Integrated Chemical Microsensor Systems in CMOS Technology (Microtechnology and MEMS)

microtechnology and mems microtechnology and mems Series Editor: H. Baltes H. Fujita D. Liepmann The series Microt...

Author: Andreas Hierlemann

126 downloads 908 Views 4MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form

DOWNLOAD PDF

microtechnology and mems

microtechnology and mems Series Editor: H. Baltes

H. Fujita

D. Liepmann

The series Microtechnology and MEMS comprises text books, monographs, and state-of-the-art reports in the very active field of microsystems and microtechnology. Written by leading physicists and engineers, the books describe the basic science, device design, and applications. They will appeal to researchers, engineers, and advanced students. Mechanical Microsensors By M. Elwenspoek and R. Wiegerink CMOS Cantilever Sensor Systems Atomic Force Microscopy and Gas Sensing Applications By D. Lange, O. Brand, and H. Baltes Micromachines as Tools for Nanotechnology Editor: H. Fujita Modelling of Microfabrication Systems By R. Nassar and W. Dai Laser Diode Microsystems By H. Zappe Silicon Microchannel Heat Sinks Theories and Phenomena By L. Zhang, K.E. Goodson, and T.W. Kenny Shape Memory Microactuators By M. Kohl Force Sensors for Microelectronic Packaging Applications By J. Schwizer, M. Mayer and O. Brand Integrated Chemical Microsensor Systems in CMOS Technology By A. Hierlemann

A. Hierlemann

Integrated Chemical Microsensor Systems in CMOS Technology With 125 Figures

123

Professor Dr. Andreas Hierlemann Physical Electronics Laboratory ETH Hoenggerberg, HPT-H 4.2, IQE 8093 Zurich Switzerland Email: [email protected]

Series Editors: Professor Dr. H. Baltes ETH Zürich, Physical Electronics Laboratory ETH Hoenggerberg, HPT-H6, 8093 Zürich, Switzerland

Professor Dr. Hiroyuki Fujita University of Tokyo, Institute of Industrial Science 4-6-1 Komaba, Meguro-ku, Tokyo 153-8505, Japan

Professor Dr. Dorian Liepmann University of California, Department of Bioengineering 466 Evans Hall, #1762, Berkeley, CA 94720-1762, USA

ISSN 1439-6599 ISBN 3-540-23782-8 Springer Berlin Heidelberg New York Library of Congress Control Number: 2004114045 This work is subject to copyright. All rights are reserved, whether the whole or part of the material is concerned, specif ically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microf ilm or in any other way, and storage in data banks. Duplication of this publication or parts thereof is permitted only under the provisions of the German Copyright Law of September 9, 1965, in its current version, and permission for use must always be obtained from Springer. Violations are liable to prosecution under the German Copyright Law. Springer is a part of Springer Science+Business Media springeronline.com © Springer-Verlag Berlin Heidelberg 2005 Printed in Germany The use of general descriptive names, registered names, trademarks, etc. in this publication does not imply, even in the absence of a specif ic statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. Typesetting: by the authors and TechBooks using a Springer LATEX macro package Cover concept: eStudio Calamar Steinen Cover production: design & production GmbH, Heidelberg Printed on acid-free paper

57/3141/jl - 5 4 3 2 1 0

Preface

This book provides a comprehensive treatment of the very interdisciplinary field of CMOS technology-based chemical microsensor systems. It is, on the one hand, targeted at scientists and engineers interested in getting first insights in the field of chemical sensing since all necessary fundamental knowledge is included. On the other hand, it also addresses experts in the field since it provides detailed information on all important issues related to realizing chemical microsensors and, specifically, chemical microsensors in CMOS technology. A large number of sensor and integrated-sensor-system implementations illustrate the current state of the art and help to identify the possibilities for future developments. Since microsensors produce “microsignals”, sensor miniaturization without sensor integration is in many cases prone to failure. This book will help to reveal the benefits of using integrated electronics and CMOS-technology for developing chemical microsensor systems and, in particular, the advantages that result from realizing monolithically integrated sensor systems comprising transducers and associated circuitry on a single chip. After a brief introduction, the fundamentals of chemical sensing are laid out, including a short excursion into the related thermodynamics and kinetics. Fabrication and processing steps that are commonly used in semiconductor industry are then abstracted. These more fundamental sections are followed by a short description of microfabrication techniques and the CMOS substrate and materials. Thereafter, a comprehensive overview of semiconductorbased and CMOS-based transducer structures for chemical sensors is given. The corresponding chemically sensitive materials and the related applications are mentioned in the context of each transducer structure. CMOS-technology is then introduced as platform technology, which allows the fabrication of microtransducers and, moreover, enables the integration of these microtransducers with the necessary driving and signal conditioning circuitry on the same chip. Several examples such as microcapacitors, microcalorimeters, microcantilevers, and microhotplates are described in great detail. In a next step, the development of monolithic multisensor arrays and fully developed microsystems with on-chip sensor control and standard interfaces is depicted. A short section on packaging shows that techniques from the semiconductor industry can also be applied to chemical microsensor packaging. The book

VI

Preface

concludes with a short outlook to future developments such as developing more complex integrated microsensor systems and interfacing biological materials such as cells with CMOS microelectronics. As with all interdisciplinary efforts, teamwork plays a central role in being successful. Therefore I am particularly grateful to many colleagues and former students, who contributed much to the work that is the topic of this book. I would like to thank Prof. Henry Baltes for giving me the opportunity and the support to enter in the field of CMOS-based sensors in his laboratory. I very much appreciated his continual interest in discovering new things and exploring new fields of science. I am also very grateful to Prof. Oliver Brand, who was always a valuable source of information on microtechnology and microfabrication. I am very much obliged to several highly motivated and excellent coworkers, whose work is amply cited in this book: Christoph Hagleitner and Kay-Uwe Kirstein, the chief circuit designers, the microhotplate group: Markus Graf, Diego Barrettino, Stefano Taschini, Urs Frey, and Martin Zimmermann, the guys working on cantilevers: Dirk Lange, Cyril Vancura, Yue Li, Jan Lichtenberg, the capacitor freaks: Andreas Koll, Adrian Kummer, the microcalorimeter people: Nicole Kerness and Petra Kurzawski, and, finally, Wan Ho Song, who did the microsensor packaging. In the outlook some first results on the combination of microelectronics and cells are mentioned. These rely on the work of Flavio Heer, Wendy Franks, Sadik Hafizovic, Robert Sunier, and Frauke Greve. I am very grateful for all their efforts, and I am looking forward to exciting new results in this research area. I am also indebted to European collaboration partners, Udo Weimar and Nicolae Barsan, University of T¨ ubingen, and to AppliedSensor GmbH, Reutlingen, who provided many of the chemically sensitive materials such as the metal oxides. The fruitful collaboration with Sensirion AG, Z¨ urich, namely Felix Mayer and Mark Hornung, is also gratefully acknowledged. Financial support for the CMOS chemical-sensor projects came from the European Union (FP5, FP6, IST-program), the Swiss Bundesamt f¨ ur Bildung und Wissenschaft (BBW), the Swiss Commission for Technology and Innovation (CTI), and the K¨ orber Foundation, Hamburg, Germany. Zurich, September 2004

Andreas Hierlemann

Contents

1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1

2

Fundamentals of Chemical Sensing . . . . . . . . . . . . . . . . . . . . . . .

9

3

Microtechnology for Chemical Sensors . . . . . . . . . . . . . . . . . . . . 3.1 Microtechnology Substrate Materials . . . . . . . . . . . . . . . . . . . . . . 3.2 Fundamental Semiconductor Processing Steps . . . . . . . . . . . . . . 3.2.1 Deposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.2 Patterning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.3 Etching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.4 Doping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3 CMOS Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.4 Microfabrication for Chemical Sensors . . . . . . . . . . . . . . . . . . . . . 3.4.1 Micromachining for Chemical Microsensors . . . . . . . . . . 3.4.1.1 Bulk Micromachining . . . . . . . . . . . . . . . . . . . . . 3.4.1.2 Surface Micromachining . . . . . . . . . . . . . . . . . . . 3.4.2 Wafer Bonding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.4.3 Sensitive-Layer Deposition . . . . . . . . . . . . . . . . . . . . . . . . .

15 16 16 17 18 19 20 20 22 22 23 25 26 26

4

Microfabricated Chemical Sensors . . . . . . . . . . . . . . . . . . . . . . . . 4.1 Chemomechanical Sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1.1 Rayleigh SAW Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1.2 Flexural-Plate-Wave or Lamb-Wave Devices . . . . . . . . . 4.1.3 Resonating Cantilevers . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2 Thermal Sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.1 Catalytic Thermal Sensors (Pellistors) . . . . . . . . . . . . . . 4.2.2 Thermoelectric or Seebeck-Effect Sensors . . . . . . . . . . . 4.3 Optical Sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.1 Integrated Optics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.2 Microspectrometers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.2.1 Fabry-Perot-Type Structures . . . . . . . . . . . . . . . 4.3.2.2 Grating-Type Structures . . . . . . . . . . . . . . . . . . . 4.3.3 Bioluminescent Bioreporter Integrated Circuits (BBIC) 4.3.4 Surface Plasmon Resonance (SPR) Devices . . . . . . . . . . 4.4 Electrochemical Sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

29 29 32 35 37 39 40 43 45 49 53 53 54 55 57 59

VIII

Contents

4.4.1 Voltammetric Sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4.2 Potentiometric Sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4.2.1 Electrochemical Cell . . . . . . . . . . . . . . . . . . . . . . 4.4.2.2 Field-Effect-Based Devices . . . . . . . . . . . . . . . . 4.4.2.2.1 MOS Field-Effect Transistors, MOSFETs, and Ion-Selective FieldEffect Transistors, ISFETs (Chemotransistors) . . . . . . . . . . . . . . 4.4.2.2.2 MOS Diode and Ion-Controlled Diode, ICD (Chemodiodes) . . . . . . . 4.4.2.2.3 MOS Capacitor and Ion-Selective Capacitor (Chemocapacitors) . . . . . . 4.4.2.2.4 Measuring Work Functions: Kelvin Probe and Suspended-Gate Field-Effect Transistor, SGFET . . . . . . . . . . . . . . 4.4.2.2.5 Light-Addressable Potentiometric Sensor, LAPS . . . . . . . . . . . . . . . . . . . 4.4.2.2.6 Field-Effect Device Fabrication . . . . 4.4.3 Conductometric Sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4.3.1 Chemoresistors . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4.3.1.1 Low-Temperature Chemoresistors . . 4.4.3.1.2 High-Temperature Chemoresistors (Hotplate Sensors) . . . . . . . . . . . . . . . 4.4.3.2 Chemocapacitors . . . . . . . . . . . . . . . . . . . . . . . . . 5

CMOS Platform Technology for Chemical Sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1 CMOS Capacitive Microsystems . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1.1 CMOS Capacitive Transducer . . . . . . . . . . . . . . . . . . . . . . 5.1.2 On-Chip Circuitry of the Capacitive Microsystem . . . . 5.1.3 Capacitive Gas Sensing . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1.3.1 Selectivity Through Sensitive Layer Thickness 5.1.3.2 Insensitivity to Low-ε Analytes . . . . . . . . . . . . . 5.1.3.3 Humidity Interference . . . . . . . . . . . . . . . . . . . . . 5.2 CMOS Calorimetric Device . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2.1 CMOS Calorimetric Transducer . . . . . . . . . . . . . . . . . . . . 5.2.2 Calorimeter Circuitry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2.3 Calorimetric Gas Sensing . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3 CMOS Integrated Resonant Cantilever . . . . . . . . . . . . . . . . . . . . 5.3.1 Resonant Cantilever Transducers . . . . . . . . . . . . . . . . . . . 5.3.1.1 Thermal Actuation . . . . . . . . . . . . . . . . . . . . . . . 5.3.1.2 Magnetic Actuation . . . . . . . . . . . . . . . . . . . . . . . 5.3.1.3 Vibration Detection . . . . . . . . . . . . . . . . . . . . . . .

60 64 64 66

67 71 72

73 75 76 76 77 78 80 83

85 88 88 90 92 94 97 98 100 100 104 105 109 110 110 113 114

Contents

5.3.1.4 Cantilever Temperature . . . . . . . . . . . . . . . . . . . 5.3.2 Microcantilever Circuitry . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3.2.1 Thermal Actuation . . . . . . . . . . . . . . . . . . . . . . . 5.3.2.2 Magnetic Actuation . . . . . . . . . . . . . . . . . . . . . . . 5.3.3 Microcantilevers as Chemical Sensors . . . . . . . . . . . . . . . 5.3.3.1 Polymer Coating . . . . . . . . . . . . . . . . . . . . . . . . . 5.3.3.2 Analyte Absorption . . . . . . . . . . . . . . . . . . . . . . . 5.3.4 Comparison of Cantilevers to Other Mass-Sensitive Devices . . . . . . . . . . . . . . . . . . . . 5.4 CMOS Microhotplate System Development . . . . . . . . . . . . . . . . 5.4.1 CMOS Microhotplates . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.4.1.1 Temperature Sensor Calibration . . . . . . . . . . . . 5.4.1.2 Thermal Microhotplate Modeling and Characterization . . . . . . . . . . . . . . . . . . . . . . 5.4.1.3 Microhotplate Heaters: Resistor and Transistor . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.4.1.4 Microhotplate Sensor Fabrication . . . . . . . . . . . 5.4.2 Hotplate-Based CMOS Monolithic Microsystems . . . . . 5.4.2.1 Analog Hotplate Microsystem . . . . . . . . . . . . . . 5.4.2.2 Analog/Digital Hotplate Microsystem . . . . . . . 5.4.2.3 Digital Hotplate Array Microsystem . . . . . . . . . 5.5 CMOS Chemical Multisensor Systems . . . . . . . . . . . . . . . . . . . . . 5.5.1 CMOS Multiparameter Biochemical Microsystem . . . . . 5.5.2 CMOS Gas-Phase Multisensor System . . . . . . . . . . . . . . 5.5.2.1 Multisystem Architecture . . . . . . . . . . . . . . . . . . 5.5.2.2 Multisystem Circuitry Components, Design and Fabrication . . . . . . . . . . . . . . . . . . . . . . . . . . 5.5.2.3 Multisystem Gas Sensor Measurements . . . . . . 5.5.2.4 Multisystem Applications and Operation Modes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.6 CMOS Chemical Microsensor and System Packaging . . . . . . . . 5.6.1 Simple Epoxy-Based Package . . . . . . . . . . . . . . . . . . . . . . 5.6.2 Chip-on-Board Package . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.6.3 Flip-Chip Package . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

IX

115 116 116 118 122 122 124 130 133 133 139 139 142 145 148 148 153 159 163 164 165 166 168 174 177 181 181 182 184

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

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193 Abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219 Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223

1 Introduction

The detection of molecules or chemical compounds is a general analytical task in the efforts of chemists to obtain qualitative and/or quantitative time- and spatially resolved information on specific chemical components [1]. Examples of qualitative information include the presence or absence of certain odorant, toxic, carcinogenic or hazardous compounds. Examples of quantitative information include concentrations, activities, or partial pressures of such specific compounds exceeding, e.g., a certain threshold-limited value (TLV), or the lower explosive limits (LEL) of combustible gases. All this information can, in principle, be obtained from either a chemical analysis system or alternatively by using chemical sensors. In both cases sampling, sample pretreatment, separation of the components and data treatment are the tasks to be fulfilled. The main components of a state-of-the-art chemical analysis or sensor system are depicted schematically in Fig. 1.1.

cA yi

xi

x1 x2 x3 . . .

comparison with calibration data

cA

result: chemical composition (quantitative or qualitative)

pattern recognition multicomponent analysis

feature extraction

data pretreatment (sensor electronics)

detection (chemical sensors )

sample conditioning (catalyst, enzyme...)

filtering (membrane)

sample uptake

sample (gas, liquid...)

Fig. 1.1. Components of a chemical analysis or sensor system. Adapted from [15]

2

1 Introduction

It is not easy to clearly distinguish between a chemical sensor and a complex analytical system. Integrated or miniaturized chromatographs or spectrometers may be denoted chemical sensors as well. However, a typical chemical sensor is, in most cases, a cheaper, smaller, and less complex device as compared to miniaturized analytical systems. A draft of the IUPAC (International Union of Pure and Applied Chemistry) provides a definition of a chemical sensor [2]: “A chemical sensor is a device that transforms chemical information, ranging from the concentration of a specific sample component to total composition analysis, into an analytically useful signal”. This rather wide definition does not require that the sensor is continuously operating and that the sensing process is reversible. But intermittently operating devices exhibiting irreversible characteristics are usually referred to as dosimeters [3]. In this context it is useful to introduce some important keywords used extensively throughout the chemical sensor literature [1, 4–11]. Reversibility Thermodynamic reversibility, strictly speaking, requires that the sensor measurand is related to a thermodynamic state function. This implies that, e.g., a certain sensor response unequivocally corresponds to a certain analyte concentration (analyte here denotes the chemical compound to be monitored). The sensor signal may not depend on the history of previous exposures or how a certain analyte concentration is reached (no memory effects or hysteresis). More details on fundamental thermodynamics of the chemical sensing process will be given in Chap. 2. Sensitivity and Cross-Sensitivity Sensitivity usually is defined as the slope of the analytical calibration curve, i.e., how largely the change in the sensor signal depends upon a certain change in the analyte concentration. Cross-sensitivity hence refers to the contributions of compounds other than the desired compound to the overall sensor response. Selectivity/Specificity Selectivity or specificity can be defined according to Janata [4] as the ability of a sensor to respond primarily to only one species in the presence of other species (usually denoted as interferants). Limit of Detection and Limit of Determination The limit of detection (LOD) corresponds to a signal equal to k-times the standard deviation of the background noise (i.e., k represents the signal-tonoise ratio) with a typical value of k = 3. Values above the LOD indicate the presence of an analyte, whereas values below LOD indicate that no analyte is detectable. The limit of determination implies qualitative information, i.e., that the signal can be attributed to a specific analyte. This in turn requires more information and, therefore, the limit of determination is always higher than the limit of detection.

1 Introduction

3

Transducer Transducer is derived from Latin “transducere”, which means to “transfer or translate”. Therefore, a device that translates energy from one kind of system (e.g., chemical) to another (e.g., physical) is termed a transducer. Biosensor Biosensors are usually considered a subset of chemical sensors that make use of biological or living material for their sensing function [10, 11]. Since this book covers mostly chemical sensors, there will not be any further diversification into chemo- and biosensors within this work. Using the above definitions, chemical sensors usually consist of a sensitive layer or coating and a transducer. Upon interaction with a chemical species (absorption, chemical reaction, charge transfer etc.), the physicochemical properties of the coating, such as its mass, volume, optical properties or resistance, reversibly change (Fig. 1.2).

analyte molecules

oscillator circuit

frequency signal, ∆f

physical measurand

data recording and processing

microbalance

driving circuitry

(physico-) chemical interaction

polymer

transducer

mass change ∆m

sensitive layer

example:

chemical sensor

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

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

4

1 Introduction

To supply the different needs in chemical sensing, a variety of transducers based on different physical principles has been devised. Following the suggestion of Janata [4, 5], chemical sensors can be classified into four principal categories according to their transduction principles: 1. 2. 3. 4.

Chemomechanical sensors (e.g., mass changes due to bulk absorption) Thermal sensors (e.g., temperature changes through chemical interaction) Optical sensors (e.g., changes of light intensity by absorption) Electrochemical sensors (e.g., changes of potential or resistance through charge transfer)

Each of those four categories of chemical sensors will be treated in great detail in Chap. 4. An overview of more recent literature on chemical sensors with regard to different transduction principles is given in [5]. Various inorganic and organic materials serve as chemically sensitive layers that can be coated onto the different transducers. Typical inorganic materials include metal oxides like tin dioxide (SnO2 ) for monitoring reducing gases such as hydrogen or carbon monoxide, or zirconium dioxide (ZrO2 ) to detect oxygen, nitrogen oxide, and ammonia. Organic layers mostly consisting of polymers such as polysiloxanes or polyurethanes are used to monitor hydrocarbons, halogenated compounds and different toxic volatile organics. A survey of typical chemically sensitive materials and their applications is given in Table 1.1. Further information on the coating materials will be provided, e.g., in the context of the different transducers in Chap. 4. Current research and development work in chemical sensors and sensitive materials evolves in three main directions: 1. Miniaturization and monolithic integration of transducers with electronics and, possibly, auxiliary sensors. 2. Search for highly selective (bio)chemical layer materials (molecular recognition, key-lock-type interactions). 3. Using arrays of sensors exhibiting different partial selectivity (polymers, metal oxides) and developing pattern recognition (odors, aromas) and multicomponent analysis methods (mixtures of gases and liquids). The latter strategy has grown very popular [12–17], especially since compact sensor arrays can presently be fabricated at low costs, and interferants, which are present in almost any practical application, can be handled. Chemical sensors meanwhile have also reached the stage of exploratory use in a variety of industrial and environmental applications, some examples being quality control or on-line process monitoring in the food-industry as well as preliminary tests in the areas of medical practice and personal (workplace) safety [18]. In particular in environmental monitoring, there is an urgent need for low-cost sensor systems detecting various pollutants at trace level.

1 Introduction

5

Table 1.1. Typical sensitive materials and applications Materials metals

Examples

Applications

Pt, Pd, Ni, Ag, Sb, Rh, . . .

inorganic gases like CH4 ,H2 , . . .

ionic compounds

electronic conductors (SnO2 , TiO2 , Ta2 O5 , In2 O3 , AlVO4 , . . . ) mixed conductors (SrTiO3 ,Ga2 O3 , perowskites, . . . ) ionic conductors (ZrO2 , LaF3 , CeO2 , nasicon, . . . )

inorganic gases (CO, NOx , CH4 . . . ) exhaust gases, oxygen, ions in water, . . .

molecular crystals

phthalocyanines (Pcs): PbPc, LuPc2 , . . .

nitrogen dioxide, volatile organics

Langmuir-Blodgett films cage compounds

lipid bilayers, polydiacetylene . . . zeolites, calixarenes, cyclodextrins, crown ethers, cyclophanes, . . .

organic molecules in medical applications, biosensing, . . . water analysis (ions), volatile organics, . . .

polymers

components of biological entities

nonconducting polymers detection of volatile organics, polyurethanes, polysiloxanes, . . . food industry (odor conducting polymers and aroma), environmental polypyrroles, polythiophenes, monitoring in gas nafion, . . . and liquid phase, . . . synthetic entities phospholipids, lipids, HIVepitopes, . . . natural entities enzymes, receptors, proteins, cells, membranes, . . .

medical applications, biosensing, water and blood analysis, pharmascreening, . . .

Key requirements for a successful chemical sensor include: • • • • • • • • • • • •

High sensitivity and low limit of detection (LOD) High selectivity to target analyte and low cross-sensitivity to interferants Short recovery and response times Large dynamic range Reversibility Accuracy, precision and reproducibility of the signal Long-term stability and reliability (self-calibration) Low drift Low temperature dependence or temperature compensation mechanisms Ruggedness Low costs (batch fabrication) and low maintenance Ease of use

6

1 Introduction

Semiconductor technology provides excellent means to effectively realize device miniaturization and to meet some of the chemical-sensor key criteria listed above (low cost, batch fabrication). The rapid development of the integrated-circuit (IC) technology during the past decades has initiated many initiatives to fabricate chemical sensors consisting of a chemically sensitive layer on a signal-transducing silicon chip [19,20]. The earliest types of chemical sensors realized in silicon technology were based on field-effect transistors (FETs) [21, 22]. Reviews of silicon-based sensors (not only chemical sensors) are given in [23–25]. In this context two more keywords have to be introduced here. Integrated Sensor A sensor is denoted an integrated sensor if the chemical sensing operation is based on a direct influence on an electric component (resistor, transistor, capacitor) integrated in silicon or another semiconductor material [11]. Smart or Intelligent Sensor The combination of interface electronics and an integrated sensor on a single chip results in a so-called “smart sensor”. At least some basic signal conditioning is usually carried out on chip. One major advantage of smart sensors is the improved signal-to-noise and electromagnetic interference characteristics [11]. In addition the connectivity problem, which occurs especially in multisensor arrays, can be eased by using on-chip multiplexers and by using bus interfaces. For more details on sensor system integration, see Chap. 5. The largely planar integrated-circuit (IC) and chemical-sensor structures processed by combining lithographic, thin film, etching, diffusive and oxidative steps have been recently extended into the third dimension using microfabrication technologies (see Chap. 3 in this book). A variety of micromechanical structures including cantilever beams, suspended membranes, freestanding bridges, gears, rotors, and valves have been produced using micromachining technology (MicroElectroMechanicalSystems MEMS ) [26–29]. MEMS technology thus provides a number of key features, which can serve to enhance the functionality of chemical sensor systems [9, 11, 26, 29–34]. Micromechanical structures (MEMS-structures) and microelectronics can be realized on a single chip allowing for on-chip control and monitoring of the mechanical functions as well as for data preprocessing such as signal amplification, signal conditioning, and data reduction [29–34]. ComplementaryMetal-Oxide-Semiconductor or CMOS-technology is the dominant semiconductor IC technology for microprocessors and Application-Specific Integrated Circuits (ASICs) and has also been used to fabricate integrated chemical microsensors. The use of CMOS technology entails a limited selection of device materials (see Sect. 3.3) and a predefined fabrication process for the CMOS part. Sensor-specific or transducer-specific materials and fabrication steps have to be introduced in most cases as post-processing after the CMOS fabrication.

1 Introduction

7

In the next chapters the fundamentals of the chemical sensing process itself will be laid out (Chap. 2) followed by a short description of microfabrication techniques and the CMOS substrate (Chap. 3). In Chap. 4, there will be an extensive treatment of the different microtransducers that are commonly used for chemical sensors. This transducer overview will be restricted to semiconductor-based and CMOS-based devices and will, for the sake of completeness, also include short abstracts on devices, which are described in much more detail in the subsequent Chap. 5 on the CMOS technology platform for chemical sensors. Chapter 5 will show the evolution from single transducers, which are integrated with the necessary driving and signal conditioning circuitry to monolithic multisensor arrays and fully developed systems with on-chip sensor control and standard interfaces. The concluding Chap. 6 will include a short glance at future developments such as combining cells and CMOS devices to develop biosensors or bioelectric interfaces.

2 Fundamentals of Chemical Sensing

The interaction of a chemical species with a chemical sensor can either be confined to the surface of the sensing layer, or it can take place in the whole volume of the sensitive coating. Surface interaction implies that the species of interest is adsorbed at the surface or interface (gas/solid or liquid/solid) only, whereas volume interaction requires the absorption of the species and a partitioning between sample phase and the bulk of the sensitive material. The different types of chemical interactions involved in a sensing process range from very weak physisorption through rather strong chemisorption to charge transfer and chemical reactions. Physisorption in this context implies that the compound is only physically ab/adsorbed (London or Van-der-Waals dispersion forces) with an interaction energy of 0–30 kJ/mol, whereas in the case of the much stronger chemisorption (interaction energy >120 kJ/mol), the particles stick to the surface by forming a chemical (usually covalent) bond. Charge transfer and chemical reactions involve, in most cases, interaction energies comparable to those of chemisorption and higher. Some of the most common interaction mechanisms and associated energies are listed in Table 2.1, for further details, see [35]. Table 2.1. Typical intermolecular interactions and energies Interaction type covalent bond ion-ion coordination, complexation, charge-transfer bonding

Typical energy [kJ/mol]

Comment

120–800

chemical reaction

250

only between ions

8–200

weak “chemical” interaction

ion-dipole

15

hydrogen bond

20

hydrogen bond: A–H· · ·B

dipole-dipole

0.3–2

between polar molecules

London dispersion (induced dipole-induced dipole)

0.1–2

physical interaction between any molecules

10


High chemical selectivity and rapid reversibility place contradictory constraints on desired interactions between chemical sensor coating materials and analytes. Low-energy, perfectly reversible (physisorptive) interactions generally lack high selectivity, while chemisorptive processes, the strongest of which result in the formation of new chemical bonds, offer selectivity, but are inherently less reversible. A practicable compromise has to be achieved with due regard to the specific application. In this context it should be noted that the commonly accepted limit of reversibility up to 20 kJ/mol refers to room temperature and will not apply to the case of, e.g., tin-dioxide–coated semiconductor sensors operated at 300◦ C to 400◦ C. On the other hand, spontaneous chemical reactions occurring at room temperature often require a tedious regeneration of, e.g., biological recognition units (enzymes). Any interaction between a coating material and an analyte is governed by chemical thermodynamics and kinetics. Thus, a fundamental thermodynamic function, the Gibbs free energy, G [J], is the most important descriptor in all chemical sensing processes: The direction of spontaneous reactions is always towards lower values of G (minimization of the Gibbs energy). The Gibbs free energy is a state function in the thermodynamic sense, i.e., its value depends only on the current state of the system and is independent of how that state has been prepared. This implies that any chemical (sensing) process described by a Gibbs energy function moves towards a dynamic equilibrium (∆G = 0, G minimal), in which both reactants and products are present but have no tendency to undergo net change. This equilibrium is reversible, i.e., an infinitesimal change in the conditions in opposite directions results in opposite changes in its state. The interaction equilibrium of an analyte, A, with a sensor coating, S, can thus be represented by: →

k

A···S . A+S⇔ ←

(2.1)

k

→

←

Here, k and k denote the rate constants of the forward reaction and the reverse reaction, which will be detailed below. Such equilibrium can be described by an equilibrium constant, K, which relates the activity, a, of reaction products (A· · ·S) to those of the reactants (A and S). This constant is thus a characteristic value for the progression of the reaction (K ≤ 1 : no reaction takes place), its numerical value depends on the system temperature. aA···S and in general : K = ani i . (2.2) K= aA · aS i The index i denotes the chemical substance, ni are the corresponding stoichiometric numbers in the chemical equation. This expression signifies that each activity (or fugacity) is raised to the power equal to its stoichiometric number, and, then, all such terms are multiplied together. Stoichiometric numbers of the products are positive and those of the reactants are negative, i.e., reactants appear as the denominator and reaction products as the


11

numerator. The activity1 , ai (fugacity, fi , for gases), which denotes the effective quantity of compound i participating in, e.g., a chemical reaction, is related to the mole fraction, xi , (partial pressure, p, for gases) of a species via: aA = γA · xA with γI ≤ 1. The activity coefficient, γ i , measures the degree of departure of a components behavior from ideal or ideally dilute behavior. The equilibrium constant, K, is also related to kinetics. For the simple → reaction in (2.1), two kinetic constants can be defined: k for the reaction ←

leading to the product A· · ·S, and k for the reaction in the opposite direction. → ← daA = − k aA aS + k aS···A . dt

(2.3)

K then represents the ratio of those two kinetic constants in equilibrium state. → k (2.4) K= ←. k Both, thermodynamics and kinetics hence affect the progress of any chemical process or reaction. Thermodynamics, namely the Gibbs free energy (minimum) or the equilibrium constant, can tell us the direction of spontaneous change and the composition at the equilibrium state, whereas kinetics tell us, whether a kinetically viable pathway exists for that change to occur, and how fast an equilibrium state will be achieved. Kinetics are important in the context of chemical sensors, since there exist chemical processes, the activation barrier of which is too high to get a reaction going, although the Gibbs free energy of the products would be below that of the reactants. Such effects can be used to advantage in tuning the selectivity of, e.g., catalytic chemical sensors (see, e.g., Sect. 4.4.3.1.2). A chemical potential has been introduced in thermodynamics. The chemical potential shows how the Gibbs energy of a system changes when a portion of a specific chemical compound is added to it or removed from it. The chemical potential of the i-th component, µi , is defined as: ∂G and ∆G = µi dni . (2.5) µi = ∂ni p,T i 1

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

12


Here ni denotes the stoichiometric number or the amount of substance in moles. Again, the stoichiometric numbers of the products are positive and those of the reactants negative. The pressure, p, and the temperature, T , are kept constant. The chemical potential can be expressed in terms of mole fractions, xi , or activities, ai , in liquids, and partial pressures, pi , or fugacities, fi , in the gas phase [35]: µi = µ0i (p, T ) + RT ln ai

or

µi = µ0i (p, T ) + RT ln fi .

(2.6)

µ0i (p, T ) here denotes the chemical potential of an appropriately defined standard state such as, e.g., “infinite dilution” or a “pure compound”; R is the molar gas constant (8.314 J/Kmol) and T denotes the temperature in [K]. So there are two terms in (2.6), a reference term and an activity-dependent term. Plugging the terms of (2.6) into (2.5), the reference terms (µ0i ) can be subsumed into ∆G0 as shown in the following equation: ni µ0i (p, T ) + RT ni ln ai = ∆G0 (p, T ) + RT ln ani i . (2.7) ∆G = i

i

i

Both, ∆G and K are characteristic descriptors for the direction of a chemical reaction. In comparing (2.2) with (2.7), it is evident that in a thermodynamic equilibrium state (∆G = 0) ∆G0 and K are interrelated via the following equation (for details, see [35]): ln K = −

∆G0 . RT

(2.8)

The more negative ∆G0 , the larger is K, or in other words, the higher the chemical potential of the reactants with regard to the products, the larger is the reaction extent, and the more spontaneous will the reaction occur in case that (already discussed) kinetic factors will not upset such predictions. According to the Gibbs fundamental equation, ∆G0 is composed of an enthalpy term, ∆H 0 , representing the reaction heat at constant pressure, and an entropy term, ∆S 0 , representing the degree of “disorder” or, thermodynamically more precise, the number of different ways in which the energy of a system can be achieved by rearranging the atoms or molecules among the states available to them (for details, see [35]): ∆G0 = ∆H 0 − T ∆S 0 .

(2.9)

For spontaneous reactions (∆G0 negative), the entropy increases and/or the enthalpy term is negative, i.e., heat is released during the chemical reaction. In the following, the thermodynamics of three prototype reactions of chemical sensors will be briefly discussed.


13

Simple Adsorption/Absorption At thermodynamic equilibrium state, the free species and the ad/absorbed species are in dynamic equilibrium, i.e., the chemical potentials of a certain compound A in gaseous and polymeric phase are identical: µgas A (p, T ) = polymer 0 (p, T ) (2.6). Absorbing all the constant terms (µi , R, T ) into a sorpµA tion constant, Ksorption , or so-called partition coefficient, the equilibrium state can be described by: asorbed (2.10) Ksorption = Afree . aA The partition coefficient is a dimensionless “enrichment factor” relating, e.g., ) to that in the probed the activity of a compound in the sensing layer (asorbed A ) and also represents a thermodynamic equilibrium gas or liquid phase (afree A constant, which is related to ∆G0 via (2.8). For surface adsorption, it is more common to relate the fractional coverage of the surface, θ, to the concentration of the analyte in the probed phase and to use different types of adsorption isotherm like Langmuir-, Freundlich-, or BET-(Brunauer-Emett-Teller) isotherms [35]. Chemical Reaction In this case, (2.2) can be applied in principle. It has to be modified with regard to the respective reaction mechanism occurring. For a simple reaction like nA A + nB B ↔ nC C + nD D, the equilibrium constant is given in analogy to (2.2): anC · anDD ani i and in particular : K = C . (2.11) K= anAA · anBB i The chemical potentials as defined in (2.6) can be used, and (2.8) holds. As already mentioned, the interaction leading to a true chemical reaction may be too strong to be reversible. Charge Transfer and Electrochemical Reaction For a reaction of type A+ + e− ↔ A, an electrochemical potential has to be introduced. The contribution of an electrical potential to the chemical potential is calculated by noting that the electrical work, We , of adding a charge, z . e (z denotes the number of elementary charges, e), to a region where the potential is φ (φ denotes the Galvani potential, which represents the bulk-to-bulk inner contact potential of two materials and is defined as the difference of the Fermi levels of these two materials), is: We = z · e · φ ; hence, the work per mole is : We = z · F · φ

(2.12)

F here denotes the Faraday constant, 96485 C/mol, which is equivalent to one mole of elementary charges. Consequently, the electrochemical potential is (compare 2.6): (2.13) µi = µ0i (p, T ) + RT ln ai + zF φ .

14


When z = 0 (neutral species), the electrochemical potential is equal to the chemical potential (2.6). Rewriting (2.7) for the electrochemical potentials leads to: ani i + zF · ∆φ . (2.14) ∆G = ∆G0 (p, T ) + RT ln i

In the equilibrium state (∆G = 0), (2.13) can be expressed in terms of K (2.2). By replacing E, the “electromotive force”, for ∆φ and by replacing E 0 , the standard cell potential, for −∆G0 /zF (a positive voltage per convention always corresponds to a negative ∆G: spontaneous reaction), the so-called “Nernst-equation” results: E = E0 −

RT ln K . zF

(2.15)

The Nernst equation now can be used to derive an expression for the potential of any electrochemical cell or, in our case, electrochemical sensor. Electrochemical reactions can be triggered by applying currents or voltages via electrodes to a sensing layer. After this short excursion into thermodynamics and kinetics, the different microfabrication techniques and the fundamentals of CMOS-devices will be detailed in Chap. 3.

3 Microtechnology for Chemical Sensors

Microtechnology and microfabrication processes are used to produce devices with dimensions in the micrometer to millimeter range. Microfabrication processes can be effectively applied to yield a single device or thousands of devices. The so-called “batch processing”, i.e., the fabrication of many devices in parallel, does not only lead to a tremendous cost reduction, but also enables the production of array structures or large device series with minute fabrication tolerances. Microfabrication processes hence significantly differ from conventional machining processes, such as drilling or milling with mechanical tools. Integrated circuit (IC) fabrication processes are the most important microfabrication processes [36–38]. The success of CMOS-technology, which is one of the enabling technologies of the information age, clearly demonstrates the efficiency of microfabrication technologies. Standard processing steps originating from semiconductor technology can be used in combination with dedicated micromachining steps to fabricate three-dimensional mechanical structures, which form the basis for the chemical microsensors detailed in Chap. 4. Key advantages of microfabricated chemical sensors include small device size and sampling volume, the possibility of batch processing, and the reproducibility of transducer/sensor characteristics due to the precise geometric control in the fabrication steps. Microfabrication techniques also can be used to either significantly improve sensor characteristics in comparison to conventionally fabricated devices, or to develop devices with new functionality that cannot be realized in conventional fabrication technology. Microsensor success stories, such as micromachined pressure sensors and accelerometers, show that microfabrication techniques are especially suitable for high-volume applications in, e.g., automotive industry. In high-volume production, the advantage of batch processing is paramount, and the high development and setup costs amortize. This chapter is organized in the following way: Microsystem substrate materials and standard processing steps originating from semiconductor technology are detailed in the first sections, followed by a short introduction to CMOS technology and a description of micromachining and layer-deposition processes that are specific to chemical microsensors and -systems.

16


3.1 Microtechnology Substrate Materials Silicon is the standard substrate material for IC fabrication and, thus, the most common substrate material in microfabrication. It is supplied as single crystal wafers with diameters from 100–300 mm. The use of silicon substrate material enables the co-integration of transducers and circuitry, which is used to advantage, e.g., in realizing CMOS-based microsystems [39]. Besides its favorable electrical properties, single crystal silicon also has excellent physical properties (mechanical strength, thermal conductivity) [40], which enable the design of micromechanical structures. Therefore, silicon is also the most common substrate material for microfabricated chemical and biosensors. A large number of micromachining techniques have been developed to structure silicon substrates (see also Sect. 3.4) [26–29, 41–43]. Glass exhibits attractive dielectric and optical properties. Glass is also supplied in wafer form in different compositions (e.g., quartz, fused silica, and borosilicate glass) and diameters. Since glass is transparent for visible light, it is particularly suited for devices with optical detection principles. Single-crystal quartz with its hexagonal lattice structure is piezoelectric and is, therefore, used, e.g., as substrate material for acoustic-wave devices (see Sect. 4.1). Last but not least, glasses are chemically inert and suitable for high-temperature applications. A number of micromachining techniques, such as isotropic wet etching or anisotropic dry etching, have been developed to structure glass. Ceramics have been extensively used as substrate material for hybrid microelectronics and in microelectronics packaging [44] . The standard material is alumina (Al2 O3 ), other materials include beryllium oxide (BeO) and aluminum nitride (AlN). Their chemical inertness, biocompatibility, and mechanical stability render ceramics a very interesting material for microsystems. Most microfabrication techniques for ceramic materials have been adapted from microelectronics packaging processes. Polymers have been more and more explored over the last years as an inexpensive substrate material. Due to the cost advantage, disposable devices, such as microfluidic arrays or microstructured biosensor assays, are often based on polymers [45]. Special processes, such as hot embossing, injection molding, laser machining, or stereolithography, have been developed to structure polymer materials even in the micrometer to nanometer range [46].

3.2 Fundamental Semiconductor Processing Steps The four basic microfabrication techniques for chemical/biosensors are identical with those used in integrated-circuit fabrication [36–38]: Deposition, patterning, doping and etching. The sequential application of these techniques to build up a device layer by layer is illustrated in Fig. 3.1.

3.2 Fundamental Semiconductor Processing Steps

wafer

17

wafer in

film deposition photolithography

doping

mask set etching

wafer out Fig. 3.1. Flow diagram of an integrated-circuit fabrication process using the four basic microfabrication techniques: Deposition, photolithography, etching, and doping. Adapted from [36]

A thin layer, such as an insulating silicon dioxide film, is deposited on a substrate. A light-sensitive photoresist layer is then deposited on top and is patterned using photolithography. The pattern is then transferred from the photoresist layer to the silicon dioxide layer by an etching process. After removing the remaining photoresist, the next layer is deposited and structured, and so on. Doping of a semiconductor material by ion implantation can be done directly after photolithography or after patterning an implantation mask (e.g., a patterned (sacrificial) silicon dioxide layer). In the following, a brief overview on the four fundamental microfabrication steps will be given. More details can be found in books on semiconductor processing [36–38]. 3.2.1 Deposition The two most common thin-film deposition methods in microfabrication are chemical vapor deposition (CVD), performed at low pressure (LPCVD), atmospheric pressure (APCVD) or plasma-enhanced (PECVD), and physical vapor deposition (PVD), such as sputtering and thermal evaporation. Typical CVD and PVD film thicknesses are a few micrometers. Other techniques include electroplating of metal films and spin- or spray-coating of polymeric films such as photoresist. Both processes can yield film thicknesses from less than 1 µm up to several hundred micrometers.

18


Dielectric layers, predominantly silicon dioxide and silicon nitride, are used as insulating material, as mask material, and for passivation. Silicon dioxide is either thermally grown on top of a silicon surface (thermal oxide) at high temperatures in an oxidation furnace (900◦ –1200◦ C), or it is deposited in a CVD system like silicon nitride. CVD oxides are deposited at temperatures between 300◦ and 900◦ C. Metal layers are used for electrical connections, as leads or electrode material, for resistive temperature sensors (thermistors) or as mirror surfaces. Metals, which are widely used in the microelectronics industry, such as aluminum, titanium, and tungsten, are routinely deposited by sputtering or by electron-beam evaporation. Depending on the application, a large number of other metals, including gold, palladium, platinum, silver or alloys, can be deposited with PVD methods. Whereas aluminum has been the standard metallization in IC-fabrication for many years, the state-of-the-art, sub-0.25 -µm CMOS technologies often feature copper metallizations due to their lower resistivity and higher electromigration resistance as compared to aluminum. Highly-doped polycrystalline silicon (polysilicon) is used as gate material for metal-oxide-semiconductor field-effect transistors (MOSFET), for electrodes and resistors, and as thermoelectric or piezoresistive material. Polysilicon is usually deposited in a LPCVD process using silane (SiH4 ) as gaseous precursor. Polymers such as photoresist are commonly deposited by spin- or spraycoating. Similar techniques are also used to coat chemical sensors with sensitive polymer films [1, 4]. 3.2.2 Patterning Photolithography is the standard process to transfer a pattern, which has been designed with a computer-assisted design (CAD) program, onto a certain material. The process sequence is illustrated in Fig. 3.2 [47]. A mask with the desired pattern is created. The mask is a glass plate with a patterned opaque layer (typically chromium) on the surface. Electron-beam lithography is then used to write the mask pattern from the CAD data. In the photolithographic process, a photoresist layer (photostructurable polymer) is spin-coated onto the material to be patterned. Next, the photoresist layer is exposed to ultraviolet (UV) light through the mask. This step is done in a mask aligner, in which mask and wafer are aligned before the subsequent exposure step is performed. Depending on whether positive or negative photoresist was used, the exposed or the unexposed photoresist areas are removed during the resist development process. The remaining photoresist acts as a protective mask during the etching process, which transfers the pattern onto the underlying material. Patterned photoresist can also be used as mask for a subsequent ion-implantation step. After the etching or ion-implantation step, the remaining photoresist is removed, and the next layer can be deposited and patterned.

3.2 Fundamental Semiconductor Processing Steps

19

thin film substrate apply photoresist

photoresist

expose photoresist

UV mask light

develop photoresist

transfer pattern

Fig. 3.2. Schematic of a photolithographic process sequence for structuring a thinfilm layer [47]

The so-called lift-off technique is a way to pattern a thin film material, which would be difficult to etch. Here, the thin film material is deposited on top of the patterned photoresist layer. In order to avoid a continuous film, the thickness of the deposited film must be less than the resist thickness. By removing the underneath photoresist, the thin film material on top is also removed by “lifting it off”, leaving a structured thin film on the substrate [28, 41, 42]. 3.2.3 Etching The two different categories of etching processes include wet etching using liquid chemicals and dry etching using gas-phase chemistry. Both methods can be either isotropic, i.e., provide the same etch rate in all directions, or anisotropic, i.e., provide different etch rates in different directions (see also Sect. 3.4 Micromachining) [37,38]. The important criteria for selecting a particular etching process encompass the material etch rate, the selectivity to the material to be etched, and the isotropy/anisotropy of the etching process. An overview on various etching chemistries used in microfabrication can be found in [48]. Wet etching is usually isotropic with the important exception of anisotropic silicon wet etching in, e.g., alkaline solution, such as potassium hydroxide (see Fig. 3.4, Sect. 3.4.1.1). Moreover, wet etching typically provides a better etch selectivity for the material to be etched in comparison to accompanying

20


other materials. An example includes wet etching of silicon dioxide using hydrofluoric-acid-based chemistry. SiO2 is isotropically etched in diluted hydrofluoric acid (HF : H2 O) or buffered oxide etch, BOE (HF : NH4 F). Typical etch rates for high-quality (thermally grown) silicon dioxide films are 0.1 µm/min in BOE. Dry etching, however, is often anisotropic, resulting in a better pattern transfer, as mask underetching is avoided (see Fig. 3.4, Sect. 3.4.1.1). Therefore, anisotropic dry etching processes, such as reactive-ion etching (RIE), of thin film materials are very common in the microelectronics industry. In an RIE system, reactive ions are generated using a plasma and are accelerated towards the surface to be etched, thus providing directional etching characteristics. Higher ion energies typically result in more anisotropic etching characteristics, but also lead to reduced etching selectivity. Though a large number of dry etching chemicals and recipes exist, mainly fluorine- or chlorine-based etching chemistry is commonly used [37, 38]. 3.2.4 Doping Doping is used to modify the electrical conductivity of semiconducting materials such as silicon or gallium arsenide [36–38]. It is hence the key process step to fabricate semiconductor devices such as diodes and transistors. In the case of silicon, doping with phosphorus or arsenic yields n-type silicon, whereas ptype silicon results from boron doping. By varying the dopant concentration of n-type silicon from 1014 to 1020 cm−3 , the resistivity at room temperature can be tuned from approximately 40 Ω cm to 7·10−4 Ω cm. Dopant atoms are introduced by either ion implantation or diffusion from a gaseous, liquid, or solid source. Ion implantation allows for introducing precisely defined quantities of dopants into the semiconductor material and is, hence, a key process of microelectronics fabrication. The substrate material, e.g., a silicon wafer, is bombarded with accelerated ionized dopant atoms in an ion implanter. The result is an approximately Gaussian distribution of the dopant atoms in the substrate wafer with a mean penetration depth controlled by the acceleration voltage. A high-temperature process is then used for annealing and activating the dopants in the case of ion implantation or for “driving-in” the dopant atoms until a desired doping profile has been achieved in the case of utilizing diffusion processes [36–38].

3.3 CMOS Technology CMOS is the dominant semiconductor technology for microprocessors, memories and application-specific integrated circuits (ASICs) [36–38]. CMOS-chips generally consist of a substrate, the transistor components, the metal layers and a passivation layer on top. The substrate is a silicon wafer, the thickness of which depends on the wafer size: 525 µm for a four-inch wafer and

3.3 CMOS Technology

21

850 µm for an eight-inch wafer. Implanted in the silicon wafer are doped regions, which form together with two polysilicon layers (e.g., transistor gate regions) and silicon oxide layers the transistor structures as defined during the CMOS process. Up to 8 metal layers consisting of aluminum (down to a feature size of 0.18 µm) or copper (0.13 µm and 0.09 µm CMOS) are used to wire the electronic components and to establish connections to the outside world (bondpads). Intermetal oxide (Si-oxide) layers are used as electrical insulator between the different metal layers. Finally, silicon nitride, silicon oxinitride, or silicon oxide layers passivate the device and protect the electronics (Fig. 3.3). The overall CMOS device is fabricated in a defined sequence of material deposition, doping, lithography and etching steps [36–38]. intermetal passivation oxide nitride metal 2 metal 1

S

D

D

n-well

n+ p+ PMOS

polysilicon

S

gate contact oxide oxide

field oxide

NMOS

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

Undoped semiconductor materials conduct electricity but not enthusiastically. Semiconductor areas that are doped become conductors of either extra electrons with a negative charge (n-doping, e.g., by phosphorus) or of positive charge carriers (p-doping, by e.g., boron). The denomination “complementary” metal oxide semiconductor is owing to the fact that both, n-channel and p-channel transistors are realized on the same substrate. The substrate is, for example, a lightly p-doped wafer material that exhibits n-doped areas (n-wells), see Fig. 3.3. More recently twin-well processes are used, in which n-wells and p-wells are implanted in the wafer substrate material. Both transistor types, n-channel (NMOS) and p-channel (PMOS) transistors are used to realize logic functions. An exemplary CMOS-chip cross-section (0.8 µm nwell, double-metal CMOS process as used for most of the devices in Chap. 5)

22


is shown in Fig. 3.3. The p-substrate exhibits n-well areas created by implantation. Heavily p-doped structures (p+) in the n-well and heavily n-doped structures (n+) in the p-substrate form transistor source and drain. The transistor gate is made of polysilicon on top of the gate oxide (SiO2 ). By applying an appropriate voltage (positive for NMOS, negative for PMOS) to the gate via the polysilicon, the substrate majority charge carriers (electrons in n-well, holes in p-substrate or p-well) are depleted in the surface area below the gate owing to the field effect, and a conducting channel is formed between source and drain (n-channel in p-substrate or p-well, p-channel in n-well). Variation of the gate voltage modulates the source-drain current. The modulation is continuous within a certain range in analog circuits, and produces only two states, “on” or “off”, in digital circuits. The metal layers (Fig. 3.3: metal 1 and metal 2, aluminum) are used to wire the transistors. Dielectric layers such as gate oxide, field oxide, contact oxide, and intermetal oxide (SiO2 ) serve as electrical insulation between conducting or semiconducting layers. The silicon nitride or other materials (oxinitride, oxide) on top serve as passivation and provide electrical and mechanical/chemical protection of the circuitry. Silicon nitride and, to a lesser extent, silicon oxide (or silicon with a native oxide layer) are durable in case of liquid exposure and are biocompatible, which principally allows for using CMOS chips also with cells or living material. CMOS aluminum, however, is neither stable in liquids or in air at higher temperatures nor is it biocompatible, and, consequently, has to be covered with noble-metal coatings such as gold or platinum for many microsystem applications. For more details on semiconductor technology see, e.g., the standard textbooks of Sze [36–38].

3.4 Microfabrication for Chemical Sensors 3.4.1 Micromachining for Chemical Microsensors Micromachined structures such as membranes and cantilevers are widely used in bio(chemical) sensors. Membranes provide, e.g., the thermal isolation required for thermal chemical sensors, whereas cantilevers can be used as resonant structures for mass-sensitive chemical sensors (see Chap. 4.1.3). In the following, the fundamental micromachining techniques are briefly reviewed. More details on micromachining techniques can be found in dedicated books on microsystem technology [25–29, 41, 42]. A recent review on microfabrication in biology and medicine can be found in [43]. The micromachining techniques are categorized into bulk micromachining [49] and surface micromachining processes [50] (see Fig. 3.4). In the case of bulk micromachining, the microstructure is formed by machining the

3.4 Microfabrication for Chemical Sensors

23

relatively thick bulk substrate material, whereas in the case of surface micromachining, the microstructure consists of thin-film layers, which are deposited on top of a substrate, and which are selectively removed in a defined sequence to yield the MEMS structure. 3.4.1.1 Bulk Micromachining One approach to enhance the functionality of IC-based devices includes micromachining the bulk substrate, which, in most cases, consists of silicon. Bulk micromachining techniques can be classified into isotropic and anisotropic etching techniques (structure geometry), or into wet and dry etching techniques (reactant phase: liquid or gaseous) [25–29, 41, 42, 49, 50]. In the case of isotropic etching, the same etch rate applies to all directions (Fig. 3.4a), whereas in the case of anisotropic etching, the substrate is preferentially etched away along certain crystal planes while it is preserved in other directions (Fig. 3.4a).

(a) bulk micromachining isotropic etching

(b) surface micromachining substrate (Si)

sacrificial layer

structural layer

microstructure

substrate (Si) anisotropic etching

substrate (Si) Fig. 3.4. Micromachining techniques: (a) Bulk micromachining, anisotropic and isotropic etching, (b) surface micromachining with sacrificial layer, structural layer and a subsequent etch step

The most common isotropic wet silicon etchant is HNA, a mixture of hydrofluoric acid (HF), nitric acid (HNO3 ), and acetic acid (CH3 COOH): Nitric acid oxidizes the silicon surface, and hydrofluoric acid etches the grown silicon dioxide layer. The acetic acid controls the dissociation of HNO3 , which provides the oxidation of the silicon. The etch rates and the resulting surface quality strongly depend on the chemical composition [41, 42, 49]. Anisotropic wet etching of silicon is the most common micromachining technique and is used to release, e.g., membranes and cantilevers for chemical and biosensors. Anisotropic silicon etchants etch single-crystal silicon at

24


different etch rates in distinct crystal directions. The etch grooves are limited by crystal planes, along which etching proceeds at slowest speed, i.e., the 111 planes of silicon. In case of 100 silicon wafers, the 111 planes are intersecting the wafer surface at an angle of 54.7◦ , so that the typical pyramid-shape etch grooves as shown in Fig. 3.4 are formed. Mask materials for anisotropic silicon etchants are silicon dioxide and silicon nitride. It is important to note, that “convex” corners of the etch mask are underetched in case of 100 silicon substrates, leading to, e.g., completely underetched cantilever structures. The etch rates in preferentially etched crystal directions such as the 100 and the 110 direction, and the ratio of the etching rates in different crystal directions strongly depend on the exact chemical composition of the etching solution and the process temperature [51]. The most common anisotropic silicon etching solution is potassium hydroxide, KOH. As an example, a 6-molar KOH solution at 95◦ C provides a 100 etch rate of 150 µm/hour and an anisotropy between 100 and 111 direction etching of 30–100:1 [52]. Since the etch rate of silicon dioxide in KOH solution is rather high (for thermal oxide approx. 1 µm/hour in 6-molar KOH solution [42]), silicon nitride films are often used as etching mask. KOH solution is very stable, yields reproducible etching results, is relatively inexpensive, and is, therefore, the most common anisotropic wet etching chemical in industrial manufacturing. The disadvantages of KOH include the relatively high SiO2 - and Al-etch rates, which require a protection of, e.g., integratedcircuit structures during etching. Etching with KOH is typically performed from the back side of the wafer, with the front side protected by a mechanical cover and/or a protective film [52]. Alternative silicon wet etchants are ammonium hydroxide compounds, such as tetramethyl ammonium hydroxide (TMAH), and ethylene diamine/ pyrochatechol (EDP) solutions. Some EDP formulations, such as EDP type S, exhibit relatively low Al- and SiO2 -etch rates, which render them suitable for releasing microstructures on the front side of CMOS-wafers [53]. More detailed discussions of wet etching of silicon can be found, e.g., in [26] and [42]. Reliable etch stop techniques are very important for achieving reproducible etching results. As already mentioned, wet anisotropic silicon etchants “stop” etching, i.e., the etch rate is reduced by at least 1–2 orders of magnitude, as soon as a 111 silicon plane or a silicon dioxide/nitride layer is reached. In addition, the etch rate is greatly reduced in highly boron-doped regions (doping concentration ≥1020 cm−3 ). The etching can also be stopped at a p-n-junction using a so-called electrochemical etch stop technique (ECE) [41]. This method has been extensively used to release silicon membranes and n-well structures (see, e.g., Chap. 5.4). ECE relies on the passivation of silicon surfaces through application of a sufficiently high anodic potential with respect to the potential of the etching solution. Isotropic dry etching of silicon is done with xenon difluoride, XeF2 . This vapor-phase etching method exhibits excellent etch selectivity with respect


25

to aluminum, silicon dioxide, silicon nitride, and photoresist, all of which can be used as etch masks. The XeF2 silicon etch rates depend on the loading (size of the overall silicon surface exposed to the etchant) with typical values of approx. 1µm/min [54]. Anisotropic dry etching of silicon is usually carried out as reactiveion-etching (RIE) with plasma-assisted etching systems. By controlling the process parameters, such as process gases and process pressure, the etching can be rendered either isotropic or anisotropic. The dry-etching anisotropy originates from experimental parameters such as the direction of the ion bombardment, and is, therefore, independent of the crystal orientation of the substrate material. Most bulk etching of silicon is accomplished using fluorine free radicals with SF6 as a typical process gas. Adding chlorofluorocarbons results in polymer deposition in parallel with etching, which leads to enhanced anisotropy. Very-high-aspect-ratio microstructures can be achieved with deep (D)RIE , a method, which has gained importance during the last years. Deep-RIE systems rely on high-density plasma sources and an alternation of etching and polymer-assisted sidewall protection steps. In a process known as the “Bosch” process [55], a mixture of trifluoromethane and argon is used for polymer deposition. Due to the ion bombardment, the polymer deposition on the horizontal surfaces can almost be prevented, while the sidewalls are passivated with a TeflonTM -like polymer. In the second process step, an SF6 -based etching chemistry provides silicon etching in the non-passivated regions, i.e., the horizontal surfaces. Both process steps are alternated, resulting in typical silicon etch rates of 2–3 µm/min with an anisotropy on the order of 30:1 [49]. Silicon dioxide and photoresist layers can be used a etch masks. Even though an exceptional anisotropy can be achieved, which is independent of the crystal orientation, one should keep in mind that deep RIE systems are by far more expensive than a simple wet-etching setup, and that only one wafer is processed at a time [56, 57]. 3.4.1.2 Surface Micromachining Surface micromachining comprises a number of techniques to produce microstructures from thin films previously deposited onto a substrate and is based on a sacrificial-layer method (Fig. 3.4b). In contrast to bulk micromachining, surface micromachining leaves the substrate intact [41,50]. A sacrificial layer is deposited and patterned on a substrate. After that, a structural thin film, in most cases polysilicon, is deposited and patterned, which will perform the mechanical or electrical functions in the final device. A selective etchant then removes exclusively the sacrificial-layer material. The thickness of the sacrificial layer determines the distance of the structural parts from the substrate surface. Common sacrificial-layer materials include silicon oxide etched by hydrogen fluoride and aluminum etched by a mixture of phosphoric, nitric and acetic acid.

26


Clamped beams, microbridges, or microchannels can be fabricated this way, microrotors and even microgears can be realized by repeated layer deposition and etching [25–29, 41]. 3.4.2 Wafer Bonding While the majority of microstructures are fabricated from a single substrate or wafer, several wafers can be joined by wafer bonding [58]. The substrates/wafers are bonded onto each other either directly or via an intermediate layer. Direct bonding techniques include silicon fusion bonding and anodic bonding. Silicon fusion bonding involves two silicon wafers, which are bonded to each other at high temperatures (T≥ 1000◦ C). Low-temperature (T < 400◦ C) fusion bonding has been demonstrated after using special cleaning procedures, but exhibits reduced bonding strength. Anodic bonding of a sodiumrich glass wafer onto a silicon wafer is accomplished by applying an electric field across the bonding interface at moderate bonding temperatures (T ≈ 300–400◦ C). Glass materials with a thermal expansion coefficient similar to that of silicon (e.g., Pyrex glass 7740) are used for anodic bonding in order to minimize thermomechanical stress. Anodic bonding and silicon fusion bonding require very clean and smooth wafer surfaces to achieve void-free bonding. The surface quality and roughness is less important if an intermediate layer is used for wafer bonding. Possible intermediate bonding layers include adhesives, low-melting-temperature glass, solder films, or metallic films such as gold. Examples will be given in Sects. 4.3.2.2 and 4.3.4. 3.4.3 Sensitive-Layer Deposition The set of microfabrication processes used for chemical/biosensors is completed by various deposition techniques for chemically or biologically sensitive layers. Chemically sensitive polymer layers or organic molecules that are used for the detection of volatile organics in air can be deposited by, e.g., dispensing, spray coating or by using self-assembled monolayers (SAMs). Metal-oxide films for the detection of, e.g., carbon monoxide and nitrogen oxides can be deposited either by a sol/gel process, by drop coating, or by sputtering [59]. Recently, microcontact printing or soft lithography [60] has been introduced as an additional method for pattern transfer. A soft polymeric stamp is used to reproduce a desired pattern directly on a substrate. Feature sizes on the order of 1 µm can be routinely achieved with this technique. The polymer stamp, often made from poly(dimethylsiloxane) (PDMS), is formed by a molding process using, e.g., a silicon master fabricated with conventional microfabrication techniques. After “inking” the stamp with the material to be printed, the stamp is brought in contact with the substrate material, and the


27

pattern of the stamp is reproduced. Surface properties of the substrate thus can be modified to, e.g., locally promote or prevent molecule adhesion. Soft lithography has been specifically developed for biological applications such as patterning cells or proteins with the help of, e.g., self-assembled monolayers (SAM) [60].

4 Microfabricated Chemical Sensors1

Chemical sensors usually consist of a sensitive layer or coating and a transducer (see Chap. 1, Fig. 1.2) [1,4–9]. Upon interaction with a chemical species (absorption, chemical reaction, charge transfer etc.), the physicochemical properties of the coating, such as its mass, volume, optical properties or resistance reversibly change. These changes in the sensitive layer properties can be detected by a variety of transducers and can be translated into electrical signals such as frequency, current, or voltage changes, which are then subjected to further data treatment and processing. As already mentioned in the introduction to this book, chemical sensors can be classified into four principal categories according to their transduction principles [4, 5]: (a) chemomechanical sensors (b) thermal sensors (c) optical sensors, and (d) electrochemical sensors. Each of those four sensor categories will be briefly introduced, and, then, specific exemplary microfabricated transducers and devices will be abstracted. The transducer overview will be restricted to semiconductor-based and CMOS-based devices and will, for the sake of completeness, also include short abstracts on devices, which are described in much more detail in the subsequent Chap. 5 on the CMOS technology platform for chemical sensors. Typical chemically sensitive materials and sensor applications will be brought up in the context of the respective transducer structures.

4.1 Chemomechanical Sensors The change in mechanical properties (e.g., mass) of a sensitive layer upon interaction with an analyte can be conveniently recorded by using micromechanical structures. Any species that can be immobilized on the sensor can, in principle, be sensed. As with most of the chemical sensors (excluding thermal sensors), the measurements are performed at a thermodynamic equilibrium state (2.1–2.4, 2.10), which is defined by the Gibbs energy minimum 1

Large parts of the material of Chap. 4 were originally published in the book: MEMS: A Practical Guide to Design, Analysis, and Applications, edited by Jan Korvink and Oliver Paul, William Andrew Publishing, Norwich, NY, 2005. Reprinted here with permission.

30

4 Microfabricated Chemical Sensors

of the system. In the simplest case such chemomechanical sensors are gravimetric sensors responding to the mass of species accumulated in a sensing layer [61–63]. Some of the sensor devices additionally respond to changes in a variety of other mechanical properties of solid or fluid media in contact with their surface such as polymer moduli, liquid density and viscosity [61–63], which will not be discussed here. The high sensitivity of gravimetric sensors provides good chemical sensitivity: mass changes in the picogram range can be detected, and ppm (parts per million) to ppb (parts per billion) detection levels have been reported for, e.g., gas and vapor sensors [61–63]. The large number of chemical species that can be present in the environment, and the difficulty in selectively and, at the same time, reversibly sorbing these species on the sensor, however, makes specific detection difficult. Most of the gravimetric sensors rely on piezoelectric materials such as quartz, lithium tantalate or niobate, aluminum nitride, zinc oxide and others. Piezoelectricity results in general from coupling of electrical and mechanical effects. The prerequisite is an anisotropic, noncentrosymmetric crystal lattice. Upon mechanical stress, charged particles are displaced and thus generate a measurable electric charge in the crystal. In turn, mechanical deformations can be achieved by applying a voltage to such a crystal (for details, see [4]). Using an alternating current (AC), the crystals can be electrically excited into a fundamental mechanical resonance mode. The resonance frequency, which is the recorded sensor output in most cases, changes in proportion to the mass loading on the crystal or device. The more mass (analyte molecules) is absorbed, e.g., in a polymer coated onto a piezoelectric substrate or transducer, the lower is the resonance frequency of the device: ∆f = −C f02 ∆m/A .

(4.1)

This equation was published by Sauerbrey in 1959 [64]. ∆f here denotes the frequency shift due to the added mass in [Hz], C is a constant, f0 is the fundamental frequency of the quartz crystal in [Hz] and ∆ m/A is the surface mass loading in [g cm−2 ]. The following equation describes the relationship between analyte gas phase concentration change, ∆c A , and the responses of mass-sensitive sensors: ∆fA = Γ · MA · K · ∆cA .

(4.2)

Here, ∆fA [Hz] denotes the frequency shift (sensor response) measured upon exposure to analyte at a concentration cA [mol/L]. MA [kg/mol] is the molar mass of the analyte vapor, K is the partition coefficient (2.10), and Γ is a gravimetric constant [L/kg·s] including, e.g., the frequency shift measured upon initial deposition of the sensitive layer, the coating density, transducer dimensions, etc. A typical signal of a gravimetric sensor is displayed in Fig. 4.1 showing the frequency shifts of a resonant cantilever coated with poly(etherurethane),

4.1 Chemomechanical Sensors

31

frequency shift [Hz]

250

500

750

1000

1250

1500

1500

1250

750

1000

500

0

250

ppm

20

-20 -40 -60

analyte: n-octane polymer: PEUT

-80 -100

0

50

100 time [min]

150

200

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

(PEUT), upon exposure to various concentrations of n-octane. At low analyte concentrations (trace level), a linear correlation between the frequency shift due to analyte absorption and the corresponding analyte concentration in the gas phase is usually observed (Fig. 4.1), provided that the sensing film on the transducer moves synchronously with the oscillating crystal surface. Significant deformations across the film thickness result in a more complex relationship between mass changes and resonant frequency due to, e.g., viscoelastic effects (concept of “acoustically thin and thick” films as detailed in [65]). The most common devices are the thickness-shear-mode resonator (TSMR) or quartz microbalance (QMB), a bulk resonator, and the Rayleigh surfaceacoustic-wave (SAW) device, both based on quartz substrates. The QMB was demonstrated to function as an organic vapor sensor by King in 1964 [66], the SAW device became popular after introducing interdigital transducers to 2

Concentration are usually given in mole-per-volume (mol/m3 ) or mass-pervolume (kg/m3 ) units. In gas sensorics ppm-units (parts per million volume or parts per million pressure) are widely used, which are strictly speaking no valid concentration units, in particular since ppm-units are dimensionless. The ppm-units are nevertheless used here owing to their popularity. Assuming the validity of the ideal-gas law, which holds true for low analyte concentrations, ppm-units can be easily converted into mass-per-volume units, e.g., (µg/L) by division through the molar volume of an ideal gas at 25◦ C (24.45 l) and multiplication with the molar mass of the analyte compound. Moreover, under normal pressure conditions 10 ppm analyte correspond to 1 Pa partial pressure of the respective analyte.

32


acoustic sensors in 1970 [67]. Since the TSMR is not semiconductor-based and not compatible with IC technology, it will not be treated here any further. Shear-horizontal-acoustic-plate-mode (SH-APM) devices devices, sheartransverse-wave device (STW) and Love-wave devices devices require quartz, lithium niobate or lithium tantalate substrates [61–63] and, hence, will not be dealt with here as well. For details and further information it is referred to a wealth of literature [61–63, 68–70]. Silicon is not a piezoelectric material. The realization of silicon-based piezoelectric transducers hence requires an additional piezoelectric layer to be patterned on the silicon. Different materials have been used such as cadmium sulfide [71], aluminum nitride [72, 73], and in particular zinc oxide (ZnO) [74–76], which will be subject to further discussion in this chapter. In the following, three semiconductor-technology-compatible types of mass-sensitive devices will be described in more detail: (1) SAW-devices on Sisubstrates with piezoelectric overlay, (2) flexural-plate-wave devices (FPWs), and (3) micromachined cantilevers. Operability in gas or liquid media, typical coating materials, target analytes, and applications will be discussed in the context of each transducer. An overview of micromachined resonant sensors is given in [70, 77]. 4.1.1 Rayleigh SAW Devices Transduction Principle and Sensing Characteristics Interdigital transducers can be used to launch and detect a surface-acoustic wave on a piezoelectric substrate [67] as schematically shown in Fig. 4.2. By applying an AC voltage to a set of interdigital transducers patterned on a piezoelectric substrate with appropriate orientation of the crystal axes, top view

Rayleigh SAW

interdigitated electrodes

excitation propagation metal electrodes

side view detection ZnO

wave propagation

particle displacement

substrate: silicon

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


33

one set of the fingers moves downwards, the other upwards, thereby creating an oscillating mechanical surface deformation. This surface deformation generates an acoustic wave, which propagates along the surface and is converted back into an electrical signal by deforming the surface in the region of the receiving transducer. The electrical signal of the receiving transducer is recorded and represents the sensor signal. For a given piezoelectric substrate, the acoustic wavelength and, thus, the operating frequency of the SAW is determined by the transducer periodicity, which is equal to the acoustic wavelength at the transducer center frequency. Typical frequencies range between 100 and 500 MHz [61–63]. Such frequencies require a sophisticated high-frequency circuit design. Therefore, a bare reference oscillator is operated together with the sensor in many cases, and the outputs are mixed to produce a difference frequency with values in the kHz-range that is recorded [74, 76]. The acoustic wave is confined to a surface region of approximately one acoustic wavelength thickness. The velocity and damping characteristics of the acoustic wave hence are extremely sensitive to changes at the transducer surface. When used in an oscillator circuit, relative changes in the wave velocity are reflected as equivalent changes in fractional oscillation frequency. A change in mass due to, e.g., absorption of a gaseous analyte in a polymeric sensing layer thus changes the device frequency according to (4.1). The acoustic (Rayleigh) wave causes an elliptical particle movement at the transducer surface (Fig. 4.2), i.e., the sensitive films deposited on top of the transducers and the piezoelectric substrate are severely deformed. Thus, additional effects such as changes in viscoelastic properties of the sensing layer can affect the sensor response [65]. Fabrication Since there exists a variety of custom-designed semiconductor and silicon processes in the literature, only the transducer-related additional fabrication steps after semiconductor processing and the sensitive-layer-processing steps will be listed in this chapter. Industrial standard IC processes like CMOS fabrication will be explicitely identified, especially since CMOS-based devices are extensively treated in Chap. 5. For more details on semiconductor and MEMS process steps, see Chap. 3. Fabrication Steps: • Optional back etching using potassium hydroxide, KOH, or ethylenediamine-pyrocatechol, EDP, to achieve a membrane structure [74, 78]). • Zinc oxide processing: Deposition mainly by sputtering techniques at 150◦ – 450◦ C. Highly oriented layers of 5–50 µm with a high degree of surface flatness [74]. • Electrode processing: Vacuum evaporation of aluminum or gold, layer thickness >200 nm.

34


• Sensitive layer: Spin or spray coating of polymers, organic layers, or biological entities. Applications Since surface-normal particle displacements occur (Fig. 4.2), and the acoustic wave velocity is larger than the compressional velocity of sound in water, the device radiates compressional waves into the liquid phase, which causes severe attenuation. Rayleigh SAW devices hence cannot be used in liquids [61–63]. Typical applications areas are environmental monitoring or personal safety devices. This includes the detection of different kinds of organic volatiles (hydrocarbons, chlorinated hydrocarbons, alcohols, etc.) by using polymeric layers [74] or porphyrins [79], and the detection of nitrogen dioxide using phthalocyanines [76]. The interaction mechanisms involve, in most cases, fully reversible physisorption and bulk/gas phase partitioning (see 2.10, 4.2). Integrated Gallium Arsenide (GaAs) SAW Sensor GaAs is a well-developed semiconductor device material for fabricating highfrequency integrated circuits, and GaAs is piezoelectric. The piezoelectric properties of GaAs and, hence, the device characteristics are similar to those of quartz except for the strong temperature dependence. An integrated GaAsSAW sensor is shown in Fig. 4.3 [80]. It consists of a 470 MHz SAW device along with a multistage amplifier (4 gain stages and impedance matching

4-gain-stage amplifier

output stage

470 MHz delay line

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


35

output stage) forming a monolithic oscillator circuit thus eliminating the need for high frequency interconnections [80]. 4.1.2 Flexural-Plate-Wave or Lamb-Wave Devices Transduction Principle and Sensing Characteristics Flexural-plate-wave devices have been introduced in 1988 [81]. Their chief advantage is their high sensitivity to added mass at a low operating frequency (typically 3–10 MHz) [82]. FPW devices feature plates that are only a few percent of an acoustic wavelength thick (typically 2–3 µm). The plates are composite structures (Fig. 4.4) consisting of a silicon nitride layer, an aluminum ground plane, a sputtered zinc oxide piezoelectric layer, all supported by a silicon substrate [61, 63, 81, 83]. FPW top view

side view silicon frame

silicon nitride membrane bottom

aluminum

ZnO

composite membrane

metal electrodes

composite membrane wave propagation

particle displacement

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

The interdigital transducers (IDTs) on these devices generate flexural waves (Lamb waves, Fig. 4.4) with retrograde elliptical particle motions as in the SAW devices. However, the velocity in the membrane is much less than in a solid substrate, and the operating frequency for a given transducer periodicity is, hence, considerably lower [61–63, 70, 81]. The Lamb waves give rise to a series of plate modes, one of which has a frequency that is much lower than those of the other possible modes. The velocity of this unique wave decreases with decreasing plate thickness. The entire thickness of the plate is set in motion like the ripples in a flag [61–63, 70, 81, 83]. The confinement of acoustic energy in the thin membrane results in a very high mass sensitivity. The sensor response (frequency shift) is proportional to the mass loading (4.1). Since the Lamb wave causes an elliptical particle movement at the transducer surface (Fig. 4.4), the sensitive films are deformed as it is the case with

36


the SAW. The frequency, however, is much lower, and, therefore, changes in viscoelastic properties of the sensing layer do not severely affect the sensor response. The sensitive layer can be deposited on either side of the membrane. Deposition on the backside (non-processed side of the wafer) has the advantage, that on-chip circuitry will not be exposed to chemicals [81–84]. Transducer Modification: Magnetically Excited FPW Magnetic excitation requires an externally applied magnetic field, but eliminates the need for a piezoelectric layer, which frequently contains elements (Zn, etc.) that pose contamination problems in IC fabrication. The device consists of a silicon nitride membrane suspended in a silicon frame. A metal meander-line transducer is patterned on the membrane surface (Fig. 4.5). Alternating current flowing in the transducer interacts with a static in-plane magnetic field to generate time-varying Lorentz forces (Fig. 4.5). These deform the membrane, exciting it into a resonant mode [70, 85]. To efficiently excite the mode, the current lines of the transducer must be positioned along lines of maximum mode displacement (Fig. 4.5). This requires a critical alignment between the top metallization pattern and the backside etch mask [70, 85]. silicon nitride membrane

I meander-line transducer

FPW

λ silicon substrate B (static in-plane magnetic field)

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

Fabrication • Evaporation of Al and Si-nitride (LPCVD) [81]. • Back etching (KOH, or EDP) to achieve a membrane structure [81, 83]. • Zinc oxide or lead zirconate titanate (PZT) [86] processing if necessary (see SAW).


37

• IDT processing: Vacuum evaporation of aluminum or gold, see SAW. • Sensitive layer: spin or spray coating of polymers, deposition of biological entities. IC process-compatible fabrication sequences for monolithic integration of the Lamb device with electronics are detailed in [83, 87]. Applications Surface-normal particle displacements occur (Fig. 4.4), but the acoustic-wave velocity is much less than the compressional velocity of sound in water. FPW devices thus can be used in the liquid phase [61, 63, 83, 88]. Typical application areas are environmental monitoring (gas and liquid phase) or biosensing in liquids. These include the detection of different organic volatiles in the gas phase (hydrocarbons, chlorinated hydrocarbons, alcohols, etc.) by using polymeric layers [81, 84, 89–91], the detection of the weight percentage of alcohol/water [83] or glycol/water [88] mixtures and the use of an FPW-based immunoassay for the detection of breast cancer antigens [92]. The interaction mechanisms involve reversible physisorption and bulk/gas phase partitioning (see 2.10, 4.2) as well as antigen/antibody binding [92]. 4.1.3 Resonating Cantilevers Transduction Principle and Sensing Characteristics Micromachined cantilevers commonly employed in atomic force microscopy (AFM) constitute a promising type of mass-sensitive transducer for chemical sensors [93–107]. The sensing principle is quite simple. The cantilever is a layered structure composed of, e.g., the dielectric layers of a standard CMOS process, silicon, metallizations, and eventually, add-on piezoelectric zinc oxide. The cantilever base is firmly attached to the silicon support. The freestanding cantilever end is coated with a sensitive layer (Fig. 4.6). The excitation of a cantilever in the resonant mode is usually performed by applying piezoelectric materials (ZnO) [106] or by making use of the bimorph effect, i.e., the different thermal expansion coefficients of the various polymeric coating

silicon frame

cantilever

Fig. 4.6. Schematic representation of a resonating cantilever

38


layer materials forming the cantilever [93–105]. This difference in material properties gives rise to a cantilever deflection upon heating. Periodic heating pulses in the cantilever base thus can be used to thermally excite the cantilever in its resonance mode at 10–500 kHz [93–95, 106]. There are two fundamentally different operation methods: (a) static mode: measurement of the cantilever deflection upon analyte-sorption-induced stress changes by means of, e.g., laser-light reflection [100, 103–105], (b) dynamic mode: excitation of the cantilever in its fundamental mode and measurement of the change in resonance frequency upon mass loading [93–95, 101, 102] in analogy to other mass-sensitive devices (4.1, 4.2). These two methods impose completely different constraints on the cantilever design for maximum sensitivity. Method (a) requires long and deformable cantilevers to achieve large deflections, whereas method (b) requires short and stiff cantilevers to achieve high operation frequencies. Method (b) is preferable with regard to integration of electronics and simplicity of the setup (feedback loop) [93–95, 101, 102, 106, 107]. Method (a) can be applied in liquids as well [101, 104], which is rather difficult using the dynamic mode. The detection of the frequency changes can be done by embedding piezoresistors in the cantilever base [93–95, 101, 102], by measuring motional capacitance changes [107], or by using optical detection by means of laser light reflection on the cantilever [97–100, 103–105]. The mass resolution of the cantilevers is in the range of a few picograms [93–96, 103–105]. This high mass-sensitivity does not necessarily imply an exceptionally high sensitivity to analytes since the area coated with the sensitive layer usually is very small (on the order of 100 × 150 µm2 ) [93]. The sensing layer is deformed upon motion of the cantilever; therefore, modulus effects are expected to contribute to the overall signal, especially since the coating thickness may exceed the thickness of the cantilever. Fabrication • • • • •

Eventually additional Al and Si-nitride (LPCVD) [93–105]. Back etching (KOH, or EDP) to achieve a membrane structure [93–107]. Zinc oxide processing if necessary [106]. Release of the cantilevers by front-side reactive-ion etching [93–103]. Sensitive layer: Spray or drop coating of polymers, deposition of biological entities.

IC process- and CMOS compatible fabrication sequences for monolithic integration of the cantilevers with electronics are detailed in [93–95,101,102,107]. Applications The application of the dynamic mode is mostly restricted to the gas phase, whereas the static mode can be used to detect analytes in liquid phase as well. Due to the bimorph effect (cantilever deformation upon heat generation), cantilevers have also been used in microcalorimetric applications [103, 108].

4.2 Thermal Sensors

39

Typical applications are environmental monitoring (gas and liquid phase) or biosensing in liquids. These include the detection of different kinds of organic volatiles (hydrocarbons, chlorinated hydrocarbons, alcohols, etc., see Fig. 4.1) or humidity in the gas phase by using polymeric layers [93–100, 102, 105], the detection of alcohol in water [101], and the hybridization and detection of complementary strands of oligonucleotides [104]. The interaction mechanisms involve reversible physisorption and bulk/gas phase partitioning (see 2.10, 4.2), as well as receptor-ligand binding [104].

4.2 Thermal Sensors Calorimetric or thermal sensors rely on determining the presence or concentration of a chemical species by measurement of an enthalpy change produced by the chemical to be detected [1,4,64,109]. Any chemical reaction (2.1, 2.11) or physisorption process (2.1, 2.10) releases or absorbs from its surroundings a certain quantity of heat (enthalpy term, ∆H 0 , in 2.9). Reactions liberating heat are termed exothermic, reactions abstracting heat are termed endothermic. This thermal effect shows a transient behavior: Continuous heat liberation/abstraction occurs only as long as the reaction proceeds. This implies that only a steady-state situation can be achieved: A chemical reaction is proceeding at a constant rate and is thus releasing/abstracting permanently a constant amount of heat. There will be, however, no heat production, and, hence, no measurable signal at thermodynamic equilibrium (∆G = 0) in contrast to mass-sensitive, optical, or electrochemical sensors. Conflicting constraints are imposed on the design of a thermal sensor: The sensor has to interact with the chemical species (exchange of matter) and thus constitutes a thermodynamically open system, but, at the same time, the sensing area should be thermally as isolated as possible. The liberation or abstraction of heat is conveniently measured as a change in temperature, which can be easily transduced into an electrical signal. All sensors aimed at thermal infrared radiation detection can, in principle, be used as chemical sensors as well. The various types of calorimetric sensors differ in the way that the evolved heat is transduced. The catalytic sensor (often denoted “pellistor”) employs platinum resistance thermometry [110–122], the thermistor employs composite oxide resistance thermometry [123–129], whereas the pyroelectric [130, 131] and Seebeck-effect [132–144] sensors utilize the respective effects to measure the temperature change. In addition, there are thermal (flow) sensors based on the different thermal conductivity of gaseous analytes [145, 146]. The micromachined cantilever enabling microthermal analysis due to the bimorph effect [103, 108] has already been mentioned in Sect. 4.1.3. Catalytic sensors, Seebeck-effect or thermoelectric sensors, pyroelectric sensors, and thermal conductivity sensors are semiconductor-technologycompatible. Since the latter is essentially a flow sensor responding to thermo-

40


physical properties of a gas (thermal conductivity, heat capacity) [145, 146], it will not be subject to further discussion here. Sensors based on the pyroelectric effect (anisotropic, noncentrosymmetric crystal lattice, permanent polarization, creation of macroscopic charges due to thermal stress in the crystal) require the deposition of pyroelectric material (lithium tantalate, zinc oxide [130], polycyclic organic compounds [131]) on the silicon chip. There are very few chemical-sensing applications reported in literature [34, 130, 131]. Therefore, this class of sensor will not be discussed any further. Thermistors are temperature-sensitive bead resistors composed of either oxide semiconductors with a negative (NTC) or a positive (PTC) temperature coefficient; that is, their resistance decreases or increases nonlinearly with temperature. PTC resistors are made, e.g., from barium or lead titanate, while NTC resistors are made from sintered transition metal oxides (titanium oxide) doped with aliovalent ions. The beads are contacted via two metallic (platinum) leads and coated with glass for chemical inertness [4, 123]. The thermistors themselves have not been fabricated in planar semiconductor technology yet, but have been integrated into silicon-based biosensors due to their small size [124, 125]. They have been used as thermal biosensors in strongly exothermic enzymatic reactions to detect urea [125] glucose [126– 128], uric acid [129], and other compounds of relevance in blood analysis. In the following, thermoelectric and catalytic calorimetric sensors will be detailed. 4.2.1 Catalytic Thermal Sensors (Pellistors) Transduction Principle and Sensing Characteristics The development of the catalytic sensor is derived from the need for a handheld detector for methane to replace the flame safety lamp in coalmines. The catalytic device measures the heat evolved during the controlled combustion of flammable gaseous compounds in ambient air on the surface of a hot catalyst by means of a resistance thermometer in proximity with the catalyst. This method is therefore calorimetric. A catalyst is a chemical compound (often a noble metal like platinum (Pt)) enabling or accelerating a chemical reaction by provision of alternative reaction paths involving intermediates with lower activation energies than the uncatalyzed mechanism (Fig. 4.7). The catalyst itself is not permanently altered by the reaction. The heated catalyst here permits oxidation of the gas at reduced temperature and at concentrations below the lower explosive limit (LEL). Three elements are necessary for this method: A catalyst, a method to heat it, and a means to measure the heat of catalytic oxidation. The term “pellistor” originally refers to a device consisting of a small platinum coil embedded in a ceramic bead impregnated with a noble metal catalyst [110]. A ceramic

4.2 Thermal Sensors

41

A*

A*cat educts A intermediates

products B

reaction: A ↔ A* ↔ B Fig. 4.7. Working principle of a catalyst: Provision of an alternative reaction path with less activation energy for the reaction A↔ A∗ ↔ B via the intermediate state A∗cat

(a) catalyst (Pd, Pt)

platinum resistors

Si-nitride passivation

(b)

silicon frame

100

m

supporting Si-oxynitride membrane

Fig. 4.8. (a) Cross-section of side-by-side microhotplates composed of a dielectric membrane on an etched silicon wafer, and platinum resistors/heaters. The device on the left has a deposited catalyst making it the active element. Redrawn from [111]. (b) Micrograph (SEM) of two meandered polysilicon microbridges. The lower meandered bridge is coated with a thin (approx. 0.1 µm) layer of platinum (CVD). In a differential gas-sensing mode, the upper uncoated filament acts to compensate changes in the ambient temperature, thermal conductivity and flow rate, while the lower filament is used to calorimetrically detect combustible gases. Reprinted from [113] with permission

bead is used since the rate of reaction (and thus the sensor signal) is directly proportional to the active surface area. Figure 4.8 shows two different micromachined designs to realize a catalytic calorimetric sensor: A meander structure on a micromachined membrane [111] and a freestanding, Pt-coated polysilicon microfilament (10 µm wide, 2 µm thick) separated from the substrate by a 2-µm air gap [112,113]. Heat losses to the silicon frame are minimized in these designs. By passing an electric current through the meander, the membrane/microbridge is heated to a temperature sufficient for the Pt surface to catalytically oxidize the combustible mixture;

42


the heat of oxidation is then measured as a resistance variation in the Pt. The combustion of methane, e.g., generates 800 kJ/mol heat, which translates into a corresponding temperature change. The structures described here are very similar to hotplate structures discussed in the electrochemical section. (Sect. 4.4.3.1.2) The temperature change of the sensor element is proportional to the combustible concentration when the device is operated in excess oxygen and in the mass-transfer-limited regime [114, 115]. The combustion of hydrogen in dry air is exemplified in Fig. 4.9 [112, 113]. The circuit maintains a constant sensor temperature by adjusting the supplied current to keep the filament resistance at a constant value. Note that the sensor response is measured at steady state, i.e., continuous combustion. In most realizations, the measuring resistor forms part of a Wheatstone bridge configuration [111, 114]. Temperature–modulated operation has been reported in [116, 117].

sensor response [V]

0.6 1.6% 1.0% 0.5% 0.1% 0.55

0.5 hydrogen on Pt-filament 0.45 200

400

600

800 1000 1200

time [s] Fig. 4.9. Sensor response of a Pt-coated filament exposed to various concentrations of hydrogen in synthetic air. Reprinted from [112] with permission

Fabrication • Back etching (KOH, or EDP) for membranes [111, 116, 118]. • Surface micromachining: Sacrificial-layer etching (HF) for the bridges [112, 113]. • Pt or catalyst processing: Sputtering [116, 117], evaporation [93–95], LPCVD [112, 113]. A processing sequence for microbridges is given in [121, 122]. Applications Main applications include monitoring and detection of flammable gas hazards in industrial, commercial and domestic environments. The lower explosive

4.2 Thermal Sensors

43

limit (LEL) is the concentration of gas in air, below which it cannot be ignited. Target gases include methane [114,118,122], hydrogen [111–115,118], propane [111], carbon monoxide [111, 119], and organic volatiles [116, 117, 120]. The detectable gas concentrations usually range between the lower-few-percent (1–5%) and the some-hundreds-of-ppm region. The interaction process is an irreversible chemical combustion reaction at high temperature liberating the respective reaction enthalpy (2.9, 2.11). 4.2.2 Thermoelectric or Seebeck-Effect Sensors Transduction Principle and Sensing Characteristics This type of sensor relies on the thermoelectric or Seebeck-effect: When two different semiconductors or metals are connected at a hot junction, and a temperature difference is maintained between this hot junction and a colder point, then an open-circuit voltage is developed between the different leads at the cold point. This thermovoltage is proportional to the difference of the Galvani potentials (inner contact potential, difference of the Fermi levels of the two materials) at the two temperatures and, thus, proportional to the temperature difference itself [132]. This effect can be used to develop a thermal sensor by placing the hot junction on a thermally isolated structure like a membrane, bridge, etc., and the cold part on the bulk chip with the thermally well-conducting silicon underneath [95, 133–135]. To achieve a higher thermoelectric voltage, several thermocouples are connected in series to form a thermopile. The membrane structure (hot junctions) is covered with a sensitive or chemically active layer liberating or abstracting heat upon interaction with an analyte. The resulting temperature gradient between hot and cold junctions then generates a thermovoltage, which can be measured. Figure 4.10 displays the schematic of a CMOS thermopile. The sensor system relies on polysilicon/aluminum thermocouples exhibiting a Seebeck coefficient of 111 µV/K. The hot junctions are in the center of the membrane, the cold junctions on the bulk wafer material. The center part (hot junctions) of the membrane is coated with a gas-sensitive layer such as a polymer. The detection process includes four principal steps: (I) absorption and partitioning or chemical reaction, (II) generation of heat, which causes (III) temperature changes to be transformed in (IV) thermovoltage changes (see, e.g., [95, 134]). Each of the four steps contributes to the overall sensor signal. The calorimetric sensor only detects changes in the heat budget at nonequilibrium state (transients) as a consequence of changes in the analyte concentration (for details, see Sect. 5.2.3). Therefore, the sensor provides a signal upon absorption (condensation heat) and desorption (vaporization heat) of gaseous analytes into the polymer [134–138], or during chemical reaction of an analyte with the sensing material [139–141]. The recorded thermovoltage change, ∆U [V], is, therefore, proportional to the derivative of the analyte concentration as a function of time, dcA /dt [mol/m3 s]:

44

4 Microfabricated Chemical Sensors cold junctions

CMOS n-well

polymer

hot junctions

analyte

dielectric membrane

p-substrate

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

dcA . (4.3) dt Here A [K·s/J] and B [V/K] are device- and coating-specific constants describing the translation of a generated molar absorption/reaction enthalpy, ∆H [J/mol], via a temperature change into a thermovoltage change. Vsens denotes the sensitive-layer volume, and K is the partition coefficient (2.10) or reaction equilibrium constant (2.2, 2.11). ∆U = A · B · Vsens · ∆H · K ·

Fabrication • Back etching (KOH, or EDP) for membranes [95, 133, 135, 142]. • Processing of the sensitive layer: Airbrush [95, 135], dispensing, spin coating, or enzyme immobilization methods [139–141]. Processing sequences for the integration of thermoelectric sensors with circuitry in a CMOS standard process are detailed in [95, 135, 142]. Sensors are commercially available from Xensor [143]. Applications Typical applications areas are environmental monitoring (gas and liquid phase) or biosensing in liquids. These include the detection of different organic volatiles in the gas phase (hydrocarbons, chlorinated hydrocarbons, alcohols etc.) by using polymeric layers [95, 135–138] or metal oxides [144], the monitoring of acid/base neutralization [139, 141], and the biosensing of glucose, urea and penicillin in the liquid phase by using suitable enzymes [134, 139–141]. The interaction mechanisms involve reversible physisorption and bulk/gas phase partitioning (see 2.10) as well as enzymatic chemical reactions (2.11) [134, 139–141].

4.3 Optical Sensors

45

4.3 Optical Sensors Light can be considered consisting of either particles (photons) or electromagnetic waves according to the principle of duality. The characteristic properties of the electromagnetic waves such as amplitude, frequency, phase, and/or state of polarization can be used to devise optical sensors [1,4,6,10,147–149]. The energy, E, of an electromagnetic wave is quantized, a quantum being termed a photon (h is Planck’s constant, 6.626 · 1034 Js, ν denotes the frequency): E =h·ν . (4.4) When light interacts with matter, several processes can take place, sometimes simultaneously. Absorption If a sample is irradiated with visible light or electromagnetic waves, the radiation can be absorbed, which results in a decrease of the intensity in the detected radiation as compared to the primary beam (Fig. 4.11). Alternatively, the radiation can be transmitted without attenuation. A prerequisite for absorption is that the absorbing matter (atom, molecule, etc.) exhibits unoccupied energy states with an energetic difference exactly equal or less than the energy of the incoming radiation quanta. The matter then absorbs the radiation energy by transition into a so-called excited state with higher internal energy. The absorption of radiation forms the base for most traditional spectroscopic methods, which are usually distinguished according to the different radiation wavelengths or frequency ranges as given in Table 4.1 [35]. Table 4.1. Different spectroscopy methods and radiation energies Radiation

Energy [J/mol]

Wavelength [m]

Transition

γ-radiation X-rays Ultraviolet (UV) Visible (VIS) Infrared (IR) Microwaves Radio waves

109 –1011 107 –109 106 –107 105 –106 102 –105 10−2 –102 1m

Nucleus excitation Core electron excitation Shell electron excitation Shell electron excitation Vibrational states Rotation states Electron spin, Nuclear spin

The absorption of monochromatic radiation (only one selected wavelength) can be quantitatively determined using the well-known Lambert-Beer relation: (4.5) I = I0 · e−ελ cA l . Here I denotes the transmitted radiation intensity at the detector, I0 the intensity of the incident radiation, ελ is the molar absorptivity at the measured

46


wavelength, cA the analyte concentration, and l the optical path length in the probed volume. Scattering Changes of the direction and/or the frequency of light are commonly denoted as scattering (Fig. 4.11). Scattering of light does not necessarily involve a transition between quantized energy levels in atoms or molecules. A randomization in the direction of light radiation occurs. Particles with sizes that are small compared to the wavelength of radiation give rise to Rayleigh scattering, while particles that are large compared to the wavelength give rise to Mie scattering [147, 148]. In both processes the particle polarization is unaltered. However, the incident radiation can promote vibrational changes (energy quantum absorption), which can alter the polarization of the irradiated particle/molecule. The frequency of the light scattered by these molecules will be different from that of the incident light and the light intensity will be much lower. Such a phenomenon is known as Raman scattering [6, 35, 150]. Fluorescence and Phosphorescence The mechanism of those two phenomena is an absorption-emission process. The wavelength or energy of the incident radiation is absorbed and promotes changes in the molecular energy states. The resulting excited state is unstable, and the molecule dissipates some of its energy to rotational and/or vibrational energy states. The molecule then can return into the ground state by emitting light at a lower frequency than the incident radiation, this process being termed fluorescence. If a more complex and slower intersystem crossing process into a triplet state, and then, a radiative transition from there to the ground state occurs, the process is called phosphorescence. For details and the respective Jablonski diagrams, which represent simplified portrayals of the relative positions of the electronic energy levels of a molecule, see for example [4, 35, 147]. Both processes, fluorescence and phosphorescence are sometimes subsumed under luminescence processes, but the term “luminescence” here will be used exclusively for chemoluminescence processes as detailed below. Fluorescence processes are extensively used in gene analysis techniques, where a defined array of single-stranded deoxyribonucleic acid (DNA) fragments is hybridized with the respective complementary strands labeled with a fluorescent marker. By illuminating the array with a laser, the sites, where the labeled DNA fragments are bound by interaction between the two complementary strands, can be detected by their positive fluorescence response [149]. This technique has been commercialized by several companies [151]. Chemoluminescence The excited state of a molecule (C∗ ) is created by a chemical reaction [4, 10, 152]; the molecule emits light during transition to the ground state according to: (4.6) A + B ⇒ C∗ ⇒ C + h · ν .

4.3 Optical Sensors

47

Chemical energy is thus directly converted into light energy in most cases without additional heat generation (cold luminescence). In the biological domain, this process is denoted bioluminescence and, e.g., occurs in glowworms. Reflection and Refraction Reflection and refraction take place when light infringes on a boundary surface between two media of distinct optical properties (refraction index). The light can either be reflected back into the original medium or be refracted (transmitted) into the adjacent medium (Fig. 4.11). Several distinct types of reflection are possible. The first is a “mirror type” or specular external reflection (Fig. 4.11) occurring at, e.g., a metal surface or generally at interfaces of media with no transmission through (evanescent waves will be treated in the context of refraction below). Another type is diffuse reflection, where the light penetrates the medium and subsequently reappears at the surface after partial absorption and multiple scattering within the medium. The optical characteristics of diffusely reflected radiation provide information on the composition of the reflecting medium [6]. Thin films (10 µm and less) on a surface can strongly affect the propagation of incident light due to reflection at each of the thin film interfaces causing a multitude of reflected, coherent beams with small phase shifts. Sensor techniques to interrogate such thin film structures include ellipsometry [153, 154] and thin-film reflectometric interference spectroscopy (RIFS), which is based upon spectral modulation of the reflectance of a thin film under white-light illumination without using the polarization information. The spectral characteristics are a function of the film thickness, and, therefore, any ad/absorption of organic matter leads to changes in the interferograms of the reflected beams [155, 156]. A variety of transducers respond to changes in the refractive index in immediate vicinity to the device surface. The propagation behavior of a wave guided by nonmetallic total internal reflection (Fig. 4.11) in a medium of high refractivity depends on the dielectric characteristics of the surrounding medium. This effect is mediated by the evanescent field, which penetrates from the optically denser guiding medium a few hundred nanometers into the optically rarer environment [149, 157, 158] (more details in Sect. 4.3.1 on integrated optics, see also Fig. 4.12). If the environment absorbs, energy will be transferred from the evanescent wave to the environment and attenuation of the traveling wave will occur (attenuated total reflection, ATR) [147, 148]. The energy of the evanescent wave can also be used to initiate fluorescence (total internal reflectance fluorescence, TIRF) [157]. Light propagating in waveguide structures without absorbing cover layers is not attenuated by environmental influences (frustrated total reflection, FTR) [147, 148]. Its propagation velocity, however, changes depending on the refractive index in the vicinity of the waveguide. Several

48


scattering

J0 light source

sample

detector

absorption

J

L

reflection and refraction

external reflection

reflected n1 n2 n 1> n2

φ1 φr

φ 1 = φr

φ2 refracted

λ

internal reflection

air

n 1> n2

standing wave

n1

metal

n2 evanescent wave

TIRF

Fig. 4.11. Schematic representation of the different processes taking place upon interaction of light with matter: Absorption, scattering, reflection/refraction, external and internal reflection. n denotes the refraction index, J the light intensity, λ the wavelength and φ the angle. For details, see text and [147–149]

setups have been proposed and will be discussed in Sect. 4.3.1 on integrated optics [147, 149, 158]. Surface plasmon resonance (SPR) is based on collective fluctuations in electron density at the surface of thin films, typically gold or silver, on a waveguide. Surface plasmon waves show the maximum of the electrical field distribution located at the waveguide/metal interface, which is exponentially decaying into the metal and the adjacent medium. SPR is detected as a strong attenuation of the reflected light beam sensitive to the medium adjacent to the metal film [159–161]. SPR will be described in more detail in Sect. 4.3.4. In comparison to other chemical sensing methods, optical techniques offer a great deal of selectivity already inherent in the various transduction mechanisms. Characteristic properties of electromagnetic waves, such as amplitude, frequency, phase, and/or state of polarization can be used to advantage. The wavelength of the radiation, e.g., can be tuned to specifically match the energy of a desired resonance or absorption process. Geometric effects (scattering) can provide additional information. Moreover, optical sensors, like any other chemical sensor, can capitalize on all the selectivity effects originating from the use of a sensitive layer. Optical sensors and classical spectroscopy methods are often very similar in methodology but differ in the arrangement of the experiment and equipment. In particular, the introduction of fiber-optic techniques has promoted the development of comparatively inexpensive optical sensor setups. In the following the focus will be exclusively on semiconductor- and MOEMS(micro-opto-electro-mechanical [162]) based transducers and their respective

4.3 Optical Sensors

49

mechanisms. The wide field of glass-based and fiber-optical techniques such as optodes [163] or micro optodes [164, 165] will not be covered. For interested readers reviews and articles by Wolfbeis and others are recommended [147,148,166–170]. The light-addressable potentiometric sensor (LAPS) [171] will be discussed in the electrochemical sensor section with the field-effect devices (Sect. 4.4.2.2.5). 4.3.1 Integrated Optics Transduction Principle and Sensing Characteristics The generation of light in silicon devices is very difficult since there is no first-order transition from the valence band to the conduction band without the involvement of a phonon (lattice vibrations) [36, 172]. Only directbandgap semiconductors like gallium arsenide (GaAs) or indium phosphide (InP) show first–order radiative electron-hole recombinations with high quantum efficiency (see section on GaAs devices below). The detection of light is possible with either silicon-based devices (photodiodes) or other semiconductor materials. Integrated optical (IO) sensors make use of guided waves or modes in planar optical waveguides. The waveguide materials usually include highrefractivity silicon dioxide or titanium dioxide and silicon nitride films on oxidized silicon wafer substrates. The guided waves or modes in planar optical waveguides include the TE (transverse electric or s-polarized, surfacenormal) and the TM (transverse magnetic or p-polarized, surface-parallel) modes. Changes in the effective refractive index of a guided mode are induced by changes of the refractive index distribution in the immediate vicinity of the waveguide surface, i.e., within the penetration depth (some hundred nanometers) of the evanescent field in the sample (Fig. 4.12a) [149, 158]. The evanescent field decays exponentially with increasing distance from the waveguide surface. Changes in the effective refractive index can be induced by absorption of an adlayer onto the surface of the waveguide from gas or liquid phase [158, 173, 174], by interaction of an analyte molecule with a recognition structure immobilized on the waveguide surface [158,175–179], or by changes of the refractive index of the medium adjacent to the waveguide in a flow-through configuration [158, 175, 176]. In the case of microporous waveguides, analyte molecule absorption or desorption directly into the pores of the wave-guiding film itself can change the waveguide refractive index [158]. A number of different IO sensors have been developed to transform the changes of the effective refractive index into readily measurable physical quantities. Grating Couplers Among the first integrated optical devices were grating coupler structures embossed with a monomode film waveguide as proposed by Lukosz [158,174– 176, 180]. A periodic grating on the surface of the waveguide can be used

50


(a) evanescent wave

(b) grating coupler

fluorophores (TIRF) evanescent sensitive field layer nl waveguide nw substrate ns < nl < nw

flow cell liquid sample

sensitive layer protection

grating waveguide substrate

ns φ

Fig. 4.12. (a) Schematic of an evanescent wave in an optical waveguide. When fluorophores are within the reach of the evanescent wave, they can be excited, and the fluorescence can be detected (TIRF). (b) Schematic of a grating coupler. A periodic grating on the surface of the waveguide is used for in- or out-coupling of radiation to/from the waveguide. The deflection angle depends on the light wavelength and the grating period and is altered by binding of an analyte on the grating. Redrawn from [158]

for in- or out-coupling of radiation (TE and TM mode) from the waveguide. In- and out-coupling are governed by the same physical laws, as the reciprocity theorem permits the reversal of the propagation direction of all light waves. The deflection angles (coupling angles) of the TE and TM modes depend on the light wavelength and the grating period (Fig. 4.12b). Binding of an analyte on the grating alters the coupling angle, which can be detected using position-sensitive detectors. The sensitivity of the grating coupler is related to the lateral dimensions of the grating region interacting with the sample. Prism couplers [180] will not be treated here. Difference Interferometer In a planar waveguide, the TE and TM mode are coherently excited by a laser. Both propagate along a common path down the same waveguide and interact with the sample within a certain length of the waveguide. The polarization-dependent interaction induces a phase difference between the two modes, which can be measured using a dedicated interferometer setup [158, 173, 174, 181, 182]. A variant of this method involves a Zeeman laser to generate two orthogonally polarized modes in a silicon nitride waveguide [183, 184].

4.3 Optical Sensors

(a) Mach-Zehnder interferometer sensitive layer

51

(b) integrated GaAs interferometer

Si-oxide Si-nitride Si-oxide

sensor area

Si substrate electric field

phase modulator

reference branch input

output

DBR laser

sensor pad

detector

L

L sensor branch

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

Two-Beam Interferometers Mach-Zehnder IO-devices (Fig. 4.13a) are monomode channel waveguides (TE or TM mode) and allow for a straightforward implementation of an interferometer structure [158,177–179,185]. A waveguide is split into an open measurement path and a protected reference path and recombined after some distance. The phase difference, introduced by analyte interaction (refractive index change) in the sensing path, is detected by interference effects. Detection limits are in the range of some picograms/mm2 [186]. Integrated Waveguide Absorbance Optodes (IWAO) A membrane inserted between two micromachined waveguides acts simultaneously as the light-guiding medium and sensing element and hence changes its spectral properties while interacting with an analyte. First results with potassium-selective optode membranes are reported in [187]. Fabrication • Patterning of silicon nitride as waveguide (LPCVD, RIE, lithography) [158, 177–179] • Deposition of an silicon oxide cladding layer (PECVD) [158, 177–179] • Deposition of the chemically sensitive layer (immobilization of biological entities) [175–179] Electro-optical modulation techniques of the sensor signal, when using zinc oxide as the optical waveguide on a chip fabricated in silicon oxinitride technology, have been reported in [188, 189].

52


∆Φ[2π]

-10

PBS

10 µg/ml avidin 10 µg/ml biotinylated protein A 0 rabbit anti-atrazine 1 serum 1:100 2

PBS

-20 PBS

3 25 µg/ml atrazine-HRP

-30 PBS

4

-40

surface mass density [ng/mm2]

0

5 0

100

200 300 time [min]

400

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

Applications Typical applications are humidity sensors [158,173,174], gas sensors [184,185] (adsorption on the device or absorption in a microporous waveguide), and environmental monitoring or biosensing. Examples include the detection of different organic solvents in the liquid phase (hydrocarbons, alcohols etc.), the monitoring of sucrose and buffer solutions [158, 173, 175, 176], and biotin/streptavidin-mediated immunosensing (Fig. 4.14) involving antibody/antigen binding experiments [158, 173–179]. The interaction mechanisms include reversible physisorption (2.10) as well as biochemical affinity reactions (2.11) [158, 175–179]. Gallium-Arsenide-Based Devices Due to their direct band gap, III-V-semiconductors offer the opportunity of fabricating and integration of lasers, waveguides, phase modulators and waveguide detectors on the same chip. GaAs/AlGaAs-based Mach-Zehnder devices with integrated light sources and detectors have been developed as shown in Fig. 4.13b [190–192]. The light source is a distributed Bragg reflector (DBR) laser, which was fabricated with a simplified grating recess technology [193, 194] and is operating on a single mode. A dielectric waveguide pad (silicon oxide, tantalum oxide) is integrated in the measurement arm of the interferometer [190–192]. The detector is a long-absorbing-length photodiode with high quantum efficiency [193]. A more recent development for optical gas sensing is the vertical-cavity surface-emitting laser (VCSEL). The cavity is formed vertically on the wafer

4.3 Optical Sensors

53

surface. Epitaxially grown Bragg mirrors serve as distributed reflectors above and below the laser’s very short active region. GaAs/AlGaAs-based VCSELs emit in the near infrared region. Oxygen sensing at 762 nm (absorption due to magnetic dipole transitions in the gas molecule without interference from other gases) was demonstrated in first experiments [195, 196]. 4.3.2 Microspectrometers 4.3.2.1 Fabry-Perot-Type Structures Transduction Principle and Sensing Characteristics A Fabry-Perot interferometer (FPI) is an optical element consisting of two partially reflecting, low-loss, parallel mirrors separated by a gap. The optical transmission characteristics through the mirrors consist of a series of sharp resonant transmission peaks occurring when the gap is equal to multiples of a half wavelength of the incident light. These transmission peaks are caused by multiple reflections of the light in the cavity. By using highly reflective mirrors, small changes in the gap (width, absorptivity) can produce large changes in the transmission response. Even though two reflective mirrors are used, transmission through the element at the peak wavelengths approaches unity. The transmission is a function of both, the gap spacing and the radiation wavelength. The devices can be used as wavelength selector or monochromator by adjusting the gap width to achieve the desired wavelength. Tunable devices with a gap width variable by electrostatic actuation using electrodes on movable micromachined parts have been reported (Fig. 4.15a) [197–200]. Such devices operate preferably in the near infrared region at wavelengths larger than 1 µm, where silicon substrates become transparent [197]. (a) tunable Fabry-Perot

(b) CMOS Fabry-Perot etalon

radiation antireflection coating

silver wafer

optical control coating (Al) electrodes

PECVD oxide

gap

fusionbonded layer corrugated support

movable mesa

n-well p+ implanted layer (SP)

Fabry-Perot cavity

aluminum

p -epilayer

}

}pnp-phototransistor

p+ substrate

Fig. 4.15. (a) Schematic of a tunable Fabry-Perot interferometer. The wavelengthdefining gap width can be changed by applying DC to the control electrodes. Aluminum is used as optical coating material. Redrawn from [197]. (b) Cross section of an integrated Fabry-Perot etalon. The gap width is determined by the thickness of the PECVD oxide, silver and aluminum are used as optical coatings. A pnp-phototransistor (p+ implanted layer/n-well/p-epilayer) is located directly underneath the etalon. Redrawn from [201]

54


A single-chip CMOS optical microspectrometer based on FPI and operating in the UV/VIS-region is reported in [201, 202]. It contains an array of 16 addressable Fabry–Perot etalons (500 × 500 µm2 ) each with a different resonant cavity length realized as a PECVD silicon oxide layer sandwiched in between an aluminum and a silver layer and placed on top an array of vertical pnp phototransistors (Fig. 4.15b) [203]. It additionally includes circuits for readout, multiplexing and driving a serial bus interface. Fabrication • • • •

Surface micromachining techniques to achieve an air gap (HF) [200, 202] Deposition of the lower mirror (evaporation and lift-off of aluminum) [201] Deposition of a silicon oxide layer of defined thickness (PECVD) [201] Deposition of the upper mirror layer (silver) [201]

Some devices are derived from air gap pressure sensors [198–200, 204, 205], the fabrication sequences of which can then be applied. The same holds for wafer-stacking techniques to achieve the FPI cavities [198,199]. The complete fabrication sequence of a monolithic CMOS VIS spectrometer is given in [201]. Applications Typical applications include gas sensors [198–200, 203–207]. The characteristic absorption wavelengths of carbon monoxide are 4.7 µm, that of carbon dioxide 4.2 µm, and that of methane or hydrocarbons 3.3 µm (IR region, molecular vibrations). Polymeric coatings in the FPI cavity have been used to detect iodine [204, 205]. Carbon dioxide sensors based on tunable FPIs (dual wavelength measurements) are commercially available from [207]. The radiation source in most cases is a light bulb or light-emitting diode (LED). 4.3.2.2 Grating-Type Structures Transduction Principle and Sensing Characteristics Micromachined diffraction gratings have been used in combination with imaging devices to set up microspectrometer arrangements [208–210], see Fig. 4.16a for a schematic. Two basic types of gratings can be easily micromachined: (1) amplitude gratings by blocking out light with an array of opaque and transparent sections, or (2) phase gratings where the light phase is modulated by variations in the grating shape [208, 209]. In the example displayed in Fig. 4.16a [210], a phase grating with a fine grating pitch creating high dispersion angles was etched into a quartz wafer. The transmission grating was then mounted directly over a charge-coupled-device (CCD) imager [208], or a CMOS imaging chip consisting of an array of custom-designed photodiodes [210]. A glass spacer is placed in the optical path between the grating and the detector. In another approach, two wafers have been micromachined so that an optical path of about 4 mm length was obtained. Light dispersed by a 32-slit diffraction grating travels along the optical path and is directed to an array of photodiodes (Fig. 4.16b) [211]. The interior of one of

4.3 Optical Sensors

(a) CMOS microspectrometer diffraction grating

photodetector array

blue

grating

glass spacer

CMOS imager chip

(b) integrated spectrometer

red

incident light

55

bevel p-wafer mirror

increasing

wavelength

n - epilayer

Fig. 4.16. (a) CMOS microspectrometer consisting of a glass and a CMOS imager chip: Different wavelengths are diffracted at different angles from the surface normal. Redrawn from [210]. (b) Schematic of an integrated spectrometer using two fusion-bonded chips, one of them carrying a grating and a photodetector array. Redrawn from [211]

the wafers is coated with a reflective film. The grating and the diode array are integrated in the second wafer, which remains uncoated. The wafers are bonded together by silicon/silicon fusion bonding. The performance of such micromachined spectrometers is comparable to that of low-end bench-top spectrometers. Diffractive optical elements, which produce analyte infrared spectra of compounds such as hydrogen fluoride (HF) for chemical sensor systems based on correlation spectroscopy, are reported in [212]. Fabrication • • • •

Patterning of gratings, quartz micromachining [208, 211] Back etching using electrochemical etch stop [211] Deposition of reflective layers [211] Wafer fusion bonding [211]

Applications Typical applications include biochemical and chemical analysis. Emission gas spectra have been recorded for carbon dioxide and helium. Fluorescence of a dye (fluorescein) was induced by a laser and detected with the microspectrometer [208]. 4.3.3 Bioluminescent Bioreporter Integrated Circuits (BBIC) Transduction Principle and Sensing Characteristics This technique employs bioluminescent bacteria placed on an applicationspecific optical integrated circuit (standard CMOS) [213–215]. The bacteria have been engineered to luminesce (4.5) when a target compound such

56


signal processing circuit

photodetectors

Fig. 4.17. Micrograph of a bioluminescent bioreporter integrated circuit (BBIC). For details, see text. Reprinted from [214] with permission

as toluene is metabolized. The integrated circuit detects, processes, and reports the magnitude of the optical signal. The microluminometer uses the p-diffusion (source and drain diffusions of p-channel MOSFETs) as the photodiode. The shallow p-diffusion has a strong response to the 490-nm bioluminescent signal. The entire sensor including all signal processing and communication functions can be realized on a single chip. The integrated circuitry contains the subunits needed to detect the optical signal, to perform analog or digital signal processing, to communicate the results, and to perform auxiliary functions (temperature, position measurement) (Fig. 4.17). Many types of bioluminescent transcriptional gene fusions have been used to develop light-emitting bioreporter bacterial strains to sense the presence, bioavailability, and biodegradation of different kinds of pollutants. The cells here were entrapped on the chip by encapsulation in natural or synthetic polymers providing a nutrient-rich hydrated environment [213–215]. Fabrication • Silicon nitride protective coating using a jet vapor-deposition technique [216] • Deposition of the cell-containing polymer (drop coating, spraying, beads) [213–215] Applications Typical applications include chemical analysis in gas or liquid phase. Depending on the integration time of the device, trace amounts of toluene and naphthalene were detected in the gas phase using suitable cell colonies (Pseudomonas putida) [213–215]. The interaction mechanism is a chemical reaction (2.11, 4.5).

4.3 Optical Sensors

57

4.3.4 Surface Plasmon Resonance (SPR) Devices Transduction Principle and Sensing Characteristics The quantum optical-electronic basis of SPR is due to the fact that the energy carried by photons of light can be “coupled” or transferred to electrons in a metal [159]. The wavelength of light, at which coupling (i.e., energy transfer) occurs, is characteristic of the particular metal and the environment, in which the metal surface is illuminated; gold being the preferred metal. The coupling can be observed by measuring the amount of light reflected by the metal surface. All the light is reflected except at the resonant wavelength, where almost all the light is absorbed (Fig. 4.18). The coupling of light into a metal surface results in the creation of a plasmon, a group of excited electrons, which behave like a single electrical entity. The plasmon, in turn, generates an electrical field, which extends about 100 nm above and below the metal surface [159]. The characteristic of this phenomenon, which makes SPR an analytical tool, is that any change in the chemical composition of the environment within the range of the plasmon field causes a change in the wavelength of light that resonates with the plasmon. That is, a chemical change results in a shift in the wavelength of light, which is absorbed rather than reflected, and the magnitude of the shift is quantitatively related to the magnitude of the chemical change [161, 217]. The phenomenon of SPR is non-specific. Different chemical changes cannot be distinguished.

light source

φ polarized light

prism

1 2

optical detection unit

reflected light

intensity

SPR monochromatic, angle variation

1

2

sensor chip metal film

angle

flow channel

Y

1 : receptor only

2 : receptor and target analyte

Fig. 4.18. Surface plasmon resonance (SPR) principle and a typical sensor response diagram: Intensity versus angle. Monochromatic light is used to excite the surface plasmon, which leads to a drastic intensity decrease at a defined reflection angle. Adsorption of an analyte changes this angle

A hybrid SPR system consisting of two micromachined silicon layers, one micromachined glass layer and an alumina substrate, is shown in Fig. 4.19 [218, 219]. Bonding was achieved using a low-temperature-curing polyimide and a solder sealing. The micromachined silicon layers contain a torsional,

58


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

all-silicon micromirror, V-grooves for optical fiber and lens, and a positionsensing photodiode (PSD). The silicon micromirror was electrostatically deflected through 9–10 degrees to direct the light beam emitted from the end of a fiber through a range of angles incident onto the metal film, setting up a surface plasmon. The position and intensity of the reflected beam was recorded with the position-sensing photodiode. The microsystem measures approximately 1 × 2 × 0.2 cm3 [218, 219]. Fabrication • Alumina substrate: Laser drilling, screen printing of thick-film conductors [218] • Anisotropic etching of the silicon micromirror and the fiber-to-lens alignment groove [218] • Filling the interior of the assembly with index-matching fluid [218, 219] • Glass slide: Deposition of the sensitive layer (drop coating, spraying, beads) [160, 161] Small SPR instruments are commercially available from Texas Instruments [220]. Applications Any pair of molecules that exhibit specific binding can be adapted to SPR measurements. These may be an antigen and antibody, a DNA probe and complementary DNA strand, an enzyme and its substrate, or a chelating agent and a metal ion. Typical applications areas are environmental monitoring (gas sensing [160,218]) or bio- and immunosensing in liquids [160,161].

4.4 Electrochemical Sensors

59

With the micromachined instrument the surface adsorption of bovine serum albumin (BSA) was tested [218, 219]. The interaction mechanisms involve specific biochemical reactions (2.11) [160, 161].

4.4 Electrochemical Sensors Electrochemical sensors constitute the largest and oldest group of chemical sensors [4, 6, 8–11, 23, 26, 34]. They rely on electrochemical or charge-transfer reactions: A+ + e− ↔ A (see 2.12–2.15 and related text in Chap. 2). Electrochemistry includes charge transfer from an electrode to a solid or liquid sample phase or vice versa. Chemical changes take place at the electrodes or in the probed sample volume, and the resulting charge or current is measured. Both, electrode reactions and charge transport in the sample are subject to changes by chemical processes, and, hence, are at the base of electrochemical sensing [4]. A key requirement for electrochemical sensors is a closed electrical circuit, though there may be no current flow (see potentiometry below). An electrochemical cell is always composed of, at least, two electrodes with two electrical connections: One through the probed sample, the other via transducer and measuring equipment. The charge transport in the sample can be ionic, electronic or mixed, while that in the transducer branch is always electronic. Electrochemical sensors are usually classified according to their electroanalytical principles [4, 8, 11]. Voltammetry Voltammetric sensors are based on the measurement of the current-voltage relationship in an electrochemical cell comprising electrodes in a sample phase. A potential is applied to the electrodes, and a current is measured, which is proportional to the concentration of the electro-active species of interest. Amperometry is a special case of voltammetry, where the potential is kept constant as a function of time. Potentiometry Potentiometric sensors are based on the measurement of the potential at an electrode, which is, in most cases, immersed in a solution. The potential is measured at equilibrium state, i.e., no current is allowed to flow during the measurement. According to the Nernst equation (2.15), the potential is proportional to the logarithm of the concentration of the electro-active species (Work function sensors will be discussed in the context of field effect devices in Sect. 4.4.2.2.4). Conductometry Conductometric sensors are based on the measurement of a conductance between two electrodes in a sample phase. The conductance is usually measured

60


by applying an AC potential with a small amplitude to the electrodes in order to prevent polarization. The presence of charge carriers determines the sample conductance. In contrast to conductometry, AC-impedance measurements are not really used for analytical applications. Impedance measurements are of special interest for membrane, electrode and electrolyte characterization. The goal of impedance measurements is to find an equivalent electronic circuit model and to correlate that model with electrochemical phenomena. Another categorization method relies on discerning the electronic components [9, 34]. There are chemoresistors, chemodiodes, chemocapacitors and chemotransistors. Within this book the electroanalytical principles will be used as the superordinated classification scheme, and the component notation will be used within this scheme. Again, the focus will be on semiconductor-based systems, and a wealth of literature on, e.g., other designs of ion-sensitive electrodes [221], or siliconcarbide-based devices operating at extremely high temperature [222–225] will be omitted. Review articles are recommended to explore further details (see, e.g., [226–231]). 4.4.1 Voltammetric Sensors Transduction Principle and Sensing Characteristics Voltammetry, in general, is the measurement of the current that flows at an electrode as a function of the potential applied to the electrode. The result of a voltammetric experiment is a current/potential curve. Amperometry is more frequently applied in chemical sensors and provides a linear current/analyte concentration relationship at a constant potential, which is predefined with regard to the target analyte. Two different electrode configurations are normally used. The two-electrode configuration [4, 8, 11] consists of a reference electrode (RE) and a working electrode (WE) (Fig. 4.20a) [11]. The disadvantage of this method is, that the RE carries current and may become polarized if it is less than one hundred times the size of the WE. Material consumption due to the current in the RE is another problem. A better approach is, therefore, the use of a three-electrode-system [4, 8, 11] in a potentiostatic configuration. An additional auxiliary electrode (AE, sometimes denoted counter electrode, CE) is introduced for current injection in the analyte (Fig. 4.20b) [11]. The reference electrode is now a true RE with a well-defined potential since no current is flowing through the RE. The potentiostat controls the current at the auxiliary electrode as a function of the applied potential. This is realized in practice with an operational amplifier (opamp) [11]. The potential is applied to the positive input of the opamp. The RE is connected to the negative input and measures the potential in the solution. The AE is connected to the output. The opamp injects a current into the solution through the AE. Due to the feedback mechanism, the current is controlled in such a way that the potential at the negative input equals


2 - electrode system (a)

61

3 - electrode system

U

+ _

(b)

U

WE RE

WE RE AE

Fig. 4.20. Schematic of a two-electrode (a) and a three-electrode configuration (b) used for voltammetric measurements. For details, see text

the potential at the positive input. The potential difference between WE and RE hence equals the applied potential. No current is flowing through the RE since the opamp has a very high input impedance. The sensor signal current is measured at the working electrode. The measured current at any given potential difference depends on the material properties, the composition and geometry of the electrodes, the concentration of the electro-active species (presumably the target analyte) and the mass transport mechanisms in the analyte phase [4, 8, 11, 34]. Among those are migration, the movement of charged particles in an electric field, convection, the movement of material by forced means like stirring or as a consequence of density or temperature gradients, and diffusion, the movement of material from high-concentration regions to low-concentration regions. The electrochemical reactions at the electrodes are normally fast in comparison to the transport and supply mechanisms. Since convection in the electrode vicinity is avoided, and migration is suppressed by, e.g., a large excess of electro-inactive salts (i.e., electro-inactive at the respective applied potential), diffusion is normally regarded to be the dominant mechanism. There are two components to the measured current, a capacitive component resulting from redistribution of charged and polar particles in the electrode vicinity, and a component resulting from the electron exchange between the electrode and the redox species (analyte) termed faradaic current [4,8,11,34]. The faradaic component is the important measurand and is, for the case of diffusion-limited conditions, directly and linearly proportional to the target analyte concentration. The limiting current (all analyte ions are immediately

62


charged or discharged upon arrival at the electrode) is then given by the Cottrell equation [4, 8, 11, 26, 34]: I∞ = ne · F · A · Ddiff

cA . Ldiff

(4.7)

Here, ne denotes the number of electrons, F is the Faraday constant, A the effective electrode area, Ddiff the diffusion coefficient, cA the target analyte concentration and Ldiff the diffusion length (for more mechanistic details see [4,8,11,26,34]). Correction terms to this equation for small electrodes have to be introduced to take into account the respective electrode geometry [4,8,26]. Cyclic Voltammetry A stationary WE is used, and a cyclic ramp potential versus some RE is applied. The potential versus time is triangular: It increases at a rate linear with time, then reverses and decreases at the same rate. The current flows as a consequence of the applied potential between the WE and an AE. This technique is used to study the electrode/sample interface [4, 8, 11, 232–235]. Stripping Voltammetry This method is used to detect heavy-metal ions at trace level by means of a mercury electrode [236–239]. The method involves an initial preconcentration phase, in which the array is held at a cathodic potential such that the metal ions from solution are reduced and amalgamated into the mercury. Then, the electrode potential is reversed to anodic, and the metals in the mercury are re-oxidized and stripped from the mercury into the solution. The charge required to strip a given metal completely from the mercury is proportional to its initial concentration in the test solution [236–239]. Fabrication • Additional silicon nitride as protective coating • Optional back etching, membrane formation and perforation for liquid electrolyte access to membrane-covered, sensitive electrodes or gas permeation [240, 241] • Deposition/patterning of metal electrodes (lift-off, thermal evaporation, sputtering) [8, 11, 240–256] • Deposition of electro-active polymers, membrane materials, hydrogels (spincasting, spraying, screen printing) [240–257] Sensor processing sequences are given in [11, 240, 241, 251, 253]. The fabrication of electrochemical sensors that are integrated with CMOS circuitry components is described in [11, 251]. Microsensor arrays with up to 1024 individually addressable elements have been reported on [254, 255]. A picture and a schematic of a CMOS-based 3-electrode amperometric sensor are shown in Fig. 4.21 [251]. The monolithic device includes the electrochemical sensor, a temperature sensor, and interface circuitry. The circuitry contains an operational amplifier as potentiostat, a switched-capacitor


63

(a)

(b)

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

current-to-voltage converter and a clock generator. Interface circuitry and the temperature sensor are realized in 3 -µm CMOS-technology. The circuitry needs a supply voltage of ±2.5 V, can apply voltages from +1 V to −1 V to the sensor and handles currents from 30 nA full scale to 1 µA full scale. The output voltage of the temperature sensor is proportional to the absolute temperature and has a sensitivity of 125 µV/K. The total sensor dimensions are 0.75 mm by 5 mm. Applications Typical applications include chemical analysis in the gas or liquid phase. If the target analyte is not an electro-active species like, e.g., glucose, oxygen, or carbon dioxide, then polymer electrolytes or enzymes (glucose oxidase) producing analyte-related ionic species are used as components of the sensitive electrode coatings. Typical target analytes in the gas phase are nitrogen oxides, hydrogen sulfide, using Nafion polymer electrolyte [241–243], as well as oxygen [248, 253, 255], and carbon dioxide using liquid electrolytes [240]. Target analytes in the liquid phase comprise dissolved oxygen [235, 251], glucose [245, 246, 251], hydrogen peroxide [240, 247], and chlorine in drinking water [244, 249] Fig. 4.22. One of the best-known voltammetric cells is the Clark cell, which is based on a two-step-reduction of oxygen via hydrogen peroxide to hydroxyl ions in aqueous solution. The Clark cell is used to measure dissolved oxygen in blood and tissue [258]. The reference electrodes in the liquid phase are in most cases silver/silver-chloride elements. The interaction mechanism in all cases is an electrochemical redox reaction (2.12–2.15).

64


100

(b)

PDMS

WE

75 CE

RE

current [pA]

(a)

50 25 0 hypochlorous acid -25 0

5

10 15 time [min]

20

25

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

4.4.2 Potentiometric Sensors Potentiometry is the direct application of the Nernst equation (2.15) through measurement of the potential between nonpolarized electrodes (WE and RE) under conditions of zero current. The measurement is carried out at thermodynamic equilibrium. In the following, it will be distinguished between two different types of potentiometric devices: • Electrochemical cell with metal electrodes. • Field-effect semiconductor devices. 4.4.2.1 Electrochemical Cell Transduction Principle and Sensing Characteristics The electrochemical cell used for potentiometric microsensors consists of two metal electrodes, a WE covered with a ion-selective membrane or gel, which preferably hosts a specific target ion, and an RE, which, in most cases, is a silver electrode covered by a thin silver chloride film [259, 260]. The WE is termed an ion-selective electrode (ISE). Both electrodes are on the same chip and are simultaneously exposed to the analyte phase (Fig. 4.23a). The ISE is, in principle, the oldest solid-state chemical sensor [261]. The design of modern micromachined nonsymmetrical ISEs, however, is completely different from that of conventional symmetrical ISEs such as a pH-glass-electrode or a


(a) electrochemical cell

(b) concentration cell electrode 1

ion-selective membrane

Ag/AgCl reference

65

electrode 2

perm-selective membrane

metal electrodes (Pt) silicon substrate solution 1

solution 2

a1 (A+)

a2 (A+)

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

lanthanium-fluoride membrane electrode. The traditional symmetric arrangements exhibit a liquid phase reservoir separated by a perm-selective membrane from the analyte phase and, thus, essentially constitute electrochemical concentration cells (Fig. 4.23b). The charge transfer processes at all interfaces (solution/solid electrolyte, solid electrolyte/metal, ionic conductor/electronic conductor) of a modern, nonsymmetrical ISE must be carefully designed. If the exchange current den2 sity of the charged species of interest is sufficiently high (i.e., >10−3 A/cm ), such interfaces are well defined, and the devices are stable. The exchange current is due to significant and continuous movement of charge carriers in both directions through the interphase region at an electrode at equilibrium (dynamic equilibrium). The magnitude of these mutually compensating currents (no net current) flowing at any zero-current potential is called exchange current. The Nernst equation (2.15) can be applied to calculate the electrochemical potential evoked by a certain analyte concentration. The design of metal-electrode potentiometric sensors is very similar to the voltammetric sensors (two-electrode configuration) described in Sect. 4.4.1. In comparison to voltammetric techniques it should be noted that the concentration dependence of the measured potential is logarithmic (2.15). Using amperometric techniques, the target ions can be selected by careful tuning of the appropriate redox potential. In potentiometry, all ions in the sample exhibiting comparable exchange current density contribute to the measured overall potential. In order to achieve selectivity to one specific ion, this target ion must provide a significantly higher exchange current density than the

66


other interfering ions. This condition can be achieved by incorporating selective binding sites or ionophores in a membrane or gel material. The sensitive coating hence has to provide selectivity. Fabrication The fabrication is very similar to that of voltammetric devices and is, therefore, not further specified here. Sensor processing sequences are given in [259, 262, 263]. The fabrication of potentiometric sensors integrated with CMOS circuitry components is described in [263, 264]. Applications Prototype applications include chemical analysis in the gas or liquid phase. Typical target analytes in the gas phase are nitrogen oxide, sulfur dioxide and carbon dioxide using sintered ceramic electrolytes (sodium, barium and silver sulfate or Nasicon) [262, 265]. Target analytes in the liquid phase comprise all kinds of ionic species like hydrogen (pH: “potentia hydrogenii”, negative decadic logarithm of the hydrogen ion concentration), potassium, ammonium, calcium, chloride, cyanide, or nitrate using ionophores in polymeric membranes [259, 260, 263, 264, 266] or chalcogenide glasses [267]. The interaction mechanism in all cases is an electrochemical chargetransfer reaction (2.12–2.15). 4.4.2.2 Field-Effect-Based Devices Transduction Principle and Sensing Characteristics The field-effect-based microfabricated potentiometric sensors, like metaloxide semiconductor (MOS) devices in electronics, rely on variations in the charge distribution within the semiconductor surface space-charge region. The three field-effect device structures commonly used include the MOS capacitor (MOSCAP), the MOS diode (MOS-diode), and the MOS fieldeffect transistor (MOSFET) [4, 8, 11, 34]. These structures are displayed in Fig. 4.24. It is important to point out, “that one of the most critical thrusts of the electronics industry’s efforts to develop commercial field-effect devices has been predicated by the need to isolate these devices from any variation in their chemical environment” [34], since the field-effect device characteristics are very sensitive to such variation. Those efforts of the electronics industry are in diametrical opposition to developing, e.g., FET-based chemical sensors. By replacing, the MOSFET metal gate with an ionic solution and a reference electrode immersed into this solution, ion-sensitive device structures have been developed, the basic structure of which is analogous to the respective MOS devices: Ion-sensitive capacitor (ISCAP), ion-controlled diode (ICD) and ion-sensitive field effect transistor (ISFET) (Figs. 4.24, 4.25). The gate region is exposed to any ion present in the solution [4, 8, 11, 34]. The device applications will be discussed in the individual device-related sections, whereas the fabrication of the family of field-effect devices will be detailed in summary at the end in Sect. 4.4.2.2.6.


(a) MOSCAP

(b) MOS-Diode

gas phase

p-silicon

gas phase metal gate (Pd)

metal (Pd) U

Si-oxide

(c) MOSFET

gas phase

metal (Pd) U

67

Ug Si-oxide

gate oxide n n source drain

p-silicon

p-silicon Ud

capacitance

current

drain voltage constant

∆Ugas

∆Ugas

∆Ugas

hydrogen no hydrogen

hydrogen no hydrogen voltage

drain current

hydrogen no hydrogen voltage

gate voltage

Fig. 4.24. Schematic representation of the different MOS field-effect devices: (a) capacitor, (b) diode, (c) transistor. Ug denotes the gate voltage, Ud the source-drain voltage. Characteristic sensor responses (voltage shift upon exposure to hydrogen) are given at the bottom

4.4.2.2.1 MOS Field-Effect Transistors, MOSFETs, and Ion-Selective Field-Effect Transistors, ISFETs (Chemotransistors) MOSFET Transduction Principle and Sensing Characteristics A MOSFET in electronics is a transistor, the source drain current (conductance) of which is modulated through an electric field perpendicular to the device surface. This electric field is generated by an isolated gate electrode and influences the charge carrier density in the conductance path between source and drain (semiconductor field effect). The MOSFET as used for chemical sensing has, e.g., a p-type silicon substrate (bulk) with two n-type diffusion regions (source and drain). The structure is covered with a silicon-dioxide insulating layer, on top of which a metal gate electrode (originally palladium [22], later other platinide metals [268, 269, 271, 272]) is deposited. When a positive voltage (with respect to the silicon) is applied to the gate electrode, holes, which are the majority carriers in the p-substrate, are depleted near the semiconductor surface. Upon applying a voltage between drain and source (Ud ), the electrons from the n-doped source can pass through this depleted surface region to the drain so that a conducting n-channel between source and drain is generated in the p-substrate near the silicon/silicon dioxide interface. The conductivity of this n-channel, i.e., the magnitude of

68


the source-drain current (Id ) can be modulated by adjusting the strength of the electrical field perpendicular to the substrate surface between gate electrode and the silicon substrate [22, 268, 269]. Applications Palladium (Pd)-gate FET structures were demonstrated to function as hydrogen sensors by Lundstr¨ om [22] and others [270,273,274]. Hydrogen molecules readily absorb on the gate metal (platinum, iridium, palladium) and dissociate into hydrogen atoms. These H-atoms can diffuse rapidly through the Pd and absorb at the metal/silicon oxide interface partly on the metal, partly on the oxide side of the interface [268,269]. Due to the absorbed species and the resulting polarization phenomena at the interface, the drain current (Id ) is altered, and the threshold voltage (Ud ) is shifted. The voltage shift (∆Ud ) is proportional to the concentration or coverage of hydrogen at the oxide/metal interface. The presence of oxygen promotes the formation of water at the gas phase/metal interface due to the catalytic reaction of atomic hydrogen with atomic oxygen. With thin catalytic metal gates (containing holes and cracks) ammonia [269,276], amines, and any kind of molecule that gives rise to polarization in a thin metal film (hydrogen sulfide, ethene, etc.) or causes charges/dipoles on the insulator surface, can be detected [278,279]. Detailed models for those processes, however, do not yet exist. Sensitivity and selectivity patterns of gas-sensitive FET devices depend on the type and thickness of the catalytic metal used, the chemical reactions at the metal surface, and the device operation temperature. For extremely high temperatures (600◦ –800◦ C), silicon carbide devices have been developed [222–225]. A paramount problem with MOSFET sensors has been long-term drift, which seems to be mitigated to some extent by the deposition of a thin alumina layer between the Pd gate and the silicon oxide [277]. The temperature should be kept constant. Alternative gate materials include polyaniline for the detection of water and ammonia [278, 279], and high-temperaturesuperconducting cuprate to monitor ammonia and nitrogen oxides [280]. ISFET Transduction Principle and Sensing Characteristics For the case of the ISFET, the gate metal electrode of the MOSFET is replaced by an electrolyte solution, which is in contact with the reference electrode, i.e., the silicon gate oxide is directly exposed to aqueous electrolyte solution Fig. 4.25 [21]. An external reference electrode is required for a stable operation of an ISFET [4, 8, 11, 34, 281–284]. Unfortunately, including such a reference is nontrivial and subject to dedicated research [34, 285–289]. An electric current (Id ) flows from the source to the drain via the channel, and, like in MOSFETs, the channel resistance depends on the electric field perpendicular to the direction of the current. It additionally depends on the potential difference across the gate oxide. Therefore, the source-drain current, Id , is influenced by the interface potential at the oxide/aqueous solution boundary. Though the electric resistance of the channel provides a measure


(a) MOSFET

(b) ISFET

gas phase metal gate (Pd) Ug

69

U

reference electrode liquid phase



p-silicon Ud

p-silicon Ud

Fig. 4.25. Schematic representation of a MOSFET (a) and an ISFET structure (b). Ug denotes the gate voltage, Ud the source-drain voltage. By replacing the metal gate of the MOSFET with an ionic solution and a reference electrode immersed into this solution, the ISFET has been developed

for the gate oxide potential, the direct measurement of this resistance gives no indication of the absolute value of this potential. However, at a defined source-drain potential (Ud ), changes in the gate potential can be compensated by a modulation of Ug . This adjustment can be carried out in a way that the changes in Ug applied to the reference electrode exactly compensate for the changes in the gate oxide potential. This is automatically performed by ISFET amplifiers with feedback, which allow for obtaining a constant source-drain current. In this particular case, the gate-source potential is determined by the surface potential at the insulator/electrolyte interface. Mechanistic studies of the processes occurring at the solution/gate oxide interface (site binding model [290]) and the oxide/semiconductor interface can be found in the literature [4, 8, 11, 34, 283, 284, 290–293]. The insulator/solution interface is assumed to represent in most cases a polarizable interface, i.e., there will be charge accumulation across the structure but no net charge passing through.Interfaces with net charge passing through are termed “faradaic”, see, e.g., [4, 8, 11, 34, 229]. Applications The gate oxide surface contains reactive Si-OH groups, which, besides providing pH sensitivity, can be used for covalent attachment of a variety of organic molecules and polymers. pH-FET (pH: “potentia hydrogenii”, negative decadic logarithm of the H+ -ion concentration).The basic ISFET is an exposed-gate-oxide FET and functions as a pH sensor [21]. The surface of the gate oxide contains OHfunctionalities, which are in electrochemical equilibrium with ions in the sample solutions (H+ and OH− ). The hydroxyl groups at the gate oxide surface can be protonated and deprotonated and, thus, when the gate oxide contacts an aqueous solution, a change of pH will change the silicon oxide surface potential. A site-binding model describes the signal transduction as a function

70


sensor signal [mV]

-1450

-1470

-1490

-1510 0

20

40

60 80 time [sec]

100

120

140

Fig. 4.26. Dynamic response of a pH-ISFET in a flow-through configuration upon repeated pH-changes by one unit (4 to 5). The signals exhibit a stable baseline and a good reproducibility. According to the Nernst equation (2.15), a pH-change of one unit causes a voltage change of 59 mV. Reprinted from [299] with permission

of the state of ionization of the amphoteric surface Si-OH groups [294, 295]. The change in the charge state of these sites leads to a variation of the dipole layer and consequently to a change in the semiconductor space charge region [4, 34, 276, 293]. Typical pH-sensitivities measured with silicon-oxide ISFETs are 37-40 mV per pH unit [295]. Sensor signals achieved with a silicon nitride membrane are given in Fig. 4.26 [299]. Inorganic gate materials for pH sensors like silicon nitride (CMOS process material) [296–302], oxynitride [303], alumina [304], tantalum oxide [304,305], and iridium oxide [306] have better properties than silicon oxide with regard to pH response, hysteresis and drift (see Fig. 4.26). In practice, these layers are deposited on top of the first layer of silicon oxide by means of chemical vapor deposition (CVD). CHEMFET The CHEMFET or chemically sensitive FET [307,308], is a modification of an ISFET with the original inorganic gate material covered by organic ionselective membranes like polyurethane, silicone rubber, polystyrene, polyamide and polyacrylates containing ionophores (Fig. 4.27). A critical point of the CHEMFET is the attachment of the sensitive membrane, which can be improved by mechanical [309] or chemical [308, 310, 311] anchoring to the surface of the gate oxide. CHEMFETs selective to K+ [308, 312–315], Na+ [316–318], Ag+ [319], transition metal cations (Pb2+ , Cd2+ ) [320–322] and some anions (NO− 3 ) [323–325] have been developed. Highly specific organic or biological compounds can be incorporated in the membrane as well: Enzymes (ENFET) [326, 327] like glucose oxidase for the detection of glucose [326, 328–330], and penicillinase for the detection


(b)

Ug

reference electrode

membrane hydrogel gate oxide n n source drain p-silicon Ud

resin

sensor response [mV]

(a) chemFET

liquid phase

71

320

a

b

a c

b

a

c

160

0 0

3

6

9

time [min] a: 10-5 M lead nitrate in acetate buffer b: 10-4 M lead nitrate in acetate buffer c: 10-2 M acetate buffer, pH: 5.5

Fig. 4.27. (a) Schematic representation of a CHEMFET, a modification of an ISFET with the original inorganic gate material covered by an organic ion-selective membrane and a hydrogel. (b) Sensor response of a CHEMFET with a lead-selective membrane upon different target analyte concentrations in an acetate buffer solution: 10−5 molar lead nitrate solution (a), 10−4 molar lead nitrate solution (b), and pure buffer solution (c). Redrawn, adapted from [322]

of penicillin [326]. Other enzymes can be used to detect pesticides and organophosphorous compounds [331–334] or aldehydes [335]. Antibodies can be immobilized for immunoreactions (IMFET), and biological entities like whole cells [336–338] or insect antennae [339] have been used (BIOFET). Differential CHEMFET Configuration CHEMFETS can be applied in a differential measurement configuration [340– 347] (Fig. 4.28a). This device configuration offers the advantage that a variety of experimental parameters and external disturbances, which affect both FET structures (e.g., light and temperature), cancel out. The inorganic gate of one of the CHEMFETS has to be rendered insensitive to pH or other target analytes by chemical surface treatment. This “insensitive” FET (sometimes denoted reference FET, REFET) should ideally show no response to the target species present in the sample phase (Fig. 4.28b) [348]. pH-sensitivity, e.g., can be suppressed by plasma-deposition of a hydrophobic polymer on the gate [349–353], or by realizing a reservoir of buffered solution with constant pH separated by a membrane from the sample phase Fig. (4.28) [348]. The applications comprise similar target analytes as for the case of CHEMFETS. 4.4.2.2.2 MOS Diode and Ion-Controlled Diode, ICD (Chemodiodes) Transduction Principle and Sensing Characteristics The chemical sensing mechanism is identical with that of the MOSFET, only the transduction method is slightly different. When the gas molecules (hydrogen) diffuse to the metal/oxide interface to form a polarization layer

72


(b)

(a) differential FET configuration gas phase nonbuffered hydrogel layer

gas-permeable buffered membrane hydrogel layer

gate oxide

Ug

n

source drain p-silicon

gate oxide

n

a

n Ud

n

drain source

b

Ug

EMF, voltage [mV]

Ag/AgCl reference electrodes

2

500

10-4 M 1

10-3 M

3

400

10-2 M 3•10-2 M

1: CO2-FET 2: insensitive FET 3: differential signal

300 0

10

20 time [min]

30

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

at the interface, the height of the energy barrier of the diode is altered. This leads to a change in either the forward voltage or the reverse current of the diode. The diode characteristics are hence shifted along the voltage axis (see Fig. 4.24), as it is the case with the MOSFET [22, 268, 269, 354–357]. The response time and sensitivity of the sensor can be improved by operation at elevated temperature. Therefore, sensing structures have been placed on thermally isolated membranes [358, 359]. The ion-controlled diode (ICD) was first described by Zemel [360]. Its operation is not too different from that of an ISFET and is analogous to that of the MOS-diode [8,34,293,360,361]. Polarization resulting from ion adsorption on the insulator (silicon oxide) causes changes in the effective forward voltage or reverse current, which can be measured. A more sophisticated structure with a through-the-chip p-n-junction, which enables the contacts on the back side of the chip to be isolated from the aqueous electrochemistry occurring at the front (“gate”) side, is described in [34,360,361]. The applications are the same as for the FETs. 4.4.2.2.3 MOS Capacitor and Ion-Selective Capacitor (Chemocapacitors) Transduction Principle and Sensing Characteristics The sensor element is a standard MOS or electrolyte insulator semiconductor (EIS) structure as shown in Fig. 4.24. Again, the target analyte species change the polarization at the metal/oxide or electrolyte/oxide interface thus affecting the flat-band voltage of the capacitor. The capacitance-voltage curve of the capacitor is shifted by a certain amount, which is proportional to the target analyte concentration in the gas or liquid phase. One can either record


73

capacitance [nF]

1000

porous silicon

800 600 400 200 0 -0.8

voltage [mV]

the capacitance as a measure of the analyte concentration or use some circuitry to keep the capacitance constant by varying the necessary bias voltage. Capacitor-type structures are straightforward to realize. The applications (MOSCAP [4,268,269,273,362,363], ISCAP/EIS [364– 367]) are similar to those of the FETs. A set of capacitance/voltage curves is shown in Fig. 4.29 [366].

-100

-300 4

5

6

7

8

pH

pH 8 pH 7 pH 6 pH 5 pH 4

-0.4

40 mV / pH

-200

0 voltage [V]

0.4

0.8

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

4.4.2.2.4 Measuring Work Functions: Kelvin Probe and Suspended-Gate Field-Effect Transistor, SGFET Transduction Principle and Sensing Characteristics The work function is defined as the minimum work required to extract an electron in vacuum from the Fermi level of a conducting phase through a surface and place it outside the reach of electrostatic forces at the so-called vacuum level [368, 369]. When two different electronic conductors are in contact, electrons flow from the material with the lower electron affinity to that with the higher one according to the difference in the (electro-)chemical potential (2.5, 2.6, 2.13) of the electrons until an equilibrium is reached. A contact or Galvani potential arises, which represents the bulk-to-bulk inner contact potential of the two materials or the difference of the Fermi levels of the two materials. If surfaces of two different materials are parallel and separated by a very thin insulator or air gap, a potential across the gap is formed, the Volta potential or outer potential, which represents the difference of the work functions of both materials. A palladium plate separated by a thin air gap (few

74


µm) from a copper plate will become positively charged, and correspondingly, the copper plate will be negatively charged. The work function of a certain material, which includes the chemical potential of the electrons (changes in electron affinity, eventual band bending due to electron transfer) and the surface dipole field (which exists even at absolutely clean surfaces) is changed upon formation of surface adsorbates, e.g., from the gas phase. Therefore, work function measurements can be used to advantage in chemical or gas sensing. The Kelvin probe relies on the displacement of one of the surfaces in a periodic oscillation. This oscillation induces charges across the surfaces and generates a sinusoidal current in the sensing plate, which is proportional to the work function difference between the sensing plate and the reference plate. This current thus directly depends on the surface chemistry of the plates. Micromachined Kelvin probes with a metal sensing film supported by a dielectric membrane and a 2.5-µm-thick silicon reference plate that is electrostatically deflected by a drive electrode (schematically depicted in Fig. 4.30a) have been fabricated and have been used to detect the surface adsorption of oxygen [370].

(a) integrated Kelvin probe heater Si-wafer

3 µm glass substrate drive electrode

temperature sensor

(b) suspended-gate FET metal gate (Pd) dielectric membrane metal sensing film electroplated contacts

movable Si-reference plane

Ug

gas phase

chemically sensitive layer

gate oxide

n

n

source

drain

p-silicon

Ud

Fig. 4.30. Schematic representation of a micromachined Kelvin probe (a) and a suspended-gate FET (b). For details, see text. Redrawn from [370] (a) and [368] (b)

The SGFET (or even the standard MOSFET) is very similar to the Kelvin probe [368, 371]. The notion of work functions was not used in the context of FETs yet, though FET transducers are sometimes denoted “work function devices”. A metal plate (suspended gate) is separated by an air gap (or in the case of the MOSFET silicon oxide) from a silicon plate (Fig. 4.30b) [368]. A variable voltage source (Ug ) is located between the back of the silicon and the gate metal. The plate distance, however, cannot be varied. Therefore, the drain-source current is used to interrogate the Volta potential. The magnitude of the drain-source current at a certain applied voltage depends on the


75

work function difference between the metal gate and the silicon. It therefore directly depends on surface adsorption or absorption chemistry in sensitive layers applied between the suspended gate and the silicon. Applications Sensitive layers applied in the gap include metal oxides to detect ammonia, carbon monoxide or nitrogen oxides [372–376], potassium iodide to detect ozone [377], palladium/polyaniline to detect hydrogen and ammonia [378], and polypyrrole to detect alcohols and volatile organics [379]. 4.4.2.2.5 Light-Addressable Potentiometric Sensor, LAPS Transduction Principle and Sensing Characteristics This type of sensor is based on the field effect as well [171], and its working principle is very closely related to that of FET devices (Fig. 4.31) [171, 380–383]. The LAPS device is a thin silicon plate (thinned down to a few µm thickness) with an approximately 100-nm-thick oxynitride layer in contact with an electrolyte solution. A potential is applied between the silicon plate and, e.g., a silver/silver chloride controlling electrode immersed in the electrolyte solution. The controlling electrode simultaneously serves as reference electrode. The sign and magnitude of the applied potential are adjusted so as to deplete the semiconductor of majority carriers at the insulator interface. Upon illumination of the plate with LEDs, hole-electron pairs are created, which can reach the depletion area (Fig. 4.31). Due to the charge separation in the depletion area, a photocurrent flows through the device. In this way, a sinusoidally modulated light beam causes a sinusoidal photocurrent, the amplitude of which depends on the width of the depletion charge

LAPS control electrode

I

chemically sensitive layer

liquid phase solution

U

insulator depletion region e-h pairs

p-silicon light

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

76


region: The larger the depletion region, the larger the photo current. The photocurrent amplitude thus depends on the absorbed species in the sensitive layer or, e.g., on the pH of the solution in contact with the insulator. The current-voltage curve is of sigmoidal shape and is shifted along the voltage axis due to chemical changes (e.g., pH-changes) at the solution/silicon oxide interface in analogy to other field-effect devices. Applications The silicon plate is straightforward to manufacture, and arrays of LEDs and different selective layers can be used on the same chip [371]. Assessment of the lateral resolution [384,385] and benchmarking against ISFETs [386] have been performed. LAP methods have been applied to the gas phase as well [380]. Applications include monitoring cell activity via pH changes resulting from cell activity and metabolism [387–390], and biosensing in liquids [381, 383, 391]. LAPS devices have been developed by Molecular Devices Inc. [392]. 4.4.2.2.6 Field-Effect Device Fabrication • Patterning of metal electrodes (lift-off, thermal evaporation, sputtering) as described for MOSFETs in [267–276] • Deposition of additional metal oxides/nitrides by LPCVD (tantalum oxide, alumina, silicon nitride) as described for MOSFETs and ISFETs in [296–306] • Optional membrane formation for temperature stabilization by back etching [358, 359] • Surface micromachining (HF etching, Al-etching) for the SGFET [374,393] • Deposition of electro-active polymers, membrane materials, hydrogels by spin-casting, spraying, screen printing, photolithography for CHEMFETs [278–280, 307–335, 349–353] The fabrication of field-effect electrochemical sensors integrated with CMOS circuitry is described in various publications [11,296,297,300,302,303,318,362, 394]. A modular chip system based on a sensing and a “service” chip has been described in [279], a flow-through ISFET in [395]. Back-side-contact ISFETs are detailed in [396, 397]. pH-ISFETS are commercially available from, e.g., Honeywell [398]. 4.4.3 Conductometric Sensors Conductometric techniques are a special case of AC-impedance techniques. Instead of the real and imaginary component of the electrode impedance at different frequencies, only the real-valued resistive component, related to the sample (sensing material) resistance, is of interest. Since complex impedances include capacitive and inductive contributions, chemocapacitors that do not rely on the field effect are included here, in the conductometric section. The section on conductometric sensors is hence organized in two parts, one on


77

resistance measurements at room temperature and elevated temperatures (chemoresistors) and the other on chemocapacitors. 4.4.3.1 Chemoresistors Transduction Principle and Sensing Characteristics Chemoresistors rely on changes in the electric conductivity of a film or bulk material upon interaction with an analyte. Conductance, G, is defined as the current, I [A], divided by the applied potential, U [V]. The unit of conductance is Ω−1 or S (Siemens). The reciprocal of conductance is the resistance, R[Ω]. The resistance of a sample increases with its length, l, and decreases with its cross-sectional area, A: R=

U 1 l 1 = = · . G I κ A

(4.8)

Conductivity or specific conductance, κ[1/Ωm], is hence defined as the current density [A/m2 ] divided by the electrical field strength [V/m]. The reciprocal of conductivity is resistivity, ρ[Ωm]. The conductivity can be thought of as the conductance of a cube of the probed material with unit dimensions [11]. Conductometric sensors are usually arranged in a metal-electrode-1/ sensitive-layer/metal-electrode-2 configuration [4]. The conductance measurement is done either via a Wheatstone bridge arrangement or by recording the current at an applied voltage in a DC mode or in a low-amplitude, lowfrequency AC mode to avoid electrode polarization. In Fig. 4.32a [399], a conductance cell (in this case metal oxides) and the respective equivalent electric circuits are depicted. The goal of conductometry is to determine the sample free molecules

adsorbed molecules

electrodes

e-

U

I

(b)

sensitive layer

esubstrate

b

d

c U

4-electrode conductance cell

catalyst

(a)

a

2-electrode conductance cell

I

a

a. contacts b. surface c. bulk d. grain boundaries substrate

electrodes

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

78


resistance (c). The lead wire resistances normally can be neglected. The electrode impedance (a) consists of two elements, the contact capacitance, and the contact resistance. By applying an AC potential, an AC current will flow through the resistor cell. If the contact capacitance is sufficiently large, no potential will build up across the corresponding contact resistance. The contact resistance should be much lower than the sample resistance and be minimized, so that the bulk contribution dominates the measured overall conductance. If surface conductivity mechanisms differing from those in the bulk occur, this can be modeled by adding an additional surface resistance (b) to the equivalent circuit. A grain boundary in the sensing material constitutes a resistance-capacitance unit (d). The conductivity depends on the concentration of charge carriers and their mobility, either of which can be modulated by analyte exposure. In contrast to potentiometry and voltammetry, conductometric measurements monitor processes in the bulk or at the surface of the sample. Any contribution of electrode processes has to be avoided. Therefore, in most cases, a four-electrode configuration is preferred over a simple two-electrode configuration (Fig. 4.32b). The outer pair of electrodes is used for injecting an AC current into the sample, the potential difference is then measured at the inner pair of electrodes. The interference of electrode impedances on the measurement results is thus excluded. 4.4.3.1.1 Low-Temperature Chemoresistors Several classes of predominantly organic materials are used for application with chemoresistors at room temperature. The chemically sensitive layer is applied over interdigitated electrodes on an insulating substrate. Electrode spacing is typically 5 to 100 µm, and the total electrode area is a few mm2 . The applied voltage ranges between 1 and 5 V. Metal-phthalocyanines constitute organic p-type semiconductors. The adsorption of oxidizing agents such as, e.g., nitrogen oxide or ozone hence decreases the resistance by increasing the number of holes in the conduction band [8]. Metal phthalocyanines at elevated operation temperatures (approx. 180◦ C) have been used to monitor nitrogen oxide [400, 401], ozone [402], hydrogen chloride [401] and even ammonia [403, 404]. Conducting Polymers such as polypyrroles, polyaniline and polythiophene exhibit a large conjugated π-electron system, which extends over the whole polymer backbone. Partial oxidation of the polymer chain then leads to electrical conductivity, because the resulting positive charge carriers (denoted polarons or bipolarons) are mobile along the chains [9]. Counteranions must be incorporated into the polymer upon oxidation to balance the charge on the polymer backbone. The conducting polymers, however, do not only react with oxidizing agents, but also respond to a wide range of organic vapors [405–408]. The underlying principle of this response is still unclear; suggestions include: (I)


79

vapor molecules could affect the charge transfer between polymer and the electrode contact, (II) analyte molecules could interact with the mobile charge carriers on the polymer chains or (III) with the counterions and thus modulate the mobility of the charge carriers, or they could (IV) alter the rate of interchain hopping in the conducting polymer [379, 409, 410]. Applications include the detection of a variety of polar organic volatiles like ethanol, methanol, components of aromas, [405–408, 411–420] and others. Conducting polymers show a high cross-sensitivity to water. Sensors are commercially available from, e.g., Osmetech and Marconi [416]. In Carbon-Black-Loaded Polymers, conducting carbon black is dispersed in non-conducting polymers deposited onto an electrode structure. The conductivity is by particle-to-particle charge percolation so that if the polymer absorbs vapor molecules and swells, the particles are, on average, further apart (Fig. 4.33), and the conductivity of the film is reduced [417].

(a) no analyte present carbon particles

resistance R1

(b) after analyte absorption polymer

resistance R2 >R1

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

Applications include monitoring organic solvents such as hydrocarbons, chlorinated compounds, and alcohols [418–421]. Sensors are commercially available from, e.g., Cyrano Sciences [422]. Hydrogels responsive to pH-changes have been applied to interdigitated electrode arrays. The hydrogel swells or shrinks to a hydration determined by the pH of the analyte solution. This leads to a corresponding increase or decrease in the mobility of the ions partitioned by the gel. The sensitivity of ion mobility to small changes in hydration causes large resistance changes [423, 424]. A conductometric variant of a Severinghaus electrode (detection of carbon dioxide via dissolution in water, formation of carbonic acid and monitoring of the pH change [4]) with a liquid reservoir has been described in [425].

80


Fabrication • Patterning of metal electrodes (lift-off, thermal evaporation, sputtering) [405–408, 418–421] • Optional membrane formation by back etching for temperature stabilization [414, 426] • Deposition of carbon-loaded polymers, membrane materials, hydrogels by spin-casting, spraying, screen printing, and photolithography [418–421,423, 424] • Deposition of conducting polymers by electrochemical deposition; sensor selectivity is modified by changing the counterion used in the polymerization process [405–408] The fabrication of an impedance device [427], and microbridges [415] integrated with CMOS circuitry components have been described. Complete processing sequences are detailed in [423, 425]. 4.4.3.1.2 High-Temperature Chemoresistors (Hotplate Sensors) There is a wealth of literature on semiconducting metal oxides and related sensors, the focus of this work, however, will be on silicon-based micromachined hotplates. A typical high-temperature chemoresistor includes an integrated heater, a thermometer and a sensing film on a thermally isolated stage such as a membrane (Fig. 4.34) [428]. Isolated micromachined structures (hotplates) exhibit very short thermal time constants on the order of milliseconds. The sensitive materials used with the hotplate sensors include widebandgap semiconducting oxides such as tin dioxide, gallium oxide, indium oxide, or zinc oxide. In general, gaseous species acting as electron donors (hydrogen) or acceptors (nitrogen oxide) adsorb on the metal oxides and form (a)

(b)

Fig. 4.34. Micrograph of a microhotplate (a), and schematic top view and side view of the device (b). The suspended plate exhibits a polysilicon heater, an aluminum plane for homogenous heat distribution and aluminum electrodes for measuring the resistance of a semiconductor metal oxide. Reprinted from [428,443] with permission


81

surface states, which can exchange electrons with the semiconductor. An acceptor molecule will extract electrons from the semiconductor and, therefore, decrease its conductivity. The opposite holds true for an electron–donating surface state. A space charge layer will thus be formed. By changing the surface concentration of donors/acceptors, the conductivity of the space charge region is modulated [4, 399, 429–431]. In addition to the above-mentioned interaction of surface adsorbates and related electronic effects, the diffusion of lattice defects from the bulk of the metal-oxide crystal also occurs (ionic conduction) at elevated temperatures (>600◦ C). The defects can act as donors or acceptors. Oxygen vacancies, e.g., act as intrinsic donors. The overall conductivity in polycrystalline samples includes contributions from the individual crystallites, the grain boundaries, insulating components such as pores, and the contacts (Fig. 4.32a). Thus, the conduction mechanism in ceramic polycrystalline samples is difficult to analyze, and a variety of empirical data has been published [4, 399, 429–431]. The most extensively investigated material, tin dioxide, is oxygen-deficient and, therefore, is an n-type semiconductor since oxygen vacancies act as electron donors. In clean air, oxygen, which traps free electrons by its electron affinity, and water are absorbed on the tin dioxide particle surface forming a potential barrier in the grain boundaries. This potential barrier restricts the flow of electrons and thus increases the resistance. When tin dioxide is exposed to reducing gases such as carbon monoxide, the surface adsorbs the gases, and some of the oxygen is removed by reaction of water and oxygen at the surface. This lowers the potential barrier, thereby reducing the electric resistance. The reaction between gases and surface oxygen depends on the sensor temperature, the gas involved, and the sensor material [4,399,429–432]. Semiconductor metal-oxide sensors usually are not very selective, but respond to almost any analyte (carbon monoxide, nitrogen oxide, hydrogen, hydrocarbons). One method to modify the selectivity pattern includes surface doping of the metal oxide with catalytic metals such as platinum, palladium, gold, and iridium [399, 430, 433]. Surface doping improves the sensitivity to reducing gases, reduces the response time and operation temperature and changes the selectivity pattern [399, 430, 433]. Most modern sensors operate in a regime, in which the overall conductivity is determined by nanocrystalline sensing materials. As a consequence of the small grain size (better surface to volume ratio), the relative interactive surface area is larger, and the density of charge carriers per volume is higher. This leads to more drastic conductivity changes and, hence, larger sensor responses in comparison to larger grains [399, 430]. Since microhotplates have a very low thermal mass, they allow for applying temperature-programmed operation modes [434–437]. By operating the device, e.g., in a cyclic thermal mode, reaction kinetics on the sensing surface

82

4 Microfabricated Chemical Sensors 1.0 ppm NO2 pure air

50 ppm CO

Rsensor[Ω]

100k

10k

close-up 1k temperature [°C]

mixture 50 ppm CO 1.0 ppm NO2

close-up

1 min

400 300 200

Fig. 4.35. Sinusoidal modulation of the operation temperature of a tin-dioxide sensor between 200◦ and 400◦ C (bottom) leads to characteristic frequency-dependent resistance features (upper part). Changes of the resistance (Rsensor ) of the micromachined sensor upon exposure to 50 ppm CO, 1 ppm NO2 and a mixture of 50 ppm CO and 1 ppm NO2 in synthetic air (50% relative humidity). Adapted from [436] with permission

are altered, producing a time-varying response signature that is characteristic for the respective analyte gas [434–437]. An example is given in Fig. 4.35 [436]. Fabrication • Additional deposition/patterning of metal electrodes using lift-off, thermal evaporation, and sputtering [438–441] • Back side (KOH) [426, 438–442] or front side (RIE, EDP, XeF2 ) [428, 443] etching for membrane formation • Deposition of metal-oxide materials by LPCVD, sol-gel processes, sputtering, and screen printing [444–449] • Sintering of the metal oxides (annealing) at elevated temperatures [450] The fabrication of hotplates on a CMOS-substrate is described in [428, 443]. For details on CMOS-based microhotplate gas sensor systems, see also Sect. 5.4. Complete processing sequences are detailed in [451, 452]. The fabrication of a 39-electrode array with a tin-dioxide-gradient coating has been reported in [453, 454]. Extensive reliability studies are reported in [455–457]. Investigations and simulations to optimize the power consumption are under way (see, e.g., [458–460]). Devices are commercially available from Figaro, Marconi, Capteur and MICS [461].


83

Applications Typical applications include the detection of hydrogen [428], oxygen [428], nitrogen oxide [438, 462], CO [439, 440], and a variety of organic volatiles [441–444, 463] using tin dioxide as sensitive layer. Thermal cycling of the hotplate structures allows for detection of gas mixtures (CO and nitrogen oxide) with a single sensor [436] (Fig. 4.35). In other applications, temperature profiles were optimized to specifically detect selected organic volatiles [443, 463, 464]. Additional sensitive materials on hotplates include, e.g., niobium and titanium oxide to detect oxygen [465], as well as metal films covered with surface oxides. Reducing gases like hydrogen donate electrons to the metal and increase the conductance, whereas electron acceptor molecules like oxygen decrease the metal conductance [451, 466, 467]. An unheated tin-dioxide oxygen sensor (slow response) has been reported in [468]. 4.4.3.2 Chemocapacitors Transduction Principle and Sensing Characteristics Chemocapacitors (dielectrometers) rely on changes in the dielectric properties of a sensing material upon analyte exposure (chemical modulation of equivalent-circuit capacitors in Fig. 4.36 by changes in the dielectric constant of the sensitive layer). Interdigitated structures that are similar to those of room-temperature chemoresistors are predominantly used [469–472].

analyte

substrate

polymer

E1

E2

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

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

84


The capacitances usually are measured at an AC frequency of a few kHz up to 500 kHz. The analyte absorption in the sensitive layer on top of the electrodes induces a change in the layer dielectric properties and, consequently, a capacitance change that can be measured. For conducting measurements at defined temperatures, sensor and reference capacitors can be placed on thermally isolated membrane structures [475–477]. For more details on capacitive sensors and CMOS capacitive microsystems, see [478–480] and Sect. 5.1. Fabrication • Deposition/patterning of metal electrodes (lift-off, thermal evaporation, sputtering) [469–474] • Optional back etching, membrane formation for temperature stabilization [475–477, 481] • Deposition of polymers (spin-casting, spraying, photolithography) [469– 480] The fabrication of capacitors integrated with CMOS circuitry components is described in [470, 472, 475, 476, 478, 479, 482, 483]. Temperature-stabilized membranes have been reported in [475–477]. Capacitive humidity sensors are commercially available from, e.g., Sensirion, Vaisala, and Humirel [484]. Applications Typical applications areas are humidity sensing using polyimide films [469– 473, 482–484], since water has a relatively high dielectric constant of 78.5 (liquid state) at 298 K leading to large capacitance changes. A variant exhibits polyimide columns sandwiched between metal electrodes [485]. More recent applications include the detection of different kinds of organic volatiles in the gas phase (hydrocarbons, chlorinated hydrocarbons, alcohols etc.) using polymeric layers [475, 476, 478, 479, 486, 487] or liquid crystals [488], and the detection of nitrogen oxide [489], sulphur dioxide [490] and carbon dioxide [491, 492] using ceramic materials. The interaction mechanisms involve reversible physisorption and bulk/gas phase partitioning (see 2.10).

5 CMOS Platform Technology for Chemical Sensors

The aim in utilizing microfabrication techniques and, in particular, CMOS technology for realizing chemical sensors was to devise more intelligent, more autonomous, more integrated, and more reliable gas sensor systems at low costs in a generic approach. Since the sensor market is strongly fragmented, i.e., there exists a large variety of applications with different needs and sensor requirements, a modular approach or “toolbox strategy” relying on a platform technology was identified as the most promising attempt to achieve major progress. Once the platform technology has been chosen, the components of the toolbox such as transducers, sensor modules, and circuit modules can be developed, some of which afterwards can be assembled into a customized system that meets the respective applications needs. The application hence dictates the system architecture and the nature of its components: As soon as the target analytes and their concentrations as well as the boundary conditions of detection are specified, the optimum transducer, the necessary driving and signal conditioning circuitry as well as interface and communication units can be selected from the toolbox, and the different modules can then be combined to arrive at a custom-designed sensor system. A multitude of development activities is necessary to obtain all the modules needed for the toolbox: (a) Design and miniaturization of transducers and directly related electronic components (potentiostats, heaters, amplifiers, etc), (b) development of digital-to-analog and analog-to-digital conversion units, interface and communication units, (c) development of additional and auxiliary functions, which are pivotal for the system performance (e.g., temperature control, temperature sensors, humidity sensors), and (d) development of dedicated microsystem packaging solutions, which are suitable for chemical or gas analysis. It is important to note that the package has to be thought of already in the initial conception phase of a microsystem, since the design and architecture of a microsystem heavily depend on the envisaged packaging concept. A fundamental issue concerns the sensor system implementation, i.e., the decision to either realize monolithic systems combining CMOS circuitry and sensor structure on the same chip (CMOS-MEMS), or to develop hybrid designs that rely on optimized sensor materials and fabrication processes

86


with external electronics. There is a number of aspects that have to be taken into account in making such a decision. Materials and Fabrication Processes For monolithic designs, the selection of materials is restricted to CMOSmaterials (silicon, polysilicon, silicon oxide, silicon nitride, aluminum, see also Sect. 3.3) and CMOS-compatible materials. There is also a limitation on available fabrication processes owing to a limited number of pre-CMOS and post-CMOS micromachining options. High-temperature steps (e.g., > 400◦ C) are detrimental to the aluminum metallization (metal oxidation, diffusion) and alter the transistor characteristics (last high-temperature step of the CMOS process is at approx. 380◦ –400◦ C). For hybrid designs, any material or the optimum sensor material can be used, and a wealth of micromachining techniques for all kinds of materials is available through prototyping services. Time to Market, Costs The production of monolithic designs relies on established industrial CMOS technology to fabricate the circuitry and the basic sensor structures. Additional fabrication equipment and related technology developments are needed for only a few sensor-specific post-processing steps. The fabrication of hybrid designs entails establishing and qualifying a reliable nonstandard production sequence. The specific production equipment and the necessary clean rooms require large investments. Sensor Response Time The response time of, e.g., a gas sensor is, in most cases, determined by the volume of the measurement chamber and the flow rate (other relevant processes include also, e.g., diffusion or dissociation). Using the monolithic approach and a suitable packaging technique (e.g., flip-chip packaging), the volume of the measurement chamber can be kept very small as a consequence of a small-size, flat and planar sensor or sensor array. In hybrid designs, the volume is depending on sensor geometries and array arrangements. Performance Capacitive or resonant microsensors perform pronouncedly better in monolithic designs owing to the fact that the influence of parasitic capacitances and crosstalk effects can be reduced by on-chip electronics (filters, amplifiers etc.). On-chip analog-to-digital conversion is another feature that helps to generate a stable sensor output that can be easily transferred to off-chip recording devices. In hybrid designs, it is sometimes very difficult to read out the rather minute analog microsensor signals.


87

Auxiliary Sensors/Smart Features Temperature or flow sensors can be monolithically co-integrated with the chemical sensors on the same chip. Calibration, control and signal processing functions as well as self-test features can be realized on chip. For hybrid designs, additional devices and off-chip components are required. Connectivity The number of electrical connections prominently contributes to the overall system costs. The monolithic implementation of a single-chip gas sensor (see Sect. 5.5 on monolithic CMOS gas sensor system later) with three micromachined sensors requires only seven connections (three for supply voltages, one for a clock signal, one for reset, and two for the serial interface). Up to 16 chips can be connected without adding any additional communication lines by implementation of a digital bus interface. A hybrid approach would require at least 30 pads for the three sensors and a total of 480 connections, if 16 sensors of each type would be combined, since there is no interface available on the sensor side. Package To package monolithic designs, IC-packaging techniques can be modified and adapted such as flip-chip-technology or simple epoxy-based packaging methods. Hybrid implementations require complex packages to reduce sensor interference (see, e.g., high-frequency acoustic-wave SAW-sensors, Sect. 4.1.1), to minimize electric crosstalk, and to optimize the critical connections. This further complicates the already difficult task of chemical sensor packaging. In summary the main disadvantages of monolithic CMOS-MEMS solution include the restriction to CMOS-compatible materials and the limited choice of micromachining processes. However, CMOS-MEMS offers on the other hand unprecedented advantages over hybrid designs especially with regard to signal quality, device performance, increased functionality and available standard packaging solutions. These advantages, in my opinion, clearly outweigh the drawbacks and limitations. Micromachined chemical sensors are not yet established on the market. In the case of well-established physical sensors such as acceleration and pressure sensors, a trend towards monolithic solutions can be identified for larger production volumes and more severe cost restrictions [493–495]. Several examples of monolithically integrated chemical sensor systems will be presented in the following sections of this chapter. The evolution from single transducers, which are integrated with the necessary driving and signalconditioning circuitry, to monolithic multisensor arrays and fully developed systems with on-chip sensor control units and standard interfaces will be shown. Microelectronics and micromechanics (MEMS-structures) have been

88


CMOS and MEMS design using CAD tool industrial CMOS process at foundry CMOS postprocessing packaging and assembly testing Fig. 5.1. Chemical microsensor fabrication sequence

realized on the same chip in all cases. The general scheme of the chemical sensor microsystem fabrication sequence is shown in Fig. 5.1. The CMOS-MEMS design is done in house, and the designs are then transferred to a foundry that performs a complete CMOS run at industrial standards (usually double-poly, double-metal 0.8-µm CMOS process as provided by, e.g., austriamicrosystems, Unterpremstätten, Austria [496]). The processed wafers come back from the foundry and undergo the post-CMOS micromachining (formation of membranes, cantilevers etc.). The wafers are then diced, the chips are tested and subsequently packaged as prototypes. The chemically sensitive coating is either applied on wafer level or to the single chips by means of spray-coating (polymers) or drop deposition (metal oxides).

5.1 CMOS Capacitive Microsystems 5.1.1 CMOS Capacitive Transducer As already mentioned in Sect. 4.4.3.2, capacitive chemical microsensors rely on changes in the dielectric properties of, e.g., a polymeric layer as a consequence of analyte absorption from the gaseous phase into the bulk polymer. The dielectric properties of the polymer with absorbed analyte differ from those of the polymeric matrix alone. A convenient way to assess the dielectric constant of a layer or film is to measure the capacitance of a polymer-filled plate-capacitor [497] or that of polymer-coated interdigitated electrodes [475–484, 498]. The latter approach has become very popular since the devices can be easily fabricated in planar technology such as CMOS. Planar polymer-coated interdigitated capacitors also provide direct access of the analyte to the sensitive layer (short sensor response time), which is more difficult to realize with modified plate capacitor designs [485]. Interdigitated capacitor designs

5.1 CMOS Capacitive Microsystems

(a) schematic

89

(b) SEM picture

polymer electrode 2

electrode 1

SiO2

Fig. 5.2. Schematic (a) and micrograph (b) of an interdigitated capacitor. The micrograph also shows the thin polymer layer, which is tightly attached to the sensor surface

have hence been chosen for the CMOS-based capacitive microsensor systems, a schematic and a micrograph of which is depicted in Fig. 5.2. The capacitors are fabricated exclusively with layers and processing steps available in the standard CMOS process sequence. Electrode 1 is made from the first aluminum metal layer while electrode 2 comprises a stack of interdigitated electrodes of the first and second metal layer, which are electrically connected on the chip via leads, albeit they are physically separated by a silicon-oxide layer (Figs. 5.2, 5.3). The quasi “three-dimensional” electrode configuration was chosen to enhance the sensitivity of the polymer-coated sensor to analyte-induced capacitance changes by maximizing the polymer volume in the regions of strong electric field (see also Fig. 4.36) [499]. The overall capacitor includes 128 electrode pairs and occupies an area of 824 µm by 814 µm. The electrode width and the interelectrode spacing are 1.6 µm. The pad-etch is used to remove the silicon nitride passivation on top of the sensing capacitor to allow for polymer coating. The electrode shapes are still visible under the polymeric coating, i.e., the polymer is tightly attached to the surface and reproduces the surface topology. A cross-sectional drawing of one electrode pair is shown in Fig. 5.3a, the corresponding equivalent circuit model, which was also used for simulations, is depicted in Fig. 5.3b [500,501]. Both electrodes (E1 and E2) exhibit considerable parasitic capacitances to the n-well (Cnwell 1 , Cnwell 2 ), which are very similar owing to the identical, though mirrored layout of the metal-1 components. The interelectrode capacitance includes two major contributions: An intermetal oxide capacitance (Cox1 ) and a composite of an oxide capacitance (Cox2 ) in series with an element that includes the polymer capacitance (Cpolymer ) shunted with the polymer resistance (Rpolymer ). The polymers used with interdigitated capacitors are mostly nonconducting (Rpolymer is very

90


E1

(a)

(b) Cox2

polymer E2 oxide n-well

E1

Rpolymer

Cpolymer Cox1 Cnwell1

E2

Cnwell2

Fig. 5.3. (a) Cross section of an electrode pair, (b) equivalent circuit model [500, 501]. For details see text

large). The two capacitors Cox2 and Cpolymer can then be substituted by a single capacitor Cpoly with Cox2 being considerably larger than Cpolymer . From finite-element simulations of a sensing capacitor coated with 10 µm of a polymer exhibiting a dielectric constant of ε = 4.8 (polyetherurethane, PEUT), the capacitances in the equivalent circuit model in Fig. 5.3 have been determined to: Cnwell1 : 17.7 pF, Cnwell2 : 18.2 pF, Cox1 : 6.4 pF, Cox2 : −8.4 pF, Cpolymer : −1.7 pF and Cpoly : 1.4 pF [479, 499–501]. The capacitance changes upon analyte absorption into the polymer are in the atto-Farad range, e.g., 4.4 aF/ppm toluene in polyetherurethane. Since the parasitic capacitances largely prevail over the capacitance of interest, Cpolymer or Cpoly , a differential readout scheme is mandatory. By using a switched-capacitor scheme with a reference and sensing capacitor of equal geometric dimensions, the reference capacitor being passivated by a sufficiently thick silicon nitride layer, the contributions of parasitic n-well capacitances and oxide capacitances cancel out to a large extent in the differential readout, since they are almost identical. In view of the analyte-induced capacitance changes in the atto-Farad range, on-chip signal conditioning circuitry is imperative, since the transfer of minute analog signals via bonds and wires is very difficult [500–502]. From the physical data and facts described above, it is evident that miniaturization without electronics integration is presumably prone to failure. 5.1.2 On-Chip Circuitry of the Capacitive Microsystem Micromachined capacitive sensors have been developed for many different applications, e.g., accelerometers, pressure sensors, fingerprint-sensors, and gas sensors [499–508]. Various read-out circuitry topologies have been developed to measure the small capacitance changes of such micromachined sensors. The majority of the designs rely on a differential readout between a sensing capacitor and a reference capacitor, the latter being not affected by the measurand. This offers the advantage that parasitic effects, such as temperature drift and ageing, affect the reference capacitor to the same extent as the


91

sensor and, hence, cancel out. As a consequence, these designs do not provide accurate information on the absolute value of the measuring capacitor. The most popular approach to read-out capacitive sensors with high resolution and good suppression of parasitics is a switched-capacitor design [509–513]. A resolution of 19 bits is needed in order to achieve a detection limit of 1 ppm for volatile organic compounds (VOCs). The bandwidth can be as low as 1 Hz. As none of the designs reported on in literature exhibits the required performance, an improved architecture based on a switched-capacitor Sigma-Delta modulator was developed. The sensor response is measured as the differential signal between a polymer-coated sensing and a nitride-passivated reference capacitor. Both, sensor and reference capacitors are split into two parts to improve the charge transfer efficiency. Sensing capacitor, (CS ), and reference capacitor, (CR ), are incorporated in the first stage of a fully differential second-order SigmaDelta-modulator (Fig. 5.4) with two switched-capacitor integrators and a subsequent comparator [501, 514, 515].

Cfb1

Cfb2

Cfb1'

Cfb2'

CR

decimation filter

digital logic

CS

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

Since the output bit stream of the Sigma-Delta modulator is proportional to the ratio (CS − CR )/(Cfb ), the four feedback capacitors (Cfb in Fig. 5.4) are realized as interdigital capacitors with the same materials as sensing and reference capacitor in order to eliminate differences in temperature behavior and ageing. Due to the small signal bandwidth, the output bit stream of the Sigma-Delta modulator is decimated using a frequency counter. For more details on the circuitry see [501, 514, 515]. A micrograph of the integrated capacitive microsystem is shown in Fig. 5.5.

92


readout circuitry

feedback capacitors

sensing capacitor

reference capacitor

Fig. 5.5. Micrograph of a capacitive sensor system including a polymer-coated sensing capacitor, a passivated reference capacitor, four interdigitated feedback capacitors, and the Sigma-Delta circuitry as detailed in the text [499, 501]

The design offers the option to place all interdigitated capacitors (sensor, reference, feedback) on micromachined membranes with integrated heaters for temperature stabilization since analyte absorption in the polymer matrix is strongly temperature-dependent (see also Chap. 2) [516]. 5.1.3 Capacitive Gas Sensing Two effects change the capacitance of a polymeric sensitive layer upon absorption of an analyte (Fig. 5.6): (i) Swelling and (ii) change of the dielectric constant due to incorporation of the analyte molecules into the polymer matrix [478, 479]. The resulting capacitance change is detected by the read-out electronics.

absorption

polarization

swelling

polymer substrate

analyte polarized analyte

electrodes

air

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


93

The partition coefficient, K, as introduced in Chap. 2 (2.10) characterizes the absorption behavior of a polymer with regard to a specific analyte. It is a chemical equilibrium constant and is defined as the ratio between the analyte concentration in the sorptive or polymer phase and that in the gas phase. For a specific analyte, K is inversely proportional to the saturation vapor pressure of this analyte and, therefore, strongly temperature-dependent. Furthermore, K depends on the analyte/polymer interaction, responsible for the partial selectivity. The analyte concentrations in both phases can be replaced by the partial pressure, pA , of the analyte and by the volume fraction of the analyte in the sensitive layer, ϕA , both used in the rest of this section. Equation (2.10) can be rewritten as ϕA =

pA M · KC · R · T ρ

(5.1)

where M and ρ denote the molar mass and the density of the analyte in liquid state, and R and T the universal gas constant and the absolute temperature [517]. As already mentioned, the two effects changing the capacitance are swelling and change of the dielectric constant. For low analyte concentrations, the swelling is linear in the amount of absorbed analyte, expressed by (5.2). (5.2) heff = h · (1 + Q · ϕA ) . Here, h and heff denote the initial polymer thickness and the resulting effective thickness after analyte absorption, respectively, Q is a dimensionless non-ideality factor of the swelling, and ϕA is the volume fraction of the analyte in the polymer. ϕA is proportional to the concentration of the analyte in the gas phase assuming the validity of Henry’s law. The proportionality factor has to be determined experimentally for every polymer/analyte combination or can be estimated with solubility parameters or linear solvation energy relationships [518]. Q = 1 represents ideal swelling, i.e., the total volume is given by the addition of the volume of the absorbed analyte in its liquid state to that of the polymer. For elastic polymers, the tendency to generate static stress while swelling increases with increasing stiffness of the polymer. The composite dielectric constant of mixtures of nonpolar liquids can be approximated for all kinds of analytes as proposed in [479, 517]. It can be generalized including Q to εeff = εpoly + ϕA · ((εA − 1) − Q · (εpoly − 1))

(5.3)

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

94


thin

thick

2 µm

6.4 µm

Fig. 5.7. Selectivity through layer thickness: Sensors coated with a thin and thick sensitive layer. The extension of the electric field lines is approximately 2 µm for this electrode geometry (1.6 µm electrode width, 1.6 µm electrode spacing) [517]

5.1.3.1 Selectivity Through Sensitive Layer Thickness The presence of two relevant physical effects gives rise to an additional mechanism for selectivity, because thin and thick sensitive layers exhibit different sensitivity patterns. For a simple interdigitated structure, the space above the device containing 95% of the field lines is within a distance of half of the electrode periodicity [480]. Therefore, a thin or thick layer is defined with respect to the electrode periodicity (see Fig. 5.7). For a layer thickness significantly less than half the periodicity, the region of strong electric field extends above the sensitive layer. Upon analyte sorption, the amount of polarizable material in the sensed region of the capacitor always increases, which results in a capacitance increase regardless of the dielectric constant of the analyte. For a layer thickness greater than half the periodicity of the electrodes, almost all electric field lines are within the polymer volume. Consequently, the capacitance is determined by the composite dielectric constant of the analyte/polymer mixture. The capacity change can, therefore, be positive or negative, depending on whether analyte or polymer has a higher dielectric constant [479] (5.3). The expected differences in responses from sensors with thin and thick polymer layers have been verified experimentally. Typical sensor response profiles from capacitors coated with a thin (0.3 µm) and thick (2.3 µm) poly(etherurethane) (PEUT) layer are plotted in Fig. 5.8. The capacitor has been alternately exposed to various concentrations of toluene and ethanol at 28◦ C and the pure carrier gas. Ethanol (ε = 24.3) has a higher dielectric constant than PEUT (ε = 4.8), and toluene a lower dielectric constant (ε = 2.36). Both analytes provide positive sensor signals with thin PEUT layers, whereas with thick layers, ethanol provides a positive signal and toluene a negative signal. The limit of detection of the capacitive microsystem at 28◦ C is approximately 8 ppm for toluene and 5 ppm for ethanol with 10 ppm analyte volume fraction corresponding to 1 Pa analyte partial pressure in the probed air [476]. With a sensor coated with a thick polymer layer, low-ε analytes can be differentiated from high-ε analytes if they are pure. In a mixture of these


95

15

sensor signal [kHz]

polymer: PEUT

0.3 µm 2.3 µm

10 toluene 600toluene – 3000 ppm

5

0 ethanol 1000 – 5000 ppm -5 0

4

8 time [h]

12

16

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

analytes, more information is needed. Zero response, e.g., from the thick layer sensor could also be generated by a certain mixture of both analytes in such a way that the positive and negative signal just cancel out. However, the combination of signals from sensors with thin and thick sensitive layers allows for determining the concentration of both analytes. The signal from the thinly coated sensor gives the total amount of analyte, whereas the signal from the thickly coated sensor indicates the mixing ratio at a given total concentration. Extensive numerical simulations (finite element method, FEM) were performed for the interdigitated capacitor with different coatings. The capacitance of the electrode geometry was simulated with polymer coatings of 17 different, logarithmically distributed thicknesses, h, in the range of 0.06–7 µm and 41 different, linearly distributed dielectric constants, εpoly , in the range of 1–5. For details on the simulations, see [479, 517]. The most important variable to calculate from the simulated data is the sensitivity. For low analyte concentrations and, hence, low volume fractions, small changes in the dielectric constant, δε, and in the polymer layer thickness, δh, are expected and the capacitance change, ∆C, can be approximated by [517]: ∆C = δh · ∂h C (h, εpoly ) + δε · δε C (h, εpoly ) .

(5.4)

The change in dielectric constant was approximated with (5.3). Together with (5.2), the capacitance change can be rewritten as [517]:


normalized sensitivity

96

εA » εP ethanol ethyl acetate 0 0

1

2

3

4

εA > εP

5

h [µm]

toluene

εA < εP

Fig. 5.9. Numerically simulated dependence of the vapor sensitivity of a capacitive sensor on the polymer layer thickness, h, for PEUT (εpoly = 4.8). The lines, from bottom to top, correspond to analytes with dielectric constants of 1.9, 2.4, 3.3, 4.7, 6.0, 10, 24, and 81. Ideal swelling was assumed (Q = 1). For better visibility, the curves are normalized so that the sensitivities for very thin layers (initial slopes of the curves) are identical [517]

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

(5.5)

The physical sensitivity, ∆C/ϕA , was then calculated for ideal swelling and for εpoly = 4.8 (PEUT). It is plotted in Fig. 5.9 as a function of the polymer layer thickness. The physical sensitivity accounts for the changes in dielectric constant and layer thickness but does not include the chemical selectivity. As explained above, sensitivities of thin layers are all positive. For even thinner layers, the sensitivities have to decrease because smaller polymer volumes absorb less analyte. For thick layers, the sign of the sensitivity depends on the dielectric constant of the analyte. Responses to analytes that have dielectric constants lower than that of the polymer (such as toluene) are positive for thin layers and show a transition from positive to negative values with increasing layer thickness. Hence, for each of these analytes, there is a critical layer thickness, where the sensitivity towards this analyte vanishes (see circle in Fig. 5.9). Consequently, the sensitivity reaches a (local) maximum somewhere between the critical layer thickness and zero. For analytes with dielectric constants higher than that of the polymer (such as ethyl acetate), the sensitivity first increases for thin layers, reaches also a maximum, decreases again and then saturates. In contrast to low-ε analytes, the saturation is at positive levels. With increasing dielectric constant of the analyte, the saturation level increases, and the maximum becomes less and less evident. For very high-ε analytes (such as water), the maximum vanishes completely. The sensitivity


physical sensitivity [pF]

8

water ethanol isopropanol

6

97

ethyl acetate toluene n-octane

4 2 0 0

1

2 3 4 5 polymer thickness [µm]

6

7

Fig. 5.10. Measured dependence of the physical sensitivity, ∆C/ϕA , on the layer thickness, h, for analytes with various dielectric constants: Low εA : n-octane (1.93) and toluene (2.36), high to very high εA : ethyl acetate (5.88), isopropanol (18.5), ethanol (24.3), and water (76.6) [517]

increases monotonically with increasing layer thickness and saturates already at relatively thin layers. To verify the simulations, the sensitivities of PEUT to various analytes were measured for layer thicknesses between 0.3 and 7.1 µm. The physical sensitivities, ∆C/ϕA , displayed in Fig. 5.10 have been calculated with independently determined partition coefficients that have been measured with thickness-shear-mode resonators (TSMRs) [517, 519]. The measurements confirm the simulations: saturation-like behavior already at relatively low thickness for water (very high ε), an increasingly clear maximum for ethanol, isopropanol, and ethyl acetate (high ε), and the transition to negative values for n-octane and toluene (low ε). 5.1.3.2 Insensitivity to Low-ε Analytes In the case of low-dielectric-constant interferants, a first method to achieve insensitivity is to choose the critical layer thickness. For each low-ε analyte, there is such a critical layer thickness, where the effect of swelling and dielectric constant change cancel out (see circles in Figs. 5.9 and 5.10). This idea, already mentioned in references [476] and [479], suffers from the required accurate adjustment of the layer thickness, which is a tedious business. Thus, another solution is presented, where the thickness adjustment is not so critical. By tuning the dielectric constant of a thick sensitive layer to be similar to that of the analyte, the sensitivity of the sensors to the respective

98


frequency shift [kHz]

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

2

1

ethanol n-octane

0

0 PEUT

25

50 75 mass fraction [%]

100 PECH

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

Ethanol with a high dielectric constant causes positive signals for both polymers and all blending ratios. n-Octane has a dielectric constant higher than that of PECH but lower than that of PEUT. Consistent with the theory, the signals of n-octane are positive for PECH and negative for PEUT. For the blended polymers, the signals show a continuous transition from positive to negative values, approximated with a quadratic function in Fig. 5.11. As can be seen, the analyte sensitivity vanishes when the dielectric constant of polymer mixture and analyte coincide (see circle in Fig. 5.11) [517]. 5.1.3.3 Humidity Interference Even more important is the elimination or reduction of the influence of humidity. Because water has a much higher dielectric constant than any polymer used, it is not possible to achieve humidity insensitivity by either of the methods discussed above. However, the thickness-dependence shown in Figs. 5.9 and 5.10 gives rise to another method, which is schematically depicted in Fig. 5.12. The sensitivity to analytes with high dielectric constant reaches saturation at low thickness, while the sensitivity to analytes with low dielectric constant

normalized sensitivity


99

humidity

0

1

2

3

4

5

h [µm] toluene n-octane

sensor 2 sensor 1

Σ∆

difference

Fig. 5.12. Eliminating humidity interference by recording difference signals using a thick and a medium thick layer: The simulated sensitivities from Fig. 5.9 are plotted as a function of the polymer layer thickness, h. Two sensors coated with a thick and a medium thick polymer layer are represented by the dashed vertical lines at the corresponding layer thickness. The difference signal, generated by connecting both sensors differentially to the Sigma-Delta-modulator, is indicated with the bars on the right hand side of the plot [517]

still varies considerably out to greater thicknesses. Hence, the signal difference of two capacitors with different layer thicknesses in the range of 1 µm to 5 µm is almost insensitive to water but retains sensitivity to n-octane and toluene, i.e., n-octane and toluene show a considerable signal gradient in that thickness region, whereas water behaves like ethanol and shows saturation. The initial idea of this method was presented already in [476] and [479]. There, the signals of two sensors were read out subsequently and then numerically subtracted. However, calculating differences of similar values is very error-prone. Accordingly, slight temporal humidity fluctuations resulted in a high noise in the signal. To overcome this obstacle, two sensors on the same chip were directly connected to the positive and negative input of the Sigma-Delta converter (Fig. 5.13). The signals were subtracted instantaneously so that any fluctuations cancelled out before being processed by the read-out circuitry– resulting in a significantly improved signal-to-noise ratio. The microsensor array used for these investigations included seven sensor and five reference capacitors, all of which are monolithically integrated with driving and read-out circuitry on a single chip (see Fig. 5.13). The reference capacitors are similar to the sensors but not coated with a polymer layer and allow for a differential measurement. A multiplexer (MUX) connects sensors and references to the read-out circuitry, a Sigma-Delta modulator (Σ∆). The Σ∆-modulator converts the analog capacitance signal to a digital output signal. The frequency of the output signal – measured with an external counter – is proportional to the difference of sensor and reference

100


S R

Σ∆

S R

R

S

S

MUX

R R

S

S

S chip size: 5.8 x 6.5 mm Fig. 5.13. Micrograph showing the sensor chip with seven sensor (S) and five reference (R) capacitors and readout circuitry: Sigma-Delta-modulator (Σ∆) and multiplexer (MUX) [476]

capacitance. The Σ∆-modulator is described in detail in references [514] and [515]. Two sensor capacitors of a chip as shown in Fig. 5.13 were coated with 1.4 and 3 µm PEUT. Figure 5.14 displays the signal of the sensor with the thin layer and the signal difference upon exposure to different concentrations of humidity, n-octane, and toluene. The difference signal is more than one order of magnitude less sensitive to water than the signal from a single sensor, whereas for n-octane and toluene, the signals stay in the same range. A second drastic improvement in signal quality was achieved by using the difference method, which has been detailed in Sect. 5.1.1. Drift, which is similar for both sensors, was also reduced by approximately an order of magnitude, which can be seen in the right part of Fig. 5.14. Especially for the humidity steps, sharp signal peaks appear in the difference signal when the analyte concentration is switched on and off. They originate from the different absorption times for the thin and the thick polymer layer, resulting in a temporarily high difference signal. The steady-state signal is not affected.

5.2 CMOS Calorimetric Device 5.2.1 CMOS Calorimetric Transducer The calorimetric transduction principle and, in particular, the thermoelectric or Seebeck-effect-based transducer has been introduced already in Sect. 4.2.2.

5.2 CMOS Calorimetric Device

∆f [kHz] humidity

40 n-octane

4.0

toluene

20

2.0

0

0.0 humidity

-20

∆f [kHz] n-octane, toluene

1.4 µm 1.4 µm - 3 µm

101

-2.0 60

120 time [min]

180

Fig. 5.14. Frequency shifts of sensors coated with PEUT upon exposure to humidity (12.5, 25, and 50% RH), n-octane (300, 600, and 1200 ppm), and toluene (400, 800, and 1600 ppm) as a function of the measurement time. The solid line is the difference signal of two sensors coated with 1.4 and 3 µm PEUT, whereas the dashed line represents the signal of the sensor with 1.4 µm PEUT. For better visibility, the signals of n-octane and toluene are displayed in a ten-times larger scale than the humidity signals. The difference signal is much less sensitive to humidity than the signal from a single sensor, but exhibits a comparable sensitivity to n-octane and toluene [517]

The expected temperature changes upon analyte interaction with the sensitive layer are in the milli-Kelvin range. Therefore, the measurement area has to be thermally insulated from the silicon substrate, which is an excellent thermal conductor. Membranes consisting of the dielectric layers of the CMOS process (silicon oxide, silicon nitride) have been previously used for infrared sensor arrays [520,521]. Dielectric membranes show a parabolic temperature profile, albeit a flat temperature profile over the sensitive area is desirable for gas-sensing applications. Such flat temperature profile can be obtained by introducing a silicon-n-well island as shown in Figs. 4.10 and 5.15, which is thermally decoupled from the bulk silicon of the chip. Predominantly the island area is afterwards coated with the sensitive layer. The temperature on the membrane can be assessed with different transducers using, e.g., the temperature coefficient of bipolar transistors or resistors. The sensitivity of these realizations is, however, not suitable to measure temperature changes in the milli-Kelvin range. The solution of choice is the utilization of thermocouples, which rely on the Seebeck-effect (see Sect. 4.2.2). The hot contacts are located on the thermally insulated n-well island, while the cold contacts are placed on the substrate. Maximum signals can be obtained using bismuth and antimony as thermocouple materials [520]. Both are not available in CMOS processes and would be difficult to deposit and to pattern. From all CMOS materials (see Sect. 3.3) the polysilicon/aluminum

102


dielectric layers

metal: Al

polysilicon heater

thermocouple }

n-well (Si-island)

polysilicon

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

reflection spot

polysilicon heater

thermocouples

distance n-well to bulk

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

thermocouple exhibits the highest Seebeck coefficient of 110 µV/K. Many thermocouples must be connected in series to achieve the desired sensitivity, i.e., measuring temperature differences in the milli-Kelvin range (Fig. 5.16). As already mentioned in Sect. 4.2.2, the detection process includes four principal steps: (I) absorption and partitioning or chemical reaction, (II) generation of heat, which causes (III) temperature changes to be transformed in (IV) thermovoltage changes. The final signal results from this sequence of processes, some of which are of chemical or physicochemical nature, i.e., depend on the involved chemical compounds (I, II), and some of which are of physical nature and are device-specific (III, IV: heat sensitivity and thermovoltage generation). The overall sensitivity is the recorded thermovoltage change ∆U [V] per change in the analyte concentration as a function of time dcA /dt [mol/m3 s] and can be written according to (4.3), Sect. 4.2.2:


103

∆U = A · B · Vsens · ∆H · K . (5.6) dcA /dt A [K·s/J] and B [V/K] denote device-specific constants describing the translation of a generated molar enthalpy ∆H [J/mol] via a temperature change into a thermovoltage change. The constant A [K·s/J] includes the feature sizes of the transducer and material properties, and B [V/K] includes the number of thermocouples and the Seebeck coefficients of the thermocouple components. Vsens denotes the sensitive-layer volume in case of a bulk effect, and K is the partition coefficient (2.10) or reaction equilibrium constant (2.2, 2.11). The overall sensitivity hence includes two contributions, the chemical and the physical sensitivity. The chemical sensitivity is correlated to the amount of heat that is generated upon an analyte concentration change. It comprises the reaction constant or partition coefficient, K, the produced enthalpy change, ∆H, and the sensitive layer volume, Vsens . This part will be treated in more detail in Sect. 5.2.3. The physical sensitivity is the thermovoltage output of the system in response to a defined heating power on the membrane. It includes the transducer features, whereas it does not depend on the detection chemistry. Some of the most important design parameters are briefly discussed. S=

• Size of the membrane: A large membrane leads to a large sensitive area and a large number of thermocouples. The mechanical stability of the membranes and the maximum allowable chip area determine the size limits. The membranes used for our devices have sizes of 650 × 650 µm2 (square membrane) and 2150 × 750 µm2 (rectangular membrane, see Figs. 5.16, 5.17). • Distance between substrate and n-well island: A large distance improves the thermal isolation of the n-well island and, therefore, the efficiency of the heating. The mechanical stability of the membrane and chip area considerations have to be taken into account. The increase of the electrical thermocouple resistance owing to the larger distance between hot and cold contacts also is an issue. • Number and width of the thermocouples: The achievable signal increases linearly with the number of thermocouples. The thermocouples, however, consist of aluminum and polysilicon, both of which exhibit comparably high thermal conductance. A large number of thermocouples thus reduces the thermal resistance of the membrane and, consequently, degrades the temperature effects. This would suggest the use of a large number of minimum-width conductors to increase the thermal resistance and make efficient use of the available area. Since polysilicon resistors exhibit low electrical conductivity, the line width cannot be reduced too much without increasing the thermal noise in the signal. In a viable compromise the square membrane design has a total of 132 thermocouples (40 kΩ electrical resistance), the rectangular membrane (Figs. 5.16,

104


sensor

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

5.17) features 300 thermocouples (143 kΩ electrical resistance). A differential strategy that has been already utilized in the case of the capacitive transducer is also applied to the calorimetric transducers as can be seen in Fig. 5.17. Such differential measurements are preferred over an absolute measurement, since the influence of ambient-medium or gas-flow temperature fluctuations is strongly reduced. 5.2.2 Calorimeter Circuitry The sensor system has to be optimized for a maximum signal-to-noise ratio (SNR) while allowing for measurements of a wide concentration range of different analytes. The thermopile converts the temperature difference between substrate and n-well island into a voltage signal. For the best overall performance, the input-referred SNR of the first amplifier and not the signal has to be maximized. The two thermopiles used here exhibit electrical resistances of 40 kΩ (square membrane) and 143 kΩ (rectangular membrane). The overall system optimization includes the input noise of the first amplifier as well as the power and area consumed by the complete readout-circuitry [501, 520]. Another important parameter is the signal bandwidth, which must be as small as possible in order to minimize the noise. The highest frequency of interest in the transient signal during absorption or desorption of analytes was experimentally determined to be approximately 400 Hz. A system and circuitry architecture based on the aforementioned considerations is shown in Fig. 5.18 [501]. It includes a differential arrangement, in which a polymer-coated sensor is connected in series to an uncoated reference. Temperature fluctuations owing to medium flow and radiation are largely cancelled out by the differential arrangement. The small signals are


sensor

+ low-noise chopper amplifier

reference

−

anti-aliasing filter

105

decimation filter 13bit

A D Sigma-Delta A/D converter

Fig. 5.18. Schematic of the calorimeter circuitry. Sensing and reference thermopiles are connected to the input stage of a low-noise chopper-stabilized instrumentation amplifier followed by an anti-aliasing filter, a Sigma-Delta A/D converter and a decimation filter [501]

amplified by a low-noise chopper amplifier. The chopping frequency is 5 kHz. The gain of the amplifier must be adjustable in order to allow for measurements over a large concentration range, i.e., for the use of a variety of different analyte-sensitive-layer combinations. The maximum gain of 6400 can be reduced by a factor of 4 or 16. The resolution is 12 bits, which is sufficient for the target applications. An anti-aliasing filter succeeding the amplifier is needed to avoid down-conversion of high frequency noise by the A/D-converter. The system has a narrow signal band from 0–400 Hz. The requirements for the stop-band are set by noise constraints, by the suppression of the signals upconverted to the chopping frequency by the demodulator at the output of the amplifier, and by clock-feedthrough. A minimum damping of 80 dB at the chopping frequency is needed to meet these requirements. The low cut-off frequency of 400 Hz would require large on-chip capacitors in the anti-aliasing/bandstop filter if a Nyquist-rate analog-digital converter would be used. By using a largely oversampled Sigma-Delta-A/D-converter, the filtering can be shifted to the digital decimation filter, where low frequencies do not increase complexity, area, and power-consumption of the design. This also simplifies the design of the anti-aliasing filter. Due to the narrow signal band, an oversampling factor of 128 can easily be realized. This makes a one-bit Sigma-Delta modulator the most suitable choice for the 12-bit A/Dconverter owing to the less stringent matching requirements. For a detailed description of each of the circuitry components, see [501, 523, 524]. 5.2.3 Calorimetric Gas Sensing Calorimetric sensors rely on determining the presence or concentration of a chemical by measurement of an enthalpy change produced by the chemical to be detected. Any chemical reaction or physisorption process releases or absorbs a certain quantity of heat from its surroundings. As already briefly mentioned in Sect. 4.2.2, the calorimetric sensor only detects changes in the heat budget at nonequilibrium state (transients) upon changes in the analyte concentration.

106


Thermodynamic equilibrium is a dynamic equilibrium state, in which reactants and products are present but have no tendency to undergo net change, i.e., as many molecules undergo the forward reaction as undergo the reverse reaction. The change in the Gibbs free energy, ∆G, is zero (see 2.1 and related text), and there is consequently no net heat production. The calorimetric sensor, therefore, only produces transient signals as is illustrated in Fig. 5.19. The initial state is thermal equilibrium with no analyte present and, hence, no temperature signal. The analyte concentration then is drastically increased, which leads to sorption/reaction heat liberation and, e.g., a temporal temperature increase on the sensor (Fig. 5.19b). As soon as equilibrium conditions, i.e., a constant analyte concentration is established, the signal returns to zero, because there is no more net enthalpy change (Fig. 5.19c).

polymer

signal time

(c) equilibrium

signal

(b) analyte absorption

signal

(a) no analyte

time

time

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

In case that an analyte concentration increase produces a positive calorimetric signal, the corresponding decrease produces a negative signal of equal intensity and vice versa (see Fig. 5.21). The signal and detection characteristics of the calorimetric transducer have some implications for its use in gas sensing. Slow changes in analyte concentrations are difficult to detect, since they continuously produce but small enthalpy changes. Therefore, a switching scheme has to be implemented, which allows for alternately exposing the sensor to analyte-loaded medium and non-contaminated medium, i.e., generates fast concentration changes or large concentration gradients. In a switched system, the amplitude of the signal, i.e., the peak height, largely depends on the time scale of the diffusion processes in the medium manifold and the diffusion of the analyte into the sensitive layer. Therefore,


107

the peak area integrated over time is the characteristic sensor signal that represents the total enthalpy change upon concentration changes (see also Fig. 5.21). For gas sensors with standard polymer coatings (deposited by spray or spin-coating), which will be discussed in more detail here, only physisorption contributes to the enthalpy change, since no chemical reaction is taking place, and no chemical bonds are formed between polymer and the analyte to be detected. The temperature change in the polymer, ∆T, upon analyte sorption arises from the overall heat production, ∆H, which, in the special case of volatile absorption in a polymeric matrix, includes two components, the heat of condensation and the heat of mixing: ∆T ∝ ∆Htotal ;

∆Htotal = ∆Hcond + ∆Hmix .

(5.7)

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

analyte molecules

polymer matrix sorption

=

condensation

+

mixing

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

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

108


thermovoltage [mV]

n-octane: 900 ppm

on

off

PDMS

time [s]

time [s]

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

to air at a temperature of 28◦ C. The sensitive layer is poly(dimethylsiloxane), PDMS. As already discussed above, two transient signals are produced, a positive one at the onset of the analyte exposure (liberation of predominantly condensation enthalpy) and a negative one upon terminating analyte exposure (abstraction of the enthalpy necessary for vaporizing the absorbed analyte). The net enthalpy changes can be approximated by integration over the peak area of the sensor signals [522, 523], which is the measurand of interest. The height of the peak maximum is strongly depending on the polymer thickness (diffusion kinetics), and the peak maximum and signal characteristics may also differ for switching on and switching off the analyte as a consequence of slightly different switching times or gas flow fluctuations. An optimized manifold with meticulously controlled gas flow, short gas paths, and fast and reproducible valve switching times is of paramount importance. The changes in the sensor response patterns upon varying the polymeric sorption matrix are displayed in Fig. 5.22. Whereas the slightly polar poly(etherurethane) shows large signals upon alcohol exposure (methanol, ethanol), the nonpolar poly(dimethysiloxane) provides hardly any signal for alcohols albeit showing large signals for nonpolar compounds such as n-octane and toluene. The sensor response patterns in Fig. 5.22 demonstrate, that sensor selectivity can be tuned by polymer variation and confirm the old alchemist rule: “Like dissolves like”, i.e., nonpolar matrices predominantly absorb nonpolar analytes, as well as polar matrices tend to absorb polar analytes. Note that the PDMS layer is with 1.5 µm thinner than the 3 µm PEUT layer and that the signals scale linearly with the polymer volume/layer thickness. In comparing Figs. 5.21 and 5.22 with capacitive sensor signals (Fig. 5.8) or cantilever signals (Fig. 4.1 and Sect. 5.3.3) it is evident, that there is

5.3 CMOS Integrated Resonant Cantilever 0.6

0.6 themovoltage [V]

109

PEUT

0.4

PDMS

0.4

0.2

0.2

0

0

-0.2

-0.2

-0.4

-0.4

-0.6

ethanol 1000-4000

toluene 500-2000

methanol 2000-3500

n-octane 250-1000 ppm

-0.6

0

50

100 150 time [min]

200

250

150

200 250 300 time [min]

350

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

significant differences. The calorimetric sensor provides two signals for each analyte exposure, which makes the concentration assessment more reliable. The measurement time is also significantly shorter since the transient evolution is on the order of 3–5 seconds, whereas it may take tens of seconds to reach equilibrium and a reliable sensor reading for equilibrium-based sensors at similar polymer layer thickness (approx. 3 µm). The calorimetric sensor signal is in addition inherently drift-free: No rapid concentration change means no sensor signal. Drawbacks of the calorimetric principle include a somewhat inferior analyte sensitivity in comparison to other polymer-based sensors (factor of approximately three to five) and the necessity of drastic gas concentration changes, i.e., a gas switching mechanism.

5.3 CMOS Integrated Resonant Cantilever As already mentioned in Sect. 4.1.3, cantilevers are mostly used as gravimetric transducers and can be operated either in a static mode by measuring the cantilever deflection or in the dynamic mode by assessing changes in the resonance behavior. The fabrication in industrial CMOS technology allows for the integration of transducers, driving circuitry, and analog or digital signal processing circuitry on the same chip [93,95]. Integrated thermal or magnetic excitation and piezoresistive detection schemes render the system independent of additional excitation elements such as piezoelectric layers [61–63,106], and independent of bulky external optical detectors [100, 103–105]. The cantilevers that will be described in more detail below are operated in the dynamic mode, and the change of the resonance frequency upon mass loading is measured. The operation features of resonant cantilevers are, as already mentioned, similar to those of other well-established mass-sensitive gas

110


sensors such as thickness-shear-mode resonators (TSMRs, quartz microbalances) [61–63, 66, 68–70] and surface-acoustic-wave (SAW) [61–63, 68–70] devices. 5.3.1 Resonant Cantilever Transducers There are basically three ways to determine the resonance frequency of a cantilever [501]: • Noise spectrum: The frequency spectrum of the thermal noise is measured and analyzed. The peak frequency is equivalent to the resonance frequency. This method does not require any excitation of the cantilever, but a bulky low-noise spectrum analyzer is needed for the measurement. • Frequency sweep of excitation signal: The cantilever is excited at different frequencies and its response is evaluated to determine the resonance frequency. A gain-phase analyzer is needed for the measurement. • Oscillator: The cantilever is used as the frequency-determining element in an oscillator loop. A cantilever actuation mechanism and a deflectionsensor are needed to realize such an oscillator. A fast feedback is required, but a simple counter can be used to determine the resonance frequency. The third option provides the most accurate frequency assessment, since the quality factor (Q-factor, ratio of center frequency and bandwidth) of the cantilever, which describes the sharpness of the system response, is enhanced by the feedback. In addition, only the third option can be realized in CMOS technology at reasonable expenses, since integrated spectrum analyzers and gain-phase meters are difficult to design. Consequently, the oscillator option was realized. In the following, two different methods to actuate the cantilever (thermal and magnetic), and the vibration detection via piezoresistors and stress-sensitive transistors will be presented. 5.3.1.1 Thermal Actuation The cantilever consists of single-crystal silicon covered by dielectric layers such as silicon oxide or silicon nitride for electrical insulation of the integrated electronic components. A cross-sectional view is shown in Fig. 5.23. Two heating resistors are integrated in the cantilever base, which are heated periodically to achieve cantilever vibration (see also Fig. 5.24). The temperature increase on the cantilever generates a bending moment due to the difference in thermal expansion coefficients of the silicon and the dielectric layers (thermal bimorph effect). The area of periodic temperature variation is closely confined to the region around the heaters due to the high excitation frequency of approx. 400 kHz. At frequencies higher than the corner frequency of approximately 1 kHz (speed of the thermal equilibration processes in the cantilever), the

5.3 CMOS Integrated Resonant Cantilever

111

Si-nitride Si-oxide polymer

n-well heater (p-diffusion) silicon frame Fig. 5.23. Cross section of a thermally actuated cantilever in CMOS technology. The heater at the cantilever base induces cantilever bending as a consequence of the bimorph effect (difference in thermal expansion coefficients of the silicon and the dielectric layers) [526]

heater

50 µm

piezo resistors

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

efficiency of the thermal actuation usually significantly drops, which reduces the cantilever oscillation amplitude and leads to a phase lag that increases with frequency. However, the created bending moment is still sufficient to cause harmonic transverse vibrations of the cantilever with an amplitude of a few nanometers. The deflection of the cantilever is proportional to the applied heating power. Therefore, a sinusoidal voltage excitation cannot be used, since all the heating power goes into the system at DC and twice the resonance frequency. A DC-offset has to be added in order to excite the fundamental resonance of the cantilever [501]. Both types of cantilevers (thermally or magnetically actuated) are formed after completion of the regular 0.8-µm CMOS process sequence [496,527]. The first step includes anisotropic wet etching of silicon from the wafer backside with KOH using an electrochemical etch-stop technique (Fig. 5.25b). After

112


this step the silicon cantilevers are still embedded in a membrane formed by the dielectric layers. The second post-processing step consists of isotropic wet etching of silicon dioxide with buffered oxide etch (BOE), which finally releases the discrete silicon cantilevers (Fig. 5.25c). The cantilevers are composed of a layer sandwich including the dielectric layers of the CMOS process and a layer of single-crystal silicon (Fig. 5.23). The silicon layer has a thickness of 5.5 µm, silicon dioxide and silicon nitride thicknesses are 1.5 µm and 1.1 µm, respectively. The cantilevers are 150 µm long and 140 µm wide. Their fundamental resonance frequency is in the range of 400 kHz, and they exhibit a Q-factor of up to 1000 in air without chemically sensitive coating. dielectric layers (a) Si

n-well

(b) heater (c)

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

The quality factor of the resonant cantilevers at atmospheric pressure is dominated by the viscous damping in the surrounding air [93]. In comparison to cantilevers employed in Scanning Probe Microscopy (SPM), the cantilevers as used for chemical sensors exhibit high quality factors. This is partly due to the fact that the cantilevers are not operated in close vicinity of a sample surface, and, hence, no squeezed-film damping occurs. The distance to the bottom of the etch cavity is 380 µm, i.e., large in comparison to the cantilever dimensions. Furthermore, the quality factor is not only a function of the resonance frequency but also of the cantilever geometry and its spring constant [528]. While SPM cantilevers for dynamic mode operation typically exhibit spring constants between 1 and 40 N/m, the cantilevers here were designed


113

to exhibit a much higher spring constant of 800 N/m. The measured quality factor of cantilevers in fundamental resonance strongly increases with increasing resonance frequency up to 200 kHz, reaching maximum values of approximately 1000 [93, 526]. This behavior is predicted by the theory of a damped harmonic cantilever beam with a damping term proportional to the beam velocity [529]. A theoretical estimation of the actual quality factors would, however, require a model for the dependence of the damping term on the cantilever dimensions. At frequencies higher than 200 kHz, a saturationlike behavior of the quality factor has been observed, indicating additional loss mechanisms, such as coupling losses into higher resonant modes or losses due to onset of acoustic radiation at higher frequencies. 5.3.1.2 Magnetic Actuation The magnetic actuation relies on Lorentz forces and requires an external magnetic field, which can be conveniently generated by including a small permanent magnet in the sensor package below the cantilever (Fig. 5.26) [530, 531]. The resulting magnetic-field vector is in plane and parallel to the cantilever. By applying an AC current such as a sinusoidal to the current loops that are patterned along its edges, cantilever oscillation is evoked in the external magnetic field by Lorentz forces. The direction of the Lorentz forces is perpendicular to the cantilever (Fig. 5.27) such producing a transverse cantilever movement. Two designs with different current paths for the electromagnetic actuation were implemented [531]. The current paths are realized using the two metal layers of the CMOS process. One design has a path with two loops and a resistance of approximately 12 Ω, the second has 8 loops and cantilever magnetic field lines

disc magnet

Si-substrate

Al-package

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

114


B

Close-up Wheatstone bridge

J

stress-sensitive transistors

X FL

50 µm cantilever

current loop

Wheatstone bridge

Fig. 5.27. Micrograph of a magnetically actuated cantilever: AC current applied to the loops and external B-field initiate a cantilever oscillation via Lorentz forces. The close-up shows the arrangement of the stress-sensitive transistors [531]

a resistance of 48 Ω. The Lorentz force, FL , acting on the cantilever can be approximated according to the following equation [531]: FL = N · I · l · Bext .

(5.8)

N represents the number of current loops, I the applied current, and l the mean length of the loop perpendicular to the external magnetic field, Bext , (see Fig. 5.27). When the magnetic field, Bext , is oriented parallel to the cantilever length, the Lorentz-force vector is perpendicular to the plane of the cantilever. The Lorentz force exerted on the cantilever critically depends on the number of current loops. With a magnetic field of 80 mT, the effective current for 2 loops is 3.5 mA, and the resulting Lorentz force amounts to 37 nN, whereas in the case of 8 current loops, the effective current is 7.1 mA and the Lorentz force 71 nN. Especially for thick layers of chemically sensitive coatings, it is therefore necessary to use cantilevers with larger numbers of current loops to reach stable oscillation conditions. 5.3.1.3 Vibration Detection The Wheatstone bridge for vibration detection is located at the clamped edge of the cantilever, where the stress induced by the deflection is maximal. The bridge consists of either four piezoresistors (Fig. 5.24) or four diode-connected PMOS transistors (Fig. 5.27) [526]. The piezoresistive effect on a slightly p-doped silicon resistor is described by the relative resistance change, ∆R/R,


115

∆R = π L σ L + πT σ T (5.9) R where πL,T are the longitudinal and transversal piezoresistive coefficients and σL,T denote the respective stress-components. The piezoresistive coefficients in parallel and perpendicular orientation with respect to the cantilever axis have opposite signs. Therefore, a differential signal can be obtained by arranging two resistors in parallel and two resistors perpendicularly to the cantilever axis in a Wheatstone-bridge configuration (see Fig. 5.24). The common-mode voltage varies only a few percent because the absolute values of the longitudinal and transversal piezoresistive coefficients are almost identical [501]. The resistances in the Wheatstone-bridge are 950 Ω. This leads to a considerable power consumption of 26 mW. As minimum-width resistors had to be chosen for the Wheatstone-bridge due to area restrictions, a good matching cannot be expected. The matching of the resistors is further deteriorated by their perpendicular orientation. This leads to a large offset-voltage of up to 20 mV. MOS-transistor detection schemes have also been developed since mechanical stress changes the carrier mobility in the channel of MOS transistors [532, 533]. Four diode-connected PMOS transistors in a bridge configuration are used to generate an output voltage that is proportional to the stress at the clamped edge of the cantilever [532]. Again, opposite signs of the stress-sensitivities are achieved by orienting the channels of the two pairs of transistors perpendicularly and in parallel to the cantilever axis (see close-up in Fig. 5.27). The diode-connected MOS transistors show a lower small-signal transconductance in comparison to diffused resistors, which reduces power and area consumption. The disadvantages include a somewhat larger noise and a reduction of the sensitivity. 5.3.1.4 Cantilever Temperature The power dissipation of the electrothermally actuated cantilever is 32 mW with 6 mW resulting from the actuation and 26 mW from the piezoresistive Wheatstone bridge. For the magnetically actuated cantilever the power dissipation of the cantilever adds up to a total of 1.3 mW [501, 531]. It results mainly from the MOS-transistor Wheatstone bridge, the power dissipation of the actuation is negligible. The Wheatstone bridge with MOS-transistors has less power dissipation since its active load (20 kΩ) is much larger than that of the design with p-diffused resistors (950 Ω). As a consequence of the high power consumption and dissipation in the case of electrothermal actuation and piezoresistive detection, the cantilever temperature is significantly enhanced as can be seen in Fig. 5.28, which displays the simulated temperature distribution on the cantilever at minimum possible power consumption (actuation: 6 mW, bridge: 1.2 mW) [526]. The cantilever temperature under realistic operation conditions can be up to 20◦ C higher than ambient temperature, which strongly affects the detection process of chemicals, such

116


0 1.7 3.5 Wheatstone bridge (1.25 mW)

thermal actuators (6 mW)

temperature variation [°C]

Fig. 5.28. Temperature distribution on a thermally actuated cantilever with piezoelectric readout [526]

as the absorption of organic volatiles in polymeric matrices. As a rule of thumb, a 10-degree temperature increase here reduces the sensor signal by 50%. Magnetic actuation and MOS-transistor readout that do not produce any significant temperature change on the cantilever are, therefore, in most cases, a better choice. 5.3.2 Microcantilever Circuitry The cantilever constitutes the frequency-determining element of an oscillator circuit, the nature of which depends on the actuation/detection principle and will be, therefore, specified for thermal and magnetic actuation. 5.3.2.1 Thermal Actuation To operate the mechanical oscillator, the displacement generated by the heating pulses and the excitation voltage variations have to be in phase for positive feedback. A static DC-component has to be superimposed to a periodic excitation voltage to achieve cantilever oscillation at the mechanical fundamental resonance (see Sect. 5.3.1.1) [501]. This DC component of the heating current produces a general temperature increase on the cantilever. The power dissipation in the piezoresistors of the Wheatstone bridge constitutes a second source of heating. Both heating effects can cause a total temperature increase of up to 20◦ C above ambient temperature in the chemically sensitive region of the cantilever with approximately 10◦ C above ambient temperature at routine operation [93]. Measurements showed that the signal from the mechanical vibration only dominates in vicinity to the resonance frequency. At low frequencies, the thermo-mechanical actuation heats the whole cantilever and creates a temperature gradient in the Wheatstone-bridge that leads to an offset. Due to


117

the large temperature coefficient of diffused resistors, the resulting signal is larger than the mechanical response for frequencies up to 200 kHz (uncoated cantilever). Low-frequency signals therefore have to be eliminated by the feedback circuit, in particular since the amplitude of the thermal crosstalk may come close to the amplitude of the mechanical signal even at resonance. For frequencies higher than 200 kHz, capacitive crosstalk is observed as a consequence of the small distance between the diffused heating resistors and the diffused resistors of the Wheatstone bridge. This small distance gives rise to parasitic capacitances through the n-well. The oscillator circuitry as shown in Fig. 5.29 was designed according to the preceding considerations to achieve optimum signal quality [501]. The output signal of the resistive Wheatstone-bridge is first amplified by a lownoise differential difference amplifier (DDA). The amplification factor has a maximum value of 35 in order to avoid saturation of the amplifier by the DC-offset of the Wheatstone-bridge. The signal is then high-pass filtered to remove the offset voltages of the bridge and the first amplifier. The high-pass filter also prevents up-conversion of the amplifier 1/f-noise and eliminates the low-frequency thermal crosstalk and related error signals. AC-coupling at the input of the first amplifier would allow for a higher gain in the first amplification stage, but the related disadvantages prevail. They include added noise and the creation of an additional path for switching interference and substrate noise to the most critical point in the circuit. cantilever high-pass RH

R1

R2

Wheatstone bridge

R3

R4

+ DDA 1st amplifier

limiter

+ DDA -

∆t

2nd amplifier

delay

Fig. 5.29. Schematic of the cantilever feedback circuitry, which includes two cascaded amplifiers, a high-pass filter, a limiter, a programmable digital delay line, and a driving stage [501]

The signal is then amplified and high-pass filtered a second time (not shown) before it is converted into a square-wave signal by the comparator. The second amplification stage is needed to achieve sufficiently large amplitudes at the input of the comparator. The minimum amplitude at the input of the comparator is determined by (a) the input offset of the comparator: An amplitude at least ten times larger than the offset-voltage (≈1 mV) is needed for the desired duty cycle of 45–55%; and (b) the noise and crosstalk at the input of the comparator.

118


The feedback loop is then closed via an inverter followed by a Schmitttrigger, which is operated as a delay element. The rise-time of the inverter is digitally adjustable. This way, the phase shift is adjusted to ensure positive feedback. A source follower at the output of the delay line is used to drive the small heating resistor. The result is an integrated oscillator operating at the cantilever resonance frequency with a short-term frequency stability better than 0.1 Hz. The integration of the feedback loop on chip (Fig. 5.30) massively improves the signal-to-noise characteristics of the sensor. The detection of mass changes of less than one picogram on the cantilever has been achieved by recording the corresponding shifts in the cantilever resonance frequency. For more details on the circuitry see [501, 526].

cantilever

feedback circuit

500 µm Fig. 5.30. Micrograph of a thermally actuated cantilever (piezoresistive detection) with monolithically integrated feedback circuitry [501]

5.3.2.2 Magnetic Actuation The magnetically actuated cantilever featuring a MOS-transistor detection scheme, 8 current loops and an overall coil resistance of 48 Ω, (see Fig. 5.27 and Sect. 5.3.1.2), has been incorporated into an integrated oscillator. Some key issues for designing the feedback circuitry shall be briefly discussed. The vibration-induced output signal of the Wheatstone bridge exhibits an amplitude of 1 mV for a bias-current of 8.3 mA (corresponds to an excitation voltage of 400 mV) and a magnetic field generated by a 100-mT electromagnet. Consequently, the feedback circuitry has to provide an amplification of at least 32 dB (air) or 63 dB (water) to achieve a loop gain of more than 0 dB. The signal amplitude at the output of the MOS-transistor bridge varies by almost an order of magnitude in dependence of the excitation current,


119

the magnetic field, the distance between the magnet and the cantilever, and the damping of the vibration through the media, in which the cantilever is operated. Additionally, there will be drift on the bridge sensitivity owing to temperature fluctuations and material aging. For operation in liquids, the high viscosity and density of the surrounding medium entails enhanced damping and, consequently, a reduced signal amplitude (e.g., 30 dB reduction). Therefore, the gain of the feedback loop has to be adjustable in order to achieve stable oscillation under all operating conditions. The tunable gain of the second amplifier is used for the coarse tuning during startup. Small variations during circuitry operation are continuously compensated by having a nonlinear transconductance in front of the output stage. The total phase shift between the excitation and the output signal of the circuitry must be close to zero degree over a wide frequency range in order to achieve stable oscillation at the mechanical resonance of the cantilever. This holds particularly true since the mechanical resonance frequency of the cantilever varies with the coating material or the surrounding medium and is additionally subjected to fabrication tolerances. The thermomechanically actuated cantilever is driven by a square-wavetype excitation, since its vibration is correlated to the heating power (see Sect. 5.3.1.1). For the magnetically actuated cantilever, a sine-wave-type signal can be used since the mechanical vibration is correlated to the excitation current. Phase noise and power consumption can then be optimized. The MOS-transistor Wheatstone bridge shows a large DC-offset of 25 mV ±5 mV. This offset must be reduced or eliminated to avoid saturation of the amplifiers in the oscillator loop. Furthermore, the excitation signal at the cantilever should not include any DC-component to prevent unwanted heating of the cantilever and to minimize the cantilever power consumption. The last stage of the feedback circuitry has to drive the low-resistance coils that feature only 48 Ω. The circuitry schematic is shown in Fig. 5.31 [531,534]. A low-noise differential difference amplifier (DDA) [535] was chosen for the first amplification stage, since it requires little area and has low power consumption. The gain defined by the feedback resistors is 30 dB to avoid saturation of the amplifier by the DC-offset of the Wheatstone bridge. Between the first and second amplification stage, a first-order high-pass filter with a cut-off frequency of 15 kHz is added to eliminate the DC-offset. The only requirement for the high-pass filter is a cut-off frequency, which is at least three times smaller than the resonance frequency of the cantilever. The second amplifier is a variable-gain amplifier based on the DDA topology. In comparison to the first DDA, the design of the second amplification stage is more flexible. Instead of using a defined feedback resistance to determine the closed-loop gain of the amplifier, a current-controlled differential linear transconductor is introduced into the feedback loop of the amplifier [536]. By changing the bias current of the transconductor in the feedback loop, the closed-loop gain can be varied. The closed-loop gain of the second amplifier

120

5 CMOS Platform Technology for Chemical Sensors cantilever differential difference high-pass amplifier (DDA) filter

variable-gain amplifier

readout analog digital

class-AB buffer

amplitude regulation

high-pass filter

Fig. 5.31. Schematic of the cantilever feedback circuitry for magnetic actuation [534]

can be adjusted from 34 dB to 43 dB by tuning this bias current in the range of 50 µA to 110 µA. It is intended to introduce a programmable bias current in the next design. After this stage, another high-pass filter, which has the same structure and same cut-off frequency as the first high-pass filter mentioned before, is added to remove the offset of the upstream amplifiers. To initiate a self-oscillation of the system, the Brownian motion of the cantilever has to be amplified. Consequently, the closed-loop gain of the system at the beginning has to be larger than 0 dB and has to be controlled so that the gain is depending on the oscillation amplitude. To this end, a nonlinear transconductance as proposed in [537] is implemented in the system to regulate the oscillation amplitude. The load of the feedback circuitry includes the coil, which is integrated on the cantilever. Compared to its resistance, the inductance of the coil can be neglected. The resistance of eight loops used in this design is only 48 Ω, which requires a comparatively large driving current. As low power consumption is an important issue for a portable sensor design, a class-AB buffer is used, which reduces the power consumption by a factor of five in comparison to a simple buffer structure. For more details on the circuitry see [531, 534]. The mass change of the cantilever upon analyte exposure leads to a frequency change. The oscillation signal can be read out as an analog voltage or as a digital square wave. In this design, a buffer drives the output signal, which is measured externally. A micrograph of the overall integrated system is shown in Fig. 5.32. A simple and affordable solution to generate the magnetic field is the integration of a permanent magnet in the chip package. Before designing the package, it was necessary to find the minimum magnetic induction needed for a stable oscillation of the cantilever feedback system. Measurements performed

5.3 CMOS Integrated Resonant Cantilever differential difference amplifier DDA

high-pass filter

121

variable-gain amplifier

500 µm

cantilever

class-AB buffer

nonlinear transconductance

high-pass filter

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

with a tunable electro-magnet showed that the minimum magnetic flux density necessary for stable oscillation is 70 mT [531]. The feedback circuitry exhibits a short-term frequency stability of better than 0.03 Hz in air. In contrast to thermal actuation, the application of which to resonant sensing in the liquid phase is difficult as a consequence of thermal losses and heat dissipation, the magnetic actuation scheme is also applicable to liquid phase dynamic measurements. Consequently the magnetically actuated cantilever system was characterized in air and in water. In both media, the system was first characterized in open-loop and, afterwards, in closed-loop operation (Fig. 5.33). In contrast to thermal actuation, the application of which to resonant sensing in the liquid phase is difficult as a consequence of thermal losses and heat dissipation, the magnetic actuation scheme is also applicable to liquid phase dynamic measurements. Consequently the magnetically actuated cantilever system was characterized in air and in water. In both media, the system was first characterized in open-loop and, afterwards, in closed-loop operation (Fig. 5.33). Under open-loop conditions at an external magnetic field of 200 mT, the cantilever oscillates at a resonance frequency of 425 kHz in air with a quality factor of 750 to 1100 (>100.000 for closed loop in air). In water, the fundamental resonance frequency of the system drops to 219 kHz with a quality factor of 23. Not only the quality factor but also the oscillation amplitude of the cantilever is reduced dramatically in water. The improvement of the quality factor of the cantilever in water as a consequence of closed-loop operation is demonstrated in Fig. 5.33 [534]. The quality factor increases from 23 in open-loop to 19,000 in closed-loop


amplitude [V]

122

closed-loop response Q = 19,000

open-loop response Q = 23

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

operation. The resonance frequency in closed-loop operation in water is 221 kHz, which is slightly different from the resonance frequency in open-loop operation. This is due to the fact, that in closed-loop operation the cantilever oscillates at a frequency, at which the total phase of the system is 0◦ , which is 221 kHz in this measurement. Figure 5.33 clearly demonstrates again the benefits of having circuitry monolithically integrated with the resonant structure. 5.3.3 Microcantilevers as Chemical Sensors To render the cantilever sensitive to chemical species, thin layers (3–6 µm) of polymers were deposited, which serve as absorption matrix to detect volatile organic compounds in air. The investigations were restricted to polymeric films, for which physisorption (no chemical interaction) and bulk dissolution of the analyte within the polymer volume are the predominant mechanisms. Upon absorption of analytes by the coating, the physical properties of the polymer film, such as its mass, change. This mass change is detected by monitoring the corresponding resonance frequency shift of the coated cantilever. 5.3.3.1 Polymer Coating The deposition of the sensitive polymeric layer, which is performed by spray coating using an airbrush or by drop deposition from polymer solutions, already changes the resonator properties of the cantilevers. Figure 5.34


123

-150

10 uncoated

uncoated

1 coated

0.1

phase [°]

amplitude [mV]

-200 -250 -300 -350 0.01 100

200 300 400

frequency [kHz]

-400 100

coated

200 300 400

frequency [kHz]

Fig. 5.34. Measured amplitude and phase of a cantilever before and after coating with PEUT (after amplification by the DDA, 30 dB). The mechanical resonance signal determines overall amplitude and phase characteristics only in vicinity to the resonance frequency. At lower and higher frequencies, thermal and capacitive crosstalk prevail [501]

shows a Bode-plot of a cantilever before and after coating with 2 µm poly (etherurethane), PEUT [501]. After coating, the resonance frequency and the amplitude at resonance decrease. A larger loop-gain is needed to compensate for the decrease in amplitude. For a harmonic oscillator, the phase shift between excitation at resonance and response is independent of the resonance frequency. Due to the thermomechanical actuation, the resonant cantilever shows an additional phase lag, which increases with frequency as was mentioned in Sect. 5.3.1.1. The phase shift at resonance is decreased after the coating procedure and depends on the polymer thickness. Due to this decrease and the fabrication-tolerance-induced fluctuation of the initial resonance frequency (

Integrated Chemical Microsensor Systems in CMOS Technology (Microtechnology and MEMS)

CMOS Hotplate Chemical Microsensors (Microtechnology and MEMS)

Piezoceramic Sensors (Microtechnology and MEMS)

Laser Diode Microsystems (Microtechnology and MEMS)

MEMS-Based Integrated Navigation

CMOS Hotplate Chemical Microsensors

CMOS-MEMS: Advanced Micro and Nanosystems

CMOS-MEMS: Advanced Micro and Nanosystems

CMOS Biotechnology (Series on Integrated Circuits and Systems)

RF MEMS Switches and Integrated Switching Circuits (MEMS Reference Shelf)

Force Sensors for Microelectronic Packaging Applications (Microtechnology and MEMS)

Bio-Medical CMOS ICs (Integrated Circuits and Systems)

Smart Material Systems and MEMS

Smart Mems And Sensor Systems

Integrated 60GHz RF Beamforming in CMOS

Reliability of Microtechnology: Interconnects, Devices and Systems

Transient-Induced Latchup in CMOS Integrated Circuits

Advanced Cmos Process Technology

The MEMS Handbook MEMS Applications

Integrated Systems of Meso-Meteorological and Chemical Transport Models

CCD Image Sensors in Deep-Ultraviolet: Degradation Behavior and Damage Mechanisms (Microtechnology and MEMS)

Electroceramic-based MEMS: fabrication-technology and applications

Smart Sensors and MEMS

Post-Processing Techniques for Integrated MEMS

CMOS Digital Integrated Circuits Analysis & Design

Design of Analog CMOS Integrated Circuits

Design of Analog CMOS Integrated Circuits (Solutions)

Design of Analog CMOS Integrated Circuits

Secure Integrated Circuits and Systems

Technology of Integrated Circuits

Clocking in Modern VLSI Systems (Integrated Circuits and Systems)

Integrated Chemical Microsensor Systems in CMOS Technology (Microtechnology and MEMS)

CMOS Hotplate Chemical Microsensors (Microtechnology and MEMS)

Piezoceramic Sensors (Microtechnology and MEMS)

Laser Diode Microsystems (Microtechnology and MEMS)

MEMS-Based Integrated Navigation

CMOS Hotplate Chemical Microsensors

CMOS-MEMS: Advanced Micro and Nanosystems

CMOS-MEMS: Advanced Micro and Nanosystems

CMOS Biotechnology (Series on Integrated Circuits and Systems)

RF MEMS Switches and Integrated Switching Circuits (MEMS Reference Shelf)

Force Sensors for Microelectronic Packaging Applications (Microtechnology and MEMS)

Bio-Medical CMOS ICs (Integrated Circuits and Systems)

Smart Material Systems and MEMS

Smart Mems And Sensor Systems

Integrated 60GHz RF Beamforming in CMOS

Reliability of Microtechnology: Interconnects, Devices and Systems

Transient-Induced Latchup in CMOS Integrated Circuits

Advanced Cmos Process Technology

The MEMS Handbook MEMS Applications

Integrated Systems of Meso-Meteorological and Chemical Transport Models

CCD Image Sensors in Deep-Ultraviolet: Degradation Behavior and Damage Mechanisms (Microtechnology and MEMS)

Electroceramic-based MEMS: fabrication-technology and applications

Smart Sensors and MEMS

Post-Processing Techniques for Integrated MEMS

CMOS Digital Integrated Circuits Analysis & Design

Design of Analog CMOS Integrated Circuits

Design of Analog CMOS Integrated Circuits (Solutions)

Design of Analog CMOS Integrated Circuits

Secure Integrated Circuits and Systems

Technology of Integrated Circuits

Clocking in Modern VLSI Systems (Integrated Circuits and Systems)

Recommend Documents