microtechnology and mems · 2019-07-19 · microtechnology and mems series editor: h. baltes h....

microtechnology and mems

microtechnology and mems

Series Editor: H. Baltes H. Fujita D. Liepmann

The series Microtechnology and MEMS comprises text books, monographs, andstate-of-the-art reports in the very active field of microsystems and microtech-nology. Written by leading physicists and engineers, the books describe the basicscience, device design, and applications. They will appeal to researchers, engineers,and advanced students.

Mechanical MicrosensorsBy M. Elwenspoek and R. Wiegerink

CMOS Cantilever Sensor SystemsAtomic Force Microscopy and Gas Sensing ApplicationsBy D. Lange, O. Brand, and H. Baltes

Micromachines as Tools for NanotechnologyEditor: H. Fujita

Modelling of Microfabrication SystemsBy R. Nassar and W. Dai

Laser Diode MicrosystemsBy H. Zappe

Silicon Microchannel Heat SinksTheories and PhenomenaBy L. Zhang, K.E. Goodson, and T.W. Kenny

Shape Memory MicroactuatorsBy M. Kohl

Force Sensors for Microelectronic Packaging ApplicationsBy J. Schwizer, M. Mayer and O. Brand

Integrated Chemical Microsensor Systems in CMOS TechnologyBy A. Hierlemann

A. Hierlemann

Integrated ChemicalMicrosensor Systemsin CMOS Technology

With 125 Figures

123

Professor Dr. Andreas HierlemannPhysical Electronics LaboratoryETH Hoenggerberg, HPT-H 4.2, IQE8093 ZurichSwitzerlandEmail: [email protected]

Series Editors:

Professor Dr. H. BaltesETH Zürich, Physical Electronics LaboratoryETH Hoenggerberg, HPT-H6, 8093 Zürich, Switzerland

Professor Dr. Hiroyuki FujitaUniversity of Tokyo, Institute of Industrial Science4-6-1 Komaba, Meguro-ku, Tokyo 153-8505, Japan

Professor Dr. Dorian LiepmannUniversity of California, Department of Bioengineering466 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 isconcerned, 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 orparts thereof is permitted only under the provisions of the German Copyright Law of September 9, 1965, inits current version, and permission for use must always be obtained from Springer. Violations are liable toprosecution under the German Copyright Law.

Springer is a part of Springer Science+Business Media

springeronline.com

© Springer-Verlag Berlin Heidelberg 2005Printed 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 andregulations and therefore free for general use.

Typesetting: by the authors and TechBooks using a Springer LATEX macro packageCover concept: eStudio Calamar SteinenCover 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 interdisciplinaryfield of CMOS technology-based chemical microsensor systems. It is, on theone hand, targeted at scientists and engineers interested in getting first in-sights in the field of chemical sensing since all necessary fundamental knowl-edge is included. On the other hand, it also addresses experts in the field sinceit provides detailed information on all important issues related to realizingchemical microsensors and, specifically, chemical microsensors in CMOS tech-nology. A large number of sensor and integrated-sensor-system implementa-tions illustrate the current state of the art and help to identify the possibilitiesfor future developments. Since microsensors produce “microsignals”, sensorminiaturization without sensor integration is in many cases prone to failure.This book will help to reveal the benefits of using integrated electronics andCMOS-technology for developing chemical microsensor systems and, in par-ticular, the advantages that result from realizing monolithically integratedsensor systems comprising transducers and associated circuitry on a singlechip.

After a brief introduction, the fundamentals of chemical sensing are laidout, including a short excursion into the related thermodynamics and kinetics.Fabrication and processing steps that are commonly used in semiconductorindustry are then abstracted. These more fundamental sections are followedby a short description of microfabrication techniques and the CMOS sub-strate and materials. Thereafter, a comprehensive overview of semiconductor-based and CMOS-based transducer structures for chemical sensors is given.The corresponding chemically sensitive materials and the related applicationsare mentioned in the context of each transducer structure. CMOS-technologyis then introduced as platform technology, which allows the fabrication ofmicrotransducers and, moreover, enables the integration of these microtrans-ducers with the necessary driving and signal conditioning circuitry on thesame chip. Several examples such as microcapacitors, microcalorimeters, mi-crocantilevers, and microhotplates are described in great detail. In a nextstep, the development of monolithic multisensor arrays and fully developedmicrosystems with on-chip sensor control and standard interfaces is depicted.A short section on packaging shows that techniques from the semiconductorindustry can also be applied to chemical microsensor packaging. The book

VI Preface

concludes with a short outlook to future developments such as developingmore complex integrated microsensor systems and interfacing biological ma-terials such as cells with CMOS microelectronics.

As with all interdisciplinary efforts, teamwork plays a central role in be-ing successful. Therefore I am particularly grateful to many colleagues andformer students, who contributed much to the work that is the topic of thisbook. I would like to thank Prof. Henry Baltes for giving me the opportunityand 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 andexploring new fields of science. I am also very grateful to Prof. Oliver Brand,who was always a valuable source of information on microtechnology and mi-crofabrication. I am very much obliged to several highly motivated and excel-lent coworkers, whose work is amply cited in this book: Christoph Hagleitnerand Kay-Uwe Kirstein, the chief circuit designers, the microhotplate group:Markus Graf, Diego Barrettino, Stefano Taschini, Urs Frey, and MartinZimmermann, the guys working on cantilevers: Dirk Lange, Cyril Van-cura, Yue Li, Jan Lichtenberg, the capacitor freaks: Andreas Koll, AdrianKummer, 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 microelectron-ics and cells are mentioned. These rely on the work of Flavio Heer, WendyFranks, Sadik Hafizovic, Robert Sunier, and Frauke Greve. I am very gratefulfor all their efforts, and I am looking forward to exciting new results in thisresearch area.

I am also indebted to European collaboration partners, Udo Weimarand Nicolae Barsan, University of Tubingen, and to AppliedSensor GmbH,Reutlingen, who provided many of the chemically sensitive materials suchas the metal oxides. The fruitful collaboration with Sensirion AG, Zurich,namely Felix Mayer and Mark Hornung, is also gratefully acknowledged.

Financial support for the CMOS chemical-sensor projects came fromthe European Union (FP5, FP6, IST-program), the Swiss Bundesamt furBildung und Wissenschaft (BBW), the Swiss Commission for Technologyand Innovation (CTI), and the Korber Foundation, Hamburg, Germany.

Zurich, September 2004 Andreas Hierlemann

Contents

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

2 Fundamentals of Chemical Sensing . . . . . . . . . . . . . . . . . . . . . . . 9

3 Microtechnology for Chemical Sensors . . . . . . . . . . . . . . . . . . . . 153.1 Microtechnology Substrate Materials . . . . . . . . . . . . . . . . . . . . . . 163.2 Fundamental Semiconductor Processing Steps . . . . . . . . . . . . . . 16

3.2.1 Deposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173.2.2 Patterning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183.2.3 Etching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193.2.4 Doping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

3.3 CMOS Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203.4 Microfabrication for Chemical Sensors . . . . . . . . . . . . . . . . . . . . . 22

3.4.1 Micromachining for Chemical Microsensors . . . . . . . . . . 223.4.1.1 Bulk Micromachining . . . . . . . . . . . . . . . . . . . . . 233.4.1.2 Surface Micromachining . . . . . . . . . . . . . . . . . . . 25

3.4.2 Wafer Bonding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263.4.3 Sensitive-Layer Deposition . . . . . . . . . . . . . . . . . . . . . . . . . 26

4 Microfabricated Chemical Sensors . . . . . . . . . . . . . . . . . . . . . . . . 294.1 Chemomechanical Sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

4.1.1 Rayleigh SAW Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . 324.1.2 Flexural-Plate-Wave or Lamb-Wave Devices . . . . . . . . . 354.1.3 Resonating Cantilevers . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

4.2 Thermal Sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 394.2.1 Catalytic Thermal Sensors (Pellistors) . . . . . . . . . . . . . . 404.2.2 Thermoelectric or Seebeck-Effect Sensors . . . . . . . . . . . 43

4.3 Optical Sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 454.3.1 Integrated Optics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 494.3.2 Microspectrometers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

4.3.2.1 Fabry-Perot-Type Structures . . . . . . . . . . . . . . . 534.3.2.2 Grating-Type Structures . . . . . . . . . . . . . . . . . . . 54

4.3.3 Bioluminescent Bioreporter Integrated Circuits (BBIC) 554.3.4 Surface Plasmon Resonance (SPR) Devices . . . . . . . . . . 57

4.4 Electrochemical Sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

VIII Contents

4.4.1 Voltammetric Sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 604.4.2 Potentiometric Sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

4.4.2.1 Electrochemical Cell . . . . . . . . . . . . . . . . . . . . . . 644.4.2.2 Field-Effect-Based Devices . . . . . . . . . . . . . . . . 66

4.4.2.2.1 MOS Field-Effect Transistors,MOSFETs,and Ion-Selective Field-Effect Transistors, ISFETs(Chemotransistors) . . . . . . . . . . . . . . 67

4.4.2.2.2 MOS Diode and Ion-ControlledDiode, ICD (Chemodiodes) . . . . . . . 71

4.4.2.2.3 MOS Capacitor and Ion-SelectiveCapacitor (Chemocapacitors) . . . . . . 72

4.4.2.2.4 Measuring Work Functions: KelvinProbeand Suspended-Gate Field-EffectTransistor, SGFET . . . . . . . . . . . . . . 73

4.4.2.2.5 Light-Addressable PotentiometricSensor, LAPS . . . . . . . . . . . . . . . . . . . 75

4.4.2.2.6 Field-Effect Device Fabrication . . . . 764.4.3 Conductometric Sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . 76

4.4.3.1 Chemoresistors . . . . . . . . . . . . . . . . . . . . . . . . . . . 774.4.3.1.1 Low-Temperature Chemoresistors . . 784.4.3.1.2 High-Temperature Chemoresistors

(Hotplate Sensors) . . . . . . . . . . . . . . . 804.4.3.2 Chemocapacitors . . . . . . . . . . . . . . . . . . . . . . . . . 83

5 CMOS Platform Technologyfor Chemical Sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 855.1 CMOS Capacitive Microsystems . . . . . . . . . . . . . . . . . . . . . . . . . . 88

5.1.1 CMOS Capacitive Transducer . . . . . . . . . . . . . . . . . . . . . . 885.1.2 On-Chip Circuitry of the Capacitive Microsystem . . . . 905.1.3 Capacitive Gas Sensing . . . . . . . . . . . . . . . . . . . . . . . . . . . 92

5.1.3.1 Selectivity Through Sensitive Layer Thickness 945.1.3.2 Insensitivity to Low-ε Analytes . . . . . . . . . . . . . 975.1.3.3 Humidity Interference . . . . . . . . . . . . . . . . . . . . . 98

5.2 CMOS Calorimetric Device . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1005.2.1 CMOS Calorimetric Transducer . . . . . . . . . . . . . . . . . . . . 1005.2.2 Calorimeter Circuitry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1045.2.3 Calorimetric Gas Sensing . . . . . . . . . . . . . . . . . . . . . . . . . . 105

5.3 CMOS Integrated Resonant Cantilever . . . . . . . . . . . . . . . . . . . . 1095.3.1 Resonant Cantilever Transducers . . . . . . . . . . . . . . . . . . . 110

5.3.1.1 Thermal Actuation . . . . . . . . . . . . . . . . . . . . . . . 1105.3.1.2 Magnetic Actuation . . . . . . . . . . . . . . . . . . . . . . . 1135.3.1.3 Vibration Detection . . . . . . . . . . . . . . . . . . . . . . . 114

Contents IX

5.3.1.4 Cantilever Temperature . . . . . . . . . . . . . . . . . . . 1155.3.2 Microcantilever Circuitry . . . . . . . . . . . . . . . . . . . . . . . . . . 116

5.3.2.1 Thermal Actuation . . . . . . . . . . . . . . . . . . . . . . . 1165.3.2.2 Magnetic Actuation . . . . . . . . . . . . . . . . . . . . . . . 118

5.3.3 Microcantilevers as Chemical Sensors . . . . . . . . . . . . . . . 1225.3.3.1 Polymer Coating . . . . . . . . . . . . . . . . . . . . . . . . . 1225.3.3.2 Analyte Absorption . . . . . . . . . . . . . . . . . . . . . . . 124

5.3.4 Comparison of Cantileversto Other Mass-Sensitive Devices . . . . . . . . . . . . . . . . . . . . 130

5.4 CMOS Microhotplate System Development . . . . . . . . . . . . . . . . 1335.4.1 CMOS Microhotplates . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133

5.4.1.1 Temperature Sensor Calibration . . . . . . . . . . . . 1395.4.1.2 Thermal Microhotplate Modeling

and Characterization . . . . . . . . . . . . . . . . . . . . . . 1395.4.1.3 Microhotplate Heaters: Resistor

and Transistor . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1425.4.1.4 Microhotplate Sensor Fabrication . . . . . . . . . . . 145

5.4.2 Hotplate-Based CMOS Monolithic Microsystems . . . . . 1485.4.2.1 Analog Hotplate Microsystem . . . . . . . . . . . . . . 1485.4.2.2 Analog/Digital Hotplate Microsystem . . . . . . . 1535.4.2.3 Digital Hotplate Array Microsystem . . . . . . . . . 159

5.5 CMOS Chemical Multisensor Systems . . . . . . . . . . . . . . . . . . . . . 1635.5.1 CMOS Multiparameter Biochemical Microsystem . . . . . 1645.5.2 CMOS Gas-Phase Multisensor System . . . . . . . . . . . . . . 165

5.5.2.1 Multisystem Architecture . . . . . . . . . . . . . . . . . . 1665.5.2.2 Multisystem Circuitry Components, Design

and Fabrication . . . . . . . . . . . . . . . . . . . . . . . . . . 1685.5.2.3 Multisystem Gas Sensor Measurements . . . . . . 1745.5.2.4 Multisystem Applications and Operation

Modes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1775.6 CMOS Chemical Microsensor and System Packaging . . . . . . . . 181

5.6.1 Simple Epoxy-Based Package . . . . . . . . . . . . . . . . . . . . . . 1815.6.2 Chip-on-Board Package . . . . . . . . . . . . . . . . . . . . . . . . . . . 1825.6.3 Flip-Chip Package . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184

6 Outlook and Future Developments . . . . . . . . . . . . . . . . . . . . . . . . 187

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193

Abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223

1 Introduction

The detection of molecules or chemical compounds is a general analytical taskin the efforts of chemists to obtain qualitative and/or quantitative time- andspatially resolved information on specific chemical components [1]. Examplesof qualitative information include the presence or absence of certain odorant,toxic, carcinogenic or hazardous compounds. Examples of quantitative infor-mation include concentrations, activities, or partial pressures of such specificcompounds exceeding, e.g., a certain threshold-limited value (TLV), or thelower explosive limits (LEL) of combustible gases.

All this information can, in principle, be obtained from either a chemi-cal analysis system or alternatively by using chemical sensors. In both casessampling, sample pretreatment, separation of the components and data treat-ment are the tasks to be fulfilled. The main components of a state-of-the-artchemical analysis or sensor system are depicted schematically in Fig. 1.1.

cA

x1

yi xi

x2x3...

comparison withcalibration data

cA

filtering (mem

brane)

detection(chem

ical sensors )

sample (gas, liquid...)

sample conditioning

(catalyst, enzyme...)

data pretreatment

(sensor electronics)

feature extraction

pattern recognitionm

ulticomponent analysis

result: chemical com

position(quantitative or qualitative)

sample uptake

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 com-plex analytical system. Integrated or miniaturized chromatographs or spec-trometers may be denoted chemical sensors as well. However, a typical chem-ical sensor is, in most cases, a cheaper, smaller, and less complex device ascompared to miniaturized analytical systems. A draft of the IUPAC (Inter-national Union of Pure and Applied Chemistry) provides a definition of achemical sensor [2]: “A chemical sensor is a device that transforms chemicalinformation, ranging from the concentration of a specific sample componentto total composition analysis, into an analytically useful signal”. This ratherwide definition does not require that the sensor is continuously operating andthat the sensing process is reversible. But intermittently operating devicesexhibiting irreversible characteristics are usually referred to as dosimeters [3].In this context it is useful to introduce some important keywords used exten-sively throughout the chemical sensor literature [1, 4–11].

ReversibilityThermodynamic reversibility, strictly speaking, requires that the sensor mea-surand is related to a thermodynamic state function. This implies that, e.g.,a certain sensor response unequivocally corresponds to a certain analyte con-centration (analyte here denotes the chemical compound to be monitored).The sensor signal may not depend on the history of previous exposures orhow a certain analyte concentration is reached (no memory effects or hystere-sis). More details on fundamental thermodynamics of the chemical sensingprocess will be given in Chap. 2.

Sensitivity and Cross-SensitivitySensitivity 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 changein the analyte concentration. Cross-sensitivity hence refers to the contribu-tions of compounds other than the desired compound to the overall sensorresponse.

Selectivity/SpecificitySelectivity or specificity can be defined according to Janata [4] as the abilityof a sensor to respond primarily to only one species in the presence of otherspecies (usually denoted as interferants).

Limit of Detection and Limit of DeterminationThe limit of detection (LOD) corresponds to a signal equal to k-times thestandard deviation of the background noise (i.e., k represents the signal-to-noise ratio) with a typical value of k = 3. Values above the LOD indicate thepresence of an analyte, whereas values below LOD indicate that no analyteis detectable.

The limit of determination implies qualitative information, i.e., that thesignal can be attributed to a specific analyte. This in turn requires moreinformation and, therefore, the limit of determination is always higher thanthe limit of detection.

1 Introduction 3

TransducerTransducer is derived from Latin “transducere”, which means to “transfer ortranslate”. Therefore, a device that translates energy from one kind of system(e.g., chemical) to another (e.g., physical) is termed a transducer.

BiosensorBiosensors are usually considered a subset of chemical sensors that make useof biological or living material for their sensing function [10, 11]. Since thisbook covers mostly chemical sensors, there will not be any further diversifi-cation into chemo- and biosensors within this work.

Using the above definitions, chemical sensors usually consist of a sensitivelayer or coating and a transducer. Upon interaction with a chemical species(absorption, chemical reaction, charge transfer etc.), the physicochemicalproperties of the coating, such as its mass, volume, optical properties orresistance, reversibly change (Fig. 1.2).

sens

itive

laye

r

analytemolecules

tran

sduc

er

data

rec

ordi

ngan

d pr

oces

singphysical

measurand

(physico-)chemical

interaction

mass change∆m

frequencysignal, ∆f

chemicalsensor

polymer micro-balance

driv

ing

circ

uitr

yoscillator

circuitexample:

Fig. 1.2. Components of a chemical sensor exemplified for the mass-sensitiveprinciple

These changes in the sensitive layer are detected by the respective trans-ducer and are translated into an electrical signal such as a frequency, current,or voltage, which is then read out and subjected to further data treatmentand processing. In Fig. 1.2, this is exemplified for the mass-sensitive principle.Analyte molecules are absorbed into a coating material (polymer) to an ex-tent governed by intermolecular forces. The change in mass of the polymericcoating in turn causes a shift in the resonance frequency of the transducer,e.g., a quartz microbalance. This frequency shift constitutes the electricalsignal that is used in subsequent data processing.

4 1 Introduction

To supply the different needs in chemical sensing, a variety of transducersbased on different physical principles has been devised. Following the sug-gestion of Janata [4, 5], chemical sensors can be classified into four principalcategories according to their transduction principles:

1. Chemomechanical sensors (e.g., mass changes due to bulk absorption)2. Thermal sensors (e.g., temperature changes through chemical interaction)3. Optical sensors (e.g., changes of light intensity by absorption)4. 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 detailin Chap. 4. An overview of more recent literature on chemical sensors withregard to different transduction principles is given in [5].

Various inorganic and organic materials serve as chemically sensitive lay-ers that can be coated onto the different transducers. Typical inorganic ma-terials include metal oxides like tin dioxide (SnO2) for monitoring reducinggases such as hydrogen or carbon monoxide, or zirconium dioxide (ZrO2) todetect oxygen, nitrogen oxide, and ammonia. Organic layers mostly consistingof polymers such as polysiloxanes or polyurethanes are used to monitor hy-drocarbons, halogenated compounds and different toxic volatile organics. Asurvey of typical chemically sensitive materials and their applications is givenin 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 sensitivematerials evolves in three main directions:

1. Miniaturization and monolithic integration of transducers with electronicsand, possibly, auxiliary sensors.

2. Search for highly selective (bio)chemical layer materials (molecular recog-nition, key-lock-type interactions).

3. Using arrays of sensors exhibiting different partial selectivity (polymers,metal oxides) and developing pattern recognition (odors, aromas) and mul-ticomponent analysis methods (mixtures of gases and liquids).

The latter strategy has grown very popular [12–17], especially since compactsensor arrays can presently be fabricated at low costs, and interferants, whichare present in almost any practical application, can be handled.

Chemical sensors meanwhile have also reached the stage of exploratoryuse in a variety of industrial and environmental applications, some exam-ples being quality control or on-line process monitoring in the food-industryas well as preliminary tests in the areas of medical practice and personal(workplace) safety [18]. In particular in environmental monitoring, there isan urgent need for low-cost sensor systems detecting various pollutants attrace level.

1 Introduction 5

Table 1.1. Typical sensitive materials and applications

Materials Examples Applications

metals Pt, Pd, Ni, Ag, Sb, Rh, . . . inorganic gases likeCH4,H2, . . .

ionic compounds electronic conductors inorganic gases(SnO2, TiO2, Ta2O5, (CO, NOx, CH4 . . . )In2O3, AlVO4, . . . ) exhaust gases, oxygen,mixed conductors ions in water, . . .

(SrTiO3,Ga2O3, perowskites, . . . )ionic conductors

(ZrO2, LaF3, CeO2, nasicon, . . . )

molecular crystals phthalocyanines (Pcs): nitrogen dioxide,PbPc, LuPc2, . . . volatile organics

Langmuir-Blodgett lipid bilayers, organic molecules in medicalfilms polydiacetylene . . . applications, biosensing, . . .cage compounds zeolites, calixarenes, water analysis (ions), volatile

cyclodextrins, crown organics, . . .ethers, cyclophanes, . . .

polymers nonconducting polymers detection of volatile organics,polyurethanes, polysiloxanes, . . . food industry (odor

conducting polymers and aroma), environmentalpolypyrroles, polythiophenes, monitoring in gas

nafion, . . . and liquid phase, . . .

components of synthetic entities medical applications,biological entities phospholipids, lipids, HIV- biosensing, water

epitopes, . . . and blood analysis,natural entities

enzymes, receptors, proteins, pharmascreening, . . .cells, membranes, . . .

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 realizedevice miniaturization and to meet some of the chemical-sensor key crite-ria listed above (low cost, batch fabrication). The rapid development of theintegrated-circuit (IC) technology during the past decades has initiated manyinitiatives to fabricate chemical sensors consisting of a chemically sensitivelayer on a signal-transducing silicon chip [19,20]. The earliest types of chemi-cal 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 introducedhere.

Integrated SensorA sensor is denoted an integrated sensor if the chemical sensing operationis based on a direct influence on an electric component (resistor, transistor,capacitor) integrated in silicon or another semiconductor material [11].

Smart or Intelligent SensorThe combination of interface electronics and an integrated sensor on a singlechip results in a so-called “smart sensor”. At least some basic signal condi-tioning is usually carried out on chip. One major advantage of smart sensorsis the improved signal-to-noise and electromagnetic interference character-istics [11]. In addition the connectivity problem, which occurs especially inmultisensor arrays, can be eased by using on-chip multiplexers and by usingbus interfaces. For more details on sensor system integration, see Chap. 5.

The largely planar integrated-circuit (IC) and chemical-sensor structuresprocessed by combining lithographic, thin film, etching, diffusive and oxidativesteps have been recently extended into the third dimension using microfabri-cation technologies (see Chap. 3 in this book). A variety of micromechanicalstructures including cantilever beams, suspended membranes, freestandingbridges, gears, rotors, and valves have been produced using micromachiningtechnology (MicroElectroMechanicalSystems MEMS) [26–29]. MEMS tech-nology thus provides a number of key features, which can serve to enhancethe functionality of chemical sensor systems [9, 11,26,29–34].

Micromechanical structures (MEMS-structures) and microelectronics canbe realized on a single chip allowing for on-chip control and monitoring ofthe mechanical functions as well as for data preprocessing such as signal am-plification, signal conditioning, and data reduction [29–34]. Complementary-Metal-Oxide-Semiconductor or CMOS-technology is the dominant semicon-ductor IC technology for microprocessors and Application-Specific IntegratedCircuits (ASICs) and has also been used to fabricate integrated chemical mi-crosensors. The use of CMOS technology entails a limited selection of devicematerials (see Sect. 3.3) and a predefined fabrication process for the CMOSpart. Sensor-specific or transducer-specific materials and fabrication stepshave to be introduced in most cases as post-processing after the CMOS fab-rication.

1 Introduction 7

In the next chapters the fundamentals of the chemical sensing processitself will be laid out (Chap. 2) followed by a short description of microfab-rication techniques and the CMOS substrate (Chap. 3). In Chap. 4, therewill be an extensive treatment of the different microtransducers that arecommonly used for chemical sensors. This transducer overview will be re-stricted to semiconductor-based and CMOS-based devices and will, for thesake of completeness, also include short abstracts on devices, which are de-scribed in much more detail in the subsequent Chap. 5 on the CMOS tech-nology platform for chemical sensors. Chapter 5 will show the evolution fromsingle transducers, which are integrated with the necessary driving and signalconditioning circuitry to monolithic multisensor arrays and fully developedsystems with on-chip sensor control and standard interfaces. The concludingChap. 6 will include a short glance at future developments such as combiningcells 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 beconfined to the surface of the sensing layer, or it can take place in the wholevolume of the sensitive coating. Surface interaction implies that the speciesof interest is adsorbed at the surface or interface (gas/solid or liquid/solid)only, whereas volume interaction requires the absorption of the species and apartitioning between sample phase and the bulk of the sensitive material. Thedifferent types of chemical interactions involved in a sensing process rangefrom very weak physisorption through rather strong chemisorption to chargetransfer and chemical reactions.

Physisorption in this context implies that the compound is only physicallyab/adsorbed (London or Van-der-Waals dispersion forces) with an interactionenergy of 0–30 kJ/mol, whereas in the case of the much stronger chemisorp-tion (interaction energy >120 kJ/mol), the particles stick to the surface byforming a chemical (usually covalent) bond. Charge transfer and chemicalreactions involve, in most cases, interaction energies comparable to those ofchemisorption and higher. Some of the most common interaction mechanismsand associated energies are listed in Table 2.1, for further details, see [35].

Table 2.1. Typical intermolecular interactions and energies

Interaction type Typical energy [kJ/mol] Comment

covalent bond 120–800 chemical reaction

ion-ion 250 only between ions

coordination, complexation, 8–200 weak “chemical” interactioncharge-transfer bonding

ion-dipole 15

hydrogen bond 20 hydrogen bond: A–H· · ·Bdipole-dipole 0.3–2 between polar molecules

London dispersion (induced 0.1–2 physical interaction betweendipole-induced dipole) any molecules

10 2 Fundamentals of Chemical Sensing

High chemical selectivity and rapid reversibility place contradictory con-straints on desired interactions between chemical sensor coating materials andanalytes. Low-energy, perfectly reversible (physisorptive) interactions gener-ally lack high selectivity, while chemisorptive processes, the strongest of whichresult in the formation of new chemical bonds, offer selectivity, but are in-herently less reversible. A practicable compromise has to be achieved withdue regard to the specific application. In this context it should be noted thatthe commonly accepted limit of reversibility up to 20 kJ/mol refers to roomtemperature and will not apply to the case of, e.g., tin-dioxide–coated semi-conductor sensors operated at 300C to 400C. On the other hand, spon-taneous chemical reactions occurring at room temperature often require atedious regeneration of, e.g., biological recognition units (enzymes).

Any interaction between a coating material and an analyte is governed bychemical thermodynamics and kinetics. Thus, a fundamental thermodynamicfunction, the Gibbs free energy, G [J], is the most important descriptor in allchemical sensing processes: The direction of spontaneous reactions is alwaystowards lower values of G (minimization of the Gibbs energy). The Gibbsfree energy is a state function in the thermodynamic sense, i.e., its valuedepends only on the current state of the system and is independent of howthat state has been prepared. This implies that any chemical (sensing) processdescribed by a Gibbs energy function moves towards a dynamic equilibrium(∆G = 0, G minimal), in which both reactants and products are presentbut have no tendency to undergo net change. This equilibrium is reversible,i.e., an infinitesimal change in the conditions in opposite directions results inopposite changes in its state. The interaction equilibrium of an analyte, A,with a sensor coating, S, can thus be represented by:

A + S

→k⇔←k

A · · ·S . (2.1)

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 de-scribed by an equilibrium constant, K, which relates the activity, a, of reac-tion products (A· · ·S) to those of the reactants (A and S). This constant isthus a characteristic value for the progression of the reaction (K ≤ 1 : noreaction takes place), its numerical value depends on the system temperature.

K =aA···S

aA · aSand in general : K =

∏i

anii . (2.2)

The index i denotes the chemical substance, ni are the corresponding stoi-chiometric numbers in the chemical equation. This expression signifies thateach activity (or fugacity) is raised to the power equal to its stoichiomet-ric number, and, then, all such terms are multiplied together. Stoichiometricnumbers of the products are positive and those of the reactants are nega-tive, 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 ef-fective quantity of compound i participating in, e.g., a chemical reaction, isrelated to the mole fraction, xi, (partial pressure, p, for gases) of a speciesvia: aA = γA · xA with γI ≤ 1. The activity coefficient, γi, measures thedegree of departure of a components behavior from ideal or ideally dilutebehavior.

The equilibrium constant, K, is also related to kinetics. For the simplereaction 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

dt= −→

k aAaS +←k aS···A . (2.3)

K then represents the ratio of those two kinetic constants in equilibriumstate.

K =→k←k

. (2.4)

Both, thermodynamics and kinetics hence affect the progress of any chemicalprocess or reaction. Thermodynamics, namely the Gibbs free energy (mini-mum) or the equilibrium constant, can tell us the direction of spontaneouschange and the composition at the equilibrium state, whereas kinetics tellus, whether a kinetically viable pathway exists for that change to occur, andhow fast an equilibrium state will be achieved. Kinetics are important in thecontext of chemical sensors, since there exist chemical processes, the activa-tion barrier of which is too high to get a reaction going, although the Gibbsfree energy of the products would be below that of the reactants. Such effectscan be used to advantage in tuning the selectivity of, e.g., catalytic chemicalsensors (see, e.g., Sect. 4.4.3.1.2).

A chemical potential has been introduced in thermodynamics. The chemi-cal potential shows how the Gibbs energy of a system changes when a portionof a specific chemical compound is added to it or removed from it. The chem-ical potential of the i-th component, µi, is defined as:

µi =(

∂G

∂ni

)p,T

and ∆G =∑

i

µi dni . (2.5)

1 The activities and activity coefficients used throughout this book are related tomole fractions for simplicity reasons. The standard states include (a) a pure com-pound or (b) infinite dilution. There exist also activities and activity coefficientsthat are related to molalities (mol/kg) or concentrations (mol/m3). The standardstate of molality is 1 mol/kg, that of concentration is 1 mol/liter. The values ofthe activity coefficients related to molalities or concentrations are significantlydifferent from those for molar fractions. A detailed discussion of this issue canbe found, e.g., in Levine, I., Physical Chemistry, 2nd edition, McGraw-Hill 1983,New York, pp. 249–258.


Here ni denotes the stoichiometric number or the amount of substance inmoles. Again, the stoichiometric numbers of the products are positive andthose of the reactants negative. The pressure, p, and the temperature, T ,are kept constant. The chemical potential can be expressed in terms of molefractions, 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 = µ0

i (p, T ) + RT ln fi . (2.6)

µ0i (p, T ) here denotes the chemical potential of an appropriately defined stan-

dard state such as, e.g., “infinite dilution” or a “pure compound”; R is themolar 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-dependentterm. Plugging the terms of (2.6) into (2.5), the reference terms (µ0

i ) can besubsumed into ∆G0 as shown in the following equation:

∆G =∑

i

niµ0i (p, T ) + RT

∑i

ni ln ai = ∆G0 (p, T ) + RT ln∏

i

anii . (2.7)

Both, ∆G and K are characteristic descriptors for the direction of a chemicalreaction. In comparing (2.2) with (2.7), it is evident that in a thermodynamicequilibrium state (∆G = 0)∆G0 and K are interrelated via the followingequation (for details, see [35]):

ln K = −∆G0

RT. (2.8)

The more negative ∆G0, the larger is K, or in other words, the higher thechemical potential of the reactants with regard to the products, the largeris the reaction extent, and the more spontaneous will the reaction occur incase that (already discussed) kinetic factors will not upset such predictions.According to the Gibbs fundamental equation, ∆G0 is composed of an en-thalpy term, ∆H0, representing the reaction heat at constant pressure, andan entropy term, ∆S0, representing the degree of “disorder” or, thermody-namically more precise, the number of different ways in which the energy ofa system can be achieved by rearranging the atoms or molecules among thestates available to them (for details, see [35]):

∆G0 = ∆H0 − T∆S0 . (2.9)

For spontaneous reactions (∆G0 negative), the entropy increases and/or theenthalpy term is negative, i.e., heat is released during the chemical reaction.

In the following, the thermodynamics of three prototype reactions ofchemical sensors will be briefly discussed.

2 Fundamentals of Chemical Sensing 13

Simple Adsorption/AbsorptionAt thermodynamic equilibrium state, the free species and the ad/absorbedspecies are in dynamic equilibrium, i.e., the chemical potentials of a cer-tain compound A in gaseous and polymeric phase are identical: µgas

A (p, T ) =µpolymer

A (p, T ) (2.6). Absorbing all the constant terms (µ0i , R, T ) into a sorp-

tion constant, Ksorption, or so-called partition coefficient, the equilibrium statecan be described by:

Ksorption =asorbed

A

afreeA

. (2.10)

The partition coefficient is a dimensionless “enrichment factor” relating, e.g.,the activity of a compound in the sensing layer (asorbed

A ) to that in the probedgas or liquid phase (afree

A ) and also represents a thermodynamic equilibriumconstant, which is related to ∆G0 via (2.8).

For surface adsorption, it is more common to relate the fractional coverageof the surface, θ, to the concentration of the analyte in the probed phase andto use different types of adsorption isotherm like Langmuir-, Freundlich-, orBET-(Brunauer-Emett-Teller) isotherms [35].

Chemical ReactionIn this case, (2.2) can be applied in principle. It has to be modified withregard to the respective reaction mechanism occurring. For a simple reactionlike nAA + nBB ↔ nCC + nDD, the equilibrium constant is given in analogyto (2.2):

K =∏

i

anii and in particular : K =

anC

C · anD

D

anA

A · anB

B

. (2.11)

The chemical potentials as defined in (2.6) can be used, and (2.8) holds. Asalready mentioned, the interaction leading to a true chemical reaction maybe too strong to be reversible.

Charge Transfer and Electrochemical ReactionFor a reaction of type A+ + e− ↔ A, an electrochemical potential has tobe introduced. The contribution of an electrical potential to the chemicalpotential is calculated by noting that the electrical work, We, of adding acharge, z.e (z denotes the number of elementary charges, e), to a regionwhere the potential is φ (φ denotes the Galvani potential, which representsthe bulk-to-bulk inner contact potential of two materials and is defined asthe 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 toone mole of elementary charges. Consequently, the electrochemical potentialis (compare 2.6):

µi = µ0i (p, T ) + RT ln ai + zFφ . (2.13)


When z = 0 (neutral species), the electrochemical potential is equal to thechemical potential (2.6). Rewriting (2.7) for the electrochemical potentialsleads to:

∆G = ∆G0 (p, T ) + RT ln∏

i

anii + zF · ∆φ . (2.14)

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 E0,the standard cell potential, for −∆G0/zF (a positive voltage per conventionalways corresponds to a negative ∆G: spontaneous reaction), the so-called“Nernst-equation” results:

E = E0 − RT

zFlnK . (2.15)

The Nernst equation now can be used to derive an expression for the po-tential of any electrochemical cell or, in our case, electrochemical sensor.Electrochemical reactions can be triggered by applying currents or voltagesvia electrodes to a sensing layer.

After this short excursion into thermodynamics and kinetics, the differentmicrofabrication techniques and the fundamentals of CMOS-devices will bedetailed in Chap. 3.

3 Microtechnology for Chemical Sensors

Microtechnology and microfabrication processes are used to produce deviceswith dimensions in the micrometer to millimeter range. Microfabricationprocesses can be effectively applied to yield a single device or thousands of de-vices. The so-called “batch processing”, i.e., the fabrication of many devicesin parallel, does not only lead to a tremendous cost reduction, but also enablesthe production of array structures or large device series with minute fabri-cation tolerances. Microfabrication processes hence significantly differ fromconventional machining processes, such as drilling or milling with mechanicaltools. Integrated circuit (IC) fabrication processes are the most important mi-crofabrication processes [36–38]. The success of CMOS-technology, which isone of the enabling technologies of the information age, clearly demonstratesthe efficiency of microfabrication technologies.

Standard processing steps originating from semiconductor technology canbe used in combination with dedicated micromachining steps to fabricatethree-dimensional mechanical structures, which form the basis for the chem-ical microsensors detailed in Chap. 4. Key advantages of microfabricatedchemical sensors include small device size and sampling volume, the pos-sibility of batch processing, and the reproducibility of transducer/sensorcharacteristics due to the precise geometric control in the fabrication steps.Microfabrication techniques also can be used to either significantly im-prove sensor characteristics in comparison to conventionally fabricated de-vices, or to develop devices with new functionality that cannot be realizedin conventional fabrication technology. Microsensor success stories, such asmicromachined pressure sensors and accelerometers, show that microfab-rication techniques are especially suitable for high-volume applications in,e.g., automotive industry. In high-volume production, the advantage of batchprocessing is paramount, and the high development and setup costs amortize.

This chapter is organized in the following way: Microsystem substratematerials and standard processing steps originating from semiconductor tech-nology are detailed in the first sections, followed by a short introduction toCMOS technology and a description of micromachining and layer-depositionprocesses 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, themost common substrate material in microfabrication. It is supplied as singlecrystal wafers with diameters from 100–300 mm. The use of silicon substratematerial enables the co-integration of transducers and circuitry, which is usedto advantage, e.g., in realizing CMOS-based microsystems [39]. Besides itsfavorable electrical properties, single crystal silicon also has excellent physicalproperties (mechanical strength, thermal conductivity) [40], which enablethe design of micromechanical structures. Therefore, silicon is also the mostcommon substrate material for microfabricated chemical and biosensors. Alarge number of micromachining techniques have been developed to structuresilicon substrates (see also Sect. 3.4) [26–29,41–43].

Glass exhibits attractive dielectric and optical properties. Glass is alsosupplied in wafer form in different compositions (e.g., quartz, fused silica,and borosilicate glass) and diameters. Since glass is transparent for visiblelight, it is particularly suited for devices with optical detection principles.Single-crystal quartz with its hexagonal lattice structure is piezoelectric andis, therefore, used, e.g., as substrate material for acoustic-wave devices (seeSect. 4.1). Last but not least, glasses are chemically inert and suitable forhigh-temperature applications. A number of micromachining techniques, suchas isotropic wet etching or anisotropic dry etching, have been developed tostructure glass.

Ceramics have been extensively used as substrate material for hybridmicroelectronics and in microelectronics packaging [44] . The standard ma-terial is alumina (Al2O3), other materials include beryllium oxide (BeO) andaluminum nitride (AlN). Their chemical inertness, biocompatibility, and me-chanical stability render ceramics a very interesting material for microsys-tems. Most microfabrication techniques for ceramic materials have beenadapted from microelectronics packaging processes.

Polymers have been more and more explored over the last years as aninexpensive substrate material. Due to the cost advantage, disposable de-vices, such as microfluidic arrays or microstructured biosensor assays, areoften based on polymers [45]. Special processes, such as hot embossing, injec-tion molding, laser machining, or stereolithography, have been developed tostructure 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 iden-tical with those used in integrated-circuit fabrication [36–38]: Deposition, pat-terning, doping and etching. The sequential application of these techniquesto build up a device layer by layer is illustrated in Fig. 3.1.

3.2 Fundamental Semiconductor Processing Steps 17

filmdeposition

photo-lithography

etching

doping

wafer

waferout

waferin

mask set

Fig. 3.1. Flow diagram of an integrated-circuit fabrication process using the fourbasic microfabrication techniques: Deposition, photolithography, etching, and dop-ing. Adapted from [36]

A thin layer, such as an insulating silicon dioxide film, is deposited ona substrate. A light-sensitive photoresist layer is then deposited on top andis patterned using photolithography. The pattern is then transferred fromthe photoresist layer to the silicon dioxide layer by an etching process. Afterremoving the remaining photoresist, the next layer is deposited and struc-tured, and so on. Doping of a semiconductor material by ion implantation canbe done directly after photolithography or after patterning an implantationmask (e.g., a patterned (sacrificial) silicon dioxide layer).

In the following, a brief overview on the four fundamental microfabricationsteps will be given. More details can be found in books on semiconductorprocessing [36–38].

3.2.1 Deposition

The two most common thin-film deposition methods in microfabrication arechemical vapor deposition (CVD), performed at low pressure (LPCVD), at-mospheric pressure (APCVD) or plasma-enhanced (PECVD), and physicalvapor deposition (PVD), such as sputtering and thermal evaporation. Typi-cal CVD and PVD film thicknesses are a few micrometers. Other techniquesinclude electroplating of metal films and spin- or spray-coating of polymericfilms such as photoresist. Both processes can yield film thicknesses from lessthan 1 µm up to several hundred micrometers.


Dielectric layers, predominantly silicon dioxide and silicon nitride, areused as insulating material, as mask material, and for passivation. Silicondioxide is either thermally grown on top of a silicon surface (thermal oxide) athigh temperatures in an oxidation furnace (900–1200C), or it is deposited ina CVD system like silicon nitride. CVD oxides are deposited at temperaturesbetween 300and 900C.

Metal layers are used for electrical connections, as leads or electrode ma-terial, for resistive temperature sensors (thermistors) or as mirror surfaces.Metals, which are widely used in the microelectronics industry, such as alu-minum, titanium, and tungsten, are routinely deposited by sputtering or byelectron-beam evaporation. Depending on the application, a large number ofother metals, including gold, palladium, platinum, silver or alloys, can be de-posited with PVD methods. Whereas aluminum has been the standard met-allization in IC-fabrication for many years, the state-of-the-art, sub-0.25 -µmCMOS technologies often feature copper metallizations due to their lowerresistivity and higher electromigration resistance as compared to aluminum.

Highly-doped polycrystalline silicon (polysilicon) is used as gate materialfor metal-oxide-semiconductor field-effect transistors (MOSFET), for elec-trodes and resistors, and as thermoelectric or piezoresistive material. Polysil-icon is usually deposited in a LPCVD process using silane (SiH4) as gaseousprecursor.

Polymers such as photoresist are commonly deposited by spin- or spray-coating. Similar techniques are also used to coat chemical sensors with sen-sitive polymer films [1, 4].

3.2.2 Patterning

Photolithography is the standard process to transfer a pattern, which hasbeen designed with a computer-assisted design (CAD) program, onto a cer-tain 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 witha patterned opaque layer (typically chromium) on the surface. Electron-beamlithography is then used to write the mask pattern from the CAD data. In thephotolithographic process, a photoresist layer (photostructurable polymer) isspin-coated onto the material to be patterned. Next, the photoresist layeris exposed to ultraviolet (UV) light through the mask. This step is done ina mask aligner, in which mask and wafer are aligned before the subsequentexposure step is performed. Depending on whether positive or negative pho-toresist was used, the exposed or the unexposed photoresist areas are removedduring the resist development process. The remaining photoresist acts as aprotective mask during the etching process, which transfers the pattern ontothe underlying material. Patterned photoresist can also be used as mask for asubsequent ion-implantation step. After the etching or ion-implantation step,the remaining photoresist is removed, and the next layer can be depositedand patterned.

3.2 Fundamental Semiconductor Processing Steps 19

substrate

UV masklight

photoresist

thin film

apply photoresist

expose photoresist

develop photoresist

transfer pattern

Fig. 3.2. Schematic of a photolithographic process sequence for structuring a thin-film 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 ontop 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. Byremoving the underneath photoresist, the thin film material on top is alsoremoved 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 usingliquid chemicals and dry etching using gas-phase chemistry. Both methodscan be either isotropic, i.e., provide the same etch rate in all directions,or anisotropic, i.e., provide different etch rates in different directions (seealso Sect. 3.4 Micromachining) [37,38]. The important criteria for selecting aparticular etching process encompass the material etch rate, the selectivity tothe material to be etched, and the isotropy/anisotropy of the etching process.An overview on various etching chemistries used in microfabrication can befound in [48].

Wet etching is usually isotropic with the important exception of anisotropicsilicon 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 betteretch selectivity for the material to be etched in comparison to accompanying


other materials. An example includes wet etching of silicon dioxide usinghydrofluoric-acid-based chemistry. SiO2 is isotropically etched in diluted hy-drofluoric acid (HF : H2O) or buffered oxide etch, BOE (HF : NH4F). Typ-ical etch rates for high-quality (thermally grown) silicon dioxide films are0.1 µm/min in BOE.

Dry etching, however, is often anisotropic, resulting in a better patterntransfer, as mask underetching is avoided (see Fig. 3.4, Sect. 3.4.1.1). There-fore, anisotropic dry etching processes, such as reactive-ion etching (RIE), ofthin film materials are very common in the microelectronics industry. In anRIE system, reactive ions are generated using a plasma and are acceleratedtowards the surface to be etched, thus providing directional etching char-acteristics. Higher ion energies typically result in more anisotropic etchingcharacteristics, but also lead to reduced etching selectivity.

Though a large number of dry etching chemicals and recipes exist, mainlyfluorine- or chlorine-based etching chemistry is commonly used [37,38].

3.2.4 Doping

Doping is used to modify the electrical conductivity of semiconducting mate-rials such as silicon or gallium arsenide [36–38]. It is hence the key process stepto fabricate semiconductor devices such as diodes and transistors. In the caseof silicon, doping with phosphorus or arsenic yields n-type silicon, whereas p-type silicon results from boron doping. By varying the dopant concentrationof n-type silicon from 1014 to 1020 cm−3, the resistivity at room temperaturecan be tuned from approximately 40 Ω cm to 7·10−4 Ωcm. Dopant atoms areintroduced by either ion implantation or diffusion from a gaseous, liquid, orsolid source. Ion implantation allows for introducing precisely defined quan-tities of dopants into the semiconductor material and is, hence, a key processof microelectronics fabrication. The substrate material, e.g., a silicon wafer,is bombarded with accelerated ionized dopant atoms in an ion implanter. Theresult is an approximately Gaussian distribution of the dopant atoms in thesubstrate wafer with a mean penetration depth controlled by the accelerationvoltage. A high-temperature process is then used for annealing and activatingthe dopants in the case of ion implantation or for “driving-in” the dopantatoms until a desired doping profile has been achieved in the case of utilizingdiffusion processes [36–38].

3.3 CMOS Technology

CMOS is the dominant semiconductor technology for microprocessors, memo-ries and application-specific integrated circuits (ASICs) [36–38]. CMOS-chipsgenerally consist of a substrate, the transistor components, the metal layersand a passivation layer on top. The substrate is a silicon wafer, the thick-ness 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 re-gions, which form together with two polysilicon layers (e.g., transistor gateregions) and silicon oxide layers the transistor structures as defined duringthe CMOS process. Up to 8 metal layers consisting of aluminum (down to afeature size of 0.18 µm) or copper (0.13 µm and 0.09 µm CMOS) are used towire the electronic components and to establish connections to the outsideworld (bondpads). Intermetal oxide (Si-oxide) layers are used as electricalinsulator between the different metal layers. Finally, silicon nitride, siliconoxinitride, or silicon oxide layers passivate the device and protect the elec-tronics (Fig. 3.3). The overall CMOS device is fabricated in a defined sequenceof material deposition, doping, lithography and etching steps [36–38].

p-substrate

n-well fieldoxide

contactoxide

gateoxide

polysiliconmetal 2 metal 1intermetal

oxidepassivation

nitride

S D D

PMOS

n+

p+NMOS

S

Fig. 3.3. Cross-section of a n-well, double-metal CMOS-chip. A p-doped siliconwafer substrate that includes n-well implantations hosts the NMOS and PMOStransistors. Two metal layers, in this case aluminum, are used for wiring the elec-tronic components (electrical isolation by intermetal oxide). On top is a siliconnitride passivation as protection layer

Undoped semiconductor materials conduct electricity but not enthusiasti-cally. Semiconductor areas that are doped become conductors of either extraelectrons with a negative charge (n-doping, e.g., by phosphorus) or of positivecharge carriers (p-doping, by e.g., boron). The denomination “complemen-tary” metal oxide semiconductor is owing to the fact that both, n-channeland p-channel transistors are realized on the same substrate. The substrateis, 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 whichn-wells and p-wells are implanted in the wafer substrate material. Both tran-sistor types, n-channel (NMOS) and p-channel (PMOS) transistors are usedto realize logic functions. An exemplary CMOS-chip cross-section (0.8 µm n-well, double-metal CMOS process as used for most of the devices in Chap. 5)


is shown in Fig. 3.3. The p-substrate exhibits n-well areas created by implan-tation. Heavily p-doped structures (p+) in the n-well and heavily n-dopedstructures (n+) in the p-substrate form transistor source and drain. Thetransistor gate is made of polysilicon on top of the gate oxide (SiO2).

By applying an appropriate voltage (positive for NMOS, negative forPMOS) to the gate via the polysilicon, the substrate majority charge car-riers (electrons in n-well, holes in p-substrate or p-well) are depleted in thesurface area below the gate owing to the field effect, and a conducting chan-nel 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-draincurrent. The modulation is continuous within a certain range in analog cir-cuits, and produces only two states, “on” or “off”, in digital circuits. Themetal layers (Fig. 3.3: metal 1 and metal 2, aluminum) are used to wire thetransistors. Dielectric layers such as gate oxide, field oxide, contact oxide, andintermetal oxide (SiO2) serve as electrical insulation between conducting orsemiconducting layers.

The silicon nitride or other materials (oxinitride, oxide) on top serve aspassivation and provide electrical and mechanical/chemical protection of thecircuitry. Silicon nitride and, to a lesser extent, silicon oxide (or silicon witha native oxide layer) are durable in case of liquid exposure and are biocom-patible, which principally allows for using CMOS chips also with cells orliving material. CMOS aluminum, however, is neither stable in liquids or inair at higher temperatures nor is it biocompatible, and, consequently, hasto be covered with noble-metal coatings such as gold or platinum for manymicrosystem applications.

For more details on semiconductor technology see, e.g., the standard text-books 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 usedin bio(chemical) sensors. Membranes provide, e.g., the thermal isolation re-quired for thermal chemical sensors, whereas cantilevers can be used as reso-nant structures for mass-sensitive chemical sensors (see Chap. 4.1.3). In thefollowing, the fundamental micromachining techniques are briefly reviewed.More details on micromachining techniques can be found in dedicated bookson microsystem technology [25–29, 41, 42]. A recent review on microfabrica-tion in biology and medicine can be found in [43].

The micromachining techniques are categorized into bulk micromachin-ing [49] and surface micromachining processes [50] (see Fig. 3.4). In thecase 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 micro-machining, the microstructure consists of thin-film layers, which are depositedon top of a substrate, and which are selectively removed in a defined sequenceto yield the MEMS structure.

3.4.1.1 Bulk Micromachining

One approach to enhance the functionality of IC-based devices includesmicromachining the bulk substrate, which, in most cases, consists of sili-con. Bulk micromachining techniques can be classified into isotropic andanisotropic etching techniques (structure geometry), or into wet and dryetching 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 pref-erentially etched away along certain crystal planes while it is preserved inother directions (Fig. 3.4a).

(a) bulk micromachining (b) surface micromachining

substrate (Si) sacrificial layer

structural layer microstructure

substrate (Si)

isotropic etching

anisotropic etching

substrate (Si)

Fig. 3.4. Micromachining techniques: (a) Bulk micromachining, anisotropic andisotropic etching, (b) surface micromachining with sacrificial layer, structural layerand a subsequent etch step

The most common isotropic wet silicon etchant is HNA, a mixture ofhydrofluoric acid (HF), nitric acid (HNO3), and acetic acid (CH3COOH):Nitric acid oxidizes the silicon surface, and hydrofluoric acid etches the grownsilicon dioxide layer. The acetic acid controls the dissociation of HNO3, whichprovides the oxidation of the silicon. The etch rates and the resulting surfacequality strongly depend on the chemical composition [41,42,49].

Anisotropic wet etching of silicon is the most common micromachiningtechnique and is used to release, e.g., membranes and cantilevers for chemi-cal and biosensors. Anisotropic silicon etchants etch single-crystal silicon at


different etch rates in distinct crystal directions. The etch grooves are lim-ited 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〉 planesare intersecting the wafer surface at an angle of 54.7, so that the typicalpyramid-shape etch grooves as shown in Fig. 3.4 are formed. Mask materialsfor anisotropic silicon etchants are silicon dioxide and silicon nitride. It isimportant to note, that “convex” corners of the etch mask are underetchedin case of 〈100〉 silicon substrates, leading to, e.g., completely underetchedcantilever structures. The etch rates in preferentially etched crystal direc-tions such as the 〈100〉 and the 〈110〉 direction, and the ratio of the etchingrates in different crystal directions strongly depend on the exact chemicalcomposition of the etching solution and the process temperature [51].

The most common anisotropic silicon etching solution is potassium hy-droxide, KOH. As an example, a 6-molar KOH solution at 95C 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 inKOH solution is rather high (for thermal oxide approx. 1µm/hour in 6-molarKOH solution [42]), silicon nitride films are often used as etching mask. KOHsolution is very stable, yields reproducible etching results, is relatively inex-pensive, and is, therefore, the most common anisotropic wet etching chemicalin industrial manufacturing. The disadvantages of KOH include the relativelyhigh SiO2- and Al-etch rates, which require a protection of, e.g., integrated-circuit structures during etching. Etching with KOH is typically performedfrom the back side of the wafer, with the front side protected by a mechanicalcover 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 typeS, exhibit relatively low Al- and SiO2-etch rates, which render them suitablefor releasing microstructures on the front side of CMOS-wafers [53]. Moredetailed discussions of wet etching of silicon can be found, e.g., in [26] and [42].

Reliable etch stop techniques are very important for achieving repro-ducible 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 mag-nitude, as soon as a 〈111〉 silicon plane or a silicon dioxide/nitride layer isreached. In addition, the etch rate is greatly reduced in highly boron-dopedregions (doping concentration ≥1020 cm−3). The etching can also be stoppedat a p-n-junction using a so-called electrochemical etch stop technique (ECE)[41]. This method has been extensively used to release silicon membranes andn-well structures (see, e.g., Chap. 5.4). ECE relies on the passivation of siliconsurfaces through application of a sufficiently high anodic potential with re-spect to the potential of the etching solution.

Isotropic dry etching of silicon is done with xenon difluoride, XeF2. Thisvapor-phase etching method exhibits excellent etch selectivity with respect


to aluminum, silicon dioxide, silicon nitride, and photoresist, all of which canbe 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 valuesof approx. 1µm/min [54].

Anisotropic dry etching of silicon is usually carried out as reactive-ion-etching (RIE) with plasma-assisted etching systems. By controlling theprocess parameters, such as process gases and process pressure, the etchingcan be rendered either isotropic or anisotropic. The dry-etching anisotropyoriginates from experimental parameters such as the direction of the ionbombardment, and is, therefore, independent of the crystal orientation ofthe substrate material. Most bulk etching of silicon is accomplished usingfluorine free radicals with SF6 as a typical process gas. Adding chlorofluoro-carbons results in polymer deposition in parallel with etching, which leads toenhanced 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 sys-tems rely on high-density plasma sources and an alternation of etching andpolymer-assisted sidewall protection steps. In a process known as the “Bosch”process [55], a mixture of trifluoromethane and argon is used for polymerdeposition. Due to the ion bombardment, the polymer deposition on the hor-izontal surfaces can almost be prevented, while the sidewalls are passivatedwith a TeflonTM-like polymer. In the second process step, an SF6-based etch-ing chemistry provides silicon etching in the non-passivated regions, i.e., thehorizontal surfaces. Both process steps are alternated, resulting in typicalsilicon 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 thoughan exceptional anisotropy can be achieved, which is independent of the crys-tal orientation, one should keep in mind that deep RIE systems are by farmore expensive than a simple wet-etching setup, and that only one wafer isprocessed at a time [56,57].

3.4.1.2 Surface Micromachining

Surface micromachining comprises a number of techniques to produce mi-crostructures from thin films previously deposited onto a substrate and isbased on a sacrificial-layer method (Fig. 3.4b). In contrast to bulk microma-chining, surface micromachining leaves the substrate intact [41,50]. A sacrifi-cial layer is deposited and patterned on a substrate. After that, a structuralthin film, in most cases polysilicon, is deposited and patterned, which willperform the mechanical or electrical functions in the final device. A selectiveetchant then removes exclusively the sacrificial-layer material. The thicknessof the sacrificial layer determines the distance of the structural parts fromthe substrate surface. Common sacrificial-layer materials include silicon oxideetched by hydrogen fluoride and aluminum etched by a mixture of phosphoric,nitric and acetic acid.


Clamped beams, microbridges, or microchannels can be fabricated thisway, microrotors and even microgears can be realized by repeated layer de-position and etching [25–29,41].

3.4.2 Wafer Bonding

While the majority of microstructures are fabricated from a single sub-strate or wafer, several wafers can be joined by wafer bonding [58]. Thesubstrates/wafers are bonded onto each other either directly or via an inter-mediate layer.

Direct bonding techniques include silicon fusion bonding and anodic bond-ing . Silicon fusion bonding involves two silicon wafers, which are bonded toeach other at high temperatures (T≥ 1000C). Low-temperature (T < 400C)fusion bonding has been demonstrated after using special cleaning proce-dures, but exhibits reduced bonding strength. Anodic bonding of a sodium-rich glass wafer onto a silicon wafer is accomplished by applying an elec-tric field across the bonding interface at moderate bonding temperatures(T ≈ 300–400C). Glass materials with a thermal expansion coefficient sim-ilar to that of silicon (e.g., Pyrex glass 7740) are used for anodic bonding inorder to minimize thermomechanical stress. Anodic bonding and silicon fu-sion bonding require very clean and smooth wafer surfaces to achieve void-freebonding. The surface quality and roughness is less important if an interme-diate layer is used for wafer bonding. Possible intermediate bonding layersinclude adhesives, low-melting-temperature glass, solder films, or metallicfilms 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 com-pleted by various deposition techniques for chemically or biologically sensitivelayers. Chemically sensitive polymer layers or organic molecules that are usedfor the detection of volatile organics in air can be deposited by, e.g., dispens-ing, spray coating or by using self-assembled monolayers (SAMs). Metal-oxidefilms for the detection of, e.g., carbon monoxide and nitrogen oxides can bedeposited either by a sol/gel process, by drop coating, or by sputtering [59].

Recently, microcontact printing or soft lithography [60] has been intro-duced as an additional method for pattern transfer. A soft polymeric stampis used to reproduce a desired pattern directly on a substrate. Feature sizeson the order of 1 µm can be routinely achieved with this technique. The poly-mer stamp, often made from poly(dimethylsiloxane) (PDMS), is formed by amolding process using, e.g., a silicon master fabricated with conventional mi-crofabrication techniques. After “inking” the stamp with the material to beprinted, the stamp is brought in contact with the substrate material, and the


pattern of the stamp is reproduced. Surface properties of the substrate thuscan be modified to, e.g., locally promote or prevent molecule adhesion. Softlithography has been specifically developed for biological applications such aspatterning 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 trans-ducer (see Chap. 1, Fig. 1.2) [1,4–9]. Upon interaction with a chemical species(absorption, chemical reaction, charge transfer etc.), the physicochemicalproperties of the coating, such as its mass, volume, optical properties orresistance reversibly change. These changes in the sensitive layer propertiescan be detected by a variety of transducers and can be translated into elec-trical signals such as frequency, current, or voltage changes, which are thensubjected to further data treatment and processing.

As already mentioned in the introduction to this book, chemical sen-sors can be classified into four principal categories according to their trans-duction principles [4, 5]: (a) chemomechanical sensors (b) thermal sensors(c) optical sensors, and (d) electrochemical sensors. Each of those four sensorcategories will be briefly introduced, and, then, specific exemplary microfab-ricated transducers and devices will be abstracted. The transducer overviewwill be restricted to semiconductor-based and CMOS-based devices and will,for the sake of completeness, also include short abstracts on devices, which aredescribed in much more detail in the subsequent Chap. 5 on the CMOS tech-nology platform for chemical sensors. Typical chemically sensitive materialsand sensor applications will be brought up in the context of the respectivetransducer structures.

4.1 Chemomechanical Sensors

The change in mechanical properties (e.g., mass) of a sensitive layer uponinteraction with an analyte can be conveniently recorded by using micro-mechanical structures. Any species that can be immobilized on the sensorcan, in principle, be sensed. As with most of the chemical sensors (excludingthermal sensors), the measurements are performed at a thermodynamic equi-librium 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 byJan 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 gravi-metric sensors responding to the mass of species accumulated in a sensinglayer [61–63]. Some of the sensor devices additionally respond to changes in avariety of other mechanical properties of solid or fluid media in contact withtheir 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 sensi-tivity: mass changes in the picogram range can be detected, and ppm (partsper million) to ppb (parts per billion) detection levels have been reportedfor, e.g., gas and vapor sensors [61–63]. The large number of chemical speciesthat 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 asquartz, lithium tantalate or niobate, aluminum nitride, zinc oxide and others.Piezoelectricity results in general from coupling of electrical and mechanicaleffects. The prerequisite is an anisotropic, noncentrosymmetric crystal lattice.Upon mechanical stress, charged particles are displaced and thus generate ameasurable electric charge in the crystal. In turn, mechanical deformationscan be achieved by applying a voltage to such a crystal (for details, see [4]).Using an alternating current (AC), the crystals can be electrically excitedinto a fundamental mechanical resonance mode. The resonance frequency,which is the recorded sensor output in most cases, changes in proportion tothe mass loading on the crystal or device. The more mass (analyte mole-cules) is absorbed, e.g., in a polymer coated onto a piezoelectric substrate ortransducer, the lower is the resonance frequency of the device:

∆f = −C f20 ∆m/A . (4.1)

This equation was published by Sauerbrey in 1959 [64]. ∆f here denotes thefrequency shift due to the added mass in [Hz], C is a constant, f0 is thefundamental frequency of the quartz crystal in [Hz] and ∆m/A is the surfacemass loading in [g cm−2]. The following equation describes the relationshipbetween analyte gas phase concentration change, ∆cA, and the responses ofmass-sensitive sensors:

∆fA = Γ · MA · K · ∆cA . (4.2)

Here, ∆fA [Hz] denotes the frequency shift (sensor response) measured uponexposure to analyte at a concentration cA [mol/L]. MA [kg/mol] is the molarmass of the analyte vapor, K is the partition coefficient (2.10), and Γ is agravimetric constant [L/kg·s] including, e.g., the frequency shift measuredupon initial deposition of the sensitive layer, the coating density, transducerdimensions, etc.

A typical signal of a gravimetric sensor is displayed in Fig. 4.1 showingthe frequency shifts of a resonant cantilever coated with poly(etherurethane),

4.1 Chemomechanical Sensors 31

-100

-80

-60

-40

-20

0

20

500 100 150 200

analyte: n-octanepolymer: PEUT

frequ

ency

shi

ft [H

z]

time [min]

250

250

500

500

750

750

1000

1000

1250

1250

1500

1500

ppm

Fig. 4.1. Typical responses of mass-sensitive sensors. Frequency shifts of apolymer-coated (poly(etherurethane), PEUT) cantilever upon exposure to differ-ent concentrations2 (250–1500 ppm) of an organic volatile: n-Octane

(PEUT), upon exposure to various concentrations of n-octane. At low ana-lyte concentrations (trace level), a linear correlation between the frequencyshift due to analyte absorption and the corresponding analyte concentrationin the gas phase is usually observed (Fig. 4.1), provided that the sensing filmon the transducer moves synchronously with the oscillating crystal surface.Significant deformations across the film thickness result in a more complexrelationship between mass changes and resonant frequency due to, e.g., vis-coelastic effects (concept of “acoustically thin and thick” films as detailedin [65]).

The most common devices are the thickness-shear-mode resonator (TSMR)or quartz microbalance (QMB), a bulk resonator, and the Rayleigh surface-acoustic-wave (SAW) device, both based on quartz substrates. The QMB wasdemonstrated to function as an organic vapor sensor by King in 1964 [66],the SAW device became popular after introducing interdigital transducers to2 Concentration are usually given in mole-per-volume (mol/m3) or mass-per-

volume (kg/m3) units. In gas sensorics ppm-units (parts per million volumeor parts per million pressure) are widely used, which are strictly speaking novalid concentration units, in particular since ppm-units are dimensionless. Theppm-units are nevertheless used here owing to their popularity. Assuming thevalidity 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) bydivision through the molar volume of an ideal gas at 25C (24.45 l) and multipli-cation with the molar mass of the analyte compound. Moreover, under normalpressure conditions 10 ppm analyte correspond to 1 Pa partial pressure of therespective analyte.


acoustic sensors in 1970 [67]. Since the TSMR is not semiconductor-based andnot compatible with IC technology, it will not be treated here any further.

Shear-horizontal-acoustic-plate-mode (SH-APM) devices devices, shear-transverse-wave device (STW) and Love-wave devices devices require quartz,lithium niobate or lithium tantalate substrates [61–63] and, hence, will notbe dealt with here as well. For details and further information it is referredto a wealth of literature [61–63,68–70].

Silicon is not a piezoelectric material. The realization of silicon-basedpiezoelectric transducers hence requires an additional piezoelectric layer to bepatterned on the silicon. Different materials have been used such as cadmiumsulfide [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 ofmass-sensitive devices will be described in more detail: (1) SAW-devices on Si-substrates with piezoelectric overlay, (2) flexural-plate-wave devices (FPWs),and (3) micromachined cantilevers. Operability in gas or liquid media, typicalcoating materials, target analytes, and applications will be discussed in thecontext of each transducer. An overview of micromachined resonant sensorsis given in [70,77].

4.1.1 Rayleigh SAW Devices

Transduction Principle and Sensing CharacteristicsInterdigital transducers can be used to launch and detect a surface-acousticwave on a piezoelectric substrate [67] as schematically shown in Fig. 4.2.By applying an AC voltage to a set of interdigital transducers patternedon a piezoelectric substrate with appropriate orientation of the crystal axes,

Rayleigh SAW top view

metal electrodes side view

substrate: silicon

ZnO

excitation

propagation

detection

interdigitated electrodes

wavepropagation

particledisplacement

Fig. 4.2. Launching, propagation and detection of a Rayleigh-type surface acousticwave by interdigitated transducers on a zinc-oxide-covered silicon substrate. Thetop view shows the electrode configuration and the wave propagation. The sideview shows the elliptical particle displacement


one set of the fingers moves downwards, the other upwards, thereby creat-ing an oscillating mechanical surface deformation. This surface deformationgenerates an acoustic wave, which propagates along the surface and is con-verted back into an electrical signal by deforming the surface in the region ofthe receiving transducer. The electrical signal of the receiving transducer isrecorded and represents the sensor signal.

For a given piezoelectric substrate, the acoustic wavelength and, thus, theoperating 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 frequen-cies require a sophisticated high-frequency circuit design. Therefore, a barereference oscillator is operated together with the sensor in many cases, andthe outputs are mixed to produce a difference frequency with values in thekHz-range that is recorded [74,76].

The acoustic wave is confined to a surface region of approximately oneacoustic wavelength thickness. The velocity and damping characteristics ofthe acoustic wave hence are extremely sensitive to changes at the transducersurface. When used in an oscillator circuit, relative changes in the wave ve-locity are reflected as equivalent changes in fractional oscillation frequency.A change in mass due to, e.g., absorption of a gaseous analyte in a polymericsensing layer thus changes the device frequency according to (4.1).

The acoustic (Rayleigh) wave causes an elliptical particle movement atthe transducer surface (Fig. 4.2), i.e., the sensitive films deposited on top ofthe transducers and the piezoelectric substrate are severely deformed. Thus,additional effects such as changes in viscoelastic properties of the sensinglayer can affect the sensor response [65].

FabricationSince there exists a variety of custom-designed semiconductor and siliconprocesses in the literature, only the transducer-related additional fabricationsteps after semiconductor processing and the sensitive-layer-processing stepswill be listed in this chapter. Industrial standard IC processes like CMOSfabrication will be explicitely identified, especially since CMOS-based de-vices are extensively treated in Chap. 5. For more details on semiconductorand MEMS process steps, see Chap. 3.

Fabrication Steps:

• Optional back etching using potassium hydroxide, KOH, or ethylene-diamine-pyrocatechol, EDP, to achieve a membrane structure [74,78]).

• Zinc oxide processing: Deposition mainly by sputtering techniques at 150–450C. Highly oriented layers of 5–50 µm with a high degree of surfaceflatness [74].

• Electrode processing: Vacuum evaporation of aluminum or gold, layerthickness >200 nm.


• Sensitive layer: Spin or spray coating of polymers, organic layers, or bio-logical entities.

ApplicationsSince surface-normal particle displacements occur (Fig. 4.2), and the acousticwave velocity is larger than the compressional velocity of sound in water, thedevice radiates compressional waves into the liquid phase, which causes severeattenuation. Rayleigh SAW devices hence cannot be used in liquids [61–63].

Typical applications areas are environmental monitoring or personalsafety devices. This includes the detection of different kinds of organicvolatiles (hydrocarbons, chlorinated hydrocarbons, alcohols, etc.) by usingpolymeric layers [74] or porphyrins [79], and the detection of nitrogen diox-ide using phthalocyanines [76].

The interaction mechanisms involve, in most cases, fully reversible physi-sorption and bulk/gas phase partitioning (see 2.10, 4.2).

Integrated Gallium Arsenide (GaAs) SAW SensorGaAs is a well-developed semiconductor device material for fabricating high-frequency integrated circuits, and GaAs is piezoelectric. The piezoelectricproperties of GaAs and, hence, the device characteristics are similar to thoseof quartz except for the strong temperature dependence. An integrated GaAs-SAW sensor is shown in Fig. 4.3 [80]. It consists of a 470 MHz SAW devicealong with a multistage amplifier (4 gain stages and impedance matching

470 MHz delay line

4-gain-stage amplifier output stage

Fig. 4.3. Micrograph of a monolithically integrated GaAs surface-acoustic-wavedevice showing the delay line, the amplifier and the output stage. Reprinted from[80] with permission


output stage) forming a monolithic oscillator circuit thus eliminating theneed for high frequency interconnections [80].

4.1.2 Flexural-Plate-Wave or Lamb-Wave Devices

Transduction Principle and Sensing CharacteristicsFlexural-plate-wave devices have been introduced in 1988 [81]. Their chief ad-vantage 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 fewpercent of an acoustic wavelength thick (typically 2–3 µm). The plates arecomposite structures (Fig. 4.4) consisting of a silicon nitride layer, an alu-minum ground plane, a sputtered zinc oxide piezoelectric layer, all supportedby a silicon substrate [61,63,81,83].

FPW

membrane bottom

side viewtop viewsiliconframe

compositemembrane

siliconnitride

aluminum ZnO

wavepropagation

particledisplacement

metalelectrodes

compositemembrane

Fig. 4.4. Schematic of a flexural-plate-wave device. The side view shows the dif-ferent layers and the membrane movement. Interdigitated electrodes are used foractuation

The interdigital transducers (IDTs) on these devices generate flexuralwaves (Lamb waves, Fig. 4.4) with retrograde elliptical particle motions asin the SAW devices. However, the velocity in the membrane is much lessthan in a solid substrate, and the operating frequency for a given transducerperiodicity is, hence, considerably lower [61–63,70,81]. The Lamb waves giverise to a series of plate modes, one of which has a frequency that is muchlower than those of the other possible modes. The velocity of this unique wavedecreases with decreasing plate thickness. The entire thickness of the plate isset in motion like the ripples in a flag [61–63, 70, 81, 83]. The confinement ofacoustic 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 trans-ducer surface (Fig. 4.4), the sensitive films are deformed as it is the case with


the SAW. The frequency, however, is much lower, and, therefore, changes inviscoelastic properties of the sensing layer do not severely affect the sensorresponse.

The sensitive layer can be deposited on either side of the membrane. De-position 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 FPWMagnetic excitation requires an externally applied magnetic field, but elim-inates the need for a piezoelectric layer, which frequently contains elements(Zn, etc.) that pose contamination problems in IC fabrication. The deviceconsists of a silicon nitride membrane suspended in a silicon frame. A metalmeander-line transducer is patterned on the membrane surface (Fig. 4.5).Alternating current flowing in the transducer interacts with a static in-planemagnetic field to generate time-varying Lorentz forces (Fig. 4.5). These de-form the membrane, exciting it into a resonant mode [70, 85]. To efficientlyexcite the mode, the current lines of the transducer must be positionedalong lines of maximum mode displacement (Fig. 4.5). This requires a crit-ical alignment between the top metallization pattern and the backside etchmask [70,85].

siliconsubstrate

silicon nitridemembrane

B (static in-plane magnetic field)

Imeander-linetransducer

λ

FPW

Fig. 4.5. Schematic representation of the magnetically excited flexural-plate-wavedevice. Lorentz forces are generated between an impressed alternating current in aserpentine conductor and a static in-plane magnetic field. Reprinted from [85] withpermission

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).


• 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 theLamb device with electronics are detailed in [83,87].

ApplicationsSurface-normal particle displacements occur (Fig. 4.4), but the acoustic-wavevelocity is much less than the compressional velocity of sound in water. FPWdevices thus can be used in the liquid phase [61,63,83,88].

Typical application areas are environmental monitoring (gas and liquidphase) or biosensing in liquids. These include the detection of different organicvolatiles in the gas phase (hydrocarbons, chlorinated hydrocarbons, alcohols,etc.) by using polymeric layers [81, 84, 89–91], the detection of the weightpercentage of alcohol/water [83] or glycol/water [88] mixtures and the use ofan FPW-based immunoassay for the detection of breast cancer antigens [92].

The interaction mechanisms involve reversible physisorption and bulk/gasphase partitioning (see 2.10, 4.2) as well as antigen/antibody binding [92].

4.1.3 Resonating Cantilevers

Transduction Principle and Sensing CharacteristicsMicromachined cantilevers commonly employed in atomic force microscopy(AFM) constitute a promising type of mass-sensitive transducer for chem-ical sensors [93–107]. The sensing principle is quite simple. The cantileveris a layered structure composed of, e.g., the dielectric layers of a standardCMOS process, silicon, metallizations, and eventually, add-on piezoelectriczinc oxide. The cantilever base is firmly attached to the silicon support. Thefreestanding cantilever end is coated with a sensitive layer (Fig. 4.6).

The excitation of a cantilever in the resonant mode is usually performedby applying piezoelectric materials (ZnO) [106] or by making use of the bi-morph effect, i.e., the different thermal expansion coefficients of the various

cantileversiliconframe

polymeric coating

Fig. 4.6. Schematic representation of a resonating cantilever


layer materials forming the cantilever [93–105]. This difference in materialproperties gives rise to a cantilever deflection upon heating. Periodic heat-ing pulses in the cantilever base thus can be used to thermally excite thecantilever 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 stresschanges by means of, e.g., laser-light reflection [100, 103–105], (b) dynamicmode: excitation of the cantilever in its fundamental mode and measurementof 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 methodsimpose completely different constraints on the cantilever design for maximumsensitivity. Method (a) requires long and deformable cantilevers to achievelarge deflections, whereas method (b) requires short and stiff cantilevers toachieve high operation frequencies. Method (b) is preferable with regard tointegration 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 piezore-sistors in the cantilever base [93–95,101,102], by measuring motional capac-itance changes [107], or by using optical detection by means of laser lightreflection on the cantilever [97–100, 103–105]. The mass resolution of thecantilevers is in the range of a few picograms [93–96, 103–105]. This highmass-sensitivity does not necessarily imply an exceptionally high sensitivityto 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, especiallysince 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 inte-gration of the cantilevers with electronics are detailed in [93–95,101,102,107].

ApplicationsThe 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 oforganic volatiles (hydrocarbons, chlorinated hydrocarbons, alcohols, etc., seeFig. 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 anddetection of complementary strands of oligonucleotides [104].

The interaction mechanisms involve reversible physisorption and bulk/gasphase 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 concen-tration of a chemical species by measurement of an enthalpy change producedby 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 surroundingsa certain quantity of heat (enthalpy term, ∆H0, in 2.9). Reactions liberatingheat are termed exothermic, reactions abstracting heat are termed endother-mic. This thermal effect shows a transient behavior: Continuous heat liber-ation/abstraction occurs only as long as the reaction proceeds. This impliesthat only a steady-state situation can be achieved: A chemical reaction isproceeding at a constant rate and is thus releasing/abstracting permanentlya constant amount of heat. There will be, however, no heat production, and,hence, no measurable signal at thermodynamic equilibrium (∆G = 0) incontrast 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 sametime, the sensing area should be thermally as isolated as possible.

The liberation or abstraction of heat is conveniently measured as a changein temperature, which can be easily transduced into an electrical signal. Allsensors aimed at thermal infrared radiation detection can, in principle, beused as chemical sensors as well. The various types of calorimetric sensorsdiffer in the way that the evolved heat is transduced. The catalytic sensor (of-ten 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 uti-lize the respective effects to measure the temperature change. In addition,there are thermal (flow) sensors based on the different thermal conductivityof gaseous analytes [145, 146]. The micromachined cantilever enabling mi-crothermal analysis due to the bimorph effect [103, 108] has already beenmentioned in Sect. 4.1.3.

Catalytic sensors, Seebeck-effect or thermoelectric sensors, pyroelectricsensors, and thermal conductivity sensors are semiconductor-technology-compatible. Since the latter is essentially a flow sensor responding to thermo-


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, noncentrosymmetriccrystal lattice, permanent polarization, creation of macroscopic charges dueto thermal stress in the crystal) require the deposition of pyroelectric material(lithium tantalate, zinc oxide [130], polycyclic organic compounds [131]) onthe silicon chip. There are very few chemical-sensing applications reported inliterature [34, 130, 131]. Therefore, this class of sensor will not be discussedany further.

Thermistors are temperature-sensitive bead resistors composed of eitheroxide semiconductors with a negative (NTC) or a positive (PTC) temperaturecoefficient; that is, their resistance decreases or increases nonlinearly withtemperature. PTC resistors are made, e.g., from barium or lead titanate,while NTC resistors are made from sintered transition metal oxides (titaniumoxide) doped with aliovalent ions. The beads are contacted via two metallic(platinum) leads and coated with glass for chemical inertness [4, 123]. Thethermistors themselves have not been fabricated in planar semiconductortechnology yet, but have been integrated into silicon-based biosensors dueto their small size [124, 125]. They have been used as thermal biosensors instrongly 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 bedetailed.

4.2.1 Catalytic Thermal Sensors (Pellistors)

Transduction Principle and Sensing CharacteristicsThe development of the catalytic sensor is derived from the need for a hand-held detector for methane to replace the flame safety lamp in coalmines. Thecatalytic device measures the heat evolved during the controlled combustionof flammable gaseous compounds in ambient air on the surface of a hot cat-alyst 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 alternativereaction paths involving intermediates with lower activation energies thanthe uncatalyzed mechanism (Fig. 4.7). The catalyst itself is not permanentlyaltered by the reaction.

The heated catalyst here permits oxidation of the gas at reduced tem-perature and at concentrations below the lower explosive limit (LEL). Threeelements are necessary for this method: A catalyst, a method to heat it, anda means to measure the heat of catalytic oxidation. The term “pellistor”originally refers to a device consisting of a small platinum coil embedded ina ceramic bead impregnated with a noble metal catalyst [110]. A ceramic


educts

products

Ainter-

mediates

A*

B

A*cat

reaction: A ↔ A* ↔ B

Fig. 4.7. Working principle of a catalyst: Provision of an alternative reaction pathwith less activation energy for the reaction A↔ A∗↔B via the intermediate stateA∗

cat

siliconframe

catalyst(Pd, Pt)

platinumresistors

supportingSi-oxynitridemembrane

Si-nitridepassivation

(a)

(b) 100 m

Fig. 4.8. (a) Cross-section of side-by-side microhotplates composed of a dielectricmembrane on an etched silicon wafer, and platinum resistors/heaters. The deviceon the left has a deposited catalyst making it the active element. Redrawn from[111]. (b) Micrograph (SEM) of two meandered polysilicon microbridges. The lowermeandered 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 compensatechanges in the ambient temperature, thermal conductivity and flow rate, whilethe lower filament is used to calorimetrically detect combustible gases. Reprintedfrom [113] with permission

bead is used since the rate of reaction (and thus the sensor signal) is directlyproportional to the active surface area.

Figure 4.8 shows two different micromachined designs to realize a catalyticcalorimetric sensor: A meander structure on a micromachined membrane [111]and a freestanding, Pt-coated polysilicon microfilament (10 µm wide, 2 µmthick) separated from the substrate by a 2-µm air gap [112,113]. Heat losses tothe silicon frame are minimized in these designs. By passing an electric currentthrough the meander, the membrane/microbridge is heated to a temperaturesufficient for the Pt surface to catalytically oxidize the combustible mixture;


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 translatesinto a corresponding temperature change. The structures described here arevery 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 com-bustible concentration when the device is operated in excess oxygen and inthe mass-transfer-limited regime [114, 115]. The combustion of hydrogen indry air is exemplified in Fig. 4.9 [112, 113]. The circuit maintains a constantsensor temperature by adjusting the supplied current to keep the filamentresistance at a constant value. Note that the sensor response is measuredat steady state, i.e., continuous combustion. In most realizations, the mea-suring resistor forms part of a Wheatstone bridge configuration [111, 114].Temperature–modulated operation has been reported in [116,117].

time [s]

sens

or r

espo

nse

[V]

hydrogen onPt-filament

200 400 600 800 1000 12000.45

0.5

0.55

0.6

1.6% 1.0% 0.5% 0.1%

Fig. 4.9. Sensor response of a Pt-coated filament exposed to various concentrationsof 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].

ApplicationsMain applications include monitoring and detection of flammable gas hazardsin industrial, commercial and domestic environments. The lower explosive


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]. Thedetectable 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 athigh temperature liberating the respective reaction enthalpy (2.9, 2.11).

4.2.2 Thermoelectric or Seebeck-Effect Sensors

Transduction Principle and Sensing CharacteristicsThis type of sensor relies on the thermoelectric or Seebeck-effect: When twodifferent semiconductors or metals are connected at a hot junction, and atemperature difference is maintained between this hot junction and a colderpoint, then an open-circuit voltage is developed between the different leadsat the cold point. This thermovoltage is proportional to the difference of theGalvani potentials (inner contact potential, difference of the Fermi levels ofthe two materials) at the two temperatures and, thus, proportional to thetemperature difference itself [132]. This effect can be used to develop a ther-mal sensor by placing the hot junction on a thermally isolated structure likea membrane, bridge, etc., and the cold part on the bulk chip with the ther-mally well-conducting silicon underneath [95, 133–135]. To achieve a higherthermoelectric voltage, several thermocouples are connected in series to forma thermopile. The membrane structure (hot junctions) is covered with a sensi-tive or chemically active layer liberating or abstracting heat upon interactionwith an analyte. The resulting temperature gradient between hot and coldjunctions then generates a thermovoltage, which can be measured.

Figure 4.10 displays the schematic of a CMOS thermopile. The sensorsystem relies on polysilicon/aluminum thermocouples exhibiting a Seebeckcoefficient 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 andpartitioning 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 non-equilibrium state (transients) as a consequence of changes in the analyte con-centration (for details, see Sect. 5.2.3). Therefore, the sensor provides a signalupon absorption (condensation heat) and desorption (vaporization heat) ofgaseous analytes into the polymer [134–138], or during chemical reaction ofan analyte with the sensing material [139–141].

The recorded thermovoltage change, ∆U [V], is, therefore, proportionalto the derivative of the analyte concentration as a function of time, dcA/dt[mol/m3s]:


p-substrateCMOSn-well

dielectricmembrane

polymer analytecold junctions

hotjunctions

Fig. 4.10. Schematic of a thermoelectric sensor. Polysilicon/aluminum thermopilesare used (hot junctions on the membrane, cold junctions on the bulk chip) to recordtemperature variations caused by analyte sorption in the polymer

∆U = A · B · Vsens · ∆H · K · dcA

dt. (4.3)

Here A [K·s/J] and B [V/K] are device- and coating-specific constants de-scribing 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).

Fabrication

• Back etching (KOH, or EDP) for membranes [95,133,135,142].• Processing of the sensitive layer: Airbrush [95, 135], dispensing, spin coat-

ing, or enzyme immobilization methods [139–141].

Processing sequences for the integration of thermoelectric sensors with cir-cuitry in a CMOS standard process are detailed in [95,135,142]. Sensors arecommercially available from Xensor [143].

ApplicationsTypical applications areas are environmental monitoring (gas and liquidphase) or biosensing in liquids. These include the detection of different or-ganic 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 ofglucose, urea and penicillin in the liquid phase by using suitable enzymes[134,139–141].

The interaction mechanisms involve reversible physisorption and bulk/gasphase 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 electromag-netic waves according to the principle of duality. The characteristic propertiesof the electromagnetic waves such as amplitude, frequency, phase, and/orstate 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 beingtermed a photon (h is Planck’s constant, 6.626 · 1034 Js, ν denotes the fre-quency):

E = h · ν . (4.4)

When light interacts with matter, several processes can take place, sometimessimultaneously.

AbsorptionIf a sample is irradiated with visible light or electromagnetic waves, the ra-diation can be absorbed, which results in a decrease of the intensity in thedetected radiation as compared to the primary beam (Fig. 4.11). Alterna-tively, the radiation can be transmitted without attenuation. A prerequisitefor absorption is that the absorbing matter (atom, molecule, etc.) exhibitsunoccupied energy states with an energetic difference exactly equal or lessthan the energy of the incoming radiation quanta. The matter then absorbsthe radiation energy by transition into a so-called excited state with higherinternal energy. The absorption of radiation forms the base for most tradi-tional spectroscopic methods, which are usually distinguished according tothe different radiation wavelengths or frequency ranges as given in Table4.1 [35].

Table 4.1. Different spectroscopy methods and radiation energies

Radiation Energy [J/mol] Wavelength [m] Transition

γ-radiation 109–1011 10−13–10−11 Nucleus excitationX-rays 107–109 10−11–10−9 Core electron excitationUltraviolet (UV) 106–107 10−9–10−7 Shell electron excitationVisible (VIS) 105–106 400–800 nm Shell electron excitationInfrared (IR) 102–105 10−6–10−4 Vibrational statesMicrowaves 10−2–102 10−4–1 Rotation statesRadio waves <10−2 >1m Electron spin, Nuclear spin

The absorption of monochromatic radiation (only one selected wave-length) can be quantitatively determined using the well-known Lambert-Beerrelation:

I = I0 · e−ελcAl . (4.5)

Here I denotes the transmitted radiation intensity at the detector, I0 the in-tensity of the incident radiation, ελ is the molar absorptivity at the measured


wavelength, cA the analyte concentration, and l the optical path length inthe probed volume.

ScatteringChanges of the direction and/or the frequency of light are commonly denotedas scattering (Fig. 4.11). Scattering of light does not necessarily involve atransition between quantized energy levels in atoms or molecules. A random-ization in the direction of light radiation occurs. Particles with sizes that aresmall compared to the wavelength of radiation give rise to Rayleigh scatter-ing, while particles that are large compared to the wavelength give rise toMie scattering [147,148]. In both processes the particle polarization is unal-tered. However, the incident radiation can promote vibrational changes (en-ergy quantum absorption), which can alter the polarization of the irradiatedparticle/molecule. The frequency of the light scattered by these moleculeswill be different from that of the incident light and the light intensity will bemuch lower. Such a phenomenon is known as Raman scattering [6, 35,150].

Fluorescence and PhosphorescenceThe mechanism of those two phenomena is an absorption-emission process.The wavelength or energy of the incident radiation is absorbed and promoteschanges in the molecular energy states. The resulting excited state is un-stable, and the molecule dissipates some of its energy to rotational and/orvibrational energy states. The molecule then can return into the ground stateby emitting light at a lower frequency than the incident radiation, this processbeing termed fluorescence. If a more complex and slower intersystem cross-ing process into a triplet state, and then, a radiative transition from thereto the ground state occurs, the process is called phosphorescence. For detailsand the respective Jablonski diagrams, which represent simplified portray-als of the relative positions of the electronic energy levels of a molecule, seefor example [4, 35, 147]. Both processes, fluorescence and phosphorescenceare sometimes subsumed under luminescence processes, but the term “lumi-nescence” here will be used exclusively for chemoluminescence processes asdetailed below.

Fluorescence processes are extensively used in gene analysis techniques,where a defined array of single-stranded deoxyribonucleic acid (DNA) frag-ments is hybridized with the respective complementary strands labeled witha fluorescent marker. By illuminating the array with a laser, the sites, wherethe labeled DNA fragments are bound by interaction between the two comple-mentary strands, can be detected by their positive fluorescence response [149].This technique has been commercialized by several companies [151].

ChemoluminescenceThe 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 accordingto:

A + B ⇒ C∗ ⇒ C + h · ν . (4.6)


Chemical energy is thus directly converted into light energy in most caseswithout additional heat generation (cold luminescence). In the biologicaldomain, this process is denoted bioluminescence and, e.g., occurs in glow-worms.

Reflection and RefractionReflection and refraction take place when light infringes on a boundary sur-face between two media of distinct optical properties (refraction index). Thelight 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 surfaceor generally at interfaces of media with no transmission through (evanescentwaves will be treated in the context of refraction below). Another type is dif-fuse reflection, where the light penetrates the medium and subsequently reap-pears at the surface after partial absorption and multiple scattering withinthe medium. The optical characteristics of diffusely reflected radiation pro-vide information on the composition of the reflecting medium [6].

Thin films (10 µm and less) on a surface can strongly affect the propa-gation of incident light due to reflection at each of the thin film interfacescausing a multitude of reflected, coherent beams with small phase shifts.Sensor techniques to interrogate such thin film structures include ellipsome-try [153, 154] and thin-film reflectometric interference spectroscopy (RIFS),which is based upon spectral modulation of the reflectance of a thin film un-der white-light illumination without using the polarization information. Thespectral characteristics are a function of the film thickness, and, therefore,any ad/absorption of organic matter leads to changes in the interferogramsof the reflected beams [155,156].

A variety of transducers respond to changes in the refractive index inimmediate vicinity to the device surface. The propagation behavior of a waveguided by nonmetallic total internal reflection (Fig. 4.11) in a medium ofhigh refractivity depends on the dielectric characteristics of the surroundingmedium. This effect is mediated by the evanescent field, which penetratesfrom the optically denser guiding medium a few hundred nanometers intothe optically rarer environment [149,157,158] (more details in Sect. 4.3.1 onintegrated optics, see also Fig. 4.12).

If the environment absorbs, energy will be transferred from the evanescentwave to the environment and attenuation of the traveling wave will occur (at-tenuated total reflection, ATR) [147,148]. The energy of the evanescent wavecan also be used to initiate fluorescence (total internal reflectance fluores-cence, TIRF) [157]. Light propagating in waveguide structures without ab-sorbing cover layers is not attenuated by environmental influences (frustratedtotal reflection, FTR) [147, 148]. Its propagation velocity, however, changesdepending on the refractive index in the vicinity of the waveguide. Several


absorption scattering

sample

light source

dete

ctor

J0 JL

evanescentwave

n2

metal

λ

air

n1

external reflection internal reflection

n1> n2

standingwave

TIRF

reflection and refraction

n1

n2

n1> n2

φ1

φ2

φr

reflected

refracted

φ1 = φr

Fig. 4.11. Schematic representation of the different processes taking place uponinteraction of light with matter: Absorption, scattering, reflection/refraction, ex-ternal 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 integratedoptics [147,149,158].

Surface plasmon resonance (SPR) is based on collective fluctuations inelectron density at the surface of thin films, typically gold or silver, on awaveguide. Surface plasmon waves show the maximum of the electrical fielddistribution located at the waveguide/metal interface, which is exponentiallydecaying into the metal and the adjacent medium. SPR is detected as a strongattenuation of the reflected light beam sensitive to the medium adjacent tothe 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 offera great deal of selectivity already inherent in the various transduction mecha-nisms. Characteristic properties of electromagnetic waves, such as amplitude,frequency, phase, and/or state of polarization can be used to advantage. Thewavelength of the radiation, e.g., can be tuned to specifically match the energyof a desired resonance or absorption process. Geometric effects (scattering)can provide additional information. Moreover, optical sensors, like any otherchemical sensor, can capitalize on all the selectivity effects originating fromthe use of a sensitive layer.

Optical sensors and classical spectroscopy methods are often very sim-ilar in methodology but differ in the arrangement of the experiment andequipment. In particular, the introduction of fiber-optic techniques has pro-moted 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


mechanisms. The wide field of glass-based and fiber-optical techniques suchas optodes [163] or micro optodes [164, 165] will not be covered. For inter-ested 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-effectdevices (Sect. 4.4.2.2.5).

4.3.1 Integrated Optics

Transduction Principle and Sensing CharacteristicsThe generation of light in silicon devices is very difficult since there is nofirst-order transition from the valence band to the conduction band with-out the involvement of a phonon (lattice vibrations) [36, 172]. Only direct-bandgap semiconductors like gallium arsenide (GaAs) or indium phosphide(InP) show first–order radiative electron-hole recombinations with high quan-tum efficiency (see section on GaAs devices below). The detection of light ispossible with either silicon-based devices (photodiodes) or other semiconduc-tor materials.

Integrated optical (IO) sensors make use of guided waves or modes inplanar optical waveguides. The waveguide materials usually include high-refractivity silicon dioxide or titanium dioxide and silicon nitride films onoxidized silicon wafer substrates. The guided waves or modes in planar op-tical waveguides include the TE (transverse electric or s-polarized, surface-normal) and the TM (transverse magnetic or p-polarized, surface-parallel)modes. Changes in the effective refractive index of a guided mode are in-duced by changes of the refractive index distribution in the immediate vicin-ity of the waveguide surface, i.e., within the penetration depth (some hundrednanometers) of the evanescent field in the sample (Fig. 4.12a) [149,158].

The evanescent field decays exponentially with increasing distance fromthe waveguide surface. Changes in the effective refractive index can be in-duced by absorption of an adlayer onto the surface of the waveguide from gasor liquid phase [158, 173, 174], by interaction of an analyte molecule with arecognition structure immobilized on the waveguide surface [158,175–179], orby changes of the refractive index of the medium adjacent to the waveguidein a flow-through configuration [158, 175, 176]. In the case of microporouswaveguides, analyte molecule absorption or desorption directly into the poresof the wave-guiding film itself can change the waveguide refractive index [158].

A number of different IO sensors have been developed to transform thechanges of the effective refractive index into readily measurable physicalquantities.

Grating CouplersAmong the first integrated optical devices were grating coupler structuresembossed 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


fluorophores (TIRF)evanescent

field

waveguide

sensitivelayer nl

nw

(a) evanescent wave

gratingwaveguide

sensitivelayer

φ

liquidsample protection

flowcell

(b) grating coupler

ns < nl < nw

substratens

substrate

Fig. 4.12. (a) Schematic of an evanescent wave in an optical waveguide. When flu-orophores are within the reach of the evanescent wave, they can be excited, and thefluorescence can be detected (TIRF). (b) Schematic of a grating coupler. A periodicgrating on the surface of the waveguide is used for in- or out-coupling of radiationto/from the waveguide. The deflection angle depends on the light wavelength andthe grating period and is altered by binding of an analyte on the grating. Redrawnfrom [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 reci-procity theorem permits the reversal of the propagation direction of all lightwaves. The deflection angles (coupling angles) of the TE and TM modes de-pend on the light wavelength and the grating period (Fig. 4.12b). Binding ofan analyte on the grating alters the coupling angle, which can be detectedusing position-sensitive detectors. The sensitivity of the grating coupler isrelated to the lateral dimensions of the grating region interacting with thesample.

Prism couplers [180] will not be treated here.

Difference InterferometerIn a planar waveguide, the TE and TM mode are coherently excited bya laser. Both propagate along a common path down the same waveguideand interact with the sample within a certain length of the waveguide. Thepolarization-dependent interaction induces a phase difference between thetwo 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 orthog-onally polarized modes in a silicon nitride waveguide [183,184].


detectorsensor pad

phase modulator

DBRlaser

LL

sensor branch

Si-nitrideSi-oxide

Si-oxidesensor areaelectric fieldSi substrate

sensitive layer

(a) Mach-Zehnder interferometer

reference branch

input output

(b) integrated GaAs interferometer

Fig. 4.13. Schematic of a conventional Mach-Zehnder interferometer (a), and an in-tegrated Mach Zehnder interferometer (light source and detector on chip) in GaAs-technology (b). The cross section shows the separate sensor (left side, open) andreference (right side, covered) branches. Redrawn from [191]

Two-Beam InterferometersMach-Zehnder IO-devices (Fig. 4.13a) are monomode channel waveguides(TE or TM mode) and allow for a straightforward implementation of aninterferometer structure [158,177–179,185]. A waveguide is split into an openmeasurement path and a protected reference path and recombined after somedistance. The phase difference, introduced by analyte interaction (refractiveindex change) in the sensing path, is detected by interference effects. Detec-tion limits are in the range of some picograms/mm2 [186].

Integrated Waveguide Absorbance Optodes (IWAO)A membrane inserted between two micromachined waveguides acts simulta-neously as the light-guiding medium and sensing element and hence changesits spectral properties while interacting with an analyte. First results withpotassium-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 zincoxide as the optical waveguide on a chip fabricated in silicon oxinitride tech-nology, have been reported in [188,189].


PBS

PBS

PBS

rabbit anti-atrazineserum 1:100

PBS

10 µg/ml avidin10 µg/ml biotinylated protein A

25 µg/mlatrazine-HRP

time [min]

surface mass density [ng/m

m2]

∆Φ[2

π]

0

1

2

3

4

5

4003002001000

-40

-30

-20

-10

0

Fig. 4.14. Sensor response (phase shift, ∆Φ) of a difference interferometer usedfor biosensing. Adsorption of avidin at the surface, affinity binding of biotinylatedprotein A to the avidin layer, binding of the rabbit anti-atrazine serum to theprotein A layer, immunoreaction of the immobilized anti-atrazine antibodies withatrazine (atrazine-horseradish peroxidase, atrazine-HRP). PBS denotes phosphatebuffer solution washing steps. Reprinted from [173] with permission

ApplicationsTypical applications are humidity sensors [158,173,174], gas sensors [184,185](adsorption on the device or absorption in a microporous waveguide), andenvironmental monitoring or biosensing. Examples include the detectionof different organic solvents in the liquid phase (hydrocarbons, alcoholsetc.), the monitoring of sucrose and buffer solutions [158, 173, 175, 176],and biotin/streptavidin-mediated immunosensing (Fig. 4.14) involving an-tibody/antigen binding experiments [158,173–179].

The interaction mechanisms include reversible physisorption (2.10) as wellas biochemical affinity reactions (2.11) [158,175–179].

Gallium-Arsenide-Based DevicesDue to their direct band gap, III-V-semiconductors offer the opportunityof fabricating and integration of lasers, waveguides, phase modulators andwaveguide detectors on the same chip. GaAs/AlGaAs-based Mach-Zehnderdevices with integrated light sources and detectors have been developed asshown 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 theinterferometer [190–192]. The detector is a long-absorbing-length photodiodewith high quantum efficiency [193].

A more recent development for optical gas sensing is the vertical-cavitysurface-emitting laser (VCSEL). The cavity is formed vertically on the wafer


surface. Epitaxially grown Bragg mirrors serve as distributed reflectors aboveand below the laser’s very short active region. GaAs/AlGaAs-based VCSELsemit in the near infrared region. Oxygen sensing at 762 nm (absorption dueto magnetic dipole transitions in the gas molecule without interference fromother gases) was demonstrated in first experiments [195,196].

4.3.2 Microspectrometers

4.3.2.1 Fabry-Perot-Type Structures

Transduction Principle and Sensing CharacteristicsA Fabry-Perot interferometer (FPI) is an optical element consisting of twopartially reflecting, low-loss, parallel mirrors separated by a gap. The opticaltransmission characteristics through the mirrors consist of a series of sharpresonant transmission peaks occurring when the gap is equal to multiples ofa half wavelength of the incident light. These transmission peaks are causedby multiple reflections of the light in the cavity. By using highly reflectivemirrors, small changes in the gap (width, absorptivity) can produce largechanges in the transmission response. Even though two reflective mirrors areused, transmission through the element at the peak wavelengths approachesunity. The transmission is a function of both, the gap spacing and the radi-ation wavelength.

The devices can be used as wavelength selector or monochromator byadjusting the gap width to achieve the desired wavelength. Tunable deviceswith a gap width variable by electrostatic actuation using electrodes on mov-able micromachined parts have been reported (Fig. 4.15a) [197–200]. Suchdevices operate preferably in the near infrared region at wavelengths largerthan 1 µm, where silicon substrates become transparent [197].

movablemesa

wafer

corrugatedsupport

controlelectrodes

opticalcoating (Al)

gap

anti-reflectioncoating

fusion-bonded

layer

(a) tunable Fabry-Perot

radiation

p -epilayer

p+ substrate

n-wellp+ implanted

layer (SP)

pnp-photo-transistor

(b) CMOS Fabry-Perot etalon

silver

aluminumFabry-Perot

cavity

PECVD oxide

Fig. 4.15. (a) Schematic of a tunable Fabry-Perot interferometer. The wavelength-defining gap width can be changed by applying DC to the control electrodes. Alu-minum is used as optical coating material. Redrawn from [197]. (b) Cross sectionof an integrated Fabry-Perot etalon. The gap width is determined by the thick-ness of the PECVD oxide, silver and aluminum are used as optical coatings. Apnp-phototransistor (p+ implanted layer/n-well/p-epilayer) is located directly un-derneath the etalon. Redrawn from [201]


A single-chip CMOS optical microspectrometer based on FPI and oper-ating in the UV/VIS-region is reported in [201, 202]. It contains an arrayof 16 addressable Fabry–Perot etalons (500 × 500 µm2) each with a differentresonant cavity length realized as a PECVD silicon oxide layer sandwiched inbetween an aluminum and a silver layer and placed on top an array of verticalpnp phototransistors (Fig. 4.15b) [203]. It additionally includes circuits forreadout, 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 forwafer-stacking techniques to achieve the FPI cavities [198,199]. The completefabrication sequence of a monolithic CMOS VIS spectrometer is given in [201].

ApplicationsTypical applications include gas sensors [198–200, 203–207]. The character-istic absorption wavelengths of carbon monoxide are 4.7 µm, that of carbondioxide 4.2 µm, and that of methane or hydrocarbons 3.3 µm (IR region,molecular vibrations). Polymeric coatings in the FPI cavity have been usedto detect iodine [204, 205]. Carbon dioxide sensors based on tunable FPIs(dual wavelength measurements) are commercially available from [207]. Theradiation source in most cases is a light bulb or light-emitting diode (LED).

4.3.2.2 Grating-Type Structures

Transduction Principle and Sensing CharacteristicsMicromachined diffraction gratings have been used in combination withimaging devices to set up microspectrometer arrangements [208–210], seeFig. 4.16a for a schematic. Two basic types of gratings can be easily mi-cromachined: (1) amplitude gratings by blocking out light with an array ofopaque and transparent sections, or (2) phase gratings where the light phaseis modulated by variations in the grating shape [208, 209]. In the exampledisplayed in Fig. 4.16a [210], a phase grating with a fine grating pitch creat-ing high dispersion angles was etched into a quartz wafer. The transmissiongrating was then mounted directly over a charge-coupled-device (CCD) im-ager [208], or a CMOS imaging chip consisting of an array of custom-designedphotodiodes [210]. A glass spacer is placed in the optical path between thegrating and the detector. In another approach, two wafers have been micro-machined so that an optical path of about 4 mm length was obtained. Lightdispersed by a 32-slit diffraction grating travels along the optical path and isdirected to an array of photodiodes (Fig. 4.16b) [211]. The interior of one of


(b) integrated spectrometer

grating photodetector array

mirror

red

blue

n - epilayer

diffractiongrating

incidentlight

increasing wavelength

CMOSimager chip

glassspacer

(a) CMOS microspectrometer

p-waferbevel

Fig. 4.16. (a) CMOS microspectrometer consisting of a glass and a CMOS im-ager chip: Different wavelengths are diffracted at different angles from the surfacenormal. Redrawn from [210]. (b) Schematic of an integrated spectrometer usingtwo 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 arrayare integrated in the second wafer, which remains uncoated. The wafers arebonded together by silicon/silicon fusion bonding. The performance of suchmicromachined spectrometers is comparable to that of low-end bench-topspectrometers.

Diffractive optical elements, which produce analyte infrared spectra ofcompounds such as hydrogen fluoride (HF) for chemical sensor systems basedon 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]

ApplicationsTypical applications include biochemical and chemical analysis. Emission gasspectra have been recorded for carbon dioxide and helium. Fluorescence ofa dye (fluorescein) was induced by a laser and detected with the microspec-trometer [208].

4.3.3 Bioluminescent Bioreporter Integrated Circuits (BBIC)

Transduction Principle and Sensing CharacteristicsThis technique employs bioluminescent bacteria placed on an application-specific optical integrated circuit (standard CMOS) [213–215]. The bacte-ria have been engineered to luminesce (4.5) when a target compound such


signalprocessing

circuit

photo-detectors

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 re-ports the magnitude of the optical signal. The microluminometer uses thep-diffusion (source and drain diffusions of p-channel MOSFETs) as the pho-todiode. The shallow p-diffusion has a strong response to the 490-nm bi-oluminescent signal. The entire sensor including all signal processing andcommunication functions can be realized on a single chip. The integratedcircuitry contains the subunits needed to detect the optical signal, to per-form analog or digital signal processing, to communicate the results, and toperform auxiliary functions (temperature, position measurement) (Fig. 4.17).

Many types of bioluminescent transcriptional gene fusions have been usedto develop light-emitting bioreporter bacterial strains to sense the presence,bioavailability, and biodegradation of different kinds of pollutants. The cellshere were entrapped on the chip by encapsulation in natural or syntheticpolymers 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]

ApplicationsTypical applications include chemical analysis in gas or liquid phase. De-pending on the integration time of the device, trace amounts of tolueneand 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).


4.3.4 Surface Plasmon Resonance (SPR) Devices

Transduction Principle and Sensing CharacteristicsThe quantum optical-electronic basis of SPR is due to the fact that the energycarried by photons of light can be “coupled” or transferred to electrons in ametal [159]. The wavelength of light, at which coupling (i.e., energy trans-fer) 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 reflectedby the metal surface. All the light is reflected except at the resonant wave-length, where almost all the light is absorbed (Fig. 4.18). The coupling oflight into a metal surface results in the creation of a plasmon, a group ofexcited electrons, which behave like a single electrical entity. The plasmon,in turn, generates an electrical field, which extends about 100 nm above andbelow the metal surface [159]. The characteristic of this phenomenon, whichmakes SPR an analytical tool, is that any change in the chemical compositionof the environment within the range of the plasmon field causes a change inthe wavelength of light that resonates with the plasmon. That is, a chemicalchange results in a shift in the wavelength of light, which is absorbed ratherthan reflected, and the magnitude of the shift is quantitatively related tothe magnitude of the chemical change [161,217]. The phenomenon of SPR isnon-specific. Different chemical changes cannot be distinguished.

SPR monochromatic, angle variationlight

source opticaldetection

unit

polarizedlight

reflectedlight

sensor chip

metal film

flow channel

Y

1

2

prism

φ

angle

inte

nsity

21

1 : receptor only

2 : receptor and target analyte

Fig. 4.18. Surface plasmon resonance (SPR) principle and a typical sensor responsediagram: Intensity versus angle. Monochromatic light is used to excite the surfaceplasmon, 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, onemicromachined glass layer and an alumina substrate, is shown in Fig. 4.19[218, 219]. Bonding was achieved using a low-temperature-curing polyimideand a solder sealing. The micromachined silicon layers contain a torsional,


Fig. 4.19. Cross-sectional diagram of the SPR microsystem, labeling individualcomponents and showing the light path. Color image available at http://www.ece.ucdavis.edu/misl/web/pages/projects/plasmon.html. Reprinted from [218] withpermission

all-silicon micromirror, V-grooves for optical fiber and lens, and a position-sensing photodiode (PSD). The silicon micromirror was electrostatically de-flected through 9–10 degrees to direct the light beam emitted from the endof a fiber through a range of angles incident onto the metal film, settingup a surface plasmon. The position and intensity of the reflected beam wasrecorded with the position-sensing photodiode. The microsystem measuresapproximately 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 align-ment 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].

ApplicationsAny pair of molecules that exhibit specific binding can be adapted to SPRmeasurements. These may be an antigen and antibody, a DNA probe andcomplementary DNA strand, an enzyme and its substrate, or a chelatingagent and a metal ion. Typical applications areas are environmental moni-toring (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 serumalbumin (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 chemicalsensors [4, 6, 8–11,23,26,34]. They rely on electrochemical or charge-transferreactions: A+ + e− ↔ A (see 2.12–2.15 and related text in Chap. 2). Elec-trochemistry includes charge transfer from an electrode to a solid or liquidsample phase or vice versa. Chemical changes take place at the electrodes orin the probed sample volume, and the resulting charge or current is measured.Both, electrode reactions and charge transport in the sample are subject tochanges by chemical processes, and, hence, are at the base of electrochemicalsensing [4].

A key requirement for electrochemical sensors is a closed electrical cir-cuit, though there may be no current flow (see potentiometry below). Anelectrochemical cell is always composed of, at least, two electrodes with twoelectrical connections: One through the probed sample, the other via trans-ducer and measuring equipment. The charge transport in the sample canbe ionic, electronic or mixed, while that in the transducer branch is alwayselectronic.

Electrochemical sensors are usually classified according to their electro-analytical principles [4, 8, 11].

VoltammetryVoltammetric sensors are based on the measurement of the current-voltage re-lationship in an electrochemical cell comprising electrodes in a sample phase.A potential is applied to the electrodes, and a current is measured, whichis proportional to the concentration of the electro-active species of interest.Amperometry is a special case of voltammetry, where the potential is keptconstant as a function of time.

PotentiometryPotentiometric sensors are based on the measurement of the potential atan electrode, which is, in most cases, immersed in a solution. The potentialis measured at equilibrium state, i.e., no current is allowed to flow duringthe measurement. According to the Nernst equation (2.15), the potential isproportional to the logarithm of the concentration of the electro-active species(Work function sensors will be discussed in the context of field effect devicesin Sect. 4.4.2.2.4).

ConductometryConductometric sensors are based on the measurement of a conductance be-tween two electrodes in a sample phase. The conductance is usually measured


by applying an AC potential with a small amplitude to the electrodes in orderto prevent polarization. The presence of charge carriers determines the sam-ple conductance. In contrast to conductometry, AC-impedance measurementsare not really used for analytical applications. Impedance measurements areof special interest for membrane, electrode and electrolyte characterization.The goal of impedance measurements is to find an equivalent electronic cir-cuit model and to correlate that model with electrochemical phenomena.

Another categorization method relies on discerning the electronic com-ponents [9,34]. There are chemoresistors, chemodiodes, chemocapacitors andchemotransistors. Within this book the electroanalytical principles will beused as the superordinated classification scheme, and the component nota-tion will be used within this scheme.

Again, the focus will be on semiconductor-based systems, and a wealth ofliterature on, e.g., other designs of ion-sensitive electrodes [221], or silicon-carbide-based devices operating at extremely high temperature [222–225] willbe omitted. Review articles are recommended to explore further details (see,e.g., [226–231]).

4.4.1 Voltammetric Sensors

Transduction Principle and Sensing CharacteristicsVoltammetry, in general, is the measurement of the current that flows at anelectrode as a function of the potential applied to the electrode. The result of avoltammetric experiment is a current/potential curve. Amperometry is morefrequently applied in chemical sensors and provides a linear current/analyteconcentration relationship at a constant potential, which is predefined withregard to the target analyte.

Two different electrode configurations are normally used. The two-electrodeconfiguration [4, 8, 11] consists of a reference electrode (RE) and a workingelectrode (WE) (Fig. 4.20a) [11]. The disadvantage of this method is, that theRE carries current and may become polarized if it is less than one hundredtimes the size of the WE. Material consumption due to the current in the REis 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 injectionin the analyte (Fig. 4.20b) [11]. The reference electrode is now a true REwith a well-defined potential since no current is flowing through the RE. Thepotentiostat controls the current at the auxiliary electrode as a function of theapplied 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 thesolution. The AE is connected to the output. The opamp injects a current intothe solution through the AE. Due to the feedback mechanism, the currentis controlled in such a way that the potential at the negative input equals


2 - electrode system 3 - electrode system

(a) (b)+_

AEWE RE

UU

WE RE

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 andRE hence equals the applied potential. No current is flowing through the REsince the opamp has a very high input impedance. The sensor signal currentis measured at the working electrode.

The measured current at any given potential difference depends on thematerial properties, the composition and geometry of the electrodes, the con-centration of the electro-active species (presumably the target analyte) andthe mass transport mechanisms in the analyte phase [4, 8, 11, 34]. Amongthose are migration, the movement of charged particles in an electric field,convection, the movement of material by forced means like stirring or as aconsequence of density or temperature gradients, and diffusion, the move-ment of material from high-concentration regions to low-concentration re-gions. The electrochemical reactions at the electrodes are normally fast incomparison to the transport and supply mechanisms. Since convection in theelectrode vicinity is avoided, and migration is suppressed by, e.g., a largeexcess of electro-inactive salts (i.e., electro-inactive at the respective appliedpotential), diffusion is normally regarded to be the dominant mechanism.There are two components to the measured current, a capacitive componentresulting from redistribution of charged and polar particles in the electrodevicinity, and a component resulting from the electron exchange between theelectrode and the redox species (analyte) termed faradaic current [4,8,11,34].The faradaic component is the important measurand and is, for the case ofdiffusion-limited conditions, directly and linearly proportional to the targetanalyte concentration. The limiting current (all analyte ions are immediately


charged or discharged upon arrival at the electrode) is then given by theCottrell equation [4, 8, 11,26,34]:

I∞ = ne · F · A · DdiffcA

Ldiff. (4.7)

Here, ne denotes the number of electrons, F is the Faraday constant, A theeffective electrode area, Ddiff the diffusion coefficient, cA the target analyteconcentration and Ldiff the diffusion length (for more mechanistic details see[4,8,11,26,34]). Correction terms to this equation for small electrodes have tobe introduced to take into account the respective electrode geometry [4,8,26].

Cyclic VoltammetryA stationary WE is used, and a cyclic ramp potential versus some RE isapplied. The potential versus time is triangular: It increases at a rate linearwith time, then reverses and decreases at the same rate. The current flowsas a consequence of the applied potential between the WE and an AE. Thistechnique is used to study the electrode/sample interface [4, 8, 11,232–235].

Stripping VoltammetryThis method is used to detect heavy-metal ions at trace level by means of amercury electrode [236–239]. The method involves an initial preconcentrationphase, in which the array is held at a cathodic potential such that the metalions from solution are reduced and amalgamated into the mercury. Then,the electrode potential is reversed to anodic, and the metals in the mercuryare re-oxidized and stripped from the mercury into the solution. The chargerequired to strip a given metal completely from the mercury is proportionalto 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 per-meation [240,241]

• Deposition/patterning of metal electrodes (lift-off, thermal evaporation,sputtering) [8, 11,240–256]

• Deposition of electro-active polymers, membrane materials, hydrogels (spin-casting, spraying, screen printing) [240–257]

Sensor processing sequences are given in [11, 240, 241, 251, 253]. The fabri-cation of electrochemical sensors that are integrated with CMOS circuitrycomponents is described in [11, 251]. Microsensor arrays with up to 1024individually addressable elements have been reported on [254,255].

A picture and a schematic of a CMOS-based 3-electrode amperometricsensor are shown in Fig. 4.21 [251]. The monolithic device includes the elec-trochemical sensor, a temperature sensor, and interface circuitry. The cir-cuitry contains an operational amplifier as potentiostat, a switched-capacitor


(a)

(b)

Fig. 4.21. Micrograph (a) and layout (b) of a CMOS-based 3-electrode ampero-metric sensor. Reprinted from [251] with permission

current-to-voltage converter and a clock generator. Interface circuitry andthe temperature sensor are realized in 3 -µm CMOS-technology. The circuitryneeds a supply voltage of ±2.5 V, can apply voltages from +1 V to −1 V tothe sensor and handles currents from 30 nA full scale to 1 µA full scale. Theoutput voltage of the temperature sensor is proportional to the absolute tem-perature and has a sensitivity of 125 µV/K. The total sensor dimensions are0.75 mm by 5 mm.

ApplicationsTypical applications include chemical analysis in the gas or liquid phase. Ifthe 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 sensi-tive electrode coatings. Typical target analytes in the gas phase are nitrogenoxides, hydrogen sulfide, using Nafion polymer electrolyte [241–243], as wellas oxygen [248, 253, 255], and carbon dioxide using liquid electrolytes [240].Target analytes in the liquid phase comprise dissolved oxygen [235,251], glu-cose [245, 246, 251], hydrogen peroxide [240, 247], and chlorine in drinkingwater [244,249] Fig. 4.22.

One of the best-known voltammetric cells is the Clark cell, which is basedon a two-step-reduction of oxygen via hydrogen peroxide to hydroxyl ions inaqueous solution. The Clark cell is used to measure dissolved oxygen in bloodand tissue [258]. The reference electrodes in the liquid phase are in most casessilver/silver-chloride elements.

The interaction mechanism in all cases is an electrochemical redox reac-tion (2.12–2.15).


(b)PDMS

CE

RE

WE

curr

ent [

pA]

time [min]

hypochlorous acid

(a)

0 5 2010 15 25

0

25

-25

100

50

75

Fig. 4.22. (a) Micrograph of an amperometric 3-electrode free-chlorine sensor witha central membrane-covered working electrode (WE) surrounded by the counterelectrode (CE, ring) and the reference electrode (RE, ring segment). A polysiloxane(PDMS) encapsulation ring guides the liquid sample phase. (b) Sensor responseupon exposure to chlorine from hypochlorous acid near the limiting threshold (ppbrange). Reprinted from [249] with permission

4.4.2 Potentiometric Sensors

Potentiometry is the direct application of the Nernst equation (2.15) throughmeasurement of the potential between nonpolarized electrodes (WE and RE)under conditions of zero current. The measurement is carried out at thermo-dynamic equilibrium.

In the following, it will be distinguished between two different types ofpotentiometric devices:

• Electrochemical cell with metal electrodes.• Field-effect semiconductor devices.

4.4.2.1 Electrochemical Cell

Transduction Principle and Sensing CharacteristicsThe electrochemical cell used for potentiometric microsensors consists of twometal electrodes, a WE covered with a ion-selective membrane or gel, whichpreferably hosts a specific target ion, and an RE, which, in most cases, is asilver electrode covered by a thin silver chloride film [259, 260]. The WE istermed an ion-selective electrode (ISE). Both electrodes are on the same chipand 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 mod-ern micromachined nonsymmetrical ISEs, however, is completely differentfrom that of conventional symmetrical ISEs such as a pH-glass-electrode or a


Ag/AgClreference

silicon substrate

ion-selectivemembrane

(a) electrochemical cell

metal electrodes (Pt)

perm-selectivemembrane

electrode 1 electrode 2

solution 1a1 (A+)

solution 2a2 (A+)

(b) concentration cell

Fig. 4.23. (a) Schematic of a potentiometric nonsymmetrical electrochemical cellwith metal electrodes. Ion-selective electrode (ISE) and RE are on the same chipexposed to the analyte. The RE is protected by a membrane. (b) Schematic of aclassical symmetrical potentiometric concentration cell. The potential is measuredbetween two half-cells containing different activities/concentrations of the sameanalyte (A+)

lanthanium-fluoride membrane electrode. The traditional symmetric arrange-ments exhibit a liquid phase reservoir separated by a perm-selective mem-brane from the analyte phase and, thus, essentially constitute electrochemicalconcentration 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 den-sity of the charged species of interest is sufficiently high (i.e., >10−3 A/cm2),such interfaces are well defined, and the devices are stable.

The exchange current is due to significant and continuous movement ofcharge carriers in both directions through the interphase region at an elec-trode at equilibrium (dynamic equilibrium). The magnitude of these mutuallycompensating currents (no net current) flowing at any zero-current potentialis called exchange current.

The Nernst equation (2.15) can be applied to calculate the electrochemicalpotential evoked by a certain analyte concentration.

The design of metal-electrode potentiometric sensors is very similar to thevoltammetric sensors (two-electrode configuration) described in Sect. 4.4.1.In comparison to voltammetric techniques it should be noted that the con-centration dependence of the measured potential is logarithmic (2.15). Usingamperometric techniques, the target ions can be selected by careful tuningof the appropriate redox potential. In potentiometry, all ions in the sampleexhibiting comparable exchange current density contribute to the measuredoverall potential. In order to achieve selectivity to one specific ion, this tar-get ion must provide a significantly higher exchange current density than the


other interfering ions. This condition can be achieved by incorporating selec-tive binding sites or ionophores in a membrane or gel material. The sensitivecoating hence has to provide selectivity.

FabricationThe fabrication is very similar to that of voltammetric devices and is,therefore, not further specified here. Sensor processing sequences are givenin [259, 262, 263]. The fabrication of potentiometric sensors integrated withCMOS circuitry components is described in [263,264].

ApplicationsPrototype applications include chemical analysis in the gas or liquid phase.Typical target analytes in the gas phase are nitrogen oxide, sulfur dioxideand carbon dioxide using sintered ceramic electrolytes (sodium, barium andsilver sulfate or Nasicon) [262,265]. Target analytes in the liquid phase com-prise all kinds of ionic species like hydrogen (pH: “potentia hydrogenii”, neg-ative decadic logarithm of the hydrogen ion concentration), potassium, am-monium, calcium, chloride, cyanide, or nitrate using ionophores in polymericmembranes [259,260,263,264,266] or chalcogenide glasses [267].

The interaction mechanism in all cases is an electrochemical charge-transfer reaction (2.12–2.15).

4.4.2.2 Field-Effect-Based Devices

Transduction Principle and Sensing CharacteristicsThe field-effect-based microfabricated potentiometric sensors, like metal-oxide semiconductor (MOS) devices in electronics, rely on variations in thecharge distribution within the semiconductor surface space-charge region.

The three field-effect device structures commonly used include the MOScapacitor (MOSCAP), the MOS diode (MOS-diode), and the MOS field-effect transistor (MOSFET) [4, 8, 11, 34]. These structures are displayed inFig. 4.24. It is important to point out, “that one of the most critical thrustsof the electronics industry’s efforts to develop commercial field-effect deviceshas been predicated by the need to isolate these devices from any variation intheir chemical environment” [34], since the field-effect device characteristicsare very sensitive to such variation. Those efforts of the electronics industryare in diametrical opposition to developing, e.g., FET-based chemical sensors.

By replacing, the MOSFET metal gate with an ionic solution and a ref-erence electrode immersed into this solution, ion-sensitive device structureshave been developed, the basic structure of which is analogous to the re-spective MOS devices: Ion-sensitive capacitor (ISCAP), ion-controlled diode(ICD) and ion-sensitive field effect transistor (ISFET) (Figs. 4.24, 4.25). Thegate region is exposed to any ion present in the solution [4, 8, 11,34].

The device applications will be discussed in the individual device-relatedsections, whereas the fabrication of the family of field-effect devices will bedetailed in summary at the end in Sect. 4.4.2.2.6.


(a) MOSCAP (b) MOS-Diode (c) MOSFET

gas phasegas phasegas phase

Si-oxide

metal (Pd)

p-silicon

Si-oxide

metal (Pd)

p-silicon

n ngate oxide

drainsource

metal gate (Pd)

p-silicon

U

Ud

UgU

voltage

capacitance

voltage

current

gate voltage

drain currentdrain voltageconstant

hydrogenno hydrogen

∆Ugas

hydrogenno hydrogen

hydrogenno hydrogen

∆Ugas∆Ugas

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-drainvoltage. 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 CharacteristicsA MOSFET in electronics is a transistor, the source drain current (conduc-tance) of which is modulated through an electric field perpendicular to thedevice surface. This electric field is generated by an isolated gate electrodeand influences the charge carrier density in the conductance path betweensource and drain (semiconductor field effect).

The MOSFET as used for chemical sensing has, e.g., a p-type siliconsubstrate (bulk) with two n-type diffusion regions (source and drain). Thestructure is covered with a silicon-dioxide insulating layer, on top of which ametal 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 thegate electrode, holes, which are the majority carriers in the p-substrate, aredepleted near the semiconductor surface. Upon applying a voltage betweendrain and source (Ud), the electrons from the n-doped source can pass throughthis depleted surface region to the drain so that a conducting n-channel be-tween source and drain is generated in the p-substrate near the silicon/silicondioxide interface. The conductivity of this n-channel, i.e., the magnitude of


the source-drain current (Id) can be modulated by adjusting the strengthof the electrical field perpendicular to the substrate surface between gateelectrode and the silicon substrate [22,268,269].

ApplicationsPalladium (Pd)-gate FET structures were demonstrated to function as hydro-gen sensors by Lundstrom [22] and others [270,273,274]. Hydrogen moleculesreadily absorb on the gate metal (platinum, iridium, palladium) and dissoci-ate into hydrogen atoms. These H-atoms can diffuse rapidly through the Pdand absorb at the metal/silicon oxide interface partly on the metal, partly onthe oxide side of the interface [268,269]. Due to the absorbed species and theresulting polarization phenomena at the interface, the drain current (Id) isaltered, and the threshold voltage (Ud) is shifted. The voltage shift (∆Ud) isproportional to the concentration or coverage of hydrogen at the oxide/metalinterface. The presence of oxygen promotes the formation of water at the gasphase/metal interface due to the catalytic reaction of atomic hydrogen withatomic oxygen.

With thin catalytic metal gates (containing holes and cracks) ammo-nia [269,276], amines, and any kind of molecule that gives rise to polarizationin a thin metal film (hydrogen sulfide, ethene, etc.) or causes charges/dipoleson the insulator surface, can be detected [278,279]. Detailed models for thoseprocesses, however, do not yet exist. Sensitivity and selectivity patterns ofgas-sensitive FET devices depend on the type and thickness of the catalyticmetal used, the chemical reactions at the metal surface, and the device op-eration temperature. For extremely high temperatures (600–800C), siliconcarbide 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 thinalumina layer between the Pd gate and the silicon oxide [277]. The tempera-ture should be kept constant. Alternative gate materials include polyanilinefor the detection of water and ammonia [278, 279], and high-temperature-superconducting cuprate to monitor ammonia and nitrogen oxides [280].

ISFET Transduction Principle and Sensing CharacteristicsFor the case of the ISFET, the gate metal electrode of the MOSFET isreplaced by an electrolyte solution, which is in contact with the referenceelectrode, i.e., the silicon gate oxide is directly exposed to aqueous electrolytesolution Fig. 4.25 [21]. An external reference electrode is required for a stableoperation of an ISFET [4,8,11,34,281–284]. Unfortunately, including such areference 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 fieldperpendicular to the direction of the current. It additionally depends on thepotential difference across the gate oxide. Therefore, the source-drain current,Id, is influenced by the interface potential at the oxide/aqueous solutionboundary. Though the electric resistance of the channel provides a measure


(a) MOSFET (b) ISFET

n ngate oxide

drainsource

referenceelectrode

p-silicon

liquid phase

Ud

gas phase

n ngate oxide

drainsource

metal gate (Pd)

p-silicon

Ud

Ug U

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 metalgate of the MOSFET with an ionic solution and a reference electrode immersedinto this solution, the ISFET has been developed

for the gate oxide potential, the direct measurement of this resistance givesno indication of the absolute value of this potential.

However, at a defined source-drain potential (Ud), changes in the gatepotential can be compensated by a modulation of Ug. This adjustment canbe carried out in a way that the changes in Ug applied to the referenceelectrode exactly compensate for the changes in the gate oxide potential.This is automatically performed by ISFET amplifiers with feedback, whichallow for obtaining a constant source-drain current. In this particular case,the gate-source potential is determined by the surface potential at the in-sulator/electrolyte interface. Mechanistic studies of the processes occurringat the solution/gate oxide interface (site binding model [290]) and the ox-ide/semiconductor interface can be found in the literature [4, 8, 11, 34, 283,284, 290–293]. The insulator/solution interface is assumed to represent inmost cases a polarizable interface, i.e., there will be charge accumulationacross the structure but no net charge passing through.Interfaces with netcharge passing through are termed “faradaic”, see, e.g., [4, 8, 11,34,229].

ApplicationsThe gate oxide surface contains reactive Si-OH groups, which, besides pro-viding pH sensitivity, can be used for covalent attachment of a variety oforganic molecules and polymers.

pH-FET (pH: “potentia hydrogenii”, negative decadic logarithm of theH+-ion concentration).The basic ISFET is an exposed-gate-oxide FET andfunctions as a pH sensor [21]. The surface of the gate oxide contains OH-functionalities, which are in electrochemical equilibrium with ions in the sam-ple solutions (H+ and OH−). The hydroxyl groups at the gate oxide surfacecan be protonated and deprotonated and, thus, when the gate oxide contactsan aqueous solution, a change of pH will change the silicon oxide surface po-tential. A site-binding model describes the signal transduction as a function


time [sec]

sens

or s

igna

l [m

V]

0 20 8040 60 100 120 140

-1490

-1470

-1510

-1450

Fig. 4.26. Dynamic response of a pH-ISFET in a flow-through configuration uponrepeated pH-changes by one unit (4 to 5). The signals exhibit a stable baseline anda good reproducibility. According to the Nernst equation (2.15), a pH-change ofone 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 di-pole layer and consequently to a change in the semiconductor space chargeregion [4, 34, 276, 293]. Typical pH-sensitivities measured with silicon-oxideISFETs are 37-40 mV per pH unit [295]. Sensor signals achieved with asilicon nitride membrane are given in Fig. 4.26 [299].

Inorganic gate materials for pH sensors like silicon nitride (CMOS processmaterial) [296–302], oxynitride [303], alumina [304], tantalum oxide [304,305],and iridium oxide [306] have better properties than silicon oxide with regardto pH response, hysteresis and drift (see Fig. 4.26). In practice, these layersare deposited on top of the first layer of silicon oxide by means of chemicalvapor deposition (CVD).

CHEMFETThe CHEMFET or chemically sensitive FET [307,308], is a modification of anISFET with the original inorganic gate material covered by organic ion-selective membranes like polyurethane, silicone rubber, polystyrene,polyamide and polyacrylates containing ionophores (Fig. 4.27). A criticalpoint of the CHEMFET is the attachment of the sensitive membrane, whichcan be improved by mechanical [309] or chemical [308, 310, 311] anchoringto 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 forthe detection of glucose [326, 328–330], and penicillinase for the detection


(b)

n ngate oxide

drainsource

referenceelectrode

p-silicon

liquid phase

Ud

Ug

(a) chemFET

hydrogelmembrane

resin

b aa c cba

a: 10-5 M lead nitrate in acetate bufferb: 10-4 M lead nitrate in acetate bufferc: 10-2 M acetate buffer, pH: 5.5

time [min]

sens

or r

espo

nse

[mV

]

0

160

320

0 3 6 9

Fig. 4.27. (a) Schematic representation of a CHEMFET, a modification of anISFET with the original inorganic gate material covered by an organic ion-selectivemembrane and a hydrogel. (b) Sensor response of a CHEMFET with a lead-selectivemembrane 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 purebuffer solution (c). Redrawn, adapted from [322]

of penicillin [326]. Other enzymes can be used to detect pesticides andorganophosphorous compounds [331–334] or aldehydes [335]. Antibodies canbe immobilized for immunoreactions (IMFET), and biological entities likewhole cells [336–338] or insect antennae [339] have been used (BIOFET).

Differential CHEMFET ConfigurationCHEMFETS can be applied in a differential measurement configuration [340–347] (Fig. 4.28a). This device configuration offers the advantage that a varietyof experimental parameters and external disturbances, which affect both FETstructures (e.g., light and temperature), cancel out. The inorganic gate ofone of the CHEMFETS has to be rendered insensitive to pH or other targetanalytes by chemical surface treatment. This “insensitive” FET (sometimesdenoted reference FET, REFET) should ideally show no response to thetarget 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 thegate [349–353], or by realizing a reservoir of buffered solution with constantpH separated by a membrane from the sample phase Fig. (4.28) [348]. Theapplications 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 CharacteristicsThe 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


Ag/AgClreferenceelectrodes

p-silicon

gas phase

Ud

(a) differential FET configuration

n ngate oxide

drainsourcen ngate oxide

drain sourceUg Ug

gas-permeablemembrane

nonbufferedhydrogel layer

bufferedhydrogel layer

time [min]

EM

F, v

olta

ge [m

V]

10-3 M

10-4 M

10-2 M3•10-2 M

1

2

3

1: CO2-FET2: insensitive FET3: differential signala b

(b)

3020100

400

500

300

Fig. 4.28. (a) Schematic of a differential FET configuration for sensing carbondioxide. The carbon-dioxide FET (formation of “carbonic acid”, pH-FET) exhibitsa nonbuffered hydrogel layer, whereas the insensitive FET has a buffered one. ThepH of the latter will not change upon carbon dioxide absorption as shown in the sen-sor responses (b). The differential configuration thus eliminates effects common toboth devices (temperature fluctuations, flow disturbances etc.). Redrawn/adaptedfrom [348]

at the interface, the height of the energy barrier of the diode is altered. Thisleads to a change in either the forward voltage or the reverse current of thediode. The diode characteristics are hence shifted along the voltage axis (seeFig. 4.24), as it is the case with the MOSFET [22, 268, 269, 354–357]. Theresponse time and sensitivity of the sensor can be improved by operationat elevated temperature. Therefore, sensing structures have been placed onthermally isolated membranes [358,359].

The ion-controlled diode (ICD) was first described by Zemel [360]. Its op-eration is not too different from that of an ISFET and is analogous to that ofthe MOS-diode [8,34,293,360,361]. Polarization resulting from ion adsorptionon the insulator (silicon oxide) causes changes in the effective forward voltageor reverse current, which can be measured. A more sophisticated structurewith a through-the-chip p-n-junction, which enables the contacts on the backside of the chip to be isolated from the aqueous electrochemistry occurringat the front (“gate”) side, is described in [34,360,361]. The applications arethe same as for the FETs.

4.4.2.2.3 MOS Capacitor and Ion-Selective Capacitor (Chemocapacitors)

Transduction Principle and Sensing CharacteristicsThe sensor element is a standard MOS or electrolyte insulator semiconduc-tor (EIS) structure as shown in Fig. 4.24. Again, the target analyte specieschange the polarization at the metal/oxide or electrolyte/oxide interface thusaffecting the flat-band voltage of the capacitor. The capacitance-voltage curveof the capacitor is shifted by a certain amount, which is proportional to thetarget analyte concentration in the gas or liquid phase. One can either record


the capacitance as a measure of the analyte concentration or use some cir-cuitry 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 curvesis shown in Fig. 4.29 [366].

voltage [V]

capa

cita

nce

[nF

]

porous silicon

volta

ge [m

V]

pH

40 mV / pH

-0.8 -0.4 0.80.400

400

200

1000

600

800

4 5 86 7

-100

-300

-200

pH 8pH 7pH 6pH 5pH 4

Fig. 4.29. Set of capacitance/voltage curves for a porous electrolyte-insulator-semiconductor (EIS) structure exposed to solutions of different pH. The insert showsthe calibration curve (pH versus voltage) Adapted from [366] with permission

4.4.2.2.4 Measuring Work Functions: Kelvin Probeand Suspended-Gate Field-Effect Transistor, SGFET

Transduction Principle and Sensing CharacteristicsThe work function is defined as the minimum work required to extract anelectron in vacuum from the Fermi level of a conducting phase through asurface and place it outside the reach of electrostatic forces at the so-calledvacuum level [368,369]. When two different electronic conductors are in con-tact, electrons flow from the material with the lower electron affinity to thatwith the higher one according to the difference in the (electro-)chemical po-tential (2.5, 2.6, 2.13) of the electrons until an equilibrium is reached. Acontact or Galvani potential arises, which represents the bulk-to-bulk innercontact potential of the two materials or the difference of the Fermi levels ofthe two materials.

If surfaces of two different materials are parallel and separated by a verythin insulator or air gap, a potential across the gap is formed, the Voltapotential or outer potential, which represents the difference of the work func-tions of both materials. A palladium plate separated by a thin air gap (few


µm) from a copper plate will become positively charged, and correspondingly,the copper plate will be negatively charged. The work function of a certainmaterial, which includes the chemical potential of the electrons (changes inelectron affinity, eventual band bending due to electron transfer) and the sur-face dipole field (which exists even at absolutely clean surfaces) is changedupon formation of surface adsorbates, e.g., from the gas phase. Therefore,work function measurements can be used to advantage in chemical or gassensing.

The Kelvin probe relies on the displacement of one of the surfaces in aperiodic oscillation. This oscillation induces charges across the surfaces andgenerates a sinusoidal current in the sensing plate, which is proportional tothe work function difference between the sensing plate and the reference plate.This current thus directly depends on the surface chemistry of the plates. Mi-cromachined Kelvin probes with a metal sensing film supported by a dielectricmembrane and a 2.5-µm-thick silicon reference plate that is electrostaticallydeflected by a drive electrode (schematically depicted in Fig. 4.30a) havebeen fabricated and have been used to detect the surface adsorption of oxy-gen [370].

(a) integrated Kelvin probe (b) suspended-gate FET

n ngate oxide

drainsource

metal gate (Pd)

p-silicon

gas phase

Ud

Ugchemically

sensitive layer

drive electrode movable Si-referenceplane

glasssubstrate

electroplatedcontacts

Si-wafer

heater temperaturesensor dielectric

membrane

3 µm metalsensing film

Fig. 4.30. Schematic representation of a micromachined Kelvin probe (a) and asuspended-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 Kelvinprobe [368,371]. The notion of work functions was not used in the context ofFETs yet, though FET transducers are sometimes denoted “work functiondevices”. A metal plate (suspended gate) is separated by an air gap (or inthe 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 andthe gate metal. The plate distance, however, cannot be varied. Therefore, thedrain-source current is used to interrogate the Volta potential. The magni-tude of the drain-source current at a certain applied voltage depends on the


work function difference between the metal gate and the silicon. It thereforedirectly depends on surface adsorption or absorption chemistry in sensitivelayers applied between the suspended gate and the silicon.

ApplicationsSensitive layers applied in the gap include metal oxides to detect ammonia,carbon monoxide or nitrogen oxides [372–376], potassium iodide to detectozone [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 CharacteristicsThis type of sensor is based on the field effect as well [171], and its work-ing 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 toa few µm thickness) with an approximately 100-nm-thick oxynitride layerin contact with an electrolyte solution. A potential is applied between thesilicon plate and, e.g., a silver/silver chloride controlling electrode immersedin the electrolyte solution. The controlling electrode simultaneously serves asreference electrode. The sign and magnitude of the applied potential are ad-justed so as to deplete the semiconductor of majority carriers at the insulatorinterface. Upon illumination of the plate with LEDs, hole-electron pairs arecreated, which can reach the depletion area (Fig. 4.31). Due to the chargeseparation in the depletion area, a photocurrent flows through the device. Inthis way, a sinusoidally modulated light beam causes a sinusoidal photocur-rent, the amplitude of which depends on the width of the depletion charge

insulator

chemicallysensitive layer

controlelectrode

p-silicon

liquid phase

U

LAPS

solution

light

e-h pairs

I

depletionregion

Fig. 4.31. Schematic representation of a light-addressable potentiometric (LAPS)device. For details, see text. Redrawn from [386,387]


region: The larger the depletion region, the larger the photo current. Thephotocurrent amplitude thus depends on the absorbed species in the sensi-tive layer or, e.g., on the pH of the solution in contact with the insulator. Thecurrent-voltage curve is of sigmoidal shape and is shifted along the voltageaxis due to chemical changes (e.g., pH-changes) at the solution/silicon oxideinterface in analogy to other field-effect devices.

ApplicationsThe silicon plate is straightforward to manufacture, and arrays of LEDs anddifferent selective layers can be used on the same chip [371]. Assessment of thelateral resolution [384,385] and benchmarking against ISFETs [386] have beenperformed. LAP methods have been applied to the gas phase as well [380].Applications include monitoring cell activity via pH changes resulting fromcell 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) asdescribed for MOSFETs in [267–276]

• Deposition of additional metal oxides/nitrides by LPCVD (tantalum ox-ide, alumina, silicon nitride) as described for MOSFETs and ISFETs in[296–306]

• Optional membrane formation for temperature stabilization by back etch-ing [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 CMOScircuitry 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 beendescribed in [279], a flow-through ISFET in [395]. Back-side-contact ISFETsare 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 atdifferent frequencies, only the real-valued resistive component, related to thesample (sensing material) resistance, is of interest. Since complex impedancesinclude capacitive and inductive contributions, chemocapacitors that do notrely on the field effect are included here, in the conductometric section. Thesection on conductometric sensors is hence organized in two parts, one on


resistance measurements at room temperature and elevated temperatures(chemoresistors) and the other on chemocapacitors.

4.4.3.1 Chemoresistors

Transduction Principle and Sensing CharacteristicsChemoresistors rely on changes in the electric conductivity of a film or bulkmaterial upon interaction with an analyte. Conductance, G, is defined as thecurrent, I [A], divided by the applied potential, U [V]. The unit of conduc-tance 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 decreaseswith its cross-sectional area, A:

R =1G

=U

I=

1κ· l

A. (4.8)

Conductivity or specific conductance, κ[1/Ωm], is hence defined as the currentdensity [A/m2] divided by the electrical field strength [V/m]. The reciprocalof conductivity is resistivity, ρ[Ωm]. The conductivity can be thought of asthe 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 mea-surement is done either via a Wheatstone bridge arrangement or by recordingthe current at an applied voltage in a DC mode or in a low-amplitude, low-frequency AC mode to avoid electrode polarization. In Fig. 4.32a [399], a con-ductance cell (in this case metal oxides) and the respective equivalent electriccircuits are depicted. The goal of conductometry is to determine the sample

IU

e- e-

freemolecules

electrodes

a. contactsb. surfacec. bulkd. grain

boundaries

substrate

adsorbedmolecules

catalyst

a adb

c

(a) (b)

2-electrodeconductance cell

4-electrodeconductance cell

I

sensitivelayer

electrodessubstrate

U

Fig. 4.32. (a) Schematic representation of a conductance cell (in this case a semi-conductor sensor with tin dioxide as sensitive layer), of the different contributions(contacts, surface, bulk and grains) to the overall conductivity and of the respec-tive equivalent circuits. Redrawn from [399]. (b) Schematic representation of atwo-electrode and four-electrode conductance cell. For details, see text


resistance (c). The lead wire resistances normally can be neglected. The elec-trode impedance (a) consists of two elements, the contact capacitance, andthe contact resistance. By applying an AC potential, an AC current will flowthrough the resistor cell. If the contact capacitance is sufficiently large, nopotential will build up across the corresponding contact resistance.

The contact resistance should be much lower than the sample resistanceand be minimized, so that the bulk contribution dominates the measuredoverall conductance. If surface conductivity mechanisms differing from thosein the bulk occur, this can be modeled by adding an additional surface re-sistance (b) to the equivalent circuit. A grain boundary in the sensing mate-rial constitutes a resistance-capacitance unit (d). The conductivity dependson the concentration of charge carriers and their mobility, either of whichcan be modulated by analyte exposure. In contrast to potentiometry andvoltammetry, conductometric measurements monitor processes in the bulkor at the surface of the sample. Any contribution of electrode processes hasto be avoided.

Therefore, in most cases, a four-electrode configuration is preferred over asimple two-electrode configuration (Fig. 4.32b). The outer pair of electrodesis used for injecting an AC current into the sample, the potential differenceis then measured at the inner pair of electrodes. The interference of electrodeimpedances on the measurement results is thus excluded.

4.4.3.1.1 Low-Temperature Chemoresistors

Several classes of predominantly organic materials are used for applicationwith chemoresistors at room temperature. The chemically sensitive layer isapplied over interdigitated electrodes on an insulating substrate. Electrodespacing is typically 5 to 100 µm, and the total electrode area is a few mm2.The applied voltage ranges between 1 and 5V.

Metal-phthalocyanines constitute organic p-type semiconductors. The ad-sorption of oxidizing agents such as, e.g., nitrogen oxide or ozone hence de-creases the resistance by increasing the number of holes in the conductionband [8]. Metal phthalocyanines at elevated operation temperatures (approx.180C) have been used to monitor nitrogen oxide [400,401], ozone [402], hy-drogen chloride [401] and even ammonia [403,404].

Conducting Polymers such as polypyrroles, polyaniline and polythiopheneexhibit a large conjugated π-electron system, which extends over the wholepolymer backbone. Partial oxidation of the polymer chain then leads to elec-trical conductivity, because the resulting positive charge carriers (denotedpolarons or bipolarons) are mobile along the chains [9]. Counteranions mustbe incorporated into the polymer upon oxidation to balance the charge onthe polymer backbone.

The conducting polymers, however, do not only react with oxidizingagents, but also respond to a wide range of organic vapors [405–408]. Theunderlying principle of this response is still unclear; suggestions include: (I)


vapor molecules could affect the charge transfer between polymer and theelectrode contact, (II) analyte molecules could interact with the mobile chargecarriers on the polymer chains or (III) with the counterions and thus modu-late the mobility of the charge carriers, or they could (IV) alter the rate ofinterchain hopping in the conducting polymer [379,409,410].

Applications include the detection of a variety of polar organic volatileslike ethanol, methanol, components of aromas, [405–408, 411–420] and oth-ers. Conducting polymers show a high cross-sensitivity to water. Sensors arecommercially available from, e.g., Osmetech and Marconi [416].

In Carbon-Black-Loaded Polymers, conducting carbon black is dispersedin non-conducting polymers deposited onto an electrode structure. The con-ductivity is by particle-to-particle charge percolation so that if the polymerabsorbs vapor molecules and swells, the particles are, on average, furtherapart (Fig. 4.33), and the conductivity of the film is reduced [417].

(a) no analyte present (b) after analyte absorption

polymercarbon particles

resistance R1 resistance R2 >R1

Fig. 4.33. Conductivity by particle-to-particle charge percolation in carbon-loadedpolymers: (a) with no analyte present, (b) during organic volatile exposure, whichcauses polymer swelling

Applications include monitoring organic solvents such as hydrocarbons,chlorinated compounds, and alcohols [418–421]. Sensors are commerciallyavailable from, e.g., Cyrano Sciences [422].

Hydrogels responsive to pH-changes have been applied to interdigitatedelectrode arrays. The hydrogel swells or shrinks to a hydration determinedby the pH of the analyte solution. This leads to a corresponding increase ordecrease in the mobility of the ions partitioned by the gel. The sensitivityof ion mobility to small changes in hydration causes large resistance changes[423, 424]. A conductometric variant of a Severinghaus electrode (detectionof carbon dioxide via dissolution in water, formation of carbonic acid andmonitoring of the pH change [4]) with a liquid reservoir has been describedin [425].


Fabrication

• Patterning of metal electrodes (lift-off, thermal evaporation, sputtering)[405–408,418–421]

• Optional membrane formation by back etching for temperature stabiliza-tion [414,426]

• Deposition of carbon-loaded polymers, membrane materials, hydrogels byspin-casting, spraying, screen printing, and photolithography [418–421,423,424]

• Deposition of conducting polymers by electrochemical deposition; sensorselectivity is modified by changing the counterion used in the polymeriza-tion process [405–408]

The fabrication of an impedance device [427], and microbridges [415] in-tegrated with CMOS circuitry components have been described. Completeprocessing 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 relatedsensors, the focus of this work, however, will be on silicon-based microma-chined hotplates. A typical high-temperature chemoresistor includes an in-tegrated heater, a thermometer and a sensing film on a thermally isolatedstage such as a membrane (Fig. 4.34) [428]. Isolated micromachined struc-tures (hotplates) exhibit very short thermal time constants on the order ofmilliseconds.

The sensitive materials used with the hotplate sensors include wide-bandgap semiconducting oxides such as tin dioxide, gallium oxide, indiumoxide, or zinc oxide. In general, gaseous species acting as electron donors (hy-drogen) 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 viewof the device (b). The suspended plate exhibits a polysilicon heater, an aluminumplane for homogenous heat distribution and aluminum electrodes for measuring theresistance of a semiconductor metal oxide. Reprinted from [428,443] with permission


surface states, which can exchange electrons with the semiconductor. An ac-ceptor molecule will extract electrons from the semiconductor and, therefore,decrease its conductivity. The opposite holds true for an electron–donatingsurface state. A space charge layer will thus be formed. By changing the sur-face concentration of donors/acceptors, the conductivity of the space chargeregion is modulated [4, 399,429–431].

In addition to the above-mentioned interaction of surface adsorbates andrelated electronic effects, the diffusion of lattice defects from the bulk of themetal-oxide crystal also occurs (ionic conduction) at elevated temperatures(>600C). The defects can act as donors or acceptors. Oxygen vacancies, e.g.,act as intrinsic donors.

The overall conductivity in polycrystalline samples includes contributionsfrom the individual crystallites, the grain boundaries, insulating componentssuch as pores, and the contacts (Fig. 4.32a). Thus, the conduction mechanismin ceramic polycrystalline samples is difficult to analyze, and a variety ofempirical data has been published [4, 399,429–431].

The most extensively investigated material, tin dioxide, is oxygen-deficientand, therefore, is an n-type semiconductor since oxygen vacancies act as elec-tron donors. In clean air, oxygen, which traps free electrons by its electronaffinity, and water are absorbed on the tin dioxide particle surface forminga potential barrier in the grain boundaries. This potential barrier restrictsthe flow of electrons and thus increases the resistance. When tin dioxide isexposed to reducing gases such as carbon monoxide, the surface adsorbs thegases, and some of the oxygen is removed by reaction of water and oxygen atthe surface. This lowers the potential barrier, thereby reducing the electricresistance. The reaction between gases and surface oxygen depends on thesensor temperature, the gas involved, and the sensor material [4,399,429–432].

Semiconductor metal-oxide sensors usually are not very selective, but re-spond to almost any analyte (carbon monoxide, nitrogen oxide, hydrogen,hydrocarbons). One method to modify the selectivity pattern includes surfacedoping of the metal oxide with catalytic metals such as platinum, palladium,gold, and iridium [399, 430, 433]. Surface doping improves the sensitivity toreducing gases, reduces the response time and operation temperature andchanges the selectivity pattern [399,430,433].

Most modern sensors operate in a regime, in which the overall conductiv-ity is determined by nanocrystalline sensing materials. As a consequence ofthe small grain size (better surface to volume ratio), the relative interactivesurface area is larger, and the density of charge carriers per volume is higher.This leads to more drastic conductivity changes and, hence, larger sensorresponses in comparison to larger grains [399,430].

Since microhotplates have a very low thermal mass, they allow for apply-ing temperature-programmed operation modes [434–437]. By operating thedevice, e.g., in a cyclic thermal mode, reaction kinetics on the sensing surface


close-up

50 ppm CO

1.0 ppm NO2

50 ppm COpure air

1 min

close-up

Rse

nsor

[Ω]

1.0 ppm NO2

mixture

1k

10k

100k

200300400

tem

pera

ture

[°C

]

Fig. 4.35. Sinusoidal modulation of the operation temperature of a tin-dioxide sen-sor between 200 and 400C (bottom) leads to characteristic frequency-dependentresistance features (upper part). Changes of the resistance (Rsensor) of the micro-machined sensor upon exposure to 50 ppm CO, 1 ppm NO2 and a mixture of 50 ppmCO 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 characteristicfor 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, thermalevaporation, 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, sputter-ing, 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 alsoSect. 5.4. Complete processing sequences are detailed in [451, 452]. The fab-rication of a 39-electrode array with a tin-dioxide-gradient coating has beenreported in [453,454]. Extensive reliability studies are reported in [455–457].Investigations and simulations to optimize the power consumption are underway (see, e.g., [458–460]). Devices are commercially available from Figaro,Marconi, Capteur and MICS [461].


ApplicationsTypical 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 thehotplate structures allows for detection of gas mixtures (CO and nitrogenoxide) with a single sensor [436] (Fig. 4.35). In other applications, temper-ature profiles were optimized to specifically detect selected organic volatiles[443,463,464].

Additional sensitive materials on hotplates include, e.g., niobium and ti-tanium oxide to detect oxygen [465], as well as metal films covered withsurface oxides. Reducing gases like hydrogen donate electrons to the metaland increase the conductance, whereas electron acceptor molecules like oxy-gen decrease the metal conductance [451, 466, 467]. An unheated tin-dioxideoxygen sensor (slow response) has been reported in [468].

4.4.3.2 Chemocapacitors

Transduction Principle and Sensing CharacteristicsChemocapacitors (dielectrometers) rely on changes in the dielectric prop-erties of a sensing material upon analyte exposure (chemical modulation ofequivalent-circuit capacitors in Fig. 4.36 by changes in the dielectric constantof the sensitive layer). Interdigitated structures that are similar to those ofroom-temperature chemoresistors are predominantly used [469–472].

substrate

analyte polymer

∆C

E 2E 1

Fig. 4.36. Schematic representation of an interdigitated capacitive sensor coveredwith a sensitive polymer layer. E1 and E2 denote the two sets of interdigitatedelectrodes, ∆C the capacitance change

In some cases, plate-capacitor-type structures with the sensitive layersandwiched between a porous thin metal film (permeable to the analyte)and an electrode patterned on a silicon support are used to increase thesensitivity by trapping the electric field [473,474].


The capacitances usually are measured at an AC frequency of a few kHzup to 500 kHz. The analyte absorption in the sensitive layer on top of the elec-trodes induces a change in the layer dielectric properties and, consequently, acapacitance change that can be measured. For conducting measurements atdefined temperatures, sensor and reference capacitors can be placed on ther-mally isolated membrane structures [475–477]. For more details on capacitivesensors 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 componentsis described in [470, 472, 475, 476, 478, 479, 482, 483]. Temperature-stabilizedmembranes have been reported in [475–477]. Capacitive humidity sensors arecommercially available from, e.g., Sensirion, Vaisala, and Humirel [484].

ApplicationsTypical 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 exhibitspolyimide columns sandwiched between metal electrodes [485]. More recentapplications include the detection of different kinds of organic volatiles inthe gas phase (hydrocarbons, chlorinated hydrocarbons, alcohols etc.) usingpolymeric layers [475, 476, 478, 479, 486, 487] or liquid crystals [488], and thedetection of nitrogen oxide [489], sulphur dioxide [490] and carbon dioxide[491,492] using ceramic materials.

The interaction mechanisms involve reversible physisorption and bulk/gasphase partitioning (see 2.10).

5 CMOS Platform Technologyfor Chemical Sensors

The aim in utilizing microfabrication techniques and, in particular, CMOStechnology for realizing chemical sensors was to devise more intelligent, moreautonomous, more integrated, and more reliable gas sensor systems at lowcosts in a generic approach.

Since the sensor market is strongly fragmented, i.e., there exists a large va-riety of applications with different needs and sensor requirements, a modularapproach or “toolbox strategy” relying on a platform technology was iden-tified as the most promising attempt to achieve major progress. Once theplatform technology has been chosen, the components of the toolbox suchas transducers, sensor modules, and circuit modules can be developed, someof which afterwards can be assembled into a customized system that meetsthe respective applications needs. The application hence dictates the systemarchitecture and the nature of its components: As soon as the target analytesand their concentrations as well as the boundary conditions of detection arespecified, the optimum transducer, the necessary driving and signal condi-tioning circuitry as well as interface and communication units can be selectedfrom the toolbox, and the different modules can then be combined to arriveat a custom-designed sensor system.

A multitude of development activities is necessary to obtain all the mod-ules needed for the toolbox: (a) Design and miniaturization of transducersand directly related electronic components (potentiostats, heaters, amplifiers,etc), (b) development of digital-to-analog and analog-to-digital conversionunits, interface and communication units, (c) development of additional andauxiliary functions, which are pivotal for the system performance (e.g., tem-perature control, temperature sensors, humidity sensors), and (d) develop-ment of dedicated microsystem packaging solutions, which are suitable forchemical or gas analysis. It is important to note that the package has to bethought of already in the initial conception phase of a microsystem, since thedesign and architecture of a microsystem heavily depend on the envisagedpackaging concept.

A fundamental issue concerns the sensor system implementation, i.e., thedecision to either realize monolithic systems combining CMOS circuitry andsensor structure on the same chip (CMOS-MEMS), or to develop hybriddesigns 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 takeninto account in making such a decision.

Materials and Fabrication ProcessesFor monolithic designs, the selection of materials is restricted to CMOS-materials (silicon, polysilicon, silicon oxide, silicon nitride, aluminum, seealso Sect. 3.3) and CMOS-compatible materials. There is also a limitation onavailable fabrication processes owing to a limited number of pre-CMOS andpost-CMOS micromachining options. High-temperature steps (e.g., > 400C)are detrimental to the aluminum metallization (metal oxidation, diffusion)and alter the transistor characteristics (last high-temperature step of theCMOS process is at approx. 380–400C).

For hybrid designs, any material or the optimum sensor material can beused, and a wealth of micromachining techniques for all kinds of materials isavailable through prototyping services.

Time to Market, CostsThe production of monolithic designs relies on established industrial CMOStechnology to fabricate the circuitry and the basic sensor structures. Addi-tional fabrication equipment and related technology developments are neededfor only a few sensor-specific post-processing steps.

The fabrication of hybrid designs entails establishing and qualifying areliable nonstandard production sequence. The specific production equipmentand the necessary clean rooms require large investments.

Sensor Response TimeThe response time of, e.g., a gas sensor is, in most cases, determined bythe volume of the measurement chamber and the flow rate (other relevantprocesses include also, e.g., diffusion or dissociation). Using the monolithicapproach and a suitable packaging technique (e.g., flip-chip packaging), thevolume of the measurement chamber can be kept very small as a consequenceof a small-size, flat and planar sensor or sensor array.

In hybrid designs, the volume is depending on sensor geometries and arrayarrangements.

PerformanceCapacitive or resonant microsensors perform pronouncedly better in mono-lithic designs owing to the fact that the influence of parasitic capacitancesand crosstalk effects can be reduced by on-chip electronics (filters, ampli-fiers etc.). On-chip analog-to-digital conversion is another feature that helpsto generate a stable sensor output that can be easily transferred to off-chiprecording devices.

In hybrid designs, it is sometimes very difficult to read out the ratherminute analog microsensor signals.

5 CMOS Platform Technology for Chemical Sensors 87

Auxiliary Sensors/Smart FeaturesTemperature or flow sensors can be monolithically co-integrated with thechemical sensors on the same chip. Calibration, control and signal processingfunctions as well as self-test features can be realized on chip.

For hybrid designs, additional devices and off-chip components are re-quired.

ConnectivityThe number of electrical connections prominently contributes to the overallsystem costs. The monolithic implementation of a single-chip gas sensor (seeSect. 5.5 on monolithic CMOS gas sensor system later) with three microma-chined sensors requires only seven connections (three for supply voltages, onefor a clock signal, one for reset, and two for the serial interface). Up to 16chips can be connected without adding any additional communication linesby implementation of a digital bus interface. A hybrid approach would re-quire at least 30 pads for the three sensors and a total of 480 connections, if 16sensors of each type would be combined, since there is no interface availableon the sensor side.

PackageTo package monolithic designs, IC-packaging techniques can be modified andadapted such as flip-chip-technology or simple epoxy-based packaging meth-ods.

Hybrid implementations require complex packages to reduce sensor in-terference (see, e.g., high-frequency acoustic-wave SAW-sensors, Sect. 4.1.1),to minimize electric crosstalk, and to optimize the critical connections. Thisfurther complicates the already difficult task of chemical sensor packaging.

In summary the main disadvantages of monolithic CMOS-MEMS solutioninclude the restriction to CMOS-compatible materials and the limited choiceof micromachining processes. However, CMOS-MEMS offers on the otherhand unprecedented advantages over hybrid designs especially with regardto signal quality, device performance, increased functionality and availablestandard packaging solutions. These advantages, in my opinion, clearly out-weigh the drawbacks and limitations.

Micromachined chemical sensors are not yet established on the market. Inthe case of well-established physical sensors such as acceleration and pressuresensors, a trend towards monolithic solutions can be identified for largerproduction volumes and more severe cost restrictions [493–495].

Several examples of monolithically integrated chemical sensor systemswill be presented in the following sections of this chapter. The evolution fromsingle transducers, which are integrated with the necessary driving and signal-conditioning circuitry, to monolithic multisensor arrays and fully developedsystems with on-chip sensor control units and standard interfaces will beshown. Microelectronics and micromechanics (MEMS-structures) have been


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 chemicalsensor microsystem fabrication sequence is shown in Fig. 5.1.

The CMOS-MEMS design is done in house, and the designs are thentransferred to a foundry that performs a complete CMOS run at industrialstandards (usually double-poly, double-metal 0.8-µm CMOS process as pro-vided by, e.g., austriamicrosystems, Unterpremstatten, Austria [496]). Theprocessed wafers come back from the foundry and undergo the post-CMOSmicromachining (formation of membranes, cantilevers etc.). The wafers arethen diced, the chips are tested and subsequently packaged as prototypes.The chemically sensitive coating is either applied on wafer level or to thesingle chips by means of spray-coating (polymers) or drop deposition (metaloxides).

5.1 CMOS Capacitive Microsystems

5.1.1 CMOS Capacitive Transducer

As already mentioned in Sect. 4.4.3.2, capacitive chemical microsensors relyon changes in the dielectric properties of, e.g., a polymeric layer as a conse-quence of analyte absorption from the gaseous phase into the bulk poly-mer. The dielectric properties of the polymer with absorbed analyte dif-fer from those of the polymeric matrix alone. A convenient way to assessthe dielectric constant of a layer or film is to measure the capacitance of apolymer-filled plate-capacitor [497] or that of polymer-coated interdigitatedelectrodes [475–484,498].

The latter approach has become very popular since the devices can beeasily fabricated in planar technology such as CMOS. Planar polymer-coatedinterdigitated capacitors also provide direct access of the analyte to the sen-sitive layer (short sensor response time), which is more difficult to realizewith modified plate capacitor designs [485]. Interdigitated capacitor designs

5.1 CMOS Capacitive Microsystems 89

(b) SEM picture(a) schematic

electrode 2

electrode 1

polymer

SiO2

Fig. 5.2. Schematic (a) and micrograph (b) of an interdigitated capacitor. Themicrograph also shows the thin polymer layer, which is tightly attached to thesensor 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 stepsavailable in the standard CMOS process sequence. Electrode 1 is made fromthe first aluminum metal layer while electrode 2 comprises a stack of inter-digitated electrodes of the first and second metal layer, which are electricallyconnected on the chip via leads, albeit they are physically separated by asilicon-oxide layer (Figs. 5.2, 5.3). The quasi “three-dimensional” electrodeconfiguration was chosen to enhance the sensitivity of the polymer-coatedsensor to analyte-induced capacitance changes by maximizing the polymervolume in the regions of strong electric field (see also Fig. 4.36) [499]. Theoverall capacitor includes 128 electrode pairs and occupies an area of 824 µmby 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 ofthe sensing capacitor to allow for polymer coating. The electrode shapes arestill visible under the polymeric coating, i.e., the polymer is tightly attachedto the surface and reproduces the surface topology.

A cross-sectional drawing of one electrode pair is shown in Fig. 5.3a, thecorresponding equivalent circuit model, which was also used for simulations,is depicted in Fig. 5.3b [500,501]. Both electrodes (E1 and E2) exhibit consid-erable parasitic capacitances to the n-well (Cnwell 1, Cnwell 2), which are verysimilar owing to the identical, though mirrored layout of the metal-1 com-ponents. The interelectrode capacitance includes two major contributions:An intermetal oxide capacitance (Cox1) and a composite of an oxide capaci-tance (Cox2) in series with an element that includes the polymer capacitance(Cpolymer) shunted with the polymer resistance (Rpolymer). The polymers usedwith interdigitated capacitors are mostly nonconducting (Rpolymer is very


Cpolymer

Cox1

Cox2

Cnwell1 Cnwell2

E2E1

RpolymerE1

polymer

E2

n-well

oxide

(a) (b)

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 asingle capacitor Cpoly with Cox2 being considerably larger than Cpolymer.

From finite-element simulations of a sensing capacitor coated with 10 µmof a polymer exhibiting a dielectric constant of ε = 4.8 (polyetherurethane,PEUT), the capacitances in the equivalent circuit model in Fig. 5.3 have beendetermined 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 changesupon analyte absorption into the polymer are in the atto-Farad range, e.g.,4.4 aF/ppm toluene in polyetherurethane. Since the parasitic capacitanceslargely prevail over the capacitance of interest, Cpolymer or Cpoly, a differen-tial readout scheme is mandatory. By using a switched-capacitor scheme witha reference and sensing capacitor of equal geometric dimensions, the refer-ence capacitor being passivated by a sufficiently thick silicon nitride layer, thecontributions of parasitic n-well capacitances and oxide capacitances cancelout to a large extent in the differential readout, since they are almost iden-tical. In view of the analyte-induced capacitance changes in the atto-Faradrange, on-chip signal conditioning circuitry is imperative, since the transferof minute analog signals via bonds and wires is very difficult [500–502]. Fromthe physical data and facts described above, it is evident that miniaturizationwithout 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 ap-plications, e.g., accelerometers, pressure sensors, fingerprint-sensors, and gassensors [499–508]. Various read-out circuitry topologies have been developedto measure the small capacitance changes of such micromachined sensors.The majority of the designs rely on a differential readout between a sens-ing capacitor and a reference capacitor, the latter being not affected by themeasurand. This offers the advantage that parasitic effects, such as tempera-ture drift and ageing, affect the reference capacitor to the same extent as the


sensor and, hence, cancel out. As a consequence, these designs do not provideaccurate information on the absolute value of the measuring capacitor.

The most popular approach to read-out capacitive sensors with highresolution and good suppression of parasitics is a switched-capacitor de-sign [509–513]. A resolution of 19 bits is needed in order to achieve a detectionlimit of 1 ppm for volatile organic compounds (VOCs). The bandwidth can beas low as 1 Hz. As none of the designs reported on in literature exhibits the re-quired performance, an improved architecture based on a switched-capacitorSigma-Delta modulator was developed.

The sensor response is measured as the differential signal between apolymer-coated sensing and a nitride-passivated reference capacitor. Both,sensor and reference capacitors are split into two parts to improve the chargetransfer efficiency. Sensing capacitor, (CS), and reference capacitor, (CR),are incorporated in the first stage of a fully differential second-order Sigma-Delta-modulator (Fig. 5.4) with two switched-capacitor integrators and asubsequent comparator [501,514,515].

CS

Cfb1'CR

Cfb1 Cfb2

Cfb2'de

cim

atio

n fil

ter

digi

tal

logi

c

Fig. 5.4. Schematic of the fully differential second-order Sigma-Delta modulatorexhibiting two switched-capacitor integrators and a subsequent comparator. Fourfeedback capacitor (Cfb) are realized as interdigital capacitors. The Sigma-Deltamodulator provides a pulse-density-modulated digital output that is decimated us-ing the frequency counter [501]

Since the output bit stream of the Sigma-Delta modulator is proportionalto 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 andreference capacitor in order to eliminate differences in temperature behaviorand ageing. Due to the small signal bandwidth, the output bit stream of theSigma-Delta modulator is decimated using a frequency counter. For moredetails on the circuitry see [501, 514, 515]. A micrograph of the integratedcapacitive microsystem is shown in Fig. 5.5.


sensingcapacitor

referencecapacitor

readoutcircuitry feedback

capacitors

Fig. 5.5. Micrograph of a capacitive sensor system including a polymer-coatedsensing capacitor, a passivated reference capacitor, four interdigitated feedback ca-pacitors, 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 heatersfor temperature stabilization since analyte absorption in the polymer matrixis 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 ab-sorption of an analyte (Fig. 5.6): (i) Swelling and (ii) change of the dielectricconstant due to incorporation of the analyte molecules into the polymer ma-trix [478, 479]. The resulting capacitance change is detected by the read-outelectronics.

polymer

substrate

electrodes

analyte

polarized analyte

air

absorption polarization swelling

Fig. 5.6. Schematic of the capacitive sensing principle showing the two relevanteffects changing the sensor capacitance: Change of the dielectric constant andswelling. The interdigitated electrodes (+,−) on the substrate (black) are coatedwith a polymer layer (gray). Big and small globes represent analyte and air mole-cules. Analyte molecules are polarized (↑) in the electric field (center). Analyte-induced polymer swelling is indicated with the dashed lines (right side) [517]


The partition coefficient, K, as introduced in Chap. 2 (2.10) characterizesthe absorption behavior of a polymer with regard to a specific analyte. It is achemical equilibrium constant and is defined as the ratio between the analyteconcentration in the sorptive or polymer phase and that in the gas phase.

For a specific analyte, K is inversely proportional to the saturation va-por pressure of this analyte and, therefore, strongly temperature-dependent.Furthermore, K depends on the analyte/polymer interaction, responsible forthe partial selectivity. The analyte concentrations in both phases can be re-placed by the partial pressure, pA, of the analyte and by the volume fractionof the analyte in the sensitive layer, ϕA, both used in the rest of this section.Equation (2.10) can be rewritten as

ϕA =pA

R · T· KC · M

ρ(5.1)

where M and ρ denote the molar mass and the density of the analyte in liquidstate, and R and T the universal gas constant and the absolute temperature[517].

As already mentioned, the two effects changing the capacitance areswelling and change of the dielectric constant. For low analyte concentra-tions, the swelling is linear in the amount of absorbed analyte, expressed by(5.2).

heff = h · (1 + Q · ϕA) . (5.2)

Here, h and heff denote the initial polymer thickness and the resulting ef-fective thickness after analyte absorption, respectively, Q is a dimensionlessnon-ideality factor of the swelling, and ϕA is the volume fraction of the an-alyte in the polymer. ϕA is proportional to the concentration of the analytein the gas phase assuming the validity of Henry’s law. The proportionalityfactor has to be determined experimentally for every polymer/analyte com-bination or can be estimated with solubility parameters or linear solvationenergy relationships [518]. Q = 1 represents ideal swelling, i.e., the total vol-ume is given by the addition of the volume of the absorbed analyte in itsliquid state to that of the polymer. For elastic polymers, the tendency togenerate static stress while swelling increases with increasing stiffness of thepolymer.

The composite dielectric constant of mixtures of nonpolar liquids can beapproximated for all kinds of analytes as proposed in [479, 517]. It can begeneralized including Q to

εeff = εpoly + ϕA · ((εA − 1) −Q · (εpoly − 1)) (5.3)

where εeff is the resulting effective dielectric constant of the polymer/analytesystem and εpoly and εA are the dielectric constants of polymer and analyte(analyte in liquid state), respectively.


thin thick6.4 µm

2µm

Fig. 5.7. Selectivity through layer thickness: Sensors coated with a thin and thicksensitive layer. The extension of the electric field lines is approximately 2 µm forthis 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 mech-anism for selectivity, because thin and thick sensitive layers exhibit differentsensitivity patterns. For a simple interdigitated structure, the space abovethe device containing 95% of the field lines is within a distance of half of theelectrode periodicity [480]. Therefore, a thin or thick layer is defined withrespect to the electrode periodicity (see Fig. 5.7).

For a layer thickness significantly less than half the periodicity, the regionof strong electric field extends above the sensitive layer. Upon analyte sorp-tion, the amount of polarizable material in the sensed region of the capacitoralways increases, which results in a capacitance increase regardless of the di-electric constant of the analyte. For a layer thickness greater than half theperiodicity of the electrodes, almost all electric field lines are within the poly-mer volume. Consequently, the capacitance is determined by the compositedielectric constant of the analyte/polymer mixture. The capacity change can,therefore, be positive or negative, depending on whether analyte or polymerhas a higher dielectric constant [479] (5.3).

The expected differences in responses from sensors with thin and thickpolymer layers have been verified experimentally. Typical sensor responseprofiles 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 hasbeen alternately exposed to various concentrations of toluene and ethanol at28C 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 pos-itive sensor signals with thin PEUT layers, whereas with thick layers, ethanolprovides a positive signal and toluene a negative signal. The limit of detectionof the capacitive microsystem at 28C is approximately 8 ppm for toluene and5 ppm for ethanol with 10 ppm analyte volume fraction corresponding to 1 Paanalyte partial pressure in the probed air [476].

With a sensor coated with a thick polymer layer, low-ε analytes can bedifferentiated from high-ε analytes if they are pure. In a mixture of these


toluene

sens

or s

igna

l [kH

z]

time [h]0 4 8 12 16

0.3 µm2.3 µm

toluene600 – 3000 ppm

ethanol1000 – 5000 ppm

polymer: PEUT

-5

0

5

10

15

Fig. 5.8. Response profiles from two capacitors coated with PEUT layers of differ-ent thickness upon exposure to 1000–5000 ppm ethanol and 500-3000 ppm tolueneas a function of time. The ratio of the dielectric constants of polymer (4.8) andanalytes (toluene: 2.36, ethanol: 24.3) controls the signs of the signals from thicklayers. For a thin polymer layer, all signals are positive [517]

analytes, more information is needed. Zero response, e.g., from the thicklayer sensor could also be generated by a certain mixture of both analytes insuch a way that the positive and negative signal just cancel out. However, thecombination of signals from sensors with thin and thick sensitive layers allowsfor determining the concentration of both analytes. The signal from the thinlycoated sensor gives the total amount of analyte, whereas the signal from thethickly coated sensor indicates the mixing ratio at a given total concentration.

Extensive numerical simulations (finite element method, FEM) were per-formed for the interdigitated capacitor with different coatings. The capaci-tance of the electrode geometry was simulated with polymer coatings of 17different, logarithmically distributed thicknesses,h, in the range of 0.06–7 µmand 41 different, linearly distributed dielectric constants, εpoly, in the rangeof 1–5. For details on the simulations, see [479, 517]. The most importantvariable to calculate from the simulated data is the sensitivity. For low an-alyte concentrations and, hence, low volume fractions, small changes in thedielectric constant, δε, and in the polymer layer thickness, δh, are expectedand the capacitance change, ∆C, can be approximated by [517]:

∆C = δh · ∂hC (h, εpoly) + δε · δεC (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]:


ethanol

ethyl acetate

toluene

h [µm]

εA < εP

εA > εP

εA » εP

0

norm

aliz

ed s

ensi

tivity

0 3 4 51 2

Fig. 5.9. Numerically simulated dependence of the vapor sensitivity of a capacitivesensor on the polymer layer thickness, h, for PEUT (εpoly = 4.8). The lines, frombottom 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, thecurves are normalized so that the sensitivities for very thin layers (initial slopes ofthe curves) are identical [517]

∆C = ϕA · [Q · h · ∂hC (h, εpoly)+ ((εA − 1) −Q · (εpoly − 1)) · ∂εC (h, εpoly)] . (5.5)

The physical sensitivity, ∆C/ϕA, was then calculated for ideal swelling andfor εpoly = 4.8 (PEUT). It is plotted in Fig. 5.9 as a function of the polymerlayer thickness. The physical sensitivity accounts for the changes in dielectricconstant and layer thickness but does not include the chemical selectivity.

As explained above, sensitivities of thin layers are all positive. For eventhinner layers, the sensitivities have to decrease because smaller polymer vol-umes absorb less analyte. For thick layers, the sign of the sensitivity dependson the dielectric constant of the analyte. Responses to analytes that havedielectric constants lower than that of the polymer (such as toluene) are pos-itive for thin layers and show a transition from positive to negative valueswith increasing layer thickness. Hence, for each of these analytes, there is acritical layer thickness, where the sensitivity towards this analyte vanishes(see circle in Fig. 5.9).

Consequently, the sensitivity reaches a (local) maximum somewhere be-tween the critical layer thickness and zero. For analytes with dielectric con-stants higher than that of the polymer (such as ethyl acetate), the sensitivityfirst increases for thin layers, reaches also a maximum, decreases again andthen saturates. In contrast to low-ε analytes, the saturation is at positivelevels. With increasing dielectric constant of the analyte, the saturation levelincreases, and the maximum becomes less and less evident. For very high-εanalytes (such as water), the maximum vanishes completely. The sensitivity


0

2

4

6

8

0 1 2 3 4 5 6 7

phys

ical

sen

sitiv

ity [p

F]

polymer thickness [µm]

ethanol tolueneethyl acetate

isopropanol

water

n-octane

Fig. 5.10. Measured dependence of the physical sensitivity, ∆C/ϕA, on the layerthickness, 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 alreadyat relatively thin layers.

To verify the simulations, the sensitivities of PEUT to various analyteswere measured for layer thicknesses between 0.3 and 7.1 µm. The physi-cal sensitivities, ∆C/ϕA, displayed in Fig. 5.10 have been calculated withindependently determined partition coefficients that have been measured withthickness-shear-mode resonators (TSMRs) [517,519].

The measurements confirm the simulations: saturation-like behavior al-ready at relatively low thickness for water (very high ε), an increasingly clearmaximum for ethanol, isopropanol, and ethyl acetate (high ε), and the tran-sition 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 achieveinsensitivity 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 dielec-tric 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 requiredaccurate adjustment of the layer thickness, which is a tedious business.

Thus, another solution is presented, where the thickness adjustment is notso critical. By tuning the dielectric constant of a thick sensitive layer to besimilar to that of the analyte, the sensitivity of the sensors to the respective


analyte almost vanishes. Insensitivity to, e.g., n-octane was achieved by blend-ing two polymers with different dielectric constants: Poly(etherurethane)(PEUT, ε = 4.8) and poly(epichlorohydrine) (PECH, ε = 1.7). Figure 5.11displays the sensor response data upon exposure to ethanol and n-octane asa function of the polymer layer composition.

0

1

2

0 25 50 75 100

freq

uenc

y sh

ift [

kHz]

ethanol

n-octane

PEUT PECHmass fraction [%]

Fig. 5.11. Frequency shifts of sensors coated with thick layers of mixtures of differ-ent PEUT and PECH content upon exposure to n-octane (1280 ppm) and ethanol(5140 ppm). When dielectric constant of polymer and analyte coincide, the sensoris blind to the respective analyte [517]

Ethanol with a high dielectric constant causes positive signals forboth polymers and all blending ratios. n-Octane has a dielectric constanthigher than that of PECH but lower than that of PEUT. Consistent with thetheory, the signals of n-octane are positive for PECH and negative for PEUT.For the blended polymers, the signals show a continuous transition from pos-itive to negative values, approximated with a quadratic function in Fig. 5.11.As can be seen, the analyte sensitivity vanishes when the dielectric constantof 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 hu-midity. Because water has a much higher dielectric constant than any polymerused, it is not possible to achieve humidity insensitivity by either of the meth-ods discussed above. However, the thickness-dependence shown in Figs. 5.9and 5.10 gives rise to another method, which is schematically depicted inFig. 5.12.

The sensitivity to analytes with high dielectric constant reaches saturationat low thickness, while the sensitivity to analytes with low dielectric constant


h [µm]no

rmal

ized

sen

sitiv

ity

humidity

n-octane

toluene

differenceΣ∆sensor 1

sensor 2

0 1 32 4 5

Fig. 5.12. Eliminating humidity interference by recording difference signals usinga thick and a medium thick layer: The simulated sensitivities from Fig. 5.9 areplotted as a function of the polymer layer thickness, h. Two sensors coated with athick and a medium thick polymer layer are represented by the dashed vertical linesat the corresponding layer thickness. The difference signal, generated by connectingboth sensors differentially to the Sigma-Delta-modulator, is indicated with the barson the right hand side of the plot [517]

still varies considerably out to greater thicknesses. Hence, the signal differenceof two capacitors with different layer thicknesses in the range of 1 µm to 5 µmis 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 thicknessregion, 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 thennumerically subtracted. However, calculating differences of similar valuesis very error-prone. Accordingly, slight temporal humidity fluctuations re-sulted in a high noise in the signal. To overcome this obstacle, two sensors onthe same chip were directly connected to the positive and negative input ofthe Sigma-Delta converter (Fig. 5.13). The signals were subtracted instanta-neously so that any fluctuations cancelled out before being processed by theread-out circuitry– resulting in a significantly improved signal-to-noise ratio.

The microsensor array used for these investigations included seven sensorand five reference capacitors, all of which are monolithically integrated withdriving and read-out circuitry on a single chip (see Fig. 5.13).

The reference capacitors are similar to the sensors but not coated with apolymer layer and allow for a differential measurement. A multiplexer (MUX)connects sensors and references to the read-out circuitry, a Sigma-Delta mod-ulator (Σ∆). The Σ∆-modulator converts the analog capacitance signal to adigital output signal. The frequency of the output signal – measured with anexternal counter – is proportional to the difference of sensor and reference


chip size: 5.8 x 6.5 mm

R

R

R

R

S

S

S

S

S

SS

R Σ∆

MUX

Fig. 5.13. Micrograph showing the sensor chip with seven sensor (S) and fivereference (R) capacitors and readout circuitry: Sigma-Delta-modulator (Σ∆) andmultiplexer (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 with1.4 and 3 µm PEUT. Figure 5.14 displays the signal of the sensor with thethin layer and the signal difference upon exposure to different concentrationsof humidity, n-octane, and toluene.

The difference signal is more than one order of magnitude less sensitive towater 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 signalquality was achieved by using the difference method, which has been detailedin Sect. 5.1.1.

Drift, which is similar for both sensors, was also reduced by approximatelyan 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 differencesignal when the analyte concentration is switched on and off. They originatefrom the different absorption times for the thin and the thick polymer layer,resulting in a temporarily high difference signal. The steady-state signal isnot affected.

5.2 CMOS Calorimetric Device

5.2.1 CMOS Calorimetric Transducer

The calorimetric transduction principle and, in particular, the thermoelectricor Seebeck-effect-based transducer has been introduced already in Sect. 4.2.2.

5.2 CMOS Calorimetric Device 101

60 120 180time [min]

∆f [k

Hz]

hum

idity

-2.0

0.0

2.0

4.0

1.4 µm1.4 µm - 3 µm

n-octane toluene

humidity

∆f [kH

z] n-octane, toluene

-20

0

20

40

Fig. 5.14. Frequency shifts of sensors coated with PEUT upon exposure to hu-midity (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 isthe difference signal of two sensors coated with 1.4 and 3 µm PEUT, whereas thedashed line represents the signal of the sensor with 1.4 µm PEUT. For better visibil-ity, the signals of n-octane and toluene are displayed in a ten-times larger scale thanthe humidity signals. The difference signal is much less sensitive to humidity thanthe signal from a single sensor, but exhibits a comparable sensitivity to n-octaneand toluene [517]

The expected temperature changes upon analyte interaction with the sensi-tive layer are in the milli-Kelvin range. Therefore, the measurement areahas to be thermally insulated from the silicon substrate, which is an excel-lent thermal conductor. Membranes consisting of the dielectric layers of theCMOS process (silicon oxide, silicon nitride) have been previously used forinfrared sensor arrays [520,521]. Dielectric membranes show a parabolic tem-perature profile, albeit a flat temperature profile over the sensitive area isdesirable for gas-sensing applications. Such flat temperature profile can beobtained by introducing a silicon-n-well island as shown in Figs. 4.10 and5.15, which is thermally decoupled from the bulk silicon of the chip. Predom-inantly the island area is afterwards coated with the sensitive layer.

The temperature on the membrane can be assessed with different trans-ducers using, e.g., the temperature coefficient of bipolar transistors or resistors.The sensitivity of these realizations is, however, not suitable to measure tem-perature changes in the milli-Kelvin range. The solution of choice is the uti-lization 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, whilethe cold contacts are placed on the substrate. Maximum signals can be ob-tained using bismuth and antimony as thermocouple materials [520]. Bothare not available in CMOS processes and would be difficult to deposit and topattern. From all CMOS materials (see Sect. 3.3) the polysilicon/aluminum


dielectric layers

n-well (Si-island) polysilicon

metal: Al

p-siliconsubstrate

polysilicon heater thermocouple

Fig. 5.15. Cross-section of a CMOS calorimetric sensor showing the thermocouples(aluminum, polysilicon) and the n-well island in the center

polysiliconheater

thermocouples

reflection spot

distancen-well to bulk

Fig. 5.16. Rectangular membrane (300 thermocouples) featuring a reflection spotfor optical layer thickness detection and a polysilicon meander heater. The thermo-couple length (distance n-well to bulk) is 200 µm

thermocouple exhibits the highest Seebeck coefficient of 110 µV/K. Manythermocouples 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 fourprincipal steps: (I) absorption and partitioning or chemical reaction, (II) gen-eration of heat, which causes (III) temperature changes to be transformed in(IV) thermovoltage changes. The final signal results from this sequence ofprocesses, some of which are of chemical or physicochemical nature, i.e., de-pend on the involved chemical compounds (I, II), and some of which are ofphysical nature and are device-specific (III, IV: heat sensitivity and thermo-voltage generation).

The overall sensitivity is the recorded thermovoltage change ∆U [V] perchange in the analyte concentration as a function of time dcA/dt [mol/m3s]and can be written according to (4.3), Sect. 4.2.2:


S =∆U

dcA/dt= A · B · Vsens · ∆H · K . (5.6)

A [K·s/J] and B [V/K] denote device-specific constants describing the trans-lation of a generated molar enthalpy ∆H [J/mol] via a temperature changeinto a thermovoltage change. The constant A [K·s/J] includes the featuresizes of the transducer and material properties, and B [V/K] includes thenumber of thermocouples and the Seebeck coefficients of the thermocouplecomponents. 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 chemicaland the physical sensitivity.

The chemical sensitivity is correlated to the amount of heat that is gen-erated upon an analyte concentration change. It comprises the reaction con-stant or partition coefficient, K, the produced enthalpy change, ∆H, andthe sensitive layer volume, Vsens. This part will be treated in more detail inSect. 5.2.3.

The physical sensitivity is the thermovoltage output of the system inresponse to a defined heating power on the membrane. It includes the trans-ducer features, whereas it does not depend on the detection chemistry. Someof the most important design parameters are briefly discussed.

• Size of the membrane: A large membrane leads to a large sensitivearea and a large number of thermocouples. The mechanical stability ofthe membranes and the maximum allowable chip area determine the sizelimits. 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 im-proves the thermal isolation of the n-well island and, therefore, the effi-ciency of the heating. The mechanical stability of the membrane and chiparea considerations have to be taken into account. The increase of the elec-trical thermocouple resistance owing to the larger distance between hotand 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 ofwhich exhibit comparably high thermal conductance. A large number ofthermocouples thus reduces the thermal resistance of the membrane and,consequently, degrades the temperature effects. This would suggest the useof a large number of minimum-width conductors to increase the thermalresistance and make efficient use of the available area. Since polysilicon re-sistors exhibit low electrical conductivity, the line width cannot be reducedtoo much without increasing the thermal noise in the signal.

In a viable compromise the square membrane design has a total of 132 thermo-couples (40 kΩ electrical resistance), the rectangular membrane (Figs. 5.16,


low-noisedifferentialamplifier

sensor

reference

Fig. 5.17. Micrograph of a differential thermopile configuration with integratedlow-noise amplifier. The measurement thermopile is coated with polymer, the ref-erence is passivated with silicon nitride [522]

5.17) features 300 thermocouples (143 kΩ electrical resistance). A differentialstrategy that has been already utilized in the case of the capacitive transduceris 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 fluctuationsis 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 dif-ferent analytes. The thermopile converts the temperature difference betweensubstrate and n-well island into a voltage signal. For the best overall perfor-mance, the input-referred SNR of the first amplifier and not the signal has tobe maximized. The two thermopiles used here exhibit electrical resistances of40 kΩ (square membrane) and 143 kΩ (rectangular membrane). The overallsystem optimization includes the input noise of the first amplifier as well asthe power and area consumed by the complete readout-circuitry [501, 520].Another important parameter is the signal bandwidth, which must be assmall as possible in order to minimize the noise. The highest frequency ofinterest in the transient signal during absorption or desorption of analyteswas experimentally determined to be approximately 400 Hz.

A system and circuitry architecture based on the aforementioned consid-erations 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 ref-erence. Temperature fluctuations owing to medium flow and radiation arelargely cancelled out by the differential arrangement. The small signals are


+

−

low-noisechopperamplifier

sensor

referenceSigma-Delta

A/D converter

anti-aliasingfilter

AD

decimationfilter

13bit

Fig. 5.18. Schematic of the calorimeter circuitry. Sensing and reference thermopilesare connected to the input stage of a low-noise chopper-stabilized instrumentationamplifier followed by an anti-aliasing filter, a Sigma-Delta A/D converter and adecimation 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 measure-ments over a large concentration range, i.e., for the use of a variety of differentanalyte-sensitive-layer combinations. The maximum gain of 6400 can be re-duced by a factor of 4 or 16. The resolution is 12 bits, which is sufficient for thetarget applications. An anti-aliasing filter succeeding the amplifier is neededto avoid down-conversion of high frequency noise by the A/D-converter.

The system has a narrow signal band from 0–400 Hz. The requirementsfor the stop-band are set by noise constraints, by the suppression of thesignals upconverted to the chopping frequency by the demodulator at theoutput of the amplifier, and by clock-feedthrough. A minimum damping of80 dB at the chopping frequency is needed to meet these requirements. Thelow cut-off frequency of 400 Hz would require large on-chip capacitors in theanti-aliasing/bandstop filter if a Nyquist-rate analog-digital converter wouldbe used. By using a largely oversampled Sigma-Delta-A/D-converter, thefiltering can be shifted to the digital decimation filter, where low frequenciesdo not increase complexity, area, and power-consumption of the design. Thisalso simplifies the design of the anti-aliasing filter. Due to the narrow signalband, an oversampling factor of 128 can easily be realized. This makes aone-bit Sigma-Delta modulator the most suitable choice for the 12-bit A/D-converter owing to the less stringent matching requirements. For a detaileddescription 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 achemical by measurement of an enthalpy change produced by the chemicalto be detected. Any chemical reaction or physisorption process releases orabsorbs a certain quantity of heat from its surroundings. As already brieflymentioned in Sect. 4.2.2, the calorimetric sensor only detects changes in theheat budget at nonequilibrium state (transients) upon changes in the analyteconcentration.


Thermodynamic equilibrium is a dynamic equilibrium state, in which re-actants and products are present but have no tendency to undergo net change,i.e., as many molecules undergo the forward reaction as undergo the reversereaction. The change in the Gibbs free energy, ∆G, is zero (see 2.1 and relatedtext), and there is consequently no net heat production. The calorimetric sen-sor, therefore, only produces transient signals as is illustrated in Fig. 5.19.The initial state is thermal equilibrium with no analyte present and, hence, notemperature signal. The analyte concentration then is drastically increased,which leads to sorption/reaction heat liberation and, e.g., a temporal temper-ature 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).

(a) no analyte (b) analyte absorption (c) equilibrium

time

sign

al

time

sign

al

time

sign

al

polymer

Fig. 5.19. Transient signal characteristics of the calorimetric sensor: (a) no ana-lyte present, (b) changing or increasing analyte concentration, (c) thermodynamicequilibrium [501]

In case that an analyte concentration increase produces a positive calori-metric signal, the corresponding decrease produces a negative signal of equalintensity and vice versa (see Fig. 5.21). The signal and detection character-istics of the calorimetric transducer have some implications for its use in gassensing.

Slow changes in analyte concentrations are difficult to detect, since theycontinuously produce but small enthalpy changes. Therefore, a switchingscheme has to be implemented, which allows for alternately exposing thesensor to analyte-loaded medium and non-contaminated medium, i.e., gener-ates 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 mediummanifold and the diffusion of the analyte into the sensitive layer. Therefore,


the peak area integrated over time is the characteristic sensor signal thatrepresents the total enthalpy change upon concentration changes (see alsoFig. 5.21).

For gas sensors with standard polymer coatings (deposited by spray orspin-coating), which will be discussed in more detail here, only physisorptioncontributes to the enthalpy change, since no chemical reaction is taking place,and no chemical bonds are formed between polymer and the analyte to bedetected. The temperature change in the polymer, ∆T, upon analyte sorptionarises from the overall heat production, ∆H, which, in the special case ofvolatile absorption in a polymeric matrix, includes two components, the heatof condensation and the heat of mixing:

∆T ∝ ∆Htotal ; ∆Htotal = ∆Hcond + ∆Hmix . (5.7)

∆Htotal here denotes the overall enthalpy change ([J/mol]). The analyte ab-sorption process in the polymeric coating can be conceptually subdivided intwo steps (Fig. 5.20) [516]: Analyte condensation from the gas into the liquidphase and subsequent mixing with the polymeric matrix. Consequently, thefirst term on the right-hand side of (5.7) is the condensation enthalpy, whichis characteristic for the analyte and independent of the sensitive coating orpolymer. The second term describes the enthalpy change upon mixing of ana-lyte and polymer. This term includes all polymer/analyte interactions and is,for many analyte/polymer combinations, small in comparison to the overallenthalpy change or condensation enthalpy. For more details see [516,524].

analytemolecules

polymermatrix

sorption = condensation + mixing

Fig. 5.20. Conceptual subdivision of the analyte absorption process: Condensationin the liquid phase and subsequent mixing with a polymeric matrix

Typical chemical sensor signals of the calorimetric transducer are dis-played in Figs. 5.21 and 5.22, which exemplify the detection of different kindsof organic volatiles in the gas phase (hydrocarbons, alcohols etc.) by usingvarious polymeric layers. Figure 5.21 (a) and (b) show the output voltageof the microsystem and the peak integrals while switching from synthetic air(nitrogen/oxygen mixture without humidity) to n-octane (900 ppm) and back


time [s]time [s]

ther

mov

olta

ge [m

V]

on

n-octane: 900 ppm

off

PDMS

Fig. 5.21. Calorimetric sensor signals: Output voltage of the microsystem andpeak integrals while switching from synthetic air to n-octane (900 ppm) and backto air at 28C with 2 µm PDMS as sensitive layer

to air at a temperature of 28C. The sensitive layer is poly(dimethylsiloxane),PDMS. As already discussed above, two transient signals are produced, a pos-itive one at the onset of the analyte exposure (liberation of predominantlycondensation enthalpy) and a negative one upon terminating analyte expo-sure (abstraction of the enthalpy necessary for vaporizing the absorbed ana-lyte). The net enthalpy changes can be approximated by integration over thepeak area of the sensor signals [522,523], which is the measurand of interest.

The height of the peak maximum is strongly depending on the polymerthickness (diffusion kinetics), and the peak maximum and signal character-istics may also differ for switching on and switching off the analyte as aconsequence of slightly different switching times or gas flow fluctuations. Anoptimized 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 poly-meric sorption matrix are displayed in Fig. 5.22. Whereas the slightly polarpoly(etherurethane) shows large signals upon alcohol exposure (methanol,ethanol), the nonpolar poly(dimethysiloxane) provides hardly any signal foralcohols albeit showing large signals for nonpolar compounds such as n-octaneand toluene. The sensor response patterns in Fig. 5.22 demonstrate, thatsensor selectivity can be tuned by polymer variation and confirm the oldalchemist rule: “Like dissolves like”, i.e., nonpolar matrices predominantlyabsorb nonpolar analytes, as well as polar matrices tend to absorb polar ana-lytes. Note that the PDMS layer is with 1.5 µm thinner than the 3 µm PEUTlayer and that the signals scale linearly with the polymer volume/layer thick-ness.

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 109

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

150 200 250 300 350

toluene500-2000

ethanol1000-4000

n-octane250-1000 ppm

methanol2000-3500

PDMS

them

ovol

tage

[V]

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0 50 100 150 200 250time [min] time [min]

PEUT

Fig. 5.22. Calorimetric response patterns of two different polymers (PEUT andPDMS) upon exposure to an identical set of organic solvents

significant differences. The calorimetric sensor provides two signals for eachanalyte exposure, which makes the concentration assessment more reliable.The measurement time is also significantly shorter since the transient evo-lution is on the order of 3–5 seconds, whereas it may take tens of secondsto reach equilibrium and a reliable sensor reading for equilibrium-based sen-sors at similar polymer layer thickness (approx. 3 µm). The calorimetric sen-sor signal is in addition inherently drift-free: No rapid concentration changemeans no sensor signal. Drawbacks of the calorimetric principle include asomewhat inferior analyte sensitivity in comparison to other polymer-basedsensors (factor of approximately three to five) and the necessity of drasticgas 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 gravimet-ric transducers and can be operated either in a static mode by measuring thecantilever deflection or in the dynamic mode by assessing changes in the res-onance behavior. The fabrication in industrial CMOS technology allows forthe integration of transducers, driving circuitry, and analog or digital signalprocessing circuitry on the same chip [93,95]. Integrated thermal or magneticexcitation and piezoresistive detection schemes render the system indepen-dent of additional excitation elements such as piezoelectric layers [61–63,106],and independent of bulky external optical detectors [100,103–105]. The can-tilevers that will be described in more detail below are operated in the dy-namic mode, and the change of the resonance frequency upon mass load-ing is measured. The operation features of resonant cantilevers are, as al-ready mentioned, similar to those of other well-established mass-sensitive gas


sensors such as thickness-shear-mode resonators (TSMRs, quartz microbal-ances) [61–63,66,68–70] and surface-acoustic-wave (SAW) [61–63,68–70] de-vices.

5.3.1 Resonant Cantilever Transducers

There are basically three ways to determine the resonance frequency of acantilever [501]:

• Noise spectrum: The frequency spectrum of the thermal noise is mea-sured and analyzed. The peak frequency is equivalent to the resonancefrequency. 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 dif-ferent frequencies and its response is evaluated to determine the resonancefrequency. A gain-phase analyzer is needed for the measurement.

• Oscillator: The cantilever is used as the frequency-determining elementin an oscillator loop. A cantilever actuation mechanism and a deflection-sensor 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 thequality factor (Q-factor, ratio of center frequency and bandwidth) of thecantilever, which describes the sharpness of the system response, is enhancedby the feedback. In addition, only the third option can be realized in CMOStechnology at reasonable expenses, since integrated spectrum analyzers andgain-phase meters are difficult to design. Consequently, the oscillator optionwas realized. In the following, two different methods to actuate the cantilever(thermal and magnetic), and the vibration detection via piezoresistors andstress-sensitive transistors will be presented.

5.3.1.1 Thermal Actuation

The cantilever consists of single-crystal silicon covered by dielectric layerssuch as silicon oxide or silicon nitride for electrical insulation of the inte-grated electronic components. A cross-sectional view is shown in Fig. 5.23.Two heating resistors are integrated in the cantilever base, which are heatedperiodically to achieve cantilever vibration (see also Fig. 5.24). The tem-perature increase on the cantilever generates a bending moment due to thedifference in thermal expansion coefficients of the silicon and the dielectriclayers (thermal bimorph effect).

The area of periodic temperature variation is closely confined to the re-gion around the heaters due to the high excitation frequency of approx.400 kHz. At frequencies higher than the corner frequency of approximately1 kHz (speed of the thermal equilibration processes in the cantilever), the


polymer

n-well

siliconframe

heater(p-diffusion)

Si-nitride Si-oxide

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 ofthe bimorph effect (difference in thermal expansion coefficients of the silicon andthe dielectric layers) [526]

heater

piezoresistors

50 µm

Fig. 5.24. Thermal actuation of the cantilever motion. The micrograph shows theheaters and the piezoresistors at the cantilever base [93]

efficiency of the thermal actuation usually significantly drops, which reducesthe cantilever oscillation amplitude and leads to a phase lag that increaseswith frequency. However, the created bending moment is still sufficient tocause harmonic transverse vibrations of the cantilever with an amplitude ofa few nanometers.

The deflection of the cantilever is proportional to the applied heatingpower. Therefore, a sinusoidal voltage excitation cannot be used, since all theheating 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 ofthe cantilever [501].

Both types of cantilevers (thermally or magnetically actuated) are formedafter completion of the regular 0.8-µm CMOS process sequence [496,527]. Thefirst step includes anisotropic wet etching of silicon from the wafer backsidewith KOH using an electrochemical etch-stop technique (Fig. 5.25b). After


this step the silicon cantilevers are still embedded in a membrane formedby the dielectric layers. The second post-processing step consists of isotropicwet etching of silicon dioxide with buffered oxide etch (BOE), which finallyreleases the discrete silicon cantilevers (Fig. 5.25c). The cantilevers are com-posed of a layer sandwich including the dielectric layers of the CMOS processand a layer of single-crystal silicon (Fig. 5.23). The silicon layer has a thick-ness of 5.5 µm, silicon dioxide and silicon nitride thicknesses are 1.5 µm and1.1 µm, respectively. The cantilevers are 150 µm long and 140 µm wide. Theirfundamental resonance frequency is in the range of 400 kHz, and they exhibita Q-factor of up to 1000 in air without chemically sensitive coating.

n-well

dielectric layers

Si

(a)

(c)

(b)

heater

Fig. 5.25. Micromachining process sequence to fabricate cantilevers: (a) thinnedCMOS wafer with Si-nitride on back side, (b) backside KOH wet etching withelectrochemical etch stop, and (c) front-side isotropic wet etching of silicon dioxideto release the cantilever [47,526]

The quality factor of the resonant cantilevers at atmospheric pressure isdominated by the viscous damping in the surrounding air [93]. In comparisonto cantilevers employed in Scanning Probe Microscopy (SPM), the cantileversas used for chemical sensors exhibit high quality factors. This is partly dueto the fact that the cantilevers are not operated in close vicinity of a samplesurface, and, hence, no squeezed-film damping occurs. The distance to thebottom of the etch cavity is 380 µm, i.e., large in comparison to the can-tilever dimensions. Furthermore, the quality factor is not only a function ofthe resonance frequency but also of the cantilever geometry and its springconstant [528].

While SPM cantilevers for dynamic mode operation typically exhibitspring constants between 1 and 40 N/m, the cantilevers here were designed


to exhibit a much higher spring constant of 800 N/m. The measured qual-ity factor of cantilevers in fundamental resonance strongly increases withincreasing resonance frequency up to 200 kHz, reaching maximum values ofapproximately 1000 [93, 526]. This behavior is predicted by the theory of adamped harmonic cantilever beam with a damping term proportional to thebeam velocity [529]. A theoretical estimation of the actual quality factorswould, however, require a model for the dependence of the damping term onthe cantilever dimensions. At frequencies higher than 200 kHz, a saturation-like behavior of the quality factor has been observed, indicating additionalloss mechanisms, such as coupling losses into higher resonant modes or lossesdue to onset of acoustic radiation at higher frequencies.

5.3.1.2 Magnetic Actuation

The magnetic actuation relies on Lorentz forces and requires an externalmagnetic field, which can be conveniently generated by including a smallpermanent magnet in the sensor package below the cantilever (Fig. 5.26)[530, 531]. The resulting magnetic-field vector is in plane and parallel to thecantilever.

By applying an AC current such as a sinusoidal to the current loops thatare patterned along its edges, cantilever oscillation is evoked in the externalmagnetic field by Lorentz forces. The direction of the Lorentz forces is per-pendicular to the cantilever (Fig. 5.27) such producing a transverse cantilevermovement. Two designs with different current paths for the electromagneticactuation were implemented [531]. The current paths are realized using thetwo metal layers of the CMOS process. One design has a path with twoloops and a resistance of approximately 12 Ω, the second has 8 loops and

Si-substrate

cantilever

Al-package

magnetic field lines

disc magnet

Fig. 5.26. Schematic of applying magnetic actuation to cantilevers: The externalmagnetic field is provided by a permanent magnet in the package [531]


cantilever

B

X FL

50 µm

J

Close-upWheatstone bridge

stress-sensitivetransistors current

loopWheatstone

bridge

Fig. 5.27. Micrograph of a magnetically actuated cantilever: AC current appliedto 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 beapproximated 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 themean length of the loop perpendicular to the external magnetic field, Bext,(see Fig. 5.27). When the magnetic field, Bext, is oriented parallel to thecantilever length, the Lorentz-force vector is perpendicular to the plane ofthe cantilever.

The Lorentz force exerted on the cantilever critically depends on the num-ber of current loops. With a magnetic field of 80 mT, the effective current for2 loops is 3.5 mA, and the resulting Lorentz force amounts to 37 nN, whereasin the case of 8 current loops, the effective current is 7.1 mA and the Lorentzforce 71 nN. Especially for thick layers of chemically sensitive coatings, it istherefore necessary to use cantilevers with larger numbers of current loops toreach stable oscillation conditions.

5.3.1.3 Vibration Detection

The Wheatstone bridge for vibration detection is located at the clamped edgeof the cantilever, where the stress induced by the deflection is maximal. Thebridge consists of either four piezoresistors (Fig. 5.24) or four diode-connectedPMOS transistors (Fig. 5.27) [526].

The piezoresistive effect on a slightly p-doped silicon resistor is describedby the relative resistance change, ∆R/R,


∆R

R= πLσL + πTσT (5.9)

where πL,T are the longitudinal and transversal piezoresistive coefficients andσL,T denote the respective stress-components. The piezoresistive coefficientsin parallel and perpendicular orientation with respect to the cantilever axishave opposite signs. Therefore, a differential signal can be obtained by arrang-ing two resistors in parallel and two resistors perpendicularly to the cantileveraxis in a Wheatstone-bridge configuration (see Fig. 5.24). The common-modevoltage varies only a few percent because the absolute values of the longitu-dinal and transversal piezoresistive coefficients are almost identical [501].

The resistances in the Wheatstone-bridge are 950 Ω. This leads to a con-siderable power consumption of 26 mW. As minimum-width resistors had tobe chosen for the Wheatstone-bridge due to area restrictions, a good match-ing cannot be expected. The matching of the resistors is further deterioratedby their perpendicular orientation. This leads to a large offset-voltage of upto 20 mV.

MOS-transistor detection schemes have also been developed since me-chanical stress changes the carrier mobility in the channel of MOS transis-tors [532, 533]. Four diode-connected PMOS transistors in a bridge config-uration are used to generate an output voltage that is proportional to thestress at the clamped edge of the cantilever [532]. Again, opposite signs of thestress-sensitivities are achieved by orienting the channels of the two pairs oftransistors perpendicularly and in parallel to the cantilever axis (see close-upin Fig. 5.27). The diode-connected MOS transistors show a lower small-signaltransconductance in comparison to diffused resistors, which reduces powerand area consumption. The disadvantages include a somewhat larger noiseand a reduction of the sensitivity.

5.3.1.4 Cantilever Temperature

The power dissipation of the electrothermally actuated cantilever is 32 mWwith 6 mW resulting from the actuation and 26 mW from the piezoresistiveWheatstone bridge. For the magnetically actuated cantilever the power dis-sipation of the cantilever adds up to a total of 1.3 mW [501, 531]. It resultsmainly from the MOS-transistor Wheatstone bridge, the power dissipationof the actuation is negligible. The Wheatstone bridge with MOS-transistorshas less power dissipation since its active load (20 kΩ) is much larger thanthat of the design with p-diffused resistors (950 Ω). As a consequence of thehigh power consumption and dissipation in the case of electrothermal actu-ation and piezoresistive detection, the cantilever temperature is significantlyenhanced as can be seen in Fig. 5.28, which displays the simulated tempera-ture distribution on the cantilever at minimum possible power consumption(actuation: 6 mW, bridge: 1.2 mW) [526]. The cantilever temperature un-der realistic operation conditions can be up to 20C higher than ambienttemperature, which strongly affects the detection process of chemicals, such


0

3.5

1.7

thermal

actuators

(6 mW)

Wheatstonebridge

(1.25 mW)

temperaturevariation [°C]

Fig. 5.28. Temperature distribution on a thermally actuated cantilever with piezo-electric readout [526]

as the absorption of organic volatiles in polymeric matrices. As a rule ofthumb, a 10-degree temperature increase here reduces the sensor signal by50%. Magnetic actuation and MOS-transistor readout that do not produceany significant temperature change on the cantilever are, therefore, in mostcases, a better choice.

5.3.2 Microcantilever Circuitry

The cantilever constitutes the frequency-determining element of an oscillatorcircuit, the nature of which depends on the actuation/detection principle andwill be, therefore, specified for thermal and magnetic actuation.

5.3.2.1 Thermal Actuation

To operate the mechanical oscillator, the displacement generated by the heat-ing pulses and the excitation voltage variations have to be in phase for positivefeedback. A static DC-component has to be superimposed to a periodic exci-tation voltage to achieve cantilever oscillation at the mechanical fundamentalresonance (see Sect. 5.3.1.1) [501]. This DC component of the heating currentproduces a general temperature increase on the cantilever. The power dissi-pation in the piezoresistors of the Wheatstone bridge constitutes a secondsource of heating. Both heating effects can cause a total temperature in-crease of up to 20C above ambient temperature in the chemically sensitiveregion of the cantilever with approximately 10C above ambient temperatureat routine operation [93].

Measurements showed that the signal from the mechanical vibration onlydominates in vicinity to the resonance frequency. At low frequencies, thethermo-mechanical actuation heats the whole cantilever and creates a tem-perature gradient in the Wheatstone-bridge that leads to an offset. Due to


the large temperature coefficient of diffused resistors, the resulting signal islarger than the mechanical response for frequencies up to 200 kHz (uncoatedcantilever). Low-frequency signals therefore have to be eliminated by the feed-back circuit, in particular since the amplitude of the thermal crosstalk maycome close to the amplitude of the mechanical signal even at resonance. Forfrequencies higher than 200 kHz, capacitive crosstalk is observed as a conse-quence of the small distance between the diffused heating resistors and thediffused resistors of the Wheatstone bridge. This small distance gives rise toparasitic capacitances through the n-well.

The oscillator circuitry as shown in Fig. 5.29 was designed according tothe preceding considerations to achieve optimum signal quality [501]. Theoutput signal of the resistive Wheatstone-bridge is first amplified by a low-noise differential difference amplifier (DDA). The amplification factor has amaximum value of 35 in order to avoid saturation of the amplifier by theDC-offset of the Wheatstone-bridge. The signal is then high-pass filtered toremove the offset voltages of the bridge and the first amplifier. The high-passfilter also prevents up-conversion of the amplifier 1/f-noise and eliminatesthe low-frequency thermal crosstalk and related error signals. AC-couplingat the input of the first amplifier would allow for a higher gain in the firstamplification stage, but the related disadvantages prevail. They include addednoise and the creation of an additional path for switching interference andsubstrate noise to the most critical point in the circuit.

R1

R4

R2

R3

RH +-

+-

∆tDDADDAWheatstone

bridge

cantilever

high-pass limiter

1stamplifier

2ndamplifier

delay

Fig. 5.29. Schematic of the cantilever feedback circuitry, which includes two cas-caded 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 (notshown) before it is converted into a square-wave signal by the comparator.The second amplification stage is needed to achieve sufficiently large ampli-tudes at the input of the comparator. The minimum amplitude at the inputof 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) isneeded for the desired duty cycle of 45–55%; and (b) the noise and crosstalkat the input of the comparator.


The feedback loop is then closed via an inverter followed by a Schmitt-trigger, which is operated as a delay element. The rise-time of the inverter isdigitally adjustable. This way, the phase shift is adjusted to ensure positivefeedback. A source follower at the output of the delay line is used to drive thesmall heating resistor. The result is an integrated oscillator operating at thecantilever resonance frequency with a short-term frequency stability betterthan 0.1 Hz. The integration of the feedback loop on chip (Fig. 5.30) massivelyimproves the signal-to-noise characteristics of the sensor. The detection ofmass changes of less than one picogram on the cantilever has been achievedby 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 detectionscheme, 8 current loops and an overall coil resistance of 48 Ω, (see Fig. 5.27and Sect. 5.3.1.2), has been incorporated into an integrated oscillator. Somekey issues for designing the feedback circuitry shall be briefly discussed.

The vibration-induced output signal of the Wheatstone bridge exhibits anamplitude of 1 mV for a bias-current of 8.3 mA (corresponds to an excitationvoltage of 400 mV) and a magnetic field generated by a 100-mT electromag-net. Consequently, the feedback circuitry has to provide an amplification ofat least 32 dB (air) or 63 dB (water) to achieve a loop gain of more than0 dB.

The signal amplitude at the output of the MOS-transistor bridge variesby almost an order of magnitude in dependence of the excitation current,


the magnetic field, the distance between the magnet and the cantilever, andthe damping of the vibration through the media, in which the cantileveris operated. Additionally, there will be drift on the bridge sensitivity ow-ing to temperature fluctuations and material aging. For operation in liquids,the high viscosity and density of the surrounding medium entails enhanceddamping and, consequently, a reduced signal amplitude (e.g., 30 dB reduc-tion). Therefore, the gain of the feedback loop has to be adjustable in orderto achieve stable oscillation under all operating conditions. The tunable gainof the second amplifier is used for the coarse tuning during startup. Smallvariations during circuitry operation are continuously compensated by havinga nonlinear transconductance in front of the output stage.

The total phase shift between the excitation and the output signal of thecircuitry must be close to zero degree over a wide frequency range in orderto achieve stable oscillation at the mechanical resonance of the cantilever.This holds particularly true since the mechanical resonance frequency of thecantilever varies with the coating material or the surrounding medium andis additionally subjected to fabrication tolerances.

The thermomechanically actuated cantilever is driven by a square-wave-type 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 canbe 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 of25 mV ±5 mV. This offset must be reduced or eliminated to avoid saturationof the amplifiers in the oscillator loop. Furthermore, the excitation signal atthe cantilever should not include any DC-component to prevent unwantedheating of the cantilever and to minimize the cantilever power consumption.

The last stage of the feedback circuitry has to drive the low-resistancecoils 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 chosenfor the first amplification stage, since it requires little area and has low powerconsumption. The gain defined by the feedback resistors is 30 dB to avoidsaturation of the amplifier by the DC-offset of the Wheatstone bridge.

Between the first and second amplification stage, a first-order high-passfilter 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 isat least three times smaller than the resonance frequency of the cantilever.

The second amplifier is a variable-gain amplifier based on the DDA topol-ogy. In comparison to the first DDA, the design of the second amplificationstage is more flexible. Instead of using a defined feedback resistance to deter-mine the closed-loop gain of the amplifier, a current-controlled differential lin-ear transconductor is introduced into the feedback loop of the amplifier [536].By changing the bias current of the transconductor in the feedback loop, theclosed-loop gain can be varied. The closed-loop gain of the second amplifier


cantilever differential differenceamplifier (DDA)

high-passfilter

variable-gainamplifier

class-ABbuffer

amplituderegulation

high-passfilter

readout

analog

digital

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 rangeof 50 µA to 110 µA. It is intended to introduce a programmable bias currentin the next design.

After this stage, another high-pass filter, which has the same structure andsame cut-off frequency as the first high-pass filter mentioned before, is addedto remove the offset of the upstream amplifiers. To initiate a self-oscillationof 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 belarger than 0 dB and has to be controlled so that the gain is depending on theoscillation amplitude. To this end, a nonlinear transconductance as proposedin [537] is implemented in the system to regulate the oscillation amplitude.

The load of the feedback circuitry includes the coil, which is integrated onthe cantilever. Compared to its resistance, the inductance of the coil can beneglected. The resistance of eight loops used in this design is only 48 Ω, whichrequires a comparatively large driving current. As low power consumption isan 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 asimple buffer structure. For more details on the circuitry see [531,534].

The mass change of the cantilever upon analyte exposure leads to a fre-quency change. The oscillation signal can be read out as an analog voltageor as a digital square wave. In this design, a buffer drives the output signal,which is measured externally. A micrograph of the overall integrated systemis shown in Fig. 5.32.

A simple and affordable solution to generate the magnetic field is the inte-gration of a permanent magnet in the chip package. Before designing the pack-age, it was necessary to find the minimum magnetic induction needed for astable oscillation of the cantilever feedback system. Measurements performed


variable-gainamplifier

differential differenceamplifier DDA

class-AB buffer nonlineartransconductance

cantilever

500 µm

high-pass filter

high-pass filter

Fig. 5.32. Micrograph of a magnetically actuated cantilever (MOS-transistor de-tection) with monolithically integrated feedback circuitry [534]

with a tunable electro-magnet showed that the minimum magnetic flux den-sity necessary for stable oscillation is 70 mT [531].

The feedback circuitry exhibits a short-term frequency stability of betterthan 0.03 Hz in air. In contrast to thermal actuation, the application of whichto resonant sensing in the liquid phase is difficult as a consequence of thermallosses and heat dissipation, the magnetic actuation scheme is also applica-ble to liquid phase dynamic measurements. Consequently the magneticallyactuated cantilever system was characterized in air and in water. In bothmedia, the system was first characterized in open-loop and, afterwards, inclosed-loop operation (Fig. 5.33).

In contrast to thermal actuation, the application of which to resonantsensing in the liquid phase is difficult as a consequence of thermal lossesand heat dissipation, the magnetic actuation scheme is also applicable toliquid phase dynamic measurements. Consequently the magnetically actuatedcantilever system was characterized in air and in water. In both media, thesystem was first characterized in open-loop and, afterwards, in closed-loopoperation (Fig. 5.33).

Under open-loop conditions at an external magnetic field of 200 mT, thecantilever oscillates at a resonance frequency of 425 kHz in air with a qualityfactor of 750 to 1100 (>100.000 for closed loop in air). In water, the funda-mental resonance frequency of the system drops to 219 kHz with a qualityfactor of 23. Not only the quality factor but also the oscillation amplitude ofthe cantilever is reduced dramatically in water.

The improvement of the quality factor of the cantilever in water as aconsequence 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


closed-loopresponseQ = 19,000

open-loopresponseQ = 23

frequency [kHz]

ampl

itude

[V]

Fig. 5.33. Open-loop and closed-loop frequency response of cantilever system inwater [534]

operation. The resonance frequency in closed-loop operation in water is221 kHz, which is slightly different from the resonance frequency in open-loopoperation. This is due to the fact, that in closed-loop operation the cantileveroscillates at a frequency, at which the total phase of the system is 0, whichis 221 kHz in this measurement. Figure 5.33 clearly demonstrates again thebenefits of having circuitry monolithically integrated with the resonant struc-ture.

5.3.3 Microcantilevers as Chemical Sensors

To render the cantilever sensitive to chemical species, thin layers (3–6 µm) ofpolymers were deposited, which serve as absorption matrix to detect volatileorganic compounds in air. The investigations were restricted to polymericfilms, for which physisorption (no chemical interaction) and bulk dissolutionof the analyte within the polymer volume are the predominant mechanisms.Upon absorption of analytes by the coating, the physical properties of thepolymer film, such as its mass, change. This mass change is detected by mon-itoring 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 spraycoating using an airbrush or by drop deposition from polymer solutions,already changes the resonator properties of the cantilevers. Figure 5.34


-400

-350

-300

-250

-200

-150

100 200 300 400

frequency [kHz]

uncoated

coated

0.01

0.1

1

10

100 200 300 400

frequency [kHz]

uncoated

coated

phas

e [°

]

ampl

itude

[mV

]

Fig. 5.34. Measured amplitude and phase of a cantilever before and after coatingwith PEUT (after amplification by the DDA, 30 dB). The mechanical resonancesignal determines overall amplitude and phase characteristics only in vicinity tothe resonance frequency. At lower and higher frequencies, thermal and capacitivecrosstalk 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 andthe amplitude at resonance decrease. A larger loop-gain is needed to com-pensate for the decrease in amplitude. For a harmonic oscillator, the phaseshift between excitation at resonance and response is independent of theresonance frequency. Due to the thermomechanical actuation, the resonantcantilever shows an additional phase lag, which increases with frequency aswas mentioned in Sect. 5.3.1.1. The phase shift at resonance is decreasedafter the coating procedure and depends on the polymer thickness. Due tothis decrease and the fabrication-tolerance-induced fluctuation of the initialresonance frequency (<5%, see Fig. 5.35), the phase at resonance cannot bepredicted accurately. The feedback circuitry therefore includes an adjustablephase-shifter in order to guarantee a positive feedback also with larger poly-mer coating thickness.

The Q-factor of the cantilever and, accordingly, the resonator stabilityare also decreasing with increasing polymer-thickness. While the chemicalsensitivity increases with polymer thickness (more sorption matrix produceslarger mass changes), the minimum detectable frequency shift is defined bythe stability of the oscillator. This leads to a trade-off for the polymer thick-ness. 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 dropsfrom 950 to 600. This is partly due to the fact that quality factor and vi-brational amplitude depend to a critical extent on the plastic or viscousproperties of the thin polymer films. Furthermore, it is observed that quality


375 400 4250

0.02

0.04

gain

fundamental frequency [kHz]

frequency:394 kHz ± 3%

Q-factor in air:980 ± 5%

Fig. 5.35. Micromachining fabrication tolerances of 4 cantilevers in an array onthe same chip (see Fig. 5.39) [526]. The resonance frequency varies between 380and 405 kHz (394 ± 3%), the Q-factor is 980 ± 5%

factor and amplitude depend strongly on the homogeneity and morphologyof the polymer film and, thus, on the quality of the deposited layer and thereproducibility of the deposition process.

Since the cantilever beam shows a decrease in resonance frequency uponsensitive-layer deposition, the polymer layer thickness can be monitored online 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 producesa frequency change of the resonant cantilever, which is essentially functioningas a balance. A schematic of the transducer is shown in Fig. 5.23 [526]. Asilicon cantilever of the thickness tSi = 6 µm is covered with silicon dioxidelayers (tOx = 2.4 µm) and a silicon nitride passivation (tNi = 1.1 µm). Thesilicon nitride passivation protects the embedded electrical components. Thecantilever is coated with a polymer layer, which is between 3 and 6 µm thick(tL). The fundamental resonance frequency of the composite beam for smalldeflections is given by [538,539]:

fo =λ2

0

2πL2

√ESiISi + EOxIOx + ENiINi + ELIL

ρmeanF. (5.10)

Here, E denotes the apparent Young’s modulus of each material, I is therespective moment of inertia and ρmean is the average specific mass densityof the composite beam. F denotes the cross-sectional area and L denotes thelength of the cantilever. λ0 is an integration constant and has a value of 1.875for the cantilever geometry under consideration.


The moment-of-inertia contribution of each layer depends on the distanceof the particular layer to the neutral layer of the beam. The neutral layer isa particular plane of the beam, where the internal mechanical stress duringvibration equals zero. When the beam shown in Fig. 5.23 bends upwards, theuppermost 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 nearthe cantilever center, which experiences neither tensile nor compressive stressduring oscillation though it is deformed through the cantilever motion [539].

The average specific mass density (cantilever materials and sensitivelayer), ρmean, and the cantilever specific mass density (only cantilever sand-wich materials), ρcant, are [93]:

ρmean =ρSitSi + ρOxtOx + ρNitNi + ρLtL

tSi + tOx + tNi + tLand

ρcant =ρSitSi + ρOxtOx + ρNitNi

tSi + tOx + tNi. (5.11)

Additional mass on the cantilever leads to a decrease in the resonance fre-quency. Therefore, negative resonance frequency shifts are detected upon ab-sorption of a gaseous analyte in the polymer. These frequency shifts originatefrom the polymer mass density increase ρL as a consequence of the additionalmass of the absorbed gas molecules.

Due to the high spring constant of the cantilever (k ≈ 800 N/m), influencesfrom changes in the elastic properties of the polymer upon gas absorption,the effects of which are more than two orders of magnitude smaller, can beneglected. Under the assumption that the cantilever is covered with a polymerlayer of uniform thickness, the average mass density change of the cantileveras a consequence of the polymer density change (polymer swelling can beneglected in a first order approximation at such low analyte concentrations[69]) is given by:

∂ρmean

∂ρL=

tLtSi + tOx + tNi + tL

=tLh

(5.12)

with h being the total thickness of the cantilever (coating and cantilevermaterials). 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

∂ρL=

∂f0

∂ρmean· ∂ρmean

∂ρL=

12· f0

ρmean· tL

h. (5.13)

Here, G represents the gravimetric sensitivity, i.e., the change in frequencydue to a change in polymer density. The corresponding mass sensitivity Gm =G · VL is determined by taking into account the polymer volume VL.


The origin of the polymer density change ∂ρL is the absorption of theanalyte gas of concentration cA. A cantilever resonance frequency shift indi-cates the corresponding mass change. The sensitivity, S, of the resonant gassensor is defined as [93]:

S =∂f0

∂cA=

∂f0

∂ρL· ∂ρL

∂cA= G · SA . (5.14)

As demonstrated earlier, the resonance frequency change results from thepolymer density change upon analyte absorption. G describes the mechanicalproperties of the cantilever device, whereas SA relates to the polymer/analyteinteractions. The mass increase, ∂mL, of the polymer with volume VL and thecorresponding polymer density increase, ∂ρL, upon an analyte concentrationchange, ∂cA, is given by:

∂ρL =∂mL

VL= Kc · ∂cA . (5.15)

The analyte sensitivity, SA, hence equals the partition coefficient, Kc, (seeChap. 2, 2.10) when mass concentrations (mass/volume) are used. Finally,the sensitivity, S [Hz/(µg/l)], of the cantilever gas sensor can be rewrittenas [93]:

S =12· f0

ρmean· tL

h· Kc . (5.16)

Variations in the material properties of the polymer barely influence the can-tilever resonance frequency in the investigated range, because the cantileveroverall elastic properties are dominated by those of the CMOS materials. Asan example, a 50% inaccuracy of the elastic modulus of 100 MPa at a poly-mer layer thickness of 12 µm results in a resonance frequency variation of only0.8%. 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-independentanalyte sensitivity SA and a thickness-dependent gravimetric sensitivity G(5.14). The total sensitivity of the resonant cantilever gas sensor, S, linearlydepends on the thickness of the applied polymer layer (thin polymeric lay-ers), like it is the case with all other polymer-based gas sensors and has beenconfirmed in a wealth of measurements [63, 70, 93, 95]. Therefore, only onegraph (Fig. 5.36) will be shown to illustrate the dependence of the sensorsignal on the layer thickness for toluene absorbed into PEUT [531].

The sensor signals increase linearly with increasing analyte concentrationas can be seen in Fig. 4.1 (Sect. 4.1) and in Fig. 5.37 [93]. Rather low analyteconcentrations (5–20 ppm n-octane) close to the detection limit were appliedto a cantilever coated with 2.8 µm PEUT.

Due to their linear analyte response characteristics, the slopes of the sen-sor response upon analyte exposure are often recorded and are given for thecase of a PEUT-coated (2.4 µm) cantilever in Table 5.1 [93].


0 500 1000 1500 2000

analyte concentration [ppm]

∆f[H

z]1.6 µm

5.6 µm

7.4 µm

14.7 µmtoluene and PEUT-500

-300

-100

0

Fig. 5.36. Response of cantilevers coated with PEUT at different thickness upontoluene absorption [531]

0 5 10 15 20 25 30 35

polymer: 2.8 µm PEUT

5 10 20

n-octane [ppm]

freq

uenc

y sh

ift [H

z]

time [min]

5

6

7

8

9

0

Fig. 5.37. Low-concentration sensor responses: 5, 10, 20 ppm (0.5, 1, 2 Pa) n-octane were dosed to a cantilever (thermal actuation), which was coated with a2.8-µm-thick PEUT layer [93]

The sensor was actuated by approximately 10 mW heating power. Linearresponses 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, deviationsmay occur as a consequence of the analyte plasticizing the polymer. The long-term 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 oftheir different molecular weights, their different gas-phase saturation vaporpressures, and due to the different partition coefficients of the analytes inPEUT. The response slope of, e.g., toluene is considerably higher than thatof ethanol, since toluene is less volatile (higher partitioning in PEUT), itsmolecular weight is higher, and its saturation vapor pressure is lower.


Table 5.1. Measured sensitivity of a cantilever coated with 2.4 µm PEUT to variousanalytes (thermal actuation, 10mW heating power) [93]

Sensitivity Saturation Vapor Pressure Molecular WeightAnalyte [Hz/ppm] at 301 K [Pa] [m.u.]

n-octane −0.05 2200 114toluene −0.1 4300 92ethyl acetate −0.02 14400 88ethanol −0.01 8700 46

A sensor signal of a magnetically actuated cantilever at high time res-olution shows the immediate response and fast recovery of the sensor as aconsequence of the relatively weak physisorption interactions (Fig. 5.38) [531].The response characteristics and analyte sensitivity do, of course, not dependon the actuation mechanism. The geometric dimensions (length, thickness)and the mechanical properties of both cantilevers are identical. The onlyimportant issue is the actuation-induced heating in the case of the thermalactuation that leads to an exponential signal reduction with increasing can-tilever temperature, whereas heating does not occur in the case of magneticactuation. A possible temperature increase on the cantilever has to be takeninto account in comparing the signals of the different cantilevers.

380810

380820

380830

0 200 400 600

time [s]

freq

uenc

y [H

z]

magnetic actuation

120 ppm toluene

5.8 µm PEUT

Fig. 5.38. Resonance frequency of a cantilever (magnetic actuation) coated with5.6 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-coatedsensor will, e.g., respond to ethanol but also to many other analytes withvarying sensitivity (see Table 5.1). Better discrimination can be achievedwith a system of more than one sensor, a so-called sensor array.

Figure 5.39 shows a micrograph of an integrated cantilever array com-prising four cantilevers, the feedback circuitry, and a multiplexer to sequen-tially address and operate the cantilevers. The resonance frequencies of the


cantilever

multiplexer

feedbackcircuitry

bondingpads

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 to980 ± 10% (see Fig. 5.35) [526]. To illustrate the use of cantilever arrays forquantitative analysis, a binary mixture of n-octane and toluene has beendosed to the sensors. The analytes show different interactions and sorp-tion properties. n-Octane is nonpolar and not polarizable, whereas toluenecontains an aromatic ring with enhanced electron density, which rendersthe molecule polarizable. Measurements with mixtures of varying n-octaneand toluene contents were recorded, some signals of which are displayed inFig. 5.40.

Three cantilevers of the array were coated with 0.6 µm of poly(dimethylsil-oxane), PDMS, 1.5 µm of poly(cyanopropylmethylsiloxane), PCPMS, and1.8 µm of poly(etherurethane), PEUT. One cantilever was left uncoated

-80

-60

-40

-20

0

20

40

freq

uenc

y sh

ift [H

z]

0 20 40 60 80 100time [min]

PDMS PCPMSuncoated PEUT

600 ppm n-octane 700 ppm n-octane

ppm toluenebinary mixture

600

600

300

300

400

400

500

700

500

700

Fig. 5.40. Responses of an array of cantilevers (Fig. 5.39) coated with differ-ent polymers (PDMS, PCPMS, PEUT) upon exposure to mixtures of n-octane(600 ppm, 700 ppm) and toluene (300–700 ppm)


to look into surface adsorption phenomena, which were found to be mar-ginal. 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 orother multicomponent analysis tools that provide quantitative determinationof both components after preceding calibration. For more details on patternrecognition and multicomponent analysis methods, see [12–17].

5.3.4 Comparison of Cantileversto Other Mass-Sensitive Devices

In this section, the sensing performance of the resonant cantilever will be com-pared to that of the most common and established mass-sensitive transduc-ers, the thickness-shear-mode resonator (TSMR) and the Rayleigh surface-acoustic-wave (SAW) device. The sensitivity equations for cantilevers andTSMRs are very similar. The cantilever equation (5.16) can be rewritten byreplacing the cantilever mass density (only cantilever sandwich materials)ρcant for ρmean in order to separate layer and transducer variables [93]:

S = − 12 ( ρcant + ρL

tLtcant

)· tLtcant

· f0Kc . (5.17)

The corresponding equation for the TSMR can be derived from the Sauerbreyequation (4.1) [64]:

S = − 1ρquartz

· tLtquartz

· f0Kc . (5.18)

A slightly different equation could be derived for the Rayleigh SAW providedthat there are no modulus contributions to the sensor response [540,541]. TheTSMR resonates in the thickness shear mode, and the polymer layer is notdeformed through the transducer movement. In the case of the SAW-devices,the polymer layer undergoes deformation, and modulus terms must be ac-counted for as soon as the polymer modulus is beyond a certain limit [65].The cantilever vibration also deforms the polymer, so that modulus contribu-tions have to be included in the equations. Due to the large spring constantof the cantilever materials (very stiff cantilever), however, it is justifiable toneglect the effect of sorption-induced polymer modulus changes on the sen-sor 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, inboth cases, proportional to the fundamental device frequency, the partitioncoefficient and the ratio of polymer layer thickness and transducer thickness.The main difference is, that the TSMR equation (5.18) does not include anyterm related to the mass density of the polymer layer, which is a consequence


of the fact, that the Sauerbrey equation only describes simple mass loadingwithout using a sensitive layer or a sorption matrix. It has additionally to benoted that (5.18) only holds true for acoustically thin layers, for which nophase lag occurs between the moving quartz surface and the outer surface ofthe polymer coating [65, 542]. Otherwise, polymer modulus terms also haveto be taken into account.

The absolute mass resolution of the cantilevers is in the range of a fewpicograms as assessed by many authors for different cantilever geometries andoperation modes (static or dynamic) [96–107, 543–546]. This high absolutemass sensitivity does, however, not necessarily imply an exceptionally highgas sensor sensitivity, since the area coated with the sensitive layer usually isvery 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 resultingoverall mass change upon analyte absorption on the cantilever is much less incomparison to the larger TSMR and SAW devices. The higher absolute masssensitivity of the cantilever is, therefore, counteracted by the small transducerdimensions.

Table 5.2 includes characteristic experimental results for TSMRs andSAWs as taken from [69] and the cantilever data compiled in this study. Al-though the fundamental resonance frequency of the cantilever is more thantwo orders of magnitude lower than that of the other devices, its performanceis comparable to those of the TSMR and the SAW delay line. Using a ratherconservative frequency noise estimate of 0.1 Hz for the cantilever and theTSMR as reported in [69] (in both cases the measured short-term frequencynoise is considerably lower [69]), the limit of detection of the cantilever is2.8 ppm n-octane (operation at 28C ambient temperature), which is closeto 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 achievedwith cantilevers are very similar to those obtained with TSMRs. Some dif-ferences in the transducer characteristics, however, merit further discussion.

Table 5.2. Comparison of mass-sensitive sensors in detecting 100 ppm (10 Pa) ofn-octane. The frequency noise (short-term) was set to 0.1 Hz for the cantilever andthe TSMR [17], and to 2Hz for the SAW device [69] in determining the limits ofdetection (LODs) [93]

f0 ∆fpoly ∆fgas ∆fgas/f0 ∆f/∆T ∆fgas/∆f/∆T LOD [ppm]Device [MHz] [kHz] [Hz] (×106) [Hz/K] [K] S/N = 3

Cantilever 0.4 −35 12.0 30 17 0.7 2.5(28C) (4 µm)TSMR 30 −72 20.6 0.7 −15 1.4 1.5(30C) (0.32 µm)SAW 80 −330 87.9 1 −160 0.6 7(30C) (1.56 µm)


The large fractional frequency shift upon gas exposure as compared tothe fundamental resonance frequency significantly contributes to the goodperformance of the cantilever. A cantilever beam shows an approximately 30times higher fractional frequency shift than TSMRs and SAWs (Table 5.2).The absolute frequency shift values of the cantilever are lower than those ofthe other devices and, therefore, frequency counters with higher precision arerequired for readout. On the other hand it is not necessary to meet the stricthigh-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 thesecond term in (5.17) and (5.18). In the case of the TSMR, the quartz platethickness is 56 µm at 30 MHz, and the coating thickness is 100–300 nm. Thethickness of SAW-coatings is in the range of 200 nm to 1 µm, the thickness ofthe quartz plate (approx. 1 mm), however, does not affect the device charac-teristics, since the penetration depth of the acoustic wave is approximatelyone acoustic wavelength (40 µm at 80 MHz, 7.3 µm at 433 MHz).

TSMR and cantilever show a sensitivity that is proportional to the reso-nance frequency as can be seen in (5.17) and (5.18). However, in the case ofthe cantilever, the inherent geometry dependence (width, length, thickness)provides much more flexibility in optimizing sensor sensitivity. It is, e.g., pos-sible to increase the mass sensitivity by reducing the cantilever thickness. Atthe same time, a high cantilever resonance frequency can be maintained byshortening the cantilever. In contrast, the only tunable parameter of a TSMRis its plate thickness.

Considering the temperature stability, the quartz is still preferable(Table 5.2), whereas the cantilever shows similar temperature dependenceas the SAW device. This is a consequence of temperature-induced varia-tion of material properties of the CMOS composite cantilever beam and thetemperature-induced changes in the relatively thick polymer layer.

A big advantage of the cantilever is its compatibility to microelectronicsand the possibility to co-integrate transducer and electronics on a single chip(miniaturization). Placing all the oscillator circuitry, the amplifiers and theanalog/digital converter in immediate vicinity to the transducer (Figs. 5.30,5.32, 5.39) drastically increases the overall device performance (as has beendemonstrated) and renders the chip less sensitive to external disturbances.Integration also simplifies the packaging because there are no sensitive analogsignals that need to be shielded from external noise sources or electromagneticinterference. Finally, integration allows for using on-chip bus-systems and forco-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 metaloxides such as tin oxide, gallium oxide or indium oxides, which have to beoperated at elevated temperature (250–600C), are frequently used materi-als in gas sensing. These oxides, most of which meanwhile exhibit sufficientlong-term stability, are very sensitive to a multitude of inorganic gases andvolatile organics and drastically change their resistance upon exposure tothose analyte gases [399,431–433,547,548].

The integration of heated structures with CMOS technology is particu-larly challenging, since the metal-oxide operating temperatures of 250C to600C are much higher than the temperature specifications of common in-tegrated circuits (between −40C and 120C). The metal oxides have to bedeposited on miniaturized heatable structures that are thermally well iso-lated from the rest of the chip. Many designs have been denoted “CMOScompatible”, which, however, in most cases means, that CMOS materialshave been used, or the sensor design can be used with a modified CMOSprocess. Only few microhotplate-sensor designs have been fabricated in anindustrial standard CMOS process with subsequent sensor-specific post-processing [428,443,549–553].

In the next sections, the CMOS-based microhotplate transducer and itscomponents as well as the fabrication and post-processing technology will bepresented. Several different microhotplate designs and realizations have beenfabricated. Thereafter, three integrated microhotplate systems, which consti-tute subsequent development steps, will be detailed with regard to circuitryimplementation and electrical and chemical sensor performance.

5.4.1 CMOS Microhotplates

Microhotplates, in most cases, consist of a thermally isolated area such as asuspended membrane featuring a heater, a temperature sensor and contactelectrodes for the sensitive layer (Fig. 5.41).

The membrane, which isolates the heated area from the bulk chip, isformed by the CMOS dielectric layers (Si-oxide and Si-nitride) that exhibitlow thermal conductivity. The membrane is released by etching away thebulk silicon. Depending on the micromachining procedure, it is possible toleave a silicon island underneath the heated area (Fig. 5.41) [551, 553]. Suchan island can serve as a heat spreader and also mechanically stabilizes themembrane. The circuitry is then arranged on the bulk chip, the temperatureof which negligibly changes upon hotplate heating as has been assessed byusing an additional temperature sensor in vicinity to the circuitry.

By realizing such microhotplates, high operation temperatures can bereached at comparably low power consumption (40–100 mW), and the de-vices feature a small thermal time constant on the order of 10 ms so that


dielectric layers(membrane)

n-well (Si-island) polysiliconheater

electrodesp-siliconsubstrate

chip temperaturesensor and

circuitry

thick-film SnO2resistor temperature

sensor

Fig. 5.41. Cross-sectional schematic of a microhotplate featuring the differentcomponents and a droplet of tin dioxide as sensitive layer [551]

temperature modulation techniques with the aim to improve sensor selectiv-ity and sensitivity can be applied [436,464].

The microhotplate design and development was guided by the followingconsiderations [551,553]:

• Membrane Layout: The membrane layout should be as symmetric aspossible to achieve good temperature homogeneity over the membrane areaand, as a consequence, low stress gradients. This includes also thermalstress owing to the mismatch of the thermal expansion coefficients of thelayer materials.

• Temperature Distribution: A homogeneous temperature distributionin the heated area is highly desirable to ensure that all sensing processeson the hotplate take place at the same defined and precisely controlledtemperature.

• Thermal Resistance: A high thermal resistance is needed to achieveminimum power consumption.

• Supply Voltage: The desired operating temperature (250C–350C) hasto be reached with a supply voltage of 5.5 V.

In view of the above guidelines, an octagonal, and, later, a circular designfor the heated area on a square membrane were selected, which are shownin Fig. 5.42 [553, 554]. A considerable part of the heat usually is dissipatedthrough the metal lines to the bulk. The octagonal and circular shape ex-hibit relatively long metal leads in comparison to a square or rectangularheated-area design, especially since the distance between the hot and thecold end of the metal leads can be increased by arranging them along themembrane diagonals. A reduction of the hotplate heated-area diameter from300 µm (Fig. 5.42) to 100 µm as shown in Fig. 5.43 further improves the de-vice thermal resistance from 4.8C/mW (octagonal) or 5.8C/mW (circular)to 10C/mW [555].


electrodes

temperaturesensor

n-well island

dielectricmembrane

heater

100 µm100 µm

(a) (b)I heat I heat

Fig. 5.42. (a) Octagonal and (b) circular hotplate designs with heaters, tempera-ture sensors and electrode structures. Both designs feature a silicon island under-neath the heated area [553,554]

electrodes

temperaturesensor

n-well island

dielectricmembrane

heater

100 µm

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 iden-tical. Thus, a symmetric heat flow through the membrane is created. Theheat is produced by a cloverleaf-shape (octogonal) or ring-shape heater (cir-cular) that extends along the edge of the heated membrane area. The circularheater and heated-area design also perfectly match the shape of the sensitivetin-dioxide droplet: Area that is not covered by the sensitive material is notheated, and, consequently, the heat losses to the ambient air are minimized.A SEM (scanning electron microscope) micrograph of a microhotplate coatedwith a SnO2-droplet is shown in Fig. 5.44 [553]. The grainy structure of thenanocrystalline material is clearly visible.

From the conventional heat conduction equation it is found that the tem-perature gradient within the heated area can be minimized by (a) using high-thermal-conductivity materials in the heated area and by (b) generating the


x300 6.00 kV 100 µm

suspendedmembrane

circularhotplate

tin oxidespot

300 µm

Fig. 5.44. Tin-dioxide-coated hotplate. The metal oxide droplet exclusively cov-ers the circular heated area (left) [555]. The grainy and porous structure of thenanocrystalline material can be seen in the SEM picture (right) [553]

heat at the location with the largest thermal losses. Both issues were takeninto account in realizing the heated-area hotplate design. First, a 5.5-µm-thickn-well silicon island, which effectively distributes the heat owing to the highthermal conductivity of silicon, was fabricated underneath the heated areain the center of the membrane [551,553]. The island also mechanically stabi-lizes the membrane, which results in lower membrane buckling. The secondfeature is a novel type of polysilicon cloverleaf-shape/ring heater includingtwo semicircular heating resistors along the edges of the heated area, whichare connected in parallel. Using this heater type, the heat is generated in thelocations with the largest heat losses via membrane and metal leads to thebulk, which ensures that the temperature in the sensitive area enclosed bythe ring heater is homogeneous [553,555].

The parallel heater configuration moreover decreases the overall heaterresistance to a nominal value of 125 Ω. With such a low resistance it is possibleto provide enough heating power to reach high temperatures (up to 400C)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 ina four-point configuration. It serves as feedback element in the temperaturecontrol loop (see Sects. 5.4.2.1–5.4.2.3) and features a nominal resistance of10 kΩ. An additional reference resistor is needed for the control circuitry [556].

Devices featuring a network of temperature sensors in the heated mem-brane area were fabricated to assess the temperature homogeneity. The tem-perature sensors were also realized as four-point arrangements. The resistanceof each of those temperature sensors can be measured individually. These ad-ditional temperature sensors were placed on the right and left side of theelectrode pair and at the upper right corner of the membrane (see markers in


Fig. 5.45a) [557]. The hotplate does neither exhibit a silicon plate underneathnor a polysilicon or metal heat spreader.

The relative discrepancy of the temperature as measured by the differentsensors (T2–T4) with respect to the sensor in the membrane center T1, whichacts as a reference, is shown in Fig 5.45b as a function of the hotplate temper-ature (T1) [557]. The values named T2 represent, e.g., the relative difference,(T2–T1)/T1, in percent [557]. One would intuitively expect, that T1 showsthe lowest temperature owing to the ring heater scheme, which would lead toa positive relative difference value for all other sensors. However, T2 shows alower temperature than T1 owing to the fact that T2 is in close proximity tothe wide metal line of the heater supply. As a consequence of the large heatflux through the metal line to the bulk silicon, the measured temperature ofT2 is lower.

1

100 µm

2

3

4

(a) (b)

-15

-10

-5

0

5

10

15

20

25

100500 150 200 250 300 350

T2T3T4

∆T/T

1 [%

]

T1 [°C]

Fig. 5.45. (a) Hotplate with temperature sensor network for temperature homo-geneity assessment. (b) Temperature distribution on the microhotplate: Relativediscrepancies [%] between temperature measured in locations 2, 3, 4 and that mea-sured 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 pos-itive signs as expected. The active region of the sensitive material is betweenthe two electrodes, so that T3 represents the temperature in the active sensorregion. Although the hotplate does not exhibit any feature for improving thetemperature homogeneity (silicon island etc.), the temperature gradient atT3 at 300C hotplate temperature is only 0.3C/µm [557].

With a polysilicon plate or a Si-island underneath, the temperature ho-mogeneity is further improved. For an identical device with a Si-island heatspreader, a relative deviation of less than 2% in the heated area was achieved[551].


The electrodes on the membrane that establish contact to the sensitivelayer consist of platinum, which is deposited on top of the CMOS-aluminum-metallization. A 50-nm titanium/tungsten layer sandwiched in between thealuminum and platinum acts as diffusion barrier and adhesion layer for thesubsequently deposited 100-nm-Pt thin film.

For advanced reliability, the temperature sensor was also implemented as aplatinum resistor especially since platinum deposition has to be done anywayto realize the electrodes [555]. A platinum temperature resistor shows lessresistance drift as compared to a polysilicon sensor, the characteristics ofwhich strongly vary over time and, in particular, with increasing exposuretime to high temperatures. Similar considerations would also apply to thepolysilicon heater but the heater resistance drift has no critical impact on thedevice performance since the circuitry controls the temperature according tothe temperature sensor input (see Sects. 5.4.2.1–5.4.2.3), so that a drift inthe 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 fora polysilicon temperature sensor requires major changes in the temperaturesensor readout circuitry [555,558]. An additional burn-in-heater for annealingthe sensitive material was also integrated, which was not used so far. The ba-sic idea was to provide a heater, which can achieve much higher temperaturesthan the heater for normal operation. Since the temperature for material sin-tering on wafer-level is limited to 400C in order not to change the transistorcharacteristics of the integrated electronics (last thermal step of the CMOSprocess is approximately 400C), higher annealing temperatures can only beapplied locally on the membranes by using the extra-heater with an externalpower source. The annealing heater is, if ever, used once, so that only theheater, which is used for operating the hotplate and which is connected tothe circuitry, was further optimized for performance.

electrodes

Pt temperatureresistor

n-well island

dielectricmembrane

heater

100 µm

CMOSmetallization

Ptmetallization

heatercontact

Fig. 5.46. Microhotplate featuring a meander-type Pt-temperature resistor [555]


5.4.1.1 Temperature Sensor Calibration

The various temperature resistors have to be calibrated in an oven by ex-posure to defined temperature steps, since it is well known that there is alarge production spread in polysilicon resistors fabricated in a CMOS process.This production spread is, on the one hand, a consequence of geometric vari-ations of the temperature sensor itself, and, on the other hand, a consequenceof fluctuations in the material properties of the polysilicon resistor. For thetemperature sensor a second-order polynomial was extracted, which includesthe 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 (27C), ∆R is the resistancechange in going from room temperature to the hotplate operating tempera-ture. The current through the temperature sensor as provided by the circuitryis constant. Therefore, the voltage can be replaced for the resistance. U0 isdetermined in the non-heated state of the microhotplate and refers to roomtemperature. By measuring the U0-value, the spread in the geometry or di-mensions of the temperature sensors is accounted for [551].

To determine the two temperature coefficients, a1 and a2, a Pt-temperatureelement is introduced into the same package with the sensor in close proxim-ity. The arrangement is then introduced in an oven and heated to a tempera-ture of up to 325C. The standard Pt-resistor serves as temperature referencein determining the two temperature coefficients of the integrated polysilicontemperature sensor. Extrapolations are not necessary in the target tempera-ture range. The resulting coefficients are then used in the measurements toassess the hotplate temperature. The spread in the coefficients of polysilicontemperature sensors within one wafer lot is in the range of 2% [555]. Thereis no significant drift of U0 during analyte exposure of typically a few hours,but in longterm measurements polysilicon temperature resistors may consid-erably drift, which is the reason for using Pt-temperature resistors on thehotplate for better reliability (Fig. 5.46).

After calibration, all temperature resistors are continuously read out dur-ing the measurements. The heating voltage is measured in a 4-point con-figuration. The power losses in the leads and in the bond wires are thuseliminated.

5.4.1.2 Thermal Microhotplate Modelingand Characterization

A key issue in designing integrated monolithic microsystems is the simulationof the microhotplate-transducer that is connected to the circuitry to ensurefull chip functionality after fabrication. This requires an adequate description


of the microhotplate and an implementation in a language that is applicableto circuitry simulations [557]. Thermal simulations of microhotplates havethree main purposes: (a) gaining information about the thermal character-istics of a given microhotplate structure, such as temperature distribution,thermal resistance and thermal time constant, (b) facilitating its optimizationprocedure, and (c) providing input parameters for the overall microsystemsimulation. The microhotplate simulation procedure includes several steps.First, the microhotplate design has to be converted into a geometry modelfor the finite-element (FEM) simulation with the aim to find a model that isas simple as possible but includes all relevant processes. The steps to arriveat the model are represented in Fig. 5.47, for details, see [557].

Fig. 5.47. Flow diagram for translating a microhotplate design into a geometrymodel for FEM-simulation using FEMLAB [557]

The picture on the bottom left-hand side shows the microhotplateschematic. The microhotplate exhibits a symmetric design so that a simu-lation of one quarter is adequate. Geometrical simplifications are introducedinto the representation of the membrane layout and structure, for which aquasi 2-dimensional membrane model is developed. Afterwards, the mem-brane model is transferred back to a 3-dimensional description of the mem-brane with a constant thickness. All material and thermal properties of thedifferent hotplate components and layers (electrodes, heaters etc.) are in-cluded in the model by then. Finally the membrane structure is combinedwith the supporting silicon cavity and the surrounding air. The result isa geometry model for the FEM-simulation [557]. The input parameters fora lumped-model description of the microhotplate are calculated from thesimulation results. An analog-hardware-description-language (AHDL)-model,


∆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 sys-tem, can be derived from the lumped-model equations [552].

There are no additional fit parameters necessary to set up the model. Themodeling results nicely coincide with experimental findings as can be seen inFig. 5.48, which shows the temperature increase on the hotplate, ∆T , versusinput power for a circular hotplate with silicon island underneath as depictedin Fig. 5.42b [555,557].

Simulations were performed in 10-mW power steps, the results of whichare plotted together with the mean value of the experimental data from threeidentical hotplates. The measured thermal resistance at room temperature is5.8C/mW ±0.2C/mW. The thermal resistance fluctuations mainly resultfrom variations in the etching process. The simulations yield a thermal resis-tance of 5.7C/mW in the heating power range between 0 and 10 mW, whichslightly deviates from the experimental results [557]. For higher temperatures,larger deviations are observed. A general trend is, that the simulated mem-brane temperatures are lower than the measured ones. For an input powerof 60 mW that produces a temperature increase of approximately 300C, thediscrepancy between measured and simulated values is still less than 5%.

To measure the thermal time constant of the microhotplate, the heaterwas driven with rectangular-shape heating pulses at a frequency of 10 Hz. Thepulse amplitude was adjusted to produce a hotplate temperature of 300C.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 andwas measured during membrane cooling (for details see [551]). Measurementswould have been distorted in case of assessing the thermal constant from thetemperature increase of the membrane during heating up, since the heaterresistance changes with temperature, i.e., initially a lot of power goes to the


heater and then falls off (power overshoot but no rectangular step function)[551, 557]. In the case of the temperature decay, only the thermal propertiesinfluence the time constant since heating immediately stops upon switchingoff the power. The modeling results again are in good agreement with themeasurements (discrepancy of 5%), which signifies that the model is suitablefor system-level simulations of microhotplate devices.

The simulation of coated microhotplates is even more difficult owing tothe fact that the material properties of the SnO2 thick-film layer are notwell known, and its deposition is not as reproducible as the industrial CMOSprocess with the subsequent micromachining. The SnO2-droplet has a typicalheight of 25 µm, and it was found that the temperature homogeneity in theheated area is improved by additional heat spreading through the thick metaloxide layer [555,557].

Since the metal oxide coating is covering only the heated area in the mem-brane center (see Fig. 5.44), no additional dissipation paths will be createdfrom the heated area to the bulk silicon, so that no significant changes in thethermal resistance have been found. The thermal time constant, however, con-siderably 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 pre-dicted. The thermal time constant of both, coated and uncoated hotplatealso depends on the quality of the etching process.

5.4.1.3 Microhotplate Heaters: Resistorand Transistor

Most conventional microhotplate heaters published to date include resistiveheating 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 heatersalong 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 theconsumed power is dissipated on the chip, which does not contribute to heat-ing the membrane. Industrial CMOS-processes offer the advantage of usingactive heating elements by, e.g., realizing a MOS-transistor-type heater on themembrane. A transistor approach has been proposed in SOI-technology [560]but has not been realized in conventional CMOS-technology so far. The powertransistor for driving the resistive heater is no more necessary, and all heat isproduced directly on the membrane and contributes to microhotplate heat-ing. Further advantages of a transistor heating-scheme include new heatercontrol modes and the fact that the full supply voltage range can be directlyapplied to the heater. As will be detailed later (Sect. 5.4.2.3), the result-ing novel hotplate type can be implemented in a smart sensor system withcomplex digital control circuitry.


microhotplate

(b) transistor heating

Ucontrol

UDD

heatingtransistor

Rsens (SnO2)

(a) resistive heating

Ucontrol

UDD

Rheat

powertransistor

Rsens (SnO2)

Fig. 5.49. Different microhotplate heating schemes: (a) resistive heater with powertransistor [551] and (b) PMOS transistor heater [559]

A cross-sectional schematic of the transistor microhotplate is shown inFig. 5.50, a micrograph in Fig. 5.51 [559]. Instead of a ring heater, there isa ring transistor, which has three components: source, drain and gate. Thegate 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 evenat those high temperatures is a consequence of the electrical isolation of then-well from the rest of the chip so that there are no leakage currents.

dielectric layers(membrane)

n-well (Si-island)

PMOS ring-transistor heater

electrodes

thick-film SnO2resistor temperature

sensor

p-diffusion

gate

SD

bulk silicon(p-doped)

Fig. 5.50. Cross-sectional schematic of a transistor-heated hotplate: D and S de-note drain and source [559]

The n-well electrons and thermally generated free electrons are confinedto 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 CMOSprocess. The heated area includes an octagonally shaped n-well silicon-island


100 µm

electrodes

Si-island(n-well)

polysilicon heater(opt. annealing)

temperaturesensor

PMOS transistor heater ring

µ

Fig. 5.51. Chip micrograph showing the membrane and the integrated PMOS-heater [559]

of 300 µm base extension underneath the dielectric layers. The n-well is notonly electrically isolated but also serves as a heat spreader. The PMOStransistor with 5 µm gate-length and 720 µm overall gate width (8 sectionsof 90 µm) is integrated in the island [559]. A special ring-shape transistorarrangement was chosen (Fig. 5.51), which leaves enough space to implementresistive temperature sensors in polysilicon. The sensor in the membrane cen-ter is used to determine the membrane temperature. An additional resistiveheater is also integrated on the heated island.

The temperature characteristics of both, resistor type heaters and tran-sistor type heaters are shown in Fig. 5.52. The graph in Fig. 5.52a dis-plays the center membrane temperature versus the applied heating powerfor two slightly different square-shape heated areas on 500 by 500 µm2 mem-branes [555]. The temperature characteristics are almost linear with increas-ing power. The nonlinearity also increases with increasing temperature. Themaximum 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 tem-perature versus the source-gate voltage, Ug, for different constant heatingvoltages, Usd (source-drain voltage) [559]. The heating characteristics are al-most linear at temperatures higher than 100C, which simplifies controllingthe membrane temperature by a rail-to-rail gate voltage. The transistor canalso 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 350Cusing a low-voltage power supply (5 V).


0 1 2 3 4 5 6

Usd = 2 V

Usd = 3.5 V

Usd = 2.5 V

Usd = 3 V

Usd = 4.5 V

Usd = 4 V

Usd = 5 V

Ug [V]-20 0 20 40 60 80 100

membrane 1 membrane 2

power [mW]

(a) (b)

0

50

100

150

200

250

300

350

400

0

100

200

300

400

500

T m

embr

ane

[°C]

T m

embr

ane

[°C

]

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 temperatureversus 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-metal0.8-µm CMOS-process as provided by austriamicrosystems [496]. A series ofpost-CMOS processes were developed to complete the transducer fabrication(Fig. 5.53) [555].

Aluminum is the only metal available in the CMOS process after removingthe CMOS passivation. The problem with Al-electrodes is, that they undergostrong oxidation during the annealing of the sensitive material, which resultsin poor or no electrical contact between the metal electrodes and the sensitivelayer. This problem was resolved by deposition of an additional Pt-layer ontop of the Al-metallization. The platinum ensures a good contact to thesensitive layer over a large temperature range.

For patterning the platinum, a lift-off process on wafer-scale was de-veloped (Fig. 5.53b–d) [555]. After deposition and the lithography of thephotoresist, a 50 nm titanium-tungsten layer was sputtered onto the Al-metallization. The titanium-tungsten layer serves as adhesion layer and actsalso as a diffusion barrier. After deposition of the titanium-tungsten, a100 nm-thick Pt layer is sputtered on top.

Afterwards, the photoresist is removed. Optical inspection showed a goodadhesion of the metal stack to the Al-metallization so that no cracks wereobserved. In a next step, a local silicon nitride passivation is applied andpatterned with the aim to protect all contacts and metal structures with theexception of the measuring electrodes (Fig. 5.53e, f) [555].


Fig. 5.53. CMOS post-processing sequence to fabricate metal-oxide-covered mi-crohotplates: Pt-electrode fabrication (a–d), local passivation (e, f) hotplate for-mation (g), and coating deposition (h) [555]

For the release of the microhotplate membrane, the wafers with the Pt-electrodes were back-side etched in aqueous 6-molar potassium hydroxidesolution (KOH) at 90C using an electrochemical etch stop technique (ECE)[561]. The etching stops at the field-oxide and at the silicon n-well, whichforms 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 methodon the micromachined substrate. One set of method relates to thin-film tech-nology and includes the direct deposition by means of conventional microtech-nological processes such as chemical-vapor deposition (CVD), sputtering orevaporation. A general difficulty is the precise patterning of these materialsat micrometer resolution (see Figs. 5.43, 5.44). Lift-off techniques are ap-plicable for a variety of metal oxides, but impose limits on the depositiontemperature [549,562].


In CVD-processes the microhotplate is heated, which leads to decompo-sition of the gaseous precursor material and locally defined deposition of athin-film layer in the heated area [563]. The hotplate temperature heavily in-fluences the film morphology resulting from CVD processes, which also holdsfor the case of sputtering. Masklessly sputtered films do not have to be pat-terned according to the authors of [564], since the resistance of the nonheatedmaterial is very high, and, therefore, gas-sensitive effects on the conductivitywill only occur in the heated area, i.e., on the microhotplate. Thermolitho-graphic patterning of sol-gel materials constitutes an alternative, which wasreported on recently [565]. Another possibility includes rheotaxial growth andthermal oxidation of tin layers (RGTO): A metallic tin layer is deposited andsubsequentially oxidized. The layer is patterned either by lift-off or by usinga shadow mask [566,567]. The result is a locally defined nanocrystalline thinfilm layer.

The other category of deposition methods on microhoplates relies on theinitial production of a metal-oxide powder. This powder is mixed with wa-ter and/or organic solvents to produce a paste, which is deposited onto thesubstrates. The deposition can either be done by spin-coating in combina-tion with etching or by thermolithographic patterning through membraneheating [568, 569]. There are also methods relying on pulverization depo-sition [570]. All these deposition technologies lead to nanocrystalline metal-oxide layers in the micrometer and submicrometer layer-thickness range. Eventhicker layers (tens of microns) are fabricated by drop-coating, screen-printingor fluidic paste deposition [450,547,571,572], in which cases post-structuringis not necessary.

For the CMOS hotplates it was important to select a metal-oxide depo-sition method, which is compatible with the CMOS-substrate, i.e., does notexceed the back-end temperature of the CMOS process of 400C in ordernot to change the transistor characteristics of the integrated electronics. Aspreviously mentioned, a “burn-in heater” is provided in most hotplate de-signs to achieve higher temperatures (up to 500C) locally on the hotplate ifnecessary.

Since nanocrystalline metal oxide materials with defined grain sizes thathave been deposited from slurries of preprocessed powders in thick-film tech-nology provide high gas sensitivity and comparably high longterm stability,this approach was chosen. The metal-oxide thick film is deposited on theelectrodes by using a drop coating method and a paste with preprocessedmetal-oxide powder exhibiting defined grain sizes in the range of 10 to20 nm [399, 547, 573, 574]. The oxide material is a Pd-doped nanocrystallineSnO2 [574], the principal gas detection mechanism of which has been detailedin Sect. 4.4.3.1.2. The whole chip (including circuitry) is heated in a belt ovenfor 20 min at 400C to anneal the oxide material, which is enough to achievesatisfactory sintering without any reduction of the material gas sensitivity.The high-temperature annealing did not produce any cracks in the membrane


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 passivationand 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 atwafer-level.

5.4.2 Hotplate-Based CMOS Monolithic Microsystems

Within this section, three subsequent development steps or different gener-ations of a monolithic hotplate-based microsystem – starting with a simpleanalog system, proceeding to an already complex analog/digital system andending with a three-hotplate digital array – will be described. Each systemwill be detailed in architecture and design, and its performance in electricaltest measurements (functional tests of the circuitry units) and upon gas ex-posure 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 micro-hotplate gas sensor platform along with circuitry for controlling the tempera-ture and for sensor signal amplification and conditioning [551]. The circuitryalso provides an electronic interface to the outside world. The membrane,which isolates the heated area from the bulk chip, is formed by the CMOSdielectric layers (Si-nitride, Si-oxide). The circuitry is arranged on the bulkchip, the temperature of which negligibly varies upon hotplate heating as willbe shown.

The block diagram in Fig. 5.54 schematically shows the chip architec-ture [551,552]. An embedded analog controller regulates the membrane tem-perature. A polysilicon resistor serves as temperature sensor, and the volt-age drop over the resistor provides the feedback signal for the temperaturecontroller. The bulk chip temperature is monitored by another temperaturesensor in close vicinity to the circuitry (see also Figs. 5.55 and 5.57). A loga-rithmic converter is implemented and is used to perform signal conditioningof the conductometric sensor signal [552]. Figure 5.55 shows a micrograph ofthe chip. The left part of the micrograph exhibits the hotplate membrane,whereas the right part shows the different circuitry components. The sensorsystem was fabricated in a double-poly, double-metal 0.8 µm CMOS processas provided by austriamicrosystems [496] with subsequent micromachiningsteps.

The circuitry can be grouped in three functional units [551,552]: (i) Mem-brane temperature control loop, (ii) bulk-chip temperature measurement, and


log converter

poly-Si heater

poly-Si temp. sensor

SnO2 sensor

membranemembrane

proportionalcontroller

temp. sensormeasurement

circuitry

driving circuitry

measurementcircuitry

controlvoltage

SnO2 resistance

membranetemperature

chiptemperature

Fig. 5.54. Block diagram of the monolithic analog-type hotplate microsystem andits components [551]

500 µm

membrane chip temperature sensor

linear-to-logconverter

proportionalcontroller

Fig. 5.55. Micrograph of the analog monolithic microsystem showing the hotplateon the left side and the circuitry components on the right [551,552]

(iii) sensitive-layer resistance measurement. The chip is supplied by meansof an external current source, which is used for biasing the membrane tem-perature sensor, the reference current of the logarithmic converter and thebulk-chip temperature measurement circuitry.

Temperature ControlThe proportional temperature controller is implemented using an opera-tional amplifier and an internal stabilization capacitor of 8 pF [552,558]. The


membrane temperature is controlled from room temperature up to 350C.The operational amplifier drives a power transistor, which provides the cur-rent to the polysilicon heater. The inputs of the operational amplifier consistof the control voltage and the voltage drop on the polysilicon temperaturesensor, which provides the feedback signal for the temperature controller.The polysilicon temperature sensor is biased with a temperature-independentcurrent [552]. The ageing effects in the polysilicon temperature sensor can-not be compensated by the electronics on chip, the circuitry will, however,allow for precise control of the preset temperature. The steady-state error ofthe temperature controller as measured in the system operation temperaturerange (ambient temperature between −40 and 120C) is less than 1% of themembrane temperature.

Bulk-Chip Temperature MeasurementThe bulk-chip temperature is assessed via a PTAT (proportional to ambi-ent temperature) temperature sensor, which makes use of the base-emittervoltage difference between a pair of diode-connected vertical pnp transistors(parasitic transistors available in the CMOS process, collectors tied to sub-strate) working at different current densities [552].

Metal-Oxide ResistanceThe resistance of the SnO2 sensitive layer can vary over a wide range (upto six decades) and is, therefore, measured using a logarithmic converterin a range from 1 kΩ to 10MΩ (see Fig. 5.58b later in this section). Therange of the logarithmic converter can be changed with the reference current(IREF). The bulk chip temperature is used to compensate for the temperaturedependence of the logarithmic converter. For more details on the circuitry andthe circuitry schematics, see [552,558].

System CharacterizationThe tracking mode and stabilization mode performance of the temperaturecontroller was assessed. The tracking mode results are displayed in Fig. 5.56a[551,552]. At measurement start, the chip was at an ambient temperature of27C, then, the control voltage was increased in steps of 10 mV to achievemembrane heating.

A control voltage of 1.80 V led, e.g., to a membrane temperature ofapproximately 350C. The measured slope of the almost perfectly lineartemperature-versus-control-voltage graph was 0.63C/mV, and the trackingerror due to noise was less than ± 0.3C.

The performance of the temperature controller in the stabilization mode isshown in Fig. 5.56b [551,552]. The ambient temperature was swept from −40

to 120C in steps of 5C, and a constant control voltage of, e.g., 1.77 V wasapplied to the membrane heater, which produced a membrane temperatureof 331C. The controller showed excellent temperature stabilization of themembrane in the investigated temperature interval with less than 1% error.The error due to noise was approximately ± 0.3C.


mem

bran

e te

mpe

ratu

re [°

C]

mem

bran

e te

mpe

ratu

re [°

C]

0

100

200

300

400

1.3 1.5 1.7 1.9control voltage [V]

328329

330

332333

-40 0 40 80 120ambient temperature [°C]

(a) tracking mode (b) stabilization mode

331

Fig. 5.56. Performance of the on-chip temperature controller in (a) the trackingmode, and (b) the stabilization mode [551,552]. A membrane temperature of 331Cwas preset in the stabilization mode

The dynamic behavior of the temperature controller was measured byapplying a fast temperature change on the membrane from 27 to 300C. Thetemperature controller was very stable and did not show any overshoot orringing [552].

The performance of the temperature sensor on the bulk chip in vicin-ity to the circuitry was also assessed. The ambient temperature was sweptfrom −40 to 120C in steps of 5C, and the control voltage was set to 0 V, i.e.,the membrane was not heated at all. A two-point calibration at temperaturesof −20 and 85C was performed. The measured temperature sensitivity wasabout 128 µV/C, and the error due to noise was less than ±1.5C [551].

An important issue concerns the hotplate-induced chip heating, since thecircuitry may not be overheated and its proper functioning is only specifiedand guaranteed in a temperature range from −40 to 120C. Therefore, thetemperature variation off-membrane on the bulk chip with respect to ambi-ent temperature upon heating the membrane was measured as displayed inFig. 5.57 [551]. The chip was kept at 27C (ambient) in a ceramic dual-in-line(DIL) package, and the control voltage was increased in steps of 10 mV, whicheffectuated a heating of the membrane from 27 to 360C. The maximum tem-perature increase on the chip was less than 6C over ambient temperature ata membrane temperature of 350C [551]. The chip temperature increase canbe further reduced by an optimized package.

Gas TestsGas test measurements were performed in a gas manifold. Vapors were gen-erated from gas bottles with calibrated target gas concentrations (e.g., CO)and then diluted as desired using computer-driven mass-flow controllers andsynthetic air (oxygen/nitrogen mixture without humidity) as carrier gas. Typ-ical experiments consisted of alternating exposures to pure synthetic air and


0

1

2

3

4

5

6

0 100 200 300 400

membrane temperature [°C]

(TC

HIP

– T

AM

BIE

NT)

[°C

]

TCHIP

package

hotplate

Fig. 5.57. Chip heating: Difference between chip temperature and ambient tem-perature as a function of the membrane temperature (ceramic DIL package) [551]

contaminated air. Exposure times of 30 minutes were followed by 30 minutespurging the chamber with synthetic air.

In a first test series, the sensor resistance was read out by a multi-meterand was plotted on a logarithmic scale. Figure 5.58a shows the sensor re-sponse (R/R0: ratio of resistance upon gas exposure to that in pure air)upon exposure to different CO concentrations at varying hotplate operationtemperature and different relative humidity levels [551]. A nonlinear responsewith increasing CO concentration is observed as it is well established in lit-erature for this type of semiconducting metal oxide (see, e.g. [399, 547]). Itis also evident from Fig. 5.58a, that the sensor response is depending on therelative humidity content of the test gas atmosphere: Higher humidity levels

(a) (b)

0 20 40 60 80 1001

10

CO concentration [ppm]

carbon monoxide (CO)

0%, 250°C0%, 300°C0%, 350°C

50%, 250°C50%, 300°C50%, 350°C

R/R

0

0.06

0.065

0.07

0.075

0.08

0 50 100 150 200

Ulo

g[V

]

time [min]

5 ppm

10 ppm

15 ppmCO (40% r.h., Tmem: 290°C)

Fig. 5.58. (a) Sensor response (R/R0: ratio of resistance upon gas exposure tothat in pure air) upon exposure to different CO concentrations at varying hotplateoperation temperature and different relative humidity levels. (b) Raw sensor signalscoming from the logarithmic converter upon CO-dosing at 40% relative humidityand 290C operation temperature [551]


generally lead to lower tin oxide resistance, i.e., increase the height of the sen-sor signal in Fig. 5.58a [551]. This is a consequence of the increased number ofCO reaction sites at the surface of the sensitive material owing to the forma-tion of hydroxy1 groups by adsorbed water [547]. It is also established thatthe optimum operation temperature for Pd-doped SnO2-thick-film sensors isin 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 plottedfor three different CO concentrations measured at 40% relative humidityand 290C operation temperature. The signals almost linearly increase withincreasing concentration as is to be expected after logarithmic conversion.From the conducted measurements (Figs. 5.58a,b) it can be concluded, thatconcentrations of even less than 1 ppm of CO are detectable with monolithicsensor systems.

5.4.2.2 Analog/Digital Hotplate Microsystem

A simplified schematic of the analog/digital chip architecture is shown inFig. 5.59 [553, 554]. The microhotplate and its components, which were dis-cussed in the previous paragraph, are represented as a block. The circuitryincludes the same three major components as in the analog system (Sect.5.4.2.1): (a) The metal oxide resistance readout, (b) the microhotplate tem-perature control loop, and (c) the temperature readout of the bulk chip car-rying the electronics [554].

linear-to-logconverter & buffer

poly-Si heater

temp. sensor

SnO2 sensor

circular membranecircular membrane

buffer & programmableoffset comp.

anti-aliasing filter & A/D converter

D/A converter &Buffer

programmableoffset comp.

temp. sensormeasurement

circuitry & buffer


programmableoffset comp.


square root circuitry &driving circuitry

digitalcontroller

&digital

interface

I2Cbus

Fig. 5.59. Detailed block schematic of the analog/digital hotplate microsystemincluding all circuitry units [553]


The resistance of the tin-dioxide layer is, again, read out by a logarith-mic converter. The temperature controller is now a digital programmablePID (Proportional-Integral-Differential) controller, which precisely controlsthe membrane temperature via the analog heater driving circuitry. The A/Dand D/A converters that connect the analog and digital part of the cir-cuitry are also shown in the block diagram. All measured data are read outvia a standard serial inter-integrated-circuit (I2C) bus interface [575]. Thedigital interface also allows for setting the membrane temperature and thecontroller parameters. Each major circuitry unit and its operation principlewill be briefly abstracted. The detailed circuitry implementation is publishedelsewhere [557,558].

Metal-Oxide Resistance MeasurementFor the analog logarithmic converter, the relation between the differentialoutput voltage, US, and the resistance of the sensing layer, RS, can be ex-pressed as

US = −kTC

e· (lnRS + c0) (5.20)

where k denotes the Boltzmann constant, TC the absolute chip temperature,e the elementary charge, and c0 a characteristic parameter in the circuitdesign. The negative sign is a consequence of the polarity of the differentialoutput voltage: Lowering the sensor resistance will lead to an increased US.

Hotplate Temperature MeasurementThe polysilicon temperature sensor in the center of the microhotplate is con-nected to a constant-current source. The voltage drop across the temperaturesensor, UT, is monitored. The corresponding voltage drop, U0, at defined am-bient temperature, T0, is initially assessed in calibration measurements. Sincethe current through the temperature sensor is constant, the following equa-tion holds [553]:

UT − U0

U0=

RT − R0

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 calibrationprocedure, 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 willbe adjusted by the digital controller so that UT corresponds to a preset ref-erence voltage, UR. Once the calibration has been done, reversing (5.22) andreplacing UR for UT in (5.21) yields the value of UR to produce a desired


membrane temperature, T . The controller additionally takes advantage of ananalog square-root circuitry, which linearizes the correlation between controlvoltage and dissipated power [556]. This linearization facilitates the calcula-tion of optimum parameters for the temperature controller.

Chip Temperature MeasurementThe bulk-chip temperature sensor is not realized as a resistor. The base-emitter voltage difference between a pair of diode-connected vertical pnp-transistors (available in the CMOS process as parasitic transistors) is relatedto the chip temperature. The relation between the absolute chip temperature,TC, and the differential voltage, UC, is given by:

UC = c1kTC

e+ Uoff . (5.23)

The factor c1 depends on the transistor design and its biasing current, andUoff is an offset voltage. Both are constants and can be assessed during atwo-point calibration procedure. The chip temperature sensor does not onlymeasure the changes in the ambient temperature, it also can be used tocompensate for the temperature dependence of the output of the on-chiplogarithmic converter according to (5.20).

The analog output signals of the logarithmic converter, US, the microhot-plate, UT, and the bulk-chip temperature sensor, UC, are converted into thedigital domain, and then can be read out via the digital I2C interface. Theinput 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 locatedin the left part of the chip with enough free space to accommodate additionalhotplates. The analog circuitry as well as the A/D and D/A converters areclearly separated and shielded from the digital circuitry for noise reasons[556]. The bulk-chip temperature sensor is located close to the analog circuitryin the center of the chip. The distance between microhotplate and circuitryis comparatively large owing to packaging requirements, as will be explainedin the packaging section (Sect. 5.6).

System CharacterizationThe tracking mode performance of the digital temperature controller is shownin Fig. 5.61 [553, 556]. The microhotplate temperature is plotted versus thedigital code of the reference voltage. The 10-bit input range from 0 to 1023corresponds to a microhotplate temperature range between 170C and 310C,so that one digit represents 0.1C. The tracking error due to noise is less than±2C [556]. The curve is almost linear with the slight nonlinearity being in-cluded in the second-order coefficient of the polysilicon temperature sensor(see (5.22)). The controllable temperature range is variable by adjusting theslope 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 pro-grammable. Such a device with on-chip controller offers several advantages in


500 µm

hotplate analog circuitry,A/D & D/A converters

digital circuitry:controller, interface

temperaturesensor

chip size: 6.7 x 4.8 mm2

Fig. 5.60. Micrograph of the analog/digital microsystem chip showing the mi-crohotplate (left), the analog circuitry (center), and the digital unit with serialinterface (right) [553,556]

150

200

250

300

350

4002000 600 800 1000

TM

[°C

]

UR [digital code]

Fig. 5.61. Tracking mode performance of the digital (PID) controller. One digitalunit represents 0.1C [556]

comparison to conventional microhotplate-based sensors. The approximatelylinear relationship between digital input code and microhotplate temperaturedirectly translates a sinusoidal input variation into a sinusoidal temperaturevariation. This does not hold for a conventional microhotplate with an uncon-trolled resistive heating element upon applying a sinusoidal heating-voltageor -current variation.


Gas-flow fluctuations or temperature changes owing to chemical reactionsalso contribute to the overall heat budget of uncontrolled microhotplates.These effects lead to a temperature variation even though a constant heatingvoltage or heating power is applied. These temperature variations inevitablyproduce changes in the sensor signal, since the metal-oxide resistance sig-nificantly changes with temperature (compare Figs. 5.58a, 5.63). In an in-tegrated microsystem, however, temperature fluctuations are immediatelycounteracted by the on-chip control circuitry, which keeps a preset temper-ature constant regardless of the power consumption. The resistance changesin the sensitive layer upon exposure to the analyte can be distinguished fromthose resulting from ambient temperature changes through the additional on-chip temperature sensor, and by monitoring the power consumption of thehotplate heater [553].

An important issue of monolithic sensor systems with integrated micro-hotplates is the overall chip heating. The temperature increase on the chip ata microhotplate temperature of, e.g., 300C was measured with the on-chiptemperature sensor and amounts to 3C, which corresponds to 1% of thehotplate temperature [553]. Such slight temperature increases, which can befurther minimized by applying suitable packaging methods, such as the use ofdie-attach materials with high thermal conductivity, will definitely not affectthe on-chip circuitry.

Gas TestsA series of chemical measurements was performed in order to test the sensorperformance. The microhotplate was heated to a defined temperature, andthe sensor signal upon carbon-monoxide exposure was recorded. The raw data(digital numbers in the 10-bit output range) were filtered with a moving-average filter averaging over 100 data points. Figure 5.62 shows a typicalsensor response [553, 554]. The sensor signal, S, is expressed in digital unitsafter the conversion of US (see 5.20) by the on-chip A/D converter. Thesensor signal output range from 200 to 280 digital units corresponds to aresistance range of 250 to 50 kΩ. As pointed out earlier, a higher sensor signalreading represents a lower resistance of the sensitive layer. Consequently, theresistance drop upon the presence of CO, a reducing gas, leads to a highersensor reading. Each analyte exposure step is 15 min at constant relativehumidity of 40% (23C humidifier temperature) and 30C chip or ambienttemperature. The microhotplate operating temperature was 275C. Low CO-concentrations of 1, 3 and 5 ppm (partial pressure 0.1–0.5 Pa CO) were dosedto the sensor. As can be seen in Fig. 5.62, a concentration of 1 ppm CO isclearly detectable. The noise in the sensor signal is ±0.5 digits [553]. Thiscorresponds to a 1-digit quantization noise of the A/D converter. At lowconcentration levels, the noise is equivalent to a CO-concentration variationof ±0.1 ppm. This value serves as input for the determination of the limitof detection, which was assessed to be to 0.2 ppm. The gas concentrationresolution also amounts to ±0.2 ppm [553].


200

225

250

275

100500 150 200

S [d

igita

l cod

e]

time [min]

CO 5 ppm 5 ppm

3 ppm3 ppm

1 ppm 1 ppm

Fig. 5.62. Low-concentration signals: 1–5 pm (0.1–0.5 Pa) CO detected with Pd-doped (0.2%) tin dioxide at 275C [553]

The sensor baseline, Sair, represents the sensor reading in humidified syn-thetic air without any analyte present. The difference between this baselinevalue before the analyte-dosing onset and the sensor signal, S, before theanalyte-dosing end, was used to determine the sensor response at a givenconcentration. The full digital output range of S from 0 to 1023 covers sensorresistances 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 expressedas [553]:

∆S = S − Sair = −(cs · lnR − cs · lnRair) = cs · ln Rair

R. (5.24)

The proportional constant, cs, includes the conversion relation of the A/Dconverter and the factor kTC/e of 5.20. Since the measurement chamber istemperature-stabilized, the chip temperature, TC, is constant. Rair is the re-sistance of the sensitive layer in synthetic air, and R is the resistance uponanalyte exposure. So (5.24) correlates ∆S to the ratio Rair /R, which is com-monly plotted as sensor signal.

The sensor responses at defined relative humidity of 40% (23C humidifiertemperature) are displayed in Fig. 5.63 for various microhotplate tempera-tures [553]. ∆S upon exposure to different CO concentrations increases withdecreasing microhotplate temperature. The highest sensitivity was achievedfor the lowest operation temperature of 225C. The sensor signal is stronglytemperature-dependent. The temperature dependence of ∆S in the temper-ature range between 250C and 300C is 0.15 digits per C at a CO con-centration of 5 ppm [553]. A temperature change of 1C at 5 ppm CO anda microhotplate temperature of 275C thus produces the same signal as aCO concentration change of 0.04 ppm. As already mentioned in the previoussection on electrical system parameters, the resolution of the temperature


1050 15

∆S 225°C∆S 250°C∆S 275°C∆S 300°C

∆S [d

igita

l cod

e]

CO concentration [ppm]

0

20

40

60

80

100

120

Fig. 5.63. Sensor responses, ∆S, upon different CO concentrations in dependenceof the hotplate temperature (constant relative humidity of 40% at 23C humidifiertemperature) [553]

controller was estimated to be ±2C [556]. These 2C correspond to a “CO-concentration uncertainty” of ±0.1 ppm, which is less than the resolution inthe concentration measurements. Consequently, the temperature uncertaintyin the controller does not significantly affect the sensor limit of detection. Itis evident from these considerations that a precise temperature control is ab-solutely indispensable to advance into low-concentration and threshold-levelmeasurements.

5.4.2.3 Digital Hotplate Array Microsystem

The microsystem features three microhotplates that are monolithically inte-grated with digital temperature controllers, readout and interface circuitry[576]. The microhotplates are heated by means of MOS-transistors (see alsoSect. 5.4.1.3) and are covered with nanocrystalline SnO2-based coatings assensitive layers. Full advantage is taken of the features offered by applyingCMOS-technology. All sensor values can be set and read out via the digitalinterface, which drastically reduces the packaging efforts, since the numberof bond wires is the same as for a single microhotplate.

The block diagram in Fig. 5.64 depicts the system architecture, empha-sizing its strong modularity [576]. Three digital PID (proportional-integral-derivative) controllers provide independent temperature regulation for eachhotplate. The MOS heating transistor is driven in pulse-density modulationby a first-order Sigma-Delta modulator [576]. Analog circuitry is required


MU

X

readouthotplate

multipliervalue

temp.controller

timervalue

I2Cinterface

controlunit

MU

X

Σ∆

3 microhotplates

Fig. 5.64. Block diagram showing the architecture of the digital hotplate arraychip [576]

only for the readout of the metal oxide resistors and the temperature sen-sors. A single input-conversion stage and multiplier are accessed from thethree controllers in time-sharing. The PID parameters, the target temper-atures, and the operation timing can be all programmed by the user via astandard serial interface [577], which also enables digital readout of the sensorvalues.

Figure 5.65 shows the fabricated chip featuring a tri-sectional floor planwith digital circuitry on the right side including the temperature controllerand the interface, analog circuitry in the center and the three micro-hotplateson the left-hand side [576]. Along the left edge, three different coatings areshown: Pure SnO2, which exhibits high partial sensitivity to NO2, SnO2 with0.2% palladium, which is used to monitor carbon monoxide, and SnO2 with3% palladium, which can be used to detect hydrocarbons such as methane.The highly Pd-doped coating appears gray as a consequence of the metalparticles. An array including those three coatings is suitable for many targetapplications.

Microsystem CharacterizationClosed-loop temperature regulation was tested in the tracking-mode: Thetemperature was sinusoidally varied on one of the hotplates (HP3), while theother two hotplates (HP1, HP2) were kept at constant temperature. The testwas performed for each of the hotplates with the remaining two hotplateskept at constant temperature (Fig. 5.66) [577].


500 µm

chip size: 5.5 x 4.5 mm2microhotplates

digital circuitry: temperature controller & interfaceA/D & D/A converters

pure SnO2

SnO2, 0.2% Pd

SnO2, 3% Pd

Fig. 5.65. Micrograph of the chip with microhotplate array and circuitry [576]

time [min]0 10 20

tem

pera

ture

[°C

]

200

250

300

350

HP 3

HP 1

HP 2

temperature modulation

Fig. 5.66. Sinusoidal temperature variation of one selected microhotplate (HP3).The other two hotplates (HP1, HP2) are kept at constant temperatures of 280Cand 300C [577]

The tracking error due to noise was ±1C. In the temperature rangeof interest, no thermal crosstalk is noticeable, as it is to be expected usingindividual temperature control loops for each hotplate. The results also showgood linearity, with a differential non-linearity less than 1/2 least significantbit (LSB) and an integral non-linearity less than 3 LSB [577].


Chemical MeasurementsFirst chemical test measurements have been conducted with the array chip.Fig. 5.67 shows the results that have been simultaneously obtained from threemicrohotplates coated with different materials at operation temperatures of280C and 300C in dry air [577]. Hotplate HP1 is covered with a Pd-dopedSnO2 layer (0.2 wt% Pd), which is optimized for CO-detection, whereas thesensitive layer on hotplate HP3 contains 3 wt% Pd, which renders this mate-rial more responsive to CH4. The material on hotplate HP2 is undoped puretin dioxide, which is sensitive to NO2.

900

800

700

600

500

400

3001.50.50 2.0 2.5 3.0 3.5

HP2: undoped 280°C

HP1: 0.2% Pd 280°C

HP3: 3% Pd 330°C

CH4[ppm]

500 15001000 15105

CO [ppm]

1000

20

25

30

35

40

45

50

sens

or r

esis

tanc

e [k

Ω]

sens

or r

esis

tanc

e [k

Ω]

time [h]

Fig. 5.67. Sensor response of three microhotplates with different sensitive layersupon 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 dopedmetal oxide (3% Pd) responds to both gases, whereas the undoped tin dioxidematerial shows hardly any signal upon CO and CH4 exposure

As is evident from Fig. 5.67, the responses of the three materials to the testgases CO and CH4 are very different. Hotplate HP1 shows larger responsesto 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 andquantification of gases, which is very important in the presence of interfer-ants or in analyzing mixtures. An array of microhotplates with individuallycontrolled temperatures, the hotplates of which are covered with differentsensitive materials, drastically increases the overall information that can beextracted from such sensors and, then, can be used as input for suitable datadeconvolution 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 –typescan provide more information than using arrays of the same transducer typeand varying the sensitive layer only. The single-type array, an example ofwhich has been shown in Sect. 5.4.2.3, has become particularly popular, asis documented in a wealth of publications [12–17]. Simultaneous use of vari-ous transducer principles is more complex (realization of different transducerand circuitry components, use of different data evaluation schemes), but,nevertheless, has proven to be beneficial in several applications [578–588]. Inthis introductory section also some non-CMOS but semiconductor-based sys-tems will be briefly sketched to illustrate the power of the multi-transducerapproach.

The range of concentrations to be monitored can be extended by, e.g., ap-plying different electrochemical sensor types: An integrated hydrogen systemconsisting of aluminum-gate FETs for low concentrations, a hydrogen sensingchemoresistor made of palladium/nickel for high concentrations and the nec-essary circuitry has been reported in [578,579]. Different transducers can alsoprovide distinct and complementary information on the analyte or analytemixtures in complex samples to be monitored. A sensor system including sev-eral ISFETs and an amperometric free-chlorine sensor for operation in waterhas been reported in [580].

Disposable electrochemical multisensor systems for fast blood analysisare marketed by, e.g., I-STAT [581]. Sodium, potassium, chloride, ionizedcalcium, pH and carbon dioxide are measured by ion-selective-electrode po-tentiometry (see Sect. 4.4.2.1). Concentrations are calculated from the mea-sured potential through the Nernst equation (2.15). Urea is first hydrolyzedto ammonium ions in a reaction catalyzed by the enzyme urease. The am-monium ions are also monitored by means of an ion-selective electrode. Glu-cose is measured amperometrically (see Sect. 4.4.1). Oxidation of glucose,catalyzed by the enzyme glucose oxidase, produces hydrogen peroxide. Theliberated hydrogen peroxide is oxidized at an electrode to produce an electriccurrent, the intensity of which is proportional to the glucose concentration.Oxygen is also measured amperometrically. The oxygen sensor is similar toa conventional Clark-electrode. Oxygen permeates through a gas permeablemembrane from the blood sample into an internal electrolyte solution whereit is reduced at the cathode. The oxygen reduction current is proportionalto the dissolved oxygen concentration. Hematocrit is determined conducto-metrically (see Sect. 4.4.3). The measured conductivity, after correction forelectrolyte concentration, is related to the hematocrit.

A significant benefit of using CMOS-technology for devising chemicalsensors or sensor systems is the possibility to co-integrate several differenttransducers along with all necessary driving circuitry on a single chip. Ad-ditional components that can be integrated include signal-conditioning cir-cuitry (amplifiers, references), multiplexers to reduce the number of output


pins, analog/digital and digital/analog converters, chip memory (calibrationvalues) or other smart features, an intra-chip communication or bus systemand a serial interface to communicate with off-chip microcontrollers or instru-ments. In the following part of Sect. 5.5, two prototype monolithic CMOSmultisensor systems will be presented, (i) a multiparameter biochemical sen-sor [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 ofions, 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-effecttransistors (ISFETs: 1–6 in Fig. 5.68), one oxygen sensor (7 in Fig. 5.68) andone conductometric sensor (8a,b in Fig. 5.68), all of which can be operated inparallel [582]. An Ag/AgCl reference electrode is also integrated on the CMOSchip to get rid of external references. The eight sensors can continuouslymonitor ions, dissolved gases and biomolecules via enzymatic reactions thatproduce charged particles. A flow channel (polyimide) restricts the liquidphase access to the sensor area.

The six ISFETs allow for direct contact of the electrolyte with the gateoxide. The gate oxide itself is either pH-sensitive (see ISFET Sect. 4.4.2.2.1),

1 2 3 5

4

6 7

heater

multiplexer and counterEPROM

potentiostaticcircuit driver

conductometriccircuit

ISFETs

8b 8a8a

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 con-ductometric sensor (8a,b). The on-chip circuitry includes an EPROM, a multiplexerand counter, a driver unit, a conductometric and potentiostatic circuit and a heater.Reprinted with permission from [582]


or the ISFET can be used as a “Severinghaus”-type pH-FET to measure dis-solved carbon dioxide (detection of carbon dioxide via dissolution in water,formation of “carbonic acid” and monitoring of the pH-change). The gateoxide can also be covered with different ion-selective membranes to achievesensitivity to a range of target ions such as potassium. All six ISFETs or onlya subset can be used. The idea was to make a standard chip to reduce man-ufacturing costs and then modify the chip with selective coatings accordingto user needs.

The integrated amperometric sensor (see Sect. 4.4.1) can be used as aClark -type oxygen sensor, which is based on a two-step-reduction of gaseousoxygen 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 isapplied to the electrodes, and the current, which depends on the solutioncomposition (concentration of charged particles or ions) is recorded.

The full system is produced in a 1.2-µm single-metal, single-poly CMOSprocess, and the chip size is 4.11 by 6.25 mm2 [582]. The chip is operatedat 5 V and hosts all driving circuitry of the sensors such as ISFET bufferamplifiers, a potentiostatic setup for the amperometric sensor, and the cir-cuitry necessary to perform a four-point conductometric measurement onchip. In addition, the chip exhibits a temperature control unit to keep thesystem temperature at a preset value (physiological conditions). This tem-perature control unit includes a temperature sensor (parasitic vertical pnpbipolar transistor) and a NMOS transistor heater. A single-bit EPROM (elec-trically programmable read-only memory) was implemented on chip to makesure that the chip is used only once and then is disposed, which is a crucialfeature in medical applications. Additional on-chip electronics include unitsto control the chip (multiplexer, demultiplexer, 4-bit Gray counter and de-coder) and units to provide the biasing and the communication to off-chipinstrumentation. Due to the high level of on-chip integration, only 5 externalconnections are needed: Two for power supply, two for bi-directional commu-nication and one for a clock signal [582].

First tests including amperometric oxygen measurements, the assessmentof potassium concentrations with ISFETs (by directly connecting the ISFETbuffer to a plotter), and conductometric measurements with a buffer solutionhave been performed [582].

5.5.2 CMOS Gas-Phase Multisensor System

The CMOS single-chip chemical microsensor system combines three differenttransducers, a mass-sensitive cantilever, a capacitive sensor, and a calori-metric sensor, all of which rely on polymeric coatings as sensitive layers todetect airborne volatile organic compounds (VOCs) [583,587,588]. The threetransducers respond to fundamentally different analyte molecule properties


E1

E2

∆C ∝ ∆cgas ∆f ∝ ∆cgas Utherm ∝ ∆cgas / ∆t

E2

E1

analytepolymer

cold junctionshot junctions

(a) capacitive sensor (b) mass-sensitive sensor (c) calorimetric sensor

Fig. 5.69. Principles of the three different types of transducers: (a) microcapacitorsensitive to changes in dielectric properties, (b) resonant cantilever sensitive tomass changes, and (c) microcalorimeter measuring the absorption or desorptionheat upon interaction of organic volatiles with the polymer [583]

(Fig. 5.69 a-c). One of them responds to the mass of sorbed molecules, an-other responds to the heat of absorption, and the third responds to the dielec-tric properties of the absorbates. The monolithic system such simultaneouslyprovides three different (“orthogonal”) sensor responses, which are used toclassify 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 signal-conditioning circuitry (sensor front ends), analog-to-digital conversion units,sensor control and power management units, and a digital interface. A pho-tograph of the microsystem chip is displayed in Fig. 5.71. The overall chipsize is 7 by 7 mm2.

The capacitive sensor as described previously in Sect. 5.1 is integratedwith a fully differential second-order Sigma-Delta-modulator and a counterto decimate the output bit stream [587,588].

The micromachined cantilever, the operation principle of which has beenalready described in Sect. 5.3.1.1 (thermal actuation), is 150 µm long andconsists of silicon as well as vapor-deposited and thermal oxide. The cantileveracts 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 Seebeck-effect with 256 polysilicon/aluminum thermocouples connected in series (500


sensors

sensor control& power

management

sens

or fr

ont e

nds

A/D

con

vers

ion

digi

tal b

us in

terfa

ce

mass-sensitive

calorimetric

capacitive

temperature

∆f

∆U

∆C

T

Fig. 5.70. Schematic of the microsystem architecture with sensors (left), drivingand signal-conditioning circuitry (sensor front ends), analog-to-digital conversionunits, sensor control and power management units (bottom), and the digital inter-face (right) [583]

referencecapacitor

sensingcapacitor resonant

cantilever

calorimetricsensor andreference

temperaturesensor

digitalinterface

decimationfilters

chopperamplifier

Σ∆ temperaturesensor

A/D convertercalorimeter

Σ∆ capacitivesensor

Fig. 5.71. Micrograph of the gas sensor system chip (size: 7 by 7 mm2). Thedifferent 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 Sigma-Delta analog/digital converter and a decimation filter [587,588].

The chip features all the sensor-specific driving circuitry and signal-conditioning circuitry (sensor frontends), which has been described in thecontext of the different transducers [501, 588]. The system chip additionallyincludes a temperature sensor since volatile absorption in polymers is stronglytemperature-dependent (a 10C temperature increase reduces the sensor sig-


nal by 50%). The temperature sensor exhibits an accuracy of 0.1C at oper-ation temperatures between −40 and +80C (see Sect. 5.5.2.2).

The analog/digital conversion is done on chip, which allows for achievinga favorable signal-to-noise ratio, since noisy connections are avoided, and arobust digital signal is generated on chip and then transmitted to an off-chip data port via an I2C serial interface [575]. The I2C bus interface (seeSect. 5.5.2.2) offers the additional advantage of having only very few signallines (essentially two) for bi-directional communication and also allows foroperating 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. Thesensors are located in the center of a metal frame, which is used to apply aflip-chip packaging technique [589] (see Sect. 5.6.3).

5.5.2.2 Multisystem Circuitry Components, Designand Fabrication

The three transducers of the multisensor chip and the specific circuitry com-ponents (sensor frontends) that have been developed to operate the trans-ducers and to read out and precondition their small signals have been alreadydescribed 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 circuitryunits 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 ofbonding pads, to control the different units on chip, and to facilitate eithersensor readout by means of a microcontroller or to enable direct displayingof the sensor data and results. Reference voltages are required for the A/D-converters and are used to bias the sensors. Finally, a temperature sensor hasto be included in order to deal with the strong temperature dependence ofthe gas absorption process in the polymeric sensitive layers.

Serial interfaceA schematic of the I2C-interface and controller architecture that is im-plemented on the multisensor chip is shown in Fig. 5.72. Interface andcontroller have been designed and simulated using very-high-speed-integrated-circuits hardware description language (VHDL). The I2C serial bus interface[575] offers some advantages in comparison to other serial interfaces like thecontroller-area-network (CAN)-bus (mostly used in the automotive indus-try) [590], the serial-peripheral-interface (SPI)-bus by Motorola [591], or thestandardized IEEE (institute of electrical and electronic engineers) 1451 businterface [592]:

• Simple, serial protocol: The I2C-protocol clearly defines the most im-portant features (addressing, word length, master/slave communication,start/stop-condition) without adding expensive overhead that requireslarge on-chip memory.


bit2

SDA SCL

I2C slave

registerbank

sensor 1register 0register 1

register x

controllersensor 1

commanddata

delayintervaldata ready sensor 1

power down

commandout

commandin

Addr. SDA SCL

command data

A/D register sensor 1

I2C master

SDASCL

addresscodingpads

bit0bit1

bit3

I2C bus

sensorparameters

data readysensor 1...3

datasensor 1...3

power downsensor 1...3

Fig. 5.72. Schematic of the I2C bus interface and the digital controller. The sensorcontroller, the register bank and the digital output register are replicated for eachtransducer, 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 anddata transmission.

• Bus-enabled: Up to 127 microsystems or multisensor chips can be con-nected on the same bus. A simple arbitration algorithm solves the problemof data collision. The I2C-bus offers two possible modes of operation [575]:• Single-master: An external unit polls the different sensor chips and col-

lects 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. Thedata-recording unit only initializes the multisensor chips and sets their timingparameters. The multisensor chips then send their data in regular intervalswithout being polled by the data-recording unit. This strategy reduces thetraffic on the bus and provides more flexibility for future extensions, such asthe introduction of threshold-limited values, i.e., a sensor triggers the system,as soon as its signal exceeds a certain threshold value.

Digital ControllerThe digital controller (Fig. 5.72) interprets the commands coming from thedata-recording unit, stores the sensor parameters and manages the access ofthe single sensors to the I2C-bus. Each sensor on the chip can be individuallyaddressed via the serial interface. A minimal set of controller commands was


implemented: “power on/off”, “read data”, and “set/read parameters”. Fordetails, see [588].

On-Chip Voltage and Current ReferencesUsing an array of monolithic multisensor systems, it is neither convenientnor cost-effective to supply each chip with a number of external precisionreference voltages or currents. Therefore, all reference voltages and currentshave to be generated on chip. The only stable and reliable reference that isavailable in a standard CMOS processes is the silicon bandgap voltage. Tak-ing the difference between the base-emitter voltages of two identical bipolartransistors biased at different currents leads to a voltage, which is propor-tional to the absolute temperature (PTAT) and is often used for temperaturesensors. On the other hand, the weighted sum of a base-emitter voltage and aPTAT-voltage generates a temperature-independent reference voltage, whichis almost equal to the silicon bandgap voltage [588].

The first designs of integrated bandgap references and temperature sen-sors were published in the late sixties and early seventies [593, 594]. A fewyears later, the first CMOS-implementations were presented. Some CMOS-designs utilize the temperature-dependence of the gate-source voltage ofMOS-transistors [595]. It is possible to generate PTAT-voltages using MOS-transistors operating in the weak inversion region [596], but it is very difficultto generate an accurate and reliable reference voltage. While the bipolar ver-sions are based on the universal silicon bandgap voltage, the MOS-referencesrely 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 tran-sistors available in CMOS processes. The standard designs [597] have limitedaccuracy owing to mismatch and offset problems. Chopping techniques can beemployed to reduce the errors as a consequence of the amplifier offset [598].If no continuous-time reference voltage is needed, switched-capacitor tech-niques and dynamic-element-matching strategies can be applied to furtherreduce mismatch-induced errors [598–600]. The references of the multisensorchip rely on the vertical pnp-transistor that is available in p-substrate CMOSprocesses. For more details on the on-chip references, see [588].

Temperature SensorThe temperature sensor relies on the linear temperature dependence of a bipo-lar transistor that is available in the CMOS process. The goal in designingthe on-chip temperature sensor was to generate all signals from a single bipo-lar transistor in order to reduce the number of necessary building blocks andto increase the accuracy of the reference voltage and its matching with themeasured voltage. This can be achieved by incorporating all elements into aswitched-capacitor Sigma-Delta-modulator (Σ∆).

The design is derived from a conventional first-order Sigma-Delta-modulator and is shown in Fig. 5.73, for details see Ref. [588]. The tempera-ture sensor exhibits an accuracy of 0.1C at operation temperatures between


−40 and +80C after calibration. The diagrams in Fig. 5.73 illustrate theprinciple of the Σ∆ temperature sensor. The input voltage, Uin, and thereference voltage, Uref , of a temperature sensor are both sums of weightedbase-emitter voltages (UBE). As the input stage of a Σ∆-modulator is anintegrator, these sums can be generated inside the Σ∆-modulator owing tothe fact that a switched-capacitor integrator sums up its input voltages ineach clock-cycle. If the clock-cycle of the modulator (clkΣ∆) is subdividedinto several phases, the single contributions (UBEx) to UPTAT and Uref canbe accumulated inside the integrator. If a switched-capacitor Σ∆-modulatoris used, the different scaling-factors of the single voltage contributions can berealized by changing the input capacitors of the switched-capacitor integra-tor in each clock phase (Fig. 5.73). The clock-cycle, clkΣ∆ of the first-orderΣ∆-modulator is subdivided into five phases. In each phase, one weightedbase-emitter voltage is added to the voltage of the integrator, Uint. The re-sulting voltage change during one clock cycle is ∆Uint = UPTAT ± Uref . Adetailed description of all components and the temperature sensor imple-mentation and performance can be found in [588].

y

T 1

I 0 I 1

clock formgenerator

clk clkΣ∆

kUint I=I0,

k=1

I=I1,k=-1

UPTAT

I=I0,k=1

I=I0,k=m

I=I1,k= -m

±Uref

clkΣ∆clk

Uint

t

y= 0

y= 1

Fig. 5.73. Schematic of the Sigma-Delta temperature sensor [588]

DesignAt present, no single commercial software package is available that can han-dle all aspects of CMOS-MEMS design. For the lower levels in the designhierarchy, dedicated software packages can be used to perform the respectivetasks under the prerequisite that the interfaces to import and export thedata are properly specified. The difficult step is the top-level design, in whichthe transducer layout, and the analog and digital circuitry have to be joined,verified and simulated [601–603].


LayoutVarious custom-made tools for circuit-design and MEMS-design are avail-able in addition to the commercially available tools [604, 605]. Furthermore,there are widely accepted standard formats to exchange data between thedifferent software packages (geometric data standard (GDS)-format, Caltechinterchange format (CIF)) and to import design fragments.

SimulationThe simulation of a complete smart sensor system (sensor elements, ana-log circuitry, digital circuitry) using finite-element (FEM)-simulation toolsis very complex and computationally expensive. Therefore, behavioral mod-els of the transducers are required. The generation of these models fromthe layout or from the results of a 3-dimensional FEM simulation is notstraightforward. Senturia [601] discusses the advantages and disadvantagesof macromodels based on lumped-circuit elements and hardware descriptionlanguages (HDLs). For some structures (mostly comb-structures for, e.g.,accelerometers), the generation of macromodels is supported by custom-designed [606, 607] and commercially available tools [608, 609]. A suitablesimulator must offer the possibility to combine macromodels of the sensorwith transistor-level netlists and various levels of HDL-representations ofanalog and digital circuitry. Furthermore, the analog and digital librariesof the CMOS-process to be used must be available. Most foundries only pro-vide parameters for the tools supplied by the large companies (CadenceTM,MentorTM, SynopsysTM), which restricts the selection to the mixed-signal sim-ulators delivered with those software packages (e.g., SPECTRE-VerilogTM,SPICE-VerilogTM, SABERTM).

Verification and Post-Layout SimulationMost 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 soft-ware package uses different rule formats, the conversion is a difficult task. Formicromachined sensors realized in CMOS-technology with additional post-CMOS processing steps, there are no rule files available. Therefore, the mostconvenient solution is to extend the rule-files that are provided for the cir-cuitry elements by rules for the micromachined structures.

Design FlowThe design flow used for the single-chip gas sensor microsystem is depicted inFig. 5.74 [588]. Various simulation tools were used to simulate and optimizethe transducers [499, 501, 523, 526, 588]. Cadence was employed to draw andsimulate the analog schematics, while SynopsysTM and ModelsimTM were usedto synthesize and simulate the digital circuitry part. The automatic layoutof the digital part was performed with CadenceTM.

The top-level design was done in CadenceTM. Simple lumped-circuit mod-els of the transducers were developed with the aim to include the transducers


FEMmodel

VHDLcode

FEMsimulation

handcalculation

transducermodel

VHDLtestbench

digitalsimulation

synthesizer

digitalschematic

top-levelanalog

schematic

analogschematics

analogsimulation

automatic layout

top-levelschematic

hand layout

top-levelmixed-signalsimulation

digitallayout

transducerlayout

transducerdesign rules

analoglayout

top-levellayout

handcalculation

design rule check,layout versus schematic

analog circuit designsensor design digital circuit design

Fig. 5.74. Design flow of the single-chip gas sensor microsystem [588]

into the top-level simulation. The transducers and the contact network, whichhas to be patterned on the wafer to enable the electrochemical etch stop, obeydesign rules that deviate from those of the circuitry. Therefore, an extensionto the standard design rule check (DRC) was developed that includes the de-finitions of the design rules of the transducers and accepts certain violationsof the CMOS design rules in the transducer vicinity. Owing to the design-ruleviolations caused by the transducers, the standard extraction of the layoutand the subsequent layout-versus-schematic (LVS) test is not possible. Thisproblem was solved by adapting the extraction rules so that the electricalfeatures of the transducers (e.g., the poly-Si resistors of the piezoresistiveWheatstone-bridge of the resonant gas-sensor) are recognized and can be ex-tracted. This enables the verification of the top-level design by comparing thefinal layout to the simulated top-level schematic and avoids wiring-errors. Formore details on the design and simulation of CMOS integrated microsystems,see [501,588].

FabricationThe circuitry and the basic sensor elements (thermocouples, heating re-sistors, piezoresistive Wheatstone-bridge, etc.) are fabricated using an un-altered industrial 0.8-µm CMOS-process provided by austriamicrosystems[496]. CMOS process steps are used wherever possible so as to reduce thenumber of post-processing steps: The pad-etch is, e.g., used to remove theSi-nitride passivation on top of the capacitive sensor electrodes [499].

The CMOS process itself (implantation steps, thermal budget, etc.) re-mains unchanged. An add-on to the basic CMOS process was developed to


enable wafer-level anisotropic etching from the back side of the wafer withan 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 withoutepitaxial layer [611]).

• Lithography modification for the metal-layers to achieve a wafer-level con-tact network to the n-wells that is needed to enable the use of the electro-chemical etch-stop technique.

The fully processed CMOS wafers are thinned to a thickness of 380 µm, anda silicon-nitride layer that serves as a mask for the subsequent KOH-etchingis deposited on the backside. The n-well-membrane precursor of the mass-sensitive cantilever and the thermally insulated island structure of the calori-metric sensor are then released simultaneously by anisotropic silicon etchingwith KOH (potassium hydroxide) from the backside of the wafer and by us-ing the electrochemical etch-stop technique that preserves the n-well of theCMOS process in predefined areas. The silicon cantilever is afterwards re-leased by using two subsequent reactive-ion-etching (RIE) steps. The wafersare then diced using a protective polymer foil covering the chips and themicrostructures. After exposure to UV-light, the foil does no more adhere tothe microstructures and can be removed without damage. Three masks areneeded for the silicon-micromachining, one for the KOH-etching and two forreleasing the cantilevers [526].

5.5.2.3 Multisystem Gas Sensor Measurements

Polymeric layers of approx. 2 µm thickness were applied to the transducers byusing 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 agas manifold. The sensors were then exposed to humidity and several volatileorganic compounds (ethanol, n-octane, trichloroethene, and toluene) at differ-ent concentrations. The total gas flow over the sensor system was 200 mL/min.The sensors were alternately exposed to VOC-loaded synthetic air and pureair at 300 s intervals.

For the simultaneous readout of several multisensor chips, a microcon-troller board (signal processing unit) was developed, which acquired the sig-nals from the microsensor chips and provided the system with stable supplyvoltages. The sensor signals were evaluated by the microprocessor, and theresults were then transmitted to a personal computer.

Simultaneously recorded sensor signals of all three transducers upon ex-posure to 1200 and 3000 ppm (parts per million) of ethanol, and 1000 and3000 ppm of toluene at 30C are displayed in Fig. 5.75 [583, 587]. The sen-sors were alternately exposed to analyte gas and pure carrier gas. The


-1

0

1

2

0 20 40 60 80 0 20 40 60 80 0 20 40 60 80

(a) capacitive

-120

-80

-40

0

(b) mass-sensitive

toluene

ethanol

1200

1000

3000

3000

freq

uenc

y sh

ift [k

Hz]

freq

uenc

y sh

ift [H

z]

time [min] time [min]

(c) calorimetric

time [min]

ther

mov

olta

ge [m

V]

-200

0

200

400

0

80

160

240

0 2 4 6 [s]

polymer:PEUT

Fig. 5.75. Sensor signals simultaneously recorded from all three polymer-coatedtransducers upon exposure to 1200 and 3000 ppm of ethanol and 1000 and 3000 ppmof toluene at 28C: (a) frequency shifts (Sigma-Delta converter output) of thecapacitor, (b) frequency shifts of the resonating cantilever, and (c) thermovoltagetransients of the calorimetric sensor. The close-up shows the development of thecalorimetric transient within 6s [583]

polymeric coating consisted of poly(etherurethane), PEUT, at approximately4 µm thickness.

Figure 5.75a shows the measured frequency signals (Sigma-Delta con-verter output) of the capacitor. Ethanol exhibiting a dielectric constant of24.5, which is larger than that of PEUT (4.8), causes a capacitance increaseand, hence, positive frequency shifts, toluene with a dielectric constant of2.4, causes a capacitance decrease and negative frequency shifts (see alsoSect. 5.1.3).

Figure 5.75b displays the cantilever response. Ethanol shows rather lowsignals as compared to toluene due to its lower molecular mass and due to itslower enrichment (partitioning) in the polymeric phase (see also Sect. 5.3.3).The extent of physisorption is in a first-order-approach proportional to theanalyte boiling temperature (ethanol: 351 K, toluene: 360 K): The lower theboiling temperature, the less enrichment in the polymer [516].

The calorimetric results in Fig. 5.75c represent a superposition of al-ready discussed partitioning and the heat budget change due to analyteab/desorption (see also Sect. 5.2.3). The absorbing analyte liberates heat(predominantly heat of condensation) causing a positive transient signal (pos-itive peak), whereas the desorbing analyte abstracts vaporization heat fromthe environment generating a negative transient signal (negative peak uponpurging, see Fig. 5.75c). The close-up in the upper left corner shows the time-resolved response upon 3000 ppm of toluene within the first 6 seconds. The


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 liq-uid phase (compare, e.g., water), whereas polymer partitioning of tolueneis stronger owing to its higher boiling point.

The signals of all three transducers linearly correlate with the analyteconcentration in the low-concentration range (less than 3% of the saturationvapor pressure at the operating temperature) [516]. As can be seen from theorthogonal responses upon gas exposure (Figs. 5.75, 5.76), different molec-ular properties or different aspects of the coating-molecule interaction aremeasured by the different platforms. Alcohols, e.g., provide comparably lowsignals on mass-sensitive transducers due to their high saturation vapor pres-sure and low molecular mass, but provide large signals on capacitors due totheir large dielectric constant (24.5). Drastic changes in thermo-voltages onthe thermopiles are measured for chlorinated hydrocarbons (not shown here)used, for example, in cooling sprays, which, in turn, have a low dielectricconstant and show only small signals on capacitors.

Another powerful feature of the system is, that even a zero response fromone 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 mi-crosystem chips coated with different polymers can be used. As an example,the sensitivities (slope of sensor signal versus analyte concentration) of a po-lar (EC) and a non-polar (PDMS) polymer to various analytes are shown inFig. 5.76 [612]. For comparability reasons, the analyte responses were nor-malized to that of n-octane, which was set to unity for each polymer andtransducer, so that three bars of unity length would represent the n-octanesignal in Fig. 5.76 for each polymer. Another possibility to normalize thesignals of the different transducers would rely on multiplying the signals orslopes with the analyte saturation vapor pressures [516].

0.0

0.5

1.0

1.5

2.0

2.5

3.0

1 2 3

CapacitorCantileverCalorimeter

0.0

0.1

0.2

0.3

0.4

0.5

0.6CapacitorCantileverCalorimeter

ethanol trichloroethenetolueneethanol trichloroethenetoluene

PDMS EC

norm

aliz

ed s

ensi

tivity

norm

aliz

ed s

ensi

tivity

Fig. 5.76. Sensitivity as normalized to that of n-octane, which was set to unity foreach transducer and polymer. The patterns of the two different polymers (PDMSand EC) differ significantly [612]


It is obvious from Fig. 5.76 that polymer variation significantly changesthe sensor system response pattern. The nonpolar PDMS preferentially ab-sorbs 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 polarso that the opposite holds true. The response characteristics owing to the dif-ferent transduction mechanisms have already been discussed in the context ofFig. 5.75, and the same considerations apply also to Fig. 5.76. The discrimi-nation capability of a microsensor system, hence, can be greatly enhanced byrealizing different transducer principles on a single chip and, additionally, byusing several chips with various polymeric coatings (see also Sect. 5.5.2.4).

Extensive measurements with the single-chip microsensor system coatedwith a variety of commercially available gas-sensitive polymers also showedgood long-term reproducibility of the sensor responses (less than 10% devia-tion within one year).

5.5.2.4 Multisystem Applications and Operation Modes

The monolithic CMOS gas sensor microsystem is targeted at identifying or-ganic solvents in transport containers or providing workplace safety in, e.g.,chemical industry. It will form part of a handheld or credit-card-size detectionunit.

Depending on the application, either single microsystem chips will be usedin standalone units (defined target gas, simple detections tasks such as ana-lyte presence), or more complex systems will be devised that include arraysof 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 hasbeen developed, which includes six single-chip multisensor systems that areassembled on a ceramic board (Fig. 5.77) and coated with different polymers.The instrument thus provides a total of 18 different (6 capacitive, 6 mass-sensitive and 6 calorimetric) chemical sensor signals. The average power con-sumption of one of the multisensor chips is approximately 100 mW.

The instrument also includes a microcontroller board (signal processingunit) developed by the University of Bologna, a miniaturized gas-flow systemdeveloped by the University of Tubingen and a battery supply. The signal-processing unit acquires the signals from the multisensor chips on the ceramicboard and provides the chips with stable supply-voltages of 2.2 and 5 V. Thesignals are evaluated by the microprocessor, and the results are then eithertransmitted to a personal computer via a standard RS232 serial interface ordirectly visualized on a liquid-crystal display (Fig. 5.78).

A photograph and a schematic of the completely assembled handheld in-strument are shown in Fig. 5.78 [588]. It comprises in addition to the sensorsa gas intake unit with pumps, valves and filters, a signal processing unit(microcontroller) and a power pack allowing for more than 24 hours of con-tinuous operation. The handheld instrument has been successfully used to


contact pads

multisensorchips

filtercapacitors

Fig. 5.77. Ceramic board carrying six multisensor chips, the electronics of whichare protected by an epoxy cover [588]. Each chip is coated with a different polymer

measurementchamber

sensorarray

micro-controller

display

RS 232

battery

filter forreference gas pump

valvesgas

inle

t gas

outle

t

Fig. 5.78. Photograph and schematic of the handheld instrument [588]

classify solvents under various environmental conditions, e.g., in laboratoryenvironment, at trade-fairs, and public exhibitions.

Operation ModesSince standalone units or handheld systems are small by nature and, there-fore, have severe constraints imposed on the use of calibration gases or fil-ter units as well as on the overall power consumption, one has to come upwith dedicated operation modes, which still enable reliable qualitative andquantitative measurements. One possibility includes the so-called “reverse


mode of operation” (RMO) [613], which has been developed for the handheldmicrosensor unit and has been adapted to the needs of the different trans-ducer principles as will be described in the following.

In classical systems, the baseline is established by purging with pure car-rier gas (typically nitrogen or synthetic air) or filtered ambient air. Purgingis, in most cases, the basic state of the system, and the sensors are exposed tothe target analytes only during a short time. The sensor signal upon analyteexposure is then compared to the carrier gas or ambient air signal, whichserves as “zero-signal”. Since neither carrying large reference gas cylindersnor 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 systemis now continuously exposed to analyte gas, and only short pulses of filteredor reference gas are used to re-establish the baseline. This operation modeinduces higher demands on the sensor stability, since large drift deterioratesthe sensor signal much more in using RMO than under “classical” operationconditions.

For the microsensor system, the reference baseline is established by5 seconds purging with filtered ambient air. The RMO-technique allowsfor performing some hundred measurement cycles using a very small filterelement.

The operation states of valves and pumps applying RMO are displayedin Fig. 5.79 [614], which additionally shows the strategy and timing of signalrecording from the different transducers and the resulting sensor signals.

Line 1 is indicating the valve status. “0” represents the basic state ofthe valve, when ambient analyte-loaded gas is directly transferred to themeasurement chamber. In state “1”, the ambient gas passes a filter unit, andanalyte molecules are removed from the gas stream: Cleaned “reference” gasis flowing over the sensors.

Line 2 represents the pump status. “0” means “pump off”, “1” meansthat the pump is operational. In line 3, the corresponding analyte gas phaseconcentrations are displayed. In the beginning of a measurement sequence,the gas composition in the measurement chamber is not defined. The pump isthen switched on for 3 seconds, and analyte gas is pumped into the measure-ment chamber. The gas remains in the chamber during two seconds. Equi-librium signals of the capacitive and mass sensitive transducers are recorded,the measurement timing of which is displayed in line 4. The resulting sensorsignals are schematically shown in line 5.

The pump is then switched on again for 3 seconds, and the valve is setto route the analyte gas through the filter. A concentration step betweenanalyte-loaded and filtered gas is thus generated.

Equilibrium state capacitive and mass-sensitive reference signals in fil-tered air are recorded during two seconds. After that, the pump is switched


pump

measurement timingcalorimeter

sensor signalcalorimeter

valve

0

1

0

1

analyteconcentrationin chamber (undefined)

3 sec 3 sec 3 sec

2 sec 2 sec

5 secline 1

line 2

line 3

measurement timingcapacitor, cantilever

line 4

line 6

line 7

sensor signalcapacitor, cantilever

line 5

Fig. 5.79. Reverse Mode of Operation (RMO) [614], as developed for a handheldunit including micromachined multisensor chips: Operation states of valves andpumps and corresponding gas concentrations in the chamber (lines 1–3), and timingof 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 operationwould not be necessary for the equilibrium-based sensors but it is necessaryto get the second calorimetric transient as shown in line 7. As already de-scribed in Sect. 5.2.3, the calorimetric sensor relies on transients and providessignals exclusively upon concentration changes. Therefore, the calorimetricrecording has to be performed at high time resolution (1 kHz) in two shortintervals covering both flanks of the concentration signal (line 3), i.e., at themaximum gradient of the analyte concentration. The two transient signals ofthe calorimetric transducer (negative upon analyte desorption, positive uponanalyte absorption) are displayed in line 7.

5.6 CMOS Chemical Microsensor and System Packaging 181

5.6 CMOS Chemical Microsensorand System Packaging

The packaging of chemical microsensors often does not receive much atten-tion despite the fact that the package largely influences the reliability andlong-term stability of a microsystem and represents a major cost factor ofcommercial microsensor systems [615]. Packaging strategies for integratedchemical microsensor systems in CMOS technology can rely on the copiousknowledge in microelectronics and IC-related packaging. However, specificand custom-designed solutions have to be developed for sensor devices, sinceit 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 stimu-lus to the sensor. The most important issues with chemical sensor packaginginclude [616]:

• Free analyte access: The analyte must have direct access to the sen-sor structures, whereas the circuitry and the electrical connections (bonds)have to be protected from the environment (temperature, radiation, hu-midity, chemicals).

• Chemically inert package: The packaging material must be chemicallyinert with respect to (sometimes) aggressive media and analytes, and it maynot outgas or release chemicals that might interfere with analyte detection.

• Microsystem-package codesign: The sensor package has to be definedduring the initial microsystem conception phase, since the electronics andtransducer layout of the microsystem and the component arrangement onthe chip are strongly interrelated with the selected packaging solution.

Three different packaging approaches featuring increasing complexity willbe presented in the following subsections of this chapter: A quite simpleapproach 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 technology-based approach (Sect. 5.6.3). The advantage of the epoxy-based packaging isits simplicity. The flip-chip packaging requires massive efforts, but allows forminimized sample volumes (gas or liquid-phase flow system), and provideshermetical sealing of the circuitry from the probed sample volume, whichleads 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 placedalong one edge of the chip, while all electrical connections (bond pads) andthe sensitive parts of the circuitry should be arranged at the other end of thechip as it is shown in Fig. 5.80. This distribution, which has to be realized


pins

TO-headerbond wires

epoxy

circuitry

sensors

chip

pads

Fig. 5.80. Schematic and photograph of an epoxy-based sensor package. The par-tial epoxy cover enables free analyte access to the sensors of the microhotplatearray

during the chip design and layout phase, is necessary since the epoxy cannotbe applied with micrometer precision and still flows and expands somewhatduring the process of drying.

The epoxy-based packaging concept and a packaged microhotplate arraychip (see also Fig. 5.65) are shown in Fig. 5.80. A commercially availablestandard gold-coated transistor-outline TO-8 socket with 16 pins has beenchosen. The chips are affixed with an electrically non-conductive die-attach.The die-attach is only applied underneath the circuitry area, so that themicrohotplate cavities are not completely sealed. Pressure changes, whichmight build up during heating can dissipate through the small slit betweenchip and package. The chip is then wire-bonded with a wedge-wedge bonderand Al-wires. For encapsulation, the bond-wires and the circuitry are coveredwith a standard epoxy (e.g., Loctite HYSOL FP 4460). The epoxy is appliedwith a dispenser and is cured at 125C. It is chemically inert and does notaffect the sensitive layer during curing. Partially covering the chips with epoxyallows, 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 theTO-socket as protection element against dust and mechanical impact. Thecap also allows for integration of metal sieves and filters, which can be animportant 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-circuitboard so that the wire bonding is directly from chip to board. The chip-on-board package has been realized for commercially available humidity sensors[484] and will be applied here to an array of six multisensor chips that have


been described in Sect. 5.5.2. A flip-chip design of the multisensor chip thatfeatures the sensors in the center of the chip within a metal frame surroundedby 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 isshown in Fig. 5.81.

filter capacitors

array contacts

capacitive

calorimetric cantilever

interface

Fig. 5.81. Photograph of the multisensor chip as designed for chip-on-board pack-aging (left) and of the packaged array (right) [588]

The transducers are all placed at one end of the chip so that the circuitrycan be protected by epoxy encapsulant. In comparing the different designs ofthe same multisensor system in Fig. 5.71 and Fig. 5.81, it is evident that thepackaging method – flip chip or chip on board – largely influences the designand layout of the microsensor system so that microsystem-package-codesignis highly desirable as has been mentioned in the introductory part of thissection (Sect. 5.6).

Six identical multisensor chips (5 mm by 7 mm) are die-attached on acommon ceramic alumina substrate (Fig. 5.81). Each of the multisensor chipsfeatures an I2C bus interface (see 5.5.2.2) so that all chips can be operatedon the same bus, and only few electrical connections are necessary. Wirebonds connect the bond pads on the chips with the electrical leads on theceramic board. The wire bonds and the circuitry are then protected withepoxy. Finally, the chemical-sensor structures of the chips are spray-coated,each chip with a different polymer using a shadow mask.


5.6.3 Flip-Chip Package

Flip-chip packaging is a technology that has been developed to allow for amultitude of connections from chip to package or board [617]. Only pads alongthe edges can be used in, e.g., standard wire bonding or packaging methods.Since the length of the edges becomes smaller and smaller with decreasingfeature size of the CMOS processes albeit the bonding pads require a certainsize and area to allow for reliable connections, a packaging and connectiontechnology was developed, which enables to make connections on the wholechip surface [617]. The chip is flipped upside down and the connections aremade simultaneously from all pads on the chip face to the correspondingpre-patterned substrate via metal solder or conductive polymers (Figs. 5.82,5.83). Flip-chip packaging is technologically more expensive than the previoustwo methods and requires a layout with a similar number of connections oneach side of the chip (see Fig. 5.83). In the case of chemical microsensors, itis not the large number of possible connections (bus interface: 7 connections)that renders this technology attractive but the possibility to define a sensorarea within a metal frame and to hermetically seal the circuitry componentsfrom the chemical environment by means of solder [589]. By using a chemi-cally inert substrate material such as alumina, a reliable and longterm-stablepackage is created.

sensor

solder ceramic

chip

metallization

opening

underfill(polymer)

sealing

Fig. 5.82. Schematic of the flip-chip process: Ceramic substrate and chip are pre-treated (metal-line patterning on the substrate, gold-bump metallization on thechip) and afterwards connected via solder and polymeric underfill [499,589]

On the chip side, the bonding pads have to be adapted for flip-chip pack-aging since the minimum area required for a reliable flip-chip connection is150 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 hasto be designed around the transducers (Figs. 5.71, 5.83), which does not serveelectrical connection purposes but is intended to provide hermetical sealingof the circuitry components outside the metal frame. The different pads have


contact pads

metal ring

ceramic substrate with solderchip with metallization

gold contacts gold frame

Fig. 5.83. Flip-chip procedure: The chip (left) is soldered face-down onto theceramic substrate. Metal sealing frames and contact pads are clearly visible [499]

to be distributed on the chip surface in a symmetric arrangement, so thatthe ceramic substrate can rest on those pads and is not tilted. The metallicframe surrounding the sensors and the pads are then covered with nickel/goldbumps.

The ceramic substrate also has to be prepared. Laser cutting is used toopen a window for the sensors in the ceramic substrate. Then, the electricalconnections are screen-printed on the ceramic. The soft solder paste for theflip-chip packaging is also applied to the ceramic using stencil printing.

The chip is then precisely positioned face-down on the ceramic substrateby means of a micromanipulator and a solder reflow is performed at 230C.The surface tension of the solder provides a very accurate alignment duringthe reflow process. Finally, an epoxy-based polymeric underfill is applied andcured at 160C. The sensors can afterwards be coated with different polymersusing a drop-coating method and the substrate openings as mask.

The flip-chip procedure is exemplified in Fig. 5.83 with a capacitive chipfeaturing three interdigitated sensor-reference pairs (top: references, center:sensors). The metal frame on the chip around the three sensors and themetal bumps at the top and the bottom of the chip can be clearly seen. Thesubstrate features three openings for demonstration purposes (one openingfor all three sensors would be sufficient) and a symmetric arrangement ofcontacts to accommodate the chip.

After the flip-chip process, the three sensing capacitors look through thewindows, albeit the references (top) and circuitry units (bottom) are sealedfrom the analyte and sample environment. The polymeric underfill mechani-cally stabilizes the packaged system.

6 Outlook and Future Developments

The field of CMOS-integrated chemical and biosensors is currently develop-ing at fast pace and is expanding as can be seen from the growing numberof publications over the last years. The two most important research anddevelopment thrusts include (i) realizing more complex monolithic sensorarrays (also including different transducer types) and integrating more andmore electronic functions on the sensor chip, and (ii) combining microelec-tronics, in particular CMOS chips, with biologically relevant molecules suchas 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 evidentfrom Chap. 5 of this book. A modular approach or “toolbox strategy” relyingon CMOS as platform technology has been described in detail. A variety ofcomponents of the toolbox such as transducers, sensor modules, and circuitmodules have been developed, which afterwards have been assembled intocustomized systems such as the digital microhotplate array (Sect. 5.4.2.3) orthe multitransducer system (Sect. 5.5.2).

Developments towards more complex microsystems in the immediate fu-ture 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 struc-tures to smaller size, e.g., micromechanical cantilevers down to the nanome-ter scale, or microhotplates to less than 100 µm diameter. Miniaturizationalso will include replacing microtransducers for rather large millimeter-sizetransducers, or devising novel sensing and actuation principles for chemicalmicrosensors. All these efforts will lead to a drastic reduction in system sizeand power consumption. A target system size of 5 by 5 mm2, and a target sys-tem power consumption of less than 100–200 mW (depending on transducertype) seem to be a realistic estimate.

188 6 Outlook and Future Developments

Another important issue in further miniaturization will concern the gas-sensitive material and its precise local deposition. Micro- and nanometer-resolution deposition methods have to be developed, which provide suffi-cient precision and reproducibility at reasonable costs and potentially highthroughput.

Further miniaturization of transducers and sensing structures stringentlyrequires a higher level of integration of electronic and circuitry compo-nents: The sensor signals will become more and more minute, so that thetransfer of low-level analog signals via bonds and interconnects will be nomore an option: The signal of, e.g., capacitive chemical microsensors (seealso Sect. 5.1) is on the order of some attoFarads (10−18 Farad). Monolithicimplementations, which combine miniaturized transducers and on-chip cir-cuitry in immediate vicinity to the signal-generating processes, represent aneffective approach to significantly reducing the influence of parasitic capaci-tances and crosstalk-effects that deteriorate low-level signals of any microsen-sor. Short on-chip interconnections between the sensor components and thesystem control help to further minimize noise.

In most cases, it will also be necessary to convert the signal on chip intothe digital domain and to transfer only digital signals to off-chip units. Datapreprocessing and A/D or D/A conversion is hence to be implemented onchip, which poses new challenges (see Sect. 5.5.2.2) with regard to designand testing: New simulation and test strategies for digital-analog co-designof 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, custom-designed interfaces, bus interfaces (see Sect. 5.5.2.2), adaptive circuits forsignal evaluation and discrimination [618, 619], or telemetry units for wire-less communication. Telemetry functions will encompass, e.g., communica-tion from implanted chips (glucose sensors) through the skin in medicalapplications [620] or the transmission of remote and distributed chemicalsensor responses to a central terminal in environmental and building con-trol scenarios by using radio frequencies [621–623] or other standards likeBluetoothTM [624].

Complete miniaturized chemical analysis systems are currently un-der development at Sandia National Laboratories and at the University ofMichigan. Sandia National Laboratories has developed both, gas and liquidphase chemical analysis systems (µ ChemLabTM) [625, 626]. The gas-phasesystem is targeted towards the field detection of chemical warfare agents andconsists of three microfabricated components (Fig. 6.1): (a) A preconcentra-tor, (b) a separation unit, and (c) a gravimetric detector. Preconcentrationis achieved by pulling an air sample across a selective absorbent coating ona thermally isolated silicon nitride membrane; a heating pulse is then usedto release the absorbed chemical species. Separation is accomplished using

6 Outlook and Future Developments 189

heater metalsilicon nitridesilicon substratethin-film adsorbent

preconcentrator accumulatesspecies of interest

gas chromatographseparates species in time

Rayleigh-SAW sensors provide sensitive detection

metal electrodes

excitation

propagation

detection

Fig. 6.1. Schematic of the µChemLabTM system as developed by Sandia NationalLaboratories (top) and micrographs of its different components (bottom). The com-ponents are not to scale. Pictures courtesy of R.W. Cernosek, Sandia NationalLaboratories

either a wall-coated or bead-packed deep-reactive-ion-etched silicon channelthat acts as a microgaschromatographic (µGC) column. The separated sam-ples are detected with a microfabricated surface-acoustic-wave (SAW) devicearray.

All key components of a microgaschromatographic column (µGC), namelya passive calibration-vapor source, a three-stage preconcentrator/focuser, a3-meter separation column, and an integrated vapor sensor array have beendeveloped at the University of Michigan [627,628]. All these components arefabricated from silicon and collectively occupy approximately 1 cm3 volume.Plumbed via a fluidic interconnection substrate in combination with fusedsilica microcapillaries, the µGC is coupled with a commercial minidiaphragmpump and minivalves to perform analyses of various mixtures of toxic volatileorganic compounds.

The second important trend is the combination of CMOS microelec-tronics with biomolecules or biological entities such as living cells orelectrogenic 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 detec-tion rather than optical detection represent a very interesting application forCMOS technology [629,630]. The sensor chip encompasses a 16×8 pixel arraywith gold sensor electrodes [630]. The pixel circuits are designed to detectand amplify sensor currents in the 100 nA to pA range that result from anelectrochemical redox cycling process. The chip is realized in 0.5-µm CMOS


technology and additionally hosts addressing circuitry, reference electrodesand potentiostatic circuitry.

Monitoring of chemical parameters in cell cultures, such as metabolicallyinduced pH-changes or oxygen contents and ionic composition in the cellenvironment, has been performed with circuit-less CMOS chips that includeISFET arrays (pH and different ions), interdigitated electrodes, an oxygenand a temperature sensor [631–634].

Arrays of electrodes or microelectrodes with associated circuitry on chipcan be used to extracellularly measure the electrical signals of electrogeniccells. Portable, cell-based biosensor systems featuring CMOS chips to inter-rogate electrogenic cells have been reported on [635–639]. Cardiac myocyteswere coupled to CMOS circuits to yield a biosensor for detecting warfareagents in portable field alarm systems. The cartridge contains a CMOS sili-con chip that incorporates a digital interface, a temperature control system,microelectrode electrophysiology sensors and analog signal buffering. Each ofthe two chambers hosting the cells on the chip features one large referenceelectrode, two stimulation electrodes, four arrays of 16 microelectrodes, tem-perature sensors and nine 16-channel multiplexers. As soon as the heart cellsare exposed to toxic chemicals they change their electrical signal character-istics (faster or slower beating frequency) or perish (no signal).

First results on coupling neurons to CMOS-based chips were publishedvery recently [640–642]. A biocompatible capacitive interface including a tita-nium oxide and zirconium oxide sandwich has been post-processed onto eachof the 16.000 pixels or sensors, which have an integrated amplification cir-cuit. Action potentials of firing neurons are thus capacitively coupled to thegates of MOS transistors that operate as input devices within a monolithicdetection system, which also features addressing circuitry, multiplexers, anda temperature sensor, but does not yet enable electrical stimulation.

A chip comprising a microelectrode array and fully integrated analog anddigital CMOS circuitry for the stimulation and recording of electrogenic cellswas reported on in [643]. This device is capable of on-chip signal filtering,which improves the signal-to-noise ratio, on-chip analog/digital conversionto prevent signal degradation, and simultaneous recording and stimulation(Fig. 6.2). Each electrode has associated circuitry comprising a stimulationbuffer and a switch, a bandpass filter and a readout buffer. The electrodeand circuitry form a repeatable unit with a 250-µm pitch. This unit can bemultiplied to form a larger array. The stimulation buffer, occupying verysmall area, uses a class-AB output stage in order to deliver large currents tothe electrode. Any stimulation waveform can be generated. Fabrication wasperformed using an industrial double-polysilicon, triple-metal 0.6-µm CMOSprocess. The platinum microelectrodes are realized during postprocessing forflexibility purposes.

First tests of the chip in Fig. 6.2 were performed with an in situ prepara-tion of cardiac myocytes (heart cells) from embryonic chicken at embryonic

6 Outlook and Future Developments 191

test structure16-electrode

array

biasing

temperaturesensor

digital control

4 A/D converters

D/A converter

30 µm

electrode (Pt)

circuitry repeating unit

500 µm

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, onedigital-to-analog converter, on-chip multiplexer, digital control unit, and temper-ature sensor. Close-up (left side): Circuitry repeating unit with buffer, switch forstimulation, bandpass filter and readout buffer [643]

day 10. The heart cells were extracted from the embryo, briefly rinsed anddirectly 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 thanthat of, e.g., chicken neurons (tens of microvolts) [643].

Needle-shaped neural probes fabricated in CMOS-technology in conjunc-tion with micromachining have been developed at the University of Michigan[644–648]. A neural probe consists of a micromachined silicon substrate withneedles supporting an array of thin-film conductors leading from recording or

U [m

V]

time [ms]

10 50 100

0.3

0.2

-0.1

0.4

-0.2

0

0.1

Fig. 6.3. Spike of cardiac myocytes from embryonic chicken [643]


interconnectingleads

stimulating orrecording sites

supporting substrate

signal-processing

circuitry

output leads

(a) (b)

Fig. 6.4. Schematic (a) and micrograph (b) of a neural probe. Fig. 6.4b shows a64-site 8-channel stimulating array delivering currents from 0 to ±128 µA with aresolution of 1 µA and 4 µsec. Reprinted with permission from [650]

stimulating sites along a shank to signal processing electronics and/or out-put leads at the rear of the device (Fig. 6.4). The recording or stimulatingsites (up to 64) are mounted along the upper surface allowing for record-ing or stimulation at many points in depth. The neural probes are aimed atrecording 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 theneural signals without significantly degrading baseline noise levels. Anotherprimary reason for on-chip electronics in such a device is to minimize thenumber of external leads, which eases connectivity and reduces the chanceof lead-related failures. In recent papers, a multichannel neural probe waspresented that also provides regio-selective chemical delivery as additionalfunction along with the stimulation and recording of cell activity [648–650].

In summary, using CMOS and CMOS-MEMS technology to devise chem-ical, biochemical and biosensors offers the possibility to select componentsfrom a wealth of electronic circuits and functions that are and will be pro-vided by the enduring huge development efforts of the IC-industry. The eco-nomic aspect, however, has to be kept in mind, i.e., the fact that CMOSsensor development is expensive (mask and processing costs) and only paysoff, if either millions or at least hundreds of thousands of pieces are sold, orthe devices provide unique functionality, which then legitimates the costlydevelopment. Applications involving large numbers of chemical or biosensorshave 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 thefield of pharmacological screening.

References

1. Gopel, W., Hesse, J., Zemel, J.N. (eds.), Sensors: A Comprehen-sive Survey, Vol. 2/3, Chemical and Biochemical Sensors, VCH-Verlagsgesellschaft,Weinheim 1991, and Baltes, H., Gopel, 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, Insti-

tute 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.,Gopel, W. Pattern Recognition and Multicomponent Analysis in Sensors Up-date, and Baltes, H., Gopel, 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, Datahandling 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. Muller, 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. Lundstrom, 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 edi-

tion, 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, Wein-

heim, 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.,

Koppel, 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. Munch, 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. Larmer, 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., Gopel, 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 Physico-Chemical 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. Bodenhofer, K., Hierlemann, A., Noetzel, G., Weimar, U., Gopel, 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. Transducers1985 (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., Gopel, 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., Gretillat, 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. ActuatorsA, 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., GuntherodtH.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., Gopel 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., Guntherodt 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.,Guntherodt 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., MeyerE., Guntherodt 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 AdvancedResearch 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 PhysicsPublishing, Bristol, UK, 1994.

110. Baker, A.R., Combustible gas-detecting, electrically heatable element, UKPatent 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 Work-shop Hilton Head Island, SC, 1996, 23–27.

113. Manginell, R.P., Smith, J.H., Ricco, A.J., Proc. 4th Annual Symposium onSmart 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. Ac-

tuators 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., Jaffrezic-

Renault, 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. Ac-

tuators 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 MechanicalSystems MEMS 2000, Myazaki, Japan, 2000, 96–101, ISBN 0-7803-5273-4.

136. Koll, A., Schaufelbuhl, 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., Schropfer, 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., Biosen-

sors, Bioelectronics, 1993, 8, 89–98.141. Kohler, J.M., Kessler, E., Steinhage, G., Grundig, B., Cammann, K.,

Mikrochim. Acta, 1995, 120, 309–319.142. Sarro, P.M., Yashiro, H., van Herwaarden, A.M., Middelhoek, S., Sens. Actu-

ators, 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 Gopel, 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 Com-prehensive Survey, Update Vol. 1, and Gopel, W., Hesse, J., Zemel, J.N. (eds.),VCH-Verlagsgesellschaft, Weinheim 1996, 1–49.

149. Brecht, A., Gauglitz, G., Gopel, W., Sensor in Biomolecular Interaction Analy-sis and Pharmaceutical Drug Screening in Sensors: A Comprehensive Sur-vey, Update Vol. 3, and Gopel, W., Hesse, J., Zemel, J.N. (eds.), VCH-Verlagsgesellschaft, 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, Sin-

gapore, 1997.153. Azzam, R.M.A., Bashara, N.M., Ellipsometry and Polarized Light, North

Holland, New York, 1988.154. Jin, G., Tengvall, P., Lundstrom, 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 Bi-

ological Application, Wiley-VCH, Weinheim, 1998.164. Klimant, I., Kuhl, M., Glud, R.N., Holst, G., Sens. Actuators B, 1997, 38–39,

29–37.165. Holst, G., Glud, R.N., Kuhl, 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 Op-

tical 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., Dubendorfer, J., Sens. Actua-

tors 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., Kashammer, J., Kramer, 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–

217.184. Ayras, P., Honkanen, S., Grace, K.M., Shrouf, K., Katila, P., Leppihalme, M.,

Tervonen, A., Yang, X., Swanson, B., Peyghambarian, N., Pure Appl. Opt.,1998, 7, 1261–1271.

200 References

185. Grace, K.M., Shrouf, K., Honkanen, S., Ayras, P., Katila, P., Leppihalme, M.,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. Actua-

tors B, 1996, 35–36, 234–240.189. Heideman, R.G., Lambeck, P.V., Sens. Actuators B, 1999, 61, 100–127.190. Maisenholder, B., Zappe, H.P., Kunz, R.E., Moser, M., Riel, P., Technical

Digest Transducers Chicago, 1997, 79–80.191. Maisenholder, B., Zappe, H.P., Kunz, R.E., Riel, P., Moser, M., Edlinger, J.,

Sens. Actuators B, 1997, 38–39, 324–329.192. Maisenholder, 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., Sochtig, J., Electronics Letters, 1994,

30, 1858–1859.195. Zappe, H.P., Hess, M., Moser, M., Hovel, 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. Actu-

ators 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., Mid-

delhoek, 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., SPIEproceedings 1998, 3328, 202–212.

216. Eres, G., Lowndes, D.H., US Patent 5 110 790, 1992.217. Liedberg, B., Nylander, C., Lundstrom, 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/spreeta221. Widmer, H., Anal. Methods Instrum., 1993, 1, 60–72.222. Vasiliev, A., Moritz, W., Fillipov, V., Bartholomaus, L., Terentjev, A., Gabus-

jan, T., Sens. Actuators B, 1998, 49, 133–138.223. Baranzahi, A., Spetz, A., Lundstrom, I., Sens. Actuators B, 1995, 26–27, 165–

169.224. Svennigstorp, H., Tobias, P., Lundstrom, 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.), CRCPress, Boca Raton, FL, 1997.

227. Wollenberger, U., Electrochemical Biosensors-Ways to Improve Sensor Perfor-mance 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 Engi-

neering in Medicine and Biology, 1994, 13, 409–419.232. Buss, G., Schoning, M.J., Luth, H., Schultze, J.W., Electrochimica Acta, 1999,

44, 3899–3910.233. Schmitt, G., Schulze, J.W., Fassbender, F., Buss, G., Luth, H., Schonung,

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 DigestTransducers 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., Esteve, D., Gue, 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., Lurick, 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.,

Bucher, S., Buhner, M., Cammann, K., Knoll, M., Sens. Actuators B, 1995,24–25, 300–303.

255. Hermes, T., Buhner, M., Bucher, S., Sundermeier, C., Dumschat, C., Bor-chardt, 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. Actu-ators B, 1994, 20, 17–22.

257. Van den Berg, A., Grisel, A., Koudelka, M., van der Schoot, B.H., Sens. Ac-tuators 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 Trans-

ducers, 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. Lundstrom, I., Sensors Act B, 1996, 56, 75–82.269. Ekedahl, L.G., Eriksson, M., Lundstrom, 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., Lundstrom, I., J. Appl. Phys., 1988, 64, 1274–1283.272. Spetz, A., Winquist, F., Sundgren, H., Lundstrom, 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., Lundstrom, I., Appl.

Phys. Lett., 1983, 43, 839–841.276. Winquist, F., Lundstrom, 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., DeWeerth, 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., Potter, 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. Ac-

tuators, B, 1997, 44, 434–440.

204 References

299. Rapp, R., Hoffmann, W., Suss, W., Ache, H.J., Golz, H., Electrochimica Acta,1997, 42, 3391–3398.

300. Cane, C., Gotz, A., Merlos, A., Gracia, I., Errachid, A., Losantos, P., Lora-Tamayo, 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., Gopel, 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. Brzozka, 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., Brzozka, 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 Rein-

houdt, D.N., Sens. Actuators B, 1993, 15–16, 195.318. Tsukuda, K., Miyahara, Y., Shibata, Y., Miyagi, H., Technical Digest Trans-

ducers, 1991, 218–221.319. Brzozka, 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., Muller, 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-Ruel, B.H., Lugtenberg, R.J., Engbersen, J.F., vanden 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. Actu-

ators B, 1999, 58, 399–408.334. Dumschat, C., Muller, 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., Offenhauser, A., Vetter, T., Weis, J., Science, 1991, 252, 1290.338. Offenhauser, A., Sprossler, C., Matsuszawa, M., Knoll, W., Biosensors Bio-

electronics, 1997, 12, 819–826.339. Schoning, M.J., Schutz, S., Schroth, P., Weissbecker, B., Steffen, A., Kordos,

P., Hummel, H.E., Luth, 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.,

Sudholter, E.J., Sens. Actuators, 1989, 18, 309–327.343. Matsuo, T., Nakajiama, H., Sens. Actuators, 1984, 5, 425–435.344. Cane, C., Gracia, I., Merlos, A., Lozano, M., Lora-Tamayo, E., Esteve, J.,

Technical Digest Transducers 1991, 225–228.345. Skowronska-Ptasinska, 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., Lundstrom, 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., Schoneberg, U., Mokwa, W., Hostika, B.J., Vogt, H., Sens. Actu-

ators B, 1992, 6, 162–164.363. Westcott, L., Rogers, G., J. Phys. E. Sci. Intrum. 1985, 18, 577–581.364. Schoning, M.J., Thust, M., Muller-Veggian, M., Kordos, P., Luth, 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. Schoning, M.J., Ronkel, F., Crott, M., Thust, M., Schultze, J.W., Kordos, P.,

Luth, H., Electrochimica Acta, 1997, 42, 3185–3193.367. Gabusjan, T., Bartolomaus, 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., Bogner, M., Fuchs, A., Winter, R., Doll, T., Eisele, I., Sens.Actuators B, 1999, 57, 35–38.

373. Bogner, 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., Bogner, 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 Bioelectron-

ics, 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., Reyes-

Barranca, 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.com399. Gopel, W., Reinhardt, G., Metal Oxide Sensors in Sensors Update Vol. 1, and

Baltes, H., Gopel, 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., Lundstrom, 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 Sen-

sors 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 Sensorsand 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. Actua-

tors 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., Gopel, 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., Gopel, 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.), El-

sevier, 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.,

Gopel, 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., Muller, 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. Gotz, A., Gracia, I., Cane, C., Lora-Tamayo, E., Horrillo, M.C., Getino, G.,

Garcıa, C., Gutierrez, 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 Tech-

nology A, 1993, 11, 1392–1395 and 1994, 12, 2149–2152.

References 209

445. Frietsch, Dimitrakopoulos, L.T., Schneider, T., Goschnick, J., Surface Coat-ings 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., Gopel, 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., Gopel,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 Work-shop, 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., Ares, L., Rodrigo, J., Gutierrez, J., Gotz, A., Gracia,

I., Fonseca, L., Cane, 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., Levy, F., Sens. Actua-tors 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., Noet-

zel, G., Weimar, U., Gopel, W., Sens. Actuators B, 1995, 24–25, 357–361.479. Steiner, F.P., Hierlemann, A., Cornila, C., Noetzel, G., Bachtold, M., Korvink,

J.G., Gopel, 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 DigestTransducers 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. Actua-

tors 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 Work-

shop, 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., Moller, 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. Ac-

tuators 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 Com-

pounds, 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 DetectionSystems 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 Trans-

ducers 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., Bodenhofer, K., Dominik, A., Gopel, 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 ofTubingen, Germany.

520. Schneeberger, K., CMOS Microsystems for Thermal Presence Detection,Ph.D. thesis ETH Zurich No. 12675, 1998.

521. Schaufelbuhl, A., Thermal Imagers in CMOS Technology, Ph.D. thesis. ETHZurich No. 14484, 2002.

522. Kerness, N., Koll, A., Schaufelbuhl, 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. thesisETH Zurich No. 14839, 2002.

524. Menolfi, C., Low-Noise CMOS Chopper Instrumentation Amplifiers for Ther-moelectric Microsensors, Ph.D. thesis ETH Zurich No. 13583, 2000.

525. Hierlemann, A., Massensensitive Detektion fluchtiger organischer Substanzenmit modifizierten Polysiloxanen, Ph.D. thesis, University of Tubingen,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. Tech-

nol. 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., Ruegg, M., Li, Y., Lange, D., Hagleitner, C., Brand, O., Hier-

lemann, 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 Engineer-

ing, 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., Guntherodt, 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., Guntherodt, 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 B2001, 78, 12–18.

547. Barsan, N., Weimar, U., Journal of Electroceramics, 2001, 7, 143–167.548. Simon, I., Barsan, 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., Cavic-

chi, R.E., Semancik, S., Montgomery, C.B., Taylor, C.J. IEEE-Sensors-Journal2002, 2, 644–655.

551. Graf, M., Barrettino, D., Zimmermann, M., Hagleitner, C., Hierlemann, A.,Baltes, H., Hahn, S., Barsan, 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 CMOStechnology, ETH Diss. 15438, 2004.

556. Barrettino, D., Graf, M., Taschini, S., Zimmermann, M., Hagleitner, C., Hi-erlemann, A., Baltes, H. Proceedings IEEE VLSI-Symposium, Kyoto, Japan2003, 157–160.

557. Graf, M., Jurischka, R., Barrettino, D., Hierlemann, A., J. Micromech. Micro-eng., 2005, 15, 190–200.

558. Barrettino, D., PhD Thesis, CMOS readout and control architectures formonolithic 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. Muller, T., Brandl, M., Brand, O., Baltes, H. Sensors and Actuators A-Physical 2000, 84, 126–133.

562. Wollenstein, J., Plaza, J.A., Cane, C., Min, Y., Bottner, H., Tuller, H.L. Sen-sors and Actuators-B-(Chemical) 2003, 93, 350–355.

563. Cavicchi, R.E., Suehle, J.S., Kreider, K.G., Shomaker, B.L., Small, J.A., Gai-tan, 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 Ac-tuators B-(Chemical) 1998, 49, 81–87.

565. Savage, N.O., Roberson, S., Gillen, G., Tarlov, M.J., Semancik, S. AnalyticalChemistry, 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., Romano-Rodriguez, A., Morante, J.R., Gopel, W. Fresenius Journal of AnalyticalChemistry 1998, 361, 110–114.

214 References

574. Kappler, J., Tomescu, A., Barsan, N., Weimar, U., Thin Solid Films 2001,391, 186–191.

575. I2C (inter integrated circuit, inter-IC) is a communication standard developedby Philips, Eindhoven, The Netherlands: http://www.semiconductors.philips.com/buses/i2c/facts.

576. Graf, M., Taschini, S., Kaser, P., Hagleitner, C., Hierlemann, A., Baltes, H.,Proc. Eurosensors, Guimaraes, Portugal (2003) 661–662.

577. Graf, M., Taschini, S., Kaser, 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., TechnicalDigest IEEE Int. Electron. Devices Meeting, San Francisco, 1992, 521–524.

579. Hughes, K.L., Miller, S.L., Rodriguez, J.L., McWorther, P.J., Sens. ActuatorsB, 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., DeCaussemaeker, 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 Jour-nal of Solid-State Circuits, 2002, 37, 1867–1878.

588. Hagleitner, C., Hierlemann, A., Baltes, H., CMOS Single-chip Gas DetectionSystems 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, Low-

Voltage, Mixed-Mode Design with CAD Tools, Voltage, Current and TimeReferences (eds: Huijsing, J.H., van de Plasche, R.J., Sansen, W.), KluwerAcademic 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 Mi-

crosystems, 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. Muller, T., Feichtinger, T., Breitenbach, G., Brandl, M., Brand, O., Baltes,

H., Proc. 11th IEEE Workshop on Micro Electro Mechanical Systems, 1998,240–245.

611. Muller, T., An industrial CMOS process family for integrated silicon sensors,Ph.D. thesis no. 13463, ETH Zurich, 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 ChemicalMicrosystems for Gas Analysis, Handbook of Machine Olfaction, Editors T.C.Pearce, S.S. Schiffman, H.T. Nagle, J.W. Gardner, Wiley VCH Weinheim, NewYork, 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 Microelec-tronics 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. IEEETransactions on Biomedical Engineering 1994, 41, 937–42.

621. Wouters, P., de Cooman, M., Lapadatu, D., Puers, R. Sensors and ActuatorsA (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-StateSensor 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 Mon-itoring 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. ProceedingsIEEE 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 Micro-electromechanical 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 ActuatorsB (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., Schulein, J., Goemann, W., Gumbrecht, W., IEEEInternational 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., Cut-tino, 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., IEEEJournal of Solid-State Circuits, 2003, 38, 2306–2317.

643. Franks, W., Heer, F., McKay, I., Taschini, S., Sunier, R., Hagleitner, C., Hi-erlemann, 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: analog-digitalAC: alternating currentADC: analo-to-digital converterAE: auxiliary electrodeAFM: atomic-force microscopyAHDL: analog hardware description languageAl: aluminumAPCVD: atmospheric-pressure chemical vapor depositionASIC: application-specific integrated circuitATR: attenuated total reflectionBBIC: bioluminescent bioreporter integrated circuitBET: Brunauer-Emett-TellerBOE: buffered oxide etchBSA: bovine serum albuminCAD: computer-assisted designCAN: controller area networkCCD: charge-coupled deviceCE: counter electrodeCHEMFET: chemically sensitive field-effect transistorCIF: Caltech interchange formatCMOS: complementary metal oxide semiconductorCVD: chemical vapor depositionD/A: digital-analogDAC: digital-to-analog converterDBR: distributed Bragg reflectorDC: directed currentDDA: differential difference amplifierDIL: dual in-lineDNA: desoxyribonucleic acidDRC: design rule checkDRIE: deep reactive-ion etchingEC: ethyl celluloseECE: electrochemical etch stopEDP: ethylenediamine/pyrochatecholEIS: electrolyte insulator semiconductor

220 Abbreviations

ENFET: enzyme-based field-effect transistorEPROM: electrically programmable read-only memoryFEM: finite-element methodFPI: Fabry-Perot interferometerFPW: flexural plate waveFTR: frustrated total reflectionGC gas chromatographyGDS: geometric data standardHDL: hardware description languageHF: hydrogen fluorideHNA: hydrofluoric acid, nitric acid, and acetic acidHRP: horseradish peroxidaseHVAC: humidity ventilation air-conditioningIC: integrated circuitI2C: inter integrated circuit, inter-ICICD: ion-controlled diodeIDT: interdigital transducer, interdigitated transducerIEEE: institute of electrical and electronics engineersIMFET: immuno field-effect transistorIO: integrated opticsIR: infraredISCAP: ion-sensitive capacitorISE: ion-selective electrodeISFET: ion-sensitive field-effect transistorIUPAC: international union of pure and applied chemistryIWAO: integrated-waveguide absorbance optodeKOH: potassium hydroxideLAPS: light-addressable potentiometric sensorLED: light-emitting diodeLEL: lower explosive limitLOD: limit of detectionLPCVD: low-pressure chemical vapor depositionLSB: least-significant bitLVS: layout versus schematicMEMS: micro-electro-mechanical systemsMOEMS: micro-opto-electro-mechanical systemsMOS: metal oxide semiconductorMOSCAP: metal oxide semiconductor capacitorMOSFET: metal oxide semiconductor field-effect transistorMUX: multiplexerNMOS: n-channel metal oxide semiconductorNTC: negative temperature coefficientOpamp: operational amplifierPBS: phosphate buffer solutionPCPMS: poly(cyanopropylmethylsiloxane)

Abbreviations 221

Pd: palladiumPDMS: poly(dimethylsiloxane)PECH: poly(epichlorohydrine)PECVD: plasma-enhanced chemical vapor depositionPEUT: poly(etherurethane)pH: potentia hydrogeniiPID: proportional integral differentialPMOS: p-channel metal oxide semiconductorppb: parts per billionppm: parts per millionPSD: photosensitive diodePt: platinumPTAT: proportional to absolute temperaturePTC: positive temperature coefficientPVD: physical vapor depositionPZT: lead zirconate titanateQ: qualityQMB: quartz microbalanceRE: reference electrodeREFET: reference field-effect transistorRGTO: rheotaxial growth and thermal oxidationRIE: reactive-ion etchingRIFS: reflectometric interference spectroscopyRMO: reverse mode of operationSAM: self-assembled monolayerSAW: surface acoustic waveSCL: serial clockSDA: serial dataSEM: scanning electron microscopySGFET: suspended-gate field-effect transistorSH-APM: shear-horizontal acoustic plate modeSi: siliconSNR: signal-to-noise ratioSOI: silicon on insulatorSPI: serial peripheral interfaceSPM: scanning-probe microscopySPR: surface plasmon resonanceSTW: shear tranverse waveTE: transverse electricTIRF: total internal reflectance fluorescenceTLV: threshold-limited valueTM: transverse magneticTMAH: tetramethyl ammonium hydroxideTO: transistor outlineTSMR: thickness-shear-mode resonator

222 Abbreviations

UV: ultravioletVCSEL: vertical-cavity surface-emitting laserVHDL: very-high-speed integrated-circuit hardware description languageVIS: visibleVOC: volatile organic compoundWE: working electrodeΣ∆ : Sigma-Delta converter

Index

absorption 9, 45absorption-emission process 46activation energy 40activity 10adsorption 9adsorption isotherm 13ammonium hydroxide 24amperometry 59amplitude grating 54analyte 2

concentrations 31condensation 107

anisotropic etching 19, 23dry 25wet 23

anodic bonding 26anti-aliasing filter 105application-specific integrated circuits

20atomic-force microscopy 37attenuated total reflection 47auxiliary electrode 60auxiliary functions 85auxiliary sensor 87

band bending 74bandgap voltage 170batch process 15bead resistor 40bimorph effect 37, 111BIOFET 71bioluminescence 55biosensor 3Bode-plot 123Bosch process 25Bragg reflector 52Brownian motion 120bulk micromachining 23

calorimetercircuitry 104gas sensing 105thermoelectric 166

cantilever 37, 166array 128capacitive crosstalk 116comparison to other mass-sensitive

devices 130deflection 38dynamic mode 38electrothermal actuation 115fabrication 111feedback circuitry 117fractional frequency shift 131magnetic actuation 113, 115, 118operation in water 122oscillator circuitry 117response

closed-loop 122open-loop 122

static mode 38temperature 115thermal actuation 110, 111thermal crosstalk 116

capacitance 84capacitance/voltage curve 73capacitive

current 61gas sensing 92microsystem circuitry 90sensor 83, 84, 166

carbon monoxide 81cardiac myocyte 190catalyst 40catalytic oxidation 40catalytic thermal sensors 40cell-based biosensor 190

224 Index

ceramics 16charge transfer 59CHEMFET 70chemical analysis system 1

miniaturized 188chemical equation 10chemical potential 11chemical reaction 13chemical sensor 4

definition 2chemical thermodynamics 10chemical vapor deposition 17chemisorption 9chemocapacitors 60, 83chemodiode 71chemoluminescence 46chemomechanical sensor 29chemoresistor 77

high-temperature 80low-temperature 78

chemotransistor 60, 67chip heating 157chlorine sensor 64chopper amplifier 105Clark cell 63

electrode 163oxygen sensor 165

CMOS 20amperometric sensor 62calorimetric transducer 100capacitive microsystem 88capacitive transducer 88chip 21design flow 172dielectric layers 101fabrication 173layout 172materials 86MEMS 85MEMS design 171microhotplate 133, 148multiparameter biochemical

microsystem 164multisensor system 163, 164operation modes 177resonant cantilever 109substrate 20technology 20

cold junction 43combustion 42compressional wave 34concentration cell 65condensation heat 43conductance 77conductance cell 77conducting polymer 78conductivity 77conductometric sensor 59, 76conductometry 59connectivity 87contact capacitance 78contact potential 13, 73contact resistance 78convection 61correlation spectroscopy 55Cottrell equation 62counter electrode 60cross-sensitivity 2current references 170current/potential curve 60cyclic voltammetry 62

decimation filter 105deep reactive-ion etching 25depletion 75deposition 17design rule check 172die-attach 182dielectric constant 88, 90dielectric properties 88dielectrometer 83difference frequency 33difference interferometer 50diffraction 54diffraction grating 54diffuse reflection 47diffusion 61diffusion length 62diffusion process 20digital controller 169digital temperature controller 155doping 20doping profile 20dosimeter 2drain 67dry etching 19dual-in-line (DIL) package 182

Index 225

electric field 89electrical potential 13electrical work 13electrochemical cell 59, 64electrochemical etch stop 146electrochemical potential 13electrochemical reaction 13electrochemical sensor 59electrode 59

impedance 78periodicity 94polarization 77

electrogenic cell 189, 190electrolyte insulator semiconductor

(EIS) 72electromagnet 118electromagnetic wave 45electromotive force 14electron affinity 73electron-beam evaporation 18electron-beam lithography 18electrophysiology 190elementary charge 13ellipsometry 47endothermic 39ENFET 70enthalpy 12, 39enthalpy change 39, 105entropy 12equilibrium 10equilibrium constant 10etch groove 24etching 19ethylene diamine/pyrochatechol 24evanescent field 47exchange current 65exothermic 39expansion coefficient 110external reflection 47

Fabry-Perot interferometer 53faradaic current 61Faraday constant 13feedback capacitors 91feedback loop 118Fermi level 13, 43field-effect 66field-effect transistor 67

ion-sensitive 66

metal oxide semiconductor 66flexural wave 35flexural-plate-wave device 35, 36

magnetic excitation 36flip-chip process 184fluorescence 46fluorophore 50four-electrode configuration 78frequency sweep 110frustrated total reflection 47fugacity 10fusion bonding 26

gallium arsenide SAW sensor 34gallium oxide 80Galvani potential 13, 43, 73gas sensor

handheld instrument 178reverse mode of operation 179

gate electrode 67Gibbs free energy 10Gibbs fundamental equation 12glass electrode 64glasses 16grain boundary 78grain size 81grating 54grating coupler 49, 50gravimetric sensor 30gravimetric transducer 109

heart cell 190heat 39, 105heat of condensation 107heat of mixing 107heater 133high-ε analyte 94hot junction 43hotplate 80hotplate microsystem analog 148

analog/digital 153digital array 159

humidity interference 98hybrid design 85hydrogel 79hydrogen 81hydrogen bond 9hydrogen sensor 68

226 Index

IMFET 71impedance 60indium oxide 80infinite dilution 12input impedance 61integrated circuit 15integrated optics 49integrated sensor 6integrated waveguide absorbance

optode 51interaction energies 9interaction mechanism 9interdigital capacitor 91interdigital transducers 32interdigitated electrodes 88internal reflection 47intersystem crossing 46ion implantation 20ion-controlled diode 71ion-selective capacitor 72ion-selective electrode 64

nonsymmetrical 64symmetrical 64

ionic conduction 81ISFET 68isotropic etching 19, 23

dry 24wet 23

Jablonski diagram 46

Kelvin probe 73, 74kinetic constant 11kinetics 10

Lamb wave 35Lamb-wave device 35Lambert-Beer relation 45layout-versus-schematic 172lift-off technique 19light 45light-addressable potentiometric sensor

75limit of detection (LOD) 2limit of determination 2linear solvation energy relationships

93lithium niobate 32lithium tantalate 32

logarithmic converter 153London dispersion 9Lorentz force 36, 113Love-wave 32low-ε analyte 94lower explosive limit 1, 40luminescence 46

Mach-Zehnder interferometer 51magnetic excitation 36magnetic field 36, 113magnetic flux 121mask aligner 18mass changes 30mass loading 30mass resolution 38mass-sensitive sensors 30membrane 133membrane buckling 136MEMS 6metal layers 21metal oxide 80, 133metal-oxide deposition 147metal-phthalocyanines 78Micro Electro Mechanical Systems,

MEMS 6microbridge 41microcontact printing 26microelectrode 190microelectrode array 190microfabrication 15microgaschromatographic column 189microhotplate 80, 133

circular 135design 134fabrication 145heater resistor 142modeling 139octagonal 135thermal simulation 139

microluminometer 56micromachining 22

post-CMOS 86pre-CMOS 86

microsensor fabrication sequence 88microsensor packaging 181microspectrometer 53microtechnology 15Mie scattering 46

Index 227

migration 61miniaturization 6, 187molar gas constant 12mole fraction 12monochromator 53monolithic CMOS-MEMS 85, 87MOS capacitor 72MOS diode 71MOSFET 67multicomponent analysis 130multisensor system 163

circuitry 168electrochemical 163gas tests 174

N-channel 67N-channel transistor 21N-type silicon 20N-well 21nanocrystalline material 81Nernst-equation 14neural probe 191neuron 190nickel/gold bumps 185nitrogen oxide 81NMOS 21noise spectrum 110NTC resistor 40

operational amplifier 60optical sensor 45optodes 49organic layers 4oscillation frequency 33oscillator 110oscillator circuitry 117outer potential 73

P-channel transistor 21P-type silicon 20package 87

chemically inert 181chip-on-board 182epoxy-based 181flip-chip 184

parasitic capacitance 89partial pressure 93partition coefficient 13, 93parts per million 31

patterning 18pellistor 40pH: “potentia hydrogenii” 66

change 70FET 69

phase grating 54phosphorescence 46photolithography 18photon 45photoresist 18phototransistor 53physical sensitivity 96physical vapor deposition 17physisorption 9, 107, 122piezoelectric materials 30, 37piezoelectricity 30piezoresistive coefficient 115piezoresistor 38, 111planar optical waveguide 49plasmon 57platform technology 85PMOS 21polarizable interface, 69polarization 46, 68, 72poly(dimethylsiloxane) 108poly(etherurethane) 90polymer 16

carbon-black-loaded 79conducting 78modulus 131swelling 92thick layer 94thin layer 94

polysilicon 21potassium hydroxide 24potentiometric sensor 59, 64potentiometry 59potentiostat 60power dissipation 115ppm-unit 31preconcentrator 188Pt-temperature resistor 138PTC resistor 40pyroelectric effect 40

quality factor 110, 112quantum efficiency 49quartz microbalance 31

228 Index

radiation 45radiation energy 45Raman scattering 46rate constant 10Rayleigh surface-acoustic-wave 31Rayleigh surface-acoustic-wave device

31, 130reactants 10reaction 10reaction rate constant 10reactive-ion etching 20reference capacitor 91reference electrode 60, 68reference oscillator 33reflection 47reflectometric interference spectroscopy

47refraction 47refraction index 47refractivity 47resistance 77resistivity 77resonance frequency 30resonating cantilever 37reverse current 72reverse mode of operation 179, 180reversibility 2

sacrificial layer 25saturation vapor pressure 93Sauerbrey equation 30, 130scattering 46Seebeck coefficient 43Seebeck-effect 43, 100selectivity 2semiconductor processing 16semiconductor technology 6sensing capacitor 91sensitive layer 3, 4sensitive materials 5sensitivity 2sensor array 4sensor requirements 5sensor system simulation 172serial interface 168Severinghaus electrode 79Severinghaus pH-FET 165SGFET 73

shear-horizontal-acoustic-plate-mode32

shear-transverse-wave 32sigma-delta modulator 91silicon 16silicon island 135silicon nitride 21silicon oxide 21smart sensor 6soft lithography 26solder 185source-drain current 68, 69specific conductance 77specific mass density 125specificity 2spectrometer 54spectroscopy 45specular external reflection 47spin-coating 18spray-coating 18sputtering 17state function 2, 10stoichiometric number 10stripping voltammetry 62surface acoustic wave 32surface dipole 74surface micromachining 25surface plasmon resonance 48, 57suspended gate 74suspended-gate field-effect transistor

73switched-capacitor 91system architecture 85

temperature coefficient 139negative 40positive 40

temperature homogeneity 134, 137temperature sensor 170

calibration 139platinum 138polysilicon 150PTAT (proportional to ambient

temperature) 150sigma-delta 170

tetramethyl ammonium hydroxide 24thermal conductivity 39thermal evaporation 17thermal mass 81

Index 229

thermal resistance 103, 134thermal sensor 39thermal time constant 133, 142thermistor 40thermocouples 102, 103thermodynamic equilibrium 10thermodynamics 10, 11thermoelectric 43thermoelectric effect 100thermopile 43thermovoltage 43, 102thickness-shear-mode resonator 31,

130three-electrode-system 60threshold-limited value 1tin dioxide 80tin dioxide, nanocrystalline 135toolbox strategy 85total internal reflection 47total internal reflectance fluorescence

47transducer 3transducer periodicity 33transient behavior 39transient signal 106transistor microhotplate 143transistor outline (TO) socket 182transistor, stress-sensitive 115transmission 47

transverse electric 49transverse magnetic 49triplet state 46two-electrode configuration 60, 78

underetching 24underfill 185

vacuum level 73vaporization heat 43vertical-cavity surface-emitting laser

52viscoelastic properties 31, 33Volta potential 73voltammetric sensor 59, 60voltammetry 59volume fraction 93

Wafer bonding 26waveguide 48wavelength 45wet etching 19Wheatstone bridge 114work function 73working electrode 60

Xenon difluoride 24

Zinc oxide 32

Printing: Strauss GmbH, Mörlenbach

Binding: Schäffer, Grünstadt

microtechnology and mems · 2019-07-19 · microtechnology and mems series editor: h. baltes h....

Documents