interrogate the Volta potential. The magnitude of the drain-source current at a certain applied voltage depends on the
4.4 Electrochemical Sensors
75
work function difference between the metal gate and the silicon. It therefore directly depends on surface adsorption or absorption chemistry in sensitive layers applied between the suspended gate and the silicon. Applications Sensitive layers applied in the gap include metal oxides to detect ammonia, carbon monoxide or nitrogen oxides [372–376], potassium iodide to detect ozone [377], palladium/polyaniline to detect hydrogen and ammonia [378], and polypyrrole to detect alcohols and volatile organics [379]. 4.4.2.2.5 Light-Addressable Potentiometric Sensor, LAPS Transduction Principle and Sensing Characteristics This type of sensor is based on the field effect as well [171], and its working principle is very closely related to that of FET devices (Fig. 4.31) [171, 380–383]. The LAPS device is a thin silicon plate (thinned down to a few µm thickness) with an approximately 100-nm-thick oxynitride layer in contact with an electrolyte solution. A potential is applied between the silicon plate and, e.g., a silver/silver chloride controlling electrode immersed in the electrolyte solution. The controlling electrode simultaneously serves as reference electrode. The sign and magnitude of the applied potential are adjusted so as to deplete the semiconductor of majority carriers at the insulator interface. Upon illumination of the plate with LEDs, hole-electron pairs are created, which can reach the depletion area (Fig. 4.31). Due to the charge separation in the depletion area, a photocurrent flows through the device. In this way, a sinusoidally modulated light beam causes a sinusoidal photocurrent, the amplitude of which depends on the width of the depletion charge
LAPS control electrode
I
chemically sensitive layer
liquid phase solution
U
insulator depletion region e-h pairs
p-silicon light
Fig. 4.31. Schematic representation of a light-addressable potentiometric (LAPS) device. For details, see text. Redrawn from [386, 387]
76
4 Microfabricated Chemical Sensors
region: The larger the depletion region, the larger the photo current. The photocurrent amplitude thus depends on the absorbed species in the sensitive layer or, e.g., on the pH of the solution in contact with the insulator. The current-voltage curve is of sigmoidal shape and is shifted along the voltage axis due to chemical changes (e.g., pH-changes) at the solution/silicon oxide interface in analogy to other field-effect devices. Applications The silicon plate is straightforward to manufacture, and arrays of LEDs and different selective layers can be used on the same chip [371]. Assessment of the lateral resolution [384,385] and benchmarking against ISFETs [386] have been performed. LAP methods have been applied to the gas phase as well [380]. Applications include monitoring cell activity via pH changes resulting from cell activity and metabolism [387–390], and biosensing in liquids [381, 383, 391]. LAPS devices have been developed by Molecular Devices Inc. [392]. 4.4.2.2.6 Field-Effect Device Fabrication • Patterning of metal electrodes (lift-off, thermal evaporation, sputtering) as described for MOSFETs in [267–276] • Deposition of additional metal oxides/nitrides by LPCVD (tantalum oxide, alumina, silicon nitride) as described for MOSFETs and ISFETs in [296–306] • Optional membrane formation for temperature stabilization by back etching [358, 359] • Surface micromachining (HF etching, Al-etching) for the SGFET [374,393] • Deposition of electro-active polymers, membrane materials, hydrogels by spin-casting, spraying, screen printing, photolithography for CHEMFETs [278–280, 307–335, 349–353] The fabrication of field-effect electrochemical sensors integrated with CMOS circuitry is described in various publications [11,296,297,300,302,303,318,362, 394]. A modular chip system based on a sensing and a “service” chip has been described in [279], a flow-through ISFET in [395]. Back-side-contact ISFETs are detailed in [396, 397]. pH-ISFETS are commercially available from, e.g., Honeywell [398]. 4.4.3 Conductometric Sensors Conductometric techniques are a special case of AC-impedance techniques. Instead of the real and imaginary component of the electrode impedance at different frequencies, only the real-valued resistive component, related to the sample (sensing material) resistance, is of interest. Since complex impedances include capacitive and inductive contributions, chemocapacitors that do not rely on the field effect are included here, in the conductometric section. The section on conductometric sensors is hence organized in two parts, one on
4.4 Electrochemical Sensors
77
resistance measurements at room temperature and elevated temperatures (chemoresistors) and the other on chemocapacitors. 4.4.3.1 Chemoresistors Transduction Principle and Sensing Characteristics Chemoresistors rely on changes in the electric conductivity of a film or bulk material upon interaction with an analyte. Conductance, G, is defined as the current, I [A], divided by the applied potential, U [V]. The unit of conductance is Ω−1 or S (Siemens). The reciprocal of conductance is the resistance, R[Ω]. The resistance of a sample increases with its length, l, and decreases with its cross-sectional area, A: R=
U 1 l 1 = = · . G I κ A
(4.8)
Conductivity or specific conductance, κ[1/Ωm], is hence defined as the current density [A/m2 ] divided by the electrical field strength [V/m]. The reciprocal of conductivity is resistivity, ρ[Ωm]. The conductivity can be thought of as the conductance of a cube of the probed material with unit dimensions [11]. Conductometric sensors are usually arranged in a metal-electrode-1/ sensitive-layer/metal-electrode-2 configuration [4]. The conductance measurement is done either via a Wheatstone bridge arrangement or by recording the current at an applied voltage in a DC mode or in a low-amplitude, lowfrequency AC mode to avoid electrode polarization. In Fig. 4.32a [399], a conductance cell (in this case metal oxides) and the respective equivalent electric circuits are depicted. The goal of conductometry is to determine the sample free molecules
adsorbed molecules
electrodes
e-
U
I
(b)
sensitive layer
esubstrate
b
d
c U
4-electrode conductance cell
catalyst
(a)
a
2-electrode conductance cell
I
a
a. contacts b. surface c. bulk d. grain boundaries substrate
electrodes
Fig. 4.32. (a) Schematic representation of a conductance cell (in this case a semiconductor sensor with tin dioxide as sensitive layer), of the different contributions (contacts, surface, bulk and grains) to the overall conductivity and of the respective equivalent circuits. Redrawn from [399]. (b) Schematic representation of a two-electrode and four-electrode conductance cell. For details, see text
78
4 Microfabricated Chemical Sensors
resistance (c). The lead wire resistances normally can be neglected. The electrode impedance (a) consists of two elements, the contact capacitance, and the contact resistance. By applying an AC potential, an AC current will flow through the resistor cell. If the contact capacitance is sufficiently large, no potential will build up across the corresponding contact resistance. The contact resistance should be much lower than the sample resistance and be minimized, so that the bulk contribution dominates the measured overall conductance. If surface conductivity mechanisms differing from those in the bulk occur, this can be modeled by adding an additional surface resistance (b) to the equivalent circuit. A grain boundary in the sensing material constitutes a resistance-capacitance unit (d). The conductivity depends on the concentration of charge carriers and their mobility, either of which can be modulated by analyte exposure. In contrast to potentiometry and voltammetry, conductometric measurements monitor processes in the bulk or at the surface of the sample. Any contribution of electrode processes has to be avoided. Therefore, in most cases, a four-electrode configuration is preferred over a simple two-electrode configuration (Fig. 4.32b). The outer pair of electrodes is used for injecting an AC current into the sample, the potential difference is then measured at the inner pair of electrodes. The interference of electrode impedances on the measurement results is thus excluded. 4.4.3.1.1 Low-Temperature Chemoresistors Several classes of predominantly organic materials are used for application with chemoresistors at room temperature. The chemically sensitive layer is applied over interdigitated electrodes on an insulating substrate. Electrode spacing is typically 5 to 100 µm, and the total electrode area is a few mm2 . The applied voltage ranges between 1 and 5 V. Metal-phthalocyanines constitute organic p-type semiconductors. The adsorption of oxidizing agents such as, e.g., nitrogen oxide or ozone hence decreases the resistance by increasing the number of holes in the conduction band [8]. Metal phthalocyanines at elevated operation temperatures (approx. 180◦ C) have been used to monitor nitrogen oxide [400, 401], ozone [402], hydrogen chloride [401] and even ammonia [403, 404]. Conducting Polymers such as polypyrroles, polyaniline and polythiophene exhibit a large conjugated π-electron system, which extends over the whole polymer backbone. Partial oxidation of the polymer chain then leads to electrical conductivity, because the resulting positive charge carriers (denoted polarons or bipolarons) are mobile along the chains [9]. Counteranions must be incorporated into the polymer upon oxidation to balance the charge on the polymer backbone. The conducting polymers, however, do not only react with oxidizing agents, but also respond to a wide range of organic vapors [405–408]. The underlying principle of this response is still unclear; suggestions include: (I)
4.4 Electrochemical Sensors
79
vapor molecules could affect the charge transfer between polymer and the electrode contact, (II) analyte molecules could interact with the mobile charge carriers on the polymer chains or (III) with the counterions and thus modulate the mobility of the charge carriers, or they could (IV) alter the rate of interchain hopping in the conducting polymer [379, 409, 410]. Applications include the detection of a variety of polar organic volatiles like ethanol, methanol, components of aromas, [405–408, 411–420] and others. Conducting polymers show a high cross-sensitivity to water. Sensors are commercially available from, e.g., Osmetech and Marconi [416]. In Carbon-Black-Loaded Polymers, conducting carbon black is dispersed in non-conducting polymers deposited onto an electrode structure. The conductivity is by particle-to-particle charge percolation so that if the polymer absorbs vapor molecules and swells, the particles are, on average, further apart (Fig. 4.33), and the conductivity of the film is reduced [417].
(a) no analyte present carbon particles
resistance R1
(b) after analyte absorption polymer
resistance R2 >R1
Fig. 4.33. Conductivity by particle-to-particle charge percolation in carbon-loaded polymers: (a) with no analyte present, (b) during organic volatile exposure, which causes polymer swelling
Applications include monitoring organic solvents such as hydrocarbons, chlorinated compounds, and alcohols [418–421]. Sensors are commercially available from, e.g., Cyrano Sciences [422]. Hydrogels responsive to pH-changes have been applied to interdigitated electrode arrays. The hydrogel swells or shrinks to a hydration determined by the pH of the analyte solution. This leads to a corresponding increase or decrease in the mobility of the ions partitioned by the gel. The sensitivity of ion mobility to small changes in hydration causes large resistance changes [423, 424]. A conductometric variant of a Severinghaus electrode (detection of carbon dioxide via dissolution in water, formation of carbonic acid and monitoring of the pH change [4]) with a liquid reservoir has been described in [425].
80
4 Microfabricated Chemical Sensors
Fabrication • Patterning of metal electrodes (lift-off, thermal evaporation, sputtering) [405–408, 418–421] • Optional membrane formation by back etching for temperature stabilization [414, 426] • Deposition of carbon-loaded polymers, membrane materials, hydrogels by spin-casting, spraying, screen printing, and photolithography [418–421,423, 424] • Deposition of conducting polymers by electrochemical deposition; sensor selectivity is modified by changing the counterion used in the polymerization process [405–408] The fabrication of an impedance device [427], and microbridges [415] integrated with CMOS circuitry components have been described. Complete processing sequences are detailed in [423, 425]. 4.4.3.1.2 High-Temperature Chemoresistors (Hotplate Sensors) There is a wealth of literature on semiconducting metal oxides and related sensors, the focus of this work, however, will be on silicon-based micromachined hotplates. A typical high-temperature chemoresistor includes an integrated heater, a thermometer and a sensing film on a thermally isolated stage such as a membrane (Fig. 4.34) [428]. Isolated micromachined structures (hotplates) exhibit very short thermal time constants on the order of milliseconds. The sensitive materials used with the hotplate sensors include widebandgap semiconducting oxides such as tin dioxide, gallium oxide, indium oxide, or zinc oxide. In general, gaseous species acting as electron donors (hydrogen) or acceptors (nitrogen oxide) adsorb on the metal oxides and form (a)
(b)
Fig. 4.34. Micrograph of a microhotplate (a), and schematic top view and side view of the device (b). The suspended plate exhibits a polysilicon heater, an aluminum plane for homogenous heat distribution and aluminum electrodes for measuring the resistance of a semiconductor metal oxide. Reprinted from [428,443] with permission
4.4 Electrochemical Sensors
81
surface states, which can exchange electrons with the semiconductor. An acceptor molecule will extract electrons from the semiconductor and, therefore, decrease its conductivity. The opposite holds true for an electron–donating surface state. A space charge layer will thus be formed. By changing the surface concentration of donors/acceptors, the conductivity of the space charge region is modulated [4, 399, 429–431]. In addition to the above-mentioned interaction of surface adsorbates and related electronic effects, the diffusion of lattice defects from the bulk of the metal-oxide crystal also occurs (ionic conduction) at elevated temperatures (>600◦ C). The defects can act as donors or acceptors. Oxygen vacancies, e.g., act as intrinsic donors. The overall conductivity in polycrystalline samples includes contributions from the individual crystallites, the grain boundaries, insulating components such as pores, and the contacts (Fig. 4.32a). Thus, the conduction mechanism in ceramic polycrystalline samples is difficult to analyze, and a variety of empirical data has been published [4, 399, 429–431]. The most extensively investigated material, tin dioxide, is oxygen-deficient and, therefore, is an n-type semiconductor since oxygen vacancies act as electron donors. In clean air, oxygen, which traps free electrons by its electron affinity, and water are absorbed on the tin dioxide particle surface forming a potential barrier in the grain boundaries. This potential barrier restricts the flow of electrons and thus increases the resistance. When tin dioxide is exposed to reducing gases such as carbon monoxide, the surface adsorbs the gases, and some of the oxygen is removed by reaction of water and oxygen at the surface. This lowers the potential barrier, thereby reducing the electric resistance. The reaction between gases and surface oxygen depends on the sensor temperature, the gas involved, and the sensor material [4,399,429–432]. Semiconductor metal-oxide sensors usually are not very selective, but respond to almost any analyte (carbon monoxide, nitrogen oxide, hydrogen, hydrocarbons). One method to modify the selectivity pattern includes surface doping of the metal oxide with catalytic metals such as platinum, palladium, gold, and iridium [399, 430, 433]. Surface doping improves the sensitivity to reducing gases, reduces the response time and operation temperature and changes the selectivity pattern [399, 430, 433]. Most modern sensors operate in a regime, in which the overall conductivity is determined by nanocrystalline sensing materials. As a consequence of the small grain size (better surface to volume ratio), the relative interactive surface area is larger, and the density of charge carriers per volume is higher. This leads to more drastic conductivity changes and, hence, larger sensor responses in comparison to larger grains [399, 430]. Since microhotplates have a very low thermal mass, they allow for applying temperature-programmed operation modes [434–437]. By operating the device, e.g., in a cyclic thermal mode, reaction kinetics on the sensing surface
82
4 Microfabricated Chemical Sensors 1.0 ppm NO2 pure air
50 ppm CO
Rsensor[Ω]
100k
10k
close-up 1k temperature [°C]
mixture 50 ppm CO 1.0 ppm NO2
close-up
1 min
400 300 200
Fig. 4.35. Sinusoidal modulation of the operation temperature of a tin-dioxide sensor between 200◦ and 400◦ C (bottom) leads to characteristic frequency-dependent resistance features (upper part). Changes of the resistance (Rsensor ) of the micromachined sensor upon exposure to 50 ppm CO, 1 ppm NO2 and a mixture of 50 ppm CO and 1 ppm NO2 in synthetic air (50% relative humidity). Adapted from [436] with permission
are altered, producing a time-varying response signature that is characteristic for the respective analyte gas [434–437]. An example is given in Fig. 4.35 [436]. Fabrication • Additional deposition/patterning of metal electrodes using lift-off, thermal evaporation, and sputtering [438–441] • Back side (KOH) [426, 438–442] or front side (RIE, EDP, XeF2 ) [428, 443] etching for membrane formation • Deposition of metal-oxide materials by LPCVD, sol-gel processes, sputtering, and screen printing [444–449] • Sintering of the metal oxides (annealing) at elevated temperatures [450] The fabrication of hotplates on a CMOS-substrate is described in [428, 443]. For details on CMOS-based microhotplate gas sensor systems, see also Sect. 5.4. Complete processing sequences are detailed in [451, 452]. The fabrication of a 39-electrode array with a tin-dioxide-gradient coating has been reported in [453, 454]. Extensive reliability studies are reported in [455–457]. Investigations and simulations to optimize the power consumption are under way (see, e.g., [458–460]). Devices are commercially available from Figaro, Marconi, Capteur and MICS [461].
4.4 Electrochemical Sensors
83
Applications Typical applications include the detection of hydrogen [428], oxygen [428], nitrogen oxide [438, 462], CO [439, 440], and a variety of organic volatiles [441–444, 463] using tin dioxide as sensitive layer. Thermal cycling of the hotplate structures allows for detection of gas mixtures (CO and nitrogen oxide) with a single sensor [436] (Fig. 4.35). In other applications, temperature profiles were optimized to specifically detect selected organic volatiles [443, 463, 464]. Additional sensitive materials on hotplates include, e.g., niobium and titanium oxide to detect oxygen [465], as well as metal films covered with surface oxides. Reducing gases like hydrogen donate electrons to the metal and increase the conductance, whereas electron acceptor molecules like oxygen decrease the metal conductance [451, 466, 467]. An unheated tin-dioxide oxygen sensor (slow response) has been reported in [468]. 4.4.3.2 Chemocapacitors Transduction Principle and Sensing Characteristics Chemocapacitors (dielectrometers) rely on changes in the dielectric properties of a sensing material upon analyte exposure (chemical modulation of equivalent-circuit capacitors in Fig. 4.36 by changes in the dielectric constant of the sensitive layer). Interdigitated structures that are similar to those of room-temperature chemoresistors are predominantly used [469–472].
analyte
substrate
polymer
E1
E2
∆C Fig. 4.36. Schematic representation of an interdigitated capacitive sensor covered with a sensitive polymer layer. E1 and E2 denote the two sets of interdigitated electrodes, ∆C the capacitance change
In some cases, plate-capacitor-type structures with the sensitive layer sandwiched between a porous thin metal film (permeable to the analyte) and an electrode patterned on a silicon support are used to increase the sensitivity by trapping the electric field [473, 474].
84
4 Microfabricated Chemical Sensors
The capacitances usually are measured at an AC frequency of a few kHz up to 500 kHz. The analyte absorption in the sensitive layer on top of the electrodes induces a change in the layer dielectric properties and, consequently, a capacitance change that can be measured. For conducting measurements at defined temperatures, sensor and reference capacitors can be placed on thermally isolated membrane structures [475–477]. For more details on capacitive sensors and CMOS capacitive microsystems, see [478–480] and Sect. 5.1. Fabrication • Deposition/patterning of metal electrodes (lift-off, thermal evaporation, sputtering) [469–474] • Optional back etching, membrane formation for temperature stabilization [475–477, 481] • Deposition of polymers (spin-casting, spraying, photolithography) [469– 480] The fabrication of capacitors integrated with CMOS circuitry components is described in [470, 472, 475, 476, 478, 479, 482, 483]. Temperature-stabilized membranes have been reported in [475–477]. Capacitive humidity sensors are commercially available from, e.g., Sensirion, Vaisala, and Humirel [484]. Applications Typical applications areas are humidity sensing using polyimide films [469– 473, 482–484], since water has a relatively high dielectric constant of 78.5 (liquid state) at 298 K leading to large capacitance changes. A variant exhibits polyimide columns sandwiched between metal electrodes [485]. More recent applications include the detection of different kinds of organic volatiles in the gas phase (hydrocarbons, chlorinated hydrocarbons, alcohols etc.) using polymeric layers [475, 476, 478, 479, 486, 487] or liquid crystals [488], and the detection of nitrogen oxide [489], sulphur dioxide [490] and carbon dioxide [491, 492] using ceramic materials. The interaction mechanisms involve reversible physisorption and bulk/gas phase partitioning (see 2.10).
5 CMOS Platform Technology for Chemical Sensors
The aim in utilizing microfabrication techniques and, in particular, CMOS technology for realizing chemical sensors was to devise more intelligent, more autonomous, more integrated, and more reliable gas sensor systems at low costs in a generic approach. Since the sensor market is strongly fragmented, i.e., there exists a large variety of applications with different needs and sensor requirements, a modular approach or “toolbox strategy” relying on a platform technology was identified as the most promising attempt to achieve major progress. Once the platform technology has been chosen, the components of the toolbox such as transducers, sensor modules, and circuit modules can be developed, some of which afterwards can be assembled into a customized system that meets the respective applications needs. The application hence dictates the system architecture and the nature of its components: As soon as the target analytes and their concentrations as well as the boundary conditions of detection are specified, the optimum transducer, the necessary driving and signal conditioning circuitry as well as interface and communication units can be selected from the toolbox, and the different modules can then be combined to arrive at a custom-designed sensor system. A multitude of development activities is necessary to obtain all the modules needed for the toolbox: (a) Design and miniaturization of transducers and directly related electronic components (potentiostats, heaters, amplifiers, etc), (b) development of digital-to-analog and analog-to-digital conversion units, interface and communication units, (c) development of additional and auxiliary functions, which are pivotal for the system performance (e.g., temperature control, temperature sensors, humidity sensors), and (d) development of dedicated microsystem packaging solutions, which are suitable for chemical or gas analysis. It is important to note that the package has to be thought of already in the initial conception phase of a microsystem, since the design and architecture of a microsystem heavily depend on the envisaged packaging concept. A fundamental issue concerns the sensor system implementation, i.e., the decision to either realize monolithic systems combining CMOS circuitry and sensor structure on the same chip (CMOS-MEMS), or to develop hybrid designs that rely on optimized sensor materials and fabrication processes
86
5 CMOS Platform Technology for Chemical Sensors
with external electronics. There is a number of aspects that have to be taken into account in making such a decision. Materials and Fabrication Processes For monolithic designs, the selection of materials is restricted to CMOSmaterials (silicon, polysilicon, silicon oxide, silicon nitride, aluminum, see also Sect. 3.3) and CMOS-compatible materials. There is also a limitation on available fabrication processes owing to a limited number of pre-CMOS and post-CMOS micromachining options. High-temperature steps (e.g., > 400◦ C) are detrimental to the aluminum metallization (metal oxidation, diffusion) and alter the transistor characteristics (last high-temperature step of the CMOS process is at approx. 380◦ –400◦ C). For hybrid designs, any material or the optimum sensor material can be used, and a wealth of micromachining techniques for all kinds of materials is available through prototyping services. Time to Market, Costs The production of monolithic designs relies on established industrial CMOS technology to fabricate the circuitry and the basic sensor structures. Additional fabrication equipment and related technology developments are needed for only a few sensor-specific post-processing steps. The fabrication of hybrid designs entails establishing and qualifying a reliable nonstandard production sequence. The specific production equipment and the necessary clean rooms require large investments. Sensor Response Time The response time of, e.g., a gas sensor is, in most cases, determined by the volume of the measurement chamber and the flow rate (other relevant processes include also, e.g., diffusion or dissociation). Using the monolithic approach and a suitable packaging technique (e.g., flip-chip packaging), the volume of the measurement chamber can be kept very small as a consequence of a small-size, flat and planar sensor or sensor array. In hybrid designs, the volume is depending on sensor geometries and array arrangements. Performance Capacitive or resonant microsensors perform pronouncedly better in monolithic designs owing to the fact that the influence of parasitic capacitances and crosstalk effects can be reduced by on-chip electronics (filters, amplifiers etc.). On-chip analog-to-digital conversion is another feature that helps to generate a stable sensor output that can be easily transferred to off-chip recording devices. In hybrid designs, it is sometimes very difficult to read out the rather minute analog microsensor signals.
5 CMOS Platform Technology for Chemical Sensors
87
Auxiliary Sensors/Smart Features Temperature or flow sensors can be monolithically co-integrated with the chemical sensors on the same chip. Calibration, control and signal processing functions as well as self-test features can be realized on chip. For hybrid designs, additional devices and off-chip components are required. Connectivity The number of electrical connections prominently contributes to the overall system costs. The monolithic implementation of a single-chip gas sensor (see Sect. 5.5 on monolithic CMOS gas sensor system later) with three micromachined sensors requires only seven connections (three for supply voltages, one for a clock signal, one for reset, and two for the serial interface). Up to 16 chips can be connected without adding any additional communication lines by implementation of a digital bus interface. A hybrid approach would require at least 30 pads for the three sensors and a total of 480 connections, if 16 sensors of each type would be combined, since there is no interface available on the sensor side. Package To package monolithic designs, IC-packaging techniques can be modified and adapted such as flip-chip-technology or simple epoxy-based packaging methods. Hybrid implementations require complex packages to reduce sensor interference (see, e.g., high-frequency acoustic-wave SAW-sensors, Sect. 4.1.1), to minimize electric crosstalk, and to optimize the critical connections. This further complicates the already difficult task of chemical sensor packaging. In summary the main disadvantages of monolithic CMOS-MEMS solution include the restriction to CMOS-compatible materials and the limited choice of micromachining processes. However, CMOS-MEMS offers on the other hand unprecedented advantages over hybrid designs especially with regard to signal quality, device performance, increased functionality and available standard packaging solutions. These advantages, in my opinion, clearly outweigh the drawbacks and limitations. Micromachined chemical sensors are not yet established on the market. In the case of well-established physical sensors such as acceleration and pressure sensors, a trend towards monolithic solutions can be identified for larger production volumes and more severe cost restrictions [493–495]. Several examples of monolithically integrated chemical sensor systems will be presented in the following sections of this chapter. The evolution from single transducers, which are integrated with the necessary driving and signalconditioning circuitry, to monolithic multisensor arrays and fully developed systems with on-chip sensor control units and standard interfaces will be shown. Microelectronics and micromechanics (MEMS-structures) have been
88
5 CMOS Platform Technology for Chemical Sensors
CMOS and MEMS design using CAD tool industrial CMOS process at foundry CMOS postprocessing packaging and assembly testing Fig. 5.1. Chemical microsensor fabrication sequence
realized on the same chip in all cases. The general scheme of the chemical sensor microsystem fabrication sequence is shown in Fig. 5.1. The CMOS-MEMS design is done in house, and the designs are then transferred to a foundry that performs a complete CMOS run at industrial standards (usually double-poly, double-metal 0.8-µm CMOS process as provided by, e.g., austriamicrosystems, Unterpremst¨atten, Austria [496]). The processed wafers come back from the foundry and undergo the post-CMOS micromachining (formation of membranes, cantilevers etc.). The wafers are then diced, the chips are tested and subsequently packaged as prototypes. The chemically sensitive coating is either applied on wafer level or to the single chips by means of spray-coating (polymers) or drop deposition (metal oxides).
5.1 CMOS Capacitive Microsystems 5.1.1 CMOS Capacitive Transducer As already mentioned in Sect. 4.4.3.2, capacitive chemical microsensors rely on changes in the dielectric properties of, e.g., a polymeric layer as a consequence of analyte absorption from the gaseous phase into the bulk polymer. The dielectric properties of the polymer with absorbed analyte differ from those of the polymeric matrix alone. A convenient way to assess the dielectric constant of a layer or film is to measure the capacitance of a polymer-filled plate-capacitor [497] or that of polymer-coated interdigitated electrodes [475–484, 498]. The latter approach has become very popular since the devices can be easily fabricated in planar technology such as CMOS. Planar polymer-coated interdigitated capacitors also provide direct access of the analyte to the sensitive layer (short sensor response time), which is more difficult to realize with modified plate capacitor designs [485]. Interdigitated capacitor designs
5.1 CMOS Capacitive Microsystems
(a) schematic
89
(b) SEM picture
polymer electrode 2
electrode 1
SiO2
Fig. 5.2. Schematic (a) and micrograph (b) of an interdigitated capacitor. The micrograph also shows the thin polymer layer, which is tightly attached to the sensor surface
have hence been chosen for the CMOS-based capacitive microsensor systems, a schematic and a micrograph of which is depicted in Fig. 5.2. The capacitors are fabricated exclusively with layers and processing steps available in the standard CMOS process sequence. Electrode 1 is made from the first aluminum metal layer while electrode 2 comprises a stack of interdigitated electrodes of the first and second metal layer, which are electrically connected on the chip via leads, albeit they are physically separated by a silicon-oxide layer (Figs. 5.2, 5.3). The quasi “three-dimensional” electrode configuration was chosen to enhance the sensitivity of the polymer-coated sensor to analyte-induced capacitance changes by maximizing the polymer volume in the regions of strong electric field (see also Fig. 4.36) [499]. The overall capacitor includes 128 electrode pairs and occupies an area of 824 µm by 814 µm. The electrode width and the interelectrode spacing are 1.6 µm. The pad-etch is used to remove the silicon nitride passivation on top of the sensing capacitor to allow for polymer coating. The electrode shapes are still visible under the polymeric coating, i.e., the polymer is tightly attached to the surface and reproduces the surface topology. A cross-sectional drawing of one electrode pair is shown in Fig. 5.3a, the corresponding equivalent circuit model, which was also used for simulations, is depicted in Fig. 5.3b [500,501]. Both electrodes (E1 and E2) exhibit considerable parasitic capacitances to the n-well (Cnwell 1 , Cnwell 2 ), which are very similar owing to the identical, though mirrored layout of the metal-1 components. The interelectrode capacitance includes two major contributions: An intermetal oxide capacitance (Cox1 ) and a composite of an oxide capacitance (Cox2 ) in series with an element that includes the polymer capacitance (Cpolymer ) shunted with the polymer resistance (Rpolymer ). The polymers used with interdigitated capacitors are mostly nonconducting (Rpolymer is very
90
5 CMOS Platform Technology for Chemical Sensors
E1
(a)
(b) Cox2
polymer E2 oxide n-well
E1
Rpolymer
Cpolymer Cox1 Cnwell1
E2
Cnwell2
Fig. 5.3. (a) Cross section of an electrode pair, (b) equivalent circuit model [500, 501]. For details see text
large). The two capacitors Cox2 and Cpolymer can then be substituted by a single capacitor Cpoly with Cox2 being considerably larger than Cpolymer . From finite-element simulations of a sensing capacitor coated with 10 µm of a polymer exhibiting a dielectric constant of ε = 4.8 (polyetherurethane, PEUT), the capacitances in the equivalent circuit model in Fig. 5.3 have been determined to: Cnwell1 : 17.7 pF, Cnwell2 : 18.2 pF, Cox1 : 6.4 pF, Cox2 : −8.4 pF, Cpolymer : −1.7 pF and Cpoly : 1.4 pF [479, 499–501]. The capacitance changes upon analyte absorption into the polymer are in the atto-Farad range, e.g., 4.4 aF/ppm toluene in polyetherurethane. Since the parasitic capacitances largely prevail over the capacitance of interest, Cpolymer or Cpoly , a differential readout scheme is mandatory. By using a switched-capacitor scheme with a reference and sensing capacitor of equal geometric dimensions, the reference capacitor being passivated by a sufficiently thick silicon nitride layer, the contributions of parasitic n-well capacitances and oxide capacitances cancel out to a large extent in the differential readout, since they are almost identical. In view of the analyte-induced capacitance changes in the atto-Farad range, on-chip signal conditioning circuitry is imperative, since the transfer of minute analog signals via bonds and wires is very difficult [500–502]. From the physical data and facts described above, it is evident that miniaturization without electronics integration is presumably prone to failure. 5.1.2 On-Chip Circuitry of the Capacitive Microsystem Micromachined capacitive sensors have been developed for many different applications, e.g., accelerometers, pressure sensors, fingerprint-sensors, and gas sensors [499–508]. Various read-out circuitry topologies have been developed to measure the small capacitance changes of such micromachined sensors. The majority of the designs rely on a differential readout between a sensing capacitor and a reference capacitor, the latter being not affected by the measurand. This offers the advantage that parasitic effects, such as temperature drift and ageing, affect the reference capacitor to the same extent as the
5.1 CMOS Capacitive Microsystems
91
sensor and, hence, cancel out. As a consequence, these designs do not provide accurate information on the absolute value of the measuring capacitor. The most popular approach to read-out capacitive sensors with high resolution and good suppression of parasitics is a switched-capacitor design [509–513]. A resolution of 19 bits is needed in order to achieve a detection limit of 1 ppm for volatile organic compounds (VOCs). The bandwidth can be as low as 1 Hz. As none of the designs reported on in literature exhibits the required performance, an improved architecture based on a switched-capacitor Sigma-Delta modulator was developed. The sensor response is measured as the differential signal between a polymer-coated sensing and a nitride-passivated reference capacitor. Both, sensor and reference capacitors are split into two parts to improve the charge transfer efficiency. Sensing capacitor, (CS ), and reference capacitor, (CR ), are incorporated in the first stage of a fully differential second-order SigmaDelta-modulator (Fig. 5.4) with two switched-capacitor integrators and a subsequent comparator [501, 514, 515].
Cfb1
Cfb2
Cfb1'
Cfb2'
CR
decimation filter
digital logic
CS
Fig. 5.4. Schematic of the fully differential second-order Sigma-Delta modulator exhibiting two switched-capacitor integrators and a subsequent comparator. Four feedback capacitor (Cfb ) are realized as interdigital capacitors. The Sigma-Delta modulator provides a pulse-density-modulated digital output that is decimated using the frequency counter [501]
Since the output bit stream of the Sigma-Delta modulator is proportional to the ratio (CS − CR )/(Cfb ), the four feedback capacitors (Cfb in Fig. 5.4) are realized as interdigital capacitors with the same materials as sensing and reference capacitor in order to eliminate differences in temperature behavior and ageing. Due to the small signal bandwidth, the output bit stream of the Sigma-Delta modulator is decimated using a frequency counter. For more details on the circuitry see [501, 514, 515]. A micrograph of the integrated capacitive microsystem is shown in Fig. 5.5.
92
5 CMOS Platform Technology for Chemical Sensors
readout circuitry
feedback capacitors
sensing capacitor
reference capacitor
Fig. 5.5. Micrograph of a capacitive sensor system including a polymer-coated sensing capacitor, a passivated reference capacitor, four interdigitated feedback capacitors, and the Sigma-Delta circuitry as detailed in the text [499, 501]
The design offers the option to place all interdigitated capacitors (sensor, reference, feedback) on micromachined membranes with integrated heaters for temperature stabilization since analyte absorption in the polymer matrix is strongly temperature-dependent (see also Chap. 2) [516]. 5.1.3 Capacitive Gas Sensing Two effects change the capacitance of a polymeric sensitive layer upon absorption of an analyte (Fig. 5.6): (i) Swelling and (ii) change of the dielectric constant due to incorporation of the analyte molecules into the polymer matrix [478, 479]. The resulting capacitance change is detected by the read-out electronics.
absorption
polarization
swelling
polymer substrate
analyte polarized analyte
electrodes
air
Fig. 5.6. Schematic of the capacitive sensing principle showing the two relevant effects changing the sensor capacitance: Change of the dielectric constant and swelling. The interdigitated electrodes (+, −) on the substrate (black ) are coated with a polymer layer (gray). Big and small globes represent analyte and air molecules. Analyte molecules are polarized (↑) in the electric field (center ). Analyteinduced polymer swelling is indicated with the dashed lines (right side) [517]
5.1 CMOS Capacitive Microsystems
93
The partition coefficient, K, as introduced in Chap. 2 (2.10) characterizes the absorption behavior of a polymer with regard to a specific analyte. It is a chemical equilibrium constant and is defined as the ratio between the analyte concentration in the sorptive or polymer phase and that in the gas phase. For a specific analyte, K is inversely proportional to the saturation vapor pressure of this analyte and, therefore, strongly temperature-dependent. Furthermore, K depends on the analyte/polymer interaction, responsible for the partial selectivity. The analyte concentrations in both phases can be replaced by the partial pressure, pA , of the analyte and by the volume fraction of the analyte in the sensitive layer, ϕA , both used in the rest of this section. Equation (2.10) can be rewritten as ϕA =
pA M · KC · R · T ρ
(5.1)
where M and ρ denote the molar mass and the density of the analyte in liquid state, and R and T the universal gas constant and the absolute temperature [517]. As already mentioned, the two effects changing the capacitance are swelling and change of the dielectric constant. For low analyte concentrations, the swelling is linear in the amount of absorbed analyte, expressed by (5.2). (5.2) heff = h · (1 + Q · ϕA ) . Here, h and heff denote the initial polymer thickness and the resulting effective thickness after analyte absorption, respectively, Q is a dimensionless non-ideality factor of the swelling, and ϕA is the volume fraction of the analyte in the polymer. ϕA is proportional to the concentration of the analyte in the gas phase assuming the validity of Henry’s law. The proportionality factor has to be determined experimentally for every polymer/analyte combination or can be estimated with solubility parameters or linear solvation energy relationships [518]. Q = 1 represents ideal swelling, i.e., the total volume is given by the addition of the volume of the absorbed analyte in its liquid state to that of the polymer. For elastic polymers, the tendency to generate static stress while swelling increases with increasing stiffness of the polymer. The composite dielectric constant of mixtures of nonpolar liquids can be approximated for all kinds of analytes as proposed in [479, 517]. It can be generalized including Q to εeff = εpoly + ϕA · ((εA − 1) − Q · (εpoly − 1))
(5.3)
where εeff is the resulting effective dielectric constant of the polymer/analyte system and εpoly and εA are the dielectric constants of polymer and analyte (analyte in liquid state), respectively.
94
5 CMOS Platform Technology for Chemical Sensors
thin
thick
2 µm
6.4 µm
Fig. 5.7. Selectivity through layer thickness: Sensors coated with a thin and thick sensitive layer. The extension of the electric field lines is approximately 2 µm for this electrode geometry (1.6 µm electrode width, 1.6 µm electrode spacing) [517]
5.1.3.1 Selectivity Through Sensitive Layer Thickness The presence of two relevant physical effects gives rise to an additional mechanism for selectivity, because thin and thick sensitive layers exhibit different sensitivity patterns. For a simple interdigitated structure, the space above the device containing 95% of the field lines is within a distance of half of the electrode periodicity [480]. Therefore, a thin or thick layer is defined with respect to the electrode periodicity (see Fig. 5.7). For a layer thickness significantly less than half the periodicity, the region of strong electric field extends above the sensitive layer. Upon analyte sorption, the amount of polarizable material in the sensed region of the capacitor always increases, which results in a capacitance increase regardless of the dielectric constant of the analyte. For a layer thickness greater than half the periodicity of the electrodes, almost all electric field lines are within the polymer volume. Consequently, the capacitance is determined by the composite dielectric constant of the analyte/polymer mixture. The capacity change can, therefore, be positive or negative, depending on whether analyte or polymer has a higher dielectric constant [479] (5.3). The expected differences in responses from sensors with thin and thick polymer layers have been verified experimentally. Typical sensor response profiles from capacitors coated with a thin (0.3 µm) and thick (2.3 µm) poly(etherurethane) (PEUT) layer are plotted in Fig. 5.8. The capacitor has been alternately exposed to various concentrations of toluene and ethanol at 28◦ C and the pure carrier gas. Ethanol (ε = 24.3) has a higher dielectric constant than PEUT (ε = 4.8), and toluene a lower dielectric constant (ε = 2.36). Both analytes provide positive sensor signals with thin PEUT layers, whereas with thick layers, ethanol provides a positive signal and toluene a negative signal. The limit of detection of the capacitive microsystem at 28◦ C is approximately 8 ppm for toluene and 5 ppm for ethanol with 10 ppm analyte volume fraction corresponding to 1 Pa analyte partial pressure in the probed air [476]. With a sensor coated with a thick polymer layer, low-ε analytes can be differentiated from high-ε analytes if they are pure. In a mixture of these
5.1 CMOS Capacitive Microsystems
95
15
sensor signal [kHz]
polymer: PEUT
0.3 µm 2.3 µm
10 toluene 600toluene – 3000 ppm
5
0 ethanol 1000 – 5000 ppm -5 0
4
8 time [h]
12
16
Fig. 5.8. Response profiles from two capacitors coated with PEUT layers of different thickness upon exposure to 1000–5000 ppm ethanol and 500-3000 ppm toluene as a function of time. The ratio of the dielectric constants of polymer (4.8) and analytes (toluene: 2.36, ethanol: 24.3) controls the signs of the signals from thick layers. For a thin polymer layer, all signals are positive [517]
analytes, more information is needed. Zero response, e.g., from the thick layer sensor could also be generated by a certain mixture of both analytes in such a way that the positive and negative signal just cancel out. However, the combination of signals from sensors with thin and thick sensitive layers allows for determining the concentration of both analytes. The signal from the thinly coated sensor gives the total amount of analyte, whereas the signal from the thickly coated sensor indicates the mixing ratio at a given total concentration. Extensive numerical simulations (finite element method, FEM) were performed for the interdigitated capacitor with different coatings. The capacitance of the electrode geometry was simulated with polymer coatings of 17 different, logarithmically distributed thicknesses, h, in the range of 0.06–7 µm and 41 different, linearly distributed dielectric constants, εpoly , in the range of 1–5. For details on the simulations, see [479, 517]. The most important variable to calculate from the simulated data is the sensitivity. For low analyte concentrations and, hence, low volume fractions, small changes in the dielectric constant, δε, and in the polymer layer thickness, δh, are expected and the capacitance change, ∆C, can be approximated by [517]: ∆C = δh · ∂h C (h, εpoly ) + δε · δε C (h, εpoly ) .
(5.4)
The change in dielectric constant was approximated with (5.3). Together with (5.2), the capacitance change can be rewritten as [517]:
5 CMOS Platform Technology for Chemical Sensors
normalized sensitivity
96
εA » εP ethanol ethyl acetate 0 0
1
2
3
4
εA > εP
5
h [µm]
toluene
εA < εP
Fig. 5.9. Numerically simulated dependence of the vapor sensitivity of a capacitive sensor on the polymer layer thickness, h, for PEUT (εpoly = 4.8). The lines, from bottom to top, correspond to analytes with dielectric constants of 1.9, 2.4, 3.3, 4.7, 6.0, 10, 24, and 81. Ideal swelling was assumed (Q = 1). For better visibility, the curves are normalized so that the sensitivities for very thin layers (initial slopes of the curves) are identical [517]
∆C = ϕA · [Q · h · ∂h C (h, εpoly ) + ((εA − 1) − Q · (εpoly − 1)) · ∂ε C (h, εpoly )] .
(5.5)
The physical sensitivity, ∆C/ϕA , was then calculated for ideal swelling and for εpoly = 4.8 (PEUT). It is plotted in Fig. 5.9 as a function of the polymer layer thickness. The physical sensitivity accounts for the changes in dielectric constant and layer thickness but does not include the chemical selectivity. As explained above, sensitivities of thin layers are all positive. For even thinner layers, the sensitivities have to decrease because smaller polymer volumes absorb less analyte. For thick layers, the sign of the sensitivity depends on the dielectric constant of the analyte. Responses to analytes that have dielectric constants lower than that of the polymer (such as toluene) are positive for thin layers and show a transition from positive to negative values with increasing layer thickness. Hence, for each of these analytes, there is a critical layer thickness, where the sensitivity towards this analyte vanishes (see circle in Fig. 5.9). Consequently, the sensitivity reaches a (local) maximum somewhere between the critical layer thickness and zero. For analytes with dielectric constants higher than that of the polymer (such as ethyl acetate), the sensitivity first increases for thin layers, reaches also a maximum, decreases again and then saturates. In contrast to low-ε analytes, the saturation is at positive levels. With increasing dielectric constant of the analyte, the saturation level increases, and the maximum becomes less and less evident. For very high-ε analytes (such as water), the maximum vanishes completely. The sensitivity
5.1 CMOS Capacitive Microsystems
physical sensitivity [pF]
8
water ethanol isopropanol
6
97
ethyl acetate toluene n-octane
4 2 0 0
1
2 3 4 5 polymer thickness [µm]
6
7
Fig. 5.10. Measured dependence of the physical sensitivity, ∆C/ϕA , on the layer thickness, h, for analytes with various dielectric constants: Low εA : n-octane (1.93) and toluene (2.36), high to very high εA : ethyl acetate (5.88), isopropanol (18.5), ethanol (24.3), and water (76.6) [517]
increases monotonically with increasing layer thickness and saturates already at relatively thin layers. To verify the simulations, the sensitivities of PEUT to various analytes were measured for layer thicknesses between 0.3 and 7.1 µm. The physical sensitivities, ∆C/ϕA , displayed in Fig. 5.10 have been calculated with independently determined partition coefficients that have been measured with thickness-shear-mode resonators (TSMRs) [517, 519]. The measurements confirm the simulations: saturation-like behavior already at relatively low thickness for water (very high ε), an increasingly clear maximum for ethanol, isopropanol, and ethyl acetate (high ε), and the transition to negative values for n-octane and toluene (low ε). 5.1.3.2 Insensitivity to Low-ε Analytes In the case of low-dielectric-constant interferants, a first method to achieve insensitivity is to choose the critical layer thickness. For each low-ε analyte, there is such a critical layer thickness, where the effect of swelling and dielectric constant change cancel out (see circles in Figs. 5.9 and 5.10). This idea, already mentioned in references [476] and [479], suffers from the required accurate adjustment of the layer thickness, which is a tedious business. Thus, another solution is presented, where the thickness adjustment is not so critical. By tuning the dielectric constant of a thick sensitive layer to be similar to that of the analyte, the sensitivity of the sensors to the respective
98
5 CMOS Platform Technology for Chemical Sensors
frequency shift [kHz]
analyte almost vanishes. Insensitivity to, e.g., n-octane was achieved by blending two polymers with different dielectric constants: Poly(etherurethane) (PEUT, ε = 4.8) and poly(epichlorohydrine) (PECH, ε = 1.7). Figure 5.11 displays the sensor response data upon exposure to ethanol and n-octane as a function of the polymer layer composition.
2
1
ethanol n-octane
0
0 PEUT
25
50 75 mass fraction [%]
100 PECH
Fig. 5.11. Frequency shifts of sensors coated with thick layers of mixtures of different PEUT and PECH content upon exposure to n-octane (1280 ppm) and ethanol (5140 ppm). When dielectric constant of polymer and analyte coincide, the sensor is blind to the respective analyte [517]
Ethanol with a high dielectric constant causes positive signals for both polymers and all blending ratios. n-Octane has a dielectric constant higher than that of PECH but lower than that of PEUT. Consistent with the theory, the signals of n-octane are positive for PECH and negative for PEUT. For the blended polymers, the signals show a continuous transition from positive to negative values, approximated with a quadratic function in Fig. 5.11. As can be seen, the analyte sensitivity vanishes when the dielectric constant of polymer mixture and analyte coincide (see circle in Fig. 5.11) [517]. 5.1.3.3 Humidity Interference Even more important is the elimination or reduction of the influence of humidity. Because water has a much higher dielectric constant than any polymer used, it is not possible to achieve humidity insensitivity by either of the methods discussed above. However, the thickness-dependence shown in Figs. 5.9 and 5.10 gives rise to another method, which is schematically depicted in Fig. 5.12. The sensitivity to analytes with high dielectric constant reaches saturation at low thickness, while the sensitivity to analytes with low dielectric constant
normalized sensitivity
5.1 CMOS Capacitive Microsystems
99
humidity
0
1
2
3
4
5
h [µm] toluene n-octane
sensor 2 sensor 1
Σ∆
difference
Fig. 5.12. Eliminating humidity interference by recording difference signals using a thick and a medium thick layer: The simulated sensitivities from Fig. 5.9 are plotted as a function of the polymer layer thickness, h. Two sensors coated with a thick and a medium thick polymer layer are represented by the dashed vertical lines at the corresponding layer thickness. The difference signal, generated by connecting both sensors differentially to the Sigma-Delta-modulator, is indicated with the bars on the right hand side of the plot [517]
still varies considerably out to greater thicknesses. Hence, the signal difference of two capacitors with different layer thicknesses in the range of 1 µm to 5 µm is almost insensitive to water but retains sensitivity to n-octane and toluene, i.e., n-octane and toluene show a considerable signal gradient in that thickness region, whereas water behaves like ethanol and shows saturation. The initial idea of this method was presented already in [476] and [479]. There, the signals of two sensors were read out subsequently and then numerically subtracted. However, calculating differences of similar values is very error-prone. Accordingly, slight temporal humidity fluctuations resulted in a high noise in the signal. To overcome this obstacle, two sensors on the same chip were directly connected to the positive and negative input of the Sigma-Delta converter (Fig. 5.13). The signals were subtracted instantaneously so that any fluctuations cancelled out before being processed by the read-out circuitry– resulting in a significantly improved signal-to-noise ratio. The microsensor array used for these investigations included seven sensor and five reference capacitors, all of which are monolithically integrated with driving and read-out circuitry on a single chip (see Fig. 5.13). The reference capacitors are similar to the sensors but not coated with a polymer layer and allow for a differential measurement. A multiplexer (MUX) connects sensors and references to the read-out circuitry, a Sigma-Delta modulator (Σ∆). The Σ∆-modulator converts the analog capacitance signal to a digital output signal. The frequency of the output signal – measured with an external counter – is proportional to the difference of sensor and reference
100
5 CMOS Platform Technology for Chemical Sensors
S R
Σ∆
S R
R
S
S
MUX
R R
S
S
S chip size: 5.8 x 6.5 mm Fig. 5.13. Micrograph showing the sensor chip with seven sensor (S) and five reference (R) capacitors and readout circuitry: Sigma-Delta-modulator (Σ∆) and multiplexer (MUX) [476]
capacitance. The Σ∆-modulator is described in detail in references [514] and [515]. Two sensor capacitors of a chip as shown in Fig. 5.13 were coated with 1.4 and 3 µm PEUT. Figure 5.14 displays the signal of the sensor with the thin layer and the signal difference upon exposure to different concentrations of humidity, n-octane, and toluene. The difference signal is more than one order of magnitude less sensitive to water than the signal from a single sensor, whereas for n-octane and toluene, the signals stay in the same range. A second drastic improvement in signal quality was achieved by using the difference method, which has been detailed in Sect. 5.1.1. Drift, which is similar for both sensors, was also reduced by approximately an order of magnitude, which can be seen in the right part of Fig. 5.14. Especially for the humidity steps, sharp signal peaks appear in the difference signal when the analyte concentration is switched on and off. They originate from the different absorption times for the thin and the thick polymer layer, resulting in a temporarily high difference signal. The steady-state signal is not affected.
5.2 CMOS Calorimetric Device 5.2.1 CMOS Calorimetric Transducer The calorimetric transduction principle and, in particular, the thermoelectric or Seebeck-effect-based transducer has been introduced already in Sect. 4.2.2.
5.2 CMOS Calorimetric Device
∆f [kHz] humidity
40 n-octane
4.0
toluene
20
2.0
0
0.0 humidity
-20
∆f [kHz] n-octane, toluene
1.4 µm 1.4 µm - 3 µm
101
-2.0 60
120 time [min]
180
Fig. 5.14. Frequency shifts of sensors coated with PEUT upon exposure to humidity (12.5, 25, and 50% RH), n-octane (300, 600, and 1200 ppm), and toluene (400, 800, and 1600 ppm) as a function of the measurement time. The solid line is the difference signal of two sensors coated with 1.4 and 3 µm PEUT, whereas the dashed line represents the signal of the sensor with 1.4 µm PEUT. For better visibility, the signals of n-octane and toluene are displayed in a ten-times larger scale than the humidity signals. The difference signal is much less sensitive to humidity than the signal from a single sensor, but exhibits a comparable sensitivity to n-octane and toluene [517]
The expected temperature changes upon analyte interaction with the sensitive layer are in the milli-Kelvin range. Therefore, the measurement area has to be thermally insulated from the silicon substrate, which is an excellent thermal conductor. Membranes consisting of the dielectric layers of the CMOS process (silicon oxide, silicon nitride) have been previously used for infrared sensor arrays [520,521]. Dielectric membranes show a parabolic temperature profile, albeit a flat temperature profile over the sensitive area is desirable for gas-sensing applications. Such flat temperature profile can be obtained by introducing a silicon-n-well island as shown in Figs. 4.10 and 5.15, which is thermally decoupled from the bulk silicon of the chip. Predominantly the island area is afterwards coated with the sensitive layer. The temperature on the membrane can be assessed with different transducers using, e.g., the temperature coefficient of bipolar transistors or resistors. The sensitivity of these realizations is, however, not suitable to measure temperature changes in the milli-Kelvin range. The solution of choice is the utilization of thermocouples, which rely on the Seebeck-effect (see Sect. 4.2.2). The hot contacts are located on the thermally insulated n-well island, while the cold contacts are placed on the substrate. Maximum signals can be obtained using bismuth and antimony as thermocouple materials [520]. Both are not available in CMOS processes and would be difficult to deposit and to pattern. From all CMOS materials (see Sect. 3.3) the polysilicon/aluminum
102
5 CMOS Platform Technology for Chemical Sensors
dielectric layers
metal: Al
polysilicon heater
thermocouple }
n-well (Si-island)
polysilicon
p-silicon substrate Fig. 5.15. Cross-section of a CMOS calorimetric sensor showing the thermocouples (aluminum, polysilicon) and the n-well island in the center
reflection spot
polysilicon heater
thermocouples
distance n-well to bulk
Fig. 5.16. Rectangular membrane (300 thermocouples) featuring a reflection spot for optical layer thickness detection and a polysilicon meander heater. The thermocouple length (distance n-well to bulk) is 200 µm
thermocouple exhibits the highest Seebeck coefficient of 110 µV/K. Many thermocouples must be connected in series to achieve the desired sensitivity, i.e., measuring temperature differences in the milli-Kelvin range (Fig. 5.16). As already mentioned in Sect. 4.2.2, the detection process includes four principal steps: (I) absorption and partitioning or chemical reaction, (II) generation of heat, which causes (III) temperature changes to be transformed in (IV) thermovoltage changes. The final signal results from this sequence of processes, some of which are of chemical or physicochemical nature, i.e., depend on the involved chemical compounds (I, II), and some of which are of physical nature and are device-specific (III, IV: heat sensitivity and thermovoltage generation). The overall sensitivity is the recorded thermovoltage change ∆U [V] per change in the analyte concentration as a function of time dcA /dt [mol/m3 s] and can be written according to (4.3), Sect. 4.2.2:
5.2 CMOS Calorimetric Device
103
∆U = A · B · Vsens · ∆H · K . (5.6) dcA /dt A [K·s/J] and B [V/K] denote device-specific constants describing the translation of a generated molar enthalpy ∆H [J/mol] via a temperature change into a thermovoltage change. The constant A [K·s/J] includes the feature sizes of the transducer and material properties, and B [V/K] includes the number of thermocouples and the Seebeck coefficients of the thermocouple components. Vsens denotes the sensitive-layer volume in case of a bulk effect, and K is the partition coefficient (2.10) or reaction equilibrium constant (2.2, 2.11). The overall sensitivity hence includes two contributions, the chemical and the physical sensitivity. The chemical sensitivity is correlated to the amount of heat that is generated upon an analyte concentration change. It comprises the reaction constant or partition coefficient, K, the produced enthalpy change, ∆H, and the sensitive layer volume, Vsens . This part will be treated in more detail in Sect. 5.2.3. The physical sensitivity is the thermovoltage output of the system in response to a defined heating power on the membrane. It includes the transducer features, whereas it does not depend on the detection chemistry. Some of the most important design parameters are briefly discussed. S=
• Size of the membrane: A large membrane leads to a large sensitive area and a large number of thermocouples. The mechanical stability of the membranes and the maximum allowable chip area determine the size limits. The membranes used for our devices have sizes of 650 × 650 µm2 (square membrane) and 2150 × 750 µm2 (rectangular membrane, see Figs. 5.16, 5.17). • Distance between substrate and n-well island: A large distance improves the thermal isolation of the n-well island and, therefore, the efficiency of the heating. The mechanical stability of the membrane and chip area considerations have to be taken into account. The increase of the electrical thermocouple resistance owing to the larger distance between hot and cold contacts also is an issue. • Number and width of the thermocouples: The achievable signal increases linearly with the number of thermocouples. The thermocouples, however, consist of aluminum and polysilicon, both of which exhibit comparably high thermal conductance. A large number of thermocouples thus reduces the thermal resistance of the membrane and, consequently, degrades the temperature effects. This would suggest the use of a large number of minimum-width conductors to increase the thermal resistance and make efficient use of the available area. Since polysilicon resistors exhibit low electrical conductivity, the line width cannot be reduced too much without increasing the thermal noise in the signal. In a viable compromise the square membrane design has a total of 132 thermocouples (40 kΩ electrical resistance), the rectangular membrane (Figs. 5.16,
104
5 CMOS Platform Technology for Chemical Sensors
sensor
low-noise differential amplifier reference Fig. 5.17. Micrograph of a differential thermopile configuration with integrated low-noise amplifier. The measurement thermopile is coated with polymer, the reference is passivated with silicon nitride [522]
5.17) features 300 thermocouples (143 kΩ electrical resistance). A differential strategy that has been already utilized in the case of the capacitive transducer is also applied to the calorimetric transducers as can be seen in Fig. 5.17. Such differential measurements are preferred over an absolute measurement, since the influence of ambient-medium or gas-flow temperature fluctuations is strongly reduced. 5.2.2 Calorimeter Circuitry The sensor system has to be optimized for a maximum signal-to-noise ratio (SNR) while allowing for measurements of a wide concentration range of different analytes. The thermopile converts the temperature difference between substrate and n-well island into a voltage signal. For the best overall performance, the input-referred SNR of the first amplifier and not the signal has to be maximized. The two thermopiles used here exhibit electrical resistances of 40 kΩ (square membrane) and 143 kΩ (rectangular membrane). The overall system optimization includes the input noise of the first amplifier as well as the power and area consumed by the complete readout-circuitry [501, 520]. Another important parameter is the signal bandwidth, which must be as small as possible in order to minimize the noise. The highest frequency of interest in the transient signal during absorption or desorption of analytes was experimentally determined to be approximately 400 Hz. A system and circuitry architecture based on the aforementioned considerations is shown in Fig. 5.18 [501]. It includes a differential arrangement, in which a polymer-coated sensor is connected in series to an uncoated reference. Temperature fluctuations owing to medium flow and radiation are largely cancelled out by the differential arrangement. The small signals are
5.2 CMOS Calorimetric Device
sensor
+ low-noise chopper amplifier
reference
−
anti-aliasing filter
105
decimation filter 13bit
A D Sigma-Delta A/D converter
Fig. 5.18. Schematic of the calorimeter circuitry. Sensing and reference thermopiles are connected to the input stage of a low-noise chopper-stabilized instrumentation amplifier followed by an anti-aliasing filter, a Sigma-Delta A/D converter and a decimation filter [501]
amplified by a low-noise chopper amplifier. The chopping frequency is 5 kHz. The gain of the amplifier must be adjustable in order to allow for measurements over a large concentration range, i.e., for the use of a variety of different analyte-sensitive-layer combinations. The maximum gain of 6400 can be reduced by a factor of 4 or 16. The resolution is 12 bits, which is sufficient for the target applications. An anti-aliasing filter succeeding the amplifier is needed to avoid down-conversion of high frequency noise by the A/D-converter. The system has a narrow signal band from 0–400 Hz. The requirements for the stop-band are set by noise constraints, by the suppression of the signals upconverted to the chopping frequency by the demodulator at the output of the amplifier, and by clock-feedthrough. A minimum damping of 80 dB at the chopping frequency is needed to meet these requirements. The low cut-off frequency of 400 Hz would require large on-chip capacitors in the anti-aliasing/bandstop filter if a Nyquist-rate analog-digital converter would be used. By using a largely oversampled Sigma-Delta-A/D-converter, the filtering can be shifted to the digital decimation filter, where low frequencies do not increase complexity, area, and power-consumption of the design. This also simplifies the design of the anti-aliasing filter. Due to the narrow signal band, an oversampling factor of 128 can easily be realized. This makes a one-bit Sigma-Delta modulator the most suitable choice for the 12-bit A/Dconverter owing to the less stringent matching requirements. For a detailed description of each of the circuitry components, see [501, 523, 524]. 5.2.3 Calorimetric Gas Sensing Calorimetric sensors rely on determining the presence or concentration of a chemical by measurement of an enthalpy change produced by the chemical to be detected. Any chemical reaction or physisorption process releases or absorbs a certain quantity of heat from its surroundings. As already briefly mentioned in Sect. 4.2.2, the calorimetric sensor only detects changes in the heat budget at nonequilibrium state (transients) upon changes in the analyte concentration.
106
5 CMOS Platform Technology for Chemical Sensors
Thermodynamic equilibrium is a dynamic equilibrium state, in which reactants and products are present but have no tendency to undergo net change, i.e., as many molecules undergo the forward reaction as undergo the reverse reaction. The change in the Gibbs free energy, ∆G, is zero (see 2.1 and related text), and there is consequently no net heat production. The calorimetric sensor, therefore, only produces transient signals as is illustrated in Fig. 5.19. The initial state is thermal equilibrium with no analyte present and, hence, no temperature signal. The analyte concentration then is drastically increased, which leads to sorption/reaction heat liberation and, e.g., a temporal temperature increase on the sensor (Fig. 5.19b). As soon as equilibrium conditions, i.e., a constant analyte concentration is established, the signal returns to zero, because there is no more net enthalpy change (Fig. 5.19c).
polymer
signal time
(c) equilibrium
signal
(b) analyte absorption
signal
(a) no analyte
time
time
Fig. 5.19. Transient signal characteristics of the calorimetric sensor: (a) no analyte present, (b) changing or increasing analyte concentration, (c) thermodynamic equilibrium [501]
In case that an analyte concentration increase produces a positive calorimetric signal, the corresponding decrease produces a negative signal of equal intensity and vice versa (see Fig. 5.21). The signal and detection characteristics of the calorimetric transducer have some implications for its use in gas sensing. Slow changes in analyte concentrations are difficult to detect, since they continuously produce but small enthalpy changes. Therefore, a switching scheme has to be implemented, which allows for alternately exposing the sensor to analyte-loaded medium and non-contaminated medium, i.e., generates fast concentration changes or large concentration gradients. In a switched system, the amplitude of the signal, i.e., the peak height, largely depends on the time scale of the diffusion processes in the medium manifold and the diffusion of the analyte into the sensitive layer. Therefore,
5.2 CMOS Calorimetric Device
107
the peak area integrated over time is the characteristic sensor signal that represents the total enthalpy change upon concentration changes (see also Fig. 5.21). For gas sensors with standard polymer coatings (deposited by spray or spin-coating), which will be discussed in more detail here, only physisorption contributes to the enthalpy change, since no chemical reaction is taking place, and no chemical bonds are formed between polymer and the analyte to be detected. The temperature change in the polymer, ∆T, upon analyte sorption arises from the overall heat production, ∆H, which, in the special case of volatile absorption in a polymeric matrix, includes two components, the heat of condensation and the heat of mixing: ∆T ∝ ∆Htotal ;
∆Htotal = ∆Hcond + ∆Hmix .
(5.7)
∆Htotal here denotes the overall enthalpy change ([J/mol]). The analyte absorption process in the polymeric coating can be conceptually subdivided in two steps (Fig. 5.20) [516]: Analyte condensation from the gas into the liquid phase and subsequent mixing with the polymeric matrix. Consequently, the first term on the right-hand side of (5.7) is the condensation enthalpy, which is characteristic for the analyte and independent of the sensitive coating or polymer. The second term describes the enthalpy change upon mixing of analyte and polymer. This term includes all polymer/analyte interactions and is, for many analyte/polymer combinations, small in comparison to the overall enthalpy change or condensation enthalpy. For more details see [516, 524].
analyte molecules
polymer matrix sorption
=
condensation
+
mixing
Fig. 5.20. Conceptual subdivision of the analyte absorption process: Condensation in the liquid phase and subsequent mixing with a polymeric matrix
Typical chemical sensor signals of the calorimetric transducer are displayed in Figs. 5.21 and 5.22, which exemplify the detection of different kinds of organic volatiles in the gas phase (hydrocarbons, alcohols etc.) by using various polymeric layers. Figure 5.21 (a) and (b) show the output voltage of the microsystem and the peak integrals while switching from synthetic air (nitrogen/oxygen mixture without humidity) to n-octane (900 ppm) and back
108
5 CMOS Platform Technology for Chemical Sensors
thermovoltage [mV]
n-octane: 900 ppm
on
off
PDMS
time [s]
time [s]
Fig. 5.21. Calorimetric sensor signals: Output voltage of the microsystem and peak integrals while switching from synthetic air to n-octane (900 ppm) and back to air at 28◦ C with 2 µm PDMS as sensitive layer
to air at a temperature of 28◦ C. The sensitive layer is poly(dimethylsiloxane), PDMS. As already discussed above, two transient signals are produced, a positive one at the onset of the analyte exposure (liberation of predominantly condensation enthalpy) and a negative one upon terminating analyte exposure (abstraction of the enthalpy necessary for vaporizing the absorbed analyte). The net enthalpy changes can be approximated by integration over the peak area of the sensor signals [522, 523], which is the measurand of interest. The height of the peak maximum is strongly depending on the polymer thickness (diffusion kinetics), and the peak maximum and signal characteristics may also differ for switching on and switching off the analyte as a consequence of slightly different switching times or gas flow fluctuations. An optimized manifold with meticulously controlled gas flow, short gas paths, and fast and reproducible valve switching times is of paramount importance. The changes in the sensor response patterns upon varying the polymeric sorption matrix are displayed in Fig. 5.22. Whereas the slightly polar poly(etherurethane) shows large signals upon alcohol exposure (methanol, ethanol), the nonpolar poly(dimethysiloxane) provides hardly any signal for alcohols albeit showing large signals for nonpolar compounds such as n-octane and toluene. The sensor response patterns in Fig. 5.22 demonstrate, that sensor selectivity can be tuned by polymer variation and confirm the old alchemist rule: “Like dissolves like”, i.e., nonpolar matrices predominantly absorb nonpolar analytes, as well as polar matrices tend to absorb polar analytes. Note that the PDMS layer is with 1.5 µm thinner than the 3 µm PEUT layer and that the signals scale linearly with the polymer volume/layer thickness. In comparing Figs. 5.21 and 5.22 with capacitive sensor signals (Fig. 5.8) or cantilever signals (Fig. 4.1 and Sect. 5.3.3) it is evident, that there is
5.3 CMOS Integrated Resonant Cantilever 0.6
0.6 themovoltage [V]
109
PEUT
0.4
PDMS
0.4
0.2
0.2
0
0
-0.2
-0.2
-0.4
-0.4
-0.6
ethanol 1000-4000
toluene 500-2000
methanol 2000-3500
n-octane 250-1000 ppm
-0.6
0
50
100 150 time [min]
200
250
150
200 250 300 time [min]
350
Fig. 5.22. Calorimetric response patterns of two different polymers (PEUT and PDMS) upon exposure to an identical set of organic solvents
significant differences. The calorimetric sensor provides two signals for each analyte exposure, which makes the concentration assessment more reliable. The measurement time is also significantly shorter since the transient evolution is on the order of 3–5 seconds, whereas it may take tens of seconds to reach equilibrium and a reliable sensor reading for equilibrium-based sensors at similar polymer layer thickness (approx. 3 µm). The calorimetric sensor signal is in addition inherently drift-free: No rapid concentration change means no sensor signal. Drawbacks of the calorimetric principle include a somewhat inferior analyte sensitivity in comparison to other polymer-based sensors (factor of approximately three to five) and the necessity of drastic gas concentration changes, i.e., a gas switching mechanism.
5.3 CMOS Integrated Resonant Cantilever As already mentioned in Sect. 4.1.3, cantilevers are mostly used as gravimetric transducers and can be operated either in a static mode by measuring the cantilever deflection or in the dynamic mode by assessing changes in the resonance behavior. The fabrication in industrial CMOS technology allows for the integration of transducers, driving circuitry, and analog or digital signal processing circuitry on the same chip [93,95]. Integrated thermal or magnetic excitation and piezoresistive detection schemes render the system independent of additional excitation elements such as piezoelectric layers [61–63,106], and independent of bulky external optical detectors [100, 103–105]. The cantilevers that will be described in more detail below are operated in the dynamic mode, and the change of the resonance frequency upon mass loading is measured. The operation features of resonant cantilevers are, as already mentioned, similar to those of other well-established mass-sensitive gas
110
5 CMOS Platform Technology for Chemical Sensors
sensors such as thickness-shear-mode resonators (TSMRs, quartz microbalances) [61–63, 66, 68–70] and surface-acoustic-wave (SAW) [61–63, 68–70] devices. 5.3.1 Resonant Cantilever Transducers There are basically three ways to determine the resonance frequency of a cantilever [501]: • Noise spectrum: The frequency spectrum of the thermal noise is measured and analyzed. The peak frequency is equivalent to the resonance frequency. This method does not require any excitation of the cantilever, but a bulky low-noise spectrum analyzer is needed for the measurement. • Frequency sweep of excitation signal: The cantilever is excited at different frequencies and its response is evaluated to determine the resonance frequency. A gain-phase analyzer is needed for the measurement. • Oscillator: The cantilever is used as the frequency-determining element in an oscillator loop. A cantilever actuation mechanism and a deflectionsensor are needed to realize such an oscillator. A fast feedback is required, but a simple counter can be used to determine the resonance frequency. The third option provides the most accurate frequency assessment, since the quality factor (Q-factor, ratio of center frequency and bandwidth) of the cantilever, which describes the sharpness of the system response, is enhanced by the feedback. In addition, only the third option can be realized in CMOS technology at reasonable expenses, since integrated spectrum analyzers and gain-phase meters are difficult to design. Consequently, the oscillator option was realized. In the following, two different methods to actuate the cantilever (thermal and magnetic), and the vibration detection via piezoresistors and stress-sensitive transistors will be presented. 5.3.1.1 Thermal Actuation The cantilever consists of single-crystal silicon covered by dielectric layers such as silicon oxide or silicon nitride for electrical insulation of the integrated electronic components. A cross-sectional view is shown in Fig. 5.23. Two heating resistors are integrated in the cantilever base, which are heated periodically to achieve cantilever vibration (see also Fig. 5.24). The temperature increase on the cantilever generates a bending moment due to the difference in thermal expansion coefficients of the silicon and the dielectric layers (thermal bimorph effect). The area of periodic temperature variation is closely confined to the region around the heaters due to the high excitation frequency of approx. 400 kHz. At frequencies higher than the corner frequency of approximately 1 kHz (speed of the thermal equilibration processes in the cantilever), the
5.3 CMOS Integrated Resonant Cantilever
111
Si-nitride Si-oxide polymer
n-well heater (p-diffusion) silicon frame Fig. 5.23. Cross section of a thermally actuated cantilever in CMOS technology. The heater at the cantilever base induces cantilever bending as a consequence of the bimorph effect (difference in thermal expansion coefficients of the silicon and the dielectric layers) [526]
heater
50 µm
piezo resistors
Fig. 5.24. Thermal actuation of the cantilever motion. The micrograph shows the heaters and the piezoresistors at the cantilever base [93]
efficiency of the thermal actuation usually significantly drops, which reduces the cantilever oscillation amplitude and leads to a phase lag that increases with frequency. However, the created bending moment is still sufficient to cause harmonic transverse vibrations of the cantilever with an amplitude of a few nanometers. The deflection of the cantilever is proportional to the applied heating power. Therefore, a sinusoidal voltage excitation cannot be used, since all the heating power goes into the system at DC and twice the resonance frequency. A DC-offset has to be added in order to excite the fundamental resonance of the cantilever [501]. Both types of cantilevers (thermally or magnetically actuated) are formed after completion of the regular 0.8-µm CMOS process sequence [496,527]. The first step includes anisotropic wet etching of silicon from the wafer backside with KOH using an electrochemical etch-stop technique (Fig. 5.25b). After
112
5 CMOS Platform Technology for Chemical Sensors
this step the silicon cantilevers are still embedded in a membrane formed by the dielectric layers. The second post-processing step consists of isotropic wet etching of silicon dioxide with buffered oxide etch (BOE), which finally releases the discrete silicon cantilevers (Fig. 5.25c). The cantilevers are composed of a layer sandwich including the dielectric layers of the CMOS process and a layer of single-crystal silicon (Fig. 5.23). The silicon layer has a thickness of 5.5 µm, silicon dioxide and silicon nitride thicknesses are 1.5 µm and 1.1 µm, respectively. The cantilevers are 150 µm long and 140 µm wide. Their fundamental resonance frequency is in the range of 400 kHz, and they exhibit a Q-factor of up to 1000 in air without chemically sensitive coating. dielectric layers (a) Si
n-well
(b) heater (c)
Fig. 5.25. Micromachining process sequence to fabricate cantilevers: (a) thinned CMOS wafer with Si-nitride on back side, (b) backside KOH wet etching with electrochemical etch stop, and (c) front-side isotropic wet etching of silicon dioxide to release the cantilever [47, 526]
The quality factor of the resonant cantilevers at atmospheric pressure is dominated by the viscous damping in the surrounding air [93]. In comparison to cantilevers employed in Scanning Probe Microscopy (SPM), the cantilevers as used for chemical sensors exhibit high quality factors. This is partly due to the fact that the cantilevers are not operated in close vicinity of a sample surface, and, hence, no squeezed-film damping occurs. The distance to the bottom of the etch cavity is 380 µm, i.e., large in comparison to the cantilever dimensions. Furthermore, the quality factor is not only a function of the resonance frequency but also of the cantilever geometry and its spring constant [528]. While SPM cantilevers for dynamic mode operation typically exhibit spring constants between 1 and 40 N/m, the cantilevers here were designed
5.3 CMOS Integrated Resonant Cantilever
113
to exhibit a much higher spring constant of 800 N/m. The measured quality factor of cantilevers in fundamental resonance strongly increases with increasing resonance frequency up to 200 kHz, reaching maximum values of approximately 1000 [93, 526]. This behavior is predicted by the theory of a damped harmonic cantilever beam with a damping term proportional to the beam velocity [529]. A theoretical estimation of the actual quality factors would, however, require a model for the dependence of the damping term on the cantilever dimensions. At frequencies higher than 200 kHz, a saturationlike behavior of the quality factor has been observed, indicating additional loss mechanisms, such as coupling losses into higher resonant modes or losses due to onset of acoustic radiation at higher frequencies. 5.3.1.2 Magnetic Actuation The magnetic actuation relies on Lorentz forces and requires an external magnetic field, which can be conveniently generated by including a small permanent magnet in the sensor package below the cantilever (Fig. 5.26) [530, 531]. The resulting magnetic-field vector is in plane and parallel to the cantilever. By applying an AC current such as a sinusoidal to the current loops that are patterned along its edges, cantilever oscillation is evoked in the external magnetic field by Lorentz forces. The direction of the Lorentz forces is perpendicular to the cantilever (Fig. 5.27) such producing a transverse cantilever movement. Two designs with different current paths for the electromagnetic actuation were implemented [531]. The current paths are realized using the two metal layers of the CMOS process. One design has a path with two loops and a resistance of approximately 12 Ω, the second has 8 loops and cantilever magnetic field lines
disc magnet
Si-substrate
Al-package
Fig. 5.26. Schematic of applying magnetic actuation to cantilevers: The external magnetic field is provided by a permanent magnet in the package [531]
114
5 CMOS Platform Technology for Chemical Sensors
B
Close-up Wheatstone bridge
J
stress-sensitive transistors
X FL
50 µm cantilever
current loop
Wheatstone bridge
Fig. 5.27. Micrograph of a magnetically actuated cantilever: AC current applied to the loops and external B-field initiate a cantilever oscillation via Lorentz forces. The close-up shows the arrangement of the stress-sensitive transistors [531]
a resistance of 48 Ω. The Lorentz force, FL , acting on the cantilever can be approximated according to the following equation [531]: FL = N · I · l · Bext .
(5.8)
N represents the number of current loops, I the applied current, and l the mean length of the loop perpendicular to the external magnetic field, Bext , (see Fig. 5.27). When the magnetic field, Bext , is oriented parallel to the cantilever length, the Lorentz-force vector is perpendicular to the plane of the cantilever. The Lorentz force exerted on the cantilever critically depends on the number of current loops. With a magnetic field of 80 mT, the effective current for 2 loops is 3.5 mA, and the resulting Lorentz force amounts to 37 nN, whereas in the case of 8 current loops, the effective current is 7.1 mA and the Lorentz force 71 nN. Especially for thick layers of chemically sensitive coatings, it is therefore necessary to use cantilevers with larger numbers of current loops to reach stable oscillation conditions. 5.3.1.3 Vibration Detection The Wheatstone bridge for vibration detection is located at the clamped edge of the cantilever, where the stress induced by the deflection is maximal. The bridge consists of either four piezoresistors (Fig. 5.24) or four diode-connected PMOS transistors (Fig. 5.27) [526]. The piezoresistive effect on a slightly p-doped silicon resistor is described by the relative resistance change, ∆R/R,
5.3 CMOS Integrated Resonant Cantilever
115
∆R = π L σ L + πT σ T (5.9) R where πL,T are the longitudinal and transversal piezoresistive coefficients and σL,T denote the respective stress-components. The piezoresistive coefficients in parallel and perpendicular orientation with respect to the cantilever axis have opposite signs. Therefore, a differential signal can be obtained by arranging two resistors in parallel and two resistors perpendicularly to the cantilever axis in a Wheatstone-bridge configuration (see Fig. 5.24). The common-mode voltage varies only a few percent because the absolute values of the longitudinal and transversal piezoresistive coefficients are almost identical [501]. The resistances in the Wheatstone-bridge are 950 Ω. This leads to a considerable power consumption of 26 mW. As minimum-width resistors had to be chosen for the Wheatstone-bridge due to area restrictions, a good matching cannot be expected. The matching of the resistors is further deteriorated by their perpendicular orientation. This leads to a large offset-voltage of up to 20 mV. MOS-transistor detection schemes have also been developed since mechanical stress changes the carrier mobility in the channel of MOS transistors [532, 533]. Four diode-connected PMOS transistors in a bridge configuration are used to generate an output voltage that is proportional to the stress at the clamped edge of the cantilever [532]. Again, opposite signs of the stress-sensitivities are achieved by orienting the channels of the two pairs of transistors perpendicularly and in parallel to the cantilever axis (see close-up in Fig. 5.27). The diode-connected MOS transistors show a lower small-signal transconductance in comparison to diffused resistors, which reduces power and area consumption. The disadvantages include a somewhat larger noise and a reduction of the sensitivity. 5.3.1.4 Cantilever Temperature The power dissipation of the electrothermally actuated cantilever is 32 mW with 6 mW resulting from the actuation and 26 mW from the piezoresistive Wheatstone bridge. For the magnetically actuated cantilever the power dissipation of the cantilever adds up to a total of 1.3 mW [501, 531]. It results mainly from the MOS-transistor Wheatstone bridge, the power dissipation of the actuation is negligible. The Wheatstone bridge with MOS-transistors has less power dissipation since its active load (20 kΩ) is much larger than that of the design with p-diffused resistors (950 Ω). As a consequence of the high power consumption and dissipation in the case of electrothermal actuation and piezoresistive detection, the cantilever temperature is significantly enhanced as can be seen in Fig. 5.28, which displays the simulated temperature distribution on the cantilever at minimum possible power consumption (actuation: 6 mW, bridge: 1.2 mW) [526]. The cantilever temperature under realistic operation conditions can be up to 20◦ C higher than ambient temperature, which strongly affects the detection process of chemicals, such
116
5 CMOS Platform Technology for Chemical Sensors
0 1.7 3.5 Wheatstone bridge (1.25 mW)
thermal actuators (6 mW)
temperature variation [°C]
Fig. 5.28. Temperature distribution on a thermally actuated cantilever with piezoelectric readout [526]
as the absorption of organic volatiles in polymeric matrices. As a rule of thumb, a 10-degree temperature increase here reduces the sensor signal by 50%. Magnetic actuation and MOS-transistor readout that do not produce any significant temperature change on the cantilever are, therefore, in most cases, a better choice. 5.3.2 Microcantilever Circuitry The cantilever constitutes the frequency-determining element of an oscillator circuit, the nature of which depends on the actuation/detection principle and will be, therefore, specified for thermal and magnetic actuation. 5.3.2.1 Thermal Actuation To operate the mechanical oscillator, the displacement generated by the heating pulses and the excitation voltage variations have to be in phase for positive feedback. A static DC-component has to be superimposed to a periodic excitation voltage to achieve cantilever oscillation at the mechanical fundamental resonance (see Sect. 5.3.1.1) [501]. This DC component of the heating current produces a general temperature increase on the cantilever. The power dissipation in the piezoresistors of the Wheatstone bridge constitutes a second source of heating. Both heating effects can cause a total temperature increase of up to 20◦ C above ambient temperature in the chemically sensitive region of the cantilever with approximately 10◦ C above ambient temperature at routine operation [93]. Measurements showed that the signal from the mechanical vibration only dominates in vicinity to the resonance frequency. At low frequencies, the thermo-mechanical actuation heats the whole cantilever and creates a temperature gradient in the Wheatstone-bridge that leads to an offset. Due to
5.3 CMOS Integrated Resonant Cantilever
117
the large temperature coefficient of diffused resistors, the resulting signal is larger than the mechanical response for frequencies up to 200 kHz (uncoated cantilever). Low-frequency signals therefore have to be eliminated by the feedback circuit, in particular since the amplitude of the thermal crosstalk may come close to the amplitude of the mechanical signal even at resonance. For frequencies higher than 200 kHz, capacitive crosstalk is observed as a consequence of the small distance between the diffused heating resistors and the diffused resistors of the Wheatstone bridge. This small distance gives rise to parasitic capacitances through the n-well. The oscillator circuitry as shown in Fig. 5.29 was designed according to the preceding considerations to achieve optimum signal quality [501]. The output signal of the resistive Wheatstone-bridge is first amplified by a lownoise differential difference amplifier (DDA). The amplification factor has a maximum value of 35 in order to avoid saturation of the amplifier by the DC-offset of the Wheatstone-bridge. The signal is then high-pass filtered to remove the offset voltages of the bridge and the first amplifier. The high-pass filter also prevents up-conversion of the amplifier 1/f-noise and eliminates the low-frequency thermal crosstalk and related error signals. AC-coupling at the input of the first amplifier would allow for a higher gain in the first amplification stage, but the related disadvantages prevail. They include added noise and the creation of an additional path for switching interference and substrate noise to the most critical point in the circuit. cantilever high-pass RH
R1
R2
Wheatstone bridge
R3
R4
+ DDA 1st amplifier
limiter
+ DDA -
∆t
2nd amplifier
delay
Fig. 5.29. Schematic of the cantilever feedback circuitry, which includes two cascaded amplifiers, a high-pass filter, a limiter, a programmable digital delay line, and a driving stage [501]
The signal is then amplified and high-pass filtered a second time (not shown) before it is converted into a square-wave signal by the comparator. The second amplification stage is needed to achieve sufficiently large amplitudes at the input of the comparator. The minimum amplitude at the input of the comparator is determined by (a) the input offset of the comparator: An amplitude at least ten times larger than the offset-voltage (≈1 mV) is needed for the desired duty cycle of 45–55%; and (b) the noise and crosstalk at the input of the comparator.
118
5 CMOS Platform Technology for Chemical Sensors
The feedback loop is then closed via an inverter followed by a Schmitttrigger, which is operated as a delay element. The rise-time of the inverter is digitally adjustable. This way, the phase shift is adjusted to ensure positive feedback. A source follower at the output of the delay line is used to drive the small heating resistor. The result is an integrated oscillator operating at the cantilever resonance frequency with a short-term frequency stability better than 0.1 Hz. The integration of the feedback loop on chip (Fig. 5.30) massively improves the signal-to-noise characteristics of the sensor. The detection of mass changes of less than one picogram on the cantilever has been achieved by recording the corresponding shifts in the cantilever resonance frequency. For more details on the circuitry see [501, 526].
cantilever
feedback circuit
500 µm Fig. 5.30. Micrograph of a thermally actuated cantilever (piezoresistive detection) with monolithically integrated feedback circuitry [501]
5.3.2.2 Magnetic Actuation The magnetically actuated cantilever featuring a MOS-transistor detection scheme, 8 current loops and an overall coil resistance of 48 Ω, (see Fig. 5.27 and Sect. 5.3.1.2), has been incorporated into an integrated oscillator. Some key issues for designing the feedback circuitry shall be briefly discussed. The vibration-induced output signal of the Wheatstone bridge exhibits an amplitude of 1 mV for a bias-current of 8.3 mA (corresponds to an excitation voltage of 400 mV) and a magnetic field generated by a 100-mT electromagnet. Consequently, the feedback circuitry has to provide an amplification of at least 32 dB (air) or 63 dB (water) to achieve a loop gain of more than 0 dB. The signal amplitude at the output of the MOS-transistor bridge varies by almost an order of magnitude in dependence of the excitation current,
5.3 CMOS Integrated Resonant Cantilever
119
the magnetic field, the distance between the magnet and the cantilever, and the damping of the vibration through the media, in which the cantilever is operated. Additionally, there will be drift on the bridge sensitivity owing to temperature fluctuations and material aging. For operation in liquids, the high viscosity and density of the surrounding medium entails enhanced damping and, consequently, a reduced signal amplitude (e.g., 30 dB reduction). Therefore, the gain of the feedback loop has to be adjustable in order to achieve stable oscillation under all operating conditions. The tunable gain of the second amplifier is used for the coarse tuning during startup. Small variations during circuitry operation are continuously compensated by having a nonlinear transconductance in front of the output stage. The total phase shift between the excitation and the output signal of the circuitry must be close to zero degree over a wide frequency range in order to achieve stable oscillation at the mechanical resonance of the cantilever. This holds particularly true since the mechanical resonance frequency of the cantilever varies with the coating material or the surrounding medium and is additionally subjected to fabrication tolerances. The thermomechanically actuated cantilever is driven by a square-wavetype excitation, since its vibration is correlated to the heating power (see Sect. 5.3.1.1). For the magnetically actuated cantilever, a sine-wave-type signal can be used since the mechanical vibration is correlated to the excitation current. Phase noise and power consumption can then be optimized. The MOS-transistor Wheatstone bridge shows a large DC-offset of 25 mV ±5 mV. This offset must be reduced or eliminated to avoid saturation of the amplifiers in the oscillator loop. Furthermore, the excitation signal at the cantilever should not include any DC-component to prevent unwanted heating of the cantilever and to minimize the cantilever power consumption. The last stage of the feedback circuitry has to drive the low-resistance coils that feature only 48 Ω. The circuitry schematic is shown in Fig. 5.31 [531,534]. A low-noise differential difference amplifier (DDA) [535] was chosen for the first amplification stage, since it requires little area and has low power consumption. The gain defined by the feedback resistors is 30 dB to avoid saturation of the amplifier by the DC-offset of the Wheatstone bridge. Between the first and second amplification stage, a first-order high-pass filter with a cut-off frequency of 15 kHz is added to eliminate the DC-offset. The only requirement for the high-pass filter is a cut-off frequency, which is at least three times smaller than the resonance frequency of the cantilever. The second amplifier is a variable-gain amplifier based on the DDA topology. In comparison to the first DDA, the design of the second amplification stage is more flexible. Instead of using a defined feedback resistance to determine the closed-loop gain of the amplifier, a current-controlled differential linear transconductor is introduced into the feedback loop of the amplifier [536]. By changing the bias current of the transconductor in the feedback loop, the closed-loop gain can be varied. The closed-loop gain of the second amplifier
120
5 CMOS Platform Technology for Chemical Sensors cantilever differential difference high-pass amplifier (DDA) filter
variable-gain amplifier
readout analog digital
class-AB buffer
amplitude regulation
high-pass filter
Fig. 5.31. Schematic of the cantilever feedback circuitry for magnetic actuation [534]
can be adjusted from 34 dB to 43 dB by tuning this bias current in the range of 50 µA to 110 µA. It is intended to introduce a programmable bias current in the next design. After this stage, another high-pass filter, which has the same structure and same cut-off frequency as the first high-pass filter mentioned before, is added to remove the offset of the upstream amplifiers. To initiate a self-oscillation of the system, the Brownian motion of the cantilever has to be amplified. Consequently, the closed-loop gain of the system at the beginning has to be larger than 0 dB and has to be controlled so that the gain is depending on the oscillation amplitude. To this end, a nonlinear transconductance as proposed in [537] is implemented in the system to regulate the oscillation amplitude. The load of the feedback circuitry includes the coil, which is integrated on the cantilever. Compared to its resistance, the inductance of the coil can be neglected. The resistance of eight loops used in this design is only 48 Ω, which requires a comparatively large driving current. As low power consumption is an important issue for a portable sensor design, a class-AB buffer is used, which reduces the power consumption by a factor of five in comparison to a simple buffer structure. For more details on the circuitry see [531, 534]. The mass change of the cantilever upon analyte exposure leads to a frequency change. The oscillation signal can be read out as an analog voltage or as a digital square wave. In this design, a buffer drives the output signal, which is measured externally. A micrograph of the overall integrated system is shown in Fig. 5.32. A simple and affordable solution to generate the magnetic field is the integration of a permanent magnet in the chip package. Before designing the package, it was necessary to find the minimum magnetic induction needed for a stable oscillation of the cantilever feedback system. Measurements performed
5.3 CMOS Integrated Resonant Cantilever differential difference amplifier DDA
high-pass filter
121
variable-gain amplifier
500 µm
cantilever
class-AB buffer
nonlinear transconductance
high-pass filter
Fig. 5.32. Micrograph of a magnetically actuated cantilever (MOS-transistor detection) with monolithically integrated feedback circuitry [534]
with a tunable electro-magnet showed that the minimum magnetic flux density necessary for stable oscillation is 70 mT [531]. The feedback circuitry exhibits a short-term frequency stability of better than 0.03 Hz in air. In contrast to thermal actuation, the application of which to resonant sensing in the liquid phase is difficult as a consequence of thermal losses and heat dissipation, the magnetic actuation scheme is also applicable to liquid phase dynamic measurements. Consequently the magnetically actuated cantilever system was characterized in air and in water. In both media, the system was first characterized in open-loop and, afterwards, in closed-loop operation (Fig. 5.33). In contrast to thermal actuation, the application of which to resonant sensing in the liquid phase is difficult as a consequence of thermal losses and heat dissipation, the magnetic actuation scheme is also applicable to liquid phase dynamic measurements. Consequently the magnetically actuated cantilever system was characterized in air and in water. In both media, the system was first characterized in open-loop and, afterwards, in closed-loop operation (Fig. 5.33). Under open-loop conditions at an external magnetic field of 200 mT, the cantilever oscillates at a resonance frequency of 425 kHz in air with a quality factor of 750 to 1100 (>100.000 for closed loop in air). In water, the fundamental resonance frequency of the system drops to 219 kHz with a quality factor of 23. Not only the quality factor but also the oscillation amplitude of the cantilever is reduced dramatically in water. The improvement of the quality factor of the cantilever in water as a consequence of closed-loop operation is demonstrated in Fig. 5.33 [534]. The quality factor increases from 23 in open-loop to 19,000 in closed-loop
5 CMOS Platform Technology for Chemical Sensors
amplitude [V]
122
closed-loop response Q = 19,000
open-loop response Q = 23
frequency [kHz] Fig. 5.33. Open-loop and closed-loop frequency response of cantilever system in water [534]
operation. The resonance frequency in closed-loop operation in water is 221 kHz, which is slightly different from the resonance frequency in open-loop operation. This is due to the fact, that in closed-loop operation the cantilever oscillates at a frequency, at which the total phase of the system is 0◦ , which is 221 kHz in this measurement. Figure 5.33 clearly demonstrates again the benefits of having circuitry monolithically integrated with the resonant structure. 5.3.3 Microcantilevers as Chemical Sensors To render the cantilever sensitive to chemical species, thin layers (3–6 µm) of polymers were deposited, which serve as absorption matrix to detect volatile organic compounds in air. The investigations were restricted to polymeric films, for which physisorption (no chemical interaction) and bulk dissolution of the analyte within the polymer volume are the predominant mechanisms. Upon absorption of analytes by the coating, the physical properties of the polymer film, such as its mass, change. This mass change is detected by monitoring the corresponding resonance frequency shift of the coated cantilever. 5.3.3.1 Polymer Coating The deposition of the sensitive polymeric layer, which is performed by spray coating using an airbrush or by drop deposition from polymer solutions, already changes the resonator properties of the cantilevers. Figure 5.34
5.3 CMOS Integrated Resonant Cantilever
123
-150
10 uncoated
uncoated
1 coated
0.1
phase [°]
amplitude [mV]
-200 -250 -300 -350 0.01 100
200 300 400
frequency [kHz]
-400 100
coated
200 300 400
frequency [kHz]
Fig. 5.34. Measured amplitude and phase of a cantilever before and after coating with PEUT (after amplification by the DDA, 30 dB). The mechanical resonance signal determines overall amplitude and phase characteristics only in vicinity to the resonance frequency. At lower and higher frequencies, thermal and capacitive crosstalk prevail [501]
shows a Bode-plot of a cantilever before and after coating with 2 µm poly (etherurethane), PEUT [501]. After coating, the resonance frequency and the amplitude at resonance decrease. A larger loop-gain is needed to compensate for the decrease in amplitude. For a harmonic oscillator, the phase shift between excitation at resonance and response is independent of the resonance frequency. Due to the thermomechanical actuation, the resonant cantilever shows an additional phase lag, which increases with frequency as was mentioned in Sect. 5.3.1.1. The phase shift at resonance is decreased after the coating procedure and depends on the polymer thickness. Due to this decrease and the fabrication-tolerance-induced fluctuation of the initial resonance frequency (<5%, see Fig. 5.35), the phase at resonance cannot be predicted accurately. The feedback circuitry therefore includes an adjustable phase-shifter in order to guarantee a positive feedback also with larger polymer coating thickness. The Q-factor of the cantilever and, accordingly, the resonator stability are also decreasing with increasing polymer-thickness. While the chemical sensitivity increases with polymer thickness (more sorption matrix produces larger mass changes), the minimum detectable frequency shift is defined by the stability of the oscillator. This leads to a trade-off for the polymer thickness. For PEUT, an optimum thickness of approx. 3–6 µm was found [526]. The resulting shift in resonance frequency is 50 kHz, and the Q-factor drops from 950 to 600. This is partly due to the fact that quality factor and vibrational amplitude depend to a critical extent on the plastic or viscous properties of the thin polymer films. Furthermore, it is observed that quality
124
5 CMOS Platform Technology for Chemical Sensors
0.04
gain
frequency: 394 kHz ± 3% Q-factor in air: 980 ± 5%
0.02
0 375
400
425
fundamental frequency [kHz] Fig. 5.35. Micromachining fabrication tolerances of 4 cantilevers in an array on the same chip (see Fig. 5.39) [526]. The resonance frequency varies between 380 and 405 kHz (394 ± 3%), the Q-factor is 980 ± 5%
factor and amplitude depend strongly on the homogeneity and morphology of the polymer film and, thus, on the quality of the deposited layer and the reproducibility of the deposition process. Since the cantilever beam shows a decrease in resonance frequency upon sensitive-layer deposition, the polymer layer thickness can be monitored on line during the coating procedure by means of frequency measurements. 5.3.3.2 Analyte Absorption When an analyte is absorbed in the polymeric film, the mass change produces a frequency change of the resonant cantilever, which is essentially functioning as a balance. A schematic of the transducer is shown in Fig. 5.23 [526]. A silicon cantilever of the thickness tSi = 6 µm is covered with silicon dioxide layers (tOx = 2.4 µm) and a silicon nitride passivation (tN i = 1.1 µm). The silicon nitride passivation protects the embedded electrical components. The cantilever is coated with a polymer layer, which is between 3 and 6 µm thick (tL ). The fundamental resonance frequency of the composite beam for small deflections is given by [538, 539]: ˆSi ISi + E ˆOx IOx + E ˆN i IN i + E ˆL IL E λ20 . (5.10) fo = 2 2πL ρmean F ˆ denotes the apparent Young’s modulus of each material, I is the Here, E respective moment of inertia and ρmean is the average specific mass density of the composite beam. F denotes the cross-sectional area and L denotes the length of the cantilever. λ0 is an integration constant and has a value of 1.875 for the cantilever geometry under consideration.
5.3 CMOS Integrated Resonant Cantilever
125
The moment-of-inertia contribution of each layer depends on the distance of the particular layer to the neutral layer of the beam. The neutral layer is a particular plane of the beam, where the internal mechanical stress during vibration equals zero. When the beam shown in Fig. 5.23 bends upwards, the uppermost layers experience compressive stress due to the cantilever bending, whereas the bottom layers experience tensile stress. The opposite holds true, when the cantilever bends downwards. The neutral layer is the plane near the cantilever center, which experiences neither tensile nor compressive stress during oscillation though it is deformed through the cantilever motion [539]. The average specific mass density (cantilever materials and sensitive layer), ρmean , and the cantilever specific mass density (only cantilever sandwich materials), ρcant , are [93]: ρSi tSi + ρOx tOx + ρN i tN i + ρL tL tSi + tOx + tN i + tL ρSi tSi + ρOx tOx + ρN i tN i = . tSi + tOx + tN i
ρmean = ρcant
and (5.11)
Additional mass on the cantilever leads to a decrease in the resonance frequency. Therefore, negative resonance frequency shifts are detected upon absorption of a gaseous analyte in the polymer. These frequency shifts originate from the polymer mass density increase ρL as a consequence of the additional mass of the absorbed gas molecules. Due to the high spring constant of the cantilever (k ≈ 800 N/m), influences from changes in the elastic properties of the polymer upon gas absorption, the effects of which are more than two orders of magnitude smaller, can be neglected. Under the assumption that the cantilever is covered with a polymer layer of uniform thickness, the average mass density change of the cantilever as a consequence of the polymer density change (polymer swelling can be neglected in a first order approximation at such low analyte concentrations [69]) is given by: tL tL ∂ρmean = = ∂ρL tSi + tOx + tN i + tL h
(5.12)
with h being the total thickness of the cantilever (coating and cantilever materials). Starting with (5.10), the shift in the resonance frequency, f0 , upon a change in the mass density of the polymer layer can be calculated: G=
f0 ∂f0 ∂ρmean 1 tL ∂f0 . = · = · · ∂ρL ∂ρmean ∂ρL 2 ρmean h
(5.13)
Here, G represents the gravimetric sensitivity, i.e., the change in frequency due to a change in polymer density. The corresponding mass sensitivity Gm = G · VL is determined by taking into account the polymer volume VL .
126
5 CMOS Platform Technology for Chemical Sensors
The origin of the polymer density change ∂ρL is the absorption of the analyte gas of concentration cA . A cantilever resonance frequency shift indicates the corresponding mass change. The sensitivity, S, of the resonant gas sensor is defined as [93]: ∂f0 ∂ρL ∂f0 = · = G · SA . (5.14) ∂cA ∂ρL ∂cA As demonstrated earlier, the resonance frequency change results from the polymer density change upon analyte absorption. G describes the mechanical properties of the cantilever device, whereas SA relates to the polymer/analyte interactions. The mass increase, ∂mL , of the polymer with volume VL and the corresponding polymer density increase, ∂ρL , upon an analyte concentration change, ∂cA , is given by: S=
∂mL = Kc · ∂cA . (5.15) VL The analyte sensitivity, SA , hence equals the partition coefficient, Kc , (see Chap. 2, 2.10) when mass concentrations (mass/volume) are used. Finally, the sensitivity, S [Hz/(µg/l)], of the cantilever gas sensor can be rewritten as [93]: ∂ρL =
f0 tL 1 · · Kc . · (5.16) 2 ρmean h Variations in the material properties of the polymer barely influence the cantilever resonance frequency in the investigated range, because the cantilever overall elastic properties are dominated by those of the CMOS materials. As an example, a 50% inaccuracy of the elastic modulus of 100 MPa at a polymer layer thickness of 12 µm results in a resonance frequency variation of only 0.8%. The influence of the polymer density is larger, but still not significant. Here, an uncertainty of 10% leads to a variation of 1.9% [93]. The sensitivity of the gas sensor is the product of the thickness-independent analyte sensitivity SA and a thickness-dependent gravimetric sensitivity G (5.14). The total sensitivity of the resonant cantilever gas sensor, S, linearly depends on the thickness of the applied polymer layer (thin polymeric layers), like it is the case with all other polymer-based gas sensors and has been confirmed in a wealth of measurements [63, 70, 93, 95]. Therefore, only one graph (Fig. 5.36) will be shown to illustrate the dependence of the sensor signal on the layer thickness for toluene absorbed into PEUT [531]. The sensor signals increase linearly with increasing analyte concentration as can be seen in Fig. 4.1 (Sect. 4.1) and in Fig. 5.37 [93]. Rather low analyte concentrations (5–20 ppm n-octane) close to the detection limit were applied to a cantilever coated with 2.8 µm PEUT. Due to their linear analyte response characteristics, the slopes of the sensor response upon analyte exposure are often recorded and are given for the case of a PEUT-coated (2.4 µm) cantilever in Table 5.1 [93]. S =
5.3 CMOS Integrated Resonant Cantilever
127
analyte concentration [ppm] 0
500
1000
1500
2000
0
∆f [ Hz]
-100
1.6 µm 5.6 µm
-300
7.4 µm
-500
14.7 µm
toluene and PEUT
Fig. 5.36. Response of cantilevers coated with PEUT at different thickness upon toluene absorption [531] 0
frequency shift [Hz]
n-octane [ppm] 5
9
10
20
8 7 6
polymer: 2.8 µm PEUT 5
0
5
10
15
20
25
30
35
time [min] Fig. 5.37. Low-concentration sensor responses: 5, 10, 20 ppm (0.5, 1, 2 Pa) noctane were dosed to a cantilever (thermal actuation), which was coated with a 2.8-µm-thick PEUT layer [93]
The sensor was actuated by approximately 10 mW heating power. Linear responses are to be expected in the low-concentration range (less than 1–2% of the saturation vapor pressure at the respective operation temperature), since Henry’s law is valid [63, 69, 525]. At higher concentrations, deviations may occur as a consequence of the analyte plasticizing the polymer. The longterm drift of the resonance frequency was subtracted from the measurements [93, 526]. A frequency shift of 49 Hz was measured for, e.g., 400 ppm toluene. The different analytes evoke different sensor responses as a consequence of their different molecular weights, their different gas-phase saturation vapor pressures, and due to the different partition coefficients of the analytes in PEUT. The response slope of, e.g., toluene is considerably higher than that of ethanol, since toluene is less volatile (higher partitioning in PEUT), its molecular weight is higher, and its saturation vapor pressure is lower.
128
5 CMOS Platform Technology for Chemical Sensors
Table 5.1. Measured sensitivity of a cantilever coated with 2.4 µm PEUT to various analytes (thermal actuation, 10 mW heating power) [93]
Analyte n-octane toluene ethyl acetate ethanol
Sensitivity [Hz/ppm]
Saturation Vapor Pressure at 301 K [Pa]
Molecular Weight [m.u.]
−0.05 −0.1 −0.02 −0.01
2200 4300 14400 8700
114 92 88 46
frequency [Hz]
A sensor signal of a magnetically actuated cantilever at high time resolution shows the immediate response and fast recovery of the sensor as a consequence of the relatively weak physisorption interactions (Fig. 5.38) [531]. The response characteristics and analyte sensitivity do, of course, not depend on the actuation mechanism. The geometric dimensions (length, thickness) and the mechanical properties of both cantilevers are identical. The only important issue is the actuation-induced heating in the case of the thermal actuation that leads to an exponential signal reduction with increasing cantilever temperature, whereas heating does not occur in the case of magnetic actuation. A possible temperature increase on the cantilever has to be taken into account in comparing the signals of the different cantilevers.
magnetic actuation
380830
120 ppm toluene
380820
5.8 µm PEUT
380810 0
200
400
600
time [s] Fig. 5.38. Resonance frequency of a cantilever (magnetic actuation) coated with 5.6 µm of PEUT upon exposure to toluene at a partial pressure of 12 Pa (120 ppm) [531]
Polymers as sensitive layers are only partially selective. A PEUT-coated sensor will, e.g., respond to ethanol but also to many other analytes with varying sensitivity (see Table 5.1). Better discrimination can be achieved with a system of more than one sensor, a so-called sensor array. Figure 5.39 shows a micrograph of an integrated cantilever array comprising four cantilevers, the feedback circuitry, and a multiplexer to sequentially address and operate the cantilevers. The resonance frequencies of the
5.3 CMOS Integrated Resonant Cantilever
129
feedback circuitry
cantilever
multiplexer bonding pads
500 µm
Fig. 5.39. Micrograph of an integrated cantilever array (chip size: 2 × 2 mm2 )
cantilevers are in the range of 380–410 kHz and differ by maximum ±10%. The same holds for the quality factor of the resonators, which amount to 980 ± 10% (see Fig. 5.35) [526]. To illustrate the use of cantilever arrays for quantitative analysis, a binary mixture of n-octane and toluene has been dosed to the sensors. The analytes show different interactions and sorption properties. n-Octane is nonpolar and not polarizable, whereas toluene contains an aromatic ring with enhanced electron density, which renders the molecule polarizable. Measurements with mixtures of varying n-octane and toluene contents were recorded, some signals of which are displayed in Fig. 5.40. Three cantilevers of the array were coated with 0.6 µm of poly(dimethylsiloxane), PDMS, 1.5 µm of poly(cyanopropylmethylsiloxane), PCPMS, and 1.8 µm of poly(etherurethane), PEUT. One cantilever was left uncoated uncoated
PDMS
PCPMS
PEUT
700
600
500
400
700
300
ppm toluene 600
500
400
binary mixture 20
300
frequency shift [Hz]
40
0 -20 -40 -60
600 ppm n-octane
700 ppm n-octane
-80 0
20
40
60
80
100
time [min]
Fig. 5.40. Responses of an array of cantilevers (Fig. 5.39) coated with different polymers (PDMS, PCPMS, PEUT) upon exposure to mixtures of n-octane (600 ppm, 700 ppm) and toluene (300–700 ppm)
130
5 CMOS Platform Technology for Chemical Sensors
to look into surface adsorption phenomena, which were found to be marginal. The sensors were exposed to two different n-octane concentrations (600 ppm, 700 ppm) and, additionally, varying toluene concentrations (300– 700 ppm). Since the four sensors respond differently to the analyte mixtures, the recorded data sets can be fed into multilinear regression algorithms or other multicomponent analysis tools that provide quantitative determination of both components after preceding calibration. For more details on pattern recognition and multicomponent analysis methods, see [12–17]. 5.3.4 Comparison of Cantilevers to Other Mass-Sensitive Devices In this section, the sensing performance of the resonant cantilever will be compared to that of the most common and established mass-sensitive transducers, the thickness-shear-mode resonator (TSMR) and the Rayleigh surfaceacoustic-wave (SAW) device. The sensitivity equations for cantilevers and TSMRs are very similar. The cantilever equation (5.16) can be rewritten by replacing the cantilever mass density (only cantilever sandwich materials) ρcant for ρmean in order to separate layer and transducer variables [93]: S = −
1 tL · f0 Kc . · tL 2 ( ρcant + ρL tcant ) tcant
(5.17)
The corresponding equation for the TSMR can be derived from the Sauerbrey equation (4.1) [64]: S = −
tL 1 · · f0 Kc . ρquartz tquartz
(5.18)
A slightly different equation could be derived for the Rayleigh SAW provided that there are no modulus contributions to the sensor response [540,541]. The TSMR resonates in the thickness shear mode, and the polymer layer is not deformed through the transducer movement. In the case of the SAW-devices, the polymer layer undergoes deformation, and modulus terms must be accounted for as soon as the polymer modulus is beyond a certain limit [65]. The cantilever vibration also deforms the polymer, so that modulus contributions have to be included in the equations. Due to the large spring constant of the cantilever materials (very stiff cantilever), however, it is justifiable to neglect the effect of sorption-induced polymer modulus changes on the sensor response, in particular since all used polymers are low-modulus (rubbery) polymers, and the analyte concentrations are very low. In comparing (5.17) and (5.18) it is obvious, that the sensitivity is, in both cases, proportional to the fundamental device frequency, the partition coefficient and the ratio of polymer layer thickness and transducer thickness. The main difference is, that the TSMR equation (5.18) does not include any term related to the mass density of the polymer layer, which is a consequence
5.3 CMOS Integrated Resonant Cantilever
131
of the fact, that the Sauerbrey equation only describes simple mass loading without using a sensitive layer or a sorption matrix. It has additionally to be noted that (5.18) only holds true for acoustically thin layers, for which no phase lag occurs between the moving quartz surface and the outer surface of the polymer coating [65, 542]. Otherwise, polymer modulus terms also have to be taken into account. The absolute mass resolution of the cantilevers is in the range of a few picograms as assessed by many authors for different cantilever geometries and operation modes (static or dynamic) [96–107, 543–546]. This high absolute mass sensitivity does, however, not necessarily imply an exceptionally high gas sensor sensitivity, since the area coated with the sensitive layer usually is very small (on the order of 100×150 µm2 or 100×500 µm2 ). As a consequence, the overall polymer volume, which absorbs the analyte, and the resulting overall mass change upon analyte absorption on the cantilever is much less in comparison to the larger TSMR and SAW devices. The higher absolute mass sensitivity of the cantilever is, therefore, counteracted by the small transducer dimensions. Table 5.2 includes characteristic experimental results for TSMRs and SAWs as taken from [69] and the cantilever data compiled in this study. Although the fundamental resonance frequency of the cantilever is more than two orders of magnitude lower than that of the other devices, its performance is comparable to those of the TSMR and the SAW delay line. Using a rather conservative frequency noise estimate of 0.1 Hz for the cantilever and the TSMR as reported in [69] (in both cases the measured short-term frequency noise is considerably lower [69]), the limit of detection of the cantilever is 2.8 ppm n-octane (operation at 28◦ C ambient temperature), which is close to that of the TSMR (1.5 ppm), and better than that of the SAW delay line (7 ppm n-octane). From all these results it looks like detection limits achieved with cantilevers are very similar to those obtained with TSMRs. Some differences in the transducer characteristics, however, merit further discussion. Table 5.2. Comparison of mass-sensitive sensors in detecting 100 ppm (10 Pa) of n-octane. The frequency noise (short-term) was set to 0.1 Hz for the cantilever and the TSMR [17], and to 2 Hz for the SAW device [69] in determining the limits of detection (LODs) [93]
Device Cantilever (28◦ C) TSMR (30◦ C) SAW (30◦ C)
f0 [MHz] 0.4 30 80
∆fpoly [kHz]
∆fgas ∆fgas /f0 ∆f/∆T ∆fgas /∆f/∆T LOD [ppm] [K] S/N = 3 [Hz] (×106 ) [Hz/K]
−35 12.0 (4 µm) −72 20.6 (0.32 µm) −330 87.9 (1.56 µm)
30
17
0.7
2.5
0.7
−15
1.4
1.5
1
−160
0.6
7
132
5 CMOS Platform Technology for Chemical Sensors
The large fractional frequency shift upon gas exposure as compared to the fundamental resonance frequency significantly contributes to the good performance of the cantilever. A cantilever beam shows an approximately 30 times higher fractional frequency shift than TSMRs and SAWs (Table 5.2). The absolute frequency shift values of the cantilever are lower than those of the other devices and, therefore, frequency counters with higher precision are required for readout. On the other hand it is not necessary to meet the strict high-frequency requirements (shielding etc.) of SAW-device operation. The cantilever transducer also tolerates larger coating thicknesses, e.g., 4 µm coating thickness at 9.5 µm cantilever thickness (Table 5.2) [93]. This positively affects the transducer sensitivity as can be taken from the second term in (5.17) and (5.18). In the case of the TSMR, the quartz plate thickness is 56 µm at 30 MHz, and the coating thickness is 100–300 nm. The thickness of SAW-coatings is in the range of 200 nm to 1 µm, the thickness of the quartz plate (approx. 1 mm), however, does not affect the device characteristics, since the penetration depth of the acoustic wave is approximately one acoustic wavelength (40 µm at 80 MHz, 7.3 µm at 433 MHz). TSMR and cantilever show a sensitivity that is proportional to the resonance frequency as can be seen in (5.17) and (5.18). However, in the case of the cantilever, the inherent geometry dependence (width, length, thickness) provides much more flexibility in optimizing sensor sensitivity. It is, e.g., possible to increase the mass sensitivity by reducing the cantilever thickness. At the same time, a high cantilever resonance frequency can be maintained by shortening the cantilever. In contrast, the only tunable parameter of a TSMR is its plate thickness. Considering the temperature stability, the quartz is still preferable (Table 5.2), whereas the cantilever shows similar temperature dependence as the SAW device. This is a consequence of temperature-induced variation of material properties of the CMOS composite cantilever beam and the temperature-induced changes in the relatively thick polymer layer. A big advantage of the cantilever is its compatibility to microelectronics and the possibility to co-integrate transducer and electronics on a single chip (miniaturization). Placing all the oscillator circuitry, the amplifiers and the analog/digital converter in immediate vicinity to the transducer (Figs. 5.30, 5.32, 5.39) drastically increases the overall device performance (as has been demonstrated) and renders the chip less sensitive to external disturbances. Integration also simplifies the packaging because there are no sensitive analog signals that need to be shielded from external noise sources or electromagnetic interference. Finally, integration allows for using on-chip bus-systems and for co-integration of “smart” and self-test features.
5.4 CMOS Microhotplate System Development
133
5.4 CMOS Microhotplate System Development As already introduced in Sect. 4.4.3.1.2, wide-bandgap semiconducting metal oxides such as tin oxide, gallium oxide or indium oxides, which have to be operated at elevated temperature (250–600◦ C), are frequently used materials in gas sensing. These oxides, most of which meanwhile exhibit sufficient long-term stability, are very sensitive to a multitude of inorganic gases and volatile organics and drastically change their resistance upon exposure to those analyte gases [399, 431–433, 547, 548]. The integration of heated structures with CMOS technology is particularly challenging, since the metal-oxide operating temperatures of 250◦ C to 600◦ C are much higher than the temperature specifications of common integrated circuits (between −40◦ C and 120◦ C). The metal oxides have to be deposited on miniaturized heatable structures that are thermally well isolated from the rest of the chip. Many designs have been denoted “CMOS compatible”, which, however, in most cases means, that CMOS materials have been used, or the sensor design can be used with a modified CMOS process. Only few microhotplate-sensor designs have been fabricated in an industrial standard CMOS process with subsequent sensor-specific postprocessing [428, 443, 549–553]. In the next sections, the CMOS-based microhotplate transducer and its components as well as the fabrication and post-processing technology will be presented. Several different microhotplate designs and realizations have been fabricated. Thereafter, three integrated microhotplate systems, which constitute subsequent development steps, will be detailed with regard to circuitry implementation and electrical and chemical sensor performance. 5.4.1 CMOS Microhotplates Microhotplates, in most cases, consist of a thermally isolated area such as a suspended membrane featuring a heater, a temperature sensor and contact electrodes for the sensitive layer (Fig. 5.41). The membrane, which isolates the heated area from the bulk chip, is formed by the CMOS dielectric layers (Si-oxide and Si-nitride) that exhibit low thermal conductivity. The membrane is released by etching away the bulk silicon. Depending on the micromachining procedure, it is possible to leave a silicon island underneath the heated area (Fig. 5.41) [551, 553]. Such an island can serve as a heat spreader and also mechanically stabilizes the membrane. The circuitry is then arranged on the bulk chip, the temperature of which negligibly changes upon hotplate heating as has been assessed by using an additional temperature sensor in vicinity to the circuitry. By realizing such microhotplates, high operation temperatures can be reached at comparably low power consumption (40–100 mW), and the devices feature a small thermal time constant on the order of 10 ms so that
134
5 CMOS Platform Technology for Chemical Sensors
chip temperature sensor and circuitry
thick-film SnO2 resistor
temperature sensor
n-well (Si-island) p-silicon substrate
electrodes
dielectric layers (membrane)
polysilicon heater
Fig. 5.41. Cross-sectional schematic of a microhotplate featuring the different components and a droplet of tin dioxide as sensitive layer [551]
temperature modulation techniques with the aim to improve sensor selectivity and sensitivity can be applied [436, 464]. The microhotplate design and development was guided by the following considerations [551, 553]: • Membrane Layout: The membrane layout should be as symmetric as possible to achieve good temperature homogeneity over the membrane area and, as a consequence, low stress gradients. This includes also thermal stress owing to the mismatch of the thermal expansion coefficients of the layer materials. • Temperature Distribution: A homogeneous temperature distribution in the heated area is highly desirable to ensure that all sensing processes on the hotplate take place at the same defined and precisely controlled temperature. • Thermal Resistance: A high thermal resistance is needed to achieve minimum power consumption. • Supply Voltage: The desired operating temperature (250◦ C–350◦ C) has to be reached with a supply voltage of 5.5 V. In view of the above guidelines, an octagonal, and, later, a circular design for the heated area on a square membrane were selected, which are shown in Fig. 5.42 [553, 554]. A considerable part of the heat usually is dissipated through the metal lines to the bulk. The octagonal and circular shape exhibit relatively long metal leads in comparison to a square or rectangular heated-area design, especially since the distance between the hot and the cold end of the metal leads can be increased by arranging them along the membrane diagonals. A reduction of the hotplate heated-area diameter from 300 µm (Fig. 5.42) to 100 µm as shown in Fig. 5.43 further improves the device thermal resistance from 4.8◦ C/mW (octagonal) or 5.8◦ C/mW (circular) to 10◦ C/mW [555].
5.4 CMOS Microhotplate System Development
(a)
I heat
(b)
n-well island
135
I heat
temperature sensor heater electrodes dielectric membrane
100 µm
100 µm
Fig. 5.42. (a) Octagonal and (b) circular hotplate designs with heaters, temperature sensors and electrode structures. Both designs feature a silicon island underneath the heated area [553, 554] 100 µm n-well island
heater
temperature sensor
electrodes
dielectric membrane
Fig. 5.43. Microhotplate featuring a circular heated area of only 100 µm diameter [555]
The width of the metal lines along each diagonal is approximately identical. Thus, a symmetric heat flow through the membrane is created. The heat is produced by a cloverleaf-shape (octogonal) or ring-shape heater (circular) that extends along the edge of the heated membrane area. The circular heater and heated-area design also perfectly match the shape of the sensitive tin-dioxide droplet: Area that is not covered by the sensitive material is not heated, and, consequently, the heat losses to the ambient air are minimized. A SEM (scanning electron microscope) micrograph of a microhotplate coated with a SnO2 -droplet is shown in Fig. 5.44 [553]. The grainy structure of the nanocrystalline material is clearly visible. From the conventional heat conduction equation it is found that the temperature gradient within the heated area can be minimized by (a) using highthermal-conductivity materials in the heated area and by (b) generating the
136
5 CMOS Platform Technology for Chemical Sensors
300 µm
circular hotplate
tin oxide spot suspended membrane x300
6.00 kV
100 µm
Fig. 5.44. Tin-dioxide-coated hotplate. The metal oxide droplet exclusively covers the circular heated area (left) [555]. The grainy and porous structure of the nanocrystalline material can be seen in the SEM picture (right) [553]
heat at the location with the largest thermal losses. Both issues were taken into account in realizing the heated-area hotplate design. First, a 5.5-µm-thick n-well silicon island, which effectively distributes the heat owing to the high thermal conductivity of silicon, was fabricated underneath the heated area in the center of the membrane [551, 553]. The island also mechanically stabilizes the membrane, which results in lower membrane buckling. The second feature is a novel type of polysilicon cloverleaf-shape/ring heater including two semicircular heating resistors along the edges of the heated area, which are connected in parallel. Using this heater type, the heat is generated in the locations with the largest heat losses via membrane and metal leads to the bulk, which ensures that the temperature in the sensitive area enclosed by the ring heater is homogeneous [553, 555]. The parallel heater configuration moreover decreases the overall heater resistance to a nominal value of 125 Ω. With such a low resistance it is possible to provide enough heating power to reach high temperatures (up to 400◦ C) with the on-chip driving circuitry at 5.5 V supply voltage. A polysilicon temperature sensor is placed in the center of the membrane. This sensor is used to assess the membrane temperature and is realized in a in a four-point configuration. It serves as feedback element in the temperature control loop (see Sects. 5.4.2.1–5.4.2.3) and features a nominal resistance of 10 kΩ. An additional reference resistor is needed for the control circuitry [556]. Devices featuring a network of temperature sensors in the heated membrane area were fabricated to assess the temperature homogeneity. The temperature sensors were also realized as four-point arrangements. The resistance of each of those temperature sensors can be measured individually. These additional temperature sensors were placed on the right and left side of the electrode pair and at the upper right corner of the membrane (see markers in
5.4 CMOS Microhotplate System Development
137
Fig. 5.45a) [557]. The hotplate does neither exhibit a silicon plate underneath nor a polysilicon or metal heat spreader. The relative discrepancy of the temperature as measured by the different sensors (T2–T4) with respect to the sensor in the membrane center T1, which acts as a reference, is shown in Fig 5.45b as a function of the hotplate temperature (T1) [557]. The values named T2 represent, e.g., the relative difference, (T2–T1)/T1, in percent [557]. One would intuitively expect, that T1 shows the lowest temperature owing to the ring heater scheme, which would lead to a positive relative difference value for all other sensors. However, T2 shows a lower temperature than T1 owing to the fact that T2 is in close proximity to the wide metal line of the heater supply. As a consequence of the large heat flux through the metal line to the bulk silicon, the measured temperature of T2 is lower. 25
(a) 2
(b)
T2 T3 T4
20 15
3
∆T/T1 [%]
10
1
5 0 -5
-10
4 100 µm
-15
0
50
100
150 200 T1 [°C]
250
300
350
Fig. 5.45. (a) Hotplate with temperature sensor network for temperature homogeneity assessment. (b) Temperature distribution on the microhotplate: Relative discrepancies [%] between temperature measured in locations 2, 3, 4 and that measured in the center (1) as a function of the hotplate temperature [557]
The temperature discrepancy between location T4 and T1 is close to 15% and that between T3 and T1 is approximately 9%, both of which exhibit positive signs as expected. The active region of the sensitive material is between the two electrodes, so that T3 represents the temperature in the active sensor region. Although the hotplate does not exhibit any feature for improving the temperature homogeneity (silicon island etc.), the temperature gradient at T3 at 300◦ C hotplate temperature is only 0.3◦ C/µm [557]. With a polysilicon plate or a Si-island underneath, the temperature homogeneity is further improved. For an identical device with a Si-island heat spreader, a relative deviation of less than 2% in the heated area was achieved [551].
138
5 CMOS Platform Technology for Chemical Sensors
The electrodes on the membrane that establish contact to the sensitive layer consist of platinum, which is deposited on top of the CMOS-aluminummetallization. A 50-nm titanium/tungsten layer sandwiched in between the aluminum and platinum acts as diffusion barrier and adhesion layer for the subsequently deposited 100-nm-Pt thin film. For advanced reliability, the temperature sensor was also implemented as a platinum resistor especially since platinum deposition has to be done anyway to realize the electrodes [555]. A platinum temperature resistor shows less resistance drift as compared to a polysilicon sensor, the characteristics of which strongly vary over time and, in particular, with increasing exposure time to high temperatures. Similar considerations would also apply to the polysilicon heater but the heater resistance drift has no critical impact on the device performance since the circuitry controls the temperature according to the temperature sensor input (see Sects. 5.4.2.1–5.4.2.3), so that a drift in the heating resistor properties has no effect on the hotplate temperature. A hotplate design with a Pt temperature resistor is shown in Fig. 5.46. It has to be noted that replacing a low-resistance Pt temperature sensor for a polysilicon temperature sensor requires major changes in the temperature sensor readout circuitry [555,558]. An additional burn-in-heater for annealing the sensitive material was also integrated, which was not used so far. The basic idea was to provide a heater, which can achieve much higher temperatures than the heater for normal operation. Since the temperature for material sintering on wafer-level is limited to 400◦ C in order not to change the transistor characteristics of the integrated electronics (last thermal step of the CMOS process is approximately 400◦ C), higher annealing temperatures can only be applied locally on the membranes by using the extra-heater with an external power source. The annealing heater is, if ever, used once, so that only the heater, which is used for operating the hotplate and which is connected to the circuitry, was further optimized for performance.
electrodes
n-well island
CMOS metallization
heater Pt metallization
Pt temperature resistor dielectric membrane
100 µm
heater contact
Fig. 5.46. Microhotplate featuring a meander-type Pt-temperature resistor [555]
5.4 CMOS Microhotplate System Development
139
5.4.1.1 Temperature Sensor Calibration The various temperature resistors have to be calibrated in an oven by exposure to defined temperature steps, since it is well known that there is a large production spread in polysilicon resistors fabricated in a CMOS process. This production spread is, on the one hand, a consequence of geometric variations of the temperature sensor itself, and, on the other hand, a consequence of fluctuations in the material properties of the polysilicon resistor. For the temperature sensor a second-order polynomial was extracted, which includes the temperature coefficients of polysilicon (a1 , a2 ) [551]: ∆T = T − T0
=
a1
∆R R0
+ a2
∆R R0
2 .
(5.19)
T0 and R0 refer to ambient or room temperature (27◦ C), ∆R is the resistance change in going from room temperature to the hotplate operating temperature. The current through the temperature sensor as provided by the circuitry is constant. Therefore, the voltage can be replaced for the resistance. U0 is determined in the non-heated state of the microhotplate and refers to room temperature. By measuring the U0 -value, the spread in the geometry or dimensions of the temperature sensors is accounted for [551]. To determine the two temperature coefficients, a1 and a2 , a Pt-temperature element is introduced into the same package with the sensor in close proximity. The arrangement is then introduced in an oven and heated to a temperature of up to 325◦ C. The standard Pt-resistor serves as temperature reference in determining the two temperature coefficients of the integrated polysilicon temperature sensor. Extrapolations are not necessary in the target temperature range. The resulting coefficients are then used in the measurements to assess the hotplate temperature. The spread in the coefficients of polysilicon temperature sensors within one wafer lot is in the range of 2% [555]. There is no significant drift of U0 during analyte exposure of typically a few hours, but in longterm measurements polysilicon temperature resistors may considerably drift, which is the reason for using Pt-temperature resistors on the hotplate for better reliability (Fig. 5.46). After calibration, all temperature resistors are continuously read out during the measurements. The heating voltage is measured in a 4-point configuration. The power losses in the leads and in the bond wires are thus eliminated. 5.4.1.2 Thermal Microhotplate Modeling and Characterization A key issue in designing integrated monolithic microsystems is the simulation of the microhotplate-transducer that is connected to the circuitry to ensure full chip functionality after fabrication. This requires an adequate description
140
5 CMOS Platform Technology for Chemical Sensors
of the microhotplate and an implementation in a language that is applicable to circuitry simulations [557]. Thermal simulations of microhotplates have three main purposes: (a) gaining information about the thermal characteristics of a given microhotplate structure, such as temperature distribution, thermal resistance and thermal time constant, (b) facilitating its optimization procedure, and (c) providing input parameters for the overall microsystem simulation. The microhotplate simulation procedure includes several steps. First, the microhotplate design has to be converted into a geometry model for the finite-element (FEM) simulation with the aim to find a model that is as simple as possible but includes all relevant processes. The steps to arrive at the model are represented in Fig. 5.47, for details, see [557].
Fig. 5.47. Flow diagram for translating a microhotplate design into a geometry model for FEM-simulation using FEMLAB [557]
The picture on the bottom left-hand side shows the microhotplate schematic. The microhotplate exhibits a symmetric design so that a simulation of one quarter is adequate. Geometrical simplifications are introduced into the representation of the membrane layout and structure, for which a quasi 2-dimensional membrane model is developed. Afterwards, the membrane model is transferred back to a 3-dimensional description of the membrane with a constant thickness. All material and thermal properties of the different hotplate components and layers (electrodes, heaters etc.) are included in the model by then. Finally the membrane structure is combined with the supporting silicon cavity and the surrounding air. The result is a geometry model for the FEM-simulation [557]. The input parameters for a lumped-model description of the microhotplate are calculated from the simulation results. An analog-hardware-description-language (AHDL)-model,
5.4 CMOS Microhotplate System Development
141
∆T ∆Tsim
Fig. 5.48. Thermal resistance of a circular microhotplate as shown in Fig. 5.42b: Simulation and measurement results [557]
which then will be used for a circuitry simulation of the complete sensor system, can be derived from the lumped-model equations [552]. There are no additional fit parameters necessary to set up the model. The modeling results nicely coincide with experimental findings as can be seen in Fig. 5.48, which shows the temperature increase on the hotplate, ∆T , versus input power for a circular hotplate with silicon island underneath as depicted in Fig. 5.42b [555, 557]. Simulations were performed in 10-mW power steps, the results of which are plotted together with the mean value of the experimental data from three identical hotplates. The measured thermal resistance at room temperature is 5.8◦ C/mW ±0.2◦ C/mW. The thermal resistance fluctuations mainly result from variations in the etching process. The simulations yield a thermal resistance of 5.7◦ C/mW in the heating power range between 0 and 10 mW, which slightly deviates from the experimental results [557]. For higher temperatures, larger deviations are observed. A general trend is, that the simulated membrane temperatures are lower than the measured ones. For an input power of 60 mW that produces a temperature increase of approximately 300◦ C, the discrepancy between measured and simulated values is still less than 5%. To measure the thermal time constant of the microhotplate, the heater was driven with rectangular-shape heating pulses at a frequency of 10 Hz. The pulse amplitude was adjusted to produce a hotplate temperature of 300◦ C. The thermal time constant resulting from the modeling is 10.1 ± 1.2 ms [555, 557]. The experimentally determined time constant was 9.7 ± 0.2 ms and was measured during membrane cooling (for details see [551]). Measurements would have been distorted in case of assessing the thermal constant from the temperature increase of the membrane during heating up, since the heater resistance changes with temperature, i.e., initially a lot of power goes to the
142
5 CMOS Platform Technology for Chemical Sensors
heater and then falls off (power overshoot but no rectangular step function) [551, 557]. In the case of the temperature decay, only the thermal properties influence the time constant since heating immediately stops upon switching off the power. The modeling results again are in good agreement with the measurements (discrepancy of 5%), which signifies that the model is suitable for system-level simulations of microhotplate devices. The simulation of coated microhotplates is even more difficult owing to the fact that the material properties of the SnO2 thick-film layer are not well known, and its deposition is not as reproducible as the industrial CMOS process with the subsequent micromachining. The SnO2 -droplet has a typical height of 25 µm, and it was found that the temperature homogeneity in the heated area is improved by additional heat spreading through the thick metal oxide layer [555, 557]. Since the metal oxide coating is covering only the heated area in the membrane center (see Fig. 5.44), no additional dissipation paths will be created from the heated area to the bulk silicon, so that no significant changes in the thermal resistance have been found. The thermal time constant, however, considerably changes owing to the additional volume of the tin dioxide droplet. The time constant for a circular coated membrane (Fig. 5.42b) is 21 ± 1 ms, which is approximately 10 ms more than for the uncoated device [555, 557]. From the model an 11-ms increase upon metal oxide deposition was predicted. The thermal time constant of both, coated and uncoated hotplate also depends on the quality of the etching process. 5.4.1.3 Microhotplate Heaters: Resistor and Transistor Most conventional microhotplate heaters published to date include resistive heating elements [551]. The microhotplates shown so far (Figs. 5.42, 5.43, 5.44, 5.45, 5.46) all feature heating resistors that are arranged as ring heaters along the boundary of the central heated area of the hotplate. Driving a resistive heater on chip requires a power transistor (Fig. 5.49). A voltage drop across this transistor occurs, and a massive fraction of the consumed power is dissipated on the chip, which does not contribute to heating the membrane. Industrial CMOS-processes offer the advantage of using active heating elements by, e.g., realizing a MOS-transistor-type heater on the membrane. A transistor approach has been proposed in SOI-technology [560] but has not been realized in conventional CMOS-technology so far. The power transistor for driving the resistive heater is no more necessary, and all heat is produced directly on the membrane and contributes to microhotplate heating. Further advantages of a transistor heating-scheme include new heater control modes and the fact that the full supply voltage range can be directly applied to the heater. As will be detailed later (Sect. 5.4.2.3), the resulting novel hotplate type can be implemented in a smart sensor system with complex digital control circuitry.
5.4 CMOS Microhotplate System Development
(a) resistive heating
UDD
143
(b) transistor heating
Rsens (SnO2)
UDD
Rsens (SnO2)
heating transistor
Rheat Ucontrol power transistor
Ucontrol microhotplate
Fig. 5.49. Different microhotplate heating schemes: (a) resistive heater with power transistor [551] and (b) PMOS transistor heater [559]
A cross-sectional schematic of the transistor microhotplate is shown in Fig. 5.50, a micrograph in Fig. 5.51 [559]. Instead of a ring heater, there is a ring transistor, which has three components: source, drain and gate. The gate is polysilicon, the source and drain consist of the p-diffusion in the n-well (see also CMOS schematic, Fig. 3.3). The functioning of the transistor even at those high temperatures is a consequence of the electrical isolation of the n-well from the rest of the chip so that there are no leakage currents. thick-film SnO2 resistor PMOS ringtransistor heater
temperature sensor
gate
D bulk silicon (p-doped)
p-diffusion
dielectric layers (membrane)
S n-well (Si-island)
electrodes
Fig. 5.50. Cross-sectional schematic of a transistor-heated hotplate: D and S denote drain and source [559]
The n-well electrons and thermally generated free electrons are confined to the silicon island. The total size of the membrane is 500 by 500 µm2 (Fig. 5.51) [559]. The membrane consists of the dielectric layers of the CMOS process. The heated area includes an octagonally shaped n-well silicon-island
144
5 CMOS Platform Technology for Chemical Sensors
100 µm µ electrodes PMOS transistor heater ring Si-island (n-well)
temperature sensor
polysilicon heater (opt. annealing)
Fig. 5.51. Chip micrograph showing the membrane and the integrated PMOSheater [559]
of 300 µm base extension underneath the dielectric layers. The n-well is not only electrically isolated but also serves as a heat spreader. The PMOS transistor with 5 µm gate-length and 720 µm overall gate width (8 sections of 90 µm) is integrated in the island [559]. A special ring-shape transistor arrangement was chosen (Fig. 5.51), which leaves enough space to implement resistive temperature sensors in polysilicon. The sensor in the membrane center is used to determine the membrane temperature. An additional resistive heater is also integrated on the heated island. The temperature characteristics of both, resistor type heaters and transistor type heaters are shown in Fig. 5.52. The graph in Fig. 5.52a displays the center membrane temperature versus the applied heating power for two slightly different square-shape heated areas on 500 by 500 µm2 membranes [555]. The temperature characteristics are almost linear with increasing power. The nonlinearity also increases with increasing temperature. The maximum temperature is reached at supply voltages of 3.8 V (membrane 1) and 4.3 V (membrane 2). Figure 5.52b shows the measured membrane temperature versus the source-gate voltage, Ug , for different constant heating voltages, Usd (source-drain voltage) [559]. The heating characteristics are almost linear at temperatures higher than 100◦ C, which simplifies controlling the membrane temperature by a rail-to-rail gate voltage. The transistor can also be easily operated in the dynamic mode by modulating the gate voltage. From both graphs it is obvious, that the hotplate can be heated up to 350◦ C using a low-voltage power supply (5 V).
5.4 CMOS Microhotplate System Development 400
500
(a)
(b)
350
Usd = 5 V
membrane 1 membrane 2
T membrane [°C]
T membrane [°C]
400
300
200
100
300 Usd = 4.5 V
250 Usd = 4 V
200
Usd = 3.5 V
150
Usd = 3 V
100
Usd = 2.5 V
50 0
-20
145
0
20
40
60
80
100
0
power [mW]
Usd = 2 V
0
1
2
3
4
5
6
Ug [V]
Fig. 5.52. (a) Temperature versus power for two resistively heated membranes. The maximum temperature is reached at supply voltages of 3.8 V (membrane 1) and 4.3 V (membrane 2) [555]. (b) Measurement of the membrane temperature versus the source-gate voltage, Ug for different source-drain voltages Usd [559]
5.4.1.4 Microhotplate Sensor Fabrication The circuitry and the basic sensor elements (electrodes, heating resistors, temperature sensor) are fabricated in an industrial double-poly, double-metal 0.8-µm CMOS-process as provided by austriamicrosystems [496]. A series of post-CMOS processes were developed to complete the transducer fabrication (Fig. 5.53) [555]. Aluminum is the only metal available in the CMOS process after removing the CMOS passivation. The problem with Al-electrodes is, that they undergo strong oxidation during the annealing of the sensitive material, which results in poor or no electrical contact between the metal electrodes and the sensitive layer. This problem was resolved by deposition of an additional Pt-layer on top of the Al-metallization. The platinum ensures a good contact to the sensitive layer over a large temperature range. For patterning the platinum, a lift-off process on wafer-scale was developed (Fig. 5.53b–d) [555]. After deposition and the lithography of the photoresist, a 50 nm titanium-tungsten layer was sputtered onto the Almetallization. The titanium-tungsten layer serves as adhesion layer and acts also as a diffusion barrier. After deposition of the titanium-tungsten, a 100 nm-thick Pt layer is sputtered on top. Afterwards, the photoresist is removed. Optical inspection showed a good adhesion of the metal stack to the Al-metallization so that no cracks were observed. In a next step, a local silicon nitride passivation is applied and patterned with the aim to protect all contacts and metal structures with the exception of the measuring electrodes (Fig. 5.53e, f) [555].
146
5 CMOS Platform Technology for Chemical Sensors
Fig. 5.53. CMOS post-processing sequence to fabricate metal-oxide-covered microhotplates: Pt-electrode fabrication (a–d), local passivation (e, f ) hotplate formation (g), and coating deposition (h) [555]
For the release of the microhotplate membrane, the wafers with the Ptelectrodes were back-side etched in aqueous 6-molar potassium hydroxide solution (KOH) at 90◦ C using an electrochemical etch stop technique (ECE) [561]. The etching stops at the field-oxide and at the silicon n-well, which forms a silicon island of 5.5 µm final thickness in the heated area (Fig. 5.53g) [555]. After the post-processing the chips are diced. The sensing layers can be categorized according to the deposition method on the micromachined substrate. One set of method relates to thin-film technology and includes the direct deposition by means of conventional microtechnological processes such as chemical-vapor deposition (CVD), sputtering or evaporation. A general difficulty is the precise patterning of these materials at micrometer resolution (see Figs. 5.43, 5.44). Lift-off techniques are applicable for a variety of metal oxides, but impose limits on the deposition temperature [549, 562].
5.4 CMOS Microhotplate System Development
147
In CVD-processes the microhotplate is heated, which leads to decomposition of the gaseous precursor material and locally defined deposition of a thin-film layer in the heated area [563]. The hotplate temperature heavily influences the film morphology resulting from CVD processes, which also holds for the case of sputtering. Masklessly sputtered films do not have to be patterned according to the authors of [564], since the resistance of the nonheated material is very high, and, therefore, gas-sensitive effects on the conductivity will only occur in the heated area, i.e., on the microhotplate. Thermolithographic patterning of sol-gel materials constitutes an alternative, which was reported on recently [565]. Another possibility includes rheotaxial growth and thermal oxidation of tin layers (RGTO): A metallic tin layer is deposited and subsequentially oxidized. The layer is patterned either by lift-off or by using a shadow mask [566, 567]. The result is a locally defined nanocrystalline thin film layer. The other category of deposition methods on microhoplates relies on the initial production of a metal-oxide powder. This powder is mixed with water and/or organic solvents to produce a paste, which is deposited onto the substrates. The deposition can either be done by spin-coating in combination with etching or by thermolithographic patterning through membrane heating [568, 569]. There are also methods relying on pulverization deposition [570]. All these deposition technologies lead to nanocrystalline metaloxide layers in the micrometer and submicrometer layer-thickness range. Even thicker layers (tens of microns) are fabricated by drop-coating, screen-printing or fluidic paste deposition [450, 547, 571, 572], in which cases post-structuring is not necessary. For the CMOS hotplates it was important to select a metal-oxide deposition method, which is compatible with the CMOS-substrate, i.e., does not exceed the back-end temperature of the CMOS process of 400◦ C in order not to change the transistor characteristics of the integrated electronics. As previously mentioned, a “burn-in heater” is provided in most hotplate designs to achieve higher temperatures (up to 500◦ C) locally on the hotplate if necessary. Since nanocrystalline metal oxide materials with defined grain sizes that have been deposited from slurries of preprocessed powders in thick-film technology provide high gas sensitivity and comparably high longterm stability, this approach was chosen. The metal-oxide thick film is deposited on the electrodes by using a drop coating method and a paste with preprocessed metal-oxide powder exhibiting defined grain sizes in the range of 10 to 20 nm [399, 547, 573, 574]. The oxide material is a Pd-doped nanocrystalline SnO2 [574], the principal gas detection mechanism of which has been detailed in Sect. 4.4.3.1.2. The whole chip (including circuitry) is heated in a belt oven for 20 min at 400◦ C to anneal the oxide material, which is enough to achieve satisfactory sintering without any reduction of the material gas sensitivity. The high-temperature annealing did not produce any cracks in the membrane
148
5 CMOS Platform Technology for Chemical Sensors
or in the tin-oxide material, as was assessed by optical inspection (see Fig. 5.44). The sensitive layer adheres well to the electrodes and the passivation and features a typical thickness of 25 µm in the center of the tin oxide drop. A degradation of or change in the circuitry performance was not observed. The thick-film preparation of the sensitive layer is fully CMOS-compatible, and the overall post-CMOS sensor processing sequence can be carried out at wafer-level. 5.4.2 Hotplate-Based CMOS Monolithic Microsystems Within this section, three subsequent development steps or different generations of a monolithic hotplate-based microsystem – starting with a simple analog system, proceeding to an already complex analog/digital system and ending with a three-hotplate digital array – will be described. Each system will be detailed in architecture and design, and its performance in electrical test measurements (functional tests of the circuitry units) and upon gas exposure to selected gases, in most cases, carbon monoxide (CO) or methane (CH4 ), will be reported on. 5.4.2.1 Analog Hotplate Microsystem The first analog monolithic sensor system combines, on a single chip, a microhotplate gas sensor platform along with circuitry for controlling the temperature and for sensor signal amplification and conditioning [551]. The circuitry also provides an electronic interface to the outside world. The membrane, which isolates the heated area from the bulk chip, is formed by the CMOS dielectric layers (Si-nitride, Si-oxide). The circuitry is arranged on the bulk chip, the temperature of which negligibly varies upon hotplate heating as will be shown. The block diagram in Fig. 5.54 schematically shows the chip architecture [551, 552]. An embedded analog controller regulates the membrane temperature. A polysilicon resistor serves as temperature sensor, and the voltage drop over the resistor provides the feedback signal for the temperature controller. The bulk chip temperature is monitored by another temperature sensor in close vicinity to the circuitry (see also Figs. 5.55 and 5.57). A logarithmic converter is implemented and is used to perform signal conditioning of the conductometric sensor signal [552]. Figure 5.55 shows a micrograph of the chip. The left part of the micrograph exhibits the hotplate membrane, whereas the right part shows the different circuitry components. The sensor system was fabricated in a double-poly, double-metal 0.8 µm CMOS process as provided by austriamicrosystems [496] with subsequent micromachining steps. The circuitry can be grouped in three functional units [551,552]: (i) Membrane temperature control loop, (ii) bulk-chip temperature measurement, and
5.4 CMOS Microhotplate System Development
149
membrane
SnO2 sensor
temp. sensor
membrane temperature
measurement circuitry
poly-Si temp. sensor
poly-Si heater
SnO2 resistance
log converter
driving circuitry
proportional controller
measurement circuitry
chip temperature
control voltage
Fig. 5.54. Block diagram of the monolithic analog-type hotplate microsystem and its components [551]
500 µm
membrane
chip temperature sensor
linear-to-log converter
proportional controller
Fig. 5.55. Micrograph of the analog monolithic microsystem showing the hotplate on the left side and the circuitry components on the right [551, 552]
(iii) sensitive-layer resistance measurement. The chip is supplied by means of an external current source, which is used for biasing the membrane temperature sensor, the reference current of the logarithmic converter and the bulk-chip temperature measurement circuitry. Temperature Control The proportional temperature controller is implemented using an operational amplifier and an internal stabilization capacitor of 8 pF [552,558]. The
150
5 CMOS Platform Technology for Chemical Sensors
membrane temperature is controlled from room temperature up to 350◦ C. The operational amplifier drives a power transistor, which provides the current to the polysilicon heater. The inputs of the operational amplifier consist of the control voltage and the voltage drop on the polysilicon temperature sensor, which provides the feedback signal for the temperature controller. The polysilicon temperature sensor is biased with a temperature-independent current [552]. The ageing effects in the polysilicon temperature sensor cannot be compensated by the electronics on chip, the circuitry will, however, allow for precise control of the preset temperature. The steady-state error of the temperature controller as measured in the system operation temperature range (ambient temperature between −40◦ and 120◦ C) is less than 1% of the membrane temperature. Bulk-Chip Temperature Measurement The bulk-chip temperature is assessed via a PTAT (proportional to ambient temperature) temperature sensor, which makes use of the base-emitter voltage difference between a pair of diode-connected vertical pnp transistors (parasitic transistors available in the CMOS process, collectors tied to substrate) working at different current densities [552]. Metal-Oxide Resistance The resistance of the SnO2 sensitive layer can vary over a wide range (up to six decades) and is, therefore, measured using a logarithmic converter in a range from 1 kΩ to 10 MΩ (see Fig. 5.58b later in this section). The range of the logarithmic converter can be changed with the reference current (IREF ). The bulk chip temperature is used to compensate for the temperature dependence of the logarithmic converter. For more details on the circuitry and the circuitry schematics, see [552, 558]. System Characterization The tracking mode and stabilization mode performance of the temperature controller was assessed. The tracking mode results are displayed in Fig. 5.56a [551, 552]. At measurement start, the chip was at an ambient temperature of 27◦ C, then, the control voltage was increased in steps of 10 mV to achieve membrane heating. A control voltage of 1.80 V led, e.g., to a membrane temperature of approximately 350◦ C. The measured slope of the almost perfectly linear temperature-versus-control-voltage graph was 0.63◦ C/mV, and the tracking error due to noise was less than ± 0.3◦ C. The performance of the temperature controller in the stabilization mode is shown in Fig. 5.56b [551,552]. The ambient temperature was swept from −40◦ to 120◦ C in steps of 5◦ C, and a constant control voltage of, e.g., 1.77 V was applied to the membrane heater, which produced a membrane temperature of 331◦ C. The controller showed excellent temperature stabilization of the membrane in the investigated temperature interval with less than 1% error. The error due to noise was approximately ± 0.3◦ C.
5.4 CMOS Microhotplate System Development
(b) stabilization mode membrane temperature [°C]
membrane temperature [°C]
(a) tracking mode 400 300 200 100 0 1.3
1.5 1.7 control voltage [V]
151
1.9
333 332 331 330 329 328 -40
0 40 80 ambient temperature [°C]
120
Fig. 5.56. Performance of the on-chip temperature controller in (a) the tracking mode, and (b) the stabilization mode [551,552]. A membrane temperature of 331◦ C was preset in the stabilization mode
The dynamic behavior of the temperature controller was measured by applying a fast temperature change on the membrane from 27 to 300◦ C. The temperature controller was very stable and did not show any overshoot or ringing [552]. The performance of the temperature sensor on the bulk chip in vicinity to the circuitry was also assessed. The ambient temperature was swept from −40 to 120◦ C in steps of 5◦ C, and the control voltage was set to 0 V, i.e., the membrane was not heated at all. A two-point calibration at temperatures of −20◦ and 85◦ C was performed. The measured temperature sensitivity was about 128 µV/◦ C, and the error due to noise was less than ±1.5◦ C [551]. An important issue concerns the hotplate-induced chip heating, since the circuitry may not be overheated and its proper functioning is only specified and guaranteed in a temperature range from −40◦ to 120◦ C. Therefore, the temperature variation off-membrane on the bulk chip with respect to ambient temperature upon heating the membrane was measured as displayed in Fig. 5.57 [551]. The chip was kept at 27◦ C (ambient) in a ceramic dual-in-line (DIL) package, and the control voltage was increased in steps of 10 mV, which effectuated a heating of the membrane from 27◦ to 360◦ C. The maximum temperature increase on the chip was less than 6◦ C over ambient temperature at a membrane temperature of 350◦ C [551]. The chip temperature increase can be further reduced by an optimized package. Gas Tests Gas test measurements were performed in a gas manifold. Vapors were generated from gas bottles with calibrated target gas concentrations (e.g., CO) and then diluted as desired using computer-driven mass-flow controllers and synthetic air (oxygen/nitrogen mixture without humidity) as carrier gas. Typical experiments consisted of alternating exposures to pure synthetic air and
152
5 CMOS Platform Technology for Chemical Sensors
(TCHIP – TAMBIENT) [°C]
6
TCHIP
hotplate
package
5 4 3 2 1 0 0
100
200
300
400
membrane temperature [°C]
Fig. 5.57. Chip heating: Difference between chip temperature and ambient temperature as a function of the membrane temperature (ceramic DIL package) [551]
contaminated air. Exposure times of 30 minutes were followed by 30 minutes purging the chamber with synthetic air. In a first test series, the sensor resistance was read out by a multi-meter and was plotted on a logarithmic scale. Figure 5.58a shows the sensor response (R/R0 : ratio of resistance upon gas exposure to that in pure air) upon exposure to different CO concentrations at varying hotplate operation temperature and different relative humidity levels [551]. A nonlinear response with increasing CO concentration is observed as it is well established in literature for this type of semiconducting metal oxide (see, e.g. [399, 547]). It is also evident from Fig. 5.58a, that the sensor response is depending on the relative humidity content of the test gas atmosphere: Higher humidity levels (a)
(b) 0.08
carbon monoxide (CO) 10
CO (40% r.h., Tmem: 290°C)
Ulog [V]
R/R0
50%, 250°C 50%, 300°C 50%, 350°C
1 0
20
40
60
0%, 250°C 0%, 300°C 0%, 350°C
80
CO concentration [ppm]
100
15 ppm
10 ppm
0.075 5 ppm
0.07
0.065 0.06
0
50
100
150
200
time [min]
Fig. 5.58. (a) Sensor response (R/R0 : ratio of resistance upon gas exposure to that in pure air) upon exposure to different CO concentrations at varying hotplate operation temperature and different relative humidity levels. (b) Raw sensor signals coming from the logarithmic converter upon CO-dosing at 40% relative humidity and 290◦ C operation temperature [551]
5.4 CMOS Microhotplate System Development
153
generally lead to lower tin oxide resistance, i.e., increase the height of the sensor signal in Fig. 5.58a [551]. This is a consequence of the increased number of CO reaction sites at the surface of the sensitive material owing to the formation of hydroxy1 groups by adsorbed water [547]. It is also established that the optimum operation temperature for Pd-doped SnO2 -thick-film sensors is in the range of 240◦ –280◦ C for CO detection [573, 574]. An exemplary output of the logarithmic converter is shown in Fig. 5.58b [551]. The difference voltage (unfiltered raw data) at the converter is plotted for three different CO concentrations measured at 40% relative humidity and 290◦ C operation temperature. The signals almost linearly increase with increasing concentration as is to be expected after logarithmic conversion. From the conducted measurements (Figs. 5.58a,b) it can be concluded, that concentrations of even less than 1 ppm of CO are detectable with monolithic sensor systems. 5.4.2.2 Analog/Digital Hotplate Microsystem A simplified schematic of the analog/digital chip architecture is shown in Fig. 5.59 [553, 554]. The microhotplate and its components, which were discussed in the previous paragraph, are represented as a block. The circuitry includes the same three major components as in the analog system (Sect. 5.4.2.1): (a) The metal oxide resistance readout, (b) the microhotplate temperature control loop, and (c) the temperature readout of the bulk chip carrying the electronics [554].
circular membrane SnO2 sensor
linear-to-log converter & buffer
anti-aliasing filter & A/D converter
programmable offset comp.
temp. sensor
buffer & programmable offset comp.
anti-aliasing filter & A/D converter
poly-Si heater
square root circuitry & driving circuitry
D/A converter & Buffer
measurement circuitry & buffer
anti-aliasing filter & A/D converter
temp. sensor
digital controller & digital interface
I 2C bus
programmable offset comp.
Fig. 5.59. Detailed block schematic of the analog/digital hotplate microsystem including all circuitry units [553]
154
5 CMOS Platform Technology for Chemical Sensors
The resistance of the tin-dioxide layer is, again, read out by a logarithmic converter. The temperature controller is now a digital programmable PID (Proportional-Integral-Differential) controller, which precisely controls the membrane temperature via the analog heater driving circuitry. The A/D and D/A converters that connect the analog and digital part of the circuitry are also shown in the block diagram. All measured data are read out via a standard serial inter-integrated-circuit (I2 C) bus interface [575]. The digital interface also allows for setting the membrane temperature and the controller parameters. Each major circuitry unit and its operation principle will be briefly abstracted. The detailed circuitry implementation is published elsewhere [557, 558]. Metal-Oxide Resistance Measurement For the analog logarithmic converter, the relation between the differential output voltage, US , and the resistance of the sensing layer, RS , can be expressed as kTC · (ln RS + c0 ) (5.20) e where k denotes the Boltzmann constant, TC the absolute chip temperature, e the elementary charge, and c0 a characteristic parameter in the circuit design. The negative sign is a consequence of the polarity of the differential output voltage: Lowering the sensor resistance will lead to an increased US . US = −
Hotplate Temperature Measurement The polysilicon temperature sensor in the center of the microhotplate is connected to a constant-current source. The voltage drop across the temperature sensor, UT , is monitored. The corresponding voltage drop, U0 , at defined ambient temperature, T0 , is initially assessed in calibration measurements. Since the current through the temperature sensor is constant, the following equation holds [553]: UT − U0 RT − R0 = . U0 R0
(5.21)
The microhotplate temperature, T , can be calculated by: T − T0 = a1
∆R R0
+ a2
∆R R0
2 (5.22)
where a1 and a2 are the temperature coefficients of the temperature sensors (compare 5.19). These coefficients have to be determined in a calibration procedure, which has been explained in the context of (5.19) in Sect. 5.4.1.1. The current through the power transistor of the heater driving circuitry will be adjusted by the digital controller so that UT corresponds to a preset reference voltage, UR . Once the calibration has been done, reversing (5.22) and replacing UR for UT in (5.21) yields the value of UR to produce a desired
5.4 CMOS Microhotplate System Development
155
membrane temperature, T . The controller additionally takes advantage of an analog square-root circuitry, which linearizes the correlation between control voltage and dissipated power [556]. This linearization facilitates the calculation of optimum parameters for the temperature controller. Chip Temperature Measurement The bulk-chip temperature sensor is not realized as a resistor. The baseemitter voltage difference between a pair of diode-connected vertical pnptransistors (available in the CMOS process as parasitic transistors) is related to the chip temperature. The relation between the absolute chip temperature, TC , and the differential voltage, UC , is given by: UC = c1
kTC + Uoff . e
(5.23)
The factor c1 depends on the transistor design and its biasing current, and Uoff is an offset voltage. Both are constants and can be assessed during a two-point calibration procedure. The chip temperature sensor does not only measure the changes in the ambient temperature, it also can be used to compensate for the temperature dependence of the output of the on-chip logarithmic converter according to (5.20). The analog output signals of the logarithmic converter, US , the microhotplate, UT , and the bulk-chip temperature sensor, UC , are converted into the digital domain, and then can be read out via the digital I2 C interface. The input value to the chip is a digital number representing the reference voltage, UR . Figure 5.60 shows the sensor chip [554,556]. The microhotplate is located in the left part of the chip with enough free space to accommodate additional hotplates. The analog circuitry as well as the A/D and D/A converters are clearly separated and shielded from the digital circuitry for noise reasons [556]. The bulk-chip temperature sensor is located close to the analog circuitry in the center of the chip. The distance between microhotplate and circuitry is comparatively large owing to packaging requirements, as will be explained in the packaging section (Sect. 5.6). System Characterization The tracking mode performance of the digital temperature controller is shown in Fig. 5.61 [553, 556]. The microhotplate temperature is plotted versus the digital code of the reference voltage. The 10-bit input range from 0 to 1023 corresponds to a microhotplate temperature range between 170◦ C and 310◦ C, so that one digit represents 0.1◦ C. The tracking error due to noise is less than ±2◦ C [556]. The curve is almost linear with the slight nonlinearity being included in the second-order coefficient of the polysilicon temperature sensor (see (5.22)). The controllable temperature range is variable by adjusting the slope of the temperature function in Fig. 5.61. The temperature offset, i.e., the minimal temperature that corresponds to the digital value 0, is also programmable. Such a device with on-chip controller offers several advantages in
156
5 CMOS Platform Technology for Chemical Sensors chip size: 6.7 x 4.8 mm2
500 µm
hotplate temperature sensor
analog circuitry, A/D & D/A converters
digital circuitry: controller, interface
Fig. 5.60. Micrograph of the analog/digital microsystem chip showing the microhotplate (left), the analog circuitry (center ), and the digital unit with serial interface (right) [553, 556] 350
TM [°C]
300
250
200
150 0
200
400 600 UR [digital code]
800
1000
Fig. 5.61. Tracking mode performance of the digital (PID) controller. One digital unit represents 0.1◦ C [556]
comparison to conventional microhotplate-based sensors. The approximately linear relationship between digital input code and microhotplate temperature directly translates a sinusoidal input variation into a sinusoidal temperature variation. This does not hold for a conventional microhotplate with an uncontrolled resistive heating element upon applying a sinusoidal heating-voltage or -current variation.
5.4 CMOS Microhotplate System Development
157
Gas-flow fluctuations or temperature changes owing to chemical reactions also contribute to the overall heat budget of uncontrolled microhotplates. These effects lead to a temperature variation even though a constant heating voltage or heating power is applied. These temperature variations inevitably produce changes in the sensor signal, since the metal-oxide resistance significantly changes with temperature (compare Figs. 5.58a, 5.63). In an integrated microsystem, however, temperature fluctuations are immediately counteracted by the on-chip control circuitry, which keeps a preset temperature constant regardless of the power consumption. The resistance changes in the sensitive layer upon exposure to the analyte can be distinguished from those resulting from ambient temperature changes through the additional onchip temperature sensor, and by monitoring the power consumption of the hotplate heater [553]. An important issue of monolithic sensor systems with integrated microhotplates is the overall chip heating. The temperature increase on the chip at a microhotplate temperature of, e.g., 300◦ C was measured with the on-chip temperature sensor and amounts to 3◦ C, which corresponds to 1% of the hotplate temperature [553]. Such slight temperature increases, which can be further minimized by applying suitable packaging methods, such as the use of die-attach materials with high thermal conductivity, will definitely not affect the on-chip circuitry. Gas Tests A series of chemical measurements was performed in order to test the sensor performance. The microhotplate was heated to a defined temperature, and the sensor signal upon carbon-monoxide exposure was recorded. The raw data (digital numbers in the 10-bit output range) were filtered with a movingaverage filter averaging over 100 data points. Figure 5.62 shows a typical sensor response [553, 554]. The sensor signal, S, is expressed in digital units after the conversion of US (see 5.20) by the on-chip A/D converter. The sensor signal output range from 200 to 280 digital units corresponds to a resistance range of 250 to 50 kΩ. As pointed out earlier, a higher sensor signal reading represents a lower resistance of the sensitive layer. Consequently, the resistance drop upon the presence of CO, a reducing gas, leads to a higher sensor reading. Each analyte exposure step is 15 min at constant relative humidity of 40% (23◦ C humidifier temperature) and 30◦ C chip or ambient temperature. The microhotplate operating temperature was 275◦ C. Low COconcentrations of 1, 3 and 5 ppm (partial pressure 0.1–0.5 Pa CO) were dosed to the sensor. As can be seen in Fig. 5.62, a concentration of 1 ppm CO is clearly detectable. The noise in the sensor signal is ±0.5 digits [553]. This corresponds to a 1-digit quantization noise of the A/D converter. At low concentration levels, the noise is equivalent to a CO-concentration variation of ±0.1 ppm. This value serves as input for the determination of the limit of detection, which was assessed to be to 0.2 ppm. The gas concentration resolution also amounts to ±0.2 ppm [553].
158
5 CMOS Platform Technology for Chemical Sensors 275
S [digital code]
CO
5 ppm
5 ppm
3 ppm
250
3 ppm
1 ppm
1 ppm
225
200
0
50
100
150
200
time [min] Fig. 5.62. Low-concentration signals: 1–5 pm (0.1–0.5 Pa) CO detected with Pddoped (0.2%) tin dioxide at 275◦ C [553]
The sensor baseline, Sair , represents the sensor reading in humidified synthetic air without any analyte present. The difference between this baseline value before the analyte-dosing onset and the sensor signal, S, before the analyte-dosing end, was used to determine the sensor response at a given concentration. The full digital output range of S from 0 to 1023 covers sensor resistances from 10 MΩ to 1 kΩ. Taking into account the conversion of US to the sensor readout value, S, and introducing (5.20), ∆S can be expressed as [553]: ∆S = S − Sair = −(cs · ln R − cs · ln Rair ) = cs · ln
Rair . R
(5.24)
The proportional constant, cs , includes the conversion relation of the A/D converter and the factor kTC /e of 5.20. Since the measurement chamber is temperature-stabilized, the chip temperature, TC , is constant. Rair is the resistance of the sensitive layer in synthetic air, and R is the resistance upon analyte exposure. So (5.24) correlates ∆S to the ratio Rair /R, which is commonly plotted as sensor signal. The sensor responses at defined relative humidity of 40% (23◦ C humidifier temperature) are displayed in Fig. 5.63 for various microhotplate temperatures [553]. ∆S upon exposure to different CO concentrations increases with decreasing microhotplate temperature. The highest sensitivity was achieved for the lowest operation temperature of 225◦ C. The sensor signal is strongly temperature-dependent. The temperature dependence of ∆S in the temperature range between 250◦ C and 300◦ C is 0.15 digits per ◦ C at a CO concentration of 5 ppm [553]. A temperature change of 1◦ C at 5 ppm CO and a microhotplate temperature of 275◦ C thus produces the same signal as a CO concentration change of 0.04 ppm. As already mentioned in the previous section on electrical system parameters, the resolution of the temperature
5.4 CMOS Microhotplate System Development
159
120 ∆S 225°C ∆S 250°C ∆S 275°C ∆S 300°C
∆S [digital code]
100 80 60 40 20 0
0
5 10 CO concentration [ppm]
15
Fig. 5.63. Sensor responses, ∆S, upon different CO concentrations in dependence of the hotplate temperature (constant relative humidity of 40% at 23◦ C humidifier temperature) [553]
controller was estimated to be ±2◦ C [556]. These 2◦ C correspond to a “COconcentration uncertainty” of ±0.1 ppm, which is less than the resolution in the concentration measurements. Consequently, the temperature uncertainty in the controller does not significantly affect the sensor limit of detection. It is evident from these considerations that a precise temperature control is absolutely indispensable to advance into low-concentration and threshold-level measurements. 5.4.2.3 Digital Hotplate Array Microsystem The microsystem features three microhotplates that are monolithically integrated with digital temperature controllers, readout and interface circuitry [576]. The microhotplates are heated by means of MOS-transistors (see also Sect. 5.4.1.3) and are covered with nanocrystalline SnO2 -based coatings as sensitive layers. Full advantage is taken of the features offered by applying CMOS-technology. All sensor values can be set and read out via the digital interface, which drastically reduces the packaging efforts, since the number of bond wires is the same as for a single microhotplate. The block diagram in Fig. 5.64 depicts the system architecture, emphasizing its strong modularity [576]. Three digital PID (proportional-integralderivative) controllers provide independent temperature regulation for each hotplate. The MOS heating transistor is driven in pulse-density modulation by a first-order Sigma-Delta modulator [576]. Analog circuitry is required
5 CMOS Platform Technology for Chemical Sensors I2C interface
control unit
timer value
3 microhotplates
temp. controller
Σ∆
hotplate
readout
MUX MUX
160
multiplier value
Fig. 5.64. Block diagram showing the architecture of the digital hotplate array chip [576]
only for the readout of the metal oxide resistors and the temperature sensors. A single input-conversion stage and multiplier are accessed from the three controllers in time-sharing. The PID parameters, the target temperatures, and the operation timing can be all programmed by the user via a standard serial interface [577], which also enables digital readout of the sensor values. Figure 5.65 shows the fabricated chip featuring a tri-sectional floor plan with digital circuitry on the right side including the temperature controller and the interface, analog circuitry in the center and the three micro-hotplates on the left-hand side [576]. Along the left edge, three different coatings are shown: Pure SnO2 , which exhibits high partial sensitivity to NO2 , SnO2 with 0.2% palladium, which is used to monitor carbon monoxide, and SnO2 with 3% palladium, which can be used to detect hydrocarbons such as methane. The highly Pd-doped coating appears gray as a consequence of the metal particles. An array including those three coatings is suitable for many target applications. Microsystem Characterization Closed-loop temperature regulation was tested in the tracking-mode: The temperature was sinusoidally varied on one of the hotplates (HP3), while the other two hotplates (HP1, HP2) were kept at constant temperature. The test was performed for each of the hotplates with the remaining two hotplates kept at constant temperature (Fig. 5.66) [577].
5.4 CMOS Microhotplate System Development
A/D & D/A converters
161
digital circuitry: temperature controller & interface
500 µm
pure SnO2
SnO2, 0.2% Pd
SnO2, 3% Pd
chip size: 5.5 x 4.5 mm2
microhotplates
Fig. 5.65. Micrograph of the chip with microhotplate array and circuitry [576]
temperature [°C]
350
temperature modulation HP 1
300
HP 2 250 HP 3 200 0
10 time [min]
20
Fig. 5.66. Sinusoidal temperature variation of one selected microhotplate (HP3). The other two hotplates (HP1, HP2) are kept at constant temperatures of 280◦ C and 300◦ C [577]
The tracking error due to noise was ±1◦ C. In the temperature range of interest, no thermal crosstalk is noticeable, as it is to be expected using individual temperature control loops for each hotplate. The results also show good linearity, with a differential non-linearity less than 1/2 least significant bit (LSB) and an integral non-linearity less than 3 LSB [577].
162
5 CMOS Platform Technology for Chemical Sensors
Chemical Measurements First chemical test measurements have been conducted with the array chip. Fig. 5.67 shows the results that have been simultaneously obtained from three microhotplates coated with different materials at operation temperatures of 280◦ C and 300◦ C in dry air [577]. Hotplate HP1 is covered with a Pd-doped SnO2 layer (0.2 wt% Pd), which is optimized for CO-detection, whereas the sensitive layer on hotplate HP3 contains 3 wt% Pd, which renders this material more responsive to CH4 . The material on hotplate HP2 is undoped pure tin dioxide, which is sensitive to NO2 . CH4[ppm] 500
1000
CO [ppm] 1000
1500
5
10
15
HP1: 0.2% Pd 280°C 50 45 800 40 700 35
600
30
500
HP3: 3% Pd 330°C
sensor resistance [kΩ]
sensor resistance [kΩ]
900
25
400 HP2: undoped 280°C
20
300 0
0.5
1.5
2.0
2.5
3.0
3.5
time [h]
Fig. 5.67. Sensor response of three microhotplates with different sensitive layers upon exposure to CO and CH4 [577]. The lightly doped tin dioxide (0.2% Pd) shows little response to methane and large responses to CO. The heavily doped metal oxide (3% Pd) responds to both gases, whereas the undoped tin dioxide material shows hardly any signal upon CO and CH4 exposure
As is evident from Fig. 5.67, the responses of the three materials to the test gases CO and CH4 are very different. Hotplate HP1 shows larger responses to CO, whereas hotplate HP3 also responds to hydrocarbons, in this case, methane. Hotplate HP2 shows hardly any signal. The single-chip array thus provides multiple inputs to identification and quantification of gases, which is very important in the presence of interferants or in analyzing mixtures. An array of microhotplates with individually controlled temperatures, the hotplates of which are covered with different sensitive materials, drastically increases the overall information that can be extracted from such sensors and, then, can be used as input for suitable data deconvolution tools (multicomponent analysis, pattern recognition) [14–17].
5.5 CMOS Chemical Multisensor Systems
163
5.5 CMOS Chemical Multisensor Systems Combinations of different chemical sensor/transducer principles and –types can provide more information than using arrays of the same transducer type and varying the sensitive layer only. The single-type array, an example of which has been shown in Sect. 5.4.2.3, has become particularly popular, as is documented in a wealth of publications [12–17]. Simultaneous use of various transducer principles is more complex (realization of different transducer and circuitry components, use of different data evaluation schemes), but, nevertheless, has proven to be beneficial in several applications [578–588]. In this introductory section also some non-CMOS but semiconductor-based systems will be briefly sketched to illustrate the power of the multi-transducer approach. The range of concentrations to be monitored can be extended by, e.g., applying different electrochemical sensor types: An integrated hydrogen system consisting of aluminum-gate FETs for low concentrations, a hydrogen sensing chemoresistor made of palladium/nickel for high concentrations and the necessary circuitry has been reported in [578,579]. Different transducers can also provide distinct and complementary information on the analyte or analyte mixtures in complex samples to be monitored. A sensor system including several ISFETs and an amperometric free-chlorine sensor for operation in water has been reported in [580]. Disposable electrochemical multisensor systems for fast blood analysis are marketed by, e.g., I-STAT [581]. Sodium, potassium, chloride, ionized calcium, pH and carbon dioxide are measured by ion-selective-electrode potentiometry (see Sect. 4.4.2.1). Concentrations are calculated from the measured potential through the Nernst equation (2.15). Urea is first hydrolyzed to ammonium ions in a reaction catalyzed by the enzyme urease. The ammonium ions are also monitored by means of an ion-selective electrode. Glucose is measured amperometrically (see Sect. 4.4.1). Oxidation of glucose, catalyzed by the enzyme glucose oxidase, produces hydrogen peroxide. The liberated hydrogen peroxide is oxidized at an electrode to produce an electric current, the intensity of which is proportional to the glucose concentration. Oxygen is also measured amperometrically. The oxygen sensor is similar to a conventional Clark-electrode. Oxygen permeates through a gas permeable membrane from the blood sample into an internal electrolyte solution where it is reduced at the cathode. The oxygen reduction current is proportional to the dissolved oxygen concentration. Hematocrit is determined conductometrically (see Sect. 4.4.3). The measured conductivity, after correction for electrolyte concentration, is related to the hematocrit. A significant benefit of using CMOS-technology for devising chemical sensors or sensor systems is the possibility to co-integrate several different transducers along with all necessary driving circuitry on a single chip. Additional components that can be integrated include signal-conditioning circuitry (amplifiers, references), multiplexers to reduce the number of output
164
5 CMOS Platform Technology for Chemical Sensors
pins, analog/digital and digital/analog converters, chip memory (calibration values) or other smart features, an intra-chip communication or bus system and a serial interface to communicate with off-chip microcontrollers or instruments. In the following part of Sect. 5.5, two prototype monolithic CMOS multisensor systems will be presented, (i) a multiparameter biochemical sensor [582] and (ii) a smart gas sensor microsystem [583]. 5.5.1 CMOS Multiparameter Biochemical Microsystem The biochemical microsensor system is aimed at continuous monitoring of ions, dissolved gases and biomolecules in liquid phase such as blood (Fig. 5.68) [582, 584], and is based on an earlier design by Gumbrecht et al. [585, 586]. The eight integrated chemical sensors comprise six ion-sensitive field-effect transistors (ISFETs: 1–6 in Fig. 5.68), one oxygen sensor (7 in Fig. 5.68) and one conductometric sensor (8a,b in Fig. 5.68), all of which can be operated in parallel [582]. An Ag/AgCl reference electrode is also integrated on the CMOS chip to get rid of external references. The eight sensors can continuously monitor ions, dissolved gases and biomolecules via enzymatic reactions that produce charged particles. A flow channel (polyimide) restricts the liquid phase access to the sensor area. The six ISFETs allow for direct contact of the electrolyte with the gate oxide. The gate oxide itself is either pH-sensitive (see ISFET Sect. 4.4.2.2.1), conductometric circuit
potentiostatic circuit
heater
driver
4 8a
EPROM
1
2
ISFETs
3
8b
5
6
7
8a
multiplexer and counter
Fig. 5.68. Micrograph of the CMOS multi-parameter biochemical sensor chip, which includes 6 ISFETS (1–6), an (amperometric) oxygen sensor (7) and a conductometric sensor (8a,b). The on-chip circuitry includes an EPROM, a multiplexer and counter, a driver unit, a conductometric and potentiostatic circuit and a heater. Reprinted with permission from [582]
5.5 CMOS Chemical Multisensor Systems
165
or the ISFET can be used as a “Severinghaus”-type pH-FET to measure dissolved carbon dioxide (detection of carbon dioxide via dissolution in water, formation of “carbonic acid” and monitoring of the pH-change). The gate oxide can also be covered with different ion-selective membranes to achieve sensitivity to a range of target ions such as potassium. All six ISFETs or only a subset can be used. The idea was to make a standard chip to reduce manufacturing costs and then modify the chip with selective coatings according to user needs. The integrated amperometric sensor (see Sect. 4.4.1) can be used as a Clark -type oxygen sensor, which is based on a two-step-reduction of gaseous oxygen in aqueous solution via hydrogen peroxide to hydroxyl ions. The conductometric sensor (see Sect. 4.4.3) consists of two parallel sensors (8a), which share one common electrode (8b). A sinusoidal AC potential is applied to the electrodes, and the current, which depends on the solution composition (concentration of charged particles or ions) is recorded. The full system is produced in a 1.2-µm single-metal, single-poly CMOS process, and the chip size is 4.11 by 6.25 mm2 [582]. The chip is operated at 5 V and hosts all driving circuitry of the sensors such as ISFET buffer amplifiers, a potentiostatic setup for the amperometric sensor, and the circuitry necessary to perform a four-point conductometric measurement on chip. In addition, the chip exhibits a temperature control unit to keep the system temperature at a preset value (physiological conditions). This temperature control unit includes a temperature sensor (parasitic vertical pnp bipolar transistor) and a NMOS transistor heater. A single-bit EPROM (electrically programmable read-only memory) was implemented on chip to make sure that the chip is used only once and then is disposed, which is a crucial feature in medical applications. Additional on-chip electronics include units to control the chip (multiplexer, demultiplexer, 4-bit Gray counter and decoder) and units to provide the biasing and the communication to off-chip instrumentation. Due to the high level of on-chip integration, only 5 external connections are needed: Two for power supply, two for bi-directional communication and one for a clock signal [582]. First tests including amperometric oxygen measurements, the assessment of potassium concentrations with ISFETs (by directly connecting the ISFET buffer to a plotter), and conductometric measurements with a buffer solution have been performed [582]. 5.5.2 CMOS Gas-Phase Multisensor System The CMOS single-chip chemical microsensor system combines three different transducers, a mass-sensitive cantilever, a capacitive sensor, and a calorimetric sensor, all of which rely on polymeric coatings as sensitive layers to detect airborne volatile organic compounds (VOCs) [583,587,588]. The three transducers respond to fundamentally different analyte molecule properties
166
5 CMOS Platform Technology for Chemical Sensors
(a) capacitive sensor polymer
E1
(b) mass-sensitive sensor
(c) calorimetric sensor
analyte
E2 hot junctions
cold junctions
E2 E1
∆C ∝ ∆cgas
∆f ∝ ∆cgas
Utherm ∝ ∆cgas / ∆t
Fig. 5.69. Principles of the three different types of transducers: (a) microcapacitor sensitive to changes in dielectric properties, (b) resonant cantilever sensitive to mass changes, and (c) microcalorimeter measuring the absorption or desorption heat upon interaction of organic volatiles with the polymer [583]
(Fig. 5.69 a-c). One of them responds to the mass of sorbed molecules, another responds to the heat of absorption, and the third responds to the dielectric properties of the absorbates. The monolithic system such simultaneously provides three different (“orthogonal”) sensor responses, which are used to classify or quantify analytes in the gas phase. 5.5.2.1 Multisystem Architecture A schematic of the microsystem architecture is displayed in Fig. 5.70 [583]. The monolithic system includes the sensors (left), driving and signalconditioning circuitry (sensor front ends), analog-to-digital conversion units, sensor control and power management units, and a digital interface. A photograph of the microsystem chip is displayed in Fig. 5.71. The overall chip size is 7 by 7 mm2 . The capacitive sensor as described previously in Sect. 5.1 is integrated with a fully differential second-order Sigma-Delta-modulator and a counter to decimate the output bit stream [587, 588]. The micromachined cantilever, the operation principle of which has been already described in Sect. 5.3.1.1 (thermal actuation), is 150 µm long and consists of silicon as well as vapor-deposited and thermal oxide. The cantilever acts as the frequency-determining element in a feedback oscillation circuit, which is entirely integrated on the chip with a counter. For more details, see [501, 587, 588]. The third transducer is a thermoelectric calorimeter based on the Seebeckeffect with 256 polysilicon/aluminum thermocouples connected in series (500
∆U capacitive ∆C temperature T
sensor control & power management
167
digital bus interface
∆f calorimetric
A/D conversion
sensors mass-sensitive
sensor front ends
5.5 CMOS Chemical Multisensor Systems
Fig. 5.70. Schematic of the microsystem architecture with sensors (left), driving and signal-conditioning circuitry (sensor front ends), analog-to-digital conversion units, sensor control and power management units (bottom), and the digital interface (right) [583]
A/D converter calorimeter calorimetric sensor and reference reference capacitor sensing capacitor Σ∆ capacitive sensor
chopper amplifier Σ∆ temperature sensor decimation filters temperature sensor resonant cantilever digital interface
Fig. 5.71. Micrograph of the gas sensor system chip (size: 7 by 7 mm2 ). The different components are indicated, Σ∆ represents Sigma-Delta converters [583]
by 500 µm2 dielectric membranes). Details have been described in Sect. 5.2. The thermovoltage is translated into a digital signal on chip using a SigmaDelta analog/digital converter and a decimation filter [587, 588]. The chip features all the sensor-specific driving circuitry and signalconditioning circuitry (sensor frontends), which has been described in the context of the different transducers [501, 588]. The system chip additionally includes a temperature sensor since volatile absorption in polymers is strongly temperature-dependent (a 10◦ C temperature increase reduces the sensor sig-
168
5 CMOS Platform Technology for Chemical Sensors
nal by 50%). The temperature sensor exhibits an accuracy of 0.1◦ C at operation temperatures between −40◦ and +80◦ C (see Sect. 5.5.2.2). The analog/digital conversion is done on chip, which allows for achieving a favorable signal-to-noise ratio, since noisy connections are avoided, and a robust digital signal is generated on chip and then transmitted to an offchip data port via an I2 C serial interface [575]. The I2 C bus interface (see Sect. 5.5.2.2) offers the additional advantage of having only very few signal lines (essentially two) for bi-directional communication and also allows for operating multiple chips on the same bus system. An on-chip digital controller (see Sect. 5.5.2.2) manages the sensor timing and the chip power budget. The sensors are located in the center of a metal frame, which is used to apply a flip-chip packaging technique [589] (see Sect. 5.6.3). 5.5.2.2 Multisystem Circuitry Components, Design and Fabrication The three transducers of the multisensor chip and the specific circuitry components (sensor frontends) that have been developed to operate the transducers and to read out and precondition their small signals have been already described in Sect. 5.1.2 (capacitor), 5.2.2 (calorimeter) and 5.3.2.1 (cantilever, thermal actuation) [501,588]. There are, however, several additional circuitry units that are needed to arrive at a smart CMOS-MEMS-based sensor system. A serial interface and a digital controller are needed to reduce the number of bonding pads, to control the different units on chip, and to facilitate either sensor readout by means of a microcontroller or to enable direct displaying of the sensor data and results. Reference voltages are required for the A/Dconverters and are used to bias the sensors. Finally, a temperature sensor has to be included in order to deal with the strong temperature dependence of the gas absorption process in the polymeric sensitive layers. Serial interface A schematic of the I2 C-interface and controller architecture that is implemented on the multisensor chip is shown in Fig. 5.72. Interface and controller have been designed and simulated using very-high-speed-integratedcircuits hardware description language (VHDL). The I2 C serial bus interface [575] offers some advantages in comparison to other serial interfaces like the controller-area-network (CAN)-bus (mostly used in the automotive industry) [590], the serial-peripheral-interface (SPI)-bus by Motorola [591], or the standardized IEEE (institute of electrical and electronic engineers) 1451 bus interface [592]: • Simple, serial protocol: The I2 C-protocol clearly defines the most important features (addressing, word length, master/slave communication, start/stop-condition) without adding expensive overhead that requires large on-chip memory.
5.5 CMOS Chemical Multisensor Systems SCL SDA
I2C bus Addr.
bit3 bit2 bit1 bit0 address coding pads
169
SDA SCL
SDA SCL
I2C slave data
I2C master
command
command
data
command command in out
register bank sensor 1 register 0 register 1
controller sensor 1 delay power down interval data ready sensor 1
register x
A/D register sensor 1 sensor parameters
data ready sensor 1...3
data sensor 1...3
power down sensor 1...3
Fig. 5.72. Schematic of the I2 C bus interface and the digital controller. The sensor controller, the register bank and the digital output register are replicated for each transducer, i.e., three times (capacitor, calorimeter, cantilever). SCL: serial clock, SDA: serial data [588]
• Small number of connections: Only two lines are needed for clock and data transmission. • Bus-enabled: Up to 127 microsystems or multisensor chips can be connected on the same bus. A simple arbitration algorithm solves the problem of data collision. The I2 C-bus offers two possible modes of operation [575]: • Single-master: An external unit polls the different sensor chips and collects the data. • Multi-master: Every single chip can initiate a data-transfer. The multi-master mode was chosen to reduce the data traffic on the bus. The data-recording unit only initializes the multisensor chips and sets their timing parameters. The multisensor chips then send their data in regular intervals without being polled by the data-recording unit. This strategy reduces the traffic on the bus and provides more flexibility for future extensions, such as the introduction of threshold-limited values, i.e., a sensor triggers the system, as soon as its signal exceeds a certain threshold value. Digital Controller The digital controller (Fig. 5.72) interprets the commands coming from the data-recording unit, stores the sensor parameters and manages the access of the single sensors to the I2 C-bus. Each sensor on the chip can be individually addressed via the serial interface. A minimal set of controller commands was
170
5 CMOS Platform Technology for Chemical Sensors
implemented: “power on/off”, “read data”, and “set/read parameters”. For details, see [588]. On-Chip Voltage and Current References Using an array of monolithic multisensor systems, it is neither convenient nor cost-effective to supply each chip with a number of external precision reference voltages or currents. Therefore, all reference voltages and currents have to be generated on chip. The only stable and reliable reference that is available in a standard CMOS processes is the silicon bandgap voltage. Taking the difference between the base-emitter voltages of two identical bipolar transistors biased at different currents leads to a voltage, which is proportional to the absolute temperature (PTAT) and is often used for temperature sensors. On the other hand, the weighted sum of a base-emitter voltage and a PTAT-voltage generates a temperature-independent reference voltage, which is almost equal to the silicon bandgap voltage [588]. The first designs of integrated bandgap references and temperature sensors were published in the late sixties and early seventies [593, 594]. A few years later, the first CMOS-implementations were presented. Some CMOSdesigns utilize the temperature-dependence of the gate-source voltage of MOS-transistors [595]. It is possible to generate PTAT-voltages using MOStransistors operating in the weak inversion region [596], but it is very difficult to generate an accurate and reliable reference voltage. While the bipolar versions are based on the universal silicon bandgap voltage, the MOS-references rely on the implantations that define the threshold voltage. Consequently, the MOS references are process-dependent and hardly reproducible. Therefore, most CMOS designs are based on the parasitic bipolar transistors available in CMOS processes. The standard designs [597] have limited accuracy owing to mismatch and offset problems. Chopping techniques can be employed to reduce the errors as a consequence of the amplifier offset [598]. If no continuous-time reference voltage is needed, switched-capacitor techniques and dynamic-element-matching strategies can be applied to further reduce mismatch-induced errors [598–600]. The references of the multisensor chip rely on the vertical pnp-transistor that is available in p-substrate CMOS processes. For more details on the on-chip references, see [588]. Temperature Sensor The temperature sensor relies on the linear temperature dependence of a bipolar transistor that is available in the CMOS process. The goal in designing the on-chip temperature sensor was to generate all signals from a single bipolar transistor in order to reduce the number of necessary building blocks and to increase the accuracy of the reference voltage and its matching with the measured voltage. This can be achieved by incorporating all elements into a switched-capacitor Sigma-Delta-modulator (Σ∆). The design is derived from a conventional first-order Sigma-Deltamodulator and is shown in Fig. 5.73, for details see Ref. [588]. The temperature sensor exhibits an accuracy of 0.1◦ C at operation temperatures between
5.5 CMOS Chemical Multisensor Systems
171
−40◦ and +80◦ C after calibration. The diagrams in Fig. 5.73 illustrate the principle of the Σ∆ temperature sensor. The input voltage, Uin , and the reference voltage, Uref , of a temperature sensor are both sums of weighted base-emitter voltages (UBE ). As the input stage of a Σ∆-modulator is an integrator, these sums can be generated inside the Σ∆-modulator owing to the fact that a switched-capacitor integrator sums up its input voltages in each clock-cycle. If the clock-cycle of the modulator (clkΣ∆) is subdivided into several phases, the single contributions (UBEx ) to UPTAT and Uref can be accumulated inside the integrator. If a switched-capacitor Σ∆-modulator is used, the different scaling-factors of the single voltage contributions can be realized by changing the input capacitors of the switched-capacitor integrator in each clock phase (Fig. 5.73). The clock-cycle, clkΣ∆ of the first-order Σ∆-modulator is subdivided into five phases. In each phase, one weighted base-emitter voltage is added to the voltage of the integrator, Uint . The resulting voltage change during one clock cycle is ∆Uint = UPTAT ± Uref . A detailed description of all components and the temperature sensor implementation and performance can be found in [588].
I0
clk clkΣ∆
I1
U int
clock form generator U
k T
int
y
y= 0 y= 1 I=I 0 , I=I 1 , k= -1 k=1 UPTAT
I=I 0 , k=1
I=I 0 , k=m ±U ref
I=I 1 , k= -m
t
1
clk
clkΣ∆
Fig. 5.73. Schematic of the Sigma-Delta temperature sensor [588]
Design At present, no single commercial software package is available that can handle all aspects of CMOS-MEMS design. For the lower levels in the design hierarchy, dedicated software packages can be used to perform the respective tasks under the prerequisite that the interfaces to import and export the data are properly specified. The difficult step is the top-level design, in which the transducer layout, and the analog and digital circuitry have to be joined, verified and simulated [601–603].
172
5 CMOS Platform Technology for Chemical Sensors
Layout Various custom-made tools for circuit-design and MEMS-design are available in addition to the commercially available tools [604, 605]. Furthermore, there are widely accepted standard formats to exchange data between the different software packages (geometric data standard (GDS)-format, Caltech interchange format (CIF)) and to import design fragments. Simulation The simulation of a complete smart sensor system (sensor elements, analog circuitry, digital circuitry) using finite-element (FEM)-simulation tools is very complex and computationally expensive. Therefore, behavioral models of the transducers are required. The generation of these models from the layout or from the results of a 3-dimensional FEM simulation is not straightforward. Senturia [601] discusses the advantages and disadvantages of macromodels based on lumped-circuit elements and hardware description languages (HDLs). For some structures (mostly comb-structures for, e.g., accelerometers), the generation of macromodels is supported by customdesigned [606, 607] and commercially available tools [608, 609]. A suitable simulator must offer the possibility to combine macromodels of the sensor with transistor-level netlists and various levels of HDL-representations of analog and digital circuitry. Furthermore, the analog and digital libraries of the CMOS-process to be used must be available. Most foundries only provide parameters for the tools supplied by the large companies (CadenceTM , MentorTM , SynopsysTM ), which restricts the selection to the mixed-signal simulators delivered with those software packages (e.g., SPECTRE-VerilogTM , SPICE-VerilogTM , SABERTM ). Verification and Post-Layout Simulation Most CMOS foundries deliver rule-files to perform design rule checks (DRC) and the extraction of the layout for the layout-versus-schematic check (LVS). Verification rules for MEMS-designs, however, are not provided. As each software package uses different rule formats, the conversion is a difficult task. For micromachined sensors realized in CMOS-technology with additional postCMOS processing steps, there are no rule files available. Therefore, the most convenient solution is to extend the rule-files that are provided for the circuitry elements by rules for the micromachined structures. Design Flow The design flow used for the single-chip gas sensor microsystem is depicted in Fig. 5.74 [588]. Various simulation tools were used to simulate and optimize the transducers [499, 501, 523, 526, 588]. Cadence was employed to draw and simulate the analog schematics, while SynopsysTM and ModelsimTM were used to synthesize and simulate the digital circuitry part. The automatic layout of the digital part was performed with CadenceTM . The top-level design was done in CadenceTM . Simple lumped-circuit models of the transducers were developed with the aim to include the transducers
5.5 CMOS Chemical Multisensor Systems
sensor design
analog circuit design
FEM simulation
analog simulation
FEM model
hand calculation
hand calculation
analog schematics
173
digital circuit design VHDL code
digital simulation
VHDL testbench
synthesizer transducer design rules
transducer layout
transducer model
top-level analog schematic
top-level schematic
digital schematic
hand layout
automatic layout
analog layout
digital layout
top-level layout
design rule check, layout versus schematic
top-level mixed-signal simulation
Fig. 5.74. Design flow of the single-chip gas sensor microsystem [588]
into the top-level simulation. The transducers and the contact network, which has to be patterned on the wafer to enable the electrochemical etch stop, obey design rules that deviate from those of the circuitry. Therefore, an extension to the standard design rule check (DRC) was developed that includes the definitions of the design rules of the transducers and accepts certain violations of the CMOS design rules in the transducer vicinity. Owing to the design-rule violations caused by the transducers, the standard extraction of the layout and the subsequent layout-versus-schematic (LVS) test is not possible. This problem was solved by adapting the extraction rules so that the electrical features of the transducers (e.g., the poly-Si resistors of the piezoresistive Wheatstone-bridge of the resonant gas-sensor) are recognized and can be extracted. This enables the verification of the top-level design by comparing the final layout to the simulated top-level schematic and avoids wiring-errors. For more details on the design and simulation of CMOS integrated microsystems, see [501, 588]. Fabrication The circuitry and the basic sensor elements (thermocouples, heating resistors, piezoresistive Wheatstone-bridge, etc.) are fabricated using an unaltered industrial 0.8-µm CMOS-process provided by austriamicrosystems [496]. CMOS process steps are used wherever possible so as to reduce the number of post-processing steps: The pad-etch is, e.g., used to remove the Si-nitride passivation on top of the capacitive sensor electrodes [499]. The CMOS process itself (implantation steps, thermal budget, etc.) remains unchanged. An add-on to the basic CMOS process was developed to
174
5 CMOS Platform Technology for Chemical Sensors
enable wafer-level anisotropic etching from the back side of the wafer with an etch stop at the n-well of the CMOS process [610, 611]: • Use of modified starting material to improve the quality of the etch pits (wafers with low oxygen content of the bulk silicon, with and without epitaxial layer [611]). • Lithography modification for the metal-layers to achieve a wafer-level contact network to the n-wells that is needed to enable the use of the electrochemical etch-stop technique. The fully processed CMOS wafers are thinned to a thickness of 380 µm, and a silicon-nitride layer that serves as a mask for the subsequent KOH-etching is deposited on the backside. The n-well-membrane precursor of the masssensitive cantilever and the thermally insulated island structure of the calorimetric sensor are then released simultaneously by anisotropic silicon etching with KOH (potassium hydroxide) from the backside of the wafer and by using the electrochemical etch-stop technique that preserves the n-well of the CMOS process in predefined areas. The silicon cantilever is afterwards released by using two subsequent reactive-ion-etching (RIE) steps. The wafers are then diced using a protective polymer foil covering the chips and the microstructures. After exposure to UV-light, the foil does no more adhere to the microstructures and can be removed without damage. Three masks are needed for the silicon-micromachining, one for the KOH-etching and two for releasing the cantilevers [526]. 5.5.2.3 Multisystem Gas Sensor Measurements Polymeric layers of approx. 2 µm thickness were applied to the transducers by using a drop-coating technique. The utilized polymers included ethyl cellulose (EC), poly(dimethylsiloxane) (PDMS), and poly(etherurethane) (PEUT). The polymer-coated single-chip microsensor system was mounted in a gas manifold. The sensors were then exposed to humidity and several volatile organic compounds (ethanol, n-octane, trichloroethene, and toluene) at different concentrations. The total gas flow over the sensor system was 200 mL/min. The sensors were alternately exposed to VOC-loaded synthetic air and pure air at 300 s intervals. For the simultaneous readout of several multisensor chips, a microcontroller board (signal processing unit) was developed, which acquired the signals from the microsensor chips and provided the system with stable supply voltages. The sensor signals were evaluated by the microprocessor, and the results were then transmitted to a personal computer. Simultaneously recorded sensor signals of all three transducers upon exposure to 1200 and 3000 ppm (parts per million) of ethanol, and 1000 and 3000 ppm of toluene at 30◦ C are displayed in Fig. 5.75 [583, 587]. The sensors were alternately exposed to analyte gas and pure carrier gas. The
5.5 CMOS Chemical Multisensor Systems
-1
3000
0
(c) calorimetric thermovoltage [mV]
1000
3000
toluene
1
(b) mass-sensitive frequency shift [Hz]
2
1200
frequency shift [kHz]
(a) capacitive
0
-40
-80
polymer: PEUT
ethanol
175
400
240 160 80 0
200
0 2 4 6 [s]
0
-200
-120 0
20 40 60 80
time [min]
0
20 40 60 80
time [min]
0
20 40 60 80
time [min]
Fig. 5.75. Sensor signals simultaneously recorded from all three polymer-coated transducers upon exposure to 1200 and 3000 ppm of ethanol and 1000 and 3000 ppm of toluene at 28◦ C: (a) frequency shifts (Sigma-Delta converter output) of the capacitor, (b) frequency shifts of the resonating cantilever, and (c) thermovoltage transients of the calorimetric sensor. The close-up shows the development of the calorimetric transient within 6s [583]
polymeric coating consisted of poly(etherurethane), PEUT, at approximately 4 µm thickness. Figure 5.75a shows the measured frequency signals (Sigma-Delta converter output) of the capacitor. Ethanol exhibiting a dielectric constant of 24.5, which is larger than that of PEUT (4.8), causes a capacitance increase and, hence, positive frequency shifts, toluene with a dielectric constant of 2.4, causes a capacitance decrease and negative frequency shifts (see also Sect. 5.1.3). Figure 5.75b displays the cantilever response. Ethanol shows rather low signals as compared to toluene due to its lower molecular mass and due to its lower enrichment (partitioning) in the polymeric phase (see also Sect. 5.3.3). The extent of physisorption is in a first-order-approach proportional to the analyte boiling temperature (ethanol: 351 K, toluene: 360 K): The lower the boiling temperature, the less enrichment in the polymer [516]. The calorimetric results in Fig. 5.75c represent a superposition of already discussed partitioning and the heat budget change due to analyte ab/desorption (see also Sect. 5.2.3). The absorbing analyte liberates heat (predominantly heat of condensation) causing a positive transient signal (positive peak), whereas the desorbing analyte abstracts vaporization heat from the environment generating a negative transient signal (negative peak upon purging, see Fig. 5.75c). The close-up in the upper left corner shows the timeresolved response upon 3000 ppm of toluene within the first 6 seconds. The
176
5 CMOS Platform Technology for Chemical Sensors
condensation/vaporization heat of ethanol is larger (ethanol: 42.3 kJ/mol, toluene: 38.0 kJ/mol at 298 K) due to directional interactions in the liquid phase (compare, e.g., water), whereas polymer partitioning of toluene is stronger owing to its higher boiling point. The signals of all three transducers linearly correlate with the analyte concentration in the low-concentration range (less than 3% of the saturation vapor pressure at the operating temperature) [516]. As can be seen from the orthogonal responses upon gas exposure (Figs. 5.75, 5.76), different molecular properties or different aspects of the coating-molecule interaction are measured by the different platforms. Alcohols, e.g., provide comparably low signals on mass-sensitive transducers due to their high saturation vapor pressure and low molecular mass, but provide large signals on capacitors due to their large dielectric constant (24.5). Drastic changes in thermo-voltages on the thermopiles are measured for chlorinated hydrocarbons (not shown here) used, for example, in cooling sprays, which, in turn, have a low dielectric constant and show only small signals on capacitors. Another powerful feature of the system is, that even a zero response from one of the transducers, if there is a measurable response from the other two, is a highly informative data point to help identify the chemical species. To further improve analyte identification/quantification, an array of microsystem chips coated with different polymers can be used. As an example, the sensitivities (slope of sensor signal versus analyte concentration) of a polar (EC) and a non-polar (PDMS) polymer to various analytes are shown in Fig. 5.76 [612]. For comparability reasons, the analyte responses were normalized to that of n-octane, which was set to unity for each polymer and transducer, so that three bars of unity length would represent the n-octane signal in Fig. 5.76 for each polymer. Another possibility to normalize the signals of the different transducers would rely on multiplying the signals or slopes with the analyte saturation vapor pressures [516]. PDMS
EC 3.0
0.5
Capacitor Cantilever Calorimeter
normalized sensitivity
normalized sensitivity
0.6
0.4 0.3 0.2 0.1
2.5 2.0 1.5 1.0 0.5 0.0
0.0
ethanol
toluene
trichloroethene
Capacitor Cantilever Calorimeter
1
ethanol
2
toluene
3
trichloroethene
Fig. 5.76. Sensitivity as normalized to that of n-octane, which was set to unity for each transducer and polymer. The patterns of the two different polymers (PDMS and EC) differ significantly [612]
5.5 CMOS Chemical Multisensor Systems
177
It is obvious from Fig. 5.76 that polymer variation significantly changes the sensor system response pattern. The nonpolar PDMS preferentially absorbs n-octane so that the response slopes of all other analytes and, especially, the polar ones (e.g., ethanol) are lower (values of less than one). EC is polar so that the opposite holds true. The response characteristics owing to the different transduction mechanisms have already been discussed in the context of Fig. 5.75, and the same considerations apply also to Fig. 5.76. The discrimination capability of a microsensor system, hence, can be greatly enhanced by realizing different transducer principles on a single chip and, additionally, by using several chips with various polymeric coatings (see also Sect. 5.5.2.4). Extensive measurements with the single-chip microsensor system coated with a variety of commercially available gas-sensitive polymers also showed good long-term reproducibility of the sensor responses (less than 10% deviation within one year). 5.5.2.4 Multisystem Applications and Operation Modes The monolithic CMOS gas sensor microsystem is targeted at identifying organic solvents in transport containers or providing workplace safety in, e.g., chemical industry. It will form part of a handheld or credit-card-size detection unit. Depending on the application, either single microsystem chips will be used in standalone units (defined target gas, simple detections tasks such as analyte presence), or more complex systems will be devised that include arrays of several chips coated with different polymers (complex gas environment, presence of interferants, accurate analyte quantification). A handheld gas sensing instrument for more complex detection tasks has been developed, which includes six single-chip multisensor systems that are assembled on a ceramic board (Fig. 5.77) and coated with different polymers. The instrument thus provides a total of 18 different (6 capacitive, 6 masssensitive and 6 calorimetric) chemical sensor signals. The average power consumption of one of the multisensor chips is approximately 100 mW. The instrument also includes a microcontroller board (signal processing unit) developed by the University of Bologna, a miniaturized gas-flow system developed by the University of T¨ ubingen and a battery supply. The signalprocessing unit acquires the signals from the multisensor chips on the ceramic board and provides the chips with stable supply-voltages of 2.2 and 5 V. The signals are evaluated by the microprocessor, and the results are then either transmitted to a personal computer via a standard RS232 serial interface or directly visualized on a liquid-crystal display (Fig. 5.78). A photograph and a schematic of the completely assembled handheld instrument are shown in Fig. 5.78 [588]. It comprises in addition to the sensors a gas intake unit with pumps, valves and filters, a signal processing unit (microcontroller) and a power pack allowing for more than 24 hours of continuous operation. The handheld instrument has been successfully used to
178
5 CMOS Platform Technology for Chemical Sensors
filter capacitors
contact pads
multisensor chips
gas inlet
filter for reference gas
pump
gas outlet
Fig. 5.77. Ceramic board carrying six multisensor chips, the electronics of which are protected by an epoxy cover [588]. Each chip is coated with a different polymer
measurement chamber
valves battery
display
sensor array
microcontroller
RS 232
Fig. 5.78. Photograph and schematic of the handheld instrument [588]
classify solvents under various environmental conditions, e.g., in laboratory environment, at trade-fairs, and public exhibitions. Operation Modes Since standalone units or handheld systems are small by nature and, therefore, have severe constraints imposed on the use of calibration gases or filter units as well as on the overall power consumption, one has to come up with dedicated operation modes, which still enable reliable qualitative and quantitative measurements. One possibility includes the so-called “reverse
5.5 CMOS Chemical Multisensor Systems
179
mode of operation” (RMO) [613], which has been developed for the handheld microsensor unit and has been adapted to the needs of the different transducer principles as will be described in the following. In classical systems, the baseline is established by purging with pure carrier gas (typically nitrogen or synthetic air) or filtered ambient air. Purging is, in most cases, the basic state of the system, and the sensors are exposed to the target analytes only during a short time. The sensor signal upon analyte exposure is then compared to the carrier gas or ambient air signal, which serves as “zero-signal”. Since neither carrying large reference gas cylinders nor using high-capacity filters is feasible in case of portable handheld units, another solution has to be found. The new concept is based on the idea to invert operation conditions. Instead of purging as standard state and short gas exposure times, the system is now continuously exposed to analyte gas, and only short pulses of filtered or reference gas are used to re-establish the baseline. This operation mode induces higher demands on the sensor stability, since large drift deteriorates the sensor signal much more in using RMO than under “classical” operation conditions. For the microsensor system, the reference baseline is established by 5 seconds purging with filtered ambient air. The RMO-technique allows for performing some hundred measurement cycles using a very small filter element. The operation states of valves and pumps applying RMO are displayed in Fig. 5.79 [614], which additionally shows the strategy and timing of signal recording from the different transducers and the resulting sensor signals. Line 1 is indicating the valve status. “0” represents the basic state of the valve, when ambient analyte-loaded gas is directly transferred to the measurement chamber. In state “1”, the ambient gas passes a filter unit, and analyte molecules are removed from the gas stream: Cleaned “reference” gas is flowing over the sensors. Line 2 represents the pump status. “0” means “pump off”, “1” means that the pump is operational. In line 3, the corresponding analyte gas phase concentrations are displayed. In the beginning of a measurement sequence, the gas composition in the measurement chamber is not defined. The pump is then switched on for 3 seconds, and analyte gas is pumped into the measurement chamber. The gas remains in the chamber during two seconds. Equilibrium signals of the capacitive and mass sensitive transducers are recorded, the measurement timing of which is displayed in line 4. The resulting sensor signals are schematically shown in line 5. The pump is then switched on again for 3 seconds, and the valve is set to route the analyte gas through the filter. A concentration step between analyte-loaded and filtered gas is thus generated. Equilibrium state capacitive and mass-sensitive reference signals in filtered air are recorded during two seconds. After that, the pump is switched
180
5 CMOS Platform Technology for Chemical Sensors
line 1
5 sec
1 valve 0
line 2
1
3 sec
3 sec
3 sec
pump 0
line 3
2 sec
2 sec
analyte concentration (undefined) in chamber
line 4 measurement timing capacitor, cantilever
line 5 sensor signal capacitor, cantilever
line 6 measurement timing calorimeter
line 7
sensor signal calorimeter
Fig. 5.79. Reverse Mode of Operation (RMO) [614], as developed for a handheld unit including micromachined multisensor chips: Operation states of valves and pumps and corresponding gas concentrations in the chamber (lines 1–3), and timing of the signal recording for the different transducers as well as resulting sensor signals (lines 4–7). For details see text
on once more with the valve set to analyte gas. The last pump operation would not be necessary for the equilibrium-based sensors but it is necessary to get the second calorimetric transient as shown in line 7. As already described in Sect. 5.2.3, the calorimetric sensor relies on transients and provides signals exclusively upon concentration changes. Therefore, the calorimetric recording has to be performed at high time resolution (1 kHz) in two short intervals covering both flanks of the concentration signal (line 3), i.e., at the maximum gradient of the analyte concentration. The two transient signals of the calorimetric transducer (negative upon analyte desorption, positive upon analyte absorption) are displayed in line 7.
5.6 CMOS Chemical Microsensor and System Packaging
181
5.6 CMOS Chemical Microsensor and System Packaging The packaging of chemical microsensors often does not receive much attention despite the fact that the package largely influences the reliability and long-term stability of a microsystem and represents a major cost factor of commercial microsensor systems [615]. Packaging strategies for integrated chemical microsensor systems in CMOS technology can rely on the copious knowledge in microelectronics and IC-related packaging. However, specific and custom-designed solutions have to be developed for sensor devices, since it is not possible to simply encapsulate the whole chip in a standard process. The package must allow for free access of the physical or chemical stimulus to the sensor. The most important issues with chemical sensor packaging include [616]: • Free analyte access: The analyte must have direct access to the sensor structures, whereas the circuitry and the electrical connections (bonds) have to be protected from the environment (temperature, radiation, humidity, chemicals). • Chemically inert package: The packaging material must be chemically inert with respect to (sometimes) aggressive media and analytes, and it may not outgas or release chemicals that might interfere with analyte detection. • Microsystem-package codesign: The sensor package has to be defined during the initial microsystem conception phase, since the electronics and transducer layout of the microsystem and the component arrangement on the chip are strongly interrelated with the selected packaging solution. Three different packaging approaches featuring increasing complexity will be presented in the following subsections of this chapter: A quite simple approach using transistor-outline (TO) sockets and epoxy encapsulant (Sect. 5.6.1), a chip-on-board-like approach (Sect. 5.6.2) and a flip-chip technologybased approach (Sect. 5.6.3). The advantage of the epoxy-based packaging is its simplicity. The flip-chip packaging requires massive efforts, but allows for minimized sample volumes (gas or liquid-phase flow system), and provides hermetical sealing of the circuitry from the probed sample volume, which leads to improved sensor reliability. 5.6.1 Simple Epoxy-Based Package Epoxy-based packages are robust and simple prototype packaging solutions, which are applicable to the whole family of CMOS-based chemical sensors. For epoxy protection, the sensors that should remain uncovered are placed along one edge of the chip, while all electrical connections (bond pads) and the sensitive parts of the circuitry should be arranged at the other end of the chip as it is shown in Fig. 5.80. This distribution, which has to be realized
182
5 CMOS Platform Technology for Chemical Sensors
sensors
pins
chip circuitry
pads bond wires
epoxy TO-header
Fig. 5.80. Schematic and photograph of an epoxy-based sensor package. The partial epoxy cover enables free analyte access to the sensors of the microhotplate array
during the chip design and layout phase, is necessary since the epoxy cannot be applied with micrometer precision and still flows and expands somewhat during the process of drying. The epoxy-based packaging concept and a packaged microhotplate array chip (see also Fig. 5.65) are shown in Fig. 5.80. A commercially available standard gold-coated transistor-outline TO-8 socket with 16 pins has been chosen. The chips are affixed with an electrically non-conductive die-attach. The die-attach is only applied underneath the circuitry area, so that the microhotplate cavities are not completely sealed. Pressure changes, which might build up during heating can dissipate through the small slit between chip and package. The chip is then wire-bonded with a wedge-wedge bonder and Al-wires. For encapsulation, the bond-wires and the circuitry are covered with a standard epoxy (e.g., Loctite HYSOL FP 4460). The epoxy is applied with a dispenser and is cured at 125◦ C. It is chemically inert and does not affect the sensitive layer during curing. Partially covering the chips with epoxy allows, on the one hand, for a good protection of the chip, on the other hand, it enables free access of the analyte to the microhotplate sensors. Eventually, a metal cap with a gas-permeable membrane can be mounted on top of the TO-socket as protection element against dust and mechanical impact. The cap also allows for integration of metal sieves and filters, which can be an important feature in tuning the sensor selectivity. 5.6.2 Chip-on-Board Package Chip on board means that the CMOS-chip is not bonded onto a dual-in-line (DIL) package or TO-socket, but is glued onto a board or printed-circuit board so that the wire bonding is directly from chip to board. The chip-onboard package has been realized for commercially available humidity sensors [484] and will be applied here to an array of six multisensor chips that have
5.6 CMOS Chemical Microsensor and System Packaging
183
been described in Sect. 5.5.2. A flip-chip design of the multisensor chip that features the sensors in the center of the chip within a metal frame surrounded by the electronics has been shown in Fig 5.71. For the chip-on-board package, another design with larger calorimetric sensors has been developed, which is shown in Fig. 5.81.
calorimetric cantilever
capacitive
interface
array contacts
filter capacitors
Fig. 5.81. Photograph of the multisensor chip as designed for chip-on-board packaging (left) and of the packaged array (right) [588]
The transducers are all placed at one end of the chip so that the circuitry can be protected by epoxy encapsulant. In comparing the different designs of the same multisensor system in Fig. 5.71 and Fig. 5.81, it is evident that the packaging method – flip chip or chip on board – largely influences the design and layout of the microsensor system so that microsystem-package-codesign is highly desirable as has been mentioned in the introductory part of this section (Sect. 5.6). Six identical multisensor chips (5 mm by 7 mm) are die-attached on a common ceramic alumina substrate (Fig. 5.81). Each of the multisensor chips features an I2 C bus interface (see 5.5.2.2) so that all chips can be operated on the same bus, and only few electrical connections are necessary. Wire bonds connect the bond pads on the chips with the electrical leads on the ceramic board. The wire bonds and the circuitry are then protected with epoxy. Finally, the chemical-sensor structures of the chips are spray-coated, each chip with a different polymer using a shadow mask.
184
5 CMOS Platform Technology for Chemical Sensors
5.6.3 Flip-Chip Package Flip-chip packaging is a technology that has been developed to allow for a multitude of connections from chip to package or board [617]. Only pads along the edges can be used in, e.g., standard wire bonding or packaging methods. Since the length of the edges becomes smaller and smaller with decreasing feature size of the CMOS processes albeit the bonding pads require a certain size and area to allow for reliable connections, a packaging and connection technology was developed, which enables to make connections on the whole chip surface [617]. The chip is flipped upside down and the connections are made simultaneously from all pads on the chip face to the corresponding pre-patterned substrate via metal solder or conductive polymers (Figs. 5.82, 5.83). Flip-chip packaging is technologically more expensive than the previous two methods and requires a layout with a similar number of connections on each side of the chip (see Fig. 5.83). In the case of chemical microsensors, it is not the large number of possible connections (bus interface: 7 connections) that renders this technology attractive but the possibility to define a sensor area within a metal frame and to hermetically seal the circuitry components from the chemical environment by means of solder [589]. By using a chemically inert substrate material such as alumina, a reliable and longterm-stable package is created.
solder
ceramic sealing
opening
chip metallization
sensor
underfill (polymer)
Fig. 5.82. Schematic of the flip-chip process: Ceramic substrate and chip are pretreated (metal-line patterning on the substrate, gold-bump metallization on the chip) and afterwards connected via solder and polymeric underfill [499, 589]
On the chip side, the bonding pads have to be adapted for flip-chip packaging since the minimum area required for a reliable flip-chip connection is 150 by 150 µm2 as compared to only 80 by 80 µm2 of a regular bonding pad. The sensors have to be placed in the center of the die, and a metal frame has to be designed around the transducers (Figs. 5.71, 5.83), which does not serve electrical connection purposes but is intended to provide hermetical sealing of the circuitry components outside the metal frame. The different pads have
5.6 CMOS Chemical Microsensor and System Packaging
chip with metallization
185
ceramic substrate with solder
metal ring
gold contacts
gold frame
contact pads
Fig. 5.83. Flip-chip procedure: The chip (left) is soldered face-down onto the ceramic substrate. Metal sealing frames and contact pads are clearly visible [499]
to be distributed on the chip surface in a symmetric arrangement, so that the ceramic substrate can rest on those pads and is not tilted. The metallic frame surrounding the sensors and the pads are then covered with nickel/gold bumps. The ceramic substrate also has to be prepared. Laser cutting is used to open a window for the sensors in the ceramic substrate. Then, the electrical connections are screen-printed on the ceramic. The soft solder paste for the flip-chip packaging is also applied to the ceramic using stencil printing. The chip is then precisely positioned face-down on the ceramic substrate by means of a micromanipulator and a solder reflow is performed at 230◦ C. The surface tension of the solder provides a very accurate alignment during the reflow process. Finally, an epoxy-based polymeric underfill is applied and cured at 160◦ C. The sensors can afterwards be coated with different polymers using a drop-coating method and the substrate openings as mask. The flip-chip procedure is exemplified in Fig. 5.83 with a capacitive chip featuring three interdigitated sensor-reference pairs (top: references, center: sensors). The metal frame on the chip around the three sensors and the metal bumps at the top and the bottom of the chip can be clearly seen. The substrate features three openings for demonstration purposes (one opening for all three sensors would be sufficient) and a symmetric arrangement of contacts to accommodate the chip. After the flip-chip process, the three sensing capacitors look through the windows, albeit the references (top) and circuitry units (bottom) are sealed from the analyte and sample environment. The polymeric underfill mechanically stabilizes the packaged system.
6 Outlook and Future Developments
The field of CMOS-integrated chemical and biosensors is currently developing at fast pace and is expanding as can be seen from the growing number of publications over the last years. The two most important research and development thrusts include (i) realizing more complex monolithic sensor arrays (also including different transducer types) and integrating more and more electronic functions on the sensor chip, and (ii) combining microelectronics, in particular CMOS chips, with biologically relevant molecules such as desoxyribonucleic acid (DNA), or biological entities such as lipid bilayers, vesicles or whole cells. The first trend towards arrays and more complex microsensor systems, which offer a lot of functions and features and are easy to use, is evident from Chap. 5 of this book. A modular approach or “toolbox strategy” relying on CMOS as platform technology has been described in detail. A variety of components of the toolbox such as transducers, sensor modules, and circuit modules have been developed, which afterwards have been assembled into customized systems such as the digital microhotplate array (Sect. 5.4.2.3) or the multitransducer system (Sect. 5.5.2). Developments towards more complex microsystems in the immediate future will include: • • • •
Further miniaturization Higher level of integration Implementation of additional functions Development of sensor-based analysis microsystems
Each of those points will be shortly elaborated. Miniaturization means the further shrinking of existing transducer structures to smaller size, e.g., micromechanical cantilevers down to the nanometer scale, or microhotplates to less than 100 µm diameter. Miniaturization also will include replacing microtransducers for rather large millimeter-size transducers, or devising novel sensing and actuation principles for chemical microsensors. All these efforts will lead to a drastic reduction in system size and power consumption. A target system size of 5 by 5 mm2 , and a target system power consumption of less than 100–200 mW (depending on transducer type) seem to be a realistic estimate.
188
6 Outlook and Future Developments
Another important issue in further miniaturization will concern the gassensitive material and its precise local deposition. Micro- and nanometerresolution deposition methods have to be developed, which provide sufficient precision and reproducibility at reasonable costs and potentially high throughput. Further miniaturization of transducers and sensing structures stringently requires a higher level of integration of electronic and circuitry components: The sensor signals will become more and more minute, so that the transfer of low-level analog signals via bonds and interconnects will be no more an option: The signal of, e.g., capacitive chemical microsensors (see also Sect. 5.1) is on the order of some attoFarads (10−18 Farad). Monolithic implementations, which combine miniaturized transducers and on-chip circuitry in immediate vicinity to the signal-generating processes, represent an effective approach to significantly reducing the influence of parasitic capacitances and crosstalk-effects that deteriorate low-level signals of any microsensor. Short on-chip interconnections between the sensor components and the system control help to further minimize noise. In most cases, it will also be necessary to convert the signal on chip into the digital domain and to transfer only digital signals to off-chip units. Data preprocessing and A/D or D/A conversion is hence to be implemented on chip, which poses new challenges (see Sect. 5.5.2.2) with regard to design and testing: New simulation and test strategies for digital-analog co-design of rather complex microsystems and chip architectures have to be elaborated. Additional functions will include subunits such as auxiliary sensors (temperature, humidity) to reduce the influence of external disturbances (temperature) and cross-interferants (humidity) on the sensor signal, customdesigned interfaces, bus interfaces (see Sect. 5.5.2.2), adaptive circuits for signal evaluation and discrimination [618, 619], or telemetry units for wireless communication. Telemetry functions will encompass, e.g., communication from implanted chips (glucose sensors) through the skin in medical applications [620] or the transmission of remote and distributed chemical sensor responses to a central terminal in environmental and building control scenarios by using radio frequencies [621–623] or other standards like BluetoothTM [624]. Complete miniaturized chemical analysis systems are currently under development at Sandia National Laboratories and at the University of Michigan. Sandia National Laboratories has developed both, gas and liquid phase chemical analysis systems (µ ChemLabTM ) [625, 626]. The gas-phase system is targeted towards the field detection of chemical warfare agents and consists of three microfabricated components (Fig. 6.1): (a) A preconcentrator, (b) a separation unit, and (c) a gravimetric detector. Preconcentration is achieved by pulling an air sample across a selective absorbent coating on a thermally isolated silicon nitride membrane; a heating pulse is then used to release the absorbed chemical species. Separation is accomplished using
6 Outlook and Future Developments
189
excitation propagation
heater metal metal electrodes
silicon nitride
detection
silicon substrate thin-film adsorbent
preconcentrator accumulates species of interest
gas chromatograph separates species in time
Rayleigh-SAW sensors provide sensitive detection
Fig. 6.1. Schematic of the µChemLabTM system as developed by Sandia National Laboratories (top) and micrographs of its different components (bottom). The components are not to scale. Pictures courtesy of R.W. Cernosek, Sandia National Laboratories
either a wall-coated or bead-packed deep-reactive-ion-etched silicon channel that acts as a microgaschromatographic (µGC) column. The separated samples are detected with a microfabricated surface-acoustic-wave (SAW) device array. All key components of a microgaschromatographic column (µGC), namely a passive calibration-vapor source, a three-stage preconcentrator/focuser, a 3-meter separation column, and an integrated vapor sensor array have been developed at the University of Michigan [627, 628]. All these components are fabricated from silicon and collectively occupy approximately 1 cm3 volume. Plumbed via a fluidic interconnection substrate in combination with fused silica microcapillaries, the µGC is coupled with a commercial minidiaphragm pump and minivalves to perform analyses of various mixtures of toxic volatile organic compounds. The second important trend is the combination of CMOS microelectronics with biomolecules or biological entities such as living cells or electrogenic cells (cells that react upon electrical stimulation and, in turn, produce electrical signals such as heart cells or neurons). Sensor arrays for DNA-analysis based on CMOS–assisted electronic detection rather than optical detection represent a very interesting application for CMOS technology [629,630]. The sensor chip encompasses a 16×8 pixel array with gold sensor electrodes [630]. The pixel circuits are designed to detect and amplify sensor currents in the 100 nA to pA range that result from an electrochemical redox cycling process. The chip is realized in 0.5-µm CMOS
190
6 Outlook and Future Developments
technology and additionally hosts addressing circuitry, reference electrodes and potentiostatic circuitry. Monitoring of chemical parameters in cell cultures, such as metabolically induced pH-changes or oxygen contents and ionic composition in the cell environment, has been performed with circuit-less CMOS chips that include ISFET arrays (pH and different ions), interdigitated electrodes, an oxygen and a temperature sensor [631–634]. Arrays of electrodes or microelectrodes with associated circuitry on chip can be used to extracellularly measure the electrical signals of electrogenic cells. Portable, cell-based biosensor systems featuring CMOS chips to interrogate electrogenic cells have been reported on [635–639]. Cardiac myocytes were coupled to CMOS circuits to yield a biosensor for detecting warfare agents in portable field alarm systems. The cartridge contains a CMOS silicon chip that incorporates a digital interface, a temperature control system, microelectrode electrophysiology sensors and analog signal buffering. Each of the two chambers hosting the cells on the chip features one large reference electrode, two stimulation electrodes, four arrays of 16 microelectrodes, temperature sensors and nine 16-channel multiplexers. As soon as the heart cells are exposed to toxic chemicals they change their electrical signal characteristics (faster or slower beating frequency) or perish (no signal). First results on coupling neurons to CMOS-based chips were published very recently [640–642]. A biocompatible capacitive interface including a titanium oxide and zirconium oxide sandwich has been post-processed onto each of the 16.000 pixels or sensors, which have an integrated amplification circuit. Action potentials of firing neurons are thus capacitively coupled to the gates of MOS transistors that operate as input devices within a monolithic detection system, which also features addressing circuitry, multiplexers, and a temperature sensor, but does not yet enable electrical stimulation. A chip comprising a microelectrode array and fully integrated analog and digital CMOS circuitry for the stimulation and recording of electrogenic cells was reported on in [643]. This device is capable of on-chip signal filtering, which improves the signal-to-noise ratio, on-chip analog/digital conversion to prevent signal degradation, and simultaneous recording and stimulation (Fig. 6.2). Each electrode has associated circuitry comprising a stimulation buffer and a switch, a bandpass filter and a readout buffer. The electrode and circuitry form a repeatable unit with a 250-µm pitch. This unit can be multiplied to form a larger array. The stimulation buffer, occupying very small area, uses a class-AB output stage in order to deliver large currents to the electrode. Any stimulation waveform can be generated. Fabrication was performed using an industrial double-polysilicon, triple-metal 0.6-µm CMOS process. The platinum microelectrodes are realized during postprocessing for flexibility purposes. First tests of the chip in Fig. 6.2 were performed with an in situ preparation of cardiac myocytes (heart cells) from embryonic chicken at embryonic
6 Outlook and Future Developments
191
test structure 16-electrode array
circuitry repeating unit
4 A/D converters digital control
30 µm D/A converter biasing temperature sensor
500 µm electrode (Pt)
Fig. 6.2. Chip for recording electrogenic cell activity: 16-electrode array (40 × 40 µm2 , 250 µm pitch), one analog-to-digital converter per row of electrodes, one digital-to-analog converter, on-chip multiplexer, digital control unit, and temperature sensor. Close-up (left side): Circuitry repeating unit with buffer, switch for stimulation, bandpass filter and readout buffer [643]
day 10. The heart cells were extracted from the embryo, briefly rinsed and directly transferred onto the electrodes of the chip. A spike is shown in Fig. 6.3. The signal is on the order of millivolts, and is considerably larger than that of, e.g., chicken neurons (tens of microvolts) [643]. Needle-shaped neural probes fabricated in CMOS-technology in conjunction with micromachining have been developed at the University of Michigan [644–648]. A neural probe consists of a micromachined silicon substrate with needles supporting an array of thin-film conductors leading from recording or
0.4
U [mV]
0.3 0.2 0.1 0 -0.1 -0.2 10
50
100
time [ms]
Fig. 6.3. Spike of cardiac myocytes from embryonic chicken [643]
192
6 Outlook and Future Developments output leads
(a)
(b)
interconnecting leads stimulating or recording sites
signalprocessing circuitry supporting substrate
Fig. 6.4. Schematic (a) and micrograph (b) of a neural probe. Fig. 6.4b shows a 64-site 8-channel stimulating array delivering currents from 0 to ±128 µA with a resolution of 1 µA and 4 µsec. Reprinted with permission from [650]
stimulating sites along a shank to signal processing electronics and/or output leads at the rear of the device (Fig. 6.4). The recording or stimulating sites (up to 64) are mounted along the upper surface allowing for recording or stimulation at many points in depth. The neural probes are aimed at recording activity from intact cortical tissue after insertion. Chronic probes have been recording signals for several months in vivo [648]. The on-chip circuitry allows for amplification and multiplexing of the neural signals without significantly degrading baseline noise levels. Another primary reason for on-chip electronics in such a device is to minimize the number of external leads, which eases connectivity and reduces the chance of lead-related failures. In recent papers, a multichannel neural probe was presented that also provides regio-selective chemical delivery as additional function along with the stimulation and recording of cell activity [648–650]. In summary, using CMOS and CMOS-MEMS technology to devise chemical, biochemical and biosensors offers the possibility to select components from a wealth of electronic circuits and functions that are and will be provided by the enduring huge development efforts of the IC-industry. The economic aspect, however, has to be kept in mind, i.e., the fact that CMOS sensor development is expensive (mask and processing costs) and only pays off, if either millions or at least hundreds of thousands of pieces are sold, or the devices provide unique functionality, which then legitimates the costly development. Applications involving large numbers of chemical or biosensors have not yet emerged but may occur in the field of automotive applications, domestic gas alarms, HVAC (humidity, ventilation, air conditioning) units, in the field of disposable medical devices (e.g., DNA or blood tests), or in the field of pharmacological screening.
References
1. G¨ opel, W., Hesse, J., Zemel, J.N. (eds.), Sensors: A Comprehensive Survey, Vol. 2/3, Chemical and Biochemical Sensors, VCHVerlagsgesellschaft,Weinheim 1991, and Baltes, H., G¨opel, W., Hesse, J. (eds.), Sensors Updates, VCH-Wiley, Weinheim. 2. Hulanicki, A., Glab, S., Ingman, F., Pure & Appl. Chem., 1991, 63, 1247. 3. Durst, R.A., Murray, R.W., Izutsu, K., Kadish, K.M., Faulkner, L.R., Draft IUPAC Report Commission V. 5. 4. Janata, J., Principles of Chemical Sensors, Plenum, New York, 1989, and Janata, J., Huber, R.J. (eds.), Solid State Chemical Sensors, Academic Press, San Diego, 1985. 5. Janata, J., Josowicz, M., Vanysek, P. DeVaney, M.D. Anal. Chem. 1998, 179R– 208R. 6. Edmonds, T.E. (ed.), Chemical Sensors, Chapman and Hall, New York, 1988. 7. Kress-Rodgers, E. (ed.), Handbook of Biosensors and Electronic Noses, CRC Press Inc., Boca Raton, FL, 1997. 8. Madou, M.J., Morrison, S.R., Chemical Sensing with Solid State Devices, Academic Press, Boston, 1989. 9. Gardner, J.W., Bartlett, P.N., Electronic Noses, Oxford University Press, 1999, and Gardner, J.W., Microsensors, Wiley, Chichester, 1994. 10. Hall, E.A., Biosensors, Open University Press, 1990. 11. Lambrechts, M., Sansen, W., Biosensors: Microelectrochemical Devices, Institute of Physics Publishing, Bristol, 1992. 12. Carey, W.P., Beebe, K.R., Kowalski, B.R. Anal. Chem. 1986, 58, 149–153. 13. Ballantine, D.S., Rose, S.L., Grate, J.W., Wohltjen, H. Anal. Chem. 1986, 58, 3058–3066. 14. Osbourn, G.C., Bartholomew, J.W., Ricco, A.J., Crooks, R.M. Acc. Chem. Res. 1998, 31, 297, and Ricco, A.J., Crooks, R.M., Osbourn, G.C. Acc. Chem. Res. 1998, 31, 289. 15. Hierlemann, A., Schweizer-Berberich, M., Weimar, U., Kraus, G., Pfau, A., G¨ opel, W. Pattern Recognition and Multicomponent Analysis in Sensors Update, and Baltes, H., G¨ opel, W., Hesse, J., (eds.), VCH: Weinheim, Germany, 1996. 16. Massart, D.L., Vandeginste, B.G.M., Deming, S.N., Michotte, Y., Kaufman, L., Data Handling in Science and Technology 2: Chemometrics: a Textbook, Amsterdam, Elsevier 1988. 17. Brereton, R.G. (ed.), Multivariate pattern recognition in chemometrics, Data handling in science and technology, Vol. 9, Elsevier, Amsterdam, New York, 1992.
194
References
18. Weetall, H.H., Biosens. Bioelectronics, 1999, 14, 237–242. 19. Van den Berg, A., van der Waal, P.D., van der Schoot, B.B., de Rooij, N.F., Sens. Mat. 1994, 6, 23–43. 20. M¨ uller, G., Deimel, P.P., Hellmich, W., Wagner, C., Thin Solid Films, 1997, 296, 157–163. 21. Bergveld, P., IEEE Trans. Biomed. Eng., 1970, BME-17, 70–71. 22. Lundstr¨ om, I., Shivaraman, S., Svensson, C., Lundkvist, L., Appl. Phys. Lett. 1975, 26, 55–57. 23. Middelhoek, S., Audet, S.A., Silicon Sensors, Academic Press Inc., London, 1989, and Middelhoek, S., Sens, Actuators A, 1994, 41–42, 1–8. 24. Sze, S.M. (ed.), Semiconductor Sensors, Wiley, New York, 1994. 25. Gad-el-Hak, M., The MEMS Handbook, CRC Press, Boca Raton, 2002. 26. Kovacs, G.T.A., Micromachined Transducers, MCB McGraw-Hill, New York, 1998. 27. Elwenspoek, M., Hansen, H., Silicon Micromachining, Cambridge University Press, Cambridge, 1998. 28. Madou, M., Fundamentals of Microfabrication, CRC Press, Boca Raton, FL, 1997. 29. Muller, R.S., Howe, R.T., Senturia, S.D., Smith, R.L., White, R.M. (eds.), Microsensors, IEEE Press, New York, 1991. 30. Tuller, H.L., Mlcak, R. Sens. Actuators B, 1996, 35–36, 255–261. 31. Picraux, S.T., McWorther, P.J., IEEE Spectrum, 1998, 12, 24–33. 32. Gardner, J.W., Varadan, V.K., Awadelkin, O.O., Microsensor, MEMS and Smart Devices, Wiley, New York, 2001. 33. Baltes, H., Brand, O., IEEE AES Systems Magazine, 1999, 14, 29–34, and Proc. SPIE, 1999, 3673, 2–10. 34. Zemel, J.N., Rev. Sci. Instrum. 1990, 61 (6), 1579–1606. 35. Atkins, P.W., Physical Chemistry, Oxford University Press, Oxford, 5th edition, 1994. 36. Sze, S.M., Semiconductor Devices: Physics and Technology, 2nd edition, Wiley, 2002. 37. Sze, S.M., (ed.), VLSI Technology, 2nd edition, McGraw-Hill, 1988. 38. Sze, S.M., Chang, C.Y. (eds.), ULSI Technology, McGraw-Hill, 1996. 39. Brand, O., Baltes, H., Sens. Actuators A, 2001, 92, 1–9. 40. Petersen, K.E., Proc. IEEE 1982, 70, 420–457. 41. Menz, W., Mohr, J., Paul, O., Microsystems Technology, Wiley-VCH, Weinheim, 2001. 42. Heuberger, A., Mikromechanik, (in German), Springer, 1991. 43. Voldman, J., Gray, M.L., Schmidt, M.A., Annu. Rev. Biomed. Eng., 1999, 1, 401–425. 44. Tummala R.R. (ed.), Fundamentals of Microsystems Packaging, McGraw-Hill, 2001. 45. Nguyen, N.T., Wereley, S.T., Fundamentals and Applications of Microfluidics, Artech House Publishing, Norwood, 2002. 46. Schift, H., David, C., Gabriel, M., Gobrecht, J., Heyderman, L.J., Kaiser,W., K¨ oppel, S., Scandella, L., Microelectronic Engineering, 2000, 53, 171–174. 47. Hierlemann, A., Brand, O., Hagleitner, C., Baltes, H., Proc. of the IEEE, Chemical and Biological Microsensors, Editors S. Casalnuovo, R.B. Brown, 91 (2003), 839–863.
References
195
48. Williams, K.R., Muller, R.S., J. of Microelectromechanical Systems, 1996, 5, 256–269. 49. Kovacs, G.T.A, Maluf, N.I., Petersen, K.E., Proc. IEEE, 1998, 86, 1536–1551. 50. Bustillo, J.M., Howe, R.T., Muller, R.S. Proc. IEEE, 1998, 86, 1552–1574. 51. Shikida, M., Sato, K., Tokoro, K., Uchikawa, D., Sens. Actuators A, 2000, 80, 179–188. 52. M¨ unch, U., Industrial CMOS Technology for Thermal Imagers, PhD Thesis No. 13801, ETH Zurich, Switzerland, 2000. 53. Baltes, H., Paul, O., Brand, O., Proc. IEEE 1998, 86, 1660–1678. 54. XACTIX, Pittsburgh, USA, http://www.xactix.com/. 55. L¨ armer, F., Schilp, A., Method of anisotropically etching silicon, German Patent No. DE 4,241,045, US patent Nr. US5501893, issued 1994. 56. STS Surface Technology Systems, Newport, UK,http://www.stsystems.com/. 57. UNAXIS Semiconductor, Balzers, FL, http://www.semiconductors.unaxis. com/. 58. Schmidt, M.A., Proc. IEEE, 1998, 86, 1575–1585. 59. Moseley, P.T., Crocker, A.J., Sensor Materials, IOP Publishing, Bristol, 1996. 60. Whitesides, G.M., Ostuni, E., Takayama, S., Jiang X., and Ingber, D.E., Annu. Rev. Biomed. Eng., 2001, 3, 335–73. 61. Grate, J.W., Frye, G.C. In Sensors Update, Vol. 2, and Baltes, H., G¨opel, W., Hesse, J. (eds.), VCH: Weinheim, FRG, 1996, pp. 37–83. 62. Grate, J.W., Martin, S.J., White, R.M. Anal. Chem., 1993, 65, 940A–948A and 987A–996A. 63. Ballantine, D.S., White, R.M., Martin, S.J., Ricco, A.J., Frye, G.C., Zellers, E.T., and Wohltjen, H., Acoustic Wave Sensors: Theory, Design, and PhysicoChemical Applications Academic Press, San Diego, 1997. 64. Sauerbrey, G.Z. Phys., 1959, 155, 206–222. 65. Martin, S.J., Frye, G.C., Senturia, S.D. Anal. Chem., 1994, 66, 2201. 66. King, W.H., Anal. Chem. 1964, 36, 1735. 67. White, R.M., Proc. IEEE, 1970, 58, 1238–76. 68. Nieuwenhuizen, M.S., Venema, A. Sens. Mater., 1989, 5, 261–300. 69. Bodenh¨ ofer, K., Hierlemann, A., Noetzel, G., Weimar, U., G¨opel, W. Anal. Chem., 1996, 68, 2210. 70. Martin, S.J., Frye, G.C., Spates, J.J., Butler, M.A., Proc. IEEE Ultrasonics Symposium (Cat. No. 96CH35993) 1996, 1, 423–434. 71. Venema, A., Thin Solid Films, 1974, 22 (2), S27–S29, and Venema, A., Dekkers, J.M., IEEE Trans. Microwave Theory Tech., 1975, MTT23 (9), 765– 767. 72. Lakin, K.M., Liu, J., Wang, K., Proc. IEEE Ultrasonics Symposium, 1974, 302–306. 73. Xia, J., Burns, S., Porter, M., Xue, T., Liu, G., Wyse, R., Thielen, C., Proc. IEEE International Frequency Control Symposium (Cat. No. 95CH35752), 1995, 879–884. 74. Chuang, C.T., White, R.M., Bernstein, J.J., IEEE Electron Device Letters, 1982, 6, 145–148. 75. Martin, S.J., Schweizer, K.S., Ricco, A.J., Zipperian, T.E., Proc. Transducers 1985 (Cat. No. 85CH2127–9), 1985, 445. 76. Nieuwenhuizen, M.S., Nederlof, A.J., Sens. Actuators B, 1992, 9, 171–176. 77. Brand, O., Baltes, H., Micromachined resonant sensors in Sensors Update Vol. 4, and Baltes, H., G¨ opel, W., Hesse, J. (eds.), VCH, Weinheim, Germany, 1998.
196
References
78. Polla, D.L., Muller, R.S., IEEE Electron Device Letters, 1986, 7, 254–256. 79. Caliendo, C., Verardi, P., Verona, E., D’Amico, A., DiNatale, C., Saggio, G., Serafini, M., Paolesse, R., Huq, S.E., Smart Mater. Struct. 1997, 6, 689–699. 80. Casalnuovo, S.A., Hietala, V.M., Heller, E.J., Frye-Mason, G.C., Baca, A.G., Wendt, J.R., Technical Digest Solid-State Sensor and Actuator Workshop, Hilton Head Island, South Carolina, 2000, 154–157. 81. Wenzel, S.W., White, R.M., IEEE Transactions on Electron Devices, 1988, 35, 735–743. 82. Wenzel, S.W., White, R.M., Appl. Phys. Lett., 1989, 54, 1976. 83. Vellekoop, M.J., Lubking, G.W., Sarro, P.M., Venema, A., Sens. Actuators A 1994, 249–263. 84. Wenzel, S.W., White, R.M., Sens. Actuators A, 1990, 21–23, 700–703. 85. Martin, S.J., Butler, M.A., Spates, J.J., Mitchell, M.A., Schubert, W.K., J. Appl. Phys. 1998, 83, 4589–4601. 86. Luginbuhl, P., Colllins, S.D., Racine, G.A., Gr´etillat, M.A., de Rooij, N.F., Brooks, K.G., Setter, N., Journal of Micromechanical Systems 1997, 6, 337– 346. 87. Vellekop, M.J., Van Rhijn, A.J., Lubking, G.W., Venema, A., Sens. Actuators A, 1991, 25–27, 699–703. 88. Martin, B.A., Wenzel, S.W., White, R.M., Sens. Actuators A, 1990, 21–23, 704–708. 89. Schubert, W.K., Adkins, D.R., Butler, M.A., Martin, S.J., Mitchell, M.A., Kottenstette, R., Wessendorf, K.O., Proc. ECS Meeting Honolulu, Hawaii, 1999, Vol. 99–23, 332–335. 90. Grate, J.W., Wenzel, S.W., White, R.M., Anal. Chem., 1991, 63, 1552–1561. 91. Grate, J.W., Wenzel, S.W., White, R.M., Anal. Chem., 1992, 64, 413–423. 92. Wang, A.W., Kiwan, R., White, R.M., Ceriani, R.L., Sens. Actuators B, 1998, 49, 13–21. 93. Lange, D., Hagleitner, C., Hierlemann, A., Brand, O., Baltes, H., Anal. Chem. 2002, 74, 3084–3095. 94. Baltes, H., Lange, D., Koll A., IEEE Spectrum, 1998, 9, 35–38. 95. Hierlemann A., Lange, D., Hagleitner, C., Kerness, N., Koll A., Brand O., Baltes, H., Sens. Actuators B, 2000, 70, 2–11. 96. Thundat, T., Chen, G.Y., Warmack, R.J., Allison, D.P., Wachter, E.A., Anal. Chem. 1995, 67, 519–521. 97. Wachter, E.A., Thundat, T., Rev. Sci. Instrum., 1995, 66, 3662–3667. 98. Lang, H.P., Berger, R., Battiston, F., Ramseyer J.P., Meyer E., Andreoli C., Brugger J., Vettiger P., Despont M., Mezzacasa T., Scandella L., G¨ untherodt H.J., Gerber C., Gimzewski J.K., Appl. Phys. A, 1998, 66, 161–64. 99. Maute M., Raible S., Prins F.E., Kern D.P., Ulmer H., Weimar U., G¨opel W., Sens. Actuators B 1999, 58, 505–511. 100. Lang, H.P., Baller, M.K., Berger, R., Gerber C., Gimzewski J.K., Battiston, F., Fornaro, P., Ramseyer J.P., Meyer E., G¨ untherodt H.J., Anal. Chim. Acta., 1999, 393, 59–65. 101. Boisen, A., Thaysen, J., Jesenius, H., Hansen, O., Ultramicroscopy, 2000, 82, 11–16. 102. Jesenius, H., Thaysen, J., Rasmussen, A.A., Veje, L.H., Hansen, O., Boisen, A., Appl. Phys. Letters, 2000, 76, 2615–2617. 103. Gimzewski, J.K., Gerber, C., Meyer, E., Schlittler, E.E., Chem. Phys. Lett. 1994, 217, 589–594.
References
197
104. Fritz, J., Baller, M.K., Lang, H.P., Rothuizen, H., Vettiger, P., Meyer E., G¨ untherodt H.J., Gerber, C., Gimzewski J.K., Science, 2000, 288, 316–318. 105. Berger, R., Delamarche, E., Lang, H.P., Gerber, C., Gimzewski, J.K., Meyer E., G¨ untherodt H.J., Science, 1997, 276, 2021. 106. Lee, S.S., White, R.M., Sens. Actuators A, 1996, 52, 41–45. 107. Britton, C.L., Warmack, R.J., Smith, S.F., Oden, P.I., Jones, R.L., Thundat, T., Brown, G.M., Bryan, W.L., DePriest, J.C., Ericson M.N., Emery, M.S., Moore, M.R., Turner, G.W., Wintenberg, A.L., Threatt, T.D., Hu, Z., Clonts, G., Rochelle, J.M., Proceedings 20th Anniversary Conference on Advanced Research in VLSI. IEEE Comput. Soc, Los Alamitos, CA, USA, 1999, 359– 68. 108. Barnes, J.R., Stephenson, R.J., Welland, M.E., Gerber, C., Gimzewski, J.K., Nature 372, 1994, 79–81. 109. Meijer, G.C.M., van Herwaarden, A.W., Thermal Sensors, Institute of Physics Publishing, Bristol, UK, 1994. 110. Baker, A.R., Combustible gas-detecting, electrically heatable element, UK Patent 892530, 1962. 111. Zanini, M., Visser, J.H., Rimai, L., Soltis, R.E., Kovalchuk, A., Hoffman, D.W., Logothetis, E.M., Bonne, U., Brewer, L., Bynum, O.W., Richard, M.A., Sens. Actuators A, 1995, 48, 187–192. 112. Manginell, R.P., Smith, J.H., Ricco, A.J., Moreno, D.J., Hughes, R.C., Huber, R.J., Senturia, S.D., Technical Digest Solid State Sensor and Actuator Workshop Hilton Head Island, SC, 1996, 23–27. 113. Manginell, R.P., Smith, J.H., Ricco, A.J., Proc. 4th Annual Symposium on Smart Structures and Materials SPIE, 1997, 273–284. 114. Dabill, D.W., Gentry, S.J., Walsh, P.T., Sens. Actuators, 1987, 11, 135–143. 115. Jones, M.G., Nevell, T.G., Sens. Actuators, 1989, 16, 215–224. 116. Aigner, R., Dietl, M., Katterloher, R., Klee, V., Sens. Actuators B, 1996, 33, 151–155. 117. Aigner, R., Auerbach, F., Huber, P., Mueller, R., Scheller, G., Sens. Actuators B, 1994, 18–19, 143–147. 118. Krebs, P., Grisel, A., Sens. Actuators B, 1993, 13–14, 155–158. 119. Gall, M., Sens. Actuators B, 1991, 4, 533–538. 120. Gall, M., Sens. Actuators B, 1993, 15–16, 260–264. 121. Mastrangelo, C.H., Thermal Application of Microbridges, Ph.D. Thesis, UC Berkeley, 1991. 122. Accorsi, A., Delapierre, G., Vauchier, C., Charlot, D., Sens. Actuators B, 1991, 4, 539–543. 123. Vandrish, G., Key Engineering Materials, 1996, 122–124, 225–232. 124. Xie, B., Danielson, B., Norberg, P., Winquist, F., Lundstroem, I., Sens. Actuators B, 1992, 127–130. 125. Danielsson, B., Mattiasson, B., Mosbach, K., Appl. Biochem. Bioeng. 1981, 3, 97–143. 126. Xie, B., Danielson, B., Technical Digest, Transducer 1995, Stockholm, Sweden, 501–503. 127. Lysenko, V., Delhomme, G., Soldatkin, A., Strikha, V., Dittmar, A., JaffrezicRenault, N., Martelet, C., Talanta, 1996, 43, 1163–1169. 128. Xie, B., Danielson, B., Winquist, F., Sens. Actuators B, 1993, 15–16, 443–447. 129. Shimohgoshi, M., Karube, I., Sens. Actuators B, 1996, 30, 17–21.
198
References
130. Polla, D.L., White, R.M., Muller, R.S., IEEE, Digest of Technical Papers, Transducers, 1985, 33–36. 131. Dorojkine, L.M., Mandelis, A., Opt. Eng. 1997, 36 (2), 473–481. 132. Van Herwaarden, A.W., Sarro, P.M., Sens. Actuators, 1986, 10, 321–346. 133. Sarro, P.M., van Herwaarden, A.W., van der Vlist, W., Sens. Actuators A, 1994, 41–42, 666–671. 134. Van Herwaarden, A.W., Sarro, P.M., Gardner, J.W., Bataillard, P., Sens. Actuators A, 1994, 43, 24–30. 135. Kerness, N., Koll, A., Schaufelbuehl, A., Hagleitner, C., Hierlemann, A., Brand, O., Baltes, H., Proc. IEEE Workshop on Micro Electro Mechanical Systems MEMS 2000, Myazaki, Japan, 2000, 96–101, ISBN 0-7803-5273-4. 136. Koll, A., Schaufelb¨ uhl, A., Brand, O., Baltes, H., Menolfi, C., Huang, H., Proc. IEEE Workshop on Micro Electro Mechanical Systems MEMS 99, Orlando, 1999, 547–551, ISBN 0-7803-5194-0. 137. Lerchner, J., Seidel, J., Wolf, G., Weber, E., Sens. Actuators B, 1996, 32, 71–75. 138. Caspary, D., Schr¨ opfer, M., Lerchner, J., Wolf, G., Thermochimica Acta 1999, 337, 19–26. 139. Lerchner, J., Wolf, A., Wolf, G., J. Thermal. Anal., 1999, 55, 212–223. 140. Bataillard, P., Steffgen, E., Haemmerli, S., Manz, A., Widmer, H.M., Biosensors, Bioelectronics, 1993, 8, 89–98. 141. K¨ ohler, J.M., Kessler, E., Steinhage, G., Gr¨ undig, B., Cammann, K., Mikrochim. Acta, 1995, 120, 309–319. 142. Sarro, P.M., Yashiro, H., van Herwaarden, A.M., Middelhoek, S., Sens. Actuators, 1988, 14, 191–201. 143. Van Herwaarden, A.W., Meas. Sci. Technol., 1992, 3, 935–937. 144. Papadopoulos, C.A., Vlachos, D.S., Avaritsiotis, J.N., Sens. Actuators B, 1996, 34, 524–527. 145. Swart, N.R., Stevens, M., Nathan, A., Karanassios, V., Sens. Materials, 1997, 9 (6), 387–394. 146. Kimura, M., Manaka, J., Satoh, S., Takano, S., Igarashi, N., Kazutoshi, N., Sens. Actuators B, 1995, 24–25, 857–860. 147. Wolfbeis, O., Boisde, G.E., Gauglitz, G., Optochemical Sensors in Sensors: A Comprehensive Survey, Vol. 2, and G¨opel, W., Hesse, J., Zemel, J.N. (eds.), VCH-Verlagsgesellschaft, Weinheim 1991, 573–646. 148. Gauglitz, G., Opto-Chemical and Opto-Immuno Sensors in Sensors: A Comprehensive Survey, Update Vol. 1, and G¨opel, W., Hesse, J., Zemel, J.N. (eds.), VCH-Verlagsgesellschaft, Weinheim 1996, 1–49. 149. Brecht, A., Gauglitz, G., G¨ opel, W., Sensor in Biomolecular Interaction Analysis and Pharmaceutical Drug Screening in Sensors: A Comprehensive Survey, Update Vol. 3, and G¨ opel, W., Hesse, J., Zemel, J.N. (eds.), VCHVerlagsgesellschaft, Weinheim 1998, 573–646. 150. Mulvaney, S.P., Keating, C.D., Anal. Chem., 2000, 72, 145R–158R. 151. http://www.affymetrix.com, http://www.apb.com, http://www.cepheid.com. 152. Blum, L.J. (ed.), Bio- and Chemoluminescent Sensors, World Scientific, Singapore, 1997. 153. Azzam, R.M.A., Bashara, N.M., Ellipsometry and Polarized Light, North Holland, New York, 1988. 154. Jin, G., Tengvall, P., Lundstr¨om, I., Arwin, H., Anal. Biochem. 1995, 232, 69–72.
References
199
155. Gauglitz, G., Brecht, A., Kraus, G., Nahm, W., Sens. Actuators B, 1993, 11, 21. 156. Piehler, J., Brecht, A., Gauglitz, G., Anal. Chem., 1996, 68, 139. 157. Klotz, A., Brecht, A., Barzen, C., Gauglitz, G., Harris, R.D., Quigley, Q.R., Wilkinson, J.S., Sens. Actuators B, 1998, 51, 181–187. 158. Lukosz, W., Sens. Actuators B, 1995, 29, 37–50. 159. Kretschmann, E., Raether, H., Z. Naturf., 1968, 23a, 2135. 160. Sensors and Actuators B, Special Issue devoted to SPR sensors, Yee, S.S. (ed.), 1999, Vol. 54 (1–2) and articles herein. 161. Homola, J., Yee, S.S., Gauglitz, G., Sens. Actuators B, 1999, 54, 3–15. 162. Ehrfeld, W., Bauer, H.D., Proc. of the SPIE, 1998, Vol. 3276, 2–14. 163. Spichiger-Keller, U.E., Chemical Sensors and Biosensors for Medical and Biological Application, Wiley-VCH, Weinheim, 1998. 164. Klimant, I., K¨ uhl, M., Glud, R.N., Holst, G., Sens. Actuators B, 1997, 38–39, 29–37. 165. Holst, G., Glud, R.N., K¨ uhl, M., Klimant, I., Sens. Actuators B, 1997, 38–39, 122–129. 166. Wolfbeis, O.S., Anal. Chem., 2000, 72, 81R–90R. 167. Dakin, J., Culshaw, B. (eds.), Optical Fiber Sensors Vols. 3 and 4, Artech House, Norwood, MA, 1997. 168. Boisde, G.E., Harmer, A. (eds.), Chemical and Biochemical Sensing with Optical Fibers and Waveguides, Artech House, Norwood, MA, 1996. 169. Dickinson, T.A., White, J., Kauer, J.S., Walt, D., Nature, 1996, 382, 697–700. 170. White, J., Kauer, J.S., Dickinson, T.A., Walt, D., Anal. Chem., 1996, 68, 2191–2202. 171. Hafeman, D.G., Parce, J.W., McConell, H.M., Science, 1988, 240, 1182–1185. 172. Sze, S.M., Physics of Semiconductor Devices, Wiley, New York, 2nd edition, 1981. 173. Lukosz, W., Stamm, C., Moser, H.R., Ryf, R., D¨ ubendorfer, J., Sens. Actuators B, 1997, 38–39, 316–323. 174. Tiefenthaler, K., Lukosz, W., J. Opt. Soc. Am. B, 1989, 6, 209–220, and Thin Solid Films, 1985, 126, 205–211. 175. Lukosz, W., Nellen, P.M., Stamm, C., Weiss, P., Sens. Actuators B, 1990, 1, 585–588. 176. Nellen, P.M., Lukosz, W., Sens. Actuators B, 1990, 1, 592–596. 177. Brosinger, F., Freimuth, H., Lacher, M., Ehrfeld, W., Gedig, E., Katerkamp, A., Spener, F., Cammann, K., Sens. Actuators B, 1997, 44, 350–355. 178. Busse, S., K¨ ashammer, J., Kr¨ amer, S., Mittler, S., Sens. Actuators B, 1999, 60, 148–154. 179. Schipper, E.F., Brugman, A.M., Dominguez, C., Lechuga, L.M., Kooyman, R.P., Greve, J., Sens. Actuators B, 1997, 40, 147–153. 180. Tiefenthaler, K., Lukosz, W., Sens. Actuators, 1988, 15, 273–284. 181. Lukosz, W., Stamm, C., Sens. Actuators A, 1991, 25, 185–188. 182. Stamm, C., Lukosz, W., Sens. Actuators B, 1993, 11, 177–181. 183. Heideman, R., Kooyman, R.P., Greve, J., Sens. Actuators B, 1993, 10, 209– 217. ¨ as, P., Honkanen, S., Grace, K.M., Shrouf, K., Katila, P., Leppihalme, M., 184. Ayr¨ Tervonen, A., Yang, X., Swanson, B., Peyghambarian, N., Pure Appl. Opt., 1998, 7, 1261–1271.
200
References
¨ as, P., Katila, P., Leppihalme, M., 185. Grace, K.M., Shrouf, K., Honkanen, S., Ayr¨ Johnston, R.G., Yang, X., Swanson, B., Electronis Letters, 1997, 33, 1651– 1653. 186. Gauglitz, G., Ingenhoff, J., Fres. J. Anal. Chem., 1994, 349, 355–359. 187. Puyol, M., del Valle, M., Garces, I., Villuendas, F., Dominguez, C., Alonso, J., Anal. Chem., 1999, 71, 5037–5044. 188. Heideman, R.G., Veldhuis, G.J., Jager, E.W.H., Lambeck, P.V., Sens. Actuators B, 1996, 35–36, 234–240. 189. Heideman, R.G., Lambeck, P.V., Sens. Actuators B, 1999, 61, 100–127. 190. Maisenh¨ older, B., Zappe, H.P., Kunz, R.E., Moser, M., Riel, P., Technical Digest Transducers Chicago, 1997, 79–80. 191. Maisenh¨ older, B., Zappe, H.P., Kunz, R.E., Riel, P., Moser, M., Edlinger, J., Sens. Actuators B, 1997, 38–39, 324–329. 192. Maisenh¨ older, B., Zappe, H.P., Moser, M., Riel, P., Kunz, R.E., Edlinger, J., Electronics Letters, 1997, 33, 986–988. 193. Hofstetter, D., Zappe, H.P., Epler, J.E., Riel, P., IEEE Photonics Technology Letters, 1995, 7, 1022–1024. 194. Hofstetter, D., Zappe, H.P., Epler, J.E., S¨ochtig, J., Electronics Letters, 1994, 30, 1858–1859. 195. Zappe, H.P., Hess, M., Moser, M., H¨ovel, R., Gulden, K., Gauggel, H.P., Monti di Sopra, F., Applied Optics, 2000, 39, 2475–2479. 196. Zappe, H.P., Sensors, 2000, 16 (9), 114–117. 197. Jerman, J.H., Clift, D.J., Technical Digest Transducers 1991, 372–375. 198. Grasdepot, F., Alause, H., Knap, W., Malzac, J.P., Suski, J., Sensor Act. B, 1996, 35–36, 377–380. 199. Alause, H., Grasdepot, F., Malzac, J.P., Knap, W., Hermann, J., Sensor Act. B, 1997, 43, 18–23. 200. Silveira, J.P., Anguita, J., Briones, F., Grasdepot, F., Bazin, A., Sens. Actuators B, 1998, 48, 305–307. 201. Correia, J.H., de Graaf, G., Kong, S.H., Bartek, M., Wolffenbuttel, R.F., Sens. Actuators A, 2000, 82, 191–197. 202. Aratani, K., French, P.J., Sarro, P.M., Poenar, D., Wolffenbuttel, R.F., Middelhoek, S., Sens. Actuators A, 1994, 43, 17–23. 203. De Graaf, G., Wolffenbuttel, R.F., Sens. Actuators A, 1998, 67, 115–119. 204. Han, J., Applied Physics Letters, 1999, 74, 445–447. 205. Han, J., Neikirk, D.P., Clevenger, M., McDevitt, J.T., SPIE Proceedings 1996, 2881, 171–179. 206. Chen, T., Sens. Actuators B, 1993, 13–14, 284–287. 207. Carson, H., Sensor Review, 1997, 17, 304–306, and http://www.vaisala.com. 208. Yee, G.M., Maluf, N.I., Hing, P.A., Albin, M., Kovacs, G.T., Sens. Actuators A, 1997, 58, 61–66. 209. Lee, S.S., Lin, L.Y., Wu, M.C., Appl. Phys. Lett., 1995, 67, 2135–2137. 210. Yee, G.M., Maluf, N.I., Kovacs, G.T., Technical Digest Transducers Sendai, 1999, 1882–1883. 211. Kwa, T.A., Wolffenbuttel, R.F., Sens. Actuators A, 1992, 31, 259–266. 212. Sinclair, M.B., Butler, M.A., Kravitz, S.H., Zubrizycki, W.J., Ricco, A.J., Optics Letters, 1997, 22, 1036–1038. 213. Simpson, M., Sayler, G., Nivens, D., Ripp, S., Paulus, M., Jellison, G., Trends in Biotechnology, 1998, 16, 332–338.
References
201
214. Simpson, M., Paulus, M., Jellison, G., Sayler, G., Applegate, B., Ripp, S., Nivens, D., Technical Digest Solid-State Sensor and Actuator Workshop, Hilton Head, 1998, 354–357. 215. Simpson, M., Sayler, G., Nivens, D., Ripp, S., Paulus, M., Jellison, G., SPIE proceedings 1998, 3328, 202–212. 216. Eres, G., Lowndes, D.H., US Patent 5 110 790, 1992. 217. Liedberg, B., Nylander, C., Lundstr¨om, I., Sens. Actuators, 1983, 4, 299–304. 218. Garabedian, R., Gonzalez, C., Richards, J., Knoesen, A., Spencer, R., Collins, S.D., Smith, R.L., Sensor Act. A., 1994, 43, 202–207. 219. Richards, J.D., Garabedian, R., Gonzalez, C., Knoesen, A., Smith, R.L., Spencer, R., Collins, S.D., Applied Optics 1993, 32, 2901–06. 220. http://www.ti.com/sc/docs/products/analog/tspr1a150100.html, and http:// www.ti.com/spreeta 221. Widmer, H., Anal. Methods Instrum., 1993, 1, 60–72. 222. Vasiliev, A., Moritz, W., Fillipov, V., Bartholom¨aus, L., Terentjev, A., Gabusjan, T., Sens. Actuators B, 1998, 49, 133–138. 223. Baranzahi, A., Spetz, A., Lundstr¨ om, I., Sens. Actuators B, 1995, 26–27, 165– 169. 224. Svennigstorp, H., Tobias, P., Lundstr¨om, I., Salomonsson, P., Martensson, P., Ekedahl, L.G., Lloyd Spetz, A, Sens. Actuators B., 1999, 57, 159–165. 225. Samman, A., Gebremariam, S., Rimai, L., Zhang, X., Hangas, J., Auner, G.W., Sens. Actuators B, 2000, 63, 91–102. 226. Fabry, P., Siebert, E., Electrochemical Sensors in The GRC Handbook of Solid State Electrochemistry, Gellings, P.J., Bouwmeester, H.J.M. (eds.), CRC Press, Boca Raton, FL, 1997. 227. Wollenberger, U., Electrochemical Biosensors-Ways to Improve Sensor Performance in Biotechnology and Engineering Reviews, Artech House, Norwood, UK, 1996, 13, 237–266. 228. Kas, J., Marek, M., Stastny, M., Volf, R., Bioelectrochem. Princ. Pract., 1996, 3, 361–453. 229. Kauffman, J.M. (ed.) Special Issue Bioelectrochem. Bioeng., 1997, 42. 230. Schultze, J.W., Tsakova, V., Electrochimica Acta, 1999, 44, 3605–3627. 231. Neuman, M.R., Buck, R.P., Cosofret, V., Lindner, E., Liu, C.C., IEEE Engineering in Medicine and Biology, 1994, 13, 409–419. 232. Buss, G., Sch¨ oning, M.J., L¨ uth, H., Schultze, J.W., Electrochimica Acta, 1999, 44, 3899–3910. 233. Schmitt, G., Schulze, J.W., Fassbender, F., Buss, G., L¨ uth, H., Sch¨ onung, M.J., Electrochimica Acta, 1999, 44, 3865–3883. 234. Sreenivas, G., Ang, S.S., Fritsch, I., Brown, W.D., Gerhadt, G.A., Woodward, D.J., nal. Chem., 1996, 68, 1858–1864. 235. Wittkampf, M., Chemnitius, G.C., Cammann, K., Rospert, M., Mokwa, W., Sens. Actuators B, 1997, 43, 40–44. 236. Desmond, D., Lane, B., Alderman, J., Hall, G., Alvarez-Icaza, M., Garde, A., Ryan, J., Barry, L., Svehla, G., Arrigan, D.W., Schniffner, L., Technical Digest Transducers Stockholm, 1995, Vol. 2, 948–951. 237. Kovacs, G.T.A., Storment, C.W., Kounaves, S.P., Sensor Act. B, 1995, 23, 41–47. 238. Belmont, C., Tercier, M.L., Buffle, J., Fiaccabrino, G.C., Koudelka-Hep, M., Anal. Chim. Acta, 1996, 329, 203–214.
202
References
239. Herdan, J., Feeney, R., Kounaves, S.P., Flannery, A.F., Storment, C.W., Kovacs, G.T.A., Environ. Sci. Technol., 1998, 32, 131–136. 240. Zhou, Z.B., Wu, Q.H., Liu, C.C., Sensor Act. B, 1994, 21, 101–108. 241. Maseeh, F., Tierney, M.J., Chu, W.S., Joseph, J., Kim, H.O., Otagawa, T., Technical Digest Transducers 1991, 359–362. 242. Buttner, W.J., Maclay, G.J., Stetter, J.R., Sensor Act. B, 1990, 1, 303–307. 243. Stetter, J.R., Hesketh, P.J., Maclay, G.J., Proc. Symposium on Chemical and Biological Sensors and Analytical Methods, 1997, ECS-Proceedings, Vol. 97– 19, 70–77. 244. Van den Berg, A., van der Waal, P.D., van der Schoot, B.H., de Rooij, N.F., Sensors Materials, 1994, 6, 23–43. 245. Steinkuhl, R., Sundermeier, C., Hinkers, H., Dumschat, C., Cammann, K., Knoll, M., Sens. Actuators B, 1996, 33, 19–24. 246. Yon Hin, B.F.Y., Sethi, R.S., Lowe, C.R., Sens. Actuators B, 1990, 1, 550–554. 247. Schwake, A., Ross, B., Cammann, K., Sens. Actuators B, 1998, 46, 242–248. 248. Dilhan, M., Est`eve, D., Gu´e, A.M., Mauvais, O., Mercier, L., Sens. Actuators B, 1995, 26–27, 401–403. 249. Van den Berg, A., Grisel, A., Verney-Norberg, E., van der Schoot, B.H., Koudelka-Hep, M., de Rooij, N.F., Sens. Actuators B, 1993, 13–14, 396–399. 250. Wittkampf, M., Cammann, K., Amrein, M., Reichelt, R., Sens. Actuators B, 1997, 40, 79–84. 251. Sansen, W., de Wachter, D., Callewaert, L., Lambrechts, M., Claes, A., Sens. Actuators B, 1990, 1, 298–302. 252. Koudelka, M., Sens. Actuators, 1986, 9, 249. 253. Hinkers, H., Sundermeier, C., L¨ urick, R., Walfort, F., Cammann, K., Knoll, M., Sens. Actuators B, 1995, 26–27, 398–400. 254. Hinkers, H., Hermes, T., Sundermeier, C., Borchardt, M., Dumschat, C., B¨ ucher, S., B¨ uhner, M., Cammann, K., Knoll, M., Sens. Actuators B, 1995, 24–25, 300–303. 255. Hermes, T., B¨ uhner, M., B¨ ucher, S., Sundermeier, C., Dumschat, C., Borchardt, M., Cammann, K., Knoll, M., Sens. Actuators B, 1994, 21, 33–37 256. Zhu, J., Wu, J., Tian, C., Wu, W., Zhang, H., Lu, D., Zhang, G., Sens. Actuators B, 1994, 20, 17–22. 257. Van den Berg, A., Grisel, A., Koudelka, M., van der Schoot, B.H., Sens. Actuators B, 1991, 5, 71. 258. Clark, L.C., Trans. Am. Soc. Artif. Int. Organs, 1956, 2, 41–48. 259. Schnakenberg, U., Lisec, T., Hintsche, R., Kuna, I., Uhlig, A., Wagner, B., Sens. Actuators B, 1996, 34, 476–480. 260. Hintsche, R., Kruse, C., Uhlig, A., Paeschke, M., Lisec, T., Schnakenberg, U., Wagner, B., Sens. Actuators B, 1995, 26–27, 471–473. 261. Nernst, W., Z. Phys. Chem., 1904, 47, 52. 262. Currie, J.F., Essalik, A., Marusic, J.C., Sens. Actuators B, 1999, 59, 235–241. 263. Goldberg, H.D., Brown, R.E., Liu, D.P., Meyerhoff, M.E., Sens. Actuators B, 1994, 21, 171–183. 264. Lauks, I., van der Spiegel, J. Sansen, W., Steyaert, M. Technical Digest Transducers, Philadelphia, 1985, 122–124. 265. Lang, T., Caron, M., Izquierdo, R., Ivanov, D., Currie, J.F., Yelon, A., Sens. Actuators B, 1996, 31, 9–12. 266. Knoll, M., Cammann, K., Dumschat, C., Sundermeier, C., Eshold, J., Sens. Actuators B, 1994, 18–19, 51–55.
References
203
267. Taillades, G., Valls, O., Bratov, A., Dominguez, C., Pradel, A., Ribes, M., Sens. Actuators B, 1999, 59, 123–127. 268. Lundstr¨ om, I., Sensors Act B, 1996, 56, 75–82. 269. Ekedahl, L.G., Eriksson, M., Lundstr¨om, I., Accounts of Chemical Research, 1998, 31, 249–256. 270. Hughes, R.C., Schubert, W.K., Zipperian, T.E., Rodriguez, J.L., Plut, T.A., J. Appl. Phys. 1987, 62, 1074–1083. 271. Spetz, A., Armgarth, M., Lundstr¨ om, I., J. Appl. Phys., 1988, 64, 1274–1283. 272. Spetz, A., Winquist, F., Sundgren, H., Lundstr¨ om, I., in Gas Sensors, and Sberveglieri, G. (ed.), Kluwer, Dordrecht, 1992, 219–279. 273. Thomas, R.C., Hughes, R.C., Journal of the Electrochemical Society, 1997, 144, 3245–3248. 274. Morita, Y., Nakamura, K., Kim, C., Sensors Act B, 1996, 33, 96–99. 275. Winquist, F., Spetz, A., Armgarth, M., Nylander, C., Lundstr¨om, I., Appl. Phys. Lett., 1983, 43, 839–841. 276. Winquist, F., Lundstr¨ om, I., Danielson, B., Anal. Chem., 1986, 58, 145–148. 277. Armgarth, M., Nylander, C., Appl. Phys. Lett., 1981, 39, 91–92. 278. Barker, P.S., Monkman, A.P., Petty, M.C., Pride, R., Synthetic Metals, 1997, 85, 1365–1366, and Sens. Actuators B, 1995, 24, 451. 279. Domansky, K., Baldwin, D.L., Grate, J.W., Hall, T.B., Li, J., Josowicz, M., Janata, J., Anal. Chem., 1998, 70, 473–481, and Polk, P.J., Smith, A.J., De Weerth, S.P., Zhou, Z., Janata, J., Domansky, K., Electroanalysis, 1999, 11, 707–711. 280. Gupta, R.P., Gergintschew, Z., Schipanski, D., Vyas, P.D., Sensors Act B, 1999, 56, 65–72, and 2000, 63, 35–41. 281. Moss, S.D., Janata, J., Johnson, C.C., Anal. Chem., 1975, 47, 2238–2234. 282. Kelly, R.G., Electrochimica Acta, 1977, 22, 1–8. 283. Bergveld, P., de Rooij, N.F., Sens. Actuators, 1981, 1, 5–15. 284. Kelly, R.G., Owen, A.E., IEE Proceedings, 1985, 132, 227–236. 285. Katsube, T. Lauks, I., Zemel, J.N., Sens. Actuators 2, 1982, 399. 286. Desmond, D., Lane, B., Alderman, J., Glennon, J.D., Diamond, D. Arrigan, D.W., Sens. Actuators B, 1997, 44, 389–396. 287. Dumschat, C., Felcmann, C., P¨otter, W., Cammann, K., Technical Digest, Transducers, Stockholm, 1995, Vol. 2, 970–972. 288. Van den Berg, A., Grisel, A., van den Vlekkert, H.H., de Rooij, N.F., Sens. Actuators, B, 1990, 1, 425–432. 289. Bousse, L.J., Bergveld, P., Geeraedts, H.J., Sens. Actuators, 1986, 9, 179–197. 290. Siu, W.M., Cobbold, R.S., IEEE Trans., 1979, ED–26, 1805–1815. 291. Buck, R.P., Hackleman, D.E., Anal. Chem., 1977, 49, 2315–2321. 292. Janata, J., Sens. Actuators, 1983, 4, 255–265. 293. Lauks, I., Sens. Actuators, 1981, 1, 261–288. 294. Bousse, L., Bergveld, P., Sens. Actuators, 1984, 6, 65. 295. Van den Berg, A., Bergveld, P., Reinhoudt, D.N., Sudholter, E.J.R., Sens. Actuators, 1985, 8, 129. 296. Hoffmann, W., Rapp, R., Sens. Actuators B, 1996, 34, 471–475. 297. Martinoia, S., Lorenzeli, L., Massobrio, G., Margesin, B., Lui, A., Sensors and Materials, 1999, 11, 279–295. 298. Yeow, T.C., Haskard, M.R., Mulcahy, D.E., Seo, H.I., Kwon. D.H., Sens. Actuators, B, 1997, 44, 434–440.
204
References
299. Rapp, R., Hoffmann, W., S¨ uss, W., Ache, H.J., G¨olz, H., Electrochimica Acta, 1997, 42, 3391–3398. 300. Can´e, C., G¨ otz, A., Merlos, A., Gr` acia, I., Errachid, A., Losantos, P., LoraTamayo, E., Sens. Actuators, B, 1996, 35–36, 136–140. 301. Sakai, T., Amemiya, I., Uno, S., Katsura, M., Sens. Actuators, B, 1990, 1, 341–344. 302. Cambiaso, A., Chiarugi, S., Grattarola, M., Lorenzelli, L., Lui, A., Margesin, B., Martinoia, S., Zanini, V., Zen, M., Sens. Actuators, B, 1996, 34, 245–251. 303. Bausells, J., Carrabina, J., Errachid, A., Merlos, A., Sens. Actuators, B, 1999, 57, 56–62. 304. Gimmel, P. Schierbaum, K.D., G¨opel, W., van den Vlekkert, H.H., de Rooij, N.F., Sens. Actuators, B, 1990, 1, 345–349. 305. Olthuis, W., Luo, J., van der Schoot, B., Bomer, J.G., Bergveld, P., Sens. Actuators, B, 1990, 1, 416–420. 306. Hendrikse, J., Olthuis, W., Bergveld, P., Sens. Actuators, B, 1998, 53, 97–103. 307. Johnson, C.C., Moss, D., Janata, J., US Patent No. 4020830, 1977. 308. Van der Wal, P., Skowronska-Ptasinska, M., van den Berg, A., Bergveld, P., Sudholter, E.J.R., and Reinhoudt, D.N., Anal. Chim. Acta, 1990, 231, 41. 309. Blackburn, G.F., and Janata, J., J. Electrochem. Soc., 1982, 129, 2580. 310. Harrison, D.J., Teclemariam, A., and Cunningham, L., Anal. Chem, 1989, 61, 246. 311. Sudholter, E.J.R., van der Wal, P., Skowronska-Ptasinska, M., van den Berg, A., Bergveld, P., and Reinhoudt, D.N., Anal. Chim. Acta, 1990, 230, 59. 312. Brz´ ozka, Z., Holterman, H.A.J., Honig, G.W.N., Verkerk, U.H., van den Vlekkert, H.H., Engbersen, J.F.J., and Reinhoudt, D.N., Sens. Actuators B, 1994, 18–19, 38. 313. Reinhoudt, D.N., Engbersen, J.F.J., Brz´ ozka, Z., van den Vlekkert, H.H., Honig, G.W.N., Holterman, H.A.J., and Verkerk, U.H., Anal. Chem., 1994, 66, 3618. 314. Van der Wal, P.D., Sudholter, E.J.R., and Reinhoudt, D.N., Anal. Chim. Acta, 1991, 245, 159. 315. Park, L.S., Hur, Y.J., Sohn, B.K., Sens. Actuators B, 1996, 57, 239–243. 316. Brunink, J.A.J., Haak, J.R., Bomer, J.G., Reinhoudt, D.N., McKervey, M.A., and Harris, S.J., Anal. Chim. Acta, 1991, 254, 75. 317. Brunink, J.A.J., Bomer, J.G., Engbersen, J.F.J., Verboom, W., and Reinhoudt, D.N., Sens. Actuators B, 1993, 15–16, 195. 318. Tsukuda, K., Miyahara, Y., Shibata, Y., Miyagi, H., Technical Digest Transducers, 1991, 218–221. 319. Brz´ ozka, Z., Cobben, P.L.H.M., Reinhoudt, D.N., Edema, J.J.H., Buter, J., and Kellogg, R.M., Anal. Chim. Acta, 1993, 273, 139. 320. Cobben, P.L.H.M., Egberink, R.J.M., Bomer, J.G., Bergveld, P., Verboom, W., and Reinhoudt, D.N., J. Am. Chem. Soc., 1992, 114, 10573. 321. Cobben, P.L.H.M., Egberink, R.J.M., Bomer, J.G., Haak, J.R., Bergveld, P., and Reinhoudt, D.N., Sens. Actuators B, 1992, 6, 304. 322. Battilotti, M., Mercuri, R., Mazzamuro, G., Giannini, I., Giongo, M., Sens. Actuators B, 1990, 1, 438–440. 323. Dumschat, C., Fromer, R., Rautschek, H., M¨ uller, H., and Timpe, H.-J., Anal. Chim. Acta, 1991, 243, 179 and Sens. Actuators B, 1990, 2, 271–276. 324. Rocher, V., Jaffrezic-Renault, N., Perrot, H., Chevalier, Y., and Le Perchec, P., Anal. Chim. Acta, 1992, 256, 251.
References
205
325. Antonisse, M.G., Snellink-Ru¨el, B.H., Lugtenberg, R.J., Engbersen, J.F., van den Berg, A., Reinhoudt, D.N., Anal. Chem., 2000, 72, 343–348. 326. Caras, S.D., Janata, J., Technical Digest Transducers, 1985, 158–161. 327. Colapicchioni, C., Barbaro, A., Porcelli, F., Sens. Actuators B, 1992, 13, 202– 207. 328. Kuriyama, T., Technical Digest Transducers, Stockholm, 1995, Vol.1, 447–449. 329. Katsube, T., Katoh, M., Maekawa, H., Hara, M., Yamaguchi, S., Uchida, N., Shiomura, T., Sens. Actuators B, 1990, 1, 504–507. 330. Kimura, J., Murakami, T., Kuriyama, T., Karube, I., Sens. Actuators, 1988, 15, 435–443. 331. Janata, J., Huber, R.J., Cohen, R., Kolesar, E.J., Aviat. Space Environ. Med., 1981, 52, 666–671. 332. Campanella, L., Colapicchioni, C., Favero, G., Sammartino, M.P., Tomassetti, M., Sens. Actuators B, 1996, 33, 25–33. 333. Wan, K., Chovelon, J.M., Jaffrezic-Renault, N., Soldatkin, A.P., Sens. Actuators B, 1999, 58, 399–408. 334. Dumschat, C., M¨ uller, H., Stein, K., Schwedt, G., Anal. Chim. Acta, 1991, 252, 7–9. 335. Vianello, F., Stefani, A., DiPaolo, M.L., Rigo, A., Lui, A., Margesin, B., Zen, M., Scarpa, M., Soncini, G., Sens. Actuators B, 1996, 37, 49–54. 336. Baumann, W.H., Lehmann, M., Schwinde, A., Ehret, R., Brischwein, M., Wolf, B., Sens. Actuators B, 1999, 55, 77–89. 337. Fromherz, P., Offenh¨ auser, A., Vetter, T., Weis, J., Science, 1991, 252, 1290. 338. Offenh¨ auser, A., Spr¨ ossler, C., Matsuszawa, M., Knoll, W., Biosensors Bioelectronics, 1997, 12, 819–826. 339. Sch¨ oning, M.J., Sch¨ utz, S., Schroth, P., Weissbecker, B., Steffen, A., Kordos, P., Hummel, H.E., L¨ uth, H., Sens. Actuators B, 1998, 47, 235–238. 340. Comte, P.A., Janata, J., Anal. Chim. Acta, 1978, 101, 247–252. 341. Van den Berg, A., Grisel, A., van den Vlekkert, H.H., de Rooij, N.F., Sens. Actuators B, 1990, 1, 395–400 and 425–252. 342. Bergveld, P., van den Berg, A., van der Wal, P.D., Skowronska-Ptasinska, M., Sudh¨ olter, E.J., Sens. Actuators, 1989, 18, 309–327. 343. Matsuo, T., Nakajiama, H., Sens. Actuators, 1984, 5, 425–435. 344. Can´e, C., Gr` acia, I., Merlos, A., Lozano, M., Lora-Tamayo, E., Esteve, J., Technical Digest Transducers 1991, 225–228. 345. Skowro˜ nska-Ptasi˜ nska, M., van der Wal, P., van den Berg, A., Bergveld, P., Sudholter, E.J.R., and Reinhoudt, D.N., Anal. Chim. Acta, 1990, 230, 67. 346. Baccar, Z.M., Jaffrezic-Renault, N., Martelet, C., Jaffrezic, H., Marest, G., Plantier, A., Sens. Actuators B, 1996, 32, 101–105. 347. Wilhelm, D., Voigt, H., Treichel, W., Ferretti, R., Prasad, S., Sens. Actuators B, 1991, 4, 145–149. 348. Shin, J.H., Lee, H.J., Kim, C.Y., Oh, B.K., Rho, K.L., Nam, H., Cha, G.S., Anal. Chem., 1996, 68, 3166–3172. 349. Nakajima, H., Esashi, M., and Matsuo, T., J. Electroanal. Chem. Soc., 1982, 129, 141. 350. Fujihira, M., Fukui, M., and Osa, T., J. Electroanal. Chem., 1980, 106, 413. 351. Tahara, S., Yoshii, M., and Oka, S., Chem. Lett., 1982, 307. 352. Matsuo, T., and Nakajima, H., Sens. Actuators, 1984, 5, 293. 353. Kimura, J., Kuriyama, T., and Kawana, Y., Sens. Actuators, 1980, 9, 373.
206
References
354. Steele, M.C., MacIver, B.A., Appl. Phys. Lett., 1976, 28, 687–788. 355. Shivamaran, M.S., Lundstr¨ om, I., Svensson, C., Hammarsten, H., Electron. Lett., 1976, 12483–484. 356. Ito, K., Surface Sci., 1979, 86, 345–352. 357. Yamamoto, M., Tonomura, S., Matsuoka, T., Tsubomura, H., Surface Sci., 1980, 92, 400–406. 358. Liu, C.C., Materials, Chemistry and Physics, 1995, 42, 87–90. 359. Wu, Q.H., Lee, K.M., Liu, C.C., Sens. Actuators B, 1993, 13–14, 1. 360. Zemel, J., US Patent 4103227, 1978. 361. Wen, C.C., Lauks, I., Zemel, J.N., Thin Solid Films, 1980, 70, 333. 362. Dura, H.G., Sch¨ oneberg, U., Mokwa, W., Hostika, B.J., Vogt, H., Sens. Actuators B, 1992, 6, 162–164. 363. Westcott, L., Rogers, G., J. Phys. E. Sci. Intrum. 1985, 18, 577–581. 364. Sch¨ oning, M.J., Thust, M., M¨ uller-Veggian, M., Kordos, P., L¨ uth, H., Sens. Actuators B, 1998, 47, 225–230. 365. Vogel, A., Hoffmann, B., Sauer, t., Wegner, G., Sens. Actuators B, 1990, 1, 408–411. 366. Sch¨ oning, M.J., Ronkel, F., Crott, M., Thust, M., Schultze, J.W., Kordos, P., L¨ uth, H., Electrochimica Acta, 1997, 42, 3185–3193. 367. Gabusjan, T., Bartolom¨ aus, L., Moritz, W., Sensors Materials, 1998, 10, 263– 273. 368. Janata, J., Josowicz, M., Anal. Chem., 1997, 69, 293A–296A. 369. Trasatti, S., Parsons, R., Pure and Applied Chem., 1986, 58, 37. 370. Bergstrom, P.L., Patel, S.V., Schwank, J.W., Wise, K.D., Sens. Actuators B, 1997, 42, 195–204, and Technical Digest, Transducers, Stockholm, 1995, Vol. 2, 993–996. 371. Josowicz, M., Janata, J., in Chemical Sensor Technology, Seiyama, T. (ed.), Elsevier, Amterdam, 1988, 153–177. 372. Scharnagl, K., B¨ ogner, M., Fuchs, A., Winter, R., Doll, T., Eisele, I., Sens. Actuators B, 1999, 57, 35–38. 373. B¨ ogner, M., Fuchs, A., Scharnagl, K., Winter, R., Doll, T., Eisele, I., Sens. Actuators B, 1998, 47, 145–152. 374. Lorenz, H., Peschke, M., Riess, H., Janata, J., Eisele, I., Sens. Actuators A, 1990, 21–23, 1023–1026. 375. Peschke, M., Lorenz, H., Riess, H., Eisele, I., Sens. Actuators B, 1990, 1, 21–24. 376. Gergintschew, Z., Kornetzky, P., Schipanski, D., Sens. Actuators B, 1996, 35– 36, 285–289. 377. Fuchs, A., B¨ ogner, M., Doll, T., Eisele, I., Sens. Actuators B, 1998, 48, 296– 299. 378. Domansky, K., Baldwin, D.L., Grate, J.W., Hall, T.B., Li, J., Josowicz, M., Janata, J., Anal. Chem., 1998, 70, 473–481. 379. Josowicz, M., Blackwood, D.J., J. Phys. Chem., 1991, 95, 493. 380. Ito, Y., Morimoto, K., Tsunoda, Y., Sens. Actuators B, 1993, 13–14, 348–350. 381. Adami, M., Sartore, M., Rapallo, A., Nicolini, C., Sens. Actuators B, 1992, 7, 343–346. 382. Adami, M., Sartore, M., Baldini, E., Rossi, A., Nicolini, C., Sens. Actuators B, 1992, 9, 25–31. 383. Owicki, J.C., Bousse, L.J., Hafeman, D.G., Kirk, G.L., Olson, J.D., Wada, H.G., Parce, J.W., Ann. Rev. Biophys. Biomol. Struct., 1994, 23, 87–113.
References
207
384. Parak, W.J., Hofmann, U.G., Gaub, H.E., Owicki, J.C., Sens. Actuators B, 1997, 63, 45–57. 385. Ito, Y., Sens. Actuators B, 1998, 107–111. 386. Fanigliulo, A., Accossato, P., Adami, M., Lanzi, M., Marinoia, S., Paddeu, S., Parodi, M.T., Rossi, A., Sartore, M., Grattarola, M., Nicolini, C., Sens. Actuators B, 1996, 32, 41–48. 387. Tanaka, H., Yoshinobu, T., Iwasaki, H., Sens. Actuators B, 1999, 59, 21–25. 388. Gavazzo, P., Paddeu, S., Sartore, M., Nicolini, C., Sens. Actuators B, 1994, 18–19, 368–372. 389. Bousse, L., Sens. Actuators B, 1996, 34, 270–275. 390. Owicki, J.C., Parce, W., Biosensors Bioelectronics, 1992, 7, 255–272. 391. Lee, W.E., Thompson, H.G., Hall, J.G., Bader, D.E., Biosensors Bioelectronics, 2000, 14, 795–804. 392. http://www.moleculardevices.com/ 393. Zhang, T.H., Petelenz, D., Janata, J., Sens. Actuators B, 1993, 12, 175–180. 394. Wong, H.S., White, M.H., IEEE Transactions on Electron Devices, 1989, 36, 479–487. 395. Sibbald, A., Shaw, J.E., Sens. Actuators, 1987, 12, 297–300. 396. Osorio-Saucedo, R., Luna-Arredondo, E.J., Calleja-Arriaga, W., ReyesBarranca, M.A., Sens. Actuators B, 1996, 37, 123–129. 397. Sakai, T., Amemiya, I., Uno, S., Katsura, M., Sens. Actuators B, 1990, 1, 341–344. 398. DuraFET, see http://www.honeywell.com 399. G¨ opel, W., Reinhardt, G., Metal Oxide Sensors in Sensors Update Vol. 1, and Baltes, H., G¨ opel, W., Hesse, J. (eds.), VCH Weinheim, 1996, 49–120. 400. Jones, T.A., Bott, B., Sens. Actuators 1984, 5, 43. 401. Bott, B. Jones, T.A., Sens. Actuators 1986, 9, 27. 402. Jones, T.A., Bott, B., Hurst, N.W., Mann, B., Surface Science, 1980, 92, 90. 403. Wohltjen, H., Barger, W.R., Snow, A.W., Jarvis, N.L., IEEE Trans. Electron. Dev., 1985, ED-32, 1170. 404. Nylander, C., Armgarth, M., Lundstr¨ om, I., Surface Science, 1980, 92, 400. 405. Bartlett, P.N., Ling-Chung, S.K., Sens. Actuators, 1989, 19, 125–140, and 141–150, as well as 1989, 20, 287–292. 406. Persaud, K., Dodd, G.H., Nature, 1982, 299, 352–355. 407. Persaud, K.C., Pelosi, P., Trans. Am. Soc. Artif. Intern. Organs, 1985, 31, 297–300. 408. Persaud, K.C., Pelosi, P., Sensor Arrays Using Conducting Polymers in Sensors and Sensory Systems for an Electronic Nose, Gardner, J.W., Bartlett, P.N. (eds.), Kluwer, Dordrecht, 1992. 409. Bartlett, P.N., Gardner, J.W., Odor Sensor for an Electronic Nose in Sensors and Sensory Systems for an Electronic Nose, Gardner, J.W., Bartlett, P.N. (eds.), Kluwer, Dordrecht, 1992. 410. Topart, P., Josowicz, M., J. Phys. Chem., 1992, 96, 7824–30. 411. Gardner, J.W., Pearce, T.C., Friel, S., Bartlett, P.N., Blair, N., Sens. Actuators B, 1994, 18–19, 240–243. 412. Gardner, J.W., Bartlett, P.N., Synth. Met., 1993, 55–57, 3665–70. 413. Cole, M., Gardner, J.W., Lim, A.W.Y., Scivier, P.K., Brignell, J.E., Sens. Actuators B, 1999, 58, 518–525. 414. Bartlett, P.N., Eliott, J.M., Gardner, J.W., Measurement and Control, 1997, 30, 273–278.
208
References
415. Gardner, J.W., Vidic, M., Ingleby, P., Pike, A.C., Brignell, J.E., Scivier, P., Bartlett, P.N., Duke, A.J., Elliott, J.M., Sens. Actuators B, 1998, 48, 289–295. 416. http://www.osmetech.plc.uk/, http://www.marconitech.com/. 417. Lonergan, M.C., Severin, E.J., Doleman, B.J., Beaber, S.A., Grubbs, R.H., Lewis, N.S., Chem. Mater., 1996, 8, 2298–2312. 418. Severin, E.J., Doleman, B.J., Lewis, N.S., Anal. Chem., 2000, 72, 658–668. 419. Doleman, B.J., Lonergan, M., Severin, E.J., Vaid, T.P., Lewis, N.S., Anal. Chem., 1998, 70, 4177–4190. 420. Patel, S.V., Jenkins, M.W., Hughes, R.C., Yelton, W.G., Ricco, A.J., Anal. Chem., 2000, 72, 1532–1542. 421. Eastman, M.P., Hughes, R.C., Yelton, W.G., Ricco, A.J., Patel, S.V., Jenkins, M.W., J. Electrochem. Soc., 1999, 146, 3907–13. 422. http://cyranosciences.com/. 423. Sheppard, N.F., Lesho, M.J., McNally, P., Francomarco, A.S., Sens. Actuators B, 1995, 28, 95–102, and Sens. Actuators B, 1996, 37, 61–66. 424. Sheppard, N.F., Tucker, R.C., Salehi-Had, S., Sens. Actuators B, 1993, 10, 73–77. 425. Varlan, A.R., Sansen, W., Sens. Actuators B, 1997, 44, 309–315. 426. Gardner, J.W., Pike, A., de Rooij, N.F., Koudelka-Hep, M., Clerc, P.A., Hierlemann, A., G¨ opel, W., Sens. Actuators B, 1995, 26,135–139. 427. Beckmann, F., Marschner, J., Hofmann, T., Laur, R., Sens. Actuators B, 1997, 62, 734–738. 428. Suehle, J.S., Cavicchi, R.E., Gaitan, M., Semancik, S., IEEE Electron Device Letters, 1993, 14, 118–120. 429. Vandrish, G., Key Engineering Materials, 1996, 122–124, 185–224. 430. Barsan, N., Schweizer-Berberich, M., G¨opel, W., Fresenius J. Anal. Chem., 1999, 365, 287–304. 431. Geistlinger, H., Sens. Actuators B, 1993, 17, 47–60. 432. Heiland, G., Kohl, D. in Chemical Sensor Technology, Seiyama, T. (ed.), Elsevier, Amsterdam, 1988, Vol. 1. 433. Kohl, D., Sens. Actuators B, 1990, 1, 158–165. 434. Lee, A.P., Reedy, B.J., Sens. Actuators B, 1999, 60, 35–42. 435. Yea, B., Osaki, T., Sugahara, K., Konishi, R., Sens. Actuators B, 1997, 41, 121–129. 436. Heilig, A., Barsan, N., Weimar, U., Schweizer-Berberich, M., Gardner, J.W., G¨ opel, W., Sens. Actuators B, 1997, 43, 45–51. 437. Corcoran, P., Lowery, P., Anglesa, J., Sens. Actuators B, 1998, 48, 448–455. 438. Sberveglieri, G., Hellmich, W., M¨ uller, G., Microsystem Technologies, 1997, 3, 183–190. 439. Demarne, V., Grisel, A., Sens. Actuators, 1988, 13, 301–313. 440. Faglia, G., Comini, E., Pardo, M., Taroni, A., Cardinali, G., Nicoletti, S., Sberveglieri, G., Microsystem Technologies, 1999, 6, 54–59. 441. Park, H.S., Shin, H.W., Yun, D.H., Hong, H-K., Kwon, C.H., Lee, K., Kim, S-T., Sens. Actuators B, 1995, 24–25, 478–481. 442. G¨ otz, A., Gr` acia, I., Can´e, C., Lora-Tamayo, E., Horrillo, M.C., Getino, G., Garc´ıa, C., Guti´errez, J., Sens. Actuators B, 1997, 44, 483–487. 443. Semancik, S., Cavicchi, R.E. Accounts of Chemical Research, 1998, 31, 279– 287. 444. Poirier, G.E., Cavicchi, R.E., Semancik, S., J. of Vacuum Science and Technology A, 1993, 11, 1392–1395 and 1994, 12, 2149–2152.
References
209
445. Frietsch, Dimitrakopoulos, L.T., Schneider, T., Goschnick, J., Surface Coatings Technology, 1999, 120–121, 265–271. 446. Semancik, S., Cavicchi, R.E., Kreider, K.G., Suehle, J.S., Chaparala, P., Sens. Actuators B, 1996, 34, 209–12. 447. Sberveglieri, G. Faglia, G., Gropelli, S., Nelli, P., Camanzi, A., Semicond. Sci. Technol., 1990, 5, 1231. 448. DiMeo, F., Cavicchi, R.E., Semancik, S., Suehle, J.S., Tea, N.H., Small, J., Armstrong, J.T., Kelliher, J.T., J. Vac. Sci. Technol. A, 1998, 16, 131–138. 449. Schweizer-Berberich, M., Zheng, J.G., Weimar, U., G¨opel, W., Barsan, N., Pentia, E., Tomescu, A., Sens. Actuators B, 1996, 31, 1–5. 450. Briand, D., Krauss, A., van der Schoot, B., Weimar, U., Barsan, N., G¨opel, W., de Rooij, N.F., Sens. Actuators B, 2000, 68, 223–233. 451. Majoo, S., Gland, J.L., Wise, K.D., Schwank, J.W., Sens. Actuators B, 1996, 35–36, 312–319. 452. Chung, W.Y., Shim, C.H., Choi, S.D., Lee, D.D., Sens. Actuators B, 1994, 20, 139–143. 453. Althainz, P., Goschnick, J., Ehrmann, S., Ache, H.J., Sens. Actuators B, 1996, 33, 72–76. 454. Althainz, P., Dahlke, A., Frietsch-Klarhof, M., Goschnick, J., Ache, H.J., Sens. Actuators B, 1995, 24–25, 366–369. 455. Swart, N.R., Nathan, A., IEEE Proc. Solid State Sensor and Actuator Workshop, Hilton Head, 1994, 119–122. 456. Bosc, J.M., Guo, Y., Sarihan, V., Lee, T., IEEE Transactions on Reliability, 1998, 47, 135–141. 457. Bosc, J.M., Odile, J.P., Microelectron. Reliab., 1997, 37, 1791–1794. 458. Pike, A., Gardner, J.W., Sens. Actuators B, 1997, 45, 19–26. 459. Lee, D-D., Chung, W-Y., Choi, M-S., Baek, J-M., Sens. Actuators B, 1996, 33, 147–150. 460. Dumitescu, M., Cobianu, C., Lungu, D., Dascalu, D., Pascu, A., Kolev, S., van den Berg, A., Sens. Actuators A, 1999, 76, 51–56. 461. http://www.figarosensor.com/, http://www.microchemical.com/ http://www.marconitech.com/, http://www.capteur.co.uk/. 462. Horrillo, M.C., Sayago, I., Ar´es, L., Rodrigo, J., Guti´errez, J., G¨ otz, A., Gr` acia, I., Fonseca, L., Can´e, C., Lora-Tamayo, E., Sens. Actuators B, 1999, 58, 325– 329. 463. Cavicchi, R.E., Suehle, J.S., Kreider, K.G., Gaitan, M., Chaparala, P., Sens. Actuators B, 1996, 33, 142–146. 464. Kunt, T.A., McAvoy, T.J., Cavicchi, R.E., Semancik, S., Sens. Actuators B, 1998, 53, 24–43. 465. Demarne, V., Balkanova, S., Grisel. A., Rosenfeld, D., L´evy, F., Sens. Actuators B, 1993, 14–15, 497–498. 466. Majoo, S., Schwank, J.W., Gland, J.L., AIChE Journal, Ceramics Processing, 1997, 43, 2760–2765. 467. Patel, S.V., Wise, K.D., Gland, J.L., Zanini-Fisher, M., Schwank, J.W., Sens. Actuators B, 1997, 42, 205–215. 468. Atkinson, J., Cranny, A., Simonis de Cloke, C., Sens. Actuators B, 1998, 47, 171–180. 469. Sheppard, N.F., Day, D.R., Lee, H.L., Senturia, S.D., Sens. Actuators, 1982, 2, 263–274.
210
References
470. Senturia, S.D., Technical Digest Transducers 1985, 198–201. 471. Glenn, M.C., Schuetz, J.A., Technical Digest Transducers 1985, 217–219. 472. Denton, D.D., Senturia, S.D., Anolick, E.S., Scheider, D., Technical Digest Transducers 1985, 202–205. 473. Delapierre, G., Grange, H., Chambaz, B., Destannes, L., Sens. Actuators, 1983, 4, 97–104. 474. Shibata, H., Ito, M., Asakursa, M., Watanabe, K., IEEE Transactions on Instrumentation and Measurement, 1996, 45, 564–569. 475. Hagleitner, C., Koll, A., Vogt, R., Brand, O., Baltes, H., Technical Digest Transducers Sendai, Japan, 1999, Vol. 2, 1012–1015. 476. Koll, A., Kummer, A., Brand, O., Baltes, H. Proc. SPIE Smart Structures and Materials, Newport 1999, Vol. 3673, 308–317. 477. Hille, P., Strack, H., Sens. Actuators A, 1992, 32, 321–325. 478. Cornila, C., Hierlemann, A., Lenggenhager, R., Malcovati, P., Baltes, H., Noetzel, G., Weimar, U., G¨opel, W., Sens. Actuators B, 1995, 24–25, 357–361. 479. Steiner, F.P., Hierlemann, A., Cornila, C., Noetzel, G., B¨achtold, M., Korvink, J.G., G¨ opel, W., Baltes, H., Technical Digest Transducers 1995, Vol. 2, 814– 817. 480. Van Gerwen, P., Laureys, W., Huyberechts, G., Op De Beeck, M., Baert, K., Suls, J., Varlan, A., Sansen, W., Hermans, L., Mertens, R., Technical Digest Transducers 1997, Vol. 2, 907–910. 481. Mutschall, D., Obermeier, E., Sens. Actuators B, 1995, 24–25, 412–414. 482. Boltshauser, T., Baltes, H., Sens. Actuators A, 1991, 25–27, 509–512. 483. Boltshauser, T., Chandran, L., Baltes, H., Bose, F., Steiner, D., Sens. Actuators B, 1991, 5, 161–164. 484. http://www.sensirion.com/, http://www.vaisala.com/, http://www. humirel. com/. 485. Kang, U., Wise, K.D., Technical Digest Solid State Sensor and Actuator Workshop, Hilton Head, 1998, 183–186. 486. Endres, H.E., Drost, S., Sens. Actuators B, 1991, 4, 95–98. 487. Casalini, R., Kilitziraki, M., Wood, D., Petty, M.C., Sens. Actuators B, 1999, 56, 37–44. 488. Dickert, F.L., Zwissler, G.K., Obermeier, E., Ber. Bunsenges. Phys. Chem., 1993, 97, 184–188. 489. Ishihara, T., Fujita, H., Takita, Y., Sens. Actuators B, 1998, 52, 100–106. 490. Lin, J., M¨ oller, S., Obermeier, E., Sens. Actuators B, 1991, 5, 219–221. 491. Balkus, K.J., Ball, L.J., Gimon-Kinsel, M.E., Anthony, J.M., Gnade, B.E., Sens. Actuators B, 1997, 42, 67–79. 492. Endres, H.E., Hartinger, R., Schwaiger, M., Gmelch, G., Roth, M., Sens. Actuators B, 1999, 57, 83–87. 493. Hierold, C., Proc. IEEE MEMS, 2000, Interlaken, CH, p. 1–6. and Infineon AG online: http://www.infineon.com. 494. Bitko, G., McNeil, A., Frank, R., Sensors Mag., July 2000. 495. Robert Bosch GmbH online: http://www.bosch.com. 496. Austriamicrosystems online: http:// www.austriamicrosystems.com. 497. Misevich, K.W., IEEE Trans. Industrial and Control Instrumentation, 16, 1969, 6–12. 498. Regtien, P.P.L., Sens. Actuators, 2, 1981, 85–95. 499. Koll, A., CMOS Capacitive Chemical Microsystems for Volatile Organic Compounds, Ph.D. thesis ETH Zurich Nr. 13460, 1999.
References
211
500. Hagleitner, C., Hierlemann, A., Baltes, H., Proceedings of IEEE Sensors 2002, First IEEE International Conference on Sensors, Orlando, FL, 2002, 1428– 1431. 501. Hagleitner, C., Hierlemann, A., Baltes, H., CMOS Single-chip Gas Detection Systems Part II, Sensors Update Vol. 12, Series Editors H. Baltes, J. Korvink, G. Fedder, Wiley VCH Weinheim, New York, 2003, 51–120. 502. Souloff, B., Advanced Microsystems for Automotive Applications 99, 1999, 262–270. 503. Tartagni, M., Guerrieri, R., IEEE J. Solid State Circuits, 33, 1998, 133–142. 504. Jachowicz, R.S., Senturia, S.D., Sens. Actuators, 2, 1981/82, 171–186. 505. Datasheet ADXL 105, Analog Devices, http://www.analog.com, 1999. 506. Sherman, S.J., Tsang, W.K., Core, T.A., Payne, R.S., Quinn, D.E., Chau, K.H.-L., Farash, J.A., Baum, S.K., IEEE Int. Electron Device Meeting, 1992, 501–504. 507. Lemkin, M., Boser, B.E., IEEE J. Solid State Circuits, 34, 1999, 456–468. 508. Van der Goes, F.M.L., Meijer, G.C.M., IEEE Trans. Instrumentation and Measurement, 45, 1996, 536–540. 509. Kung, J.T., Lee, H.-S., Howe, R.T., IEEE J. Solid State Circuits, 23, 1988, 972–977. 510. Brandl, M., Kempe, V., Proc. IEEE MEMS’01, 2001, 6–9. 511. Cho, S.T., Wise, K.D., Sens. Actuators A, 36, 1993, 47–56. 512. Gregorian, R., Temes, G.C., Analog MOS Integrated Circuits for Signal Processing, John Wiley & Sons, 1986. 513. Malcovati, P., CMOS Thermoelectric Sensor Interfaces, Ph.D. thesis ETH Zurich Nr. 11424, 1996. 514. Hagleitner, C., Koll, A., Vogt, R., Brand, O., Baltes, H., Proc. IEEE Transducers 1999, Sendai, Japan, 1999, 1012–1015. 515. Kawahito, S., Koll, A., Hagleitner, C., Baltes, H., Tadokoro, Y., Trans. IEE of Japan, 1999, 119–E, 3, 138–142. 516. Hierlemann, A., Ricco, A.J., Bodenh¨ofer, K., Dominik, A., G¨opel, W., Anal Chem. 2000, 72, 3696–3708. 517. Kummer, A., Hierlemann, A., Baltes, H., Anal Chem., 2004, 76, 2470–2477. 518. Grate, J.W., Abraham, M.H. Sensors and Actuators B-Chemical 1991, 3, 85– 111. 519. Frank, M. personal communication, K-values recorded with TSMRs by M. Frank, Institute of Physical and Theoretical Chemistry, University of T¨ ubingen, Germany. 520. Schneeberger, K., CMOS Microsystems for Thermal Presence Detection, Ph.D. thesis ETH Zurich No. 12675, 1998. 521. Schaufelb¨ uhl, A., Thermal Imagers in CMOS Technology, Ph.D. thesis. ETH Zurich No. 14484, 2002. 522. Kerness, N., Koll, A., Schaufelb¨ uhl, A., Hagleitner, C., Hierlemann, A. Brand, O., Baltes, H., Proc. IEEE MEMS 2000, 2000, 96–101. 523. Kerness, N., CMOS-Based Calorimetric Chemical Microsensors, Ph.D. thesis ETH Zurich No. 14839, 2002. 524. Menolfi, C., Low-Noise CMOS Chopper Instrumentation Amplifiers for Thermoelectric Microsensors, Ph.D. thesis ETH Zurich No. 13583, 2000. 525. Hierlemann, A., Massensensitive Detektion fl¨ uchtiger organischer Substanzen mit modifizierten Polysiloxanen, Ph.D. thesis, University of T¨ ubingen, Germany, 1996.
212
References
526. Lange, D., Brand, O., Baltes, H., CMOS Cantilever Sensor Systems, Springer, Berlin, 2002. 527. Hierlemann, A., Baltes, H., Analyst, 2003, 128, 15–28. 528. Blom, F.R., Bouwstra, S., Elwenspoek, M., Fluitman, J.H., J. Vac. Sci. Technol. B 1992, 10, 19–26. 529. Buser, R.A., Resonant Sensors in Sensors, A Comprehensive Survey, Vol. 7, Bau, H.H., de Rooij, N.F., Kloeck, B. (eds.), VCH Weinheim, 1994, 205–284. 530. Lange, D., Hagleitner, C., Herzog, C., Brand, O., Baltes, H., Sensors and Actuators A 103, 2003, 150–155. 531. Vancura, C., R¨ uegg, M., Li, Y., Lange, D., Hagleitner, C., Brand, O., Hierlemann, A., Baltes, H., Proc. IEEE Transducers, Boston, MA, USA, 2003, 1355–1358. 532. Huijsing, J.H., IEEE Trans. Instrumentation and Measurement, 25, 9/1976, 227–231. 533. Kansy, R.J., IEEE J. Solid State Circuits, 15, 1980, 373–375. 534. Li, Y., Vancura, C., Hagleitner, C., Lichtenberg, J., Brand, O., Baltes, H., Proc. IEEE Sensors, Toronto, Canada, 2003, 244–245. 535. Saeckinger, E., Guggenbuehl, W., IEEE J. of Solid-State Circuits, 1987, Vol. 22. No. 2, p. 287–294. 536. Gano, A.J., Proc. 6th IEEE Int. Conference on Circuits and Systems, Paphos, 1999, pp. 785–788. 537. Kummenacher, F., IEEE J. of Solid-State Circuits, 1988, Vol. 23, No. 3, pp. 750–758. 538. Weaver, W., Timoshenko, S.P., Young, D.H. Vibration Problems in Engineering, Wiley, New York, 1990. 539. Gere, J.M., Timoshenko, S.P. Mechanics of Materials, 3rd ed., PWS-Kent, Boston, 1990. 540. Wohltjen, H. Sens. Actuators, 1984, 5, 307–325. 541. Ballantine, D.S., Wohltjen, H., Anal. Chem., 1989, 61, 705–712. 542. Hierlemann, A., Zellers, E.T., Ricco, A.J., Anal. Chem. 2001, 73, 3458–3466. 543. Baller, M., Lang, H.P., Fritz, J., Gerber C., Gimzewski J.K., Drechsler, U., Rothuizen, H., Despont, M., Vettiger P., Battiston, F., Ramseyer J.P., Fornaro, P., Meyer, E., G¨ untherodt, H.J., Ultramicroscopy, 2000, 82, 1–9. 544. Battiston, F., Ramseyer, J.P., Lang, H.P., Baller, M., Gerber, C., Gimzewski, J.K., Meyer, E., G¨ untherodt, H.J., Sens. Actuators B, 2001, 77, 122–131. 545. Britton, C.L., Jones, R.L., Oden, P.I., Hu, Z., Warmack, R.J., Smith, S.F., Bryan, W.L., Rochelle, J.M. Ultramicroscopy 2000, 82, 17–21. 546. Kim, B.H., Prins, F.E., Kern, D.P., Raible, S., Weimar, U. Sens. Actuators B 2001, 78, 12–18. 547. Bˆ arsan, N., Weimar, U., Journal of Electroceramics, 2001, 7, 143–167. 548. Simon, I., Bˆ arsan, N., Bauer, M., Weimar, U., Sensors and Actuators B, 2001, 73, 1. 549. Chan, P.C.H., Yan, G-Z., Sheng, L-Y., Sharma, R.K., Tang, Z., Sin, J.K.O., Hsing, I-M., Wang, Y., Sens. Actuators B, 2002, 82, 277. 550. Afridi, M.Y., Suehle, J.S., Zaghloul, M.E., Berning, D.W., Hefner, A.R., Cavicchi, R.E., Semancik, S., Montgomery, C.B., Taylor, C.J. IEEE-Sensors-Journal 2002, 2, 644–655. 551. Graf, M., Barrettino, D., Zimmermann, M., Hagleitner, C., Hierlemann, A., Baltes, H., Hahn, S., Bˆ arsan, N., Weimar, U. IEEE Sensors Journal, 2004, 4, 9–16.
References
213
552. Barrettino, D., Graf, M., Zimmermann, M., Hagleitner, C., Hierlemann, A., Baltes, H. Analog-Integrated-Circuits-and-Signal-Processing 2004, 39, 275– 287. 553. Graf, M., Barrettino, D., Taschini, S., Hagleitner, C., Hierlemann, A., Baltes, H., Anal. Chem. 2004, 76, 4437–4445. 554. Graf, M., Barrettino, D., Taschini, S., Hagleitner, C., Hierlemann, A., Baltes, H., Proc. IEEE MEMS, Kyoto, Japan, 2003, 303–306. 555. Graf, M. PhD Thesis, Microhotplate-based chemical sensor systems in CMOS technology, ETH Diss. 15438, 2004. 556. Barrettino, D., Graf, M., Taschini, S., Zimmermann, M., Hagleitner, C., Hierlemann, A., Baltes, H. Proceedings IEEE VLSI-Symposium, Kyoto, Japan 2003, 157–160. 557. Graf, M., Jurischka, R., Barrettino, D., Hierlemann, A., J. Micromech. Microeng., 2005, 15, 190–200. 558. Barrettino, D., PhD Thesis, CMOS readout and control architectures for monolithic hotplate and cantilever systems, ETH Diss. 15412, 2004. 559. Graf, M., Barrettino, D., Taschini, S., Hagleitner, C., Hierlemann, A., Baltes, H., Proc. Eurosensors 16, Prague, Czechia, Editors: Sanisteir, J., Ripka, P., 2002, Vol. 2, 249–250. 560. Gardner, J.W., Covington, J.A., Udrea, F., Dogaru, T., Lu, C.C., Milne, W., Proc. Transducers 01, Munich, Germany, 2001, 1688–1691. 561. M¨ uller, T., Brandl, M., Brand, O., Baltes, H. Sensors and Actuators APhysical 2000, 84, 126–133. 562. W¨ ollenstein, J., Plaza, J.A., Cane, C., Min, Y., B¨ottner, H., Tuller, H.L. Sensors and Actuators-B-(Chemical) 2003, 93, 350–355. 563. Cavicchi, R.E., Suehle, J.S., Kreider, K.G., Shomaker, B.L., Small, J.A., Gaitan, M., Chaparala, P. Applied Physics Letters 1995, 66, 812–814. 564. Sheng, L.Y., Tang, Z.N., Wu, J., Chan, P.C.H., Sin, J.K.O. Sensors and Actuators B-(Chemical) 1998, 49, 81–87. 565. Savage, N.O., Roberson, S., Gillen, G., Tarlov, M.J., Semancik, S. Analytical Chemistry, 2003, 75, 4360–4367. 566. Demarne, V., Grisel, A. Sensors and Actuators B-(Chemical) 1993, 15, 63–67. 567. Faglia, G., Comini, E., Cristalli, A., Sberveglieri, G., Dori, L. Sensors and Actuators B-(Chemical) 1999, 55, 140–146. 568. Chung, W.Y., Lim, J.W., Lee, D.D., Miura, N., Yamazoe, N. Sensors and Actuators B-(Chemical) 2000, B64, 118–123. 569. Cavicchi, R.E., Walton, R.M., Aquino Class, M., Allen, J.D., Panchapakesan, B. Sensors and Actuators-B-(Chemical) 2001, B77, 145–154. 570. Jimenez, I., Cirera, A., Cornet, A., Morante, J.R., Gracia, I., Cane, C. Sensors and Actuators B-(Chemical) 2002, 84, 78–82. 571. Vincenzi, D., Butturi, M.A., Guidi, V., Carotta, M.C., Martinelli, G., Guarnieri, V., Brida, S., Margesin, B., Giacomozzi, F., Zen, M., Pignatel, G.U., Vasiliev, A.A., Pisliakov, A.V. Sensors and Actuators-B-(Chemical) 2001, B77, 95–99. 572. Heule, M., Gauckler, L.J. Sensors and Actuators B-(Chemical) 2003, 93, 100– 106. 573. Kappler, J., Barsan, N., Weimar, U., Dieguez, A., Alay, J.L., RomanoRodriguez, A., Morante, J.R., Gopel, W. Fresenius Journal of Analytical Chemistry 1998, 361, 110–114.
214
References
574. Kappler, J., Tomescu, A., Barsan, N., Weimar, U., Thin Solid Films 2001, 391, 186–191. 575. I2 C (inter integrated circuit, inter-IC) is a communication standard developed by Philips, Eindhoven, The Netherlands: http://www.semiconductors. philips.com/buses/i2c/facts. 576. Graf, M., Taschini, S., K¨aser, P., Hagleitner, C., Hierlemann, A., Baltes, H., Proc. Eurosensors, Guimaraes, Portugal (2003) 661–662. 577. Graf, M., Taschini, S., K¨aser, P., Hagleitner, C., Hierlemann, A., Baltes, H., Proc. IEEE-MEMS 2004, Maastricht, NL, 351–354. 578. Rodriguez, J.L., Hughes, R.C., Corbett, W.T., McWorther, P.J., Technical Digest IEEE Int. Electron. Devices Meeting, San Francisco, 1992, 521–524. 579. Hughes, K.L., Miller, S.L., Rodriguez, J.L., McWorther, P.J., Sens. Actuators B, 1996, 37, 75–81. 580. Van den Berg, A., van der Wal, P.D., van der Schoot, B., de Rooij, N.F., Sens. Materials, 1994, 6, 23–43. 581. http://www.i-stat.com/ 582. Lauwers, E., Suls, J., Gumbrecht, W., Maes, D., Gielen, G., Sansen, W., IEEE Journal of Solid-State Circuits, 2001, 36, 2030–2038. 583. Hagleitner, C., Hierlemann, A., Lange, D., Kummer, A., Kerness, N., Brand, O., Baltes, H., Nature, 2001, 414, 293–296. 584. Witvrouw, A., Van Steenkiste, F., Maes, D., Haspeslagh, L., Van Gerwen, P., De Moor, P., Sedky, S., Van Hoof, C., de Vries, A.C., Verbist, A., De Caussemaeker, A., Parmentier, B., Baert, K. Microsystem Technologies 2000, 6, 192–199. 585. Gumbrecht, W., Peters, D., Schelter, W., Erhardt, W., Henke, J., Steil, J., Sykora, U., Sens. Actuators B, 1994, 18–19, 704–708. 586. Schelter, W., Gumbrecht, W., Montag, B., Sykora, U., Erhardt, W., Sens. Actuators B, 1992, 6, 91–95. 587. Hagleitner, C., Lange, D., Hierlemann, A., Brand, O., Baltes, H., IEEE Journal of Solid-State Circuits, 2002, 37, 1867–1878. 588. Hagleitner, C., Hierlemann, A., Baltes, H., CMOS Single-chip Gas Detection Systems Part I, Sensors Update Vol. 11, Series Editors H. Baltes, J. Korvink, G. Fedder, Wiley VCH Weinheim, New York, 2003, 101–155. 589. Koll, A., Kawahito, S., Mayer, F., Hagleitner, C., Scheiwiller, D., Brand, O., Baltes, H., Proc. of SPIE Smart Structures and Materials 1998, 3328, 223–232. 590. http://www.interfacebus.com/Design Connector CAN.html. 591. http://www.epanorama.net/links/serialbus.html#spi. 592. http://www.motion.aptd.nist.gov/. 593. Hilbiber, D.F., Digest of Technical Papers ISSCC, 1964, 32–33. 594. Widlar, R.J., IEEE J. Solid State Circuits, 6, Feb. 1971, 2–7. 595. Blauschild, R.A., Tucci, P., Muller, R.S., Meyer, R.G., Digest of Technical Papers ISSCC 1978, 1978, 50–51. 596. Williams, J., EDN, Apr. 1991, 180–181. 597. Meijer, G.C.M., in Analog Circuit Design: Low-Noise, Low-Power, LowVoltage, Mixed-Mode Design with CAD Tools, Voltage, Current and Time References (eds: Huijsing, J.H., van de Plasche, R.J., Sansen, W.), Kluwer Academic Publishers, 1996, 243–268. 598. A. Bakker, High Accuracy CMOS Smart Temperature Sensors, Ph.D. thesis, TU Delft, 2000.
References
215
599. Song, B.S., Gray, P.R., IEEE J. Solid State Circuits, 18, Dec. 1983, 634–643. 600. Tuthill, M., IEEE J. Solid State Circuits, 33, July 1998, 1117–1122. 601. Senturia, S.D., CAD Challenges for Microsensors, Microactuators, and Microsystems, Proc. IEEE, 86, 1998, 1611–1626. 602. Mukherjee, T., Fedder, G.K., J. VLSI Signal Processing, 21, 1999, 233–249. 603. Marshall, J.C., Parameswaran, M., Zaghloul, M.E., Gaitan, M., IEEE Circuits and Devices Magazine, 8, 1992, 10–17. 604. Magic: http://www.research.compaq.com/wrl/projects/magic/magic.html. 605. Ballistic: http://www.eecg.toronto.edu/∼gdt/. 606. Fedder, G.K., Jing, Q., IEEE Trans. Circuits and Systems II, 46, 1999, 1309– 1315. 607. Sugar: www.bsac.eecs.berkeley.edu/∼cfm. 608. Coventor, Cary, NC, USA, http://www.coventor.com. 609. MEMSCAP, Bernin, France, http://www.memscap.com. 610. M¨ uller, T., Feichtinger, T., Breitenbach, G., Brandl, M., Brand, O., Baltes, H., Proc. 11th IEEE Workshop on Micro Electro Mechanical Systems, 1998, 240–245. 611. M¨ uller, T., An industrial CMOS process family for integrated silicon sensors, Ph.D. thesis no. 13463, ETH Z¨ urich, 1999. 612. Kurzawski, P., Lazic, I., Hagleitner, C., Hierlemann, A., Baltes, H., Proc. IEEE Transducers, Boston, MA, USA, 2003, 1359–1362. 613. Krauss, A., Lange, D., Weimar, U., Hierlemann, A., Brand, O., Baltes, H., Proc. InternationaI Symposium on Electronic Noses, ISOEN, Brighton, 2000, 35–36. 614. Hierlemann, A., Weimar, U., Baltes, H., Handheld and Palmtop Chemical Microsystems for Gas Analysis, Handbook of Machine Olfaction, Editors T.C. Pearce, S.S. Schiffman, H.T. Nagle, J.W. Gardner, Wiley VCH Weinheim, New York, 2003, 201–229. 615. Brand, O., Baltes, H., Proc. Micro System Technologies 2001, 2001, 37–42. 616. Brand, O., Baltes, H., Waelti, M., IEEE European Symp. on Reliability of Electron Devices, Failure Physics and Analysis (ESREF) 2000 and Microelectronics Reliability, 40, 2000, 1255–1262. 617. Lau, J.H., Flip-chip technologies, Mc Graw Hill, New York 1995. 618. Wilson, D.M., DeWeerth, S.P. Sensors and Materials 1998, 10, 169–83. 619. Apsel, A., Stanford, T., Hasler, P., 1998, Proceedings of the 1998 IEEE International Symposium on Circuits and Systems, ISCAS ’98. (Cat. No.98CH36187), 1998, Vol. 3, 107–110. 620. Shults, M.C., Rhodes, R.K., Updike, S.J., Gilligan, B.J., Reining, W.N. IEEE Transactions on Biomedical Engineering 1994, 41, 937–42. 621. Wouters, P., de Cooman, M., Lapadatu, D., Puers, R. Sensors and Actuators A (Physical) 1994, A41, 198–206. 622. Puers, B., Wouters, P., De Cooman, M., Sensors and Actuators A 1993, 260–7. 623. Bult, K., Burstein, A., Chang, D., Dong, M., Fielding, M., Kruglick, E., Ho, J., Lin, F., Lin, T.H., Kaiser, W.J., Mukai, R., Nelson, P., Newburg, F.L., Pister, K.S.J., Pottie, G., Sanchez, H., Stafsudd, O.M., Tan, K.B., Ward, C.M., Yung, G., Xue, S., Marcy, H., Yao, J., Technical Digest. Solid-State Sensor and Actuator Workshop, Hilton Head, USA, 1996, 205–210. 624. http://www.bluetooth.com/.
216
References
625. Mowry, C.D., Morgan, C.H., Manginell, R.P., Kottenstette, R.J., Lewis, P.R., Frye-Mason, G.C., Baca, Q. in Chemical and Biological Early Warning Monitoring for Water, Food, and Ground, J.L. Jensen and L.W. Burggraf (eds.), 2002, Vol. 4575, pp. 83–90. 626. Kottenstette, R.J., Frye-Mason, G.C., Mowry, C.D., Lewis, P.R., Manginell, R.P., Adkins, D.R., Dulleck, Jr., G.R., Martinez, D., Matzke, C.M., Morgan, C.H., Anderson, L.F., Proceedings µTAS 2001, Monterrey, CA, USA, Oct. 21–25. 627. Potkay, J.A., Driscoll, J.A., Agah, M., Sacks, R.D., Wise, K.D. Proceedings IEEE Micro Electro Mechanical Systems MEMS, 2003, Kyoto, Japan, 395– 398. 628. Wei-Cheng-Tian, Pang, S.W., Chia-Jung-Lu, Zellers, E.T., Journal of Microelectromechanical Systems, 2003, 12, 264–72. 629. Caillat, P., David, D., Belleville, M., Clerc, F., Massit, C., Revol Cavalier, F., Peltie, P., Livache, T., Bidan, G., Roget, A., Crapez, E. Sensors and Actuators B (Chemical) 1999, B61, 154–62. 630. Thewes, R., Hofmann, F., Frey, A., Holzapfl, B., Schienle, M., Paulus, C., Schindler, P., Eckstein, G., Kassel, C., Stanzel, M., Hintsche, R., Nebling, E., Albers, J., Hassmann, J., Sch¨ ulein, J., Goemann, W., Gumbrecht, W., IEEE International Solid-State Circuits Conference, ISSCC, 2002, san Francisco, USA, Digest of Technical Papers, 2002, 350–351. 631. Baumann, W.H., Lehmann, M. Schwinde, A., Ehret, P., Brischwein, M., Wolf, B., Sens. Actuators B, 1999, 55, 77. 632. Lehmann, M., Baumann, W., Brischwein, M., Gahle, HJ., Freund, I., Ehret, R., Drechsler, S., Palzer, H., Kleintges, M., Sieben, U., Wolf, B., Biosens. Bioelectron. 2000, 15, 117. 633. Lehmann, M., Baumann, W., Brischwein, M., Gahle, HJ., Freund, I., Ehret, R., Drechsler, S., Palzer, H., Kleintges, M., Sieben, U., Wolf, B., Biosens. Bioelectron. 2001, 16, 195. 634. Wolf, B., Kraus, M., Brischwein, M., Ehret, R., Baumann, W., Lehmann, M., Bioelectrochemistry and Bioenenergetics 1998, 46, 215–225. 635. Gilchrist, K.H., Barker, V.N., Fletcher, L.E., DeBusschere, B.D., Ghanouni, P., Giovangrandi, L., Kovacs, G.T.A., Biosens. Bioelectron. 2001, 16, 557. 636. Stenger, D.A., Gross, G.W., Keefer, E.W., Shaffer, K.M., Andreadis, J.D., Ma, W., Pancrazio, J.J., Trends in Biotechnology, 2001, 19, 304. 637. Kovacs, G.T.A., Proceedings of the IEEE, 2003, 91, 915–930. 638. Pancrazio, J.J., Bey, R.P., Jr., Loloee, A., Manne, S., Chao, H.C., Howard, L.L., Milton Gosney, W., Borkholder, D.A., Kovacs, G.T.A., Manos, P., Cuttino, D.S., Stenger, D.A. Biosensors & Bioelectronics 1998, 13, 971–9. 639. DeBusschere, B.D., Kovacs, G.T.A., Biosensors & Bioelectronics 2001, 16, 543–556. 640. Fromherz, P., ChemPhysChem, 2002, 3, 276–284. 641. Fromherz, P., Neuroelectronic Interfacing, in Nanoelectronics and Information Technology, Waser, R. (ed), Wiley VCH, Berlin, 2003. 642. Eversmann, B., Jenkner, M., Hofmann, F., Paulus, C., Brederlow, R., Holzapfl, B., Fromherz, P., Merz, M., Brenner, M., Schreiter, M., Gabl, R., Plehnert, K., Steinhauser, M., Eckstein, G., Schmitt-Landsiedel, D., Thewes, R., IEEE Journal of Solid-State Circuits, 2003, 38, 2306–2317. 643. Franks, W., Heer, F., McKay, I., Taschini, S., Sunier, R., Hagleitner, C., Hierlemann, A., Baltes, H., Biosens. Bioelectronics 2004, 20/2, 358–366.
References
217
644. Najafi, K., Wise, K.D., Sensors and Actuators A 1990, 21–23, 589–591. 645. Qing, B., Gingerich, M.D., Wise, K.D., Technical Digest. Solid-State Sensor and Actuator Workshop, Hilton Head, USA, 1998, 15–18. 646. Csicsvari, J., Henze, D.A., Jamieson, B., Harris, K.D., Sirota, A., Bartho, P., Wise, K.D., Buzsaki, G., Journal of Neurophysiology, 2003, 90, 1314–1323. 647. Bai, Q., Wise, K.D., IEEE Transactions on Biomedical Engineering, 2001, 48, 911–920. 648. Wise, K.D., Sensors and Materials, 1998, 10, 385–395. 649. Chen, J., Wise, K.D., IEEE Transactions on Biomedical Engineering, 1997, 44, 760–769. 650. Wise, K.D., Anderson, B.J., Hetke, J.F., Kipke, D.R., Najafi, K., Proc. of the IEEE, Biomedical applications for MEMS and microfluidics, editors: Lee, A., Fair, R.B., 2004, 92, 76–97.
Abbreviations
A/D: AC: ADC: AE: AFM: AHDL: Al: APCVD: ASIC: ATR: BBIC: BET: BOE: BSA: CAD: CAN: CCD: CE: CHEMFET: CIF: CMOS: CVD: D/A: DAC: DBR: DC: DDA: DIL: DNA: DRC: DRIE: EC: ECE: EDP: EIS:
analog-digital alternating current analo-to-digital converter auxiliary electrode atomic-force microscopy analog hardware description language aluminum atmospheric-pressure chemical vapor deposition application-specific integrated circuit attenuated total reflection bioluminescent bioreporter integrated circuit Brunauer-Emett-Teller buffered oxide etch bovine serum albumin computer-assisted design controller area network charge-coupled device counter electrode chemically sensitive field-effect transistor Caltech interchange format complementary metal oxide semiconductor chemical vapor deposition digital-analog digital-to-analog converter distributed Bragg reflector directed current differential difference amplifier dual in-line desoxyribonucleic acid design rule check deep reactive-ion etching ethyl cellulose electrochemical etch stop ethylenediamine/pyrochatechol electrolyte insulator semiconductor
220
Abbreviations
ENFET: EPROM: FEM: FPI: FPW: FTR: GC GDS: HDL: HF: HNA: HRP: HVAC: IC: I2 C: ICD: IDT: IEEE: IMFET: IO: IR: ISCAP: ISE: ISFET: IUPAC: IWAO: KOH: LAPS: LED: LEL: LOD: LPCVD: LSB: LVS: MEMS: MOEMS: MOS: MOSCAP: MOSFET: MUX: NMOS: NTC: Opamp: PBS: PCPMS:
enzyme-based field-effect transistor electrically programmable read-only memory finite-element method Fabry-Perot interferometer flexural plate wave frustrated total reflection gas chromatography geometric data standard hardware description language hydrogen fluoride hydrofluoric acid, nitric acid, and acetic acid horseradish peroxidase humidity ventilation air-conditioning integrated circuit inter integrated circuit, inter-IC ion-controlled diode interdigital transducer, interdigitated transducer institute of electrical and electronics engineers immuno field-effect transistor integrated optics infrared ion-sensitive capacitor ion-selective electrode ion-sensitive field-effect transistor international union of pure and applied chemistry integrated-waveguide absorbance optode potassium hydroxide light-addressable potentiometric sensor light-emitting diode lower explosive limit limit of detection low-pressure chemical vapor deposition least-significant bit layout versus schematic micro-electro-mechanical systems micro-opto-electro-mechanical systems metal oxide semiconductor metal oxide semiconductor capacitor metal oxide semiconductor field-effect transistor multiplexer n-channel metal oxide semiconductor negative temperature coefficient operational amplifier phosphate buffer solution poly(cyanopropylmethylsiloxane)
Abbreviations
Pd: PDMS: PECH: PECVD: PEUT: pH: PID: PMOS: ppb: ppm: PSD: Pt: PTAT: PTC: PVD: PZT: Q: QMB: RE: REFET: RGTO: RIE: RIFS: RMO: SAM: SAW: SCL: SDA: SEM: SGFET: SH-APM: Si: SNR: SOI: SPI: SPM: SPR: STW: TE: TIRF: TLV: TM: TMAH: TO: TSMR:
palladium poly(dimethylsiloxane) poly(epichlorohydrine) plasma-enhanced chemical vapor deposition poly(etherurethane) potentia hydrogenii proportional integral differential p-channel metal oxide semiconductor parts per billion parts per million photosensitive diode platinum proportional to absolute temperature positive temperature coefficient physical vapor deposition lead zirconate titanate quality quartz microbalance reference electrode reference field-effect transistor rheotaxial growth and thermal oxidation reactive-ion etching reflectometric interference spectroscopy reverse mode of operation self-assembled monolayer surface acoustic wave serial clock serial data scanning electron microscopy suspended-gate field-effect transistor shear-horizontal acoustic plate mode silicon signal-to-noise ratio silicon on insulator serial peripheral interface scanning-probe microscopy surface plasmon resonance shear tranverse wave transverse electric total internal reflectance fluorescence threshold-limited value transverse magnetic tetramethyl ammonium hydroxide transistor outline thickness-shear-mode resonator
221
222
Abbreviations
UV: VCSEL: VHDL: VIS: VOC: WE: Σ∆ :
ultraviolet vertical-cavity surface-emitting laser very-high-speed integrated-circuit hardware description language visible volatile organic compound working electrode Sigma-Delta converter
Index
absorption 9, 45 absorption-emission process 46 activation energy 40 activity 10 adsorption 9 adsorption isotherm 13 ammonium hydroxide 24 amperometry 59 amplitude grating 54 analyte 2 concentrations 31 condensation 107 anisotropic etching 19, 23 dry 25 wet 23 anodic bonding 26 anti-aliasing filter 105 application-specific integrated circuits 20 atomic-force microscopy 37 attenuated total reflection 47 auxiliary electrode 60 auxiliary functions 85 auxiliary sensor 87 band bending 74 bandgap voltage 170 batch process 15 bead resistor 40 bimorph effect 37, 111 BIOFET 71 bioluminescence 55 biosensor 3 Bode-plot 123 Bosch process 25 Bragg reflector 52 Brownian motion 120 bulk micromachining 23
calorimeter circuitry 104 gas sensing 105 thermoelectric 166 cantilever 37, 166 array 128 capacitive crosstalk 116 comparison to other mass-sensitive devices 130 deflection 38 dynamic mode 38 electrothermal actuation 115 fabrication 111 feedback circuitry 117 fractional frequency shift 131 magnetic actuation 113, 115, 118 operation in water 122 oscillator circuitry 117 response closed-loop 122 open-loop 122 static mode 38 temperature 115 thermal actuation 110, 111 thermal crosstalk 116 capacitance 84 capacitance/voltage curve 73 capacitive current 61 gas sensing 92 microsystem circuitry 90 sensor 83, 84, 166 carbon monoxide 81 cardiac myocyte 190 catalyst 40 catalytic oxidation 40 catalytic thermal sensors 40 cell-based biosensor 190
224
Index
ceramics 16 charge transfer 59 CHEMFET 70 chemical analysis system 1 miniaturized 188 chemical equation 10 chemical potential 11 chemical reaction 13 chemical sensor 4 definition 2 chemical thermodynamics 10 chemical vapor deposition 17 chemisorption 9 chemocapacitors 60, 83 chemodiode 71 chemoluminescence 46 chemomechanical sensor 29 chemoresistor 77 high-temperature 80 low-temperature 78 chemotransistor 60, 67 chip heating 157 chlorine sensor 64 chopper amplifier 105 Clark cell 63 electrode 163 oxygen sensor 165 CMOS 20 amperometric sensor 62 calorimetric transducer 100 capacitive microsystem 88 capacitive transducer 88 chip 21 design flow 172 dielectric layers 101 fabrication 173 layout 172 materials 86 MEMS 85 MEMS design 171 microhotplate 133, 148 multiparameter biochemical microsystem 164 multisensor system 163, 164 operation modes 177 resonant cantilever 109 substrate 20 technology 20
cold junction 43 combustion 42 compressional wave 34 concentration cell 65 condensation heat 43 conductance 77 conductance cell 77 conducting polymer 78 conductivity 77 conductometric sensor 59, 76 conductometry 59 connectivity 87 contact capacitance 78 contact potential 13, 73 contact resistance 78 convection 61 correlation spectroscopy 55 Cottrell equation 62 counter electrode 60 cross-sensitivity 2 current references 170 current/potential curve 60 cyclic voltammetry 62 decimation filter 105 deep reactive-ion etching 25 depletion 75 deposition 17 design rule check 172 die-attach 182 dielectric constant 88, 90 dielectric properties 88 dielectrometer 83 difference frequency 33 difference interferometer 50 diffraction 54 diffraction grating 54 diffuse reflection 47 diffusion 61 diffusion length 62 diffusion process 20 digital controller 169 digital temperature controller 155 doping 20 doping profile 20 dosimeter 2 drain 67 dry etching 19 dual-in-line (DIL) package 182
Index electric field 89 electrical potential 13 electrical work 13 electrochemical cell 59, 64 electrochemical etch stop 146 electrochemical potential 13 electrochemical reaction 13 electrochemical sensor 59 electrode 59 impedance 78 periodicity 94 polarization 77 electrogenic cell 189, 190 electrolyte insulator semiconductor (EIS) 72 electromagnet 118 electromagnetic wave 45 electromotive force 14 electron affinity 73 electron-beam evaporation 18 electron-beam lithography 18 electrophysiology 190 elementary charge 13 ellipsometry 47 endothermic 39 ENFET 70 enthalpy 12, 39 enthalpy change 39, 105 entropy 12 equilibrium 10 equilibrium constant 10 etch groove 24 etching 19 ethylene diamine/pyrochatechol 24 evanescent field 47 exchange current 65 exothermic 39 expansion coefficient 110 external reflection 47 Fabry-Perot interferometer faradaic current 61 Faraday constant 13 feedback capacitors 91 feedback loop 118 Fermi level 13, 43 field-effect 66 field-effect transistor 67 ion-sensitive 66
53
metal oxide semiconductor 66 flexural wave 35 flexural-plate-wave device 35, 36 magnetic excitation 36 flip-chip process 184 fluorescence 46 fluorophore 50 four-electrode configuration 78 frequency sweep 110 frustrated total reflection 47 fugacity 10 fusion bonding 26 gallium arsenide SAW sensor 34 gallium oxide 80 Galvani potential 13, 43, 73 gas sensor handheld instrument 178 reverse mode of operation 179 gate electrode 67 Gibbs free energy 10 Gibbs fundamental equation 12 glass electrode 64 glasses 16 grain boundary 78 grain size 81 grating 54 grating coupler 49, 50 gravimetric sensor 30 gravimetric transducer 109 heart cell 190 heat 39, 105 heat of condensation 107 heat of mixing 107 heater 133 high-ε analyte 94 hot junction 43 hotplate 80 hotplate microsystem analog analog/digital 153 digital array 159 humidity interference 98 hybrid design 85 hydrogel 79 hydrogen 81 hydrogen bond 9 hydrogen sensor 68
148
225
226
Index
IMFET 71 impedance 60 indium oxide 80 infinite dilution 12 input impedance 61 integrated circuit 15 integrated optics 49 integrated sensor 6 integrated waveguide absorbance optode 51 interaction energies 9 interaction mechanism 9 interdigital capacitor 91 interdigital transducers 32 interdigitated electrodes 88 internal reflection 47 intersystem crossing 46 ion implantation 20 ion-controlled diode 71 ion-selective capacitor 72 ion-selective electrode 64 nonsymmetrical 64 symmetrical 64 ionic conduction 81 ISFET 68 isotropic etching 19, 23 dry 24 wet 23 Jablonski diagram
46
Kelvin probe 73, 74 kinetic constant 11 kinetics 10 Lamb wave 35 Lamb-wave device 35 Lambert-Beer relation 45 layout-versus-schematic 172 lift-off technique 19 light 45 light-addressable potentiometric sensor 75 limit of detection (LOD) 2 limit of determination 2 linear solvation energy relationships 93 lithium niobate 32 lithium tantalate 32
logarithmic converter 153 London dispersion 9 Lorentz force 36, 113 Love-wave 32 low-ε analyte 94 lower explosive limit 1, 40 luminescence 46 Mach-Zehnder interferometer 51 magnetic excitation 36 magnetic field 36, 113 magnetic flux 121 mask aligner 18 mass changes 30 mass loading 30 mass resolution 38 mass-sensitive sensors 30 membrane 133 membrane buckling 136 MEMS 6 metal layers 21 metal oxide 80, 133 metal-oxide deposition 147 metal-phthalocyanines 78 Micro Electro Mechanical Systems, MEMS 6 microbridge 41 microcontact printing 26 microelectrode 190 microelectrode array 190 microfabrication 15 microgaschromatographic column 189 microhotplate 80, 133 circular 135 design 134 fabrication 145 heater resistor 142 modeling 139 octagonal 135 thermal simulation 139 microluminometer 56 micromachining 22 post-CMOS 86 pre-CMOS 86 microsensor fabrication sequence 88 microsensor packaging 181 microspectrometer 53 microtechnology 15 Mie scattering 46
Index migration 61 miniaturization 6, 187 molar gas constant 12 mole fraction 12 monochromator 53 monolithic CMOS-MEMS 85, 87 MOS capacitor 72 MOS diode 71 MOSFET 67 multicomponent analysis 130 multisensor system 163 circuitry 168 electrochemical 163 gas tests 174 N-channel 67 N-channel transistor 21 N-type silicon 20 N-well 21 nanocrystalline material 81 Nernst-equation 14 neural probe 191 neuron 190 nickel/gold bumps 185 nitrogen oxide 81 NMOS 21 noise spectrum 110 NTC resistor 40 operational amplifier 60 optical sensor 45 optodes 49 organic layers 4 oscillation frequency 33 oscillator 110 oscillator circuitry 117 outer potential 73 P-channel transistor 21 P-type silicon 20 package 87 chemically inert 181 chip-on-board 182 epoxy-based 181 flip-chip 184 parasitic capacitance 89 partial pressure 93 partition coefficient 13, 93 parts per million 31
patterning 18 pellistor 40 pH: “potentia hydrogenii” 66 change 70 FET 69 phase grating 54 phosphorescence 46 photolithography 18 photon 45 photoresist 18 phototransistor 53 physical sensitivity 96 physical vapor deposition 17 physisorption 9, 107, 122 piezoelectric materials 30, 37 piezoelectricity 30 piezoresistive coefficient 115 piezoresistor 38, 111 planar optical waveguide 49 plasmon 57 platform technology 85 PMOS 21 polarizable interface, 69 polarization 46, 68, 72 poly(dimethylsiloxane) 108 poly(etherurethane) 90 polymer 16 carbon-black-loaded 79 conducting 78 modulus 131 swelling 92 thick layer 94 thin layer 94 polysilicon 21 potassium hydroxide 24 potentiometric sensor 59, 64 potentiometry 59 potentiostat 60 power dissipation 115 ppm-unit 31 preconcentrator 188 Pt-temperature resistor 138 PTC resistor 40 pyroelectric effect 40 quality factor 110, 112 quantum efficiency 49 quartz microbalance 31
227
228
Index
radiation 45 radiation energy 45 Raman scattering 46 rate constant 10 Rayleigh surface-acoustic-wave 31 Rayleigh surface-acoustic-wave device 31, 130 reactants 10 reaction 10 reaction rate constant 10 reactive-ion etching 20 reference capacitor 91 reference electrode 60, 68 reference oscillator 33 reflection 47 reflectometric interference spectroscopy 47 refraction 47 refraction index 47 refractivity 47 resistance 77 resistivity 77 resonance frequency 30 resonating cantilever 37 reverse current 72 reverse mode of operation 179, 180 reversibility 2 sacrificial layer 25 saturation vapor pressure 93 Sauerbrey equation 30, 130 scattering 46 Seebeck coefficient 43 Seebeck-effect 43, 100 selectivity 2 semiconductor processing 16 semiconductor technology 6 sensing capacitor 91 sensitive layer 3, 4 sensitive materials 5 sensitivity 2 sensor array 4 sensor requirements 5 sensor system simulation 172 serial interface 168 Severinghaus electrode 79 Severinghaus pH-FET 165 SGFET 73
shear-horizontal-acoustic-plate-mode 32 shear-transverse-wave 32 sigma-delta modulator 91 silicon 16 silicon island 135 silicon nitride 21 silicon oxide 21 smart sensor 6 soft lithography 26 solder 185 source-drain current 68, 69 specific conductance 77 specific mass density 125 specificity 2 spectrometer 54 spectroscopy 45 specular external reflection 47 spin-coating 18 spray-coating 18 sputtering 17 state function 2, 10 stoichiometric number 10 stripping voltammetry 62 surface acoustic wave 32 surface dipole 74 surface micromachining 25 surface plasmon resonance 48, 57 suspended gate 74 suspended-gate field-effect transistor 73 switched-capacitor 91 system architecture 85 temperature coefficient 139 negative 40 positive 40 temperature homogeneity 134, 137 temperature sensor 170 calibration 139 platinum 138 polysilicon 150 PTAT (proportional to ambient temperature) 150 sigma-delta 170 tetramethyl ammonium hydroxide 24 thermal conductivity 39 thermal evaporation 17 thermal mass 81
Index thermal resistance 103, 134 thermal sensor 39 thermal time constant 133, 142 thermistor 40 thermocouples 102, 103 thermodynamic equilibrium 10 thermodynamics 10, 11 thermoelectric 43 thermoelectric effect 100 thermopile 43 thermovoltage 43, 102 thickness-shear-mode resonator 31, 130 three-electrode-system 60 threshold-limited value 1 tin dioxide 80 tin dioxide, nanocrystalline 135 toolbox strategy 85 total internal reflection 47 total internal reflectance fluorescence 47 transducer 3 transducer periodicity 33 transient behavior 39 transient signal 106 transistor microhotplate 143 transistor outline (TO) socket 182 transistor, stress-sensitive 115 transmission 47
transverse electric 49 transverse magnetic 49 triplet state 46 two-electrode configuration
229
60, 78
underetching 24 underfill 185 vacuum level 73 vaporization heat 43 vertical-cavity surface-emitting laser 52 viscoelastic properties 31, 33 Volta potential 73 voltammetric sensor 59, 60 voltammetry 59 volume fraction 93 Wafer bonding 26 waveguide 48 wavelength 45 wet etching 19 Wheatstone bridge 114 work function 73 working electrode 60 Xenon difluoride Zinc oxide
32
24
Printing: Strauss GmbH, Mörlenbach Binding: Schäffer, Grünstadt