Semiconductor Gas Sensors

SEMICONDUCTOR GASSENSORS

This page intentionally left blank

SEMICONDUCTOR GASSENSORS

Woodhead Publishing Series inElectronic and Optical Materials

Second Edition

Edited by

RAIVO JAANISOUniversity of Tartu, Tartu, EstoniaOOI KIANG TANNanyang Technological University, Singapore

Woodhead Publishing is an imprint of ElsevierThe Officers’ Mess Business Centre, Royston Road, Duxford, CB22 4QH, United Kingdom50 Hampshire Street, 5th Floor, Cambridge, MA 02139, United StatesThe Boulevard, Langford Lane, Kidlington, OX5 1GB, United Kingdom

Copyright © 2020 Elsevier Ltd. All rights reserved.

No part of this publication may be reproduced or transmitted in any form or by any means,electronic or mechanical, including photocopying, recording, or any information storage andretrieval system, without permission in writing from the publisher. Details on how to seekpermission, further information about the Publisher’s permissions policies and ourarrangements with organizations such as the Copyright Clearance Center and the CopyrightLicensing Agency, can be found at our website: www.elsevier.com/permissions.

This book and the individual contributions contained in it are protected under copyright by thePublisher (other than as may be noted herein).

NoticesKnowledge and best practice in this field are constantly changing. As new research andexperience broaden our understanding, changes in research methods, professional practices, ormedical treatment may become necessary.

Practitioners and researchers must always rely on their own experience and knowledge inevaluating and using any information, methods, compounds, or experiments described herein.In using such information or methods they should be mindful of their own safety and the safetyof others, including parties for whom they have a professional responsibility.

To the fullest extent of the law, neither the Publisher nor the authors, contributors, or editors,assume any liability for any injury and/or damage to persons or property as a matter of productsliability, negligence or otherwise, or from any use or operation of any methods, products,instructions, or ideas contained in the material herein.

Library of Congress Cataloging-in-Publication DataA catalog record for this book is available from the Library of Congress

British Library Cataloguing-in-Publication DataA catalogue record for this book is available from the British Library

ISBN: 978-0-08-102559-8

For information on all Woodhead Publishing publications visit ourwebsite at https://www.elsevier.com/books-and-journals

Publisher: Matthew DeansAcquisition Editor: Kayla Dos SantosEditorial Project Manager: Ali Afzal-KhanProduction Project Manager: Debasish GhoshCover Designer: Matthew Limbert

Typeset by TNQ Technologies

http://www.elsevier.com/permissions

https://www.elsevier.com/books-and-journals

Contents

Contributors xi

Part One Basics

1. Fundamentals of semiconductor gas sensors 3

Noboru Yamazoe and Kengo Shimanoe

1.1 Introduction 4

1.2 Classification of semiconductor gas sensors 5

1.3 Resistor-type sensors: empirical aspects 6

1.4 Resistor-type sensors: theoretical aspects 14

1.5 Future trends 34

References 37

2. Conduction mechanism in semiconducting metal oxidesensing films: impact on transduction 39

N. Barsan, M. Huebner and U. Weimar

2.1 Introduction 39

2.2 General discussion about sensing with semiconducting metal

oxide gas sensors 41

2.3 Sensing and transduction for p- and n-type semiconducting

metal oxides 47

2.4 Investigation of the conduction mechanism in semiconducting

metal oxide sensing layers: studies in working conditions 57

2.5 Conduction mechanism switch for n-type SnO2–based sensors

during operation in application-relevant conditions 66

2.6 Conclusion and future trends 67

References 67

3. The effect of electrode-oxide interfaces in gas sensor operation 71

Sung Pil Lee and Chowdhury Shaestagir

3.1 Introduction 72

3.2 Electrode materials for semiconductor gas sensors 74

3.3 Electrode-oxide semiconductor interfaces 95

3.4 Charge carrier transport in the electrode-oxide semiconductor

interfaces 104

v j

3.5 Gas/solid interactions in the electrode-oxide semiconductor interfaces 119

3.6 Conclusions 124

References 125

4. Introduction to semiconductor gas sensors: a blockscheme description 133

Arnaldo D’Amico and Corrado Di Natale

4.1 Introduction 133

4.2 The sensor blocks 135

4.3 Metal oxide semiconductor capacitor: the case of the

hydrogen gas sensitivity of Pd-SiO2-Si 142

4.4 Light-addressable potentiometric sensor 144

4.5 Metal oxide semiconductor field-effect transistor 148

4.6 Metal oxide semiconductors 151

4.7 Conclusions 156

References 156

Part Two Materials

5. One- and two-dimensional metal oxide nanostructures forchemical sensing 161

E. Comini and D. Zappa


5.2 Deposition techniques 162

5.3 Conductometric sensor 169

5.4 Transduction principles and related novel devices 170


References 175

6. Hybrid materials with carbon nanotubes for gas sensing 185

Thara Seesaard, Teerakiat Kerdcharoen and

Chatchawal Wongchoosuk


6.2 Synthesis of carbon nanotube 192

6.3 Preparation of carbon nanotubedmetal oxide sensing films 194

6.4 Sensor assembly 199

6.5 Characterization of carbon nanotube–metal oxide materials 200

6.6 Sensing mechanism of carbon nanotube–metal oxide

gas sensors 205

vi Contents

6.7 Fabrication of electrodes and CNT/polymer nanocomposites

for textile-based sensors 206

6.8 Sensor assembly for textile-based gas sensors 210

6.9 Characterization of CNT/polymer nanocomposites sensing

materials on textile substrate 212

6.10 Sensing mechanism of CNT/polymer nanocomposites sensing

materials on fabric substrate 215

6.11 Conclusion 216

Acknowledgments 217

References 217

7. Carbon nanomaterials functionalized with macrocycliccompounds for sensing vapors of aromatic VOCs 223

Pierrick Clément and Eduard Llobet


7.2 Cyclodextrins 226

7.3 Calixarenes and derivatives 229

7.4 Deep cavitands 230

7.5 Conclusions 232

Acknowledgments 235

References 235

8. Luminescence probing of surface adsorption processesusing InGaN/GaN nanowire heterostructure arrays 239

Konrad Maier, Andreas Helwig, Gerhard M€uller and Martin Eickhoff

8.1 Adsorptiondkey to understanding semiconductor gas sensors 239

8.2 III-nitrides as an emerging semiconductor technology 243

8.3 Photoluminescent InGaN/GaN nanowire arrays 243

8.4 Optical probing of adsorption processes 245

8.5 Experimental observations of PL response 246

8.6 Analysis of adsorption phenomena 250

8.7 Molecular mechanism of adsorption 261

8.8 Conclusions and outlook 266

References 267

9. Rare earth–doped oxide materials for photoluminescence-basedgas sensors 271

V. Kiisk and Raivo Jaaniso


9.2 Sm3þ:TiO2 277

9.3 Eu3þ:ZrO2 288

Contents vii

9.4 Tb3þ:CePO4 294

9.5 Pr3þ:(K0.5Na0.5)NbO3 298

9.6 Conclusion 299

References 300

Part Three Methods and integration

10. Recent progress in silicon carbide field effect gas sensors 309

M. Andersson, A. Lloyd Spetz and D. Puglisi


10.2 Background: transduction and sensing mechanisms 312

10.3 Sensing layer development for improved selectivity of SiC gas sensors 327

10.4 Dynamic sensor operation and advanced data evaluation 332

10.5 Applications 335

10.6 Summary 338

Acknowledgments 217

References 339

11. Semiconducting direct thermoelectric gas sensors 347

F. Rettig and R. Moos


11.2 Direct thermoelectric gas sensors 353


References 381

12. Dynamic operation of semiconductor sensors 385

Andreas Sch€utze and Tilman Sauerwald


12.2 Dynamic operation of metal oxide semiconductor gas sensors 388

12.3 Dynamic operation of gas-sensitive field-effect transistors 398

12.4 Conclusion and outlook 404

References 408

13. Micromachined semiconductor gas sensors 413

D. Briand and J. Courbat


13.2 A brief history of semiconductors as gas-sensitive devices 414

13.3 Microhotplate concept and technologies 416

13.4 Micromachined metal oxide gas sensors 425

viii Contents

13.5 Complementary metal oxide semiconductor–compatible

metal oxide gas sensors 437

13.6 Micromachined field-effect gas sensors 442

13.7 Nanostructured gas sensing layers on microhotplates 445

13.8 Semiconductor gas sensors on polymeric foil and their additive

manufacturing 450

13.9 Manufacturing, products, and applications 454

13.10 Conclusion 458

References 459

14. Integrated CMOS-based sensors for gas and odor detection 465

P.K. Guha, S. Santra and J.W. Gardner


14.2 Microresistive complementary metal oxide semiconductor

gas sensors 467

14.3 Microcalorimetric complementary metal oxide semiconductor

gas sensor 469

14.4 Sensing materials and their deposition on complementary

metal oxide semiconductor gas sensors 472

14.5 Interface circuitry and its integration 475

14.6 Integrated multisensor and sensor array systems 480


Useful web addresses 485

References 486

Index 489

Contents ix


Contributors

M. AnderssonLink€oping University, Link€oping, Sweden

N. BarsanUniversity of T€ubingen, T€ubingen, Germany

D. BriandEcole Polytechnique Fédérale de Lausanne, Neuchatel, Switzerland

Pierrick ClémentMicrosystems Laboratory, �Ecole Polytechnique Féderale de Lausanne (EPFL), Lausanne,Switzerland

E. CominiDepartment of Information Engineering, University of Brescia, Brescia, Italy

J. CourbatFormely Ecole Polytechnique Fédérale de Lausanne, Neuchatel, Switzerland

Arnaldo D’AmicoDepartment of Electronic Engineering, University of Rome Tor Vergata, Roma, Italy

Corrado Di NataleDepartment of Electronic Engineering, University of Rome Tor Vergata, Roma, Italy

Martin EickhoffInstitute of Solid State Physics, University of Bremen, Bremen, Germany

J.W. GardnerUniversity of Warwick, Coventry, United Kingdom

P.K. GuhaIndian Institute of Technology, Kharagpur, West Bengal, India

Andreas HelwigAirbus Group Innovations, Munich, Germany

M. HuebnerUniversity of T€ubingen, T€ubingen, Germany

Raivo JaanisoUniversity of Tartu, Tartu, Estonia

Teerakiat KerdcharoenDepartment of Physics and NANOTEC Center of Excellence, Faculty of Science, MahidolUniversity, Ratchathewi, Bangkok, Thailand

V. KiiskUniversity of Tartu, Tartu, Estonia

Sung Pil LeeKyungnam University, Changwon, Kyungnam, Korea

xi j

Eduard LlobetMINOS-EMaS, Department of Electronic Engineering, Universitat Rovira i Virgili,Tarragona, Spain

A. Lloyd SpetzLink€oping University, Link€oping, Sweden

Konrad MaierAirbus Group Innovations, Munich, Germany

R. MoosUniversity of Bayreuth, Bayreuth, Germany

Gerhard M€ullerDepartment of Applied Sciences and Mechatronics, Munich University of AppliedSciences, Munich, Germany

D. PuglisiLink€oping University, Link€oping, Sweden

F. RettigUniversity of Bayreuth, Bayreuth, Germany

S. SantraIndian Institute of Technology, Kharagpur, West Bengal, India

Tilman SauerwaldLab for Measurement Technology, Department Systems Engineering, Saarland University,Saarbr€ucken, Germany

Andreas Sch€utzeLab for Measurement Technology, Department Systems Engineering, Saarland University,Saarbr€ucken, Germany

Thara SeesaardDepartment of Physics, Faculty of Science and Technology, Kanchanaburi RajabhatUniversity, Muang District, Kanchanaburi, Thailand

Chowdhury ShaestagirIntel Corporation, Hillsboro, OR, United States

Kengo ShimanoeFaculty of Engineering Sciences, Kyushu University, Kasuga, Fukuoka, Japan

U. WeimarUniversity of T€ubingen, T€ubingen, Germany

Chatchawal WongchoosukDepartment of Physics, Faculty of Science, Kasetsart University, Chatuchak, Bangkok,Thailand

Noboru YamazoeFaculty of Engineering Sciences, Kyushu University, Kasuga, Fukuoka, Japan

D. ZappaDepartment of Information Engineering, University of Brescia, Brescia, Italy

xii Integrated CMOS-based sensors for gas and odor detectionContributors

PART ONE

Basics

1j


CHAPTER ONE

Fundamentals of semiconductorgas sensorsNoboru Yamazoe, Kengo ShimanoeFaculty of Engineering Sciences, Kyushu University, Kasuga, Fukuoka, Japan

Contents

1.1 Introduction 41.2 Classification of semiconductor gas sensors 51.3 Resistor-type sensors: empirical aspects 6

1.3.1 Sensing materials and devices 61.3.1.1 Sensing materials 61.3.1.2 Sensitizers 81.3.1.3 Device structure 91.3.1.4 Fabrication 10

1.3.2 Gas sensing characteristics 111.3.2.1 Response and response transients 111.3.2.2 Operating temperature 121.3.2.3 Disturbances to gas response 13

1.3.3 Semiconductor oxygen sensors 131.4 Resistor-type sensors: theoretical aspects 14

1.4.1 Receptor function and transducer function 141.4.2 Response to oxygen (base air resistance) 181.4.3 Response to inflammable gases 221.4.4 Response to oxidizing gases 231.4.5 Extensions 251.4.6 Nonresistive sensors 271.4.7 Field-effect transistor-type gas sensors 27

1.4.7.1 Principle 271.4.7.2 Solid electrolyte-gate field-effect transistor 281.4.7.3 Oxide semiconductor-gate field-effect transistor 291.4.7.4 Dielectric material-gate field-effect transistor 31

1.4.8 Oxygen concentration cell type sensors 311.4.9 Other gas sensors 32

1.4.9.1 Metaleinsulatoresemiconductor capacitor type sensors 321.4.9.2 Diode-type sensors 33

Semiconductor Gas Sensors, Second EditionISBN: 978-0-08-102559-8https://doi.org/10.1016/B978-0-08-102559-8.00001-X

© 2020 Elsevier Ltd.All rights reserved. 3 j

https://doi.org/10.1016/B978-0-08-102559-8.00001-X

1.5 Future trends 341.5.1 Needs and seeds 341.5.2 Basic approaches desired 351.5.3 Challenges 36

References 37

1.1 Introduction

Semiconductor gas sensors using metal oxides such as SnO2 werepioneered by two research groups in Japan.1,2 These sensors were soonput on the market as gas leak alarms and proved to be indispensable in keep-ing people safe from the distressing circumstances resulting from gas leaks. Atthe same time, their success had worldwide impact on researchers, creatingawareness of the importance of gas sensors or chemical sensors moregenerally. Great effort has subsequently been made in the development ofnew gas sensors, including those using silicon semiconductor devices andsolid electrolytes devices. If the definition of a semiconductor gas sensor isa sensor into which a semiconductor material is incorporated, there is avariety of semiconductor gas sensors of varying structures, made of differentmaterials and involving various working principles.

This introduction describes the fundamental aspects of the varioussemiconductor gas sensors that have been developed so far, or that are pro-posed. First, they are classified into five types, based on the constitutionalprinciple of sensor devices (Section 1.2). The structure of devices, theirworking principles, and sensing mechanisms are described in subsequentsections. However, the greatest space is devoted to describing experimentalknowledge and the theory of gas response of the sensors based on resistors,which have been made full use of and which still have potential for furtherdevelopment. It has long been queried why sensors of this type arepromoted with regard to their sensitivity, as the constituent oxides aresmaller than in other types of device,3 though a semiempirical analysis hasbeen attempted.4,5 This issue was recently resolved by developing a newtheory on the receptor function of small-sized oxides.6,7 As revealed inthe new theory, small semiconductors are depleted of electrons in two stagesby a process of ionosorption of oxygen or oxidizing gases, resulting in theappearance of regional depletion followed by volume depletion. Gasresponse can be sufficiently understood based on the same theory. It isshown that the theory gives an important clue to understanding the gas

4 Noboru Yamazoe and Kengo Shimanoe

response of oxides attached to potentiometric gas sensors (Section 1.5). Thechapter closes with personal observations regarding semiconductor gassensors (Section 1.6).

1.2 Classification of semiconductor gas sensors

Generally speaking, a gas sensor is composed of a receptor and a trans-ducer, as illustrated in Fig. 1.1. The former is provided with a material or amaterials system which, on interacting with a target gas, either induces achange in its own properties (work function, dielectric constant, electrodepotential, mass, etc.) or emits heat or light. The transducer is a device totransform such an effect into an electrical signal (sensor response). Theconstruction of a sensor is determined by the transducer used, with thereceptor appearing to be implanted within it. From this perspective, a semi-conductor gas sensor can be defined as a sensor in which a semiconductormaterial is used as a receptor and/or transducer.

There are two groups of semiconductors: oxide and nonoxide (typically,silicon). Nonoxide semiconductors cannot work as a receptor because theyare coated with a protective insulation layer, but they can provide a trans-ducer in the form of MIS FETs (metaleinsulatoresemiconductor field-effect transistor) and MIS capacitors. In contrast, oxide semiconductorscan work as both a receptor and a transducer (mostly in the form of a resistor)

Figure 1.1 Gas sensor as constituted of a receptor and a transducer. R ¼ resistance,E ¼ electromotive force, I ¼ current, Vth ¼ threshold voltage (FET), Cp ¼ capacitance.

Fundamentals of semiconductor gas sensors 5

owing to their chemical and physical stability in hostile environments atelevated temperatures.

Table 1.1 shows various examples of semiconductor gas sensors classifiedaccording to the types of transducer used and subclassified by the kinds ofreceptor used, together with the kinds of signal output (response), typicalsensor devices, and the gases targeted. The transducers are seen to beavailable in the forms of resistors, diodes, MIS capacitors, MIS FETs, oroxygen concentration cells. For each type of sensor thus classified, devices,sensing principles, and the important features of semiconductor gas sensorsare now described.

1.3 Resistor-type sensors: empirical aspects

Of the various types of sensor, resistor sensors have received the great-est investigation and have proven their feasibility in practice. These sensorsare often called “oxide semiconductor gas sensors.” There are two subtypes:surface sensitive and bulk sensitive. This section is devoted to surface-sensitive resistor sensors, except for Section 1.3.3 which briefly discussesbulk sensitive resistor sensors. It is noted that books and review articleshave been published about oxide semiconductor gas sensors.8e10

1.3.1 Sensing materials and devices1.3.1.1 Sensing materialsA surface-sensitive resistor sensor works on a very simple principle; onexposure to a target gas in air at an elevated temperature, its resistance eitherdecreases or increases as a function of the partial pressure of the gas. Of themany metal oxides, n-type oxides (SnO2, In2O3,WO3, ZnO, and g-Fe2O3)and p-type oxides (CuO and Co3O4) exhibit significant gas sensingproperties. Mainly because of stability issues, however, SnO2, In2O3, andWO3 have been adopted as the sensor materials utilized in practice. Inpractice, even these oxides are frequently loaded or mixed empiricallywith several foreign materials as a sensitizer (PdO, Pt, Fe2O3, etc), a skeletonmaterial (alumina), or a binder (silica).

When an n-type oxide is used, resistance decreases on exposure toinflammable or reducing gases in the air (inorganic: H2, CO, NH3, H2S,NO, etc; organic: CH4, propane, alcohols, odorants, etc.), while it increaseson exposure to oxidative gases (NO2, ozone, N2O, etc.). Apart from suchredox-active gases, CO2 and water vapor have been known to affect theresistance to a greater or lesser degree. Exploitation of the effects of CO2

has led to the development of a semiconductor CO2 sensor.11


Table 1.1 Classification of semiconductor gas sensors according to the types of transducers and receptors used.Transducer Response signal Receptor Device (example) Target

Resistor Resistance Oxides Porous SnO2 (surface-sensitive)

A variety of gases

Sintered TiO2 (bulk-sensitive)

Air/fuel ratio (carengine)

Diode Bias current Oxides Pd-TiO2 (single crystal) H2

Metaleinsulatoresemiconductor(MIS) capacitor

Bias potential shift Pd Pd-gate capacitor H2, NH3

MIS field-effecttransistor (FET)

Threshold voltage shift Pd Pd-gate FET H2, NH3

Ionic Proton H2

conductors conductor gate FETNaNO2-gate FET NO2

Oxides WO3-gate FET NO2

Dielectrics Cellulose-gate FET HumidityOxygen concentrationcell

Cell voltage Oxides Pt/zirconia/oxide/Pt A variety of gases

Note: “Oxides” stands for semiconductive metal oxides.

Fundamentals

ofsem

iconductorgas

sensors7

1.3.1.2 SensitizersGas sensing properties, especially gas responses, are known to be oftenimproved significantly when constituent oxides are loaded with smallamounts of appropriately chosen foreign materials. Examples are SnO2-PdO (CO, propane, etc.), SnO2-Pt and/or PdO (methane), SnO2-Co3O4

(CO), SnO2-CuO (H2S), SnO2-Ag2O (H2), In2O3-PdO (CO, odorantgases), WO3-Au (NH3), SnO2-La2O3-Pt (ethanol), SnO2-CaO (ethanol),In2O3-Fe2O3 (ozone), SnO2-Fe2O3 (NO2), TiO2-Cr2O3 (NO), etc. Inthis list, the materials following the oxide semiconductors are sensitizersand the target gases are shown in parentheses. As suggested from the largevariation in sensitizers, the mechanisms of sensitization involved are not sosimple.

It is useful to know that the dispersion of the sensitizers, except Pt, al-ways causes the resistances of the device in base air to increase. This suggeststhat those interact with the oxides and increase the work function of theoxides. In view of heterogeneous catalysis, Pt, PdO, CuO, Ag2O,Co3O4, and Au are well-known oxidation catalysts to reducing gases.Therefore, such catalytic activity is relevant to the sensitizing actions. Itshould be noted, however, that the mere promotion of oxidation reactionscannot contribute to gas response unless it has something to do with thesurface properties of the oxides. In this sense, the sensitizers, except Pt,undergo redox changes such as PdO þ H2 / Pd þ H2O, Pd þ (1/2)O2 / PdO, and the changes of their redox state on exposure to targetgases can possibly induce changes in device resistance (gas response)through electronic interactions with oxides (electronic sensitization). Inthe case of Pt, on the other hand, it seems that the target gas (methane)is partially oxidized on Pt to HCHO or CO, which then reacts with theadsorbed oxygen of the oxide (chemical sensitization).

La2O3 and CaO, which have no such catalytic oxidation activity, modifythe acid-base properties of the oxide surface more basic; on the acidicsurface, ethanol undergoes dehydration (no consumption of O�),C2H5OH / C2H4 þ H2O; on the basic surface, it undergoes oxidativedehydrogenation, C2H5OH þ 2 O� / C2H4O þ H2O. It is thus under-stood that, in this case, the selectivity of reaction paths is changed by thesensitizers. As shown above, Fe2O3 promotes response to oxidizing gases,though the mechanism of promotion is not yet clear.

There can be no doubt that sensitizers are very important for practicaldevices. Unfortunately, however, little basic research has been carried outon sensitizers and sensitizing actions.


1.3.1.3 Device structureSensor devices are fabricated into a resistor in which a porous stack of thesensing materials is attached with a heater and a resistance measuring probe(usually a pair of metal electrodes). Various structures have been devised inpractice, as shown in Fig. 1.2. Originally fabrication was a sintered blockstructure (about 0.5 cm in size) with a pair of Pt coil electrodes inserted(a); one of the coils also served as a heater. This was followed by a thinalumina tube within a heavy coating (b); a pair of wire electrodes was woundon the tube and a heater was set inside it. Currently in wide use is a thick filmstructure (c), screen-printed on an alumina substrate with a pair of elec-trodes, and a heater printed in advance. A microversion of this structure,

Figure 1.2 Device structures adopted for resistor-type sensors in practice. (a) Sinteredblock, (b) thin alumina tube-coated layer, (c) screen printed thick film, (d) small beadinserted with coil and needle electrodes, (e) small bead inserted with a single coil(heater and electrode), (f) practical sensor element assembling sensor device, metalcap, and filter.


known as a MEMS (microelectromechanical system) sensor, is currentlyunder development, as will be described later. Apart from these standardstructures, bead-shaped structures have been devised for practical use. Asmall bead made of sensing materials (about 0.5 mm in size) is insertedwith a coil and needle electrodes in (d); the coil also works as a heater. Asimilar bead is inserted with a single coil (heater) in (e), the so-called “hotwire” type; a change in the resistance of the sensing materials affects thecomposite resistance between the two terminals of the inserted coil, whichis measured precisely on a bridge circuit as gas response.

For actual use, each device is bonded to the connector pins and putinside a metal cap with a hole(s) on top to remove the risk of triggeringgas explosions. In addition, an adsorbent such as active carbon (often referredto as a “filter”) is placed in a layer immediately behind the hole to removeunpleasant gases, as shown in (f).

1.3.1.4 FabricationImportant guidelines for device fabrication collected empirically can besummarized as follows:1. Crystallite sizes of oxide semiconductors should be as small as possible.2. Sensitizers should be dispersed as finely as possible.3. Sensing layer thickness and porosity should also be optimized to improve

selectivity and durability.According to these guidelines, fabrication of devices is carried out care-

fully. It starts with the preparation of a fine powder of oxide semiconductor(crystallite size around 10 nm in diameter) through what is known as a “wet”process. This is the precipitation of a precursor of the oxide from an aqueoussolution of its metal salt(s), followed by the gentle washing, drying, andcalcination of the precursor before its conversion to the final powder. Thepowder is loaded with a small amount of a sensitizer and then convertedinto slurry (paste) by milling it using water or organic vehicles, togetherwith any other necessary additives. The slurry is finally deposited over theelectrodes (block or bead type) or on the substrate (thick film type), and,after drying, the deposit is sintered under specific conditions to stabilizethe porous microstructure.

It is noted that all of the semiconductor gas sensors so far in use are of thethick film (or layer) type, prepared through the wet processes discussedabove. Thin film type devices, especially those fabricated via physicalmethods such as sputtering, have frequently shown interesting sensing per-formances in the short term, but little use is currently made of these devices.


1.3.2 Gas sensing characteristics1.3.2.1 Response and response transientsThe behavior of resistance on switching between base air and gas ambient isillustrated in Fig. 1.3(a). On switching to an inflammable gas ambient, theresistance reduces from a value in air (Ra) to a stationary value (Rg), whileit goes back to Ra on switching back. Empirically, gas response is definedas the ratio Ra/Rg (normalized conductance). The rate of response orrecovery is expressed empirically in terms of the time (s) needed for a90% full response or recovery. In the case of oxidizing gases such as NO2,which increase the resistance, gas response is defined as Rg/Ra (normalizedresistance).

The dependence of Rg on the partial pressure of target gas (Pg) isknown empirically to fall on linear correlations on logarithmic scales12;that is, Rg ¼ cPa

g , where a and c are constants (power law), as shown in

Fig. 1.3(b). Accordingly, gas response also follows power law,Ra=Rg ¼ cPa

g (inflammable gases) or Rg=Ra ¼ cPag (oxidizing gases).

Figure 1.3 Response and recovery transients. (a) On switching on and off an inflam-mable gas in air, (b) linear correlation observed between resistance (Rg) and partialpressure of the gas (Pg) on logarithmic scales (power law).


The power index, a, is almost fixed depending on the kinds of target gas,taking values roughly equal to 1/2 to many inflammable gases (H2, CO,etc), 1 to NO2, and 1/2 for O3. It is noted that the resistance under exposureto varying partial pressure of oxygen (PO2) follows the power equation with

a ¼ 1=2, namely, RO2 ¼ c'''P1=2O2

. The power indices are related to themodes of interaction between the gases and the surface of oxide semi-conductors, as will be discussed later.

Sensitivity is usually defined as a slope of the correlation betweengas response and Pg. In the event that power law holds well, however,this definition is meaningless because sensitivity is dependent on Pg unlessa is unity. This difficulty is overcome if Pg is replaced by Pa

g in the abovedefinition. The slope (sensitivity) is then nothing but the proportionalityconstant of the power equation. Sensitivity is determined by the physico-chemical constants of semiconductor, target gas, and oxygen.

1.3.2.2 Operating temperatureResponse and response transients are sensitive to the operating temperature.The rates of response and recovery naturally increase with increasing tem-perature. On the other hand, response shows different behavior dependingon whether the gas is inflammable or oxidizing. For an inflammable gas,response goes through a maximum on increasing temperature, resultingin a well-known bell-shaped correlation between the response and temper-ature. This dependence appears because the rate constant of the surfacereaction between gas and adsorbed oxygen (kR) increases exponentiallywith a rise in temperature, while the Knudsen diffusion coefficient of thegas (DK) does so sublinearly. In the lower temperature region, kR < DK isheld so that kR is an exclusive determinant for gas response. In highertemperatures, on the other hand, the relation is inversed, kR > DK, andthe response is attenuated by the gas diffusion and reaction effect.13,14 Inthis temperature range, the gas is consumed significantly by diffusion fromthe surface to the inside of the porous sensing layer. The effective partialpressure of the gas in the inner region where the resistance is actuallymeasured can be significantly lower than the nominal value outside. Theratio of the actual gas response to the ideal (free of attenuation) is knownas the “utility factor” (U). U remains unity in lower temperatures, whilein higher temperatures it decreases rather sharply with increasing tempera-ture, increasing diffusion length (sensing layer thickness), and decreasingpore size. It follows that the response maximum and the temperature atthat point vary not only by the kinds of gas and oxide semiconductor


but also by the device structure (layer thickness, in particular) and the sensingmaterials adopted in processing. Strictly speaking, there is a further possiblereason for the decrease of gas response at high temperature: oxygenadsorption is decreased with increasing temperature. Therefore, if the partialpressure of the inflammable gas is too great, adsorbed oxygen is consumed(resistance reaches minimum) such that gas response will decrease withincreasing temperature, reflecting the temperature dependence of theadsorbed oxygen. This discussion is valid for a small partial pressure of gas.

Oxidizing gases such as NO2, on the other hand, are adsorbed on theoxide semiconductor particles. The amount of adsorption, and thereforethe gas response, increases as the temperature drops. Operating temperatureis then determined as a compromise between gas response and rates ofresponse and recovery.

1.3.2.3 Disturbances to gas responseGas response reacts to disturbances to varying degrees. There are two kindsof disturbance: a drift of base air resistance (Ra) and a modulation of gasresponse (Rg) by coexistent gases. As for the former, Ra shifts downwardquickly on increasing the partial pressure of coexistent water vapor(PH2O), a phenomenon known as a “short-term effect of water vapor.”Apart from this phenomenon, PH2O seems to be related to a long-term driftof Ra; it is known that Ra undergoes seasonal changes; that is, it goes up insummer and goes down in winter. Unfortunately, these two types of driftsare yet to be clarified in detail. Practically, attempts have been made tocorrect the long-term drift partly by means of software.

The disturbance brought about by a modulation of gas response can besimplified if both the target gas and coexistent gas are inflammable, as in thecase of sensing CO in the coexistence of H2. The strength of the disturbancecan be estimated if the sensitivity to each gas is known. To mitigate interfer-ences by coexistent gases, nonstandard modes of sensor operation have beenadopted in some cases for sensing CO and alcohol in the breath.

1.3.3 Semiconductor oxygen sensorsAt sufficiently high temperatures, where the bulk diffusion of componention oxides is activated to a significant degree, oxide semiconductors areknown to change nonstoichiometry, and thus electronic conductivitychanges depending on PO2. On exposure to a mixture of inflammable gasand air, sensors using such oxides change resistance depending on thecomposition of the mixture. What is responsible for the change in resistance


is not the reducing gas itself but PO2 in the ambient after the reducing gas hasbeen oxidized completely. Resistor-type oxygen sensors working on thisprinciple have been proposed by using oxides such as TiO2, Nb2O5, andMgO-CoO. Among them, one using TiO2 has been successfully incorpo-rated into car engine exhaust control systems in practice. The sensor,fabricated into a well-sintered block of TiO2 with a pair of electrodesinserted, is exposed to car engine exhausts at high temperature(e.g., 1073K). As the resistance decreases or increases stepwise as air/fuel(A/F) ratio crosses the border between lean burn and rich burn, it can beutilized for A/F ratio control. Its share in the market is somewhat small,however, compared with that of its competitor, zirconia oxygen sensors.

1.4 Resistor-type sensors: theoretical aspects

For resistive-type gas sensors, a porous assembly of fine particles(mostly grains) of oxide semiconductors should function as a receptor anda transducer. It has long been accepted that grains act as a receptor to gases,while the contacts between the grains act as the transducer which transformsthe gas reception into a change in device resistance. However, an under-standing of the receptor function and the transducer function involvedhad remained far from being satisfactory until basic approaches to thembegan very recently. This section focuses on recent advances in the basic(theoretical) approaches, though the studies are still in progress.

1.4.1 Receptor function and transducer functionOxide semiconductors are known to exhibit unique interactions with somesorts of gases, resulting in the ionosorption of the gases. In the event that thegas in a problem situation has a large electron affinity, such as O2 and NO2,the host semiconductor supplies electrons to the gas to allow it to beadsorbed as anionic species such as O�, O2� or NO2

�. In the event thatthe gas is low in ionization potential, such as NO, on the other hand, thegas donates electrons to the semiconductor to be adsorbed as cationic spe-cies, such as NOþ. The electrons supplied or given up in these ionosorptionprocesses are transferred from the bulk of the semiconductor to the surface,or vice versa, accompanied by a change in energy band structure (bandbending) of the semiconductor. It is well-known that electron transferfrom the bulk of n-type semiconductor results in the formation of anelectron-depleted layer in the semiconductor. No doubt, an oxide semicon-ductor sensor, when placed in air, is subjected to the adsorption


(ionosorption) of oxygen, and its resistance in air (air base) is determinedusually from the equilibrium of oxygen adsorption. As very recentlyrevealed with SnO2 sensors, oxygen is adsorbed mainly in the form ofO2� in extremely dry air, whereas in the presence of low humidity (0.1%in volume and above), the adsorption in that form is suppressed almostcompletely by water vapor to be replaced by the adsorption in anotherform (O�). In practice, it can thus be assumed as a good approximationthat the latter form (O�) prevails over the former (O2�) under usual sensoroperating conditions. The sensor is utilized for detecting a target gascoexistent in air by means of a change in the resistance of the device. Targetgases fall into two groups: gases which undergo ionosorption (such as NO2)and inflammable gases (such as H2, CO, and C3H8). In cases whereionosorption takes place in addition to that of oxygen, the energy bandstructure changes accordingly. Usually, however, serious interference oftenoccurs between the ionosorption of the gas and that of oxygen, reducing theresultant change in energy band structure. The key to designing a sensorsensitive to such a gas is discovering how to mitigate such interference.Inflammable gases react with the anionic adsorbates of oxygen. As a resultof the reaction, electrons of the adsorbates are returned to the semi-conductor, causing the energy band structure to revert to one thatcorresponds to smaller amounts of oxygen adsorbates. Obviously, responseto a gas in this group will be enhanced as the consumption of the oxygenadsorbates is made more efficient.

Here, it is of central importance to show how the qualitative understand-ing mentioned above can be converted into more quantitative ones. Forsimplicity, let us assume that a sensor device is a porous stack of uniformgrains of an n-type oxide semiconductor. It is accepted that each grain playsthe role of a receptor, while that of the transducer is played by each contactbetween grains; that is the most resistive part in the device, so it determinesthe resistance of the whole device. However, further understanding has beenless than straightforward. For some considerable time, efforts were made tounderstand the receptor and the transducer functions based on the surfacespace charge layer model and the double Schottky barrier model, as shownby (a) and (b) in Fig. 1.4, respectively. These models, (a) and (b), wereguessed at by many researchers as analogies from a metal semiconductorcontact diode (see, for instance, Ref. 9). It was assumed that the thicknessof the depletion layer (w) should increase as oxygen adsorption as anionicspecies (typically O�) increases, while it should decrease as the adsorbedoxygen is consumed with an inflammable gas (H2). Correspondingly, the


double Schottky barrier formed across the contact between grains shouldchange its height, inducing changes in contact resistance and, hence,resistance in the device. Unfortunately, these models were unable to givequantitative information regarding gas response. Shortcomings of the modelswere made clear recently by our basic approaches, as described below.

The receptor model (a) assumes implicitly that the semiconductor grainsare sufficiently large. In reality, however, they are very tiny (typically about10 nm in diameter), so the space charge layer can easily extend over theentire area of grains; that is, w grows to grain radius (a), at PO2 significantlybelow that in air, PO2 (a). Obviously, a new process of electron depletion hasto take place afterward until the grains reach electrostatic equilibriumwith oxygen adsorption at PO2(a). A method proposed here is one in whichelectron depletion is achieved by shifting the Fermi level downward by pkT, as shown in Fig. 1.5.6,7 Here, p is the Fermi level shift as expressed inthe unit of kT, where kT is thermal energy. The electrons supplied to theadsorbates in this stage are squeezed out of the grains by increasing p. Todistinguish the electron depletion of this type (accompanied by a changein p) from the conventional type one (accompanied by a change in w), theseare denoted as volume depletion and regional depletion, respectively. Thevalue of p or w is determined uniquely for given conditions of gas adsorptionand semiconductor grains. Importantly, p or w depends on a when the

Figure 1.4 Diagrams of electron depletion for oxide grains and the resistance of contactbetween grains. (a) Space charge layer model, (b) double Schottky barrier model,(c) regional and volume depletion model, (d) surface conductive grains contact model.


conditions are otherwise fixed. As shown in Fig. 1.4(c), small oxides areusually in a state of regional depletion at low PO2; while those that areusually in a state of volume depletion in base air (the whole area beingdepleted) and their electronic states are controlled by p. It is noted, however,that more rigorous discussion should be extended in terms of reduced radius(n) rather than of radius (a), as discussed later.

The double Schottky barrier model (Fig. 1.4(b)) also turned out to becompletely misleading. It focused attention on the electron transport pathrunning through the centers of contacting grains. In reality, however, thereare a tremendous number of other transport paths running on the surface ofgrains, which are free of potential barriers, as shown in Fig. 1.4(d). Theelectron transport through the contact can thus be achieved by migrationor tunneling of the surface electrons, indifferent to the bulk electrons inside.The contact resistance and the device resistance (R) are then inverselyproportional to the surface density of electrons, [e]S, as long as the grainsare uniform. Device resistance (R) as normalized by that at flat band state

Ec EcPO2 = 0

PO2(I) PO2(I)

PO2(II)

PO2(II)

PO2(III)

PO2(I) PO2(I)

PO2(II)

PO2(II)

PO2(III)

PO2(III)

PO2(III)

O–(I) O–(I)

O–(II)O–(II)

O–(III)

O–(III)

qV(r

)

qV(r

)

p(III)kT)

p(III)kT)

p(II)kT)

0–a –a/2 a/2ar

0r

–a a0

Nd

n

0

Nd

n

00–a/2 a/2

r

(a)

(c) (d)

(b)

r

Figure 1.5 Energy band diagrams: (a) and (b) distributions of conduction electrons;(c) and (d) for two kinds of grains different in radius (a or a/2) at steps of increasing PO2.


(R0), called “reduced resistance,” is expressed by using the donor density ofsemiconductor (ND) as follows:

RR0

¼ ND��e�

�S (1.1)

1.4.2 Response to oxygen (base air resistance)Let us consider a case where oxygen is adsorbed as O� on an oxide grain ofradius a. The adsorption equilibrium is written as follows:

O2þ 2e� ¼ 2O�0KO2PO2

�e�

�2S ¼

�O��2 (1.2)

Here, KO2 is the adsorption constant and [O�] the surface concentration ofO�. Note that [e]S is a variable of the grain. At the same time, we have toconsider the electrostatic equilibrium of the grain. Assuming that there is nosurface state other than O�, [e�]S and [O

�] can be expressed as a function ofp, respectively, for volume depletion as follows:15

�O��¼ �QSC

q¼ NDLD

n�n3

�� AðnÞexpð� pÞ

o(1.3)

½e� �S ¼ ND exp

��16

n2 � p

(1.4)

Here, QSC is the total surface charge density of the grain, which isassumed to be ascribed solely to [O�] in this case. q is the elementary chargeof proton. LD is the Debye length defined as LD ¼ (εkT/q2ND)

1/2, where εis permittivity, and n is reduced radius defined as n ¼ a/LD. A(n) stands forthe number of free electrons remaining in the conduction band at p ¼ 0 asnormalized by NDLD and the surface area of the grain. Assuming Boltz-mann’s distribution law for the tailing of electrons, it is given by thefollowing integral:

AðnÞ ¼�1n2

Z n

oR2exp

��16

R2

dR

There are three simultaneous equations, Eqs. (1.2)e(1.4), correlatingamong three variables, [e�]S, [O

�], and p. It is thus possible to determineeach variable as a function of KO2PO2. The solution for [e�]S is transformedinto normalized resistance through Eq. (1.1).

ND

½e��S¼ RR0

¼ cðnÞ þ�Sa

ðKO2PO2Þ1=2 (1.5)


S is the shape factor for the semiconductor crystals used; i.e., S ¼ 3for spheres, 2 for columns, and 1 for plates. Constant c(n) is given byc(n) ¼ (3/n) exp (n2/6) A(n); it increases from unity as n increases; first,gradually when n is small and then exponentially afterward.

The correlations given by Eq. (1.5) are illustrated in Fig. 1.6, wherereduced resistance (R/R0) is related to ðKO2 PO2Þ1=2 for variously sizedgrains (LD is assumed to be 3 nm). The linear correlations coincide withthe power index (1/2) to PO2, as previously mentioned. Its slope is givenby (3/a), indicating that sensitivity to oxygen increases as a decreases. Asalso indicated in Fig. 1.6, the correlation is bent in the initial region ofPO2 for larger grains where regional depletion takes place. Remarkably, itcan be shown that R/R0 is almost independent of a in the regional area.Such correlations have, in fact, been confirmed experimentally. It is alsonoted that, under a particular condition, oxygen adsorption to form anotherspecies (O2�) also takes place, which is demonstrated by the lineardependence of R/R0 on P1=4

O2in the stage of volume depletion.

Notably, the sensor response is related to the kind and amount of oxygenspecies adsorbed on the surface of the metal oxide semiconductors. Theadsorption equilibrium for O� and O2� can be discussed as follows,respectively.

O� formation: O2 þ 2e� ¼ 2O� (R1)

ðK1PO2Þ1=2½e�S ¼�O�� (1.6)

Figure 1.6 Reduced resistance (R/R0) as correlated with (KO2PO2)1/2/LD for devices using

oxide grains different in reduced size (n).


O2� formation: O2 þ 4e� ¼ 2O2� (R3)

ðK2PO2Þ1=2½e�2S ¼�O2�� (1.7)

Here, K1 and K2 are the oxygen adsorption equilibrium constants of O�

and O2�, and [O�] and [O2�] are the concentrations of O� and O2� ions,respectively. In this case, the relationship between the oxygen partial pres-sure and electric resistance is explained using the following Eq. (1.8):16

RR0

¼ 12

�c þ 3

aðK1Þ1=2$P1=2

O2

þ(14

�c þ 3

aðK1Þ1=2$P1=2

O2

2

þ�6ND

a

ðK2Þ1=2$P1=2

O2

)1=2

(1.8)

Here, c is a constant. The equilibrium constants K1 and K2 indicate theoxygen adsorption ability on the metal oxide surface as O� and O2�. On thebase of volume depletion, the relationship between the oxygen adsorptionand the electric resistance was investigated. For SnO2, the oxygen adsorp-tion species in dry and wet atmospheres using the relationship betweenthe electric resistance and oxygen partial pressure (PO2) was reported.

17,18

In short, O2� and O� adsorb on the surface in dry and wet atmospheres,respectively. In addition, the amount of O� was decreased remarkably byadsorption of OH group, resulting that it brings about the deteriorationon gas response. Fig. 1.7 shows the relationships between the electricresistance and oxygen partial pressure in dry and wet atmospheres on neatSnO2 (14 nm in diameter) operated at 300 and 350�C. In dry atmosphere,the operating temperature gives different relationship. At 350�C, the electricresistance is directly proportional to P1=4

O2. On the other hand, however, that

at 300�C does not show linearity to both P1=4O2

and P1=2O2

. This means thatboth adsorption species (O2� and O�) coexist on surface of SnO2. In wetatmosphere including 0.1 vol% water vapor, the electric resistances atboth temperatures decreased as compared with those in dry atmosphere,

but increased in proportion to P1=2O2

. The properties can be understood bymoisture effect. Moisture acted as an inhibitor to oxygen adsorption inform of O� and O2�, whereas the moisture admitted at elevated tempera-ture acted as a promoter to increase the adsorptive strength of O� sites.An increase of the O� site population as well as the existence of thresholdpressure for oxygen adsorption on the same sites suggests the formation of


a sort of surface hydrate, dehydration of which seems to leave O� sitesbehind. The response to oxygen can be understood satisfactorily by usingadsorption constant of oxygen, threshold pressure of oxygen, and semi-conductor properties of tin oxide in Eq. (1.8).

Oxygen adsorption species are different in materials such as receptorloading, surface modification, In2O3, and WO3. Table 1.2 shows oxygenadsorption species on each material at 350�C. Interestingly, oxygenadsorption species of Pd-loaded, Sb-doped, and Fe3þ or Zr4þ-modifiedSnO2 are O2� although the sensors are operated in wet atmosphere. Inaddition, oxygen on neat In2O3, which is slightly stronger in basicity thanSnO2, acts only as O2�. However, WO3 seems not to have an oxygenadsorption species because surface lattice oxygen is easily formed and activefor redox reaction.

1

0.5

1.5

00 0.2 0.4 0.6 0.8

0 0.2 0.4 0.6 0.8 1

PO21/2/atm 1/2

PO21/4/atm 1/4

PO21/2/atm 1/2

1

300°C

300°C

O2– and O–

350°C

350°C

O2–R

esis

tanc

e/10

5 Ω

Res

ista

nce/

104 Ω

1

0.5

00 0.2 0.4 0.6 0.8 1

(a)

(b)

Figure 1.7 Response of neat SnO2 device to oxygen in dry (a) and wet (b) atmospheresat 300 and 350�C.


1.4.3 Response to inflammable gasesSimple inflammable gases such as H2 and CO react with adsorbed oxygen(O�) in one step, while the supply for the O� consumed is the ambient.In a steady state, the following reactions proceed at an equal rate:

O2þ 2e�/2O� ðR1Þ H2 þO�/H2Oþ e� (R2)

When the rate of the reverse reaction of (R1) is negligible, the surfacedensity of O� at the steady state is expressed as follows:

�O�� ¼ �ðk1PO2ðaÞÞ

ðk2PH2Þ�

e��2S (1.9)

Here, PO2(a) and PH2 are partial pressures of oxygen in air and hydrogen,respectively, while k1 and k2 are the rate constants of (R1) and (R2),respectively. Eq. (1.9) is a constraint connecting [e�]S and [O�] in this case.Then, the equations for [e�]S, [O

�], and p can be solved as previouslyperformed. By using Eq. (1.1), reduced resistance under exposure to H2,Rg/R0 is derived for volume depletion as follows:

Rg

R0¼ cðnÞ=2þ

�cððnÞ=2Þ2 þ

�3ND

a

ðk1PO2ðaÞÞðk2PH2Þ

1=2

(1.10)

This equation is consistent with the power law (�1/2) to H2. Reducedresistance in air, Ra/R0, is obtained by substituting PO2(a) for PO2 inEq. (1.5). The conventional response to H2, Ra/Rg ¼ (Ra/R0)/(Rg/R0),can then be derived, which is expressed as follows when Ra/Rg >>1:

Ra

Rg¼

�Sðk2=k�1ÞðaNDÞ

1=2

P1=2H2

ðvolume depletionÞ (1.11)

S is the shape factor and equals 3 for spheres; ke1 is the rate constant ofthe reverse reaction of (R1), ke1 ¼ k1/KO2. The response is thus shown to be

Table 1.2 Oxygen adsorption species for each sensor material.

Sensor materialsAtmosphere

(dry or wet air) Temperature (350�C)

Neat SnO218) Dry O2�

Wet O�

Pd-loaded SnO219) Dry Wet O2�

Sb-doped SnO220) Dry Wet O2�

Fe3þ, Zr4þ-modified SnO221) Dry Wet O2�

Neat In2O322) Dry Wet O2�

WO323)

Pd-WO323)

Dry Wet Surface lattice oxygen(O2� adsorption, nonreactive)


linear to P1=2H2

, which accords with the experimental data as shown inFig. 1.9, where (Ra/Rg)

2 is correlated with PH2 instead. The proportionalityconstant (sensitivity) is promoted with increasing rate constant ratio (k2/ke1)and by decreasing grain radius (a) and donor density (ND). The effects ofgrain size can thus be rationalized theoretically. However, it should be notedthat there are many other inflammable gases which react with O� in morecomplex ways. Treatments of the responses to those gases have yet to beundertaken.

As mentioned in Section 1.4.2., two types of oxygen adsorption areobserved on surface of SnO2. In these cases, the response to inflammablegases can be understood as shown in Fig. 1.8. In the case of O� (case 1),the electric resistance of SnO2 element is low because of one electronreaction to oxygen atom (xa(O

�)). The electric resistance decreases byreaction of O� adsorbed on SnO2 to H2. In the case of O2� (case 2), theelectric resistance is high because it is the reaction of two electrons, so thechange in electric resistance is larger than that in the case 1. The reactionsin wet and dry atmospheres correspond to the case 1 and 2, respectively.

1.4.4 Response to oxidizing gasesLet NO2 be an example of an oxidizing gas. It is adsorbed on the grains toform NO2

� as follows:

NO2þ e� ¼ NO2� KNO2PNO2

�e�

�S ¼ ½NO2�� (1.12)

Reaction of O2– and H2 (case2)

Reaction of O– and H2 (case1)

O2 + 4e

O2 + 2e

2O2–

2O–

H2 + O2–

H2 + O–

H2O + 2e–

H2O + e–Res

ista

nce

Xa(O2–+O–)

Xa(O–)

Case 2Case 1

0 0.5 1.0PH

2

0( x 10–3)

Figure 1.8 Schematic illustration of the gas response profile as related with O� andO2� adsorption species.


KNO2 and PNO2 are the equilibrium adsorption constant and partial pressureof NO2, respectively. In base air, oxygen is adsorbed, too, according toEq. (1.2), so that there are two kinds of adsorbates accommodating electronstransferred from the grain. Through the same procedure used in the previoussections, it can be derived that reduced resistance to NO2 in the stage ofvolume depletion is expressed as follows:

Rg

R0¼ cðnÞ þ

�Sa

ðKO2PO2ðaÞÞ1=2 þ

�Sa

KNO2PNO2 (1.13)

There is thus linear correlation between resistance and PNO2, with itsslope being inversely proportional to the grain size (a).

Gas response (Rg/Ra) is derived from Eqs. (1.13) and (1.5), if the grainsare already in the stage of volume depletion in air and Rg/Ra >>1, to be asfollows:

Rg

Ra¼

nKNO2

.ðKO2PO2ðaÞÞ1=2

oPNO2 (1.14)

The response is independent of a in this case because the dependence ofRg/R0 and Ra/R0 on a is canceled out. If regional depletion prevails inair, however, a totally different situation arises. Now, Ra/R0 is almostindependent of a, as stated previously, so that Eq. (1.14) is replaced, approx-imately, by Eq. (1.15). The response is then inversely proportional to a.

Rg

Ra¼

�R0

Ra

�Sa

KNO2PNO2 (1.15)

Figure 1.9 Correlations between gas response (Ra/Rg)2 and partial pressure of

hydrogen (PH2) as observed with SnO2 grains of 12 and 16 nm in diameter at 573K.


Devices based on WO3 have been found to be sensitive to NO2, asshown in Fig. 1.9, where granular and lamellar crystals of WO3 are usedin Fig. 1.10(a) and (b) respectively. The lamellar crystals with smaller a areseen to be particularly sensitive to NO2, in agreement with Eq. (1.15), beingcapable of detecting NO2 at 10 ppb. The response is linear to PNO2 at loweroperating temperatures. It is suggested that this material allows NO2 to beadsorbed efficiently, while keeping O2 adsorption at a minimal level(KNO2 >> KO2), thus imposing the situation rationalized by Eq. (1.15).

1.4.5 ExtensionsThe theory of gas response can be applied or extended to the analyses ofother related phenomena of gas sensors, though such work is still in its earlystages. For example, the rates of response and recovery have been formulatedtheoretically.24 It has also been derived that a type of sensitization takes placewhen semiconductor grains are dispersed with an additive that deprivesthem of conduction electrons and thus affects the reduction of the effectiveradius of the grains.25

The theory also provides a useful tool to understand the nature and rolesof the metaleoxideesemiconductor contacts involved in semiconductorgas sensors.26 Under conditions where oxide grains are covered with asufficiently large density of adsorbates (surface states), their energy band

100 200

150

100

50

00 50 100 150 200 250

80

60

40

20

00 400200 600 800 1000

200°C 200°C

250°C

300°C

300°C

400°C

Sen

sor r

espo

nse

(Rg /

Ra)

Sen

sor r

espo

nse

(Rg /

Ra)

NO2 concentration / ppb NO2 concentration / ppb

(a) (b)

Figure 1.10 Correlations between gas response (Rg/Ra) and partial pressure of nitrogendioxide as observed with WO3-based devices at various temperatures. (a) Granular WO3

as pyrolyzed from ammonium tungstate, (b) lamellar WO3 with crystallites of about13 nm in size prepared through a colloidal process.


structure is known to remain unaltered, even when they are brought incontact with a metal (pinning). Instead, contact potential (dCP, in volts) isgenerated across the contact to compensate the work function differencein between, in addition to the conduction band edge difference (dEC)appearing in between, as shown in Fig. 1.11. The expression for volumedepletion of dCP is as follows:

qdCP¼ qð4m � 4sÞ; q4s ¼ q4s;0 þ ðdRDðnÞþ pÞkT (1.16)

Here, qfm and qfs are the work function values of the metal and semi-conductor (f in volts), respectively, and fs,0 is the value of fs at flat bandstate. The expression dRD(n) kT gives the total lowering of the Fermi levelduring regional depletion, which is given by dRD(n) ¼ n2/(2S), where S isthe shape factor. For a resistor-type sensor, dCP acts as a directional barrier todrifting electrons. It reduces the drift mobility of the electrons travelingagainst it and eventually increases the resistance of the contact involved. Itfollows that the contacts between electrode metal and oxide grains are moreresistive and more gas-sensitive than the other usual contacts between oxidegrains. This has been confirmed to be the case with a narrow gap electrodesattached device, in which the aperture between electrodes was as small as1 mm or below.27

Metalesemiconductor contact also appears to play a key role in thepotentiometric gas sensors attached with oxide semiconductors. In thesedevices, gas response seems to reflect the change of contact

Figure 1.11 Energy band diagrams of oxide grain and metal electrode before and aftercontact under exposure to base air. Note: The band diagrams remain unaltered(pinning) while contact potential is generated in between upon contact.


potential imposed by switching from base air to the target gas ambient,dCP(g) � dCP(a), as described later. Through Eq. (1.16) and other relations,it is correlated with the gas response of resistor-type sensors in ideal cases asfollows:

dCPðgÞ� dCPðaÞ ¼ 4SðgÞ � 4sðaÞ ¼�RTF

In

�Rg

Ra

(1.17)

Here, R and F are gas constant and Faraday constant, respectively, andRT/F ¼ kT/q.

1.4.6 Nonresistive sensorsVarious nonresistive gas sensors using semiconductors have been proposed.As described below, these sensors, constructed based on various principles,provide useful information to learn how receptor function and transducerfunction are generated and combined together into gas sensors, thoughmost of the sensors are yet to be exploited further for use in practice.

1.4.7 Field-effect transistor-type gas sensors1.4.7.1 PrincipleThe typical structure and characteristic of Pd-gate FET gas sensors are illus-trated in Fig. 1.12(a) and (b). As is well-known, a FET, usually attached witha normal metal gate, is a device for controlling drain current by gate voltageapplied. Under well-controlled conditions, drain current starts to flow whengate voltage (V) exceeds a threshold voltage (Vth) and, on a further increasein V, it increases proportionally to (V � Vth)

2, as shown in Fig. 1.12(b). It isendowed with gas sensing ability when the metal gate is attached with anadequate foreign material. If the new gate system modulates the electricalfield underneath depending on the gas ambient, the drain current of thedevice at a fixed gate voltage will change accordingly. Alternatively, actualdevices focus attention to Vth and its shift is taken as gas response.

The FET gas sensor first proposed was Pd-gate FET; Pd particles weredispersed in the gate region.28 It responded to H2 and NH3 in air at423K. Reportedly, the H atoms dissociated from these molecules aredissolved into Pd metal and polarize in the vicinity of the border to theunderlying insulator layer (SiO2) to modulate the electrical field underneath.However, with no supporting evidence having been found, this speculationshould be reconsidered. Later, various materials were introduced successfullyinto the gate. Those are typified in three groups: i.e., solid electrolytes, oxidesemiconductors, and dielectrics. As observed, combinations of these


materials with the gate metal form gas-sensitive functional systems: half cell,metalesemiconductor contact, or capacitor, respectively.

1.4.7.2 Solid electrolyte-gate field-effect transistorThree-phase contact between metal, solid electrolyte, and gas is known toact as an active site for electrochemical reactions (half cell reaction). If thesolid electrolyte is a proton conductor, for instance, the following reactiontakes place in the presence of H2, and the half cell equilibrium is expressedby the following Nernst equation:

H2¼ 2Hþ þ 2e�;FM �FSE ¼ ��RT2F

In PH2 þ Constant (1.18)

The electrical potentials of metal and solid electrolyte are FM and FSE,respectively. The constant is determined by the kinds of materials involved.The same half cell is formed when the proton conductor is placed betweenthe gate metal and the insulator layer of the FET. Thus, FSE is raised by an

VG

VDS

Pd

SiO2

p-Sin n

ID

ΔV

ID

ΔV

With H2 Without H2

Vth VG

(a)

(b)

Figure 1.12 (a) Structure of Pd-gate FET. (b) Drain current (ID) characteristics observed:VG, gate voltage, VDS (sourceedrain voltage).


amount as indicated by Eq. (1.18) higher than FM, which is now controlledexternally as gate voltage. This means that, at a fixed gate voltage, FSE

increases and, hence, the electrical field underneath also increases withincreasing PH2. In the alternative mode of operation, Vth shifts down asPH2 increases, following Nernst’s equation. Such behavior has beenconfirmed experimentally with an antimonic acid layer attached device,which responded well to H2 diluted in N2 at room temperature.29 Theresponse to H2 in air deviated considerably from this behavior because ofthe occurrence of mixed potential.

Similarly, devices sensitive to NO2 or CO2 can be fabricated by attachingNaNO2 (Naþ ionic conductor) or Li2CO3-based composite salt (Liþ ionicconductor) to the gate, respectively.30,31 The response mechanisms involvedcan be understood in the same way. In the NO2 device, for example, the halfcell reaction is expressed as follows:

NO2þ e� þNaþ ¼ NaNO2;FM � FSE ¼�RTF

InPNO2 þ Constant

(1.19)

FSE � FM should shift down and so Vth should shift up, with increasingPNO2. This behavior has been confirmed experimentally, as shown inFig. 1.13. The device was fairly sensitive, responding to a few tens ppbNO2 in air, showing a Nernst slope fairly close to that of the one-electron reaction expected.

In conventional electrochemistry, a half cell is always combined withanother (reference half cell), and its electrochemical equilibrium is investi-gated through the cell voltage (EMF). In contrast, the half cell of the presentdevice is combined with an FET underneath and its electrochemicalequilibrium is investigated through Vth.

1.4.7.3 Oxide semiconductor-gate field-effect transistorOxide semiconductors have been introduced into the gate of FET. A typicalexample would be theWO3-gate FET, which was sensitive to NO2 in air, asshown in Fig. 1.14.32 Obviously, the high sensitivity originates fromthe excellent receptor function of WO3 to NO2. In current devices,metalesemiconductor contact is made between the gate metal and fineWO3 crystals, and the resulting contact potential seems to play a decisiverole. Owing to the contact potential, FS � FM goes up or down with achange in PNO2, according to Eq. (1.16). Here, FS is the electrical potentialof WO3. In the same way as the previous devices were treated, the gas


Gate voltage (VGS)

Sourceelectrode

Source-drain voltage

P-type

NNN-channel

VG

NO2

NaNO2 + WO3 Drain electrode

Ta2O5 / SiO2

400

350

300

250

20010 100 1000

NO2 concentration / ppb

130°CVDS = 3VID = 200mA

VG

/ m

V

Air

78.9 mV / decade(n = 1.0)

(a)

(b)

A

Figure 1.13 NaNO2-gate field-effect transistor (FET) sensor. (a) Construction of NaNO2-gate FET sensor, (b) NO2 sensing characteristics observed.

500

400

300

200

10010 100 1000

NO2 concentration / ppb

150°C

180°C

92.9 mV /decade(n = 0.9)

119.6 mV /decade(n = 0.8)

Air

Air

VG

/ m

V

Figure 1.14 NO2 sensing characteristics as observed with WO3-gate field-effecttransistor (FET) sensor.


response in threshold voltage mode is derived by using Eq. (1.17), given asfollows:

VthðgÞ�VthðaÞ ¼�RTF

�1n PNO2 þ

�1n

�R0

Ra

�Sa

KNO2

�(1.20)

When PNO2 is sufficiently large, Eq. (1.18) is seen to be very similar toEq. (1.19), with the response linearly correlated with PNO2 on a semi-logarithmic scale with the same Nernst slope. However, the constantsappearing in both the equations have totally different meanings from eachother. The constant in Eq. (1.20) mainly reflects the sensitivity of thereceptor function of the grains to NO2. It determines the position ofthe semilogarithmically linear correlation along the vertical axis and, so,the lower detection limit of PNO2.

1.4.7.4 Dielectric material-gate field-effect transistorWhen a layer of dielectric material is introduced beneath the gate metal, acapacitor is formed on the top of the FET, its capacitance varying dependingon the dielectric constant and layer thickness of the material. The presenceof the capacitor naturally imposes modulation of the electrical fieldunderneath, which is otherwise controlled by the gate voltage only. If thedielectric layer is porous and capable of absorbing a polar molecule gaseffectively to change its dielectric constant, the resulting device is madesensitive to the gas through the change in capacitance; Vth moves furtheraway from the air level as the gas partial pressure increases. Based on thisprinciple, the devices sensitive to polar gases (such as water vapor andethanol gas) have been fabricated fairly successfully by using dielectricmaterials such as cellulose and its derivatives.33

1.4.8 Oxygen concentration cell type sensorsAn oxygen concentration cell is constructed by using stabilized zirconia (anO2� ionic conductor) and it is known to work well as an oxygen sensor. Ifan oxide semiconductor such as SnO2 is deposited between the sensingelectrode (Pt) and zirconia (Fig. 1.15(a)), the device is also made sensitiveto various reducing and oxidizing gases other than oxygen.34 The response(EMF) to such a nonoxygen gas, starting from 0 in base air, increases ordecreases linearly with the increasing logarithm of the partial pressure ofthe gas (Fig. 1.15(b)), while EMF to a fixed gas ambient varies somewhatdrastically with the kind and size of the oxides used. For a considerabletime, such a response to nonoxygen gases has been considered to be


ascribable to the mixed potential generated at the zirconia/oxide semi-conductor interface and, for this reason, such devices have been called“mixed potential” type sensors. The mixed potential is postulated to begenerated to H2 in air, for instance, through the following pair of redoxreactions:

O2þ 4e�/2O2�; O2� þH2/H2Oþ 2e�

However, it is hard to understand why the response is promoted by adecreasing size of oxides (grain size effect) based on this theory.

Basic approaches to this group of sensors are highly desired to reveal thefundamental mechanism of gas sensing involved.

1.4.9 Other gas sensorsThis section describes types of semiconductor gas sensors that have not beenmentioned so far.

1.4.9.1 Metaleinsulatoresemiconductor capacitor type sensorsThe structure of a MIS capacitor and its capacitance versus applied voltagecharacteristics are shown in Fig. 1.16(a) and (b),35 respectively. A MIScapacitor is obtained if a MIS FET is deprived of the sourceedrain currentchannel (see Fig. 1.12). To provide the MIS capacitor with gas sensing

Figure 1.15 Oxygen concentration cell type gas sensors attached with oxide semicon-ductors. (a) Device structure, (b) responses to reducing gas.


ability, foreign materials (solid electrolytes, oxide semiconductors, or dielec-trics) are placed under the metal layer, in the same way as in a MIS FET. Inthese devices, the capacitance depends on the voltage applied to the metallayer (relative to the semiconductor), whereas at a fixed voltage it changeson switching from air base to gas ambient. To keep the capacitance thesame, the applied voltage is obliged to shift up or down on changes in theambient, and this shift is taken as the response of the device to the gas.

1.4.9.2 Diode-type sensorsNon-ohmic contact between metal and semiconductor shows a rectifyingproperty, which is utilized in what is known as a “metalesemiconductorcontact diode” (Schottky diode). Many researchers have attempted to applythe same principle to gas sensors. Various combinations between metals(Pt, Pd, Ag, etc.) and oxide semiconductors (TiO2, ZnO, etc.) have beenchosen to fabricate diode devices. In many cases, the resulting devices

Sensingphase

Ta2O5

Si

Gold mesh (top electrode)

SiO2

Au (backside electrode)

310

300

290

280

270

260

250

240200 400 600 800 1000

Applied voltage to p-silicon / mV

1 ppm NO2

Air

Cap

acita

nce

/ pF

(a)

(b)

Figure 1.16 Metaleinsulatoresemiconductor (MIS) capacitor. (a) Structure of MIScapacitor, (b) capacitance versus applied voltage characteristic obtained.


showed a reducing gas-dependent rectifying property; in H2 containingambient, forward current density was promoted conspicuously withincreasing PH2, while reverse current density was also promoted as well,which was unexpected. Such gas-dependent behavior is of sufficient interestfrom a standpoint of developing gas sensors. At the same time, however, itsuggests the need to reconsider the gas sensing mechanism involved. Amatter of concern is whether the contacts formed there are, in fact, of thenon-ohmic type, as expected. It has been recognized in other semicon-ductor sensors that the same contact is achieved through generating contactpotential instead of undergoing electron transfer, as stated previously. Thecontact potential can be responsible not only for the rectifying propertybut also for the promotion of current density, forward as well as reverse,with increasing PH2. Therefore, further careful investigations are neededinto this type of gas sensor.

1.5 Future trends

Semiconductor gas sensors will become more and more important inthe future. Seeds and needs for them, basic approaches needed, and chal-lenges desired are described below as a personal view of the present authors.

1.5.1 Needs and seedsThere is a great variety of gases around us of different properties, origin, andconcentration. Some are hazardous and should be kept under control, whileothers may be vital for life or symptomatic of health conditions. Gas sensorsare needed for various purposes: safety, amenity, energy saving,health, foods, environmental protection, and so on. As is well-known, theapplication of gas sensors in practice began with inflammable gas alarms toprotect people from fatal gas hazards such as gas explosions, incompletecombustion accidents, and exposure to poisonous gases. Fire alarms usinga semiconductor gas sensor in combination with a smoke or thermaldetector and breath alcohol checkers for preventing drunken driving arealso examples of gas sensors used for safety purposes. For the purposes ofamenity and energy saving, air quality sensors have been installed in aircleaners, while a pair of sensors sensitive to CO and NO2 has been incorpo-rated into a car autodamper system. Odor sensors and breath odor checkersalso belong to this category.

Gas sensors are important in other categories, too, though theirdevelopment is more difficult because the target gases concerned are usually


of very low concentrations. For example, volatile organic compounds areone of the urgent targets; if generated in houses, those may cause sick housesyndrome, while some of them are even carcinogenic. Various hazardousgases frequently used in factories, laboratories, and hospitals should becontrolled with the use of gas sensors to protect the health of people work-ing there. Sensing of bioactivity-related gases is also important in health andfoods. Detection of disease-related gases is drawing increasing attention formedical purposes. Sensory monitoring of air pollutants has been a deepconcern to many researchers but, unfortunately, for a variety of reasonsthis is yet to receive attention. Semiconductor gas sensors, which areendowed with high sensitivity compared with other gas sensors, are, inprinciple, the best suited for such applications, though a great deal of effortshould be put into substantiating new frontiers for gas sensors.

As a new seed in gas sensors, microsensors fabricated by using MEMStechnology, known as “MEMS sensors,” have recently been exploitedextensively, aiming at realizing battery-driven gas sensors. As shown inFig. 1.17, the gas sensing layer (about 100 � 100 mm wide and a few tensnm thick) is deposited on a diaphragm, which is suspended over a cavitycreated within a silicon chip. Electrodes and a heater are printed on thediaphragm beforehand. As a typical feature of such a microdevice,the sensing layer temperature can be changed quickly (within 30 ms), sothat the device can be compatible with temperature-programmed operation.This feature would seem to bring about new intelligent functions to gassensors. Temperature-programmed gas response diagrams, for instance,may be useful for the identification of target gases.

1.5.2 Basic approaches desiredSemiconductor gas sensors have so far been developed on the basis ofexperience and intuition. Tremendous efforts have been devoted to

Figure 1.17 Microelectromechanical system gas sensor.


discovering new sensing materials, new ways of materials processing, newtypes of device, new targets for gas sensing, and so on, putting emphasison gas sensing performances. This approach, however, is not always so effec-tive for further advances of gas sensors. With the receptor function of smalloxide semiconductors having been clarified, there is now a keen need forapproaches shedding light on the more basic side of gas sensors. The knowl-edge thus accumulated will be useful in establishing guidelines for designingsemiconductor gas sensors. Matters for further investigation include thefollowing:• establishing methods to characterize and control semiconductive proper-

ties, especially the donor density, of oxides;• seeking quantitative correlations between sensitivity data and catalytic

oxidation data for a series of inflammable gases;• seeking quantitative correlations between semiconductor properties and

gas sensing properties for oxides;• basic analyses of the existing state and the roles of sensitizers;• basic analyses of the effects of mixing one oxide semiconductor with

another.Preparation of discrete nanocrystals of oxide semiconductors has become

increasingly popular recently. Sensors using nanocrystals of exoticmorphology have been fabricated and often shown to exhibit interestinggas sensing properties. Unfortunately, however, origins of such interestingproperties have received scant investigation from a basic standpoint, makingit difficult to draw on information useful in the design of gas sensors. In fact,nanosized crystallites have already been utilized in practical gas sensors. Itwould be informative to undertake a critical evaluation of the differencesbrought about by such a change in morphology.

1.5.3 ChallengesThere are subjects of research which are worth challenging to progress theinnovation of semiconductor gas sensors. Some examples are listed below:1. Elucidation of control of water vapor effects: Disturbances by water

vapor have been a major origin of errors in gas response. Eliminationof them upgrades the quality of gas sensing.

2. Verification of ultrasensitive gas sensors: New frontiers of gas sensorapplications often demand that they cope with reducing gases atsub-ppm levels. It is necessary, first, to prove that such high-sensitivesensors can be devised.


3. Contact potential-conscious sensor design: Gas response of a resistor-type sensor seems to be promoted significantly by contact potential if aproperly designed composite gas sensing layer is used.

4. Exploration to make FET type and oxygen concentration cell type gassensors more flexible in operating temperature: FET based on siliconcannot function at temperatures higher than c.180�C, whereas the cellusing zirconia cannot function at temperatures lower than c.550�C;neither is able to work in the most important temperature range for gassensing. Exploration for new semiconductors and new solid electrolytesis desired to eliminate these limitations.

References1. Seiyama T, Kato A, Fujiishi K, Nagatani M. Anal Chem 1962;34:1502.2. Taguchi N. Published patent application in Japan. 1962. S37-47677, Oct.3. Xu C, Tamaki J, Miura N, Yamazoe N. Sensor Actuator B Chem 1991;3:147.4. Rothschild A, Komen Y. J Electroceram 2004;13:697.5. Rothschild A, Komen Y. J Appl Phys 2004;95:6374.6. Yamazoe N, Shimanoe K. J Electrochem Soc 2008;155:J85.7. Yamazoe N, Shimanoe K. J Electrochem Soc 2008;155:J93.8. Watson J. Sensor Actuator 1984;5:29.9. Madou M, Morrison SR. Chemical sensing with solid state devices. Boston: Academic Press;

1989.10. Korotcenkov G. Chemical sensors: fundamentals of sensing materials. New Jersey:

Momentum Press; 2011.11. Yoshioka T, Mizuno N, Iwamoto M. Chem Lett 1991;20:1249.12. Clifford PK, Tuma DT. Sensor Actuator 1982/1983;3:233.13. Sakai G, Matsunaga N, Shimanoe K, Yamazoe N. Sensor Actuator B Chem 2001;80:125.14. Yamazoe N, Shimanoe K. Sensor Actuator B Chem 2011;154:277.15. Yamazoe N, Shimanoe K. Sensor Actuator B Chem 2011;158:28.16. Yamazoe N, Suematsu K, Shimanoe K. Sensor Actuator B Chem 2012;163:128.17. Shimizu Y, Egashira M. MRS Bull 1999;24:18.18. Yamazoe N, Suematsu K, Shimanoe K. Sensor Actuator B Chem 2013;176:443.19. Ma N, Suematsu K, Yuasa M, Kida T, Shimanoe K. ACS Appl Mater Interfaces 2015;7:

5863.20. Suematsu K, Sasaki M, Ma N, Yuasa M, Shimanoe K. ACS Sens 2016;1(7):913.21. Suematsu K, Uchino H, Mizukami T, Watanabe K, Shimanoe K. J Mater Sci 2019;

54(4):3135.22. Sun Y, Suematsu K, Watanabe K, Nishibori M, Hu J, Zhang W, Shimanoe K.

J Electrochem Soc 2018;167:B275.23. Hua Z, Yuasa M, Kida T, Yamazoe N, Shimanoe K. Chem Lett 2014;43:1435.24. Yamazoe N, Shimanoe K. Sensor Actuator B Chem 2010;150:132.25. Yamazoe N, Shimanoe K. Thin Solid Films 2009;517:6148.26. Yamazoe N, Shimanoe K, Sawada C. Thin Solid Films 2007;515:8302.27. Tamaki J, Niimi J, Ogura S, Konishi S. Sensor Actuator B Chem 2006;117:353.28. Lundstr€om I, Shivaraman MS, Svensson C, Lundkvist L. Appl Phys Lett 1975;26:55.29. Miura N, Harada T, Yoshida N, Shimizu Y, Yamazoe N. Sensor Actuator B Chem 1995;

24e5:499.30. Nakata S, Shimanoe K, Miura N, Yamazoe N. Sensor Actuator B Chem 2001;77:512.


31. Shimanoe K, Goto K, Obata K, Nakata S, Sakai G, Yamazoe N. Sensor Actuator B Chem2004;102:14.

32. Nakata S, Shimanoe K, Miura N, Yamazoe N. Electrochemistry 2003;71:503.33. Karube I, Tamiya E, Sode K, Yokoyama K, Kitagawa Y, Suzuki H, Asano Y. Anal

Chim Acta 1988;213:69.34. Lu G, Miura N, Yamazoe N. J Electrochem Soc 1996;143:L154.35. Zamani C, Shimanoe K, Yamazoe N. Sensor Actuator B Chem 2005;109:216.


CHAPTER TWO

Conduction mechanism insemiconducting metal oxidesensing films: impact ontransductionN. Barsan, M. Huebner, U. WeimarUniversity of T€ubingen, T€ubingen, Germany

Contents

2.1 Introduction 392.2 General discussion about sensing with semiconducting metal oxide gas sensors 412.3 Sensing and transduction for p- and n-type semiconducting metal oxides 47

2.3.1 Modeling of conduction for p- and n-type semiconducting metal oxides innormal conditions (operation in air)

49

2.3.2 Modeling of the conduction for n-type semiconducting metal oxide:extension to low oxygen concentrations

53

2.4 Investigation of the conduction mechanism in semiconducting metal oxidesensing layers: studies in working conditions

57

2.4.1 Sample preparation and experimental conditions 572.4.2 Conduction mechanism of p-type CuOdexperimental results 582.4.3 Conduction mechanism of n-type SnO2dexperimental results 62

2.5 Conduction mechanism switch for n-type SnO2ebased sensors during operationin application-relevant conditions

66

2.6 Conclusion and future trends 67References 67

2.1 Introduction

Chemoresistive gas sensors based on semiconducting metal oxides(SMOXs) are very successful, being sold in millions, in applications as diverseas the detection of explosive gas leakages in residential premises, or the con-trol of air intake in car interiors.1 There is a continuous effort to extend theirapplications in markets as different as indoor air quality or consumer goods(AMS, Austria http://ams.com/eng/Products/Environmental-Sensors/

Semiconductor Gas Sensors, Second EditionISBN: 978-0-08-102559-8https://doi.org/10.1016/B978-0-08-102559-8.00002-1


http://ams.com/eng/Products/Environmental-Sensors/Gas-Sensors

https://doi.org/10.1016/B978-0-08-102559-8.00002-1

Gas-Sensors and Sensirion, Switzerland https://www.sensirion.com/en/environmental-sensors/gas-sensors/multi-pixel-gas-sensors/). After theinitial publication of gas sensitive effects on germanium by2; metal oxideswere identified as possible sensitive materials by3e5 and were brought tothe market by,6 who founded the largest manufacturer of SMOX sensors:Figaro Engineering (Figaro, Osaka, Japan, http://www.figarosensor.com/).

The success of this type of device is based on their good priceeperformance ratio; they are• inexpensive (the price range is a few euros per sensor);• easy to use (there is a direct relationship between the concentration of the

target gas and the sensor resistance);• very sensitive (generally being able to measure down to a few ppm, or

even a few hundred ppb);• very stable (with reported life times extending into decades);• easy to integrate in arrays for more ambitious analytical tasks; and• reasonably low power consumption when realized on micromachined

membranes using a pulsed temperature mode (realized by batteryoperation).The gas detection with SMOX-based gas sensors is, in principle, simple:

in air, at temperatures between 150 and 400�C, oxygen is adsorbed on thesurface of the metal oxides by trapping electrons from the bulk with theoverall effect of increasing the resistance of the sensor (for n-type materials)or decreasing it (for p-type materials). The additional occurrences of gases inthe atmosphere that react with the preadsorbed oxygen, or directly with theoxide, determine the relative changes of the sensor resistance (sensor signals).From this very naïve picture, one can already get the idea that one has toexamine two aspects: the surface reaction taking place between the materialand the gases (called the “receptor function”) and the transduction of it intothe corresponding changes in the electrical resistance of the sensor. Thiscontribution examines the influence of the conduction mechanism on thetransduction of surface reactions into sensor signals. Section 2.2 presentsthe understanding of the functioning of SMOX-based sensors, and Section2.3 examines the main differences brought about by the type of conductionof the material. Section 2.4 presents examples of applying simultaneouswork function and conductance measurements to the theoretical study ofthe conduction mechanisms. In Section 2.5, based on experimental resultsobtained in more realistic conditions (exposure to CO in humid air) andby using the findings from the theoretical modeling we demonstrate thatalso in practical application a switch of the conduction mechanism is

40 N. Barsan et al.

http://ams.com/eng/Products/Environmental-Sensors/Gas-Sensors

https://www.sensirion.com/en/environmental-sensors/gas-sensors/multi-pixel-gas-sensors/

https://www.sensirion.com/en/environmental-sensors/gas-sensors/multi-pixel-gas-sensors/

http://www.figarosensor.com/

possible. This chapter closes with Section 2.6 which offers a set of conclu-sions and an outlook for future studies.

2.2 General discussion about sensing withsemiconducting metal oxide gas sensors

All SMOX-based gas sensors are realized by depositing a sensing layerover an insulating substrate provided with electrodes and a heater. The elec-trodes are used for the readout of sensor resistance; the heater raises the tem-perature of the SMOXs sufficiently high to allow for their fast andreproducible operation, generally between 150 and 400�C. An example ispresented in Fig. 2.1.

In this example, the sensing layer, in the form of a thick porous film, isdeposited by applying screen-printing technology onto a planar aluminasubstrate equipped with interdigitated Pt electrodes on its front, for thereadout of the electrical resistance, and a Pt heater on its reverse, whichallows the sensor to operate at well-controlled temperatures. All commercialsensors are based on thick porous layers, for reasons that will be given below;in addition to screen printing, other coating technologies (e.g., drop coating)

Sensor device

Sensing layer

4.2 mm

Cross section

Pt-electrode

porous sensing layer

3.5 mm

layer morphologywith electrodes

porous layer withlarge grains

7 m

m

25.4

mm

500 μm 1 μm

Figure 2.1 Design of the sensor substrate used at the University of T€ubingen; theporous thick film sensing layer is deposited on to an alumina substrate, providedwith interdigitated Pt electrodes and a Pt heater on the backside allows the operationat well-controlled temperatures.

Conduction mechanism in semiconducting metal oxide sensing films: impact on transduction 41

can be applied. Although the working principles of such devices seem quitesimple, the sensing processdwhich includes surface reactions, correspond-ing charge transfer processes and their translation into variations of the elec-trical resistance of the sensordis very complex. Fig. 2.2 presents a diagram ofthe various elements involved in the simple case of CO detection with ann-type SMOX (e.g., an SnO2-based gas sensor).

O2(gas)

O2

CO(gas)CO2(gas)

– O–

2O–CO–

2

e–e–e–

O– O– O–O–

O–O–O–O–

O– O– O–O–O– O– O–

O–

O–O–O–O–O–

O–O–O–

EVac

EF

O–O–

O– O–O– O–

O–O–O–

O–

O–O–O–

O–O–O–O–

O– O–O–

O–O–O–O–

∼I

∼I

R

R

qVc qVs

Figure 2.2 Sketch representing how the surface reactions are transduced into ameasurable signal. Due to the chemisorption of atmospheric oxygen, a depletion layerat the surface of the grains is formed. The presence of reducing gases like CO reducesthe negative charge trapped at the surface under formation of CO2. The measurableresult is a decrease in the sensor’s resistance (R). The surface reaction and the corre-sponding conduction situation are indicated by the arrows.

42 N. Barsan et al.

Fig. 2.2 demonstrates how, due to the chemisorption of atmosphericoxygen, a depletion layer is formed on the surface of the grains comprisingthe sensing layer. The presence of reducing gases, such as CO, reduces thenegative charge trapped on the surface by the formation of CO2. Themeasurable result is a decrease in the sensor’s resistance (R). The surfacereaction and corresponding conduction are indicated by the arrows.

The grains of the sensing layer are loosely sintered together; in theexample, it is considered that any influence of the surface does not extendinto the whole grain, so one can consider that there are two distinctly sepa-rate areas: a space charge layer on the surface and, unaffected by exposure togas, the bulk. In dry air, atmospheric oxygen interacts with the surface ofSnO2, acceptor levels are created, and electrons from the conductionband are trapped at these levels, forming molecular and/or atomic oxygenions. Consequently, the depletion layer appears on the surface of the grains;in the energy band representation, this is formalized as a bending of theupward band, meaning that the electrons need more energy to reach the sur-face (against the electric field of the negatively charged surface). Hence, theconduction in the sensing layer is controlled by the back-to-back Schottkybarriers formed between the grains. It is generally accepted that the CO isreacting with preadsorbed oxygen, forming CO2 that disperses in the atmo-sphere.7 These surface reactionsdthe ionosorption of oxygen and its con-sumption by the presence of COdare the chemical basis of sensing; theydescribe the receptor function of the sensitive material. The charge transfer,associated with the surface chemical reactions, determines the measuredeffect, namely the resistance change: the reaction of CO with the iono-sorbed oxygen decreases the surface negative charge, the consequence beinga reduction of the energy barrier height between the grains. That enables aprogressively greater number of electrons to flow from one electrode to theother, which translates into a reduction of resistance in the sensor. Assumingthat the intrinsic characteristics of the material remain constant, the relation-ship between the change of the surface charge and the change of the resis-tance depends on the morphology of the thick film layer. A useful criterionfor classification takes into consideration the accessibility of the sensinglayer’s bulk to gases, and Fig. 2.3 illustrates a simple distinction betweencompact and porous layers.

In the case of a compact layer, gas interaction only takes place on thegeometric surface; the flow of current is only influenced by the thicknessof the depletion layer on the surface of the layer. For porous layers, the


gas can penetrate into the entire layer and, in that way, each individual grainis affected by the surrounding gaseous composition. The current is conse-quently determined by the barriers between all the grains.

For compact layers, the bulk is not accessible to gases and the interactiononly takes place on the geometric surface (the as-formed electron depletedlayer is colored light gray, in contrast to the electron-rich bulk regioncolored dark gray. Here, the assumption that the constant material propertiesdo not depend on the process by which the layer is formed ensures that, forboth type of layer, there are surface and bulk zones). The electrical currenttherefore flows parallel to the surface and the conduction process takes placein the lower resistive bulk area, with the consequence that it is only indi-rectly influenced by the modulation of the low resistive cross-section area.This explains why the relative resistance changes for such kinds of layerare low.

In the case of porous layers, the gaseous species can penetrate into thebulk, which makes the active surface much deeper. Here, the electrical cur-rent is forced to cross the surface by passing from one grain to the next and,accordingly, is directly influenced by the energy barriers between the grains(grain boundary model). These are the main reasons why the best results forn-type metal oxide-based gas sensors are obtained by using porous thick filmlayers, where the conduction mechanism is controlled by the back-to-backSchottky barriers. Furthermore, the dimensions of the grains in porous layers(d) have to be taken into consideration. There are two casesddepending onthe relation between the dimension of the grains, d, and the Debye length,

Gas

Current flowCurrent flow

Porous layer

ProductGas

Compact layer

Product

Figure 2.3 Schematic drawing showing the difference between a porous and acompact layer. In case of a compact layer, the gas interaction only takes place at thegeometric surface; the current flow is only influenced by the thickness of the depletionlayer at the layers surface. For porous layers, the gas can penetrate into the whole layerand by that every single grain is influenced by the surrounding gaseous composition.The current is consequently determined by the barriers between all the grains.

44 N. Barsan et al.

LDdwhich show a different dependency between the conductance and thetarget gas concentration:C Case 1: grains large enough to have an unaffected bulk area (d >> LD);C Case 2: grains smaller than, or comparable to, the Debye length

(d � LD).A detailed discussion about the modeling of the two cases is given in

Ref. 8 and 9.It must be noted that because of the access of the gases to the whole vol-

ume, the interaction can take place in different parts of the sensor device;meaning that, in principle, it is possible to have contributions from the entiresensor and not only from the sensitive material. Fig. 2.4 presents a scanningelectron microscopy diagram showing the cross section of an SnO2 porousthick film sensor and the different contact possibilities. In addition to thegrainegrain contacts (a), Fig. 2.4 shows additional interfaces that can playa role in sensing and transduction: the graineelectrodeeAl2O3 substratecontact (c) and/or the graineAl2O3 substrate contact (b). The most obviouscontribution might be related to (c), due to the fact that the current needs togo through the electrodes and that, due to the noble metal nature of theseelectrodes, there is a possibility of catalytic effects. The insulating nature ofthe inert substrate means that (b) is a significantly less probable influencing

(a)

(b)

(c)

Porous layer withgrain boundary model

porous thick film layerSnO

Pt-electrode

AI O -substarte

Grain - AI O -substrate contact

50 μ

m

ra

Figure 2.4 The different possibilities of gas interaction in the case of porous layers. Thegas penetrates into the layer and the interaction can therefore take place at the grainegrain boundaries (a), the graineAl2O3esubstrate contact (b), and at the grainelectrodeeAl2O3 substrate contact (c).


factor. Furthermore, it was shown that the electrical contribution of theelectrodeeSMOX interface is a series resistance that does not change undertarget gas exposure,9 the influence of which it is possible to minimize bymaking sure that the number of grainegrain contacts between the electrodesis much greater than 2. This simply implies that the use of gaps between theelectrodes is much greater than the grain size. The proven chemical effect ofthe electrodes10,11 has to be considered as an influencing factor for theaverage grainegrain sensing unit in the layer.

Discussion in the following sections will focus on the part of the conduc-tion process/mechanism in p- and n-type SMOXs. A theoretical discussion(modeling of the conduction) about the different dependencies between thesurface chemistry and the corresponding resistance of the layer will be per-formed. By the use of measurement techniques undertaken during workingconditions, the validity of the models will be proven experimentally. In do-ing so, one needs to use a parameter that is directly linked to both surfacereactivity and conduction.

The ideal candidate is the change in the surface band bending (qDV) ofthe SMOX because its magnitude under gas exposure is a measure of surfacereactivity and also controls the electrical transport from one electrode to theother. Consequently, one needs a technique which is able to directly mea-sure its changes on gas exposure; the well-established Kelvin probemethod12 for the measurement of work function changes (DF) was selectedfor this purpose. Work function changes can be caused by changes in theband bending (qDV), the electron affinity (Dc), or the bulk position ofthe Fermi level (D(EC,B�EF)[ electrochemical potential), as shown inFig. 2.5 which describes the influence of CO reaction in air on the workfunction of SnO2.

DF¼ qDV þ Dcþ D�EC;B � EF

�(2.1)

The latter contribution can be excluded in the temperature range inwhich SMOX sensors are usually operated because the thermal energy isnot sufficiently high for bulk reactions with the atmospheric gases (e.g., ox-ygen bulk diffusion) to take place.

Electron affinity depends on the concentration of surface dipoles, gener-ally linked to surface species related to the reaction with water vapor.13 Byensuring that the electron affinity is constant (Dc ¼ 0), one obtains directaccess to the changes in the band bending (qDV) by using the Kelvin probetechnique, which can be achieved by keeping the system in very dry condi-tions. The corresponding changes in the resistance can be easily measured

46 N. Barsan et al.

with the sensor device being used. Consequently, one can measure thedependence of the resistance on the surface band bending in different con-ditions and, therefore, the conduction mechanism can be identified.

2.3 Sensing and transduction for p- and n-typesemiconducting metal oxides

In the field of SMOX-based gas sensors, by far the most studied ma-terial is the n-type SnO2. Moreover, most of the commercial sensors mar-keted today are based on it, generally in combination with noble metaladditives.14,15 The other material used in commercial sensors in applicationsinvolving the detection of oxidizing gases is WO3, which is also an n-typesemiconductor. This is intriguing because since the early 1980s considerableefforts were directed toward finding alternative materials with a better ordifferent sensing performance, among them p-type oxides, such as Cr2O3

EV,S

EV,B

EV,S

Ea

ED,S

EC,S qVCO

EVac

qΔV

ΔΦ

CO in air

EF

ED,B

EC,B

x

EV,B

x

Ea

ED,S

EC,S

qVair

χ Φair χΦco

EC,B

ED,BEF

EVac

O-

Figure 2.5 Energy band representation of SnO2 showing the different contributions tothe work function for the example of CO sensing in dry air conditions. Following sym-bols are used for the different parameters: EVac h vacuum level; EC,S(B) h conductionband at the surface (in the bulk); ED,S(B) h donor levels at the surface (bulk); Ea h sur-face acceptor levels; EF h Fermi level; EV,S(B) h valence band at the surface (bulk); c helectron affinity; Fair(CO) h work function in air (on CO exposure); qVair(CO) h bandbending in air (on CO exposure); qDV ¼ change of band bending; DF h change ofwork function; and x h distance from the surface.


and CuO (see 16e20 for reports on their sensing performance to differentgases: H2, O2, EtOH, CO, NO2, etc.). Much lower sensor signals wereconsistently observedddefined for n-type SMOX sensors as shown inEq. (2.2) and for p-type SMOX sensors in Eq. (2.3)dwhen comparedwith SnO2-based sensors in spite of well-known high surface reactivity(R denotes the electrical resistance of the sensor, G denotes the electricalconductance).

Sn¼ Rair

Rgas¼ Ggas

Gair(2.2)

Sp¼Rgas

Rair¼ Gair

Ggas(2.3)

An example for this observation is given in Fig. 2.6, where the EtOHsensing behavior of p-type Cr2O3 (open symbols) is compared with thatof undoped SnO2 (filled symbols) exposed to CO. For both materials,huge changes in the work function (DF, continuous lines with squares)on target gas exposure (EtOH and CO) were measured, indicating a highsurface reactivity (change of band bending). In the case of n-type SnO2,these changes are translated into rather “large” sensor signals (dotted linewith filled circles). For p-type Cr2O3, similar changes resulted in muchlower levels of signal (dotted line with open circles). The reason for this

CO conc. [ppm]

EtoH conc. [ppm]

Sen

sor s

igna

l

0

0,00

–0,05

–0,10

–0,15

–0,20

–0,25

20 40 60 80 100

0 10 20 30 40 500

2

4

6

8

10

12

14

ΔΦ [e

V] ΔΦ

ΔΦS

S

(Cr2O3)(Cr2O3)(SnO2)(SnO2)

Figure 2.6 Comparison between p-type Cr2O3 and n-type SnO2. For similar changes inthe work function on exposure to EtOH and CO, respectively, the n-type material showsmuch higher changes in the resistance (sensor signal S).

48 N. Barsan et al.

was recently unveiled when it was demonstrated that large changes in thesurface band bending (qDV) do not result in large changes in resistance(sensor signals) because of the conduction mechanism.21,22

To understand and explain this huge difference in the relationshipbetween the surface chemistry and the changes in conductance, one mustlook in more detail at the conduction processes in these materials. The focuswill be, first, on the modeling of the conduction for p-type metal oxides todetermine the relationship between resistance and band bending in normaloperational conditions. Subsequently, n-type materials will be examined forconditions in which the conduction mechanism changes.

2.3.1 Modeling of conduction for p- and n-typesemiconducting metal oxides in normal conditions(operation in air)

Even if the initial modeling were based on results obtained on Cr2O3,21 the

large grain size of that material would have the effect of making the weightof surface phenomena even less significant because of the possible conduc-tion contribution of the bulk. Accordingly, to simplify the case under inves-tigation, CuO was used as a prototype p-type metal oxide for the gas sensingperformance in response to CO; its grain size making it feasible to considerthat the surface plays the dominant role in conduction.23 The validity of thefindings, although, is not limited to CuO. The p-type semiconductingbehavior of CuO is related to the presence of acceptor levelsdattributedto copper vacanciesdin the band gap, which determine the appearance ofholes in the valence band. The adsorption of oxygen on the surface ofCuO is considered to be at the origin of CO sensing and can be describedby Eq. (2.4):

12Oair

2 þ SA4O�ðadÞ þ hþ (2.4)

where Oair2 represents atmospheric oxygen, SA an adsorption site for oxygen,

O�ðadÞ the resulting chemisorbed oxygen species, and hþ the created hole in

the valence band. The interaction of atmospheric O2 with the surface of theSMOX determines the formation of acceptor levels, and the electrontransfer from the valence band to the surface leads to the formation ofionosorbed oxygen species resulting in upward band bending. The nega-tively charged surface is compensated by an increased hole concentration inthe valence band that determines the formation of an accumulation layer.This is a very important difference when one compares the case of p-type


materials with n-type materials: in p-type, the conductivity of the surfaceincreases because of the adsorption of atmospheric oxygen. This situation,represented by energy bands, is shown on the left-hand side of Fig. 2.7. Theeffect of CO exposure, very similar to the general reaction mechanism forSnO2 explained in Section 2.2, is the consumption of ionosorbed oxygenspecies that determines the reduction of negative charge trapped at thesurface:

COgasþOeðadÞ þ hþ/COgas

2 þ SA (2.5)

where COgas represents the carbon monoxide in the gas phase, OeðadÞ pre-

adsorbed oxygen species, hþ a hole in the valence band, COgas2 the formed

product, and SA a free adsorption site for oxygen. As a consequence, thehole concentration near the surface decreases; in energy terms, this situationis described by a decrease in the surface band bending. Hence, the con-ductivity at the surface decreases and the overall sensor resistance increases.

Fig. 2.8 illustrates the differences in the conduction mechanisms between“similar” porous layers of p-type and n-type materials. The “similarity”dmeaning comparable grain size, morphology, and parameters of thedepletion/accumulation layersdis considered to focus on the conductionmechanism only. The left-hand side of Fig. 2.8 describes an n-typeSMOX for a porous layer consisting of loosely sintered grains with a radiuslarger than the Debye length (not fully depleted). In this case, one can

EC.S EC.S

EV.S EV.S

EV.B

qΔV

ΔΦ

ΦCO

EC.B

EV.B

EF

EC.B

EF

EVac EVac

EaEa

χ χΦair

qVairqVco

CO

X X

Figure 2.7 Energy band representation for a p-type semiconducting metal oxide ma-terial. The reaction of CO as target gas with the p-type material causes a decrease inthe band bending (qDV) and therefore changes in the work function (DF). The usedsymbols for the different parameters are equal to Fig. 2.5.

50 N. Barsan et al.

differentiate between a gas sensitive surface depletion layer (upward bandbending, light gray) with a large electrical resistance and an unaffectedbulk region (dark gray) with a lower electrical resistance. The electriccurrent through the layer from one electrode to the other is thereforedetermined by the concentration of electrons (nS) having sufficient energyto overcome the potential barrier (qVS) between the grains (back-to-backSchottky barriers). The dependence between this concentration and thesurface band bending can be described by a Boltzmann distribution byassuming that the Schottky approximation is valid:

ns ¼ nb exp

�� qVS

kT

�(2.6)

where nb represents the electron density in the bulk and kT the thermalenergy (z0.05 eV). For the conductance, one can consequently write

Gnfexp

�� qVS

kT

�(2.7)

As shown in Fig. 2.7, the upward band bending in the case of p-typeMOXs determines the formation of an accumulation layer for holes.Accordingly, the conductivity in the surface space charge layer increasesin comparison with the bulk, and conduction takes place differentlycompared with that described by the depletion layer. The current willnow flow through the accumulation parallel to the surface and also through

O O O O O O O O

OOOOOOOO

O O O O O O O O O

OOOOOOOO

OO O O O O O O

OOOOOOOO

O O

OO O

O O O O O

OOOOO

n-typeSMOX p-typeSMOX

Depletion layer Accumulation layer

Figure 2.8 Cartoon-like illustration of the conduction processes and the correspondingenergy band representation for an n-type semiconducting metal oxide (SMOX) material(left) described by a depletion layer and for a p-type SMOXmaterial (right) with an accu-mulation layer.


the bulk; this situation can be described by two resistors in parallel. The lattercontribution from the bulk depends on the nature of the material and themorphology of the layer. It is obvious that larger grains have a higherbulk influence, which makes the surface effectdand, therefore, thesensingdless important. In this case (e.g., for Cr2O3 with rather largegrains), a complex relationship between the conductance/resistance andthe band bending was obtained.21 A much simpler relationship is devisedby ensuring that the grains are quite small (e.g., as in the case of CuO(z25 nm)),23 so that the contribution to the conductance of the bulk canbe ignored. Hence, one can assume that the conduction process is nowdominated by the average hole concentration in the accumulation layer(GpfepS).

Considering that also in this situation the Boltzmann statistics are valid,the average hole concentration epS can be easily calculated by using a one-dimensional approach.

eps¼ 1x0$

Zx00

pb exp

�qV ðxÞkT

�dx (2.8)

where pb represents the hole density in the bulk and x0 the width of the spacecharge layer. To evaluate the integral in Eq. (2.8), one must solve thePoisson equation for the accumulation layer if the conductance is deter-mined by the holes. After the first integration of Poisson’s equation, oneobtains24

dV ðxÞdx

¼ �ffiffiffiffiffiffiffiffiffiffiffiffi2kTpbεε0

r$exp

�qV ðxÞ2kT

�(2.9)

where V represents the potential at a certain point x and εε0 the relativepermittivity of the material. By using the definition of the Debye length�LD ¼ ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

kTεε0=q2pbp �

and the boundary conditions that V¼ VS for x ¼ 0,

the following relationship between the distance from the surface, x, and thepotential V(x) is obtained for the second integration:

x¼ffiffiffi2

p$LD$

exp

�� qVS

2kT

�� exp

�� qV ðxÞ

2kT

�(2.10)

Now, the integral in Eq. (2.8) can be calculated by changing the variablesand by using Eq. (2.10). For the average concentration of holes in the surfacespace charge layer, one obtains

52 N. Barsan et al.

eps ¼ pb exp

�qVS

2kT

�(2.11)

Hence, one can write for the dependence of the conductance and thesurface band bending:

Gpfexp

�qVS

2kT

�(2.12)

Comparing the latter expression with the corresponding one for n-typeporous thick film layers (Eq. 2.7), one can clearly observe that the same sur-face chemistry (same change in the surface band bending) is translated differ-ently into a change of the conductance/resistance depending on what kindof material is used. Considering the sensor signal as the relative change of thesensor resistance due to exposure to the target gas, Sn,p, one obtains that thesignal of a p-type is simply the square root of the signal for the n-type MOX:

Sp ¼ffiffiffiffiffiSn

p(2.13)

This fact clearly shows why the use of p-type materials as chemoresistivegas sensors is not optimal. Although these materials may be highly reactive tothe target gases, the output electrical signal is quite low, as already shown inFig. 2.6 for Cr2O3.

To calculate the band bending changes from the changes in the resis-tance, one has to use the following dependency:

qDV ¼ � 2kT$ln

�Rgas

Rair

�(2.14)

More details about the calculations can be found in Ref. 23.The classical, state-of-the-art preparation technology for SMOX-based

gas sensorsdthick, porous sensing layersdis not the best choice forp-type materials. In their case, the direct readout of the changes in the surfaceband bending would be more efficient; in the case of a resistive readout,thin, compact films with electrodes deposited on the top would be moreappropriate.

2.3.2 Modeling of the conduction for n-type semiconductingmetal oxide: extension to low oxygen concentrations

As already mentioned, for the case of n-type SnO2 porous layers, the con-duction mechanism in an oxygen containing background is determined bythe appearance of a depletion layer at the surface of the grains. The negativesurface charge related to ionosorbed oxygen species is compensated by a


positive space charge layer near the surface, resulting in an upward bandbending. In cases where the dimension of the grains is larger than the Debyelength, one can distinguish between a rather resistive space charge layer onthe surface and a bulk area with a lower resistance. The conduction is there-fore controlled by the barrier height on the surface of the grains. Only theelectrons which have sufficient energy to overcome the back-to-backSchottky barriers between the grains can move from one electrode toanother. The diagram of the conduction process in the depletion layer inFig. 2.9 helps achieve a better understanding. The dependence of theconductance and the surface band bending can be easily expressed byEq. (2.7), assuming that the Schottky approximation is valid and that thebulk donors are fully ionized.

By decreasing the amount of oxygen in the background (or by increasingthe concentration of the reducing gases), one consequently also decreases theinitial upward band bending on the surface of the grains; in certain condi-tions, it is possible to reach a flat band situation. That means that there areno energy differences between the surface and the bulk, as well as the factthat the concentration of the free charge carriers is constant. A similar situ-ation is possible also in air when the Debye length exceeds the grain size:fully depleted grains. There, the position of the Fermi level relative to theminimum of the conduction band does not correspond to the Fermi levelof the bulk conditions.9

If the oxygen concentration in the atmosphere were lowered, the flatband situation could most probably be reached in the absence of oxygen,if one considers that the full height of the upward band bending is only

Depletion layer n-type SMOX Accumulation layer n-type SMOX

O– O– O–O– O– O– O–

O– O–O– O–O–

O–O–O–O–O–

O– O–O–O–O–O–

O–

E E

qVs

–qVsEc.s Ec.s

EF

EF

Figure 2.9 Cartoon-like presentation of the conduction processes for n-type SnO2 incase of depletion layerecontrolled and accumulation layerecontrolled model. SMOX,semiconducting metal oxide.

54 N. Barsan et al.

determined by the ionosorption of oxygen, it means that there are nointrinsic electron traps on the surface. If the band bending is furtherdecreased (e.g., by exposure to reducing gases that will form surfacedonors25), one would record a downward band bending at the surface,meaning the formation of an accumulation layer and, therefore, the conduc-tion mechanism will change. This situation is presented in the right-handside of Fig. 2.9. The easiest way for the electrical current to travel fromone electrode to another is now through the accumulation layer on the sur-face (lowest resistivity). Hence, one can also assume, as in the case of p-typeMOXs, that the conductance is simply proportional to the average electronconcentration in the accumulation layer (Gn f ~nS). The challenge is now tofind a relationship which describes the dependence between the changes inthe band bending and the changes in the resistance for the accumulationlayer.

The procedure is very similar to that shown previously for the p-typematerial and is described in detail in Ref. 26. By assuming that also in theaccumulation layer conditions the Boltzmann statistics are valid and thatone can use a one-dimensional approach, one can write for the average elec-tron concentration:

enS ¼ 1x0$

Zx00

nb exp

�� qV ðxÞ

kT

�dx (2.15)

The only difference compared with the accumulation layer for thep-type SMOX is that the electrons are the charge carriers and that the accu-mulation is therefore described by a downward band bending. Again, hereone has to solve the Poisson equation24; with its solution, the averageelectron concentration ~nS can be calculated. One obtains

enS ¼ nb$exp

�� qVS

2kT

�(2.16)

The dependence of the conductance on the surface band bending in thecase of the accumulation layer can therefore be described by

Gnfexp

�� qVS

2kT

�(2.17)


By comparing Eqs. (2.17) and (2.7), it becomes obvious that the impactof the surface band bending on the conductance is very different dependingon the condition (depletion vs. accumulation layer). If the conduction isdescribed by the accumulation layer, the relative change of the resistanceis simply the square root of the relative change in relation to the depletionlayer model, assuming that in both cases one measures the same surface bandbending changes:

Sacc ¼ffiffiffiffiffiffiffiffiSdep

p(2.18)

The latter expression implies that the “largest” signals are obtained in thecase of the depletion layer model. The effect of the surface chemistry on theresistance changes becomes weaker where the conduction moves intothe accumulation layer.

The analytical solution (Eq. 2.16) obtained for the average electronconcentration in the accumulation layer is only valid in conditions whereone can use the Boltzmann statistics. If the conduction band edge at the sur-face crosses the Fermi level, this assumption is no longer valid. One then hasto use the FermieDirac statistics instead of the Boltzmann statistics to deter-mine the dependence of the electron concentration on the surface bandbending. The average electron concentration in the accumulation in thiscase can be calculated numerically.26 By comparing the trend of the analyt-ical and the numerical solutions, both describing how the average electronconcentration in the accumulation layer depends on the surface bandbending, the following statements can be made:• both trends are similar (same slope) up to around 0.3 eV after the crossing

of the surface conduction band edge with the Fermi level position;• with further increasing of the surface band bending, the trends are

strongly divergence, which reflects the lack of appropriateness of theBoltzmann approximation.The slope of the “correct” numerical solution is getting lower (a smaller

coefficient than (2kT)�1 in the exponent in Eq. [2.17]), indicating that theinfluence of the band bending on the resistance is becoming weaker.

The latter calculations in Eqs. (2.15) to (2.17) show that the conductionmechanism for an n-type SnO2 sensor might change from one controlled bya depletion layer to one dominated by transport through the accumulationlayer, depending on the operational conditions. This fact has to be borne inmind, especially if the concentration range to be explored is very large.

56 N. Barsan et al.

2.4 Investigation of the conduction mechanism insemiconducting metal oxide sensing layers: studies inworking conditions

The ideas and models presented in Section 2.3 are applied to SnO2

and CuO as model systems for n- and p-type SMOXs.

2.4.1 Sample preparation and experimental conditionsThe n- type semiconducting SnO2 powder was synthesized by a conven-tional wet chemistry solegel procedure (SnCl4(aq) and NH3(aq)) followedby a calcination treatment at 1000�C for 8 h.27 For the highly crystallinep-type CuO nanoparticles, a soft chemistry route was employed using amixture of copper acetate, oleic acid, and trioctylamine.23 The investigatedporous thick film layers were obtained by using the automatic screen-printing procedure. Therefore, the powders were mixed with an appropriateamount of an organic vehicle to obtain a homogenous paste which wassubsequently printed onto the alumina substrates. For the electrical readout,the substrates are provided with interdigitated Pt electrodes, and a Pt heateron the backside allowed the operation at well-controlled temperatures (seeFig. 2.1). To remove the residual organic solvent, the sensors were finallyheated in a moving belt oven (SnO2: 400e600�C; CuO: 300e450�C).

To investigate the conduction mechanism in the sensing layers, simulta-neous DC resistance and work function change measurements were taken inworking conditions using the Kelvin probe technique (McAllister KP6500K Probe). The latter is a noncontact, nondestructive method whichmeasures the changes in the contact potential differences (CPDs) betweenthe sensor and a vibrating reference electrode. Variations in the CPDinduced by the changes in the surrounding atmosphere (e.g., CO exposure)represent the changes in the material’s work function.12

DCPD ¼ eDF=q (2.19)

By choosing conditions in such a way that possible contributions fromthe electron affinity (Dc) to the work function changes can be ignored(very dry conditions excluding influences of surface dipoles from humidity),the changes in the surface band bending can be directly measured as changesin the contact potential difference (DCPD). In these circumstances, one candirectly correlate the changes in the surface band bending with the corre-sponding sensor resistance change. It is important to note that both measuredparameters are average values corresponding to an average sensing unit of the


layer, which includes all influences from grain size dispersion, influence ofsubstrate and electrode, etc.

2.4.2 Conduction mechanism of p-type CuOdexperimentalresults

Simultaneous DC resistance ([_Keithley_2000] multimeter) and work func-tion change measurements on exposure to CO (10, 30, 50, 70, and100 ppm) of a CuO-based porous thick film gas sensor were performed at150�C in dry air conditions. The time dependencies of the resistance andthe CPD are presented in Fig. 2.10. One observes that the resistanceincreases on CO exposure, whereas the work function decreases. This indi-cates that there are fewer holes in the accumulation layer and a decrease ofthe upward band bending occurs (see also Fig. 2.7). One expects that, in dryconditions, CO reacts with preadsorbed oxygen, resulting in the cancella-tion of a hole and the formation of CO2 as described by Eq. (2.5).

In these circumstances, the measured changes in the work function onexposure to CO should only be caused by changes in the band bending.To prove this assumption, the different contributions which may causechanges in the work function on increasing CO concentrations are shownin Fig. 2.11. The changes of the work function (DF, dark gray with stars)are directly measured, and the changes in the band bending (qDV, blackline with dots) are extracted from the resistance changes by using therelationship for the dependence of the resistance and the surface band

0.3

0.2

0.1

0.0

0 2 4 6 6 1010k

100k

ResistanceCPD

Time (h)

Res

ista

nce

(Ω)

CP

D (V

)

10pp

m C

O

30pp

m C

O

50pp

m C

O

70pp

m C

O

100p

pm C

O

Figure 2.10 Simultaneous contact potential differences (CPDs) and electrical resistancechanges of a CuO sensordoperated at 150�Cddue to exposure to different concentra-tions of CO (10, 30, 50, 70, and 100 ppm) in dry air conditions.

58 N. Barsan et al.

bending (Eq. 2.14). The changes of the electron affinity (Dc, open squares)are calculated according to the following equation:

Dc¼DF� qDV (2.20)

One clearly observes that no changes occur in the electron affinity duringthe reaction of CO with CuO in dry air conditions; all changes in the workfunction are caused by changes in the band bending.

The fitting curve describing the dependence between the sensor signal(Sp ¼ Rgas/Rair) and the corresponding changes in the work function/band bending is given in Fig. 2.12. The value for the slope, as obtained, fairlyaccurately reflects the dependence gained from the theory (experimentalvalue of 2.000 � 0.034 and theoretical value of 2).

The experimental results show a good match to the theory. In addition,it was becoming clear that the use of p-type materials as chemoresistive gassensing materials is not optimal. The same surface chemistry (same bandbending) results in a much lower sensor signal for the p-type materialcompared with n-type materials, where the conduction is described by adepletion layer model (upward band bending). Furthermore, the reactionof CO with preadsorbed oxygen species is supported because no changesin the electron affinity are observed.

To extend the findings to more realistic orientated conditions, similarexperiments in a background of 50% relative humidity (25�C) were

0.02

0.00

–0.02

–0.04

–0.060 10 00 30 40 50 60 70 80 90 100 110

CO conc. (ppm)

qΔV

Δχ

∆E (e

V)

∆Φ

Figure 2.11 Changes of band bending (qDV), work function (DF), and possiblechanges of the electron affinity (Dc) with increasing CO concentrations (10, 30, 50,70, and 100 ppm) of the CuO sensor in dry air condition operated at 150�C. DE repre-sents the changes of all contributions in the unit of eV.


performed. Fig. 2.13 presents the different contributions: the measuredchanges in the work function (stars), the calculated band bending changes(black with circles) from the measured resistance changes, and the extractedelectron affinity (squares) changes on CO exposure in the humid back-ground. There, the situation is completely different compared with the

2

1

0.0 0.1 0.2 0.3 0.4 0.5 0.6

∆Φ/2kT

∆Φ

Sen

sor s

igna

l

Sp = exp2.000 + 0.034KT–

Figure 2.12 Fitting curve describing the dependence between the sensor signal (Sp)and the corresponding changes in the work function. The as-obtained value for theslope reflects quite well the dependence gained from the theory.

qΔV

Δχ∆Φ

∆E (e

V)

0.00

–0.02

–0.04

–0.06

–0.08

0 20 40 60 80 100

CO conc. (ppm)

Figure 2.13 Changes of band bending (qDV), work function (DF), and possiblechanges of the electron affinity (Dc) with increasing CO concentrations (10, 30,50, 70, and 100 ppm) of the CuO sensor in 50% relative humidity (25�C) operated at150�C.

60 N. Barsan et al.

experiments in dry air: significantly larger changes in the work function areobserved, whereas, at the same time, the decrease in the band bending ismuch smaller. This implies that, on CO exposure in the presence ofhumidity, a strong decrease occurs in electron affinity. For a better under-standing of the difference observed between Figs. 2.11 and 2.13, the influ-ence of the humidity itself and its effect on the different contributions isdepicted in Fig. 2.14. The exposure to water vapor not only determines alarge decrease in the band bending (increasing resistance, circles) but alsoan increase in the electron affinity (squares), which indicates a change inthe concentration of surface dipoles. The reaction of water with the surfaceof CuO can be expressed as follows (see also25):

OeðadÞ þH2O

gas þ 2CuCu þ hþ42�CuþCu �OHe

�þ SA (2.21)

A water molecule from the atmosphere (H2Ogas) reacts with preadsorbed

oxygen ions (OeðadÞ) and two Cu sites (2CuCu) on the surface under the for-

mation of two terminal hydroxyl groups�2�CuþCu�OHe

��. The appear-

ance of the two terminal hydroxyl groups is responsible for the increase inthe electron affinity (formation of local surface dipoles) and the cancellationof a hole (hþ) determines the decrease in the band bending. SA is the freedadsorption site for chemisorbed oxygen.

Consequently, there is competition between CO and H2O for oxygenions as reaction partners in the presence of humidity. This explains the

qΔV

Δχ∆Φ

∆E (e

V) –0.03

0.00

0.03

0.06

–0.06

–0.09

–0.12

–0.150 20 40 60 80

Relative humidity (%)

Figure 2.14 Changes of band bending (qDV), work function (DF), and possiblechanges of the electron affinity (Dc) of the CuO sensor on exposure to humidity levels(10%, 30%, 50%, and 70% relative humidity @25�C) at an operation temperature of150�C.


observed behavior in Fig. 2.13. The effect of CO exposure in humidconditions is reduced (smaller sensor signals); and the buildup of the dipolesis hindered due to fewer adsorption sites for water vapor, which determinesthe monitored decrease in the electron affinity.

This example demonstrates how one can identify the sensing mechanismof CO and CuO in the presence of humidity by using working conditionDC resistance and work function change measurements in combinationwith appropriate modeling of the conduction.

2.4.3 Conduction mechanism of n-type SnO2dexperimentalresults

The experiments were performed on an SnO2-based gas sensor operated at300�C. Investigations were made into the influences of CO and H2 indifferent oxygen backgrounds, and that of oxygen itself on the resistanceand the band bending.

In Fig. 2.15, the measured changes of the resistance and the CPD duringstepwise increasing oxygen concentrations from 0 (N2 atmosphere) up to2500 ppm are shown. One observes a steep increase in both resistance andwork function at the lower concentrations due to the adsorption of oxygen,resulting in ionosorbed oxygen species on the surface; a form of saturationoccurs as oxygen reaches 2000 ppm.

ResistanceCPD

Time (h)

CP

D (V

)

Res

ista

nce

(Ω)

1M

100K

300p

pm O

2

1000

ppm

O2

500p

pm O

2

2000

ppm

O2

2500

ppm

O2

100p

pm O

2

0.00

–0.03

–0.06

–0.09

–0.12

0 5 10 15 20 25

Figure 2.15 Simultaneous contact potential differences (CPDs) and electrical resistancechanges of an SnO2 sensordoperated at 300�Cdwith stepwise increasing oxygenamount (100, 300, 500, 1000, 2000, and 2500 ppm).

62 N. Barsan et al.

Fig. 2.16 illustrates the behavior of the resistance and the CPD on expo-sure to four CO pulses (10, 30, 70, and 100 ppm) in the absence(Fig. 2.16(a)) and in the presence (Fig. 2.16(b)) of 22,000 ppm of oxygen.A huge drop is observed in both the resistance and the work function dueto exposure to CO; except for the first pulse (10 ppm), the equilibrium statein the work function is reached in the allotted 3 h of CO exposure. Therecovery process, however, is very slow for both parameters; the baselinecould not be reached again in the 3 h allowed for recovery. The presenceof oxygen in the background determines a higher baseline resistance (forma-tion of ionosorbed oxygen) and a decrease in the relative changes of both theresistance and the work function in comparison with the absence of oxygen.The response and recovery times in the presence of oxygen are considerablymore rapid.

Fig. 2.17 illustrates the time dependence of the resistance and the CPD ofa similar experiment using five pulses of H2 (10, 20, 30, 50, and 100 ppm)instead of CO in the absence of oxygen (Fig. 2.17(a)) and in a backgroundof 22,000 ppm of oxygen (Fig. 2.17(b)). In the case of hydrogen, the drop inthe resistance and in the work function in the absence of oxygen is muchmore dramatic (the material becomes almost conductive). The responseand recovery times are more rapid compared with CO. In the presence ofoxygen, as expected, increases in the baseline resistance and the decreaseof the signals were observed.

Fig. 2.18 presents an overview of all the results obtained by simultaneousDC resistance and work function change measurements, including similar

ResistanceCPD

Time (h)

Res

ista

nce

(Ω)

Res

ista

nce

(Ω)

10M

1M

100k

10k

1k

100

10M

1M

100k

10k

1k

100

CP

D (V

)

CP

D (V

)

10pp

m C

O

30pp

m C

O

70pp

m C

O

100p

pm C

O

10pp

m C

O

30pp

m C

O

70pp

m C

O

100p

pm C

O

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.0

–0.1

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.0

–0.10 5 10 15 20 25 30 35 40

Time (h)0 5 10 15 20 25 30 35

(a) (b)

Figure 2.16 Simultaneous contact potential differences (CPDs) and electrical resistancechanges of an SnO2 sensordoperated at 300�Cdduring exposure to four pulses of CO(10, 30, 70, and 100 ppm) in the absence of oxygen (a) and in a background of22,000 ppm of oxygen (b).


experiments in a background of 200 ppm of oxygen. There, the resistanceand the corresponding band bending changesdextracted from the changesin the work functiondare plotted semilogarithmically. As a reference, thesituation in nitrogen was chosen (qDV ¼ 0). The existence of three differentareas with a seamless transfer in between each other is obvious. Each of themcan be accurately fitted by a proper exponential dependency of the resistanceand the corresponding band bending. The calculated slopes are ((1.0 � 0.1)kT)�1, ((2.35 � 0.12)kT)�1, and ((3.80 � 0.05)kT)�1, respectively.

ResistanceCPD

Time (h) Time (h)R

esis

tanc

e (Ω

)

Res

ista

nce

(Ω)

10M

1M

100k

10k

1k

100

10

10M

1M

100k

10k

1k

100

10

CP

D (V

)

CP

D (V

)

1.4

1.2

1.0

0.8

0.6

0.4

0.2

0.0

–0.2

1.4

1.2

1.0

0.8

0.6

0.4

0.2

0.0

–0.20 5 10 15 20 25 30 0 5 10 15 20 25 30

(a) (b)

10pp

m H

20pp

m H

30pp

m H

50pp

m H

100p

pm H

10pp

m H

20pp

m H

30pp

m H

50pp

m H

100p

pm H

Figure 2.17 Simultaneous contact potential differences (CPDs) and electrical resistancechanges of an SnO2 sensordoperated at 300�Cdduring exposure to five pulses of H2

(10, 20, 30, 50, and 100 ppm) in the absence of oxygen (a) and in a background of22,000 ppm of oxygen (b).

10M

1M

100k

10k

1k

100

10

Res

ista

nce

(Ω)

q∆V (eV)

O2H2 (0ppm O2)

H2 (200ppm O2)

H2 (22000ppm O2)

CO (0ppm O2)

CO (200ppm O2)

CO (22000ppm O2)

0.2 0.0 –0.2 –0.4 –0.6 –0.8 –1.0 –1.2 –1.4

Flat band situation

R ∼ expqVS

1.0kT

R ∼ expqVS

2.35kT

R ∼ exp qVS

3.80kT

Figure 2.18 Dependence of the resistance and the corresponding band bendingchanges (qDV). The flat band situation is denoted as the situation in a dry N2 back-ground. The three different regions/models are shown with the lines.

64 N. Barsan et al.

The trend observed in the experimental data is in line with the theoret-ical modeling presented in Section 2.3.2. The conduction moves from amechanism controlled by the presence of the depletion layer (theoreticalvalue: 1; experimental value: 1.0 � 0.1) to one controlled by transportthrough the accumulation layer where the Boltzmann statistics are still valid(theoretical value: 2; experimental value: 2.35 � 0.12) to the extreme case inwhich the Fermi level extends deep into the conduction band on the sur-face. It could be demonstrated in the theory26 that, in this area, the valueshould increase above 2, reflecting that the influence of the surface bandbending on the resistance decreases. This decrease is also supported by theexperimental/phenomenological parameter showing a value of 3.80 � 0.05.

The latter results correlate the measured resistances with the measuredchanges in the work function obtained from several individual measure-ments in different combinations of O2, CO, and H2. The fact that the exper-imental points are sitting on the same curve combining the individualconduction mechanisms indicates that the reactions involving these gaseshave very similar effects. The upward band bending is determining a con-duction mechanism dominated by the surface depletion layer; the down-ward band bending changes the conduction mechanism to one dominatedby accumulated electrons in the surface layer. In the latter case, the deeperthe conduction band edge on the surface falls below the Fermi level (higherconcentrations of CO and H2 in the absence of oxygen), the weaker the ef-fect of the band bending on the sensor signal becomes.

The switch from one conduction model to the other takes place directlyin a dry nitrogen atmosphere; this indicates that, under these conditions, aflat band situation is present (absence of active intrinsic surface traps).

The significance of the results presented is not only limited to the con-ditions used here. In higher oxygen backgrounds (synthetic air), one couldalso find resistance changes on exposure to reducing gases (high concentra-tions) of a few orders of magnitude, which could result in a switch betweenthe different conduction mechanisms. This could explain why it is some-times so difficult to describe the dependence of the full sensor responseon the target gas concentration over a large concentration range with a singlecurve. In the following chapter, experiments in more realistic conditionswere used to further examine such a possible switch of the conductionmechanism.28


2.5 Conduction mechanism switch for n-typeSnO2ebased sensors during operation in application-relevant conditions

Again for this set of experiments, similar SnO2 sensors as described inSection 2.4.1 were used at 300�C and exposed to the following conditions:N2; dry air; 0.3e250 ppm CO in dry air; humid air with 6% r.h. @25�C;0.3e250 ppm CO in humid air with 6% r.h. @25�C; humid air with20% r.h. @25�C; humid air with 50% r.h. @25�C, 0.3e250 ppm CO inhumid air with 50% r.h. @25�C. Fig. 2.19 represents the time dependenceof the resistance during exposure to different humidity levels and CO. Thebaseline resistance in pure N2drepresenting the flat band conditiondisindicated by a dotted line.

One can observe that during exposure to CO in humid air the resistancedecreases much below the value corresponding to the absence of oxygen(N2 background). This means a change in the conduction mechanismfrom the one controlled by the depletion layer to the one controlled bythe accumulation layer. The consequences are significant for both sensingcomponents: transduction and reception.

In the case of the transduction, it means that the conductance is no moreproportional to the surface concentration of free charge carriers but to theaverage one over the whole accumulation layer. This means that the samechange in the band bending will result in a reduced conductance change

1M

100k

10k

1k

1000

100

10

1

0,10,0

0 0 10 20 30Time [h]Time [h]

40 5020 40 60 80 100 120 140 160 180 200

Res

ista

nce

[Ω]

Con

cent

ratio

n C

O [p

pm]

1M

100k

Air

Air

CO

in d

ry a

ir

Air

6% r.

h.

Air

20%

r.h.

Air

50%

r.h.

CO

in a

ir 6%

r.h.

CO

in a

ir 20

% r.

h.

CO

in a

ir 50

% r.

h.

Undoped SnO2

Depletion layer

Accumulationlayer

N2

Figure 2.19 DC electrical resistance measurement of an undoped SnO2 sensor (poly-crystalline thick film sensing layer) during exposure to N2, and 0e250 ppm CO indifferent background conditions. Right: the profile of CO exposure as a function oftime, which is valid for all background conditions.

66 N. Barsan et al.

and hence a lower sensor signal in case of an accumulation layer (seeEq. 2.18).

2.6 Conclusion and future trends

The contribution presented here highlights the importance of theconduction mechanism in the SMOX sensing layers for the performanceof the corresponding gas sensors. It basically demonstrates that high surfacereactivity and the considerable charge transfer processes associated with it arenot sufficient for “large” sensor signals. These depend to a large extent onthe way in which the surface changes are translated into measurable changesof the electrical resistance of the sensor, which depend on the conductionmechanism.

The proposed conduction models, which are based on simple assump-tions and confirmed by the experimental results, explain the weaker perfor-mance of the devices based on p-type materials when compared with thosebased on n-type materials. They also open up new opportunities for inves-tigation in combination with working condition characterizationtechniques.

Future work will concentrate on applying the models for more compli-cated and realistic operational conditions in the direction indicated by theCuO investigation presented in Section 2.4.2. The understanding of theeffect of the presence of humidity in the ambient atmosphere as well asthe understanding of the effect of surface dopants and bulk doping are ofcrucial importance.

References[1] Barsan N, Koziej D, Weimar U. Metal oxide-based gas sensor research: how to? Sensor

Actuator B Chem 2007;121(1):18e35. https://doi.org/10.1016/j.snb.2006.09.047.[2] Brattain W, Bardeen J. Surface properties of germanium. Bell Telephone Syst Tech Publ

Monogr 1953;2086:1e41.[3] Bielanski A, Deren J, Haber J. Electric conductivity and catalytic activity of semicon-

ducting oxide catalysts. Nature 1957;179(4561):668e9. https://doi.org/10.1038/179668a0.

[4] Heiland G. Zum Einfluss von Wasserstoff auf die elektrische Leitf€ahigkeit von ZnO-Kristallen. Z Phys 1954;138:459e64. https://doi.org/10.1007/BF01327362.

[5] Seiyama T, Kato A, Fujiishi K, Nagatani M. A new detector for gaseous componentsusing semiconductive thin films. Anal Chem 1962;34:1502f. https://doi.org/10.1021/ac60191a001.

[6] Taguchi N. US patent No. 3631436. 1971.[7] Henrich VE, Cox PA. The surface science of metal oxides. Cambridge University Press;

1994. 0e521e44389-X.


https://doi.org/10.1016/j.snb.2006.09.047

https://doi.org/10.1038/179668a0

https://doi.org/10.1038/179668a0

https://doi.org/10.1007/BF01327362

https://doi.org/10.1021/ac60191a001

https://doi.org/10.1021/ac60191a001

[8] Barsan N. Conduction models in gas-sensing SnO2 layers: grain-size effects andambient atmosphere influence. Sensor Actuator B Chem 1994;17(3):241e6. https://doi.org/10.1016/0925e4005(93)00873-W.

[9] Barsan N, Weimar U. Conduction model of metal oxide gas sensors. J Electroceram2001;7(3):143e67. https://doi.org/10.1023/A:1014405811371.

[10] Dutraive MS, Lalauze R, Pijolat C. Sintering catalytic effects and defect chemistry inpolycrystalline tin dioxide. Sensor Actuator B Chem 1995;26(1e3):38e44. https://doi.org/10.1016/0925-4005(94)01552-S.

[11] Weimar U, Morante JR, Schweizer-Berberich M, Barsan N, Goepel W. Electrode ef-fects on gas sensing properties of nanocrystalline SnO2 gas sensors. In: Conference pro-ceedings EUROSENSORS XI, Warschau (Poland); 1997. ISBN 83-908335-0-6,1377e80.

[12] Oprea A, Barsan N, Weimar U. Work function changes in gas sensitive materials: fun-damentals and applications. Sensor Actuator B Chem 2009;142(2):470e93. https://doi.org/10.1016/j.snb.2009.06.043.

[13] Barsan N, Weimar U. Understanding the fundamental principles of metal oxide basedgas sensors; the example of CO sensing with SnO2 sensors in the presence of humidity.J Phys Condens Matter 2003;15(20):R813e39. PII S0953e8984(03)33587e8.

[14] Ihokura K, Watson J. Stannic oxide gas sensors. In: Principles and applications. BocaRaton: CRC Press; 1994.

[15] Williams DE. Semiconducting oxides as gas-sensitive resistors. Sensor Actuator B Chem1999;57(1e3):1e16. https://doi.org/10.1016/S0925e4005(99)00133e1.

[16] Kim YS, Hwang IS, Kim SJ, Lee CY, Lee JH. CuO nanowire gas sensors for air qualitycontrol in automotive cabin. Sensor Actuator B Chem 2008;135(1):298e303.

[17] Li Y, Liang J, Tao Z, Chen J. CuO particles and plates: synthesis and gas-sensorapplication. Mater Res Bull 2008;43(8e9):2380e5. https://doi.org/10.1016/j.materresbull.2007.07.045.

[18] Miremadi BK, Singh RC, Chen Z, Morrison SR, Colbow K. Chromium oxide gassensors for the detection of hydrogen, oxygen and nitrogen oxide. Sensor Actuator BChem 1994;21(1):1e4. https://doi.org/10.1016/0925e4005(93)01208-L.

[19] Shimizu Y, Nakashima N, Hyodo T, Egashira M. NOx sensing properties of varistor-type gas sensors consisting of micro p-n junctions. J Electroceram 2001;6:209e17.https://doi.org/10.1023/A:1011448513611.

[20] Zhang J, Liu J, Peng Q, Wang X, Li Y. Nearly monodisperse Cu2O and CuO nano-spheres: preparation and applications for sensitive gas sensors. Chem Mater 2006;18(4):867e71. https://doi.org/10.1021/cm052256f.

[21] Barsan N, Simion C, Heine T, Pokhrel S, Weimar U. Modeling of sensing and trans-duction for p-type semiconducting metal oxide based gas sensors. J Electroceram 2010;25(1):11e9. https://doi.org/10.1007/s10832e009e9583-x.

[22] Pokhrel S, Simion CE, Quemener V, Barsan N, Weimar U. Investigations of conduc-tion mechanism in Cr2O3 gas sensing thick films by ac impedance spectroscopy andwork function changes measurements. Sensor Actuator B Chem 2008;133(1):78e83.https://doi.org/10.1016/j.snb.2008.01.054.

[23] Huebner M, Simion CE, Tomescu-Stanoiu A, Pokhrel S, Barsan N, Weimar U. In-fluence of humidity on CO sensing with p-type CuO thick film gas sensors. SensorActuator B Chem 2011a;153:347e53. https://doi.org/10.1016/j.snb.2010.10.046.

[24] Morrison SR. The chemical physics of surfaces. New York: Plenum; 1977. ISBN0e306e30960e2.

[25] Huebner M, Pavelko RG, Barsan N, Weimar U. Influence of oxygen backgrounds onhydrogen sensing with SnO2 nanomaterials. Sensor Actuator B Chem 2011b;154(2):264e9. https://doi.org/10.1016/j.snb.2010.01.049.

68 N. Barsan et al.

https://doi.org/10.1016/0925–4005(93)00873-W

https://doi.org/10.1016/0925–4005(93)00873-W

https://doi.org/10.1023/A:1014405811371

https://doi.org/10.1016/0925-4005(94)01552-S

https://doi.org/10.1016/0925-4005(94)01552-S



https://doi.org/10.1016/S0925–4005(99)00133–1

https://doi.org/10.1016/j.materresbull.2007.07.045


https://doi.org/10.1016/0925–4005(93)01208-L

https://doi.org/10.1023/A:1011448513611

https://doi.org/10.1021/cm052256f

https://doi.org/10.1007/s10832–009–9583-x




[26] Barsan N, Huebner M, Weimar U. Conduction mechanisms in SnO2 based polycrys-talline thick film gas sensors exposed to CO and H2 in different oxygen backgrounds.Sensor Actuator B Chem 2011;157(2):510e7. https://doi.org/10.1016/j.snb.2011.05.011.

[27] Diéguez A, Romano-Rodríguez A, Morante JR, Kappler J, Barsan N, Goepel W.Nanoparticle engineering for gas sensor optimisation: improved sol-gel fabricatednanocrystalline SnO2 thick film gas sensor for NO2 detection by calcination, catalyticmetal introduction and grinding treatments. Sensor Actuator B Chem 1999;60(2e3):125e37. https://doi.org/10.1016/S0925e4005(99)00258e0.

[28] Barsan N, Rebholz J, Weimar U. Conduction mechanism switch for SnO2 base sensorsduring operation in application relevant conditions; implications for modeling ofsensing. Sensor Actuator B Chem 2015;207:455e9.




https://doi.org/10.1016/S0925–4005(99)00258–0


CHAPTER THREE

The effect of electrode-oxideinterfaces in gas sensoroperationSung Pil Lee1, Chowdhury Shaestagir21Kyungnam University, Changwon, Kyungnam, Korea2Intel Corporation, Hillsboro, OR, United States

Contents

3.1 Introduction 723.2 Electrode materials for semiconductor gas sensors 74

3.2.1 Metals and conduction 743.2.2 Influence of electrode materials 77

3.2.2.1 Silver 833.2.2.2 Gold 833.2.2.3 Platinum 843.2.2.4 Palladiumesilver 843.2.2.5 Platinumesilver 843.2.2.6 Platinumegold 843.2.2.7 Palladiumegold 85

3.2.3 Electrode configuration 853.2.4 Electrode geometry 90

3.3 Electrode-oxide semiconductor interfaces 953.3.1 Ideal contact of metal and oxide semiconductor 953.3.2 Contacts with surface states and an interfacial layer 993.3.3 Image force effects on the barrier height 103

3.4 Charge carrier transport in the electrode-oxide semiconductor interfaces 1043.4.1 Electric field and capacitance in the metal-semiconductor interface 1043.4.2 Transport mechanism across the junction barrier 1093.4.3 Tunneling effects in the oxide-semiconductor interface 1123.4.4 Structure of the interfacial layer 115

3.4.4.1 State 0: the clean semiconductor surface 1173.4.4.2 Stage 1: the dilute limit 1183.4.4.3 Stage 2: monolayer formation 1183.4.4.4 Stage 3: addition monolayers and interdiffusion 118



https://doi.org/10.1016/B978-0-08-102559-8.00003-3

3.5 Gas/solid interactions in the electrode-oxide semiconductor interfaces 1193.5.1 Dipole formation in the interfacial layer 1193.5.2 Effects of hydrogen adsorption in the Schottky barrier junction 1203.5.3 Adsorption of other gases in the Schottky barrier junction 122

3.6 Conclusions 124References 125

3.1 Introduction

There is a growing demand for gas sensors for efficient use of energyand raw materials, as well as to reduce environmental pollution despiteincreasingly complex manufacturing processes. Taguchi sensor is the mostwell-known gas sensor1,2 that detects reducing gases, whereas an oxygensensor based on an ion-conducting sensor is the second most famoustype.3e6 Research and development for gas sensors is conducted in twostages. The first stage is to develop a new sensor whose application is empir-ically optimized. The characteristics of sensitivity, selectivity, long-termdrift, and reliability are defined, although its operation mechanism is notfully understood.7e11 The second stage is to modify, optimize, andstandardize the system of the developed sensor.12e15 The developed sensorscan measure change or values of current,16,17 impedance,18,19 capacitance,20

frequency,21 potential difference,22,23 and electromotive force.24 In addi-tion, the correlation between the sensor structure and electrode is veryimportant to expressively depict these phenomenological parameters thatcharacterize the sensor.

The semiconductor gas sensor is not an energy conversion (generator-type) but energy control (modulator-type) sensor (see, e.g., Fraden, Hand-book of Modern Sensors). The physical properties of a sensing materialchange on exposure to gas molecules, and external electric energy transmitsthe change as a sensor signal. This implies that, in most cases, the electrodeof the semiconductor gas sensor is similar to that of an electronic device,which delivers current flow or electric power supply without loss or supplieselectric energy from external power sources to the device. Thus, in conven-tional electronic devices, the electrode only connects the device and externalcircuit. Accordingly, a strong mechanical adhesion and small contactresistance are the most significant factors; in addition, durability, chemicalresistance, reliability, and cost should be considered.25e28 However, theelectrode of a semiconductor gas sensor not only measures the electric

72 Sung Pil Lee and Chowdhury Shaestagir

properties of the sensor but also measures the catalytic properties of thesensing material. The ohmic electric contact made between the device andthe electrode material is acceptable; however, the semiconductor gas sensorsometimes requires a rectifying contact between the sensing material and elec-trode. A rectifying contact would create a dipole in the interfacial zone of ametal and semiconductor triggered by gas adsorption, reducing a potentialbarrier from time to time or leading to complex phenomena such as fieldemission or tunneling effect due to thermionic field emission.29e35 In specialcases where the semiconductor gas sensor is applied to cars or in the aerospaceindustry, the electrode material should be able to operate above 600�C.6,36

The surface and interface science for semiconductor gas sensors havebeen extensively studied since Seiyama et al.1 reported that the chargecarriers in the surface of oxide semiconductor in contact with a gas variedaccording to the gas concentration. In addition, gas sensing mechanism,37e39

gas sensor technology,40 the semiconductor junction for gas sensors,41

practical hydrogen sensors,42 and gas sensor design43 have been reviewedby several researchers.

The electrode materials and geometry have advanced considerably in thelast few decades. The physics of the energy barrier in an electrodeesemiconductor interface could be significantly compared with the energybarrier in the contact of a doped semiconductor. Charge transfer duringthe chemical reaction of gas in the electrode/semiconductor interfaces leadsto a uniform Fermi level instead of energy band bending. This chemicalreaction in the interfaces would affect the conductance of the sensor, aswell as chemical reaction in the semiconductor surface. To design reliablesemiconductor gas sensors requires the understanding of electrodeesemiconductor interfaces and control of the geometric and electronicstructures of electrodes.

The aim of this chapter is to describe and review the interface chemistryand transition theory of the electrode-oxide semiconductor layer in gassensor operation. Section 3.2 deals with criteria for selecting the metaland semiconductor materials used in the fabrication of gas sensor. Thechemistry and physics of barrier formation in the metal-oxide semicon-ductor interfacial layer are outlined in Section 3.3. The recent investigationsinto the charge carrier transport model, including the tunnel effect in theelectrode-oxide semiconductor interface, are discussed in Section 3.4.Section 3.5 surveys research and development works that have been under-taken on the gas/solid interactions in the electrodeesemiconductorinterfaces.

The effect of electrode-oxide interfaces in gas sensor operation 73

3.2 Electrode materials for semiconductor gassensors

3.2.1 Metals and conductionUnderstanding of the behavior of electrons in solids is one of the keys

to understanding electrode materials. The electron theory of solids is capableof explaining electrical, optical, magnetic, thermal, and chemical propertiesof materials. In other words, electron theory provides important fundamen-tals for a technology which is often considered to be the basis for moderncivilization. Electrical conduction involves the motion of charges in amaterial under the influence of an applied electric field. A material can begenerally classified as a conductor if it contains a large number of freeelectrons or mobile charges carriers. In metals, due to the nature of metallicbonding, the valence electrons from the atoms form a sea of electrons thatare free to move within the metal and are therefore called “conductionelectrons.” This is especially true for pure metals, where atom size and pack-ing are uniform and nothing is present to dissipate the free motion of elec-trons. Alloying disrupts the uniformity of the structure and reduces theelectrical conductivity. An increase in temperature also disrupts the structurebecause of lattice vibration and results in a decrease in electrical conductivity.

Good electrical conductors, such as metals, are also known to be goodthermal conductors. The conduction of thermal energy from higher tolower temperature regions in a metal involves the conduction electronscarrying the energy. Consequently, there is an innate relationship betweenthe electrical and thermal conductivities, which is supported by theory andexperiments.44

The conductivity, s, of different materials spans about 25 orders ofmagnitude, as shown in Fig. 3.1. This is a largest-known variation in aphysical property. It is generally accepted that, in metals and alloys, theelectronsdparticularly the outer or valence electronsdplay an importantrole in electrical conduction. Before making use of the electron theory,we need to remind of some fundamental equations of physics pertaining

10–18 10–16 10–14 10–12 10–10 10–8 10–6 10–4 10–2 102 104 106 1081

QuartzPorcelain

RubberNaCl Mica

Glass Si Ge MnFe

AgCuGaAs

Insulators Semiconductors Metals

σ [1/Ω·m]

Figure 3.1 Room temperature conductivity of various materials.


to electrical conduction. These laws have been extracted from experimentalobservations.45

Ohm’s law

V ¼R$I (3.1)

relates the potential difference V (in volts) with the electrical resistance R (inohms) and the electrical current I (in amperes). A differential form of Ohm’slaw is

J ¼ s$E (3.2)

which links current density, J ¼ I/Adi.e., the current per unit area (A/m2),with conductivity s and electric field strength

E¼VL

(3.3)

The resistance of a conductor can be calculated from its physical dimen-sions by

R¼LrA

(3.4)

where L is the length of the conductor, A is its cross-sectional area, and r isthe specific resistance or resistivity. The conductivity is in inverse proportionto the resistivity:

r¼ 1s

(3.5)

Fig. 3.2 shows the net flow of electrons in a conductor cross-sectionalarea A in the presence of an applied field Ex. Notice that the direction ofelectron motion is opposite to that of the electric field Ex and of conven-tional current because the electrons experience a Coulombic force eEx, inthe x direction, because of their negative charge. We know that the conduc-tion electrons are actually moving around randomly in the metal, but we

--

- --

-

-

--

--

-

-

--

--

-

---

-dxν

Ix

Δx

Ex

A

-

Figure 3.2 Drift of electrons in a conductor in the presence of an applied electric field.


will assume that, as a result of the application of the electric field Ex, they allacquire a net velocity in the x direction. Otherwise, there would be no netflow of charge through area A.

The average velocity of the electrons in the x direction at time t isdenoted as vdx(t). This is called the “drift velocity,” which is the instanta-neous velocity vx in the x direction averaged over many electrons,(w1028 m�3); that is

vdx¼ 1N

½vx1þ vx2þ vx3þ , , , þ vxN � (3.6)

where vxi is the x direction velocity of ith electron and N is the number ofconduction electrons in the metal. Suppose that n is the number of electronsper unit volume in the conductor (n ¼ N/V). In time Dt, electrons move adistance Dx ¼ vdxDt, so the total charge Dq crossing the area A is enADx.This is valid because all the electrons within distance Dx pass through A;thus, n(AvdxDt) is the total number of electrons crossing A in time Dt.

The current density in the x direction is

Jx¼ DqADt

¼ enAvdxDtADt

¼ envdx (3.7)

This general equation relates Jx to the average velocity vdx of theelectrons. It must be appreciated that the average velocity at one timemay not be the same as at another time because the applied field, forexample, may be changing: Ex ¼ Ex(t).

JxðtÞ¼ envdxðtÞ (3.8)

To relate the current density Jx to the electric field Ex, we must examinethe effect of the electric field on the motion of the electrons in theconductor. To do so, we will consider copper crystal. The copper atomhas a single valence electron in its 4s subshell, and this electron is looselybound. The solid metal consists of positive ion cores, Cuþ, at regular sitesin the face-centered cubic crystal structure.2 The valence electrons detachthemselves from their parent atoms and wander around freely in the solid,forming a kind of electron cloud or gas. These mobile electrons are freeto respond to an applied field, creating a current density Jx. The valenceelectrons in the electron gas are therefore conduction electrons. If theelectric field strength is not very high, then the drift velocity of conductionelectrons is proportional to the electric field strength, vdxðtÞ ¼ mExðtÞ (m emobility), and the Ohm’s law (3.2) follows from Eq. (3.8) with the conduc-tivity given by s ¼ enm.


3.2.2 Influence of electrode materialsMany researchers have long studied the interaction between the electrodeand sensor materials, as well as the impact of the electrode materials onthe sensing behavior.46e59 The types of electrode materials used for semi-conductor gas sensors are classified into bulk, thick film, and thin film.The bulk type is rarely used for the semiconductor gas sensor. The thickfilm type, made by screen printing from a paste, and the thin film type,made by vacuum deposition, are employed in many cases. The impact ofthe electrode on the properties of gas sensors based on tin oxide has beenstudied mainly by comparing various electrode materials such as Au, Pt,Ag, and Pd.29,46e54

Toohey60 summarized the study on the interaction between electrodesand sensor materials and the influence of electrode materials in sensingbehavior. The most common electrode material in practical and experimentalsensors appears to be platinum, although gold and silver are occasionally used.Chemically, Pt and Au are relatively inert. As pure metals, they can besputtered or evaporation coated onto a substrate, and both are available inink formulations for screen printing. Ball or wedge bonding with platinum,gold, or aluminum wire allows the sensor device to be packaged in aconventional semiconductor header. A study of SnO2 and SnO2eMn2O3

hydrogen sensors with gold, palladium, and platinum electrodes showedthat changing from platinum to gold could produce many-fold increase insensitivity and a shift in peak sensitivity temperature from w375 tow450�C (Fig. 3.3). Also, while the pure tin dioxide sensors had linear IeVcharacteristics under all conditions, the mixed oxide devices showed nonlin-earity for high hydrogen concentrations with palladium or gold electrodes,but did not so when platinum was used. This suggests that the electrodeesensor contact was an appreciable component of the total sensor impedance.47

0

200 250 300 350T (°C)

400 450 500 550

20

40

60

80 Pd

Pt

Au

RN

2/Rg

100

120

140

Figure 3.3 Sensitivity influence of three electrodes on a SnO2eMn2O3 (10:1) sensor forhydrogen gas.


Capone et al.46 analyzed the impacts of two different interdigitatedelectrode geometries on the sensitivity of two different electrode materials(Au and Pt) for CO gas. These studies revealed that the Au electrode hada lower stability level than the Pt electrode. With regard to temperature,the sensitivity of CO was the highest at approximately 300�C for the Auelectrode and at 450�C for Pt electrode, and it was observed to decreaseslightly at low temperatures. In addition, a pure tin oxide sensor hasdisplayed linear currentevoltage properties under all conditions, whereas asensor with an additive has shown nonlinear properties. The Pd and Auelectrodes had nonlinear characteristics, but the Pt electrode had linearcharacteristics for high hydrogen concentrations. These studies reportedthat the electrodeesemiconductor contact exerts substantial influence onthe entire sensor impedance.47

Saukko et al.48 studied the influence of electrode materials on theproperties of a tin oxideebased gas sensor. The energy barrier betweenthe electrode and the sensing semiconductor could be significant comparedto the energy barriers between the semiconductor grains. Then, chemicalreactions between the gas atmosphere and metalesemiconductor interfacewould strongly affect the overall conductance of the sensor. When SnO2

thick film gas sensors that use Au and Pt as electrode materials were testedfor hydrogen and CO gas, it was observed that the Pt electrode was moresensitive to H2, whereas the Au electrode was more sensitive to CO.Durrani49 used Ag, Al, Au, and Pt to study the effect of electrode materialon the SnO2-based CO thin film gas sensor. Pt and Au showed higherresponse than Ag or Al when the electrode material was below the sensingmaterial. In addition, Gourari et al.,47 Pijolat,50 and Bertrand et al.51 havestudied Pt and Au as electrode materials in SnO2 gas sensors.

Schottky-type sensors, in which the metal and semiconductor are incontact, are most widely used as hydrogen sensors. When the gas is notadsorbed in Schottky-type sensors, the energy band of the semiconductorbends upward or downward by the difference in the Fermi level betweenthe metal and the semiconductor in the thermal equilibrium state. Such asituation arises when there is a thin insulator layer between the metal andsemiconductor as well.28,42 In general, Pd is used as the electrode materialfor Schottky-type H2 sensors. When hydrogen molecules adsorb onto Pd,which is a catalytic metal, they dissociate into hydrogen ions. Some of theseions permeate Pd, spread toward the metalesemiconductor interface, formdipoles, and then change the metal’s work function and, hence, the Schottkybarrier height. This change in the Schottky barrier height causes a shift in the


currentevoltage (IeV) characteristics and thus the response can be measuredas the change in voltage when the diode is operated at constant bias current.

In metaleinsulatoresemiconductor field effect transistor (MISFET)etypehydrogen sensors, the threshold voltage in the gate layer changes based on thehydrogen concentration, resulting in a change in the drain current. Hydrogensensors that use Schottky diodes were proposed for the first time by Lund-strom et al.61 and Steele and MacIver59 Both diodes used palladium as themetal, and the semiconductor substrates used were n-Si and CdS, respec-tively. In 1979, Ito62 had predicted that Schottky diodes consisting of similarmetals and ionic semiconductors (such as SnO2, In2O3, KTaO3, ZnO, etc.)would also be sensitive to hydrogen. Comparative studies between Schottkydiodes using Pd and Pt as the catalytic metals indicated the superior perfor-mance of Pt in terms of speed of response and sensitivity to hydrogen.63,64

In addition to Pd and Pt, other hydrogen-sensitive metals and alloys hadbeen proposed including Ru,65 Ni,66 Au,67 Ag,68 IrPt, and PdAg.69 Songet al.69 tested the response to hydrogen gas of AlGaN/GaN Schottky diodeswith Pt, IrPt, and PdAg from 200 to 800�C. From 200 to 300�C, PdAgdiodes exhibited significantly higher sensitivity compared with Pt and IrPtdiodes. Above 400�C, however, IrPt and Pt diodes showed higher sensitivity,while the sensitivity of PdAg diodes degraded because of the poor thermalstability (Fig. 3.4). Studies on the electrode effects of semiconductor gassensors are summarized in Table 3.1.

One of the main disadvantages regarding the metal oxideebased gassensors is a gradual loss of stability and reliability: the problems of agingand drift of the sensors. Important factors in selecting an electrode material

0200 600400

Temperature (°C)800

1

2

Sen

sitiv

ity (S

) 3

PtIrPtPdAg

4

Figure 3.4 The comparison of hydrogen sensitivity in AlGaN/GaN Schottky diodes withdifferent catalytic metals.


for a gas sensor include the long-term stability, heat resistance, chemicalresistance, and adhesion to a substrate. Long-term investigations willdetermine the usability of the sensors. According to Meixner and Lampe73

the main reasons for inadequate long-term stability are the changes of themetal oxide and the metal electrode, instability of the wire contacts, andinteraction with an unsuitable sensor casing.

The degradation of contacts is mainly due to the diffusion at theelectrode and oxide interface or the interaction of electrode with thesurrounding atmosphere.74 As an electrode material for semiconductor gassensors, Ag is stable in air and used over a wide temperature range. However,Ag has a low long-term stability disadvantage and the degradation ofcontacts. Ag can easily move or migrate at temperatures above 300�C. Auis also one of the most popular electrode materials owing to its high electricconductivity and reliability. However, it has the disadvantage of easilydiffusing into the substrate (especially silicon) at a relatively low temperature.On the other hand, Pt is the most stable electrode material, with littledegradation. However, it is expensive and has poor substrate adhesion. Toimprove the adhesion to the substrate, a “glue layer” of Cr, Ti, or W isneeded between the electrode and the substrate. For good adhesion, Hoefer

Table 3.1 Studies for electrode effects of semiconductor gas sensors.Electrode materials Sensing materials Target gases References

Au, Pd, Pt SnO2 H2 Gourari et al.47

Pd, Pt ZnO, SnO2, In2O3,KTaO3

H2 Ito62

Ru SiC H2 Basu et al.65

Ni Si H2 Salehi and Nazerian66

Au ZnO H2 Pandis et al.67

Pt, IrPt, PdAg Al GaN-GaN H2 Song et al.69

Au, Pt SnO2 H2, CO Saukko et al.48

Au, Pt SnO2 H2, CO Rank et al.53

Au, Pt SnO2 CO Capone et al.46

Ag, Al, Au, Pt SnO2 CO Durrani49

Au, Pt SnO2 CO Bertrand et al.51

Au Fe2O3eIn2O3 CO Golovanov et al.70

Ag, Au ZnO CO, NO2 Lin et al.55

Au WO3 NO2 Tamaki et al.71

Au SnO2 NO2 Shaalan et al.72

Au, Pt SnO2 Benzene Pijolat50

Pt, Au, PteAu SnO2 H2O Ylinampa et al.52

Al WO3 Cl2 Bender et al.56


et al.75 used Ta, whereas Michel et al.76 used TiN as the glue layer. Sozzaet al.77 also reported that the Ti/Pt layer can prevent the rather fastdegradation as compared to the Ti/Au layer or the Ti/Pd/Au layer. Caponeet al.78 studied the influence of the aging of the Ti/Au interdigitated elec-trical contacts on the responses of pure and Ni-, Os-, Pt-, and Pd-dopedSnO2 thin films. They found that the use of Ti/Pt electrical contacts, whichwere more stable than Ti/Au or Ti/Pd/Au structures, could reduce one ofthe possible causes of aging that produced the drift of the sensor responses.

Some semiconductor gas sensors use a conductive polymer as theelectrode material. Most organic polymers are electrically nonconductive,but conductive polymers can be produced by providing a channel forelectrons to travel along polymer chains or to jump from chain to chain.79

Such conductive polymers include polyaniline, polyacetylene, polypyrrole,poly(p-phenylene), polythiophene, and poly(p-phenylenevinylene), amongwhich polyaniline (PANI) is the most widely used.80e82 Fig. 3.5 showsexamples of conductive polymers. Polyaniline has received significant atten-tion as it has a high electrical conductance of 103 S/cm and has beenreported to have metallic properties. According to the synthesis method,polyaniline can be divided into the following states (Fig. 3.6):(i) completely oxidized state (PB: 1 � y ¼ 0, quinoid); (ii) intermediateoxidation state (EB: 1 � y ¼ 0.5); and (iii) completely deoxidized state(LB: 1 � y ¼ 1, benzenoid). EB is generally easily produced using anoxidizer such as (NH4)2S2O8 to oxidize aniline directly in the presence ofprotonic acid. LB can be easily obtained by applying a reducing agentsuch as hydrazine hydrate to EB, whereas PB can be produced using anoxidant such as m-chloroperoxybenzoic acid.83,84

Polyaniline Polyacetylene

Polypyrrole Poly(p-phenylene)

Poly(p-phenylene vinylene)Polythiophene

HN

N

n

n

n

H

SCH CH

Figure 3.5 Various conductive polymers.


Conductive polymers have a high conductance of approximately 103 S/cm as compared to that of an ITO electrode. Thus, they can enable theproduction of thin films through spin coating, which is much moreeconomical and convenient than evaporation or sputtering. However,conductive polymers have some disadvantages, such as a property changeduring the production of doped polymer composition, the use of a nonvol-atile solvent (m-cresol), and their color. For realizing a flexible sensor systemin future, the metal electrode materials must be replaced by organicmaterials. Among organic materials, monomolecular pentacene has thehighest level of charge transfer.80 Pentacene, however, has a disadvantagein terms of its manufacturing process, such as it is impossible to affect vacuumevaporation. Polythiophene derivatives are used as conductive polymers toreplace pentacene, and they have high electric field mobility; however, theyshow a relatively low on-and-off ratio. Polyaniline and polypyrrole alsohave a low on-and-off ratio, but the ratio can be enhanced because theconductivity level of nanostructured polyaniline can be more easily adjustedthan the doping level as compared to polymers. The replacement of theelectrode material, which is the core part of flexible devices, is veryimportant. Many experts expect that when the replacement technology isaccomplished, the development of flexible devices will reach the stage ofcompletion. Polyaniline can also be applied to a variety of fields, such asthe electrode material of sensors, insulation layer of O-TFT, and channelmaterial of an electrical transport layer. Notably, for polyaniline, the electri-cal conduction can be much more easily controlled than for carbon nano-tubes by adjusting the protonic acid doping levels and by otherappropriate methods.80

H HN N

y 1–yNN

N N N N

HN

HN

HN

HN

Leuco-emeraldine Base (LB)

Emeraldine Base (EB)

Pernigraniline Base (PB)

Figure 3.6 Polyaniline bases according to oxidation states.


Recently, the field of printed electronics has been receiving considerableattention because of the development of semiconductor fabrication technol-ogy.85 It is important for semiconductor gas sensors to make electrodes byprinting the materials. The most commonly used materials for electrodesare precious metals, such as silver, gold, platinum, and palladium. The alloysof these metals are also widely used.

3.2.2.1 SilverSilver inks were probably the first thick film inks to be developed and wereused mainly in the construction of capacitors. There are several points infavor of silver as a conductive ink. Not only is it the least expensive metalwhich is compatible with the normal thick film process, but also it maybe made to have good bond strength and high conductivity and is easilywetted by tin-lead solder. Although its leach resistance to this solder ispoor, this is easily overcome by slight modification of the solder. Silvercompositions are also compatible with several resistor and dielectric systems.The major disadvantage of silver is its strong tendency to migrate over thesurface of insulants and resistors when subjected to electrical fields underconditions of high humidity; this may lead to the lowering of resistanceor complete short circuits in other thick film components.

3.2.2.2 GoldGold inks may be constituted to have high conductivity, similar to that ofsilver, and produce films which are stable under all normal service condi-tions. They are compatible with most dielectric and some resistor systems,although they are not generally suitable for terminating the palladium/silvertype of resistors. The main disadvantages of gold are its high cost and itsunsuitability for solder joining. Tin-lead solders are unsuitable for usewith gold conductors and cannot easily be modified, as is the case of silver,although special solders such as gold-tin alloys may be used. Gold is normallyused in circuits where high conductivity and reliability are required and inapplications where silicon devices are to be eutectically bonded, or whereultrasonic bonds have to be made, to gold or aluminum wires. A furtheruse of gold is in the closure of hermetic packages, where lids may be sealedto gold metallization using a solder alloy consisting of gold (80%)/tin (20%).The advantage of gold in this situation is that, in a neutral atmosphere, thesolder seal may be made without the use of flux.


3.2.2.3 PlatinumPlatinum inks are the most expensive of thick film conductive inks, but theyare occasionally used where extreme resistance to molten solder and to boldstrength degradation by solder is required.

3.2.2.4 PalladiumesilverPalladiumesilver alloys are perhaps the most widely used conductor compo-sitions. They are less expensive than gold alloys, are compatible with mostdielectric and resistor systems, and are suitable for ultrasonic wire bonding.The sheet resistivity is typically in the region 0.01e0.04 U/sq and, althoughthis figure is considerably higher than the resistivity of pure metal conduc-tors, it is lower than the figures for gold alloy conductors. The addition ofpalladium to silver greatly reduces the rate of dissolution of the metal inmolten solder. Increasing the palladium content thus provides greater leachresistance, but at the expense of solderability and conductivity. It alsoincreases the cost. It is common practice, therefore, for ink manufacturersto produce a range of palladiumesilver alloys of different palladium contentso that the best compromise may be chosen for any particular application.Palladiumesilver pastes can be fired with excellent initial bond strength tothe substrate, but this rapidly degrades if the circuits are stored at elevatedtemperature (above 70�C) when the conductors are tinned. A furtherdisadvantage is the possibility of silver migration under conditions of highhumidity. The rate of migration is, however, considerably reduced by thepresence of the palladium.

3.2.2.5 PlatinumesilverIncreasing world demand for palladium and wild fluctuations in its pricehave recently persuaded some ink manufacturers to add platinumesilveralloys to their range of conductors as alternatives to palladiumesilver. Sheetresistance ranges from 0.01 to 0.04 U/sq with increasing platinum content,and in this and most other respects the two ranges are found to be generallyequivalent. Platinumesilvers are not, however, recommended for hybridapplications involving ultrasonic aluminum wire bonding.

3.2.2.6 PlatinumegoldPlatinumegold systems possess many of the advantages of both gold and plat-inum. They have excellent solderability combined with outstanding resistanceto solder leaching and are also suitable for both wire and die bonding. Theyare compatible withmost other thick filmmaterials and, while the initial bondstrength tends to be lower than that of palladiumesilver alloys, they have


much greater resistance to solder bond strength degradation. The chief disad-vantages lie in their high cost and rather high electrical resistivity(0.08e0.1 U/sq).

3.2.2.7 PalladiumegoldThis alloy system has generally similar properties to platinumegold and isless costly. The solder leach resistance and solder aging are, however, inferiorto the more expensive material. Resistivity is of the order 0.04e0.10 U/sq.Table 3.2 summarizes the properties of metal inks for semiconductor gassensors.

3.2.3 Electrode configurationThe electrodes used for gas sensors should be in contact with the substrates,and their electrical properties should be easily measured. The followingconditions are thus required:(1) They should be chemically and mechanically stable on the substrates.(2) The connection to the lead-out terminals should be easy.(3) The sensing film should not be damaged during electrode formation.(4) They must have a geometry that is suitable for sensor construction.

The two-electrode type configuration, in which the gas sensing materialis positioned between two metal electrodes, is the most widely applied onein semiconductor gas sensors. Occasionally, a third electrode is used as aheater for the sensors. Toohey60 explained the electrode types used in semi-conductor gas sensors. Two-electrode configurations are used for gassensors, as shown in Fig. 3.7. In type (a), the Pt electrode is formed on analumina cylinder, which is applied to a Figaro sensor, and then the sensingmaterials are deposited on it and sintered. In type (b), a tablet made of anoxide semiconductor is sintered and then the electrode is formed on bothsides. In type (c), two combs face each other to create an interdigitatedgeometry on the substrate. The transmission type (d) sensor is formed, tofabricate a surface acoustic wave filter that measures the frequency changes.The interdigitated geometry is the most widely accepted geometry for theelectrodes of a gas sensor, as it enables a wide contact area between theelectrodes within the limited area. In addition, it forms the electrodes firstand then deposits the sensing materials on them, thereby causing no damageto the sensing materials.

A one-electrode configuration that differs from the two-electrode typethat has been previously used for semiconductor gas sensors has been devel-oped. Korotcenkov43 reviewed the design and type of the one-electrode


configuration for semiconductor gas sensors. One electrode acts as both theheater and the measuring terminal, unlike the two electrode setup. Asdemonstrated in Fig. 3.8, one-electrode gas sensors can be formed byapplying the metal oxide in the form of a bead on the electrode material

Table 3.2 Properties of printing metals and alloys for electrode of semiconductorgas sensors.Materials Electrical properties Advantages Disadvantages

Silver - High conductivity- Compatible with

resistor anddielectric system

- Resistivity:1.59 ⅹ 10�8 U m

- Least expensive- Good bond strength

- Tendency tomigrate over thesurface of insulantsand resistors underhigh humidity

Gold - High conductivityand reliability


- Alloy with tin maybe made withoutthe use of flux

- High cost- Unsuitability for

solder joining

Platinum - Use where extremeresistance to moltensolder and to bondstrength degradationby solder is required


- Available wire, flatplate, and tube

- Large range of size- Useable in high

temperature

- Most expensive

Palladiumesilver

- Compatible withresistor anddielectric system

- Sheet resistance:0.01e0.04 U/sq

- Suitable forultrasonic wirebonding

- The possibility ofsilver migrationunder highhumidity

Platinumesilver

- Alternatives to PdeAg


- Not recommendedfor hybridapplicationsinvolving ultrasonicwire bonding

Platinumegold

- Compatible withmost thick filmmaterials


- Excellentsolderability

- Suitable for bothwire and diebonding

- High cost- Rather high

electrical resistivity

Palladiumegold

- Similar properties toPteAu


- Less expensive thanPteAu

- Inferior solder leachresistance and solderaging than PteAu


or by shunting the electrode through a coating. For the design ofone-electrode semiconductor sensors, materials such as SnO2,

86e89

In2O3,90 and Fe2O3

70,91 have mainly been used. Fig. 3.9 shows schemati-cally a standard two-electrode sensor and one-electrode sensors in planardesign. The schematic electrical circuits of these sensors are also shown inFig. 3.9, where RPt is a coil resistance of Pt spiral, RMeO is interturnresistance of metal oxide ceramics, and RS is a total resistance of the sensor.For operation of the one-electrode sensors, impedance matching should beperformed between the shunting semiconductor resistance and the electroderesistance, as shown in Fig. 3.9. One can adjust the electrode resistance byvarying the electrode thickness, distance between the electrodes, orresistance of the sensing materials by adding additives to the oxide semicon-ductor or by modifying the thickness.43

Electrode Electrode

Alumina tubeCatalyst

CatalystPt wire

Pt wire Wire

Wire

Heater

Heater

Sensingmaterial

Sensor electrode

Heater electrode

Heater

Contact padSensingmaterial

SubstrateInterdifitedelectrode

Signal

Sensingfilm

Output

Interdigitedelectrode

Piezoelectricsubstrate

(a)

(d) (e)

(b) (c)

Figure 3.7 Two electrode configuration used in gas sensors: (a) cylinder, (b) disk,(c) parallel plates, (d) interdigit, and (e) surface acoustic wave (SAW) line.

Bonding pad

Ceramic bead

Supporter

Lead wire

Pt heater(electrode)

ElectrodeSensing material

(a) (b)

Figure 3.8 One-electrode configuration: (a) ceramic bead surrounding Pt electrodeand (b) Pd electrode on alumina substrate.


Faglia et al.92 used four-probe array analysis in the gas detection system todistinguish between the grain contribution and the contact contribution.This suggested that the contact contribution was very important for COdetection, while the material contributes to CH4 detection in tin oxidegas sensors. For four-electrode semiconductor sensor design, CrTiO3

93

and WO3/TiO294 have mainly been used.

It is clear that the material and geometry of the electrodes can influencegas sensor behavior. Many researchers are investigating the fabrication of gassensors using nanoparticulate materials as the sensitive layer. While it ispossible to use “normal”-sized electrodes with widths and separations ofseveral microns for these devices, it is of interest to examine the changesin response which are obtained when nanoelectrodes are used; i.e., contactsof comparable dimensions to a single particle around 5 nm. Potential advan-tages of nanoelectrodes include the following60:(i) the possibility of addressing single nanodots;(ii) the ability to vary the relative contributions of electrodeedot and

dotedot contacts to the total sensor resistance;(iii) where a nanodot film consists of conducting and nonconducting

particles, decreasing the electrode size could increase sensitivity byaround an order of magnitude or more by “softening” the percolationthreshold;95

(iv) small electrode systems use less sensor chip area; however, producingstructures of these sizes is problematic.

Conductive gassensing metal oxide

Conductive gassensing metal oxidePt contact Pt planar heater

I (const)

Pt planar heaterDielectric substrate Dielectric substrateCapsulating layer

RMeORMeO RS

RPtRPt

Standard solid stateconductometric sensor

One-electrodesemiconductor sensor

(a) (b)

Figure 3.9 Planar constructions of standard (two-electrode) and one-electrode semi-conductor sensors.


As all electronic parts become integrated and intelligent, it is also inevi-table to make small and integrated gas sensors. So far, many researchers haveused conventional MISFETs61,96e105 or microelectromechanical systems(MEMSs)106e112 to manufacture semiconductor gas sensors. Gas detectionwith such technologies depends on the varying conductivity owing to gasadsorption and the reaction on the metaleinsulatoresemiconductor (MIS)structure surface or varying work functions of the MISFET resulting fromcatalytic reaction in the gate electrode. However, sensor stability is notensured yet, although MIS gas sensors are increasingly needed. As the gateelectrode is exposed, unintended reactions between the gate electrode andmaterials near it reduce the sensor sensitivity or selectivity with time andit takes longer to respond to the gas molecules.

The gate electrode of the CO gas sensors with the MIS structure needs toapply a voltage for the device so as to form a channel and also carry outcatalytic actions.105 Thus, the electrode can be made porous so that thearea where the adsorbed gases contact the sensing materials increases insteadof the gate covering the surface of the sensing materials. Janata andJosowic113 created a suspended microgrid on a FET gate to extend thelifetime of the sensing gas. In these devices, the gate metal is preceded byan additional space, which, in the case of GasFET, is permeable to gases.The suspended grid above the gate insulator is made of Pt or Au. Applyinga Pd layer to this creates a hydrogen sensor. If a conductive polymer layer,such as polypyrrole, is deposited on the metal grid, then the sensor is sensi-tive to alcohols. In both cases, the reaction of the gas with the surface of thesuspended metal grid or with the surface of the insulator causes a change inthe electric field that is detected in the modified drain current.114 Leeet al.17,115 have tried to deposit a porous metal gate for humidity sensitivefield effect transistors (HUSFETs) that can sense humidity. Here, a thingold film of approximately 100 Å through which water molecules couldpenetrate was deposited on the active layer before a pattern was formedusing lift-off techniques. When water molecules meet carbon nitridethrough the porous gold layer of the gate in Fig. 3.10, adsorbed watermolecules on the carbon nitride are able to form dipole and to reorient freelyunder an applied gate voltage, resulting in an increase in the dielectricconstant.17 Thereafter, Fukuda et al.105 applied porous Pt as gate electrodematerials to improve the sensitivity of the MOSFET-type hydrogen sensors.When a porous electrode was used, the sensor detected 22 ppm of H2 gas inless than 2 min, thus indicating a remarkable gas detecting performance. Itssensitivity level was enhanced by approximately 10 times as compared to


that of a nonporous Pt surface because of the catalytic property of the porousPt surface. For the purpose of detection of negative ions in the air, Leeet al.116 have suggested a nanoFET sensor that uses a TieAl layer as theelectrode for the source and drain, while using a floated Ti/Au layer asthe electrode on the gate oxide.

3.2.4 Electrode geometryMany researchers have studied the influence of the geometry and position ofelectrodes on the sensitivity and selectivity of sensors.60,71,72,117e121

The width of digits in interdigitated electrodes or the space between theelectrodes can affect the sensor performance. In other words, when theelectrode spacing is narrow, the current between electrodes flows only inthe film area right above it. On the contrary, when the spacing is wide,the current flows both horizontally and vertically throughout the film,thereby sampling a wider area.60,118 In addition, the electrodeesemiconductor interface itself can cause a change in the device sensitive resis-tance. When the width/gap ratio of the electrode is changed, the influenceof both the interface and the film resistance on sensitivity can be relativelyreduced.

Sensor

Reference

Reference passivation

Source padMental 2Mental 1

Gate pad

Drain pad

OxideOxideOxide

Source

Si3N4

Drain

SiO2

CNxPorous Au

N+

N+

N+

N+

Locos

Figure 3.10 Design of differential humidity sensitive field effect transistors with porousAu gate.


Vilanova et al.119 studied the influence of electrode position, electrodegap, and active layer thickness for high, medium, and poor catalytic activitysensor/gas pairs. The purely geometric effect arises because the film conduc-tance does not change instantly or uniformly when the gas ambient changes:the gas must diffuse through the film, reacting with the particle surfaces as itdoes so. A numerical simulation indicated, for example, that where a sensoris highly sensitive to the test gas, sensitivity increased with electrode spacingwhen the electrodes were underneath the film but decreased with spacingwhen the electrodes were deposited on top of the film. In contrast, whenan electrode was placed above the sensing film, the sensor sensitivitydecreases as the spacing between the electrodes increased. The result ofinjecting a highly reactive gas was same as the result of injecting a lowreactive gas when the gap between the electrodes decreased and becamesmaller than the film’s thickness. Fig. 3.11 shows that the sensitivity dependson electrode spacing for a sensor whose electrode is placed below the sensorfilm. In this case, the detection level for even a highly reactive gas wasobserved to be the same as that of a low-reactive gas. On the contrary,when the electrode gap is sufficiently wide, the detection level of even alow-reactive gas was observed to be the same as that of a highly reactivegas. Therefore, the gas detection performance of a sensor with an electrodeplaced below its film is better when the electrode width is wide andelectrode spacing is narrow.

1E–011E–02

1E–01

1E+00

1E+01

1E+02

1E+03

1E+04

1E+00 1E+01 1E+02 1E+03

Bottom

ΔG/G

0

W (μm)Figure 3.11 Sensitivity versus electrode gap for electrodes placed bottom.


Gardner120 derived expressions which defined the response of a pair ofplanar conductometric gas sensors according to the electrode thicknessand an electrode gap. The steady-state conductance in air, Go, of a homo-geneous film of conductivity so and thickness L lying on semiinfiniteelectrodes can be found by integrating the current density over a closedsurface; hence,

G0¼ s0bp

ln

2641þ

�1þ w2

L2

�1=2

w2L

375 (3.9)

where w is the separation of the electrodes and b is the length of theelectrodes. It is assumed that the edge effects can be neglected (b [ w).When a gas of concentration C0 is introduced, it can diffuse into the porousfilm and react at sites dispersed uniformly throughout the film. Thesereaction sites modify the local conductivity of the film according to somefunction that depends on the local gas concentration Cx. The steady-stateresponse R (fractional change in conductance) of the sensor by integratingthe concentration-dependent current density can be given by

R¼GðCxÞ �Go

Gop¼ sðCxÞ � so

so

¼R x=L¼1x=L¼0 FðCxÞ

h�xL

�2 þ� w2L

�2i1=2d�xL

�

ln

("1þ

�1þ w2

4L2

�1=2#,� w

2L

�) (3.10)

In the case of the narrow-gap sensor, the baseline conductance Gon andthe response of the narrow-gap sensor Rn become

Gon¼ sobp

ln

�4Lwn

�

where wn/L � 1

Rn¼R x=L¼1x=L¼0 FðCxÞ

��xL

�2 þ�wn

2L

�2�12

d�xL

�

ln

�4Lwn

� (3.11)


In the case of the wide-gap sensor, the baseline conductance Gow can besimplified by using the first term in Maclaurin expansion of the logarithmicfunction to give

Gowz2p

sobLww

(3.12)

The sensing electrodes behave like a parallel-plate structure. As the elec-tric field inside the film is nearly constant and independent of the distance x,the response of the wide-gap sensor Rw can be reduced to

Rwz

Z x=L¼1

x=L¼0FðCxÞd

�xL

�(3.13)

Under uniform gas profile (type I), the steady-state response function isgiven by the power law

FIðCoÞ¼ k2Cno 0 < n < 1 (3.14)

where Co is the external gas concentration, k2 is the gas-sensitivity parameterfor a semiconducting oxide material and exponent n normally lies between0.3 and 0.9. The steady-state response of this type of sensor pair is asstraightforward as the concentration profile (Fig. 3.12(a)). When a semi-conducting oxide material is doped with catalyst of high activity (moving-boundary gas profile, type II), it is probable that the gas entering the film is

Increasing value of C0

Increasing value of C0

Air Active film Substrate

Air Active film Substrate

1.0 0.5 0.0 1.0 0.5 0.0–0.5

Distance in film, x/L Distance in film, x/L

Gas

con

cent

ratio

n, C

x

Gas

con

cent

ratio

n, C

x

(a) (b)

Cp

Figure 3.12 Gas profile in the active film: (a) uniform and (b) moving boundary.


rapidly consumed and so does not penetrate all the way into the film(Fig. 3.12(b)).120 The response function can be written as

FIIðCxÞ¼�k2C

n0 xb � x � L

0 0 � x < xb(3.15)

The response of the wide-gap sensor becomes simply the fraction of thefilm penetrated by the gas multiplied by the isotherm; hence,

Rw z k2Cno

Z 1

xb=Ld�xL

�¼ k2C

no

�1� xb

L

�(3.16)

The response of the narrow-gap sensor can also be found from Eqs.(3.11) and (3.15):

Rnzk2Cno

ln

(2

,"xbLþ�x 2b

L2 þ w 2n

4L2

�1=2#)

ln

�4Lwn

� (3.17)

Tamaki et al. studied the effects of gap size differences between elec-trodes.71 Microgap electrodes (0.1e1.5 mm) were formed on the siliconsubstrate using the MEMS process, and WO3 films were deposited on theseelectrodes. When the gap size was larger than 0.8 mm, there was no changein sensitivity to NO2; however, when it was smaller than 0.8 mm, the sensi-tivity tended to increase and was expected to increase further in the range ofless than 0.1 mm. They explained that by the number of grains of WO3 inthe microgap and the resistance change at the boundaries. Shaalan et al.72

also fabricated microgap electrodes (1e30 mm) by dc sputtering and FIBtechniques and deposited SnO2 nanowires on them using the suspensiondropping method. They suggested that the interface between the electrodesand the sensing area plays a very important role in the sensing mechanism ofSnO2 gas sensors. Comparison between the small gap and large gapelectrodes showed that the small gap electrode had the advantage of reli-ability and high sensitivity to low NO2 concentration, whereas the largegap electrode had relatively high sensitivity for high concentrations.

Hoefer et al.121,122 used an array of electrodes of differing width andseparation to examine contact resistance effects in tin dioxide sensors. Thetransmission line method they used involves measuring the total resistanceof a semiconductor sample as a function of electrode separation. It wasshown that the modified sheet resistance displayed greater sensitivity to


CO and NO2 than either the sheet resistance itself or the contact resistance.In this case, wide electrodes with narrow spacing would produce the mostsensitive detection. An array of electrodes varying in width and spacing,but all using the same sensing material, could be used to resolve a mixtureof CO, CH4, NO2, and water vapor into separate measurements of eachcomponent by first determining the relative sensitivity of the total resistanceof each electrode pair to the individual gases.123

3.3 Electrode-oxide semiconductor interfaces

3.3.1 Ideal contact of metal and oxide semiconductorThe potential barrier which forms when a metal is contacted with an

oxide semiconductor arises from the separation of charges at the metal-oxidesemiconductor interface, such that a high-resistance region devoid of mobilecarriers is created in the oxide semiconductor. This is similar with SchottkyeMott model that explains the barrier height of metal-semiconductorcontact.124 According to this model, the barrier results from the differencein the work functions of the two substances. The energy band diagramsin Fig. 3.13 illustrate the process of barrier formation. Fig. 3.13(a) showsthe electron energy band diagram of a metal of work function Fm(¼efm)and an n-type semiconductor of work function Fs(¼efs), which is smallerthan Fm. The work function of a metal is defined as the amount of energyrequired to raise an electron from the Fermi level to the vacuum level. Thevacuum level is the energy level of an electron just outside the metal withzero kinetic energy and is the reference level in Fig. 3.13(a). The workfunction Fm has a volume contribution due to the periodic potential ofthe crystal lattice and a surface contribution due to the possible existenceof a dipole layer at the surface. The work function Fs of the semiconductoris defined similarly and is a variable quantity because the Fermi level in thesemiconductor varies with the doping. An important surface parameterwhich does not depend on doping is the electron affinity c defined as theenergy difference of an electron between the vacuum level and the loweredge of the conduction band. The work functions Fm and Fs and theelectron affinity c are usually expressed in electron volts (eV).

Note that the semiconductor shown in Fig. 3.13(a) does not contain anycharges at the surface so that the band structure of the surface is the same asthat of the bulk and there is no band bending. Fig. 3.13(b) shows the energyband diagram after the contact is made and equilibrium has been reached.When the two substances are brought into intimate contact, electrons


from the conduction band of the semiconductor, which have higher energythan the metal electrons, flow into the metal till the Fermi level on the twosides is brought into coincidence. As the electrons move out of the semicon-ductor into the metal, the free electron concentration in the semiconductorregion near the boundary decreases. As the separation between the conduc-tion band edge Ec and Fermi level EF increases with decreasing electronconcentration and in thermal equilibrium EF remains constant throughout,the conduction band edge Ec bends up as shown in Fig. 3.13(b). Theconduction band electrons which cross over into the metal leave a positivecharge of ionized donors behind, so the semiconductor region near themetal gets depleted of mobile electrons. Thus, a positive charge is establishedon the semiconductor side of the interface and the electrons, which crossover into the metal form a thin sheet of negative charge contained within

Vacuum level

EF

EF EF

Ec

EFEFi

E

Ec

eΦB0

eΦn

eVbi

eΦm

eχ eΦs

ν

Eν

(a)

(b)

Depletionregionxn = W

Figure 3.13 Energy band diagram of metal contact to n-type semiconductor with Fm

> Fs: (a) neutral materials separated from each other and (b) thermal equilibrium sit-uation after the contact has been made, where xn is the penetration of the depletionregion into the n material, W is the width of the depletion region, and EFi is intrinsicFermi level.


the ThomaseFermi screening distance from the interface (z0.5 Å).Consequently, an electric field is established from the semiconductor tometal in Fig. 3.13(b). Note that the width of space charge layer in thesemiconductor is appreciable because the donor concentration in thesemiconductor is several orders of magnitude smaller than the electronconcentration in the metal.41

Let us now investigate how much the energy bands in the semicon-ductor will bend upward. It should be evident that because the band gapof the semiconductor is not changed by making contact with the metal,the valence band edge Ev will move up parallel to the conduction bandedge Ec. Also, the vacuum level in the semiconductor will follow thesame variations as Ec. This is because the electron affinity of the semicon-ductor is assumed to remain unchanged even after the metal contact ismade. Thus, for metalesemiconductor system in thermal equilibrium, theimportant point which determines the barrier height is that the vacuum levelmust remain continuous across the transition region. Hence, the vacuumlevel from the semiconductor side must approach the vacuum level onthe metal side gradually to preserve the continuity. The amount of bandbending, then, is just equal to the difference between the two vacuum levels,which is equal to the difference between the two work functions. Thisdifference is given by eVbi ¼ (Fm � Fs), where Vbi is expressed in voltsand is known as “contact potential difference” or the built-in potential ofthe junction: eVbi obviously is the potential barrier, in which an electronmoving from the semiconductor into the metal has to surmount. However,the barrier looking from the metal toward the semiconductor is different andis given by124,125

FB0¼ðFm�cÞ (3.18)

As Fs ¼ c þFn, we have

eVbi¼FB0 � Fn (3.19)

where Fn ¼ (Ec � EF) represents the penetration of the Fermi level in theband gap of the semiconductor and e is the electronic charge. Eq. (3.18) wasstated by Schottky126 and, independently, by Mott.127

The exact shape of the potential barrier can be calculated from the chargedistribution within the space charge layer. In most cases, the height fB0 of thepotential barrier is orders of magnitude larger than the thermal voltage, kT/e,and the space charge region in the semiconductor becomes a high-resistivitydepletion region devoid of mobile carriers. The shape of the barrier is then


determined from the donor distribution in the semiconductor. Schottkyassumed the semiconductor to be uniformly doped up to the metal interface,which gives rise to a uniform charge density in the depletion region. Theelectric field strength for this constant space charge rises linearly with distancefrom the edge of the space charge layer and the resulting parabolic barrier isknown as a “Schottky barrier.”126 Mott127 assumed a thin layer of semicon-ductor, devoid of any charge, sandwiched between a uniformly dopedsemiconductor and the metal. The electric field strength in thin region isconstant and the potential increases linearly across this region. This type ofbarrier is known as the “Mott barrier.” The Mott barrier is encounteredin situation where a thin layer of low-doped, nearly intrinsic semiconductoris interposed between a metal and a heavily doped semiconductor.

The above description applies only to n-type semiconductor whosework function is less than the metal work function Fm. The electron energyband diagrams for an n-type semiconductor with Fm < Fs are shown inFig. 3.14. Fig. 3.14(a) shows the energy bands for separated materials. Afterthe contact is made, electrons flow from the metal into the conduction band

eΦm

eΦs

eΦneΦBn

EFEFi

Ec

EFEF

EF

Ec

Eν

Eν

(a)

(b)

Figure 3.14 Energy band diagram of metal contact to n-type semiconductor with Fm

> Fs: (a) neutral materials separated from each other and (b) contact under thermalequilibrium.


of the semiconductor, leaving behind a positive charge on the metal andcausing an accumulation of electrons on the semiconductor side of theboundary. When equilibrium is reached, the Fermi level in the semicon-ductor is raised by an amount (Fs � Fm) as shown in Fig. 3.14(b). Theaccumulation layer charge in the semiconductor is confined to a thicknessof the order of Debye length and is, essentially, a surface charge. As theconcentration of electrons in the metal is very large, the positive chargeon the metal side is also a surface charge contained within a distance of about0.5 Å from the metal-semiconductor interface. It is clear that no depletionregion is formed in the semiconductor and there is no potential barrier forthe electron flow either from the semiconductor toward the metal or inthe opposite direction. The electron concentration is increased in the regionnear the interface and the highest resistivity region in the system is the bulksemiconductor region.

The foregoing discussion has shown that, in case of n-type semicon-ductor, a metal-semiconductor contact is rectifying if Fm > Fs and isnonrectifying if Fm < Fs. The opposite is true for a metal p-type semicon-ductor contact. Table 3.3 gives the most preferred experimental values ofwork function for important metals.125,127e133

3.3.2 Contacts with surface states and an interfacial layerThe periodicity of the crystal lattice terminates at the surface of a semicon-ductor. In a crystal, the surface atoms have neighbors only on the semicon-ductor side; on the vacuum side, there are no neighbors with whom thesurface atoms can make bonds. Thus, each of the surface atoms has onebroken bond in which only one electron is present and the other is missing.The broken bonds are known as “dangling bonds.”Dangling bonds give rise

Table 3.3 Work function of some important metals.Metal Work function (eV) Metal Work function (eV)

Pt 5.65 Zn 4.33Ni 5.25 Al 4.28Pd 5.12 Ag 4.26Au 5.1 Pb 4.25Cu 4.65 Ta 4.25W 4.55 Cd 4.22Cr 4.5 Ga 4.2Hg 4.49 In 4.12Sn 4.42 Zr 4.05Ti 4.33 Cs 2.14


to localized energy states at the surface of the semiconductor with energylevels lying in the forbidden gap. These surface states are usually continu-ously distributed in the band gap and are characterized by a neutral levelF0. The position of this neutral level is such that, when there is no bandbending in the semiconductor, the states are occupied by electrons up toF0, making the surface electrically neutral. The states below F0 aredonor-like because they are neutral when occupied and are positive whenempty. Obviously, the states above F0 behave as acceptor-like. On cleansurface of semiconductor, the density of surface states equals the density ofsurface atoms. Adsorbed layers of foreign atoms may considerably reducethis density by completing the broken bonds. The surface states modifythe charge in the depletion region and, thus, affect the barrier height.Fig. 3.15(a) shows the electron energy band diagram of an n-type semicon-ductor under flat band condition. When a metal is now brought in contactwith the semiconductor and equilibrium is reached, the Fermi level in thesemiconductor must change by an amount equal to the contact potentialby exchanging charge with the metal. If the density of surface states at the

(a)

(b)

EF

EF EF

Ec

Eν

Eν

EF

Ec

Surface state

Surface state

Metal Semiconductor

Figure. 3.15 Energy band diagram of the band bending interface in the presence ofsurface states for a metal/n-type semiconductor: (a) under flat band condition and(b) after contact formation.


semiconductor surface is very large, then the charge exchange takes placelargely between the metal and the surface states, and the space charge inthe semiconductor remains almost unaffected. As a result, the barrier heightin Fig. 3.15(b) becomes independent of the metal work function. In thiscase, the barrier height is said to be pinned by surface states.134 Consideringa continuum of interface states, a phenomenological formula for Schottkybarrier height can be formulated.135,136

FB0¼ SðsÞcþ F0ðsÞ (3.20)

whereF0(s) represents the contribution of the surface states and the interfaceindex S(s) ¼ dFB0/dc gives the dependence of barrier height on the metalelectronegativity.

The interface between metals and nonmetals has been classified into fourbroad types according to the resulting interfacial atomic configuration:137

(1) The nonmetal is an insulator (or a semiconductor) and the metal isphysisorbed on its surface.

(2) The nonmetal is a highly polarizable semiconductor and the metalmakes a weak chemical bond but does not react with it to form abulk compound.

(3) The highly polarizable semiconductor reacts with the metal and formsone or more chemical compounds.

(4) A thin film of oxide is left during the surface preparation of a highlypolarizable semiconductor which prevents an intimate contactbetween the metal and the conductor. The film is referred to as inter-facial layer.The type (1) interface is an ideal Schottky barrier contact, in which the

barrier height varies directly with the metal work function in accordancewith Eq. (3.18). The type (2) interface approximates to a Bardeen barrier,131

provided that the surface states are assumed to be distributed in space insidethe semiconductor to allow a potential drop across this region. In the cleancontacts of this type, one would expect the barrier height to show a weakdependence on Fm. The type (3) interface represents a case of strongchemical bonding between the metal and the semiconductor and, hence,we would expect the barrier height to depend on some quantity relatedto chemical or metallurgical reactions at the interface. The type (4) contactis the one which is most frequently encountered in actual metal-semiconductor devises.

In most metal-semiconductor contacts, the semiconductor surface beforemetal deposition is prepared by chemical cleaning and a thin insulating oxide


layer is invariably left on the surface of the semiconductor. The thickness ofthis interfacial layer depends on the method of surface preparation and, for agood Schottky contact, must be less than about 20 Å. The energy banddiagram of a contact with interfacial oxide layer is shown in Fig. 3.16. Inthis figure, potential drops linearly across the interfacial oxide layer becausethis layer is assumed to be an ideal insulator devoid of any charge. It has alsobeen assumed that in the lower edge of the conduction band, the insulatorlies below the vacuum level. When the interfacial layer is thin enough(<30 Å), the potential drop across it is negligibly small compared to thatin the semiconductor depletion region. Such a thin layer is transparent tothe electrons as the electrons can tunnel through it in either direction. Forthese reasons, the barrier height FB0 and the contact potential differenceVbi remain almost unaffected by the presence of a thin interfacial layer.

In this state, the charge in the depletion region vanishes and the chargeon the metal side is balanced by the charge in the interface states on the semi-conductor side. This flat band barrier height is given by138

F*B0¼C1ðFm�cÞ þ ð1�C1ÞðEg �F0Þ ¼ C1Fm þ C2 (3.21)

where

C1¼ εi

εi þ e2tDs

Here, εi ¼ εrε0 is the permittivity of the insulating layer, t is its thickness,e is the electronic charge, andDs is the density of interface states per unit area

eΦB0

eΦ0

EFEF

Ec

Eν

Metal Semiconductor

t Insulating interfacial film

Figure 3.16 Electron energy band diagram of a metal-semiconductor contact with sur-face states and interfacial layer.


per eV. The position of the neutral level F0 is measured from the top of thevalence band. From Eq. (3.21) it is seen that asDs tends to zero, C1 becomesunity and F*

B0 tends to the Schottky limit of Eq. (3.18).The expression for the flat band barrier height is modified when there is a

charge present in the interfacial oxide layer. Considering the case of then-type semiconductor, if there is a fixed charge Qox per unit area withinthe oxide layer, then Eq. (3.21) is replaced by

F*B0¼C1ðFm�cÞ þ ð1�C1ÞðEg �F0Þ � C1tQox

εi(3.22)

3.3.3 Image force effects on the barrier heightThe other effect that makes the barrier height depend on the electric field inthe depletion region is the lowering of the image force barrier. This is notdependent on the presence of the interfacial oxide layer and occurs evenwhen such a layer is absent. Lowering of the image force barrier can beunderstood by referring to Fig. 3.17. When an electron is at a distance xfrom the metal, there exists an electric field perpendicular to the metalsurface. This field may be calculated by assuming a hypothetical positiveimage charge e located at a distance (ex) inside the metal. The force ofattraction F between the electron and its image charge is e2/4pεd(2x)

2,and the electron has a negative potential energy F$x ¼ ee2/16pεdx relativeto that of an electron at infinity, as shown by the dotted curve in Fig. 3.17.This potential energy must be added to the barrier energy eeEx to obtainthe total energy of the electron. It can be seen from Fig. 3.17 that the

Metal Dielectric

eΦB0

eΦBm

eεΔΦ

E(x)

x

x = 0 x = 0

EF

xm

x

(b)(a)

Figure 3.17 (a) Image charge and electric field lines at a metal-dielectric interface and(b) electron energy diagram showing the image force lowering of the barrier.


maximum energy occurs at a distance xm from the metal surface, and it canalso be shown that the magnitude DFB0 (¼eDfB0) of the barrier lowering isgiven by26

DFB0¼"

e3Nd

8p2ε2dεs

ðVbi� V Þ#1=4

(3.23)

Here, Nd is the donor concentration in the semiconductor and V is theapplied voltage. The image force permittivity εd may be different from thestatic permittivity εs of the semiconductor. This is because the electronapproaches the barrier with the thermal velocity (z107 cm/s) and, if itstransit time through the barrier region is small compared to the dielectricrelaxation time, then the semiconductor does not get fully polarized.However, it is found that in practical situations, the electron transit timethrough the barrier region is sufficiently large to justify εd ¼ εs.

The image force lowering of the barrier results from the field producedby an electron and will be absent when there is no electron present in thesemiconductor conduction band near the top of the barrier. Hence, whenthe barrier height is measured by a method which does not requiremovement of the electron over the barrier, the obtained value of FB0 isnot lowered by the image force.

3.4 Charge carrier transport in the electrode-oxidesemiconductor interfaces

3.4.1 Electric field and capacitance in the metal-semiconductor interface

The electric field and potential distribution in the depletion region of aSchottky barrier junction depend on the barrier height, the applied voltage,and the impurity concentration. These dependences are frequently neededand can be obtained by the solution of a one-dimensional Poisson equation.Fig. 3.18(a) shows the energy band diagram of a reverse-biased Schottkybarrier junction made on an n-type semiconductor. We assume the semicon-ductor to be nondegenerated and uniformly doped and divide it into a spacecharge region and a neutral region devoid of any space charge. At any point inthe semiconductor, the Poisson equation can be written as124,125

d2fdx2

¼ � eεs½Nd þ pðxÞ� nðxÞ� (3.24)


where εs is the semiconductor permittivity, Nd is the donor concentration,and n(x) and p(x) are the electron and the hole concentrations at any point xin the semiconductor, respectively. It is assumed that all the donors areionized. Taking the potential f to be zero in the neutral bulk region of thesemiconductor at the edge of the space charge layer, one can write

nðxÞ¼ n0 exp

�efðxÞkT

�; pðxÞ ¼ p0 exp

�� efðxÞ

kT

�(3.25)

where n0 and p0 represent the equilibrium electron and hole concentrationsin the neutral semiconductor. Substituting the values of n(x) and p(x) in Eq.(3.24), one obtains

d2fdx2

¼ � eεs

�Nd � n0 exp

�efðxÞkT

�þ p0 exp

�� efðxÞ

kT

�(3.26)

A closed form solution of this equation is not possible. An additionalsimplifying assumption made in the analysis is the so-called “depletionapproximation.” In this approximation, the free carrier concentrations areassumed to fall abruptly from their equilibrium values n0 and p0 in thebulk neutral region to a negligibly small value in the barrier space charge

Metal Semiconductor

eΦB0

EF EF

Ec

Ev

E

Em

eVd

Vd

W

x

x

(a)

(b)

(c) (x)φ

Figure 3.18 Electric field and potential distribution in the metal-semiconductor inter-face: (a) energy band diagram, (b) electric field, and (c) potential distribution.


region. In reality, this transition occurs smoothly over a distance in whichthe bands bend by about 3kT, but the calculations made using the depletionapproximation are sufficiently accurate for most purposes. Thus, using thedepletion approximation, Eq. (3.26) can be written as

d2f

dx2¼ � e

εsNd 0 < x < W

¼ 0 x > W

(3.27)

whereW represents the width of the depletion region. Integrating Eq. (3.27)with respect to x and using the condition that df/dx ¼ 0 at x ¼W, weobtain the electric field ε(x) in the depletion region:

εðxÞ¼ � d4dx

¼ εm

�1� x

w

�(3.28)

where εm ¼ �eNdεsW is the maximum electric field which occurs at x ¼ 0. A

second integration with the boundary condition f ¼ 0 at x ¼W leads to thefollowing relation:

fðxÞ¼ eNd

2εsW 2

�1� x

W

�2(3.29)

Thus, the potential varies parabolically with the distance in the depletionregion and has its maximum value f(0) ¼ Vd given by

Vd ¼ðVbi�V Þ ¼ eNd

2εsW 2 (3.30)

where V is the externally applied voltage. For a forward bias V¼ VF and fora reverse bias V ¼ eVR. The negative sign in the above equation shows thatthe potential at x ¼ 0 is negative with respect to that at x ¼ W. Thedepletion region width W is obtained from the above relation:

W ¼�2εseNd

jVbi � V j�1=2

(3.31)

The width of depletion region at zero bias is obtained by putting V ¼ 0.From Eq. (3.31), it can be seen that W decreases below its value W0 in caseof a forward bias and increases above W0 in the event of a reverse bias.Figs. 3.18(b) and (c) show the electric field and the potential distributionsfor the Schottky barrier junction.

In the depletion approximation, we have neglected the electron and holeconcentrations in comparison to the donor concentration Nd. In a strong


n-type semiconductor, the hole concentration is negligible but at the edgeof the depletion region x ¼W the electron concentration n(W) ¼ n0 ¼ Nd

and decreases exponentially with the decreasing potential f(x). It should beclear that the concentration n(x) becomes negligibly small compared withNd when ef(x) in Eq. (3.25) is e4kT or less. Thus, the depletion approxi-mation is valid only when the potential drop Vd across the depletion regionis large compared to about 4kT/e. A change in the voltage across theSchottky barrier junction causes a change in the width of the depletionregion, and this change is accomplished by the movement of charge carriersinto the space charge layer or out of this region. This change in the depletionregion charge gives rise to a capacitance. Referring to Fig. 3.18(a) andignoring the charge in the surface states, there are three sources of chargein the barrier region. First, there is the charge Qd in the depletion region,which results from the movement of electrons out of the semiconductorinto the metal. Second, there is a charge Qm on the metal surface, whichis caused by the electrons that have crossed from the semiconductor intothe metal. Finally, if the band bending is sufficiently large, a charge Qh

will occur due to holes which exist in the semiconductor region immedi-ately adjacent to the metal contact.95 Electrical neutrality in the junctionregion requires thatQd þ Qm þ Qh ¼ 0, where each of these charges repre-sents charge per unit area of the junction. Suppose now that the bias acrossthe junction is increased by a small amount DVd. The space charge layercapacitance C0 per unit area is defined by the relation:

C0 ¼ dQd

dVd¼ � d

dVdðQmþQhÞ (3.32)

Let us ignore the effect of minority carriers and take Qh ¼ 0 so thatQd ¼ eQm. Applying Gauss law at the metal-semiconductor boundary,we obtain

εsEm¼Qd (3.33)

The maximum electric field strength is assumed to occur at x ¼ 0. Thefield Em has been calculated in Eq. (3.28), assuming the depletion approxi-mation. A more accurate expression for εm is obtained by integrating Eq.(3.26), assuming that band bending is small so that p(x) is negligible every-where and

d2fdx2

¼ � eεs

�Nd � n0 exp

�efðxÞkT

�(3.34)


Multiplying both sides of this relation by 2df/dx and integrating fromx ¼ 0 to x ¼W, with f(0) ¼ eVd and f(W) ¼ 0 and assuming Nd ¼ n0,we obtain �

dfdx

�2

x¼0¼ ε

2m ¼ 2e

εsNd

�Vd � kT

e

�(3.35)

where Vd ¼ (Vi � V) is the voltage drop across the depletion region and thebuilt-in voltage Vi is taken to be positive. The depletion region charge Qd

per unit area is then given by

Qd ¼ εsεm ¼�2eεsNd

�Vd � kT

e

��1=2(3.36)

and the depletion region capacitance C can be written as

C¼AdQd

dVd¼ A

�eεsNd

2½Vbi � ðkT=eÞ � V �1=2

(3.37)

where A represents the area of the Schottky barrier contact. As charge Qd

varies with the voltage in a nonlinear manner, the capacitance is a nonlinearfunction of voltage and can be defined only as a differential capacitancepresented to a small change DVd in the voltage of the depletion region.

If an interfacial oxide layer is present between the metal and thesemiconductor, a portion of the applied voltage appears across this layerand modifies the dependence of the depletion region charge Qd on theapplied voltage. The capacitances of the interfacial layer and the depletionregion are effectively in series, and the overall capacitance in general maybe a complicated function of the interfacial layer parameters and the appliedvoltage. However, when the interfacial layer is thin (about 30 Å or less),electrons can tunnel through the layer from the metal to the semiconductorand the resulting Schottky barrier is nearly ideal. The currentevoltagecharacteristics of the nearly ideal diode are described in terms of an idealityfactor n given by

1n¼�vFB

evV

�(3.38)

where vFB/vV represents the change in the barrier height with appliedvoltage because of the presence of the interfacial layer. If the states at theoxide-semiconductor interface are uniformly distributed in energy, then n isnearly independent of the applied voltage.


To obtain the CeV characteristics for the nearly ideal case, we substituteeVi ¼ (FB � Fn) in Eq. (3.36) and write

Qd ¼ ½2εiεsNdðFB �Fn � kT � eV Þ�1=2 (3.39)

and

C¼A

dQd

dV

¼ eA

�εsNd

2½FB0 � Fn � kT � eV �1=2�

1� vFB

evV

�(3.40)

which, after substituting for vFB0/vV in terms of the ideality factor n,becomes

C¼A

dQd

dV

¼ eAn

�εsNd

2½FB �Fn � kT � eV �1=2

(3.41)

The barrier height FB for a constant n can be written as

FB¼FB0 þ vFB

vVV ¼ FB0 þ

�1� 1

n

�eV (3.42)

where FB0 represents the zero bias barrier height. Combining Eqs. (3.41)and (3.42), one obtains

1C2¼

2nA�2

e2εsNd½nðFB0�Fn� kT Þ� eV � (3.43)

From Eq. (3.43), one observes that the effect of interfacial layer is to scaleup both the slope and the intercept. Thus, the 1/C2 versus voltage V plotunderestimates the dopant concentration and results in a larger interceptV0 over its values in ideal case without interfacial layer. For a nearly idealSchottky barrier, n is 1.1 or less so that the value of Nd is not changedmuch but V0 is significantly increased.

3.4.2 Transport mechanism across the junction barrierFig. 3.19 schematically depicts these processes for a forward biased Schottkybarrier made on an n-type semiconductor. The inverse processes occurunder reverse bias. The current flows in a Schottky barrier because of chargetransport from the semiconductor to the metal, or vice versa. There are fourdifferent mechanisms by which the carrier transport can occur:(1) thermionic emission over the barrier;(2) tunneling through the barrier;


(3) carrier recombination (or generation) in the depletion region; and(4) carrier recombination in the neutral region of the semiconductor,

which is equivalent to the minority carrier in junctions.125

Referring to Fig. 3.19, an electron emitted over the barrier fromsemiconductor into the metal must move through the high field depletionregion. In traversing this region, the motion of the electron is governed bythe drift and the diffusion processes. The emission of electrons into the metalis controlled by the density of available states in the metal. Thus, the twoprocessesdthe emission over the barrier and the drift and diffusion in thedepletion regiondare effectively in series and whichever offers the higherresistance determines the current. In their original treatment, Wanger,139

Schottky, and Spenke140 assumed that the current was limited by the driftand diffusion processes. The diffusion theory leads to the following expres-sion for the diode current:34

I ¼ eANcmEm exp

��FB

kT

��exp

�eVkT

�� 1

�(3.44)

where A is the diode cross-sectional area, Nc is the effective density of statesin the conduction band of the semiconductor, m is the electron mobility, andall other symbols have their usual meanings. As the maximum field Em in Eq.(3.44) is voltage-dependent, the preexponential factor in this equation doesnot saturate as it should in an ideal Schottky diode. Subsequent work byBethe showed that the diode current is limited by thermionic emission overthe barrier and is not in agreement with Eq. (3.44). The difference betweenthe two mechanisms is shown by the position of the quasi-Fermi level in thedepletion region. According to the diffusion theory, the electrons are inequilibrium with the lattice even when the junction is forward biased, so

Δe B

eVF

eVd

EF

xmEcEF

Ev

(a)(b)

(c) (d)

φ

e Bφ

Figure 3.19 Energy band diagram of a forward biased Schottky barrier junction onn-type semiconductor showing different transport processes.


that their quasi-Fermi level coincides with the metal Fermi level at theinterface, as shown by dotted curve in Fig. 3.19. In the thermionic emissiontheory, on the other hand, the electrons entering the metal have energyhigher than the metal electrons and their quasi-Fermi level is almosthorizontal through the depletion region, as shown by the dashed curve.

The effect of drift and diffusion in the depletion region is assumed to benegligible in the thermionic emission theory and the barrier height isassumed to be large compared to kT. From Fig. 3.19, it is obvious thatonly those electrons whose kinetic energy exceeds the height of the poten-tial barrier will be able to reach the top of the barrier. Assuming that theelectrons have a Maxwellian distribution of velocities, the number ofelectrons n* per unit area, which have sufficient energy to move over thebarrier from the semiconductor into the metal, is given by

n * ¼ n0 exp

��eðVbi � V ÞkT

�(3.45)

where n0 represents the electron concentration in the neutral semiconductoroutside the depletion region and V is the voltage applied to thesemiconductor.

For a nondegenerate semiconductor,

n0¼Nc exp

��Fn

kT

�(3.46)

and as FB ¼ eVi þ Fn from Eq. (3.45) we obtain

n*¼Nc exp

�� ðFB � eV Þ

kT

�(3.47)

If these electrons are assumed to have an isotropic distribution of veloc-ities, then from the kinetic theory the flux of electrons incident on thebarrier is

�n*v

� 4. Supposing that all the incident electrons cross over into

the metal and none is reflected back, the current ISM due to passing ofelectrons from the semiconductor to the metal is given by

ISM ¼ eAv4Nc exp

��FB � eV

kT

��(3.48)

where v is the average thermal velocity of electrons in the semiconductor.Although the electrons flow from the semiconductor into the metal, thecurrent ISM flows from the metal to the semiconductor and is taken to bepositive in Fig. 3.19.


For unbiased junction under thermal equilibrium, no net current canflow. Consequently, the current given by Eq. (3.48) must be balanced byan opposite current IMS due to crossing of electrons from the metal intothe semiconductor making I ¼ ISM þ IMS ¼ 0 and

IMS ¼ � eAv4Nc exp

��FB

kT

�(3.49)

In the presence of an applied bias V, the barrier for electron flow fromthe metal to semiconductor remains practically unchanged at fB and so isthe current IMS ¼ eI0. The current ISM, however, is given by Eq. (3.48)and, combining this equation with Eq. (3.49), we obtain

I ¼ I0

�exp

�eVkT

�� 1

�(3.50)

For a Maxwellian distribution, the average velocity v ¼ �8kT

pm*

�1=2and substituting Nc ¼ 2(2pm*kT/h2)3/2, the current I0 can be written as

I0¼ART 2 exp

��FB

kT

�(3.51)

where

R¼ 4pm*ek2

h3

is the Richardson constant for thermionic emission from the metal into thesemiconductor with electron effective mass m*, h is the Plank’s constant, andA is the diode area.

3.4.3 Tunneling effects in the oxide-semiconductor interfaceThe thermionic emission diffusion theory of Crowell and Sze141 assumesthat the electron distribution function remains Maxwellian in the barrierdepletion region and that the classical drift and diffusion equations can beused throughout this region. Near the top of the barrier, the electric fieldis very high and the distribution function changes considerably within amean free path. Under these conditions, it is no longer possible to splitthe current into drift and diffusion components. Moreover, we should notexpect the distribution function near the top of the barrier to be Maxwellianand isotropic. As the electrons entering the metal are assumed not to return,


the distribution function near the boundary must be anisotropic. However,the transport equations derived from the first-order solution of the Boltz-mann equation are found to be inadequate near the top of the barrier anda new set of transport equations has been proposed. These considerationsshow that the problem of hot electrons in a rapidly varying field has notbeen solved to date and we do not yet have an exact theory of the Schottkybarrier currentevoltage characteristics. It is, therefore, surprising that in spiteof the drastic assumptions the existing theory has been so successful indescribing the IeV characteristics.

Besides the diffusion and thermionic emission mechanisms, electrons canalso be transported across the barrier by quantum mechanical tunneling. Thetwo ways in which tunneling can occur in a Schottky barrier junction areshown in Fig. 3.20 for (1) forward bias and (2) reverse bias. The semicon-ductor in Fig. 3.20 is assumed to be doped to degeneracy such that the Fermi

n

e B

Δ

eVF

eVR

TFE

TFE

FE

FE

EF

EF

Em

EF

Ec

EF

Ec

(a)

(b)

φ

φ

e Bφ

e Bφ

Figure 3.20 Field emission and thermionic field emission tunneling through a Schottkybarrier on n-type semiconductor: (a) forward bias and (b) reverse bias.


level lies above the bottom of the conduction band. Because of heavydoping, the depletion region is very thin and, at low temperatures, electronswith energy close to the Fermi level can tunnel from the semiconductor intothe metal. This process is known as “field emission” (FE). At higher temper-atures, a significant number of electrons are able to rise high above the Fermilevel, where they see a thinner and lower barrier. These electrons thus cantunnel into the metal before reaching the top of the barrier. This tunnelingof thermally excited electrons is known as “thermionic field emission”(TFE). As the number of electrons decreases rapidly with energy abovethe Fermi level, and the barrier thickness and height also decreases, thereexists an energy Em at which the contribution of TFE becomes maximum.If temperature is gradually raised still further, a limit is reached at whichpractically all the electrons are able to reach the top of the barrier and therm-ionic emission predominates.

Tunneling through a Schottky barrier has been analyzed theoretically byPadovani and Stratton142 and by Crowell and Rideout.143 The main resultsof their study are described below. Field emission in the forward directionoccurs only in degenerate semiconductors and, except for very low forwardbiases, the IeV characteristic in the presence of tunneling can be describedby the relation

I ¼ Is exp

�eVE0

�(3.52)

where

E0¼E00 cot h

�E00

kT

�

and

E00¼ eh4p

�Nd

m*εs

�1=2

where m* is the electron effective mass and h is Planck’s constant. Thepreexponential factor Is in Eq. (3.52) is only weakly dependent on voltageand is a complicated function of barrier height, parameters of thesemiconductor, and the temperature. The ratio kT/E00 is a measure of therelative importance of TE and tunneling. At low temperature, Eoo maybecome large compared to kT and we have E0 z E00, then the slope of lnIversus theV plot is constant and independent of T. This is the case for FE. Athigh temperatures, where E00 � kT, we get E0 ¼ kT and the slope of the


lnI versus the V plot is e/kT, which corresponds to TE. For intermediatevalues of temperature, the slope can be written as e/nkT with

n¼E00

kTcoth

�E00

kT

�(3.53)

3.4.4 Structure of the interfacial layerA barrier of metal/semiconductor contact is a limiting situation which canbe described as two infinite half planes of material, a metal and a semicon-ductor, brought into contact. A typical technical barrier would result fromcontacting metal on the semiconductor after a series of in-ambient prepara-tions. An ideal barrier results from depositing the metal in a carefullycontrolled way, where precautions are taken to keep the interface atomicallyclean, that is, the only atoms present are those of the initial semiconductorand the desired metal. The procedure for cleaning the surface may stronglyinfluence the outcome by altering the surface structure stoichiometry or byintroducing surface defects. Two regimes are(a) Building up the metal layer by layer by evaporation deposition(b) Pressing two bulk pieces together to form a point contact

In both cases, the technical interface would have some amount of extra-neous material trapped at the interface. In actual situations, this materialmight consist of a native oxide of 10e15 Å thick.26 The interplay of a va-riety of effects then determines the interfacial composition, interfacial width,and the number of interfacial states. The complexity of the general problemarises from the diversity of concurrent effects that can be at work in specificmaterial situations. Schottky barriers are fabricated with the layer-by-layerapproach or the deposition of the metal on the semiconductor. Before treat-ing the details of the barrier evolution stages, it is worth describing the detailsof the overall process to motivate the division into the stages, enumerated inTable 3.4, into which the evolution can be decomposed.144

The details of the formation of the interfacial layer are the focus of thissection without concurrently emphasizing the processes that generate inter-face stated. This separation better allows the individual aspects to be identi-fied, but should not imply a decoupling of the two phenomena. Theinterface layer can have a multiplicity of zones with a division betweenmetallic and semiconducting regions, as depicted schematically inFig. 3.21.137,138,145e148

On the metal side, the alloyed zone may be of sufficient width tobecome the metal forming the barrier. There is an insulating layer present,


either intentionally or unintentionally, between the metal and the semicon-ductor. This layer can be a purposefully deposited layer or a thermally grownlayer. The layer could also inadvertently result from the act of junction for-mation. There can be an interfacial or transition layer, extending into thesemiconductor due to out-diffusion of constituents of the semiconductor,in-diffusion of metal atoms, chemical interactions, and semiconductorsurface damage. The presence of insulating layer can have two effects onbarrier formation: one is a geometrical effect (it can simply further separatethe charge in the metal from the charge in the semiconductor, giving a largedipole length) and the other is a potentially significant modification of thedipole arrangement. If the interfacial layer contains no charge, then its effecton barrier formation would be geometrical. However, the presence of an

Table 3.4 Enumeration of the layer-by-layer stages of Schottky barrier formation.

0. Clean surfacea. Orderedb. Disordered

1. Dilute limita. Chemisorptionb. Reaction

2. Metal nucleationa. Eutectic formationb. Compound formationc. Interdiffusion and interaction

3. Asymptotic overlayer

Metal Semiconductor

Bulk 1 1 2 Bulk

Interfacial zones

Figure 3.21 Schematic depiction of metal-semiconductor interfacial zones.


interfacial layer together with interface states has more than a geometricaleffect. It can strongly modify the dipole arrangement because these localizedstates can hold charge. Such interface states can be extrinsic arising fromdefects caused by cross diffusion, chemical interaction, and semiconductorsurface damage and rearrangement. Interface states can also be intrinsic:they can be a basic feature of the semiconductor surface or they can arisefrom the extension of metal electron wave function into the semiconductorgap. As interface states can store charge, they can modify the field in theinterface layer. It is seen that their presence, together with the interface layer,can modify Vb andFB. Unlike the simple geometrical case, it is now possibleto increase Vb and FB or decrease them, depending on the charge stored inthe interface states.145

On the semiconductor side, the interface states are distributed within thetwo zones, one being derived from the evanescent tail of the metal wavefunction extending into the semiconductor and the remaining interfacelayer being the second zone. The interface states in zone 1 in many caseshave the most significant effect on determining the barrier height. The firstzone is at most 10 Å thick from estimates available in literature.145,146 Thesecond zone can be very large depending on the thermal history of thesample and can also significantly affect the applied bias performance ofthe barriers.

This new interfacial metal can have different characteristics from theoriginally deposited metal.

3.4.4.1 State 0: the clean semiconductor surfaceThe layer-by-layer evolution starts from the clean semiconductor surface.The clear semiconductor surfaces can be divided into two types of regime:ordered and disordered. In the case of cleaved surfaces, disordering usuallyresults from structured damage induced by the cleaving process employed.The noncleavage faces are typically cleaned by a sputtering technique whichusually leaves the surface disordered and, in the case of compound semicon-ductors, nonstoichiometric. Thermal annealing will usually restore theordering. The most perfect noncleavage surfaces are those grown in situwith MBE techniques.

Most vacuum-semiconductor interfaces reconstruct or relax so that theyshow lateral periodicities or coordinates different from a strict truncation ofthe bulk lattice. A variety of driving forces for these reconstructions havebeen identified, but in general they can be summed up as deriving fromthe surface lowering its energy by remaining semiconducting. The


reconstruction has a strong influence on the spectrum of intrinsic surfacestates and their energy position. The character of surface states observedwith electron spectroscopy is very sensitive to the presence of adsorbatesor overlayers.

3.4.4.2 Stage 1: the dilute limitThe initial stage of interface development constitutes the dilute limit. Thedilute limit is characterized by a density of metal atoms small enough thata continuous metal film does not form. The structural aspects of the dilutelimit are complex and depend strongly on the semiconductor duringdeposition. Changes in the surface structure at this stage of deposition areknown as “impurity stabilized.” Deposition rate effects are not well-explored but have been demonstrated with respect to nucleation of singlecrystal metals. An important aspect of the dilute limit is that the heat ofcondensation of the metal atom may be sufficient to dislodge semiconductorsurface atoms. If there is a tendency to surface reaction, then this alsostrongly influences this stage.

3.4.4.3 Stage 2: monolayer formationThe monolayer formation stage plays the most important part in deter-mining the characteristics of the interface. As the density of atoms increases,metal nuclei begin to form. The nucleation of the metal film is a verycomplex subject. Characteristic of the monolayer formation stage is theproduction of interdiffusion due to the heat released as the metal nucleates.This heat augments electromigration due to the heat released as the presenceof the dipole. In some cases, the release of this heat can drive other reactionsover their activation barrier. The approach to monolayer formation mayhave several substages at which ordered metal overlayers may result. Thestructural aspects of the monolayer stage are strongly dependent on thedeposition temperature and subsequent annealing history.

3.4.4.4 Stage 3: addition monolayers and interdiffusionThe final characteristics of the interface can continue to be driven, asadditional monolayers are added. Strain fields due to lattice mismatch andgrain boundary phenomena can play an important role. Usually the interfacereaches a stable configuration with 3e10 monolayers for most metalson semiconductor. Interfaces are often metastable, however, and the inter-diffusion is accelerated at increased temperature or under biased conditions.


The application of core level photoemission to this problem confirmed whathad already been deduced with Auger spectroscopy, but on a finer dimen-sional scale. An overriding fact of metal-semiconductor interfaces is theintrinsic aspects of interfacial interdiffusion and interaction. These two as-pects of the chemistry often tend to overwhelm issues that might be deducedfrom the clean separate surfaces. The interdiffusion can result in interfacelayers with radically different characteristics from the supposed components.

3.5 Gas/solid interactions in the electrode-oxidesemiconductor interfaces

3.5.1 Dipole formation in the interfacial layerIt is known that a number of metalsde.g., Pd or Ptdadsorb and

dissolve hydrogen and that the adsorbed atoms change the work functionof the metal surfaces. By using a thin metal film as the catalytic metal elec-trode of the Schottky devices, hydrogen and other gases that react with thecatalytic metal can be detected. Hydrogen molecules are adsorbed and disso-ciated on the outer metal surface. Then, the atomic hydrogen permeatesthrough the bulk lattice toward the metal-semiconductor, causing a pertur-bation at this interface that gives rise to a change in the sensor outputsignal.41,42,149e152 Two hypothetical steps were proposed to explain thedetection mechanism:42

(1) the hydrogen atom is polarized at the metal-semiconductor interface,which gives rise to a dipole layer, or

(2) an excess of charge states at the metal-semiconductor interface is createdin the presence of hydrogen and reduces the Schottky barrier height.Fig. 3.22 shows the band diagrams without and with hydrogen adsorp-

tion. The hydrogen-polarized layer, arising from the intrinsic electric field ofthe diode, redistributes the charges in the depletion region and abates the

EF EF

Ec EcEF EF

eΦB eΦB

Eν Eν

Metal Semiconductor Metal Semiconductor

++

––

(a) (b)

Figure 3.22 Energy band diagram of a Schottky junction without (a) and withhydrogen adsorption (b).


degree of band bending. Hence, the corresponding current is correlatedwith the number of hydrogen atoms adsorbed at the metal-semiconductorinterface. When atomic hydrogen has been formed on the outer metalsurface, which is exposed to the ambient, an equilibrium is reached betweenthe hydrogen concentration at this metal surface and that at the metal-semiconductor interface. Based on this model, the change in the IeVcharacteristics and the decrease in the Schottky barrier height are stronglyrelated to the hydrogen concentration.

3.5.2 Effects of hydrogen adsorption in the Schottky barrierjunction

According to the dipole model, originally developed by Lundstrom,61,96 thehydrogen sensitivity of catalytic gate devices is based on the dissociation ofmolecules on the catalytic metal surface and the diffusion through the metalfilm to form a polarized layer at the metal-insulator interface. The polarizedlayer gives rise to a shift of capacitanceevoltage curve of a MOS capacitor orthe currentevoltage curve of a Schottky diode. Catalytic gate devices alsorespond to hydrogen-containing molecules such as hydrocarbons, providedthat the molecules are also dissociated on the catalytic metal surface. Oxygenatoms are also dissociated on the catalytic metal surface. Water formationwith oxygen atoms from oxygen-containing molecules consumes hydrogenand, therefore, decreases the sensor response. In other words, catalytic metalgates have a direct response to hydrogen and hydrocarbons, as well as anindirect response to oxygen molecules, whose effect is to decrease the directresponse.

The hydrogen response of field effect devices in normal air depends onseveral significant steps. First, the hydrogen molecule has to dissociate on thecatalytic metal surface. This dissociation occurs in competition with adsorp-tion of other molecules in the ambientdin particular, oxygen molecules. Asthe oxygen concentration is normally around 20%, a certain minimumhydrogen concentration is needed to give a significant hydrogen concentra-tion on the catalytic metal surface. For nonporous Pd, this concentration istypically around 1 ppm if a couple of mV response is needed, while it ishigher for nonporous Pt. When atomic hydrogen has formed on the metalsurface, an equilibrium between the hydrogen concentration at the metalsurface and the hydrogen concentration at the metaleinsulator interface isreached by diffusion of hydrogen atoms through the metal film. Avery important part of this process is the surface reactions occurring at the


metaleair interface. Hydrogen is supplied by the dissociation of hydrogenmolecules from the gas phase, but there is also a back reaction due to thesurface reactions between adsorbed hydrogen atoms and adsorbed oxygenatoms, resulting in water molecules that desorb rapidly from the surface atnormal operation temperatures. In addition to this back reaction, adsorbedhydrogen atoms can also associate and desorb again as hydrogen molecules.These reactions are summarized in Fig. 3.23.153

When hydrogen atoms are adsorbed to the surface adsorption sites (s) ofthe metal, reaction is expressed as151

H2; gasþ 2s42Hs (3.54)

Because oxygen is present in the air ambience, it can also be adsorbed onthe metal surface and react with the adsorbed hydrogen to form OHs, whichcan then react with Hs to form water (H2O). The corresponding reactionsare expressed as

O2; gasþ 2s42Os (3.55)

OsþHs4OHs þ s (3.56)

HsþOHs4H2Oþ 2s (3.57)

2HsþOs4H2Oþ 3s (3.58)

Except some of the dissociated hydrogen atoms reacting with the oxygenmolecules, surface oxygen, and hydroxyl group, the other adsorbed

H2 H2

H H H H H H H O O O

H2OO2

Pd

ΔVSiO2

Si

Figure 3.23 Surface reactions, diffusion, and trapping on the catalytic metal surface ofa Pd-MIS device.


hydrogen atoms diffuse through the metal into the metal-oxide interfaceand then further penetrate into the oxide film.

siþHs4sþHsi (3.59)

sO þHs4sþHso (3.60)

where si and so are the hydrogen adsorption sites in the interface and theoxide layer, respectively. The adsorption sites at the interface are located atthe oxygen atoms on the oxide surface,153 and the absorption sites in theoxide layer are located at the silicon or oxygen atoms (SiO2 devices).104

Because the oxygen atoms existing at the oxide surface provide more trapsites for forming the hydrogen dipoles, a large number of hydrogen atomscould then be adsorbed at the metal-oxide, thereby acting as a dipole layer.Consequently, the resultant barrier height FB;H2 is reduced and becomessmaller than fB,air because of the formation of dipoles. The hydrogen atomsabsorbed within the oxide layer react with the oxide atoms and subsequentlyrelease the electrons, which raises the Fermi level of the oxide. Then, theoxide barrier VO; air between metal and oxide is gradually reduced down toVOB;H2 during the reaction process.

3.5.3 Adsorption of other gases in the Schottky barrierjunction

A semiconductor gas sensor generally consists of a thin or thick film ofparticulate semiconductor material, whose macroscopic conductance ismeasured using a pair of electrodes with dimensions much larger than theparticle size. Gas phase species can become adsorbed on the particles. If phys-isorption occurs, there is little direct effect on the macroscopic conductivity,although occupation of surface sites by the physisorbed species may blockadsorption of other species. If chemisorptions occurs, with charge transferbetween the adsorbate and the semiconductor, the charged surface layerproduces bending of the conduction and valence band edges. When parti-cles are essentially separate, with no grain intergrowth, the band bendingleads to a potential barrier between neighboring particles and a change inthe macroscopic conductance relative to that of a pristine film with noadsorbate. Where some particle intergrowth has occurred, the depletionregion induced by chemisorption narrows the conductive regions (necks)between grains, again leading to a decrease in macroscopic conductance,


relative to the film with no adsorption.45 In either case, it will be seen thatchanges in the number of species adsorbed, or in their relative charges, willlead to changes in the macroscopic conductance, as depicted in Fig. 3.24.46

It is the basis of the gas sensing mechanism in conductometric (“chemresis-tor”) sensors based on tin dioxide and other semiconductors.

For the detection of CO gas, the operation of sensor requires oxygen as aconstituent of the background gas. The reaction of CO with the adsorbedoxygen and/or oxygen in the oxide semiconductor causes the reactant prod-uct, CO2, to be desorbed from the device.38,154 The desorption and reactionof CO gas at steady-state proceeds via the following reaction at the metal-oxide semiconductor interface:

2e� þO242Oeads (3.61)

COþOeads4CO2;des þ e� (3.62)

Oxygen can dissociate on the metal and oxygen atoms so formed mayspill over onto the oxide semiconductor surface, a well-known

Gas adsorption,charge transfer

Surface statesand chemical reactions Catalytic effects

Electrode Electrode

Grain boundary

Gas

Energy Current flow

V

eVs

W

x

CO2CO2CO CO CxHy

H2O

H2O

OH–OH–

O–O–

e– e–

Figure 3.24 Schematic of a conductometric gas sensor with electrodes placed high upin the sensor face.


phenomenon in catalysis.155,156 The adsorption (reduction) of the chemi-sorbed Oe ion at the metal-oxide semiconductor interface increases(decreases) the resistance of the oxide semiconductor. In MIS gas (CO,NOx, etc.) sensors which have the metal-oxide semiconductor gate layer,the behavior and gas adsorption mechanism are somewhat different fromthose of conventional noble gate hydrogen sensors. The chemisorptionprocess leads to the creation of dipole moments and changes of work func-tion, thus directly shifting the threshold voltage due to the change of carriersin the inversion layer. The consequent change in voltage across the dipolelayer can be solved using Poisson’s equation, the result of which is givenby105

DVth¼ �Nqd=ε (3.63)

whereNq is the density of the adsorption sites, d is the dipole moment of thedipole layer, and ε is the permittivity. The change in voltage across thedipole bilayer leads to a shift of Vth in the negative gate bias direction. In theIeV characteristics of the MISFET, IDsat can be expressed as

IDsat ¼CðVG � DVthÞ22

(3.64)

where C is a constant determined by the device design.

3.6 Conclusions

Identifying the phenomena that occur between the electrode andsemiconductor of the semiconductor gas sensor has been a very interestingtopic for decades. One-electrode sensors using micromachining technologyare widely applied, replacing the traditional interdigit two-electrode type. Asa planar design is applied to the one-electrode sensor that is produced withthe micromachining technology, large-scale integration is appropriate. Inaddition, it has advantages of low power consumption and productioncost. For the microsensor represented as the FET type, materials used forthe gate electrode and the type of configuration are essential. When it workssimply as a device that supplies voltage to the gate, poly Si or aluminum canbe used. When the gate electrode should perform a catalytic function orallow gas penetration, special materials are used or a unique design is applied.Many studies also attempt to apply conductive polymer to the gas sensor.The conductive polymer electrode using polypyrrole (PPy) andpolyaniline (PANI) is manufactured with casting, layer-by-layer deposition,spin-coating, or LB techniques.


The phenomena at the electrode-semiconductor interface, accompaniedby potential barrier formation, can be explained by the charge transporttheory; nevertheless, several different physical situations occur, dependingon the types and shapes of sensors. It is not one single mechanism thatwill account for all the different material situations encountered. Althoughthe electrode of gas sensors is known to affect sensing mechanism of semi-conductor gas sensors, it is still unclear how much the electrode influencesthe sensors. The anticipation that the electrode will improve the propertiesof gas sensors lead scholars researching on gas sensors to choose an optimumelectrode material and modify electrode geometry according to its sensorconstruction. In the near future, totally new mechanisms that cannot beexplained with current model may be proposed and applied for nano-sized gas sensors currently being researched.

References1. Seiyama T, Kato A, Fulishi K, Nagatani M. A new detector for gaseous components

using semiconductive thin films. Anal Chem 1962;34:1502e3.2. Taguchi N. Gas detecting device. US Patent 3695848. 1972.3. L�opez-G�andara C, Ramos FM, Cirera A. Ysz-based oxygen sensors and the use of

nanomaterials, a review from classical models to current trends. J Sens 2009:1e15.4. Moos R. A brief overview on automotive exhaust gas sensors based on

electroceramics. Int J Appl Ceram Technol 2005;2:401e13.5. Riegel J, Neumann H, Wiedenmann HM. Exhaust gas sensors for automotive emis-

sion control. Solid State Ion 2002;152e153:783e800.6. Ivers-Tiff�ee E, H€ardtl KH, Menesklou W, Riegel J. Principles of solid state oxygen

sensors for lean combustion gas control. Electrochim Acta 2001;47:807e14.7. Anderson T, Ren F, Pearton S, Kang BS, Wang HT, Chang CY, Lin J. Advances in

hydrogen, carbon dioxide, and hydrocarbon gas sensor technology using GaN andZnO-based devices. Sensors 2009;9:4669e94.

8. Chang YC, Bai H, Li SN, Kuo CN. Bromocresol green/mesoporous silica adsorbentfor ammonia gas sensing via an optical sensing instrument. Sensors 2011;11:4060e72.

9. Wang Y, Tong MM, Zhang D, Gao Z. Improving the performance of catalytic com-bustion type methane gas sensors using nanostructure elements doped with rare earthcocatalysts. Sensors 2011;11:19e31.

10. Wongchoosuk C, Wisitsoraat A, Phokharatkul D, Tuantranont A, Kerdcharoen T.Multi-walled carbon nanotube-doped tungsten oxide thin films for hydrogen gassensing. Sensors 2010;10:7705e15.

11. Fine GF, Cavanagh LM, Afonja A, Binions R. Metal oxide semi-conductor gas sensorsin environmental monitoring. Sensors 2010;10:5469e502.

12. Lee SP. Micro-humidity sensor system using CMOS technology. Sens Lett 2011;9:1e6.

13. Song K, Wang Q, Liu Q, Zhang H, Cheng Y. A wireless electronic nose system usinga Fe2O3 gas sensing array and least squares support vector regression. Sensors 2011;11:485e505.

14. Lilienthal AJ, Loutfi A, Duckett T. Airborne chemical sensing with mobile robots.Sensors 2006;6:1616e78.


15. Gonzalez-Jimenez J, Monroy JG, Blanco JL. The multi-chamber electronic nosedanimproved olfaction sensor for mobile robotics. Sensors 2011;11:6145e64.

16. Endres HE, G€ottler W, Hartinger R, Drost S, HellmichW,M€uller G, Braunm€uhl CB,Krenkow A, Perego C, Sberveglieri G. A thin-film SnO2 sensor system for simulta-neous detection of CO and NO2 with neural signal evaluation. Sens Actuators B1996;36:353e7.

17. Lee SP. Synthesis and characteristics of carbon nitride for micro humidity sensors.Sensors 2008;8:1508e18.

18. Chen D, Lei S, Chen Y. A single polyaniline nanofiber field effect transistor and its gassensing mechanisms. Sensors 2011;11:6509e16.

19. Sch€uler M, Sauerwald T, Sch€utze A. Metal oxide semiconductor gas sensor self-testusing Fourier-based impedance spectroscopy. J Sens Syst 2014;3:213e21.

20. Kang KS, Lee SP. CO gas sensors operating at room temperature. J Mater Sci 2003;38:4319e23.

21. Hagleitner C, Lange D, Hierlemann A, Brand O, Baltes H. CMOS single-chip gasdetection system comprising capacitive, calorimetric and mass-sensitivemicrosensors. IEEE J Solid State Circuits 2002;37:1867e78.

22. Liu J, Wang W, Li S, Liu M, He S. Advances in SAW gas sensors based on thecondensate-adsorption effect. Sensors 2011;11:11871e84.

23. Marr I, Reiß S, Hagen G, Moos R. Planar zeolite film-based potentiometric gassensors manufactured by a combined thick-film and electroplating technique. Sensors2011;11:7736e48.

24. Chinvongamorn C, Pinwattana K, Praphairaksit N, Imato T, Chailapakul O. Amper-ometric determination of sulfite by gas diffusion-sequential injection with boron-doped diamond electrode. Sensors 2008;8:1846e57.

25. Pascoe KJ. Properties of materials for electrical engineers. London, UK: John Wiley & Sons;1973.

26. Rhoderick EH. Metal-semiconductor contacts. Oxford, UK: Clarendon Press; 1978.27. Sze SM. Physics of semiconductor devices. New York (NY), USA: John Wiley & Sons;

1981.28. Sze S. Semiconductor devices. New York (NY), USA: John Wiley & Sons; 2002.29. Zhou X, Xu Y, Cao Q, Niu S. ‘Metalesemiconductor ohmic contact of SnO2-based

ceramic gas sensors. Sens Actuators B Chem 1997;41:163e7.30. Schierbaum KD. Engineering of oxide surfaces and metal/oxide interfaces for chem-

ical sensors, recent trends. Sens Actuators B Chem 1995;24e25:239e47.31. Zhang W, Vascincelos EA, Uchida H, Katsube T, Nakatsubo T, Nishioka Y. A study

of silicon Schottky diode structures for NOx gas detection. Sens Actuators B 2000;65:154e6.

32. Irokawa Y. Hydrogen sensors using nitride-based semiconductor diodes, the role ofmetal/semiconductor interfaces. Sensors 2011;11:674e95.

33. Schalwig J, M€uller G, Karrer U, Eickhoff M, Ambacher O, Stutzmann M, G€orgens L,Dollinger G. Hydrogen response mechanism of PteGaN Schottky diodes. Appl PhysLett 2002;80:1222e4.

34. Petty MC. Conduction mechanisms in Pd/SiO2/n-Si Schottky diode hydrogendetectors. Solid State Electron 1986;29:89e97.

35. Matsuo K, Negoro N, Kotani J, Hashizume T, Hasegawa H. Pt Schottky diode gassensors formed on GaN and AlGaN/GaN heterostructure. Appl Surf Sci 2005;244:273e6.

36. Soo MT, Cheong KY, Noor AFM. Advances of SiC-based MOS capacitor hydrogensensors for harsh environment applications. Sens Actuators B 2010;151:39e55.

37. Morrison SR. Selectivity in semiconductor gas sensors. Sens Actuators B 1987;12:425e40.


38. Morrison SR. Mechanism of semiconductor gas sensor operation. Sens Actuators B1987;11:283e7.

39. Barsan N, Huebner M, Weimar U. Conduction mechanism in semiconducting metaloxide sensing film, impact on transition. In: Jaaniso R, Tan OK, editors. Semiconductorgas sensors. Cambridge, UK: Woodhead Publishing; 2013. p. 64e113.

40. Liu X, Cheng S, Liu H, Hu S, Zhang D, Ning H. Review a survey on gas sensingtechnology. Sensors 2012;12:9635e65.

41. Potje-Kamloth K. Semiconductor junction gas sensors. Chem Rev 2008;108:367e99.42. Hubert T, Boon-Brett L, Black G, Banach U. Hydrogen sensorsda review. Sens Ac-

tuators B 2011;157:329e52.43. Korotcenkov G. ‘Practical aspects in design of one-electrode semiconductor gas

sensors’, Status report. Sens Actuators B 2007;121:664e78.44. Kasap S. Principles of electrical engineering materials and devices. New York: McGraw-Hill;

1997.45. Hummel R. Electronic properties of materials. New York: Springer-Verlag; 1985.46. Capone S, Siciliano P, Quaranta F, Rella R, Epifani M, Vasanelli L. Moisture influ-

ence of geometry effect of Au and Pt electrodes on CO sensing response of SnO2microsensors based on sol-gel thin film. Sens Actuators B 2001;77:503e11.

47. Gourari H, Lumbreras M, Van Landschoot R, Schoonman J. Electrode nature effectson stannic oxide type layers prepared by electrostatic spray deposition. Sens Actuators B1999;58:365e9.

48. Saukko S, Latto V. Influence of electrode material on properties of SnO2-based gassensor. Thin Solid Films 2003;436:137e40.

49. Durrani SM. The influence of electrode metals and its configuration on the response oftin oxide thin film CO sensor. Talanta 2006;68:1732e5.

50. Pijolat C. Etude des proprietes physico-chimiques et des proprietes electriques du dioxyde d’etainen fonction de l’atmosphere gazeuse environnante, Application a la detection selective des gaz[Ph.D. thesis]. Polytechnique de Grenoble; 1986.

51. Bertrand J, Koziej D, Barsan N, Viricelle JP, Pijolat C, Weimar U. Influence of thenature of the electrode on the sensing performance of SnO2 sensors, Impedancespectroscopy studies. In: Proceedings of eurosensors XX, Goeteborg, Sweden, 17e20September 2006; 2006. p. 100e1.

52. Ylinampa A, Lantto V, Lepp€avuori S. Some differences between Au and Pt electrodesin SnO2 thick-film gas sensors. Sens Actuators B 1993;13e14:602e4.

53. Rank S, Hafner S, Barsan N, Weimar U. The impact of the nature of the electrodematerial on SnO2 thick Film sensor performance, Influence on oxygen adsorption.Process Eng 2012;47:514e7.

54. Barsan N, Schweizer-Berberich M, G€opel W. Fundamental and practical aspects in thedesign of nanoscaled SnO2 gas sensors, a status report. J Anal Chem 1999;365:287e304.

55. Lin HM, Tzeng SJ, Hsiau PJ, Tsai WI. Electrode effects on gas sensing properties ofnanocrystalline zinc oxide. Nanostruct Mater 1998;10:465e77.

56. Bender F, Kim C, Mlsna T, Vetelino JF. Characterization of a WO3 thin film chlorinesensor. Sens Actuators B 2001;77:281e6.

57. Sofian M, Oussama ME, Imad AA, Marsha CK. Semiconducting metal oxide basedsensors for selective gas pollutant detection. Sensors 2009;9:8158e96.

58. Ahmad IA. Metal/metal-oxide nanoclusters for gas sensor applications. J Nanomater2016;16:1e717.

59. Steele M, MacIver B. Palladium/cadmium-sulfide Schottky diodes for hydrogendetection. Appl Phys Lett 1976;28:687e1687.

60. Toohey MJ. Electrodes for nano dot-based gas sensors. Sens Actuators B 2005;105:232e50.


61. Lundstrom I, Shivaraman MS, Svensson S. A hydrogen sensitive Pd-gate MOStransistors. J Appl Phys 1975;46:3876e81.

62. Ito K. Hydrogen sensitive Schottky barrier diodes. Surf Sci 1979;86:345e52.63. Cheng C, Tsai Y, Lin K, Chen H, Hsu W, Chuang H, Chen C, Liu W. Hydrogen

sensing characteristics of Pd- and Pt-Al0.3Ga0.7As metal semiconductor(MS) Schottkydiodes. Semicond Sci Technol 2004;19:778e82.

64. Kim J, Gila B, Chung G, Abernathy C, Pearton S, Ren F. Hydrogen sensitive GaNSchottky diodes. Solid State Electron 2003;47:1069e73.

65. Basu S, Roy S, Jacob C. Ruthenium as Schottky metal for SiC-based high temperaturehydrogen sensors. Mater Technol Hydrog Econ 2004;801:193e8.

66. Salehi A, Nazerian V. Characterization of magnetic Ni/n-Si Schottky contact forhydrogen gas sensing applications. Sens Actuators B 2007;122:572e7.

67. Pandis C, Brilis N, Bourithis E, Tsamakis D, Ali H, Krishnamoorthy S, Iliadis A,Kompitsas M. Low-temperature hydrogen sensors based on Au nanoclusters andSchottky contacts on ZnO films deposited by pulsed laser deposition on Si andSiO2 substrates. IEEE Sens J 2007;7:448e54.

68. Hossein-Babaei F, Purahmad M. Wide dynamic range hydrogen sensing using silver-rutile Schottky diode. In: Proceedings of the third international conference on sensing technol-ogy, Taipei, Taiwan, 30 Novembere03 December 2008; 2008. p. 72e5.

69. Song J, Lu W, Flynn J, Brandes G. AlGaN/GaN Schottky diode hydrogen sensorperformance at high temperatures with different catalytic metals. Solid State Electron2005;49:1330e4.

70. Golovanov V, Liu CC, Kiv A, Fuk D, Ivanovskaya M. Microfabricated one-electrodeIn2O3 and Fe2O3 composite sensors. Solid State Phys 2009;13:68e73.

71. Tamaki J, Miyaji A, Makinodan J, Ogura S, Konishi S. Effect of micro-gap electrodeon detection of dilute NO2 using WO3 thin film microsensors’. Sens Actuators B 2005;108:202e6.

72. Shaalan NM, Yamazaki T, Kikuta T. Influence of morphology and structure geometryon NO2 gas-sensing characteristics of SnO2 nanostructures synthesized via a thermalevaporation method. Sens Actuators B 2011;153:11e6.

73. Meixner H, Lampe U. Metal oxide sensors. Sens Actuators B 1996;33:198e202.74. Korotcenkov G. Handbook of gas sensor materials. New York (NY), USA: Springer;

2013.75. Hoefer U, K€uhner G, Schweizer W, Sulz G, Steiner K. CO and CO2 thin-film SnO2

gas sensors on Si substrates. Sens Actuators B 1994;22:115e9.76. Michel HJ, Leiste H, Halbritter J. Structural and electrical characterization of PVD-

deposited SnO2 films for gas esensor application. Sens Actuators B 1995;24e25:568e72.

77. Sozza A, Dua C, Kerlain A, Brylinski C, Zanoni E. Long-term stability of Ti-Pt-Aumetallization system for Schottky contact and first-level metallization on SiCMESFET. Microelectron Reliab 2004;44:1109e13.

78. Capone S, Epifani M, Francioso L, Kaciulis S, Mezzi A, Siciliano P, Taurino A. Influ-ence of electrodes ageing on the properties of the gas sensors based on SnO2. SensActuators B 2006;115:396e402.

79. Lee J, Kim E. Effect of structural and morphological changes on the conductivity ofstretched PANI-DBSA/HIPS film. Bull Korean Chem Soc 2011;32:2661e5.

80. Skotheim TA.Handbook of conducting polymers. New York (NY), USA: Marcel Dekker;1986.

81. Ray S, Easteal AJ, Cooney RP, Edmonds NR. Structure and properties ofmelt-processed PVDF/PMMA/polyaniline blends. Mater Chem Phys 2009;113:829e38.


82. Singh SK, Gupta RK, Singh RA. Optical, mechanical, and electrical studies ofpolymer composites based on charge transfer complex of phenothiazine-iodine withpoly(vinyl chloride). Synth Met 2009;159:2478e85.

83. Zhang DH. On the conductivity measurement of polyaniline pellets. Polym Test 2007;26:9e13.

84. Ansari R, Raofie F. Removal of mercuric ion from aqueous solutions using sawdustcoated by polyaniline. Eur J Chem 2006;3:35e43.

85. Holmes P, Loasby R. Handbook of thick film technology. Glasgow, UK: ElectrochemicalPub.; 1976.

86. Song K, Joo B, Choi N, Lee Y, Lee S, Huh J, Lee D. A micro hot-wire sensors for gassensing applications. Sens Actuators B 2004;102:1e6.

87. Korotcenkov G, Boris I, Brinzari V, Lychkovsky Y, Karkotsky G, Golovanov V,Cornet A, Rossinyol E, Rodriguez J, Cirera A. Gas sensing characteristics of one-electrode gas sensors on the base of doped In2O3ceramics. Sens Actuators B 2004;103:13e22.

88. Jones E. The pellistor catalytic gas sensor. In: Moseley P, Tofield B, editors. Solid-stategas sensors. Bristol, UK: Institute of Physics; 1987. p. 17e31.

89. Fukui K. Detection and measurements of odour by sintered tin oxide. Sens Actuators B1991;5:27e32.

90. Malchenko S, Lychkovski Y, Baykov M. In2O3-based gas sensors. Sens Actuators B1993;13e14:159e61.

91. Malchenko S, Lychkovski Y, Baykov M. One electrode semiconductor sensors fordetection of toxic and explosive gases in air. Sens Actuators B 1992;7:505e6.

92. Faglia G, Comini E, Sberveglieri G, Rella R, Siciliano P, Vasanelli L. Square andcollinear four probe array and Hall measurements on metal oxide thin film gassensors. Sens Actuators B 1998;53:69e753.

93. Williams D. Semiconducting oxides as gas-sensitive resistors. Sens Actuators B 1999;57:1e16.

94. Gerlich M, Kornely S, Fleischer M, Meixner H, Kassing R. Selectivity enhancementof a WO3/TiO2 gas sensor by the use of a four-point electrode structure. Sens ActuatorsB 2003;93:503e8.

95. Bl€aser D, R€uhl T, Diehl C, Ulrich M, Kohl D. Nanostructured semiconductor gassensors to overcome sensitivity limitations due to percolation effects. Phys A 1999;266:218e23.

96. Lundstrom I. Hydrogen sensitive MOS structures, Part 1, Principles and applications.Sens Actuators B 1981;1:403e26.

97. Abe H, Esashi M, Matsuo T. ISFET’s using inorganic gate thin films’. IEEE TransElectron Devices 1979;26:1932e9. ED.

98. Sibbald A, Whalley P, Covington A. A miniature flow-through cell with a four-junction chemFET integrated circuit. Anal Chem Acta 1984;159:47e62.

99. Bergveld P. Development of an ion-sensitive solid state device for neuro-physiologicalmeasurements. IEEE Trans Biomed Eng 1990:70e1. BME-17.

100. Mariucci L, Fortunato G, Pecora A, Bearzotti A, Carelli P, Leone R. Hydrogenatedamorphous silicon technology for chemical sensing thin film transistors. Sens ActuatorsB 1992;6:29e33.

101. Litovchenko VG, Gorbanyuk TI, Efremov AA, Evtukh AA. Effect of macrostructureand composition of the top metal electrode on properties of MIS gas sensors.Microelec-tron Reliab 2000;40:821e4.

102. Qiu YY, Azeredo-Leme C, Alcacer LR, Franca JE. A CMOS humidity sensor withon-chip calibration. Sens Actuators A 2001;92:80e7.

103. Gu L, Huang Q, Qin M. A novel capacitive-type humidity sensor using CMOS fabri-cation technology. Sens Actuators B 2004;99:491e8.


104. Eriksson M, Salomonsson A, Lundstr€om I. The influence of the insulator surface prop-erties on the hydrogen response of field-effect gas sensors. J Appl Phys 2005;98.0349031e0439036.

105. Fukuda H, Kasama K, Nomura S. Highly sensitive MISFET sensors with porous PteSnO2 gate electrode for CO gas sensing applications. Sens Actuators B 2000;64:163e8.

106. Trinchi A, Kandasamy S, Wlodarski W. High temperature field effect hydrogen andhydrocarbon gas sensors based on SiC MOS devices. Sens Actuators B 2008;133:705e16.

107. Johnson CL, Wise KD, Schwank JW. A thin-film gas detector for semiconductor pro-cess gases. IEDM Tech Dig 1988:662e3.

108. Suehl JS, Cavicchi RE, Gaitan M, Semancik S. Tin oxide gas sensor fabricated usingCMOS micro-hotplates and in situ processing. IEEE Electron Device Lett 1993;4:118e20.

109. Chan PCH, Yan G, Sheng LY, Sharma RK, Tang Z, Sin JKO, Hsing IM, Wang YY.An integrated gas sensor technology using surface micro-machining. In: Proceedings ofthe MEMS 2001, Interlaken, Switzerland, 21e25 January, 2001; 2001. p. 543e6.

110. Yan G, Tang Z, Chan PCH, Sin JKO, Hsing IM, Wang Y. An experimental study onhigh-temperature metallization for micro-hotplate-based integrated gas sensors. SensActuators B 2002;86:1e11.

111. Dai CL, Xiao FY, Juang YZ, Chiu CF. An approach to fabricating microstructures thatincorporate circuits using a post-CMOS process. J Micromech Microeng 2005;15:98e103.

112. Dai CL. A capacitive humidity sensor integrated with micro heater and ring oscillatorcircuit fabricated CMOS-MEMS technique. Sens Actuators B 2007;122:375e80.

113. Janata J, Josowicz M. Suspended gate field effect transistors modified with polypyrroleas alcohol sensor. Anal Chem 1986;58:514e7.

114. Hauptman P. Sensors, principles and applications. Hertfordshire, UK: Prentice Hall; 1991.115. Lee SP, Park KJ. Humidity sensitive field effect transistors. Sens Actuators B 1996;

35e36:80e4.116. Lee KN, Seo YT, Yoon S, Lee MH, Kim YK, Seong WK. Chemical gating experi-

ment of a nano-field-effect transistor sensor using the detection of negative ions in air.Sens Actuators B 2016;236:654e8.

117. Jain U, Harker A, Stoneham A, Williams D. Effect of electrode geometry on sensorresponse. Sens Actuators B 1990;2:111e4.

118. Williams D, Pratt K. Theory of self-diagnostic sensor array devices using gas-sensitiveresistors. J Chem Soc Faraday Trans 1995;91:1961e6.

119. Vilanova X, Llobet E, Brezmes J, Calderer J, Correig X. Numerical simulation of theelectrode geometry and position effects on semiconductor gas sensor response. Sens Ac-tuators B 1998;48:425e31.

120. Gardner J. Intelligent gas sensing using an integrated sensor pair. Sens Actuators B 1995;26e27:261e6.

121. Hoefer U, Steiner K, Wagner E. Contact and sheet resistances of SnO2 thin films fromtransmission-line model measurements. Sens Actuators B 1995;26:59e63.

122. Hoefer U, B€ottner H, Felske A, K€uhner G, Steiner K, Sulz G. Thin-film SnO2 sensorarrays controlled by variation of contact potentialeeA suitable tool for chemometricgas mixture analysis in the TLV range. Sens Actuators B 1997;44:429e33.

123. Reeves G, Harrison H. Obtaining the specific contact resistance from transmission linemodel measurements. IEEE Electron Device Lett 1982;3:111e3.

124. Neamen D. Semiconductor physics and devices. New York (NY), USA: McGraw-Hill;2003.


125. Tyagi M. Physics of Schottky barrier junction. In: Sharma B, editor. Metal-semiconductor Schottky barrier junction and their application, NY, USA, Plenum, NewYork; 1984.

126. Schottky W. Naturwissenchaften 1938;26:843.127. Mott N. Note on the contact between a metal and an insulator or semiconductor. Proc

Camb Philol Soc 1938;34:568e72.128. Islam M, Hakim M. Electron affinity and work function of polycrystalline SnO2 thin

film. J Mater Sci Lett 1986;5:63e5.129. Feucht D. Heterojunctions in photovoltaic devices. J Vac Sci Technol 1977;14:57e64.130. Gutmann S, Wolak M, Conrad M, Beerbom M, Schlaf R. Effect of UV and X-ray

radiation on the work function of TiO2 surfaces. J Appl Phys 2010;107:103705.131. Halas S, Durakiewicz T. Work functions of elements expressed in terms of the Fermi

energy and the density of free electrons. J Phys Condens Matter 1999;10:10815e25.132. Methfessel M, Hennig D, Scheffler M. Trends of the surface relaxations, surface en-

ergies, and work functions of the 4d transition metals. Phys Rev 1992;B13:4816e29.133. Akbi M. A method for measuring the photoelectric work function of contact materials

versus temperature. IEEE Trans Compon Packag Manuf Technol 2014;4:1293e302.134. Kal W, Frese Jr . Simple method for estimating energy levels of solids. J Vac Sci Technol

1979;16:1042e4.135. Bardeen J. Surface states and rectification at a metal-semiconductor contact. Phys Rev

1947;71:717e27.136. Milnes A, Feucht. Heterojunctions and metal-semiconductor junctions (chap. 6 and chap. 7).

New York: Academy Press; 1972.137. Mead C. Metal-semiconductor surface barriers. Solid State Electron 1966;9:1023e32.138. Andrew J, Phillips J. Chemical bonding and structure of metal-semiconductor

interfaces. Phys Rev Lett 1969;22:1433e6.139. Wagner C. Theory of current rectifier. Phys Z 1931;32:641e5.140. Schottky W, Spenke E. Quantitative treatment of the space charge and boundary-

layer theory of the crystal rectifier. Wiss Veroff Siemens-Werken 1939;18:225e91.141. Crowell C, Sze S. Current transport in metal-semiconductor barrier. Solid State Electron

1966;9:1035e48.142. Padovani F, Stratton R. Field and thermionic-field emission in Schottky barriers. Solid

State Electron 1966;9:695e707.143. Crowell C, Rideout V. Normalized thermionic-field emission in metal-

semiconductor barriers. Solid State Electron 1969;12:89e105.144. Bachrach RZ. Interface chemistry and structure of Schottky barrier formation. In:

Sharma B, editor. Metal-semiconductor Schottky barrier junction and their application. NewYork, USA: Plenum; 1984.

145. Heine V. Theory of surface states. Phys Rev 1965;A138:1689e96.146. Louie SG, Chelikowsky JR, Cohen ML. Ionicity and the theory of Schottky barriers.

Phys Rev B 1977;15:2154e62.147. Cowley A, Sze S. Surface states and barrier height of metal-semiconductor systems.

J Appl Phys 1965;36:3212e20.148. Wada O, Majerfeld A, Robson P. InP Schottky contacts with increased barrier height.

Solid State Electron 1982;25:381e7.149. Lundstrom I, Svensson C. In: Janata J, Huber J, editors. In solid state chemical sensors.

New York: Academic Press; 1985.150. Lee C, Yan J. Sensing mechanisms of Pt/b-Ga2O3/GaN hydrogen sensor diodes. Sens

Actuators B 2010;147:723e9.151. Ruths P, Ashok S, Fonash S, Ruths J. A study of Pd-Si MIS Schottky-barrier diode

hydrogen detector. IEEE Trans Electron Devices 1981;ED-28:1003e9.


152. Johansson M, Lundstrom I, Ekedahl L. Bridging the pressure gap for palladium metal-insulator-semiconductor hydrogen sensors in oxygen containing environments. J ApplPhys 1998;84:44e51.

153. Lundstr€om I, Sundgren H, Winquist F, Eriksson M, Krantz-Rulcker C, Lloyd-Spetz A. Twenty-five years of field effect gas sensor research in Linkoping. Sens Actu-ators B 2011;157:329e52.

154. Eggins B. Chemical sensors and biosensors. New York: Wiley; 2002. p. 49.155. Yanagida H. Gas sensing characteristics of porous zinc oxide ceramic with and without

platinum catalyst. In: Proc. int. meet. on chemical sensors; 1983. p. 95e100.156. Pajonk G, Teichner S, Germain J. Studies in surface science and catalysis. New York:

Elsevier; 1983. p. 17.


CHAPTER FOUR

Introduction to semiconductorgas sensors: a block schemedescriptionArnaldo D’Amico, Corrado Di NataleDepartment of Electronic Engineering, University of Rome Tor Vergata, Roma, Italy

Contents

4.1 Introduction 1334.2 The sensor blocks 135

4.2.1 The response curve 1364.2.2 Sensitivity 1374.2.3 Resolution 1384.2.4 Example of the evaluation of resolution 139

4.3 Metal oxide semiconductor capacitor: the case of the hydrogen gas sensitivity ofPd-SiO2-Si

142

4.4 Light-addressable potentiometric sensor 1444.5 Metal oxide semiconductor field-effect transistor 1484.6 Metal oxide semiconductors 151

4.6.1 SnO2 bands 1514.6.2 Band diagram modulation 153

4.7 Conclusions 156References 156

4.1 Introduction

Several human activities require the measure of the concentration ofgases and volatile compounds. Industrial processes and pollution controlhave been the traditional applications of gas sensors.1 In the last years, a num-ber of novel and fascinating applications emerged, not least the detection ofthe volatile metabolites as biomarker for several diseases.2

In this context, analytical instruments are ready to provide the necessaryaccuracy for monitoring gases and vapors. However, although the great effortsof miniaturization of traditional instruments, such as gas chromatographs and



https://doi.org/10.1016/B978-0-08-102559-8.00004-5

mass spectrometers,3,4 these are still bulky, expensive, and require trained andskill technicians for proper operation and for their data analysis andinterpretation.

Solid-state sensors have been always considered the solution to over-come some of these drawbacks and to facilitate gas analysis avoiding themandatory use of dedicated laboratories.

Furthermore, solid-state sensors can take advantage of the micro/nanoe-lectronic technologies; thus, they can be cheap and reliable and furthermorereadily available for the latest information technologies such as the so-calledInternet of things.5 Semiconductor gas sensors are solid-state devices basedon semiconductor materials and/or semiconductor devices. Then, whenwe talk of semiconductors, we may intend either those materials (inorganicor organic) that show some intrinsic sensitivity to gases or the devices thatconstitute the basis for transducers such as, for instance, those based on fieldeffects where changes in the electronic properties of the gate can controleither the output current or voltage.

The parameters of devices and materials depend on the quality andquantity of the gases and vapors to which they are in contact. Semiconductormaterials, particularly in transducer systems, offer the perspective of integra-tion of the sensors in the microelectronic realm taking advantage of theintegrated electronic interfaces for signal processing.

In this paper we will consider a variety of sensors. Before describing theworking principles of sensors, it is worthy to introduce some basic definitionsof sensor properties.6,7 These definitions define a common language necessaryfor an objective comparison among sensors.

Ideal gas sensors should be sensitive, selective, and stable. Additional prop-erties are reversibility, accuracy, response time, and linearity. For many appli-cations, additional properties such as low cost, small size, unsensitivity toradiation and temperature, and robustness are also required. Finally, sensorsshould always be compatible with standard preprocessing electronics.

In practice, being almost impossible to satisfy at once all these constraintsan acceptable compromise is always sought satisfying only a part of theabove-mentioned features.

In this paper, we will describe the sensors through a block scheme whichhelps to decompose the sensor principle in a number of elementary stepseach endowed with its own sensitivity. The scheme might help to comparedifferent sensors emphasizing where the different performance arises. Thus,instead of paying attention to the basic theories, widely available in literature

134 Arnaldo D’Amico and Corrado Di Natale

and in many chapters of this book, this chapter stays focused on the basicintrinsic sensitivities.

To facilitate the comprehension of the basic mechanisms of the sensorsbased on field effects, we will consider as an example the chemical sensitivityof palladium films with respect to the hydrogen gas.8 This example pointsout the basic properties of the devices and it can be easily generalized tosemiconductor devices made of either inorganic or organic materials.

4.2 The sensor blocks

Fig. 4.1 shows a complete block scheme of a general sensor system.Not only the whole system can be considered as a sensor but also elementsor subsets of the scheme can be identified as sensors itself. The wholeappraisal of such a scheme is mandatory to understand the role of devices.

Sensors are the interface between the electronic circuits and the outerworld. It is simple to understand that the properties of sensor signals, namelythe electric signals depending on the outer world quantities, are determinedboth by the features of the sensor as a device and by the electric circuit atwhich the sensor is connected.

sensorY=f(M)

electronicinterface amplifier filter A/D

conversion

μP

storage communication

actuation

Outer worldmeasurand

ener

gy

N

v,iv,iv,iY

display

Figure 4.1 General scheme of a sensor system. The intrinsic sensor is the interfacebetween the electronic system and the outer world.

Introduction to semiconductor gas sensors: a block scheme description 135

Fig. 4.1 shows the general scheme of a sensor-based electronic system.The intrinsic sensor element provides the interface with the outer world.The parameter Y is the electric parameter affected by the interaction withthe gas. This may be, for instance, the resistivity, dielectric constant, orwork function. The quantity Y itself is not observable until it is transformedby the interface circuit into a voltage or a current signal. The signal can thenbe amplified and filtered to facilitate its measurement. The measurement isusually performed through an analog to digital conversion, and the digitalsignal is processed to actuate some action on the outer world (e.g., a regu-lation system), stored, communicated, or displayed. Finally, it is important tokeep in mind that each part of the system needs energy.

All steps of the processing chain contribute to the overall sensitivity. Eachblock is characterized by a proper input/output relationship and the combi-nation of all of them makes the total sensitivity.

4.2.1 The response curveThe response is any quantity useful to represent the state of a sensor as adevice or of the system at which the sensor is connected or of a sensor systemthat can be simple (formed by only one block) or rather complex made bymore than one block as described in Fig. 4.1.

In some cases, the response is defined in reference to a baseline value thatis the response in absence of stimulus. In such a case, the response can begiven as the difference, ratio, or relative change with respect to the baseline.

Independently from the definition, the response is a function of theconcentration of the gas. The functional relationship depends on the natureof the intrinsic sensing element and in case of a sensor system on the circuitparameters.

The response curve (see Fig. 4.2) is usually represented in a Cartesian planeidentified by two axes carrying the sensor signal and the gas concentration(generally called measurand), respectively. Because of the unavoidable noise,the origin of the plot cannot be reached, but rather the curve stops in corre-spondence of the noise level of the output signal.

The response curve is the result of the sensor calibration. This operationcorresponds in measuring the sensor signal when the device is exposed toknown concentrations. The fitted experimental points give rise to theanalytical response curve that is normally used to estimate the concentrationonce the output of the sensor is measured.


4.2.2 SensitivityIt is defined as the derivative of the response curve with respect to themeasurand:

S ¼ dVout

dM(4.1)

The sensitivity is the slope of the response curve (Fig. 4.2), and becausethe response is limited, the increase of sensitivity narrows the range ofconcentration. In case of gas sensors, a large sensitivity implies the capabilityto measure tiny changes of concentration.

Where the response curve is flat, the sensitivity is null; this means that inthis region any small change of the measurand does not change the outputsignal.

Fig. 4.3 shows a subset of blocks of Fig. 4.1; it represents the steps leadingfrom the concentration to the measured digital quantity. Taking intoconsideration these steps, it is convenient to express the partial sensitivitiesaccording to the nomenclature given in the same figure.

It appears clearly that the overall sensitivity (the sensitivity of all thechain) is only partially related to the sensitivity of the first block, which isthe most important. Rather the whole sensitivity includes also many stepsof the electronic processing of the signal. The sensitivity of the intermediateblocks may be called the internal sensitivities. As shown along this chapter,the internal sensitivity can be further divided in many others, depending onthe complexity of the sensor system.

Vout, max

Vout, min=Vnoise

Vout

measurand MmaxMmin

Figure 4.2 Example of a typical response curve, when the sensor response is theoutput voltage signal (Vout).


Fig. 4.3 shows elements of a typical sensor system. The intrinsic sensor ischaracterized by a parameter that is affected by the measured M, the sensoris connected to an interface circuit and then any variation of Y produces avariation of a signal v0, the signal can be amplified giving rise to a largersignal va, noisy components could be removed by a filter and vf is thefiltered signal, and finally the analog signal can be converted in a number(N) by an analog to digital converter. The definition of sensitivity can beapplied to any block of the chain. Note that in case of the amplifier, thesensitivity is an alternate definition of the gain and the two quantities arecoincident when the output of the amplifier is linearly proportional tothe input. The total sensitivity (dN/dM) is then obtained with as the prod-uct of the sensitivity of each block.

4.2.3 ResolutionThe resolution is the necessary consequence of the peculiar nature of thequantity represented in the vertical axis of the response curve. Indeed, thesensor signal is the result of a measurement and it is subject to measurementerrors. Errors of measurement are contributed by both the finite accuracy ofthe measurement instrument (in modern electronic systems this is ruled bythe analog to digital conversion) and by the electronic noise. The electronicnoise fixes the lower limit for the measurement error.

The resolution (Mres) is the smallest measurable change of the measurand.Given a sensor signal Vout, the corresponding resolution is

Mres¼ limVout/Vnoise

Vout

S¼ Vnoise

S(4.2)

S = dNdM

= dNdv f

×dv fdva

× dvadv0

× dv0dY

× dYdM

= dfADdv f

× dfFdva

× dfAdv0

× dfCdY

× dfSdM

Figure 4.3 The total sensitivity is given by the product of the sensitivities (derivatives ofthe transfer functions) of all the blocks of the chain leading from the measurand to themeasured quantity.


The sensitivity is calculated in the neighbor of the measurand at which Vout

is measured. Of course, in case of a nonlinear response curve, both theresolution and the sensitivity are functions of the measurand.

Vout can be considered at the output of each block in Fig. 4.3. As eachelectronic block adds its own noise, shorter the block chain, better theresolution.

It is worth pointing out that noise and sensitivity measurements are ofparamount importance for the sensor characterization to evaluate the reso-lution of a given sensor. Different types of noise may be encountered whenwe are dealing with sensors, and not all sensors have the same noise, whilesome sensors may show more than one kind of noise. In this context, therelevant parameter for the noise characterization is the noise spectral densitys( f ). The root mean square of the output signal (Vrms) is the measurablemanifestation of the noise. The relationship between S( f ) and the rootmean square of the output signal is

Vrms ¼

264Z f2

f1sðf Þdf

3751=2

(4.3)

The integral is calculated in the frequency interval practically defined bythe measurement time. Shorter the measurement time, wider the frequencyinterval and then the noise contribution to the sensor signal.

The most important types of noise are listed in Table 4.1.9 Excess noisesare manifested as an additional contribution to the current and then theyoccur only in biased materials.

It is worth mentioning that in gas sensors, additional sources of noisecome from the fluctuation of the concentration at the sensor surface ofthe compounds at which the sensor is exposed and from the fluctuationof adsorption/desorption processes. These terms contribute to the overallspectral density of sensor noise as an additional flicker-like noise.10

4.2.4 Example of the evaluation of resolutionTo fix the ideas about the relationship between sensitivity and resolution,let us consider the simple circuit in Fig. 4.4 where a resistance temperaturedetector (RTD) is connected to a simple readout circuit. The relationshipbetween temperature and resistance can be easily generalized to any resis-tive gas sensor.

Let us consider the sensor represented by a linearized response curve


RSðTÞ¼R0$½1þ a$ðT � T0Þ� (4.4)

R0 is the resistance known at the reference temperature T0 and a is thetemperature coefficient.

The block scheme of the sensor in Fig. 4.4 is shown in Fig. 4.5. TheRTD is made of a material whose resistivity is affected by changes of tem-perature. The transducer circuit that generates the sensitive signal is made bythe current source I0, the signal is then amplified by the noninverting ampli-fier, and, in case, it can be converted into a digital quantity.

The overall sensitivity of Vout with respect to the temperature can bewritten as

Table 4.1 Characteristics of the most typical noises found in semiconductors. c, c0,and c00 are constants; g is a factor close to 2; a is a factor close to 1; d is a factorranging from 2 3; b is a factor ranging from 0.8 to 3. q is the electron charge and k isthe Boltzmann constant. s is the recombination time of generationerecombination(GR) processes. Finally, u ¼ 2pf, i is current, sv is spectral density of voltage, and si isspectral density of current.Denomination Definition Spectral density

Thermal noise Manifestation of thethermal motion ofelectric charges(electrons in solids); itsmagnitude isproportional to theresistance.

sv ¼ 4$k$T$R

Shot noise Excess noise typical forspace charge regionsand therefore presentin all junction devices(e.g., diodes).

si ¼ 2$q$i

Flicker noise Excess noise likely dueto a continuousdistribution of traps,typical forsemiconductors.

sv ¼ c Vg

ua

Burst noise Excess noise emergingin semiconductorslikely due to impurityatoms.

sn ¼ c0V d

ub

GR noise Excess noise due to GRprocesses insemiconductors.

sn ¼ c00 s1þu2s2


dVout

dT¼ dVout

dVin

dVin

dRS

dRS

dT¼ A$ST$SR ¼

�1þ R2

R1

�$I0$R0a (4.5)

where A is the amplification, ST is the sensitivity of the transducer circuit,and SR is the intrinsic sensitivity of the thermistor.

The resolution can be directly calculated from the definition.

Tres ¼ Vnoise

A$ST$SR(4.6)

Note that the magnitude of noise depends on the single block character-istics; for instance, the amplifier amplifies both the signal and the noise andthen it is irrelevant in improving the resolution, rather it worsens the reso-lution because it adds its own noise to the total signal.

This last expression makes evident that the noise of the output signal andthe sensitivity of each block are necessary to evaluate the sensors performance.On the other hand, the knowledge of the intrinsic sensor (the temperaturesensor in this case), although fundamental for the final performance, does

R(T)

ρ(T)

Transducercircuit Amplifier A/D

conversion

ΔT

ρ+Δρ

(R+ΔR) (R+ΔR)*I (R+ΔR)*I*AN

Figure 4.5 Block diagram of the sensor systems based on the circuit of Fig. 4.4.

R2

R1

I0 RS(T) Vout (T)

Figure 4.4 RS is a resistive sensor, a resistance temperature detector (RTD) in theexample, connected to a noninverting amplifier.


not provide any knowledge about the actual performance until the electronicblocks are defined and characterized.

4.3 Metal oxide semiconductor capacitor: the case ofthe hydrogen gas sensitivity of Pd-SiO2-Si

The first solid-state gas sensor that we take into consideration is themetal oxide semiconductor (MOS) capacitor schematically shown inFig. 4.6. Let us consider the case when the gate metal is palladium.11 Thecatalytic properties of palladium to favor the dissociative adsorption ofhydrogen gas and the following diffusion of atomic hydrogen are known,as well the H2 sensing properties of this device.

The basic structure is formed by a stack of four regions: an ohmic con-tact, p-type silicon, silicon dioxide, and the thin film of palladium. In thedepletion mode, the total capacitance Ct is given by the series of the oxideand the depletion layer capacitances.

In a reference atmosphere where hydrogen gas is not present, the differ-ence between the work functions of Pd and Si (p-type) is such to generate adepletion layer.12 As a consequence of H2 adsorption, the depletion layersize changes. This results in a variation of the depletion capacitance andthen in the total capacitance.

The intrinsic mechanism of sensitivity is reassumed in Fig. 4.7.The global sensitivity is the derivative of the total capacitance with

respect to the hydrogen concentration.It is worth pointing out that each block is characterized by a proper

response. In fact the first block is supposed to be the slowest being due tochemical reactions at the palladium surface and to the diffusion process of

P

Figure 4.6 Schematic of Pd-OS capacitor. The capacitance is the series of the oxidecapacitance (fixed) and the depletion layer capacitance (variable). The depletion layersize changes as a consequence of the exposure to H2.


hydrogen atoms from the palladium surface to the Pd/SiO2 interface. Theelectronic time constant should not be neglected; however, the overallresponse time is mainly controlled by the diffusion processes, which arerather slow being the diffusion rate of atomic hydrogen through palladiumof the order of 5 ms/Å.

From Fig. 4.7, the total sensitivity St can be expressed as

S¼ dVout

d½H2� ¼dVout

dCtot

dCtot

dxd

dxddVFB

dVFB

dDFdDFdd

ddd½H2� (4.7)

From the above expression, all the physical terms contributing to theoverall sensitivity are visible. The last term dd/d[H2] is the intrinsic contri-bution of palladium, while all the other terms, summarized in dCtot/dd, arerelated to the MOS structure, and then they depend on the oxide thick-ness, permittivity of oxide, semiconductor, and the doping concentrationin silicon, and the area of the MOS structure. Eventually, the first termdepends on the particular circuit used to convert the capacitance changesinto a voltage change.

Fig. 4.8 shows an example of a circuit that can be used to measure thevalue of the capacitor C as a function of the applied DC voltage V0.

The MOS capacitor is biased by the sum of a DC voltage (V0) and ACvoltage (vi); then the total applied voltage, if R1 ¼ R2, is VS¼ �V0 � Vm

cos(ut). It is important that Vm � V0. In this way, the AC signal is a smallperturbation necessary to extract a signal proportional to C, but C dependsalmost completely on V0. Of course vi must be as small as possible but of a

Δδ ΔΦ Δ VFB

Δ xd

Δ VoutΔCtot

Δ(H2)

Figure 4.7 The exposure to hydrogen gas concentration ([H2]) results in a dissociativeadsorption of H2 at the palladium surface. It produces a concentration of atomichydrogen that quickly diffuses toward the oxide surface. The layer of atomic hydrogenforms a dipole layer (d) at the oxide surface that can be interpreted as a change of thework function difference (DF) between palladium and silicon, the consequence ofwhich is a change of the flat band voltage (VFB) that is related to the size of the deple-tion layer (xd) and then to the total capacitance (DCtot). Finally, a proper transducercircuit can convert the capacitance changes into a voltage change (DVout).


sufficient level to give a measurable output with at least a signal to noise rationot less than about 6.

The output signal is given by

Vout¼R3CtotdVs

dt¼ R3CtotuVm sinðutÞ (4.8)

The above circuit can be used to measure the relationship between Ctot.

and V0; this is an outmost characteristic of MOS devices, and for sensingpurposes, all circuit parameters can be optimized and kept fixed to measurethe changes of Ctot.

The scheme shown in Fig. 4.7 describes the chain of sensitivity when theMOS capacitor is not biased in inversion. In case of inversion, the depletionlayer size is fixed, rather in this case, the change of the flat band voltageaffects the threshold voltage at which the inversion layer is formed. Thiscondition is exploited in the metal oxide semiconductor field-effect tran-sistor (MOSFET) device discussed below.

4.4 Light-addressable potentiometric sensor

The intrinsic photoconductivity of semiconductors is exploited inMOS structures to give rise to an interesting device called light-addressable potentiometric sensor (LAPS).13

As shown in Fig. 4.9, the output voltage across a load resistance RL isgenerated by an internal equivalent voltage generator which is developed

R2

R3

R1

R1V0

Figure 4.8 Example of a circuit to measure the capacitance of metal oxide semicon-ductor structure as a function of the applied voltage V0.


once pulsed light of suitable time width and of specific wavelength isabsorbed in the semiconductor.

In a first approximation, the electrical equivalent circuit of LAPS isshown in Fig. 4.10.

Vin is the internal voltage source generated across the depletion regionwhose magnitude is due, by the product of the depletion region impedanceand the current of the electrons and holes produced by the adsorbed pho-tons. Electrons and holes are separated by the built-in electric field locatedinside the space charge region; the capacitor is the series of the oxide andthe depletion capacitances, while RL is the load resistance. When the deviceis shined by pulses of light, the output voltage corresponds to the derivative

ε+qε

−qε

P

O

D

S

Figure 4.9 Principle of light-addressable potentiometric sensor. Photons absorbedin silicon give rise to electronehole couples. The couples generated in the space chargeregion are separated by the built-in electric field. The displacement charges the metaloxide semiconductor (MOS) capacitor and results in a signal observable across a loadresistor in series to the MOS capacitor. Keeping constant the photon flux, the signalis proportional to the volume of the space charge region and then it is affected bywork function changes.

ViRL

Figure 4.10 Light-addressable potentiometric sensor equivalent circuit.


of the voltage pulses (and not current pulses); the time constant is equal toCRL (see Fig. 4.11).

The gas sensing mechanism is illustrated by the block scheme inFig. 4.12.

The gas sensitivity is due to the same processes occurring in the MOScapacitor; the difference here is that the size of the depletion layer modulatesthe amount of electronehole pairs that are separated by the built-in electricfield. In practice, instead of measuring the change of the MOS capacitor,here a voltage proportional to the flux of impinging photons and to the

Lightintensity

V0

Time

Time

Figure 4.11 Typical output voltage of a light-addressable potentiometric sensor in caseof pulsed light.

Δδ ΔΦ ∆ VF B

Δxd

n − pCouples

i n ; i p

vout

Δ(H2)

Figure 4.12 Block diagram of light-addressable potentiometric sensor.


volume of the space charge region are generated. Actually the sensor is sen-sitive to both light intensity and gas concentration and the device becomesselective to gas only once the intensity of light has been kept constant.

S¼ dVout

d½H2� ¼dVout

dqdqdxd

$dxddVFB

dVFB

dDFdDFdd

ddd½H2� (4.9)

The quantity dq namely the amount of photo-induced electronehole pairsis proportional to the intensity of the impinging light. Nevertheless, thesensitivity of the LAPS is proportional to the intensity of light. In otherwords, the sensitivity lies in the number of electronehole couples related tothe adsorbing volume defined by the depletion layer width and the sectionarea of the device. It is worth to consider that the light pulse is a probe of theextension of the space charge region.

The dependence from the light introduces an additional noise because ofthe fluctuation of the light source; this is equivalent to a shot noise propor-tional to the root square of the light intensity.

In terms of sensitivity improvement, the key quantity is dxd/dVFB. Thedoping concentration of the semiconductor can actually determine a largervariation of the depletion region with respect to changes of the flat bandvoltage.

LAPS sensors can be easily integrated in arrays.14 Fig. 4.13 shows a picto-rial example. Each element of the array could be made of a different gatematerial or could be probed by light of different wavelengths. The lightcan be addressed sequentially to each element of the matrix. Then-multiple repetition of the scan and the storage data may allow a significantnoise reduction evaluated by a factor of

ffiffiffin

pwhere n is the number of light

scans per each element.

LightPt

W

Au

Pd

Figure 4.13 Pictorial view of an array of light-addressable potentiometric sensor.


4.5 Metal oxide semiconductor field-effect transistor

MOSFETs are among the most reliable and versatile transducers forgas sensor applications. In electronics the MOSFET structure has beencontinuously modified and improved along the years and it is at the basisof both analog and digital electronics.12

More than four decades ago, the MOSFET structure was transformedinto chemical sensors replacing the gate with either an electrolyte or a cat-alytic metal15,16 because that MOSFET has been the basis for the develop-ment of several sensors and biosensors for a plethora of different applications.

Being the MOSFET an extension of the MOS capacitor, also in this casethe sensitivity is activated anytime, the interaction between the gas and gateproduces an amount of charges or dipoles at the gate SiO2 interface.

This layer of charges or dipoles generates an additive gate voltage andthen a variation of the drain source current.

Fig. 4.14 shows a schematic drawing of a MOSFET. In its first imple-mentation as gas sensor, the gate was a film of hydrogen-sensitivepalladium.12

The charge control equation of the MOSFET in the quasi-linear regionis approximately12

Metal Metal

Oxide OxideMetalOxide

InversionDepletion

VS

VG

VB

VD

n+ n+

p+p-type Si

Metal

Figure 4.14 General scheme of a metal oxide semiconductor field-effect transistordevice.


IDS ¼mn$Cox$wL

�ðVGS � VT Þ$VDS � V 2

DS

2

�(4.10)

In the saturation region, the current becomes largely independent fromVDS:

IDS ¼mn$Cox$w2L

ðVGS � VT Þ2 (4.11)

The above relationship holds beyond the pinch-off point and neglectsthe effective channel length change due to VDS.

In the above formulas, VGS and VDs are the voltages applied between thegate and source and drain and source, respectively, and IDS is the currentfrom drain to source. mn is the effective electrons mobility in the channel,Cox is the oxide capacitance, w and L are the width and the length of thechannel, respectively, and VT is the threshold voltage. This is a key param-eter of the device because, at first approximation, only when VGS > VT thechannel is formed and the current can flow.

The threshold voltage depends on all the relevant quantities of the MOSstructure such as the flat band voltage and the oxide capacitance.

The saturation condition is obtained from Eq. 4.10 under the conditionthat the current IDS does not depend on VDS, namely dIDS/dVDS ¼ 0. Thiscondition is achieved when VDS ¼VGS � VT.

Eqs. (4.10) and (4.11) are plotted in Fig. 4.15. This is the so-called outputcharacteristics of the MOSFET. The MOSFET should be polarized to fixthe quiescent point in the saturation region.

The block scheme of the sensing mechanism of the Pd-FET is shown inFig. 4.16. It corresponds to the block scheme of MOS capacitor and LAPS,except that changes in the flat band voltage in this case affect the thresholdvoltage. Then if the biasing parameters of the MOSFET are kept constant,the IDS changes.

The original Pd-FET structure was modified to extend the sensitivity toother species than hydrogen and to use more sensitive materials. A first inter-esting development consisted in the use of ultrathin metal films as gate. Inthis condition, the metal film is not homogeneous but rather it is character-ized by a number of cracks that leave the oxide exposed to the ambient air.17

In this way, the analyte does not necessitate traveling through the metal butit can reach directly the oxide surface. This offered the possibility to usemore catalytic metal and to expand the sensitivity to other analytes suchas ammonia.18 Additionally, a cracked gate layer can also accommodate


nonconductive organic layers as chemically sensitive material. This opportu-nity has been exploited to develop porphyrins functionalized MOSFETs.19

The MOSFET structure was also fabricated with organic semiconduc-tors.20 These devices are usually made as thin film transistor architecture,and the small thickness of the organic semiconductor enables the accumula-tion mode and the sensitive materials as the organic semiconductor.21,22

0 1 2 3 4 5Drain-source voltage [V]

0

0.005

0.01

0.015

0.02

0.025

0.03

0.035

0.04

0.045

0.05D

rain

cur

rent

[mA

]Vgate= 1VVgate=1.5VVgate= 2VVgate=2.5VVgate= 3VVgate=3.5VVgate= 4VVgate=4.5VVgate= 5VVgate=5.5V

Figure 4.15 Currentevoltage curves of the metal oxide semiconductor field-effecttransistor where the gate voltage is kept as a parameter.

Δδ ΔΦ ∆ ςΦΒ

∆ςουτ

∆ ςΤ

Δ ΙΔΣ

Δ(H2)

Figure 4.16 Block diagram of Pd field-effect transistor (FET). The adsorption ofhydrogen gas changes the flat band voltage of the metal oxide semiconductor(MOS) structure, and the flat band voltage affects the threshold voltage and then thecurrentevoltage characteristics of the device.


4.6 Metal oxide semiconductors

MOS are probably the more diffused gas sensors. A number ofdifferent metal oxides can be actually used; the first of these sensors wasmade of ZnO23 but the most popular of these materials them became thetin oxide (SnO2).

24 In the following section, the working principles ofSnO2 are illustrated.

25

Imperfect stoichiometric ratio between tin and oxygen occurs in realmaterials; the vacancy of oxygen atoms leaves a quota of electrons of tinatoms not engaged in covalent bonds and weakly bond to their atoms. Atroom temperature, this results in a light doping of the material that showsan n-type character.

The surface chemistry of tin oxide is complex, and here the simplestreactions involving oxygen are described.

The principle of operation of SnO2-based semiconductor gas sensor ismainly related to the change in conductivity occurring when reducing gases(such as CO, CH4) interact with chemisorbed oxygen ions.

The adsorption of O2 at the surface of tin oxide results in charged speciesthat subtract electrons from the semiconductor conductance band. As aconsequence, a surface depletion layer appears and the surface conductivityis strongly reduced. The reduction of conductivity is manifested in polycrys-talline materials, where the current moves from grain to grain and then it issubject to the built-in potentials associated to the surface depletion layers.

The interaction with reducing gases removes the surface oxygen species,electrons are released in the conduction band and the intergrain potentialbarriers are also lowered, and as a consequence, the conductivity increases.

The above-described process is temperature-activated and normal oper-ation temperatures are between 150 and 600�C, depending on metal oxideand gas.

Surface addition of nanoparticles deposited can greatly improve thesensitivity and the selectivity.26 More recently, organic surface functionali-zation was also introduced.27,28

4.6.1 SnO2 bandsAt thermal equilibrium (no voltage applied), many situations are possible atthe SnO2 surface. As an example we consider the case of a SnO2-based COsensor for which conduction and valence band bandings are mainly becauseof the following charge modifiers: traps (due to surface defects), additives,


adsorbed oxygen and oxygen ions formation, and formation of CO2 in pres-ence of CO. Fig. 4.17 shows the block schemes of the sensing principle.

Here, we consider two main cases, taking into account that mixed situ-ations involving not all the modifiers may be possible.

Fig. 4.18 shows a typical cross section of a SnO2 sensor with two ohmiccontacts, a layer of polycrystalline SnO2 of a given thickness and an electricinsulator but thermal conductor substrate and a heater on the other face.Fig. 4.19 shows the band diagrams of a monodimensional sequence ofadjacent grains, in thermal equilibrium. The equilibrium condition isreached through a rearrangement of electrons across the interfaces betweengrains. In Fig. 4.19, the charge, in the deep depletion approximation, andthe electric field are shown.

When a voltage is applied across the electrodes, we can suppose that thepotential is distributed only across the depletion layer. This is a typicalassumption when electronic devices are modeled.12 Then the voltagedrop reduces the barrier height from one side of the contact between grain,and it results in a net current flowing in the material (see Fig. 4.20).

Clearly, the bulk region of the grains does not participate to the gas sensi-tivity or to the conduction. In nanocrystalline materials, the bulk is actuallyeliminated and the whole material is exploited for sensing and conductionpurposes.

Δφ Δσ–ID T

Figure 4.17 Block diagrams of metal oxide semiconductor gas sensors in case of CO, orany other reducing gas, detection. Oxygen ions adsorbed at the metal oxide surfaceinteract with airborne CO molecules, and as a consequence, the depletion layer sizeand the surface barrier are decreased. In a polycrystalline material, it results in a changeof the total conductivity. A transducer circuit, such as that shown in Fig. 4.4, transformsthe change of conductivity into a voltage signal.

Electriccontact

Electriccontact

Heater

Insulating substrate

Polycrystalline material

Figure 4.18 Cross section of a typical polycrystalline sensor.


4.6.2 Band diagram modulationEach step of the sensing principles illustrated in Fig. 4.17 in the block schemechanges the band diagram of the material at the metal oxideeair interface.Fig. 4.21 shows four steps leading from the equilibrium configuration ofinert material to the consequences of the interaction with a reducing gas.Here again the carbon monoxide is explicitly mentioned.

Fig. 4.21(a) shows the band diagram of inert material. This condition ismet either in vacuum or at a low temperature. Low temperature means lessthan about 150�C, namely when the adsorption of airborne oxygen is notfavored. At high temperature and in air, the adsorption of oxygen can takeplace. Oxygen can be adsorbed both as atomic and molecular species.

(a)

(b)

(c)

Figure 4.19 (a) Band diagram, (b) charge distribution, and (c) electric field in a mono-dimensional arrangement of metal oxide grains. Note that the chain is terminated atboth sides by a junction with the metal electrode. Fm and FSnO2 are the work functionsof the metal electrode and the sensor material, respectively. fi is the Schottky barrierbetween SnO2 and the metal electrode, fi is the intergrain potential barrier. ND isthe electrons density in the conduction band corresponding to the donor concentra-tion and xd is the depletion layer size at the intergrain interface.


In Fig. 4.21(b), the case of atomic oxygen adsorption is shown. The molecularoxygen undergoes a dissociative adsorption onto the metal oxide surface andtwo atomic oxygens are adsorbed. The bond is provided by two electronswhich are displaced from the conduction band to the oxygen atoms. As aconsequence, the surface region of the semiconductor is depleted of electrons,the bands bend upward, and a surface potential (qfS) and a work functionchange appear. Surface oxygen can further react, at the optimal temperature,with a reducing gas molecule (such as CO). The consequence is the formationof a volatile CO2 molecule and the release of an electron in the conductionband. This elicits a reduction of the surface barrier (qf0

S) and the workfunction.

The full picture is much more complex because of the multiple oxygenspecies, each adsorbed at different energy. Furthermore, the presence ofadditional species in air, e.g., water vapor, makes the involved chemistrymore complex. However, the above description provides a sufficientintroduction to the main phenomena involved in the gas sensitivity ofMOS.

E

V=0

V>0

V>0

Fermi level

S

E

E

φ

SφφS – VA

φS – VA

Figure 4.20 Changes in the conduction band diagram of the junction between twoadjacent grains at the equilibrium (V ¼ 0) and under bias positive and negative. VA isthe portion of the total applied voltage across a single junction. Owing to the numer-osity of grains, VA is small enough to ensure the quasi-equilibrium condition. Boththermionic and tunnel currents can be simultaneously present. Both the contributionsgive a current inversely proportional to the exponential of the barrier height.


In air at low tempearture In air at high tempearture In air and reducing gas athigh temperature: step 1

In air and reducing gas athigh temperature: step 2

(a) (b) (c) (d)

Figure 4.21 Band diagram modulation in the different steps of reducing gas detection. (a) In air at low temperature, (b) in air at high tem-perature, (c) in air and reducing gas at high temperature (step 1), and (d) in air and reducing gas at high temperature (step 2).

Introductionto

semiconductor

gassensors:a

blockschem

edescription

155

4.7 Conclusions

In this chapter, a general introduction to the topic of semiconductorgas sensors has been provided. General phenomena related to the sensingproperties of semiconductor materials and semiconductor devices havebeen introduced, discussing the properties of MOS (resistors) and MOSdevice (capacitors) and the related FET.

The discussion has been maintained at a general level focusing theattention on the basic processes responsible of the gas sensitivity. For thisreason, block schemes have been introduced to help the reader to localizethe sensitivity sources. The main purpose of these diagrams is to introduce adecomposition of the global sensitivity into elementary phenomena tomake evident where the sensitivity emerges, which are the steps to improvethe sensor, and finally how to modify the sensor to extend its property.

We would like to propose this approach as a general method to presentsensor properties. This in particular is necessary for novel sensors where ablock scheme with the partial sensitivity values enables the comparisonwith other similar or, sometimes, identical sensors. A last point is concernedwith the overall response time which is made up by the contribution of eachblock and then the knowledge of the response time of each elementaryelement is necessary for the development of more performant sensors.

References1. Lee D. IEEE Sens J 2001;1:214e24.2. Di Natale C, Paolesse R, Martinelli E, Capuano R. Anal Chim Acta 2014;824:1e17.3. Lu C, Whiting J, Sacks R, Zellers E. Anal Chem 2003;75:1400e9.4. Black W, Stocks B, Mellors J, Engen J, Ramsey J. Anal Chem 2015;87:6286e387.5. Potyrailo R. Chem Rev 2016;116:11877e923.6. D’Amico A, Di Natale C. IEEE Sens J 2001;1:183e90.7. D’Amico A, Di Natale C, Sarro P. Sens Actuators B 2015;207:1060e8.8. Conrad H, Ertl G, Latta E. Surf Sci 1974;41:435e46.9. Van der Ziel A. Noise in solid state devices and circuits. (New York, USA): J. Wiley; 1986.10. Falconi C, Di Natale C, Martinelli E, D’Amico A, Zampetti E, Gardner J, Van Vliet C.

Sens Actuators B 2012;174:577e85.11. Poteat T, Lalevic B. IEEE Trans El Dev 1982;29:123e9.12. Sze S, Ng K. Physics of semiconductor devices. 3rd ed. J. Wiley; 2006.13. Bratov A, Abramova N, Ipatov A. Anal Chim Acta 2010;678:149e59.14. Hu N, Ha D, Wu C, Zhou J, Kirsanov D, Legin A, Wang P. Sens Actuators A 2012;187:

50e6.15. Bergveld P. IEEE Trans Bio-Medical Eng 1970;17:70e1.16. Lundstrom I, Shivaraman S, Svensson C, Lundkvist L. Appl Phys Lett 1975;26:55e70.17. Spetz A, Helmerssson U, Enquist F, Armgarth M, Lundstr€om I. Thin Solid Films 1989;

177:77e93.18. Spetz A, Armgarth M, Lundstr€om I. J Appl Phys 1988;64:1274e83.


19. Andersson M, Holmberg M, Lundstrom I, Lloyd-Spetz A, Martensson P, Paolesse R,Falconi C, Proietti E, Di Natale C, D’Amico A. Sens Actuators B 2001;77:567e71.

20. Guillaud G, Al Sadoun M, Maitrot M, Simon J, Bouvet M. Chem Phys Lett 1990;167:503e6.

21. Torsi L. Dodalabapour Anal Chem 2005;382A:381A.22. Mabeck I, Malliaras G. Anal Bioanal Chem 2006;384:343e53.23. Seyama T, Kato A, Fujiishi K, Nagatani M. Anal Chem 1962;34.24. Barsan N, Weimar U. J Electroceramics 2001;7:143e67.25. Yamazoe N, Shimanoe K. Sens Actuators B 2011;158:28e34.26. Fine G, Cavanagh L, Afonja A, Binions R. Sensors 2010;10:5469e502.27. Sivalingam Y, Martinelli E, Catini A, Magna G, Pomarico G, Basoli F, Paolesse R, Di

Natale C. J Phys Chem C 2012;116:9151e7.28. Hijazi M, Rieu M, Stambouli V, Tournier G, Viricelle J, Pijolat C. Sens Actuators B

2018;256:440e7.



PART TWO

Materials

159j


CHAPTER FIVE

One- and two-dimensional metaloxide nanostructures forchemical sensingE. Comini, D. ZappaDepartment of Information Engineering, University of Brescia, Brescia, Italy

Contents

5.1 Introduction 1615.2 Deposition techniques 162

5.2.1 Two-dimensional nanostructures 1635.2.2 One-dimensional nanostructures 166

5.2.2.1 Vapor phase growth methods 1675.2.2.2 Liquid phase growth methods 1685.2.2.3 Template-assisted methods 169

5.3 Conductometric sensor 1695.3.1 Device integration 170

5.4 Transduction principles and related novel devices 1705.5 Conclusion and future trends 174References 175

5.1 Introduction

Metal oxides have very different electrical properties from metals,semiconductors, to insulators and are used in many different areas such assensors, superconductors, magnets, and lighting. In relation to chemicalsensing applications, the ability of metal oxides to change their electricalconductivity with the composition of the surrounding atmosphere hasbeen known for almost 60 years.1 In October 1968, the first generation ofcommercial devices was produced on a large scale by TGS (Taguchi GasSensor, now Figaro Engineering Inc.) in Japan. These sensors were madeof SnO2 thick films and were used for the detection of explosive gases.

Over the years, the demand for cheap, small, low power consuming butreliable solid-state chemical sensors has continued to grow. Consequently,



https://doi.org/10.1016/B978-0-08-102559-8.00005-7

significant research efforts have been made worldwide to improve on the“3Ss” (sensitivity, selectivity, and stability), mainly with empirical approaches,but also through some basic theoretical research and spectroscopy studies.

The metal oxide that has been paid the greatest attention for chemicalsensing is SnO2

2; however, other n-type semiconducting oxidesdsuch asTiO2,

3 In2O3,4 WO3,

5 ZnO,6 Fe2O3,7dhave been proposed and studied.

On the contrary, p-type oxidesdlike CuO8 and NiO,9dare not extensivelyinvestigated yet, mainly due to the lower performances expected.10 Further-more, the use of mixed oxides in form of heterostructures, as well as theaddition of noble metals or other functional materials,11 has been studiedto improve not only the sensitivity but also the selectivity and the stability.One of the most important articles in metal oxide chemical sensing waspublished in 1991 by Yamazoe.12 It was shown that, as the crystallite sizewas reduced, there was a huge improvement in sensor performance. Thisshifted the research community focus to investigate the performance ofmaterials with the smallest crystallite size, while ensuring that their propertiesremain stable during long-term, high-temperature operation, which isnecessary for metal oxide chemical sensing.

Another important discovery that changed the field of chemical sensingwas the synthesis of single crystal one-dimensional oxide nanowires. Thesematerials have great potential thanks to their reduced lateral dimensions andsingle crystal habits, both for fundamental study and for potential nanodeviceapplications, i.e., the third generation of metal oxide gas sensors. The firsttechnique used to fabricate these nanowires was the simple evaporation ofthe desired commercial metal oxide powders at high temperatures, followedby condensation at lower temperatures on the substrates. This technologywas developed back in the 1960s and is still the most widespread growthprocess, even if nowadays there are many other techniques that enable thesynthesis of nanostructures of different shape and size.

5.2 Deposition techniques

The technique of layer deposition and coating has a variety of indus-trial applications, such as protective layers, sensors, resistive films, andcatalyzers. We will briefly present the different deposition techniques as afunction of the nanostructures obtained for, two-dimensional nanostructures (thin films),, one-dimensional nanostructures (nanowires, nanorods, etc.).

162 E. Comini and D. Zappa

5.2.1 Two-dimensional nanostructuresThere are several preparation methods that have been optimized andproposed during recent years. We can, for example, distinguish betweenphysical vapor deposition (PVD) techniques, chemical vapor deposition(CVD) techniques, and techniques that do not require a vacuum.

In the case of PVD techniques, the source material to be deposited is inthe solid phase. An intermediate vapor phase is then formed and finally thesolid phase is deposited on the substrate. The method used for the vaporiza-tion of the source material distinguishes between the different depositiontechniques. For example, heat transfer is used in thermal and electronbeam evaporation (EBE), while bombardment by energetic ions is used insputtering processes.

The thermal evaporation (TE) technique is the oldest PVD process. Inthis technique, the substrate is kept at a short distance from the sourcematerial, which is heated until it vaporizes. The vapor then condenses onthe substrate.13,14

The equipment required to apply all these PVD methods works atpressures lower than ambient pressure: this is necessary to control thecomposition of the deposited material. To improve its purity, the meanfree path of the particles must be greater than the distance between thesource and the substrate. In the specific case of TE, this also allows a loweroperational temperature for the vaporization. A conventional depositionsystem consists of a vacuum chamber, a mechanical roughing pump, ahigh vacuum pump, a heated crucible, and a substrate holder.

In the case of TE, unintentional doping in the film can result from thehigh temperature of vaporization that has to be reached by the sourcematerial. Particular attention must be paid to the material composition ofthe crucible containing the source material and to the possible alloys thatit can form with the source material at the evaporation temperature.

EBE is similar to TE, except that the source material is vaporized by theheat transferred from an electron beam accelerated toward the source. Onthe impingement of the high-energy electrons, their kinetic energy isconverted into heat and the source material can reach temperaturesexceeding 3000�C, causing a local melting where the beam is focused.The advantage of this technique is that the electron beam can be focusedon specific areas of the source material, so that the interaction betweensource material and support materials can be reduced.15e17

One- and two-dimensional metal oxide nanostructures for chemical sensing 163

Another method based on evaporation is pulsed laser deposition. Thehigh photon flux incident on the target induces an essentially instantaneoustemperature increase that causes the evaporation of the target material.18e20

The experimental setup consists of a vacuum chamber in which there is thetarget material and a window through which the laser beam is focused ontothe target. The ejected material that arises from a target after laser irradiationis called an “ablation plume” (i.e., a plasmalike substance containing freeelectrons and ions, neutral particles, molecular fragments, and chemicalreaction products). The physical process of laser ablation is extremelycomplicated, and there are several key parameters involved, such as beamenergy density, the laser pulse duration, and the laser wavelength.

One of the most commonly used PVD techniques for industrial applica-tions is sputtering. The sputtering process was discovered by W.R. Grove in1852 while studying the discharge in a tube containing gas, but its use inindustry and research began in recent decades.

In this process, the surface of the sputtering target is bombarded withgaseous ions under high voltage acceleration. Atoms or entire moleculesof the target material are ejected and can reach the substrate. There is nomelting of the material: the ejection of the particles from the source material(target) is a result of the momentum transferred from the incoming particle.The conventional setup for sputtering is a vacuum chamber where theworking gas is introduced, a high negative voltage is applied to the target,and the positively charged ionized atoms are accelerated toward it.21e23

To sputter conductive materials, direct current sputtering is used,whereas for nonconductive materials, radio frequency must be appliedduring the sputtering process to prevent the target from charging up dueto the bombardment from positively charged ions. Another configurationis magnetron sputtering, where a magnetic field is added beneath the targetto deflect and confine electrons, allowing for lower working pressures. Insputtering, the phase transition is obtained mechanically rather thanchemically or thermally, so virtually any material can be deposited. Theenergy of the ejected molecules is higher with respect to TE and EBE,thus improving the crystallinity and adhesion of the thin film.

The second group of deposition techniques is CVD: a chemical reactiontransforms the molecules in the gas phase, known as the “precursors,” intosolid films or powders on the substrate. There are several configurationssuch as, low pressure chemical vapor deposition (LPCVD),, atmospheric pressure chemical vapor deposition,


, plasma enhanced chemical vapor deposition,, photochemical vapor deposition,, laser chemical vapor deposition, and, metal organic chemical vapor deposition.

In CVD, the reactant gases are diluted in carrier gases and introducedinto the reaction chamber at room temperature, while the deposition surfaceis heated. The energy necessary to start the desired chemical reaction can besupplied as thermal energy with resistive, radiant or inductive heating, or asphoton energy or glow discharge plasma. Depending on the workingconditions, the reactant may experience homogeneous chemical reactionsin the vapor phase before hitting the surface. Otherwise, when the reactantgases approach the surface, they slow down and heterogeneous reactionsoccur on the surface, forming the deposited material. Gaseous reactionby-products are then transported by the carrier gas out of the reactionchamber.

Whichever heating method is employed, CVD has to provide a volatileprecursor containing the elements that compose the deposited film, trans-port the precursor toward the substrate surface, enhance or reduce reactionsin the gas phase, and provide the surface reaction needed to form the film.The setup consists of a reaction chamber; gas/vapor delivery lines; theenergy source; vacuum systems (LPCVD); an exhaust system; and gasflow, pressure and temperature monitoring systems.24e26 Hazardous vaporsare also frequently used and may be produced by chemical reactions. Thus,safety equipment may be necessary. The advantages of CVD films are goodadhesion, good step coverage, and high versatility of materials, but adrawback is the formation of hazardous and corrosive by-products.

Beyond PVD and CVD, there are techniques from the liquid/solutionphase: such as solegel, spray coating, spin coating, electrochemical deposi-tion, and liquid phase epitaxy. The solegel process is the most widely usedmethod for the deposition of metal oxide for gas sensors.

The solegel process generally involves the transition from a liquid solinto a solid gel phase. Inorganic metal salts or metal organic compounds,such as metal alkoxides, may be used as precursors. The solegel depositionprocess usually has four steps:, colloidal particles are dispersed in a liquid (sol),, deposition of sol solution on the substrate by spraying, dipping, orspinning,

, polymerization of the particles in the sol by stabilizing component’sremoval (gel), and


, heat treatment to pyrolyze the remaining organic or inorganic compo-nents forming the final film.The advantages of this process are the production of high-purity metal

oxides, a highly controllable composition, the low temperature deposition,and a simple and economic experimental setup.27e29 However, disadvan-tages such as weak adhesion and low wear-resistance limit its full industrialexploitation.

5.2.2 One-dimensional nanostructuresA nomenclature for one-dimensional materials has not yet been established.Different and creative names have been presented in the literaturedsuch asnanotubules, -whiskers, -fibers, -fibrils, -cables, -castles, etc.daccording tothe morphology on the nanostructures. Terms such as “nanowires” or“nanorods” are probably the most common in the literature for structureswith two dimensions not exceeding a few hundreds of nanometers.

In the past 2 decades, the number of synthesis techniques forone-dimensional nanostructures has grown exponentially. These techniquescan be divided into top-down and bottom-up approaches. The first involvewhittling down the size of materials from the bulk size to nanometer scalevia standard microfabrication technologiesdas for example lithography,exfoliation, and lift-off processesdand allows the preparation ofwell-organized nanowires.30e33 However, the crystalline quality of fabri-cated nanomaterials is not excellent, and the manufacturing cost forlarge-scale production is usually very high. The second approach, on thecontrary, consists of self-assembly of atomic or molecular building blocks,by using synthesis techniques like vapor phase transport, solution-basedtechniques, or template growth.34 The advantages are the fine control inshape, morphology, structure, high purity, and crystallinity, together withthe low cost of the experimental equipment. The main drawback is thechallenging integration of the nanostructures on planar substrates, neededfor the exploitation of their useful properties. Several one-dimensional oxidenanostructures with different properties and morphologies have beenfabricated using bottom-up synthetic routes. Most of these structures couldnot have been prepared easily and economically using top-down tech-nologies. These bottom-up techniques can be generally classified in threedifferent types: (i) vapor phase growth methods, (ii) liquid phase growthmethods, and (iii) template-assisted methods. A few morphologies of thesenew nanostructures with potential as chemical sensing devices are sum-marized schematically in Fig. 5.1.


5.2.2.1 Vapor phase growth methodsAmong bottom-up techniques, vapor phase deposition is probably the mostwidely used, thanks to its simplicity and versatility, and is mainly based on acontrolled condensation of a vaporized metal oxide material. To obtainone-dimensional structures, there has to be a preferential growth direction,i.e., a faster growth rate in a particular direction. Even though the exactmechanism responsible for one-dimensional growth in the vapor phase isstill not clearly understood, vapor phase methods have been exploredand are extensively used by many research groups to synthesizeone-dimensional materials. The main advantage is its simplicity in termsof the procedure and the experimental setup used.

In general, the vapor phase is obtained by evaporation of metal oxidepowder (PVD), chemical reduction, or other precursor-based reactions(CVD). TE, laser ablation, or evaporation by ion, electron, and molecularbeams could be used to evaporate the materials or the precursors. The vaporsare then transported and condensed onto the substrate’s surface held atlower temperatures. By controlling the supersaturation of the vapor,one-dimensional materials can be easily obtained. When the growth ofthe nanowire crystal directly originates from the condensation from thevapor phase without the use of a catalyzer, the term ‘vaporesolid growth’is typically used. Defect-free nanowires can be produced using this tech-nique; however, there is still no consensus on the growth mechanism. Ifthe growth originates from the condensation onto catalyst particles, whichare liquid at such high temperatures, the growth is typically defined as a“vapor liquid solid” (VLS) process. The mechanism for VLS was proposedby Wagner in 1964. Under deposition conditions, the catalyzer has toform a liquid solution with the desired material. It should also have a low

Figure 5.1 Schematic representation of some morphologies of one-dimensionalnanostructures. From left to right: nanowire, longitudinal heterojunction, core-shellheterojunction, nanotube, nanofiber, nanorod, and hierarchical heterostructure.


vapor pressure and be chemically inert. In the process, the vapor diffuses intothe liquid catalyzer and, as the concentration becomes too high, the growthspecies precipitate to form the nanowire. The liquid phase is a preferentialcondensation site, and this causes a higher growth rate of the VLS withrespect to the VS. Furthermore, by controlling the dimension and dispersionof the catalyzer, control can be achieved over the diameter of the nanowire.Among all vapor phase methods, the VLS process is the most successfullyused and cited for generating nanowires of different oxides such asZnO,35 SnO2,

36 In2O3,37 NiO,38 TiO2,

39 and many more,40 with single-crystalline structures and in considerable amounts.

5.2.2.2 Liquid phase growth methodsWet chemistry is another widely diffused approach for fabricatingone-dimensional metal oxide nanostructures. There are many experimentaltechniques for the preparation of nanowires from the liquid phase. A consid-erable research effort has been expended in developing template-freemethods for the deposition of one-dimensional nanostructures in a liquidenvironment; the most important procedures are hydrothermalmethods,41e48 electrospinning,49e61 sonochemical,62e68 electrochemicalanodization, and electrodeposition and surfactant-assisted synthesis.69

The hydrothermal process has been a well-known procedure for materialsynthesis since the 1970s. It begins with an aqueous mixture of soluble metalsalt (metal and/or metaleorganic) precursors, then the solution is placed inan autoclave at a high temperature (between 100 and 300�C) and underrelatively high pressure (>1 atm) conditions. ZnO nanorods,69e75

CuO,76,77 ceria,78,79 titania,80,81 MnO2,82,83 and Co3O4,

84 have beenprepared by using wet chemical hydrothermal approaches.

Electrospinning exploits an electrical charge to force the formation ofmats of fine fibers.50,85 A solid fiber is produced as the electrified jet iscontinuously stretched due to the electrostatic repulsions between thesurface charges and the evaporation of solvent. A number of oxides havebeen fabricated as fibrous structures: Al2O3, CuO, NiO, TiO2, SiO2,V2O5, ZnO, Co3O4, Nb2O5, MoO3, and MgTiO3.

86e97 However, theone-dimensional nanostructures produced by electrospinning are, ingeneral, polycrystalline.

The electrochemical method is a relatively simple and effective wayto prepare one-dimensional semiconductor nanostructures by anodicoxidation (anodization) or electrodeposition. In case of anodization, themetal substrate is immersed in an electrolyte solution. The substrate is


then oxidized by a controlled electric field to form porous or tubular oxidestructures. The electrochemical anodization method is probably best knownfor the preparation of metal oxide nanotubes, but can be also used to preparevertically aligned nanowires.98e101

5.2.2.3 Template-assisted methodsThere are several references reporting on template-assisted approaches fornanofabrication such as Hulteen and Martin.102 They are regarded as oneof the pioneer groups for functional nanowire array fabrication. With theuse of a periodic structured template, one-dimensional nanostructures canbe prepared, thanks to the confinement effect of the porous template.The templates can be prepared easily with anodization. Control of the aspectratio and the area density of one-dimensional nanostructures can beachieved by changing the diameter and length of the template and bychanging the anodization voltage.103e105

The nanostructures can be deposited into the nanopores by electrodepo-sition or solegel deposition methods. The advantages of being low cost andrepeatable, together with their potential compatibility with silicon technol-ogies, make these nanostructure synthesis procedures interesting. Despite itssimplicity, template-based growth is characterized by the production ofpolycrystalline nanowires, which can limit their potential for both funda-mental studies and applications.

5.3 Conductometric sensor

The semiconducting properties of metal oxides are due to deviationfrom stoichiometry. In most oxides, such as tin oxide, oxygen vacanciesare responsible for the n-type behavior.106,107 The normal workingcondition for a chemical sensor in the presence of air is at relatively hightemperatures (500e800K). At these temperatures, the metal oxide conduc-tion is electronic and there are ionized oxygen vacancies. Oxygen in suchconditions is chemisorbed on the metal oxide surface, capturing chargecarriers from the conduction band and producing a space charge area nearthe surface. Chemical sensing is achieved in most cases by oxidationreactions between chemical species and chemisorbed oxygen, causing adecrease in the surface barrier, leading to a change in conductance. Otherchemical species, such as nitrogen oxide or water vapor, may chemisorbdirectly on the metal oxide surfaces by trapping or releasing electrons.


5.3.1 Device integrationDevice integration is very easy and well-established for thin films, which maybe easily patterned or deposited on the final transducers. In the case of one-dimensional nanostructures, instead, some open issues still remain.11,108

One-dimensional nanostructures should be grown directly on thetransducers, but, depending on the deposition conditions, this may notalways be possible due to high temperatures, pressures, or the aggressiveambient required for their preparation. In these cases, they have to be trans-ferred afterward. The easiest way to transfer is by drop coating,109,110 butother techniques such as dielectrophoresis,111,112 or roll transfer,113 maybe used, which are more compatible with industrial-scale manufacturing.Single-nanowire devices are still not exploited for mass production, dueto the very precise integration process required. For such devices, nanoma-nipulation114 of the single metal oxide nanowire can be used.

The problem that remains in all cases is the low mechanical and electricalstability of the contact achieved between the metal oxide and the metallicelectrodes or the substrate. To obtain stable devices, which can work forvery long time, there must be a good and reliable electrical contact, withthe lowest contact resistance possible. This is because the metal semi-conductor junction forming at the interface between the metal oxide andthe metal may play a role in chemical sensing. This is even more importantfor single-nanowire devices, because the junction is in series with thenanowire resistance; for multiple-nanowire devices, instead, it is connectedto a large number of resistances and thus less prominent.

New lithographic techniques have been proposed for the integration ofthe vapor phase growth process with device fabrication.115e119 Concerningchemical sensing, a high temperature lift-off procedure for the integration ofa nanowire network on sensing transducers was developed by using siliconoxide as a sacrificial layer.120 This allows a clean patterning and assures thepresence of uniform surfaces for the deposition of contacts.

For single-nanowire devices, highly expensive techniques (such as afocused ion beam, or a series of nanolithographic tools) could be used,ranging from proton and electron beam nanolithography,121e123 in whichpatterned substrates are obtained under the application of a charged particlebeam, to nanoimprint lithography.124,125

5.4 Transduction principles and related novel devices

When a sensing material is exposed to a specific atmosphere, it mayinteract with it in many different ways, which result in a change of some


of its physical properties, i.e., electrical, optical, magnetic, and evenstructural. Among these, electrical properties are for sure the most commonand the easiest to be detected. The interaction of the sensing material withsurrounding atmosphere can be transduced as a change of resistance, imped-ance, or work function. The easiest measurable parameter is the sensorresistance in DC conditions. It may be measured by a voltamperometrictechnique at constant bias but, in commercial chemical sensors, the sensingfilm is usually inserted inside a voltage divider.

A typical kinetic response of conductance as a function of an introduc-tion of a concentration step is shown in Fig. 5.2. After the reducing speciesis introduced at time t1, the sensor conductance Gi increases to Gf, in thetime needed to reach the new thermodynamic equilibrium of the surfacereactions. If the metal oxide is not stable, or if there is an irreversiblechemisorption, a steady state may not be reached.

Response time is the time necessary for the electrical conductance toreach a threshold value (usually 90%) of the difference between Gf andGi. Recovery time is the time necessary for the conductance to recover toa level expressed as a percentage fraction (usually 90%) of GfeGi. Concern-ing the response of chemical sensors, the linearity hypothesis is not verified,and the response when working with gas mixtures cannot be deduced by thesuperimposition principle, with a simple sum of the individual response.

Con

cent

ratio

n (p

pm)

Concentration (ppm)Conductance (S)

Con

duct

ance

(S)

Gf

Gi

t1 t2Time (S)

Figure 5.2 Conductance variation of the sensor produced by the introduction of a stepconcentration of a reducing gas.


The sensor response toward a reducing species and an n-type metal oxidemay be defined as the relative change in conductance:

Gf =Gi

For an oxidizing species and an n-type metal oxide, there is an increase inthe resistance and the sensor response may be defined as the relative changeof resistance:

Rf =Ri

The calibration curve can be obtained after measuring the responseat different concentrations in the same operational conditions. The cali-bration curve is generally reported in a bilogarithmic scale because therelation between concentration and conductance follows a power law(Fig. 5.3).

Impedance is another possible transduced signal, and it can be measuredby a spectroscopic analyzer or by LCR (L ¼ inductance, C ¼ capacitance,R ¼ resistance) bridges. It may be useful to identify the different contribu-tions to the sensor response (grain boundaries, bulk and contact) but,due to higher costs, there are no commercial devices based on thistransduction.

Most of the sensing performances reported in the literature are based onmeasurements of individual devices in artificial environments that do notreproduce field conditions. In some studies, the carrier gas is nitrogen insteadof synthetic air, and no humidity or interfering gases are introduced. That iswhy it is very difficult, and sometimes impossible, to make a fair comparison

Concentration (ppm)

Res

pons

e

Figure 5.3 Calibration curve of the response of a chemical sensor toward a chemicalspecies.


of all the results reported in literature, or to speculate on sensing perfor-mances in a real environment.

Few comparative studies between nanowire and polycrystalline chemicalsensors have been reported.126e129 Sysoev reported that even if the nanopar-ticles had a higher response to 2-propanol vapors at first, after some days ofoperation the response of the nanoparticles decreases to the stable responseof nanowires.128 This was ascribed to the irreversible sintering processin the nanoparticles that occurs due to high temperature operation.Kumar compared different morphologies of ZnO nanostructures, and hehighlighted that one-dimensional ZnO nanomaterials provide a prospectivebase due to their crystallinity for their applications as durable conducto-metric gas sensors compared with nanoparticles and thin films.129

The research on one-dimensional nanostructures is not as advanced asthat on two-dimensional nanostructures, due to the difficulties in fabricatingthe device. Nevertheless, to exploit the unique possibilities of thesestructures, the focus has to be on peculiar properties that can lead to essentialadvances in functional devices. For example, the self-heating property canbe used for the development of fully autonomous chemical sensors.130,131

Self-heating of a single nanowire is due to the dissipated power (Jouleeffect) induced by the bias current applied in conductometric measurements.Nanowires, with their small mass, can be heated up to several hundreds ofdegrees with a few tens of microwatts. Moreover, the thermal responsetime of these devices is extremely fast (in milliseconds range). This makesit possible to even observe the kinetics of the interactions between the gasmolecules and the metal oxide. By combining low power electronics withcontinuous or pulsed self-heating of nanowires, it will be possible to reducepower consumption to the microwatt range, or even lower.130,131

Another most interesting approach proposed to improve chemical inter-actions and reduce the operating temperature is optical excitation. Hightemperatures limit the application of chemical sensors to nonexplosive andinflammable environments: the use of standard metal oxideebased devicesat 200�C or more is not recommended in presence of free hydrogen, forexample. As metal oxide semiconductors absorb photons with an energyabove their bandgap, free carriers are produced in the space charge area.The excess electrons are swept away from the surface, while excess holesare swept toward it due to the electrical field in the space charge area,with a decrease in the surface band bending. Several years ago, the effectof photoactivation on the sensing performances was demonstrated for


two-dimensional nanostructure metal oxide chemical sensors.132e135

The first report on the possibility of using optical excitation on one-dimensional nanostructure sensing devices was published by Law et al.136

After several years, the response of optically excited single-nanowire deviceswas shown to be comparable with devices that were thermally activated, inthe optimal experimental conditions.137,138 Many metal oxide materialsused for gas sensing applications have a wide bandgap (3.6e3.9 eV forSnO2), therefore is necessary to use UV light to excite these sensingmaterials. Thanks to the advances in LED fabrication, now UV LEDs(325 nm for example) are quite cheap and could be easily integrated intosensing systems.139

Conductometric devices are by far the most explored ones, but othertransducing mechanisms have been investigated also. Field effect transistors(FETs) incorporating metal oxide nanowires have been fabricated,combining the advantages of conductometric devices with the possibilityto further tune the sensing properties by channel modulation.These NW-FET devices have been largely used as biosensors, thanks tothe possibility to functionalize the surface with specific receptors.140,141

A novel electrical transduction mechanism was recently exploited, firstlyon 2D thin films142,143 and then on quasieone-dimensional metal oxidenanowires.144 Instead of measuring the relative change in the conductanceof the material, surface ionizationebased devices measure the ionic currentbetween the surface of the metal oxide material and a counter electrode, inthe presence of ionized gas molecules. It was demonstrated that these devicesmight easily discriminate, for example, amines and hydrocarbons with aminefunctional groups, which enable sensors for illicit drug monitoring to bemade.

Optical properties are also influenced by the interaction of metal oxidesurface with the surrounding atmosphere. For example, a reversible modifi-cation of static photoluminescence efficiency of ZnO nanowires wasobserved on exposure to low concentrations of nitrogen dioxide.145 Asimilar behavior was detected on TiO2/SnO2 nanoparticles on ammoniaexposure.146


Significant efforts have been made to develop and test new metaloxides, especially in the form of nanowires, nanoparticles, and nanotubes.However, their application as chemical sensors still faces problems such as


device stability over time, selectivity and long-term drift due to stoichiom-etry changes, and coalescence of crystallites, especially for nanoparticles. Thenotion of preparing multipurpose devices has been replaced by the develop-ment of sensors tailored for specific and focused applications.

The improvement of computer systems and of imaging and spectro-scopic techniques will provide powerful tools for the better understandingof chemical sensing mechanisms and help to optimize sensor design.One-dimensional nanostructures have a greater surface-to-volume ratio,better stoichiometry, and a higher degree of crystallinity compared withtwo-dimensional nanostructures. They also have reduced instability associ-ated with grain coalescence. These factors make one-dimensional metaloxides very promising for the better understanding and the developmentof a new generation of chemical sensors.

References1. Brattain WH, Bardeen J. Surface properties of germanium. Bell Syst Tech J 1953;32:

1e41. https://doi.org/10.1002/j.1538-7305.1953.tb01420.x.2. Yuliarto B, Gumilar G, Septiani NLW. SnO2 nanostructure as pollutant gas sensors:

synthesis, sensing performances, and mechanism. Adv Mater Sci Eng 2015. https://doi.org/10.1155/2015/694823. Article number 694823.

3. Galstyan V, Comini E, Faglia G, Sberveglieri G. TiO2 nanotubes: recent advances insynthesis and gas sensing properties. Sensors 2013;13(11):14813e38. https://doi.org/10.3390/s131114813.

4. Ilin A, Martyshov M, Forsh E, Forsh P, Rumyantseva M, Abakumov A, Gaskov A,Kashkarov P. UV effect on NO2 sensing properties of nanocrystalline In2O3. SensorActuator B Chem 2016;231:491e6. https://doi.org/10.1016/j.snb.2016.03.051.

5. Wang C, Sun R, Li X, Sun Y, Sun P, Liu F, Lu G. Hierarchical flower-like WO3nanostructures and their gas sensing properties. Sensor Actuator B Chem 2014;204:224e30. https://doi.org/10.1016/j.snb.2014.07.083.

6. Zhu L, Zeng W. Room-temperature gas sensing of ZnO-based gas sensor: a review.Sens Actuat A Phy 2017;267:242e61. https://doi.org/10.1016/j.sna.2017.10.021.

7. Tao Y, Gao Q, Di J, Wu X. Gas sensors based on a-Fe2O3 nanorods, nanotubes andnanocubes. J Nanosci Nanotechnol 2013;13(8):5654e60. https://doi.org/10.1166/jnn.2013.7559.

8. Kim H-J, Lee J-H. Highly sensitive and selective gas sensors using p-type oxide semi-conductors: overview. Sensor Actuator B Chem 2014;192:607e27. https://doi.org/10.1016/j.snb.2013.11.005.

9. Tonezzer M, Dang LTT, Tran HQ, Iannotta S. Multiselective visual gas sensor usingnickel oxide nanowires as chemiresistor. Sensor Actuator B Chem 2018;255:2785e93.https://doi.org/10.1016/j.snb.2017.09.094.

10. Barsan N, Simion C, Heine T, Pokhrel S, Weimar U. Modeling of sensing and trans-duction for p-type semiconducting metal oxide based gas sensors. J Electroceram 2010;25(1):11e9. https://doi.org/10.1007/s10832-009-9583-x.

11. Nikoobakht B, Wang X, Herzing A, Shi J. Scalable synthesis and device integration ofself-registered one-dimensional zinc oxide nanostructures and related materials. ChemSoc Rev 2013;42(1):342e65. https://doi.org/10.1039/c2cs35164a.


https://doi.org/10.1002/j.1538-7305.1953.tb01420.x

https://doi.org/10.1155/2015/694823

https://doi.org/10.1155/2015/694823

https://doi.org/10.3390/s131114813

https://doi.org/10.3390/s131114813



https://doi.org/10.1016/j.sna.2017.10.021

https://doi.org/10.1166/jnn.2013.7559

https://doi.org/10.1166/jnn.2013.7559




https://doi.org/10.1007/s10832-009-9583-x

https://doi.org/10.1039/c2cs35164a

12. Yamazoe N. New approaches for improving semiconductor gas sensors. Sensor ActuatorB 1991;5:7. https://doi.org/10.1016/0925-4005(91)80213-4.

13. Kanitkar P, Kaur M, Sen S, Joshi A, Kumar V, Gupta SK, Yakhmi JV. Growth andgas-sensing studies of metal oxide semiconductor nanostructures. Int J Nanotechnol2010;7(9e12):883e906. https://doi.org/10.1504/IJNT.2010.034696.

14. Yaacob MH, Yu J, Latham K, Kalantar-Zadeh K, Wlodarski W. Optical hydrogensensing properties of nanostructured Pd/MoO(3) films. Sens Lett 2011;9(1):16e20.https://doi.org/10.1166/sl.2011.1410.

15. Chen SE, Lu HH, Brahma S, Huang JL. Effects of annealing on thermochromic prop-erties of W-doped vanadium dioxide thin films deposited by electron beamevaporation. Thin Solid Films 2017;644:52e6. https://doi.org/10.1016/j.tsf.2017.05.052.

16. Wongchoosuk C, Wisitsoraat A, Tuantranont A, Kerdcharoen T. Portable electronicnose based on carbon nanotube-SnO(2) gas sensors and its application for detection ofmethanol contamination in whiskeys. Sensor Actuator B Chem 2010;147(2):392e9.https://doi.org/10.1016/j.snb.2010.03.072.

17. Lu C, Chen Z. High-temperature resistive hydrogen sensor based on thin nanoporousrutile TiO(2) film on anodic aluminum oxide. Sensor Actuator B Chem 2009;140(1):109e15. https://doi.org/10.1016/j.snb.2009.04.004.

18. Durrani SMA, AI-Kuhaili MF. Effect of biasing voltages and electrode metals andmaterials on the sensitivity of electron beam evaporated HfO2 thin film CO sensor.Mater Chem Phys 2008;109(1):56e60. https://doi.org/10.1016/j.matchemphys.2007.10.034.

19. Trucchi DM, Zanza A, Bellucci A, Marotta V, Orlando S. Photoconductive andphotovoltaic evaluation of In(2)O(3)-SnO(2) multilayered thin-films deposited on sil-icon by reactive pulsed laser ablation. Thin Solid Films 2010;518(16):4738e42. https://doi.org/10.1016/j.tsf.2009.12.072.

20. Belysheva TV, Gerasimov GN, Gromov VF, Trakhtenberg LI. The sensor properties ofFe2O3 center dot In2O3 films: the detection of low ozone concentrations in air. Russ JPhys Chem 2008;82(10):1721e5. https://doi.org/10.1134/S0036024408100142.

21. Mitu B, Marotta V, Orlando S. Multilayered metal oxide thin film gas sensors obtainedby conventional and RF plasma-assisted laser ablation. Appl Surf Sci 2006;252(13):4637e41. https://doi.org/10.1016/j.apsusc.2005.07.102.

22. Al-Hardan N, Abdullah MJ, Aziz AA. Impedance spectroscopy of undoped and Cr-doped ZnO gas sensors under different oxygen concentrations. Appl Surf Sci 2011;257(21):8993e7. https://doi.org/10.1016/j.apsusc.2011.05.078.

23. Shaalan NM, Yamazaki T, Kikuta T. Effect of micro-electrode geometry on NO(2)gas-sensing characteristics of one-dimensional tin dioxide nanostructure microsensors.Sensor Actuat B-Chem 2011;156(2):784e90. https://doi.org/10.1016/j.snb.2011.02.039.

24. Batista C, Teixeira V, Ribeiro RM. Pulsed DC reactive magnetron sputtering of va-nadium dioxide thermochromic thin films. Mater Technol 2011;26(1):35e9. https://doi.org/10.1179/175355511X12941605982307.

25. Haireche S, Boumeddiene A, Guittoum A, El Hdiy A, Boufelfel A. Structural,morphological and electronic study of CVD SnO2:Sb films. Mater Chem Phys 2013;139(2e3):871e6. https://doi.org/10.1016/j.matchemphys.2013.02.046.

26. Beardslee JA, Mebust AK, Chaimowitz AS, Davis-VanAtta CR, Leonard H,Moersch TL, Afridi MY, Taylor CJ. Using precursor chemistry to template vanadiumoxide for chemical sensing. Chem Vap Depos 2010;16:206e10. https://doi.org/10.1002/cvde.201004286.

27. Bekermann D, Rogalla D, Becker HW, Winter M, Fischer RA, Devi A. Volatile,monomeric, and fluorine-free precursors for the metal organic chemical vapor depo-sition of zinc oxide. Eur J Inorg Chem 2010;9:1366e72. https://doi.org/10.1002/ejic.200901037.


https://doi.org/10.1016/0925-4005(91)80213-4

https://doi.org/10.1504/IJNT.2010.034696

https://doi.org/10.1166/sl.2011.1410

https://doi.org/10.1016/j.tsf.2017.05.052




https://doi.org/10.1016/j.matchemphys.2007.10.034




https://doi.org/10.1134/S0036024408100142

https://doi.org/10.1016/j.apsusc.2005.07.102




https://doi.org/10.1179/175355511X12941605982307

https://doi.org/10.1179/175355511X12941605982307


https://doi.org/10.1002/cvde.201004286

https://doi.org/10.1002/cvde.201004286

https://doi.org/10.1002/ejic.200901037

https://doi.org/10.1002/ejic.200901037

28. Neri G. Non-conventional sol-gel routes to nanosized metal oxides for gas sensing:from materials to applications. Sci Adv Mater 2010;2(1):3e15. https://doi.org/10.1166/sam.2010.1062.

29. Breedon M, Spizzirri P, Taylor M, du Plessis J, McCulloch D, Zhu JM, Yu LS, Hu Z,Rix C, Wlodarski W, Kalantar-zadeh K. Synthesis of nanostructured tungsten oxidethin films: a simple, controllable, inexpensive, aqueous sol-gel method. Cryst GrowthDes 2010;10(1):430e9. https://doi.org/10.1021/cg9010295.

30. Sun Z, Liao T, Kou L. Strategies for designing metal oxide nanostructures. Sci ChinaMater 2017;60:1e24. https://doi.org/10.1007/s40843-016-5117-0.

31. Ramadan S, Kwa K, King P, O’Neill A. Reliable fabrication of sub-10 nm siliconnanowires by optical lithography. Nanotechnol 2016;27(42):425302. https://doi.org/10.1088/0957-4484/27/42/425302.

32. Marrian CRK, Tennant DM. Nanofabrication. J Vac Sci Technol A 2003;21:S207e15.https://doi.org/10.1116/1.1600446.

33. Candeloro P, Comini E, Baratto C, Faglia G, Sberveglieri G, Kumar R,Carpentiero A, Di Fabrizio E. SnO2 lithographic processing for nanopatterned gassensors. J Vac Sci Technol B 2005;23:2784e8. https://doi.org/10.1116/1.2110371.

34. Zhou Y, Li J-T, Sun S-G. Synthesis-cum-assembly toward hierarchicalnanoarchitectures. Coord Chem Rev 2017;352:291e305. https://doi.org/10.1016/j.ccr.2017.09.018.

35. Serrano A, Arana A, Gald�amez A, Dutt A, Monroy BM, G€uell F, Santana G. Effect ofthe seed layer on the growth and orientation of the ZnO nanowires: consequence onstructural and optical properties. Vacuum 2017;146:509e16. https://doi.org/10.1016/j.vacuum.2017.03.010.

36. Terasako T, Kohno K, Yagi M. Vapor-liquid-solid growth of SnO2 nanowires utiliz-ing alternate source supply and their photoluminescence properties. Thin Solid Films2017;644:3e9. https://doi.org/10.1016/j.tsf.2017.05.053.

37. Tuzluca FN, Yesilbag YO, Ertugrul M. Synthesis of In2O3 nanostructures withdifferent morphologies as potential supercapacitor electrode materials. Appl Surf Sci2018;427:956e64. https://doi.org/10.1016/j.apsusc.2017.08.127.

38. Kaur N, Comini E, Zappa D, Poli N, Sberveglieri G. Nickel oxide nanowires: vaporliquid solid synthesis and integration into a gas sensing device. Nanotechnol 2016;27(20):205701. https://doi.org/10.1088/0957-4484/27/20/205701.

39. Lee JC, Park KS, Kim TG, Choi HJ, Sung YM. Controlled growth of high-qualityTiO2 nanowires on sapphire and silica. Nanotechnol 2006;17(17):4317e21. https://doi.org/10.1088/0957-4484/17/17/006.

40. Klamchuen A, Suzuki M, Nagashima K, Yoshida H, Kanai M, Zhuge F, He Y,Meng G, Kai S, Takeda S, Kawai T, Yanagida T. Rational concept for designingvapor-liquid-solid growth of single crystalline metal oxide nanowires. Nano Lett2015;15(10):6406e12. https://doi.org/10.1021/acs.nanolett.5b01604.

41. Candeloro P, Carpentiero A, Cabrini S, Di Fabrizio E, Comini E, Baratto C, Faglia G,Sberveglieri G, Gerardino A. SnO2 sub-micron wires for gas sensors. Micro Eng 2005;78e79:178e84. https://doi.org/10.1016/j.mee.2004.12.024.

42. Liu B, Zeng HC. Hydrothermal synthesis of ZnO nanorods in the diameter regime of50 nm. J Am Chem Soc 2003;125:4430e1. https://doi.org/10.1021/ja0299452.

43. Wang JSM, Gao L. Wet chemical synthesis of ultralong and straight single-crystallineZnO nanowires and their excellent UV emission properties. J Mater Chem 2003;13:2551e4. https://doi.org/10.1039/B307565F.

44. Guo M, Diao P, Cai SM. Hydrothermal growth of well-aligned ZnO nanorod arrays:dependence of morphology and alignment ordering upon preparing conditions. J SolidState Chem 2005;178:1864e73. https://doi.org/10.1016/j.jssc.2005.03.031.


https://doi.org/10.1166/sam.2010.1062

https://doi.org/10.1166/sam.2010.1062

https://doi.org/10.1021/cg9010295

https://doi.org/10.1007/s40843-016-5117-0

https://doi.org/10.1088/0957-4484/27/42/425302

https://doi.org/10.1088/0957-4484/27/42/425302

https://doi.org/10.1116/1.1600446

https://doi.org/10.1116/1.2110371

https://doi.org/10.1016/j.ccr.2017.09.018

https://doi.org/10.1016/j.ccr.2017.09.018

https://doi.org/10.1016/j.vacuum.2017.03.010

https://doi.org/10.1016/j.vacuum.2017.03.010



https://doi.org/10.1088/0957-4484/27/20/205701

https://doi.org/10.1088/0957-4484/17/17/006

https://doi.org/10.1088/0957-4484/17/17/006

https://doi.org/10.1021/acs.nanolett.5b01604

https://doi.org/10.1016/j.mee.2004.12.024

https://doi.org/10.1021/ja0299452

https://doi.org/10.1039/B307565F

https://doi.org/10.1016/j.jssc.2005.03.031

45. Cao MH, Wang YH, Guo CX, Qi YJ, Hu CW, Wang EB. A simple route towardsCuO nanowires and nanorods. J Nanosci Nanotechnol 2004;4:824e8. https://doi.org/10.1166/jnn.2004.822.

46. Zhou KB, Wang X, Sun XM, Peng Q, Li YD. Enhanced catalytic activity of ceriananorods from well-defined reactive crystal planes. J Catal 2005;229:206e12.https://doi.org/10.1016/j.jcat.2004.11.004.

47. Dharmadhikari DV, Athawale AA. Studies on structural and optical properties of rareearth copper oxides synthesized by template free hydrothermal method. Mater Sci EngB: Solid-State Mater Adv Technol 2018;229:70e8. https://doi.org/10.1016/j.mseb.2017.12.012.

48. Einarsrud MA, Grande T. 1D oxide nanostructures from chemical solutions. Chem SocRev 2014;43:2187e99. https://doi.org/10.1039/C3CS60219B.

49. Yuan ZY, Su BL. Titanium oxide nanotubes, nanofibers and nanowires. Colloids SurfA Physicochem Eng Asp 2004;241:173e83. https://doi.org/10.1016/j.colsurfa.2004.04.030.

50. Formhals A. Process and apparatus for preparing artificial threads. 1934. US patent, 1 975504.

51. Dai H, Gong J, Kim H, Lee D. A novel method for preparing ultra-fine alumina-borate oxide fibres via an electrospinning technique. Nanotechnol 2002;13:674e7.https://doi.org/10.1088/0957-4484/13/5/327.

52. Viswanathamurthi P, Bhattarai N, Kim HY, Lee DR, Kim SR, Morris MA. Prepara-tion and morphology of niobium oxide fibres by electrospinning. Chem Phys Lett2003;374:79e84. https://doi.org/10.1016/S0009-2614(03)00702-4.

53. Shao C, KimHY, Gong J, Ding B, Lee DR, Park SJ. Fiber mats of poly(vinyl alcohol)/silica composite via electrospinning. Mater Lett 2003;57:1579e84. https://doi.org/10.1016/S0167-577X(02)01036-4.

54. Guan H, Shao C, Chen B, Gong J, Yang X. A novel method for making CuO super-fine fibres via an electrospinning technique. Inorg Chem Commun 2003;6:1409e11.https://doi.org/10.1016/j.inoche.2003.08.021.

55. Guan H, Shao C, Chen B, Gong J, Yang X. Preparation and characterization of NiOnanofibres via an electrospinning technique. Inorg Chem Commun 2003;6:1302e3.https://doi.org/10.1016/j.inoche.2003.08.003.

56. Yang X, Shao C, Guan H, Li X, Gong J. Preparation and characterization of ZnOnanofibers by using electrlospun PVA/zinc acetate composite fiber as precursor. InorgChem Commun 2004;7:176e8. https://doi.org/10.1016/j.inoche.2003.10.035.

57. Ding B, Kim H, Kim C, Khil M, Park S. Morphology and crystalline phase study ofelectrospun TiO2-SiO2 nanofibres. Nanotechnol 2003;14:532e7. https://doi.org/10.1088/0957-4484/14/5/309.

58. Viswanathamurthi P, Bhattarai N, Kim HY, Lee DR. Vanadium pentoxide nanofibersby electrospinning. Scr Mater 2003;49:577e81. https://doi.org/10.1016/S1359-6462(03)00333-6.

59. Tavakoli Foroushani F, Tavanai H, Ranjbar M, Bahrami H. Fabrication of tungstenoxide nanofibers via electrospinning for gasochromic hydrogen detection. Sens ActuatB 2018;268:319e27. https://doi.org/10.1016/j.snb.2018.04.120.

60. Von Reitzenstein NH, Bi X, Yang Y, Hristovski K,Westerhoff P. Morphology, struc-ture, and properties of metal oxide/polymer nanocomposite electrospun mats. J ApplPolym Sci 2016;133(33):43811. https://doi.org/10.1002/app.43811.

61. Niu C, Meng J, Wang X, Han C, Yan M, Zhao K, Xu X, Ren W, Zhao Y, Xu L,Zhang Q, Zhao D, Mai L. General synthesis of complex nanotubes by gradient elec-trospinning and controlled pyrolysis. Nat Commun 2015;6:7402. https://doi.org/10.1016/j.snb.2017.05.161.


https://doi.org/10.1166/jnn.2004.822

https://doi.org/10.1166/jnn.2004.822

https://doi.org/10.1016/j.jcat.2004.11.004

https://doi.org/10.1016/j.mseb.2017.12.012

https://doi.org/10.1016/j.mseb.2017.12.012

https://doi.org/10.1039/C3CS60219B

https://doi.org/10.1016/j.colsurfa.2004.04.030

https://doi.org/10.1016/j.colsurfa.2004.04.030

https://doi.org/10.1088/0957-4484/13/5/327

https://doi.org/10.1016/S0009-2614(03)00702-4

https://doi.org/10.1016/S0167-577X(02)01036-4

https://doi.org/10.1016/S0167-577X(02)01036-4

https://doi.org/10.1016/j.inoche.2003.08.021



https://doi.org/10.1088/0957-4484/14/5/309

https://doi.org/10.1088/0957-4484/14/5/309

https://doi.org/10.1016/S1359-6462(03)00333-6

https://doi.org/10.1016/S1359-6462(03)00333-6


https://doi.org/10.1002/app.43811



62. Dharmaraj N, Park HC, Lee BM, Viswanathamurthi P, Kim HY, Lee DR. Prepara-tion and morphology of magnesium titanate nanofibres via electrospinning. InorgChem Commun 2004;7. https://doi.org/10.1016/j.inoche.2003.12.033. 431e133.

63. Xu CK, Xu GD, Liu YK, Wang GH. A simple and novel route for the preparation ofZnO nanorods. Solid State Commun 2002;122:175e9. https://doi.org/10.1016/S0038-1098(02)00114-X.

64. Xu CK, Zhao XL, Liu S, Wang GH. Large-scale synthesis of rutile SnO2 nanorods.Solid State Commun 2003;125:301e4. https://doi.org/10.1016/S0038-1098(02)00826-8.

65. Xu CK, Xu GD, Wang GH. Preparation and characterization of NiO nanorods bythermal decomposition of NiC2O4 precursor. J Mater Sci 2003;38:779e82. https://doi.org/10.1023/A:1021856930632.

66. Gao T, Li QH, Wang TH. Sonochemical synthesis, optical properties, and electricalproperties of core/shell-type ZnO nanorod/CdS nanoparticle composites.ChemMater2005;17:887e92. https://doi.org/10.1021/cm0485456.

67. Miao JJ, Wang H, Li YR, Zhu JM, Zhu JJ. Ultrasonic-induced synthesis of CeO2nanotubes. J Cryst Growth 2005;281:525e9. https://doi.org/10.1016/j.jcrysgro.2005.04.058.

68. Kim DS, Kim JC, Kim BK, Kim DW. One-pot low-temperature sonochemical syn-thesis of CuO nanostructures and their electrochemical properties. Ceram Int 2016;42(16):19454e60. https://doi.org/10.1016/j.ceramint.2016.09.044.

69. Wang Y, Liu P, Zhu K, Wang LJ. Hierarchical bilayered hybrid nanostructural arraysof NiCo2O4 micro-urchins and nanowires as a free-standing electrode with highloading for high-performance lithium-ion batteries. Nanoscale 2017;9(39):14979e89. https://doi.org/10.1039/c7nr03979d.

70. Kumar RV, Koltypin Y, Xu XN, Yeshurun Y, Gedanken A, Felner I. Fabrication ofmagnetite nanorods by ultrasound irradiation. J Appl Phys 2001;89:6324e8. https://doi.org/10.1063/1.1369408.

71. Sundararajan M, Sakthivel P, Fernandez AC. Structural, optical and electrical proper-ties of ZnO-ZnS nanocomposites prepared by simple hydrothermal method. J AlloyComp 2018;768:553e62. https://doi.org/10.1016/j.jallcom.2018.07.245.

72. Han SY, Akhtar MS, Jung I, Yang OB. ZnO nanoflakes nanomaterials via hydrother-mal process for dye sensitized solar cells. Mater Lett 2018;230:92e5. https://doi.org/10.1016/j.matlet.2018.07.083.

73. Choi SC, Sohn SH. Controllable hydrothermal synthesis of bundled ZnO nanowiresusing cerium acetate hydrate precursors. Phys E Low-dimens Syst Nanostruct 2018;104:98e100. https://doi.org/10.1016/j.physe.2018.07.014.

74. Chen X, Shen Y, Zhang W, Zhang J, Wei D, Lu R, Zhu L, Li H, Shen Y. In-situgrowth of ZnO nanowire arrays on the sensing electrode via a facile hydrothermalroute for high-performance NO2 sensor. Appl Surf Sci 2018;435:1096e104. https://doi.org/10.1016/j.apsusc.2017.11.222.

75. Sun Y, Ndifor-Angwafor NG, Riley DJ, Ashfold MNR. Synthesis and photolumines-cence of ultra-thin ZnO nanowire/nanotube arrays formed by hydrothermal growth.Chem Phys Lett 2006;431:352e7. https://doi.org/10.1016/j.cplett.2006.09.100.

76. Wang Y, Wang D, Yan B, Chen Y, Song C. Fabrication of diverse CuO nanostruc-tures via hydrothermal method and their photocatalytic properties. J Mater Sci MaterElectron 2016;27(7):6918e24. https://doi.org/10.1007/s10854-016-4645-8.

77. Filipi�c G, Cvelbar U. Copper oxide nanowires: a review of growth. Nanotechnol 2012;23(19):194001. https://doi.org/10.1088/0957-4484/23/19/194001.

78. Liao Y, He L, Zhao M, Ye D. Ultrasonic-assisted hydrothermal synthesis of ceriananorods and their catalytic properties for toluene oxidation. J Environ Chem Eng2017;5(5):5054e60. https://doi.org/10.1016/j.jece.2017.09.037.



https://doi.org/10.1016/S0038-1098(02)00114-X

https://doi.org/10.1016/S0038-1098(02)00114-X

https://doi.org/10.1016/S0038-1098(02)00826-8

https://doi.org/10.1016/S0038-1098(02)00826-8

https://doi.org/10.1023/A:1021856930632

https://doi.org/10.1023/A:1021856930632

https://doi.org/10.1021/cm0485456

https://doi.org/10.1016/j.jcrysgro.2005.04.058

https://doi.org/10.1016/j.jcrysgro.2005.04.058

https://doi.org/10.1016/j.ceramint.2016.09.044

https://doi.org/10.1039/c7nr03979d

https://doi.org/10.1063/1.1369408

https://doi.org/10.1063/1.1369408

https://doi.org/10.1016/j.jallcom.2018.07.245

https://doi.org/10.1016/j.matlet.2018.07.083


https://doi.org/10.1016/j.physe.2018.07.014



https://doi.org/10.1016/j.cplett.2006.09.100

https://doi.org/10.1007/s10854-016-4645-8

https://doi.org/10.1088/0957-4484/23/19/194001

https://doi.org/10.1016/j.jece.2017.09.037

79. Karl Chinnu M, Vijai Anand K, Mohan Kumar R, Alagesan T, Jayavel R. Formationand characterisation of CeO2 and Gd:CeO2 nanowires/rods for fuel cell applications.J Exp Nanosci 2015;10(7):520e31. https://doi.org/10.1080/17458080.2013.845916.

80. Banerjee AN, Anitha VC, Joo SW. Improved electrochemical properties ofmorphology-controlled titania/titanate nanostructures prepared by in-situ hydrothermalsurface modification of self-source Ti substrate for high-performance supercapacitors. SciRep 2017;7(1):13227. https://doi.org/10.1038/s41598-017-11346-2.

81. Tang Y, Ren H, Huang J. Synthesis of porous TiO2 nanowires and their photocata-lytic properties. Front Optoelectron 2017;10(4):395e401. https://doi.org/10.1007/s12200-017-0735-3.

82. Yuan ZY, Ren TZ, Du G, Su BL. A facile preparation of single-crystalline a-Mn2O3nanorods by ammonia-hydrothermal treatment of MnO2. Chem Phys Lett 2004;389:83e6. https://doi.org/10.1016/j.cplett.2004.03.064.

83. Wang B, Qiu J, Feng H, Wang N, Sakai E, Komiyama T. Preparation of MnO2/car-bon nanowires composites for supercapacitors. Electrochim Acta 2016;212:710e21.https://doi.org/10.1016/j.electacta.2016.07.066.

84. Rui X, Tan H, Sim D, Liu W, Xu C, Hng HH, Yazami R, Lim TM, Yan Q.Template-free synthesis of urchin-like Co3O4 hollow spheres with good lithium stor-age properties. J Power Sources 2013;222:97e102. https://doi.org/10.1016/j.jpowsour.2012.08.094.

85. Ngadiman NHA, Yusof NM, Idris A, Misran E, Kurniawan D. Development of high-ly porous biodegradable g-Fe2O3/polyvinyl alcohol nanofiber mats using electrospin-ning process for biomedical application. Mater Sci Eng C 2017;70:520e34. https://doi.org/10.1016/j.msec.2016.09.002.

86. Vidyadharan B, Aziz RA, Misnon II, Anil Kumar GM, Ismail J, Yusoff MM, Jose R.High energy and power density asymmetric supercapacitors using electrospun cobaltoxide nanowire anode. J Power Sources 2014;270:526e35. https://doi.org/10.1016/j.jpowsour.2014.07.134.

87. Archana PS, Gupta A, Yusoff MM, Jose R. Tungsten doped titanium dioxide nano-wires for high efficiency dye-sensitized solar cells. Phys Chem Chem Phys 2014;16(16):7448e54. https://doi.org/10.1039/c4cp00034j.

88. Lai D, Chen Y, Yang D, Guo W, Yu Y. Alumina (Al2O3) nanofibers fromelectrospinning. Adv Mater Res 2012;476e478:379e82. https://doi.org/10.4028/www.scientific.net/AMR.476-478.379.

89. Leindecker GC, Alves AK, Bergmann CP. Synthesis of niobium oxide fibers by elec-trospinning and characterization of their morphology and optical properties. Ceram Int2014;40(PB):16195e200. https://doi.org/10.1016/j.ceramint.2014.07.054.

90. Shahhosseininia M, Bazgir S, Joupari MD. Fabrication and investigation of silica nano-fibers via electrospinning.Mater Sci Eng C 2018;91:502e11. https://doi.org/10.1016/j.msec.2018.05.068.

91. Harilal M, Krishnan SG, Pal B, Reddy MV, Ab Rahim MH, Yusoff MM, Jose R.Environment-Modulated crystallization of Cu2O and CuO nanowires by electrospin-ning and their charge storage properties. Langmuir 2018;34(5):1873e82. https://doi.org/10.1021/acs.langmuir.7b03576.

92. Hosseini SR, Ghasemi S, Kamali-Rousta M, Nabavi SR. Preparation of NiO nano-fibers by electrospinning and their application for electro-catalytic oxidation ofethylene glycol. Int J Hydrog Energy 2017;42(2):906e13. https://doi.org/10.1016/j.ijhydene.2016.09.116.

93. Sabzehmeidani MM, Karimi H, Ghaedi M. Electrospinning preparation of NiO/ZnOcomposite nanofibers for photodegradation of binary mixture of rhodamine B andmethylene blue in aqueous solution: central composite optimization. Appl OrganometChem 2018;32(6):e4335. https://doi.org/10.1002/aoc.4335.


https://doi.org/10.1080/17458080.2013.845916

https://doi.org/10.1038/s41598-017-11346-2

https://doi.org/10.1007/s12200-017-0735-3

https://doi.org/10.1007/s12200-017-0735-3

https://doi.org/10.1016/j.cplett.2004.03.064

https://doi.org/10.1016/j.electacta.2016.07.066

https://doi.org/10.1016/j.jpowsour.2012.08.094


https://doi.org/10.1016/j.msec.2016.09.002




https://doi.org/10.1039/c4cp00034j

https://doi.org/10.4028/www.scientific.net/AMR.476-478.379

https://doi.org/10.4028/www.scientific.net/AMR.476-478.379

https://doi.org/10.1016/j.ceramint.2014.07.054



https://doi.org/10.1021/acs.langmuir.7b03576

https://doi.org/10.1021/acs.langmuir.7b03576

https://doi.org/10.1016/j.ijhydene.2016.09.116

https://doi.org/10.1016/j.ijhydene.2016.09.116

https://doi.org/10.1002/aoc.4335

94. Boyadjiev SI, Kéri O, B�ardos P, Firkala T, G�aber F, Nagy ZK, Baji Z, Tak�acs M,Szil�agyi IM. TiO2/ZnO and ZnO/TiO2 core/shell nanofibers prepared by electro-spinning and atomic layer deposition for photocatalysis and gas sensing. Appl Surf Sci2017;424:190e7. https://doi.org/10.1016/j.apsusc.2017.03.030.

95. Wang Y, Zhang J, Liu L, Zhu C, Liu X, Su Q. Visible light photocatalysis of V2O5/TiO2 nanoheterostructures prepared via electrospinning. Mater Lett 2012;75:95e8.https://doi.org/10.1016/j.matlet.2012.01.074.

96. Dorneanu PP, Airinei A, Olaru N, Homocianu M, Nica V, Doroftei F. Preparationand characterization of NiO, ZnO and NiOeZnO composite nanofibers by electro-spinning method. Mater Chem Phys 2014;148(3):1029e35. https://doi.org/10.1016/j.matchemphys.2014.09.014.

97. Kanjwal MA, Barakat NAM, Sheikh FA, Khil MS, Kim HY. Physiochemical charac-terizations of nanobelts consisting of three mixed oxides (Co3O4, CuO, and MnO2)prepared by electrospinning technique. J Math Sci 2008;43(16):5489e94. https://doi.org/10.1007/s10853-008-2835-3.

98. Varghese OK, Paulose M, Grimes CA. Long vertically aligned titania nanotubes ontransparent conducting oxide for highly efficient solar cells. Nat Nanotechnol 2009;4:592e7. https://doi.org/10.1038/nnano.2009.226.

99. Singh DP, Srivastava ON. Applied potential dependent growth of SnO2 nanostruc-tures by anodic oxidation of tin. Adv Sci Lett 2012;16:255e60. https://doi.org/10.1166/asl.2012.2188.

100. Park J, Kim K, Choi J. Formation of ZnO nanowires during short durations of poten-tiostatic and galvanostatic anodization. Curr Appl Phys 2013;13(7):1370e5. https://doi.org/10.1016/j.cap.2013.04.015.

101. Faid AY, Allam NK. Stable solar-driven water splitting by anodic ZnO nanotubularsemiconducting photoanodes. RSC Adv 2016;6(83):80221e5. https://doi.org/10.1039/c6ra18747a.

102. Hulteen JC, Martin CR. A general template-based method for the preparation ofnanomaterials. J Mater Chem 1997;7:1075e87. https://doi.org/10.1039/A700027H.

103. Chen YH, Shen YM, Wang SC, Huang JL. Fabrication of one-dimensional ZnOnanotube and nanowire arrays with an anodic alumina oxide template via electro-chemical deposition. Thin Solid Films 2014;570(PB):303e9. https://doi.org/10.1016/j.tsf.2014.03.014.

104. Tian Y, Li Z, Dou S, Zhang X, Zhang J, Zhang L, Wang L, Zhao X, Li Y. Facile prep-aration of aligned NiO nanotube arrays for electrochromic application. Surf CoatingTechnol 2018;337:63e7. https://doi.org/10.1016/j.surfcoat.2017.12.054.

105. Su FY, Zhang WD. Fabrication and photoelectrochemical property of In2O3/ZnOcomposite nanotube arrays using ZnO nanorods as self-sacrificing templates. MaterLett 2018;211:65e8. https://doi.org/10.1016/j.matlet.2017.09.085.

106. Epifani M, Prades JD, Comini E, Pellicer E, Avella M, Siciliano P, Faglia G, Cirera A,Scotti R, Morazzoni F, Morante JR. The role of surface oxygen vacancies in the NO2sensing properties of SnO2 nanocrystals. J Phys Chem C 2008;112(49):19540e6.https://doi.org/10.1021/jp804916g.

107. Godinho KG, Walsh A, Watson GW. Energetic and electronic structure analysis ofintrinsic defects in SnO2. J Phys Chem C 2009;113(1):439e48. https://doi.org/10.1021/jp807753t.

108. Joshi RK, Schneider JJ. Assembly of one dimensional inorganic nanostructures intofunctional 2D and 3D architectures. Synthesis, arrangement and functionality. ChemSoc Rev 2012;41(15):5285e312. https://doi.org/10.1039/c2cs35089k.

109. Yoshida T, Shinohara K, Itohara D, Fujita Y. Effects of thermal pressing on ZnOnanoparticle layers deposited by drop casting. e-J Surf Sci Nanotechnol 2016;14:175e8. https://doi.org/10.1380/ejssnt.2016.175.






https://doi.org/10.1007/s10853-008-2835-3

https://doi.org/10.1007/s10853-008-2835-3

https://doi.org/10.1038/nnano.2009.226

https://doi.org/10.1166/asl.2012.2188

https://doi.org/10.1166/asl.2012.2188

https://doi.org/10.1016/j.cap.2013.04.015

https://doi.org/10.1016/j.cap.2013.04.015

https://doi.org/10.1039/c6ra18747a

https://doi.org/10.1039/c6ra18747a

https://doi.org/10.1039/A700027H



https://doi.org/10.1016/j.surfcoat.2017.12.054


https://doi.org/10.1021/jp804916g

https://doi.org/10.1021/jp807753t

https://doi.org/10.1021/jp807753t

https://doi.org/10.1039/c2cs35089k

https://doi.org/10.1380/ejssnt.2016.175

110. Ko YH, Kim S, Yu JS. Drop-cast and dye-sensitized ZnO nanorod-based visible-lightphotodetectors. Phys Status Solidi Rapid Res Lett 2013;7(9):659e63. https://doi.org/10.1002/pssr.201307160.

111. Freer EM, Grachev O, Duan X, Martin S, Stumbo DP. High-yield self-limitingsingle-nanowire assembly with dielectrophoresis. Nat Nanotechnol 2010;5:525e30.https://doi.org/10.1038/nnano.2010.106.

112. Raychaudhuri S, Dayeh SA, Wang D, Yu ET. Precise semiconductor nanowire place-ment through dielectrophoresis. Nano Lett 2009;9(6):2260e6. https://doi.org/10.1021/nl900423g.

113. Chang Y-K, Hong FC-N. The fabrication of ZnO nanowire field-effect transistors byroll-transfer printing. Nanotech 2009;20:195302. https://doi.org/10.1088/0957-4484/20/19/195302.

114. Ramírez FH, Taranc�on A, Casals O, Rodríguez J, Romano-Rodríguez A,Morante JR, Barth S, Mathur S, Choi TY, Poulikakos D, Callegari V, Nellen PM.Fabrication and electrical characterization of circuits based on individual tin oxidenanowires. Nanotechnol 2006;17:5577e83. https://doi.org/10.1088/0957-4484/17/22/009.

115. Kim Y-K, Kim GT, Ha JS. Simple patterning via adhesion between a buffered-oxideetchant-treated PDMS stamp and a SiO2 substrate. Adv Funct Mater 2007;17:2125e32.https://doi.org/10.1002/adfm.200700217.

116. Maury P, Peter M, Mahalingam V, Reinhoudt DN, Huskens J. Patterned self-assembled monolayers on silicon oxide prepared by nanoimprint lithography and theirapplications in nanofabrication. Adv Funct Mater 2005;15:451e7. https://doi.org/10.1002/adfm.200400284.

117. Sun Y, Khang D-Y, Hua F, Hurley K, Nuzzo RG, Roger JA. Photolithographic routeto the fabrication of micro/nanowires of III-V semiconductors. Adv Funct Mater 2005;15:30e40. https://doi.org/10.1002/adfm.200400411.

118. Morag A, Jelinek R. “Bottom-up” transparent electrodes. J Colloid Interface Sci 2016;482:267e89. https://doi.org/10.1016/j.jcis.2016.07.079.

119. Hong LY, Lin HN. NO gas sensing at room temperature using single titanium oxidenanodot sensors created by atomic force microscopy nanolithography. Beilstein J Nano-technol 2016;7(1):1044e51. https://doi.org/10.3762/bjnano.7.97.

120. Vomiero A, Ponzoni A, Comini E, Ferroni M, Faglia G, Sberveglieri G. Direct inte-gration of metal oxide nanowires into an effective gas sensing device. Nanotechnol2010;21:145502. https://doi.org/10.1088/0957e4484/21/14/145502.

121. Griffith S, Mondol M, Kong DS, Jacobson JM. Nanostructure fabrication by directelectron beam writing of nanoparticles. J Vac Sci Technol B 2002;20:2768e72.https://doi.org/10.1116/1.1526697.

122. Donarelli M, Ferroni M, Ponzoni A, Rigoni F, Zappa D, Baratto C, Faglia G,Comini E, Sberveglieri G. Single metal oxide nanowire devices for ammonia andother gases detection in humid atmosphere. Procedia Eng 2016;168:1052e5. https://doi.org/10.1016/j.proeng.2016.11.338.

123. Morante JR. Chemical to electrical transduction mechanisms from single metal oxidenanowire measurements: response time constant analysis. Nanotechnol 2013;24(44):444004. https://doi.org/10.1088/0957-4484/24/44/444004.

124. Mårtensson T, Carlberg P, Borgstrom M, Montelius L, Seifert W, Samuelson L.Nanowire arrays defined by nanoimprinting lithography. Nano Lett 2004;4:699e702. https://doi.org/10.1021/nl035100s.

125. Ma P, Xu Z, Wang M, Lu L, Yin M, Chen X, Li D, Ren W. Fast fabrication of TiO2hard stamps for nanoimprint lithograph. Mater Res Bull 2017;90:253e9. https://doi.org/10.1016/j.materresbull.2017.03.010.


https://doi.org/10.1002/pssr.201307160

https://doi.org/10.1002/pssr.201307160

https://doi.org/10.1038/nnano.2010.106

https://doi.org/10.1021/nl900423g

https://doi.org/10.1021/nl900423g

https://doi.org/10.1088/0957-4484/20/19/195302

https://doi.org/10.1088/0957-4484/20/19/195302

https://doi.org/10.1088/0957-4484/17/22/009

https://doi.org/10.1088/0957-4484/17/22/009

https://doi.org/10.1002/adfm.200700217




https://doi.org/10.1016/j.jcis.2016.07.079

https://doi.org/10.3762/bjnano.7.97

https://doi.org/10.1088/0957–4484/21/14/145502

https://doi.org/10.1116/1.1526697

https://doi.org/10.1016/j.proeng.2016.11.338


https://doi.org/10.1088/0957-4484/24/44/444004

https://doi.org/10.1021/nl035100s



126. Sberveglieri G, Baratto C, Comini E, Faglia G, Ferroni M, Pardo M, Ponzoni A,Vomiero A. Semiconducting tin oxide nanowires and thin films for chemical warefareagents detection. Thin Solid Films 2009;517:6156. https://doi.org/10.1016/j.tsf.2009.04.004.

127. Ponzoni A, Baratto C, Bianchi S, Comini E, Ferroni M, Pardo M, Vezzoli M,Vomiero A, Faglia G, Sberveglieri G. Metal oxide nanowire and thin films basedgas sensors for chemical warefare simulant detection. IEEE Sens J 2008;8:735.https://doi.org/10.1109/JSEN.2008.923179.

128. Sysoev VV, Schneider T, Goschnick J, Kiselev, Habicht W, Hahn H, Strelcov E,Kolmakov A. Percolating SnO2 nanowire network as a stable gas sensor: direct com-parison of long-term performance versus SnO2 nanoparticle films. Sens Actuat B 2009;139:699. https://doi.org/10.1016/j.snb.2009.03.065.

129. Kumar R, Al-Dossary O, Kumar G, Umar A. Zinc oxide nanostructures for NO2gasesensor applications: a review. Nano-Micro Lett 2014;7(2):1e24. https://doi.org/10.1007/s40820-014-0023-3.

130. Ramírez FH, Taranc�on A, Casals O, Arbiol AJ, Rodríguez R, Morante JR. Highresponse and stability in CO and humidity measures using a single SnO2 nanowire.Appl Phys Lett 2008;93:123110. https://doi.org/10.1016/j.snb.2006.09.015.

131. Prades JD, Hern�andez-Ramírez F, Fischer T, Hoffmann M, M€uller R, L�opez N,Mathur S, Morante JR. Quantitative analysis of CO-humidity gas mixtures withself-heated nanowires operated in pulsed mode. Appl Phys Lett 2010;97(24):243105.https://doi.org/10.1063/1.3515918.

132. Comini E, Cristalli A, Faglia G, Sberveglieri G. Light enhanced gas sensing propertiesof indium oxide and tin dioxide sensors. Sens Actuat B 2000;65:260. https://doi.org/10.1016/S0925-4005(99)00350-0.

133. Comini E, Faglia G, Sberveglieri G. UV light activation of tin oxide thin films forNO2 sensing at low temperatures. Sens Actuat B 2001;78:73e7. https://doi.org/10.1016/S0925-4005(01)00796-1.

134. Comini E, Ottini L, Faglia G, Sberveglieri G. SnO2 RGTO UV activation for COmonitoring. IEEE Sens J 2004;4(1):17e20. https://doi.org/10.1109/JSEN.2003.822216.

135. de Lacy Costello BPJ, Ewen RJ, Ratcliffe NM, Richards M. Highly sensitive roomtemperature sensors based on the UV-LED activation of zinc oxide nanoparticles.Sens Actuat B 2008;134(2):945e52. https://doi.org/10.1016/j.snb.2008.06.055.

136. Law M, Kind H, Messer B, Kim F, Wang P. Photochemical sensing of NO2 withSnO2 nanoribbon nanosensors at room temperature. Angew Chem Int Ed 2002;41(13):2405e8. https://doi.org/10.1002/1521-3773(20020703)41:13<2405::AID-ANIE2405>3.0.CO;2-3.

137. Prades JD, Jimenez-Diaz R, Hernandez-Ramirez F, Cirera A, Romano-Rodriguez A,Morante JR. Equivalence between thermal and room temperature UV light-modulated responses of gas sensors based on individual SnO2 nanowires. Sens ActuatB 2009;140(2):337e41. https://doi.org/10.1016/j.snb.2009.04.070.

138. Prades JD, Jimenez-Diaz R, Manzanares M, Hernandez-Ramirez F, Cirera A,Romano-Rodriguez A, Mathur S, Morante JR. A model for the response towardsoxidizing gases of photoactivated sensors based on individual SnO2 nanowires. PhysChem Chem Phys 2009;11(46):10881e9. https://doi.org/10.1039/b915646a.

139. Espid E, Taghipour F. UV-LED photo-activated chemical gas sensors: a review. CritRev Solid State Mater Sci 2017;42(5):416e32. https://doi.org/10.1080/10408436.2016.1226161.

140. Ahmad R, Mahmoudi T, Ahn M-S, Hahn Y-B. Recent advances in nanowires-basedfield-effect transistors for biological sensor applications. Biosens Bioelectron 2018;100:312e25. https://doi.org/10.1016/j.bios.2017.09.024.




https://doi.org/10.1109/JSEN.2008.923179


https://doi.org/10.1007/s40820-014-0023-3

https://doi.org/10.1007/s40820-014-0023-3


https://doi.org/10.1063/1.3515918

https://doi.org/10.1016/S0925-4005(99)00350-0

https://doi.org/10.1016/S0925-4005(99)00350-0

https://doi.org/10.1016/S0925-4005(01)00796-1

https://doi.org/10.1016/S0925-4005(01)00796-1




https://doi.org/10.1002/1521-3773(20020703)41:13<2405::AID-ANIE2405>3.0.CO;2-3





https://doi.org/10.1039/b915646a

https://doi.org/10.1080/10408436.2016.1226161

https://doi.org/10.1080/10408436.2016.1226161

https://doi.org/10.1016/j.bios.2017.09.024

141. Nehra A, Pal Singh K. Current trends in nanomaterial embedded field effecttransistor-based biosensor. Biosens Bioelectron 2015;74:731e43. https://doi.org/10.1016/j.bios.2015.07.030.

142. Hackner A, Habauzit A, Muller G, Comini E, Faglia G, Sberveglieri G. Surface ioni-zation gas detection on platinum and metal oxide surfaces. IEEE Sens J 2009;9(12):1727e33. https://doi.org/10.1109/JSEN.2009.2030705. 5290390.

143. Bouxin B, Maier K, Hackner A, Mueller G, Shao F, Prades JD, Hernandez-Ramirez F, Morante JR. On-chip fabrication of surface ionisation gas sensors. SensorActuator B Chem 2013;182:25e30. https://doi.org/10.1016/j.snb.2013.02.049.

144. Ponzoni A, Zappa D, Comini E, Sberveglieri V, Faglia G, Sberveglieri G. Metal oxidenanowire gas sensors: application of conductometric and surface ionizationarchitectures. Chem Eng Trans 2012;30:31e6. https://doi.org/10.3303/CET1230006.

145. Baratto C, Todros S, Faglia G, Comini E, Sberveglieri G, Lettieri S, Santamaria L,Maddalena P. Luminescence response of ZnO nanowires to gas adsorption. SensorActuator B Chem 2009;140(2):461e6. https://doi.org/10.1016/j.snb.2009.05.018.

146. Singh N, Pandey V, Singh N, Malik MM, Haque FZ. Application of TiO2/SnO2nanoparticles in photoluminescence based fast ammonia gas sensing. J Opt (India)2017;46(3):199e203. https://doi.org/10.1007/s12596-017-0404-3.






https://doi.org/10.3303/CET1230006

https://doi.org/10.3303/CET1230006


https://doi.org/10.1007/s12596-017-0404-3

CHAPTER SIX

Hybrid materials with carbonnanotubes for gas sensingThara Seesaard1, Teerakiat Kerdcharoen2,Chatchawal Wongchoosuk31Department of Physics, Faculty of Science and Technology, Kanchanaburi Rajabhat University, MuangDistrict, Kanchanaburi, Thailand2Department of Physics and NANOTEC Center of Excellence, Faculty of Science, Mahidol University,Ratchathewi, Bangkok, Thailand3Department of Physics, Faculty of Science, Kasetsart University, Chatuchak, Bangkok, Thailand

Contents

6.1 Introduction 1866.2 Synthesis of carbon nanotube 192

6.2.1 Arc discharge 1926.2.2 Laser ablation 1926.2.3 Chemical vapor deposition 193

6.3 Preparation of carbon nanotubedmetal oxide sensing films 1946.3.1 Spin-coating 1946.3.2 Drop-coating 1976.3.3 Screen-printing 1976.3.4 Dip-coating 1986.3.5 Electron beam (E-beam) evaporation 198

6.4 Sensor assembly 1996.5 Characterization of carbon nanotubeemetal oxide materials 200

6.5.1 Raman spectroscopy 2006.5.2 X-ray diffraction 2016.5.3 Scanning electron microscope 2036.5.4 Transmission electron microscopy 204

6.6 Sensing mechanism of carbon nanotubeemetal oxide gas sensors 2056.7 Fabrication of electrodes and CNT/polymer nanocomposites for textile-based

sensors206

6.7.1 Preparation of textile-based electrode 2066.7.1.1 Crocheting technique 2066.7.1.2 Embroidery technique 2076.7.1.3 Screen printing technique 208

6.7.2 Preparation of CNT/polymer nanocomposite sensing materials 2096.8 Sensor assembly for textile-based gas sensors 210

6.8.1 Immersion-coating technique 2106.8.2 Drop-coating technique 211



https://doi.org/10.1016/B978-0-08-102559-8.00006-9

6.9 Characterization of CNT/polymer nanocomposites sensing materials on textilesubstrate

212

6.9.1 Scanning electron microscopy 2126.9.1.1 Fabric-based embroidered gas sensors 2126.9.1.2 Fabric-based screen-printed gas sensors 213

6.10 Sensing mechanism of CNT/polymer nanocomposites sensing materials onfabric substrate

215

6.11 Conclusion 216Acknowledgments 217References 217

6.1 Introduction

Until recently, classical methods such as human sensory evaluation,1

gas chromatography,2 and mass spectrometry3 have been the only availabletechniques for assessing the odors of objects, products, and the environment.Although the methods are reliable and accurate, practical utilization of theseinstruments is time-consuming, complicated, and costly. The advent ofchemical gas sensors and the electronic nose (e-nose) in the 1990s4,5 hasopened new opportunities for applications in many areas never seen before,especially for real-time, on-site, and rapid measurements. Since then, chem-ical gas sensors and the e-nose have been adopted as standard tools withwhich to complement, or even replace, traditional analytical instrumentsin many areas ranging from quality control of foods6,7 and beverages,8,9

environment protection10 to public safety.11 In general, chemical gas sensorscan be classified into four types12 based on their transduction principles:optical, thermal, electrochemical, and gravimetric. Among these techniques,electrochemical transduction has so far dominated applications of chemicalgas sensors in the measurement systems, because the interface setup ismore straightforward than other transduction methods.13e16

At present, most commercial chemical gas sensors adopt this technology,and metal oxide (MOX) semiconductors offer the most favored sensor archi-tecture due to their low-cost, high sensitivity, and simplicity in function.17

One could easily combine several functional elements in the same device,such as the sensitive layer, signal converter, and control electronics. Despitethe simple working principles ofMOX gas sensors, the gas sensing mechanismat the microscopic level is very complex and is still not adequately under-stood.18,19 Gas sensors made of the same MOX materials can have different

186 Thara Seesaard et al.

properties depending on the fabrication techniques and preparationconditions. It is firmly believed that catalytic reduction/oxidation at themicroscopic surface underlies the chemoresistive property of MOXs.20 Thesereactions are governed by the electronic structure, chemical composition,crystal structure, and relative orientation of the oxide surface to the analytemolecules, thereby allowing their gas sensing properties to be tuned bymodifying such parameters. The most successful approach to optimizingthe gas sensing properties of MOXs is to modify the microscopic structureby reducing grain size and modifying various crystallite parameters.21

Among MOX materials, tin oxides have been the most frequently usedsolid-state gas sensors. The sensitivity and selectivity of these materials canbe tuned on the basis of structural engineering. Tin oxides have a rich setof structural parameters that can be modified. For example, tin oxidenanocrystals obtained from a spray pyrolysis experiment can have manycrystallographic planes such as (110), (111), (200), (101), (011), (�1,�1,2),etc.22 Such crystallographic parameters are sensitive to the change in grainsize responsible for the different gas sensing properties of films prepared usingdifferent conditions.23 MOXs can be doped by a small amount of metals, suchas Sn, Pd, Cu, Nb, etc., to modify structural and electronic properties. It wasfound that doping tin oxide with Sn, In, and Nb leads to a decrease in thegrain size down to the nanometer range.24 Kawamura et al. found that theinterplay between different crystal growth directions can be controlled bythe addition of impurities.25 Besides pure metals, MOXs can be doped ormixed with organic materials, leading to so-called “hybrid” chemical gassensors. The field of hybrid chemical gas sensors is still in the infant stage.26

Combining both hard and soft materials into a single film is quite challengingdue to complications both in the preparation and fabrication processes.27 Inthis chapter, we are specifically interested in the hybridized MOX gas sensorsbased on MOX and carbon nanotube (CNT) composites.28,29 For moreinformation about the modification of the MOXs with other additives, otherchapters in this book or the current reference section should be consulted.30

MOXs are a very robust technology mostly adopted commercially forsemiconductor gas sensors. The stability and durability (an average life of5 years) have been gladly welcomed by industries, including the environ-ment, security, petrochemicals, and agriculture. However, MOXs havecertain disadvantages that limit their applicability in many areas, such as inmobile devices where energy consumption is a major concern. Reducingthe operating temperature (around 250e400 �C for most MOXs) toroom temperature has become a topic of research interest worldwide.

Hybrid materials with carbon nanotubes for gas sensing 187

Because the sensing mechanism of MOXs is based on the surface reactivityof the materials to incoming analyte gases where electron transfer will play amajor role,31 engineering the conductivity of the surface can lead todesirable sensing properties. Doping with impurities has been a successfultechnique for the modification of MOX surfaces. Apart from metal dopantsas mentioned in the previous section, CNTs have several advantages overother composite materials.29 The CNT is very conductive and the gassensing can be performed at room temperature. Consequently, mixingCNT with MOXs would result in increasing surface conductivity andreducing the operating temperature. The highly specific area of CNTswill also enhance the active surface of MOXs, leading to enhancedsensitivity and selectivity. MOX and CNT composite materials can beclassified into two groups, depending on which material is the greater inthe composition: MOX-decorated CNTs and CNT-doped MOXs.1. MOX-decorated CNTs. In this case, the CNT is functionalized by attach-

ing MOX nanoparticles, either by bonded or nonbonded interaction,onto the sidewall of the CNT.32,33 The most common method forachieving strong interaction of MOX nanoparticles on the CNT is tooxidize CNTs by strong acids to introduce carboxyl or hydroxyl groupson the CNT surface. Such functional groups can directly interact withthe oxygen of the MOX nanoparticles via hydrogen bonding. Thebonded interactions between such functional groups with metal atomsthrough the pair of electrons on the oxygen are also possible. VariousMOX nanocrystals coated onto CNTs have been investigated for theirfunctionality, such as ZnO,34 TiO2,

35 SiO2,36 SnO2,

37 MnO2,38 and

Fe2O3.39 Such functionalization creates novel properties which extend

the applicability of CNT into many new areas, such as capacitors,photocatalysts, and batteries. In gas sensing applications, MOX-decorated CNTs have shown enhanced sensitivity, improved responseand recovery times, and a dramatic reduction in operating temperature.Several ambient gases have been reporteddfor example, CO, NO2,NH3, and ethanol (see reference 29 and comprehensive reference listtherein).

2. CNT-doped MOXs. In this case, CNTs are embedded within the MOXmatrix. This chapter will focus on this type of CNT and MOX hybridmaterial. The MOX/CNT thin films can be prepared using varioustechniques, such as spin-coating, drop-coating, dip-coating, and electronbeam evaporation, details of which will be given in the followingsections.


Although the MOX hybrid materials were actively developed by manyresearchers for the last several decades, but in fact the MOX gas sensors havecertain restrictions that do not support the development in wearable sensingtechnologies. A desire to overcome these restrictions has given impetus toresearchers to focus on searching for new gas sensing materials and emergingtechnologies in the field of textile-based gas sensors. CNT/polymernanocomposite materials have received special attention from researchersworldwide, as it can operate at room temperature, low power consumptionrequirements, and offer several design possibilities with the e-textiles.40 Inaddition, CNT/polymer nanocomposite materials are also investigated asalternative gas sensing materials for several applications such as healthcare,41

agriculture,42 industry,43 and environment.44e46 Recently, researchers havedeveloped a wearable smart technology to change both a design and afabrication process by using electronic circuit and digital componentsembedded in clothing known as e-textile or electronic textile. E-textiletechnology revolts fiber and textile manufacturing and has special propertiesthat cannot be found in a traditional fabric such as in communication,perceptive functions, and conduct energy.47,48 An example of wearablee-textile innovation resulting from the combination of textile and electronictechnology providing an ideal platform for wearable health tech devices isthe biometric smart shirt; a body metric system that monitors cardiac,breathing patterns49 and body activity tracker of baby sleep positions withseveral sensors embedded in the fabric.50 There are also smart sock trackcadences which can access to real-time biometric data to indicate walkingabnormalities, gait analysis, and weight distribution of the body evenlyaround the foot while walking and running.51 Consequently, we will beable to track the progress of effectiveness of exercise. However, it mightbe possible to integrate electronic components into textiles or fabrics inthe early stages of e-textile technology because the device connectivitycomponent is unfashionable due to untidiness of the electrical wiring andinflexibility of the cables when woven into clothing. It is likely that hazardmay occur especially while wearing an unwieldy and inconvenient garment.Then, the researchers try to figure out the appropriate method to mergeelectronics with clothing and jewelry. As a result, these devices will besafe to use and real-time application monitoring will be easily built. It ispossible to have a more perfect form of communication without any cablesdue to the wireless platform for new frontier of e-textile applications such ase-textile fashion, finishing and decoration, military applications, and smartclothing.52


In addition, there are new developments in electrical properties oftextile materials, including a conductive thread, metallic finish yarn, silveryarn, polyester powder coating/metal finishes, and synthetic fabricsconsisting of fibers with high electrical conductivity.53 The innovativee-textile devices and textile materials (fabrics, yarns, and threads) havebeen designed with different types of fabrication techniques and processessuch as luminous fabric and light-emitting fiber which were developed todecorate the building and automotive parts.54 Besides, Philips research andthe Institute of Textile Institute TITV Greiz, Germany, has collaborativelearning in developing e-textile technology for health and quality of life,which is known by the name of photonic textiles. Photonic textileswere created by embedding LEDs into plastic or film and then woveninto the fabric to make them soft and flexible to allow unlimited shapes,which can increase the user interaction by combining the functionalityof sensors and communication devices.55 Additionally, PLACE-it Project(Platform for Large Area Conformable Electonics by Integration) hasdeveloped an optoelectronic which is extremely thin, lightweight, andcan be easily applied for using in health and medicine, with some applica-tions, such as skin treatments and measuring the circulation of the bloodfrom all parts of the body. Moreover, it can also be used in product designsuch as lamps, curtains, advertising, and fashion.56 E-textile can not onlyintegrate electronics directly into the textile substrates but also makeelectronic components from fibers and textiles. For example, researchesfrom NC STATE UNIVERSITY College of Textiles have demonstratedthat the development of lithium-ion battery provides better performanceby using MnOx/C nanofibers instead of graphite in the anode of thebattery, called 18650 cells.57 An interesting smart textile project, theProeTEX (PROtection E-TEXtiles), is the research cooperation of Euro-pean countries which gives greater importance to the development ofe-textile-based MicroNano technology and wearable system for rescueworkers and firemen.58,59 Recent research in electronic textile technologyfield has found out that the key component of this technology is theelectronic circuit part on textile for controlling and processing equipment.Thus, new fabrication process and materials in the development ofelectronic component on textiles substrate providing a consistent qualityof the e-textile work piece in every production is a core basis for creatinga wearable electronic textile system. Currently, there has been increasinginterest in using screen printing technology as a manufacturing process in


the development of electronic commerce research to construct a perpen-dicular wiring structure on textile substrate,60,61 which is possible toproduce electronic components and circuits on textile that cannot beachieved by other methods such as wet chemical or photolithographictechnology. Electronic circuit board in industrial electronic sector weredeveloped by using a screen printing process for the reason that this processcan produce many items in a relatively short period of time and they can bereproduced at low cost with regular quality of the work piece and precisepattern control.

Although the e-textiles were actively developed by many researchers inthe last decade, the innovative e-textile products were found mainly infashionable clothing and decoration. While there are a lot of researches tosupport learning and creating innovation-oriented research for healthcareapplications most of which are the innovations that can aid interpretationof measurement resulting in terms of physiological parameters andbiokinetics such as respiration, movement, touch, brain waves, heart rate,breathing, cardiac activity, and body temperature.62 It can be seen thatamong a variety of smart textile innovations, there is still space availablefor the creation of an innovative e-textile for molecular detection usingnanomaterial and nanohybrid materials. To make e-textile technologymore perfect, researchers have been conducting research, which is relatedto the development of textile-based gas sensors used as wearable electronicnoses.63 Research has been carrying on in the direction of e-textile develop-ment so as to be widely accepted. Sniffing e-textile innovation was designedto be more fashionable and comfortable for using in daily life so that thewearer would be able to work beyond the original limit. At first, the circuitboard was developed from a rigid material and then it was changed to softmaterial such as fabric, rubber, and plastic to make it flexible. In addition,fabric substrate is also friendly to human skin and esthetically acceptable.For researchers, the biggest challenge is to overcome those restrictions.The details of research are related to the e-textile technology and clothingis the first priority that we emphasize on because it is what you can wearthroughout your whole life. Moreover, clothing can not only be easilyadapted to biological function that actually happens but also give comfortto the physical mobility. In this chapter, we will highlight the mostimportant fabrication process of flexible CNTebased textile gas sensors,characteristics of CNT/polymer nanocomposites for multifunctionalitysensing, and the main progress in gas sensing.


6.2 Synthesis of carbon nanotube

CNTs are generally produced using three main techniques: arcdischarge, laser ablation, and chemical vapor deposition (CVD). Eachtechnique can be modified to suit the specific research purpose.

6.2.1 Arc dischargeArc discharge was the first technique recognized for producing multiwalledcarbon nanotubes (MWCNTs)64 and single-walled carbon nanotubes(SWCNTs).65,66 The arc discharge technique generally involves the useof two high-purity graphite electrodes as anode and cathode. The elec-trodes are vaporized by the passage of a DC current (w100 A) throughthe two high-purity graphite electrodes separated (w1e2 mm) in400 mbar of helium atmosphere. After arc discharging for a period oftime, a carbon rod is built up at the cathode. The native method willmainly produce MWCNTs, rather than SWCNTs. However, with theaddition of a metal catalyst, such as Fe, Co, Ni, Y, or Mo, on either theanode or the cathode, SWCNTs can also be produced. The quantityand quality (such as length, diameter, purity, etc.) of the nanotubesobtained depend on various parameters, such as the surface density ofthe metal catalysts, inert gas pressure, type of gas, plasma arc, temperature,current, and system geometry.

6.2.2 Laser ablationSmalley and colleagues produced CNTs using the laser ablation technique in1995.67 For the laser ablation technique, a high-power laser is used tovaporize carbon from a graphite target at high temperature. BothMWCNTsand SWCNTs can be produced with this technique. To generate SWCNTs,metal particles must be added as catalysts to the graphite targets, similar to thearc discharge technique. The quantity and quality of CNTs produceddepend on several factors, such as the amount and type of catalysts, laserpower and wavelength, temperature, pressure, type of inert gas, and the fluiddynamics near the carbon target. The laser was focused onto a carbon targetcontaining 1.2% of cobalt/nickel with 98.8% of graphite composite that wasplaced in a 1200 �C quartz tube furnace under an argon atmosphere(w666.61 mbar). These conditions were achieved for production ofSWCNTs in 1996 by Smalley’s group.68 In such a technique, argongas carries the vapors from the high temperature chamber into a cooledcollector positioned downstream. The nanotubes will self-assemble from


carbon vapors and condense on the walls of the flow tube. The diameterdistributions of SWCNTs that result from this method vary by about1.0e1.6 nm. CNTs produced by laser ablation were purer (up to 90%purity) than those produced by the arc discharge process and have a verynarrow distribution of diameters.

6.2.3 Chemical vapor depositionThe use of the CVD technique to produce MWCNTs was first reported byEndo and his research group in 1993.69 Three years later, Dai in Smalley’sgroup successfully adapted CO-based CVD to produce SWCNTs.70 TheCVD technique can be achieved by taking a carbon source in the gas phaseand using an energy source, such as plasma or a resistively heated coil, totransfer energy to a gaseous carbon molecule. The CVD process employshydrocarbons as the carbon sources, including methane, carbon monoxide,and acetylene. The hydrocarbons flow through the quartz tube placed insidean oven at a high temperature (w720 �C). At high temperature, the hydro-carbons are broken down to hydrogen and carbon radicals, producing purecarbon clusters. The carbon then diffuses to the substrate, which is heatedand coated with a catalyst (usually a first-row transition metal such as Ni,Fe, or Co) where CNTs will be formed if the proper parameters are main-tained. The advantages of the CVD process are low power input, lowertemperature range, relatively high purity, and, most importantly, the possi-bility of scaling up the process. This method can produce both MWCNTsand SWCNTs depending on the temperaturedproduction of SWCNTswill occur at a higher temperature than MWCNTs. It should be notedthat SWCNTs can be classified into metallic and semiconducting typesdepending on their diameters and chiralities. Synthesis of SWCNTs alwaysproduces a mixture of metallic and semiconducting SWCNTs which is oneof the crucial problems for the development of SWCNT-based electronicapplications. Nowadays, several methods, including amine extraction,71

DNA separation,72 modified free solution electrophoresis,73 density-gradient ultracentrifugation,74 polymer wrapping,75 etc., have beenemployed to separate semiconducting and metallic SWCNTs. The mostrecent progress on the structure separation of SWCNTs can be found in aliterature.76 In this chapter, most of reviewed papers did not mention tothe type of SWCNTs. Therefore, it can be assumed to a mix of semicon-ducting and metallic types and effects of the types on gas sensing propertiesare neglected.


6.3 Preparation of carbon nanotubedmetal oxidesensing films

Sensing film is the heart of a gas sensor device. The key to success indeveloping a gas sensor device is a technique capable of preparing a sensingfilm that exhibits high selectivity and sensitivity to a desired target gas,long-term stability, good repeatability, rapid response, small size, and lowpower consumption. Until now, it has been widely known that mostcommercially available gas sensors are still based on pure MOX gas sensors(i.e., SnO2 and WO3). Such gas sensors have been successfully used inmany applications, but they still suffer from poor selectivity and high powerconsumption. These disadvantages can be crucial obstacles for the develop-ment of future advanced technology, such as wearable sensing devices.Recently, doping of CNTs into MOX has attracted considerable attentionbecause hybrid SWCNTs/SnO2 sensors exhibit high sensitivity and a goodrecovery property in detecting NO2 at room temperature.77 The hybridCNT/MOX gas sensors based on thin-film nanostructures are summarizedin Table 6.1.

As shown in Table 6.1, there are five methods to deposit hybrid CNTs/MOXs onto electrodes.

6.3.1 Spin-coatingSpin-coating is a method for applying liquid-based coatings onto a rotatingsubstrate. A typical spin-coating process consists of four basic stages,89 asshown in Fig. 6.1. The coating liquid material is applied to the top ofsubstrate in the deposition stage. The amount of applied liquid dependson the viscosity of the liquid and the size of the substrate to be coated. Inthe acceleration stage, liquid is spread across the wafer by centrifugal force.The spinning speed is set at a specific value depending on the desired filmthickness. The coated substrate is then spun at a higher speed. The liquidflows radially outward, whereas excess liquid flows to the perimeter andleaves as droplets. In the final stage, evaporation of the solvent takes overas the primary mechanism of thinning. The thickness of the dry film (Lfilm)with an approximation of constant evaporation and no liquid remaining atthe end of the process can be written as follows:90,91

Lfilm ¼"3b0

�x0A� xIA

�ek2r

#1=3�1� x0A

�u�1=2 (6.1)


Table 6.1 List of hybrid CNT/MOX gas sensors.Target gas Detection range Operating temperature Sensing material Fabrication technique References

NO2 25e1000 ppm 25 �C SWCNTs-SnO2 Spin-coating 77NO2 100e500 ppb 25 �C MWCNTs-SnO2 Drop-coating 78CO 10e50 ppm 150 �C MWCNTs-WO3 Drop-coating 78CH2O 0.03e10 ppm 250 �C MWCNTs-SnO2 Screen-printing 79NH3 60e800 ppm 25 �C MWCNTs-SnO2 Spin-coating 80EtOH/MeOH 100e1000 ppm 250 �C MWCNTs-SnO2 E-beam evaporation 81H2 5000e50,000 ppm 250 �C MWCNTs-WO3 E-beam evaporation 82NO2 100e1000 ppb 25 �C MWCNTs-WO3 Drop-coating 83NH3 20 ppm 25 �C CNTs-ZnO Spin-coating 84CH3COCH3 1 vol.% 25 �C MWCNTs-TiO2 Screen-printing 85NH3 1 vol.% 25 �C MWCNTs-TiO2 Dip-coating 86O2 10 ppm 350 �C CNTs-TiO2 Drop-coating 87H2 4% in air 25 �C SWCNTs-Co3O4 Spin-coating 88

Hybrid

materials

with

carbonnanotubes

forgas

sensing195

ek ¼ cDgpAMA

RTrb1=2g

(6.2)

where b0 is kinematic viscosity; x0A represents the initial concentration ofsolvent in the coating liquid; xIA represents the mass fraction of solvent in thecoating liquid that would be in equilibrium with the mass fraction of solventin the bulk gas; u is the spin speed; c denotes the ratio of kinematic viscosityand mass diffusivity of the ambient gas; Dg is the binary diffusivity of thesolvent in the ambient gas; pA is the vapor pressure of the pure solvent attemperature (T);MA is the molecular weight of the solvent; and R, r, and bgdenote the universal gas constant, liquid density, and kinematic viscosity ofthe ambient gas, respectively.

To prepare the CNT/MOX liquids for spin-coating, the SWCNTbundles were dispersed in the organometallic solutions (Sn [OOCCH(C2H5)-C4H9]2,aq; tin (II) 2-ethylhexanoate w90% in 2-ethylhexanoicacid)77 by ultrasonic vibration. Alternatively, MWCNT bundles withSnO2 nanoparticles and cetyltrimethyl ammonium bromide can bedispersed in water.80 In the case of synthesis of CNT-ZnO,84 CNT in amixture of ethanol and water was dropped into triethanolamine andZnCl2 solution at a specific temperature. For SWCNT/Co3O4 thin films,88

these can be formed by spin-coating a metalepolymer complex(Cox(C2H5N)n), as a product of the reaction of CoSO4$7H2O and polye-thylenimine in water, onto an SWCNT thin film. After deposition, theCo3O4 and SWCNT thin films were annealed at a high temperature toform Co3O4/SWCNT composite.

ωω

dω dt ≠ 0

(a)

(c) (d)

(b)

Figure 6.1 The four basic stages of spin coating: (a) deposition, (b) acceleration, (c) flowdomination, and (d) evaporation.


6.3.2 Drop-coatingDrop-coating is a simple method for preparing films and employs a precisionpipette to put solution onto the substrate. The film thickness is controlled bythe amount and concentration of solution deposited on the substrate. Thistechnique is closely related to spin-coating. Therefore, it can be used asan alternative when spin-coating is not possibledfor example, when thesolvents are not sufficiently volatile to evaporate during a spin-coating pro-cess or not sufficiently viscous to produce a thick film.92 Preparation of someCNT/MOX sensors78,83,87 successfully uses this method for depositingfilms. Before a drop-coating process, hybrid CNT/MOX solution is usuallyprepared using an adapted solegel method for obtaining well-dispersedCNT in the MOX matrix. For example, CNT/TiO2 was obtained byadding CNTs to TiO2 prepared from titanium isopropoxide(IV) Ti[OCH(CH3)2]4 precursors in a dry nitrogen atmosphere. An adequatemixture of the two components was obtained by dissolving them in glycerol(employed as an organic vehicle) and stirring the resulted solution in anultrasonic bath at a specific temperature.78

6.3.3 Screen-printingScreen-printing is a commonly used industrial technique for fast andinexpensive deposition of films over large areas. The principle of screen-printing93 is shown in Fig. 6.2.

Emulsion

Screen

PasteSqueegee

Snapp-off

Substrate

Printing

Levelling

Wet film

Figure 6.2 The screen printing process.


A pattern is photographically defined on a stainless steel screen by meansof an emulsion layer. A paste of the material to be screen-printed is pressedthrough the screen using a foam applicator (squeegee). After leveling, theprinted wet film is dried at a specific temperature. The thickness of thescreen-printed film depends on the viscosity of the paste, the pressure andspeed of the squeegee, the snap-off distance between the screen and thesubstrate, and the mesh number of the screen. In a roller squeegee system,Fox and his colleague employed a numerical model to estimate depositionthickness at different half-tone coverage (more detail can be found inRef. 94). Preparation of a paste to fabricate an MWCNT/TiO2 gas sensorwas reported by Sanchez et al.85 The MWCNT/TiO2 composites wereprepared by solegel techniques using titanium tetraisopropoxide[Ti(C3H6OH)4] as the precursor and 2-propanol as the solvent. The mixturewas added in HCl and heated at a specific temperature. The solid output wasmixed with few drops of Triton-X and propylene glycol to prepare a pastefor the screen-printing method.

6.3.4 Dip-coatingDip-coating can be described as a process where a substrate is dipped into asolution. It is then withdrawn from the solution at a controlled speed undercontrolled temperature and atmospheric conditions. The coating thickness isprimarily affected by the withdrawal speed, fluid viscosity, fluid density, andsurface tension. If the withdrawal speed is chosen such that the sheer rateskeep the system in the Newtonian regime, the coating thickness (LDip)can be calculated by the LandaueLevich equation:95,96

LDip ¼ 0:94ðhnÞ2=3

g1=6LV ðrgÞ1=2

(6.3)

where h denotes fluid viscosity, v represents the withdrawal speed, gLV is theliquidevapor surface tension, r is fluid density, and g is gravity.

In the case of MWCNT/TiO2 gas sensors, S�anchez and Rinc�on86

employed dip-coating based on a solegel solution. The solegel solutioncontaining Ti-isopropoxide and acid-treated MWCNTs was either precip-itated or kept as a sol by adjusting the pH and surfactant concentration.

6.3.5 Electron beam (E-beam) evaporationThe electron beam (E-beam) evaporation process is a physical vapor depo-sition that yields a high deposition rate from 0.1 to 100 mm/min at relativelylow substrate temperatures. The E-beam process offers extensive possibilities


for controlling film structure and morphology, with desired properties suchas dense coating, high thermal efficiency, low contamination, highreliability, and high productivity. The deposition chamber is evacuated toa pressure of 1.33 � 10�5 mbar or lower. The material to be evaporatedis in the form of ingots or a compressed solid. The E-beam can be generatedfrom electron guns by thermionic emission, field electron emission, or theanodic arc method. The electron beam is accelerated to a high kineticenergy and focused toward the starting material. The kinetic energy ofthe electrons is converted into thermal energy that will increase the surfacetemperature of the materials, leading to evaporation and deposition onto thesubstrate. The deposition rate depends on the starting material and E-beampower. The deposited film thickness can be measured in situ by a quartzcrystal monitor. The evaporation of CNTs with MOXs (i.e., SnO2 andWO3) is a relatively new concept. A plausible mechanism for CNT/MOX coevaporation can be drawn as follows81,82: the E-beam is used tobombard the surface of the starting materials (i.e., CNT/SnO2 or CNT/WO3). The MOXs (such as SnO2 or WO3) are evaporated at a temperatureof w1500 �C in a high vacuum, while CNT fragments that are small andvery light are carried into the vapor by surrounding SnO2 or WO3 mole-cules. It should be noted that CNTs themselves are not decomposed duringevaporation because this temperature is well below CNT sublimation point(>3000 �C) in a high vacuum condition. When CNT molecular fragmentsarrive at the substrate, SnO2 or WO3 vapor is condensed and coated aroundthem. As the substrate cools down, CNTs remain in the lattice of MOXsdue to physicochemical binding between the MOXs and CNTs.

6.4 Sensor assembly

A typical sensor structure is displayed in Fig. 6.3. The sensing film isdeposited on top of a substrate between the electrodes. The heater is also in-tegrated on the reverse of the substrate. It should be noted that a heater unitmay not be necessary if the sensor will be operating at room temperature.

ElectrodeSensingmaterial Electrode

Substrate

Heater

Figure 6.3 A simple sensor structure.


Apart from the sensing film, the electrodes also play an important role ingas sensing response. For instance, the electrode material, gap sizes, andelectrode structure can affect the sensor response.97e99 Mishra and Agar-wal97 reported that the sensitivity of the thick-film SnO2 sensor for H2

and CO is much higher when silver electrodes are used instead of goldelectrodes (about 65.5% and 42.6%, respectively). Tamaki et al. foundthat sensitivity was increased with decreasing gap size.98 The performanceof the sensor was improved by using interlacing electrodes.99 Therefore,the design of a gas sensor structure is necessary for fabricating a high-performance hybrid CNT/MOX sensing device.

6.5 Characterization of carbon nanotubeemetaloxide materials

To confirm the structure and quality of produced CNT and MOXhybrid materials, there are four characterization techniques that are normallyused. These techniques are described in the following subsections.

6.5.1 Raman spectroscopyRaman spectroscopy is a spectral measurement based on inelastic scatteringof monochromatic radiation. When a molecule is irradiated with an intensemonochromatic light (usually a laser source), photons excite the moleculefrom the ground state to a virtual energy state. The photons are reemittedwhen the molecule relaxes. The frequency of the reemitted photons shiftsin comparison with the original monochromatic light frequency. This shiftprovides information about vibrational, rotational, and other low frequencytransitions in molecules. Information from Raman spectroscopy is summa-rized in Fig. 6.4.

Analysis

Properties

Characteristicraman

frequencies

Composition ofmaterial

Changes infrequency oframan peak

Stress/strainstate

Polarization of raman peak

Crystalsymmetry and

orientation

Width oframan peak

Qulaity ofcrystal

Intensity of raman peak

Amount ofmaterial

Figure 6.4 Information from Raman spectroscopy.


Raman spectroscopy was used to confirm the existence of CNTs in anMOX film. Raman spectra of an SWCNT/Co3O4 film

88 are displayed inFig. 6.5. The peak of crystalline Co3O4 can be clearly observed at694 cm�1 for the A1g mode, while it appears as two significant peaks forSWCNTsdnamely D-band and G-band at 1350 and 1590 cm�1, respec-tively. It should be noted that the intensity of the D-band(w1300e1500 cm�1) is a qualitative metric of SWCNT defects holdingsignificant information on the crystalline quality, while the G-band(w1500e1605 cm�1) is derived from the in-plane vibration usually existingin graphite and useful for measuring SWCNT graphene sheet folding. Foranalysis of a CNT/MOX sensing film, the Raman shift for the MOXs(i.e., SnO2, WO3, TiO2, etc.), D-band, and G-band should be observed.

6.5.2 X-ray diffractionX-ray is a high-energy electromagnetic radiation having energies rangingfrom w200 eV to 1 MeV. The X-ray diffraction (XRD) is based on theelastic scattering of monochromatic X-rays. It is usually used to characterizethe chemical composition and crystallographic structure of materials byplotting the angular positions and intensities of the resultant diffracted peaksof radiation satisfied with Bragg’s law conditions. The diffraction intensitycan be written as follows:100

IðhklÞa ¼ I0l3

64pr

�e2

mec2

�2MðhklÞV 2a

��FðhklÞa��2�1þ cos2ð2qÞcos2ð2qmÞ

sin2 q cos q

�hkl

na

ms

(6.4)

400 800 1200 1600

Inte

nsity

(a.u

.)

Raman shift (cm–1)

CO3O4

G-band

D-band

Figure 6.5 Raman spectra of single-walled carbon nanotube/Co3O4 film (upper line),Co3O4 thin film (middle line), and the SiO2/Si substrate (lower line).88


where I(hkl)a is the intensity of the reflection of hkl in phase a, I0 is theincident beam intensity, l denotes the X-ray wavelength, r denotes thedistance from the specimen to the detector, (e2/mec

2)2 represents the squareof the classical electron radius,M(hkl)a is the multiplicity of reflection of hkl inphase a,Va is the volume fraction of phase a, F(hkl)a is the structure factor forreflection hkl of phase a (i.e., the vector sum of scattering intensities of allatoms contributing to that reflection), 2qm represents the diffraction angle ofthe monochromator, va is the volume of the unit cell of phase a, and ms isthe linear absorption coefficient of the specimen.

The XRD patterns of MWCNT/SnO280 are shown in Fig. 6.6. In

general, an XRD pattern of CNT locates near the (002), (100), (110), and(112) reflections of graphite. The prominent peak (2q z 26�) can be attrib-uted to the (002) reflection of carbon. In this case, the most intense twopeaks of MWCNTs correspond to (002) and (100), while only SnO2 inthe crystalline phase can be indexed from the patterns for SnO2. It can beobserved that the characteristic peaks of MWCNT/SnO2 composites arequite similar to the patterns of SnO2. From this observation, it may be hy-pothesized that the MWCNTs are well-embedded in the SnO2 matrix orthere are no MWCNTs in the SnO2 matrix. However, almost all CNT/MOX films from other studies have a similar pattern. Peaks of CNT are usu-ally absent for the CNT/MOX composite films in the XRD analysis. Othertechniques may need to confirm the existence of CNTs in MOX films.

Figure 6.6 X-ray diffraction patterns of (a) SnO2, (b) multiwalled carbon nanotubes(MWCNTs), and (c) SnO2/MWCNTs composites.80


6.5.3 Scanning electron microscopeThe scanning electron microscope (SEM) employs a focused beam of high-energy electrons to generate a variety of signals at the surface of sample. Thetypes of signals produced from the interaction of the high-energy electronswith the sample include secondary electrons, back-scattered electrons,characteristic X-rays, and other photons of various energies. These signalscan be used to examine many characteristics of the samples, such as surfacetopography and morphology and crystallographic information and compo-sition. The basic principle of SEM is shown in Fig. 6.7.

The SEM surface morphology of a CNT/WO3 film prepared byE-beam evaporation is displayed in Fig. 6.8. One can see that the sensingfilm prepared by this technique is highly homogeneous, with grain sizesranging from 40 to 80 nm. It should be noted that the surface morphologyof other films (including pure SnO2, pure WO3, and CNT/SnO2) preparedby E-beam evaporation81,82 is in accordance with observations on the nano-crystalline CNT/WO3 film. With SEM resolution, a CNT structure cannotbe observed on the thin film surface. In cases of CNT/MOX films preparedby other methods (i.e., spin-coating),77 the morphologies of the pure MOX

Figure 6.7 Principle of scanning electron microscope.


and hybrid CNT/MOX are also very similar. Thus, it is quite difficult toobserve the CNTs on the surface. In the previous studies, it was suggestedthat CNTs are mostly embedded in the MOX-based matrix.

6.5.4 Transmission electron microscopyTransmission electron microscopy (TEM) provides a much higher spatialresolution than SEM. TEM can facilitate study of the inner structure andanalysis of the features on an atomic scale (in the range of a few nanometers).Although the TEM technique involves electrons to produce enlarged im-ages similar to the SEM technique, the working principle of TEM is some-what different from SEM. In general, TEM uses high E-beam energies inthe range of 60e350 keV to pass through a thin sample to project an imageonto a fluorescent screen. The sample for TEM is usually required to besliced into an extremely thin section (<100 nm) and pretreated with heavymetals (staining) before visualization. The image resolution of TEM (d) interms of the classic Rayleigh criterion for the visible-light microscope canbe given approximately by Eq. (6.5):

d ¼ 0:61lm sin b

(6.5)

where l denotes the wavelength of the radiation, m represents the refractiveindex of the viewing medium, and b is the semiangle of collection of themagnifying lens.

Figure 6.8 Scanning electron microscopic image of multiwalled carbon nanotubeseWO3 thin film on Si substrate.


TEM characterization can be used to confirm CNT inclusion in MOXfilms. A typical high-resolution TEM (HRTEM) image of a CNT/WO3

composite prepared by E-beam evaporation is shown in Fig. 6.9. It shouldbe noted that HRTEM uses both transmitted and scattered beams to createan interference image. HRTEM observation clearly shows that a singleMWCNT fragment is, indeed, embedded into the nanocrystalline WO3

layer (see Fig. 6.9(a)). The diameter of CNTs and the crystal size of WO3

were estimated to be in the range of w20e50 nm (see Fig. 6.9(a)) and3e10 nm (see Fig. 6.9(b)), respectively. By comparison with pure WO3,the doping of CNT does not change the phase or surface morphology ofthe film, but it may help form nanochannels in the MOX films, leadingto the enhancement of the sensitivity and reduction of the operatingtemperature.

6.6 Sensing mechanism of carbon nanotubeemetaloxide gas sensors

It is widely known that many MOXs (such as WO3, SnO2, and TiO2)are n-type semiconductors, while CNT is a p-type semiconductor. CNT/MOX gas sensors can be either p-type or n-type semiconductors, dependingon the quantity of CNTs and the operating temperature. The CNT/MOXgas sensor behaves as an n-type semiconductor if the electrical conductivityof the gas sensor increases when reducing gases (i.e., H2, CO, or NH3) areabsorbed by its surface. In the case of the p-type semiconductor, the electri-cal conductivity of the sensor increases in the presence of an oxidizing gas

Figure 6.9 Typical high-resolution transmission electron microscopic image of (a) CNT/WO3 film and (b) nanocrystalline WO3 prepared by E-beam evaporation technique.


(O3, NOX, etc.). Various oxygen species chemisorbed at the surface (such asO2�, O2

�, and O�) are available for catalytic reactions with gas, dependingon the temperature at the MOX surface. Of the oxygen species, O� iscommonly chemisorbed at the operating temperature range of 200e400�C, while O2� and O2

� are mostly contributed at low temperature.101

The main sensing mechanism of CNT/MOX gas sensors can be describedby using the model of a potential barrier to electronic conduction at thegrain boundary, as shown in Fig. 6.10.

From many studies, it was found that CNTs are embedded in the MOXlayer leading to the formation of pen heterojunctions. Therefore, there aretwo depletion layers to interact with gas, as shown in Fig. 6.10. Twodepletion layers are the region on the surface of the MOX and the interfacebetween the CNTs and MOX. The depletion layers at the pen heterojunc-tions can be modulated. The potential barriers at the interfaces or inside theMOX may be changed. This change of the depletion layer in the penheterojunctions of CNT/MOX was used to explain the enhanced responseof the film at low operating temperatures due to the amplification effects ofjunction structure combined with the gas reaction. Moreover, the formationof CNTs in the MOXmatrix can also introduce nanochannels. These nano-channels play an important role in gas diffusion. The gas molecules can easilytransport into the gas sensing layers, leading to increasing sensitivity.81,82

6.7 Fabrication of electrodes and CNT/polymernanocomposites for textile-based sensors

The fabrication of textile-based sensors is preceded by two processes:(1) preparation of textile-based electrodes and (2) preparation of CNT-polymer nanocomposite materials.

6.7.1 Preparation of textile-based electrodeTextile-based electrodes were fabricated using three different techniques,crocheting, embroidery, and screen printing technique, which will bedescribed in the following sections.

6.7.1.1 Crocheting techniqueThis technique was used to make thread-based electrodes. In the first step,100% cotton threads, size #40 thread (lustrous 3-ply) was used as theelectrodes.102 The thread-based electrode was designed to look like a


“dumbbell” comprising thread-based sensing materials connected with snappoppers metal electrodes at both ends. To increase the active surface of thesensor, a cotton thread was manually woven into a continuous crochet chainwith the length of approximately 10 mm by using crochet hooks. Snappoppers metals were used as electrodes for making electrical contact withan external circuit as shown in Fig. 6.11(a)e(c).

6.7.1.2 Embroidery techniqueThis technique is used to make fabric-based electrodes. In the first step,conductive thread was used as the interdigitated electrodes on embroideryon the fabric substrates along the lines to produce comblike patterns asshown in Fig. 6.12 The embroidered interdigitated electrodes were impor-tant components of the wearable sensor to function as sensing areas, whosesurface area was approximately 0.5 � 1.5 cm2 in case of fabric-basedembroidered electrode (pattern 1). Conductive thread and snap fastenerswere used to connect the chemical gas sensors with the external circuitport, in which external devices can be plugged into the active area to fulfillthe measurement.103

O2–

NO– NO–O– O–

NO2– SnO2

NO2–

O2–

O2–

O2–

NO2–

NO2–

NO2–NO2

–

Pot

entia

lSWCNTs

Depletion layer

In o

xidi

zing

gas

(NO

2)

In a

ir

d1

d2d4

d3

Ev

Ef

Ec

DistanceGrain boundary

Figure 6.10 Model of a potential barrier to electronic conduction at grain boundary forhybrid single-walled carbon nanotube (SWCNT)/SnO2 sensors.77 The dashed andcontinuous lines show the potential barriers of hybrid SWCNT/SnO2 in environmentof air and environment of oxidizing gas (i.e., NO2), respectively.


6.7.1.3 Screen printing techniqueThe screen printing process with conductive silver ink is commonly used inthe electronic industry, especially for complicated electronic circuitry andin the electronic components. In the past decade, several groups ofresearchers conducted experiments printing onto different kinds of substratessuch as paper, polyimide, fabric, and glass.104 Thereafter, researchers have

Figure 6.11 Fabrication process of thread-based electrodes using crocheting tech-nique. (a) and (b), cotton thread and snap poppers metal were used to make an elec-trode by manually weaving thread into a continuous crochet chain and (c) the size ofcomplete crochet electrode model has a length of about 10 mm.

Fabric-based embroidered electrode (Pattern 1)

Fabric-based embroidered electrode (Pattern 2)

Embroidery technique

5 mm

2 mm

1 mm

15 mm30

mm

50 mm

S1 S2 S3 S4

S5 S6 S7 S8

Figure 6.12 Fabrication process of fabric-based embroidered electrodes using embroi-dery technique.


developed electronic textiles by using screen printing techniques andconductive silver ink to create flexible printed circuits and electronic deviceson the textile substrate such as antennas,105 planar electrodes, and planarcircuits.106,107

Results of research can describe the methodologies of screen printing fordeveloping a fabric-based screen-printed electrode as fabric-based screen-printed gas sensors for smelling shirt system.108 The screen-printed electrodearray fabrication is conducted by the screen printing technique of whichprocesses are shown in Fig. 6.13 (Step 1e4). After completing screen print-ing with conductive silver ink, the fabric-based screen-printed electrode ar-rays were oven-dried at 120 �C for 15 min (Step 5). Finally, nine metalpoppers were attached on the screen-printed electrode arrays to serve as aconnector for making electrical connections.

6.7.2 Preparation of CNT/polymer nanocomposite sensingmaterials

The preparation of CNT/polymer nanocomposite gas sensing material forimmersion (dip)-coating and drop-coating process are described in detailas shown in Fig. 6.14. Starting with the polymer dissolving process, thefirst step was to dissolve each polymer (3 mg) completely in 1 mL of propersolvent. Then, different types of CNTs such as the MWCNTs, the

Fabric substrate

Fabric-based screen printed electrodes

Screen printing process

Screen printed electrode(before heated)Heated 120 °C (15 min)

S1 S2

45°S3

5

4

1 2 3

S4

S5S6

S75 mm

1 mm

10 mm

S8

1 2

3

4

56

7

8

Figure 6.13 Fabrication process of fabric-based screen-printed electrodes using screenprinting techniques.


carboxylic functionalized single-walled carbon nanotubes (SWCNTs-COOH), and the hydroxyl functionalized single-walled carbon nanotubes(SWCNTs-OH) was blended into each polymer solution to obtain a highconductivity with a percent loading for polymer:CNTs of 70:30 (step 2).Next, this solution was stirred for 30 min, followed by 30 min of continuousultrasonic vibration at room temperature 25 �C. This process (steps 3e4 inFig. 6.14) was repeated three times to ensure sample uniformity.

6.8 Sensor assembly for textile-based gas sensors

Textile-based gas sensors were fabricated using two differentmanufacturing techniques: immersion (dip)-coating technique and drop-coating technique.

6.8.1 Immersion-coating techniqueFor coating the crochet thread with the nanocomposite sensing materials,the solution of CNT/polymer was prepared according to section 11.7.2.Then the “dumbbell” was immersed for 3 h in the solution of coatingmaterial under controlled temperature (35 �C). Then it was kept at roomtemperature (about 25 �C) under constant stirring for 2 days. Finally, theresulted thread-based gas sensors were baked in the oven at 100 �C for3 h to remove any residual solvents. The immersion (dip-)-coating processand the appearance of the thread-based gas sensor are shown in Fig. 6.15.

1 2 3

5 4

Polymer solutions Polymer/CNTs solutions

Polymer/CNTs sensing materials

Stirred (30 min)

Sonicated (30 min)

Figure 6.14 Preparing sensing materials for drop-coating process.


6.8.2 Drop-coating techniqueThe CNT/polymer sensing solution mixtures, as prepared by the processdescribed in section 11.7.2, were drop-coated onto the fabric-based embroi-dered and screen-printed interdigitated electrodes. The electrical resistanceof fabric-based gas sensor was measured during the drop-coating process.It was found that the electrical resistance of each sensor was approximately1e20 kU. Then, the fabricated sensors were baked in an oven at a controlledtemperature of 80 �C for 1 h to eliminate any remaining solvents as shownin Fig. 6.16. Finally, the shape and features of the fabric-based gas sensorsfabricated by both techniques (embroidery and screen printing techniques)are shown in Fig. 6.17.

Step 1 Step 2 Step 3

Thread-based gas sensor

Immersion conditions Stirring conditions Bake conditions- Time (3 h)- Temperature (35 °C)

- Time (3 h)- Temperature (100 °C)

- Time (2 days)- Temperature (25 °C)

Figure 6.15 Immersion-coating technique for thread-based gas sensor fabrication.

S1 S2

45°

S3

S4

S5S6

S7

S8

1 2

3

4

56

7

8

10 mm

5 mm

1 m

m

Fabric-based embroidered electrode

Fabric-based screen printed electrode

Heated 80 °C (1 h)

Fabric-based embroideredgas sensors

Fabric-based screenprinted gas sensors

Figure 6.16 Drop-coating technique for fabric-based gas sensor fabrication.


6.9 Characterization of CNT/polymernanocomposites sensing materials on textilesubstrate

6.9.1 Scanning electron microscopy6.9.1.1 Fabric-based embroidered gas sensorsSEM was performed to investigate the microstructure of the fabric-basedembroidered chemical gas sensors. Fig. 6.18 (a) and (b) show the surfaceof the conductive thread (functioning as interdigitated electrodes) embroi-dered on the cotton satin fabric substrate at a magnification of 30� and100�, respectively.

Fabric-based embroideredgas sensors

Fabric-based screen printedgas sensors

(a) (b)

Figure 6.17 Photographs of the fabric-based gas sensors (a) fabricated by embroiderytechnique and (b) by screen printing technique.

Figure 6.18 Scanning electron microscopy pictures of (a) the surface of the conductivethread (functioning as interdigitate electrodes) embroidered on the cotton fabric sub-strate at a magnification of 30� and (b) the surface of the cotton fabric substrate at amagnification of 100�.63


In Fig. 6.19 (a) and (b), the cross section of the conductive thread andthe cotton satin fabrics as coated by a thick film of the CNT/polymernanocomposite materials are displayed at 300� and 600� magnification.It can be seen that the nanocomposite materials have penetrated into thefabric and coated around individual fibers within the thread throughoutthe full thickness of the cotton satin fabrics. The rough and porous natureof the fabric surface helps to increase the percolation of the analytegases into the sensing materials, thereby enhancing the sensing responseto specific gases.

6.9.1.2 Fabric-based screen-printed gas sensorsIn Fig. 6.20(a)e(d), the microstructure of the fabric-based screen-printed gassensors as investigated using SEM is presented: (a) shows at a magnificationof �50 the conductive silver ink confirming that the ink was well dispersedon the cotton fabric surface to create the patterned film electrodes; (b) thecross section of the thick film of conductive silver ink covered on the surfaceof cotton fabric at 150� magnification; (c) the screen-printed electrode sur-face as coated by a thick film of the CNT/polymer nanocomposite materialsat 40�magnification; and (d) the morphology at 600�magnification of thesensing materials infiltrating into the space between the fibers creating goodadhesion at the fiberematrix interface.

Figure 6.19 Scanning electron microscopy pictures of (a) the cross section of theconductive thread and the cotton fabrics as coated by a thick film of the CNT/polymernanocomposites materials 300� and (b) the cross section at 600� magnification.63


The adhesion force between molecules, between the cotton fabricsubstrate, and nanosized particles of silver in the conductive ink andeven the CNT/polymer nanocomposite networks can be explained bythe texture and surface morphology effects on adhesion stabilization ingas sensing films deposited by drop-coating process. Hence, the cottonfabric substrate is a porous structure; the conductive silver ink and the

Figure 6.20 Scanning electron microscopy micrographs of the fabric-based screen-printed gas sensors. (a) the conductive silver ink covering the fabric surface withoutany crack and creating the patterned of electrodes at 50 time magnifications, (b) thecross-section of the thickness around 100 micrometer of conductive silver ink coveringon the fabric surface at 150 time magnifications, (c) the cotton fabric and conductivesilver ink coated by sensing materials at 40 time magnifications and (d) the morphologyof the sensing materials spreading through the electrode surface at 600 timemagnifications.


CNT/polymer nanocomposite slurry percolated through the interfiberpores by capillary force. Therefore, the roughness and porous structureon the sensing areas are important to the efficacy of the sensor responseto volatile gases. In addition, the chemical properties of CNT/polymernanocomposite sensing materials and cotton fabric substrate adhere morestrongly because of the composition of cotton fabric substrate which ismainly of cellulose. The cellulose fiber and the CNT/polymer nanocom-posite networks have increased adhesive interactions between moleculeforces by van der Waals forces.109

6.10 Sensing mechanism of CNT/polymernanocomposites sensing materials on fabric substrate

Fig. 6.21 shows the microstructure of the fabric-based gas sensors andthe mechanism of sensing in the materials coated on the fiber surface. SEMimage at 600� magnification of the gas sensor sheet which was a permeableporous elastic fabric whose surface is covered with a CNT/polymer sensingfilm is shown. Certain nanocomposite film materials used as sensitive snifferinfiltrated into the space between the fibers and created a good adhesion onthe fiberematrix interface. In this case, the sensing capability of the sensor

Gas molecules

Polymer/CNTs compositematerials coated on fabric fibers

Polymer swelling due to chemical sorptionof gas/liquid molecules

CNT PolymerRf > Ri

Reference resistance (Ri) Resistance gas exposure (Rf)

Desorption/recovery

Res

ista

nce

Xs(0)

Baseline

Reference gas Odorant Odorant off

Time

Sensor response Xs(t)

Figure 6.21 Mechanism of carbon nanotube (CNT)/polymer sensing materials coatedon the fiber surface.


was not only affected by polymer swelling phenomena but also enhanced bythe increasing surface of the fabric. Gas molecules can percolate both intomicroporous structure of the polymer which was coated on the surfaceand within the fine structure of the fabric. The sensing mechanism ofCNT/polymer gas sensors can be described mainly by the polymer swellingbehavior resulting in the change of the electronic pathway on CNTsnetwork.110 Moreover, another sensing principle that may possibly causethe change in the electronic property of the sensors is the electron transfer-ring capability between gas molecules and CNTs.

6.11 Conclusion

The unique structure and electronic properties of CNTs provide atremendous potential for construction of not only CNTs and MOX hybridmaterials but also textile sensors in the field of gas sensing applications.Advantages for mixing CNTs in MOXs for gas sensors are the reductionof operating temperature and enhancement of sensitivity and selectivitydue to the amplification effects of pen heterojunctions with the gasreaction, formation of nanochannels for gas diffusion, high specific surfacearea, and increase of charge carriers on the surface. As a result of theseadvantages, the hybrid CNT/MOX gas sensor may be used instead ofthe popular commercial MOX gas sensors (such as TGS gas sensors) inthe near future. Moreover, CNT/polymer nanocomposites also selectedto use as gas sensing materials in the field of textile sensor and wearabletechnology as it can operate at room temperature and low energyconsumption during operation which is suitable for wearable device.The integration of textiles and electronics for wearable technology hascaused the wide variety of smart textile innovations. However, innovativee-textile that exists today is not found function for molecular detectionusing nanomaterial and nanohybrid materials. Therefore, establishing asniffing e-textile innovation is the biggest challenge to overcome somerestrictions imposed by the basic function of clothing. In addition, it isvery important to design and develop sniffing e-textile which has to bemore comfortable, flexible, bendable, and washable. Moreover, textilegas sensor innovation can not only be easily adapted to biological functionthat actually happens but is also fashionable and esthetically acceptable,which is a great opportunity for CNT research and development in termsof the textile gas sensor technology in the future.


AcknowledgmentsFinancial support from the Thailand Research Fund/Mahidol University to T.S. through theRoyal Golden Jubilee Ph.D. Program (Grant No. PHD/0178/2554) is acknowledged. T.K.acknowledges Mahidol University and National Nanotechnology Center (NANOTEC).C.W. acknowledges Kasetsart University Research and Development Institute (KURDI).

References1. Granitto PM, Biasioli F, Endrizzi I, Gasperi F. Discriminant models based on sensory

evaluations: single assessors versus panel average. Food Qual Prefer 2008;19:589e95.2. Fitzgerald G, James KJ, MacNamara K, Stack MA. ‘Characterisation of whiskeys using

solid-phase microextraction with gas chromatographyemass spectrometry’.J Chromatogr A 2000;896:351e9.

3. Vinaixa M, Vergara A, Duran C, Llobet E, Badia C, Brezmes J, Vilanova X,Correig X. Fast detection of rancidity in potato crisps using e-noses based on massspectrometry or gas sensors. Sensor Actuator B 2005;106:67e75.

4. Shurmer HV, Gardner JW, Corcoran P. Intelligent vapour discrimination using acomposite 12-element sensor array. Sensor Actuator B 1990;1:256e60.

5. Shurmer HV, Gardner JW. Odour discrimination with an electronic nose. SensorActuator B 1992;8:1e11.

6. Tikk K, Haugen J-E, Andersen HJ, Aaslyng MD. ‘Monitoring of warmed-overflavour in pork using the electronic nose e correlation to sensory attributes andsecondary lipid oxidation products’. Meat Sci 2008;80:1254e63.

7. Santonico M, Pittia P, Pennazza G, Martinelli E, Bernabei M, Paolesse R, D’Amico A,Compagnone D, Natale CD. Study of the aroma of artificially flavoured custards bychemical sensor array fingerprinting. Sensor Actuator B 2008;133:345e51.

8. Ragazzo-Sanchez JA, Chalier P, Chevalier D, Calderon-Santoyo M, Ghommidh C.Identification of different alcoholic beverages by electronic nose coupled to GC. SensorActuator B 2008;134:43e8.

9. Yu H, Wang J. Discrimination of LongJing green-tea grade by electronic nose. SensorActuator B 2007;122:134e40.

10. Kuske M, Romain AC, Nicolas J. Microbial volatile organic compounds as indicatorsof fungi. Can an electronic nose detect fungi in indoor environments? Build Environ2005;40:824e31.

11. Zhang S, Xie C, Zeng D, Zhang Q, Li H, Bi Z. A feature extraction method and asampling system for fast recognition of flammable liquids with a portable E-nose.Sensor Actuator B 2007;124:437e43.

12. James D, Scott SM, Ali Z, O’Hare WT. Chemical sensors for electronic nose systems.Microchim Acta 2005;149:1e17.

13. Choopun S, Hongsith N, Mangkorntong P, Mangkorntong N. Zinc oxide nanobeltsby RF sputtering for ethanol sensor. Physica E 2007;39:53e6.

14. Sriyudthsak M, Supothina S. Humidity-insensitive and low oxygen dependencetungsten oxide gas sensors. Sensor Actuator B 2006;113:265e71.

15. Anukunprasert T, Saiwan C, Traversa E. ‘The development of gas sensor for carbonmonoxide monitoring using nanostructure of NbeTiO2’. Sci Technol Adv Mater2005;6:359e63.

16. Watcharaphalakorn S, Ruangchuay L, Chotpattananont D, Sirivat A, Schwank J.Polyaniline/polyimide blends as gas sensors and electrical conductivity response toCO-N2 mixtures. Poly Inter 2005;54:1126e33.

17. Korotcenkov G. Metal oxides for solid-state gas sensors: what determines our choice?Mater Sci Eng B 2007;139:1e23.


18. Korotcenkov G. Gas response control through structural and chemical modification ofmetal oxide films: state of the art and approaches. Sensor Actuator B 2005;107:209e32.

19. Surnev S, Ramsey MG, Netzer FP. Vanadium oxide surface studies. Prog Surf Sci 2003;73:117e65.

20. Batzill M, Diebold U. The surface and materials science of tin oxide. Prog Surf Sci 2005;79:47e154.

21. Brinzari V, Korotcenkov G, Schwank J, Lantto V, Saukko S, Golovanov V. Morpho-logical rank of nano-scale tin dioxide films deposited by spray pyrolysis fromSnCl4•5H2O water solution. Thin Solid Films 2002;408:51e8.

22. Korotcenkov G, Cornet A, Rossinyol E, Arbiol J, Brinzari V, Blinov Y. ‘Facetingcharacterization of SnO2 nanocrystals deposited by spray pyrolysis from SnCl4e5H2Owater solution’. Thin Solid Films 2004;471:310e9.

23. Korotcenkov G, Macsanov V, Tolstoy V, Brinzari V, Schwank J, Faglia G. Structuraland gas response characterization of nano-size SnO2 films deposited by SILD method.Sensor Actuator B 2003;96:602e9.

24. Szezuka D, Werner J, Oswald S, Behr G, Wetzing K. XPS investigations of surfacesegregation of doping elements in SnO2. Appl Surf Sci 2001;179:301e6.

25. Kawamura F, Takahashi T, Yasui I, Sunagawa I. Impurity effect on [1 1 1] and [1 1 0]directions of growing SnO2 single crystals in SnO2eCu2O flux system. J Cryst Growth2001;233:259e68.

26. Judeinstein P, Sanchez C. ‘Hybrid organiceinorganic materials: a land ofmulti-disciplinarity’. J Mater Chem 1996;6:511e25.

27. Matsubara I, Hosono K, Murayama N, Shin W, Izu N. Organically hybridized SnO2gas sensors. Sensor Actuator B 2005;108:143e7.

28. Wisitsoraat A, Tuantranont A, Thanachayanont C, Patthanasettakul V, Singjai P.Electron beam evaporated carbon nanotube dispersed SnO2 thin film gas sensor.J Electroceramics 2006;17:45e9.

29. Eder D. Carbon nanotube e inorganic hybrids. Chem Rev 2010;110:1348e85.30. Wang C, Yin L, Zhang L, Xiang D, Gao R. Metal oxide gas sensors: sensitivity and

influencing factors. Sensors 2010;10:2088e106.31. Schierbaum KD, Weimar U, G€opel W, Kowalkowski R. Conductance, work func-

tion and catalytic activity of SnO2-based gas sensors. Sensor Actuator B 1991;3:205e14.32. ZhangW-D, Xu B, Jiang L-C. Functional hybrid materials based on carbon nanotubes

and metal oxides. J Mater Chem 2010;20:6383e91.33. Goldoni A, Petaccia L, Lizzit S, Larciprete R. Sensing gases with carbon nanotubes: a

review of the actual situation. J Phys Condens Matter 2010;22:013001.34. Khanderi J, Hoffmann RC, Gurlo A, Schneider JJ. Synthesis and sensoric response of

ZnO decorated carbon nanotubes. J Mater Chem 2009;19:5039e46.35. Moriguchi I, Hidaka R, Yamada H, Kudo T, Murakami H, Nakashima N.

A mesoporous nanocomposite of TiO. and carbon nanotubes as a high-rateLi-intercalation electrode material. Adv Mater 2006;18:69e73.

36. Whitsitt EA, Moore VC, Smalley RE, Barron AR. LPD silica coating of individualsingle walled carbon nanotubes. J Mater Chem 2005;15:4678e87.

37. Chen Y, Zhu C, Wang T. The enhanced ethanol sensing properties of multi-walledcarbon nanotubes/SnO2 core/shell nanostructures. Nanotechnology 2006;17:3012e7.

38. Chen J, Zhang W-D, Ye J-S. Nonenzymatic electrochemical glucose sensor based onMnO2/MWNTs nanocomposite. Electrochem Commun 2008;9:1268e71.

39. Zhao X, Johnston C, Grant PS. A novel hybrid supercapacitor with a carbon nanotubecathode and an iron oxide/carbon nanotube composite anode. J Mater Chem 2009;19:8755e60.

40. Haisong Q, Jianwen L, Edith M. Smart cellulose fibers coated with carbon nanotubenetworks. Fibers 2014;2:295e307.


41. Panida L, Enrico S, Natthapol W, Reinhard RB, Teerakiat K. A novel wearable elec-tronic nose for healthcare based on flexible printed chemical sensor array. Sensors 2014;14:19700e12.

42. Ni Luh WS, Brian Y. Reviewdthe development of gas sensor based on carbonnanotubes. J Electrochem Soc 2016;163:B97e106.

43. Timmer B, Olthuis W, van den Berg A. Ammonia sensors and their applicationsdareview. Sensor Actuator B Chem 2005;107:666e77.

44. Mittal M, Kumar A. Carbon nanotube (CNT) gas sensors for emissions from fossil fuelburning. Sensor Actuator B Chem 2014;203:349e62.

45. K€arkkanen A, Avarmaa T, Jaaniso R. “Gas sensing properties of SWCNT and TeflonAF composites,” abstracts of the sixth IEEE conference on sensors. IEEE Sensors 2007:279e80. Atlanta, 28e31 October 2007.

46. Paabo K, Avarmaa T, Jaaniso R, Kodu M, M€aeorg U, Saar A. Synthesis and gas-sensing properties of phenylhydrazine-functionalised single-wall carbon nanotubesin polymer matrix. Cent Eur J Chem 2013;11:945e52.

47. Lymberis A. Wearable smart systems: from technologies to integrated systems. ConfProc IEEE Eng Med Biol Soc 2011:3503e6.

48. Chan M, Est�eve D, Fourniols JY, Escriba C, Campo E. Smart wearable systems: cur-rent status and future challenges. Artif Intell Med 2012;56:137e56.

49. Young-Dong L, Wan-Young C. Wireless sensor network based wearable smart shirtfor ubiquitous health and activity monitoring. Sensor Actuator B Chem 2009;140:390e5.

50. Zhihua Z, Tao L, Guangyi L, Tong L, Yoshio I. Wearable sensor systems for infants.Sensors 2015;15:3721e49.

51. Perrier A, Vuillerme N, Luboz V, Bucki M, Cannard F, Diot B, Colin D, Rin D,Bourg J-P, Payan Y. Smart diabetic socks: embedded device for diabetic footprevention. IRBM 2014;35:72e6.

52. Lymberis A, Paradiso R. Smart fabrics and interactive textile enabling wearable per-sonal applications: R&D state of the art and future challenges. Conf Proc IEEE EngMed Biol Soc. 2008:5270e3.

53. Matteo S, Alessandro C. Wearable electronics and smart textiles: a critical review. Sen-sors 2014;14:11957e92.

54. Jing S, Chunghin C, Xiaoming T. Luminous fabric devices for wearable low-levellight therapy. Biomed Opt Express 2013;4:2925e37.

55. Cherenack K, Koen VO, Liesbeth VP. Smart photonic textiles begin to weave theirmagic. Laser Focus World 2012;48:63.

56. Van Os K, Cherenack K. Wearable textile-based phototherapy systems. Stud HealthTechnol Inform 2013;189:91e5.

57. Zhan L, Liwen J, Mariah DW, Xiangwu Z. Electrodeposited MnOx/carbon nano-fiber composites for use as anode materials in rechargeable lithium-ion batteries.J Power Sources 2010;195:5025e31.

58. Hertleer C, Rogier H, Member S, Vallozzi L, Langenhove LV. A textile antenna foroff-body communication integrated into protective clothing for firefighters. IEEETrans Adv Packag 2009;57:919e25.

59. Hertleer C, Rogier H, Vallozzi L, Declercq F. A textile antennas based on high-performance fabrics. In: Proceedings of 2nd European conference on antennas and propagation;Edinburgh, UK; 2007. p. 1e5.

60. Marianna G, Aleksandrs V, Inese P, Ausma V. Screen printed sensor for enuresis alarmsystem. Mater Sci Text Cloth Technol 2013;3:12e5.

61. Gordon P, Russel T, Steve B, John T. The development of screen printed conductivenetworks on textiles for biopotential monitoring applications. Sensor Actuator Phys2014;206:35e41.


62. Pandian PS, Mohanavelu K, Safeer KP, Kotresh TM, Shakunthala DT, Parvati G,Padaki VC. Smart Vest: wearable multi-parameter remote physiological monitoringsystem. Med Eng Phys 2008;30:466e77.

63. Thara S, Panida L, Teerakiat K. Development of fabric-based chemical gas sensors foruse as wearable electronic noses. Sensors 2015;15:1885e902.

64. Iijima S. Helical microtubules of graphitic carbon. Nature (London) 1991;354:56e8.65. Iijima S, Ichihashi T. Single-shell carbon nanotubes of 1-nm diameter.Nature (London)

1993;363:603e5.66. Bethune DS, Kiang CH, De Vries MS, Gorman G, Savoy R, Vazquez J, Beyers R.

Cobalt-catalysed growth of carbon nanotubes with single-atomic-layer walls. Nature(London) 1993;363:605e7.

67. Guo T, Nikolaev P, Thess A, Colbert DT, Smalley RE. Catalytic growth of single-walled nanotubes by laser vaporization. Chem Phys Lett 1995;243:49e54.

68. Thess A, Lee R, Nikolaev P, Dai H, Petit P, Robert J, Xu C, Lee YH, Kim SG,Rinzler AG, Colbert DT, Scuseria GE, Tom�anek D, Fischer JE, Smalley RE. Crystal-line ropes of metallic carbon nanotubes. Science 1996;273:483e7.

69. Endo M, Takeuchi K, Igarashi S, Kobori K, Shiraishi M, Kroto HW. The productionand structure of pyrolytic carbon nanotubes. J Phys Chem Solids 1993;54:1841e8.

70. Dai H, Rinzler AG, Nikolaev P, Thess A, Colbert DT, Smalley RE. Single-wall nano-tubes produced by metal-catalyzed disproportionation of carbon monoxide. ChemPhys Lett 1996;260:471e5.

71. Maeda Y, Kimura S, Kanda M, Hirashima Y, Hasegawa T, Wakahara T, Lian YF,Nakahodo T, Tsuchiya T, Akasaka T, Lu J, Zhang XW, Gao ZX, Yu YP,Nagase S, Kazaoui S, Minami N, Shimizu T, Tokumoto H, Saito RJ. Large-scale sep-aration of metallic and semiconducting single-walled carbon nanotubes. Am Chem Soc2005;127:10287e90.

72. Zheng M, Jagota A, Semke ED, Diner BA, McLean RS, Lustig SR, Richardson RE,Tassi NG. DNA-assisted dispersion and separation of carbon nanotubes. Nat Mater2003;2:338e42.

73. Ihara K, Endoh H, Saito T, Nihey F. Separation of metallic and semiconductingsingle-wall carbon nanotube solution by vertical electric field. J Phys Chem C 2011;115:22827e32.

74. Arnold MS, Green AA, Hulvat JF, Stupp SI, Hersam MC. Sorting carbon nanotubesby electronic structure using density differentiation. Nat Nanotechnol 2006;1:60e5.

75. Chen FM, Wang B, Chen Y, Li LJ. Toward the extraction of single species of single-walled carbon nanotubes using fluorene-based polymers. Nano Lett 2007;7:3013e7.

76. Cui J, Yang D, Zeng X, Zhou N, Liu H. Recent progress on the structure separationof single-wall carbon nanotubes. Nanotechnology 2017;28:452001e15.

77. Wei BY, Hsu MC, Su PG, Lin HM, Wu RJ, Lai HJ. A novel SnO2 gas sensor dopedwith carbon nanotubes operating at room temperature. Sensor Actuator B 2004;101:81e9.

78. Espinosa EH, Ionescu R, Chambon B, Bedis G, Sotter E, Bittencourt C, Felten A,Pireaux JJ, Correig X, Llobet E. Hybrid metal oxide and multiwall carbon nanotubefilms for low temperature gas sensing. Sensor Actuator B 2007;127:137e42.

79. Wang J, Liu L, Cong SY, Qi JQ, Xu BK. An enrichment method to detect low con-centration formaldehyde. Sensor Actuator B 2008;134:1010e5.

80. Hieu NV, Thuya LTB, Chien ND. Highly sensitive thin film NH3 gas sensor oper-ating at room temperature based on SnO2/MWCNTs composite. Sensor Actuator B2008;129:888e95.

81. Wongchoosuk C, Wisitsoraat A, Tuantranont A, Kerdcharoen T. Portable electronicnose based on carbon nanotube-SnO2 gas sensors and its application for detection ofmethanol contamination in whiskeys. Sensor Actuator B 2010;147:392e9.


82. Wongchoosuk C, Wisitsoraat A, Phokharatkul D, Tuantranont A, Kerdcharoen T.Multi-walled carbon nanotube-doped tungsten oxide thin films for hydrogen gassensing. Sensors 2010;10:7705e15.

83. Sedlackova K, Ionescu R, Balazsi C. TEM investigations on CNT added hexagonalWO3 films for sensing applications. Nano 2008;3:223e7.

84. Zhang H, Du N, Chen B, Li D, Yang D. Carbon nanotube-ZnO nanosphere heter-ostructures: low-temperature chemical reaction synthesis, photoluminescence, andtheir application for room temperature NH3 gas sensor. Sci Adv Mater 2009;1:13e7.

85. Sanchez M, Guirado R, Rincon ME. ‘Multiwalled carbon nanotubes embedded insolegel derived TiO2 matrices and their use as room temperature gas sensors’.J Mater Sci Mater Electron 2007;18:1131e6.

86. S�anchez M, Rinc�on ME. ‘Sensor response of solegel multiwalled carbon nanotubes-TiO2 composites deposited by screen-printing and dip-coating techniques’. SensorActuator B 2009;140:17e23.

87. Llobet E, Espinosa EH, Sotter E, Ionescu R, Vilanova X, Torres J, Felten A, Pireaux JJ,Ke X, Tendeloo GV, Renaux F, Paint Y, Hecq M, Bittencourt C. ‘Carbon nanotubeeTiO2 hybrid films for detecting traces of O2’. Nanotechnology 2008;19:375501.

88. Li W, Jung H, Hoa ND, Kim D, Hong S-K, Kim H. Nanocomposite of cobalt oxidenanocrystals and single-walled carbon nanotubes for a gas sensor application. SensorActuator B 2010;150:160e6.

89. Stern KH. Metallurgical and ceramic protective coatings. Chapman & Hall, Springer Pub-lisher; 1996. ISBN-10: 041-254-4407; ISBN-13: 978-041-254-440-8.

90. Hellstrom SL. Basic models of Spin coating. Lecture note, Stanford University; 2007.91. 91a Bornside DE, Macosko CW, Scriven LE. Spin coating of a PMMA/chloroben-

zene solution. J Electrochem Soc 1991;138:317e20.91b Bornside DE, Macosko CW,Scriven LE. Spin coating: one-dimensional model. J Appl Phys 1989;66:5185.

92. Zhang L, Fang X, Ye C. Controlled growth of nanomaterials. World Scientific Publishing;2007. ISBN-10: 981-256-7283; ISBN-13: 978-981-256-728-4.

93. Burgelman M. Thin film solar cells by screen printing technology. Proc WorkshopMicrotech Therm Prob Electron 1998:129e35.

94. Fox IJ, Bohan MFJ, Claypole TC, Gethin DT. Film thickness prediction in halftonescreen-printing. J P Mech Eng 2003;217:345e9.

95. Landau LD, Levich BG. Acta physiochim, vol. 17. U R S S; 1942. p. 42e54.96. Schmidt H, Mennig M. Wet coating technologies for glass. Tutorial: Institut f€ur Neue

Materialien; 2000.97. Mishra VN, Agarwal RP. Effect of electrode material on sensor response. Sensor Actu-

ator B 1994;22:121e5.98. Tamaki J, Miyaji A, Makinodan J, Ogura S, Konishi S. Effect of micro-gap electrode

on detection of dilute NO2 using WO3 thin film microsensors. Sensor Actuator B 2005;108:202e6.

99. Jain U, Harkera AH, Stoneham AM, Williams DE. Effect of electrode geometry onsensor response. Sensor Actuator B 1990;2:111e4.

100. 100a Connolly JR. Introduction quantitative X-ray diffraction methods. Spring2010.100b Chung FH. Quantitative interpretation of X-ray diffraction patterns. I.Matrix-flushing method of quantitative multicomponent analysis. J Appl Crystallogr1974;7:519e25.

101. Cheong HW, Lee MJ. Sensing characteristics and surface reaction mechanism ofalcohol sensors based on doped SnO2. J Ceram Process Res 2006;7:183e91.

102. Thara S, Panida L, Sasiprapa S, Chayanin K, Teerakiat K. A novel creation of thread-based ammonia gas sensors for wearable wireless security system. In: The 2014 11th in-ternational conference on electrical engineering/electronics, computer, telecommunications and in-formation technology. ECTI-CON; 2014. p. 1e4.


103. Thara S, Panida L, Teerakiat K. Wearable electronic nose based on embroideredamine sensors on the fabric substrates. In: The 2012 9th international conference on electricalengineering/electronics, computer, telecommunications and information technology.ECTI-CON; 2012. p. 1e4.

104. Saleem K, Leandro L, Ravinder SD. Technologies for printing sensors and electronicsover large flexible substrates: a review. IEEE Sens J 2015;15:1e4.

105. Tomas B, Richard L, Jan R. Screen printed antennas on textile substrate. In: Electronicssystem-integration technology conference (ESTC). vols. 1e4; 2014.

106. Hyejung K, Yongsang K, Binhee K, Hoi-Jun Y. A wearable fabric computer byplanar-fashionable circuit board technique. In: Conference: sixth international workshopon wearable and implantable body sensor networks. Berkeley, CA, USA: BSN; 2009.

107. Kai Y, Chris F, Russel T, Steve B, John T. Screen printed fabric electrode array forwearable functional electrical stimulation. Sensor Actuator 2014;213:108e15.

108. Thara S, Panida L, Sasiprapa S, Teerakiat K. On-cloth wearable E-nose for monitoringand discrimination of body odor signature. In: The 2014 IEEE ninth international con-ference on intelligent sensors, sensor networks and information processing. ISSNIP; 2014.p. 1e5.

109. Randjbaran E, Zahari R, Majid DL, Sultan TH, Mazlan N. ‘Reasons of adding carbonnanotubes into composite systems ereview paper’. Mech Mech Eng 2017;21:549e68.

110. Lorwongtragool P, Wisitsoraat A, Kerdcharoen T. An electronic nose for aminedetection based on polymer/SWNT-COOH nanocomposite. J Nanosci Nanotechnol2011;11:10454e9.


CHAPTER SEVEN

Carbon nanomaterialsfunctionalized with macrocycliccompounds for sensing vaporsof aromatic VOCsPierrick Clément1, Eduard Llobet21Microsystems Laboratory, �Ecole Polytechnique Féderale de Lausanne (EPFL), Lausanne, Switzerland2MINOS-EMaS, Department of Electronic Engineering, Universitat Rovira i Virgili, Tarragona, Spain

Contents

7.1 Introduction 2237.2 Cyclodextrins 2267.3 Calixarenes and derivatives 2297.4 Deep cavitands 2307.5 Conclusions 232Acknowledgments 235References 235

7.1 Introduction

BTEX is a term describing a set of chemicals closely related tobenzene. This set consists of benzene itself, toluene (i.e., methyl benzene),ethylbenzene, and xylenes. BTEX compounds are aromatic volatile organiccompounds (VOCs). They are colorless, sweet-smelling liquids undernormal temperature and pressure. However, their moderate to high vaporpressures imply that they can evaporate easily.

There are natural sources of BTEX compounds. For example, theseappear in gas emissions from volcanoes and forest fires, are present in crudeoil, and can be found near natural gas and petroleum deposits. However, themain emissions of BTEX into environment are of anthropogenic nature.Primary releases of these compounds occur through emissions from combus-tion engines, mostly from vehicles and also from aircraft and petroleum cokeovens. Incidentally, BTEX compounds are also found in the smoke of



https://doi.org/10.1016/B978-0-08-102559-8.00007-0

cigarettes. Petrochemical industry is one of the main emission sources andBTEX compounds are among the most abundantly produced chemicalsin the world. For instance, these compounds are created or used duringthe processing of petroleum products and the manufacturing of many chem-ical products such as paints, lacquers, thinners, solvents, adhesives, or inks.The manufacturing of rubber and plastics, cosmetics, and pharmaceuticalproducts is also a source of BTEX.

Although BTEX can be briefly bound to soils and sediments or bespilled in sea water, most releases eventually end up in the atmosphere(e.g., in land reclamation), where they may react with other pollutantsand contribute to the formation of photochemical smog. As any otherVOCs, BTEX compounds also play a role in the formation of ground-level ozone which can damage crops and exacerbate respiratory conditionsin humans (e.g., asthma).

The most common human exposure to BTEX compounds results fromcontaminated air breathing, particularly in areas of heavy motor vehicletraffic, petrol stations, motor vehicle repair stations, roadside works, andthrough cigarette smoke. Exposure to BTEX at normal environmentalconcentrations, and even to higher concentrations over a short period oftime, is unlikely to cause significant health damage. However, long-termexposure to higher concentrations (usually only experienced in occupa-tional settings) can be toxic to the liver, kidneys, central nervous system,and eyes. Table 7.1 shows exposure levels for BTEX compoundsestablished by the US safety and health administrations.

Table 7.1 Exposure levels to BTEX compounds as indicated by OSHAa and NIOSHb,January 2018.Compound TWAc (ppm) STELd (ppm) IDLHe (ppm)

Benzene 1 5 500Toluene 100 150 500Ethylbenzene 100 125 800Xylenes 100 150 900aOSHA: Occupational Safety and Health Administration (USA).bNIOSH: The National Institute for Occupational Safety and Health (USA).cTWA (time-weighted average): Employer shall assure that no employee is exposed to an airborneconcentration of the pollutant in excess of the TWA value as an 8-hour TWA.dSTEL (short-term exposure limit): The employer shall assure that no employee is exposed to anairborne concentration of the pollutant in excess of the STEL value as averaged over any 15-minperiod.eIDLH: Immediately dangerous to life and health exposure level.

224 Pierrick Clément and Eduard Llobet

BTEX compounds feature similar structures but quite different toxico-logical properties. Indeed, benzene is listed among the most harmfulVOCs because it is recognized as a human carcinogen by the US Environ-mental Protection Agency and by the European Commission.1,2 Long-termexposures to relatively low concentrations of benzene over months or yearslead to severe hemotoxic effects such as aplastic anemia and pancytopeniaand to acute nonlymphocytic leukemia.3e7 According to the Directive2008/50/EC of the European Parliament and of the Council of May2008, the limit value for the annual average exposure to benzene is5 mg m�3 (1.6 ppb).8

Nowadays, several methods for detecting traces of BTEX in air are inuse. Most of them involve pumping of the sample and subsequent analysisby employing colorimetric detector tubes or gas chromatography(GC-FID, GC-MS). These methods are bulky, expensive, and do notallow implementation for a continuous monitoring of BTEX traces. Inthe last few years, preconcentration methods and GC equipment havebeen improved in terms of miniaturization and with a limit of detection(LOD) reaching the ppb level for benzene.9e11 However, such systemsare still limited by their long response time, high power consumption,and high cost. Alternatively, the use of portable photoionization detector(PID) has been reported as well, but PID devices are not selective andgive a total reading for VOCs. The only option to make PID more selec-tive for BTEX in general and benzene in particular is to utilize a single-use,disposable, and rather expensive filter at the inlet port of the device thatwould result in a dramatic cost increase of running benzene measurements.The industries in which normal activity may result in active exposure oftheir workers to BTEX compounds (specially to benzene) would clearlybenefit from affordable, portable, highly sensitive, and selective detectorsable to run continuous measurements.

During the last decades, a great effort has been done aiming to investi-gate different strategies to improve the selectivity of chemical sensors. Toreach that milestone, sensors have been equipped with bioreceptorsemploying specific antigeneantibody-type binding interactions (inspiredby nature) where the size, shape, and charge allow the selective detectionof biological target species such as proteins, bacteria, viruses, and DNA.Additionally, this new generation of sensors mainly operates in aqueousmedia where the energy cost for the molecular recognition is reducedbecause of the water molecules that temporally “occupy” the bioreceptorprior the recognition. For gaseous species, the recognition is different, and

Carbon nanomaterials functionalized with macrocyclic 225

nonspecific interactions dominate their affinity with the medium. Scienti-fic groups have further exploited the possibility to synthesize molecularreceptors that could eventually mimic the specificity of biological receptorsreproducing the concept of binding site complementarity and shape recog-nition. Macrocyclic compounds such as cyclodextrins (CDs), calixarenes,and cavitands have been widely employed because of their commonpresence of cavities with molecular dimensions, which can act as molecularreceptors.

In the last years, many research groups have reported the development ofsensors employing carbon nanomaterials such as carbon black, carbonnanofibres, carbon nanotubes, and graphene.12 These materials are veryattractive because they allow for developing simple, chemoresistive devicesoperating at low temperatures, even at room temperature. Although carbonnanomaterials are particularly sensitive to their local chemical environmentof the gas phase, their functionalization seems essential if the aim is toselectively detect a few target gases or vapors. Indeed, different approacheshave been reported for functionalizing carbon nanomaterials in view oftailoring their gas sensing properties. Most of these functionalizationstrategies consist of creating controlled defects, decorating the outersidewalls of nanofibres, nanotubes, or the surface of graphene with metalor metal oxide nanoparticles, grafting functional groups such as carbonyl,carboxyl, or amine groups or more complex molecules such as macrocycliccompounds.13 Here we will review the approach of employing macrocycliccompounds grafted to carbon nanomaterials for developing gas sensors andsensor systems, with special emphasis in the results achieved for selectivelydetecting BTEX compounds. In such an approach, carbon nanomaterialsplay the role of transducing element (able to collect and transport efficientlyelectronic charge) and the grafted macrocyclic compounds are the selectivemolecular receptors, i.e., implement a receptor function in the gas sensor.

7.2 Cyclodextrins

CDs aremacrocyclic oligosaccharideswhich contain a hydrophobic cav-ity presentinghydroxyl groups at both rims thatmake themwater-soluble.Themost common CDs consist of six, seven, or eight a-D-glucopyranose unitsconjoined through a-(1/4)-glycosidic bonds and are, respectively, nameda-, b-, g-CDs. They are, therefore, suitable to capture hydrophobic guestsin aqueous media, where numerous hosteguest complexes have beenreported. Nevertheless, in the solidegas interface, the selectivity is mainly


driven by London dispersion interactions, size, and shape fit. Selectivity can beincreased by modifying the chemical groups on both rims. Table 7.2 summa-rizes some applications of modified CDs in gas sensing.

Only few examples of carbon nanomaterials functionalized with CDshave been reported so far. Duarte and coworkers18 developed a conductivepolymer nanocomposite (CPC) chemoresistor based on linear and branchedpolyamides synthesized from bifunctional and heptafunctional b-CDmono-mers and (Z) octadec-9-enedioic-N-hydroxysuccinimide ester bearing amultiwalled carbon nanotube (MWCNT) conducting architecture. Thelatter sensor was formed through a spray deposition of the carbon nanotubesand the CD polymers (dispersed separately in an organic solvent) layer bylayer on interdigitated ceramic substrate. The same group has demonstratedthe ability of CPC-based gas sensor to reversibly detect polar and nonpolarVOCs with an expected LOD to lay in the low ppb range. Furthermore,polyamide synthesized from b-CD(NH2)2(OH)19 is shown to be selectivetoward propanol in nitrogen gas carrier. This happens because of the stronghydrophilic character that the 19 hydroxyl moieties offer to the compoundmaking it able to generate many hydrogen bonds with polar protic solvents.

Employing the same principle, Nag and coworkers19 developed aquantum resistive chemical vapor sensor based on an array of b-CD func-tionalized reduced graphene oxide (RGO). Pyrene adamantane was usedto noncovalently tether the CDs to the RGO by self-assembly. This inno-vative connection allows the pep stacking of pyrene with graphene inone end and the inclusion of adamantane in CD cavity in the other end,preserving the accessibility of the analytes to functional sites (route a).They also compared a parallel route (route b) by simply noncovalentlybinding perbenzylated CD with RGO (RGO@PBCD) (Fig. 7.1).

The CD-modified graphene was sprayed layer by layer onto interdigi-tated electrodes controlling the resistance of the device. The authorsdemonstrated the selective detection of benzene as low as 400 ppb with asignal-to-noise ratio of 88 with the RGO@PBCD without sensitivity tohumidity in nitrogen carrier gas.

Following the strategy of employing a sensor array rather than a singlesensor, Pi-Guey Su and coworkers20 designed an array of quartz crystalmicrobalance (QCM) sensors allowing the differentiation of NH3

(1000e5000 ppm), CO (1500e7500 ppm), and NO2 (10e50 ppm) fromtheir tertiary mixture. This discrimination was possible by treating the dataof the sensors by principal component analysis. Graphene oxide (GO),b-CD functionalized GO, and N-substituted pyrrole derivative-based films


Table 7.2 Examples of modified cyclodextrins (CDs) with their gas sensing properties.

Modified CD Transducer Selective to Interferent(s)Limit ofdetection References

2,6-Per-O-(t-butyl-dimethylsilyl)-a-CD

Quartz crystalmicrobalance

Benzene Methane, propane, butane,pentane, eethyne, ammonia,nitrobenzene,and toluene

0.088 mgdm�3 inair

[14]

Polyaniline-b-CD composite Resistive Toluene Benzene N/C [15]Potassium iodide and a-CD Optical Ozone Humidity Several ppb

in air[16]

g-CD and potassium ions Gassorption analyzer

Formaldehyde N/C N/C [17]

228Pierrick

Clém

entand

EduardLlobet

were used as sensitive layers for gas sensing. The b-CD was noncovalentlyattached to the GO through pep stacking by simply mixing them. Eachmaterial was deposited on the QCM by spin coating.

7.3 Calixarenes and derivatives

Calixarenes are similar to the CDs’ cyclic structure and have generatedinterest over the last century because of their easy and tunable synthesis asmacrocyclic molecular receptors. They are composed of phenolic unitsjoined at meta position through methylene bridges. Calix[n]arenes are pref-erably synthesized with n ¼ 4, 6, 8 (building blocks) units. Hence, theypossess variable cavity dimension with the possibility to functionalize theirupper and/or lower rim to tailor their affinity with a target guest moleculethrough different noncovalent interactions such as pep stacking, cation-pand CH-p interaction, and hydrogen bonding.

A review on calixarene-based materials for gas sensing applicationwritten by Chilin Zou and coworkers highlights the new development in

Pyrene-adamantane

OHO

OH

OH

O

O

OHCOOH

COOH

COOH

COOH

HOOC

Perbenzylated cyclodextrin

RGO@PBCDRGO@PYAD-CD

R

N2H4 @ 100°C24 h reflux

Figure 7.1 Two routes synthesis of functionalized cyclodextrin (CD) to reducedgraphene. Reproduced from Nag S, Duarte L, Bertrand E, Celton V, Castro M, Choudh-ary V, Guegan P, Feller JeF. Ultrasensitive QRS made by supramolecular assembly offunctionalized cyclodextrins and graphene for the detection of lung cancer VOC bio-markers. J Mater Chem B 2014;2:6571e9. © Royal Society of Chemistry, 2014, withpermission.


the field of monitoring and detection of hazardous gases.21 Table 7.3 showssome examples of gas sensing applications employing modified calixarenes.

Despite the numerous examples that can be found in the literatureabout the ability of calixarenes for trapping gas molecules with highaffinity, only one work on calixarene-functionalized carbon nanomaterialsfor gas sensing has been reported recently. Baysak and coworkers27 reportthe use of single-walled carbon nanotubes (SWCNTs), the sidewalls ofwhich were noncovalently functionalized with pyrene bearing calix[4]pyr-role. Sensors were implemented as chemoresistors by coating a filter paperwith calixarene-functionalized SWCNTs contacted with two planarelectrodes. Fast response and higher affinity for acetone (20e500 ppm)compared with other VOCs were reported. Nevertheless, recent reports,where the calixarene is attached to the carbon nanomaterial via pepstacking28e30 (including pyrene modification of the calixarene) or covalentbonding31,32 or incorporated in a composite,33 have demonstrated selectiverecognition of target analytes, but in aqueous media only.

7.4 Deep cavitands

Derived from resorcinarene scaffolds, deep cavitands have been widelystudied for their synthetic versatility and selective complexation with targetmolecules. Notably, deep cavitands can be designed by, respectively, tuningtheir bridging group connected to the phenolic moieties of the resorcinar-ene. As a result, it is possible to control the dimensions, shape, and bindinggroups of the formed cavity.

Cram and coworkers were the pioneers to study cavitands as potentialmolecular receptors via the hosteguest strategy.34 Dalcanale andcoworkers did a subsequent work by modifying the bridging group ofthe resorcinarene to monitor VOCs in air. They have recently publisheda review highlighting their progress.35 Briefly, in their last study, theyfound out that rigidifying the cavity of the quinoxaline cavitand (QxCav)introducing four ethylenedioxy bridges at the upper rim (EtQxBox)improves the interaction with aromatic guests compared to the conforma-tional mechanism of the QxCav.36

This subtle modification allows additional interaction with toluene,ethylbenzene, and xylene guest than with benzene because of the upperrim that is too far to interact with benzene. They further implementedthe EtQxBox as preconcentrator coupled to miniaturized PID, and by using


Table 7.3 Modified calixarenes with their gas sensing properties.Modified calixarene Transducer Selective to Interferent(s) Limit of detection References

5,11,17,23-Tetrakis(tert-butyl)-25-carboxymethoxy-26,27,28-tris(ethoxycarbonylmethoxy)calix[4]arene polymer

Optical(colorimetric)

NO2/N2O4 N/C N/C [22]

Calix[4]arenes derivatives Quartz crystalmicrobalance(QCM)

N/C Chloroform,benzene,toluene, andethanol

N/C [23]

25,27-(Dipropylmorpholinoacetamido)-26,28-dihydroxycalix[4]arene

QCM, surfaceplasmonresonance

N/C Dichloromethane,chloroform,benzene, andtoluene

1.48 ppm fordichloromethanein air

[24]

Calix[4]azacrown Luminescence Tetrahydrofuran Acetone, methanol,dichloromethane,ethyl acetate,cyclohexane,n-hexane,benzene,toluene, trifluoroaceticacid, and petroleumether

N/C [25]

Calix[4]arenesderivatives

QCM Methylenechloride

Acetone, acetonitrile,carbon tetrachloride,chloroform, N,N-dimethylformamide,1,4-dioxane, ethanol,ethyl acetate, dioxane,xylene, toluene,methanol,n-hexane, and water

54.1 ppmin air

[26]

a smart temperature program, benzene is selectively desorbed and its LODreaching the ppb level. This approach is illustrated in Fig. 7.2.

Recently, Llobet and coworkers studied the possibility to couple thepromising gas sensing properties of cavitands with MWCNTs as resistivegas sensors.37 They first grafted gold nanoparticles on oxygen plasmaetreatedMWCNTs where the thioether-legged QxCav is further tethered on gold bya self-assembled monolayer approach (QxCaveAu-MWCNTs). Thisfunctionalization process is illustrated in the upper part of Fig. 7.3.

On a sensing event, a charge transfer is observed between the cavitand andthe Au-MWCNTs changing the general conductivity of the system. Fig. 7.4illustrates several response and recovery cycles for the sensors to increasingbenzene concentrations in the ppb range. The sensor showed clearly highersensitivity for benzene than for other aromatic and nonaromatic VOCs,with an LOD of 600 ppt in dry air. Nevertheless, a nonnegligible cross-sensitivity with NO2 and ambient humidity was observed. However, thiscan be overcome, at least partially, by adding a sensor employing bareAu-MWCNTs, which are extremely sensitive to NO2 and poorly responsiveto benzene. The use of a filter at the inlet of the detector would helpremoving ambient moisture and thus the undesired cross-sensitivity effectof humidity.

7.5 Conclusions

The covalent or noncovalent functionalization of carbon nanomaterialswith macrocycles opens fascinating opportunities for advancing towardmolecular recognition in the gas phase. CDs, calixarenes, and deep cavitandshave been employed because they present cavities with molecular dimen-sions, which can act as molecular receptors. Nowadays, it is possible to finelycontrol the dimensions, shape, and binding groups of the formed cavity. Inother words, macrocycles are becoming engineered scaffolds that pointtoward mimicking the specificity of biological receptors, reproducing theconcept of binding site complementarity and shape recognition.

In the development of gas sensors employing macrocycles, two mainapproaches can be identified. On the one hand, macrocycle compoundsare employed to functionalize carbon nanomaterials in view of obtainingnew functional adsorbent materials, i.e., more efficient and with improvedselectivity. These adsorbents are then employed as coatings in gravimetrictransducers or in the miniaturized preconcentration units of gas detectors.On the other hand, macrocycle-functionalized carbon nanomaterials are


O O O

O O O O O O OO

O O OO O

N N N N N N N N

RRR REtQxBox

Sample in

PumpVa

lve

Air in

MEMS cartridge

EtQxBoxcavitand

mesh

MEMStemperature

controlelectronic

PIDacquisitionelectronic

PID70

60

50

40

30

10

20

0

5 6 7 8 9 10 11 12 13

Time (min)

PID

sig

nal (

mV

)

Filte

r

ReferanceB = 20 ppbT = 20 ppbB+T = 20 + 20 ppb

200

150

100

50

ME

MS

car

tridg

e te

mpe

ratu

re (°

C)

R=C6H13

(a) (b) (c)

Figure 7.2 Representation of EtQxBox cavitand structure (a) with a scheme of the benzene monitoring device (b) and typical responses ofthe photoionization detector (PID) to a temperature ramp (c). Reproduced from Khaled E, Khalil M, el Aziz GA. Calixarene/carbon nanotubesbased screen printed sensors for potentiometric determination of gentamicin sulphate in pharmaceutical preparations and spiked surface watersamples. Sensor Actuator B Chem 2017;244:876e84. © American Chemical Society, 2013, with permission.

Carbon

nanomaterials

functionalizedwith

macrocyclic

233

O-MWCNT Au-MWCNT N2⊂cav-Au-MWCNT

(a)

(b)

Figure 7.3 (a) Fabrication scheme of the QxCaveAu-MWCNTs hybrid nanomaterialwhere Au-RF sputtering is followed by the self-assembly of thioether-legged QxCavmonolayer on the Au-NP surface. (b) Schematic representation of the sensing eventof a benzene molecule. Adapted from Clément P, Korom S, Struzzi C, Parra EJ,Bittencourt C, Ballester P, Llobet E. Deep cavitand self-assembled on Au NPs-MWCNT ashighly sensitive benzene sensing interface. Adv Funct Mater 2015;25:4011e20. © JohnWiley and Sons, 2015, with permission.

651.75

651.50651.25651.00650.75650.50650.25

650.00649.75649.50

0 1000 2000 3000 4000 5000t(s)

2,5

ppb

Air

5 pp

b

Air Air Air Air

10 p

pb

20 p

pb

40 p

pb

R(Ω

)

Figure 7.4 Response and recovery cycles to successively increasing concentrations ofbenzene for a chemoresistive sensor employing a QxCaveAumultiwalled carbon nano-tubes hybrid nanomaterial. Adapted from Clément P, Korom S, Struzzi C, Parra EJ,Bittencourt C, Ballester P, Llobet E. Deep cavitand self-assembled on Au NPs-MWCNT ashighly sensitive benzene sensing interface. Adv Funct Mater 2015;25:4011e20. © JohnWiley and Sons, 2015, with permission.


also used in simple chemiresistive gas sensors in which the carbon nanoma-terial is merely an efficient means of conducting free charge carriers and thereceptor function is played by the macrocycles attached.

The techniques employed in the synthesis of macrocycles and in thefunctionalization of carbon nanomaterials are well-known and allow forimplementing solution processing of gas sensitive devices. This impliesthat such techniques are suitable for the mass production of both hybridnanomaterials and sensors at low production costs, allowing cost-effectivecommercialization.

Some of the reported hybrid nanomaterials show remarkable sensitivityand selectivity to aromatic VOCs and, in particular, quinoxaline-bridgedcavitand-functionalized MWCNT sensors show very high sensitivity towardlow levels of benzene in dry air (i.e., experimentally tested down to2.5 ppb), with a theoretical lower detection limit of 600 ppt. In addition,both detection and baseline recovery are run at room temperature, whichimplies that sensors operate at low power consumption.

However, sensors still suffer from cross-sensitivity issues, namely toambient moisture and to oxidizing species such as ozone or nitrogendioxide. Some solutions exist already to tackle such problems, such as usingfilters for trapping water at the inlet of the detector system or using an arrayof sensors with partial selectivity and chemometrics. However, thesesolutions are suboptimal because filters are cumbersome, may alter theprofile of VOCs, or become saturated, and using sensor arrays and patternrecognition adds complexity, makes calibration and recalibration moredifficult, and increases overall cost. Should these cross-sensitivity issuesbe ameliorated by further increasing the performance of functionalmaterials, macrocyclic compound functionalized carbon nanomaterialsmay soon be integrated in a new generation of inexpensive, handheldanalyzers or wearable detectors for BTEX compounds with potentialapplications in workplace safety or environment monitoring.

AcknowledgmentsE. L. is supported by the Catalan Institution for Research and Advanced Studies (ICREA), viathe 2018 Edition of the ICREA Academia Award.

References1. United States Environmental Protection Agency. Drinking water contaminants. 2009.

http://water.epa.gov/drink/contaminants/index.cfm#Organic.2. European Commission. Air quality standards. 2015. http://ec.europa.eu/environment/

air/quality/standards.htm.


http://water.epa.gov/drink/contaminants/index.cfm#Organic

http://ec.europa.eu/environment/air/quality/standards.htm

http://ec.europa.eu/environment/air/quality/standards.htm

3. Baan R, Grosse Y, Straif K, Secretan B, El Ghissassi F, Bouvard V, Benbrahim-Tallaa L,Guha N, Freeman C, Galichet L. A review of human carcinogensdpart F: chemicalagents and related occupations. Lancet Oncol 2009;10:1143e4.

4. Fass �O A, Arduino G. Environmental regulation in the European union. Encyclopedia ofEnvironmetrics; 2012.

5. Dougherty D, Garte S, Barchowsky A, Zmuda J, Taioli E. NQO1, MPO, CYP2E1,GSTT1 and GSTM1 polymorphisms and biological effects of benzene exposuredaliterature review. Toxicol Lett 2008;182:7e17.

6. Hester RE, Harrison RM. Volatile organic compounds in the atmosphere. Royal Society ofChemistry; 1995.

7. DeCaprio AP. The toxicology of hydroquinone-relevance to occupational and envi-ronmental exposure. CRC Crit Rev Toxicol 1999;29:283e330.

8. Official journal of the European union,50/EC of the European parliament and of the council of 21May 2008 on ambient air quality and cleaner air for Europe. 2008. http://ec.europa.eu/environment/air/quality/legislation/existing_leg.htm.

9. Sun J, Guan F, Cui D, Chen X, Zhang L, Chen J. An improved photoionizationdetector with a micro gas chromatography column for portable rapid gas chromatog-raphy system. Sensor Actuator B Chem 2013;188:513e8.

10. Liaud C, Nguyen N, Nasreddine R, Le Calvé S. Experimental performances study of atransportable GC-PID and two thermo-desorption based methods coupled to FID andMS detection to assess BTEX exposure at sub-ppb level in air. Talanta 2014;127:33e42.

11. Jian R-S, Huang Y-S, Lai S-L, Sung L-Y, Lu C-J. Compact instrumentation of a m-GCfor real time analysis of sub-ppb VOC mixtures. Microchem J 2013;108:161e7.

12. Llobet E. Gas sensors using carbon nanomaterials: a review. Sensor Actuator B Chem2013;179:32e45.

13. Wang F, Swager TM. Diverse chemiresistors based upon covalently modifiedmultiwalled carbon nanotubes. J Am Chem Soc 2011;133:11181e93.

14. Lai CSI, Moody GJ, Thomas JDR, Mulligan DC, Stoddart JF, Zarzycki R. Piezoelec-tric quartz crystal detection of benzene vapour using chemically modified cyclodextrins.J Chem Soc Perkin Trans 1988;2:319e24.

15. Subramanian E, Jeyarani BML, Murugan C, Padiyan DP. Crucial role of undoped/dopedstate of polyaniline‒b‒cyclodextrin composite materials in determining sensor functionality towardbenzene/toluene toxic vapor. 2016.

16. Izumi K, UtiyamaM,Maruo YY. A porous glass-based ozone sensing chip impregnatedwith potassium iodide and a-cyclodextrin. Sensor Actuator B Chem 2017;241:116e22.

17. Wang L, Liang X-Y, Chang Z-Y, Ding L-S, Zhang S, Li B-J. Effective formaldehydecapture by green cyclodextrin based metal-organic framework. ACS Appl MaterInterfaces 2018;10:42e6.

18. Duarte L, Nag S, Castro M, Zaborova E, Ménand M, Sollogoub M, Bennevault V,Feller J-F, Guégan P. Chemical sensors based on new polyamides biobased on(Z) octadec-9-enedioic acid and b-cyclodextrin. Macromol Chem Phys 2016;217:1620e8.

19. Nag S, Duarte L, Bertrand E, Celton V, Castro M, Choudhary V, Guegan P, Feller J-F.Ultrasensitive QRS made by supramolecular assembly of functionalized cyclodextrinsand graphene for the detection of lung cancer VOC biomarkers. J Mater Chem B2014;2:6571e9.

20. Su P-G, Chuang T-Y. Simple and rapid differentiation of toxic gases using a quartzcrystal microbalance sensor array coupled with principal component analysis. SensorActuator Phys 2017;263:1e7.

21. Kumar S, Chawla S, ZouMC. Calixarenes based materials for gas sensing applications: areview. J Inclusion Phenom Macrocycl Chem 2017;88:129e58.


http://ec.europa.eu/environment/air/quality/legislation/existing_leg.htm

http://ec.europa.eu/environment/air/quality/legislation/existing_leg.htm

22. Gusak AS, Ivanova EA, Prokhorova PE, Rusinov GL, Verbitskiy EV, Morzherin YY.Synthesis and use of polymer-immobilized calix[4]arene derivatives as molecularcontainers for nitrous gases. Russ Chem Bull 2014;63:1395e8.

23. Ozmen M, Ozbek Z, Buyukcelebi S, Bayrakci M, Ertul S, Ersoz M, Capan R.Fabrication of LangmuireBlodgett thin films of calix[4]arenes and their gas sensingproperties: investigation of upper rim para substituent effect. Sensor Actuator B Chem2014;190:502e11.

24. Acikbas Y, Bozkurt S, Halay E, Capan R, Guloglu ML, Sirit A, Erdogan M. Fabricationand characterization of calix[4] arene LangmuireBlodgett thin film for gas sensingapplications. J Inclusion Phenom Macrocycl Chem 2017;89:77e84.

25. Oueslati I, Paixao JA, Shkurenko A, Suwinska K, Seixas de Melo JS, Batista deCarvalho LAE. Highly ordered luminescent calix[4]azacrown films showing anemission response selective to volatile tetrahydrofuran. J Mater Chem C 2014;2:9012e20.

26. Temel F, Tabakci M. Calix[4]arene coated QCM sensors for detection of VOCemissions: methylene chloride sensing studies. Talanta 2016;153:221e7.

27. Baysak E, Yuvayapan S, Aydogan A, Hizal G. Calix[4]pyrrole-decorated carbonnanotubes on paper for sensing acetone vapor. Sensor Actuator B Chem 2018;258:484e91.

28. Sun Y, Mao X, Luo L, Tian D, Li H. Calix [4] arene triazole-linked pyrene: clicksynthesis, assembly on graphene oxide, and highly sensitive carbaryl sensing in serum.Org Biomol Chem 2015;13:9294e9.

29. Yang L, Ran X, Cai L, Li Y, Zhao H, Li C-P. Calix [8] arene functionalizedsingle-walled carbon nanohorns for dual-signalling electrochemical sensing of aconitinebased on competitive host-guest recognition. Biosens Bioelectron 2016;83:347e52.

30. Yang L, Xie X, Cai L, Ran X, Li Y, Yin T, Zhao H, Li C-P. p-sulfonated calix [8] arenefunctionalized graphene as a “turn on” fluorescent sensing platform for aconitinedetermination. Biosens Bioelectron 2016;82:146e54.

31. Dionisio M, Schnorr JM, Michaelis VK, Griffin RG, Swager TM, Dalcanale E.Cavitand-functionalized SWCNTs for N-methylammonium detection. J Am ChemSoc 2012;134:6540e3.

32. Adarakatti PS, Malingappa P. Amino-calixarene-modified graphitic carbon as a novelelectrochemical interface for simultaneous measurement of lead and cadmium ions atpicomolar level. J Solid State Electrochem 2016;20:3349e58.

33. Khaled E, Khalil M, el Aziz GA. Calixarene/carbon nanotubes based screen printedsensors for potentiometric determination of gentamicin sulphate in pharmaceuticalpreparations and spiked surface water samples. Sensor Actuator B Chem 2017;244:876e84.

34. Cram DJ, Karbach S, Kim HE, Knobler CB, Maverick EF, Ericson JL, Helgeson RC.Host-guest complexation. 46. Cavitands as open molecular vessels form solvates. J AmChem Soc 1988;110:2229e37.

35. Pinalli R, Pedrini A, Dalcanale E. Environmental gas sensing with cavitands.Chem Eur J2018;24:1010e9.

36. Trzci�nski JW, Pinalli R, Riboni N, Pedrini A, Bianchi F, Zampolli S, Elmi I, Massera C,Ugozzoli F, Dalcanale E. In search of the ultimate benzene sensor: theEtQxBox solution. ACS Sens 2017;2:590e8.

37. Clément P, Korom S, Struzzi C, Parra EJ, Bittencourt C, Ballester P, Llobet E. Deepcavitand self-assembled on Au NPs-MWCNT as highly sensitive benzene sensinginterface. Adv Funct Mater 2015;25:4011e20.



CHAPTER EIGHT

Luminescence probing of surfaceadsorption processes usingInGaN/GaN nanowireheterostructure arraysKonrad Maier1, Andreas Helwig1, Gerhard M€uller2, Martin Eickhoff31Airbus Group Innovations, Munich, Germany2Department of Applied Sciences and Mechatronics, Munich University of Applied Sciences,Munich, Germany3Institute of Solid State Physics, University of Bremen, Bremen, Germany

Contents

8.1 Adsorptiondkey to understanding semiconductor gas sensors 2398.2 III-nitrides as an emerging semiconductor technology 2438.3 Photoluminescent InGaN/GaN nanowire arrays 2438.4 Optical probing of adsorption processes 2458.5 Experimental observations of PL response 246

8.5.1 General response behavior 2468.5.2 Response to oxidizing gases 2468.5.3 Response to H2O vapor 2488.5.4 Response to reducing gases 249

8.6 Analysis of adsorption phenomena 2508.6.1 Concentration and temperature dependence of the PL response 2508.6.2 Competitive adsorption of air constituents 2538.6.3 Competition between quenching and enhancing H2O adsorbates 259

8.7 Molecular mechanism of adsorption 2618.8 Conclusions and outlook 266References 267

8.1 Adsorptiondkey to understandingsemiconductor gas sensors

Gas sensors based on metal oxides (MOXs) are a very widely studiedclass of sensors.1e4 The most commonly employed transduction mechanismis the detection of electrical resistance modulations of MOX materials as



https://doi.org/10.1016/B978-0-08-102559-8.00008-2

these are exposed to reactive gases while being heated to elevated temper-atures. Transduction mechanisms which convert a partial pressure of reactivegas, pgas, in the ambient air into an electrical output signal are twofold: thefirst effect is reductions in surface conductance that take place as stronglyoxidizing, i.e., electron withdrawing gases, adsorb on n-type oxides. Widelystudied examples are O2, NO2, and O3. The second important effect are en-hancements in surface conductance that take place as reducing gases adsorbon heated n-type oxides and as these interact with coadsorbed oxygen ionspecies. In these surface reactions, neutral combustion products such asH2O and CO2 are generated and electrons, initially trapped on oxygenion species, are returned to the semiconductor adsorbent. Examples ofreducing gases that follow this second line of detection are CO, H2, and ahuge range of hydrocarbons.

Experimental parameters that can be derived from resistance measure-ment and gas exposure tests are the relative resistance response Rres(pgas)and the gas sensitivity Sres(pgas):

Rres�pgas

� ¼ R0 � Rgas

R0; (8.1)

Sres�pgas

� ¼ dRres

dpgas. (8.2)

In these equations, pgas is the partial pressure of the reactive gas in the airambient,Rres the magnitude of the relative resistance response to the reactivegas, and Sres the corresponding gas sensitivity. R0, in turn, is the baselineresistance of the sensor under clean-air conditions and DRgas ¼ R0 � Rgas

the negative change in the sensor resistance under gas exposure.Although these processes are qualitatively well understood, a quantitative

analysis of those microscopic mechanisms that ultimately lead to the exper-imentally determined functions Rres(pgas) and Sres(pgas) has remained difficultand continues to be challenging as the electrical conductivity can depend incomplex and manifold ways on the material characteristics of the sensitivelayers. In the analysis of these functions, it has been customary to breakdown the experimentally observed sensitivity Sres(pgas) into a transducerand a receptor part5e7:

Sres�pgas

�¼ dRres

dpgas¼ dRres

dNOminus

dNOminus

dpgas

¼ TrðRres;NOminusÞRec�NOminus; pgas

�; (8.3)

240 Konrad Maier et al.

whereNOminus stands for the areal density of adsorbed oxygen ion species onthe sensor surface. Whereas the receptor part, Rec(NOminus, pgas), is normallyconsidered to follow simple mass-action laws or common adsorptionisotherms, the transducer part, Tr(Rres, NOminus), can take on different formsdepending on the crystal size, the stoichiometry, and the morphology of thesensing layers.5e7 Depending on the height and widths of the intergrainpotential barriers, electron transport across grain boundaries can proceedthrough thermal activation and/or tunneling steps. Furthermore, as crystalsizes often have dimensions in the range of nanometers, different types ofcarrier depletion can abound in the bulk crystal grains, ranging fromregionally depleted to critically and volume-depleted.5e7 In nanocrystallineand porous layers, finally, the electrical transport over macroscopic distancescan often take the form of percolation paths, which makes the interpretationof resistive response data even more complex.8,9

In recent years, another form of chemical response has been studied,which can be observed in MOX materials grown in nanowire form.10e12

This kind of response involves luminescence light induced by UV light sour-ces and detection of longer-wavelength visible light. The interesting aspect isthat reactive gases that adsorb at the nanowire surfaces may act as recombi-nation centers which reduce the luminescence intensity below its clean-airbaseline level. As this optically induced chemical sensitivity does not rely onthe presence of thermally activated charge carriers, the optical response canin general be observed at more moderate temperatures than the resistiveresponse of MOX materials. Even more important is that the generationand emission of luminescence light is a local phenomenon that does notdepend in a similarly complex way on the morphology and intergraintransport of photogenerated charge carriers as in resistively readout sensors.Measurements of the optical response of MOX semiconductors thereforehold promise to shed more light on those processes that are related to theadsorption of gases on MOX surfaces.

In this chapter, we report on gas detection experiments performed onarrays of ternary group III-nitride nanowires, namely InGaN nanowiresformed on GaN nanowire templates. Such InGaN/GaN heterostructurenanowire arrays (NWAs) are attractive for luminescence studies as thebandgap of InGaN alloys can be varied over a large range extending fromEg z 0.7 eV (InN) up to Eg z 3.5 eV (GaN) with the bandgap alwaysremaining direct.13 Because of this bandgap variability, InGaN/GaNNWAs can be grownwhich allow the PL excitation and the PL emission lightto be absorbed and emitted in well-separated spectral ranges. This spectral

Luminescence probing of surface adsorption processes using InGaN/GaN nanowire 241

separation allows comparatively simple equipment to be used which is ofconcern considering future gas sensing applications.14 Other aspects whichmake InGaN alloys interesting for fundamental investigations into gas sensingmechanisms are that InGaNmaterials retain their luminescence up to temper-atures extending well into the operation temperature range of resistivelyreadout MOX gas sensors. Most importantly, increasing evidence emergesthat InGaN surfaces carry native oxides,15e21 which make InGaN/GaNnanowires behave similar as more conventional MOX nanowires.

On the pages below, we will show that the PL response of InGaN/GaNNWAs, RPL(pgas, T), quite universally takes the form of products consistingof a temperature-dependent recombination term asat(T) and a Langmuiradsorption isotherm, qL(pgas, T):

RPL�pgas;Eads;T

�¼ PLgas � PL0PL0

¼ asatðTÞ qL�pgas;Eads;T

�. (8.4)

Whereas the first term depends on the physical parameters that controlthe radiative recombination processes in the volume of the InGaN/GaNnanowires and of the nonradiative ones at their surfaces, the second termdepends on those parameters that control the chemistry of the adsorptionprocesses at the transducer surface. These are the adsorbate partial pressurein the ambient air, pgas, the binding energy of the adsorbates on thetransducer surface, Eads, and the temperature T at which the adsorption takesplace.

In this chapter, our concern is on the function qL(pgas, Eads, T) and how itvaries under different conditions of gas exposure and sensor operationconditions. Making use of the excellent luminescence properties ofInGaN/GaN NWAs, we have been able to study the adsorption behaviorof several classes of gases on InGaN surfaces with native surface oxidesover wide concentration and temperature ranges. In this way, we haveobtained information on adsorbates and adsorbate-binding energies, whichhitherto could not be observed and discussed with clarity on other kinds ofgas-sensitive materials.14,22e27 Examples discussed here include the compet-itive adsorption of oxidizing air constituents on III-nitride surfaces and thesurprisingly complex adsorption behavior of H2O, which can form bothquenching and enhancing adsorbates on such surfaces, depending onexperimental conditions. We further provide evidence that enhancing wateradsorbates produced in surface oxidation reactions provide an indirectapproach to detect otherwise unreactive hydrocarbon species.


8.2 III-nitrides as an emerging semiconductortechnology

With the silicon semiconductor technology approaching its physicallimitations, the interest of the semiconductor community has turned towide-bandgap materials such as silicon carbide (SiC),28 diamond,29 andgallium nitride (GaN).30 Among these materials, GaN is outstanding becauseof its ability of forming alloys with indium (In) and aluminum (Al). Because ofthis capability, III-nitride materials form a continuous series of alloys withdirect bandgaps ranging from 0.7 up to 6.2 eV.13 With this potential athand, UV-LEDs and lasers have become available and III-nitride materialshave rapidly developed into a reliable materials base for the rapidly developingfield of solid-state lighting technology.31 Another attractive feature of theIII-nitride alloy system is the spontaneous and piezoelectric polarizationphenomena that can be observed in AlGaN/GaN heterostructures.32 Theselatter phenomena have been important for the realization of high-electronmobility transistors (HEMTs) which have become important building blocksin the fields of high-temperature and high-frequency electronics.33 Inaddition to these mainstream applications, AlGaN/GaN HEMTs have alsoreceived increasing attention in the field of chemical and biochemical sensors.Recent reviews of this work can be found in Ref. 34,35.

8.3 Photoluminescent InGaN/GaN nanowire arrays

While so far most of the work on III-nitride-based chemical sensors hasused conventional chemical-to-electrical transducer principles, very littlework has been reported yet which employs the excellent optoelectronicproperties of the III-nitride material system. This chapter attempts to closethis gap, reporting on experiments performed on III-nitride materials grownin the form of NWAs on silicon substrates. Whereas our previous work hasconcentrated on the growth, structural, and physical characterization ofsuch materials,23 our concern here is presenting an in-depth study of thosephotoluminescence (PL) changes that take place as NWAs are illuminatedby low-cost ultraviolet light sources while being simultaneously exposed todifferent chemical environments.14,22e27

Fig. 8.1 shows the investigated nanowire heterostructures, summarizing atthe same time information about their geometrical size, their crystallographicorientation, and their luminescence properties. The nanowires shown inFig. 8.1(a) were grown on (111) silicon substrates using plasma-assisted


molecular beam epitaxy with the growth proceeding along the (0001) axis.Substrate temperatures during the growth of GaN sections were 720�Cand 500�C during the InGaN sections. The nanowires typically consist of arelatively long GaN base (w500 nm), followed by a shorter InGaN section(w200 nm) and a thin GaN cap layer (w20 nm). The scanning electronmicroscopy (SEM) image in Fig. 8.1(b) shows that the self-assembled growthmode leads to irregular arrays of nanowires with hexagonal cross sections. Gasaccess to the lateral side walls is enabled by the relatively large wire-to-wiredistances which are in the same order of magnitude as the diameters of thenanowires. The cross-sectional SEM shows that the thickness of the nano-wires tends to increase toward the growth surface, which is partly causedby the lateral overgrowth of the previously deposited wire sections. Thelower-bandgap InGaN sections therefore are likely to carry a thin GaNcoating, which further tends to become covered by a thin layer of nativeoxide as the NWAs are exposed to the air ambient.15e21 Fig. 8.1(c) showsspectra of luminescence light as emitted in response to excitation light from

GaN top (20 nm)InGaN layer(200 nm)

GaN base(500 nm)

Si (111)substrate

(000

1)-

Side view Top view

300 nm

100 100

10 10

1 1

0.1 0.12.0 2.2 2.4 2.6 2.8 3.0 3.2

Energy (eV)

PL

inte

nsity

(a.u

.)

PL

max

imum

(a.u

.)

4 K

TNW

375 K

10 100 1000T (K)

λex=325 nm

(a)

(c)

(b)

(d)

Figure 8.1 (a) Heterojunction nanowires on Si substrates with axial stacking of GaN andInGaN sections, (b) cross-sectional and top view scanning electron microscopy (SEM)pictures of nanowire arrays grown using plasma-assisted molecular beam epitaxy; (c,d) spectral distribution of photoluminescence (PL) light and temperature dependenceof PL intensity.


a He-Cd laser at 325 nm and as observed at different NWA temperatures.Fig. 8.1(d), finally, shows the variation of PL intensity with temperature inthe range from 4K up to 375K.

8.4 Optical probing of adsorption processes

The measurement system for the characterization of the optochemicalresponse is displayed in Fig. 8.2(a) and the spectral characteristics of its opticalcomponents in Fig. 8.2(b). The light of a near-UV LED light source(l w 365 nm), used for luminescence excitation, is reflected onto theInGaN/GaN NWA placed on the ground plate and focused onto theNWA by a lens system. The same lens system captures the luminescencelight (green arrow in Fig. 8.2(a)) and focusses it onto the detector windowof a compact photomultiplier tube integrated into the top lid of the sensorsystem. Appropriate filters were used to separate the UV excitation and thegreen luminescence light. To allow measurements at different transducertemperatures, the NWA samples were mounted on a ceramic heatersubstrate carrying a screen-printed platinum meander on its backside. ThisPt heater meander simultaneously served as a Pt thermometer and as anindicator for the adjusted heater temperature. With this sample holder,maximum temperatures in the range of 150 �C could be reached. The totalvolume of the sensor chamber amounted to approximately 5 cm3. With thetotal gas flow rate of 500 sccm, this allowed for gas exchange times in theorder of one second.

Figure 8.2 (a) Experimental arrangement for measuring the photoluminescenceemission spectra of GaN/InGaN nanowire transducers under variable gas flows andat different nanowire array operation temperatures; (b) spectral characteristic of theoptical components indicated in (a). PMT, photomultiplier tube.


The PL response tests toward different gases were carried out using acustom-designed gas test rig with a set of mass flow controllers as shownin Fig. 8.3. All gas/vapor mixtures delivered from this test gas rig wereguided through a downstream mass flow controller into the measurementchamber to maintain a constant gas flow rate of 500 sccm independent ofgas composition. In this way, any effects potentially emerging from variableair flows were ruled out.

8.5 Experimental observations of PL response

8.5.1 General response behaviorInGaN/GaN nanooptical probesdlike MOXsdexhibit a nonselective

broad-range gas response. When exposing InGaN/GaN NWAs to differentgases, three different kinds of PL response behaviors can be observed. InFig. 8.4, these responses are schematically represented as functions of timearising from rectangular boxlike gas exposure profiles. In addition, this figureshows how gas response values were evaluated from such transients.

8.5.2 Response to oxidizing gasesAmong all reactive gases, O2 is the one that occurs in highest concentrations(w20% in N2) in ambient air. As a reference point to all other kinds of gasresponse, it is therefore necessary to investigate how InGaN/GaN nanoopti-cal probes respond to changes in gas concentration from pure N2 to syntheticair (SA: 20% O2 and 80% N2) and back again. Fig. 8.5 shows that suchchanges produce quenching PL responses, showing that O2 adsorbates form

Figure 8.3 Schematics of the gas test rig, featuring test gas cylinders and a vapor satu-ration bottle for producing high concentrations of alcohols, either diluted in SA or in N2.


NT

R

P P Q

T

(a) (b) (c) –

– ––

–

Figure 8.4 Extraction of gas response data from photoluminescence (PL) responsetransients of nanowire array transducers: (a) pure quenching response, (b) enhancingresponse, and (c) transition from quenching to enhancing response.

G

T ( ) T ( )

G

(a) (b)

Figure 8.5 (a) Photoluminescence (PL) response to an O2 concentration pulse with aconcentration of 20% O2/N2 as applied in an inert background of 100% dry N2 followedby a 300 ppb NO2 pulse in a synthetic air background (20% O2/80% N2). (b) PLresponses to 330 ppb O3 as applied in a background of dry synthetic air.


surface recombination centers which enhance the surface recombination ve-locity beyond its native level in inert N2. As O2 is a strongly oxidizing gas witha positive electron affinity of EO2 ¼ 0.448 eV,36,37 adsorbed O2 molecules arelikely to trap electrons photogenerated in the InGaN/GaN nanowires, thusforming negatively charged adsorbates. Such negatively charged centers causean upward band bending and thus attract photogenerated holes into theelectron-trapping adsorbates. As the energy released during such surfacerecombination processes can be dissipated either inside the nanowiresthemselves or transferred as kinetic energy to the desorbing molecules, theseprocesses are likely to be nonradiative.

NO2 and O3 are even more strongly oxidizing gases with electron affin-ities of ENO2 ¼ 2.273 eV and EO3 ¼ 2.103 eV.36,37 Because of their highreactivity, both gases are harmful to human health. Normal environmentalconcentrations of NO2 are below 1 ppm and below 100 ppb in the case ofO3. Being strongly oxidizing gases, both compete with the much higherconcentrations of O2 for photogenerated electrons from the NWA bulkand for adsorption sites on the InGaN surfaces. As shown on the right-hand side of Fig. 8.5, sub-ppm concentrations of NO2 and O3 producereductions in the PL intensity additionally those reductions already inducedby the background O2. Quite interestingly, the NO2- and O3-inducedreductions are of similar size as the ones produced by the huge backgroundconcentration of 20% O2 ¼ 2 � 105 ppm O2. We will see in the followingthat NO2 and O3 have a capability of displacing existing O2 adsorbates fromtheir binding sites, thus forming adsorbate sites with strongly enhancedrecombination velocities.

8.5.3 Response to H2O vaporH2Owas found to be outstanding among a large variety of gases as it producesPL intensity changes in inert backgrounds of dry N2 without any interventionof reactive O2. Moreover, H2O proved to be exceptional as it can form bothquenching (Q) and enhancing (E) adsorbates, which convert into each otherduring a single vapor exposure.26 This capability of H2O of forming Q- andE-adsorbates is shown in Fig. 8.6. In this figure, PL responses are shown asH2O vapor pulses with 30% relative humidity were applied under differentNWA operation conditions. In the left-hand panels, a low LED excitationlight intensity of 7 mWwas applied, while vapor sensing tests were performedat room temperature and at 120 �C. In the right-hand panels, results areshown in which the same experiment was repeated, but at a higher LEDintensity of 200 mW. An overall look at these data shows that on onset of


each H2O vapor pulse, the PL is initially sharply quenched and that thisquenching fades away with increasing speed as the NWA temperature and/or the LED light intensity are raised. While in the limits of low temperatureand low LED power almost purely quenching responses are observed, almostpurely enhancing responses are found in the limit of high temperature(120�C) and high LED light intensity. Another interesting feature occursimmediately after termination of the H2O exposure pulses. There, PLovershoots are observed which fade away with increasing speed, again asNWA temperatures and/or LED light powers are raised.

8.5.4 Response to reducing gasesIn contrast to O2 and H2O, reducing gases such as H2 and hydrocarbons donot have any intrinsic capability of producing sizable PL changes.25,27 InFig. 8.7, an InGaN/GaN NWA was exposed to increasing concentrationsof ethanol (EtOH), either applied in a background of inert N2 or in amore reactive background of synthetic air. Additionally, room temperatureand elevated temperature (T ¼ 120 �C) exposures are compared. The verysmall quenching responses toward EtOH, when applied in inert N2 back-grounds, show that EtOH adsorbatesdlike the oxidizing gases discussedabovedintrinsically form recombination centers. Considering the smallresponses of DPL� 1% at concentrations in the order of 104 ppm, theEtOH-derived recombination centers are orders of magnitude less efficient

R

T ( ) T ( )

L H

P Opt = 0.7 mW

P Opt = 200 mW

–

–

–

(a) (b)

Figure 8.6 Photoluminescence response to water vapor pulses (gray boxes) as appliedat different nanowire array temperatures and at different LED light powers used forexcitation.


than those derived fromO2 and even more fromNO2 or O3.When EtOH isapplied in reactive backgrounds of synthetic air, however, EtOH consistentlyshows sizable enhancing responses, particularly at elevated temperatures.27

Similar data as for ethanol were also obtained for other alcohols and foraliphatic hydrocarbons, not carrying any functional groups.27 Like EtOH,enhancing PL responses of any sizable magnitude consistently were onlyobserved when the reducing gases were diluted in synthetic air and whenthe InGaN/GaN NWAs were operated at elevated temperatures. Overall,these latter data indicate that reducing gases are indirectly detected via theconsumption of quenching oxygen adsorbates and via the formation ofenhancing water adsorbates. CO2 molecules, which are also likely to formduring surface oxidation processes, did not produce any sizable PL response.27

8.6 Analysis of adsorption phenomena

8.6.1 Concentration and temperature dependence ofthe PL response

Having established the main qualitative features of the gas response,we now turn to the discussion how this response varies with the gas concen-tration and the temperature of the InGaN/GaN NWA transducers. Fig. 8.8

T ( ) T ( )

(a)

(c)

(e)

(b)

(d)

(e)

Figure 8.7 Photoluminescence response of an InGaN/GaN nanowire array towardethanol exposure pulses applied in backgrounds of N2 (a, c) and in synthetic air (b,d). Panels (e, f) show the timing of the ethanol vapor flows.


summarizes response data for some of the major air constituents and aircontaminants (O2, NO2, and O3) as well as for H2O and EtOH. To ensurecomparable conditions, all data were acquired at high LED excitation lightintensities (w200 mW). The test gases O2 and H2O were diluted in dry N2

to show that these intrinsically form recombination-enhancing (O2) andrecombination-reducing (H2O) adsorbates without any interference withother reactive gases. This, of course, does not exclude the possibility thatO2 and H2Omolecules might be competing with the much more numerousN2 molecules for adsorption sites on the InGaN/GaN surfaces. All othergases were diluted in dry synthetic air, which means that the main compe-tition of these test gases is with the O2 molecules in the synthetic air.Whereas data points stand for measured response data, the full lines representfits to Langmuir adsorption isotherms (Eq. 8.5).

As a common trend, one can observe that all gas responses increase as theInGaN/GaN operation temperatures are raised. The observed magnitudes

RR

R

–––––

–

–

–

–

–

–

– – ––

(a)

(b)

(c)

(d)

(e)

(f)

C C

RR

R

- -

Figure 8.8 Concentration and temperature dependencies of the photoluminescence(PL) response of an InGaN/GaN nanowire array to PL-reducing (O2, NO2, and O3) andPL-enhancing gases (H2O and EtOH). Panel f shows the variation of the PL responsewith LED light intensity. Data points represent measured data; full lines represent fitsto the Langmuir adsorption and recombination model.25


of response, however, differ in sign with all oxidizing gases producingnegative, i.e., quenching, and H2O and EtOH positive, i.e., enhancingPL responses. Qualitatively, all gases show very weak concentration depen-dencies with trends toward saturation at very high gas concentrations. Thefull lines through the individual data sets show that all data can be fitted toLangmuir isotherms38e40 by choosing optimum values for the adsorptionenergy Eads and for the saturated gas response asat:

RPL�pgas;T ;M

� ¼ asatðTÞ

2664 pgas

pgas þ P00ðT ;MÞexp�� Eads

kBT

�3775:(8.5a)

In the square bracket term, pgas stands for the analyte partial pressure andT for the InGaN/GaNNWA temperature and P00 for the Langmuir desorp-tion pressure of the analyte gas:

P00ðT ;MÞ ¼ kB TVQðT ;MÞ . (8.5b)

VQ(T, M), finally, is the quantum volume of the adsorbates:

VQðT ;MÞ ¼�

h2

2 pM kB T

�32

; (8.5c)

with M standing for the adsorbate molecular mass and h and kB for Planck’sand Boltzmann’s constants. For the gases considered in this chapter, P00 takeson values in the order of P00 z 1011 Pa24,38 Although all data can be fittedto such Langmuir isotherms, experimental verification is not easy to performas the predicted concentration dependencies are very weak, varying in aquasilogarithmic manner around the centers of the sensitivity windows, i.e.,around those partial pressures p1/2 at which the InGaN/GaN NWAs attainhalf of their saturation responses. Experimental verification requires verylarge variations in the test gas concentrations, which for physical and tech-nical reasons cannot be easily produced in all cases. The best evidence for theLangmuir hypothesis could be obtained in the case of EtOH, where the testgas concentration could be varied over five orders of magnitude (Fig. 8.8(e)).


Reference to Fig. 8.8 shows that the positions of the sensitivity windows,i.e., those partial pressures p1/2 at which 50% of the saturation PL response, isobserved, depends on the respective species. Equating the right-hand side ofEq. (8.5a) toRPL(p, T)¼ asat(T)/2, the partial pressures p1/2 can be evaluatedat which the gas response attains half of its maximum possible response:

p1=2 ¼ P00exp

�� Eads

kBT

�: (8.6)

In Fig. 8.9, this quantity is plotted as a function of the Langmuir adsorp-tion and recombination (LAR) adsorption energy Eads. Comparison of thetheoretical results (full lines) with the values of Eads (data points) extractedfrom the data of Fig. 8.8 shows that relatively modest variations in adsorp-tion energy lead to considerable shifts in the positions of the sensitivitywindows.

8.6.2 Competitive adsorption of air constituentsFocusing on the positions of the sensitivity windows in Fig. 8.8, it is evidentthat increasing NWA temperatures mainly impact the magnitudes of thesaturated responses, asat, but hardly the positions of the sensitivity windows.This latter effect is demonstrated in more detail in Fig. 8.10(a) where wehave replotted the O2 response data with the maximum response at eachtemperature scaled to unity. Also shown in this figure is a series of Langmuir

–

–

–

Figure 8.9 Correlation between adsorption energy Eads and center concentration p1/2of the sensitivity windows. The full lines show p1/2 versus Eads relationships as calculatedfrom the Langmuir adsorption and recombination model25 for O2 and NO2 (O3), respec-tively. The data points stand for pairs of p1/2-Eads values that had been extracted fromthe experimental data.


isotherms which had been calculated based on the assumption that theO2-related recombination centers bind to the InGaN surfaces with a uniqueand species-specific adsorption energy of Eads O2 ¼ 0.66 eV, allowing at thesame time the NWA temperature to vary over the full experimental range,i.e., 25 �C �T �150 �C. While this variation in temperature shifts theexpected positions of the LAR sensitivity windows by no less than fourorders of magnitude, all experimental data approximately fall ontoone and the same isotherm corresponding to an adsorbent temperature ofabout 80�C. As this huge discrepancy exceeds any experimental scatterin the determination of PL responses, this effect is discussed in further detail.

Mathematically, this discrepancy can be resolved by assuming that theadsorption energy is not a species-dependent constant but rather atemperature-dependent quantity as shown in Fig. 8.10(b). Overall, thesedata show that the adsorption energies for all species investigated varylinearly with temperature, increasing from values close to zero at cryogenictemperatures to values around 1 eV at temperatures around 150�C.Comparing the adsorption energies for the different species at any fixedtemperature, it is found that adsorbate-binding energies increase in the orderof H2O, O2, EtOH, NO2, and O3.

R

–

–

–

–

(a) (b)

C

Figure 8.10 (a) (Data points) Normalized O2 response as observed at nanowire arraytemperatures ranging in between 25 �C and 150 �C; (full lines): Langmuir adsorptionand recombination isotherms calculated assuming a constant and species-specificoxygen binding energy of Eads O2 ¼ 0.66 eV. (b) Eads as a function of adsorbent temper-ature T. The data points were evaluated from the photoluminescence response data ofFig. 8.8.


At first sight, the data in Fig. 8.10(b) could be interpreted in the way thatall investigated adsorbates undergo physisorption at low temperature andremain in this weakly bound state as long as temperatures are low enoughto prevent relaxation into more deeply bound chemisorption states. Thelinear increase in Langmuir adsorption energies at T > 50K further suggeststhat there are no unique chemisorption states but that there are rathercontinua of deeper and deeper bound chemisorption states which can bereached at the expense of surmounting higher and higher reaction barriers.Although such a scenario cannot be excluded, we draw attention to asecond, more realistic scenario that is able to explain the observed linearincrease in adsorption energies with temperature.25 Here, we considerthat the InGaN/GaN NWA transducer surfaces are never exposed to asingle gas alone and that the different kinds of molecules therefore competefor a limited number of adsorption sites. Competitive adsorption phenom-ena are well-known and generally recognized in the fields of catalysis andchromatography.41e44 To the best of our knowledge, however, competitiveadsorption has not been seriously considered in the field of gas sensing yet.

To introduce the concept of competitive adsorption, we start from theusually considered idealized situation of a single gas interacting with anadsorbent. In such an idealized scenario, Langmuir adsorption leads to asurface coverage q with pgas standing for the gas pressure and T for thetemperature of the adsorbent38e40:

q�pgas;T ;M

� ¼ pgas

pgas þ P00ðT ;MÞexp�� Eads

kBT

� : (8.7)

The only parameter in this equation is the adsorption energy Eads whichmeasures the strength of adsorption on the adsorbent under study. Asdiscussed in text books,39 Eq. (8.7) represents the steady-state solution ofthe differential equation, which describes the kinetics of adsorption:

dq�pgas;T

�dt

¼ radsðTÞ$�1� q�pgas;T

��pgas � rdesðT Þ$q�pgas;T�

(8.8)

with rads and rdes standing for the adsorption and desorption rate constants ofthe analyte in question. In this equation, the term (1 � q) considers that anyadsorption site on the adsorbent surface can only be occupied once.

The idealized case of an adsorbent being exposed to a single gas can beeasily generalized. Considering the simplest case of two gas species, A and B,


competing for a single kind of adsorption sites, Eq. (8.8) turns into a systemof two coupled equations for the surface coverages qA and qB:

dqAdt

¼ rads;AðTÞ$ð1� qA � qBÞ$pA � rdes;AðTÞ$qA; (8.9a)

and

dqBdt

¼ rads;BðTÞ$ð1� qA � qBÞ$pB � rdes;BðTÞ$qB. (8.9b)

Under steady-state conditions, one then obtains for the surface coveragesof adsorbates A and B:

qA�pA; pB;T

�¼ pA

pAþ P00AðTÞexp�� EA

kBT

� �1þ pB

P00BðT Þ exp

�EB

kBT

�� ;(8.10a)

and

qB�pA; pB;T

�¼ pB

pBþ P00BðTÞexp�� EB

kBT

� �1þ pA

P00AðT Þ exp�EA

kBT

�� .(8.10b)

Both entities, obviously, follow similar isotherms as in the single-adsorbate case. Differences, however, are caused by additional terms thatappear in the denominators, which describe the interaction of both adsor-bates on the adsorbent surface. Both equations can be reduced to the familiarsingle-component Langmuir form by replacing the species-dependentadsorption energies in the denominator, EA,B by species-dependent effectiveadsorption energies, which contain contributions of the competing species aswell. Focusing on species A, one obtains

EA;eff ð pB;TÞ¼ � kBT ln

�exp

�� EA

kBT

�þ pB

P00BðT Þ exp

�EB�EA

kBT

��;

(8.11)

and similarly, for EB,eff(pA, T). Obviously, this latter expression does not onlydepend on the adsorption strength of adsorbate A but it is also sensitive to thedifference in adsorption energies of both species. For EA ¼ EB, Eeff,A effec-tively reduces to Eeff,A ¼ EA as realistic partial pressures pB of component Balways remain far smaller than the desorption pressure P00B of species B.


As an example of competitive adsorption, we show in Fig. 8.11a PLresponse data of an optimized InGaN/GaN NWA sample with pronouncedsensitivity toward O2. There, the observed quenching response of O2 couldbe followed over three orders of magnitude in O2 concentration andexplained as a competition with the more numerous but weaker-bindingN2 background molecules, which are supposed to be nonquenching. Asshown there, the observed O2 response can be explained by a competitionof N2 and O2 adsorbates in case the adsorption energies EO2 and EN2 arefixed at EO2 ¼ 0.95 eV and EN2 ¼ 0.75 eV, respectively. According tothis result, the N2 adsorbates win this competition if O2 abounds in smallconcentrations. The InGaN/GaN surface is then almost completely coveredwith nitrogen. At O2 concentrations around 100 ppm, the stronger-bindingO2 molecules start to displace the nitrogen adsorbates, until at approximately3000 ppm O2, nitrogen and oxygen adsorbates are present in equal concen-trations. Finally, at normal ambient air concentrations of O2, the surface isalmost completely covered with O2. As O2 adsorbates tend to adsorb inionized form, this exchange of N2 for O2 adsorbates obviously can onlyinvolve a small fraction of 0.1%e1% of all geometrically available surfacestates as the Weisz limitation39,40 needs to be respected.

With the values of EO2 and EN2 being fixed by the O2 concentrationdependence of Fig. 8.119(a), the temperature dependence of the apparent

C

( )–

(a) (b)

Figure 8.11 (a) Variation of the surface coverages of N2 and O2 as the O2 concentrationin the N2 background is raised. Data points represent photoluminescence responsedata obtained on an InGaN/GaN nanowire array kept at a temperature of about 120�C; (b) variation of the effective adsorption energies with temperature for the fourinvestigated air constituents. Data points are the Eads(T) values obtained from Langmuiradsorption and recombination fits.


adsorption energy for O2 can be evaluated from Eq. (8.11). With the back-ground nitrogen pressure of pN2 ¼ 8$104 Pa, the data in Fig. 8.11(b) areobtained. There, the calculated trend for EO2;eff

�T�, is compared to the

LAR adsorption energies for O2 already displayed in Fig. 8.10(b). Consid-ering the calculated EO2;eff

�T�trend, it is seen that the effective binding

energy should first increase linearly with temperature and then turn tosaturation as the temperature is further raised. Within the much more limitedtemperature range in which EO2 values could be determined fromO2 calibra-tion curves, the predicted linear trend is reasonably well confirmed. Alsoincluded in Fig. 8.11(b) are those LAR adsorption data for H2O, EtOH,NO2, and O3 that have already been reported above, alongside with theirrespective fits to Eq. (8.11). Obviously, the linear increases in LAR adsorptionenergies for these other gases can also be reasonably well approximated by thecompetitive adsorption model, provided the high-temperature limits ofadsorption energies for these other gases are correctly scaled relative to theinitially determined value of EO2. As Eq. (8.11) only depends on thedifference in adsorption energies of the respective analyte with its maincompetitor, these saturation values do not necessarily represent the exactquantum-chemical binding energies of the individual adsorbates. In principle,however, these values would become experimentally accessible in case PLresponse measurements could be extended into the range of temperaturesat which saturation in LAR-binding energies can be observed.

Another interesting observation is that the saturation values of Eads deter-mined in Fig. 8.11(b) are linearly correlated to the electron affinity of the testgases (Fig. 8.12). As the electron affinity of molecules measures the energygained in capturing a free electron on a molecule X, i.e.,

X þ e/ X�; (8.12)

this correlation emphasizes the role of photogenerated electrons in enablingthe formation of adsorbate states on InGaN surfaces. Whereas the positiveelectron affinities of O2, NO2, and O3 mean that free electrons can becomeexothermally bound to these gases, the negative electron affinities of N2 andH2O mean that electron attachment is endothermic, and that therefore theformation of negatively charged N2

� and H2O� ions is unlikely. Whereas

free water does not attract electrons, electron capture, however, becomespossible when water is disintegrated into H and OH fragments.

As already mentioned above, the saturation values of Eads,sat listed inFig. 8.12 do not necessarily represent the true quantum chemical bindingenergies of the adsorbates. As any possible shift in the initial value of the


N2-binding energy would shift all other binding energies by the sameamount, the correlation in Fig. 8.12 would remain unaltered.

8.6.3 Competition between quenching and enhancing H2Oadsorbates

The overview presented in Section 8.5 has shown that water vapor influ-ences the PL response of InGaN/GaN NWAs in a more complex mannerthan most other gases investigated. Whereas O2, NO2, and O3 simplygive rise to quenching PL responses, water vapor can form both quenching(Q) and enhancing (E) adsorbates. Most interestingly, our data suggest thatinitially quenching adsorbates tend to transform into enhancing ones asIII-nitride surfaces are continually exposed to water vapor. Given enoughtime, Q-adsorbates will therefore always tend to transform intoE-adsorbates, which represent the equilibrium form of H2O adsorption.

In the following, we present evidence that such QeE transformationbehavior is another manifestation of a competitive adsorption process thatcan occur at UV-illuminated III-nitride surfaces. Following the idea thatH2O can form two different kinds of competing adsorbates, rateEqs. (8.9a,b) can be used again, with species A standing for Q- and speciesB for E-adsorbates. As the competition between both is a time-dependentone, time-dependent solutions to Eqs. (8.9a,b) need to be fitted to experi-mentally observed PL transients. Such fits then yield values for the adsorp-tion and desorption rate constants of the two kinds of H2O adsorbates as wellas activation energies for the kinetic processes of adsorption and desorption.

E ( )– –

Figure 8.12 Saturated values of Eads as a function of the electron affinity36 of the testgases.


The time rates of change of surface coverages qQ and qE ultimatelydepend on the rate rH2O with which H2Omolecules collide with the InGaNsurfaces:

rH2O�pH2O;T

� ¼ pH2Offiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2pMH2OkBT

p Aads (8.13)

Here, pH2O stands for the H2O partial pressure in the gas phase,MH2O forthe water molecular mass, and Aads for the effective area of the adsorptionsites. Furthermore, as not every gas-kinetic collision will lead to an adsorp-tion event, we write the two adsorption rates as products of the gas-kineticcollision rate, rH2O(pH2O,T), a species-specific sticking factor (sQ,sE), and aBoltzmann factor containing the species-specific reaction barriers(εads,Q, εads,E):

rads;Q�pH2O;T

� ¼ sQrH2O�pH2O;T

�exp

�� εads;Q

kBT

�; (8.14a)

rads;E�pH2O;T ; Popt

� ¼ sErH2O�pH2O;T

�exp

�� εads;E

kBT

��Popt

Popt;max

�g

:

(8.14b)

Moreover, as the formation of enhancing adsorbates is accelerated by UVillumination, we introduce in rads,E an additional factor that depends on apower g of the UV input optical power Popt. Similarly, we assume thatdesorption of both species requires thermal activation as well:

rdes;QðTÞ ¼ r0;Q exp

�� εdes;Q

kBT

�; (8.15a)

rdes;EðTÞ ¼ r0;E exp

�� εdes;E

kBT

�. (8.15b)

With the solutions for qQ and qE obtained, the PL response under theinfluence of water vapor adsorption becomes26

RPL H2O¼ qQA

�1� RQ

Rnr

�|fflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflffl}

¼ :aQ

þ qEA

�1� RE

Rnr

�|fflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflffl}

¼ :aE

(8.16)

In this final equation, RQ and RE are the nonradiative recombinationrates through quenching and enhancing adsorbates, while Rnr is the nonra-diative recombination rate under clean-air conditions; A is a common scalefactor.


In Fig. 8.13(a), we show measurements of the water vaporeinduced PLresponse as measured at a relative humidity of 30%, but at different NWAtemperatures and at different LED light exposure levels. In Fig. 8.13(b),these results are compared with those PL responses that were calculatedusing the above set of equations. For simplicity, we assumed ideal boxlikeH2O exposure pulses as shown in the bottom panels in Fig. 8.13. As canbe seen, the simulated PL data reasonably well reproduce the observedfeatures of the actual PL response. Water vapor exposure pulses, in partic-ular, can be seen to produce initially quenching responses which more orless rapidly turn into enhancing ones as the exposures are being maintained.Another encouraging aspect is that the PL overshoots that occur after thetermination of each H2O exposure pulse can also be reproduced.

More insight into the underlying microscopic processes can be gained bylooking at those values of model parameters that are revealed by the abovefits. Turning to the adsorption parameters first, we find that the stickingprobabilities for forming Q-adsorbates are quite low (sQ¼3.5$10�2), andeven lower for E-adsorbates (sE¼2.4$10�6). More interestingly, the forma-tion of Q-adsorbates does not seem to require thermal energy (εads,Qz 0 eVper adsorbate), while E-ones require sizable amounts of activation energy(εads,E z 0.4 eV per adsorbate). Furthermore, the formation ofE-adsorbates is found to increase with the square root of the UV LEDpower, i.e., g ¼ 1/2. Another interesting result is that both kinds of adsor-bates seem to desorb without any significant input in thermal energy(εdes,Qy εdes,E z 0 eV). Both kinds of adsorbates, however, desorb atvery different temperature-independent rates (rdes,Q z 10 Hz vs. rdes,E z5$10�2 Hz). Overall, our kinetic considerations show that there is hardlyany energy involved in the adsorption and desorption of quenchingadsorbates, whereas sizable amounts of thermal energy and, even moreUV light energy, are required for the formation of enhancing adsorbates.It is therefore reasonable to associate Q-adsorbates with physisorbed andE-adsorbates with chemisorbed water molecules, the dividing line betweenphysi- and chemisorption being the typical strength of a hydrogen bridgebond (EH z 0.4 � 0.5eV).36

8.7 Molecular mechanism of adsorption

Our experiments have revealed that molecular adsorption at InGaNsurfaces can both quench and enhance the native PL of InGaN/GaNNWAs. The fact that adsorption can modify the native PL in both directions


suggests that the native PL is limited by intrinsic surface sites and that theirtrapping and recombination cross sections are being altered as adsorbatesbind to them. Our gas sensing tests have revealed equilibrium adsorptionenergies Eads ¼ εdes � εads of several groups of adsorbates on these sitesand some additional information about the kinetic parameters εads and εdes

has been obtained in the case of water adsorption. The microscopic natureof the intrinsic surface sites and the ways in which adsorbates bind to thesesites, however, have not yet been positively identified. Identification of thesemolecular entities clearly remains a subject of future research. A possibleapproach to attain such information is to apply diffuse reflection Fouriertransform spectroscopy (DRIFTS)45,46 to InGaN/GaN NWAs as these areexposed to different analytes. To motivate such research, we close thischapter with some ideas concerning the possible nature of the intrinsicsurface sites and about the ways adsorbates might bond to these.

Starting point of our considerations is the nature of chemical bonds atnonpolar III-nitride surfaces. The bulk equilibrium lattice sites of Ga(In)and N, which in Kr€ogereVink notation38e40 are denoted by GaGa (InIn)and NN, all feature tetrahedral coordination. Within the bulk, the tetrahe-dral coordination of the constituent atoms of InGaN/GaN nanowires isenabled by an electron transfer from N toward Ga or In atoms. In contrast

M S

P opt = 0.7 mW

Popt = 200 mWPopt = 200 mW

P opt = 0.7 mW

T ( ) T ( )

()

(a) (b)

Figure 8.13 (a) Experimental photoluminescence (PL) response data as observed un-der widely different experimental conditions. For each level of excitation-light intensity,the variation of the PL-light intensity in response to the humidity profiles (gray-filledboxes) is shown for three different temperatures; (b) values of PL response as obtainedfrom the competitive adsorption model.


to the Kr€ogereVink notation in which all ground state equilibrium config-urations have formal charge zero, we denote these bulk equilibrium sites byGa4

� (In4�) and N4

þ. In this notation, which has become known underdifferent names such as octet or 8-N rules,47,48 the superscript (�) denotesthe ionic charge of the constituent atoms and the subscript the number ofcovalent bonds to neighbor atoms required to arrive either at completelyfilled or empty valence shells. The advantage of the 8-N notation is thatit most visibly keeps track of those valence electron transfers that take placeduring electronically driven coordination changes. In the realm of solid-statephysics, 8-N considerations have proved to be useful to understandelectronically driven coordination changes in the bulk of amorphoussemiconductors.48e50 Here, we employ such considerations to the analysisof electronically driven coordination changes at molecules adsorbed onphotoexcited semiconductor surfaces.

Moving from the bulk of InGaN/GaN nanowires to their surfaces, thetetrahedral coordination of the bulk atoms cannot be continued. Due to alack of bond partners, the InGaN/GaN constituents cannot form four cova-lent bonds at the surface, which results either in a high density of danglingbonds at the surface or in the formation of threefold coordinated N-Ga(In)sites with reconstructed surface bonds. Following the above notation, thesethreefold coordinated surface sites can be labeled as N3

0, Ga30, and In3

0. Apair of such reconstructed sites at a nonpolar

�1100

�lateral nanowire surface

is sketched in Fig. 8.14(a). In principle, such a pair can transform from itsreconstructed ground state into an activated excited state once a photogener-ated electronehole pair becomes trapped at this pair. The transformation

B A

G

P

RP

(N30) (N4

+)

(N4+)(Ga4

–) (Ga4–)

(Ga4–)(Ga3

0)

Ga

Ga Ga

Ga

N

N

N

N

+ –

(a) (b)/

Figure 8.14 (a) Tetrahedral bulk bonding and surface reconstruction at lateral�1100

�surfaces. The coloring indicates formation of Lewis acidebase pairs; (b) photoactivatedstate of III-nitride surface following photogeneration in the bulk. The circles with plusand minus signs denote a photogenerated electronehole pair.


shown in Fig. 8.14(b) requires that an electronehole pair, initially delocalizedacross the entire InGaN well volume, becomes sharply localized at a pair ofGa(In) and N atoms. Such localization is necessary to make the trappedelectronic charge chemically active in the sense of enabling local redox reac-tions as those shown in Fig. 8.14. Such localization is a relatively improbableprocess. In amorphous semiconductors, such localization phenomena occur inthe small fraction floc < 10�3 of localized band-tail states which is a smallfraction of the total number of electronic band states. The other requirementfor enabling local coordination changes is the lack of geometrical constraintsthat would otherwise prevent electronically destabilized chemical bonds todisintegrate and to reform in a different manner. In amorphous semiconduc-tors, local coordination changes are enabled by density-deficient regions likevoidlike defects and/or by the diffusional motion of bonded hydrogen.48e50

All such changes are clearly impossible within the bulk of a fully coordinatedcrystalline material as in the interior of an InGaN/GaN nanowire. A crystal-line sensor surface, however, is a far less constrained environment. Wetherefore conjecture that a small fraction of the total number of Ga(In)-Nsurface sites may actually support local transformations as those shown inFig. 8.14(b). In gas sensors and in heterogeneous catalysis, the number densityof charged surface sites is limited by the Weisz limitation.39,40 Physically, thislimitation arises because surface charge densities in the order of 1012cm�2 willgenerate electrical fields comparable to the breakdown field of the underlyingsemiconductor. This well-established limit suggests that again only a smallfraction fWZL < 10�3 of all Ga(In)-N surface sites will actually be able tosupport bond reconfigurations as those shown in Fig. 8.14.

Accepting such a possibility of local reconfiguration, the followingpicture emerges: turning to the reconstructed ground state in Fig. 8.14(a)first, we note that N being a potential electron donor and Ga an electronacceptor, a surface N-Ga pair can be considered as a Lewis acid/base(LAB) pair. On photoexcitation (hn) such pairs can trap electronehole pairsas shown in Fig. 8.14b forming N4

þ and Ga4� (or In4�, respectively):

N03þGa034Nþ

4 þGa�4 ; (8.17)

When such a process occurs in vacuum or in an inert atmosphere, thepair of charged dangling bonds will discharge after a time sr as the electrontrapped on the Gae4 site tunnels back to its neighboring Nþ

4 site. When theenergy difference between initial and final states can be dissipated in theform of phonons, the ensuing recombination process will be radiation-lessand result in PL quenching.


As InGaN/GaN surfaces exposed to ambient air are likely to form thinlayers of natural oxide,15e21 similar reconfigurations are also conceivable onoxidized surfaces (Fig. 8.15). Like on nonoxidized surfaces, pairs of neigh-boring oxygen and gallium atoms can form LAB pairs which can switchbetween reconstructed ground states and electronically excited states withtwo dangling bond radicals pointing out of the surface:

O02þGa034Oþ

3 þ Ga�4 (8.18)

When InGaN/GaN NWAs are operated in backgrounds containingoxygen or other kinds of reactive gases, the charged dangling bonds pointingout of electronically excited sensor surfaces can form cross-linking bonds andthus allow reactive gases to form adsorbates. Such chemisorption bonds alterthe trapping and recombination cross sections of the native LAB pairs andthus promote changes in the luminescence output which become experi-mentally observable in the form of a gas response. These latter possibilitiesare sketched in Fig. 8.16, indicating that adsorbate bonding can both quench(O2, NO2, and O3) and enhance (H2O) the native PL response.

The above considerations about adsorbate bonding describe a possiblescenario that can account in a qualitative manner for the gas response datadescribed in earlier sections of this chapter. As adsorbate bonding is acomplex and exciting field of research with many unresolved issues andchallenges ahead, we expect that this picture will become modified asadditional spectroscopic evidence becomes available. A key enabling factorfor attaining such information is that the PL probing of gas response canpotentially be combined with other forms of operando spectroscopy, inparticular DRIFTS.45,46

B AG

(O20)

(N4+)

(O3+)

(N4+)

Ga

Ga

Ga

Ga

GaGa(Ga30) (Ga4

–)

N N NN

O OP

R

+ –

P

(a) (b)

Figure 8.15 (a) Lewis acidebase pair formed by Ga and embedded oxygen at lateral�1100

�surfaces; (b) photoactivated state of an oxidized III-nitride surface after

photoexcitation.


8.8 Conclusions and outlook

InGaN/GaN nanowire structures proved to be very efficient nanoop-tical probes for investigating adsorption processes on semiconductor surfaces.Because PL emission does not depend in a similarly complex way on thetransport of photoexcited electronehole pairs as in resistively readout gassensors, luminescence probing provides a more direct approach to the obser-vation of adsorption processes on semiconductor surfaces. Looking towardfuture perspectives, improvements in material’s quality and the applicationof nanowire heterostructures with strong carrier confinement as well asrefinements in the optical readout periphery will be able to extend the acces-sible temperature range of the optochemical transducers to temperatures upand beyond the 200 �C range where resistively readout MOX gas sensors areroutinely operated and where thermally activated surface oxidation andsurface combustion processes take place. As combinations of PL responsewith DRIFTS measurements appear to be experimentally feasible, PLadds another useful tool to the toolbox of in operando spectroscopies.45,46

We are therefore confident that InGaN/GaN NWA will provide deeperinsights into the microscopic processes underlying adsorption at oxide andnonoxide surfaces in the near future.

ae

w

G P

O W

T

N N N N

NNNNGa Ga Ga Ga

GaGaGaGa

+ ++–

–– O

O O

H

H

H

H

(a) (b) (c) (d)

Figure 8.16 Lewis acidebase (LAB) pair alternating between ground (a) and electron-ically excited state (b) giving rise to thermal quenching in vacuum or in inert gas atmo-spheres; (c) LAB modified by dissociative oxygen adsorption featuring an enhancednonradiative recombination rate; and (d) LAB with surface bonds passivated by waterfragments featuring a reduced nonradiative recombination rate.


Concerning future applications in the field of gas sensors, it should be keptin mind that InGaN/GaN NWAs can also be grown on transparent sapphireor GaN substrates. In this way, the electrical readout periphery can becompletely removed from the spot of sensing where the sensitive materialsare in contact with the medium to be sensed. By employing light fibers forcarrying UV excitation light to the NWAs and luminescence light back tothe light detector, InGaN/GaN nanooptical probes can also be operated inenvironments heavily affected by electromagnetic interference.14 Anotherinteresting aspect is that by means of selective area growth, III-nitride nano-wires can be grown in the form of regular arrays with wire-to-wire distancesin the order of visible light wavelengths. With the help of such sparse arraysnot only a better media access is enabled but also a more efficient lightcoupling can be achieved by exploiting photonic crystal effects.51

References1. Moseley PT, Norris JOW, Williams DE. Techniques and mechanisms in gas sensing. Adam

Hilger; 1991.2. Ihokura K, Watson J. The stannic oxide gas sensor principles and applications. Boca Raton:

CRC Press; 1994.3. Korotcenkov G. Handbook of gas sensor materials; properties, advantages and shortcomings for

applications volume 1: conventional approaches. 2013.4. Comini E, Faglia G, Sberveglieri G. Solid state gas sensing. 1st ed. Boston, MA: Springer

Science & Business Media; 2009. https://doi.org/10.1007/978-0-387-09665-0.5. Yamazoe N, Shimanoe K. Receptor function and response of semiconductor gas sensor.

J Sensors 2009;2009:1e21. https://doi.org/10.1155/2009/875704.6. Hua Z, Li Y, Zeng Y, Wu Y. A theoretical investigation of the power-law response of

metal oxide semiconductor gas sensors I: Schottky barrier control. Sensor Actuator BChem 2018;255:1911e9. https://doi.org/10.1016/J.SNB.2017.08.206.

7. Hua Z, Qiu Z, Li Y, Zeng Y, Wu Y, Tian X,Wang M, Li E. A theoretical investigationof the power-law response of metal oxide semiconductor gas sensors II: size and shapeeffects. Sensor Actuator B Chem 2018;255:3541e9. https://doi.org/10.1016/J.SNB.2017.09.189.

8. Ulrich M, Bunde A, Kohl C-D. Percolation and gas sensitivity in nanocrystalline metaloxide films. Appl Phys Lett 2004;85:242e4. https://doi.org/10.1063/1.1769071.

9. Sauerwald T, Russ S. Percolation effects in metal oxide gas sensors and related systems.In: Kohl C-D, Wagner T, editors.Gas sens. Fundam. Berlin Heidelberg: Springer; 2013.https://doi.org/10.1007/5346.

10. Faglia G, Baratto C, Sberveglieri G, Zha M, Zappettini A. Adsorption effects of NO2 atppm level on visible photoluminescence response of SnO2 nanobelts. Appl Phys Lett2005;86:011923. https://doi.org/10.1063/1.1849832.

11. Comini E, Baratto C, Faglia G, Ferroni M, Sberveglieri G. Single crystal ZnOnanowires as optical and conductometric chemical sensor. J Phys D Appl Phys 2007;40:7255e9. https://doi.org/10.1088/0022-3727/40/23/S08.

12. Pallotti DK, Passoni L, Gesuele F, Maddalena P, Di Fonzo F, Lettieri S. Giant O2-induced photoluminescence modulation in hierarchical titanium dioxidenanostructures. ACS Sens 2017;2:61e8. https://doi.org/10.1021/acssensors.6b00432.


https://doi.org/10.1007/978-0-387-09665-0

https://doi.org/10.1155/2009/875704

https://doi.org/10.1016/J.SNB.2017.08.206



https://doi.org/10.1063/1.1769071

https://doi.org/10.1007/5346

https://doi.org/10.1063/1.1849832

https://doi.org/10.1088/0022-3727/40/23/S08

https://doi.org/10.1021/acssensors.6b00432

13. Palmer DW. Properties of III-nitride semiconductors. 2018. http://www.semiconductors.co.uk/nitrides.htm.

14. Paul S, Helwig A, M€uller G, Furtmayr F, Teubert J, Eickhoff M, Sumit P.Opto-chemical sensor system for the detection of H2 and hydrocarbons based onInGaN/GaN nanowires. Sensor Actuator B Chem 2012;173:120e6. https://doi.org/10.1016/j.snb.2012.06.022.

15. Schalwig J, M€uller G, Karrer U, Eickhoff M, Ambacher O, Stutzmann M, G€orgens L,Dollinger G. Hydrogen response mechanism of PteGaN Schottky diodes. Appl PhysLett 2002;80:1222e4. https://doi.org/10.1063/1.1450044.

16. Weidemann O, HermannM, Steinhoff G,Wingbrant H, Lloyd Spetz A, StutzmannM,Eickhoff M. Influence of surface oxides on hydrogen-sensitive Pd:GaN Schottkydiodes. Appl Phys Lett 2003;83:773. https://doi.org/10.1063/1.1593794.

17. Shalish I, Shapira Y, Burstein L, Salzman J. Surface states and surface oxide in GaNlayers. J Appl Phys 2001;89:390e5. https://doi.org/10.1063/1.1330553.

18. Garcia MA, Wolter SD, Kim T-H, Choi S, Baier J, Brown A, Losurdo M, Bruno G.Surface oxide relationships to band bending in GaN. Appl Phys Lett 2006;88:013506.https://doi.org/10.1063/1.2158701.

19. Winnerl A, Garrido JA, Stutzmann M. GaN surface states investigated by electrochem-ical studies. Appl Phys Lett 2017;110:101602. https://doi.org/10.1063/1.4977947.

20. Wang L, Bu Y, Li L, Ao J-P. Effect of thermal oxidation treatment on pH sensitivity ofAlGaN/GaN heterostructure ion-sensitive field-effect transistors. Appl Surf Sci 2017;411:144e8. https://doi.org/10.1016/j.apsusc.2017.03.167.

21. Kehagias T, Dimitrakopulos GP, Becker P, Kioseoglou J, Furtmayr F, Koukoula T,H€ausler I, Chernikov A, Chatterjee S, Karakostas T, Solowan H-M, Schwarz UT,Eickhoff M, Komninou P. Nanostructure and strain in InGaN/GaN superlatticesgrown in GaN nanowires. Nanotechnol 2013;24:435702. https://doi.org/10.1088/0957-4484/24/43/435702.

22. Teubert J, Becker P, Furtmayr F, Eickhoff M. GaN nanodiscs embedded in nanowiresas optochemical transducers. Nanotechnol 2011;22:275505. https://doi.org/10.1088/0957-4484/22/27/275505.

23. Teubert J, Paul S, Helwig A, M€uller G, Eickhoff M. Group III-nitride chemical nano-sensors with optical readout. In: Kohl C-D, Wagner T, editors. Gas sens. Fundam.Springer Berlin Heidelberg; 2014. p. 311e38. https://doi.org/10.1007/5346_2014_58.

24. Maier K, Helwig A, M€uller G, Becker P, Hille P, Sch€ormann J, Teubert J, Eickhoff M.Detection of oxidising gases using an optochemical sensor system based on GaN/InGaN nanowires. Sensor Actuator B Chem 2014;197:87e94. https://doi.org/10.1016/j.snb.2014.02.002.

25. Maier K, Helwig A, M€uller G, Hille P, Teubert J, Eickhoff M. Competitive adsorptionof air constituents as observed on InGaN/GaN nano-optical probes. Sensor Actuator BChem 2017;250:91e9. https://doi.org/10.1016/j.snb.2017.04.098.

26. Maier K, Helwig A, M€uller G, Hille P, Teubert J, Eickhoff M. Photoluminescenceprobing of complex H2O adsorption on InGaN/GaN nanowires. Nano Lett 2017;17:615e21. https://doi.org/10.1021/acs.nanolett.6b03299.

27. Maier K, Helwig A, M€uller G, Eickhoff M. Photoluminescence detection of surfaceoxidation processes on InGaN/GaN nanowire arrays. ACS Sens 2018;3(11):2254e60. https://doi.org/10.1021/acssensors.8b00417.

28. She X, Huang AQ, Lucia O, Ozpineci B. Review of silicon carbide power devices andtheir applications. IEEE Trans Ind Electron 2017;64:8193e205. https://doi.org/10.1109/TIE.2017.2652401.


http://www.semiconductors.co.uk/nitrides.htm

http://www.semiconductors.co.uk/nitrides.htm



https://doi.org/10.1063/1.1450044

https://doi.org/10.1063/1.1593794

https://doi.org/10.1063/1.1330553

https://doi.org/10.1063/1.2158701

https://doi.org/10.1063/1.4977947


https://doi.org/10.1088/0957-4484/24/43/435702

https://doi.org/10.1088/0957-4484/24/43/435702

https://doi.org/10.1088/0957-4484/22/27/275505

https://doi.org/10.1088/0957-4484/22/27/275505

https://doi.org/10.1007/5346_2014_58

https://doi.org/10.1007/5346_2014_58




https://doi.org/10.1021/acs.nanolett.6b03299


https://doi.org/10.1109/TIE.2017.2652401

https://doi.org/10.1109/TIE.2017.2652401

29. Rath P, Ummethala S, Nebel C, Pernice WHP. Diamond as a material for monolith-ically integrated optical and optomechanical devices. Phys Status Solidi 2015;212:2385e99. https://doi.org/10.1002/pssa.201532494.

30. Flack TJ, Pushpakaran BN, Bayne SB. GaN technology for power electronic applica-tions: a review. J Electron Mater 2016;45:2673e82. https://doi.org/10.1007/s11664-016-4435-3.

31. Chao E. Analysis of LED technologies for solid state lighting markets e technical report No.UCB/EECS-2012-138. 2012. https://www2.eecs.berkeley.edu/Pubs/TechRpts/2012/EECS-2012-138.html.

32. Ambacher O, Majewski J, Miskys C, Link A, Hermann M, Eickhoff M, Stutzmann M,Bernardini F, Fiorentini V, Tilak V, Schaff B, Eastman LF. Pyroelectric properties ofAl(In)GaN/GaN hetero- and quantum well structures. J Phys Condens Matter 2002;14:3399e434. https://doi.org/10.1088/0953-8984/14/13/302.

33. Fletcher ASA, Nirmal D. A survey of gallium nitride HEMT for RF and high powerapplications. Superlattice Microst 2017;109:519e37. https://doi.org/10.1016/J.SPMI.2017.05.042.

34. Pearton SJ, Kang BS, Kim S, Ren F, Gila BP, Abernathy CR, Lin J, Chu SNG. GaN-based diodes and transistors for chemical, gas, biological and pressure sensing. J PhysCondens Matter 2004;16:R961e94. https://doi.org/10.1088/0953-8984/16/29/R02.

35. Pearton SJ, Ren F. Gallium nitride-based gas, chemical and biomedical sensors. IEEEInstrum Meas Mag 2012;15:16e21. https://doi.org/10.1109/MIM.2012.6145256.

36. NIST. Chemistry webBook. 2016. http://webbook.nist.gov/chemistry/.37. Wu T-Y. Electron affinity of boron, carbon, nitrogen, and oxygen atoms. Phys Rev E

1955;100:1195e6. https://doi.org/10.1103/PhysRev.100.1195.38. Kittel C, Kroemer H. Thermal physics. 2nd ed. New York: W. H. Freeman and

Company; 1998.39. Henzler M, G€opel W. Oberfl€achenphysik des Festk€orpers. 2nd ed. Wiesbaden: Teubner

Verlag; 1994.40. Morrison SR. The chemical physics of surfaces. Boston, MA: Springer US; 1977. https://

doi.org/10.1007/978-1-4615-8007-2.41. Gun’ko VM. Competitive adsorption. Theor Exp Chem 2007;43:139e83. https://

doi.org/10.1007/s11237-007-0020-4.42. Martin RJ, Al-Bahrani KS. Adsorption studies using gas-liquid chromatographydII.

Competitive adsorption. Water Res 1977;11:991e9. https://doi.org/10.1016/0043-1354(77)90157-9.

43. Jacobson JM, Frenz JH, Horvath CG. Measurement of competitive adsorptionisotherms by frontal chromatography. Ind Eng Chem Res 1987;26:43e50. https://doi.org/10.1021/ie00061a009.

44. Poplewska I, Piatkowski W, Antos D. Effect of temperature on competitive adsorptionof the solute and the organic solvent in reversed-phase liquid chromatography.J Chromatogr A 2006;1103:284e95. https://doi.org/10.1016/j.chroma.2005.11.038.

45. Weckhuysen BM. Operando spectroscopy: fundamental and technical aspects ofspectroscopy of catalysts under working conditions. Phys Chem Chem Phys 2003;5:1.https://doi.org/10.1039/b309654h.

46. Degler D, Barz N, Dettinger U, Peisert H, Chassé T, Weimar U, Barsan N. Extendingthe toolbox for gas sensor research: operando UV/vis diffuse reflectance spectroscopyon SnO2-based gas sensors. Sensor Actuator B Chem 2016;224:256e9. https://doi.org/10.1016/J.SNB.2015.10.040.

47. Langmuir I. The arrangement of electrons in atoms and molecules. J Am Chem Soc 1919;41:868e934. https://doi.org/10.1021/ja02227a002.


https://doi.org/10.1002/pssa.201532494

https://doi.org/10.1007/s11664-016-4435-3

https://doi.org/10.1007/s11664-016-4435-3

https://www2.eecs.berkeley.edu/Pubs/TechRpts/2012/EECS-2012-138.html

https://www2.eecs.berkeley.edu/Pubs/TechRpts/2012/EECS-2012-138.html

https://doi.org/10.1088/0953-8984/14/13/302

https://doi.org/10.1016/J.SPMI.2017.05.042

https://doi.org/10.1016/J.SPMI.2017.05.042

https://doi.org/10.1088/0953-8984/16/29/R02

https://doi.org/10.1109/MIM.2012.6145256

http://webbook.nist.gov/chemistry/

https://doi.org/10.1103/PhysRev.100.1195

https://doi.org/10.1007/978-1-4615-8007-2

https://doi.org/10.1007/978-1-4615-8007-2

https://doi.org/10.1007/s11237-007-0020-4

https://doi.org/10.1007/s11237-007-0020-4

https://doi.org/10.1016/0043-1354(77)90157-9

https://doi.org/10.1016/0043-1354(77)90157-9

https://doi.org/10.1021/ie00061a009

https://doi.org/10.1021/ie00061a009

https://doi.org/10.1016/j.chroma.2005.11.038

https://doi.org/10.1039/b309654h



https://doi.org/10.1021/ja02227a002

48. Mott NF. Electrons in disordered structures. Adv Phys 1967;16:49e144. https://doi.org/10.1080/00018736700101265.

49. M€uller G, Kalbitzer S, Mannsperger H. A chemical-bond approach to doping, compen-sation and photo-induced degradation in amorphous silicon. Appl Phys A Solid Surf1986;39:243e50. https://doi.org/10.1007/BF00617268.

50. Robertson J. Mott lecture: how bonding concepts can help understand amorphoussemiconductor behavior. Phys Status Solidi 2016;213:1641e52. https://doi.org/10.1002/pssa.201532875.

51. Winnerl J, Hudeczek R, Stutzmann M. Optical design of GaN nanowire arrays forphotocatalytic applications. J Appl Phys 2018;123:203104. https://doi.org/10.1063/1.5028476.


https://doi.org/10.1080/00018736700101265

https://doi.org/10.1080/00018736700101265

https://doi.org/10.1007/BF00617268



https://doi.org/10.1063/1.5028476

https://doi.org/10.1063/1.5028476

CHAPTER NINE

Rareearthedopedoxidematerialsfor photoluminescence-basedgas sensorsV. Kiisk, Raivo JaanisoUniversity of Tartu, Tartu, Estonia

Contents

9.1 Introduction 2729.1.1 The concept of PL-based gas sensing 2729.1.2 Advantages of rare eartheactivated inorganic sensor materials 2739.1.3 Overview of the progress 275

9.2 Sm3þ:TiO2 2779.2.1 Introduction 2779.2.2 Preparation and characterization of samples 2789.2.3 PL-based oxygen sensing 2809.2.4 Sensing mechanism and its mathematical model 2829.2.5 Multivariable sensing with TiO2:Sm

3þ 2869.3 Eu3þ:ZrO2 288

9.3.1 Introduction 2889.3.2 Preparation and characterization of samples 2899.3.3 Oxygen sensing 2909.3.4 Sensing mechanism 291

9.4 Tb3þ:CePO4 2949.4.1 Introduction 2949.4.2 Preparation and characterization of samples 2959.4.3 Gas sensing and its mechanism 295

9.5 Pr3þ:(K0.5Na0.5)NbO3 2989.5.1 Introduction 2989.5.2 Synthesis 2989.5.3 Oxygen sensing 298

9.6 Conclusion 299References 300



https://doi.org/10.1016/B978-0-08-102559-8.00009-4

9.1 Introduction

9.1.1 The concept of PL-based gas sensingElectrical (such as electrochemical or conductometric) gas sensors are

widely used, but in many cases, optical sensing has distinct advantages. Notonly it is more tolerant to electrical noise and independent of material’s elec-trical properties (including their long-term drifts due to, e.g., growth of con-tacting necks between the particles in granular materials) but potentiallyallows remote access to the sensing volume and operation in imagingmode.1 Fluorescence, or more generally photoluminescence (PL), is oneof the most sensitive optical signals (compared with absorption or Ramanscattering), although usually not from the gas molecules themselves butfrom some dedicated PL centers in the solid, which either directly or indi-rectly interact with the gas molecules. In addition, as opposed to absorption,PL is a two-step process (excitation and emission), so that external influenceshave more chances to intercept the emission of a photon.

Conventional chemiresistive gas sensing involves measurement of theelectrical conductance (or resistance) of a thin semiconductor layer, typicallya metal oxide (MOX) film (Fig. 9.1). Usually, the material is simulta-neously heated and/or optically stimulated to decrease the response time(e.g., stimulate recovery from gas adsorption). The setup forluminescence-based sensing looks somewhat similar, but instead of electricalconductivity, one measures the intensity of the secondary emission from thematerial (i.e., PL). In this case, photostimulation (also called photoexcitation)is mandatory. Note that, while the electrical conductivity is an integratedquantity characterizing the bulk of the material, the optical signal could

Photostimulation (optional) Photostimulation

Sensor film

Heater Heater

A

Luminescence

Figure 9.1 The principle of luminescence-based gas sensing (right) compared with thatof conductometric sensing (left).

272 V. Kiisk and Raivo Jaaniso

be detected (in principle) separately from each point of the illuminatedmaterial.

As a well-known example, optical oxygen sensing is frequently done byutilizing oxygen-sensitive PL probes embedded into an oxygen permeablepolymer or porous solegel matrix. These probes are mostly organic mole-cules with energy levels resonant with the triplet oxygen levels and a longexcited state lifetime.1 Accordingly, the sensing effect is based on PLquenching via the so-called collisional energy transfer process between theprobe and the oxygen molecules. Hence, the fluorescence intensity andits decay time monotonically decrease with increasing oxygen concentration(partial pressure) in a predictable manner as quantified by the SterneVolmerrelationship. Suitable selection of the probe allows sensitive operation over aspecific (partial) pressure range of the target gas. Equivalently, one can alsoconsider the fluorophores as pressure sensors.2

In principle, the PL signal can be collected from tiny volumes (comparedwith the volume required for infrared optical absorption by gas molecules3).The system can be rather tightly integrated using miniature solid-state exci-tation sources and detectors. In this regard, luminescent gas sensing is com-parable and compatible with the chemiresistive gas sensing. Semiconductorscan exhibit simultaneous gas-sensitive electrical conductivity and PL signals.Moreover, solid matrices can accommodate several different PL centers (orseveral emissive transitions of single PL center), which can be distinguishedspectrally. These electrical and optical signals can complement each otherproviding further possibilities to improve the specificity, accuracy, or dy-namic range of gas sensors. PL can also be employed as a sensitive analyticaltool for probing sophisticated processes initiated by surface reactions in gassensors or photocatalysts.4,5

9.1.2 Advantages of rare eartheactivated inorganic sensormaterials

There are numerous organic luminescent probes available, mainly based onpolycyclic aromatic hydrocarbons and metaleligand complexes.1 However,organic fluorophores are susceptible to photobleaching and cannot with-stand temperatures beyond a few hundred centigrades (at least not in achemically reactive environment). Such conditions are frequently encoun-tered in industrial applications (for example, in the thermal treatment ofpolymer packaging for food or healthcare products). Thermal quenchingof the fluorescence is sometimes evident already at room temperature. Thereis accordingly an interest to develop inorganic gas-sensitive luminophores

Rare earthedoped oxide materials for photoluminescence-based gas sensors 273

which show stable operation in more aggressive environments and underintense optical excitation.

In nanocrystalline form, many inorganic phosphors exhibit certainambient sensitivity. For example, recent studies revealed that the intensityof the intrinsic luminescence of many common MOX nanopowders(ZnO, TiO2, SnO2, WO3) notably reacts to the change of ambient environ-ment.6,7 Other works have attempted to harness the defect-related or exci-tonic PL (mostly of ZnO, SnO2, TiO2, and MgO) using a more refinedmorphology of the nanomaterial, such as nanostructured films8e10 or nano-wires/-rods/-belts,11e17 occasionally achieving ppm-level detection of haz-ardous gases9,11e16 or sensing of oxygen over a wide pressure range.8,17,18

However, the presence and emission properties of lattice defects (such asF-centers) and excitons can be drastically dependent on the quality of thematerial.

On the other hand, such dielectric or semiconductor matrices with rela-tively wide energy gap can accommodate various impurities acting as PLcenters with well-defined luminescence properties. Of the several distinctclasses of impurity ions, trivalent rare earth (RE) ions constitute highly pho-tostable impurity centers possessing predictable narrow excitation and emis-sion bands and long fluorescence lifetimes (w1 ms).19,20 Especially in aregular crystalline surrounding, the RE ions exhibit a series of sharp spectrallines typical for the 4fe4f transitions. These features simplify the detection ofthe sensor signal, monitoring either PL decay kinetics or PL intensities atseveral different emission wavelengths resulting in a ratiometric response.Because of peculiarities of their energy level schemes, the main emissiontransition of several RE ions (such as Eu3þ and Tb3þ) is also quite resistantto cross relaxation and thermal quenching. The main quenching mecha-nism, multiphonon relaxation, becomes apparent only at relatively hightemperatures (assuming low phonon frequencies of the host medium21).This is also the basis for the use of RE-activated refractory materials for op-tical sensing of high temperatures, such as encountered in the thermal barriercoatings of gas turbines.20

In many cases of the studied RE-doped materials (as reviewed in thefollowing), it is believed that the gas sensing stems from a redox reaction,i.e., charge transfer between some lattice ions or defects (including the acti-vator itself) and the adsorbate molecules. In particular, the sensing mecha-nism may involve trapping or release of free charge carriers in the energybands of the semiconductor matrix leading effectively to a long range inter-action. This is necessary as the PL center is usually located inside the


nanocrystal. The chemical change induced by gas adsorption can intercepteither the excitation or emission path (or both) of a PL center. Unlike theoxygen sensors based on direct quenching of PL by gas molecules, thismechanism can cause either increase or decrease of the PL intensity as thematerial is exposed to the gas. The excitation path can be easily affectedonly if the luminescence is sensitized (through energy transfer from thehost or other impurities).

There are also inorganic materials where luminescence itself originatesfrom the surface (such as the recombination of electronehole pairs insome semiconductors), and the emission center can more directly interactwith the adsorbed or even the gaseous oxygen. For example, Nagai andNoguchi reported already in 1978 that the PL of cleaved n-InP (110) surfacereversibly reacted to O2, H2, N2, and H2O.22 More recently, oxygen wasfound to have a remarkably strong quenching effect on the PL attributedto radiative decay of excitons localized at the corners or edges of MgO nano-cubes.18,23 For InGaN/GaN nanowires, this type of gas response is describedin Chapter 8.

9.1.3 Overview of the progressOne of the earliest works studied porous Eu2O3-gAl2O3 composites (pow-der compacts), where the 325 nm laser irradiation in vacuum decreased theEu3þ PL intensity while the same laser irradiation in oxygen gas recoveredthe PL.24 Such kind of response clearly opposes that of organic fluorophores.The effect was attributed to Eu3þ/Eu2þ valence change coupled to creationor annihilation of oxygen vacancies at gAl2O3 and Eu2O3 particle surfaces.The transitions were clearly photoinduced, so that the material was proposedas erasable photomemory. At room temperature, under a laser power densityof just 32 mW/cm2, the response time was less than 10 min.

One of the most interesting gas-sensitive RE-activated material, Sm3þ-doped nanocrystalline TiO2 (anatase), was first reported in 200525 and wasmore rigorously studied in recent years in the form of solegel-preparednanopowder.7,26e28 PL of the Sm3þ ions was found to be reasonablyresponsive to oxygen gas even at room temperature (with a response timeof a few minutes). Similarly to Eu2O3-gAl2O3, the PL significantlyimproved in an oxygen-rich environment, but the mechanism is quitedifferent and based on an indirect influence of adsorbed oxygen on the fluo-rescence quantum yield of Sm3þ. Moreover, the ultraviolet excitation isinitially absorbed by the matrix which thereafter transfers the energy tothe RE emission center,29,30 further complicating the interpretation of the


sensing mechanism.28 It is remarkable that the PL is oxygen-sensitive over apressure range spanning at least four orders of magnitude and reaching thetrace concentrations.

Recently, somewhat similarly prepared TiO2:Eu3þ nanopowder was

reported also exhibiting a decreased Eu3þ PL (under 325 nm excitation)with reduced ambient oxygen pressure.31 However, in this case, the Eu3þ

emission lines were relatively broad and there remains a possibility thatthe ions are prevalently located at the surface. Most recent studies (unpub-lished data) have shown that the Sm3þ:TiO2 PL is quite sensitive not only tooxygen but also to NH3 and H2O. More interestingly, PL of Nd3þ:TiO2

appears to exhibit a similarly strong, but reversed oxygen-sensing behavior,where the mechanism is quite different and connected to the excitationefficiency, which is again indirectly affected by gas adsorption. The similarquenching effect of oxidizing gases (O2 or NO2) has been already reportedfor the intrinsic (band-to-band excited) PL of TiO2,

4,8 ZnO,9,10,16 andSnO2

11,16 nanostructured films or nanobelts.Microspectroscopic studies of rather small (mean diameter w10 nm)

crystallites of TiO2:Eu3þ demonstrated a case where a significant fraction

of the RE ions were located at surface.32 The surface and interior Eu3þ cen-ters could be spectroscopically distinguished by the ratio of the 5D0/

7F1and 5D0/

7F2 transitions, because the latter is hypersensitive to local sym-metry. While exciting the nanoparticles with 405 nm laser in argon atmo-sphere, activation of certain light-emitting defect sites (presumably formedat surface) was observed. The ratio of the two Eu3þ emission bands wasreversibly changed as well. It was proposed that energy transfer from freeand trapped excitonic states to Eu3þ ions takes place, but the trapped exci-tons (at the defect sites) can only excite surface-located Eu3þ ions.

In 2010, a patent was issued claiming multiple methods of optical oxygensensing based on the measurements of the PL intensity of Eu3þ ions doped inZrO2 nanocrystallites not bigger than 60 nm.33 Working temperature wasproposed in the range of 0e350�C. Recently also several peer-reviewed ar-ticles were published reporting similar sensing effect from solegel-preparedZrO2:Eu

3þ particles operated at 300�C.34,35 The studies showed that, atleast for low Eu3þ concentrations, the response mechanism of ZrO2:Eu

3þ

is quite similar to that of TiO2:Sm3þ. Interestingly, it was possible to control

the magnitude and even the sign of the relative PL response by codopingwith niobium, where presumably the combined effect of Nb5þ and Eu3þ

ions controls the number of vacancies or other charge-compensating defectsin the material.34,35


Similar kind of oxygen sensing mechanism was more explicitly demon-strated on the basis of CePO4:Tb

3þ nanocrystals.36 This material utilizes thefact that several RE ions in solids can exist as redox couples, in this caseCe3þ/Ce4þ. The change of valence was induced by exposing the materialto the oxidizing or reducing atmosphere at 200�C. The activator ion(Tb3þ) did not change its valence, but both its excitation and emission prop-erties were affected by the valence change of cerium.

A complementary case of PL-based oxygen sensing was demonstratedwith Pr3þ-doped (K0.5Na0.5)NbO3.

37 In addition to the response observedin the absolute PL intensity, the researchers also recorded reasonably strongratiometric response. The Pr3þ ion has several emitting levels in the excitedstate, and in the (K0.5Na0.5)NbO3 host these levels seem to be differentlyaffected by adsorbed oxygen.

Ratiometric sensor material can also be realized in a more straightforwardmanner by utilizing several differently gas-sensitive constituents. Forinstance, a recent work prepared a hybrid material containing Sm3þ:TiO2

nanoparticles (discussed above) attached to the metal-organic frameworkBio-MOF-1 containing Tb3þ ions.38 The PL of Sm3þ was enhanced,whereas that of Tb3þ was quenched by oxygen, leading to a strong ratio-metric response.

Another quite complex composite system based on Sr4Al14O25:Eu2þ,Dy3þ (belonging to a family of persistent phosphors39) was recentlystudied.40 Samples containing the phosphor particles along with plasmoni-cally active silver nanoparticles (in resonance with Eu2þ excitation) dispersedin a polymer matrix were prepared in various morphologies. Oxygenquenched the Eu2þ luminescence up to w4 times, where the oxygen con-centration dependence could be described by a modified SterneVolmerlaw. The sensing mechanism probably involves autoionization of the excitedEu2þ ions as an intermediate step, but one should recognize that numerousenergy and charge transfer processes have been proposed to describe the PLmechanisms of persistent phosphors in general.39

9.2 Sm3D:TiO2

9.2.1 IntroductionTitanium dioxide (TiO2) is the most common oxide of titanium with

a wide range of applications, including pigments, photocatalysts, and gassensors. Crystalline TiO2 has two common polymorphs. Bulk crystals ofTiO2 usually possess the thermodynamically stable rutile phase, whereas


TiO2 nanocrystals develop mostly in the metastable anatase phase. Nano-crystalline anatase gradually transforms to rutile at temperatures rangingfrom 600 to 1200�C.41 The phase transition is kinetically defined, and thetransition temperature depends on crystallite size and purity of the mate-rial.41,42 For instance, doping with RE ions results in a higher transition tem-perature, about 900�C for a typical solegel-prepared material.43

The bandgap of anatase is 3.2 eV,44 or a bit higher for nanocrystallites lessthan 30 nm in diameter.45 This implies that already near-UV light is stronglyabsorbed by thin TiO2 films. Moreover, several impurity ions (such as Nd3þ,Sm3þ, Eu3þ, and Yb3þ) have been found to emit intense visible or NIR PLafter being excited through an energy transfer from the TiO2 host.

30,46 Itwas proposed that Sm3þ and Nd3þ are particularly suitable dopants inanatase TiO2 because their energy levels are located close to the middle ofthe TiO2 bandgap so that both hole and electron trapping efficiently takesplace, whereas autoionization is negligible.30 These features facilitate theuse of doped thin films of TiO2 for special luminescence applications suchas optical gas sensing, where widespread violet light sources (e.g., light-emitting diodes) can be used for PL excitation combined with compact pho-todetectors (e.g., photodiodes) for monitoring the PL.

9.2.2 Preparation and characterization of samplesMost of the gas sensing studies of Sm3þ:TiO2 have been carried out on solegel-prepared films or powders.25e28,47 The less-pronounced effect was alsoobserved in the case of an atomic layer deposited sample.25 The solegel pro-cess is based on the gelation of a colloidal suspension (“sol”) and formation ofa continuous inorganic network in the liquid phase (“gel”), as a result of hy-drolysis and polycondensation reactions in a solution containing an organicprecursor and water. The organic precursor is typically Ti(OC4H9)4 (tita-nium butoxide) dissolved in butanol. RE impurity is incorporated by addingproper amount of corresponding salt to the mixture (e.g., SmCl3•6H2O).Gel powder is obtained by dripping the solution to distilled water while stir-ring. A white precipitate is formed which is dried in an evaporator.

RE-related PL of the as-prepared gel is typically very weak due toquenching of the PL by OH-groups and organic residues. The materialneeds to be heat-treated at temperatures as high as 800�C to optimize thePL.48 Up to this temperature, the anatase phase is mostly preserved whereascrystallinity is drastically improved. To reduce agglomeration of the powderparticles, the crystallized powder can be dispersed in distilled water using anultrasound probe. The suspension is then transferred to a glass or silica


substrate by drop coating. The resulting material consists of a hierarchicalstructure of agglomerated nanocrystallites with an average grain size of about40 nm (Fig. 9.2). BET measurements showed a broad distribution of poresizes ranging from 25 to 150 nm.

The emission spectrum of RE-activated TiO2, when excited with a UVlight, generally consists of sharp lines due to the trivalent RE ion as well as abroad emission band peaking at 550 nm (Fig. 9.3). The latter is believed tobe intrinsic to anatase-type TiO2, originating either from self-trapped orbound excitons or defect states.44,49 The intrinsic emission becomes quitestrong at cryogenic temperatures50 and possibly also under an intense laserexcitation as the RE emitter becomes more easily saturated (due to a longlifetime of the excited state). Occasionally, the intrinsic emission itself ex-hibits rather strong ambient sensitivity,8 but in this series of studies, it wasnearly constant and could potentially be used as a reference.

Sm3þ ion has only one major emitting level 4G5/2 and the four terminalmanifolds 6HJ cause the four emission bands in the visible spectral range. Thespectral fine structure is due to the regular crystalline surrounding (crystal-field splitting of the 6HJ states). The intensity of the Sm

3þ lines strongly de-pends on the concentration of oxygen in the ambient environment, but the

Figure 9.2 Scanning electron micrograph of solegel-prepared annealed TiO2:Sm3þ

nanoparticles deposited on fused silica substrate. Adapted from Eltermann M, Utt K,Lange S, Jaaniso R. Sm3þ doped TiO2 as optical oxygen sensor material. Opt Mater 2015;51:24e30. https://doi.org/10.1016/j.optmat.2015.11.020.


https://doi.org/10.1016/j.optmat.2015.11.020

spectral fine structure and intensity ratio of individual lines is unaffected. Itindicates that the crystallographic surrounding of the emitting ions remainsunchanged during changes in gas composition, and only one crystalline siteof Sm3þ is present. Other Sm3þ sites in TiO2 may exist as well, but those arenot revealed under indirect excitation.48

9.2.3 PL-based oxygen sensingThe simplest type of gas sensing response is observed in the Sm3þ PL inten-sity (Fig. 9.4). In each measurement cycle, the sample was exposed to theoxygenenitrogen mixture for 10 min and then to pure nitrogen for10 min before the next exposure to oxygen. The final PL intensity inoxygen-containing ambient monotonically increases with increasing O2

concentration, an exactly opposite behavior to the PL-based sensorsdescribed by SterneVolmer law. The PL intensity has certain nonlineardependence from O2 concentration. The change of the latter is observedover four orders of magnitude (from about 100 ppm trace level up to normalpressure). Moreover, the response is reasonably fast even at room tempera-ture. As shown in the inset of Fig. 9.4, the characteristic response time is un-der 1 min and recovery time about 5 min. Note that the response time of thematerial itself should be even smaller, as the measured response includes theinstrumental time of changing the gas composition.

500 550 600 650 700 750

Wavelength (nm)

Inte

nsity

(a.u

.)

6H7/2

6H5/2

4G5/2

6H9/2

6H11/2

100 % N2

60 % N2 + 40 % O2

(a)

(b)

Figure 9.3 Photoluminescence emission spectra of the TiO2:Sm3þ nanoparticles

excited with 355 nm laser in different ambient environments. Adapted from Elter-mann M, Utt K, Lange S, Jaaniso R. Sm3þ doped TiO2 as optical oxygen sensor material.Opt Mater 2015;51:24e30. https://doi.org/10.1016/j.optmat.2015.11.020.



At the same time, a systematic trend was also observed in the Sm3þ PLdecay kinetics (using a nanosecond pulsed excitation source). The decaybecame faster as ambient O2 concentration decreased (Fig. 9.5). The effectis quite pronounced and uniform over the wide concentration range so thatone can easily employ the PL decay to define the sensor response indepen-dently of the absolute PL intensity. However, the decays strongly diverge

01

2

3

4

5

6

Rel

ativ

e P

L in

tens

ity I/

I 0

100

% O

2

2010

% O

240

1 %

O2

Time (m60 8

0.1

% O

2

min)

~1min

90 %

80 100

0.01

% O

2

~5min

90 %

0

Figure 9.4 Typical temporal response curve of Sm:TiO2 photoluminescence (PL) inten-sity for a set of gas exchange cycles covering a wide oxygen concentration range of100e0.01 vol% (at 25�C and normal pressure).28 The strongest Sm3þ emission bandaround 615 nm has been integrated (Fig. 9.3, with background subtracted).

0 1000 400030002000 0 1000 400030002000Time (μs) Time (μs)

λex = 355 nmλem = 612 nm

λex = 355 nmλem = 612 nm

100% O210% 1% 0.1%

0.01% 100% N2

100% O210% 1% 0.1% 0.01%

100% N2

PL

inte

nsity

(cou

nts)

PL

inte

nsity

(cou

nts)

102

103

104

105

102

103

104

(a) (b)

Figure 9.5 Sm3þ photoluminescence (PL) decay kinetics of two TiO2:Sm3þ samples

with 3% (A) and 0.5% (B) Sm concentration.28 The colored dots represent experimentaldata, solid black lines represent theoretical fits (Eq. 9.3), and the dashed straight linerepresents an exponential decay with characteristic time 300 ms.


from a simple exponential decay reflecting the complex nature of theexcitationeemission route of Sm3þ as discussed in the following sections.

9.2.4 Sensing mechanism and its mathematical modelAnalysis of the PL decay kinetics has turned out to be most informative to un-derstand the PL-based gas sensing mechanism inMOXs. The natural decay ofSm3þ ions should be about 300 ms as typically detected under direct excitationwith visible light.48 Such exponential decay is represented by the dashed linesin Fig. 9.5. The tail of the host-sensitized PL decay is clearly slower and in-dicates a delayed excitation of the Sm3þ ions (i.e., the excitation energy istrapped in the host matrix for a relatively long time). However, the initialpart of the decay is even faster in oxygen-deficient ambient gas, indicatingthat the gas sensing effect is mostly linked to PL quenching. Hence, it isconcluded that the excited state of the Sm3þ ion (energy donor) is depopu-lated not only by natural decay but also via energy transfer to certain latticedefects (energy acceptors). In turn, the number of such lattice defects iscontrolled by adsorption processes at the surface. It is plausible that the defectsare oxygen vacancies or other intrinsic TiO2 lattice defects, which only in aspecific charge state possess excitation energy matching the emission ofSm3þ. A qualitative scheme of the processes is depicted in Fig. 9.6.

A corresponding mathematical model is built as follows. Assuming thatthe acceptors are randomly distributed in the material, the PL decay shouldfollow the well-known law51

uðtÞ ¼ e�k0t�ðc=c0Þðk0tÞb (9.1)

Here k0 is the rate constant for natural decay (¼ 1=300 ms). The powerindex b depends on the type of interaction between the donors andacceptors and on the Euclidean dimensionality of the acceptor space. Themost common case is dipolar interaction in 3D spacedthe F€orster’s energytransfer with b ¼ 1=2. Parameter c is the acceptor concentration, while theconstant c0 characterizes the energy transfer strength and is inversely pro-portional to the third power of the characteristic (F€orster’s) energy transferradius.

The delayed excitation of Sm3þ ions can be attributed to trapping of theinitial excitation (polaron or some kind of exciton) in the TiO2 host for anextended period before the energy is transferred to a nearby unexcited Sm3þ

ion. The binding energy of these trap states must be relatively small (com-parable with the thermal vibration energy kBT ), as the resulting escape timesare still as small as several milliseconds at room temperature. Indeed, several


studies suggest the existence of such shallow traps in TiO2.52e54 Assuming

that the delayed population occurs at a constant rate k (corresponding tothe presence of a single type of trap level), the decay law (1) can be gener-alized as follows:

IðtÞ¼ I0uðtÞ þ I1ke�kt5uðtÞ; (9.2)

where the first term describes the instantly excited PL centers and the secondterm those PL centers excited with a delay. The 5 symbol marks convo-lution. The convolution reflects the fact that different PL centers start todecay at different time delays from the laser pulse, and the convolution sumsover all possible delay times from 0 to t. It was found that the experimentaldecay curves were accurately reproduced only by assuming the presence oftraps with different depths, leading to a distribution rðkÞ of delayed popu-lation rates. Eq. (9.2) is now further generalized to the equation

IðtÞ¼ I0uðtÞ þ I1

ZN0

dk rðkÞke�kt5uðtÞ; (9.3)

Radiative processesNon-radiative processesEnergy transferSurface electron trapping

UVexcitation

Shallowtraps

Valence band

Conduction band

PL center(Sm3+ ion)

PL

Quenchingdefect C

rystallite surface

O2–

a b

Figure 9.6 A simplified energy diagram of the TiO2:Sm3þ sensor material. The Sm3þ ion

is exited either instantly (blue dashed line) or in a delayed manner (red dashed line). Oncethe excitation has reached the Sm3þ ion, it can (a) emit a photon or (b) be quenched bya defect. This defect can be “switched off” by electron transfer to surface oxygen spe-cies. PL, photoluminescence. Adapted from Eltermann M, Kiisk V, Berholts A, Dolgov L,Lange S, Utt K, Jaaniso R. Modeling of luminescence-based oxygen sensing by redox-switched energy transfer in nanocrystalline TiO2:Sm

3þ. Sensor Actuator B Chem 2018;265:556e64.


Presumably, the result is not very sensitive to the choice of rðkÞ. Goodfitting of the decays (solid lines in Fig. 9.5) can be obtained by assumingthermally activated release of carriers from traps with an exponential distri-bution of trap depths.28 The parameter b may also require adjustment forhigh Sm3þ concentrations (where cross relaxation becomes dominant),segregation of the impurity (due to annealing at high temperatures55), orrather small radius of the crystallites (where the acceptor space cannot beconsidered infinite).

One can now study the effect of oxygen adsorption on the differentmodel parameters derived from the fitting. It was found that only the relativeacceptor concentration c =c0 changed substantially and decreased systemati-cally with increasing O2 content in the ambient gas (Fig. 9.7). So the con-centration change of the acceptor centers seems to be the main factorconnecting the changes in O2 concentration and PL intensity of Sm3þ.

Note that within the frames of the proposed model, the delayed excita-tion complicates only the decay function (Eq. 9.3) and increases the averagelifetime but does not alter the stationary PL signal and its dependence on theoxygen pressure. This is because each excitation, which is initially trapped inthe host, will finally end up exciting a Sm3þ ion. In the case of F€orster’s en-ergy transfer, one can explicitly derive the stationary PL intensity (i.e., areaunder the decay curve)28:

S ¼ I0 þ I1k0

�1� ffiffiffi

pp

q exp�q2�erfcðqÞ�;

where 2q ¼ c=c0. As one can see, the equation does not include any detailson the traps causing the delayed PL and the intensities I0 and I1 are simplyadded.

Further elaboration of the model requires assumptions on O2 adsorptionand related processes. Adsorption of gaseous O2 on MOX surfaces is gener-ally accompanied by a charge transfer and formation of various negativelycharged oxygen species at the surface (O2

e being dominant at room temper-ature56,57). The electron involved in the charge transfer could be taken fromthe acceptor defects (either directly or through the conduction band ofTiO2). Different adsorption mechanisms exist depending on the defectivestate and coverage of the surface (including the presence of hydroxylgroups).58,59 Therefore, the adsorption isotherm can be quite complex. Itwas found28 that the changes in the derived acceptor concentration c =c0


can be connected to the changes of O2� surface density, assuming that the

latter follows the T�oth isotherm:

cadscsat

¼� ðKxÞm1þ ðKxÞm

�1=m;

where x is gaseous O2 concentration (or partial pressure) and K ;m are pa-rameters of the isotherm. This isotherm is a generalization of the Langmuirand Freundlich isotherms and is believed to be more suitable for a broadpressure range and heterogeneous substrates.

We note that photoadsorption and -desorption of O2 on clean or hydrox-ylated TiO2 surfaces has also been known for a long time57,60,61 and maydetermine some essential aspects of the gas sensing, such as response timeor drift of the PL signal. Impact of both photochemistry and humiditymust be taken into account in the future studies and applications of this sensor.

0 10 20 30 40 50Time (h)

SIP

L

SE

CS

EP

L

9.1

9.2

9.3

8.0

8.5

9.0

–14.5

–14.0

–13.5

[O2]

(%)

100

101

Figure 9.7 The temporal behavior of extrinsic (Sm3þ) photoluminescence (PL) (SEPL),intrinsic PL (SIPL), and electrical conductivity (SEC) of the TiO2 nanopowder, while beingexcited by 365 nm LED and subjected to a randomly changing oxygen concentration.7

All signals are represented as logarithms, whereas data point color encodes the actualoxygen concentration varied between 0.21% and 21%. The experiment was done atroom temperature.


9.2.5 Multivariable sensing with TiO2:Sm3þ

The photoexcited charge carriers in TiO2 are known to cause quite signif-icant photoconductivity, which also responds to the ambient environ-ment.62 It would be intriguing to register simultaneously the correlatedresponses of photoconductivity and PL, to improve selectivity or accuracyof the sensor and gain further insight into the complex energy and chargetransfer processes. Indeed, a recent study7 showed that the electrical conduc-tance of a thin layer of solegel-prepared TiO2:Sm powder (similar to theone discussed above) had almost one-to-one correspondence to the lumi-nescence of Sm3þ ions, while the ambient oxygen concentration (in a flow-ing O2/N2 mixture) was randomly varied (Fig. 9.7). Considering theanticipated conduction mechanism (through the double-Schottky barriersformed between contacting nanoparticles), it is likely that adsorption ordesorption of oxygen causes not only recharging of the acceptor defects(quenching Sm3þ PL) but also changes the extent of the space charge layerson crystallite surfaces, leading to a respective change in electricalconductance.

In the cited study, the oxygen sensitivity of the intrinsic broadband emis-sion (with a spectrum similar to the one shown in Fig. 9.3) was also recorded(Fig. 9.7). The response of the intrinsic PL is much slower and has possibly adifferent mechanism.8 Nevertheless, it can deliver complementary informa-tion as a sensor signal.

Figure 9.8 depicts the pairwise correlations between the signals. Interest-ingly, there are almost no occurrences where different O2 concentrations

SIP

L

SIP

L

SE

C

SECSEPL SEPL

[O2](%)100 101

Figure 9.8 The three signals from Fig. 9.7 plotted pairwise against each other.7


map to the same coordinates in these two-dimensional spaces. Thespace spanned by the Sm3þ PL intensity and photoconductivity is quitesqueezed. However, the defect-related PL intensity in combinationwith one of the remaining two signals covers a wide space and could poten-tially more reliably predict the ambient O2 concentration and suppress thesensor drift (which is clearly present in the individual signals shown inFig. 9.7).

Mapping combinations of the three signals (features) to correspondingO2 concentrations (target) is a typical machine learning problem. Forinstance, if two sensor signals S1 and S2 are involved, then the predictedoxygen concentration p½O2� ¼ f ðS1; S2Þ, where f is a convenient mathe-matical function involving a number of free parameters that can be “trained”to fit the data. Because of the rather smooth relationships between lnðp½O2�Þand the logarithms of the measured signals (Fig. 9.8), it is justified to considerlnðp½O2�Þ as a function of S1 and S2 and expand the unknown function intoa Taylor series:

lnðp½O2�Þ ¼ a0 þ a1S1 þ a2S2 þ a11S21 þ a12S1S2 þ a22S22 þ. (9.4)

The coefficients a0, ai, ai;j, etc. (which represent the partial derivatives)can be considered as the free parameters of the model. They can be deter-mined (trained) by minimizing a cost function taken as the residual sum ofsquared errors between the logarithms of the predicted and the correspond-ing true O2 concentrations. The advantage of the approach is that due to thelinear dependence of the model on the parameters and the particular choiceof the cost function, the optimization is an ordinary least square (OLS) prob-lem. Regardless of the number of variables, OLS leads to a system of linearequations for the optimal parameter values, so that one can directly find theglobal optimum without a complex and time-consuming iterative processrequired for training a neural network.

Models with different orders and signal combinations were tested, wherethe initial 32 h (60%) of data was used for training and the remaining datapoints were used to evaluate the mean relative error for the model predic-tion. It turned out that in most cases a linear relation (first-order polyno-mials) provided the best results. As an example, Fig. 9.9 demonstrates howfusing electrical conductance and intrinsic PL yield significant improvementof precision in predicting oxygen concentration. More than 4 times ofreduction in the relative error was achieved.


9.3 Eu3D:ZrO2

9.3.1 IntroductionZirconium dioxide (ZrO2) is a well-known refractory material pri-

marily used in hard ceramics, gemstones, abrasives, and optical coatings.At ambient conditions, the stable phase of nominally pure zirconia is themonoclinic one (m-ZrO2, naturally found as the mineral baddeleyite). Athigh temperatures, a martensitic transformation first into tetragonal (t-ZrO2) and then into cubic (c-ZrO2) polymorph takes place.63 SubstitutingZr4þ with aliovalent impurities (such as Ca2þ or Y3þ) induces charge-compensating defects (such as anion vacancies), which due their structuraldistortion can stabilize the metastable t-ZrO2 or c-ZrO2 phases at room tem-perature.64 Especially the cubic phase possesses a remarkable ionic conduc-tivity facilitating its use in fuel cells and high-temperature oxygen sensors.Tetragonal phase may also develop in a nanozirconia (e.g., powders withcrystallite size less than about 30 nm) as a result of excess surface or strainenergies.65,66

The bandgap of m-ZrO2 is around 5.7 eV,67 which corresponds to op-tical transparency down to 220 nm. Nevertheless, even photons with longerwavelength (w280 nm) can excite certain intense broadband PL, usuallyattributed to oxygen vacancies or residual impurities.68 This broadbandPL sometimes shows an oxygen sensitivity,69e71 but the effect is somewhat

0 10 20 30 40 50Time (h)

–1.0

–0.5

0.0

0.5

1.0 ECEC+IPL

Rel

ativ

e er

ror

Figure 9.9 The relative errors of two different calibration functions obtained throughordinary least squares optimization.7 Model “EC” involves only photoconductivity,whereas “ECþIPL” combines conductivity and intrinsic PL using Eq. (9.4). Only thepoints to the left of the red line were used for training, whereas the remaining datapoints served for testing the model. In this experiment, the oxygen concentrationwas varied between 0.21% and 2.1%.


controversial, possibly because several different PL centers contribute to thewide emission spectrum. Moreover, this PL requires a deep UV excitationand exhibits a strong temperature dependence. However, the wide bandgapcan accommodate various optical impurities.

Because of charge difference (and sometimes also size mismatch), triva-lent RE ions do not naturally substitute into MeO2-type crystalline hosts(compared with hosts like Y2O3). The large amount of charge-compensating defects and the phase stability issues may limit the perfor-mance and maximum useable concentration of RE ions for luminescenceapplications. At least in the case of ZrO2:Er

3þ, it has been reported thatniobium (Nb) codopant improves the (upconverted) luminescence perfor-mance as well as stability of the dominant monoclinic phase.72,73 It wasshown that in ZrO2:Eu,Nb, the niobium was incorporated as Nb5þ, at leastclose to the surface.34 Hence, it is believed that a comparable amount of Nbcodopant compensates the charge difference of the RE3þ and Zr4þ ions.Moreover, the ionic radii of Zr4þ, Eu3þ, and Nb5þ are 78, 101, and 69pm, respectively (due to Shannon,74 for coordination number 7, as in m-ZrO2). Hence, there is a chance that Nb5þ also compensates for the latticedistortion induced by the RE3þ ion.

9.3.2 Preparation and characterization of samplesZirconia nanopowders can be prepared by using various solegel routes. Theparticular materials used for the gas sensing experiments were prepared byusing solegel combustion technique, where glycine was used as fuel andnitric acid as the oxidant.72,75 ZrCl4 was dissolved in methanol, whereasNb2O5 and Eu2O3 were dissolved in nitric acid. Appropriate amounts ofthe MOX solutions and glycine were mixed. The resulting mixture wasevaporated on a hot plate while stirring at 90e100�C and concentrated untila gel consistence was obtained. The gel was heated to 300e350�C in anopen oven for 2 h to promote combustion to eliminate the nitric oxide.The obtained black powder was finally annealed at 1200�C for 2 h resultingin a white powdered material. Scanning electron microscopy showedstrongly agglomerated particle-like formations with diameters rangingfrom 200 to 600 nm (Fig. 9.10). Only a submicron porous network isresolved.

X-ray diffraction (XRD) and Raman scattering analyses showed conclu-sively the impact of impurity content on the crystal structure. At low impu-rity concentrations, the material was mostly monoclinic althoughnonnegligible amount (3e6 at%) of the tetragonal phase was also present.


By contrast, 8at% of europium impurity could stabilize the material in thetetragonal phase, leaving no traces of m-ZrO2. However, a similar amountof Nb codopant strongly suppressed the formation of t-ZrO2 so thatm-ZrO2 became again dominant. XRD analysis also established crystallitesizes of about 50 and 25 nm in the monoclinic and stabilized tetragonalphases, respectively. Lower heat treatment temperatures can give smallercrystallites (which can be advantageous for gas sensing), but their phasepurity and luminescence performance are compromised.

Owing to the large bandgap of ZrO2, the PL of Eu3þ can be excitedeither directly, through the O2�/Eu3þ charge transfer absorption band(around 250 nm for m-ZrO2), or the energy transfer after band-to-bandexcitation.76,77 There is one prevalent Eu3þ site in both m-ZrO2 andt-ZrO2 with different symmetry properties, resulting in a clear distinctionof the spectral fine structures. In the oxygen sensing studies, direct excita-tion at 395 nm was used (intra-4f transition 5D0/

7FJ). The obtained PLspectra of Eu3þ correlate well to the dominant crystal structure (Fig. 9.11).

9.3.3 Oxygen sensingSimilarly to TiO2:Sm

3þ, the PL of ZrO2:Eu3þwas also found to be sensitive

to ambient oxygen.34,35 However, the temperature, the content of Nbcodopant, annealing conditions, and even excitation laser intensity all influ-enced the size and sign of the response. Most of the oxygen response studieswere conducted at 300�C because the response was stronger at elevatedtemperatures.

Figure 9.10 Scanning electron microscopy micrographs of the solegel-derived ZrO2:Eu,Nb powder (annealed at 1200�C) at two different magnifications.34


All samples systematically responded to the change in oxygen concentra-tion (Fig. 9.12). The best performance was demonstrated by the sample con-taining 1.48at% Eu, which was overcompensated by Nb (2.74at%). In thiscase, the PL intensity became stronger as oxygen concentration increased.By contrast, the responses of the materials containing only Eu (either 2 or8at%) were reversed. Usually the stable behavior shown in Fig. 9.12 wasrecorded only during a repeated cycle of gas exposures. Fig. 9.13 showsan extended measurement. At the beginning, one can recognize certainslow background process with a characteristic time constant w100 min,which affects the absolute PL intensity. The signal becomes stable afterone to two cycles.

9.3.4 Sensing mechanismSeveral attempts were made to identify the sensing mechanism from the PLdecay kinetics, using either pulsed or modulated laser for excitation. At leastin the case of ZrO2:Eu(1.48at),Nb(2.74at%), it is quite certain that quench-ing of Eu3þ fluorescence by random acceptors are involved, similarly to

T=23°Cλexc=395 nm

Eu 2

Eu 8

Eu 8.68 Nb 8.12

PL

inte

nsity

(arb

. uni

ts)

580 590 600 610 620 630 640 650 660Wavelength (nm)

Figure 9.11 Photoluminescence (PL) emission spectra of solegel-derived ZrO2:Eu andZrO2:Eu,Nb powders annealed at 1200�C. The concentration of dopants (in at%) isshown by the numbers after Eu and Nb. Adapted from Kiisk V, Puust L, M€andar H,Ritslaid P, Bite I, Jankovica D, Sildos I, Jaaniso R. Phase stability and oxygen-sensitivephotoluminescence of ZrO2:Eu,Nb nanopowders. Mater Chem Phys 2018;214:135e42.


TiO2:Sm3þ. As Eu3þ is directly excited, no delayed luminescence is

observed in this case. The effect of changing the atmosphere was quite pro-nounced in the case of pulsed excitation (see the curves taken at 300�C inFig. 9.14). In an oxygen enriched ambient, nearly perfect single exponentialdecay of Eu3þ was observed with a time constant about 1.1 ms. This can beattributed to the natural lifetime of the 5D0 excited state.

λexc=395 nmT=300°C

Eu (2)

Eu (8)

Eu (8.68) Nb (8.12)

Eu (2.12) Nb (1.87)

Eu (1.48) Nb (2.74)

100%

N2

100%

N2

100%

N2

100%

N2

100%

N2

100%

N2

50%

N2

50%

O2

75%

N2

25%

O2

88%

N2

12%

O2

94%

N2

6%O

2

100%

O2

Time (min)

Inte

nsity

0 10 20 30 40 50 60 70 80 90 100 110 120

0,4

0,8

0,6

1,0

0,4

0,8

0,6

1,0

0,4

0,8

0,6

1,0

0,4

0,8

0,6

1,0

0,4

0,8

0,6

1,0

Figure 9.12 Temporal behavior of the Eu3þ PL intensity of the ZrO2:Eu and ZrO2:Eu,Nbpowders in response to the ambient oxygen concentration changes at 300�C.35


Nor

mal

ized

inte

nsity

Time (min)0

0.0

0.2

0.4

0.6

0.8

1.0

50 100 150 200 250 300 350 400

Figure 9.13 Typical behavior of ZrO2:Eu3þ PL intensity during three successive cycles of

oxygen exposure.35 The first two cycles are identical to those in Fig. 9.12 whereas thelast cycle contains alternating exposures to 100% O2 and 100% N2 to test for thestability of the response.

λexc=464 nmλdet=615 nm

O2 (300°C)O2 (23°C)N2 (23°C)N2 (300°C)

0 2000 4000 6000 8000 10000 12000 14000Time (μs)

10–3

10–2

10–1

100

101

PL

inte

nsity

(arb

. u.)

Figure 9.14 Eu3þ PL decay kinetics of the ZrO2:Eu(1.48at),Nb(2.74at%) powder indifferent ambient conditions.34 The intensities of the decay curves are normalized tothe same number of excitation pulses.


For ZrO2:Eu,Nb materials with other impurity concentrations or anneal-ing conditions, the interpretation is less clear. Although the reported PLdecays are qualitatively in agreement with the corresponding stationary PLresponses, several different types of kinetics were observed, of which a fewdistinct variants are depicted in Fig. 9.15. The more strongly curved PLdecays of the uncompensated ZrO2:Eu(2at%) are in agreement with theassumption that oxygen vacancies behave as PL quenching centers, yet thegas response is unexpectedly reversed, compared with ZrO2:Eu(1.48at%),Nb(2.74at%). Even more interestingly, quite a different sensor responsewas observed at high doping levels, where some Eu3þ emitters are effectivelyswitched on or off by the change of ambient gas, resulting in the apparentvertical shift of the PL kinetics. However, it is unlikely that one phosphormaterial can show several completely unrelated gas sensing mechanisms.Assuming that the PL quenching centers and electron donors are oxygenvacancies, there might be an interplay between vacancies in different chargestates and how these vacancies are positioned with respect to Eu3þ sites.

9.4 Tb3D:CePO4

9.4.1 IntroductionOrthophosphates, such as LnPO4 (where Ln represents a trivalent

lanthanide ion), constitute another popular class of host materials for REemitters. RE-rich LnPO4 occurs naturally as varieties of the mineral mona-zite possessing a monoclinic structure. Rhabdophane is a hydrated, hexago-nal form of CePO4, transforming irreversibly to the monazite phase afterheating at about 800�C.78,79 Monazite is thermally stable.80

Nitrogen

Nitrogen

Oxygen

Oxygen

0,0 0,1 0,2

Eu (2)

Eu (8.68) Nb (8.12)1000°CP

L in

tens

ity (a

rb. u

nits

)

0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7Time (ms)Time (ms)

Figure 9.15 Eu3þ PL decay kinetics of a ZrO2:Eu and ZrO2:Eu,Nb powder in 100% O2

and 100% N2, demonstrating different sensing mechanisms as well as the reversal ofthe response. Adapted from Kiisk V, Puust L, M€andar H, Ritslaid P, Bite I, Jankovica D,Sildos I, Jaaniso R. Phase stability and oxygen-sensitive photoluminescence of ZrO2:Eu,Nbnanopowders. Mater Chem Phys 2018;214:135e42.


In practical phosphors, Ce3þ is almost always used as a sensitizer forTb3þ.81 However, in nanocrystalline LnPO4:Ce

3þ,Tb3þ or CePO4:Tb3þ

phosphors, Ce3þ is easily oxidized to Ce4þ, unless the synthesis is speciallyelaborated to protect Ce3þ.82,83 Ce4þ quenches the luminescence ofTb3þ, whereas energy transfer from Ce4þ to Tb3þ is impossible.82,84 Suchkind of disadvantages for conventional phosphors can be favorable for sensorapplications.

9.4.2 Preparation and characterization of samplesDifferently from TiO2:Sm and ZrO2:Eu, the gas sensing experiments withCePO4:Tb

3þ were conducted on a material consisting of rather fine andregularly shaped nanorods.36 The material was prepared by a simple aqueousroute based on the standard Schlenk technique, using CeCl3, TbCl3, andH3PO4 as precursors (note that there are also other aqueous routes resultingin LnPO4 nanorods or -wires

79). The as-prepared product was annealed at300�C in a reducing atmosphere for 2 h to remove the structural water. Theconcentration of Tb was 10at%, where presumably Tb3þ substitutes forCe3þ ions in the crystal lattice.

The synthesis resulted in single-crystalline (rhabdophane-type) nanorodswith a quite regular shape, having a width of 10e20 nm. The size andmorphology resulted in a large surface-to-volume ratio with a specific sur-face area of 176 m2 g�1.

The 4f/5d electronic transitions of Ce3þ result in several overlappingabsorption peaks in the deep UV region. In the particular material, most effi-cient excitation of PL occurred over 250e300 nm (Fig. 9.16). The broad PLband of Ce3þ ion is also located in the UV region, but is quite weak, as mostof the excitation energy is transferred to Tb3þ ions emitting green light(quantum yield as high as 50% for the Tb3þ emission was reported).

9.4.3 Gas sensing and its mechanismGas sensing experiments were conducted under quite aggressive conditions.At 200�C, the material was exposed alternately to highly oxidizing (100%oxygen) or reducing (95% N2/5% H2) atmospheres. When exposed to ox-ygen, the PL consistently decreased to a negligible level with a characteristicresponse time of a few minutes (Fig. 9.17). When exposed to 95% N2/5%H2, the PL recovered in a similar manner. For comparison, microcrystalsof CePO4:Tb prepared by solid-state reaction were also measured, andeven such bulk material showed a measurable PL response (although the ef-fect was very small,w5%). These observations suggest that, due to the small


300 350 400 450 500 550 600 650Wavelength / nm

Inte

nsity

/ a.

u.

200 250 300 350 400Wavelength / nm

5D4-7F6

5D4-7F5

5D4-7F45D4-7F3

Ce(III)

Figure 9.16 Photoluminescence (PL) emission spectrum of CePO4:Tb3þ nanorods. PL

excitation spectrum is shown in the inset. Adapted from Di W, Wang X, Xinguang R.Nanocrystalline CePO4:Tb as a novel oxygen sensing material on the basis of its redoxresponsive reversible luminescence. Nanotechnol 2010;21:075709.

100

80

60

40

20

00 20 40 60 80 100

Time / min

Nor

mal

ized

inte

nsity

/ a.

u.

O2 N2+H2

Figure 9.17 Development of the Tb3þ PL intensity of nanocrystalline CePO4:Tb whenalternately subjected to 100% O2 and 95% N2/5% H2 atmospheres. Adapted from DiW, Wang X, Xinguang R. Nanocrystalline CePO4 :Tb as a novel oxygen sensing material onthe basis of its redox responsive reversible luminescence. Nanotechnol 2010;21:075709.


size of the nanocrystals, the chemical change induced by the ambient gas hasan exhaustive effect on the Tb3þ centers, resulting in the very high contrastobserved in the PL intensity.

In this case, the chemical effect of the ambient gas was explicitly identi-fied using X-ray photoelectron spectroscopy (XPS). XPS spectrum of thesample exposed to oxygen showed clearly an additional peak characteristicof Ce4þ ion. The peaks due to Ce3þ were diminished but remained quitestrong. Hence, the reduced energy transfer from Ce3þ to Tb3þ cannot fullyexplain the drastic suppression of the PL intensity. Measurement of PL decaykinetics showed that the PL of Tb3þ was additionally quenched (Fig. 9.18),probably by the created Ce4þ centers. All described effects were found to bereversible. This gas sensing mechanism assumes direct reaction betweenCe3þ/4þ ions and gas molecules and is therefore limited to rather smallnanocrystals.

Again, the sensor response can be defined based on either PL intensity orPL decay time. For this material, the dependence on oxygen concentrationwas more linear (compared with TiO2:Sm and ZrO2:Eu), making such anoptical sensor appropriate for operation at high oxygen concentrations.

Inte

nsity

(a.u

.)0.1

0.01

1E–30.000 0.005 0.010 0.015 0.020 0.025

Time (s)

abcd

Figure 9.18 Photoluminescence decay kinetics of CePO4:Tb nanocrystals (from the 5D4

level of Tb3þ) for the original sample (a) and those exposed to 40% (b), 80% (c), and100% (d) oxygen atmospheres at 200�C for 5 min, respectively. Solid curves describeexponential (a) or biexponential fits (aec). Adapted from Di W, Wang X, Xinguang R.Nanocrystalline CePO4 :Tb as a novel oxygen sensing material on the basis of its redoxresponsive reversible luminescence. Nanotechnol 2010;21:075709.


9.5 Pr3D:(K0.5Na0.5)NbO3

9.5.1 IntroductionThe recently reported oxygen-sensitive PL of Pr3þ:(K0.5Na0.5)

NbO337 represents an interesting case. First, the material consists of micro-

crystals (rather than nanocrystals). Second, the activator was praseodymiumion (Pr3þ), which is one of the few RE ions exhibiting several emittinglevels. Therefore, the potentially dual optical response can be obtained. Itis common to utilize such emitters for optical temperature sensing wherepopulations of the emitting levels are in thermal equilibrium (this has alsobeen realized for the particular material85), but the approach is novel inthe context of gas sensors.

The host, potassiumesodium niobate, is otherwise known as a promisinglead-free piezoelectric ceramic.86 The optical bandgap of the bulk materialwith perovskite (orthorhombic) structure has been estimated to bew4.3 eV.87 Some synthesis routes result in bandgap values as small as3.2 eV.88

The main levels of Pr3þ-producing emission in the visible range are 3P0and 1D2. The latter gives an emission band at w600 nm due to the1D2/

3H4 transition. A series of emission bands originating from the 3P0level are observed in the visible range, the most prominent atw500 nm be-ing due to the 3P0/

3H4 transition. Ratio of the emissions depends onexcited state dynamics and is determined by the host, excitation route,and concentration of the activators.89

9.5.2 SynthesisThe samples with 0.5at% of Pr3þ were fabricated via a conventional solid-state reaction, which is commonly used for the synthesis of various phos-phors. K2CO3, Na2CO3, Nb2O5, and Pr6O11 were ball-milled with theaddition of alcohol and then calcined in an alumina crucible at 880�C.The obtained material was remilled, mixed with polyvinyl alcohol, andpressed into pellets, which were finally sintered at 1100�C. As a result,(K0.5Na0.5)NbO3 microcrystals (1e4 mm in size) with the orthorhombiccrystal structure and cubic morphology were obtained. The pellets wereadditionally annealed either in argon or oxygen at 950�C.

9.5.3 Oxygen sensingUsing excitation at 325 nm, the PL intensity in 100% N2 and 100% O2

atmospheres at 1 bar was compared.37 Notable change of the PL intensity


was observed only at elevated temperatures, using the material annealedin argon. At 98�C, both 3P0/

3H4 and1D2/

3H4 emissions were present,but only the latter significantly (more than 2�) responded to the switchingof ambient atmosphere. At 165�C, the response of the 1D2/

3H4 emissionbecame even stronger (more than 3�), whereas the 3P0/

3H4 emission wasquenched and probably unusable. Therefore, several detection protocols arefeasible with this material, including dual and ratiometric response.

At 165�C, the 1D2/3H4 emission systematically responded to oxygen

concentration down to 2%. At high oxygen concentrations, the responsewas quite quick (about 1e2 min), but slowed down at lower concentra-tions. The overall behavior resembles that of TiO2:Sm

3þ (see Section9.2). Although PL decay kinetics were not recorded in this case, thepresented data are compatible with the mechanism involving resonant en-ergy transfer to certain defects affected by oxygen adsorption. In particular,the high relative response achieved with micron-sized crystals indicates thatthe effective interaction range is quite large. The thickness of the electrondepletion layer resulting from oxygen adsorption is either comparable tothe penetration depth of the 325 nm excitation light or extendsthroughout the crystal.

The weak response of the 3P0/3H4 emission may indicate that the tran-

sition is not in a good resonance with the energy acceptor (defect). Never-theless, the authors propose the involvement of a secondary sensingmechanism, where oxygen adsorption (through electron trapping) affectsthe position of the intervalence charge transfer state.

9.6 Conclusion

The existing results have convincingly demonstrated that the lumines-cence of trivalent RE ions doped into MOX nano- or even microcrystalscan exhibit a pronounced and systematic response to an ambient gas, wherethe sensing mechanism is fundamentally different from the SterneVolmertype collisional quenching of organic fluorophores. The mechanism is usu-ally based on a fluorescence quenching process of the RE emitter coupled toa gas adsorptionerelated redox process. In some cases, both excitation andemission routes are affected.

Because of this multistage sensing mechanism, the gas concentrationdependence of the response is inherently more complex (albeit in many casesempirically very close to a power law). The response of several oxygen-sensitive nanophosphors (TiO2:Sm

3þ, ZrO2:Eu3þ) is such a function of


oxygen concentration which allows reaching the trace levels (100 ppm) andspanning a wide dynamic range.

Real-world applications necessitate further studies to evaluate andimprove the stability, accuracy, and specificity (including reducing the influ-ence of humidity) and diversify the response (i.e., also detect reactive gasesother than oxygen). Some advancement may stem from proper engineeringof surface morphology and functionalization. On the other hand, materialscontaining multiple RE emitters (such as TiO2:Sm

3þ,Nd3þ) or RE ionsexhibiting several emitting levels (such as Pr3þ) are potentially capable ofthe dual optical response. Moreover, intrinsic luminescence and (photo)conductivity can complement RE luminescence, providing multivariableoutput from a single sensor.

References1. Wang X, Wolfbeis OS. Optical methods for sensing and imaging oxygen: materials,

spectroscopies and applications. Chem Soc Rev 2014;43:3666e761.2. Gregory JW, Sakaue H, Liu T, Sullivan JP. Fast pressure-sensitive paint for flow and

acoustic diagnostics. Annu Rev Fluid Mech 2014;46:303e30. https://doi.org/10.1146/annurev-fluid-010313-141304.

3. Hodgkinson J, Ralph PT. Optical gas sensing: a review. Meas Sci Technol 2013;24:012004.

4. Pallotti DK, Passoni L, Maddalena P, Di Fonzo F, Lettieri S. Photoluminescence mech-anisms in anatase and rutile TiO2. J Phys Chem C 2017;121:9011e21. https://doi.org/10.1021/acs.jpcc.7b00321.

5. Cho B, HahmMG, Choi M, Yoon J, Kim AR, Lee Y-J, Park S-G, Kwon J-D, Kim CS,Song M, Jeong Y, Nam K-S, Lee S, Yoo TJ, Kang CG, Lee BH, Ko HC, Ajayan PM,Kim D-H. Charge-transfer-based gas sensing using atomic-layer MoS2. Sci Rep 2015;5:8052.

6. Zhyrovetsky VM, Popovych DI, Savka SS, Serednytski AS. Nanopowder metal oxidefor photoluminescent gas sensing. Nanoscale Res Lett 2017;12:132. https://doi.org/10.1186/s11671-017-1891-5.

7. Eltermann M, Kiisk V, Lange S, Jaaniso R. Multivariable oxygen sensing based onphotoconductivity and photoluminescence of TiO2 nanoparticles. Sensor Actuator BChem 2019 (in print).

8. Pallotti DK, Passoni L, Gesuele F, Maddalena P, Di Fonzo F, Lettieri S. Giant O2-induced photoluminescence modulation in hierarchical titanium dioxidenanostructures. ACS Sens 2017;2:61e8. https://doi.org/10.1021/acssensors.6b00432.

9. Valerini D, Cretì A, Caricato AP, Lomascolo M, Rella R, Martino M. Optical gassensing through nanostructured ZnO films with different morphologies. Sensor ActuatorB Chem 2010;145:167e73. https://doi.org/10.1016/j.snb.2009.11.064.

10. Sanchez-Valencia JR, Alcaire M, Romero-G�omez P, Macias-Montero M, Aparicio FJ,Borras A, Gonzalez-Elipe AR, Barranco A. Oxygen optical sensing in gas and liquidswith nanostructured ZnO thin films based on exciton emission detection. J PhysChem C 2014;118:9852e9. https://doi.org/10.1021/jp5026027.

11. Faglia G, Baratto C, Sberveglieri G, Zha M, Zappettini A. Adsorption effects of NO2 atppm level on visible photoluminescence response of SnO2 nanobelts. Appl Phys Lett2005;86:011923. https://doi.org/10.1063/1.1849832.


https://doi.org/10.1146/annurev-fluid-010313-141304

https://doi.org/10.1146/annurev-fluid-010313-141304

https://doi.org/10.1021/acs.jpcc.7b00321


https://doi.org/10.1186/s11671-017-1891-5

https://doi.org/10.1186/s11671-017-1891-5



https://doi.org/10.1021/jp5026027

https://doi.org/10.1063/1.1849832

12. Setaro A, Bismuto A, Lettieri S, Maddalena P, Comini E, Bianchi S, Baratto C,Sberveglieri G. Optical sensing of NO2 in tin oxide nanowires at sub-ppm level. SensorActuator B Chem 2008;130:391e5. https://doi.org/10.1016/j.snb.2007.09.015.

13. Liu X, Du B, Sun Y, Yu M, Yin Y, Tang W, Chen C, Sun L, Yang B, Cao W,Ashfold MNR. Sensitive room temperature photoluminescence-based sensing ofH2S with novel CuOeZnO nanorods. ACS Appl Mater Interfaces 2016;8:16379e85.https://doi.org/10.1021/acsami.6b02455.

14. Comini E, Baratto C, Faglia G, Ferroni M, Sberveglieri G. Single crystal ZnO nano-wires as optical and conductometric chemical sensor. J Phys D Appl Phys 2007;40:7255.

15. Aad R, Simic V, Cunff LL, Rocha L, Sallet V, Sartel C, Lusson A, Couteau C,Lerondel G. ZnO nanowires as effective luminescent sensing materials for nitroaromaticderivatives. Nanoscale 2013;5:9176e80. https://doi.org/10.1039/C3NR02416D.

16. Bismuto A, Lettieri S, Maddalena P, Baratto C, Comini E, Faglia G, Sberveglieri G,Zanotti L. Room-temperature gas sensing based on visible photoluminescence proper-ties of metal oxide nanobelts. J Opt A Pure Appl Opt 2006;8:S585e8. https://doi.org/10.1088/1464-4258/8/7/S45.

17. Liu X, Sun Y, Yu M, Yin Y, Yang B, Cao W, Ashfold MNR. Incident fluence depen-dent morphologies, photoluminescence and optical oxygen sensing properties of ZnOnanorods grown by pulsed laser deposition. J Mater Chem C 2015;3:2557e62. https://doi.org/10.1039/C4TC02924K.

18. Gheisi AR, Neygandhi C, Sternig AK, Carrasco E, Marbach H, Thomele D, Diwald O.O2 adsorption dependent photoluminescence emission from metal oxide nanoparticles.Phys Chem Chem Phys 2014;16:23922e9. https://doi.org/10.1039/c4cp03080j.

19. Eliseeva SV, B€unzli J-CG. Lanthanide luminescence for functional materials and bio-sciences. Chem Soc Rev 2009;39:189e227. https://doi.org/10.1039/B905604C.

20. Chambers MD, Clarke DR. Doped oxides for high-temperature luminescence and life-time thermometry. Annu Rev Mater Res 2009;39:325e59. https://doi.org/10.1146/annurev-matsci-112408-125237.

21. Gabriel S, ElhananW. Energy gap law in the solvent isotope effect on radiationless tran-sitions of rare earth ions. J Chem Phys 1975;62:208e13.

22. Nagai H, Noguchi Y. Ambient gas influence on photoluminescence intensity from InPand GaAs cleaved surfaces. Appl Phys Lett 1978;33:312e4. https://doi.org/10.1063/1.90351.

23. Siedl N, Koller D, Sternig AK, Thomele D, Diwald O. Photoluminescence quenchingin compressed MgO nanoparticle systems. Phys Chem Chem Phys 2014;16:8339e45.https://doi.org/10.1039/c3cp54582b.

24. Mochizuki S, Araki H. Reversible photoinduced spectral transition in Eu2O3-gAl2O3composites at room temperature. Phys B Condens Matter 2003;340e342:913e7.

25. Reedo V, Lange S, Kiisk V, Lukner A, Tatte T, Sildos I. Influence of ambient gas on thephotoluminescence of sol-gel derived TiO2:Sm

3þfilms. In: Rosental A, editor. Optical

materials and applications: 4th international conference on advanced optical materials and devices.SPIE; 2005. pp. 59460F 1e6. SPIE 2005 5946 59460F.pdf.

26. Eltermann M, Lange S, Utt K, Joost U, Kink I, Kiisk V, Sildos I. TiO2 nano-powders aspotential low-temperature optical gas sensors. Key Eng Mater 2014;605:368e71.https://doi.org/10.4028/www.scientific.net/KEM.605.368.

27. Eltermann M, Utt K, Lange S, Jaaniso R. Sm3þ doped TiO2 as optical oxygen sensormaterial. Opt Mater 2015;51:24e30. https://doi.org/10.1016/j.optmat.2015.11.020.

28. Eltermann M, Kiisk V, Berholts A, Dolgov L, Lange S, Utt K, Jaaniso R. Modeling ofluminescence-based oxygen sensing by redox-switched energy transfer in nanocrystal-line TiO2:Sm

3þ. Sensor Actuator B Chem 2018;265:556e64.29. Luo W, Li R, Chen X. Host-sensitized luminescence of Nd3þ and Sm3þ ions incorpo-

rated in anatase titania nanocrystals. J Phys Chem C 2009;113:8772e7.



https://doi.org/10.1021/acsami.6b02455

https://doi.org/10.1039/C3NR02416D

https://doi.org/10.1088/1464-4258/8/7/S45

https://doi.org/10.1088/1464-4258/8/7/S45

https://doi.org/10.1039/C4TC02924K

https://doi.org/10.1039/C4TC02924K

https://doi.org/10.1039/c4cp03080j

https://doi.org/10.1039/B905604C

https://doi.org/10.1146/annurev-matsci-112408-125237

https://doi.org/10.1146/annurev-matsci-112408-125237

https://doi.org/10.1063/1.90351

https://doi.org/10.1063/1.90351

https://doi.org/10.1039/c3cp54582b

https://doi.org/10.4028/www.scientific.net/KEM.605.368


30. Chakraborty A, Debnath GH, Saha NR, Chattopadhyay D, Waldeck DH,Mukherjee P. Identifying the correct hosteguest combination to sensitize trivalentlanthanide (guest) luminescence: titanium dioxide nanoparticles as a model hostsystem. J Phys Chem C 2016;120:23870e82. https://doi.org/10.1021/acs.jpcc.6b08421.

31. Almeida NAF, Rodrigues J, Silva P, Emami N, Soares MJ, Monteiro T, Lopes-da-Silva JA, Marques PAAP. Pressure dependent luminescence in titanium dioxide parti-cles modified with europium ions. Sensor Actuator B Chem 2016;234:137e44. https://doi.org/10.1016/j.snb.2016.04.157.

32. Tachikawa T, Ishigaki T, Li J-G, Fujitsuka M, Majima T. Defect-mediated photolumi-nescence dynamics of Eu3þ-doped TiO2 nanocrystals revealed at the single-particle orsingle-aggregate level.Angew Chem Int Ed 2008;47:5348e52. https://doi.org/10.1002/anie.200800528.

33. qojkowski W, Ga1azka K, Opali�nska A, Chudoba T, �Swiderska-�Sroda A, Millers D,Grigorjeva L, Smits K. Method of measuring of oxygen content in gas. 2012. WO2012110967 A1 (EP2686669 A1).

34. Puust L, Kiisk V, Eltermann M, M€andar H, Saar R, Lange S, Sildos I, Dolgov L,Matisen L, Jaaniso R. Effect of ambient oxygen on the photoluminescence of solegel-derived nanocrystalline ZrO2:Eu,Nb. J Phys D Appl Phys 2017;50:215303.https://doi.org/10.1088/1361-6463/aa6c48.

35. Kiisk V, Puust L, M€andar H, Ritslaid P, Bite I, Jankovica D, Sildos I, Jaaniso R. Phasestability and oxygen-sensitive photoluminescence of ZrO2:Eu,Nb nanopowders. MaterChem Phys 2018;214:135e42.

36. Di W, Wang X, Xinguang R. Nanocrystalline CePO4 :Tb as a novel oxygen sensingmaterial on the basis of its redox responsive reversible luminescence. Nanotechnol2010;21:075709.

37. Tang W, Sun Y, Wang S, Du B, Yin Y, Liu X, Yang B, Cao W, Yu M. Pr3þ-Doped(K0.5Na0.5)NbO3 as a high response optical oxygen sensing agent. J Mater Chem C 2016;4:11508e13. https://doi.org/10.1039/C6TC04216C.

38. Weng H, Xu X-Y, Yan B. Novel multi-component photofunctional nanohybrids forratio-dependent oxygen sensing. J Colloid Interface Sci 2017;502:8e15. https://doi.org/10.1016/j.jcis.2017.04.081.

39. Van den Eeckhout K, Smet PF, Poelman D. Persistent luminescence in Eu2þ-dopedcompounds: a review. Materials 2010;3:2536e66.

40. Aydin I, Ertekin K, Demirci S, Gultekin S, Celik E. Sol-gel synthesized Sr4Al14O25:Eu2þ/Dy3þ blueegreen phosphorous as oxygen sensing materials. Opt Mater 2016;62:285e96. https://doi.org/10.1016/j.optmat.2016.10.019.

41. Shannon RD, Pask JA. Kinetics of the anatase-rutile transformation. J Am Ceram Soc1965;48:391e8. https://doi.org/10.1111/j.1151-2916.1965.tb14774.x.

42. Reidy DJ, Holmes JD, Morris MA. The critical size mechanism for the anatase to rutiletransformation in TiO2 and doped-TiO2. J Eur Ceram Soc 2006;26:1527e34.

43. Setiawati E, Kawano K. Stabilization of anatase phase in the rare earth; Eu and Sm iondoped nanoparticle TiO2. J Alloys & Comp 2008;451:293e6.

44. Tang H, Prasad K, Sanjines R, Schmid PE, Lévy F. Electrical and optical properties ofTiO2 anatase thin films. J Appl Phys 1994;75:2042e7.

45. Almquist CB, Biswas P. Role of synthesis method and particle size of nanostructuredTiO2 on its photoactivity. J Catal 2002;212:145e56. https://doi.org/10.1006/jcat.2002.3783.

46. Frindell KL, Bartl MH, Robinson MR, Bazan GC, Popitsch A, Stucky GD. Visible andnear-IR luminescence via energy transfer in rare earth doped mesoporous titania thinfilms with nanocrystalline walls. J Solid State Chem 2003;172:81e8.






https://doi.org/10.1002/anie.200800528

https://doi.org/10.1002/anie.200800528

https://doi.org/10.1088/1361-6463/aa6c48

https://doi.org/10.1039/C6TC04216C





https://doi.org/10.1006/jcat.2002.3783

https://doi.org/10.1006/jcat.2002.3783

47. Dolgov L, Eltermann M, Lange S, Kiisk V, Zhou L, Shi J, Wu M, Jaaniso R. Au/SiO2nanoparticles in TiO2:Sm

3þfilms for improved fluorescence sensing of oxygen. J Mater

Chem C 2017;5:11958e64. https://doi.org/10.1039/C7TC03704J.48. Kiisk V, �Savel M, Reedo V, Lukner A, Sildos I. Anatase-to-rutile phase transition of

samarium-doped powder detected via the luminescence of Sm3þ. Physics Procedia2009;2:527e38.

49. Tang H, Berger H, Schmid PE, Lévy F. Photoluminescence in TiO2 anatase singlecrystals. Solid State Commun 1993;87:847e50.

50. Lange S, Sildos I, Kiisk V, Aarik J. Energy transfer in the photoexcitation of Sm3þ-implanted TiO2 thin films. Mater Sci Eng B 2004;112:87e90.

51. Blumen A, Manz J. On the concentration and time dependence of the energy transferto randomly distributed acceptors. J Chem Phys 1979;71:4694e702. https://doi.org/10.1063/1.438253.

52. Tang H, Lévy F, Berger H, Schmid PE. Urbach tail of anatase TiO2. Phys Rev B 1995;52:7771e4.

53. Yamakata A, Ishibashi T, Onishi H. Time-resolved infrared absorption spectroscopy ofphotogenerated electrons in platinized TiO2 particles. Chem Phys Lett 2001;333:271e7.https://doi.org/10.1016/S0009-2614(00)01374-9.

54. Antila LJ, Santomauro FG, Hammarstr€om L, Fernandes DLA, S�a J. Hunting for theelusive shallow traps in TiO2 anatase. Chem. Commun. 2015;51:10914e6. https://doi.org/10.1039/C5CC02876K.

55. Setiawati E, Kawano K, Tsuboi T, Seo HJ. Studies on thermal migration of Eu iondoped into TiO2 nanoparticles. Jpn J Appl Phys 2008;47:4651.

56. Intrinsic defects of TiO2(110): interaction with chemisorbed O2, H2, CO, and CO2.Phys Rev B 1983;28:3427.

57. Linsebigler AL, Lu G, Yates JT. Photocatalysis on TiO2 surfaces: principles, mecha-nisms, and selected results. Chem Rev 1995;95:735e58.

58. Batzill M, Diebold U. Surface studies of gas sensing metal oxides. Phys Chem Chem Phys2007;9:2307e18. https://doi.org/10.1039/b617710g.

59. Henrich VE. The surfaces of metal oxides. Rep Prog Phys 1985;48:1481.60. Munuera G, Rives-Arnau V, Saucedo A. Photo-adsorption and photo-desorption of

oxygen on highly hydroxylated TiO2 surfaces. Part 1.drole of hydroxyl groups inphoto-adsorption. J Chem Soc, Faraday Trans 1979;1(75):736e47. https://doi.org/10.1039/F19797500736.

61. Yates Jr JT. Photochemistry on TiO2: mechanisms behind the surface chemistry. Surf Sci2009;603:1605e12.

62. Nelson J, Eppler AM, Ballard IM. Photoconductivity and charge trapping in porousnanocrystalline titanium dioxide. J Photochem Photobiol A Chem 2002;148:25e31.

63. Graeve OA. Zirconia. In: Ceramic and glass materials: structure, properties and processing.Springer; 2008. p. 169e97.

64. Fabris S, Paxton AT, Finnis MW. A stabilization mechanism of zirconia based on ox-ygen vacancies only. Acta Mater 2002;50:5171e8. https://doi.org/10.1016/S1359-6454(02)00385-3.

65. Garvie RC. The occurrence of metastable tetragonal zirconia as a crystallite size effect.J Phys Chem 1965;69:1238e43.

66. Djurado E, Bouvier P, Lucazeau G. Crystallite size effect on the tetragonal-monoclinictransition of undoped nanocrystalline zirconia studied by XRD and Ramanspectrometry. J Solid State Chem 2000;149:399e407. https://doi.org/10.1006/jssc.1999.8565.

67. Balog M, Schieber M, Michman M, Patai S. Chemical vapor deposition and character-ization of HfO2 films from organo-hafnium compounds. Thin Solid Films 1977;41:247e59.


https://doi.org/10.1039/C7TC03704J

https://doi.org/10.1063/1.438253

https://doi.org/10.1063/1.438253

https://doi.org/10.1016/S0009-2614(00)01374-9

https://doi.org/10.1039/C5CC02876K

https://doi.org/10.1039/C5CC02876K

https://doi.org/10.1039/b617710g

https://doi.org/10.1039/F19797500736

https://doi.org/10.1039/F19797500736

https://doi.org/10.1016/S1359-6454(02)00385-3

https://doi.org/10.1016/S1359-6454(02)00385-3

https://doi.org/10.1006/jssc.1999.8565

https://doi.org/10.1006/jssc.1999.8565

68. Kiisk V, Puust L, Utt K, Maaroos A, M€andar H, Viviani E, Piccinelli F, Saar R, Joost U,Sildos I. Photo-, thermo- and optically stimulated luminescence of monoclinic zirconia.J Lumin 2016;174:49e55. https://doi.org/10.1016/j.jlumin.2015.12.020.

69. Fidelus JD, Lojkowski W, Millers D, Smits K, Grigorjeva L. Advanced nanocrystallineZrO2 for optical oxygen sensors. In: 8th IEEE conference on sensors; 2009. p. 1268e72.

70. Mochizuki S, Fujishiro F. The photoluminescence properties and reversible photoin-duced spectral change of CeO2 bulk, film and nanocrystals. Phys. Stat. Sol. (b) 2009;246:2320e8. https://doi.org/10.1002/pssb.200844419.

71. Mochizuki S, Saito T. Defect-effects on the photoluminescence of ZrO2 bulk, film andnanocrystals. Phys B Condens Matter 2012;407:2911e4.

72. Smits K, Sarakovskis A, Grigorjeva L, Millers D, Grabis J. The role of Nb in intensityincrease of Er ion upconversion luminescence in zirconia. J Appl Phys 2014;115:213520. https://doi.org/10.1063/1.4882262.

73. Smits K, Olsteins D, Zolotarjovs A, Laganovska K, Millers D, Ignatans R, Grabis J.Doped zirconia phase and luminescence dependence on the nature of chargecompensation. Sci Rep 2017;7:44453. https://doi.org/10.1038/srep44453.

74. Shannon RD. Revised effective ionic radii and systematic studies of interatomic dis-tances in halides and chaleogenides. Acta Crystallogr n.d.;32:751e67.

75. Smits K, Jankovica D, Sarakovskis A, Millers D. Up-conversion luminescence depen-dence on structure in zirconia nanocrystals. Opt Mater 2013;35:462e6.

76. Li L, Yang HK, Moon BK, Choi BC, Jeong JH, Jang K-W, Lee HS, Yi SS. Structure,charge transfer bands and photoluminescence of nanocrystals tetragonal and monoclinicZrO2:Eu. J Nanosci Nanotechnol 2011;11:350e7.

77. Lange S, Kiisk V, Aarik J, KirmM, Sildos I. Luminescence of ZrO2 and HfO2 thin filmsimplanted with Eu and Er ions. Phys Status Solidi 2007;4:938e41.

78. Synthesis and characterization of mixed-morphology CePO4 nanoparticles. J Solid StateChem 2007;180:840e6. https://doi.org/10.1016/j.jssc.2006.12.009.

79. Fang Y-P, Xu A-W, Song R-Q, Zhang H-X, You L-P, Yu JC, Liu H-Q. Systematicsynthesis and characterization of single-crystal lanthanide orthophosphate nanowires.J Am Chem Soc 2003;125:16025e34. https://doi.org/10.1021/ja037280d.

80. Hikichi Y, Nomura T, Tanimura Y, Suzuki S, Miyamoto M. Sintering and propertiesof monazite-type CePO4. J Am Ceram Soc 1990;73:3594e6. https://doi.org/10.1111/j.1151-2916.1990.tb04263.x.

81. Kamiya S, Mizuno H. Phosphors for lamps. In: Yen WM, Shionoya S, Yamamoto H,editors. Phosphor handbook. 2nd ed. CRC Press; 2007.

82. Zhu H, Zhu E, Yang H, Wang L, Jin D, Yao K. High-brightness LaPO4:Ce3þ, Tb3þ

nanophosphors: reductive hydrothermal synthesis and photoluminescent properties.J Am Ceram Soc 2008;91:1682e5. https://doi.org/10.1111/j.1551-2916.2008.02320.x.

83. Buissette V, Moreau M, Gacoin T, Boilot J-P. Luminescent core/shell nanoparticleswith a rhabdophane LnPO4-xH2O structure: stabilization of Ce3þ-dopedcompositions. Adv Funct Mater 2006;16:351e5. https://doi.org/10.1002/adfm.200500285.

84. Riwotzki K, Meyssamy H, Kornowski A, Haase M. Liquid-phase synthesis of dopednanoparticles: Colloids of luminescing LaPO4:Eu and CePO4:Tb particles with a nar-row particle size distribution. J Phys Chem B 2000;104:2824e8. https://doi.org/10.1021/jp993581r.

85. TangW,Wang S, Li Z, Sun Y, Zheng L, Zhang R, Yang B, CaoW, YuM. Ultrahigh-sensitive optical temperature sensing based on ferroelectric Pr3þ-doped (K0.5Na0.5)NbO3. Appl Phys Lett 2016;108:061902. https://doi.org/10.1063/1.4941669.

86. Wu J, Xiao D, Zhu J. Potassiumesodium niobate lead-free piezoelectric materials: past,present, and future of phase boundaries. Chem Rev 2015;115:2559e95. https://doi.org/10.1021/cr5006809.


https://doi.org/10.1016/j.jlumin.2015.12.020

https://doi.org/10.1002/pssb.200844419

https://doi.org/10.1063/1.4882262

https://doi.org/10.1038/srep44453

https://doi.org/10.1016/j.jssc.2006.12.009

https://doi.org/10.1021/ja037280d



https://doi.org/10.1111/j.1551-2916.2008.02320.x



https://doi.org/10.1021/jp993581r

https://doi.org/10.1021/jp993581r

https://doi.org/10.1063/1.4941669

https://doi.org/10.1021/cr5006809

https://doi.org/10.1021/cr5006809

87. Rani J, Patel PK, Adhlakha N, Singh H, Yadav KL, Prakash S. Mo6þ modified(K0.5Na0.5)NbO3 lead free ceramics: structural, electrical and optical properties.J Mater Sci Technol 2014;30:459e65. https://doi.org/10.1016/j.jmst.2013.10.022.

88. Jiang H, Su TT, Gong H, Zhai YC. Direct preparation of K0.5Na0.5NbO3 powders.Cryst Res Technol 2011;46:85e9. https://doi.org/10.1002/crat.201000501.

89. Boutinaud P, Mahiou R, Cavalli E, Bettinelli M. Red luminescence induced by inter-valence charge transfer in Pr3þ-doped compounds. J Lumin 2007;122e123:430e3.https://doi.org/10.1016/j.jlumin.2006.01.198.


https://doi.org/10.1016/j.jmst.2013.10.022

https://doi.org/10.1002/crat.201000501

https://doi.org/10.1016/j.jlumin.2006.01.198


PART THREE

Methods and integration

307j


CHAPTER TEN

Recent progress in siliconcarbide field effect gas sensorsM. Andersson, A. Lloyd Spetz, D. PuglisiLink€oping University, Link€oping, Sweden

Contents

10.1 Introduction 30910.2 Background: transduction and sensing mechanisms 312

10.2.1 Transducer platform 31310.2.2 Transduction mechanisms 31610.2.3 Sensing mechanisms 318

10.2.3.1 General 31810.2.3.2 Detection of hydrogen-containing gases 32010.2.3.3 Detection of nonhydrogen-containing gases 324

10.3 Sensing layer development for improved selectivity of SiC gas sensors 32710.3.1 New material combinations 32710.3.2 Tailor-made sensing layers for oxygen 32810.3.3 Tailoring layers for CO2 and NOx 329

10.4 Dynamic sensor operation and advanced data evaluation 33210.5 Applications 335

10.5.1 Sensor packaging 33510.5.2 Applications and field tests 336

10.6 Summary 338Acknowledgments 339References 339

10.1 Introduction

Chemical sensors based on silicon field effect transistors (Si-FETs) wereintroduced in the 1970s when, first, the ion-sensitive FET for pH measure-ments1 and, in 1974, the hydrogen-sensitive metal oxide semiconductor(MOS) FET2,3 were invented. After more than 40 years of research anddevelopment on chemical gas sensors, today the field effect transistor gassensor based on silicon carbide (SiC-FET) is recognized as the most suitablefor detection of a variety of different gas molecules at operating temperatures



https://doi.org/10.1016/B978-0-08-102559-8.00010-0

from about 200 �C to more than 600 �C.4e8 Regardless of Si or SiC as thesemiconductor in the field effect gas sensor, a catalytically active gate materialsuch as palladium (Pd), platinum (Pt), or iridium (Ir) provides its gas sensitivity.Besides transistors, MOS capacitors and Schottky diodes with catalytic gatecontacts have been developed for gas-sensing purposes, the basic sensingmechanism being common to all the different field effect sensor devices.On exposure of the sensors to a certain substance or gas mixture, the interac-tion between the gas and gate contact changes the electrical field across theMOS structure, in turn modulating the current through, or the capacitanceover, the device.

The introduction of the first metal insulator semiconductor (MIS) gassensor devices based on SiC in the early 1990s9,10 opened up for newapplications of field effect sensors. In 1999, at the International Conferenceof Silicon Carbide and Related Materials in North Carolina, USA, the firstgas sensor based on a SiC-FET was presented,11 and one of the main resultsof this development is devices with excellent long-term stability.7 Thewide bandgap of SiC (3.26 eV for the commonly used polytype 4H)permits operational temperatures beyond the limit of approximately 200�C for Si-based sensors without suffering intrinsic conduction effects.Extending the range of sensor operation temperatures allowed explorationof gas metal interactions and catalytic reactions occurring above 200 �C,facilitating detection of many more compounds. SiC is also chemically inert,preventing device degradation caused by high temperature or reactions withother materials or substances. SiC-based field effect sensors have thereforebeen utilized in high temperature (up to 600 �C) and corrosive applicationssuch as combustion control in car exhausts and small- and medium-scalepower plants,12e15 monitoring of ammonia (NH3) slip from diesel exhaustand flue gas after treatment systems,16,17, as well as for indoor air qualitycontrol.18e20 Commercial sensor systems based on SiC are availablethrough an SME launched in 2007 (SenSiC AB, Kista, Stockholm,Sweden, www.sensic.se). Monitoring the regeneration of nitrogen oxides(NOx) storage catalysts has also been suggested as an application suitable forfield effect sensors based on SiC.21 Olga Casals et al. investigated SiC-basedMIS capacitors with Pt/TaOx gate metal in an atmosphere with highrelative humidity, 45% RH. Detection of 1 part per million (ppm)hydrogen (H2) at 260 �C, 2 ppm carbon monoxide (CO) at 240 �C, and20 ppm ethene (C2H4) at 320 �C in nitrogen (N2) was possible even atthis high humidity level, and variation in the humidity (15%e45%) didnot influence the response. The authors conclude that the SiC sensors

310 M. Andersson et al.

http://www.sensic.se

are especially suitable for monitoring exhaust gases from hydrogen orhydrocarbon-based fuel cells.22

Field effect devices based on other wide bandgap semiconductorsdsuch asdiamond,23 gallium nitride (GaN), and aluminum gallium nitride(AlGaN)24,25dhave also been demonstrated for gas-sensing purposes. Chenet al. fabricated a Schottky diode based on GaN on sapphire substrate witha Pd nanoparticleemodified top layer on the Pd gate contact. The detectionlimit for H2 in air is less than 0.8 ppm at 25 �C; however, there is an influenceof the humidity level at this low temperature.26 Chou et al. fabricated twoPd/AlGaN/GaN Schottky diodes on sapphire, one with pyramid-like Pdnanoparticles on top and one without.27 The pyramid nanoparticlesimproved the sensitivity to H2 with a detection limit of 10 parts per billion(ppb) in air at 27 �C. Guo et al. reported ultralow electrostatic detection oftrinitrotoluene from 0.1 parts per trillion (ppt) to 10 ppb in buffer solutionusing an AlGaN/GaN high electron mobility transistor with gold nanopar-ticles functionalized with cysteamine.28 Offermans et al. demonstratedAlGaN/GaN two-dimensional electron gas (2DEG) devices, which are pro-cessed as suspended membranes on Si29 and operated with ultralow power.Without additional sensing material, detection of nitrogen dioxide (NO2)concentrations between 11 and 20 ppb in single ppb steps is demonstratedat 250 �C with low influence of humidity, and at 275 �C ammonia(1e12 ppm in humid atmosphere) is detected in the opposite direction. Byadding Pt as the gate contact, H2 is detected at 150 �C for a concentrationrange of 300e3000 ppm. When applying a sensing layer of a pH-sensitivepolymer that retains water, CO2 formed charged species in the liquid phaseand could be detected at 25 �C in the range 1000e6000 ppm.

Weng et al. fabricated MISiC capacitors with a gate contact of Pd/TiO2

on top of oxidized SiC (Pd/TiO2/SiO2/SiC). At 325 �C, in a mixture of H2

and oxygen (O2), the response to H2 is lower as compared with the responseto H2 in N2. On the other hand, a mixture of hydrogen sulfide (H2S) and O2

gives a larger response as compared with H2S in N2.30 This is due to the Claus

reaction,31 according to which the oxygen in the presence of titanium dioxide(TiO2) reacts with the sulfur in the H2S molecule and accordingly bothhydrogen atoms are released and may participate in the detection process.

Nakagomi et al. compared Schottky diodes based on the polytype 4H-SiCwith a lowdoped epilayer andb-gallium(III) oxide (b-Ga2O3) with Pt-sensingelectrodes and Ni/Pt or Ti/Al/Pt/Au as the ohmic contacts on the rear side.The Pt-Ga2O3 showed a lower detection limit for hydrogen in oxygen, at400 �C of a few tens ppm of hydrogen. Nonstoichiometry conditions of

Recent progress in silicon carbide field effect gas sensors 311

the Ga2O3 surface or of the Pt-Ga2O3 interface is suggested as the reason forthis.32 In their next paper,33 two devices in series are processed on 1 mmb-Ga2O3 on sapphire, one with Pt gate electrode, the other device withoutgate electrode and with ohmic contacts as above. It is demonstrated that thisdesign allows stable hydrogen detection in oxygen atmosphere from about40 ppm even with temperature fluctuations as large as 150 �C in the temper-ature range 400e550 �C. For lower temperatures, the resistivity of the Ga2O3

is too large and for higher temperatures the Ga2O3 itself is sensitive tohydrogen. The crystal structure b-Ga2O3 is an n-type material, with aband gap equal to 4.9 eV. Thin films of b-Ga2O3 were deposited on topof p-type nickel(II) oxide (NiO) substrates whereby an interfacial layerof g-Ga2O3 was found between the two materials.34

The last decade has seen the development of new devices, new materialcombinations, and new operation modes of SiC-based field effect sensors,and, with the advent of epitaxially grown graphene on SiC, as well as withthe integration of a number of 2D materials for gas sensing applications,35 thefield is expanding even further. Very promising possibilities for ultrasensitivedetection of gaseous compounds are offered by epitaxially grown grapheneon SiC-based sensor structures,36e39 as reported in the first edition of thisbook.40 Nanoparticle decoration of the graphene surface has considerablyimproved selectivity, sensitivity, and speed of response of graphene sensors,while the intrinsic properties of graphene were retained.41e43 This areanow also expands into development of novel 2D materials on SiC otherthan graphene for gas- and liquid-phase sensing applications.44,45

Based on theoretical modeling and material research, selectivity and sensi-tivity toward various gases are currently also being improved by the develop-ment of new combinations of gas-sensitive layers. Recent trends, reviewed inthis chapter, include simplification of device designs to reduce fabrication costsand increase stability, aswell as novel designs to facilitate sensor packaging, highreusability, and thus an efficient product development. Especially high temper-ature applications require advanced packaging solutions and a novel approachusing low temperature cofired ceramic (LTCC) is presented. Dynamic sensoroperation through temperature and gate bias cycling is another recent line ofdevelopment that makes use of advanced data evaluation to enhance stability,as well as selectivity toward certain substances.

10.2 Background: transduction and sensingmechanisms

In this section, the basic physical principles and electrical operation ofthe transducer platform, the FET device, are given. Moreover, a description


of the sensing mechanisms, when the devices are used as gas sensors, is givenin general and for hydrogen- and nonhydrogen-containing gases. For thissection, we also refer to Ref. 7.

10.2.1 Transducer platformThe MIS capacitor represents the heart of most field effect sensor devices,and the physics of MIS capacitors has been widely studied and treated indetail in well-known semiconductor physics and other sensor books.46e48

Here, we will only give the basic physical principles regarding the metalinsulator semiconductor field effect transistor (MISFET), because this isthe ultimate transducer for commercial sensor devices.

MISFET devices may be distinguished in normally off or enhancement typeand normally on or depletion type devices. Normally off means that with zeroapplied gate bias no channel between drain and source is created, whereasnormally on means that a channel already exists at zero applied gate bias.More details can be found in Ref. 49.

A schematic of the enhancement typeMISFETdevice under different con-ditions and its corresponding currentevoltage (I/V) characteristics is shown inFig. 10.1. The channel conductance, determined by its dimensions, themobility of the electrons, and the inversion charge density of electrons canbe modulated by the gate bias, VGS. When no gate bias is applied(VGS ¼ 0), there is no conductive path from source to drain, therefore nocurrent flows through the conducting channel (Fig. 10.1(a)). As soon as agate bias is applied (VGS > 0), the channel (n-type inversion layer) developsallowing electrons to flow between the source and drain terminals in responseto a drain bias (VDS). For a gate bias larger than the threshold voltage VT

(VGS > VT) and small VDS, the device operates in the so-called linear region(Fig. 10.1(b)). As the drain-source voltage increases, the voltage drop acrossthe insulator near the drain terminal decreases. This means that the inducedinversion charge density near the drain decreases, the channel depth (i.e.,the thickness of the inversion channel) near the drain terminal is reduced,and the slope of the I/V curve decreases. The point at which the channeldepth at the drain is reduced to zero is called pinch-off and represents the onsetof saturation (Fig. 10.1(c)). Here, the voltage drop across the insulator at thedrain is equal to the threshold voltage (VDS,sat ¼ VGSeVT). Beyond thepinch-off point, the drain-source current remains constant, resulting in aflat I/V curve (Fig. 10.1(d)). This region is called saturation region.


Transistor-based sensor devices are commonly operated in saturationmode. The drain current (ID,sat) versus gate voltage (VGS) relationship forthe saturation region is described quantitatively by

ID;sat ¼ Wmnεins

2Ldins½VGS � VT �2 (10.1)

Insulator

Insulator

Insulator

Insulator

G

G

G

G

D

D

D

D

S

S

S

S

n+

n+

n+

n+

n+

n+

n+

n+

p-type epilayer

p-type epilayer

p-type epilayer

p-type epilayer

n-type substrate

n-type substrate

n-type substrate

n-type substrate

+VGSVGS < VT

VGS < VT, small VDS

VGS > VT, VDS = VD,sat

VGS > VT, VDS > VDS,sat

+VGS

+VGS

+VGS

+VDS

VDS

VDS

VDS

VDS

VGS2

VGS1

VGS2 VGS1

+VDS

+VDS

+VDS

ID

ID

ID

ID

>

VDS,sat (VGS2)

VDS,sat (VGS1)

(a)

(b)

(c)

(d)

Figure 10.1 Enhancement type metal insulator semiconductor field effect transistor(MISFET) device under different operating conditions and corresponding I/V curves.MISFET operated (a) in equilibrium condition (VGS ¼ 0), (b) in the linear region, (c) atthe onset of saturation, and (d) beyond saturation.


VT ¼ 2dins½eNaεsFF�1=2εins

�Qssdinsεins

þ Fms þ 2FF (10.2)

where W and L are the channel width and length, respectively; mn is thechannel electron mobility; εins and dins are the insulator permittivity andthickness, respectively; VT is the threshold voltage; e is the elementarycharge; Na the bulk doping concentration; εs the semiconductor permit-tivity; QSS the insulator charge density; Fms the metal-to-semiconductorwork function difference; and FF is the Fermi potential, which is thepotential difference between the Fermi level and the intrinsic Fermi level.

Regarding the transistor-based sensor devices, enhancement and depletiontype MISFET transistors are both used. A detailed study of the differencebetween enhancement and depletion type SiC-FET gas sensors can be foundin Ref. 7. However, the depletion type MISFET has the advantages of oper-ation at zero or very low applied gate voltage, less influence of temperaturefluctuations, and generally more stable operation of the sensors, therefore itis preferable as a gas sensor.

Concerning the depletion type MISFET, even when no bias is appliedto the gate terminal (VGS ¼ 0), there is a current flow, i.e., a conductivepath from source to drain exists. The threshold voltage of this device isdefined by the difference between the built-in voltage across the gatemetal/insulator/SiC stack and the pinch-off voltage. The latter is theapplied voltage which cuts off the conducting path between source anddrain and is dependent on the thickness, ds, and the doping level, Nd, ofthe n-type active layer:

Vp ¼ eNdds2εs

(10.3)

Eq. (10.3) shows the pinch off voltage of the depletion type MISFET, εsis the permittivity of the semiconductor.

For a more in-depth treatment of field effect device theory and opera-tion, see, for instance Refs. 7, 46e48.

The design of the device parameters influences the size of the gasresponse. It has been demonstrated, in the case of a SiC-FET with porousIr as the gate material, that a decrease of the gate length from 40 to20 mm results in a factor two increase of the sensor response to CO in 3%oxygen at an operating temperature of 200 �C.7 Other studies showedthat optimizing the thickness of the gate dielectric almost doubled the gasresponse to ammonia. In addition, optimizing the device for lower field


strength between the different terminals of the device increased the long-term performance.

10.2.2 Transduction mechanismsParameters such as device dimensions, electron mobility, permittivity, anddoping concentration are inherent to the choice of materials, the design,and the processing of field effect sensor devices. Once fabricated, the valuesof these are fixed but the charges located in or at the surface of the insulator,QSS, the metal-to-semiconductor work function difference, Fms, and anyinternal gate voltage drop, VGSint, added to the externally applied gatebias, VGSext, can also have an influence on the drain current, ID. Any changein the values of one or more of these parameters will change the I/Vcharacteristics of the FET devices.

Thus, if the interactions between the gas and gate materials on exposureto a certain substance lead to the introduction of an internal gate voltagedrop, a change in gate insulator charge, and/or a change in gate metalwork function, the substance could be detected through a change in draincurrent (see Fig. 10.2). This requires the injection of charge to or chargeseparation at the gate contact/insulator interface, or species capable ofchanging the metal work function to adsorb on the inner surface of thegate contact material. Examples of changes in ID/VGS characteristics for agas-induced internal voltage drop are given in Fig. 10.2(c).

When atoms or molecules adsorb on a surface, there is most often somekind of charge transfer between the adsorbates and the surface, and thus aseparation of charge, as well as a change in work function of the material.One kind of field effectebased gas sensor, the suspended gate FET(SGFET), utilizes this latter phenomenon.50 The design of SGFETs includesa very small air gap between the gate contact material and the insulator, justlarge enough to facilitate rapid diffusion of gas molecules to the gate contactsurface facing the insulator. Any change in drain current, i.e., sensor signal,on gas exposure is directly related to the change in work function of the gatematerial, resulting from adsorption of one or more gaseous substances to itssurface. As mentioned, a common mode of operation of transistor-basedfield effect sensors is to keep the drain current constant and measure theresulting drain-source voltage drop as a sensor signal. Connecting the tran-sistor’s drain and gate (enhancement-type devices) or source and gate(depletion-type devices) terminals, when operating the device as a gassensor, makes it a simple two-terminal device (e.g. Ref. 7). In the other


mode of operation, the drain current is measured as the sensor signal at a con-stant drain-source voltage. The choice of the operation mode, at a constantdrain current or at a constant drain-source voltage, as well as of the electricaloperating point along the currentevoltage (I/V) curve of the device alsoinfluences the size of the gas response. As an example, we demonstratedthat operating a SiC-FET sensor, with porous Ir on top of a dense thinfilm of tungsten trioxide (WO3) as the sensing layer, at a constant drain-source voltage and measuring the drain current as the sensor signal, gave asensor response to 100 ppb benzene which was in the saturation region abouttwice that in the linear region, see Fig. 10.3(a).8 In Fig. 10.3(b), the sameoperationmode is used for a SiC-FETwith a porous Ir gate, and the detection

p-type epilayer p-type epilayer

Insulator Insulator

H2NH3

CONO

O2

HH HHH+ + + + +

_ _ _ _ _O– O–

O–O–O–

O–O–

H2O CO2NH3H2 CONO NO2O2

eVins

eVGS,ext

eVins

EF eΦseΦs

EF

Ev

Ec

Ev

Ec

EFi EFi

MM I IS S

eVGS,int

__ +

+

Gas

Exposure

VGS = 2V

VGS = 5V

VGS = 4V

VGS = 3V

VDS [V]1 2 3 4 5

ID

VGS = 2V

VGS = 3V

VGS = 4V

VGS= 1V

VDS[V]1 2 3 4 5

VGS,ext = VDS ID VGS,ext + VGS,int = VDS

Metal Insulator Insulator SemiconductorSemiconductor Metal

(a)

(b)

(c)

Figure 10.2 (a) Examples of reactions on the catalytic metal gates are displayed, as wellas the effect of hydrogen and oxygen anion adsorption on the number of charge car-riers in the channel. (b) Corresponding changes of the energy band diagram, air/ inertatmosphere to the left and hydrogen exposure to the right and (c) the change in I/Vcharacteristics following hydrogen exposure.


limits for formaldehyde, benzene, and naphthalene are shown as a function ofrelative humidity. For naphthalene, the performance of the gas mixing systemsets the limit to 0.5 ppb.

10.2.3 Sensing mechanisms10.2.3.1 GeneralWork function changes and the creation of internal voltage drops are merelythe general mechanisms behind the conversion of chemical interactions be-tween the gas and the sensor device into an electrical output. Voltage drops

Figure 10.3 (a) Sensor response to 10, 50, and 100 ppb of benzene (C6H6) at 300 �C, indry air, and under operation at the linear (upper signal) and saturation (bottom signal)regions of the transistor. (b) Detection limit as a function of relative humidity forformaldehyde (CH2O), benzene (C6H6), and naphthalene (C10H8). For C10H8, thedetection limit can only be stated to be below 0.5 ppb, because our gas mixing systemcannot provide C10H8 concentrations below 0.5 ppb.


can be introduced and work function changes can be achieved in a numberof different ways. To be useful for specific applications, the sensors must,however, be able to distinguish between different gas mixtures and/or quan-tify one or more substances with good resolution. The sensitivity and selec-tivity toward the substance(s) of interest are important figures of merit for aspecific sensor, as are detection limit, speed of response, and stability.

The sensor’s sensitivity and selectivity to the analyte of interest are largelydetermined by the specific interactions between the various ambient gaseoussubstances and the gate materials exposed to the surrounding gas. Theseinteractions include adsorption and reactions of atoms and molecules onthe surfaces of the gate materials, as well as desorption from the same surfaces.

In general, adsorption and desorption are dependent on, for example, theambient temperature, the partial pressure of the substance, the desorptionenergy, and the sticking coefficient. The sticking coefficient gives the probabil-ity for adsorption of a molecule incident on an empty adsorption site and isdependent on temperature and activation energy for adsorption. It is thereforedifferent for differentmolecules, surface compositions, and crystal orientations.Furthermore, the adsorption of molecules on the sensor surface may be director, via precursor states, it may be dissociative or nondissociative and there maybe interactions between adsorbed species on the surface. All these details of theadsorption will affect the equilibrium state of the molecules on the sensor sur-face. In addition, other constituents of the surrounding gas matrix may adsorbto the surface and affect the coverage of the target substance in different ways(e.g., by reducing or blocking adsorption of this substance or removing it fromthe surface through chemical reactions).

At the steady state, equilibrium usually develops between the adsorption,chemical surface reactions, and desorption of different substances in thesurrounding gas matrix. An overview of the surface processes and examplesof their influence on the device characteristics is given inFig. 10.2. Consideringthe operational temperature of the sensor to be constant, and the sticking co-efficients, interaction, and desorption energies to be inherent to the moleculesand the surface, the steady-state condition on the surface is dependent on thepartial pressures of the gas matrix constituents and, therefore, reflects thecomposition of the surrounding gas. Several different gas matrices may, how-ever, give rise to the same equilibrium surface conditions for a certain surfaceand operational temperature. Conversely, a different surface, or a change inoperational temperature, may give rise to a different equilibrium surface con-dition for the same gas matrix, highlighting the importance and possibilities


regarding the choice of gate material and sensor operational temperature, asalso exemplified below.

10.2.3.2 Detection of hydrogen-containing gasesHydrogen, H2, adsorbs dissociatively on catalytic gate metals such as Pd, Pt,and Ir. In the presence of hydrogen alone, the steady-state surface coverageof hydrogen atoms follows the simple Langmuir relation51,52 and is onlydependent on ambient hydrogen pressure. Normally, also other substancesare present in the surrounding atmosphere and may affect the equilibriumcoverage of hydrogen in different ways. Notably, oxygen also adsorbs disso-ciatively on commonly used gate metals at sensor operating temperatures(200e600 �C)ethe recombination and desorption rates, however, beingvery low below 300 �C. At this temperature, oxygen can basically beremoved at any appreciable rate only through reaction with other atomsor molecules, such as chemisorbed hydrogen in the formation of water. Innormal air, the pressure-dependent hydrogen coverage for every gate mate-rial is determined by the adsorption and desorption characteristics ofhydrogen and its reaction with adsorbed oxygen. Variations in hydrogenand oxygen partial pressures thus lead to a change in hydrogen coverage.

The generated hydrogen atoms are, to some extent, also withdrawn fromthe surface by rapid diffusion through the metal contact to the metal/insu-lator interface. Due to the very rapid diffusion of the hydrogen atoms, thesurface coverage and interface hydrogen concentration are in equilibrium.As concluded from infrared spectroscopy,53 the hydrogen atoms adsorb tooxygen atoms in the surface of oxidic insulators, forming hydroxyl groups(OH) on the oxide accompanied by substantial charge transfer. BecauseOH groups have a large dipole moment, the interface layer of dipoles intro-duces a sharp potential step at the interface, earlier referred to as an internalvoltage drop, Vint. This voltage drop adds to the externally applied bias,resulting in a shift of the I/V characteristics of the sensor, DVint, as illustratedin Fig. 10.2(c), and is given by:

DVint ¼ nH $r

ε0(10.4)

where r is the dipole moment of an OH group, ε0 is the permittivity of freespace, and nH is the number of hydrogen atoms per unit area at the interface,which is related to the coverage of hydrogen on the metal surface.54,55 Thesize of the I/V shift is thus a measure of the ambient partial pressure ofhydrogen, in relation to other gases such as oxygen. A corresponding energy


band diagram illustrating the effect of this dipole layer can be found inFig. 10.2(b).

From a sensor response point of view, it has been shown that dipole forma-tion is the dominant effect regarding hydrogen detection. Work functionchanges due to adsorption on the metal side of the metal/insulator interfaceonly have a minor influence on the sensor signal, introducing a small shift inthe I/V or capacitanceevoltage (C/V) characteristics in the opposite directionto that generated by dipole formation.56 Further evidence for the importanceof an oxidic insulator surface has been obtained from sensors based on bothSiC and GaN Schottky diodes,57,58 for which the hydrogen response consid-erably improved on the introduction of a thin oxide between the metal andthe semiconductor. When comparing the hydrogen response from deviceswith different insulator materials (e.g., Al2O3, Ta2O5, SiO2), the responsecorrelates well with the insulator surface density of oxygen atoms,59 furtheremphasizing the role of the oxygen as adsorption sites for the hydrogen(see Fig. 10.4(a and b)). The choice of insulator thus influences the hydrogensensitivity of field effect devices, as well as their dynamic range. This was alsostudied by Roy et al. for capacitive SiC sensors employing either hafnium(IV)oxide (HfO2) or TiO2 as the dielectric and Ti/Pd as the catalytic contact. Bymultiple linear regression, real-time gas concentration in a mixture of differentgas species could be monitored using different catalytic gate metals anddifferent insulators in a sensor array. This work points out that also defectsin the insulator surface play a role for gas detection.60 Ofrim et al. also usedSiC capacitors as hydrogen sensors utilizing silicon dioxide (SiO2), TiO2,and zinc oxide (ZnO) as the gate insulator, whereby the TiO2-based capac-itive sensor showed superior performance.61

In the case of other molecules containing hydrogen, the same basic princi-ples as for hydrogen apply if free hydrogen atoms can be generated on adsorp-tion. At temperatures of approximately 600 �C or above, field effect sensorswith catalytic metal gates, here Pt, exhibit a binary response to hydrocarbons,irrespective of hydrocarbon identity (see Fig. 10.4(c)).62 As long as the oxygenconcentration is such that complete oxidation of the hydrocarbons can takeplace on the gate metal, the high reaction rates keep the surface fairly cleanfrom hydrocarbons and, to a large extent, oxygen covered. Any hydrocarbonssticking to the surface are oxidized directly on adsorption without generationof any free hydrogen atoms. When increasing the hydrocarbon concentration


beyond the stoichiometric hydrocarbon to oxygen ratio, the hydrocarbonsreduce the gate metal surface and effectively deplete it of oxygen. Dissociation,rather than oxidation, is the dominating process, producing free hydrogenatoms which can reach the interface and induce an internal voltage drop.

Time (min)

Res

pons

e (V

)

Res

pons

e (V

)

Hydrogen concentration (ppm)

SiO

Al O Al OSi NSiOAl O inertSi N inertSiO inertTa O inert

600

500

400

300

200

100

00 1 2 3 4 5 6

α

ΔV (m

V)

T=600 ± 4°C

butane + 0.5% propane/4%O2 butane + 0.5% propane/4%O2 butane + 1% propane/6%O2 butane/1%O2butane/2%O2butane/4%O2ethane/1%O2ethane/2%O2ethane/4%O2

0

0

–100

–100

–200

–200

–300

–300

–400

–400

–500

–500

–600

–600

–700

–700

–800

–800

3% O

10% O

100 150 200 250 300 350 400Temperature (°C)

25 ppm50 ppm75 ppm100 ppm125 ppm150 ppm175 ppm200 ppm225 ppm250 ppm

Si N

Sen

sor r

espo

nse

(mV

)

(a)

(c)

(b)

(d)

Figure 10.4 The figure displays, in (a) and (b), the response to hydrogen in the range10 ppm to 1% in air or N2 (indicated as inert in (b)) at 140 �C for Pd/Pt gate field effectsensors with various insulator materials. In (c) and (d), the response DV to saturatedhydrocarbons at 600 �C as a function of equivalence ratio a are given, as well as theresponse to unsaturated hydrocarbons at temperatures of 100e400 �C and concentra-tions well below the equivalence ratio. The equivalence ratio is defined as the ratio ofthe actual fuel-to-oxidizer ratio to the stoichiometric fuel-to-oxidizer ratio. (a) and(b) are reprinted with permission from the Eriksson M, Salomonsson A, Lundstr€om I, BriandD, Åbom AE. The influence of the insulator surface properties on the hydrogen response offield-effect gas sensors. J Appl Phy 2005;98(3):34903e8 © 2012 American Institute ofPhysics. (c) is reprinted with permission from Baranzahi A, Lloyd Spetz A, Glavmo M,Carlsson C, Nytomt J, Salomonsson P, Jobson E, H€aggendal B, Mårtensson P, Lundstr€om I.Response of metal-oxide-silicon carbide sensors to simulated and real exhaust gases.Sensor Actuator 1997;B43:52e9. © 1997 Elsevier. (d) is reprinted with permission from theAndersson M, Everbrand L, Lloyd Spetz A, Nystr€om T, Nilsson M, Gauffin C, Svensson H. AMISiCFET based gas sensor system for combustion control in small-scale wood fired boilers.Proc IEEE Sensors 2007:962e5. Atlanta, USA Oct 28e31. © 2007 IEEE.


At temperatures below 300 �C, certain hydrocarbonsde.g., unsaturatedhydrocarbons such as ethene (C2H4) and propene (C3H6)dmay still reducethe catalytic metal surface and produce free hydrogen even in the presenceof excess amounts of oxygen.63,64 The underlying reason is the higher stickingprobability of these hydrocarbons compared with oxygen and the lower ratesof oxidation at lower temperatures. However, for decomposition of saturatedhydrocarbons on the catalytic sensor surface in an atmosphere of excessoxygen, all decomposed hydrogen atoms end up as water molecules, whichdesorb from the sensor surface. Therefore, no free hydrogen atoms are gener-ated and no sensor response is obtained from these substances for conditions ofexcess oxygen.64,65 Pt gate sensors operated at 200e300 �C therefore alsoexhibit a binary switch in sensor response to unsaturated hydrocarbons, theswitch point being dependent on oxygen concentration and temperature(due to the temperature dependence of the sticking coefficients) (seeFig. 10.3(d)). Binary switch behavior was also reported by Kahng et al. usinga SiC capacitor with a dense Pt gate for hydrogen sensing in ultrahigh vacuum(UHV).66 The binary behavior is due to the competition between hydrogenoxidation and diffusion to the metal/oxide interface. It was also concludedthat oxygen is needed to restore the sensor baseline after exposure tohydrogen.

Another hydrogen-containing substance which has attracted a great deal ofinterest in the field of high-temperature gas sensors is ammonia (NH3).Ammonia has not been observed to dissociate on adsorption on Pt at temper-atures below approximately 225 �C.67 Furthermore, there is some evidenceof oxygen-mediated dissociation occurring on Pd-MOS sensor devices,68

which leads to direct oxidation of adsorbed NH3. The view that no freehydrogen atoms are generated on the Pt surface accords with the observationsfrom sensors with dense, homogeneous Pt gates, for which no NH3 responseis obtained. In case of a discontinuous/porous gate metal (see Fig. 10.2), whenexposing parts of the oxide to the ambient atmosphere, the field effect devicesexhibit similar sensing characteristics as for hydrogen.69,70

The generally accepted view emphasizes the importance of the threephase boundaries between oxide, metal, and the gas phase as the site forammonia dissociation to create OH groups on the surface of the oxide.71

At the metal/oxide border, hydrogen from an ammonia molecule may bedirectly transferred to oxygen atoms in the surface of the oxide, possiblyas a proton, the charged complex being stabilized by its proximity to themetal. Fourier transform infrared spectroscopic measurements on a modelsystem consisting of a Pt impregnated SiO2 powder revealed the formation


of OH groups at temperatures above 225 �C on exposure to NH3.72 The

amount of OH groups formed correlated well with the Pt loading(coverage), which has been interpreted as the formed OH groups beinglocated close to the metal/oxide border. Local response measurementsperformed on capacitive field effect sensor devices by laterally resolvedphotocurrent measurements provided similar results, relating the generationof OH groups to the metal/oxide border. Furthermore, these investigationsalso indicated the possibility for diffusion of hydrogen/protons into themetal/oxide interface underneath the metal,70,71 inducing the same kindof internal voltage drop as in the case of hydrogen exposure. Hydrogendetection sites underneath the (Pt) metal at the metal/insulator interfacewas systematically studied by Åbom et al. by scratch adhesion measurements,transmission electron microscopy, and atomic force microscopy studies ofripped off metal films.73 The size of the semi-inert hydrogen responseincreased with roughness of the Pt metal surface facing the insulator, whichshowed a blocking effect of Pt metal in direct contact with the insulator(SiO2).

As previously mentioned, oxygen also adsorbs dissociatively on Pt andnegatively charged oxygen atoms may spillover to exposed areas of the oxidesurface in devices with a discontinuous (porous) gate contact. At the steadystate, an equilibrium between oxygen coverage on the Pt surface and concen-tration of oxygen anions on the oxide surface would then develop. It has beensuggested that the response of porous Pt gate sensors to reducing substancessuch as hydrogen, hydrocarbons, and ammonia may partly originate fromthe reverse spillover of oxygen anions and their removal on the Pt surfacethrough reactions with adsorbed hydrogen, hydrocarbon, and ammoniamolecules.74,75 It should be noted that the removal of negative chargesfrom the oxide surface has the same effect on the I/V or C/V characteristicsof field effect sensors as the voltage drop introduced by OH group formation.

10.2.3.3 Detection of nonhydrogen-containing gasesCarbon monoxide (CO) is an example of a reducing, nonhydrogen-containing substance for which the interaction with metal (e.g., Pt) gate fieldeffect sensors may cause a substantial change in the I/V or C/V characteris-tics of a device. Without being able to generate any free hydrogen onadsorption, the CO sensitivity has been stipulated to be at least partly causedby the removal of oxygen anions,74 as discussed above, and/or the reductionof a surface platinum oxide.76e78 The CO response also correlates well withthe CO oxidation characteristics on silica-supported Pt.78,79 At the point


where the oxidation rate suddenly drops when increasing the CO/O2 ratioor decreasing the temperature, the sensor signal exhibits a binary switch froma small to a large response (see also Fig. 10.5).79,80 In analogy with the pre-viously discussed case regarding hydrocarbons, the higher sticking probabil-ity of CO compared with oxygen at lower temperatures leads to the Pt

0–100 –200–300–400–500–600–700

0–100 –200–300–400–500–600–700

100 150 200 250 300 350 400

0 ppm250 ppm500 ppm750 ppm1000 ppm1250 ppm

Temperature (C°C)

CO3% O2

10% O2

0.12

0.06

0.00

0.12

0.06

0.00

2400 2000 1600

Wavenumber (cm–1)

Log

(1/R

) (a.

u.)

20912064 1839

Sen

sor s

igna

l (V

)

Sen

sor s

igna

l (V

)

Par

tial p

ress

ure

(Tor

r)

Par

tial p

ress

ure

(Tor

r)

CO COH H

2,00E–009 2,00E–009

1,00E–009 1,00E–009

0,00E+009 0,00E+009

Addition of 500 ppm H to the CO pulse Addition of 500 ppm H to the CO pulse1,61,4 1,41,2 1,21,0 1,00,8 0,80,6 0,60,4 0,4

125125

250250

500500

375 375625 625750 750875875

100010001125

11251250

1250CO concentration in each test gas pulse (ppm) CO concentration in each test gas pulse (ppm)

5% O

Sen

sor r

espo

nse

(mV

)

10% O

(a)

(c)

(b)

(d)

(a) (a)

(b) (b)

Figure 10.5 In (a), the CO/O2 and temperature-dependent binary switch of theresponse of Pt gate field effect sensors toward CO is exemplified, whereas(b) displays the disruption of the adsorbed CO layer on the Pt surface on hydrogenexposure. The spectral peaks at 1839, 2091, and 2064 cm�1 (upper panel; no H2 expo-sure) correspond to CO adsorbed on Pt, whereas the peaks at wave numbers slightlybelow 2400 cm�1 (lower panel; exposure to 500 ppm H2 in otherwise the same condi-tions as in the upper panel) represents gaseous CO2. In (c) and (d), the sensor responsetoward CO in the range of 125e1250 ppm in the absence/presence of hydrogen(500 ppm) is given for two different oxygen concentrations (lower panels), as well asthe downstream H2 and CO2 partial pressures (upper panel). (a) is reprinted withpermission from the Andersson M, Everbrand L, Lloyd Spetz A, Nystr€om T, Nilsson M,Gauffin C, Svensson H. A MISiCFET based gas sensor system for combustion control insmall-scale wood fired boilers. Proceedings of the IEEE international conference on sensors,Proc IEEE Sensors 2007:962e5. Atlanta, USA Oct 28e31. © 2007 IEEE. (b) is reprinted withpermission from the Becker E, Andersson M, Eriksson M, Lloyd Spetz A, Skoglundh M. Studyof the sensing mechanism towards carbon monoxide of platinum-based field effect sen-sors, IEEE Sens J 2011;11(7):1527e34.© 2011 IEEE. (c) and (d) are reprinted with permissionfrom the Darmastuti Z, Pearce R, Lloyd Spetz A, Andersson M. The influence of gate biasand structure on the CO sensing performance of SiC based field effect sensors. Proc IEEESensors 2011:133e6. Limerick, Ireland, October 28e31, 2011. © 2011 IEEE.79


surface being practically covered with CO (unless the CO/O2 ratio is toosmall), almost excluding oxygen adsorption, also at CO concentrationswell below the oxygen concentration. With no or very little oxygen onthe surface, the CO oxidation rate is very low. At higher temperature orhigher oxygen concentration, the poisoning of the sensor by adsorbedCO on the sensor surface recovers as the CO is removed and the Pt surfacerapidly reverts to being dominated by adsorbed oxygen. A large response ofPt gate field effect devices to CO therefore correlates with a surfacecompletely covered by CO, whereas a small CO response is encounteredwhenever the Pt surface is oxygen dominated. However, not only porousPt gate contacts exhibit these characteristics. Dense films without anyexposed oxide areas also show the same binary switch in sensor signal.81

Furthermore, on introduction of hydrogen at a constant concentration,the large CO response of Pt gate sensors can either increase or decrease,depending on the CO/O2 ratio and temperature (see Fig. 10.5(c and d)).It has also been concluded that the presence of hydrogen can break theself-poisoning of the CO oxidation (see Fig. 10.5(b)).78 This indicates thathydrogen may be able to penetrate/adsorb on a Pt surface covered byCO and, if the CO concentration in relation to the oxygen concentrationis small, disrupt the CO coverage.

If, instead, the CO/O2 ratio is too high in comparison with the hydrogenconcentration, or the Pt surface temperature is too low, the surface remainscovered by CO and, effectively, depleted of oxygen. Without any oxygenon the surface, there is no risk of hydrogen adsorbing on the Pt surface beingoxidized. A much higher proportion of hydrogen atoms can therefore reachthe interface. As a consequence, a CO-covered surface will exhibit a verymuch larger sensitivity detecting even small concentrations of hydrogen,suggesting the CO response partly being mediated through an increased sensi-tivity to the background concentration of hydrogen which is present in all gasmixtures. Further support for the influence of hydrogen on the CO responseis given from UHV studies on Si-based field effect devices.82,83

As exemplified above, the application-specific performance of a sensor isthus influenced by adsorption, reactions between adsorbed species, diffusionof species on the surface, and desorption characteristics of the individual sub-stances which are present in the gas mixture. These characteristics depend onthe materials interacting with the substances, the structure of the materials,and the operating temperature; therefore, the selectivity and sensitivity tothe gases of interest can be influenced by the choice of gate materials, theirstructure, and temperature. For the development of sensors for new


applications, it is therefore important to gain knowledge about gasesolidinteractions and sensor mechanisms to be able to tailor devices with goodselectivity and sensitivity to the target substance(s).

10.3 Sensing layer development for improvedselectivity of SiC gas sensors

The ability of hydrogen atoms to diffuse through the commonly usedgate materials renders most of the field effect sensors developed so far toexhibit sensitivity to hydrogen. In addition, nitride-based insulators have atendency to oxidize over time, providing the necessary sites for hydrogenadsorption.59 In developing sensors for specific applications, the issue ofcross-sensitivity to hydrogen and substances containing hydrogen thereforehas to be considered. For most applications, this cross-sensitivity has been alimitation for the development of field effectebased devices for sensing ofsubstances that do not contain hydrogen, such as oxygen, nitrogen oxides,and sulfur oxides. To widen the areas of application for field effect sensorsby increasing selectivity toward other substances than hydrogen orhydrogen-containing gases, a line of development has been the introductionof new material combinations. However, also the nature of the transducerinfluences the gas response. SiC-FET devices were studied together withquartz crystal microbalance (QMB) sensors which employed the sameporphyrin-based sensing layers. While the SiC-FET device responds to thecharging of the gate introduced by the interaction of gases with the porphyrinlayer, the QMB device responds (changes of the operating frequency) to thechange in total mass of the device due to gas molecules absorbing in thesensing layer.84 Therefore, the combination of the SiC-FET device withthe QMB device gives more information about a certain gas mixture. InSection 10.4, we will introduce temperature cycling operation mode andadvanced data evaluation to improve selectivity and sensitivity of one sensorworking as a virtual sensor array.

10.3.1 New material combinationsFrom theoretical considerations and experimental results, there are indica-tions suggesting that hydrogen terminationdand, thus, OH groupformationdis energetically unfavorable on most magnesium oxide (MgO)surfaces.79,85 It has also been postulated that hydrogen adsorption at theinsulator/metal interface of the MgO/Pt system would occur on the metal,rather than on the insulator side of the interface. Experimental results point


in the same direction, showing that field effect sensors based on MgO/Ptstructures exhibit no or very little response/sensitivity to hydrogen. Thevery small hydrogen-induced response is also in the opposite direction tothe normal hydrogen response of SiO2/Pt structures, as briefly discussedearlier, indicating hydrogen adsorption to the metal side of the interface.56

Furthermore, sensitivity to CO of devices comprising dense Pt gate filmson top of MgO is extremely low or nonexistent, providing further indica-tions for the response to CO of SiO2/Pt-based sensors at least partly beingmediated by an increase in sensitivity toward background hydrogen.

10.3.2 Tailor-made sensing layers for oxygenWith the introduction of MgO as the top part of the insulating layer in fieldeffect sensors, the cross-sensitivity to hydrogen or substances containinghydrogen can thus be markedly reduced. This has also been shown for fieldeffect devices with other gate contacts than Pt. By using conducting oxidesas gate materialdsuch as iridium oxide (IrO2) or ruthenium oxide (RuO2),for which the work function changes as a function of oxidation state86dithas been shown that the sensitivity toward oxygen, and thereby the gas-sensing abilities of field effect sensors, can also be retained when MgO isused as the insulator87 (see Fig. 10.6). This realizes oxygen sensors with

Senso signal (V)

Senso signal (V)

0,30

0,400,350,300,250,200,150,100,050,00

0,25

0,20

0,15

0,10

0,05

0,000 30 60 90 120 150 180

Time (min)

1 2 5 10 15 20% O2

0 30 60 90Time (min)

(a)

(b)

Figure 10.6 This figure displays, in (a), the response of a Pt/IrO2/MgO sensor toward1%, 2%, 5%, 10%, 15%, and 20% O2 in a background of 0.1% O2 in N2 at 500 �C. In(b), the response to 200, 500, and 1000 ppm CO and 100, 250, and 500 ppm propenefor the same sensor is given. Reprinted with permission from the Proceedings of the IEEEInternational Conference on Sensors, Christchurch, New Zealand, October 2009. p. 2031e5.©2009 IEEE (Andersson et al., 2009).


no need for reference gas, unlike the lambda sensor (in the United States,universal heated exhaust gas oxygen, UHEGO).88,89

Partial oxidation or reduction on exposure to different oxygen concen-trations at elevated temperatures changes the work function of the gatematerial at the gate material/insulator boundary and thereby, as discussedin Section 10.2, also the C/V or I/V characteristics of the device. Similarsensors employing ruthenium oxide nanoparticles deposited on SiO2 asgate material, on the other hand, exhibit more or less the same responsecharacteristics to hydrogen and substances containing hydrogen as Pt/SiO2 and Ru/SiO2 structures.

90

10.3.3 Tailoring layers for CO2 and NOx

Not only cross-sensitivity issues have been addressed in the development ofnew sensing materials and material combinations but also possible solutionsfor the detection of substances (e.g., CO2 and NO2), which have not beenpossible to detect with the field effect sensors developed so far, have beeninvestigated. Ion-conducting materials sandwiched between a porous metalgate contact and the insulator have been studied since 2000.91 On exposureof such structures to the target gas (e.g., O2), the target gas adsorbs on themetal gate surface, picking up charges from the metal and thereby formingthe corresponding ions (e.g., by formation of oxygen anions Oe). At thethree phase boundaries between the metal, ion conductor, and gas phase,these ions spillover to, and can be incorporated in, the material at vacantpositions. Most often, but not always, the material is partly composed bythe same atoms/ions as the target gas for detection. In the case of oxygen,the ionic conductor is normally an oxide, such as zirconium oxide(ZrO2), commonly doped by another element, e.g., yttrium, to createmore oxygen vacancies.89

At elevated temperatures, the ions start to becomemobile,moving throughthematerial fromhigh to low concentration by diffusion through vacancies. Asa result, charges are introduced into the electronically nonconducting materialand to the interface between the ion-conducting and -insulating layers,thereby, as described in Section 10.2, changing the C/V or I/V characteristicsof the device. This diffusion is counteracted by the drift due to the electricalfield created between the interface and the gate electrode, the latter beingheld at a constant potential. At equilibrium, the net ion current is zero andthe potential drop across the ion-conducting layer, DV, in simple terms istheoretically given by the Nernst relation (Eq. (10.5)):


DV ¼ VoffsetðTÞ þ�RTnF

�: In

� ½A�surface½A�interface

�(10.5)

where Voffset is the potential difference inherent to the material or materialscombination, R is the molar gas constant, T is the temperature, n is thenumber of electrons transferred per reaction, F is the Faraday constant, and[A]surface and [A]interface are the concentrations of species A at the surface andinterface, respectively. For details, see, for instance.92,93

In conjunctionwith SiC-based field effect sensors, this concept was mainlyused earlier for the detection of oxygen. Examples of ion-conductingmaterialsintroduced between the porous gate contact and the insulator includeZrO2,

94,95 cerium dioxide (CeO2),96 and perovskite compounds such as

barium tin oxide (BaSnO3).97,98 ZrO2 has been shown to work fairly well

for oxygen assessments at high temperatures (600 �C and above), whereasdevices based on the ionic conductor lanthanum trifluoride (LaF3) exhibit agood oxygen sensitivity at lower temperatures, down to room tempera-ture.99,100 Furthermore, the combination of MgO (as insulator) and LaF3 infield effect devices has also given indications on the possibility for oxygensensing with markedly reduced cross-sensitivity to hydrogen or substancescontaining hydrogen. Possibilities for oxygen sensing at lower operating tem-peratures, as compared with the abovementioned MgO/conducting oxidecombinations, have thus been demonstrated.87

Furthermore, the concept has also been utilized for the development offield effect devices sensitive to carbon dioxide (CO2) and nitrogen dioxide(NO2). A solid electrolyte was introduced, which facilitated incorporationof carbonates (CO3

2�) or nitrites/nitrates (NO2�/NO3

�) in theion-conducting layer. In combination with a catalytic gate electrode, fromwhich the carbonates/nitrates are generated on adsorption of CO2/NO2

(through the reaction with adsorbed oxygen anions), field effect sensor de-vices based on the same principles as oxygen sensors can be realized.101,102

In analogy with the oxygen sensors, the transfer of charge, in the form ofCO3

2� or NO3�, from the gate electrode into the electronically noncon-

ducting solid electrolyte, gives rise to a change in the C/V or I/V character-istics of the device. Earlier investigations have also shown promising resultsregarding NO2 sensing with Si-based MIS capacitor structures employingsodium nitrite (NaNO2) as the ion-conducting material.101

For detection of CO2 with a carbonate-based electrolyte, device oper-ating temperatures of at least 400 �C are required. Therefore, SiC-baseddevices are chosen, see Section 10.1, for the development of a SiC field


effect CO2 sensor based on the binary lithium carbonate (Li2CO3)/bariumcarbonate (BaCO3) solid electrolyte (see Fig. 10.7). The binary ionconductor exhibits, in addition to good sensitivity to CO2, an excellentstability also under humid conditions. Fig. 10.7 shows a device with electro-lyte deposited on top of MgO and a highly porous Pt gate electrode withpromising results regarding CO2 monitoring.103 In this case, MgO alsoacts as a passivation layer, preventing lithium ions (Liþ) from diffusinginto the insulating layer during processing and operation of the device.

Perovskites are used as NOx storage materials in catalytic converters fordiesel engine exhaust after treatment.104 Strontium titanate (SrTiO3) hasbeen employed as gate material in SiC-FET devices for NOx detection.Single digit ppm detection was demonstrated between 550 and 600 �C,while at lower temperature, i.e., 530 �C, the response to NOx was some-what lower but compensated by improved selectivity to NH3.

105

Auxiliary layer

Porous Pt sensingelectrode (20nm)

Pt(400 nm)Ti(10 nm)

MgO(100 nm)SiOx(25 nm)Si3N4(25 nm)SiO2(50 nm)

n-SiC epi (5μm)

4H-SiC (n-type)

Ni(100 nm)Ti(10 nm)Pt(400 nm)

Bonding pad

LCR meter

Backsideohmic contact

T = 400°CC = 337 pF

AirAir

1%

10%20%

100 mV

20 min

600

550

500

450

4001 10

CO2 concentration/%

n = 1.5

Air base

Sen

sor s

igna

ls (V

G) m

V

(a) (b)

(c)

102

Figure 10.7 In (a), a schematic drawing of the MISiC field effectebased CO2 sensor witha binary carbonate (Li2CO3-BaCO3) auxiliary layer is given; (b) and (c) show the responsecharacteristics during CO2 exposure at 400 �C for the device in (a).


10.4 Dynamic sensor operation and advanced dataevaluation

To improve selectivity toward certain gaseous substances for whichdetection, discrimination, and quantification otherwise might be difficultdue to interference from other gases, the remedy has often been the introduc-tion of more sensors, each with its own cross-sensitivity pattern. Normally,the combination of sensors and sensitivity patterns is very complicated,involving a large number of different kinds of sensors60 or similar sensorsoperated at different temperatures. The large number of sensor signals andtheir individual cross-sensitivities make necessary to reduce dimensionalityby using multivariate statistical data analysis and pattern recognition methodsto retrieve the desired information.

The most common method to reduce dimensionality is principal compo-nent analysis (PCA).106,107 Multivariate methods such as PCA have, forexample, been used in conjunction with SiC-based field effect devices tomonitor the combustion process in biomass fueled power plants13 for theestimation of ammonia concentration in typical flue gases65 and for fastlambda control of a gasoline engine.108

Another example of a multivariate analysis method is linear discriminantanalysis (LDA).109,110 In analogy with PCA, new variables (discriminantfunctions) are introduced as linear combinations of the original variables.Whereas PCA is an unsupervised method, in LDA the assignment of sensorobservations into predefined groupsee.g., corresponding to concentrationsof a certain target gaseis a prerequisite already when constructing the newvariables. The linear combinations of sensor signals are calculated such thatthe distances between the centers of predefined groups are maximized in thenew projected data set, while minimizing the scatter among observationswithin the different groups. This makes LDA a supervised method.

As was discussed earlier, the interactions between a certain gate materialand the substances of the surrounding gas matrix are temperature-dependent. Different substances show different temperature dependence,which is the reason why operation of a sensor at different temperatures canprovide more information about the gas matrix composition, or the concen-tration of a specific gas in a background of other gases. Instead of an array ofsensors, each of them operated at a different temperature, the operation of onesensor in a cycled temperature operation mode can provide just as much, oreven more, information. In this way, not only the application of more tem-peratures is simplified but there is also the benefit of automatically obtaining


information from nonequilibrium conditions, when changing from one tem-perature to another, aiding in the discrimination between gases and concen-trations. The mean value of the sensor signal at different temperatures, as wellas the derivatives of the signal corresponding to temperature changes, can thenbe extracted and treated by multivariate statistical methods (just as for the caseof signals from many individual sensors). Another advantage related to the useof one sensor as a virtual sensor array includes a reduction of drift problemsand, overall, a better control of the sensor signal and its stability over time.

This approach has been developed using commercial resistive-type MOSsensorsdfor example, for early fire detection in coal mines.111,112 Thisconcept is now also applied to field effect sensor devices based on SiC fordetection and quantification of NO2, SO2, discrimination between differentgases (such as H2, NH3, and CO), and different concentrations for both Ptand Ir gate field effect sensors.113,114,115 It was also possible to discriminatethree different volatile organic compound (VOC) molecules, formaldehyde(50, 100, 150 ppb), benzene (1, 3, 5 ppb), and naphthalene (5, 20, 35 ppb)from each other in a mixture of them in humid air using an Ir-gated SiC-FET and a 1-min temperature cycle.19

Gate bias ramping of SiC-FET devices introduced hysteresis in the sensorsignal, the shapeofwhich revealedmore information about the gasmixture un-der testing.Therefore, gate bias cycling is another alternative for dynamicmodeoperation. The interaction between the various gaseous substances and the gatematerial is not only dependent on their identity and temperature but also on thegate potential. Temperature cycled operation combined with gate bias cyclingimproved the resolution when discriminating and quantifying NO2, CO, andNH3.

116 Mixtures of four gases (NH3, CO, NO, and CH4) at two differentconcentrations (250 and 500 ppm) could be discriminated by employing LDAevaluation.117 Apart from the ambient condition, the shape of the hysteresisvaried also with rate of the bias sweep and, of course, the temperature.This was assumed to show the existence of at least two competing chemicalprocesses taking place on the sensor surface, which are also sensitive to thelevel of the applied gate bias. Fig. 10.8 shows an example of combined tem-perature and bias cycled operation (TCO-GBCO), feature extraction, anddiscrimination of NH3 and CO in a background of dry N2.

Bastuck et al. investigated the complementary effects fromusing bothMOSsensor devices and SiC-FET sensors in advanced operation modes. The MOSsensors were used in TCOmode, with and without a preconcentrator system,and the SiC-FET sensors were operated in a combined temperature cycledoperationegate bias cycled operation (TCO-GBCO) mode.118


The development of two different sensor operating modes may also open-up possibilities regarding self-diagnostic sensor systems.Comparison of the datafrom two independent methodsde.g., temperature and bias cyclingdmayincrease the chances for fault detection and self-diagnosis of the sensor. Inthe event of a sensor malfunction, it is not likely that the outcome of twoseparate evaluation schemes would be similar, the discrepancy between themtherefore indicating problems. The concept has been demonstrated for aresistive-typemetal oxide,MOX, semiconductor sensor utilizing simultaneoustemperature cycling, and electrical impedance spectroscopy measurements.119

6

5

4

3

2

1

0

–1

–2

–30 10 20 30 40 50 60

Time (s)

320

310

300

290

280

270

260

250

240

230Te

mpe

ratu

re (°

C)

400

300

200

100

–2 –1 0 1 2 3 4 5

CO (500 ppm) (300°C)

VGS (V)

VG

S (V

)

I OS (μ

A)

0%

0%

O2–

O2–

0% r.h., 777 mV/s

0% r.h., 777 mV/s

3

2

1

0

–1

–2

–3

–30 –20 –10 0 10First discriminant function (99.89%)

VGS

Tset

Treal

Dry nitrogenNH3 (500 ppm)

CO (500 ppm)

250/300°C

Sec

ond

disc

rimin

ant f

unct

ion

(0.1

1 %

)

(a) (b)

(c)

Figure 10.8 (a) TCO-GBCO combined cycle. (b) Features possible to extract from a hys-teresis curve: maximum horizontal and vertical width (red arrows), maxima positions(red dots), and the enclosed area by the curve (light red). (c) Linear discriminant analysisdiscrimination between the dry N2 background, NH3, and CO. The hysteresis featuresused are horizontal and vertical width and area of ramps from �2 to 5 V, i.e.,777 mVs�1 (9 s) at 250 and 300 �C together with the signal mean value of each cycle(over 0e10 s).117


10.5 Applications

Except for long-term stable sensors, field measurements require suitablepackaging of the sensors and functional electronics. In the following section,we will review important improvement in the packaging of SiC-FET gassensors.

10.5.1 Sensor packagingThe transistor outline (TO) header is an industrial standard that has been usedfor several decades to provide a mechanical basis for the installation of elec-tronic and optical components such as semiconductors and laser diodes, whileat the same time providing power to the components with the aid of pins.The SiC-FET gas sensor research improved considerably when TO headerswere introduced for microelectronic packaging applications. Fig. 10.9 showsa SiC-FET sensor device mounted on a ceramic (Al2O3) substrate, with a thinresistive-type Pt heater wire on the backside, together with a Pt100 temper-ature sensor. The leads of the heater substrate and temperature sensor are spotwelded to a gold-plated 16-pin TO8 header, whereas the sensors’ electricalcontacts are connected to the pins of the TO8 header by gold wire bonding.Such sensor packaging enables operation temperatures even above 600 �C,with good control of temperature and data acquisition. As an example ofhigh temperature applications, TO headers have been used in engine exhaustsystems and in flue gas channels in bioheaters. Over the last years, an innova-tive packaging technology based on LTCC has been developed, improvingthe performance of the SiC-FET sensors and widening the range of possibleapplications, see Fig. 10.9 (top right). This technology is characterized by her-metically sealed modules processed from sheets of unsintered LTCC, which

Figure 10.9 Four-inch diameter SiC wafer with about 2000 sensor chips commerciallyprocessed. Close-up of the wafer shows single transistor devices. After dicing the chipsare mounted in a ceramic package (top right) or in a 16-pin TO8 header (bottom right).


are provided by cavities and vias by laser cutting and electrical contact byscreen printing. The sheets are then stacked and finally sintered in an ovenat 850 �C, which renders a ceramic component.120 Nowak et al. presentedLTCC packaging of a SiC-based hydrogen sensor, which is glued on thescreen-printed contacts.121 Sobocinski et al. demonstrated for the first timea SiC-FET sensor chip introduced in the LTCC stack and cofired in onesingle step to a packaged device, which do not need any glue or bondwires.122 This requires LTCC sheets, which do not shrink in the x-y direction(Hereaus Gmbh) during sintering.123 The electrical and sensing properties ofthe SiC-FET gas sensor are retained after the sintering process at 850 �C.124

10.5.2 Applications and field testsThe outstanding properties, e.g., in terms of long-term stability and high tem-perature performance of the SiCmaterial in gas sensor devices aremanifested ina range of successful applications and field tests. Loloee et al. demonstrated therobustness of SiC-based sensors using the same Pt-SiO2-SiC capacitive devicesfor continuous hydrogen monitoring in a coal gasification plant during 5 daysand, after that, during 20 days in the laboratory.125 One SiC transistor devicehas also been operated in a small bioheater for more than 42 months. Controlof the inlet air to the bioheater by two SiC-FET gas sensors and a temperaturesensor increased the efficiency of the combustion of the wood fuel andconsiderably decreased the emissions of CO and hydrocarbons.7,14 Successfulmonitoring of ammonia in the exhausts of a diesel engine equipped withselective catalytic reduction system was demonstrated already 200516 andthe sensors were successfully tested in two diesel trucks.4

Not only emissions from vehicles and industrial plants are a threat to ourhealth, even indoor environments in private and public buildings need to becontrolled. Indoor air pollution is one of the top five environmental risks topublic health which significantly affect quality of life and economy. The listof top-10 gases in air pollution includes the so-called VOCs, a wide class ofcarbonehydrogen-containing chemicals which are normally found in manyproducts of common use, e.g., tobacco smoke, paints, detergents, glues, con-struction materials, and pressed-wood products. In 2010, the World HealthOrganization (WHO) released guidelines for a range of hazardous VOCs,e.g., formaldehyde, benzene, and naphthalene, which are frequently foundin indoor environments in concentrations of health concern. Formaldehyde,regarded as the most prevalent VOC, is classified as a probable human carcin-ogen with a recommended exposure limit of 81 ppb during 30 min of expo-sure. Benzene is classified as a known human carcinogen at any level ofexposure. Naphthalene is reported as carcinogenic in animal experiments


and a possible human carcinogen, the exposure limit for this substance is set to1.9 ppb as an average annual level.126 Developing sensor systems specificallyselective for such gases at the low ppb or even sub-ppb levels has become amarket demand priority. Recently, it was demonstrated that the SiC-FETdevices detect formaldehyde and naphthalene at concentrations below therecommended exposure limits, i.e., 10 ppb CH2O and below 0.5 ppbC10H8 under 60% RH. Moreover, the SiC-FET device has proven to detectbenzene down to 0.2 ppb in 20% RH and 1e3 ppb in 60% RH, see alsoFig. 10.3.8,18 In the last couple of years, different field test campaigns havebeen carried out in the framework of local collaborations or Europeanprojects (e.g., SENSIndoor, Key-VOCs). As an example, an experiment atan elementary school was carried out during a period of 3 months for specificdetection of formaldehyde. A commercial formaldehyde monitor (FM-801,Graywolf) and a carbon dioxide concentration, temperature, and relativehumidity transmitter (tSense Touch Screen CO2 þ RH/T Transmitter,SenseAir AB) were used as reference instruments. The FM-801 formaldehydemeter provides a measurement range from 20 ppb to 1 ppm and records onedata point every 30 min. By using the SiC-FET-based sensor system indynamic operation mode, continuous monitoring is significantly improvedallowing data point recording approximately every minute (depending onthe temperature cycle used). Fig. 10.10 displays a temperature cycle of 80 s(four temperatures) and extraction of virtual sensor signals in the school tests.20

Figure 10.10 Temperature cycle (blue, solid line), temperature (blue, dashed line), andthe raw sensor signal (black line) during one cycle. The mean of four different areasis computed and the middle of each one, marked by a colored dot, can be regardedas a virtual sensor (left panel). In the right panel, the data of the virtual sensors isextracted from the raw sensor signal.


In Fig. 10.11, measurements from 11 days are shown. Using a multivariateregression model based on partial least squares regression (PLSR) on the sensordata, the experiment demonstrated a very good correlation between the SiC-FET sensor and the FM-801 meter. The formaldehyde builds up at night andduring weekends while the ventilation is switched off. The highest peak forthe reference instrument was 34 ppb (August 21), while the computedPLSR1 signal from the SiC-FET sensor data had a peak of 24.5 ppb (August28). In summary, the formaldehyde always stayed well below the thresholdvalue of 80 ppb. The data evaluation also revealed some possible cross-sensitivity of the SiC-FET sensor to other common VOCs that are emittedby breath (e.g., acetone and isoprene), which is an area to be furtherinvestigated.20

10.6 Summary

The SiC-FET devices as high-temperature gas sensors are commer-cially available in sensor systems for combustion control, e.g., in small-and medium-scale power plants. Research and development has realizedtailor-made sensing layers for, e.g., oxygen and carbon dioxide detection.Detection of toxic indoor gases, VOCs, below legally restricted levels hasbeen demonstrated. Temperature and bias cycled operation modes together

CO2 conc

r.h.

Amb. temp.

Form. conc.

PLSR1

Ventilation

Aug 20 Aug 21 Aug 22 Aug 23 Aug 24 Aug 25 Aug 26 Aug 27 Aug 28 Aug 29 Aug 302016

Figure 10.11 Sensors signals from reference instruments from a period of 11 days,standardized and shifted for visualization. The virtual sensors have been used to buildthe PLSR1 signal (red), smoothed with a window size of 22 (w30 min) for visualization.The bottom blue line represents the normal schedule of the ventilation of the school.The start of each day, i.e., midnight, is marked on the x-axis, and night from 6 p.m. to 6a.m. is marked by darker areas. Note that Aug 20/21 and Aug 27/28 there is no school(weekends).


with advanced data evaluation based on multivariate statistics improvedselectivity and sensitivity in complex gas mixtures. Field test campaignshave demonstrated the suitability of using the SiC-FET sensor as a selectiveformaldehyde sensor.

AcknowledgmentsGrants are acknowledged from the VINN Excellence Center in research and innovation onFunctional Nanoscale Materials (FunMat), the Swedish Governmental Agency for Innova-tion Systems (VINNOVA #621-2012-4497), and the Swedish Research Council (VR#621-2012-4497). The authors also acknowledge funding from the European Union’s Sev-enth Programme for research, technological development, and demonstration under grantagreement No. 604311 (SENSIndoor), and from the COST Action TD1105 (EuNetAir).A.L.S. acknowledges the Swedish Government Strategic Research Area in Materials Scienceon Functional Materials at Link€oping University (Faculty Grant SFO-Mat-LiU No. 2009-00971).Dr Ruth Pearce is acknowledged for the contribution in the first edition of this book chap-ter.40 The epitaxial graphene sensor area has grown into an independent research area, exem-plified in the introduction.

References1. Bergveld P. Development of an ion sensitive solid state device for neuro-physiological

measurements. IEEE Trans Biomed Eng 1970;17:70e1.2. Lundstr€om I, Shivaraman S, Svensson C, Lundkvist L. A hydrogen-sensitive MOS

field-effect transistor. Appl Pys Lett 1975;26(2):55e7.3. Lundstr€om KI, Shivaraman MS, Svensson CM. A hydrogen-sensitive Pd-gate MOS

transistor. J Appl Phys 1975;46:3876e81.4. Lundstr€om I, Sundgren H, Winquist F, Eriksson M, Krantz-R€ulcker C, Lloyd-

Spetz A. Twenty-five years of field effect gas sensor research in Link€oping. Sensor Actu-ator B 2007;121:247e62.

5. Trinchi A, Kandasamy S, Wlodarski W. High temperature field effect hydrogen andhydrocarbon gas sensors based on SiCMOS devices. Sensor Actuator B Chem 2008;133:705e16.

6. Lloyd Spetz A, Skoglundh M, Ojam€ae L. FET gas-sensing mechanism, experimentaland theoretical studies. ch. 4. In: Comini E, Faglia G, Sberveglieri G, editors. Solid stategas sensing, New York. Norwell MA, USA: Springer; 2009, ISBN 978-0-387-09664-3.p. 153e79.

7. Andersson M, Pearce R, Lloyd Spetz A. New generation SiC based field effect tran-sistor gas sensors. Sensor Actuator B Chem 2013;179:95e106. https://doi.org/10.1016/j.snb.2012.12.059.

8. Puglisi D, Eriksson J, Andersson M, Huotari J, Bastuck M, Bur C, Lappalainen J,Schuetze A, Lloyd Spetz A. Exploring the gas sensing performance of catalyticmetal/metal oxide 4H-SiC field effect transistors. Mater Sci Forum 2016;858:997e1000.

9. Hunter GW, Neudeck P, Jefferson GD, Madzar GC, Liu CC, Wu QH. The devel-opment of hydrogen sensor technology at NASA Lewis research center. In: Proceedingsof the 4th annual space system health management technology conference, Cincinatti, USA;1992.




10. Spetz A, Arbab A, Lundstr€om I. Gas sensors for high temperature operation based onmetal oxide silicon carbide (MOSiC) devices. In: Proceedings of Eurosensors VI, SanSebastian, Spain; 1992. p. 19e23.

11. Svenningstorp H, Unéus L, Tobias P, Lundstr€om I, Ekedahl L-G, Lloyd Spetz A. Hightemperature gas sensors based on catalytic metal field effect transistors. Mater Sci Forum2000;338e342:1435e8.

12. Wingbrant H, Svenningstorp H, Salomonsson P, Tengstr€om P, Lundstr€om I, LloydSpetz A. Using a MISiCFET device as a cold start sensor. Sensor Actuator B 2003;93:295e303.

13. Unéus L, Artursson T, Mattsson M, Ljung P, Wigren R, Martensson P, Holmberg M,Lundstr€om I, Lloyd Spetz A. Evaluation of on-line flue gas measurements by MISiC-FET and metal oxide sensors in boilers. IEEE Sens J 2005;5(1):75e81.

14. Andersson M, Everbrand L, Lloyd Spetz A, Nystr€om T, Nilsson M, Gauffin C,Svensson H. A MISiCFET based gas sensor system for combustion control in small-scale wood fired boilers. Proc IEEE Sensors 2007:962e5. Atlanta, USA Oct 28e31.

15. Andersson M. SiC based field effect sensors e from basic understanding to tailoring ofdevices for high temperature applications IX International workshop on semicon-ductor gas sensors. SGS 2015. 13e16 December, Zakopane, Poland, 2015.

16. Wingbrant H, Svenningstorp H, Salomonsson P, Kubinski D, Visser JH, L€ofdahl M,Lloyd Spetz A. Using a MISiC-FET sensor for detecting NH3 in SCR systems. IEEESens J 2005;5(5):1099e105.

17. Andersson M, Wingbrant H, Petersson H, Unéus L, Svenningstorp H, L€ofdahl M,Holmberg M, Lloyd Spetz A. Gas sensor arrays for combustion control. In:Ranch S, editor. Encyclopedia of sensors, vol. 4. USA: American Scientific Publishers;2006. p. 139e54.

18. Puglisi D, Eriksson J, Bur C, Schuetze A, Lloyd Spetz A, Andersson M. Catalyticmetal-gate field effect transistors based on SiC for indoor air quality control. J SensSens Syst (JSSS) 2015;4:1e8.

19. Bur C, Bastuck M, Puglisi D, Sch€utze A, Lloyd Spetz A, Andersson M. Discriminationand quantification of volatile organic compunds in the ppb-range with gas sensitiveSiC-FETs using multivariate statistics. Sensor Actuator B Chem 2015;214:225e33.

20. Bastuck M, Puglisi D, M€oller M, Reimringer W, Sch€utze A, Lloyd Spetz A,Andersson M. Low-cost chemical gas sensors for selective formaldehyde quantificationat ppb-level in the field. In: Proc. AMA Conf. 2017 sensor 2017 and IRS2; 2017.p. 702e7.

21. Schalwig J, Ahlers S, Kreisl P, Bosch v, Braunm€uhl C, M€uller G. A solid-state gassensor array for monitoring NOx storage catalytic converters. Sensor Actuator B2004;101:63e71.

22. Casals O, Becker T, Godignon P, Romano-Rodriguez A. SiC-based MIS gas sensorfor high water vapor environments. Sensor Actuator B Chem 2012;175:60e6.

23. Gurbuz Y, Kang WP, Davidson JL, Kerns DV. High-temperature tolerant diamonddiode for carbon monoxide gas detection. J Appl Phys 1998;84:6935e6.

24. Schalwig J, Kreisl P, Ahlers S, M€uller G. Response mechanism of SiC-based MOSfield-effect gas sensors. IEEE Sens J 2002;2(5):394e402.

25. Eickhoff M, Schalwig J, Weidemann O, G€orgens L, Neuberger R, Hermann M,Steinhoff G, Baur B, M€uller G, Ambacher O, Stutzmann M. Electronics and sensorsbased on pyroelectric Al-GaN/GaN heterostructures. Phys Status Solidi C: Conf 2003;0(6):1908e18.

26. Chen H-I, Cheng C-H, Chen W-C, Liu I-P, Lin K-W, Liu W-C. Hydrogen sensingperformance of a Pd nanoparticle/Pd film/GaN-based diode. Sensor Actuator B Chem2017;247:514e9.


27. Chou P-C, Chen H-I, Liu I-P, Chen W-C, Chen C-C, Liou J-K, Lai C-J, Liu W-C.On a Schottky diode-type hydrogen sensor with pyramid-like Pd nanostructures. Int Jof Hydrogen Energy 2015;40:9006e12.

28. Guo Y, Wang X, Miao B, Li Y, Yao W, Xie Y, Li J, Wu D, Pei R. An AuNPs-functionalized AlGaN/GaN high electron mobility transistor sensor for ultrasensitivedetection of TNT. RSC Adv 2015;5:98724e9.

29. Offermans P, Si-Ali A, Brom-Verheyden G, Greens K, Lenci S, van Hove M,Decoutere S, Van Schaijk R. Suspended AlGaN/GaN membrane devices withrecessed open gate areas for ultra-low-power ari quality monitoring. Conf Proc IEEESensor 2015;15:871e4. IEDM.

30. Weng MH, Mahapatra R, Wright NG, Horsfall AB. Role of oxygen in high temper-ature hydrogen sulfide detection using MISiC sensors. Meas Sci Technol 2008;19:024002 (5 pp.).

31. Davydov AA, Marshneva VI, Shepotko ML. Metal oxides in hydrogen sulphideoxidation by oxygen and sulphur dioxide I. The comparison study of the catalytic ac-tivity. Mechanism aof the interactions between H2S and SO2 on some oxides. ApplCatal A 2003;244:93e100.

32. Nakagomi S, Ikeda M, Kokubun Y. Comparison of hydrogen sensing properties ofSchottky diodes based on SiC and b-Ga2O3 single crystal. Sens Lett 2011;9(2):616e20.

33. Nakagomi S, Sai T, Kokubun Y. Hydrogen gas sensor with self temperature compen-sation based on b-Ga2O3 thin film. Sensor Actuator B 2013;187:413e9.

34. Nakagomi S, Kubo S, Kokubun Y. The orientational relationship between mono-clinic b-Ga2O3 and cubic NiO. J Cryst Growth 2016;445:73e7.

35. Yakimova R, Virojanadara C, Gogova D, Syv€aj€arvi M, Siche D, Larsson K,Johansson LI. Analysis of the formation conditions for large area epitaxial grapheneon SiC substrates. Mater Sci Forum 2010;645e648:565e8.

36. Pearce R, Yakimova R, Eriksson J, Hultman L, Andersson M, Lloyd Spetz A.Development of FETs and resistive devices based on epitaxially grown single layer gra-phene on SiC for highly sensitive gas detection. Mater Sci Forum 2012;717e720:680e90.

37. Pearce R, Eriksson J, Iakimov T, Hultman L, Lloyd Spetz A, Yakimova R. On thediffering sensitivity to chemical gating of single and double layer epitaxial grapheneexplored using scanning kelvin probe microscopy. ACS Nano 2013;7(5):4647e56.

38. Eriksson J, Pearce R, Iakimov T, Vironajadara C, Gogova D, Andersson M,Syv€aj€arvi M, Lloyd Spetz A, Yakimova R. The influence of substrate morphologyon thickness uniformity and unintentional doping of epitaxial graphene on SiC.Appl Phys Lett 2012;100:241607.

39. Eriksson J, Puglisi D, Vasiliauskas R, Lloyd Spetz A, Yakimova R. Thickness unifor-mity and electron doping in epitaxial graphene on SiC. Mater Sci Forum 2013;740e742:153e6.

40. Andersson M, Lloyd Spetz A, Pearce R. Recent trends in silicon carbide (SiC) and gra-phene based gas sensors. In: Janiso R, Tan OK, editors. Semiconductor gas sensors. 1st ed.Philadelphia: Woodhead publishing; 2013. p. 117e58.

41. Eriksson J, Puglisi D, Kang Y-H, Yakimova R, Lloyd Spetz A. Adjusting the elec-tronic properties and gas reactivity of epitaxial graphene by thin sufacemetallization. Physica B 2014;439:105e8.

42. Eriksson J, Puglisi D, Strandqvist C, Gunnarsson R, Ekeroth S, Ivanov IG,Helmersson U, Uvdal K, Yakimova R, Lloyd Spetz A. Modified epitaxial grapheneon SiC for extremely sensitive and selective gas sensors. Mater Sci Forum 2016;858:1145e8.


43. Rodner M, Puglisi D, Yakimova R, Eriksson J. A platform for extremely sensitive gassensing: 2D materials on silicon carbide. In: TechConnect briefs. Materials for energy, effi-ciency and sustainability, 2; 2018, ISBN 978-0-9988782-3-2. p. 101e4.

44. Shtepliuk I, Eriksson J, Khranovskyy V, Iakimov T, Lloyd Spetz A, Yakimova R.Monolayer graphene/SiC Schottky barrier diodes with improved barrier height uni-formity as a sensing platform for the detection of heavy metals. Beilstein J Nanotechnol2016;7:1800e14.

45. Rodner M, Bahonjic J, Mathisen M, Ekeroth S, Helmersson U, Ivanov IG,Yakimova R, Eriksson J. Performance tuning of gas sensors based on epitaxialgraphene on silicon carbide. Mater Des 2018;153:153e8.

46. Sze SM. Physics of semiconductor devices. New York: John Wiley & Sons; 1981.47. Lundstr€om I. Field effect chemical gas sensors, In: G€opel W, Hesse J, Zemel J.N,

editors. Sensors A comprehensive survey, 1991;2(1). Wienheim: VCH VerlagsgesellschaftmbH p. 467-528.

48. Dimitrijev S. Understanding semiconductor devices. New York: Oxford University Press;2000.

49. Bur C. Selectivity enhancement of gas sensitive field effect transistors by dynamicoperation. Link€oping Stud Sci Technol 2015:24e31. Dissertation No. 1644. ISBN:978-91-7519-119-5.

50. Senft C, Galonska T, Widanarto W, Eisele I, Frerichs HP, Wilberts M. Stability ofFET-based hydrogen sensors at high temperatures. Proc IEEE Sensors Conf 2007:189e92. 4388368.

51. Fogelberg J, Eriksson M, Dannetun H, Petersson L-G. Kinetic modeling of hydrogenad/absorption in thin films on hydrogen sensitive field effect devices: observation oflarge hydrogen dipoles at the Pd-SiO2 interface. J Appl Phys 1995;78(2):988e96.

52. Ekedahl L-G, Eriksson M, Lundstr€om I. Hydrogen sensing mechanisms of metal-insulator interfaces. Acc Chem Res 1998;31(5):249.

53. Wallin M, Gr€onbeck H, Lloyd Spetz A, Eriksson M, Skoglundh M. Vibrational anal-ysis of H2 and D2 adsorption on Pt/SiO2. J Phys Chem B 2005;109:9581e8.

54. ErikssonM, Lundstr€om I, Ekedahl L-G. A model of the Temkin isotherm behavior forhydrogen adsorption at Pd-SiO2 interfaces. J Appl Phys 1997;82(6):3143e6.

55. Usagawa T, Kikuchi Y. Device characteristics for Pt-Ti-O gate Si-MISFETs hydrogengas sensors. Sensor Actuator B 2011;160(1):105e14.

56. Eriksson M, Ekedahl L-G. Hydrogen adsorption states at the Pd/SiO2 interface andsimulation of the response of a Pd metal-oxide-semiconductor hydrogen sensor.J Appl Phys 1998;83(8):3947e51.

57. Schalwig J, M€uller G, Eickhoff M, Ambacher O, Stutzmann M. Group III-nitridebased gas sensors for combustion monitoring. Mater Sci Eng B 2002;93:207e14.

58. Weidemann O, Hermann M, Steinhoff G, Wingbrant H, Lloyd Spetz A,Stutzmann M, Eickhoff M. Influence of surface oxides on hydrogen-sensitivePd-GaN Schottky diodes. Appl Phys Lett 2003;83(4):773e5.

59. Eriksson M, Salomonsson A, Lundstr€om I, Briand D, Åbom AE. The influence of theinsulator surface properties on the hydrogen response of field-effect gas sensors. J ApplPhys 2005;98(3):34903e8.

60. Roy SK, Furnival BJD, Wood NG, Vassilevski KV, Wright NG, Horsfakk AB,O�Malley CJ. SiC gas sensor array for extreme environment. Proc IEEE Sensors2013:250e3. 978-1-4673-4642-9/13.

61. Ofrim B, Udrea F, Brezeanu G, Hsieh AP-S. Hydrogen sensor base on silicon carbid(SiC) MOS capacitor. Proc IEEE Sensors 2012:367e70.

62. Baranzahi A, Lloyd Spetz A, Glavmo M, Carlsson C, Nytomt J, Salomonsson P,Jobson E, H€aggendal B, Mårtensson P, Lundstr€om I. Response of metal-oxide-siliconcarbide sensors to simulated and real exhaust gases. Sensor Actuator 1997;B43:52e9.


63. Burch R,Watling TC. Kinetics and mechanism of the reduction of NO by C3H8 overPt/Al2O3 under lean-burn conditions. J Catal 1997;169:45e54.

64. Burch R, Watling TC. The effect of sulphur on the reduction of NO by C3H6 andC3H8 over Pt/Al2O3 under lean-burn conditions. Appl Catal, B 1998;17:131e9.

65. Andersson M, Ljung P, Mattsson M, L€ofdahl M, Lloyd Spetz A. Investigations on thepossibilities of a MISiCFET sensor system for OBD and combustion control utilizingdifferent catalytic gate materials. Top Catal 2004;30/31:365e8.

66. Kahng YH, Lu W, Tobin RG, Loloee R, Ghosh RN. The role of oxygen inhydrogen sensing by a platinum-gate silicon carbide sensor: an ultrahigh vacuumstudy. J Appl Phys 2009;105:064511.

67. Chilton TH. The manufacture of nitric acid by the oxidation of ammonia. In:Chemicalengineering progress monograph series, vol. 3. New York: American Institute of ChemicalEngineers; 1960. 56.

68. Fogelberg J, Lundstr€om I, Petersson L-G. Ammonia dissociation on oxygen coveredpalladium studied with a hydrogen sensitive Pd-MOS device. Phys Scripta 1987;35:702e5.

69. Spetz A, Armgarth M, Lundstr€om I. Optimization of ammonia-sensitive structureswith platinum gates. Sensor Actuator 1987;11:349e65.

70. L€ofdahl M, Utaiwasin C, Carlsson A, Lundstr€om I, Eriksson M. Gas responsedependence on metal gate morphology of field-effect devices. Sensor Actuator B2001;80:183e92.

71. L€ofdahl M. Spatially resolved gas sensing, Link€oping studies in science and technology.Sweden: Link€oping; 2001. p. 115e27. Dissertation no. 696.

72. Wallin M, Byberg M, Gr€onbeck H, Skoglundh M, Eriksson M, Lloyd Spetz A.Vibrational analysis of H2 and NH3 adsorption on Pt/SiO2 and Ir/SiO2 model sensors.In: Proceedings of IEEE international conference on sensors, Atlanta, USA; 2007. p. 1315e7.

73. Åbom AE, Haasch RT, Hellgren N, Finnegan N, Hultman L, Eriksson M. Character-ization of the metaleinsulator interface of field-effect chemical sensors. J Appl Phys2003;93(12):9760e8.

74. Schalwig J, M€uller G, Karrer U, Eickhoff M, Ambacher O, Stutzmann M, G€orgens L,Dollinger G. Hydrogen response mechanism of Pt-GaN Schottky diodes. Appl PhysLett 2002;80(7):1222e4.

75. Yamaguchi T, Kiwa T, Tsukada K, Yokosawa K. Oxygen interference mechanism ofplatinum-FET hydrogen gas sensor. Sensor Actuator 2007;136(1):244e8.

76. Dean VW, Frenklach M, Phillips J. Catalytic etching of platinum foils and thin films inhydrogen-oxygen mixtures. J Phys Chem 1988;92:5731e8.

77. Nakagomi S, Tobias P, Baranzahi A, Lundstr€om I, Mårtensson P, Lloyd Spetz A.Influence of carbon monoxide, water and oxygen on high temperature catalyticmetal-oxide silicon carbide structures. Sensor Actuator B 1997;45:183e91.

78. Becker E, Andersson M, Eriksson M, Lloyd Spetz A, Skoglundh M. Study of thesensing mechanism towards carbon monoxide of platinum-based field effect sensors.IEEE Sens J 2011;11(7):1527e34.

79. Darmastuti Z, Pearce R, Lloyd Spetz A, Andersson M. The influence of gate bias andstructure on the CO sensing performance of SiC based field effect sensors. Proc IEEESensors 2011:133e6. Limerick, Ireland, October 28-31, 2011.

80. Andersson M, Lloyd Spetz A, Pearce R. Tunable gas alarms for high temperatureapplications based on 4H-SiC MISFET devices. In: Proceedings of the international con-ference on silicon carbide and related materials, Cleveland, ICSCRM 2011, OH, USA,September 11e16; 2011. p. 365.

81. Andersson M, Lloyd Spetz A. Tailoring of SiC based field effect gas sensors forimproved selectivity to non-hydrogen containing species. In: Proc. IMCS13, Perth,Australien 12e14 July; 2010. p. 369.


82. Eriksson M, Ekedahl L-G. The influence of CO on the response of hydrogen sensitivePd-MOS devices. Sensor Actuator B 1997;42:217e23.

83. Medlin JW, McDaniel AH, Allendorf MD, Bastasz R. Effects of competitive carbonmonoxide adsorption on the hydrogen response of metal-insulator-semiconductorsensors: the role of metal film morphology. J Appl Phys 2003;93(4):2267e74.

84. Di Natale C, Buchholt K, Martinelli E, Paolesse R, Pomarico G, D’Amico A,Lundstr€om I, Lloyd Spetz A. Investigation of quartz microbalance and ChemFETtransduction of molecular recognition events in a metalloporphyrin film. Sensor Actu-ator B Chem 2009;135:560e7.

85. Chizallet C, Costentin G, Che M, Delbecq F, Sautet P. Revisiting acido-basicity onthe MgO surface by periodic density functional theory calculations: role of surface to-pology and ion coordination on water dissociation. J Phys Chem B 2006;110:15878e86.

86. Sang YH, Ho WJ, Jong-Lam L. IrO2 Schottky contact on n-type 4H-SiC. Appl PhysLett 2003;82(26):4726e8.

87. Andersson M, Lloyd Spetz A. Tailoring of field effect gas sensors for sensing ofnon-hydrogen containing substances from mechanistic studies on model systems. In:Proceedings of the IEEE international conference on sensors, Christchurch, New Zealand,October; 2009. p. 2031e5.

88. Logothetis EM, Visser JH, Soltis RE, Rimai L. Chemical and physical sensors based onoxygen pumping with solid-state electrochemical cells. Sensor Actuator B 1992;9:183e9.

89. Visser JH, Soltis RE. Automotive exhaust gas sensing systems. IEEE Trans Instrum Mea-surement 2001;50(6):1543e50.

90. Salomonsson A, Petoral Jr RM, Uvdal K, Aulin C, K€all P-O, Ojam€ae L, Strand M,Sanati M, Lloyd Spetz A. Nanocrystalline ruthenium oxide and ruthenium in sensingapplications e an experimental and theoretical study. J Nanoparticle Res 2006;8:899e910. https://doi.org/10.1007/s11051-005-9058e1.

91. Lloyd Spetz A, Nakagomi S, Wingbrant H, Andersson M, Salomonsson A, Roy S,Wingqvist G, Katardjiev I, Eickhoff M, Uvdal K, Yakimova R. New Materials forChemical and Biosensors, Mater Manuf Process 2006;21:253e6.

92. Reinhardt G, Mayer R, R€osch M. Sensing small molecules with amperometricsensors. Solid State Ionics 2002;150:79e92.

93. Garzon FH, Mukundan R, Lujan R, Brosha EL. Solid state ionic devices for combus-tion gas sensing. Solid State Ionics 2004;175:487e90.

94. Miyahara Y, Tsukada K, Miyagy H. Field-effect transistor using a solid electrolyte as anew oxygen sensor. J Appl Phys 1987;63(7):2431e4.

95. Tobias P, Macak K, Helmersson U, Lundstr€om I, Lloyd Spetz A. Zirconia based ox-ygen sensor without the need of a reference electrode. In: Proceedings of the 8th interna-tional meeting on chemical sensors, Basel, Switzerland; 2000. p. 149.

96. Jacobsén S, Helmersson U, Ekedahl L-G, Lundstr€om I, Mårtensson P, Lloyd Spetz A.Pt/CeO2 SiC Schottky diodes with high response to hydrogen and hydrocarbons. In:Proceedings of transducers ’01 and Eurosensors XV, Munich, Germany; 2001. p. 832e5.

97. Cerd�a J, Arbiol J, Dezanneau G, Díaz R, Morante JR. Perovskite-type BaSnO3 pow-ders for high temperature gas sensor applications. Sensor Actuator B 2002;84:21e2.

98. Cerd�a J, Morante JR, Lloyd Spetz A. New tunnel Schottky SiC devices using mixedconduction ceramics. Mater Sci Forum 2003;433(6):949e52.

99. Krause S, Moritz W, Grohmann I. Improved long term stability for an LaF3 basedoxygen sensor. Sensor Actuator B 1994;18(19):148e54.

100. Vasiliev A, Moritz W, Fillipov V, Bartholom€aus L, Terentjev A, Gabusjan T. Hightemperature semiconductor sensor for the detection of fluorine. Sensor Actuator B1998;49:133e8.


https://doi.org/10.1007/s11051-005-9058–1

101. Zamani C, Shimanoe K, Yamazoe N. A new capacitive-type NO2 gas sensorcombining an MIS with a solid electrolyte. Sensor Actuator B 2005;109:216e20.

102. Kida T, Kishi S, Yuasa M, Shimanoe K, Yamazoe N. Planar NASICON-based CO2sensor using BiCuVOx/Perovskite-type oxide as a solid-reference electrode.J Electrochem Soc 2008;155:J117e21.

103. Inoue H, Andersson M, Yuasa M, Kida T, Lloyd Spetz A, Shimanoe K. CO2 sensorcombining a metal-insulator silicon carbide (MISiC) capacitor and a binary carbonate.Electrochem Solid State Lett 2010;14(1):J4e7. https://doi.org/10.1149/1.3512998.

104. Abrahamsson B, Gr€onbeck H. NOx adsorption on ATiO3(001) perovskite surfaces.J Phys Chem C 2015;119:18495e503.

105. M€oller P, Andersson M, Lloyd Spetz A, Puustinen J, Lappalainen J, Eriksson J. NOxsensing with SiC field effect transistors. Mater Sci Forum 2016;858:993e6.

106. Wold S, Sj€ostr€om M, Eriksson L. PLS-regression: a basic tool of chemometrics. Che-mometr Intell Lab Syst 2001;58:109e30.

107. Chatfield C, Collins AJ. Introduction to multivariate analysis. London: Chapman & Hall;1980.

108. Larsson O, G€oras A, Nytomt J, Carlsson C, Lloyd Spetz A, Artursson T, Holmberg M,Lundstr€om I, Ekedahl L-G, Tobias P. Estimation of air fuel ratio of individual cylindersin SI engines by means of MISiC sensor signals in a linear regression model. In: SAEtechnical paper series, paper 2002e01e0847, Detroit, Michigan, USA; 2002.

109. Duda RO, Hart PE, Stork DG. Pattern classification. New York: Wiley; 2000.110. Gutierrez-Osuna R. Pattern analysis for machine olfaction. IEEE Sens J 2002;2(3):

189e202.111. Lee AP, Reedy BJ. Temperature modulation in semiconductor gas sensing. Sensor

Actuator B 1999;60:35e42.112. Reimann P, Horras S, Sch€utze A. Field-test system for underground fire detection

based on semiconductor gas sensors. In: Proceedings of the IEEE international conferenceon sensors, 2009, Christchurch, New Zealand; 2009. p. 659e64.

113. Bur C, Reimann P, Sch€utze A, Andersson M, Lloyd Spetz A. Increasing the selectivityof Pt-gate SiC field effect gas sensors by dynamic temperature modulation. In: Proc.IEEE int. Conf. on sensors, Waikoloa, USA 1e4 November; 2010. p. 1267e72.https://doi.org/10.1109/ICSENS.2010.5690598.

114. Bur C, Reimann P, Sch€utze A, Andersson M, Lloyd Spetz A. New method forselectivity enhancement of SiC field effect gas sensors for quantification of NOx,microsystem technologies/smart sensors. Actuator MEMS 2012;18(7):1015e25.https://doi.org/10.1007/s00542e012e1434-z. Springer-Verlag, Berlin.

115. Darmastuti Z, Bur C, Lindqvist N, Andersson M, Sch€utze A, Lloyd Spetz A.Hierarchical methods to improve the performance of the SiC e FET as SO2; sensorsin flue gas desulphurization systems. Sensor Actuator B 2015;206:609e16.

116. Bur C, BastuckM, Lloyd Spetz A, AnderssonM. Selectivity enhancement of SiC-FETgas sensors by combining temperature and gate bias cycled operation using multivariatestatistics. Sensor Actuator B Chem 2014;193:931e40.

117. Bastuck M, Bur C, Lloyd Spetz A, Andersson M, Sch€utze A. Gas identification basedon bias induced hysteresis of gas-sensitive SiC field effect transistor. J Sens Sens Syst(JSSS) 2014;3:9e19.

118. Bastuck M, Reimringer W, Conrad T, Sch€utze A. Dynamic multi-sensor operationand read-out for highly selective gas sensor systems. Proc Eng 2016;168:1685e8.

119. Reimann P, Dausend A, Sch€utze A. A self-monitoring and self-diagnosis strategy forsemiconductor gas sensor systems. In: Proceedings of IEEE international conference onsensors. Italy: Lecce; 2008. 192e5.


https://doi.org/10.1149/1.3512998

https://doi.org/10.1109/ICSENS.2010.5690598

https://doi.org/10.1007/s00542–012–1434-z

120. Sobocinski M, Zwiertz R, Juuti J, Jantunen H, Golorika L. Electrical and electrome-chanical characteristics of LTCC embedded piezoelectric bulk actuators. Adv ApplCeram 2010;109(3):135e8.

121. Nowak D, Kulczak D, Januszkiewicz M, Dziedzic A. High temperature LTCCpackage for SiC-based gas sensor. Opt Appl 2009;XXXIX(4):701e4.

122. Sobocinski M, Khajavizadeh L, Andersson M, Lloyd Spetz A, Juuti J, Jantunen H.Performance of LTCC embedded gas sensors. Procedia Eng 2015;120:253e6.

123. Rabe T, Schiller WA, Hochheimer T, Modes C, Kipka A. Zero shrinkage of LTCCby self-constrained sintering. Int J Appl Ceram Technol 2005;2(5):374e82.

124. Lloyd Spetz A, Sobocinski M, Halonen N, Puglisi D, Juuti J, Jantunen H,Anderson M. LTCC, new packaging approach for toxic gas and particle detection.Proceedia Eng 2015;120:484e7.

125. Loloee R, Chorpening B, Beer S, Ghosh RN. Hydrogen monitoring for power plantapplications using SiC sensors. Sensor Actuator B 2008;129:200e10.

126. WHO (World Health Organization). Regional office for Europe. WHO guidelines forindoor air quality: selected pollutants, ISBN 978 92 890 0213 4.


CHAPTER ELEVEN

Semiconducting directthermoelectric gas sensorsF. Rettig, R. MoosUniversity of Bayreuth, Bayreuth, Germany

Contents

11.1 Introduction 34711.1.1 Motivation for research on direct thermoelectric gas sensors 34711.1.2 Thermoelectric power 34911.1.3 Direct and indirect thermoelectric gas sensors 35011.1.4 Early research activities 353

11.2 Direct thermoelectric gas sensors 35311.2.1 Measurement techniques 35311.2.2 Modeling and simulation of thermoelectric gas sensors 35711.2.3 Measurements and results 36711.2.4 Ionic direct thermoelectric gas sensors 378

11.3 Conclusion and future trends 380References 381

11.1 Introduction

This introductory section will describe the various issues that aremotivating research on direct thermoelectric gas sensors (DTEGs) beforepresenting a brief introduction to thermoelectric power for the generalreader. The principles of direct and indirect thermoelectric gas sensors arethen outlined, while early research work is reviewed in the final subsection.

11.1.1 Motivation for research on direct thermoelectric gassensors

Gas sensors play an important role in many applications and have beenextensively developed during the past few decades. This is especially thecase for applications in monitoring automotive exhaust gases (lambda probe)and air quality (AQ sensors). Although the lambda probe itself cannot reducepolluting emissions from automobiles, it allows the adjustment of a



https://doi.org/10.1016/B978-0-08-102559-8.00011-2

stoichiometric mixture of air and fuel.1 A modern concept with two lambdaprobes even allows detection of a defect in a three-way catalyst.2 AQsensors can monitor the air quality in houses and cars,3 as well as detectingconcentrations of unburnt hydrocarbons,4 an important point in fire preven-tion. Other applications include alerting people when harmful gases are inthe ambient atmosphere.5

Since the 1960s, many research activities have been addressed to resistive(also known as “conductometric”) gas sensors. Since the development ofTaguchi’s sensor based on SnO2,

6 many semiconducting materials havebeen investigated and analyzed. Besides SnO2, the most prominent exam-ples are TiO2;

7,8 SrTiO3;9,10 SrTixFe1exO3ed;

11,66 WO3;12 Ga2O3;

13

Cr2O3;14e16 or ZnO.17 Many of these materials are discussed elsewhere

in this book. The appeal of resistive gas sensors is the relative simplicity ofmanufacturing resistive sensors combined with an uncomplicated principlefor taking measurements. Some of these materials have been tested inautomotive exhausts (e.g., 8,9,13,18). However, harsh environments are achallenge for resistive gas sensors, as poisoning or deterioration of the gassensitive materials by aggressive components such as SO2 or NOx may occuror abrasion of the gas sensitive layer by particle-containing high-velocity gasstreams may cause irreversible harm to the gas sensors.19 This is easy tounderstand because each geometric changedwhich may occur, for instance,by abrasiondcan have a marked effect on the resistance and cause flawedconcentration readings. Protective layers were proposed to overcome theseproblems.20,21

For application in exhausts, potentiometric or amperometric gas sensorsbased on ion conduction membranes of yttria-stabilized zirconia (YSZ)are typically used.1,22 Such sensors provide sufficient stability against harshenvironments. The potentiometric principle offers the possibility ofmeasuring a path-independent quantitydthe electrical potential difference(voltage). In theory, abrasion does not significantly affect the sensor signals.These advantages come at the cost of a more complicated design wherepotentiometric and amperometric gas sensors are concerneddfor instance,the classical lambda probe requires an air Ref. 23 or a pumped Ref. 24.

Semiconducting DTEGs do not have the disadvantages of resistiveor potentiometric gas sensors. The measurand is a path-independentthermovoltage and no gas reference is required. The typical materialsused in resistive gas sensors can also be utilized for DTEGs. In this chapter,it will be shown that intrinsic semiconducting materials have an enhancedsensitivity compared with classical p- or n-type conducting materials.

348 F. Rettig and R. Moos

These advantages are the main drivers for the research and developmentof DTEGs.

In the next section, a short introduction is given to the term “thermo-electric power” or “Seebeck coefficient,” and some early research activitieson DTEGs are summarized. The main part of this chapter deals withmeasurement techniques, modeling, and simulation of DTEGs based onsemiconducting oxides. Recent results obtained with some semiconduct-ing materials, together with a consideration of ionic DTEGs, completethe main part of this chapter. The chapter concludes with a discussionon the disadvantages and drawbacks of DTEGs and possible future researchtopics.

11.1.2 Thermoelectric powerThe intention of this section is to give the inexperienced reader a shortintroduction to the physical background of thermopower, also known as“thermoelectric power.” The reader is referred to Thermoelectricity25 for amore detailed analysis.

The simplest treatment of the thermopower for semiconductors isbased on the fact that the velocity of electrons increases with increasingtemperature. Let us assume a wire that is divided along its length into alow-temperature section and a high-temperature section. In cross section,at the point in the middle where the two sections meet, all electrons thatpass the interface are counted. After a certain time, more electrons willhave traveled from the high-temperature section to the low-temperaturesection than vice versa. As a result, an electrical voltage evolves betweenboth sections to compensate the driving force. In the book of Thermo-electricity25 a detailed calculation is given for a semiconductor assuming aBoltzmann distribution of the electrons and a temperature gradient in acertain direction. The thermovoltage is caused by the thermal diffusion ofthe charge carriers in the temperature gradient in the different sections ofthe wire. The calculation clearly demonstrates that temperature-dependent contact voltages do not play a role in the measured thermovolt-age. As a result, the thermopower (Seebeck coefficient), h, of pure n-type,pure p-type, and mixed nep-type conductors can be expressed, respectively,by Eqs. (11.1e11.3) (e.g., 26):

hn�conductor ¼ � kBe

�lnNC

nþ Ae

�(11.1)

Semiconducting direct thermoelectric gas sensors 349

hp�conductor ¼kBe

�lnNV

pþ Ah

�(11.2)

h ¼ snhn þ sphp

sn þ sp(11.3)

In these equations, kB is the Boltzmann constant, e is the electron charge,NC andNV are the effective densities of states in the conduction band and inthe valence band, Ae and Ah are the transport constants representing thescattering mechanism, n and p are the concentrations of electrons in theconduction band and holes in the valence band, and sn and sp are theconductivities of the electrons and the holes.

The second, more general, treatment is based on nonequilibriumthermodynamics. Fluxes and forces are connected by a matrix. The diagonalelements (the main effects) of this matrix are well-knowndfor example, thediffusion coefficient (which is the connection between particle fluxes undera concentration gradient) or the thermal conductivity (which relates thetemperature gradient with the heat flux). One of the nondiagonal elementsis the Seebeck coefficient (thermopower, h), which relates a temperaturegradient with a particle flux. Based on this, general equations are obtainedthat describe the heat and particle flow in a thermal and concentrationprofile:

divðs , gradVthermoelectricþ sh , gradTÞ ¼ 0

divðshT , gradVthermoelectricþ k , gradTÞz0 (11.4)

Here, s is the conductivity, Vthermoelectric is the thermovoltage, h theSeebeck coefficient, T the temperature, and k is the thermal conductivity fora vanishing electrical field. The second equation is here set to zero, as Jouleheating as a second-order effect does not play a significant role. More detailscan be found in Refs. 27 or 28.

11.1.3 Direct and indirect thermoelectric gas sensorsThermoelectric gas sensors can be divided more or less arbitrarily into directand indirect thermoelectric gas sensors. Until recently, research has beenmainly addressed to indirect gas sensors, and DTEGs have rarely beenstudied. Fig. 11.1 explains the working principle for both types of thermo-electric gas sensor.

Indirect thermoelectric gas sensors use the heat of an exothermic reactionthat stems from a combustible analyte. The temperature on a catalytically


coated area of a (usually planar) substrate increases with the concentration ofthe analyte.29 The temperature difference between catalytically inactiveareas on the substrate is usually measured either by thermocouples or bythermopiles. Therefore, this type of sensor is called an “indirect thermoelec-tric gas sensor.” Its measurement principle is similar to pellistor sensors.30

Indirect thermoelectric gas sensor

Direct thermoelectric gas sensor (absolute temperature measurement)

Direct thermoelectric gas sensor (relative temperature measurement)

Gas sensitive film(gsf)

Catalyticactive coating

Thermoelectric material

Thermo couples (tc)

Pt

Pt

Pt

Pt

Au

Au

C3H8 + 5O2

3CO2 + 4H2O

V

Vgsf

Vgsf

VΔT, ΔT

T2

T2

T2

T1

T1

T1

–ΔH

ΔT

Gas sensitive film(gsf)

(a)

(b)

(c)

Figure 11.1 Principle of the setup of (a) indirect and (b) and (c) direct thermoelectricgas sensors. The difference between (b) and (c) is the method of measuring thetemperature difference DT. Reprinted from Rettig F. (2008), Direkte thermoelektischeGassensoren (in German) [PhD thesis]. University of Bayreuth, Shaker-Verlag; 2008 withpermission from Shaker-Verlag.


The thermoelectric material itself should not be catalytically active. Thecatalytically active coating of the thermoelectric material defines both thesensitivity and the selectivity of the sensors. As an example, the reader isreferred to Refs. 29 or 31, where further information on indirect thermo-electric gas sensors is given.

In contrast, in DTEGs, the Seebeck coefficient (thermopower, h) of thegas sensitive material itself changes when the concentration of the analytevaries in the ambient atmosphere. The density of the free electrons and/or defect electrons (holes)dor, in other words, the Fermi leveldis directlyaffected by a changing analyte gas concentration. There are several possiblephysical effects to explain how the Fermi level can be dependent on the gasphase. Chemisorption, for instance, may occur following the reaction:

O2þ 2e�42O�ads (11.5)

This chemisorption process captures electrons from the gas sensitivematerial. Therefore, a space charge region evolves from the interface of thematerial and the gas phase. In porous structures, this interface is typically thegrain surface. For an n-type semiconductor such as SnO2, electrons aredepleted in this region, resulting in an increased resistance not only for thewhole grain but also for the whole gas sensitive film. If reducing gases arepresent, they may consume the chemisorbed oxygen and the electron is trans-ferred back to the gas sensitive material, the space charge regions vanish partly,and the resistance of the gas sensitive material decreases, often by decades.

Typically, at higher temperatures, one finds effects in which the bulk ofthe material is involved. The electron concentration in the bulk material canbe modulated by ex- or incorporation of oxygen according to

12O2þ 2e� þ V ••4Ox

Oo (11.6)

Then, the oxygen partial pressure of the surrounding gas atmosphere isthe driver for a change in the electron concentration. These examples clearlyshow that DTEGs are based on the same physical principles as conducto-metric gas sensors, as in both cases the analyte concentration modulatesthe electron density. However, the measurand is different. In conducto-metric devices, the material property “conductivity” changes and, hence,the resistance of a sensor varies with the analyte concentration. In contrast,the determination of thermopower is more complicated, as not only thethermovoltage has to be measured but also a known temperature differencehas to be applied or, at least, measured.


11.1.4 Early research activitiesThe concept of DTEGs is not a recent one. Several authors worked on thistopic in the 1980s, but no systematic studies were conducted at that time.Pisarkiewicz and Stapinski32 reported on the change of the Seebeck coeffi-cient of SnO2 when applying reducing gases. The effect was attributed to amodulation of the depletion layer at the grain surfaces affecting the Fermilevel. Siroky33 used a thermoelectric gas sensor based on SnO2 to detectflammable gases. Here, it was considered to measure thermopower andconductivity in parallel. Mizsei34 also explained the change of the thermo-electric power of palladium-activated tin oxide SnO2 in the presence of H2

with the affected depletion layer. Moos35 described a method for measuringthe oxygen content of a gas by using the thermoelectric effect of a bulkmaterial. Ionescu36 reported on a SnO2 gas sensor with increased selectivityusing simultaneous measurement of resistance and the Seebeck coefficient.In 2000, Liess and Steffes37 presented a DTEG based on In2O3, and Smulkoet al.38 used thermoelectric voltage fluctuations for gas sensing. However, itshould be noted that this early work was scattered research, without aholistic consideration of the material and the correspondingly requisitetransducers and evaluating systems. Additionally, these early approachesdid not classify their devices as DTEGs.

11.2 Direct thermoelectric gas sensors

The following sections present the optimization of the transducers andthe gas sensitive materials as well as results for different DTEGs. Section 7.2.1covers the measurement technique for DTEGs. Section 7.2.2 describes thetheoretical design of transducers and gas sensitive materials to enable thedesign of accurate, fast, and long-term stable DTEGs. Section 11.2.3presents results for different DTEGs with different materials. Section11.2.4 describes ionic DTEGs as alternatives to semiconducting oxidematerials.

11.2.1 Measurement techniquesCompared with the relatively simple resistance measurement, DTEGsrequire a more sophisticated setup. The measurand thermopower (Seebeckcoefficient) is defined by

hgsf ¼ hPt �DVgsf

DT(11.7)


In Eq. (11.7), hgsf is the Seebeck coefficient of the gas sensitive film, DVgsf

is the measured thermovoltage of the gas sensitive film, and DT is thetemperature difference at the junctions between the gas sensitive layer andthe conductor tracks. Owing to the fact that the conductor tracks also add athermovoltage, the thermopower of the gas sensitive layer has to becorrected by the thermopower of the conductor track material(here platinum), hPt.

There are two ways to determine the temperature difference betweenthe junctions of the conductor tracks and the gas sensitive film. Fig. 11.1depicts both possibilities: in (b), the temperature difference, DT, is directlydetermined, whereas in (c), the temperatures at both junctions, T1 andT2, are measured separately and the temperature difference DT is calculated.Because of the fact that not only the temperature difference but also thetemperature of the gas sensor has to be controlled precisely, the optionpresented in Fig. 11.1(c) is advantageous. Combinations of both optionsare also possible.

The thermovoltages of metallic thermocouples are usually easy tomeasure, although the voltages are in the microvolt range. In contrast, theinternal resistances of semiconducting oxides are by orders higher. Suchhigh ohmic voltage sources are difficult to measure.39,40 Therefore, a trans-ducer for a DTEG has to be developed to ensure an accurate performance.The transducer which will be discussed below allows a maximum internalresistance of the gas sensitive layer of about 1 MU.

According to Eq. (11.7), it would be possible to apply the temperaturedifference statically, but a temperature modulation technique enables oneto measure the thermovoltages of the sensor material more precisely. Inaddition, some plausibility checks are possible. Furthermore, it is well-known that materials may decompose slowly in a temperature gradientdue to the Soret effect.41

Fig. 11.2 shows the design of a thermoelectric gas sensor device manu-factured according to planar ceramic multilayer technology. The heaterbrings the tip of the sensor to operation temperature by applying a heatervoltage,Vheater. The modulation heater generates the temperature differencefor the gas sensitive layer. For this purpose, a sinusoidal modulation voltage,Vmodu, is applied. The equipotential layer will be explained in the nextparagraph. The gas sensitive layer and two thermocouples are located onthe top of the sensor. With the help of the thermocouples, the temperatures


T1 and T2 are determined. The thermovoltage of the gas sensitive film, Vgsf,is measured over the platinum legs of the thermocouples. According toEq. (11.7), the thermopower of the gas sensitive layer can be calculated.More information on the challenges faced in manufacturing such sensorscan be found in Ref. 27.

Fig. 11.3 shows experimental data relating to DTEGs: (a)e(d) stem froma sensor without an equipotential layer. The temperature difference inFig. 11.3(a) is clearly sinusoidal, whereas the thermovoltage, Vgsf, of thegas sensitive layer differs significantly from sinusoidal behavior. The distortedsignal, Vgsf, prevents a linear regression with DT. The reason for the distor-tion becomes apparent from Fig. 11.3(d), where a Fourier analysis of thethermovoltage of the gas sensitive layer, Vgsf, is shown. Besides the expected

Gas sensitive filmand thermocouples

Insulation layer

Equipotential layer

Insulation layer

Modulation heater

Substrate

Heater

elpuocomrehTelpuocomrehT

Au

Gas sensitive film 4 mm

Equipotential ring Pt

Pt

Vgsf

Vmodu

Vheater

T1 T2

Figure 11.2 Setup of the direct thermoelectric gas sensors presented in this chapter.The applied and measured voltages are also indicated. Reprinted from Rettig F., MoosR. Temperature-modulated direct thermoelectric gas sensors: thermal modeling and resultsfor fast hydrocarbon sensors. Meas Sci Technol 2009;20(6):0652059 with permission fromIOP-Publishing Ltd.


signal of 10 mHz, an additional signal at 5 mHz is found. This is thefrequency of the applied modulation voltage, Vmodu. According toEq. (11.8), the temperature modulation frequency, fDT, is twice themodulation frequency, fmodu:

P ¼ V 2modu

R¼ V 2

0;moud

R$cos2ðpfmodutÞ

¼ V 20;modu

2R$ð1þ cosð4pfmodutÞÞ ¼

V 20;modu

2R$ð1þ cosð2pfDT tÞÞ

(11.8)

The 5 mHz signal is an interference of the modulation voltage, Vmodu,with the thermopower due to a small residual conductivity of the substratematerial at elevated temperatures of several hundred degree Celsius. Thefinite resistance of the substrate material and the modulation voltage, inthe range of several volts, in combination with the relatively high resistanceof the gas sensitive layer and the small thermovoltages, indicates that theamplitudes of thermovoltage and the disturbing voltage are in the same

4020–2–4–6–8–10–12

420

–2–4–6–8

–10–12

642086420 50–3–6–9 02510121–

–2

–4

–6

–8

–10

–12

5.0–0

–1.0

–1.5

–2.0

–2.5

–3.0

–3.5

–0.54

2

02

1

0

–1.0

–1.5

–2.0

–2.5

–3.0

–3.5–8 –6 –4 –2 0 5 10 15 20

–1–2–3–4–5–6–7–8

0 100 200 300 400

0 100t (s)

t (s)

f (mHz)

f (mHz)

200 300 400

ΔT (K

)ΔT

(K)

ΔT (K

)ΔT

(K)

ΔT (K)

ΔT (K)

Vgs

f (m

V)

Vgs

f (m

V)

Vgs

f (m

V)

Vgs

f (m

V)

(a)

)g()f()e(

(h)

)c()b(

(d)

Figure 11.3 Measured and analyzed signals of a direct thermoelectric gas sensor (a),(b), (c), (d) without equipotential layer and (e), (d), (f), (g), (h) with equipotential layer.The time resolved signals of the temperature difference, DT, and the measured thermo-voltage, Vgsf, are shown in (a) and (e). The DT(Vgsf) diagram is given in (b) and (f). Theresult of the Fourier analysis of DT and Vgsf can be found in (c), (d), (g), and (h).


order. The low resistance of the thermocouples is a short circuit for thedisturbing modulation voltage and, therefore, the disturbing voltagecollapses, in contrast to the situation where the gas sensitive layer has a signif-icantly higher resistance. Therefore, the low resistance of the thermocouplescompared with the high resistance of the gas sensitive layer is the reason whythe temperatures measured by the thermocouples remain almost unaffectedby the modulation.

It would be possible to implement a Fourier analysis to calculate thethermopower of the gas sensitive layer from the distorted voltage of thegas sensitive layer,43 but an improved design of a DTEG with an additionalequipotential layer offers the possibility to measure almost sinusoidal signalsof the temperature difference, DT, and the thermovoltage of the gassensitive layer, Vgsf. Fig. 11.3(e)e(h) shows the results of a sensor withan equipotential layer. Both the Fourier analysis of Vgsf and DT showcontributions of nearly one frequency at 10 mHz (Fig. 11.3(g) and (h))dthat is, the signals are almost sinusoidal (Fig. 11.3(e)) and, therefore, the slopeof a linear regression can be used to determine the thermopower of the gassensitive film (Fig. 11.3(f)).

More details on the design of the equipotential layer can be found inRef. 43. The higher the internal resistance of the gas sensitive material,the more necessary the equipotential layer becomes to improve the signalquality of the DTEGs. Although the modulation technique improvesaccuracy, there is a fundamental drawback of the temperature modulation(Fig. 11.3(e)): at least one complete modulation period is required to obtaina first reading for the thermopower of the material. With a frequency of10 mHz, the response time of such a sensor is about 100 s. Even if thematerial itself responds much more rapidly to changing analyte concentra-tions, the measurement technique impedes a faster sensor response. Thenext section deals with a solution to the problem of reducing the responsetime and presents a simulation of the electrical and thermoelectric materialproperties.

11.2.2 Modeling and simulation of thermoelectric gassensors

According to Ref. 44, the thermal time constant (relaxation time after asudden temperature step) of typical gas sensors manufactured in thick-filmtechnology on planar alumina substrates is around 10 s. This thermal timeconstant is valid for large temperature steps. It is mainly driven by theconvection coefficient of the gas sensor substrate. However, for temperature


modulation, large temperature differences are not necessary. A temperaturedifference of 20 �C suffices for a DTEG. To reduce the response time of thesensor, the temperaturemodulation frequency has to be increased significantly.

The idea of a higher modulation frequency can be explained by a simplemodel known from Earth sciences. If one applies a sinusoidal temperaturechange to the ground, it is interesting to consider the depth to which groundtemperature modulations are present. If one considers, for instance, a dailytemperature change (hot days and cold nights), the penetration depth ofthe thermal wave is around 10 cm. For yearly temperature modulations(hot summers and cold winters), the thermal wave can penetrate to a depthof around 1 m. That is the reason why water pipes should be installed at leastto this depth, otherwise the water would freeze. As a result, the frequency ofthe thermal modulation has a major influence on the penetration depth.45

Based on this concept, a one-dimensional model has been built up, whichdescribes the thermal behavior of a DTEG. A more detailed explanationof the model can be found in Ref. 42.

The setup of the one-dimensional model is shown in Fig. 11.4. Becauseof the fact that the ratio of the cross section to the perimeter is small, it ispossible to introduce surface-related convective heat loss as a volume heatloss into the one-dimensional partial differential equation:

rcpvTvt

� kv2T

vx2¼ 1

2dq0ð1þ expð2jpftÞÞ � hconv$Psensor

AsensorðT� TaÞ

for x < d = 2

(11.9)

rcpvTvt

� kv2T

vx2¼ �hconv$Psensor

AsensorðT� TaÞ

for x � d=2

(11.10)

Modulation heaterSymmetry plane

hsensor

bsensor

xdxd/20

Figure 11.4 Setup for the one-dimensional model for the thermal simulation of a directthermoelectric gas sensor. Reprinted from Rettig F., Moos R. Temperature-modulateddirect thermoelectric gas sensors: thermal modeling and results for fast hydrocarbonsensors. Meas Sci Technol 2009;20(6):0652059 with permission from IOP-Publishing Ltd.


Eqs. (11.9) and (11.10) were used to simulate the thermal behavior of thesensor. The density of the material is r, cp is the heat capacity, T is thetemperature, t the time, k is the thermal conductivity, x is the coordinateaccording to Fig. 11.4, d is the lateral length of the heater, _q is the heatgeneration of the modulation heater per volume, f the modulationfrequency, hconv the convection coefficient, Psensor the perimeter of thesensor (Psensor ¼ 2hsensor þ 2bsensor), Asensor the cross section of the sensor(Asensor ¼ hsensor $ bsensor), and Ta the ambient temperature. These partialdifferential equations can be divided into a static differential equation andinto a differential equation for a stationary harmonic solution. Details canbe found in Refs. 27 or 42. The solution to these equations is

TM;SðxÞ ¼ G2K

8>>>>><>>>>>:

1� exp

�� ffiffiffiffi

Kp d

2

�cosh

� ffiffiffiffiK

px�

for 0 < x � d2

sinh

� ffiffiffiffiK

p d2

�exp

�� ffiffiffiffiK

px�

for x >d2

with G ¼ q02dk

and K ¼ hconvPsensorkAsensor

þ j2pf rcp

k

(11.11)

For the static part of the solution, TS, f is zero, otherwise the actualtemperature modulation frequency, f, is used to calculate the amplitudeand the phase angle of the complex temperature distribution, TM (x). Theconsidered thermal model of the sensor was verified by comparison withthe measured temperature distribution of a real sensor. The factor G relatesthe volume source of heat _q with the geometry of the modulation heater. Kis, in general, the decay constantddue to the fact that it is a complexnumber, a decaying thermal wave is the result.

Fig. 11.5(b) shows the temperature difference amplitude, DTM (x), withthe modulation frequency as a parameter. The experimental data wereobtained by a line scan using an infrared camera. The mean applied electricalpower was the same for all different modulation frequencies f. The statictemperature distribution, Ts(x), therefore agrees for all temperaturemodulation frequencies (Fig. 11.5(c)). This is validated by the thermal modelof the sensor and by the measured static temperature distribution. Asexpected, the amplitude of the harmonic thermal distribution decreaseswith increasing frequency (Fig. 11.5(b)). The length of the gas sensitive layerwas designed to be 4 mm. If the modulation heater was placed at one end of


the gas sensitive layer, the temperature at the other end of the gas sensitivelayer would barely be affected by the harmonic part of the thermalmodulation. The behavior is described best with the parameter “penetrationdepth” of a thermal wave lT. Fig. 11.5(a) shows lT as a function of themodulation frequency. A modulation frequency of fz 1 Hz results in apenetration depth of the thermal wave lT of around 1e2 mm. Therefore,a temperature modulation frequency of 1 Hz requires a modulation heaterto be placed at a distance of less than 1 mm from one end of the gas sensitivelayer. Otherwise, only a small amplitude of the thermal wave will reach thegas sensitive layer.

A sensor was fabricated according to the design illustrated in Fig. 11.2.First, the platinum modulation heater was screen-printed and fired; then,the insulation layer was applied onto the modulation heater layer. Afterthe equipotential layer and a further insulation layer, the heater and itsconductor tracks were applied. Finally, the thermocouples were

f = 0.31 Hzf = 1 Hzf = 3.16 Hzf = 10 Hzf = 31.6 Hz

fmod (Hz) x (mm)

T s (°

C)

ΔTM

(°C

)

k = 1.4 Wm –1 K –1

Standard parameters

hsensor = 650 μm

l T (m

m)

Model

102

101

100

10–110–3 10–2 10–1 100 101

25

20

15

10

5

02802602402202001801600 1 2 3 4 5 6 7

(a) (b)

(c)

Figure 11.5 Results of the thermal modeling of a direct thermoelectric gas sensor. Thedecay constant (penetration depth) IT of the thermal is shown in (a). The good agree-ment of both the harmonic part and the static part between model and measurementis demonstrated in panels (b) and (c). (c) Reprinted from Rettig F., Moos R. Temperature-modulated direct thermoelectric gas sensors: thermal modeling and results for fasthydrocarbon sensors. Meas Sci Technol 2009;20(6):0652059 with permission fromIOP-Publishing Ltd.


screen-printed on the top and fired, and an insulation layer was screen-printed and fired. As a result, the distance between the modulation heaterand the thermocouple was around 40e60 mm, while the other thermo-couple was 4 mm away. Experimental results with different DTEGs arediscussed in the next section.

The most important part of DTEGs is, of course, the gas sensitive material.Typical gas sensitive materials for classical conductometric gas sensors werenot developed and optimized for the application as DTEGs. The Seebeckcoefficients of these semiconducting oxide materials change with theconcentration of free charge carriers, as shown in Eqs. (11.1)e(11.3). Themeasurand resistance (conductance) is always a positive valuedin contrastto the thermopower, which can have either positive or negative signs. Thisoffers the opportunity to use different materials for DTEGs. Moos35 describeda direct thermoelectric oxygen sensor with an intrinsic bulk material to beoperated above 600 �C. Another approach is considered here for simulation:gas sensitive materials for temperatures around 400e600 �C. In thistemperature range, bulk incorporation of oxygen in these materials is avery slow process and can be ignored.46

Simulation results obtained for semiconductor materials are shown in thenext paragraphs. Fig. 11.6 depicts the geometrical model. The model has atwo-dimensional rotational symmetry. The dark gray area is considered forthe simulation. Each grain is described by its radius RK. The neck radius,RH,describes the interconnection with adjacent grains. Chemisorption takesplace at the grain surface. A space charge region develops from the grain

Grad TRK RH

2ne–

2On–

O2Chemisorption

r

x

Figure 11.6 Model with rotational symmetry for the analysis of materials for directthermoelectric gas sensors. The dark gray area was simulated with the commercialFEM-software Comsol Multiphysics. Reprinted from Rettig F., Moos R. Morphologydependence of thermopower and conductance in semiconducting oxides with spacecharge regions. Solid State Ionics 2008b;179(39):2299e2307 with permission fromElsevier.


surface to the middle of the grain. The overall thermoelectric andconductive properties of the grain were calculated as follows:

First, the isothermal PoissoneBoltzmann equation was solved48 fromEq. (11.12):

divðgradFÞ¼ � e2niε0εrkBT

�p0niexpð�FÞ � n0

niexpðFÞ �NA �ND

ni

�(11.12)

Here, F is the reduced potential, which is given by

F ¼ efkBT

(11.13)

where e is the elementary charge, 4 is the electrical potential, kB is theBoltzmann constant, and T is the temperature. In Eq. (11.12), ni is theintrinsic charge carrier concentration, ε0 $ εr is the dielectric constant, p0 andn0 are the hole concentration in the valence band and the electronconcentration in the conduction band, respectively, and NA and ND are theacceptor and the donor concentration in the material. Eq. (11.12) is valid formaterials in which electrons and holes are the mobile charge carriers. Aprecondition for the validity of the equation is that the conduction band andthe valence band are only weakly occupied so that the Boltzmann statistic forthe charge carriers can be applied. For the simulation, Eq. (11.12) wasnormalized regarding the space coordinates; details can be found in Ref. 47.The Debye length gives an idea of the extent of the space charge regions inthe grain. For a p-type semiconductor, the Debye length LD,p can becalculated by

LD;p ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiε0εrkBTe2p0

s(11.14)

The normalized nonlinear partial differential equation (Eq. 11.12) wasimplemented for the rotational symmetric geometry according toFig. 11.6 into the commercial finite element method program ComsolMultiphysics. The solution of this equation is the distribution of the reducedpotential in a gas sensitive grain.

An example solution is shown in Fig. 11.7. A reduced surface potentialFof �5 was applied to the grain surface. At a temperature of 400 �C, thisreduced potential corresponds to an electrical potential 4 of around300 mV. The grain radius was RK ¼ 200 nm, and the neck radius was


RH ¼ 80 nm. The material itself was slightly donor-doped (ND/ni ¼ 10 andNA/ni ¼ 0.1) and the Debye length was 20 nm. The middle of the grain isbarely influenced by the reduced surface potential, whereas almost thewhole neck region is influenced by the reduced surface potential.

Using Eqs. (11.15) and (11.16), the charge carrier concentrationdistributions in the grain can be calculated49:

n ¼ n0 expðFÞ (11.15)

p ¼ p0 expð�FÞ (11.16)

As a result, Fig. 11.8 shows the electron concentration (top) and the holeconcentration (bottom) in the different areas of the grain. At each point ofthe grain, n $p ¼ n2i is valid. However, as the electron concentration cannever be lower than zero, the space charge region of the electrons extendsmuch more toward the grain center than the space charge region of the holes.

The reduced potential F, the electron concentration n/ni, and the holeconcentration p/ni are extracted on the r-axes from Fig. 11.8 and plotted inFig. 11.9. Starting from the grain surface, an inversion area with a length ofabout 30 nm can be seen. In this area, the hole concentration is larger thanthe electron concentration, although the material is slightly donor-doped.This inversion plays a major role for the enhanced sensitivity of materialsfor DTEGs.

r

x0

–5 –4 –3 –2 –1 0

r = RK

Φ

Figure 11.7 Obtained simulation data for the reduced potential F. A reduced surfacepotential was set to F ¼ �5 for a grain radius of RK ¼ 200 nm and a neck radius ofRH ¼ 80 nm. The material was slightly donor-doped (ND/ni ¼ 10 and NA/ni ¼ 0.1). TheDebye length was 20 nm. Reprinted from Rettig F., Moos R. Morphology dependence ofthermopower and conductance in semiconducting oxides with space charge regions. SolidState Ionics 2008b;179(39):2299e2307 with permission from Elsevier.


nni

r = RK

r = RK

n0

ni

n0

ni

pni

pni

ni

p0

ni

= 10

= 0.1

x

x

14

12

10

8

6

4

2

0

Figure 11.8 Calculated electron concentration n/n0 and hole concentration p/p0 for agrain radius of RK ¼ 200 nm and a neck radius of RH ¼ 80 nm based on the results ofFig. 11.7. The material is slightly donor-doped (ND/ni ¼ 10 and NA/ni ¼ 0.1). The Debyelength is 20 nm. Reprinted from Rettig F., Moos R. Morphology dependence of thermo-power and conductance in semiconducting oxides with space charge regions. Solid StateIonics 2008b;179(39):2299e2307 with permission from Elsevier.

0

–2

–4

15

10

5

00 50 100 150 200

Φ

Φ

r (nm)

p/ni

n/ni

n/n i

,p/n

iInve

rsio

n ar

ea

Figure 11.9 Calculated course of the electron concentration n/ni and the hole concen-tration p/ni for a grain radius of RK ¼ 200 nm and a neck radius of RH ¼ 80 nm. Thecurves are extracted from Fig. 11.8 on the r-axis. The material is slightly donor-doped(ND/ni ¼ 10 and NA/ni ¼ 0.1). The Debye length is 20 nm. Reprinted from Rettig F. Dir-ekte thermoelektische Gassensoren (in German) [PhD thesis]. University of Bayreuth,Shaker-Verlag; 2008 with permission from Shaker-Verlag.


Based on Fig. 11.8, the local Seebeck coefficients and the localconductivities can be calculated by applying Eqs. (11.1)e(11.3) and(11.17)e(11.19).

sn ¼ emnn (11.17)

sp ¼ empp (11.18)

s¼ sn þ sp (11.19)

The calculation of the local properties is possible for homogeneoussemiconductors; for inhomogeneous semiconductors, the presumptionshave to be checked. The space charge region extends to about 20 nm.Therefore, the local properties change significantly in this length scale.The mean free path of the charge carriers has to be significantly lowerthan the width of the space charge region, as otherwise the charge carriersare not able to achieve the local equilibrium when traveling through thecrystal. This assumption is definitively not valid for so-called “lifetime”semiconductors,50 where the restoration of the electronehole equilibriumtakes a considerable time. However, in small polaron conductors, theelectrons are more or less localized to single atoms and, hence, the assump-tion for the localized properties can be valid. For some oxide materials, apolaron-type conduction mechanism can be considered. Data for chargecarrier lifetime or charge carrier mean free paths in oxides are rare. Oneof the few data available is published by Barsan;51 the mean free path ofthe electrons in SnO2 is lower than the Debye length by at least a factorof 25.

divðs $ grad4iÞ ¼ 0 (11.20)

divðs $ gradVniþ sh $ gradTÞ ¼ 0 (11.21)

divðshT $ gradVniþ k $ gradTÞ ¼ 0 (11.22)

Using the local properties and Eqs. (11.20)e(11.22) (from Ref. 28), theeffective (overall) conductivity and the effective (overall) thermopower arecalculated. The starting point for the calculation was the reduced potentialF applied on the grain surface. This surface potential is the result of thechemisorption of oxygen on the grain surface. By integration of the spacecharge region, the surface charge concentration can be calculated. Thissurface charge concentration is a (nonlinear) function of the chemisorbedoxygen and, therefore, a measure for the concentration of oxygen in theambient atmosphere. The chemisorption itself can be describeddfor


example, by a Wolkenstein isotherm.52 However, the correlation betweenthe adsorbed amount of oxygen and the ambient oxygen concentration willnot be covered here. It is sufficient to take a look at the surface chargeconcentration to extract some interesting results regarding new materialsfor DTEGs.

Fig. 11.10 shows the calculated effective thermopower, heff, as a functionof the normalized surface charge concentration, which is (as explained in thelast paragraph) a measure for the ambient oxygen concentration. It wascalculated by dividing the thermovoltage difference from the left and theright grain boundaries by the temperature difference at the left and the rightgrain boundary. The normalization factor is the intrinsic charge carrierdensity, ni. Each element of Fig. 11.10 shows the thermopower for a differ-ently doped material with the grain size as a parameter. The slope of thecurve is a measure for the sensitivity of the material. A steep slope indicates

Intrinsic Slightly donor-doped Donor-doped1.2

1.01.0

0.5

0.0

–0.5

–1.0

–0.9

–1.0

–1.1

–1.2

0 15 30 450.0 0.2 0.4 0.6

0.80.60.40.20.0

1.30 0.900.850.800.750.700.650.600.55

0 100 200 300

1.251.201.151.101.051.000.95

0 1 2 3

0.0 0.1 0.2 0.3

50 nm200 nm1000 nm

Slightly acceptor-doped Acceptor-doped

η eff.

(mV

K–1

)η e

ff.(m

VK

–1)

NQ·ni–1 (μm) NQ·ni

–1 (μm)

NQ·ni–1 (μm)NQ·ni

–1 (μm)NQ·ni–1 (μm)

Figure 11.10 The effective thermopower, heff, as a function of the surface chargeconcentration NQ$ni

�1. In each panel, the doping concentration is varied (intrinsic:ND/ni ¼ 0.1, NA/ni ¼ 0.1; slightly donor-doped: ND/ni ¼ 10, NA/ni ¼ 0.1; donor-doped:ND/ni ¼ 1000, NA/ni ¼ 0.1; slightly acceptor-doped: ND/ni ¼ 0.1; NA/ni ¼ 10; acceptor-doped: ND/ni ¼ 0.1, NA/ni ¼ 1000). The reduced surface potential varies from 0 to �5.Each curve is plotted for different grain radii. Reprinted from Rettig F. Direkte thermo-elektische Gassensoren (in German) [PhD thesis]. University of Bayreuth, Shaker-Verlag;2008 with permission from Shaker-Verlag.


a large sensitivity. As illustrated in Fig. 11.10, acceptor-doped materialsexhibit a low sensitivity, donor-doped materials show a better sensitivity,while intrinsic or slightly donor-doped materials are maximally sensitive.In addition, the grain size also influences the sensitivity. The grain radiusof RK z 50 nm is near the Debye length and, therefore, the space chargeregion influences almost the whole grain. For large grains(RK z 1000 nm), only the surface regions are affected. Then, the effectivethermopower is insensitive to changes of the ambient oxygen concentration.The slightly donor-doped material with a grain radius of RK z 50 nm hasthe maximum sensitivity. The reason is obvious from Fig. 11.9: if oxygenis adsorbed, an inversion layer is built from the surface of the grain. If asemiconductor changes from n-type to p-type semiconducting behavior,the sign change in thermopower (the sensitivity) reaches its maximal value.However, if one considers the measurand “resistance change” of a slightlydonor-doped or intrinsic material, one finds almost no sensitivity, as theconductivity always has a positive sign. As a conclusion for this section,the semiconducting materials for DTEGs should be intrinsic or slightlydonor-doped for a maximal sensitivity, if they are based on the chem-isorption of oxygen. For bulk materials, where the entire stoichiometry ofthe material is changed, intrinsic materials may have also some advantagesregarding sensitivity because, in this case, the sign of the thermopoweralso changes.26,53 However, many of the oxide materials that are typicallyused for gas sensors show a significant ionic conductivity at and aroundthe intrinsic minimum.54 It is supposed that the ionic contribution willinteract with the thermopower of electrons and holes.

11.2.3 Measurements and resultsThis section deals with real transducers and gas sensitive materials forDTEGs. Using the results from previous sections, accurate, rapid, andlong-term stable DTEGs with increased sensitivity can be designed.

A first experiment to demonstrate the advantages of DTEGs is shown inFig. 11.11.55 Here, instead of a planar sensor, a small porous ceramicbrick-shaped sample of SrTi0.6Fe0.4O3ed was measured, as described below.The thermopower, h, and the resistance, R, were measured simultaneously(details in Ref. 56). The samples were kept at 700, 800, and 900 �C for 7 heach. Within these 7 h (duration of each run), the oxygen partial pressurewas varied stepwise, and the final values of R and h were plotted. It isinteresting to observe that the resistance characteristics of the material shiftedfrom run to run, presumably because of the sintering process of the sample.


This large shift amounts to an error of approximately one decade in pO2.The thermopower, h, however, remains constant. This experiment clearlyhighlights the advantage of the DTEGs, which is based on the geometryindependence of the potential difference measurement.

As a first planar approach, a DTEG for hydrocarbons based on SnO2 ispresented. In this case, the entire sensor was still heated in a tube furnaceand only the temperature modulation was applied by a planar structure.Therefore, the manufacturing procedure of this sensor was simplercompared with the sensor presented in Fig. 11.2. Fig. 11.12(a) shows thedesign of the sensor. The upper part of the sensor containing the thermo-couples and the gas sensitive layer was joined with the lower part withthe modulation heater by the wet screen-printed equipotential layer. Afterdrying and firing, the sensor was complete.

The sensor was first tested with propane and then a significant part of thegas sensitive layer was milled out. The milled-out portion of the gas sensitivelayer can be seen in Fig. 11.12(b). After milling out the gas sensitive layer,the sensing response was measured again.

Fig. 11.13 shows the results. The thermopower of the gas sensitive layer,hSnO2, is barely influenced after milling out a portion of the gas sensitive

First runSecond runThird run

First runSecond runThird run

2.2 002(a) (b)

180

160

140

120

100–3 –2 –1 0–3 –2 –1 0

2.1

2.0

1.9

1.8

1.7

1.6

1.5

1.4

log pO2 gol)rab( pO2 (bar)

log(

R/ Ω

)

T = 800°CMaterial STF:SrTi0.6Fe0.4O3–δ

η STF

40(μ

VK

–1)

Figure 11.11 (a) Resistance R and (b) thermopower (Seebeck coefficient), h, of a porousSTF (SrTi0.6Fe0.4O3ed) specimen when exposed to different oxygen partial pressures.From Moos R, Izu N, Rettig F, Reiß S, Shin W, Matsubara I. Resistive oxygen gas sensorsfor harsh environments. Sensors 2011;11(4):3439e3465.


layer (Fig. 11.13(a)), but the resistance RSnO2 isdas expected from thepreceding discussiondsignificantly increased (Fig. 11.13(b)). If a propaneconcentration of 100 ppm were present in the ambience of the gas sensor,a milled-out DTEG would measure a concentration of 80 ppm propane.However, a milled-out conductometric gas sensor (Fig. 11.13(b)) wouldonly measure a concentration of 30 ppm propane.

One sees also a drift of both measurands: thermopower and resistance. Atfirst glance, they correlate. Such an assumption can be checked with a

Gas sensitive layerand reference

Alumina substrate

Equipotential layer Pt

Alumina substrate

Modulation heater

SnO2 layer

Milled out partPt-conductor tracks

Reference Equipotential ring

(a)

(b)

Figure 11.12 SnO2-based direct thermoelectric gas sensor (a) in thick-film technology.(b) The sensor was measured before and after milling out a part of the gas sensitivelayer. Reprinted from Rettig F, Moos R. Direct thermoelectric hydrocarbon gas sensorsbased on SnO2. IEEE Sens J 2007b;7:1490e1496 with permission from IEEE. © 2007 IEEE.


Jonker diagram.58 In such a plot, the thermopower, hSnO2, and the conduc-

tivity (or the resistance, RSnO2) are plotted against each other. If, after millingout areas of the film, the data points stay on the same curve as the pointsbefore milling out, the assumption would be correct. However, this is notthe case, as the curves in Fig. 11.13(c) are clearly shifted to the right afterthe milling process. The measurand thermopower is not a function of thegeometry of the gas sensitive layer. This might be advantageous for abrasivegas streams. More details of this sensor can be found in Ref. 57.

Fig. 11.14 shows photographs of two DTEGs with SnO2 as the gassensitive films. The sensors were manufactured as shown in Fig. 11.2. Thegas sensitive layer was applied with a brush to ensure low internal resistanceof the gas sensitive layer (because of the geometry independency of themeasurand thermopower, the geometry does not play a role!). The sensorsalso had a heater on the reverse side. It heated the entire sensor tip to theoperational temperature of 400 �C. The temperature modulation was

cC3H8 (ppm)

cC3H8 (ppm)

–380

–400

–420

–440

–460

–480

–500

–520

–380–400–420–440–460–480–500–520

0 20 40 60 80 100

0 20 40 60 80 100

Before milling outAfter milling out

η SnO

2( μ

VK

–1)

η SnO

2( μ

VK

–1)

RS

nO2

(Ω)

106

10–6 10–5 10–4

105

RSnO2(Ω–1)–1

(a) (b)

(c)

Figure 11.13 Measurement results of a direct thermoelectric gas sensor as shown inFig. 11.12, with SnO2 as the gas sensitive material at 400 �C with 1% oxygen withdifferent propane concentrations (balance nitrogen). The curves indicate (a) the resultsof the thermopower hSnO2, (b) the resistance, RSnO2, and (c) the Jonker diagram, before(circles) and after (triangles) milling out a part of the gas sensitive layer. Reprinted fromRettig F, Moos R. Direct thermoelectric hydrocarbon gas sensors based on SnO2. IEEE SensJ 2007b;7:1490e1496 with permission from IEEE. © 2007 IEEE.


applied by the modulation heater with a modulation frequency of 0.312 Hz.A continuous regression was used to extract the thermopower of the gassensitive layer.

Fig. 11.15 shows the results for both sensors. The propane concentrationprofile is shown in Fig. 11.15(a), and the measured thermopower of thegas sensitive layer is plotted in Fig. 11.15(c). Fig. 11.15(b) shows thecorresponding error of the thermopower, DhSnO2

�hSnO2

, determined bythe error of the regression analysis. The characteristics of the sensors canbe found in Fig. 11.15(d). Both sensors behave nearly identically, in thetransient diagram and in the sensor characteristics. For a resistive gas sensor,identical behavior would be surprising, especially when the gas sensitivelayer was applied using such a poorly reproducible technique. The errorsDhSnO2

�hSnO2

from the continuous regression are usually below 1%.When considering the modulation frequency of 0.312 Hz, this low erroris remarkable. The error can be used to check if the sensor is workingcorrectly. If the error is above a certain level for a certain time, the sensorhas to be checked. This clearly shows that accurate, rapid, and reliableDTEGs can be designed. More details about this sensor can be found inRef. 42.

As shown above, bulk materials also show a promising oxygen gasedependent thermoelectric behavior. Therefore, two DTEGs were prepared

Pt

Au

SnO2

Figure 11.14 Photograph of direct thermoelectric gas sensors with SnO2 as the gassensitive layer. Reprinted from Rettig F, Moos R. Temperature-modulated direct thermo-electric gas sensors: thermal modeling and results for fast hydrocarbon sensors. Meas SciTechnol 2009;20(6):0652059 with permission from IOP-Publishing Ltd.


with inks made from the same ceramic powder with which the brick-likespecimens of Fig. 11.11 were made. Fig. 11.16(b) shows a photograph oftwo sensors with SrTi0.6Fe0.4O3ed as the oxygen gas sensitive material.The design of the sensor (Fig. 11.16(a)) was modified for this materialbecause the diffusion barrier material SrAl2O4 is needed to prevent aninteraction of the gas sensitive material SrTi0.6Fe0.4O3ed

18 with the aluminasubstrate during firing. The SrAl2O4 has to be fired at 1300 �C for goodadhesion. The printed insulation layer introduced in Fig. 11.2 is not suitablefor such high firing temperatures. For the upper part of the sensor, theSrAl2O4 was printed and fired first on the alumina substrate. Then, thePt-conductor tracks were printed and fired. Afterward, SrTi0.6Fe0.4O3ed

(a)

(b)

(c)(d)

1.61.20.80.40.0

600500400300200100

0

– 350

– 400

– 450

– 500

– 550

– 600

– 650

0 500t (s)1000 1500 4503001500

SnO

( μV

K)

c CH

(ppm

)

CC3H8 (ppm)

(%)

ΔηS

nOη

SnO

ηS

nO (μ

VK

)

350

400

450

500

550

600

650

First sensor

Second sensor

η

Figure 11.15 Results obtained from a direct thermoelectric gas sensor with SnO2 (seeFig. 11.14) at 400 �C and 1% oxygen with different propane concentrations (balancenitrogen). The temperature modulation frequency was 0.312 Hz. The curves indicate(c) the results of the thermopower hSnO2, (b) the relative error of the thermopowerDhSnO2

�hSnO2

, and (d) the characteristics of the two gas sensors shown in Fig. 11.13.The transient propane profile is shown in (a). Reprinted from Rettig F, Moos R.Temperature-modulated direct thermoelectric gas sensors: thermal modeling and resultsfor fast hydrocarbon sensors. Meas Sci Technol 2009;20(6):0652059 with permission fromIOP-Publishing Ltd.


paste was either screen-printed or applied with a brush. Finally, for theupper substrate, the Au-conductor tracks and the equipotential layer werescreen-printed and fired in a separate step. The lower alumina substratewas printed with the heater, followed by the modulation heater. The heaterconductor tracks and one insulation layer were screen-printed and fired in asingle step. Both substrates of the sensor were joined by applying a secondwet screen-printed insulation layer on the upper substrate. After dryingand firing, the sensor was ready for the measurements. Details on thecomplex manufacturing process of this sensor can be found in Ref. 27.Please note the different methods by which the gas sensitive layer wasapplied: with a brush (Fig. 11.16(b), upper sensor, first sensor) and by screenprinting (Fig. 11.16(b), lower sensor, second sensor).

Both sensors were heated to their operational temperature of 700 �C bythe Pt heater. A temperature modulation frequency of 0.156 Hz was applied

SrTi0.6Fe0.4O3

Pt-conductor tracksSrTi0.6Fe0.4O3 layerAu-conductor tracksSrAl2O4 layerSubstrate Al2O3Equipotential layer Au

Insulation layer

Modulation heater Pt

Substrate Al2O3

Heater Pt

Heater conductor tracks Au

(a)

(b) Pt

Au

Figure 11.16 Direct thermoelectric gas sensors with SrTi0.6Fe0.4O3ed as the gassensitive material: (a) design and (b) photograph. Reprinted from Rettig F. Direktethermoelektische Gassensoren (in German) [PhD thesis]. University of Bayreuth, Shaker-Verlag; 2008 with permission from Shaker-Verlag.


to the modulation heater. The thermopower was determined by acontinuous regression analysis over two periods. Owing to this lowmodulation frequency, the sensor response time is limited to 12.8 s.

The sensors were tested with different oxygen/nitrogen mixtures.Fig. 11.17 shows the results of both sensors as described in Fig. 11.16.The oxygen partial pressure was varied stepwise from pure nitrogen topure oxygen (Fig. 11.17(a)). Both sensors behaved almost identically despitethe fact that the geometry of the gas sensitive layer differed significantly(Fig. 11.16(b)). Both sensors reached their equilibrium state at each oxygen

101

10–4

10–3

10–2

10–1

8

6

4

2

0

220

200

180

160

140

120

100

80

220

200

180

160

140

120

100

8080 400 800

t (s)

1200

pO (b

ar)

pO (bar)

μVK dec

100

10–3 10–2 10–1 100

ΔηS

TFη

STF

(%)

η STF

(μV

K)

ηS

TF (μ

VK

)

First sensor

Second sensor

(a)

(b)

(c)(d)

Figure 11.17 Thermopower results of a direct thermoelectric gas sensor with aSrTi0.6Fe0.4O3ed gas sensitive film at 700 �C in different oxygen concentrations (balancenitrogen). The temperature modulation frequency was 0.156 Hz. The black and the graycurves indicate (c) the results of the thermopower, hSTF, (b) the relative error of thethermopower, DhSTF/hSTF, (d) the characteristics of the two gas sensors shown inFig. 11.16. The transient oxygen profile is shown in (a). Reprinted from Rettig F.Direkte thermoelektische Gassensoren (in German) [PhD thesis]. University of Bayreuth,Shaker-Verlag; 2008 with permission from Shaker-Verlag.


concentration within 50 s. The relative error of the measured thermopower(Fig. 11.17(b)) was usually below 2%. However, the gas sensitive materialhad a relative low sensitivity toward oxygen. The slope in a half-logarithmic plot is about �28 mVK�1 per decade pO2. The material isknown to have a slope of about �0.2 in a double logarithmic plot of theresistance versus the oxygen partial pressure. Therefore, the expected slopeof a DTEG would be about 40 mVK�1 per decade.59 As already mentioned,the hole concentration changes in different oxygen concentrations. As aresult, both the thermopower and the conductivity change. For a p-typematerial, like the one shown here, the slope in a logelog plot of resistanceversus oxygen partial pressure can be transferred to a slope in a plot of thethermopower and the logarithmic oxygen partial pressure.

In Fig. 11.11, this value was almost found. The reason for this deviationfrom the theoretically expected slope is not clear. However, if one comparesFigs. 11.11 and 11.17 more closely, one finds that also the absolute value ofthe thermopower in Fig. 11.17 is lower than that in Fig. 11.11, e.g., h (1 bar,Fig. 11.11) z 108 mVK�1, whereas h (1 bar, Fig. 11.17) z 85 mVK�1.This behavior was also found in Ref. 60. In this case, it was assumed thatthe temperature difference was not measured at the same point where thethermovoltage was determined. In other words, the temperature difference(DT in Eq. 11.4) was determined too large; hence, the thermovoltage, aswell as the slope in the h (pO2) plot, appeared too small. Therefore, inRef. 60, a geometry correction factor was introduced and proven. Assumingthe geometrical situation and a linear temperature gradient, the slope and theabsolute thermopower could be increased by a factor of 1.1.27 However, thevalues of the bulk sample are not achieved. The different sintering temper-ature for the screen-printed layer (1100 �C) and the bulk sample (1300 �C)might be a reason. Also, the ionic thermopower of the material54 maycontribute to the deviation from theory because, despite the material beinga p-type semiconductor, ionic conductivity contributes to the electricalcharge transport in a nonnegligible way. Overall, SrTi0.6Fe0.4O3ed can beused for DTEGs; however, the sensitivity is low.

The last sensor presented in this section has much higher sensitivity.According to the previous section, intrinsic gas sensitive thermoelectricmaterials exhibit a larger sensitivity. As intrinsic material, Fe2O3 was used,as this material is known to have intrinsic semiconducting properties atlow temperatures.61,62 This intrinsic semiconducting behavior makes thematerial unsuitable for conductometric gas sensors, as there is almost nochange in conductivity with changing ambient pO2. However, it is an ideal


candidate for DTEGs. The design of the sensor (Fig. 11.18(a)) is quite similarto that shown in Fig. 11.2. The conductivity of an intrinsic semiconductingmaterial is typically low because only the intrinsic charge carriers contributeto the electrical conduction. The developed transducer for a DTEG allows amaximum internal resistance of the gas sensitive material of about 1 MU. Ifthis range is exceeded, the measured thermovoltage becomes too noisybecause of disturbance from ambient voltages. For Fe2O3, the internalresistance of the gas sensitive layer is higher than 1 MU with a gas sensitivelayer 4 mm long. For this reason, the distance between the two thermo-couples was reduced to about 1.2 mm. As a result, the internal resistanceis sufficiently low for an accurate evaluation of the thermopower.Unfortunately, however, the reproducibility of different DTEGS suffersbecause of the small distance between the thermocouples. The point wherethe temperature difference is determined is typically not the same point atwhich the thermovoltage of the gas sensitive layer is read out. The gassensitive layer was applied to the transducer with a brush (Fig. 11.18(b)).Fig. 11.18(c) shows a photograph of the complete gas sensor.

The sensor was heated up to 580 �C and a temperature modulationfrequency of 0.312 Hz was applied. The continuous linear regression to

Gas sensitive layer and thermocouples

Insulation layer

Equipotential layer

Insulation layer

Modulation heater

Alumina substrate

Main heater

Cover layer

Thermocouple 1

Gas sensitive layer

Thermocouple 2

6 m

m

50 mm

(a)

(c)

(b)

Figure 11.18 Direct thermoelectric gas sensors with Fe2O3 as the gas sensitive mate-rial: (a) design and (b) photograph. Reprinted from Rettig F, Moos R. a-iron oxide: anintrinsically semiconducting oxide material for direct thermoelectric oxygen sensors.Sensor Actuator B Chem 2010;145(2):685e690 with permission from Elsevier.


determine the thermopower was carried out for two periods; therefore, theresponse time of the sensor is limited to a regression time of 6.4 s. The tem-perature difference on the gas sensitive layer varied from�15 toþ5 �Cwithrespect to the sensor temperature of 580 �C. Within this temperature differ-ence range, the thermovoltage, DVgsf, correlated almost linearly with thetemperature difference, DT. The temperature at both ends of the gas sensi-tive layer and the thermovoltage of the gas sensitive layer were determinedwith a 90 ms interval. Fig. 11.19 shows the results of a typical measurementrun. The oxygen concentration was varied from pure nitrogen to pure ox-ygen in seven steps (Fig. 11.19(a)). The determined thermopower, hFe2O3

,varied from about �400 mVK�1 to �50 mVK�1 (Fig. 11.19(c)). The sensorreaches its equilibrium state within a few seconds. A more detailed timelyanalysis of the sensor response concluded that the measured response time

10

10101010

(a)

(b)

(c)(d)

5

34

210

–100

0

–200

–300

–400

–500

–100

0

–200

–300

–400

–500150 300 450 600

t (s)

p O (b

ar)

pO (bar)

10

10 10 10 10

|Δη

FeO

| (μV

K)

ηFe

O (μ

VK

)

ηFe

O (μ

VK

)

85 μVK dec

First sensor

Second sensor

Third sensorFirst sensor

Figure 11.19 Experimental results of the direct thermoelectric gas sensor with Fe2O3

as an oxygen sensitive layer at 580 �C (sensor from Fig. 11.18). The temperaturemodulation frequency was 0.312 Hz. Fig. 11.19(a) shows the transient oxygen profile,(b) shows the absolute error of the thermopower, DhFe2O3, (c) depicts the transientresult of the thermopower, hFe2O3, and (d) gives the characteristics of the three gassensors. Reprinted from Rettig F, Moos R. a-iron oxide: an intrinsically semiconductingoxide material for direct thermoelectric oxygen sensors. Sensor Actuator B Chem 2010;145(2):685e690 with permission from Elsevier.


(t63%) of 16 s can be assigned to the gas exchange in the test chamber. Theabsolute error of the thermopower

��DhFe2O3

�� (Fig. 11.19(b)) is usually below2 mVK�1. The error exceeds this upper limit only after the stepwise of theoxygen concentration changes. The sensor is quite rapid, even the irregular-ities in the gas dosing at t ¼ 400 s are partially determined by the gas sensor.The major advantage of this gas sensor is its far higher sensitivity comparedwith the sensor shown in Fig. 11.17 (compare Fig. 11.19(d) withFig. 11.17(d)). The sensitivity reaches a value of about 85 mVK�1 per decadeand is about three times higher compared with the sensor based onSrTi0.6Fe0.4O3ed. More details on this sensor can be found in Ref. 63.

It was shown in this section that it is possible to manufacture accurate,rapid, and sensitive DTEGs. The design of the DTEGs can be developedknowledge-based. As intrinsic materials show the best sensitivity, theinternal resistance of the gas sensitive layers has to be considered, and theinsulation and equipotential layers have to be applied. An appropriatetemperature modulation frequency needs to be selected to achieve goodresults.

11.2.4 Ionic direct thermoelectric gas sensorsThe DTEGs introduced in the preceding sections were based on semi-conducting oxide materials. However, other materials besides electronicconductors can be employed as materials for DTEGs. Several years ago, itwas shown that the thermopower h of an electrochemical cell with Pt elec-trodes separated by an oxygen ion conductor follows Eq. (11.23) (e.g., 64):

h¼ SðTÞ � kB4e

� lnðpO2Þ �Q•

O2�

2eT� hPt (11.23)

In Eq. (11.23), Q•O2� is the heat of transport of the oxygen ions, S(T) is the

entropy term, and hPt is the Seebeck coefficient of platinum. In rough first-order approximation, all these three terms can be considered as constantwith respect to pO2. Then, a theoretical sensitivity s of the thermoelectriccell can be derived:

s¼ dhd logðpO2Þ ¼ �kB

4eln10 (11.24)

According to Eq. (11.24), the sensitivity should be s z �50 mVK�1 perdecade pO2, which is in the same order of magnitude as that ofsemiconductor materials.


The first implementation of such a sensor device is reported in Ref. 60,in which 8 mol% Y2O3-stabilized zirconia was used for the gas sensitivematerial. The sensor setup was similar to the sensors described above;however, an additional Pt-cermet was applied to get a high exchange rateat the YSZePt interface. It becomes clear from the results in Fig. 11.20that the sensitivity reaches the expected value. Astonishingly, but inaccordance with Eq. (11.24), almost no temperature dependency of h wasobserved. This indicates that the three pO2-independent terms either havea negligible temperature dependency or their temperature dependenciescompensate each other. Additionally, no cross-sensitivities to NO, H2,H2O, CO, CO2, or HC are observed. However, the long-term stabilityof this sensor has to be improved.

The perovskite-type proton conductor BaCe0.95Y0.05O3ed has also beenconsidered for ionic direct thermoelectric gas sensors.65 Although ahydrogen-dependent thermopower could be measured, the different mobilespecies (ions, electrons, holes) allow the material to apply only in certainatmospheres with defined oxygen and hydrogen partial pressures.

600

550

500

ηY

SZ

(µV

K–1)

450

–3.0 –2.5 –2.0 –1.5log pO2 (bar)

–1.0 –0.5 0.0 0.5

–50 μVK–1 per decade

700°C750°C800°C

Figure 11.20 Results of an ionic direct thermoelectric gas with yttria-stabilized zirconiaas an oxygen sensitive layer. Note the almost nonexistent temperature dependence ofthe sensor signal. Reprinted from R€oder-Roith U, Rettig R, R€oder T, Janek J, Moos R, SahnerK. Thick-film solid electrolyte oxygen sensors using the direct ionic thermoelectric effect.Sensor Actuator B Chem 2009;136(2):530e535 with permission from Elsevier.



DTEGs are an alternative to resistive gas sensors. Accurate, rapid, andlong-term stable gas sensors have been presented in this chapter. Themain advantage of DTEGs is the measurand “thermopower” or “Seebeckcoefficient.” In contrast to conductometric gas sensors, the measurandthermopower is not influenced by changes in the geometry of the gassensitive layer. A damage of the gas sensitive layer directly influences theresistance, but the thermopower remains virtually unaffected. An exampleof such an abrasion-resisting gas sensor is shown in Fig. 11.20.

Besides the sensors discussed, DTEGs have been developed with specialrespect to the measurement principle. First, an adequate transducer has beendeveloped. For the DTEGs presented, a temperature modulation techniquewas chosen to determine the thermopower. The advantage of this techniqueis improved accuracy and that the signals of the temperature differences andthe thermovoltages can be analyzed either by regression analysis or by aFourier analysis. Disturbing voltages are filtered out by these signal analyses.The disadvantage of the temperature modulation is the long response timeof the sensor, which is determined here by the regression or the Fourieranalysis. The problem can be overcome by rapid temperature modulationwith a modulation heater placed within a distance of about 60 mm fromone end of the gas sensitive layer. The thermal behavior of a DTEG hasbeen modeled; both the thermal model and measurement of the thermalproperties of the DTEGs agree very well.

The gas sensitive layer of a DTEG determines the performance of thegas sensor. A general analysis of materials with chemisorption has been intro-duced. The model is based on semiconductor and thermoelectric equations.The partial differential equations (here PoissoneBoltzmann equations) havebeen solved by Comsol Multiphysics. The solution has been used tocalculate the isothermal and the nonisothermal properties of a gas sensitivegrain. The results of the simulation concluded that small grains are generallyadvantageous because of higher sensitivity. Furthermore, materials with onlyan intrinsic charge carrier density should have the largest sensitivitycompared with n-type or p-type semiconducting materials.

Four different gas sensors based on SnO2, SrTi0.6Fe0.4O3ed, Fe2O3, andYSZ demonstrated the potential of DTEGs. The classical material SnO2 wastested with propane, while the other materials were used as oxygen sensitivematerials. The temperature modulation frequency for the different materialswas 0.312 Hz. The response time of the sensors was determined by the signal


analyses of thermovoltage and temperature difference. It amounted to 6.4 s.All the sensors showed a reproducible and rapid behavior. For the DTEGs, itwas shown that it is possible to mill out a significant part of the gas sensitivelayer without affecting the measured thermopower significantly. With anadapted design, DTEGs may be a good alternative to resistive gas sensors.

Besides the encouraging results, DTEGs still have great potential forfurther improvements. The temperature modulation frequency of0.312 Hz is not sufficient for all applications. For fast responding devices,the temperature modulation has to be in the range of 100 Hz.Manufacturing technology needs to be adjusted to achieve this. Micro-machined ceramics or silicon hot plate gas sensors may be preferred becausesuch a high-temperature modulation frequency needs quite small structuresthat may not be feasible with conventional ceramic thick-film technology.

Until now, only a few materials have been studied for DTEGapplication. The research focus for gas sensitive materials is usually onresistive materials. The possibility of intrinsic materials with an enhancedsensitivity (when applied in the DTEG mode) is an important propertythat needs to be addressed in the future. Intrinsic (low conducting) behaviorcan be improved, when materials with a high mobility of charge carriers areused. Then, films with a low internal resistance and a high sensitivity can beobtained. Furthermore, the aspect of utilizing ion conducting materialsshould be more emphasized because a high selectivity can be expectedbecause of the distinct ion conduction.

References1. Riegel J, Neumann N, Wiedenmann HM. Exhaust gas sensors for automotive emission

control. Solid State Ionics 2002;152:783e800.2. Moos R. A brief overview on automotive exhaust gas sensors based on electroceramics.

Int J Appl Ceram Technol 2005;2:401e13.3. Denk I, Ingrisch K, Weigold T, Baumann K, Weiblen K, Bauer M, Zeppenfeld A,

Schuhmann B. New CO/NOx-sensor system for automotive climate control in a smallsize housing with high mounting flexibility. Sensor 99 Proc 1999:339e44.

4. Williams DE. Semiconducting oxides as gas-sensitive resistors. Sensor Actuator B Chem1999;57:1e16.

5. Yamazoe N. Toward innovations of gas sensor technology. Sensor Actuator B Chem2005;108:2e14.

6. Taguchi N. Gas detecting element and making of it. 1970. US patent specification,US3644795.

7. Logothetis EM, Kaiser WJ. TiO2 film oxygen sensors made by chemical vapourdeposition from organometallics. Sensor Actuator 1983;4:333e40.

8. Takami A. Development of titania heated exhaust-gas oxygen sensor.AmCeram Soc Bull1988;67(12):1956e60.


9. Gerblinger J, Meixner H. Fast oxygen sensors based on sputtered strontium-titanate.Sensor Actuator B Chem 1991;4(1e2):99e102.

10. Sch€onauer U. Response-times of resistive thick-film oxygen sensors. Sensor Actuator BChem 1991;4(3e4):431e6.

11. Moos R, Menesklou W, Schreiner HJ, H€ardtl KH. Materials for temperatureindependent resistive oxygen sensors for combustion exhaust gas control. Sensor ActuatorB Chem 2000;67(1e2):178e83.

12. Cantalini C, Pelino M, Sun HT, Faccio M, Cantucci S, Lozzi L, Passacantando M.Cross sensitivity and stability of NO2 sensors from WO3 thin film. Sensor Actuator BChem 1996;35(1e3):112e8.

13. Lampe U, Fleischer M, Meixner H. Lambda measurement with Ga2O3. Sensor ActuatorB Chem 1994;17(3):187e96.

14. Jayaraman V, Gnanasekar KI, Prabhu E, Gnanasekaran T, Periaswami G. Preparationand characterisation of Cr2exTixO3þd and its sensor properties. Sensor Actuator B1999;55(2e3):175e9.

15. Niemeyer D,Williams DE, Smith P, Pratt KFE, Slater B, Catlow CRA, Stoneham AM.Experimental and computational study of the gas-sensor behaviour and surfacechemistry of the solid-solution Cr2exTixO3 (x < ¼ 0.5). J Mater Chem 2002;12(3):667e75.

16. Ruiz AM, Sakai G, Cornet A, Shimanoe K, Morante JR, Yamazoe N. Cr-doped TiO2gas sensor for exhaust NO2 monitoring. Sensor Actuator B Chem 2003;93(1e3):509e18.

17. Chou SM, Teoh LG, Lai WH, Su YH, Hon MH. ZnO/Al thin film gas sensor fordetection of ethanol vapor. Sensors 2006;6:1420e7.

18. Moos R, Rettig F, H€urland A, Plog C. Temperature-independent resistive oxygenexhaust gas sensor for lean-burn engines in thick-film technology. Sensor Actuator BChem 2003;93(1e3):43e50.

19. Rettig F, Moos R, Plog C. Poisoning of temperature independent resistive oxygen sen-sors by sulfur dioxide. J Electroceram 2004;13(1e3):733e8.

20. Gerblinger J, Lampe U, Meixner H. German patent specification. 1993. DE4339737C1.21. Rettig F, Moos R, Plog C. Sulfur adsorber for thick-film exhaust gas sensors. Sensor

Actuator B Chem 2003;93(1e3):36e42.22. Ivers-Tiffee E, H€ardtl KH, Menesklou W, Riegel J. Principles of solid state oxygen

sensors for lean combustion gas control. Electrochim Acta 2001;47(5):807e14.23. Somov SI, Guth U. A parallel analysis of oxygen and combustibles in solid electrolyte

amperometric cells. Sensor Actuator B Chem 1998;47:131e8.24. Baunach T, Sch€anzlin K, Diehl L. Sauberes Abgas durch Keramiksensoren. Phys J 2006;

5(5):33e8.25. Heikes RR, Ure RW. Thermoelectricity. Interscience Publishers; 1961.26. Choi GM, Tuller HL, Goldschmidt D. Electronic-transport behavior in single-

crystalline Ba0.03Sr0.97TiO3. Phys Rev B 1986;34(10):6972e9.27. Rettig F. Direkte thermoelektische Gassensoren (in German). PhD thesis, University of

Bayreuth, Shaker-Verlag; 2008.28. Nagy PB, Nayfeh AH. On the thermoelectric magnetic field of spherical and cylindrical

inclusions. J Appl Phys 2000;87:7481e90.29. Shin W, Matsumiya M, Izu N, Murayama N. Hydrogen-selective thermoelectric gas

sensor. Sensor Actuator B Chem 2003;93(1e3):304e8.30. Willet M. Recent developments in catalytic gas sensors. In: 1st international workshop an

smart gas sensors technology an application. Germany: Freiburg im Breisgau; 2005.31. Balducci A, D’Amico A, Di Natale C, Marinelli M, Milani E, Morgana ME, Pucella G,

Rodriguez G, Tucciarone A, Verona-Rinati G. High performance CVD-diamond-based thermocouple for gas sensing. Sensor Actuator B Chem 2005;111:102e5.


32. Pisarkiewicz T, Stapinski T. Influence of gas atmosphere on thermopower measure-ments in tin oxide thin-films. Thin Solid Films 1989;174:277e83.

33. Siroky K. Use of the Seebeck effect for sensing flammable-gas and vapors. SensorActuator B Chem 1993;17(1):13e7.

34. Mizsei J. H2-induced surface and interface potentials on pd-activated SnO2 sensor films.Sensor Actuator B Chem 1995;28(2):129e33.

35. Moos R. Method and apparatus for detecting the oxygen content of a gas. US patentspecification; 1998US6368868.

36. Ionescu R. Combined Seebeck and resistive SnO2 gas sensors, a new selective device.Sensor Actuator B Chem 1998;48(1e3):392e4.

37. Liess M, Steffes H. The modulation of thermoelectric power by chemisorption: a newdetection principle for microchip chemical sensors. J Electrochem Soc 2000;147(8):3151e3.

38. Smulko JM, Ederth J, Li YF, Kish LB, Kennedy MK, Kruis FE. Gas sensing by thermo-electric voltage fluctuations in SnO2 nanoparticle films. Sensor Actuator B Chem 2005;106(2):708e12.

39. Keem JE, Honig JM. Seebeck measurements and their interpretation in high-resistivitymaterials case of semiconducting V2O3. Phys Status Solidi A-Appl Res 1975;28(1):335e43.

40. Keithley. Low level measurements handbook. Keithley Instruments Inc; 2004.41. Timm H, Janek J. On the Soret effect in binary nonstoichiometric oxide-skinetic

demixing of cuprite in a temperature gradient. Solid State Ionics 2005;176:1131e43.42. Rettig F, Moos R. Temperature-modulated direct thermoelectric gas sensors:

thermal modeling and results for fast hydrocarbon sensors. Meas Sci Technol 2009;20(6). 065205 9.

43. Rettig F, Moos R. Direct thermoelectric gas sensors: design aspects and first gas sensors.Sensor Actuator B Chem 2007;123(1):413e9.

44. Simon T, Barsan N, Bauer M, Weimar U. Micromachined metal oxide gas sensors:opportunities to improve sensor performance. Sensor Actuator B Chem 2001;73(1):1e26.

45. Kittel C, Kr€omer H. Thermal physics. W. H. Freeman; 1980.46. Jamnik J, Kamp B,Merkle R,Maier J. Space charge influenced oxygen incorporation in

oxides: in how far does it contribute to the drift of Taguchi sensors? Solid State Ionics2002;150(1e2):157e66.

47. Rettig F, Moos R. Morphology dependence of thermopower and conductance insemiconducting oxides with space charge regions. Solid State Ionics 2008b;179(39):2299e307.

48. Maier J. Physical chemistry for ionic material: ions and electrons in solids. John Wiley & SonsLtd; 2004.

49. Tsch€ope A. Interface defect chemistry and effective conductivity in polycrystallinecerium oxide. J Electroceram 2005;14:5e23.

50. Henisch HK. Semiconductor contacts. Oxford University Press; 1984.51. Barsan N. Conduction models in gas-sensing SnO2 layers: grain-size effects and ambient

atmosphere influence. Sensor Actuator B Chem 1994;17(3):241e6.52. Wolkenstein T. Electronic processes on semiconductor surfaces during chemisorption. New

York: Consultants Bureau; 1991.53. Yoo HI, Song CR. Thermoelectricity of BaTiO3þd. J Electroceram 2001;6:61e74.54. Rothschild A, Menesklou W, Tuller HL, Ivers-Tiffee E. Electronic structure, defect

chemistry, and transport properties of SrTi1exFexO3ey solid solutions. Chem Mater2006;18:3651e9.

55. Moos R, Izu N, Rettig F, Reiß S, ShinW,Matsubara I. Resistive oxygen gas sensors forharsh environments. Sensors 2011;11(4):3439e65.


56. Rettig F, Sahner K, Moos R. Thermopower of LaFe1exCuxO3ed. Conf Proc Solid StateIonics 2005;15:569. Baden-Baden.

57. Rettig F, Moos R. Direct thermoelectric hydrocarbon gas sensors based on SnO2. IEEESens J 2007b;7:1490e6.

58. Jonker GH. Application of combined conductivity and Seebeck-effect plots for analysisof semiconductor properties. Philips Res Rep 1968;23(2):131e8.

59. Moos R, H€ardtl KH. Defect chemistry of donor doped and undoped strontium titanateceramics between 1000 �C and 1400 �C. J Am Ceram Soc 1997;80:2549.

60. R€oder-Roith U, Rettig R, R€oder T, Janek J, Moos R, Sahner K. Thick-film solidelectrolyte oxygen sensors using the direct ionic thermoelectric effect. Sensor ActuatorB Chem 2009;136(2):530e5.

61. Gurlo A, Barsan N, Oprea A, Sahm M, Sahm T, Weimar U. An n- to p-typeconductivity transition induced by oxygen adsorption on Fe2O3. Appl Phys Lett2004a;85(12):2280e2.

62. Gurlo A, SahmM, Oprea A, Barsan N,Weimar U. A p- to n-transition on Fe2O3-basedthick film sensors studied by conductance and work function change measurements.Sensor Actuator B Chem 2004b;102(2):291e8.

63. Rettig F, Moos R. a-iron oxide: an intrinsically semiconducting oxide material fordirect thermoelectric oxygen sensors. Sensor Actuator B Chem 2010;145(2):685e90.

64. Ahlgren EO, Poulsen FW. Thermoelectric power of stabilized zirconia. Solid State Ionics1995;82:193e201.

65. R€oder-Roith U, Rettig F, Sahner K, R€oder T, Janek J, Moos R. Perovskite-typeproton conductor for novel direct ionic thermoelectric hydrogen sensor. Solid StateIonics 2011;192(1):101e4.

66. Williams D, Tofield B, McGeehin P. Oxygen sensors. European patent specification;1985. EP00062994.


CHAPTER TWELVE

Dynamic operation ofsemiconductor sensorsAndreas Sch€utze, Tilman SauerwaldLab for Measurement Technology, Department Systems Engineering, Saarland University, Saarbr€ucken,Germany

Contents

12.1 Introduction 38512.2 Dynamic operation of metal oxide semiconductor gas sensors 388

12.2.1 Temperature-cycled operation 39012.2.2 Field effect 39612.2.3 Optical excitation 398

12.3 Dynamic operation of gas-sensitive field-effect transistors 39812.3.1 Temperature-cycled operation for SiC-FET sensors 39912.3.2 Gate biasecycled operation 40112.3.3 Current compensation mode 40312.3.4 Combined methods 403

12.4 Conclusion and outlook 404References 408

12.1 Introduction

Semiconductor gas sensors offer a range of advantages. Especially theirhigh sensitivity and robust long-term performance combined with low costmake them attractive for various applications. On the other hand, they alsopose considerable challenges due to their typically low selectivity and poorstability, i.e., baseline drift and changes in sensitivity. Note that high robust-ness and poor stability are not always a contradiction. For instance, metaloxide semiconductor (MOS) gas sensors have a proven lifetime of severaldecades for detection of explosive gas leaks, i.e., high concentrationsexceeding typical ambient variations, and for monitoring air quality ormore correctly sudden changes in air quality in cars just to mention twomain applications. In both cases, the application does not require a stablebaseline or constant sensitivity for determination of a gas concentration.



https://doi.org/10.1016/B978-0-08-102559-8.00012-4

Stability problems can be overcome with suitable measurement setups.In physical sensors, differential measurements are often used to suppresstemperature cross sensitivity (e.g., force and pressure sensors using Wheat-stone bridges) or determine the sensitivity (e.g., magnetic sensors withbuilt-in excitation coils). In chemical sensors, pellistors and mass-sensitivedevices often use the same approach with a differential setup combiningone gas-sensitive and one inert transducer.

Selectivity, on the other hand, is typically not a problem for physicalsensors simply due to the much smaller number of relevant factorsinfluencing the sensor signal (or, more abstract, due to the lower dimension-ality of the input space). In chemical sensors, the number of relevant factors(or the dimensionality of the input space) is huge: each molecule is basicallyan independent factor and even seemingly simple environments like indoorair or breath often contain hundreds of gas components with relevant targetgases covering a wide concentrations range from ppb level, e.g., benzene, upto several percent, e.g., humidity. To increase the selectivity of chemicalsensor systems, biomimetic approaches are often used to emulate the(mammalian) nose.1 The most important approach is the use of multisensorarrays combining various more or less specific sensors and using patternrecognition methods to interpret the resulting signal pattern.2,3 Note,however, that both the nose and multisensor arrays still have severelimitations: the response is strongly nonlinear, different gases or gas mixturescan lead to identical results and the odor response changes due to accommo-dation, saturation, or poisoning effects. One should also point out that theresponse spectrum of noses has been adapted by evolution: it tends to blankout molecules such as H2, CO, CH4, or H2O, not because these cannot bedetecteddour nose detects other small molecules such as NH3, CH2O(formaldehyde), and especially H2S quite sensitivelydbut probably due totheir low specificity or information content. Note that multisensor arrayscan come in different forms from actual n physical sensors to a single sensorelement with multiple electrodes. Furthermore, this can include electricalmultiparameter readout (EMR) methods, i.e., current or voltage sweepsor measurement at different frequencies up to impedance spectroscopy: inthese cases, additional information about the nonohmic behavior of thesensor is obtained which can reflect the interaction of the gas atmospherewith the sensor surface and thus provide additional information for gasidentification.

Many methods are used to interpret the response patterns of multisensorarrays.4e6 Especially artificial neural networks have been widely used to

386 Andreas Sch€utze and Tilman Sauerwald

emulate the data processing in our brain when recording odors, but inprinciple all methods used in pattern analysis, e.g., for image interpretationor speech recognition, can be also applied to multisensor arrays. Systematicdata evaluation is based on four typical steps: (i) data preprocessing, e.g., toreduce noise or eliminate offset; (ii) feature extraction to extract informationfrom the raw data; (iii) feature selection or more generally dimensionalityreduction to limit the data to relevant information; and (iv) classificationand/or quantification to interpret the data. In addition, as patternrecognition is based on learning from examples rather than model-basedinterpretation, validation is required to ensure that patterns are not over-interpreted. While achieving impressive selectivity for a wide range ofapplications, multisensor arrays often exacerbate the limited stability of gassensor systems: if a single sensor in an array has changed due to drift orpoisoning or is otherwise not in its calibrated state, the pattern interpretationwill lead to different, often completely false results.

Another very powerful approach to achieve differential measurements isbased on dynamic excitation of the sensor, i.e., the sensor is changing fromone state to another due to external variations. For chemical sensor, thesimplest and already quite powerful approach is to measure the sensor signalwith and without the influence of the target gas or gas mixture or at leastusing a controlled variation.7,8 The time response will provide the relativechange of the sensor signal, i.e., the conductance for MOX sensors. Inaddition, the time constant or slope can be observed which reflects therate of interaction between sensor and gas and components of gas mixturescan be identified by their different time constants. Furthermore, the initialslope of the sensor response is often linear over a much larger concentrationrange than the steady-state response as nonlinearity is caused by the limitednumber of interaction sites on the sensor surface limiting the signal at highercoverage. While this method is quite useful for lab evaluation of sensors, i.e.,to elucidate the interaction mechanism, its use for practical application islimited due to the requirement of a reference or zero air leading to morecomplex and costly systems. Note that this approach is also inspired bynature to emulate the sniffing of dogs where the dynamic also providesadditional information. Although this method, also referred to as transientresponse or breathing mode,9,10 has gained increasing interest, this chapterwill focus on dynamic sensor operating modes that can be directly controlledelectronically to allow full control over the operating mode and make use ofthe information gain in signal evaluation.

Dynamic operation of semiconductor sensors 387

The first, but still most relevant and widely studied, approach fordynamic operation is temperature modulation,11e13 also referred to astemperature-cycled operation (TCO) due to its inherent repetitive nature.Temperature as the most important physical parameter influencing chemicalinteraction is a natural candidate for dynamic operation of chemical sensors.While it can in principle be applied to any chemical sensor principle, MOXsensors are prime candidates due to their operation at elevated temperatureusing integrated heaters. It has also been shown for pellistor-typesensors,14,15 and gas-sensitive field-effect transistors (GasFETs),16 especiallythose based on SiC due to their operation at elevated temperatures.17 Otherdynamic operating modes are based on the field effect/polarization18e20 andon optical excitation.21,22 Note that, as for the temperature, these parametersare most often used to simply determine the optimal operating point forstatic operation, but all actually change the equilibrium on the sensor surfaceand thus provide additional information especially during nonequilibriumstates. The general approach is discussed using various terms, i.e., modula-tion,12,13 transient analysis,23,24 dynamic response,25,26 programming,27 orpulsed operation.28,29 We prefer the term dynamic operation as this clearlyimplies the active variation of a control parameter (temperature, bias voltage,illumination, etc.) by the sensor electronics allowing application-specificoptimization of the sensor system performance. The following sectionswill discuss dynamic operating modes for MOX and GasFET sensors toshow that not only selectivity can be improved but that stability and evensensitivity benefit from this approach.

12.2 Dynamic operation of metal oxidesemiconductor gas sensors

MOS gas sensors detect redox reactions of gases on the semiconductorsurface. These can be direct reactions of the gas with surface states(EleyeRideal mechanism) or indirect reactions requiring the adsorptionof the gas followed by a subsequent reaction (LangmuireHinshelwoodmechanism). Any redox reaction causes a change in surface charge andtherefore in the conductance of the sensor. To understand the impact ofsurface charge on the conduction, the conduction mechanism needs to bediscussed. In general, a surface charge has to be compensated by an oppositecharge in the semiconductor. In the case of n-type semiconductors, e.g., tindioxide, SnO2, the most widely used material for MOX sensors, negativesurface charges will be compensated by a positive space charge layer.


The space charge is accompanied by an electrostatic bending of all electronicbands and especially a bending of the conduction band. The height of theband bending Vs and the corresponding energy Eb ¼ qVs can be calculatedby Poisson’s equation as a function of the density of negative surface chargeNs, the density of positive charges in the semiconductor given by thenumber of (ionized) donors Nd, the permittivity of the semiconductor εr ;the dielectric constant ε0, and the elementary charge q (Eq. 12.1).

Eb ¼ q2N2s

2εrε0Nd(12.1)

Please note that Eq. (12.1) was derived as 1D solution (plane surface)using the Schottky approximation. Because of the band bending, thenumber of electrons at the surface ns is significantly reduced; it can beestimated by a Boltzmann equation based on the implicit assumption thatall donors are ionized (Eq. 12.2).

ns ¼ Nd exp

�� Eb

kbT

�(12.2)

The conductance model for these surface effects obviously depends onthe morphology of the sensor film. For common sensors with a granularthick film, it is often assumed that the conductance is completely dominatedby the surface charge and the resulting energy barrier. Therefore, theconductance (Eq. 12.3) can be modeled by a single grain to graincontact.30e32

G¼G0$e� Eb

kbT (12.3)

here G0 denotes the conductance of the granular film for the (hypothetical)flat band case.

Despite the large variety of possible reaction processes on the surface,the surface coverage in an equilibrium can typically be calculated by amass action law. An important case, the reaction of reducing gases on anMOX sensor surface, has been studied intensively.32e34 In this case, areducing gas R reacts (Eq. 12.4) with ionosorbed oxygen, which is itselfadsorbed at a surface site s (Eq. 12.5), in a direct reaction.

O�x þR#ROx þ e� (12.4)

sþ 12O2 þ e�#O�

ads (12.5)


Of course, Eqs. (12.4) and (12.5) only show a simplified reaction schemefor a direct reaction of the reducing gas with a single type of ionosorbedoxygen. A more general discussion of reaction schemes can be found inthe studies by Ref. 33 and Ref. 34. Moreover, even an additional electrontransfer to the semiconductor yielding an accumulation layer has beenreported at high excess of reducing gas at the surface.35,36

However, in most cases, i.e., in air with only a small concentration ofreducing gas, the number of free electrons at the surface is limiting theadsorption of the oxygen, which is itself dependent on the band bendingcaused by oxygen adsorption. This limitation, also known as Fermi levelpinning, is the reason that the actual density of adsorbed oxygen is onlychanging very little at the equilibrium of reactions (12.4) and (12.5). Onthe other hand, even small changes in surface charge cause quite largechanges in the MOS sensor conductance (cf. Eq. 12.3). The sensor responseSeq (the subscript eq is indicating the equilibrium condition) defined asthe change of conductance given by the conductance under reducing gasGgas divided by the conductance in air Gair can be estimated in equilibriumcondition by a power law.32,33

Seq¼Ggas

Gair� 1 ¼ kcngas (12.6)

The exponent n is equal to 0.5 for the basic reaction scheme shown inEqs. 12.4 and 12.532 and can have other rational values for other, morecomplex reaction schemes.33,34

12.2.1 Temperature-cycled operationIt is obvious that temperature has a strong impact on numerous effects inMOX sensors, among others on the rate constants of the redox reactions.For most redox processes, an activation energy needs to be overcome beforethe reaction can take place. Gases can therefore be discriminated withrespect to their activation energy. In 1974, the principle was utilized forthe first time in a procedure to selectively detect carbon monoxide andhydrocarbons, mainly methane, in firedamp.11 In the following years,many studies using TCO were published12,25,37e41 reporting the use invarious application fields such as fire detection, air quality sensing, and inthe detection of emissions. In an earlier work, we have demonstrated theselectivity achieved with TCO41 for detection of three hazardous VOCs(volatile organic compounds), benzene, formaldehyde, and naphthalene,at very low concentrations. The compounds were tested at two


concentrations, one below and one above the recommended threshold limitvalues in air: the tested concentrations were 0.5 and 5 ppb for benzene,10 and 100 ppb for formaldehyde, and 2 and 20 ppb for naphthalene. Allconcentrations were tested at various ambient humidity levels and in thepresence of 0, 500, or 2000 ppb ethanol as a typical, nonhazardous VOC.Three different MOS sensors (GGS 1330, GGS 2330, and GGS 5330, allfrom UST, Umweltsensortechnik, Geschwenda, Germany) have beenused each with a specific temperature cycle. For each sensor, a set of 40features were calculated from the sensor response during the temperaturecycle, describing the shape of sensor conductance. Using these shape-describing features, a linear discriminant analysis (LDA) was performed.LDA is a supervised training method that maximizes the distance betweendifferent groups with respect to the scattering within the groups.4 Theresulting LDA projections for the each of the three sensors and forthe data fusion of all sensors are shown in Fig. 12.1. The discrimination ofthe three toxic gases from each other and from air is quite successful forall three sensors, but there is still some overlap. For the validation of theresults, leave-one-out cross validation (LOOCV) is performed4 indicatingin all cases a correct classification of over 95% of the measurements. Datafusion of the three sensors improves the discrimination further (Fig. 12.1,lower right). No overlap is observed even in the 2D-scatterplot and,consequently, LOOCV results in 100% correct classification.

Many investigations use a somewhat haphazard way of defining thetemperature profile of TCO rather than a consistent optimization process,due to the numerous effects that make the derivation of a universal modelvery challenging. However, in the last few years, the advantages of amodel-based optimization became obvious. To allow an objective compar-ison of various temperature profiles, the definition of the sensor responseis extended to the concept of a quasi-static sensor response SqsðtÞ at awell-defined time t within each temperature cycle (Eq. 12.7).

SqsðtÞ¼GgasðtÞGairðtÞ � 1 (12.7)

It was reported that the quasi-static sensor response can in some cases beestimated by a power law similar to the equilibrium condition,42,43 in somecases showing a large improvement of the sensor response compared withoperation at constant temperature. In other cases, the quasi-static sensorresponse has a completely different characteristic with the gas concentra-tion44,45 due to the fact that the surface coverage in the TCO can be far


from equilibrium. Looking at the numerous effects of temperature onthe sensor conductance, we start with the fact that the conductance isthermally activated (Eq. 12.3). As this is due to the statistic distribution ofthe conduction band electrons, this thermal activation is following anytemperature variation almost immediately. More complicated are the redoxprocesses on the surface, which need significant time to reach an equilibriumdue to their activation energy. Nakata et al. have proposed that the reactionsshould then be expressed by several temperature-dependent rateconstants26,46,47 which should be considered in the evaluation of gas sensor.To this end, they presented an FFT (fast Fourier transformation)-basedmethod for feature extraction, which should implicitly reflect the differentrate constants.

An explicit determination of the reaction rates was enabled with theintroduction of micromachined membrane sensors, which allow heating

Figure 12.1 LDA projection for the discrimination of hazardous VOCs in air (at varioushumidities and against changing ethanol background up to 2 ppm). (a) data of GGS1330 sensor; (b) data of GGS 2330 sensor; (c) data of GGS 5330 sensor; (d) fusion ofall sensor data (modified after [41]).


and especially cool down of the sensing layer on a very short time scale of afew milliseconds. Using these devices, Ding et al.48 proposed and tested a setof rate equations for the modeling of the surface charge, which in principleallow the modeling of the temperature cycle. Following this approach, Bauret al. developed a method to optimize the sensor response within TCOusing a rate equation model.44,49 In this model, only a single negativelycharged surface state is assumed to simplify the rate equations to one adsorp-tion and one desorption term (Eq. 12.8).

dNs

dt¼ kans exp

�xDH2kbT

�� kdNs (12.8)

The adsorption is determined by the concentration of electrons and theenthalpy of adsorptionDH for the negative charge (ionosorbed oxygenO�Þand by a rate constant ka for adsorption. The desorption is given by thedensity of surface charge Ns and a rate constant kd. The increase of sensorresponse is because the activation energy EbNðT Þ caused by the bandbending (Eq. 12.1) in equilibrium is strongly temperature dependent. Astrong increase of EbNðTÞ and therefore of NsNðT Þ was observed withtemperature (Fig. 12.2).

A high temperature period in the temperature cycle can thereforeprovide a surface with a large excess of surface oxygen, which will cause apredominant reduction of the sensor accompanied by strong sensor responseduring any following low temperature period. The principle of this mode of

1

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2100 150 200 250 300 350 400 450

Temperature (°C)

Act

ivat

ion

ener

gy (e

V)

Figure 12.2 Activation energy in equilibrium over temperature (modified after [49]).


operation is shown in Fig. 12.3. The operating mode is divided into foursections (ieiv):(i) Equilibration during high temperature period to achieve a high

negative surface charge (oxidation). The energy barrier increaseswith the surface coverage according to Eq. (12.1). The equilibriumlevel is then denoted as Eb;Thigh;N.

(ii) Rapid decrease of temperature to preserve excess surface charge. Thelogarithmic conductance decreases with a constant slope due to itsthermal activation (Arrhenius’ law).

(iii) Detection period at low temperature with predominant surfacereduction. The change of the energy barrier is detected with respectto the gas concentration (the new equilibrium is denoted as Eb;Tlow;N,but this equilibrium does not necessarily have to be reached within thetemperature cycle).

(iv) (Rapid) increase of temperature to restart oxidation period.Because of the high excess of ionosorbed oxygen, the rate equation at the

beginning of the low temperature period (iii) can be simplified to Eq. 12.9.

dNs

dt¼ � kdNs with kd ¼ kdes þ

Xj

k jred (12.9)

dEBdt

0x–

dlnGd(1 T) = –

Eb,Tlow, ∞

∞

∞

kB

Δt

Δt Δt

0

0

dlnGd(1 T) = –

Eb,Thigh,∞

kB

0x–

dEBdt

0x–Δt

0x–

Loga

rithm

of c

ondu

ctan

ce

Inverse temperature

Figure 12.3 Principle of a TCO for the optimization of the sensor response (modifiedafter [44]). The light grey area represents the electron-depleted region at the grainsurface. The chemisorbed oxygen is represented by dashes.


The total rate constant kd contains all mechanisms for surface reduction,i.e., the desorption of ionosorbed oxygen kdes and the redox reactions of theionosorbed oxygen k jred with different reducing gases j. In this mode,the sensor response can be very high. In fact, for 1 ppm ethanol in air, weobserved an increase of the response compared with isothermal operationby a factor of 1000.49 Note that this boost in sensitivity is strongly dependingon the gas concentration as can be observed by comparing the pulse peakheight and the near steady-state response at the end of the plateaus inFig. 12.4. For selective detection of gases, i.e., the original use of TCO,the temperature cycle has to combine various “low” temperature periodsat different temperatures to make use of the change of k jred with temperaturefor gas identification (Fig. 12.4).

The sensor response shows a distinct maximum on each plateau.For ethanol, which is easily oxidized, the response is highest at the lowestoperating temperature of 130 �C, whereas for benzene, a relatively stablemolecule, the highest response is observed between 220 and 270 �C.

At the beginning of the low temperature period, the sensor response isobviously no longer following a power law, as the surface coverage decreaseslinearly with the applied gas concentration (Eq. 12.9).50 Together withEqs. (12.1) and (12.3), this leads to Eq. (12.10) showing the exponentialdependency of the sensor response with respect to k jred:

SqsðtiÞ¼ expð2 � kd � tiÞexpð2 � kdes � tiÞ ¼ exp

0@X

j

k jred

1A � ti (12.10)

Figure 12.4 Sensor response to ethanol (solid lines) and benzene (dashed) at variousconcentrations during the temperature cycle shown at the bottom (modified after [49]).


Please note that Eq. (12.10) uses the approximation of negligible changesof Ns (cf. Eq. 12.9) and is therefore a valid approximation only for a shortperiod after the temperature change. A more generalized method for thedetermination of the rate constants and their use as sensor signal can be foundin Ref. 50. For a direct reaction of the reducing gas (EleyeRidealmechanism), k jred is proportional to the gas concentration c jgas (a

j being aproportionality constant) and therefore the sensor response is simply given by

SqsðtiÞ¼ exp

0@X

j

a jc jgas

1A � ti (12.11)

For a LangmuireHinshelwood reaction mechanism, this dependencymight be more complex depending on the adsorption isotherm of thereducing compound. For low concentrations, however, a linear isothermis often appropriate (Henry isotherm). In this case, Eq. (12.11) is alsoapplicable for this type of reaction. Obviously, this approach to TCOoptimization is especially suitable for selectively measuring small concentra-tions of reducing gases, which is, e.g., required for determination of airquality. For the detection of benzene in pure air, this method has beensuccessfully tested yielding very good results. In this experiment, threelow temperature plateaus were used and on each the rate constant kdðT Þwas calculated as model-based feature, which were then used as (nearlylinear) virtual sensors in a multilinear quantification algorithm (partial leastsquare regression) as shown in Fig. 12.5.

With small variations, Eq. (12.11) can also be used for the description ofchanging gas concentrations, e.g., for gas pulses.52 (Please note that forclarity only a single gas component is represented in Eq. 12.12.)

SqsðtÞ ¼ exp

0@Z t

0

a � cgas dt01A (12.12)

In this sense, the principle of temperature modulation can also be usedfor systems with discontinuous gas application like sensor preconcentratoror gas chromatographic systems.52,53

12.2.2 Field effectA dynamic change of surface charge and surface reactions can also beobtained through the variation of electrical fields. An external electrical


field, e.g., caused by a suspended electrode, will cause compensating chargeson the sensor54 which can favor the adsorption or desorption of ionosorbedspecies such as NO2 depending on the polarity of the electrode. Dynamicvariation of perpendicular electrical fields is mostly used in GasFET setups(cf. Section 12.3.2) as the gate electrode provides an easy access to thisparameter. However, in some types of MOS sensors, a simple modulationof the electrical field for the sensor readout can be used to obtain severalvirtual sensors.55e57 Tungsten oxide can often contain mobile donors,which can be polarized even with small electrical fields. The concentrationof the donors is directly linked to the band bending (at constant surfacecharge) according to Eq. (12.1). Donor-accumulated regions thereforeinitially show a smaller band bending and a lower resistance. After the donoraccumulation and the change in band bending, a change in oxygen surfacecoverage according to Eq. (12.2) would be expected with a certain increaseof the band bending. Using a four electrode tungsten oxide sensor, thisaccumulation and surface relaxation has been observed by a decrease andsubsequent increase of the resistance (as well as the expected inversebehavior at the donor depletion region).56 It was demonstrated that a cyclicvariation of readout voltages can even be used to generate gas-specificsignals, i.e., to boost selectivity, as shown in Fig. 12.6.57

10 10

8 8

6 6

4 4

2 2

0 0

0 02 24 46 8 8610 10Concentration set point /ppb Concentration set point /ppb

Sen

sor s

yste

m re

adou

t / p

pb

Sen

sor s

yste

m re

adou

t / p

pb

Training date Training date

RegressionRegressionTest date

Figure 12.5 Quantification of benzene using a three-sensor array with responseoptimized temperature cycles for the detection of benzene at (sub-)ppb level. Thequantification was performed using the pre-processed sensor data (rate constants)and PLSR. On the left: training data with 10% and 40% RH; on the right: training andtest data including various untrained benzene concentrations and all benzeneconcentrations at 25 % RH.51


12.2.3 Optical excitationOptical excitation induces a broad variety of reactions that can be used fordynamic operation. Irradiation with UV light typically causes the generationof electronehole pairs that can act as counterparts for a redox reaction at thesurface. Especially holes, which are typically not abundant in n-type semi-conductors, play an important role in the recombination with ionosorbed,oxidized species such as NO2 or O3. Thus, optical excitation is often usedto activate the desorption process for improving the sensor kinetics at lowtemperatures and even at room temperature for various sensor materialsranging, e.g., from tin oxide,58 titanium oxide,59 and indium oxide60e63

to organic semiconductors.64 However, only very few authors actually useoptical excitation to create dynamic signals.61,62,64 When used in dynamicoperation, the optical excitation was shown to have a significant potentialfor selective detection of gases.64

12.3 Dynamic operation of gas-sensitive field-effecttransistors

Gas-sensitive field-effect devices, especially transistors (GasFETs),offer additional modes of interaction between the semiconductor sensor

Figure 12.6 Two virtual sensors obtained by feature extraction of resistance changeduring a voltage pulse. The total duration of the bias cycle is 300 s with two oppositevoltage pulses of �600 mV for 30 s each followed by 120 s relaxation periods [58].


and the atmosphere. While the signal in MOS sensors is only resulting fromcharge transfer between gas and surface (chemisorption) or redox reactionson the surface, i.e., chemical processes, GasFETs can also detect gas due tophysisorption due to the field effect of polar molecules. The first GasFETsdemonstrated by Lundstr€om65 with solid palladium gates were only ableto detect hydrogen and hydrogen-containing gases due to ionizedhydrogen, i.e., protons, diffusing through the gate. The response and appli-cation spectrum were greatly expanded with novel technological approachessuch as suspended or perforated gates and by making use of silicon carbide(SiC) as semiconductor instead of Si. SiC allows operation at elevatedtemperatures due to its much larger bandgap as well as in harsh environ-ments due to its chemical inertness (cf. Chapter 8). However, with theseapproaches, GasFETs face similar challenges as MOS sensors, i.e., limitedselectivity and stability due to the broad response spectrum and drift orpoisoning effects caused by slow or irreversible chemical processes on thesurface. On the other hand, GasFETs offer additional potential for dynamicoperation: not only temperature modulation for SiC-FET sensors similarto MOX sensors but also variation of the gate bias VGS as a direct way toinfluence the response spectrum and sensitivity of GasFET sensors as shownin Fig. 12.7.16

While dynamic modes of operation have been widely utilized for MOXsensors already for over 40 years,11 first results utilizing temperature and gatebias modulation for GasFETs were published around 200066,67 and2010,16,20 respectively. Both methods, individually and in combination,were systematically studied by C. Bur.17

12.3.1 Temperature-cycled operation for SiC-FET sensorsSimilar to MOS sensors, TCO has proven to be a powerful tool forimproving the selectivity of SiC-FET sensors.68,69 Fig. 12.8 showsone example for the discrimination and quantification of hazardous VOC(benzene, formaldehyde, and naphthalene), e.g., for indoor air qualitycontrol and demand-controlled ventilation.70 Other potential applicationsinclude exhaust gas monitoring, i.e., selective quantification of NO and/or NH3 for SCR systems in diesel engines68,69 or for SO2 in power plantmonitoring and control.71,72

The principal processes on the sensor surface are similar to those occur-ring on MOS sensors, i.e., temperature-induced changes in ionosorption ofoxygen and other redox species as well as adsorption of polar molecules.However, the mechanism cannot be modeled today with a simple approach


as for the MOS sensors, cf. Section 12.2.1. This is due, on the one hand, tothe more complex structure of the sensor surface with catalyst clusters, openinsulator surface, and the highly relevant three-phase boundaries betweencatalyst, surface, and atmosphere.73 On the other hand, the much largerthermal time constants of SiC-FET sensors (several seconds comparedwith approximately 10 ms for microstructured MOX sensors) do not allowexperimental differentiation between temperature changes and the inducedgas adsorption and redox reactions.

VDS VGS

n-type active layerp-type buffer layer

n-type 4H-SiC substrate

D S

0 1 2 3 4 5

0

100

200

300

400

500

20 % O2, 0 % r.h.

VGS= 2 V

VGS= 0 VI DS

(μA

)

VDS (V)

Background0

NH3 (500 ppm)CO (500 ppm)

VGS= –2V

200 °C

(a)

(b)

Figure 12.7 (a) Schematic cross-section of a SiC-FET; (b): typical IV-curves of the SiC-FETin pure dry air (black) and with 500 ppm ammonia (NH3, orange, dashed) and carbonmonoxide (CO, green, dashed) for different values of the applied gate potential VGS. Instatic mode, this can be used to maximize the sensor response (VGS¼ 2 V) or to improveselectivity (here: NH3 vs. CO at VGS ¼ 0 V) (modified after [16]).


12.3.2 Gate biasecycled operationTypical measurements of GasFETs are performed by setting two electricalparameters of the transistor, e.g., the gate bias VGS and the drain-sourcevoltage VDS (cf. Fig. 12.7), at constant values and measuring the resulting

TD

T

SAS

D

FNB

BenzeneNaphthalene

CB A

T

T

T

Below threshod

FNB

Formaldehyde

(a) (c)

(b) (d)

Figure 12.8 (a) temperature cycle, actual temperature and resulting sensor signalVDS (constant current mode: IDS ¼ 45 mA) in air for a SiC-FET (cf. Fig. 12.7);(b) normalized difference signal of the sensor response for 100 ppb formaldehyde, 20ppb naphthalene, and 4.5 ppb benzene at 40 % RH in air; (c) discrimination of benzene,naphthalene, and formaldehyde based on LDA: each group contains three differentconcentrations above the ventilation threshold of the respective gas; the fourth groupcontains pure air as well as one concentration of each gas below the threshold. All groupscontain data at 20 % and 40 % RH. Using a Mahalanobis distance classifier the achievedleave-one-out cross-validation rate of nearly 90 % indicates the almost perfect discrimi-nation of all gases and background. (d) Quantification of naphthalene based on LDA: theplot shows gas concentration vs. value of first discriminant function with a second orderfit for the training data (0, 5 and 40 ppb, solid symbols); evaluation data (2.5, 10 and 20ppb) are marked by open symbols. The indicated line could be used as a threshold limitvalue, e.g. for ventilation control (modified after [71]).


third parameter, in this case the drain-source current IDS. The controlparameters are selected to optimize the sensitivity or the selectivity versusrelevant interfering gases. Additional information is obtained by dynamicallychanging one control parameter, e.g., the gate bias, during the measurementto probe different operating points of the transistor, resulting in a measure-ment signal over the control parameter cycle, i.e., gate biasecycledoperation, Fig. 12.9. One seemingly obvious advantage would be thepossibility to use faster cycles by employing fast changes of the electricalcontrol parameter. We found, however, that the changes induced on the

(a) (b)

(c)

Figure 12.9 Hysteresis of DIDS¼ IDS(gas) - IDS(background) for CO (green, triangles) andNH3 (orange, circles) in pure nitrogen at 50 % RH at 187 �C (a) and 265 �C (b) and in airat 50 % RH at 265 �C (c). The shape of the hysteresis is greatly influenced by the back-ground gas, the humidity and the sensor temperature; the observed cross-over pointsindicate multiple processes taking place on and in the sensor (modified after [16]).


sensor surface by variation of the gate bias lead to very slow equilibration withtime constants in the order of several 10 minutes16 The resulting complexhysteresis curves, Fig. 12.9, indicate multiple processes competing on andin the sensor, i.e., ionosorption on catalyst and insulator, spill-over from cata-lyst to insulator andddue to the extremely long time constantsdprobablydiffusion of ions into the sensor layers. This latter assumption is alsocorroborated by the influence of the insulator material on the sensor perfor-mance.74 Gate bias cycling therefore does not only allow a significant increasein the selectivity16 but can also be applied for experimental studies to achievea better understanding of the relevant processes determining the sensorbehavior and, thus, for improving the sensor performance further.

12.3.3 Current compensation modeAnother approach for dynamic operation is achieved by keeping both VDS

and IDS constant by a closed-loop control of the gate bias VGS. In effect, thismeans that gas-induced changes on the sensor surface are compensated by achange of the gate bias. This method was recently tested in a preliminarystudy which found that the signal to noise ratio is only slightly decreasedin this operating mode if suitable electronics with sufficiently high resolutionare used for measuring the current and setting the gate voltage.75,76 Again,this method can be used to improve our understanding of the processes onthe sensor surface as, ideally, the resulting voltage difference DVGS directlyreflects the change of the control voltage of the transistor, i.e., the effectivecharge caused by the additional gas adsorption on the surface allowing directcomparison of different gases, gas mixtures, and concentrations. In addition,the current compensation mode proved to result in a more linear sensorresponse allowing quantification of ammonia in the range from 0 to30 ppm with constant uncertainty as shown in Fig. 12.10.

12.3.4 Combined methodsDynamic operating modes can be combined to achieve even better perfor-mance for the overall system. The results shown in Fig. 12.10 are actuallyobtained by combining a simple temperature cycle (linear increase from180 to 270 �C in 25 s followed by a linear temperature decrease back to180 �C in 35 s), thus combining TCO and CCM. For this example,TCO primarily achieves the desired selectivity for discriminating differentgases, while CCM improves the quantification due to the improved linearityof the response.


Similarly, TCO and GBCO can be combined resulting in improvedperformance as demonstrated for the discrimination of carbon monoxide(CO), nitrogen dioxide (NO2), and ammonia (NH3) independent ofconcentration and also for quantification of the individual gases.77

Fig. 12.11 demonstrates that a combined cycle achieves better results thanTCO or GBCO separately for CO quantification. In the same study, wecould also show that, while there is some drift of the sensor, suitable featurescan be identified allowing stable gas discrimination and quantification bycombining calibration data from different states of aging using multivariatestatistics.

12.4 Conclusion and outlook

Dynamic operation is a powerful and versatile tool for improving theperformance of semiconductor gas sensor systems with respect to the three

0 10 20 30Concentration / ppm

0

5

10

15

Sen

sor r

espo

nse

/ µA

0

2

4

6

rel.

sens

or re

spon

se /

%

Standard operation

R2lin=0,976

Rlog2 =0,995

0 10 20 30Concentration / ppm

0

0.05

0.1

0.15

0.2

Sen

sor r

espo

nse

/ V

0

2

4

6

8

rel.

sens

or re

spon

se /

%

Current compensation mode

R2lin=0.994

(a)

(b)

Figure 12.10 Quantification of NH3 with a porous Pt-gate SiC-FET sensor comparingstandard operation (a) to current compensation mode (CCM, b); while standard opera-tion results in a logarithmic calibration curve, CCM achieves a linear response andhigher relative signals for high concentrations (modified after [77]).


S, sensitivity, selectivity, and stability. Combined with novel materials basedon nanotechnologies and novel substrates/transducers realized with micro-technologies, this approach is one key for addressing new applications withlow cost gas sensors, e.g., in environmental monitoring, health, safety, andsecurity. Compared with conventional multisensor arrays, which require

1086420

-2-4-6-8

86420

-2-4-6-8

-10

1086420

-2-4-6-8

-10-15 -10 -5 0 5 10 15 20 25 30 35 40

1st discriminant function

FeaturesT-cycleparts A and C

FeaturesT-cycle & GB-cyclesparts A to D

Validation: 99.07 %Wilks' Lambda:0.0157

FeaturesGB-cycle

part D

Validation: 98.14 %Wilks' Lambda: 0.0023

Validation: 98.14 %Wilks' Lambda: 0.0082

@200ºC

0 ppm

ppm

200 400

600 800

Figure 12.11 LDA showing CO quantification using features from temperature cyclingin the temperature range 200e 260�C only at VGS¼ 0 (top), from gate bias variation (-1to þ2 V) only at 200�C (centre) and from the combined cycle (bottom). While theGB-cycle data yield better class separation, T-cycling achieves a higher leave-one-outcross-validation rate. The best results are achieved by combining the features [78].


frequent recalibration due to their limited stability, the dynamic operationapproach provides much better stability due to the inherently differentialnature of the measurement comparing different sensor states instead ofmeasuring one specific state only. On the other hand, many similaritiesare obvious, especially in signal processing: both methods provide multipleraw measurement values which are interpreted with suitable patternrecognition methods. Due to this similarity, the term virtual multisensorwas coined to emphasize the fact that a single physical sensor provides theinformation. Note, however, that this term does not only apply to dynamicsensor operation but also to EMR methods like electrical impedancespectroscopy (EIS).

All virtual multisensor methods require advanced electronics to make fulluse of their potential, both to control the dynamic excitation and to read-out the sensor response with the required electrical and temporal resolution.Note that this means that virtual multisensor systems are not necessarilycheaper than multisensor arrays, because the increased cost of the electronicscan greatly outweigh the reduced cost of a single versus multiple sensorelements. On the other hand, dynamic operation is more versatile as theoperating mode (in addition to the data analysis) can be adapted to differentapplication requirements, in effect shifting from hardware to softwaresolutions. This makes dynamic operation very interesting for novel“generic” application scenarios like gas sensors integrated into smartphones:the target application can be set via software (“there’s an app for that”)allowing different use cases to be addressed with the same hardware.

The primary benefit of dynamic operation generally lies in the acquisi-tion of characteristic time constants in addition to the obtained electricalvalues, which provide further information about the ambient atmosphere.This would seem to suggest that dynamic operation would result in slowerresponse and recovery times for the gas sensor system. However, theopposite is actually the case as could be demonstrated for an applicationdiscriminating fuel vapors to prevent false fueling of cars. While the sensorsignal of an isothermally operated MOS sensor requires several 10 s to reachsteady state in gasoline or diesel vapor atmosphere, the shape of the TCOsensor response could already be evaluated after one temperature cyclewith a duration of only 2 s.78 Note that this also allows changing the tem-perature cycle during the sensor operation, i.e., to improve the classificationperformance after a first classification step.37 Of course, the achievablemeasurement rate depends on the time constants to be measured, in thiscase the interaction between gas molecules and gas-sensitive layer.


This would suggest to use increased temperatures to speed up the chemicalprocesses to achieve a higher measurement rate. However, the temperaturealso changes the dominating processes on the sensor surface, which limits thepotential for tuning the response times with the operating temperature tomatch the acquisition electronics. Note that EMR methods allow muchfaster operation (e.g., Fourier-based impedance spectroscopy achieving acomplete spectrum over a wide frequency range in only 16 ms79), but obtainonly steady-state information and therefore provide less information aboutthe ambient. For instance, interaction of methane (CH4) with a semicon-ductor surface will be minimal at low temperature due to the high reactionenthalpy and thus identification and quantification will be quite difficult,while for CO the opposite is true.

This suggests that combinations of different methods are very attractiveto operate the sensor under the best conditions for the required informationand/or to extract as much information as possible. Combined operation hasbeen demonstrated for EIS/TCO for MOX sensors79 and for GBCO/TCO77 as well as CCM/TCO76 for GasFETs. Note that a combinationof complementary methods can also allow sensor self-monitoring: byevaluating the sensor response separately with two complementary methods,a sensor that is no longer performing as calibrated can be identified bydifferent predictions resulting from the two methods.6 This can provide asimple and cost-efficient alternative to regular field tests of sensor systemsand might be a key for acceptance of semiconductor sensor systems in safetyrelevant applications like fire or gas leak detection.

Finally, novel research results propose to manipulate the gas supply to thesensor with micro preconcentrators (mPC) integrated with gas sensorelements in a microcontainment.53 While gas adsorption/desorption iscontrolled by the mPC temperature similar to standard sampling and thermaldesorption techniques, the gas transport is based on diffusion only thusavoiding mechanical pumps and valves. Because of the compatibility ofthe mPC and sensor technologies, both being based on microhotplates,this greatly expands the performance spectrum of the integrated gas sensorsystem. Benefits are the increase of the sensitivity by the increased targetgas concentration in the release peak and the improvement of selectivityby suppressing permanent gases, especially H2 and CO, which are notadsorbing on the mPC. Moreover, these systems are also offering improvedstability by providing an internal zero air reference: after the release peak, thetarget gas concentration drops to practically zero in the microsystem asnearly all gas molecules adsorb on the mPC.80


References1. Schiffman SS, Pearce TC. Introduction to olfaction: perception, anatomy, physiology,

and molecular biology. In: Pearce TC, Schiffman SS, Nagle HT, Gardner JW, editors.Handb. Mach. Olfaction e Electron. Nose Technol. Wiley-VCH; 2002.

2. Persaud K, Dodd G. Analysis of discrimination mechanisms in the mammalian olfactorysystem using a model nose. Nature 1982;299:352e5. https://doi.org/10.1038/299352a0.

3. Persaud KC. Electronic gas and odour detectors that mimic chemoreception in animals.Trends Anal Chem 1992;11:61e7. https://doi.org/10.1016/0165-9936(92)80079-L.

4. Gutierrez-Osuna R. Pattern analysis for machine olfaction: a review. IEEE Sens J 2002;2:189e202. https://doi.org/10.1109/JSEN.2002.800688.

5. Marco S, Gutierrez-Galvez A. Signal and data processing for machine olfaction andchemical sensing: a review. IEEE Sens J 2012;12:3189e214. https://doi.org/10.1109/JSEN.2012.2192920.

6. Reimann P, Sch€utze A. Sensor arrays, virtual multisensors, data fusion, and gas sensordata evaluation. In: Wagner T, Kohl CD, editors. Gas Sens. Fundam. Springer Ser.Chem. Sensors biosensors, vol. 15; 2014.

7. Matsunaga N, Sakai G, Shimanoe K, Yamazoe N. Diffusion equation-based study ofthin film semiconductor gas sensor-response transient. Sensor Actuator B Chem 2002;83:216e21. https://doi.org/10.1016/S0925-4005(01)01043-7.

8. Sch€utze A, Pieper N, Zacheja J. Quantitative ozone measurement using a phthalocya-nine thin-film sensor and dynamic signal evaluation. Sensor Actuator B Chem 1995;23:215e7. https://doi.org/10.1016/0925-4005(94)01281-L.

9. Helwig A, Beer S, M€uller G. Breathing mode gas detection. Sensor Actuator B Chem2013;179:131e9. https://doi.org/10.1016/j.snb.2012.07.088.

10. Helwig A, M€uller G, Sberveglieri G, Faglia G. Gas response times of nano-scale SnO2gas sensors as determined by the moving gas outlet technique. Sensor Actuator B Chem2007;126:174e80. https://doi.org/10.1016/j.snb.2006.11.032.

11. Eicker H.Method and apparatus for determining the concentration of one gaseous component in amixture of gases. 1977. US patent US4012692A, http://www.google.tl/patents/US4012692.

12. Lee AP, Reedy BJ. Temperature modulation in semiconductor gas sensing. Sensor Actu-ator B Chem 1999;60:35e42. https://doi.org/10.1016/S0925-4005(99)00241-5.

13. Llobet E. Temperature-modulated semiconductor gas sensors. In: Grimes CA,Dickery EC, Pishko MV, editors. Encycl. Sensors. Los Angeles: American Scientific Pub-lishers; 2006. p. 131e52.

14. Engel M, Baumbach M, Kammerer T, Sch€utze A. Preparation of microstructured pel-listors and their application for fast fuel vapor discrimination. In: 17th IEEE Int. Conf.Micro electro Mech. Syst. Maastricht MEMS 2004 Tech. Dig.; 2004. p. 268e71. https://doi.org/10.1109/MEMS.2004.1290574.

15. Fricke T, Sauerwald T, Sch€utze A. Study of pulsed operating mode of a microstructuredpellistor to optimize sensitivity and poisoning resistance. Proc IEEE Sensors 2014:661e4.https://doi.org/10.1109/ICSENS.2014.6985085.

16. Bastuck M, Bur C, Lloyd Spetz A, Andersson M, Sch€utze A. Gas identification based onbias induced hysteresis of a gas-sensitive SiC field effect transistor. J Sensor Sens Syst2014;3:9e19. https://doi.org/10.5194/jsss-3-9-2014.

17. Bur C. Selectivity enhancement of gas sensitive field effect transistors by dynamic operation.Dissertation Link€oping University and Saarland University , Link€oping University Elec-tronic Press/Shaker Verlag; 2015. http://liu.diva-portal.org/smash/record.jsf?pid¼diva2:791934&dswid¼-3121.


https://doi.org/10.1038/299352a0

https://doi.org/10.1038/299352a0

https://doi.org/10.1016/0165-9936(92)80079-L




https://doi.org/10.1016/S0925-4005(01)01043-7

https://doi.org/10.1016/0925-4005(94)01281-L



http://www.google.tl/patents/US4012692

http://www.google.tl/patents/US4012692

https://doi.org/10.1016/S0925-4005(99)00241-5

https://doi.org/10.1109/MEMS.2004.1290574

https://doi.org/10.1109/MEMS.2004.1290574


https://doi.org/10.5194/jsss-3-9-2014

http://liu.diva-portal.org/smash/record.jsf?pid=diva2:791934&dswid=-3121





18. Fischer S, Pohle R, Fleischer M, Moos R. Method for reliable detection of differentexhaust gas components by pulsed discharge measurements using standard zirconiabased sensors. Procedia Chem 2009;1:585e8. https://doi.org/10.1016/j.proche.2009.07.146.

19. Sch€utze A. Pr€aparation und Charakterisierung von Phthalocyanin-Schichten zum Nachweis oxi-dierender und reduzierender Gase. Aachen: Dissertation, Justus-Liebig-Universit€at Gießen,Shaker Verlag; 1995.

20. Pohle R, von Sicard O, Fleischer M, Frerichs HP, Wilbertz C, Freund I. Gate pulsedreadout of floating gate FET gas sensors. Procedia Eng 2010;5:13e6. https://doi.org/10.1016/j.proeng.2010.09.036.

21. Carotta MC, Cervi A, Fioravanti A, Gherardi S, Giberti A, Vendemiati B, Vincenzi D,Sacerdoti M. A novel ozone detection at room temperature through UV-LED-assistedZnO thick film sensors. Thin Solid Films 2011;520:939e46. https://doi.org/10.1016/j.tsf.2011.04.173.

22. Herr�an J, Fern�andez-Gonz�alez O, Castro-Hurtado I, Romero T, Mandayo GG,Casta~no E. Photoactivated solid-state gas sensor for carbon dioxide detection at roomtemperature. Sensor Actuator B Chem 2010;149:368e72. https://doi.org/10.1016/j.snb.2010.06.050.

23. Vergara A, Benkstein KD, Semancik S. Thermally-assisted transient analysis forreducing the response time of microhotplate gas sensors. In: 2013 IEEE sensors; 2013.p. 1e4. https://doi.org/10.1109/ICSENS.2013.6688156.

24. Gutierrez-Osuna R, Gutierrez-Galvez A, Powar N. Transient response analysis fortemperature-modulated chemoresistors. Sensor Actuator B Chem 2003;93:57e66.https://doi.org/10.1016/S0925-4005(03)00248-X.

25. Kato Y, Yoshikawa K, Kitora M. Temperature-dependent dynamic response enablesthe qualification and quantification of gases by a single sensor. Sensor Actuator B Chem1997;40:33e7. https://doi.org/10.1016/S0925-4005(97)80196-7.

26. Nakata S, Ozaki E, Ojima N. Gas sensing based on the dynamic nonlinear responses of asemiconductor gas sensor: dependence on the range and frequency of a cyclictemperature change. Anal Chim Acta 1998;361:93e100. https://doi.org/10.1016/S0003-2670(98)00013-0.

27. Kunt TA, McAvoy TJ, Cavicchi RE, Semancik S. Optimization of temperatureprogrammed sensing for gas identification using micro-hotplate sensors. Sensor ActuatorB Chem 1998;53:24e43. https://doi.org/10.1016/S0925-4005(98)00244-5.

28. Schweizer-Berberich M, Zdralek M, Weimar U, G€opel W, Viard T, Martinez D,Seube A, Peyre-Lavigne A. Pulsed mode of operation and artificial neural networkevaluation for improving the CO selectivity of SnO2 gas sensors. Sensor Actuator BChem 2000;65:91e3. https://doi.org/10.1016/S0925-4005(99)00333-0.

29. Fischer S, Pohle R, Magori E, Sch€onauer-Kamin D, Fleischer M, Moos R. Pulsedpolarization of platinum electrodes on YSZ. Solid State Ionics 2012;225:371e5.https://doi.org/10.1016/j.ssi.2012.03.020.

30. Morrison SR. Semiconductor gas sensors. Sensor Actuator 1981;2:329e41. https://doi.org/10.1016/0250-6874(81)80054-6.

31. Heiland G, Kohl D. Problems and possibilities of oxidic and organic semiconductorgas sensors. Sensor Actuator 1985;8:227e33. https://doi.org/10.1016/0250-6874(85)85005-8.

32. Madou MJ, Morrison SR. Chemical sensing with solid state devices. San Diego: AcademicPress; 1989.

33. Barsan N,Weimar U. Conduction model of metal oxide gas sensors. J Electroceram 2001;7:143e67. https://doi.org/10.1023/A:1014405811371.

34. Yamazoe N, Shimanoe K. Theory of power laws for semiconductor gas sensors. SensorActuator B Chem 2008;128:566e73. https://doi.org/10.1016/j.snb.2007.07.036.


https://doi.org/10.1016/j.proche.2009.07.146

https://doi.org/10.1016/j.proche.2009.07.146








https://doi.org/10.1016/S0925-4005(03)00248-X

https://doi.org/10.1016/S0925-4005(97)80196-7

https://doi.org/10.1016/S0003-2670(98)00013-0

https://doi.org/10.1016/S0003-2670(98)00013-0

https://doi.org/10.1016/S0925-4005(98)00244-5

https://doi.org/10.1016/S0925-4005(99)00333-0

https://doi.org/10.1016/j.ssi.2012.03.020

https://doi.org/10.1016/0250-6874(81)80054-6

https://doi.org/10.1016/0250-6874(81)80054-6

https://doi.org/10.1016/0250-6874(85)85005-8

https://doi.org/10.1016/0250-6874(85)85005-8

https://doi.org/10.1023/A:1014405811371


35. Heiland G. Homogeneous semiconducting gas sensors. Sensor Actuator 1981;2:343e61.https://doi.org/10.1016/0250-6874(81)80055-8.

36. Barsan N, H€ubner M, Weimar U. Conduction mechanisms in SnO2 based polycrystal-line thick film gas sensors exposed to CO and H2 in different oxygen backgrounds.Sensor Actuator B Chem 2011;157:510e7. https://doi.org/10.1016/j.snb.2011.05.011.

37. Gramm A, Sch€utze A. High performance solvent vapor identification with a two sensorarray using temperature cycling and pattern classification. Sensor Actuator B Chem 2003;95:58e65. https://doi.org/10.1016/S0925-4005(03)00404-0.

38. Polese D, Martinelli E, Catini A, D’Amico A, Di Natale C. Self-adaptive thermal mod-ulation of gas sensors. Procedia Eng 2010;5:156e9. https://doi.org/10.1016/j.proeng.2010.09.071.

39. Paczkowski S, Paczkowska M, Dippel S, Schulze N, Sch€utz S, Sauerwald T, Weiß A,Bauer M, Gottschald J, Kohl C-D. The olfaction of a fire beetle leads to new conceptsfor early fire warning systems. Sensor Actuator B Chem 2013;183:273e82. https://doi.org/10.1016/j.snb.2013.03.123.

40. Jaegle M,W€ollenstein J, Meisinger T, B€ottner H, M€uller G, Becker T, Braunm€uhl CB-V. Micromachined thin film SnO2 gas sensors in temperature-pulsed operation mode.Sensor Actuator B Chem 1999;57:130e4. https://doi.org/10.1016/S0925-4005(99)00074-X.

41. Leidinger M, Sauerwald T, Reimringer W, Ventura G, Sch€utze A. Selective detectionof hazardous VOCs for indoor air quality applications using a virtual gas sensor array.J Sensor Sens Syst 2014;3:253e63. https://doi.org/10.5194/jsss-3-253-2014.

42. Bastuck M, Leidinger M, Sauerwald T, Sch€utze A. Improved quantification ofnaphthalene using non-linear partial least squares regression. In: 16th Int. Symp. OlfactionElectron. Nose, Dijon, Fr.; 2015. http://arxiv.org/abs/1507.05834.

43. Sauerwald T, Baur T, Sch€utze A. Strategien zur Optimierung des temperaturzyklischenBetriebs von Halbleitergassensoren. In: XXVIII. Messtechnisches Symp. des Arbeitskreisesder Hochschullehrer f€ur Messtechnik. Aachen: Shaker Verlag; 2014. p. 65e74. https://doi.org/10.5162/AHMT2014/3.1.

44. Schultealbert C, Baur T, Sch€utze A, B€ottcher S, Sauerwald T. A novel approachtowards calibrated measurement of trace gases using metal oxide semiconductorsensors. Sensor Actuator B Chem 2017;239:390e6. https://doi.org/10.1016/j.snb.2016.08.002.

45. Baur T, Sch€utze A, Sauerwald T. Detektion von kurzen Gaspulsen f€ur die Spurenga-sanalytik (detection of short gas pulses for trace gas analysis). Tm e Tech Mess 2017;84(S1):88e92. https://doi.org/10.1515/teme-2017-0035.

46. Nakata S, Akakabe S, Nakasuji M, Yoshikawa K. Gas sensing based on a nonlinearresponse: discrimination between hydrocarbons and quantification of individualcomponents in a gas mixture. Anal Chem 1996;68:2067e72. https://doi.org/10.1021/ac9510954.

47. Nakata S, Takemura K, Neya K. Non-linear dynamic responses of a semiconductor gassensor: evaluation of kinetic parameters and competition effect on the sensor response.Sensor Actuator B Chem 2001;76:436e41. https://doi.org/10.1016/S0925-4005(01)00652-9.

48. Ding J, McAvoy TJ, Cavicchi RE, Semancik S. Surface state trapping models for SnO2-based microhotplate sensors. Sensor Actuator B Chem 2001;77:597e613. https://doi.org/10.1016/S0925-4005(01)00765-1.

49. Baur T, Sch€utze A, Sauerwald T. Optimierung des temperaturzyklischen Betriebs vonHalbleitergassensoren (optimization of temperature cycled operation of semiconductorgas sensors). Tm e Tech Mess 2015;82:187e95. https://doi.org/10.1515/teme-2014-0007.


https://doi.org/10.1016/0250-6874(81)80055-8


https://doi.org/10.1016/S0925-4005(03)00404-0





https://doi.org/10.1016/S0925-4005(99)00074-X

https://doi.org/10.1016/S0925-4005(99)00074-X


http://arxiv.org/abs/1507.05834

https://doi.org/10.5162/AHMT2014/3.1

https://doi.org/10.5162/AHMT2014/3.1



https://doi.org/10.1515/teme-2017-0035

https://doi.org/10.1021/ac9510954

https://doi.org/10.1021/ac9510954

https://doi.org/10.1016/S0925-4005(01)00652-9

https://doi.org/10.1016/S0925-4005(01)00652-9

https://doi.org/10.1016/S0925-4005(01)00765-1

https://doi.org/10.1016/S0925-4005(01)00765-1



50. Schultealbert C, Baur T, Sch€utze A, Sauerwald T. Facile quantification and identifica-tion techniques for reducing gases over a wide concentration range using a MOS sensorin temperature-cycled operation. Sensors (Switzerland) 2018;18:744. https://doi.org/10.3390/s18030744.

51. Sauerwald T, Baur T, Leidinger M, Reimringer W, Spinelle L, Gerboles M, Kok G,Sch€utze A. Highly sensitive benzene detection with metal oxide semiconductor gas sen-sors e an inter-laboratory comparison. J Sensor Sens Syst 2018;7:235e43. https://doi.org/10.5194/jsss-7-235-2018.

52. Baur T, Schultealbert C, Sch€utze A, Sauerwald T. Novel method for the detection ofshort trace gas pulses with metal oxide semiconductor gas sensors. J Sensor Sens Syst2018;7:411e9. https://doi.org/10.5194/jsss-7-411-2018.

53. Leidinger M, Rieger M, Sauerwald T, Alépée C, Sch€utze A. Integratedpre-concentrator gas sensor microsystem for ppb level benzene detection. SensorActuator B Chem 2016;236:988e96. https://doi.org/10.1016/j.snb.2016.04.064.

54. B€ogner M, Doll T. Advanced gas sensing e introduction to the electroadsorptive effectand its application. In: Doll T, editor. Adv. Gas Sens. e electroadsorptive Eff. Relat. Tech.Springer; 2003. p. 2e37.

55. Liess M. Electric-field-induced migration of chemisorbed gas molecules on a sensitivefilmda new chemical sensor. Thin Solid Films 2002;410:183e7. https://doi.org/10.1016/S0040-6090(02)00209-2.

56. Sauerwald T, Skiera D, Kohl D. Field induced polarisation and relaxation of tungstenoxide thick films. Thin Solid Films 2005;490:86e93. https://doi.org/10.1016/j.tsf.2005.04.009.

57. Sauerwald T, Skiera D, Kohl CD. Selectivity enhancement of gas sensors using non-equilibrium polarisation effects in metal oxide films. Appl Phys A Mater Sci Process2007;87:525e9. https://doi.org/10.1007/s00339-007-3980-2.

58. Fort A, Gregorkiewitz M, Machetti N, Rocchi S, Serrano B, Tondi L, Ulivieri N,Vignoli V, Faglia G, Comini E. Selectivity enhancement of SnO2 sensors by meansof operating temperature modulation. Thin Solid Films 2002;418:2e8. https://doi.org/10.1016/S0040-6090(02)00575-8.

59. Zhang S, Lei T, Li D, Zhang G, Xie C. UV light activation of TiO2 for sensingformaldehyde: how to be sensitive, recovering fast, and humidity less sensitive. SensorActuator B Chem 2014;202:964e70. https://doi.org/10.1016/j.snb.2014.06.063.

60. Klaus D, Klawinski D, Amrehn S, Tiemann M, Wagner T. Light-activated resistiveozone sensing at room temperature utilizing nanoporous In2O3 particles: influence ofparticle size. Sensor Actuator B Chem 2015;217:181e5. https://doi.org/10.1016/j.snb.2014.09.021.

61. Klawinski D, Meixner D, Kohl C, Wagner T. Cyclic optical activation of semiconduct-ing gas sensors: influence of cycling frequency. In: Proc. Sensors 2017, AMA Conf. 2017-05-30 e 2017-06-01 Nuremberg, ger; 2017. p. 731e3. https://doi.org/10.5162/sensor2017/P5.11.

62. Gonzalez O, Roso S, Vilanova X, Llobet E. Enhanced detection of nitrogen dioxide viacombined heating and pulsed UV operation of indium oxide nano-octahedra. Beilstein JNanotechnol 2016;7:1507e18. https://doi.org/10.3762/bjnano.7.144.

63. Chinh ND, Quang ND, Lee H, Thi Hien T, Hieu NM, Kim D, Kim C, Kim D. NOgas sensing kinetics at room temperature under UV light irradiation of In2O3nanostructures. Sci Rep 2016;6:35066. https://doi.org/10.1038/srep35066.

64. Sch€uler M, Sauerwald T, Sch€utze A, Gaudillat P, Suisse J-M, Bouvet M. Selectivequantification of humidity and ammonia by optical excitation of molecularsemiconductor-doped insulator (MSDI) sensors. 2015 IEEE sensors e Proc. 2015:p1e4. https://doi.org/10.1109/ICSENS.2015.7370407.


https://doi.org/10.3390/s18030744

https://doi.org/10.3390/s18030744





https://doi.org/10.1016/S0040-6090(02)00209-2

https://doi.org/10.1016/S0040-6090(02)00209-2



https://doi.org/10.1007/s00339-007-3980-2

https://doi.org/10.1016/S0040-6090(02)00575-8

https://doi.org/10.1016/S0040-6090(02)00575-8




https://doi.org/10.5162/sensor2017/P5.11

https://doi.org/10.5162/sensor2017/P5.11

https://doi.org/10.3762/bjnano.7.144

https://doi.org/10.1038/srep35066


65. Lundstr€om I, Shivaraman S, Svensson C. A hydrogen-sensitive Pd-gate MOS transistor.J Appl Phys 1975;46:3876e81. https://doi.org/10.1063/1.322185.

66. Ekl€ov T, Mårtensson P, Lundstr€om I. Enhanced selectivity of MOSFET gas sensors bysystematical analysis of transient parameters. Anal Chim Acta 1997;353:291e300.https://doi.org/10.1016/S0003-2670(97)87788-4.

67. Briand D,Wingbrant H, Sundgren H, Van der Schoot B, Ekedahl LG, Lundstr€om I, DeRooij NF. Modulated operating temperature for MOSFET gas sensors: hydrogen re-covery time reduction and gas discrimination. Sensor Actuator B Chem 2003;93:276e85. https://doi.org/10.1016/S0925-4005(03)00230-2.

68. Bur C. New method for selectivity enhancement of SiC field effect gas sensors for quantification ofNOx. Diploma thesis. Saarland University; 2012.

69. Bur C, Reimann P, Andersson M, Sch€utze A, Lloyd Spetz A. Increasing the selectivityof Pt-gate SiC field effect gas sensors by dynamic temperature modulation. IEEE Sens J2012:1267e72. https://doi.org/10.1109/JSEN.2011.2179645.

70. Bur C, Bastuck M, Puglisi D, Sch€utze A, Lloyd Spetz A, Andersson M. Discriminationand quantification of volatile organic compounds in the ppb-range with gas sensitiveSiC-FETs using multivariate statistics. Sensor Actuator B Chem 2015;214:225e33.https://doi.org/10.1016/j.snb.2015.03.016.

71. Lloyd Spetz A, Huotari J, Bur C, Bjorklund R, Lappalainen J, Jantunen H, Sch€utze A,Andersson M. Chemical sensor systems for emission control from combustions. SensorActuator B Chem 2013;187:184e90. https://doi.org/10.1016/j.snb.2012.10.078.

72. Darmastuti Z, Bur C, M€oller P, Rahlin R, Lindqvist N, AnderssonM, Sch€utze A, LloydSpetz A. SiC-FET based SO2 sensor for power plant emission applications. Sensor Actu-ator B Chem 2014;194:511e20. https://doi.org/10.1016/j.snb.2013.11.089.

73. Eriksson M, Petersson LG. Spillover of hydrogen, oxygen and carbon monoxide inoxidation reactions on SiO2 supported Pd. Surf Sci 1994;311:139e52. https://doi.org/10.1016/0039-6028(94)90485-5.

74. Eriksson M, Salomonsson A, Lundstr€om I, Briand D, Åbom AE. The influence of theinsulator surface properties on the hydrogen response of field-effect gas sensors. J ApplPhys 2005;98:1e7. https://doi.org/10.1063/1.1994941.

75. Daut C. Signalkompensation mittels Gate-bias bei gassensitiven SiC-Feldeffekttransistoren.Bachelor thesis. Saarland University; 2017.

76. Bastuck M, Daut C, Sch€utze A. Signalkompensation mittels gate-potential bei gas-sensitiven Feldeffekttransistoren. In: 13. Dresdner sensor-symposium; 2017. p. 277e82.https://doi.org/10.5162/13dss2017/P4.03.

77. Bur C, Bastuck M, Lloyd Spetz A, Andersson M, Sch€utze A. Selectivity enhancementof SiC-FET gas sensors by combining temperature and gate bias cycled operation usingmultivariate statistics. Sensor Actuator B Chem 2014;193:931e40. https://doi.org/10.1016/j.snb.2013.12.030.

78. Kammerer T, Engel M, Sch€utze A. An intelligent fuel sensor based on a micro-structured gas sensor. In: Proc. IEEE sensors 2003 (IEEE Cat. No.03CH37498), vol.2; 2003. p. 1064e9. https://doi.org/10.1109/ICSENS.2003.1279106.

79. Sch€uler M, Sauerwald T, Sch€utze A. Metal oxide semiconductor gas sensor self-test us-ing fourier-based impedance spectroscopy. J Sensor Sens Syst 2014;3:213e21. https://doi.org/10.5194/jsss-3-213-2014.

80. Leidinger M, Reimringer W, Alépée C, Rieger M, Sauerwald T, Conrad T, Sch€utze A.Gas measurement system for indoor air quality monitoring using an integrated pre-concentrator gas sensor system. In: Conf. Mikro-Nano-Integration e 6. GMM-workshop10/05/2016 e 10/06/2016 Duisburg, Deutschl; 2016. p1e6. In: https://www.vde-verlag.de/proceedings-en/454278027.html.


https://doi.org/10.1063/1.322185

https://doi.org/10.1016/S0003-2670(97)87788-4

https://doi.org/10.1016/S0925-4005(03)00230-2






https://doi.org/10.1016/0039-6028(94)90485-5

https://doi.org/10.1016/0039-6028(94)90485-5

https://doi.org/10.1063/1.1994941

https://doi.org/10.5162/13dss2017/P4.03






https://www.vde-verlag.de/proceedings-en/454278027.html

https://www.vde-verlag.de/proceedings-en/454278027.html

CHAPTER THIRTEEN

Micromachined semiconductorgas sensorsD. Briand1, J. Courbat21Ecole Polytechnique Fédérale de Lausanne, Neuchatel, Switzerland2Formely Ecole Polytechnique Fédérale de Lausanne, Neuchatel, Switzerland

Contents

13.1 Introduction 41313.2 A brief history of semiconductors as gas-sensitive devices 41413.3 Microhotplate concept and technologies 416

13.3.1 Concept and thermal design 41613.3.2 Microhotplate realization and performance 41813.3.3 Microhotplate reliability 421

13.4 Micromachined metal oxide gas sensors 42513.4.1 Thin gas-sensitive films 42513.4.2 Thick gas-sensitive films 42813.4.3 Temperature modulation 43213.4.4 Packaging 435

13.5 Complementary metal oxide semiconductorecompatible metal oxide gassensors

437

13.6 Micromachined field-effect gas sensors 44213.7 Nanostructured gas sensing layers on microhotplates 44513.8 Semiconductor gas sensors on polymeric foil and their additive manufacturing 450

13.8.1 Semiconductor gas sensors on polymeric foil 45013.8.2 Printing semiconductor gas sensors 452

13.9 Manufacturing, products, and applications 45413.10 Conclusion 458References 459

13.1 Introduction

Metal oxide gas sensors based on screen printing thick layers onalumina substrates to form a platinum heater and electrodes, and to patternthe thick metal oxide gasesensitive film, have been commercialized for afew decades. At the beginning of the 1980s, micromachining of silicon



https://doi.org/10.1016/B978-0-08-102559-8.00013-6

took considerable strides and led to the emergence of new microelectrome-chanical systems (MEMS) devices. The use of microfabrication techniques torealize microsensors and MEMS devices has brought different advantagesthan miniaturization, such as batch processing, formation of arrays, reducedpower consumption, and new modes of operation. Some work has beenundertaken by micromachining anodic alumina1,2 but the extensive devel-opments were carried out based on silicon micromachining.3

This chapter therefore focuses on silicon micromachined semiconductorgas sensors. After a brief history of silicon hotplates and metal oxide gassensors, more information will be provided on the microhotplate concept,realization, and reliability. The core of this chapter comprises a section onmicromachined thin- and thick-film metal oxide gas sensors addressingtemperature modulation. Some highlights are given concerning comple-mentary metal oxide semiconductor (CMOS) and silicon on insulator(SOI) implementation of metal oxide gas sensors and micromachinedfield-effect gas sensors. Finally, trends on the integration of nanostructuredgas sensing materials on micromachined transducers and on semiconductorgas sensors on polymeric foil, and their additive fabrication, are highlighted.

13.2 A brief history of semiconductors asgas-sensitive devices

In 1952, Brattain and Bardeen reported on the change of the semicon-ducting properties of germanium with a variation of the partial pressure ofoxygen in the surrounding atmosphere.4 Seiyama published 10 years laterresults demonstrating the gas sensing effect on metal oxides.5 Taguchibrought metal oxide semiconductor gas sensors to market using an aluminaceramic tube mounted with the metal oxide and electrodes and a heater coilpassing through it. He founded in 1969 the company Figaro EngineeringInc., which is still today the largest manufacturer of semiconductor gassensors worldwide. Nowadays, the commercially available devices aremostly manufactured using screen printing on small and thin ceramicsubstrates exhibiting a power consumption of 0.2e1 W. In 1988, Demarneet al. demonstrated and patented the first thin-film metal oxide gas sensorsbased on a micromachined silicon substrate. The microhotplate was madeof a thermally insulating silicon oxide membrane. It embedded a gold heater.Gold electrodes were patterned on top and covered with a thin tin dioxidefilm. The device operated with a significantly reduced power consumptionof about 100 mW to reach 300 �C, a value still much lower than commer-cially available devices on alumina substrates.

414 D. Briand and J. Courbat

Motorola licensed the technology and put effort into developing massproduced metal oxide gas sensors using silicon micromachining(Fig. 13.1). Polysilicon heaters were introduced in an oxideenitridemembrane, using gold electrodes as before. They ceased work on the chem-ical sensor in 1998, but the technology was taken over by MicroChemicalSystems SA in Switzerland and has evolved to be aligned with the develop-ments reported by other research and industrial groups. Micromachinedthick-film semiconductor gas sensors were introduced by drop-coatingthe metal oxide on a thin dielectric membrane with platinum used bothfor heaters and electrodes, offering improved performances and robustness.6

This technology has been exploited because then by AppliedSensor GmbH(Section 6.4.2). Temperature modulation was introduced as a mode ofoperation due to the low thermal mass of the microhotplates. This modeof operation is now mainly applied to applicative scenarios to minimizepower consumption; to reduce the influence of humidity, for example, toenhance the discrimination capabilities of these sensors; and to improve theirstability over time (Section 6.4.3).

Since 2000, the field has been evolving toward the use of SOI wafers, theimplementation of these sensors in CMOS technology and on polymericsubstrates, and the identification of suitable modes of operation for differentapplications. The field is now strongly focusing on nanomaterials,7 especially

SnO2 layer

Si/SiO2 diaphragmBulk Si/SiO2

1 mm

Pt electrode

Poly-Si heater

Metalheader

15 m

m

10 mm Mesh

Nylon capcharcoalFilterMesh

Metal canGold wire

Sensor die

Figure 13.1 Diagram of the MGS 1100 sensor from Motorola. Micromachined sensorelement is illustrated on the left, and the sensor housing on the right. The sensitivefilms were obtained by rheotaxial growth and thermal oxidation of tin layers depositedon the silicon oxideenitride membrane. From Simon I, Barsan N, Bauer M, Weimar, U.Micromachined metal oxide gas sensors: opportunities to improve sensor performance.Sens Actuators B 2001;73:1e26.

Micromachined semiconductor gas sensors 415

nanostructured metal oxides, but one can question whether this would bethe solution to the main problems remaining with thin- and thick-filmdevices. Despite the extensive work carried out in this regard, little has trans-ferred to and been exploited by industry so far. However, since 2010,different companies have been gaining interest in micromachined semicon-ductor gas sensors, such as AMS in Austria, Bosch in Germany, Figaro inJapan, and Sensirion AG in Switzerland. Microhotplates being a matureand robust technology, the main issue remains of the synthesis of performingmaterials and their effective integration into a robust manufacturing process.One trendy approach is the use of digital printing, i.e., inkjet, to depositmetal oxide nanoparticles in solution.

Research and developments since the end of the 1980s has reported ahuge set of metal oxide materials and hotplate combinations. Because oflimitations of space, it has been necessary to be selective regarding thework to be presented in this chapter, which is far from exhaustive. Moredetails on the different configurations of alumina- and silicon-type metaloxide gas sensors can be found in Ref. 3.

13.3 Microhotplate concept and technologies

Silicon micromachining has been used to generate thermally insulatedheating elements suspended on a dielectric membrane. By patterningmetallic electrodes (Au, Pt) on top of the membrane, these structures havebeen applied as low-power transducers in metal oxide gas sensors. Thissection provides information on the design, fabrication, characteristics, andreliability of microhotplates used in semiconductor gas sensors.

13.3.1 Concept and thermal designThe operation of a metal oxide gas sensor relies on the change in resistanceof an n- or p-type semiconducting layerdmainly SnO2dwhen exposed toreducing or oxidizing gases.

A diagram of a typical cross-sectional view of a silicon micromachinedmetal oxide (MOX) sensor is presented in Fig. 13.2. Their developmenthas evolved toward silicon substrates to produce devices suitable for commer-cialization due to their low cost, low-power consumption, and high reli-ability. To lower the resistivity of the gas-sensitive film, as well as toimprove the kinetics of the chemical reactions, the metal oxide layer is heatedwith a microheater. The heated area is usually embedded in a thin dielectricmembrane to improve the thermal insulation and to reduce the power


consumption of the device, which is typically in the order of a few tens ofmilliwatts at 300 �C, and its thermal time constant (few to tens of millisec-onds). Thermal programming allows kinetically controlled selectivity.

Fig. 13.3 illustrates the heat losses that occur in a microhotplate whenoperating. The thermal energy, Q, generated by the Joule effect in themicroheater, is given by

DQ¼R$I2$Dt (13.1)

where I is the current flowing through the heater with a resistance R duringDt time. This heat is dissipated in the device and in the surrounding envi-ronment by three means:• conduction in the device;• convection in the surrounding media (typically air); and• radiation.

Dielectricmembrane

Gas sensitive layerElectrodes

Heater

Si

Figure 13.2 Cross-sectional diagram of a micromachined metal oxide gas sensor.

Convection

Conduction

Tamb

Radiation

Thot

Figure 13.3 Heat losses in a microheating device: conduction, convection, andradiation.


Thus, the heat generated by the microheater is equal to the sum of theheat lost by conduction in the device,Qcond, by convection in the air,Qconv,and by radiation, Qrad:

R $ I2$Dt ¼ DQcond þ DQconv þ DQrad (13.2)

The thermal design of microhotplates is mainly based on finite elementsimulation with the objective of optimizing the power consumption andobtaining a uniform temperature distribution over the active area. A precisemodel to evaluate the uniformity of power consumption and temperatureover the heated area requires many empirical parameters to be known ormeasured accurately.8,9 Different heater layouts have been published, mainlymeander or spiral shapes6,10 spiral shapes exhibiting better spatial tempera-ture uniformity.11,12 Improvement in temperature uniformity was alsoattempted by using a plate heater as shown by Cakir et al.13. A maximumtemperature variation of 7% was reached in the sensor-active area usingan ITO-based heater. Also the implementation of an array of sensors on asingle membrane/heater has been considered to decrease size/cost and over-all power consumption.

13.3.2 Microhotplate realization and performanceMicrohotplates are made using a combination of thin-film and siliconmicromachining processes. There are two main kinds of micromachinedsilicon substrates: closed membrane and bridge membrane. They consistof a suspended thin dielectric membrane, made of silicon nitride and/orsilicon oxide, that is released using silicon micromachining on either theobverse or reverse faces. The typical thickness of the membranes is from0.5 to 2 mm. Closed membranes have lateral dimensions of about0.5e1 mm, with approximately half the length being used as the activearea. Edge effects can be minimized by using circular membranes.14 Thetypical lateral dimensions of bridge membranes lie between 100 and200 mm. A silicon plug/island or a highly thermal conductive material,such as silicon carbide, can be implemented to improve uniformity oftemperature. Diagrams of these structures are presented in Fig. 13.4. Abridge membrane exhibits lower power consumption due to better thermalinsulation from the silicon substrate, whereas a closed membrane is moreconvenient for patterning the sensing element. In addition, silicon micro-electronics components can be integrated on the thermally insulated areaof the device. Amor et al.15 integrated temperature-measurement diodesand metal oxideesemiconductor field-effect transistor (MOSFET) under


their microheater that could be heated up to 335 �C. N-MOSFET andp-MOSFET showed good properties up to 280 and 240 �C, respectively.

Microhotplates with a bridge-membrane design based on CMOS-compatible processes were proposed by Cavicchi et al.16. The architectureof the hotplate is presented in Fig. 13.5. During the 2000s, the Swiss FederalInstitute of Technology Zurich (ETHZ), Switzerland, came up withdifferent generations of CMOS micromachined metal oxide gas sensors

Sensing material (thin or thick film)

Sensing material (thin or thick film)

Heater + thermometerElectrodes

Heater + thermometer

~ 1–2 μm

~ 1–1.5 mm

~ 1–1.5 mm

~ 100–200 μm

~ 1–2 μm

~ 400 μm

~ 400 μm Si

Si

SiSi

Si

Si plug

Electrodes

Pit

Anisotropic etching

Sacrificial etching

Suspension beamsActive area

Active area

(a) (b)

(c) (d)

Membrane

Figure 13.4 Diagram of a suspended membraneetype gas sensor; (a and b) reverse ofsilicon micromachining; (c and d) obverse surface micromachiningd(a and c) top view,(b and d) side view. Adapted from Simon I, Barsan N, Bauer M, Weimar, U. Micromachinedmetal oxide gas sensors: opportunities to improve sensor performance. Sens Actuators B2001;73:1e26.


with integrated driving and readout circuitries.17 The heat necessary for thechemical reactions between the gaseous environment and the sensing layerwas provided by the Joule effect through a field-effect transistor (FET) orpolysilicon resistor. For improved reliability, platinum and tungsten arepreferred as heater material at the time of writing. More details on the heaterperformances are provided in Section 6.3.3, on reliability. The heater andthermometer, which are needed to control the sensor operation tempera-ture, can be either a dual purpose unit or two separate components. Poly-silicon and platinum have often been used; microelectronic components,such as a forward bias silicon pen junction as a temperature sensor, can beconsidered when silicon is available on the membrane.

With a resolution in the micrometer range for the photolithographicpatterning of the electrodes, the gas-sensitive area can be significantlyreduced in comparison with screen printing on ceramic substrates.Regarding the electrode material, platinum is favored because it showsvery good chemical stability and can provide higher gas responses.18 Thetwo main approaches for the deposition of the gas-sensitive sensing layer

(a)

Suspendedstructure

SnO2 oxide film

Film contactsInsulating SiO2

Insulating SiO2

Doped polysiliconheater

(b)

50 μm

Figure 13.5 Obverse of CMOS silicon micromachined hotplate: (a) optical picture;(b) diagram. Courtesy of Dr Steve Semancik, NIST, USA.


are either thin- or thick-film techniques. A thin film is usually realized byevaporation or sputtering; a thick film is deposited by screen printing, spraypyrolysis, or drop coating.3 Once deposited, these materials usually requireannealing at high temperatures (350e800 �C) in an oxygen-containingatmosphere to modify the morphology (e.g., grain size) and microstructure(e.g., porosity, surface-to-volume ratio). The parameters of this annealingstep have to be carefully selected to be compatible with the microhotplateitself. Some temperature limitations occur with microhotplates based on aCMOS-compatible process.

Several micromachined hotplates for metal oxide gas sensors have beenreported in the literature. However, robust and established technologies allmake use of the closed-membrane design in combination with platinum asthe electrode material. Recent papers show that platinum is now mainlyused as a heater material with tungsten applied in CMOS-compatibledevices. The characteristics of some representative examples are summarizedin Table 13.1. The optimization of the micromachined platform is very closeto the optimum achievable, with a minimum active areadand, therefore,power consumptiondreached. According to the resolution of the photo-lithographic process, it is becoming difficult to further reduce the size ofthe hotplates and yet retain an exploitable sensing layer and heater resistancevalues. The next steps are toward using nanopatterning techniques,self-heated metal oxide nanostructures, and printing on flexible polymericsubstrates, as presented in Sections 6.7 and 6.8.

13.3.3 Microhotplate reliabilityOperating at a relatively high temperature, the electrothermomechanicalreliability of micromachined hotplates is an important aspect for metal oxidegas sensors.

Numerical thermomechanical studies have been performed to improvethe robustness of the membrane, addressing buckling and stress concentra-tion.19 Thermomechanical reliability depends on the design and materialsused. In general, the membranes made of dielectric materials deposited at ahigher temperature (e.g., low-pressure chemical vapor depositiondLPCVD)are more robust. Attempts with SiN deposition were also performed byPECVD, however, they revealed to be less robust.20 The membrane is usuallyformed of a stress-compensated stack of thin films of silicon nitride, siliconoxynitride, and/or silicon oxide. A heater embedded in between LPCVDlow-stress silicon nitride thin films has proven to be robust.6,21 This dielectricmaterial is, however, not commonly available in MEMS foundries. Different


Table 13.1 Comparison of various microhotplate designs that have been reported in the literature.

Year AuthorActive area(1000 mm2)

Hotplate area(1000 mm2)

Power at300 �C (mW)

Power/heaterarea (mW/1000 mm2)

CMOSYes/No

Membrane(M)/bridge (B)

Materialmembrane orbridge Heater

1990 Dibbern 202.5 1822.5 55 0.27 No M Oxynitride NiFe1993 Suehle 10 40 40 4.00 Yes B CMOS films Poly-Si1995 Zanini 722.5 1440 90 0.12 No M Oxynitride Pt1995 Gardner 472.5 3596.4 40 0.08 No M Nitride Pt1996 Aigner 300 1000 35 0.12 No M Nitride Pt1996 Lee 10 1000 18 1.80 No M Oxynitride Poly-Si Pt1997 Gotz 250 1210 55 0.22 No M Oxynitride Poly-Si1998 Guidi 562.5 2250 67 0.12 No M Nitride Pt1998 Astie 230 3240 125 0.54 No M Si/SiO Poly-Si1999 Horrillo 250 1210 38 0.15 No M Nitride Poly-Si2001 Udrea 90 250 100 1.11 Yes M CMOS films FET2001 Benn 40 NA 8.6 0.215 No B SiC SiCeN2002 Afridi 10 44 27.5 2.75 Yes B CMOS films Poly-Si2002 Briand 202.5 1000 50 0.25 No M Nitride Pt/FET2002 Mo 6.4 25.6 6 0.94 No B Oxynitride Pt2002 Chan 14.4 57.6 60 4.17 No B Oxynitride Poly-Si2003 Lee 31.4 1000 30 0.95 No M Nitride Pt2003 Tsamis 10 NA 15 1.50 No B Porous Si Poly-Si Pt2004 Fujres 10 NA 7.5 0.75 No B Nitride Pt2004 Baroncini 250 1000 20 0.08 No M Nitride Pt2004 Laconte 57.6 409.6 13 0.22 No M Oxynitride Poly-Si2005 Graf 70.7 250 50 0.71 Yes M CMOS films Poly-Si2005 Lee 3990 NA 73 0.02 No B Nitride Pt

422D.Briand

andJ.C

ourbat

2006 Elmi 20.1 NA 6 0.30 No B Oxynitride Pt2006 Belmonte 160 NA 30 0.19 No B Oxynitride Pt2007 Guo 36.1 90 23 0.64 No B Oxynitride Pt2008 Barborini 1000 NA 24 0.02 No B Oxynitride Pt2008 Briand 250 2250 60 0.24 No M Polyimide Pt2008 Ali 17.67 250 14 0.79 Yes M CMOS films W2008 Ali 0.452 70.7 6 13.27 Yes M CMOS films W

Notes: Where exact values are not given, they have been deduced from the information given in the particular paper. NA: not available.Adapted from Ali SZ, Udrea F, MilneWI, Gardner JW. Tungsten based SOI microhotplates for smart gas sensors. J Microelectromech S 2008;17(6):1408e1417; in whichall references can be found.

Microm

achinedsem

iconductorgas

sensors423

techniques have been implemented to improve mechanical stability of themembrane. Iwata et al.22 added SU-8 structures to reinforce bridges of asuspended membrane at a cost of a higher power consumption.

Accelerated aging tests have also been developed to determine andanalyze the failure mechanisms by thermally cycling the device, by rampingup the power until breakdown, or by operating it at temperatures higherthan their normal use.23e25 Cracks in the dielectric membrane, electromi-gration, and electroestress migration have been identified as the main causesof failure.26 At high temperatures, the migration of the platinum atoms inthe heater meander was linked to the mechanical stress in the dielectricmembrane. They usually occur in location of high temperature gradientand/or high current densities. Reduction of current density accumulationbetween two different conductive materials has been achieved by27.Platinum was used for the heater, while conductive tracks were made ofAlCu. Current density at the metal junction could be reduced by 20% byforming a slope of 45 � at the end of the AlCu line, reducing failure likeli-hood of the electrical connection. State-of-the-art technologies can allowtemperature cycling up to several millions of cycles before failure, enablingtemperature modulation of the sensor (Section 6.4.3).

The heater material is a crucial point for the stability of this type of deviceduring operation. Driven by CMOS compatibility, poly-Si was first used butit suffers from an inappropriate drift of its electrical resistivity at high temper-ature.28 Platinum is the material that has been implemented for the heaterfor improved reliability. It is used in most micromachined metal oxidesensors on the market at the time of writing, not only for the heater butalso for the electrodes. Courbat et al.29 showed that adding a small amountof another refractory metal (such as iridium) to the platinum can improve itsresistance to electromigration. However, Mo exhibited superior perfor-mances to platinum, allowing higher operational temperatures30 and lowheater resistance drift.20

TiNda CMOS-compatible materialdhas been applied as a heatingelement showing relatively better performances than platinum.31 FETshave also been implemented as heaters in CMOS technology but this re-quires a silicon area in or underneath the membrane.32 A very low-powermicromachined hotplate platform was designed using SOI technology anda robust tungsten heater.8 This device is on the market in the products port-folio from Cambridge CMOS Sensors Ltd in the United Kingdom, whichwas acquired in 2016 by AMS, Austria. One constraint is the obligation towork with the thin films available in the CMOS process. Depending on the


process, the CMOS dielectric stack of films is not always optimum and post-processes can be necessary. For instance, this can involve the deposition ofthe metallic electrodes (Pt, Au), or a passivation and stress compensatingdielectric thin film.

13.4 Micromachined metal oxide gas sensors

In the main, two types of metal oxide gas-sensitive films have beenintegrated into micromachined hotplate transducers: thin and thick films.The different developments will be presented in this section. The integrationof a third type of structurednanowires, into which considerable efforts arebeing made at the time of writingdwill be presented in Section 6.7.1, Trendsand perspectives. Other chapters in this book address in detail the synthesis,sensing mechanisms, and properties related to these different sensing films.

In this section, for better readability and to allow comparison betweenresults, all responses are given as Rgas/Rair if Rgas > Rair or as Rair/Rgas ifRair > Rgas, where Rair is the baseline resistance of the sensors in air andRgas is its resistance when exposed to the analyte under examination.

13.4.1 Thin gas-sensitive filmsFirst, micromachined gas sensors were obtained using thin-film depositiontechnologies. That technique, used for semiconductor manufacturing, isavailable in most cleanrooms with evaporation or sputtering machines.The motivation at that time was to produce MEMS-based metal oxidegas sensors using thin-film technology only, being a disruptive technologycompared with the thick-film technologies used on alumina.

The first silicon micromachined thin-film metal oxide gas sensor wasdeveloped at CSEM SA, Switzerland by Demarne et al.21; this was commer-cialized at the beginning of the 1990s by Microsens SA in Switzerland. Itconsisted in a SiO2 membrane embedding a gold-based meander-shapedheater. A thin film of SnOx was sputtered and patterned by lift-off. Twoconfigurations were proposed, without and with a silicon plug to makethe temperature reached in the active area of the device more uniform.To attain 300 �C, the supplied powers were, respectively, 104 and183 mW. Motorola also showed a significant interest in the developmentof commercial micromachined thin-film gas sensors for CO detection.33

They ceased their activities in that field at the end of the 1990s. Develop-ment was pursued by MiCS (MicroChemical Systems SA) in Switzerland,now part of the SGX Sensortech group.


Other techniques have been used for the fabrication of thin-film metaloxide gas sensors. At NIST in the United States,16,34 produced gas sensorsby chemical vapor deposition (CVD). By applying a current and thus heat-ing the hotplate, sensing films could be deposited locally (i.e., only on heatedactive areas) using an adequate organometallic precursor. SnO2 and ZnOfilms were obtained with tetramethyltin and diethylzinc in an oxygen atmo-sphere. They were deposited onto different seed layers, which played asignificant role in terms of gas selectivity.

Besides CVD and sputtering from a target of the desired material, thinfilms were obtained by sputtering or evaporation through the rheotaxialgrowth and thermal oxidation (RGTO) process. This method consists indepositing thin layers of a metal, followed by its thermal oxidation in anoxygen-rich atmosphere. Tin oxide layers of 350 nm in thickness wereobtained with this technique from sputtered Sn by Faglia et al.11 for thedesign of CO sensors. The highest sensitivity to CO was obtained at anoperating temperature of 400 �C. Responses of between two and threewere obtained when the device was exposed to 25 ppm of CO, the alarmlevel in many countries. With the same technique,35 grew SnO2 films onvery low power hotplates. A temperature of 300 �C was reached with asupply power of 6 mW. The sensor had a response of 7.5 when exposedto 100 ppb of NO2 at 200 �C and 5 under 10 ppm of CO at 450 �C.

In 2003, the European Aeronautic Defence and Space company in Ger-many developed gas sensors based on silicon technology to replace thick-film devices, which were usually based on alumina substrates and had ahigh level of power consumption.36 A main drawback Muller et al. identi-fied in Si-based devices was their fragile membrane. Therefore, they builttheir devicesdan array of three hotplatesdfrom SOI to keep the top Silayer as a robust suspended membrane. A fabrication yield of 100% wasachieved with a top Si layer thicker than 5 mm. Typical power consumptionswere in the range of 50 e80 mW to reach an operating temperature of300 �C. The active area of the device could be operated at different temper-atures and functionalized through thin- and/or thick-film technology.Friedberger et al.37 evaporated Sn and obtained SnO2 by RGTO. Thesensing film had good sensitivity toward hydrocarbon and hydrogen, buta very low response to CO.

W€ollenstein et al.38 developed an array combining several gas sensinglayers by successive photolithography steps and sputtering or e-beam evap-oration. A device with four different metal oxide layers could be produced.The films had to be deposited in a specific order, depending on the


temperature required for stabilization. The layer with the highest annealingtemperature was deposited first. Titanium-doped chromium oxide wasproduced by successively evaporating Cr and Ti layers, which were subse-quently annealed at 850 �C. ZnO films were obtained by direct current(DC) magnetron sputtering from a Zn target combined with an Ar/O2

plasma. Pt-doped SnO2 films were obtained by radio frequency magnetronreactive sputtering from a SnO2 target followed by the deposition of a fewtens of a nanometer of Pt. The sintering of ZnO and SnO2 films occurred ata temperature of 700 �C and could be performed simultaneously. As forZnO, WO3 was sputtered from a W target in an Ar/O2 plasma with alow deposition rate to ensure proper oxidation of the material. The lastmaterial that could be deposited was V2O5. It was performed by e-beamevaporation of vanadium under controlled oxygen pressure. To reach a fullyoxidized film, the evaporation was followed by an additional oxidationtreatment at 500 �C in synthetic air. The silicon wafer was then bondedto a micromachined glass component acting as a structural element. Toreduce power consumption as much as possible, the reverse of the Si waferwas wet etched in a KOH solution. Etching stopped at the dielectric thinfilms and at the highly doped Si layer. Gas sensing measurements arepresented in Fig. 13.6. The sensors were operated at about 200 mW to reach

1M

100k

10k

600500400300

200

10010 15 20

Time (hours)

Sen

sor r

espo

nse

(Ω)

25

H2

100 ppm 50 ppm 2 ppm 100 ppm

NO2 NH3CO

30 35

WO3

CTOZnOV2O5

SnO2

Figure 13.6 Gas measurement obtained from the sensor array. The sensing materialsexhibited different behavior toward the analytes. From W€ollenstein J, Plaza JA, CanéC, Min Y, B€ottner H, Tuller HL. A novel single chip thin film metal oxide array. SensActuators B 2003;93:350e355


a temperature of 400 �C. They were exposed to H2, CO, NO2, and NH3 astesting gases. Discrimination can be made between them because somematerial resistive variation was observed only for specific gases. ZnO wasthe only layer exhibiting a response to NO2 and V2O5 to NH3.

In the mid 2010s, there has been a renewed interest in using pulsed laserdeposition to produce metal oxide films with various morphologies onmicromachined silicon hotplates.39 SGX SensorTech SA in Switzerlandand Bosch in Germany have notably evaluated this technique to manufac-ture thin film metal oxide gas sensors integrated on MEMS hotplates.

13.4.2 Thick gas-sensitive filmsIn the mid-1990s, thick filmebased metal oxide sensors began to attractattention. There were issues regarding the stability and reproducibility ofmetal oxide thin films. New deposition methods brought from outsidethe semiconductor industry were useddmainly pipetting, drop coating,and screen printing. The first combination of a thick-film sensing layer com-bined with a microhotplate was carried out by Barsan (see 40), by pipettingpure SnO2, 0.2% Pt-doped SnO2, or 0.2% doped SnO2 on gold electrodespatterned on micromachined hotplates. A polycrystalline structure wasobtained by sintering the SnO2 layers at 600 �C in air. A power supply of60 mW was needed to operate the sensor at 400 �C. The pure SnO2-basedsensors showed the best sensitivity to organic solvents. It exhibited a resis-tance variation of 32% when exposed to 25 ppm of n-octane. The sensor’sresponse and recovery times were, respectively, 40 and 60 s. Drop coating ofPd-doped SnO2 pastes was first introduced by41. The sensing material wasdeposited on micromachined hotplates for the discrimination of CO,NO2, and their binary mixtures. Briand et al.6 used this technique for thedeposition of 2% Pd-doped SnO2 paste42 on interdigitated Pt electrodes.The diameter of the drop was 400 mm with a thickness of a few tens ofmicrons. It was deposited on a membrane of 1 � 1 mm2 and 1 mm thick.The sensing material could be annealed on a chip using the sensor’s heater.For operating the device, a temperature of 300 �Cwas reached with a powersupply of 70 mW. The device showed a response of 2.2 and 1.4, respec-tively, to 10 ppm of CO and 2000 ppm of CH4. Despite their high thick-ness, drop coating has led to highly stable, reproducible sensors with verygood sensitivity. These results led to the large-scale commercialization ofdrop-coated metal oxide gas sensors by AppliedSensor GmbH, Germany,for the automotive market.43 The microhotplate technology developedby Briand et al.6 has been combined with much thinner optimized SnO2


and WO3 films, having a thickness of less than 5 mm (Fig. 13.7). Typical gasresponses are displayed in Fig. 13.8. The metal oxide drop was then furtherreduced by using capillaries for its deposition and could reach a diameter ofabout 20 mm.44 Smaller hotplates can be thus used, leading to a potentialfurther decrease in power consumption.

Drop coating was then used by many other groups. Among others,45 andlater,46 fromMorante’s group in Spain used it for the deposition of SnO2 andBaSnO3. Espinosa et al.47 in Italy deposited drops made of 1% Pt-dopedWO3, 1% Pt-doped SnO2, 1% Pd-doped SnO2, and 1% Au-doped SnO2

on a suspended microhotplate with a diameter of 80 mm. It required about8 mW to reach an operating temperature of 400 �C. As test gas, the sensingfilms were tested with ethylene, acetaldehyde, ethanol, and ammonia.

A second technique widely used for the deposition of thick-film metaloxides on alumina substrates is screen printing. Looking at the success metby the thick drop-coated films, screen printing was reconsidered. Vincenziet al.48 screen-printed Pd-doped SnO2 paste onto micromachined micro-hotplates. The paste also contained a glass frit (a low melting temperature

(a)

(b)Sensing layer Heating electrodes

Sensing electrodes

Figure 13.7 (a) SEM image of a drop-coated metal oxide gas sensor from AppliedSen-sor GmbH. (b) Three-dimensional schematic drawing of the sensor structure. FromBlaschke M, Tille T, Robertson P, Mair S, Weimar U, Ulmer H. MEMS gas-sensor array formonitoring the perceived car-cabin air quality. IEEE Sens J 2006;6(5):1268e1308.


glass) to increase its viscosity and improve adhesion to the substrate. Partic-ular care had to be taken to avoid breaking the SiO2/Si3N4 membraneduring film deposition. This was achieved by using a special stencil, whichreduced pressure on the membrane. The sensing film was 250 � 350 mm2

and had a thickness of about 40 mm. The film was then fired at 650 �Cfor 1 h, using the sensor’s heater. For gas detection, the devices operatedat 400 �C with a power of 30 mW and were evaluated with CO, CH4,and NO2. Fairly low responsesd1.2 for 50 ppm of CO, 1.03 for1000 ppm of CH4, and 1.7 for 0.1 ppm of NO2dwere obtained. It wasascribed to the glass frit, which insulated the SnO2. To avoid breaking themembranes during screen printing,49 deposited a 5 mm thick, undopedSnO2 sensing film before releasing the membrane. It led to a significantlyimproved yield of 95% after encapsulation of the sensors. They showedresponses of about 3e25 ppm ethanol and to 625 ppm of ammonia and8e62.5 ppm of acetone. From the same group,50 screen printed SnO2

and WO3 pastes on micromachined transducers. When exposed to CO,

CONO2Butanol

0

60

50

40

30

20

10

0

Con

cent

ratio

n (p

pm)

(a)

(b)

50 100 150 200 250Time (min)

300 350 400 450 500

Pd3Pt02U

0

10–2

10–3

10–4

10–5

Log

cond

uctiv

ity (S

)

10–6

50 100 150 200 250

Time (min)300 350 400 450 500

Figure 13.8 Gas response (b) of 3% Pd-doped SnO2 (Pd3), 0.2% Pt-doped SnO2 (Pt02),and undoped SnO2 (U) to changes in concentrations of CO, NO2, and butanol (a). FromBlaschke M, Tille T, Robertson P, Mair S, Weimar U, Ulmer H. MEMS gas-sensor array formonitoring the perceived car-cabin air quality. IEEE Sens J 2006;6(5):1268e1308.


low responses were obtained by SnO2 and no response was observed withWO3. In the case of an exposure to 1 ppm of NO2, responses of 3.63with SnO2 at 250 �C and 8.91 with WO3 at 200 �C were measured. More-over, Ivanov et al. sputtered the same materials so as to investigate andcompare the sensing properties of thin- and thick-film metal oxide layers.The results revealed that thick-film gas sensing layers have a higher degreeof sensitivity than thin-film layers. This is due to the nature of the depositedfilm, which is more compact in the case of thin films, thus reducing thesurface-to-volume ratio.18

SnO2 screen printing paste contains a binder to control the rheologicalproperties and to ensure a good adhesion of the film to the substrate.Glasses bring problems of SnO2 percolation and thus reduce the conduc-tivity. Remedy to this issue,51 evaluated different inks with an optionalorganic binder, instead of a mineral binder, and with Sn alkoxide, whichlead to the formation of SnO2 during thermal annealing. Sensor filmswith a low conductance were obtained when no binder was used becauseof numerous cracks in the layer. The presence of both the organic binderand the alkoxide gave good results in terms of paste adhesion and conduc-tivity, but the pattern resolution achieved was limited. However, nowa-days, screen printing resolution down to 20 mm has been demonstratedin the field of printed electronics and better results could be expected formetal oxide pastes.

Beside drop coating and screen printing, a further technologydflamespray pyrolysis (FSP)dshowed promising results. The deposition techniqueconsists in spraying liquid precursors, which form a flame. The precursorsreact in the gas phase with the subsequent particle formation. This methodallows a good control on morphologydamorphous or crystallinedas wellas doping. Films with thicknesses of a few micrometers which do not requireany annealing can be obtained. Sahm et al.52 used this method for the depo-sition of SnO2 on alumina substrates. Gas measurements were performed.The SnO2 sensing film showed a good response to low concentrations ofNO2 (below 200 ppb) and propanal, and a low response to CO, which istypical for undoped SnO2 films. K€uhne et al.53 used the same method forthe deposition of Pt-doped SnO2 onto micromachined hotplates. Thesensing film was patterned through a shadow mask. The transducer coatedwith the sensing film is presented in Fig. 13.9(a). The devices operated at250 �C with a power supply of about 25 mW. It showed a good responsetoward ethanol concentrations between 25 and 100 ppm, as illustrated inFig. 13.9(b).


13.4.3 Temperature modulationMetal oxide gas sensors can be operated in two modes: constant temperature(i.e., isothermal) and temperature-modulated modes. In constant tempera-ture mode, the selectivity can be enhanced by using an array of sensorscovered with different materials or dopants38,54; or by operating at different

100 μm

25

50

100

50

25

01

2

3

4

5

(a)

(b)

5 10 15Time (min)

Sen

sor r

esis

tanc

e (M

Ω)

20 25 30

Figure 13.9 (a) SEM image of a Pt-doped SnO2 film deposited by flame spray pyrolysis;(b) gas response to ethanol. Values given are EtOH concentrations (ppm); 90% r.h.(20 �C); Tm ¼ 250 �C. (b). Adapted from K€uhne S, Graf M, Tricoli A, Mayer F, Pratsinis SE,Hierlemann A. Wafer-level flame-spray-pyrolysis deposition of gas-sensitive layers onmicrosensors. J Micromech Microeng 2008;18:035040.


temperatures.16,41 However, the use of several sensors considerably increasesthe complexity and the power consumption of the system. Additionally, adrawback with constant temperature operation is that a mixture of oxidizingand reducing gases can offset each other and no signal variation will beobserved.18 With the micromachining of the devices, their thermal responsetimes were drastically reduced to the millisecond range. This allowed theiroperation in a pulsed or cycled temperature mode to avoid the interferenceof humidity and allowed the discrimination of several gases with one singlesensor. This measurement technique was first introduced by55. They applieda sine signal to the sensor heater and measured the response of the SnO2 gassensing layer when exposed to different analytes. They observed thatmethane and propane gave a higher response with a heater at its maximumtemperature, while CO is better measured in a cooling state. Each gas can beidentified by a specific temporal response pattern, which depends on itschemical reaction with the gas-sensitive material.56

Major investigations related to temperature-modulated micromachinedmetal oxide gas sensors were performed in Semancik’s group. Rattonet al.57 applied a sawtooth signal shape to the heater to reach temperaturesup to 550 �C. The behavior of methanol, ethanol, acetone, and formalde-hyde was studied. The sensor signal was processed through the GrameSchmidt approach, fast Fourier transform (FFT), Haar wavelet transform,or the Granger approach to reduce the number of coefficients describingthe signal and to retain as much relevant information as possible. Best resultswere achieved with the Haar transform, which efficiently compressed theinformation while removing noise and drift effects. Kunt et al.58 used thesame device to discriminate methanol and ethanol using temperature mod-ulation. Both gases responded differently to the temperature change, as canbe seen in Fig. 13.10. In this study, they optimized the temperature profileto improve response selectivity between these two gases.

The sensitivity can be further improved by taking advantage of theunsteady state of the number of oxygen species at the surface of the metaloxide when its temperature is changing. Llobet et al.59 showed that thetransient response of thermally cycled metal oxide sensors decreases thesensor’s response to humidity and to the drift in the resistance of the gas-sensitive layer. Several options of temperature variations have been pre-sented in the literature to improve selectivity. Different waveforms atdifferent frequencies have been applied to the heater of the gas sensor toachieve thermal cycling of its temperature. The sensor response can bethen analyzed by signal processing. FFT was used by60. They applied a


sine wave and its second harmonic to the sensor heater to improve theselectivity of a SnO2 semiconductor gas sensor. Depending on the phaseshift of the second signal compared with the first, discrimination betweenalcohols, hydrocarbons, and aromatic compounds could be performed.Fig. 13.11 shows the sensor response to ethanol, ethane, and toluene asrepresentative examples of these gas families. Llobet et al.59 used discretewavelet transform and an artificial neural network to measure and discrim-inate CO, NO2, and their mixture. The wavelet technique gave better re-sults than FFT in terms of data compression and tolerance to noise and driftin the sensor response.

A system based on simpler electronics relies on pulsing the temperature(i.e., the heater is only switched on and off). Depending on the duty cycle,it allows a significant reduction in power consumption.44 Among othertechniques, this was used by Faglia et al.11 for the detection of CO withan Au-doped SnO2 film. They used a square signal with a period between0.5 and 180 s. The heater was powered for 100 ms, which was sufficient toreach a steady state. Beside a reduction in power consumption, Faglia et al.

2

1

0

–1

–296 98 100 102 104

Time (s)

Nor

mal

ized

con

duct

ance

Tem

pera

ture

(°C

)

106 108 110

380360340320

280260

300

96 98 100 102 104Time (s)

106 108 110

(a)

(b)

Figure 13.10 (a) Conductance response of a SnO2 gas sensing layer to ethanol (dashedline) and methanol (solid line) for temperature pulses of 275 and 380 �C (b). From KuntTA, McAvoy TJ, Cavicchi RE, Semancik S. Optimization of temperature programmed sensingfor gas identification using micro-hotplate sensors. Sens Actuators B 1998;53:24e43


observed an increase in sensor response, compared with DCmeasurements,for periods up to 20 s. Therefore, such a method can allow a reduction inpower consumption while improving sensing performances.

13.4.4 PackagingSilicon micromachined semiconductor gas sensors are mainly packagedusing standard metallic transistor outline (TO) headers as support, andwire bonding is used for their electrical connection. Typically, a metalliccap with a grid is fixed to the TO header with a hydrophobic gas permeablemembrane on top of it. A filtering agent can be also included in the package.

The use of silicon microfabrication techniques brings not only the abilityto process the sensors at wafer level but also, as demonstrated in Raibleet al.61 in 2006, the encapsulation and testing of the sensors at wafer level.This concept allows liquid-tight sealing of gas sensor devices, which protects

1.0(a)

(b)

(c)

0.5

400 450 500 550

(a-1)1.0

0.5

400 450 500 550

(a-2)1.0

0.5

400 450 500 550

(a-3)1.0

0.5

400 450 500 550

(a-4)

0.4

0.2

G (m

S)

T (K)

400 450 500 550

(b-1)1.4

0.2

400 450 500 550

(b-2)1.4

0.2

400 450 500 550

(b-3)1.4

0.2

400 450 500 550

(b-4)

1.0

0.5

0400 450 500 550

(c-1)1.0

0.5

0400 450 500 550

(c-2)1.0

0.5

400 450 500 550

(c-3)1.0

0.5

0400 450 500 550

(c-4)

Figure 13.11 Sensor temperature versus sensor conductance of the sensor when a sinewave was applied to the heater without the second harmonic (dashed line) and withthe second harmonic (solid line) with a phase shift of (1) 0 rad, (2) p/2 rad, (3) p rad,and (4) 3p/2 rad. The sample gases tested were 1000 ppm of (a) ethanol, (b) ethane,(c) toluene. From Nakata S, Okunishi H, Nakashima Y. Distinction of gases with a semi-conductor sensor under a cyclic temperature modulation with second-harmonic heating.Sens Actuators B 2006;119:556e561.


them during production (e.g., wafer dicing) and later in the application,while still allowing the target gases to reach the sensing layer. The basis ofwafer-level packaging is the combination of a structured Pyrex wafer witha micromachined substrate wafer. Thereafter, thick-film SnO2 layers aredeposited and stabilized before a diffusion membrane is attached, which sealsthe wafer stack as shown in Fig. 13.12. The wafer stack is finally diced intoindividual sensor elements which can be mounted on printed circuit boardusing different interconnection methods, such as chip on board, flip-chip,tape-automated bonding, and so on (Fig. 13.13).

Briand et al.62 reported on a higher level integration of wafer-level pack-aged micromachined metal oxide gas sensors. The concept was based on theinsertion of the metal oxide drop into the micromachined cavity in thesilicon substrate with the platinum electrodes at its bottom. Using this

Thick-film SnO2 layer(a)

(b)

Diffusion filtermembrane

Pyrex filter support

Optional base toease pick and place

Micro-machinedsubstrateMicro-machinedhotplate membraneBonding pads

Figure 13.12 (a) Diagram of the wafer-level packaged metal oxide sensor; (b) opticalpicture of an individual sensor area with the Pyrex rim and the metal oxide drop beforethe fixation of the gas permeable membrane. From Raible S, Briand D, Kappler J, de RooijNF. Wafer level packaging of micromachined gas sensors. IEEE Sens J 2006;6(5):1232e1235.


approach, the Pyrex rim was no longer necessary and the gas permeablemembrane could be fixed directly onto the silicon substrate to close thecavities containing the drop-coated metal oxide film (Fig. 13.14). For a200 mm-wide deep reactiveeion-etched (DRIE) membrane, a powerconsumption of 15 mW was reached at 300 �C. DRIE technology alsoallows the reduction of the chip size to a minimum, compared withKOH etching. Following the trends in the field of sensor packaging andmounting, surface-mount devices are appearing on the market using aplastic, molded package as a cost-effective approach, as it is described forthe gas sensor product from Sensirion AG, Switzerland, in Section 6.9.

13.5 Complementary metal oxidesemiconductorecompatible metal oxide gas sensors

CMOS-compatible and SOI-based microhotplates used as transducersfor metal oxide gas sensors were reported, respectively, by Suehle et al.63 andLaconte et al.64 They addressed the realization of the hotplates themselves ina CMOS-compatible process with an integrated poly-Si heater. But the realbenefit of this technology comes with the integration of the completedriving and readout electronics on the sensor chip. Beside the potentialreduction of power consumption and the cost of the sensor system, thenumber of bonding wires can be decreased, as can the packaging. The inte-gration of the electronic circuitry can also improve signal response fidelitydue to on-chip signal processing and amplification and conditioning of small

Figure 13.13 Chip on board wafer-level packaged metal oxide gas sensors on printedcircuit board. From Raible S, Briand D, Kappler J, de Rooij NF. Wafer level packaging ofmicromachined gas sensors. IEEE Sens J 2006;6(5):1232e1235.


(a) Gas permeable membrane

Si Si

SnO2 SiO2

Si3N4

200–600 μm

390 μm

Electrodes Heater

(b)

(c)

Figure 13.14 (a) Diagram of the wafer-level packaged semiconductor gas sensors withthe metal oxide film deposited in the silicon micromachined cavity; (b) and (c) opticalimages of the DRIE etched 200 mm-wide drop-coated metal oxide device (b) before and(c) after encapsulation at the wafer level. From Briand D, Guillot L, Raible S, Kappler J, deRooij NF. Highly integrated wafer level packaged MOX gas sensors. Proceedings of theTransducers’07 conference. Lyon, France, June 10e14; 2007. p. 2401e2404.


sensor signals. Benefits can be brought to the operation of the sensor byallowing the implementation of driving, signal conditioning, and compen-sation strategies. However, if the yield of the formation of the sensing layeron the sensor chip is not sufficiently high, the failure cost will be significantlyhigher, together with the loss of the electronics.

Four main concerns need to be addressed when integrating metal oxidesensors in a CMOS-compatible process:• The dielectric membrane of the microhotplates will be composed of

CMOS dielectric films. It can be formed through a silicon microma-chining postprocess either on the back or the front.

• The standard electrically conductive materials are doped polysilicon andaluminum, which are not suitable to be used as heaters (Section 6.3.3) orelectrodes (oxidation of Al) for the sensor. Implementing platinum, thecommonly used material, as heater and electrode material involvespostprocessing steps. Another approach for the heater is to use tungstenwhich can be available in CMOS technology.

• The postdeposition of the metal oxide sensing layer needs to be CMOScompatible, and its postdeposition annealing is limited in terms of tem-perature and time.

• Once the CMOSmetal oxide sensor chip is available, the miniaturizationof the device brings different issues to the CMOS electronics design. Werefer the reader to the comprehensive review published by Gardneret al.65; for more information about electronics circuitry design.Afridi et al.66 have reported on an array of four bridge-type front micro-

machined hotplates with postprocessed gold electrodes and including inter-face electronics. The metal oxide films, tin oxide and titanium dioxide, weredeposited using an LPCVD process when operating the microhotplates atdifferent temperatures. A decoder was used to select a given microheaterand sensing resistive layer, with a bipolar transistor or a MOSFET switch,respectively. The signal-to-noise ratio was improved using an on-chip oper-ational amplifier.

ETH Zurich, in Switzerland, has extensively developed CMOS-compatible metal oxide gas sensors with on-chip integrated circuitry.12,67

Postprocessing was used to include platinum electrodes on the hotplatecoated with a drop-coated Pd-doped tin oxide film. Annealing of the metaloxide film was performed at a maximum temperature of 400 �C, whichprevented any degradation of the device. Fig. 13.15 presents an advancedanalog/digital monolithic sensor system.17 Its fabrication was performedusing an industrial CMOS process followed by postprocessing steps for


(a)

(b)

(c)

Referenceresistor

Temperaturesensor

Temperaturesensor

n-wellSi-island

Dielectricmembrane

Ringheater

Electrodes

100 μm

500 μm

6.8 4.7 mm2

Nanocrystalline SnO2thick-film layer

Si-island

Micro-hotplate

Bulk-chiptemperature

sensorAnalog

circuitry andconverters

Digital circuitry(controller

and interface)

Figure 13.15 Monolithic metal oxideebased gas sensor system in CMOS technology:(a) close-up of the microhotplate (left) and SEM micrograph of a microhotplatewith a drop-coated nanocrystalline thick-film layer (right); (b) cross-sectional diagramof the heated area of the sensor chip; (c) micrograph of the CMOS-based overall sensorsystem chip featuring microhotplates and circuitry.17


the patterning of the platinum electrodes, the release of the membrane bysilicon micromachining, and the deposition of the sensing layer. Dielectricthin films available in the CMOS process were used for the thermally insu-lated membrane, electrical insulation, and passivation. The active areafeatured a circular-shape resistive heater, a temperature sensor, and elec-trodes to contact the sensing layer. In Fig. 13.15(c), the microhotplate,the analog circuitry (including analog-to-digital and digital-to-analogconverters), and the digital circuitry are distinguishable. The digital partincluded a programmable digital temperature controller and a digitalinterface. This enabled control of the sensor temperature, as well as areadout of the temperature of the hotplate and the gas sensor signal. A log-arithmic converter connected to the resistance layout of the sensitive layernot only allowed a first-order signal linearization but also helped to addressthe large variation range of the metal oxide resistance from 1 kU to100 MU. A stand-alone version of the monolithic sensor system (includingthree transistor-heated microhotplates32 with fully digital temperaturecontrollers and a digital interface) was developed to take complete advan-tage of this technology. In 2017, Sensirion AG, Switzerland, has released aCMOS compatible metal oxide gas sensor product for which more detailscan be found in Section 6.9.

Robust high-temperature tungsten-based SOI microhotplates werereported by Ali et al.8 and have been successfully commercialized byCambridge CMOS Sensors Ltd. in the United Kingdom. The hotplatesare fabricated using a standard SOI CMOS process in a commercial foundry,followed by a DRIE postprocessing step to release the dielectric silicon di-oxide closed-type membrane. The process was performed on 150 mm SOIwafers with a 0.25 mm-thick silicon device layer sitting on a 1 mm-thick boxoxide layer used as etch stop during the DRIE of silicon. The silicon devicelayer is very thin and can be removed from the whole membrane area forbetter thermal insulation. One of the tungsten metal layers was used as heaterand exhibited very stable behavior at a high temperature of 500 �C. Anultralow power consumption of 12 mW and a fast transient time of 2 msto reach 600 �C were reported. Fig. 13.16 presents a diagram of this device.The complete integration of the CMOS electronic circuitry with the sensorelement is still to be demonstrated.


13.6 Micromachined field-effect gas sensors

The field-effect gas sensing principle was first demonstrated by Prof.Lundstr€om in 1975 by replacing the standard aluminum gate of a MOSFETwith a catalytic metal, such as palladium, for the detection of hydrogen.68 Byheating up the device, hydrogen molecules dissociate in hydrogen atoms,which diffuse through the catalytic metal, reaching the metaledielectric inter-face of the FET devices. Electric dipoles are created, which induce a changein the IeV curve characteristics of the FET device. By tuning the catalyticgate material of the device, a series of gases (mainly containing hydrogenatoms) can be sensed using the FET as a transducer.69 Extensive literaturecan be found on the topic and AppliedSensor GmbH is now commercializingthe technology mainly for application in the fuel cell market. Modulating thetemperature is also of interest for this sensing principle, and some work hasbeen undertaken in that direction. However, low power and low thermalmass devices are desirable for this purpose.70 These devices have also beendeveloped on silicon carbide for applications in harsh environment.71

At the end of the 1990s, in the framework of the European projectChemical Imaging for Automotive Applications (CIA), reducing the powerconsumption of GasFETs was identified as being of interest to the automo-tive market. Developments have been undertaken by Briand et al.72 toachieve the thermal insulation of a GasFETs array based on the microhot-plate concept. At that time, the technology was further developed for itsintegration into an electronic nose by Nordic Sensors Technologies,Sweden (now AppliedSensor).

Tungstenplateheater

Metal heatspreading

Gassensingmaterial

PassivationSilicon heatspreadingplate

Silicondioxide

Buried silicondioxide

Substrate

P P PN NN

PMOS NMOS

CMOSSensor

Figure 13.16 Design of tungsten SOI chip: gas sensor and integrated CMOS circuitry.From Ali SZ, Udrea F, Milne WI, Gardner JW. Tungsten based SOI microhotplates forsmart gas sensors. J Microelectromech S 2008;17(6):1408e1417.


Basically, using silicon micromachining, an array of four GasFETs devices,with different catalytic layers (Pd, Ir, Pt), were located on a silicon islandthermally insulated from the silicon chip frame by a thin-film dielectricmembrane made of silicon nitride;73 Fig. 13.17. A two-step wet silicon aniso-tropic etching in KOH was developed to achieve a 10 mm-thick silicon plugunderneath the dielectric membrane, in which the electrical componentswere located. A doped silicon resistor used as heater and a diode used as atemperature sensor were integrated into the design, as shown in Fig. 13.17.

Dielectric membrane (0.5 μm thick) LPCVD + PECVD nitride 1.8 mm

(a)

(b) Heater

MOSFET

MOSFET

MOSFET

Diode

Ground

Al

400 μm 900 μm 10 μm

SiSi

MOS MOS

MOS MOSFET D EDOID

D

Heater

Figure 13.17 (a) Diagram of cross-sectional view of the low-power MOSFET array gassensor. The electronic components are located in a silicon island insulated from the sil-icon chip frame by a dielectric membrane made of two nitride layers; (b) Optical imageof the low-power micromachined MOSFET gas sensor showing the electronic devices inthe silicon island suspended by a dielectric membrane. From Briand D, van der Schoot B,de Rooij NF, Sundgren H, Lundstr€om I. A low-power micromachined MOSFET gas sensor. JMicroelectromech S 2000b;9(3):303e308


Processing, however, remained heavy, with many photolithographicsteps. Power consumption was significantly reduced to 90 mW for an oper-ational temperature of 170 �C. But the most interesting feature was the fastmodulation of the temperature. A thermal time constant of less than 100 mscould be reached with sensing devices configured in this way. Modificationsof the kinetics of the gas reactions with the sensing film occurred whenmodulating the temperature. They depended on the sensor “history,” onthe nature of the gaseous atmosphere, and on the type of materials used asthe catalytic film. Reduction of the recovery time of the device wasachieved by performing a temperature pulse following the gas exposure,and the discrimination of gases in a mixture using temperature cycling(100e200 �C) was especially valuable, with an effective resolution at a tem-perature modulation of “low” frequency (0.1 Hz) and large amplitude.24,25

The data were Fourier transformed before the evaluation was made usingprincipal components analysis plots. Discrimination was shown for gaseousmixtures of hydrogen and ammonia (10e100 ppm) in air (Fig. 13.18).

2.04

2.03

2.02

2.01

2 0 50 100Time (s)

Sen

sor s

igna

l (V

)

150 200

1.965

1.96

1.955

1.95

1.945 0 50 100Time (s)

Sen

sor s

igna

l (V

)

150 200

1.75

1.74

1.73

1.72

0 50 100Time (s)

Sen

sor s

igna

l (V

)

150 200

1.73

1.72

1.71

1.7

0 50 100Time (s)

Sen

sor s

igna

l (V

)

150 200

(b)

(d)

(a)

(c)

Figure 13.18 Ir-MOSFET voltage (continuous line) at a temperature variation between150 and 200 �C at 0.01 Hz (dashed line) in the presence of (a) pure synthetic air; (b)100 ppm H2; (c) 10 ppm NH3; (d) 10 ppm NH3; and 100 ppm H2. From Briand D, Tom-assone G-M, de Rooij NF. Accelerated ageing of micro-hotplates for gas sensing applications.Proceedings of IEEE sensors 2003 conference. (Toronto, Canada); 2003a. p. 1314e1317;Briand D, Wingbrant H, Sundgren H, van der Schoot B, Ekedahl L-G, Lundstr€om I, de Rooij, N.F. Modulated operating temperature for MOSFET gas sensors: hydrogen recovery timereduction and gas discrimination. Sens Actuators B 2003b;93:276e285.


13.7 Nanostructured gas sensing layers onmicrohotplates

Nanowires are seen as a solution with which to improve the sensitivity,selectivity, stability, and response time of metal oxide gas sensors. Meieret al.74 grew SnO2 nanowires of 100 nm in diameter by the vaporesolidgrowth method. For testing, they were deposited onto micromachined hot-plates and contacted with a focused ion beam scanning electron microscope(FIB-SEM), as shown in Fig. 13.19. Because of their diameter being similarto the Debye length, a completely depleted conduction channel can beobtained. Maximum response to CO and NH3 occurred at about 260 �C.SnO2 nanoparticles can also be grown by solegel method. Li et al.75 achievedSnO2 nanomaterial by precipitating SnCl4$5H2O from an aqueous solution.

100 μm

Fourelectrodes

(a)

(b)

SnO2nanowire

500 nm

Pt FIBcontacts

Etc

hed

pit

Pol

ysili

con

heat

er

NW

Figure 13.19 (a) Optical image of microhotplates with nanowires, NW. (b) SEM imagesof SnO2 nanowire on a microhotplate. From Meier DC, Semancik S, Button B, Strelcov E,Kolmakov A. Coupling nanowire chemiresistors with MEMS microhotplate gas sensingplatforms. Appl Phys Lett 2007;91:063118.


The obtained powder could then be doped by adding TiO2 or carbon nano-tubes. The nanopowders were deposited on microhotplate membrane. Adroplet of deionized water was first drop coated on the membrane. It wasfollowed by scattering SnO2-based powder on the substrate. The powderwas mixed with water to obtain a paste, which was later dried. Sensing filmsof approximately 200 mm in thickness were produced. The sensors weretested against ethanol at 300 �C. They showed, however, a poor selectivitytoward methanol, acetone, formaldehyde, NH3, and toluene. Similarly,76 ob-tained Au-doped SnO2 nanocomposites. They first precipitated SnCl2$2H2Oto get SnO powder. It was then mixed to HAuCl4 to obtain Au nanoparticlesattached to the surface of SnO2 mixture of nanoparticles and nanowires. Thelatter was maskless deposited by DPN (dip-pen nanolithography), whichallowed confining the sensing material to the electrode area of a commercialmicrohotplate. Concentrations of ethanol between 100 and 1000 ppm couldbe detected at 400 �C. The sensor revealed, however, to be sensitive tohumidity and showed fair selectivity toward toluene and acetone.

Materials other than SnO2 also exhibited good gas sensing performances.Ryu et al.77 fabricated In2O3 nanowires by a laser ablation method. Thenanowires were then sonicated in isopropanol to obtain a suspension, whichwas deposited onto microhotplates. When operating at 275 �C, responses(R/R0) of 1.6e50 ppm of ethanol, of 2e100 ppm of CO, and of0.5e50 ppm of H2 were measured. In addition, the micromachined gassensor exhibited a short gas response time of about 22 s.

Vapor phase growth is a technique that can be used for producing ratherhigh quantity of nanomaterials. Marasso et al.78 used it to form ZnO nano-tetrapods from a metallic Zn seed. The ZnO nanostructures were dispersedin a solvent before their precipitation on the membrane of a hotplate bycentrifugation. The deposited structure exhibited a good adhesion to thesubstrate avoiding any firing process. The obtained sensors revealed amaximum response to ethanol and methane at 400 �C and to H2S andNO2 at 300 �C.

An alternative method for growing nanotubes is through hydrothermalprocess. It involves crystallizing material from an aqueous solution attemperature typically between 80 and 90 �C. Such method was used byShao et al.79 to obtain ZnO nanowires from a ZnO seed layer. Their diam-eters were between 50 and 300 nm for a length of about 6 mm. They werethen drop coated onto a commercial microhotplate. An AC signal wasapplied between electrodes to align the nanowires. A subsequent annealingwas performed at 400 �C. They showed good response to NH3 when


heated at 350 �C. Lee et al.80 obtained ZnO nanowires on a microhotplatethrough a lift-off process. A photoresist mask was patterned and the substratewas immersed in an aqueous solution for hydrothermally growing the nano-wires. Once the process was completed, the photoresist was stripped. Chenet al.81 grew ZnO nanowires in situ on the electrodes of a microhotplate.Zinc acetate was first drop coated onto the electrodes. After drying, a seedfilm of zinc acetate crystallites was formed. It was followed by hydrothermalprocess to grow grass-like nanowires. They could be then used as seed layerfor a second hydrothermal process to obtain branch structures onto them.These nanostructures showed a very good sensitivity toward H2S whenheated at 300 �C with a limit of detection of 3 ppb.

Other materials can be grown by hydrothermal processes. For instance,82

obtained hexagonal WO3 nanorods of 80e150 nm in diameter and4e5 mm in length from sodium tungstate. The obtained nanowires couldbe decorated with Au or Pt nanoparticles. Au-doped nanowires had anenhanced sensitivity toward H2S with a concentration detection as low as5 ppb. Doping additionally reduced response time to 1 ppm of H2Scompared with undoped WO3 wires from 300 s down to 30e40 s. Inkjetprinting can be used to pattern hydrothermally grown nanowires, thusavoiding shadow-masking or photoresist patterning. Krainer et al.83 depos-ited a suspension of WO3 nanowires with a commercially available inkjetprinter on microhotplate membrane. Once deposited, the deposited drop-lets were annealed at 400 �C for 12 h. Sub-ppm concentrations of H2Scould be detected at 250 �C independently of the relative humidity level.

Nanotubes can also be grown by CVD processes. Recently84, showedthat AACVD (Aerosol-Assisted Chemical Vapor Deposition) techniquewas suitable for growing WO3 nanoneedles. The nanoneedles could befunctionalized with Au and/or Pt nanoparticles. The method involvestemperatures between 350 and 600 �C, which are compatible withMEMS-based devices. The patterning is typically made through a shadowmask. The fabricated sensors showed good discrimination between ethanol,hydrogen, and CO when heating between 100 and 300 �C. These gases areof particular relevance in proton-exchange fuel cells. AACVD was alsoreported to be used for growing Cu2O-decorated WO3 nanoneedles byAnnanouch et al.85 in one-step process on microhotplate. The resultingsensor showed a response of 27.5 to 5 ppm of H2S when heated at390 �C with a limit of detection of approximately 300 ppb. In addition,the device exhibited a selectivity against H2, CO, NH3, C6H6, and NO2.The same author reported later PdO nanoparticle-decorated WO3


nanoneedle with a two-step AACVD process.86 Their integration on amicrohotplate is illustrated in Fig. 13.20. It aimed at H2 detection in renew-able energy source. Exposure to 500 ppm of H2 led to a sensor response of1670 when heated at 150 �C. The sensor response was defined as the ratio ofthe sensor resistance in air to the analyte of interest for reducing gases and theopposite for oxidizing gases. The response decreased above that temperatureand provided unreproducible results. Additionally, the sensor had a goodselectivity against NH3, C6H6, and CO.

Nanowires can be grown directly from a substrate. For instance,87 grewCuO nanowires directly from 600 nm-thick Cu structures placed on theelectrodes of a microhotplate. The latter was heated at approximately335 �C using its buried heater. Growth occurred in a gas test chamberwith synthetic air. This process resulted in 1 mm long nanowires with adiameter of about 20 nm. As the sensors were mounted on PCB, nanowiregrowth could be electrically monitored as well as CO sensing capabilities.Because gas measurement occurred in the very same chamber, the sensors

Figure 13.20 WO3 film morphology on a micromachined hotplate observed by SEMimages at low (a and b) and high (c) magnification. (d) Cross section of WO3 nanonee-dles. Reprinted with permission from Annanouch FE, Haddi Z, Ling M, Di Maggio F, VallejosS, Vilic T, Zhu Y, Shujah T, Umek P, Bittencourt C, Blackman C, Llobet E. Aerosol-asssitedCVD-grown PdO nanoparticle-decorated tungsten oxide nanoneedles extremely sensitiveand selective to hydrogen. ACS Appl Mater Interfaces 2016;8:10413e10421. Copyright(2016) American Chemical Society.


could be assessed without being exposed to ambient environment. Theyshowed responses (R/R0*100) of 6.4% and 27.6% to CO concentrationsof, respectively, 1 and 30 ppm when operating at 325 �C. The sensorperformances dropped once exposed to humid environment because ofhydroxylation of the CuO surfaces.

A main issue toward reducing the power consumption of metal oxide gassensors is their operating temperature, which is reduced in some cases byusing nanostructures. Previous examples used microhotplates to reach theoptimum thermal operating conditions. In an alternative move,88 addressedthis problem by directly using the probing current applied to the nanowiresas the heat source. This significantly simplified the device by avoiding theneed for the integration of a heater into the hotplate. Moreover, it reducedthe heated area and, consequently, power consumption. Currents in therange of 0.1e300 nA were flowing through an SnO2 nanowire to heat itup to 300 �C. The measured power consumption was 30 mW, two to threeorders of magnitude lower than “standard” micromachined metal oxide gassensors, making them compatible with energy harvesting systems. Very fastsensors were obtained with response times in the millisecond range. Theyhad a good response to CO and NO2. Si nanowires were used as gas sensorsby89. The devices could operate at room temperature, drastically reducingtheir power consumption. They could be thus transferred on polyethyleneterephthalate (PET) plastic foil as substrate (Fig. 13.21). A response of about

50 μm

Figure 13.21 SEM image of an array of SNAP nanowire sensors. Each device (horizontalstrip) is contacted by two Ti electrodes (oriented vertically) that extend to larger pads(top and bottom image edges). Inset: digital photograph of the flexible sensor chip.From McAlpine MC, Ahmad H, Wang D, Heath JR. Highly ordered nanowire arrays onplastic substrates for ultrasensitive flexible chemical sensors. Nat Mater 2007;6:379e384.


2 was obtained under an exposure of 2 ppm of NO2. The detection of NO2

concentrations as low as 20 ppb was possible. The device response time wasup to few minutes, depending on the gas concentration. Purge cycles withvacuum and fresh air were necessary for the sensor to recover after an expo-sure to NO2. The nanowires could be functionalized with alkane-, alde-hyde-, and amino-silane to improve selectivity and allow differentiationof a binary mixture of acetone and hexane.

The fabrication of nanowires has been mastered and they have shown tobe suitable for gas sensing. However, several issues remain for their large-scaleuse in commercial devices and for the achievement of reproducible results. Itmainly concerns the precise location of the nanowires on a specific area andtheir electrical contact. From an operation point of view, to benefit from theirlow operational temperature for gas detection, sensitivity to humidity andslower desorption kinetics will need to be addressed in some cases.

13.8 Semiconductor gas sensors on polymeric foiland their additive manufacturing

13.8.1 Semiconductor gas sensors on polymeric foilThe use of plastic substrates, since 2008, has been seen as a solution to

further decreasing sensor cost and manufacturing complexity, comparedwith devices manufactured on silicon or ceramic substrates. Plastic addition-ally shows other benefits, such as compatibility with large-scale fabrication(roll-to-roll), printing compatibility, lightweight, and conformality. Suchdevices aim at new applications where low cost is a prerequisite: smartsensing labels, wearable devices, consumer goods, distributed systems, andso on. However, metallic oxide films are usually annealed at high tempera-ture, and the main challenge of processing them and operating them on plas-tic substrates is the limited thermal budget. Nanowires, the FSP depositiontechnique, and low sintering temperature nanoparticle inks are potentialcandidates for integration at a relatively low temperature onto polymerictransducing platforms of performing metal oxide materials.

Briand et al.90 were the first to demonstrate the use of polyimide (PI) as asubstrate for the fabrication of plastic-based metal oxide gas sensors. Twotypes of devices were fabricated by standard microfabrication equipment.The first solution consisted in using silicon as the substrate, which wasspin coated with a PI layer. Once the bulk silicon was dry etched, a PI mem-brane embedding a Pt-based heater and with electrodes on top was released.The second solution was based on the use of a commercially available PI foil


as the substrate. A Pt heater was patterned and covered with a photosensitivespin-coatable PI layer used as a dielectric film to electrically insulate the elec-trodes on top. In both configurations, the interdigitated electrodes weredrop coated with a Pd-doped SnO2 layer as the gas sensing film with amaximum annealing temperature at 450 �C. These devices showed goodgas sensing performances but suffered from excessive power consumptionwhen operating at 325 �C: 82 mW for devices on silicon and 130 mWfor the device on PI foil. To reduce power consumption,9 investigatedthe miniaturization of drop-coated metal oxide gas sensors on PI foil. Theirtransducers were optimized in terms of power consumption and tempera-ture uniformity through electrothermal simulations. Devices from 100 mmdown to 15 mmwere produced. With the idea of reducing power consump-tion further, the PI foil could be dry etched in an O2/CF4 plasma to obtainclosed and suspended membranes about 3 mm thick. The deposition of themetal oxide layer (Pd-doped SnO2) was carried out with micropipettes.44

The smallest droplet had a diameter of 20 mm (Fig. 13.22(a)). A powerconsumption as low as 6 mW was required to reach 300 �C with a15 mm-wide heater with a closed membrane in a continuous operatingmode.With a simplified fabrication process avoiding the bulk micromachin-ing of the PI foil, only 10 mW was necessary with a heater of the same size.These sensors could operate for more than 1 year at 200 �C.91 The sensorsworked in both continuous and pulsed modes, which decrease the powerconsumption to the sub-mW level. The devices showed to be effectivefor the detection of CO (Fig. 13.22(b)), CH4, and NO2. Furthermore, amethod for the encapsulation of chemical sensors at foil level was demon-strated.92 It consisted in a prepatterned rim made of a dry photoresist filmlaminated onto the PI substrate containing the gas sensors. They werecovered with a water-repellent gas permeable membrane.

ZnO nanowires were grown on PI-based microhotplates by93. Zn wassputtered onto the substrate through a shadow mask and then oxidizedfor 12 h at 300 �C. Such a relatively low temperature was required to avoiddamaging the plastic foil. The ZnO nanotubes showed a response towardNO2. PET foils were used by McAlpine et al.89 as the substrate onto whichnanotubes were deposited (see Section 6.7 for more information). Thedevice showed itself to be suitable for measuring NO at room temperature.

The operation of metal oxide gas sensors on plastic foil was successfullydemonstrated. However, to make them fully compatible with large-scalefabrication techniques, i.e., printing, additional work is required. This topicis addressed in the next section.


13.8.2 Printing semiconductor gas sensorsRecently, since 2010, with the emergence of printing techniques, newdeposition methods compatible with large area manufacturing have beenapplied to gas sensing materials. Inkjet-printed pure and doped SnO2 wasperformed on silicon and alumina substrates.94 The use of inkjet printingfacilitated doping by the consecutive printing of SnO2 and a dopant. Apure SnO2-based sensor exhibited a response of about 7e50 ppm of ethanoland 55 when exposed to 50 ppm of H2S when operating, respectively, at425 and 179 �C. However, their printed layers required annealing at

6(b)

(a)

5

4

3

2

120 30 40 50

CO concentration (ppm)

Che

mre

sist

ive

resp

onse

(—)

100 μm - 28 mW50 μm - 18 mW25 μm - 11 mW15 μm - 10 mW

60 70 80 90 100

Gas sensinglayer

Heaterconnections

50 μm

20 μmElectrodeconnections

Figure 13.22 (a) Optical image (top view) of metal oxide gas sensor on PI; (b) gasresponse to CO for several sensor sizes when operating at 250 �C. Adapted fromCourbat J, Briand D, Yue L, Raible S, de Rooij NF. Drop-coated metal-oxide gas sensor onpolyimide foil with reduced power consumption for wireless applications. Sens Actuators B2012;161:862e868.


550 �C, making them incompatible with plastic substrates. This drawbackwas counteracted by Peter et al.95, who developed a titanium-dopedchromium oxide ink that did not require any firing. The adhesion to the sil-icon substrate and the film stability was improved by sintering the printedlayer at 400 �C. This temperature is, however, compatible with a highperformance polymer such as some PIs. Moreover, being an additive tech-nique, inkjet printing is of significant interest with regard to the localpatterning of different sensing films on one substrate. In the case of arrays,all sensing material can be deposited simultaneously, simplifying fabricationof the device.

Kukkola et al.96 used another technique compatible with roll-to-roll pro-cessing: gravure printing. They depositedWO3 sensing films on interdigitatedelectrodes patterned on Kapton HN PI foil from DuPont. However, thefabrication of an integrated heating element was not addressed in this study.For gas response measurement, the sensor was placed in a heated gas cell at200 �C. A gas response was obtained for a concentration of 5 ppm of NO.

A coplanar architecture was reported by Ramírez et al.97 in 2018 toimplement in one single layer the electrodes and the heating element ofprinted gas sensors. The design includes two electrodes and three contacts.One of the electrodes works as heating element and, simultaneously, drainsthe sensing current. Compared with other coplanar topologies, thisapproach simplifies the transducers processing to a single printing step,avoiding the use of an interdielectric layer between heater and electrodes.This cost-effective architecture and process was applied to the fabricationof heated transducers for metal oxide gas sensors. The two electrodeswere made by inkjet printing of gold on PI foil. For the validation of theconcept, a Pt-loaded WO3 sensing layer was grown on top of these trans-ducers printed with the proposed topology. This simple architecture hasstrong potential for the realization of fully printed resistive gas sensors andcan be implemented as well in cleanroom processed transducers.

The first fully inkjet-printed tin dioxide (SnO2) gas sensor was reportedby Rieu et al.98 in 2016. Gold electrodes and heater were inkjet printed oneach side of a PI substrate. A SnO2-based solegel ink was inkjetted onto theelectrodes. A final annealing at 400 �C allowed to synthetize the SnO2

sensing film. The device was operated at temperatures between 200 and300 �C using the integrated heater. The proper operation of the fully printedmetal oxide gas sensors was validated under exposure to carbon monoxideand nitrogen dioxide, in dry and wet air. In 2018, Khan et al. have reportedon a low-power metal oxide gas sensor using aerosol jet printing to reduce


the area of the hotplate transducer to 500 � 500 mm2. Aerosol jet was usedto print the gold heater and electrodes and the interdielectric layer made ofPI. The transducer consumes 78 mW at an operating temperature of200 �C. Inkjet printing was used to coat the transducers electrodes withPd-doped tin dioxide nanoparticles.

13.9 Manufacturing, products, and applications

Large volume manufacturing of semiconductor gas sensors has begunin the early 2000s with the company MiCS (MicroChemical Systems SA) inSwitzerland, now part of the SGX Sensortech group, and AppliedSensorGmbH in Germany, bought by AMS AG in Austria, both addressing theautomotive industry with metal oxide sensors for air quality monitoring.43

In the 2010s, micromachined metal oxide sensors targeting the air indoorquality monitoring market have been also developed. Other companiessuch as Figaro Engineering in Japan, the pioneer in the field of metal oxide,start-up Cambridge CMOS Sensor (CCS) in United Kingdom, SensirionAG in Switzerland, and large companies Bosch Sensortec in Germany andAMS in Austria are now proposing MEMS-based metal oxide sensor prod-ucts. AMS has acquired AppliedSensor GmbH and Cambride CMOSSensor to increase its technology portfolio. AMS, Bosch, and Sensirionare proposing environmental sensing solutions made of a variety of sensors,combining metal oxide sensors with temperature, humidity, pressure, opti-cal CO2 sensors, and particle sensors, among others.

Figaro Engineering Inc. investigated the potential commercialization ofmicromachined metal oxide gas sensors.99 The device is based on a sus-pended membrane etched from the front for minimizing the powerconsumption. They dispensed different metal oxide materials that wereannealed with the integrated heater on the chip. The layer thicknesseswere between below 1 mm to about 50 mm, depending on the gas tobe detected. This research and development work has led to a new product,the TGS8100, for the detection of air contaminants, such as hydrogen(1e30 ppm) and ethanol, for air quality and appliance control. Thesensor comes in a surface mount package with a footprint of2.5 � 3.2 � 0.99 mm3.102 It consumes 15 mW with an applied heatervoltage of 1.8V and circuit voltage of 3.0V DC pulse. It exhibits high sensi-tivity to cigarette smoke, cooking odors, and gaseous air contaminants withapplication examples such as indoor air quality monitors, air cleaners, venti-lation control, and kitchen range hood control.


Bosch Sensortec BME680 Integrated Environmental Unit is an envi-ronmental sensor for mobile applications and wearables. BME680 com-bines a metal oxide gas sensor for VOCs monitoring with air pressure,humidity, and ambient air temperature sensing functions within a singlepackage. The combo MEMS solution enables multiple new capabilitiesfor portable and mobile devices such as air quality measurement, homeautomation, and other applications for the Internet of Things (IoT). Thesensor comes in a 3.0 � 3.0 mm2 footprint package with I2C and SPIcommunication interfaces. Applications include smart homes, smart officesand buildings, smart energy, smart transportation, HVAC, elderly care, andsport/fitness.

More and more devices in our surroundings are being equipped withsensors to monitor environmental parameters such as air pollution. In partic-ular, mobile platforms such as wearables and mobile phones offer newopportunities for sensing applications. Such a combination enables forexample monitoring of personal exposure to outdoor or indoor air pollut-ants such as NOx or volatile organic compounds that affect our healthand well-being. These new applications pose a number of requirementsfor gas sensing technologies such as high sensitivity, good long-term stability,low power consumption, small package size, and low production costs.

Sensirion’s multipixel gas sensor SGP (Sensirion Gas sensor Platform,Fig. 13.23) combines three key innovations that are crucial for the wide-spread integration of MOX-based gas sensors in mobile and IoT applica-tions: long-term stability through siloxane resistance, a fully digital gasmeasurement solution monolithically integrated on one chip, and the inte-gration of several sensing elements in one sensor.100 The SGP offers a com-plete gas sensor system integrated into compact DFN package of

Figure 13.23 The SGP multipixel gas sensor. Courtesy of Sensirion AG.


2.45 � 2.45 � 0.9 mm3 size. Sensirion’s CMOSens technology allows tocointegrate analog and digital electronics together with a microhotplateand the sensing elements on a single chip as shown in the block diagramin Fig. 13.24. Four MOX sensing elements based on layers of metal oxidenanoparticles are deposited on a microhotplate (Fig. 13.25). The resistanceof each sensing element can be measured separately by readout electrodes.A heater and a temperature sensor are also integrated on the hotplate toactively control its operating temperature. This guarantees a stable opera-tion of the sensor, independent of ambient temperature. The signals from

Mems

HotplateT-sensor

Heatercontroller

Analog Digital

Signal processing

Systemcontroller

On-chipmemory

Analog frontend

VDDH

VDD

VSS

SDA

SCL

I 2C interface

Figure 13.24 Block diagram of the SGP multipixel gas sensor platform. Courtesy ofSensirion AG.

Read-out electrode

Heater

Sensor element: «pixel»

Figure 13.25 Micrograph of the SGP showing the four sensing elements, the readoutelectrodes, and the heater. Courtesy of Sensirion AG.


the four sensor elements are measured by a highly optimized amplifiercovering a measurement range of eight orders of magnitude. This is crucialfor covering a wide variety of metal oxide sensing materials as well asdifferent gases and gas concentrations with a single hardware platform.The signals are further processed in the digital signal processing stagewith algorithms, e.g., for averaging, baseline compensation, and humiditycompensation. In addition, individual calibration parameters are writtenduring production into an on-chip memory. This allows to convert thesensor raw signals into calibrated output signals, for example concentrationsof volatile organic compounds. All these features greatly simplify the inte-gration of the SGP into different applications. The output signal candirectly be used by customers as air quality indication without furtherprocessing.

The combination of several MOX sensing elements on one chip bringstwo important advantages. First, it allows for measuring gas concentrationsof several gases such as outdoor air pollutants and VOCs with one sensor.This greatly reduces cost and footprint in comparison with solutions usingseveral sensor chips. Second, the combination of signals from differentsensing elements can also be used to improve the selectivity with respectto the target gas. Traditional metal oxideebased gas sensors suffer frompoor long-term stability when they are operated in atmospheres containingeven very low concentrations of siloxanes, which are silicon-containingcompounds found in many products of our everyday life such as cosmetics,cleaning agents, or plastic parts. The degradation caused by siloxanes typi-cally results in a significant loss of sensitivity to VOCs and other gases aswell as in a strong increase of response time.101 The degradation processand therefore the sensor life time depends on the siloxane concentration.This problem is in particular pronounced in applications like mobile phones,where the sensor is constantly exposed to high siloxane concentrationsdegassing from various components of the mobile phone. The core technol-ogy of the SGPdMOXSensdprovides the sensor with a unique robustnessagainst contamination by siloxanes. This is achieved by a combination ofoptimization of the sensing material, operation mode, and the combinationof signals from different sensing elements. The siloxane resistance signifi-cantly improves the long-term stability and accuracy of the SGP. TheSGP offers a unique combination of integration, multipixel platform, andlong-term stability that not only leverages MOX-based gas sensing into anew area but also opens up completely new gas sensing applications likemobile phones, wearables, and IoT devices.


With respect to MEMS-based MOX gas sensors, the recent years haveshown a clear trend toward utilizing sensors in the consumer space. Thishas led to further cost and power reduction via miniaturization and moreadvanced, low-cost packaging solutions, e.g., mold packages. The smallersubstrate sizes gave rise to challenges with respect to deposition processesand choice of MEMS processes. The latter are nowadays more and moretransferred to standard CMOS foundries and materials such as tungstenare replacing noble metals in hotplate structures. Eventually, this trendmay lead to 3D-integrated or monolithic devices.

A major challenge for MOX gas sensors production remains the device-to-device variation which is aggravated by the shrinking device sizes, result-ing in the need to have very stable processes for both MEMS wafermanufacturing and MOX deposition for high volume production.

13.10 Conclusion

Micromachined semiconductor gas sensors based on silicon microhot-plate technology is now a mature technology with a few examples of deviceson the market, mainly based on thick-film metal oxides (notably SnO2 andWO3). Since the end of the 1980s, the technology has evolved significantlyand offers very good models for their design and robust processes for theirfabrication. Various efforts have led to devices that perform very well at oper-ational temperatures above 500 �C, with homogeneous temperature distribu-tion over the sensing area and minimum power consumption. Powerconsumption for continuous operation is in the order of a few mW, andsub-mW consumption can be reached using a pulsing mode of operation.These platforms can now welcome many different types of semiconductinggas sensing materials, with various formations of device array, with the veryinteresting possibility of modulating the operational temperature and inte-grating the electronics with the sensor silicon chip. The concept of microhot-plates has been extended to field-effect gas sensors also with reduced powerconsumption and thermal cycling capabilities. Trends and perspectives aremainly in relation to nanotechnology-based devices, with the integration ofnanostructured gas sensing films on conventional microhotplates and espe-cially on polymeric-based microhotplates. New processing methods are alsobeing investigated for the integration of metal oxide sensing layers ontomicrohotplate devices, such as FSP, nanowire synthesis, and the printing ofmetal oxide sensing layers, mainly using inkjet. Finally, fully printed versionof metal oxide gas sensors has been demonstrated on large-area polymeric foil.


References1. Mardilovich P, Routkevitch D, Govyadinov A. Hybrid micromachining and surface

microstructuring of alumina surface. In: Conf. Proc. of microfabricated systems and MEMSV, 198th ECS meeting; 2000. p. 33e42.

2. Vasiliev A, Pavelko RG, Gogish-Klushin SY, Kharitonov DY, Gogish-Klushina OS,Sokolov AV, Pisliakov AV, Samotaev NN. Alumina MEMS platform for impulsesemiconductor and IR optic gas sensors. Sens Actuators B 2008;132(1):216e23.

3. Simon I, Barsan N, Bauer M, Weimar U. Micromachined metal oxide gas sensors: op-portunities to improve sensor performance. Sens Actuators B 2001;73:1e26.

4. Brattain WH, Bardeen J. Surface properties of germanium. Bell Syst Tech J 1952;32:1.5. Seiyama T, Kato A, Fujushi K, Nagatani M. A new detector for gaseous components

using semiconductive thin films. Anal Chem 1962;34:1502e3.6. Briand D, Krauss A, van der Schoot, Weimar U, Barsan N, G€opel W, de Rooij NF.

Design and fabrication of high-temperature micro-hotplate for drop-coated gassensors. Sens Actuators B 2000a;68:223e33.

7. Ho GW. Gas sensors with nanostructures oxide semiconductor materials. Sci Adv Mat2011;3(2):150e68.

8. Ali SZ, Udrea F, Milne WI, Gardner JW. Tungsten based SOI microhotplates forsmart gas sensors. J Microelectromech S 2008;17(6):1408e17.

9. Courbat J, Canonica M, Teyssieux D, Briand D, de Rooij NF. Design and fabricationof micro-hotplates made on a polyimide foil: electrothermal simulation and character-ization to achieve power consumption in the low mW range. J Micromech Microeng2010a;21:015014.

10. Elmi I, Zampolli S, Cardinali GC. Optimization of a wafer-level process for the fabri-cation of highly reproducible thin-film MOX sensors. Sens Actuators B 2008a;131:548e55.

11. Faglia G, Comini E, Cristalli A, Sberveglieri G, Dori L. Very low power consumptionmicromachined CO sensors. Sens Actuators B 1999;55:140e6.

12. Graf M, Barrettino D, Zimmermann M, Hierlemann A, Baltes H, Hahn S, Barsan N,Weimar U. CMOS monolithic metal-oxide sensor system comprising a microhotplateand associated circuitry. IEEE Sens J 2004a;4(1):9e16.

13. Cakir MC, Caliskan D, B€ut€un B, }Ozbay E. Planar indium tin oxide heater forimproved thermal distribution for metal oxide micromachined gas sensor. Sensors2016;16:1612.

14. Presmanes L, Thimont Y, el Younsi I, Chapelle A, Blanc F, Talhi C, Bonningue C,Barnabé A, Menini P, Tailhades P. Integration of P-CuO thin sputtered layers ontomicrosensor platforms for gas sensing. Sensors 2017;17:1409.

15. Amor S, André N, Gérard P, Ali SZ, Udrea F, Tounsi F, Mezghani B, Francis LA,Flandre D. Reliable characteristics and stabilization of on-membrane SOIMOSFET-based components heated up to 335�C. Semicond Sci Technol 2017;32:014001.

16. Cavicchi RE, Suehle JS, Kreider KG, Gaitan M, Chaparala P. Fast temperature pro-grammed sensing for micro-hotplate gas sensors. IEEE Electron Device Lett 1995;16(6):286e8.

17. Graf M, Gurlo A, Barsan N, Weimar U, Hierlemann A. Microfabricated gas sensorsystems with sensitive nanocrystalline metal-oxide films. J Nano Res 2006a;8(6):823e39.

18. Barsan N, Schweizer-Berberich M, G€opel W. Fundamental and practical aspects in thedesign of nanoscaled SnO2 gas sensors: a status report. Fresnius J Anal Chem 1999;365(4):287e304.


19. Puigcorbé J, Vogel D, Michel B, Vil�a A, Gr�acia I, Cané C, Morante JR. Thermal andmechanical analysis of micromachined gas sensors. J Micromech Microeng 2003;14:548e56.

20. Rao LLR, Singha MK, Subramaniam KM, Jampana N, Asokan S. Molybdenummicroheaters for MEMS-based gas sensor applications: fabrication, electro-thermo-mechanical and response characterization. IEEE Sens J 2017;17(1):22e9.

21. Demarne V, Grisel A. An integrated low-power thin-film CO gas sensor on silicon.Sens Actuators 1988;13:301e13.

22. Iwata T, Soo WPC, Matsuda K, Takahashi K, Ishida M, Sawada K. Design, fabrica-tion, and characterization of bridge-type micro-hotplates with an SU-8 supportinglayer for a smart gas sensing system. J Micromech Microeng 2017;27:024003.

23. Bosc J-M, Guo Y, Sarihan V, Lee T. Accelerated life testing for micromachined chem-ical sensors. IEEE Trans Reliab 1998;47(2):135e41.

24. Briand D, Tomassone G-M, de Rooij NF. Accelerated ageing of micro-hotplates forgas sensing applications. In: Proceedings of IEEE sensors 2003 conference, Toronto, Canada;2003. p. 1314e7.

25. Briand D, Wingbrant H, Sundgren H, van der Schoot B, Ekedahl L-G, Lundstr€om I,de Rooij NF. Modulated operating temperature for MOSFET gas sensors: hydrogenrecovery time reduction and gas discrimination. Sens Actuators B 2003b;93:276e85.

26. Briand D, Beaudoin F, Courbat J, de Rooij NF, Desplats R, Perdu P. ‘Failure analysisof micro-heating elements suspended on thin membranes ‘. Microelectron Reliab 2005;45:1786e9.

27. Lahlalia A, Filipovic L, Selberherr S. Modeling ans simulation of novel semiconductingmetal oxide gas sensors for wearable devices. IEEE Sens J 2018;18(5):1960e70.

28. Ehmann M, Ruther P, von Arx M, Paul O. Operation and short-term drift ofpolysilicon-heated CMOS microstructures at temperatures up to 1200K. J MicromechMicroeng 2001;11(4):397e401.

29. Courbat J, Briand D, de Rooij NF. Reliability improvement of suspended platinum-based micro-heating elements. Sens Actuators A 2008;142:284e91.

30. Mele L, Santagata F, Iervolino E, Mihailovic M, Rossi T, Tran AH, Schellevis H,Creemer JF, Sarro PM. A molybdenum MEMS microhotplate for high-temperatureoperation. Sens Actuators A 2012;188:173e80.

31. Creemer JF, Briand D, Zandbergen HW, van der Vlist W, de Boer CR, de Rooij NF,Sarro PM. Microhotplates with TiN heaters. Sens Actuators A 2008;148:416e21.

32. Graf M, M€uller SK, Barrettino D, Hierlemann A. Transistor heater for microhotplate-based metal-oxide microsensors. IEEE Electron Device Lett 2005;26(5):295e7.

33. Lyle RP, Hughes HG, Walters D. Micromachined silicon CO gas sensors. ElectrochemSoc Proc 1997;97(5):188e98.

34. Semancik S, Cavicchi RE, Wheeler MC, Tiffany JE, Poirier GE, Walton RM,Suehle JS, Panchapakesan B, DeVoe DL. Microhotplate platforms for chemical sensorresearch. Sensor Actuator B 2001;77:579e91.

35. Elmi I, Zampolli S, Cozzani E, Mancarella F, Cardinali GC. Development of ultra-low-power consumption MOX sensors with ppb-level VOC detection capabilitiesfor emerging applications. Sens Actuators B 2008b;135:342e51.

36. M€uller G, Friedberger A, Kreisl P, Ahlers S, Schulz O, Becker T. A MEMS toolkit formetal-oxide-based gas sensing systems. Thin Solid Film 2003;436:34e45.

37. Friedberger A, Kreisl P, Rose E, M€uller G, K€uhner G, W€ollenstein J, B€ottner H.‘Micromechanical fabrication of robust low-power metal oxide gas sensors ‘. Sens Ac-tuators B 2003;93:345e9.

38. W€ollenstein J, Plaza JA, Cané C, Min Y, B€ottner H, Tuller HL. A novel single chipthin film metal oxide array. Sens Actuators B 2003;93:350e5.


39. Huotari J, Kekkonen V, Haapalainen T, Leidinger M, Sauerwald T, Puustinen J,Liimatainen J, Lappalainen J. Pulsed laser deposition of metal oxide nanostructuresfor highly sensitive gas sensor applications. Sens Actuators B 2016;236:978e87.

40. Gardner JW, Pike A, de Rooij NF, Koudelka-Hep M, Clerc PA, Hierlemann A,G€opel W. Integrated array sensor for detecting organic solvents. Sens Actuators B1995;26e7:135e9.

41. Heilig A, Barsan N, Weimar U, Schweizer-Berberich M, Gardner JW, G€opel W. Gasidentification by modulating temperatures of SnO2-based thick film sensors. Sens Ac-tuators B 1997;43:45e51.

42. Kappler J, Barsan N, Weimar U, Dieguez A, Alay JL, Romano-Rodriguez A,Morante JR, G€opel W. Correlation between XPS, Raman and TEM measurementsand the gas sensitivity of Pt and Pd doped SnO2 based gas sensors. Fresenius J AnalChem 1998;361(2):110e4.

43. Blaschke M, Tille T, Robertson P, Mair S, Weimar U, Ulmer H. MEMS gas-sensorarray for monitoring the perceived car-cabin air quality. IEEE Sens J 2006;6(5):1268e308.

44. Courbat J, Briand D, Yue L, Raible S, de Rooij NF. Drop-coated metal-oxide gassensor on polyimide foil with reduced power consumption for wireless applications.Sens Actuators B 2012;161:862e8.

45. Puigcorbé J, Vil�a A, Cerd�a J, Cirera A, Gr�acia I, Cané C, Morante JR. Thermo-mechanical analysis of micro-drop coated gas sensors. Sens Actuators A 2002;97e8:379e85.

46. Cerd�a Belmonte J, Puigcorbé J, Arbiol J, Vil�a A, Morante JR, Sabaté N, Gr�acia I,Cané C. High-temperature low-power performing micromachined suspendedmicro-hotplate for gas sensing applications. Sens Actuators B 2006;114:826e35.

47. Espinosa E, Ionescu R, Zampolli S, Elim I, Cardinali GC, Abad E, Leghrib R,Ramírez JL, Vilanova X, Llobet E. Drop-coated sensing layers on ultra low powerhotplates for an RFID flexible tag microlab. Sens Actuators B 2010;144:462e6.

48. Vincenzi D, Butturi MA, Guidi V, Carotta MC, Martinelli G, Guarnieri V, Brida S,Margesin B, Giacomozzi F, Zen M, Pignatel GU, Vasiliev AA, Pisliakov AV. Devel-opment of a low-power thick-film gas sensor deposited by screen-printing techniqueonto a micromachined hotplate. Sens Actuators B 2001;77:95e9.

49. Llobet E, Ivanov P, Vilanova X, Brezmes J, Hubalek J, Malysz K, Gr�acia I, Cané C,Correig X. Screen-printed nanoparticles tin oxide films for high-yield sensormicrosystems. Sens Actuators B 2003;96:94e104.

50. Ivanov P, Stankova M, Llobet E, Vilanova X, Brezmes J, Gr�acia I, Cané C, Calderer J,Correig X. Nanoparticle metal-oxide films for micro-hotplate-based gas sensorsystems. IEEE Sens J 2005;5(5):798e809.

51. Viricelle J-P, Pijolat C, Riviere B, Rotureau D, Briand D, de Rooij NF. Compati-bility of screen-printing technology with micro-hotplate for gas sensor and solid oxidefuel cell development. Sens Actuators B 2006;118:263e8.

52. Sahm T, M€adler L, Gurlo A, Barsan N, Pratsinis SE, Weimar U. Flame spray synthesisof tin dioxide nanoparticles for gas sensing. Sens Actuators B 2004;98:148e53.

53. K€uhne S, Graf M, Tricoli A, Mayer F, Pratsinis SE, Hierlemann A. Wafer-level flame-spray-pyrolysis deposition of gas-sensitive layers on microsensors. J Micromech Microeng2008;18:035040.

54. Panchapakesan B, Cavicchi R, Semancik S, DeVoe DL. Sensitivity, selectivity and sta-bility of tin oxide nanostructures on large area arrays of microhotplates. Nanotechnology2006;17(2):415e25.

55. Sears WM, Colbow K, Consadori F. General-characteristics of thermally cycled tinoxide gas sensors. Semicond Sci Technol 1989;4(5):351e9.


56. Lee AP, Reedy BJ. Temperature modulation in semiconductor gas sensing. Sens Ac-tuators B 1999;60:35e42.

57. Ratton L, Kunt T, McAvoy T, Fuja T, Cavicchi R, Semancik S. A comparative studyof signal processing techniques for clustering microsensor data (a first step towards anartificial nose). Sens Actuators B 1997;41:105e20.

58. Kunt TA, McAvoy TJ, Cavicchi RE, Semancik S. Optimization of temperature pro-grammed sensing for gas identification using micro-hotplate sensors. Sens Actuators B1998;53:24e43.

59. Llobet E, Brezmes J, Ionescu R, Vilanova X, Al-Khalifa S, Gardner JW, Barsan N,Correig X. Wavelet transform and fuzzy ARTMAP-based pattern recognition forfast gas identification using a micro-hotplate gas sensor. Sens Actuators B 2002;83(1e3):238e44.

60. Nakata S, Okunishi H, Nakashima Y. Distinction of gases with a semiconductor sensorunder a cyclic temperature modulation with second-harmonic heating. Sens ActuatorsB 2006;119:556e61.

61. Raible S, Briand D, Kappler J, de Rooij NF. Wafer level packaging of micromachinedgas sensors. IEEE Sens J 2006;6(5):1232e5.

62. Briand D, Guillot L, Raible S, Kappler J, de Rooij NF. Highly integrated wafer levelpackaged MOX gas sensors. In: Proceedings of the Transducers’07 conference, Lyon, France,June 10e14, 2007; 2007. p. 2401e4.

63. Suehle JS, Cavicchi RE, Gaitan M, Semancik S. Tin oxide gas sensor fabricated usingCMOS micro-hotplates and in-situ processing. IEEE Electron Device Lett 1993;14:118e20.

64. Laconte J, Dupont C, Flandre D, Raskin J-P. SOI CMOS compatible low powermicroheater optimization for the fabrication of smart gas sensors. IEEE Sens J 2004;4(5):670e80.

65. Gardner JW, Guha PK, Udrea F, Covington J. CMOS interfacing for integrated gassensors: a review. IEEE Sens J 2010;10(12):1833e48.

66. Afridi MY, Suehle JS, Zaghoul ME, Berning DW, Hefner AR, Cavicchi RE,Semancik S, Montgomery CB, Taylor CJ. A monolithic CMOS Microhotplate-based gas sensor system. IEEE Sens J 2002;2(6):644e55.

67. Barrettino D, Graf M, Taschini S, Hafizovic S, Hagleitner C, Hierlemann A. CMOSmonolithic metal-oxide gas sensor microsystems. IEEE Sens J 2006;6(2):276e86.

68. Lundstr€om I, Shivaraman S, Svensson C, Lundkvist L. A hydrogen-sensitive MOSfield-effect transistor. Appl Phys Lett 1975;26:55e7.

69. Lundstr€om I, Sundgren H, Winquist F, Eriksson M, Krantz-R€ulcker C, Lloyd-Spetz A. Twenty-five years of field effect gas sensor research in Link€oping. Sens Actu-ators B 2007;121(1):247e62.

70. Kreisl P, Helwig A, M€uller G, Obermeier E, Sotier S. Detection of hydrocarbon spe-cies using silicon MOS field-effect transistors operated in a non-stationarytemperature-pulse mode. Sens Actuators B 2005;106:442e9.

71. Lloyd-Spetz A, Baranzahi A, Tobias P, Lundstr€om I. High temperature sensors basedon metal-insulator-silicon carbide devices. Phys Status Solidi 1997;162(1):493e511.

72. Briand D, Sundgren H, van der Schoot B, Lundstr€om I, de Rooij NF. Thermally iso-lated MOSFET for gas sensing application. IEEE Electron Device Lett 2001;22(1):11e3.

73. Briand D, van der Schoot B, de Rooij NF, Sundgren H, Lundstr€om I. A low-powermicromachined MOSFET gas sensor. J Microelectromech S 2000b;9(3):303e8.

74. Meier DC, Semancik S, Button B, Strelcov E, Kolmakov A. Coupling nanowirechemiresistors with MEMS microhotplate gas sensing platforms. Appl Phys Lett2007;91:063118.


75. Li M, Yan W, Zhu H, Guo Z, Tang Z. Fabrication and characterization of a low po-wer consumption ethanol gas sensor based on a suspended micro-hotplate. RSC Adv2015;5:51953e60.

76. Santra S, Sinha AK, De Luca A, Ali SZ, Udrea F, Guha PK, Ray SK, Gardner JW.Mask-less deposition of Au-SnO2 nanocomposites on CMOS MEMS platform forethanol detection. Nanotechnology 2016;27:125502.

77. Ryu K, Zhang D, Zhou C. High-performance metal oxide nanowire chemical sensorswith integrated micromachined hotplates. Appl Phys Lett 2008;92:093111.

78. Marasso SL, Tommasi A, Perrone D, Cocuzza M, Mosca R, Villani M, Zappettini A,Calestani D. A new method to integrate ZnO nano-tetrapods on MEMS micro-hotplates for large scale gas sensor production. Nanotechnology 2016;27:385503.

79. Shao F, Fan JD, Hern�andez-Ramírez F, F�abrega C, Andreu T, Cabot A, Prades JD,L�opez N, Udrea F, De Luca A, Ali SZ, Morante JR. NH3 sensing with self-assembled ZnO-nanowire mHP sensors in isothermal and temperature-pulsed mode.Sens Actuators B 2016;226:110e7.

80. Lee J, Kim J, Im JP, Lim SY, Kwon JY, Lee S-M, Moon SE. MEMS-based NO2 gassensor using ZnO Nano-rods for low-power IoT application. J Korean Phys Soc 2017;70(10):924e8.

81. Chen Y, Xu P, Xu T, Zheng D, Li X. ZnO-nanowire size effect induced ultra-highsensing response to ppb-level H2S. Sens Actuators B 2017;240:264e72.

82. Tak�acs M, Z�amb�o D, De�ak A, Pap AE, D€ucs€o C. WO3 nano-rods sensitized with no-ble metal nano-particles for H2S sensing in the ppb range. Mater Res Bull 2016;84:480e5.

83. Krainer J, Deluca M, Lackner E, Sosada F, Wimmer-Teubenbacher R, Koeck A,Gspan C, Rohracher K, Wachmann E, Schrems M. CMOS integrated tungsten oxidenanowire networks for ppb-level hydrogen sulfide sensing. In: Proceedings of the IEEEsensors conference, Orlando, USA, October 30 e November 3, 2016; 2016. p. 1e3.

84. Vallejos S, Umek P, Stoycheva T, Annanouch F, Llobet E, Correig X, De Marco P,Bittencourt C, Blackman C. Single-step deposition of Au- and Pt-nanoparticle-func-tionalized tungsten oxide nanoneedles synthetized via aerosol-assisted CVD, and usedfor fabrication of selective gas microsensor arrays. Adv Funct Mater 2016;23:1313e22.

85. Annanouch FE, Haddi Z, Vallejos S, Umek P, Guttmann P, Bittencourt C, Llobet E.Aerosol-asssited CVD-grown WO3 nanoneedles decorated with copper oxide nano-particles for the selective and humidity-resilient detection of H2S. ACS Appl Mater In-terfaces 2015;7:6842e51.

86. Annanouch FE, Haddi Z, Ling M, Di Maggio F, Vallejos S, Vilic T, Zhu Y, Shujah T,Umek P, Bittencourt C, Blackman C, Llobet E. Aerosol-asssited CVD-grown PdOnanoparticle-decorated tungsten oxide nanoneedles extremely sensitive and selectiveto hydrogen. ACS Appl Mater Interfaces 2016;8:10413e21.

87. Steinhauer S, Chapelle A, Menini P, Sowwan M. Local CuO growth on microhot-plate: in situ electrical measurements and sensing applications. ACS Sens 2016;1:503e7.

88. Prades JD, Jimenez-Diaz R, Hernandez-Ramirez F, Cirera A, Romano-Rodriguez A,Morante JR. Harnessing self-heating in nanowires for energy efficient, fully autono-mous and ultra-fast gas sensors. Sens Actuators B 2010;144:1e5.

89. McAlpine MC, Ahmad H, Wang D, Heath JR. Highly ordered nanowire arrays onplastic substrates for ultrasensitive flexible chemical sensors. Nat Mater 2007;6:379e84.

90. Briand D, Colin S, Courbat J, Raible S, Kappler J, de Rooij NF. Integration of MOXgas sensors on polyimide hotplates. Sens Actuators B 2008;130:430e5.

91. Courbat J, Briand D, de Rooij NF. Reliability of micro-hotplates on polyimide foil.In: Proceedings Transducers’11 conference, Beijing, China; 2011. p. 338e41.


92. Courbat J, Briand D, de Rooij NF. Foil level packaging of a chemical gas sensor.J Micromech Microeng 2010b;20:055026.

93. Zappa D, Briand D, Comini E, Courbat J, de Rooij NF, Sberveglieri G. Zinc oxidenanowires deposited on polymeric hotplates for low-power gas sensors. In: Proceedingsof the eurosensors 2012 conference, Cracow, Poland; 2012.

94. Shen W. Properties of SnO2 based gas-sensing thin films prepared by ink-jet printing.Sens Actuators B 2012;166e7:110e6.

95. Peter C, Kneer J, W€ollenstein J. Inkjet printing of titanium doped chromium oxide forgas sensing application. Sens Lett 2011;9:1e5.

96. Kukkola J, Jansson E, Popov A, Lappalainen J, M€aklin J, Halonen N, T�oth G,Shchukarev A, Mikkola J-P, Jantunen H, Kord�as K, Hast J, Hassinen T, Sunnari A,Jokinen K, Haverinen H, Sliz R, Jabbour G, Fabritius T, Myllyl€a R, Vasiliev A,Zaretskiy N. Novel printed nanostructured gas sensors. Procedia Eng 2011;25:896e9.

97. Ramírez JL, Fatima EA, Llobet E, Briand D. Architecture for the efficientmanufacturing by printing of heated, planar, resistive transducers on polymeric foilfor gas sensing. Sens Actuators B 2018;258:952e60.

98. Rieu M, Camara M, Tournier G, Viricelle J-P, Pijolat C, de Rooij NF, Briand D.Fully inkjet printed SnO2 gas sensor on plastic substrate. Sens Actuators B 2016;236:1091e7.

99. Ishibashi N, Kaneyasu K. Development and application of semiconductor gas sensorusing MEMS technology. In: Presented at 3rd GOSPEL workshop: gas sensors based onsemiconducting metal oxide e new directions, T€ubingen, Germany, November 30eDecember1, 2009; 2009.

100. R€uffer D, Hoehne F, B€uhler J. New digital metal-oxide (MOx) sensor platform. Sen-sors 2018;18:1052.

101. Ehrhardt J-J, Colin L, Jamois D. Poisoning of platinum surfaces by hexamethyldisilox-ane (HMDS): application to catalytic methane sensors. Sensor Actuator B Chem 1997;40:117e24.

102. Khan S, Briand D. All-printed low-power metal oxide gas sensors on polymericsubstrates. Flexible and Printed Electronics 2019;4:015002.


CHAPTER FOURTEEN

Integrated CMOS-based sensorsfor gas and odor detectionP.K. Guha1, S. Santra1, J.W. Gardner21Indian Institute of Technology, Kharagpur, West Bengal, India2University of Warwick, Coventry, United Kingdom

Contents

14.1 Introduction 46514.2 Microresistive complementary metal oxide semiconductor gas sensors 46714.3 Microcalorimetric complementary metal oxide semiconductor gas sensor 46914.4 Sensing materials and their deposition on complementary metal oxide

semiconductor gas sensors472

14.5 Interface circuitry and its integration 47514.6 Integrated multisensor and sensor array systems 48014.7 Conclusion and future trends 483Useful web addresses 485References 486

14.1 Introduction

Gas sensors are increasingly becoming an important part of oureveryday lives. They can be found in our homes (e.g., monitoring the levelof CO in air from gas-fired boilers), in our workplace (e.g., checking thelevels of toxic gases and odors in offices), and in hospitals (e.g., monitoringanesthetic and respiratory gases during operations). There has been anincreasing demand for improved workplace safety for certain industries(e.g., working in coal mines) through tougher government legislationdeven in developing countries such as China and India. Moreover, thereare also some emerging niche markets (e.g., sensors for notebook computers,tablets, and even mobile phones), which require very low-cost (<$5) andlow-power (milliwatt or less) gas/odor sensors in high volumes (tens of mil-lions of units per year). However, the average cost of a gas sensor availabletoday is in excess of $25 and well over $70 for optical ones (for methane andCO2). This is because the production of discrete gas sensors tends to be



https://doi.org/10.1016/B978-0-08-102559-8.00014-8

semiautomated and in batches of relatively small volumes (hundreds tothousands rather than millions). In addition, these sensors require associatedinterface electronics using discrete components to monitor the sensor outputand drive heaters. This increases the cost even further of the sensor modulewith packaging to often in excess of $100.

According to a recent market research report, the gas sensor market willreach around $1.3 billion by 2023, at a compound annual growth rate of6.8% between 2017 and 2023 (“Gas Sensors Market e Global Forecast2023” by Markets and Markets). The major factors driving the growth aredue to the continuous development of miniaturized sensors, increasingenforcement of health and safety regulations by government, and increasingawareness of indoor/outdoor air quality control among users. Furthergrowth will be generated for semiconductor-based gas sensors (as opposedto electrochemical or optical sensors) if solid-state gas sensors can be devel-oped on a complementary metal oxide semiconductor (CMOS) chip com-bined with the associated circuitry using standard microelectronic andmicroelectromechanical processes at a reasonable cost. Success in thisendeavor must reduce the volume cost of sensors to below $1e2 and, hence,access new high volume markets. Current silicon-based gas sensors can cost$50 each in low volumes, which is unacceptable to the mainstream con-sumer market.

One of the most common and successful types of gas sensor sold is theelectrochemical gas sensor that operates at ambient temperatures; however,although these are ultra-low power, they are bulky (w1 cm3) and it is notpossible to integrate this type of sensor and the necessary circuits on a siliconchip. This is because they require a significant volume of liquid electrolytesto operate for a sufficient lifetime and employ non-CMOS catalytic mate-rials, such as platinum or silver electrodes. Another successful gas sensor isbased on nondispersive infrared absorption, but low-cost versions havepoor limits of detection (w100s ppm), require high power for the source,and require expensive non-CMOS optical band-pass filters. They are alsoeven more costly than electrochemical sensors as stated above at $75 ormore with electronics and packaging.

In this chapter, we discuss the two main types of low-cost semiconductorgas microsensors: (1) resistive and (2) calorimetric. Recent research reportsthat it is possible to integrate these types of gas sensor onto a singleCMOS die and, hence, they have the potential for costs low enough toopen up mass markets in the near future. Here we address the key issues asso-ciated with the integration of these gas sensors onto a single CMOS die. The

466 P.K. Guha et al.

main challenges are to deposit/grow gas-sensitive materials (CMOS, inparticular, imposes many constraints on material integration); isolate thesensing area from interface electronics present in the same silicon die; andisolate the nature of the electronic circuits required to interface the sensorsand power the microheater. We also briefly discuss the potential for low-cost arrays of gas sensors, such as those required for complex odor detection,namely electronic nose (e-nose) applications. Finally, we consider the futureof smart CMOS gas sensors with regard to low-cost, low-power devicesneeded for the emerging Internet of things (IoT).

14.2 Microresistive complementary metal oxidesemiconductor gas sensors

Resistive gas sensors are based on monitoring a change in the electricalconductivity of a sensing material (primarily, nonstoichiometric n-type and p-type semiconducting metal oxides, such as SnO2, WO3, ZnO, TiO2, CuO,and NiO) at high operating temperatures compared with standard CMOStechnology (<þ125 �C) (typically, 200e500 �C) in the presence of the haz-ardous/odorous gas. These are perhaps one of the earliest types of availablegas sensor on the market. For example, Taguchi gas sensors (TGS) werecommercialized in the early 1970s and are still produced today by FigaroInc., Japan. The TGS comprises a mm-sized platinum heater coil inside aceramic tube several mm long. A thick layer of porous tin oxide film isattached by hand over thick platinum electrodes on the ceramic cylinderand sintered at a high temperature (e.g., 650 �C) to achieve the appropriatecrystalline grain structure. Fig. 14.1 shows the structure of a Taguchi-type

Lead wire

Lead wireElectrode

Ceramic tube

Sintered tin oxideHeater coil

Figure 14.1 Basic structure of a Taguchi-type tin oxide gas sensor (Figaro Engineering,Japan).

Integrated CMOS-based sensors for gas and odor detection 467

tin oxide gas sensor. The electrical resistance of the tin oxide coating changesin the presence of any gas that reacts with chemisorbed oxygen species (O�,O2

�, etc) at high temperature (typically 250e450 �C), and this is measuredvia the electrodes and a basic potential divider circuit with a matched refer-ence resistor. The reasons for the high-temperature operation of metal oxidegas sensors are (1) the reaction with chemisorbed oxygen sites (and catalyticdopants) changes and can be more specific at higher temperatures, (2) thereaction kinetics are generally much faster, although the sensitivity maydecrease too, and (3) the interference from water vapor is reduced above150 �C.1 Platinum wire resistive heaters are used to heat up the gas sensingmaterial to high operating temperatures for the reaction to be efficient butrequire several hundred milliwatts of power. The key challenges associatedwith resistive CMOS sensor design are to reduce the size and power con-sumption significantly and to achieve a stable and uniform temperatureover the sensing area.

New-generation microresistive sensors are usually formed on a thin ther-mally insulated silicon oxide or nitride membrane so as to reduce the powerloss; this thin membrane-like structure is often referred to as a “siliconmicrohotplate.” Early researchers (beforew2005) studied different materialsfor designing the electrically conducting microheater structures, such asdoped polysilicon, doped single crystal silicon, aluminum, and platinum.2e4.Platinum heaters are probably the most stable material at 400 �C, but thismaterial is not available in a CMOS foundry; hence, post-CMOS depositionis necessary. Within the CMOS process, polysilicon has a poor long-termstability at high temperature because of electromigration, grain boundarymovement, and crack propagation, while aluminum shows electromigrationat high temperatures (aluminum has a melting point of 660 �C, prolongheating over 350 �C softens the material and can cause electromigration).Metal oxide semiconductor field-effect transistor heaters can be controlledby applying varying bias voltage at the gate; however, very high temperatureoperation is not reliable in silicon on insulator (SOI) CMOS because of thetriggering of a parasitic bipolar transistor.5

In the literature, many researchers have claimed that their design is“CMOS compatible,” where in most of the cases “CMOS compatible”means that either the sensor design can be made using a modified CMOSprocess or CMOS materials have been used in the fabrication. However,no standard CMOS process has been used to produce their respective de-vices or components.6 There have been very few groups worldwide whoactually used commercial CMOS-MEMS foundry to fabricate sensor/circuit


single chip solution, e.g., Prof. Gardner’s group at Warwick University,Prof. Udrea’s group at Cambridge University, Prof. Henry Baltes’s groupat ETH Zurich, Dr. Semancik, and others at NIST to name a few.

Microheaters made from pþ silicon perform better than nþ silicon andmuch better than doped polysilicon in long-term operation (using industrialCMOS foundry).7 Some commercial foundries now support high-temperature metals (after 2005) along with aluminum metallization in thesame wafer; so, high-temperature metal (e.g., tungsten) can be used forthe heater and aluminum (lower resistivity compared with tungsten) for cir-cuit metallization. Work led by a group from Cambridge University usingSOI CMOS wafer reported the superiority and reliability of a tungsten heat-er.8 Here, the heater is embedded in an oxide/nitride membrane, hence,thermally isolated from the rest of the chip, which ensures reliable on-chip circuit operation. The membranes were formed at the wafer level(from MEMS foundry) using a backside deep reactive ion etch (i.e., apost-CMOS MEMS process) of the handle silicon. The heaters can heatup the membrane and to, say, 300 �C only require 6 mW of power. Theinterdigitated electrode (for resistance measurement of the sensing material),formed by the top metal layer, was exposed at the same step of bond padopening. The cross-sectional view of the sensor and a top view of a fabri-cated device are shown in Fig. 14.2(a) and (b). Other research groupsdsuchas Baltes (ETH, Zurich) and de Rooij (IMT, Neuchatel)dhave reportedequally power-efficient non-SOI microheaters fabricated inside siliconplugs.9,10 Microhotplates can also be made using front etch of silicon, wherethe membrane is suspended via bridges.11 Such microhotplates are more po-wer efficient (because the hotplate is surrounded by air) compared to back-etched one but may be less stable.

14.3 Microcalorimetric complementary metal oxidesemiconductor gas sensor

Microcalorimetric sensors detect heat or the change in enthalpygenerated from the chemical reaction that takes place between the sensingmaterial and the gas molecules. This kind of sensor generally functions athigher temperatures than resistive gas sensors (around 500 �C or more),and there is a need to integrate the microheater under the catalytic sensingarea, as in the case of microresistors. One of the popular calorimeters is calledthe “pellet resistor” or “pellistor,” which was designed and patented byEnglish Electric Valve (EEV) Ltd in the early 1970s (www.e2v.com).


http://www.e2v.com

Conventional pellistors (shown in Fig. 14.3) are made of a platinum coil in aporous alumina bead (diameter of a few millimeters) along with anembedded metal catalyst (e.g., palladium). The platinum coil not only heatsup the bead to enable the reaction but also acts as a resistance thermometerto measure the change in temperature of the catalytic surface relative to a

Metalmicro-heater Interdigitated

electrodeSensing layer

Siliconnitride

Silicon-dioxide

Buried oxideHandle silicon

Handle silicon etched by DRIE

Gas sensor area Circuit area

(a)

(b)

Figure 14.2 (a) Cross-sectional view of complementary metal oxide semiconductormicrohotplate-based gas sensor (drawing not to scale). (b) An optical microscope im-age of the fabricated microhotplate device. Adapted from Santra S, Ali SZ, Guha PK,Covington JA, Zhong G, Robertson J, Milne WI, Gardner JW, Udrea F. Post-CMOS wafer levelgrowth of carbon nanotubes for low-cost microsensorsda proof of concept. Nanotechnol2010a;21:485301. 7pp.


second noncatalytic reference beaddin a potential divider circuit. The latestdesigns of calorimeters are called microcalorimeters and usually use mem-brane technology such as microresistive sensors (except these do not needthe sensing electrodes). Two identical sensors (one having an active sensingmaterial and the other playing the role of a reference) and a differentialamplifier are often used to measure the difference in temperature causedby the chemical reaction, as shown in Fig. 14.4. Two accurate temperature

Figure 14.3 A conventional commercial pellistor (City Technology Ltd, UK, now part ofHoneywell).

Figure 14.4 A differential transducing circuit for microcalorimeter. Adapted fromUdrea F, Gardner JW, Setiadi D, Covington JA, Dogaru T, Lu CC, Milne WI. Design andsimulations of SOI CMOS micro-hotplate gas sensors. Sensor Actuator B 2001;78:180e90.


sensors placed within the membranes are used to detect the temperature inthe respective microcalorimetric sensors.4

14.4 Sensing materials and their deposition oncomplementary metal oxide semiconductor gassensors

Modern solid-state (resistive) gas sensors are generally based on wide-band-gap semiconducting metal oxideesensing materials. Resistive gas sen-sors available in the market (e.g., TGS 812, Figaro Engineering) generallyuse doped tin oxide as the sensing material. Solid thin film metal oxidesare not effective for microsensors (the smaller surface area causes lower sensi-tivity). Thus, nanomaterials are suitable candidates for gas detection becauseof their exceptionally high surface area and dimension compared with thedepletion region. So, there is a possible increase in sensitivity even for aminiaturized sensor area using new nanomaterials with even smaller grainsizes or dimensions and hence greater total surface areas.

Metal oxideebased sensing layers are popular because they give reason-ably good response toward different gases and volatile organic compounds(VOCs). However, they have two major drawbacks: (1) metal oxidesrequire elevated temperature to achieve good sensitivity (hence consumelarge power) and also (2) they respond to more than one gas and VOC(hence have poor selectivity).

The first issue can be sorted out by using more power-efficient micro-hotplates. Fast microhotplates (because of their low thermal mass) can beused in a pulse mode to reduce power consumption even further. In thelast few years, there have been reports of two-dimensional (2D) materialsas sensing layer, e.g., reduced graphene oxide,12 molybdenum disulphide(MoS2),

13 tungsten disulphide (WS2),14 etc. Such materials can interact

with chemical analytes at near room temperature. Thus, they are promisingcandidate materials for very low-power resistive sensors. However, thesensitivity of these materials is usually poor compared with high-temperature n-type metal oxides. This can be improved if 2D layered mate-rials are mixed with metal oxides (to get the advantage of both), i.e., largeresponse at near room temperature.15

The second issue can be improved by various techniques, e.g., (1) func-tionalizing the sensing layer by metal nanoparticles,16 (2) using mixed metaloxides,17 or (3) analyzing the sensor array data through pattern recognitiontechnique (e.g., principle component analysis, artificial neural network).


However, growing nanomaterials directly on CMOS wafer is very chal-lenging; the CMOS substrate cannot sustain prolonged high-temperatureheating (because high temperature can cause rediffusion in active area ofCMOS transistor and also deformation in metal tracks because of aluminumelectromigration), which is often necessary during the growth process andannealing of the nanomaterials. Again, it is recommended not to use anyharsh chemicals and environments (e.g., long plasma exposure might dam-age fragile microhotplate structure) for nanomaterial synthesis on CMOS.Therefore, a low temperature growth process (<500 �C) with CMOS-friendly chemicals is highly desirable for sensing material synthesis. Onesuch method has been reported for growing zinc oxide nanowires on a fullyprocessed CMOS microhotplate.18 A scanning electron microscopy pictureof grown nanowires on an SOI device is shown in Fig. 14.5. Recently, alocal growth technique has also been reported for carbon nanotubes(CNTs)dagain, on microhotplate area.19 Here, microheaters are used toform catalytic islands and, hence, CNTs are grown only on the heater areaswithout exposing the whole CMOS chip to problematic high temperatures.This idea has been extended to grow CNTs at the wafer level in specifiedareas of multiple chips simultaneously (scheme shown in Fig. 14.6), whichlater can be diced and thus serve the purpose of post-CMOS batch fabrica-tion of sensing material.

(a)

(b)

Figure 14.5 (a) Scanning electron microscopy image of the zinc oxide nanowires on acomplementary metal oxide semiconductor microhotplate. (b) Higher magnificationview of nanowires. Adapted from Santra S, Guha PK, Ali SZ, Hiralal P, Unalan HE,Covington JA, Amaratunga GAJ, Milne WI, Gardner JW, Udrea F. ZnO nanowires grown onSOI CMOS substrate for ethanol sensing. Sensor Actuator B 2010b;146:559e65.


There are many possible ways of depositing nanomaterials onto a CMOSsubstratedfor example, chemical vapor deposition (CVD) is possible pro-vided the process temperature can be reduced to below 600 �C. Onetype of CVD is aerosol-assisted CVD; in this process, precursors are trans-ported to the substrate by means of a liquid/gas aerosol, which can be gener-ated ultrasonically. This technique is suitable for use with nonvolatileprecursors. Alternatively, very thin layers of metal oxides can be grown byusing an atomic layer deposition process. Another promising recentapproach is flame spray pyrolysis because it enables the control of the sizeof the metal oxide nanocrystals in a high-temperature flame before deposi-tion on a wafer close to room temperature. However, the above-mentionedhigh-temperature synthesis techniques might not always be effective torealize highly sensitive sensing layer. This is because a high-temperature syn-thesis technique can reduce the number of active defect sites in the nano-structure. Thus, simple low-cost chemical routes (e.g., modified hummersmethod for graphene oxide synthesis, liquid exfoliation for 2D material syn-thesis, and hydrothermal techniques for metal oxide synthesis) might be abetter approach (simpler and lower cost) to get nanostructures with ampledefect sites.

Techniques such as spray coating and even inkjet printing can be veryuseful for depositing chemically grown nanomaterials on specified areas;particularly, printers with high precision are already available in the market

Figure 14.6 Scanning electron microscopy image of carbon nanotubes, which werelocally grown simultaneously using microheaters placed at adjacent chips. Adaptedfrom Santra S, Ali SZ, Guha PK, Covington JA, Zhong G, Robertson J, Milne WI, Gardner JW,Udrea F. Post-CMOS wafer level growth of carbon nanotubes for low-cost microsensorsdaproof of concept. Nanotechnol 2010a;21:485301. 7pp.


(Sonoplot, Tesscorn). The advantage of this approach is that one can avoidseveral steps of conventional lithography and that use of very high temper-atures and harsh chemicals is not required. However, sometimes, nanoma-terials are not easy to dissolve or form stable suspensions in a liquid; in thiscase, it may be necessary to use a high-speed sonicator to avoid the nozzleclogging up in the printer or spray-coater. Also, use of properly viscousnanomaterials reduces the chances of nozzle clogging. Again, in situ growthof nanomaterial directly on the device gives better adhesion compared tospray coating/inkjet printing. Therefore, in latter cases, one might need tocarry out surface treatment (e.g., roughening of device surface throughmild oxygen plasma) before sensing layer deposition.

14.5 Interface circuitry and its integration

There are two possible ways of interfacing gas sensors with their asso-ciated interface circuitry: the hybrid approach and the monolithic approach:• Hybrid approach: this method employs separate chips for the sensors and

the circuits. So, one can reuse the same circuit chip, even if there is aproblem with the sensor device (e.g., breaking of the membrane), andenhance manufacturing yield through screening. The other mainadvantage is that there is no restriction on the type of material used andthe process steps for the fabrication of the sensors; hence, both the ma-terial and the fabrication process can be adjusted to optimize the overallperformance of the sensor devices. However, parasitic capacitances orinductances associated with long interconnect and bonding wires areundesirable and can cause higher levels of signal degradation/noise. Inaddition, the hybrid approach is more expensive than a single-chipimplementation when considering high volumes. Sometimes thisapproach is called a system in a package.

• Monolithic approach: in this method, the sensor and the circuits are on thesame silicon chip. The monolithic approach is cost-effective in high unitvolumes, making it commercially attractive. From the performance pointof view, this system offers advantages such as a significant reduction in thenoise and cross-talk levels as the number (and length) of wires and bondsis substantially reduced and there should be much better matchingbetween different sensors. However, a fault in a sensor will result in thefailure of the complete chip, even if the circuitry is working properly.This approach is also called a system on a chip.


Apart from the above two methods of integration, there has been recentdevelopment to integrate sensors on top of CMOS substrate, namely three-dimensional (3D) integration. In this approach, different building blocks,e.g., analog front end, digital electronics, memory, wireless module, andmultisensor arrays, can be fabricated on separate silicon substrates fromsame/different foundries and integrated through-silicon vias to achieve 3Dstacking. This will reduce delay (because of reduction in long metal trackspresent in a complex monolithic chip), add additional functionality, increaseflexibility (because of possibility of fabrication from different foundries), andof course significantly reduce silicon footprint. However, it is extremelychallenging to integrate so many functionalities, proper thermal manage-ment, and package of entire system. A range of methods of 3D stackinghave been explored recently, starting from monolithic stacking to diebonding and wafer bonding, to name a few.

A common problem with any microsensor is that they tend to generatesmall electrical signals at the sensor output, so the front-end circuitry needsto be of very low noise (i.e., a good analog design is required). Present sub-micron CMOS processes offer very low rail-to-rail supply (say, 0e1.8 V);this is good for low power but makes front-end design much harder becauseof sensor drift (drift occurs because of aging and incomplete release of analyteat the end of a measurement, etc.).

As chemoresistive sensors are based on a change in conductivity at highoperating temperatures, so main circuit blocks required for interfacing thesesensors are driving circuits for microheaters, temperature control units, andsensing material interface circuits.• Driving circuits for microheaters: this generally requires a stable, accurate,

controllable current or voltage source circuit with large current (tensof milliamps) driving capability.

• Temperature control unit: A bell-shaped temperature dependence of thegas sensor response is most commonly observed experimentally witha distinct maximum depending on the sensing material used andcorresponding gas analytes; thus the temperature of the gas sensingmaterial plays a vital role in improving the selectivity of the sensor.Hence, accurate control of the temperature of the gas sensing material isextremely important. In most commercial gas sensors, the change inheater resistance with temperature is used to set or control the operatingtemperature of the heater itself. In one such reported work, Bota et al.20

used the same resistor element as the heater and temperature sensor.They used a pulse width modulation technique where, during the off


mode of the heater, a small current was driven through the heater tomeasure its temperature (as shown in Fig. 14.7). However, a separatetemperature sensor (in the form of silicon/polysilicon resistor or silicondiode) can increase the flexibility in circuit design and, hence, thereading and control of the precise temperature of the sensing area.There are several ways to control the heater temperaturedfor example,the heating can be controlled in oneoff mode (sometimes called a“bangebang” controller), proportional mode, orproportionaleintegralederivative (PID) mode. The oneoff controlleris the simplest form: if the microhotplate is cooler than a setpointtemperature, the heater is turned on at maximum power; once it ishotter than the setpoint temperature, the heater is switched offcompletely. However, bangebang control can give rise to instabilityunless controlled carefully. It might also damage the heater because ofvery rapid changes in voltage (hence, temperature). In this respect, on-chip digital PID controller is a much better option.21 It usually has rapidrespond time without overswing and steady state error; hence, smoothtemperature control of the heater is possible.

Figure 14.7 Diagram of a temperature control circuit, in which the same heater is usedfor measuring the temperature. Adapted from Bota SA, Dieguez A, Merino JL, Casanova R,Samitier J, Cane C. A monolithic interface circuit for gas sensor arrays: control and mea-surement. Analog Integr Circuits Signal Process 2004;40:175e84. (VDD, supply voltage.Vref, reference voltage).


• Sensing material interface circuit: the design of this interfacing circuit is oneof the main challenging components of a resistive gas sensor. This isbecause of the following:(a) The circuit must handle the precision and dynamic range that is

demanded by the gas sensing element. The baseline resistancevaries enormously depending on the sensing material used, e.g.,sensing materials such as metal oxides can have sheet resistances fromkU per square up to even GU per square.

(b) Often, there is drift in the baseline resistance of the sensing material;this can be due to a variety of reasons (e.g., the material is notthermodynamically stable, material has been polluted by the tar-geted chemical analytes). The circuits should be able to compensatefor this drift to give accurate changes in resistance in the presence ofthe target gas.

The literature reports many different solutions to the challengesmentioned briefly above. A very simple scheme is to use a resistor dividercircuit or a Wheatstone bridge technique. But these are not ideal whenseeking full integration, as they need either trimming or variable resistorsfrom outside the chip to match the sensor resistor. Also, simple voltagedivider will require a resistor bank circuit along with very large value of re-sistors to cover wide range of sensing material resistance, which willconsume a large silicon area. One of the popular approaches found in theliterature is the resistance-to-frequency conversion. The challenge withsuch a scheme is to retrieve/isolate parasitic capacitances associated withthe sensing material because this is a high-frequency process and, hence,resistance measurement will be contaminated with sensing material capaci-tance. Two such schemes have been reported by De Marcellis et al.22 andGrassi et al.23,24 In the scheme reported by De Marcellis et al., the generatedpulse provides both sensing resistance values and parasitic capacitances asso-ciated with the sensing material. In the scheme reported by Grassi et al.,(shown in Fig. 14.8), the parasitic capacitor role can be avoided by isolatingthe sensor resistor from the oscillator portion of the circuit. This wasachieved by using a sensing material resistor at the reference arm of a currentmirror and then using the mirror current to charge and discharge a capacitor.The capacitor voltage is then fed into comparators to compare with high(VH) and low voltage (VL) levels to generate a digital pulse. This resultantsquare wave time period has the information of the sensing material resistor.The scheme achieved (simulation result) a worst-case precision of about


0.5% over the range of five decades (1 kUe100 MU), as shown inTable 14.1.

A relatively simple scheme has been reported by Barrettino et al.25 Alarge dynamic range of sensing material was compressed by using two diodes(logarithmic converter approach); however, the scheme generated only 8-bit resolution (the circuit is shown in Fig. 14.9). One popular baseline driftremoval approach has been reported by Koickal et al.26,27 and is shown inFig. 14.10. In this scheme, during the setup phase, each sensor is drivenby a small value current source and the voltage across the sensor is digitallystored using a simple counting analog-to-digital converter (ADC). Thisstored value is converted back to an analog signal (using a digital-to-analogconverter) and then subtracted from the sensor signal, thus removing the

Vref+

+

+

+

––

–

–

VDD

Sensingmaterial

Flipflop

VH

VL

CNTy

CNTy

CNTy

CNTx

CNTx

CNTxQ

Figure 14.8 Resistance controlled oscillator circuit for gas sensing material interface. Qand Q are two complementary signals at the output of the flip flop. Adapted from GrassiM, Malcovati P, Baschirotto A. A 141-dB dynamic range CMOS gas-sensor interface circuitwithout calibration with 16-bit digital output word. IEEE J Solid State Circuits 2007;42:1543e1554.

Table 14.1 Simulation precision results for sensors resistance materials.Rnominal, applied Rmeasured, simulated Linear error (%)

1 kU 1.0183 kU �0.30881100 kU 102.27 kU þ0.131481 MU 1.0212 MU �0.01593100 MU 101.90 MU �0.04878


baseline resistance. The scheme achieved good sensor stability, and the driftof baseline resistance was found to be less than 5 mV.

More advanced signal processing circuitry is best performed using digitalblocks and microcontroller unit. Digital blocks are easier to design andconsumed less power. Delta-sigma ADC is often used to reduce noise ofthe sensor signal. For smart sensor system, low-power microcontroller inter-facing is very useful. This will help in controlling sensor signal, data process-ing (e.g., pattern recognition operation), and also transferring sensor data tomobile platform.

14.6 Integrated multisensor and sensor array systems

There is an increasing desire to measure more than one gas simulta-neously. For example, in a boiler combustion process, there could be arequirement to measure hydrogen or methane, as well as CO, CO2, andeven oxygen. Alternatively, in the case of automotive gases, there is a desire

Figure 14.9 Logarithmic compression circuit for gas sensing material interface. (Vcontrolis a voltage applied from outside, which will fix the voltage across the sensing materialand thus fix the current at the reference arm of the current source. Vout1 and Vout2 aretwo output terminals of the circuit.) Adapted from Barrettino D, Graf M, Zimmermann M,Hierlemann A, Baltes H, Hahn S, Barsan N, Weimar U. A smart single-chip micro-hotplate-based chemical sensor system in CMOS-technology. Int Symp Circuits Syst 2002;2:157e60.


to measure CO, NO2, and unburnt hydrocarbons. In other words, there isan increasing demand to have multiegas sensing systems with perhaps be-tween four and eight sensors (including ambient humidity, ambient temper-ature, and even barometric pressure).

In some ways, this is the same requirement faced when developing elec-tronic nose technology. Sensor-based electronic noses are based on an arrayof nonspecific gas sensors coupled with a pattern recognition technique.28,29

Electronic noses tend to be large, expensive instruments. For example, theFox 4000 electronic nose (Alpha MOS, France) comprises 18 power-hungry Taguchi-like metal oxideeresistive gas sensors (see Fig. 14.11).

Consequently, there is a significant market pull for sensor arrays contain-ing selective gas sensing materials (multigas devices) and partially selective gas

Figure 14.10 Diagram for baseline cancellation circuit. (Clk, clock; DAC, digital toanalog converter; R1, resistance; Vout, output voltage of the circuit.) Adapted fromKoickal TJ, Hamilton A, Tan SL, Covington JA, Gardner JW, Pearce TC. Analog VLSI circuitimplementation of an adaptive neuromorphic olfaction chip. IEEE Trans Circuit Syst I:Regular Paper 2007;54:60e73.


sensing materials (electronic noses). The two main challenges in sensorarrayebased technology is cost and power consumption. First, the cost ofsensor arrays will be significant; a set of four gas sensors will probably cost$100. This, with the addition of discrete circuitry, display, packaging, etc.,leads to prices of $500e$1000 and, so, a limited market take-up. Second,the power consumption of a set of six metal oxide gas sensors will be 1e2W. This is not compatible with their implementation of battery-operateddevices (such as mobile phones and tablets)dneither would it be suitablefor automotive applications, where low power is still a major demand.

The power consumption of SOI-based gas sensing is much lower thanother technologies and enables individual devices to operate at below10 mWDC8; through the pulsing of the heater, it is possible to obtain a po-wer consumption of less than 1 mW per device. In addition, the use of deepreactive ion etching permits the integration of at least four microhotplates ona single 1 mm by 1 mm die. For example, Fig. 14.12 shows 1 mm squaresilicon dies with two and four separate microhotplates on them. Thefour-element chip can run four gas sensors at different temperatures withlow-power consumption (power consumption of less than 25 mW at400 �C). The small thin SOI membranes provide a thermal responsetime of about 5 ms and so these devices can be operated in a pulsedmode and, hence, reduce the power consumption by a factor of 10 whilestill sampling every second.

Figure 14.11 Fox 4000 electronic nose (Alpha MOS, France).


The utilization of SOI CMOS technology enables on-chip annealing at650 �C and operating temperatures of up to 600 �C, thus allowing the use ofnot only metal oxideeresistive materials (e.g., SnO2 or WO3) but also cat-alytic metals (e.g., palladium or platinum). With a CMOS or BiCMOS pro-cess comprising analog components, it is now possible to combine a set offour sensors with analog sensing/drive circuitry and a digital controller ina die of only 1 mm by 2 mm, thus permitting the production of smart gassensors in high volumes for only $1e2. In practice, the cost of mountingand wire bonding the chip onto, say, a TO-5 package is likely to costmore than the silicon die itself!


In this chapter, we have discussed technological developments towardthe reduction in the size, power, and, most importantly, cost of both resistiveand calorimetric gas (and odor) sensors. A recent CambridgeeWarwickUniversity spin-out Cambridge CMOS Sensors Ltd in the United Kingdom(acquired by ams in June 2016 and now called ams Sensors UK Ltd) believesthat it is possible to fabricate a set of gas sensors and associated circuitry on asingle silicon die at a cost of $1e2 in volume production and with a powerconsumption of a few milliwatts. This can be achieved through the use ofhigh-temperature tungsten interconnects in an SOI CMOS process anddeep reactive ion etching on the reverse of the silicon wafers. The low

Figure 14.12 Multiple microhotplate-based gas sensors on a single silicon die (Cam-bridge CMOS Sensors, UK now ams Sensors UK Ltd).


thermal mass of the SOI membrane results in a thermal response time of justa few milliseconds. This, in turn, makes the technology suitable for usingmore advanced signal processing methods that modulate the device temper-ature. For example, it has been shown that switching the operating temper-ature between set points and measuring the time-series signals can result in asingle microsensor measuring more than one gas!30 The method is based onthe principle that different gases react at different rates and so the dynamicalinformation can be used not only to identify different gases but also simplemixtures of gases.

The future of smart gas sensors may also be based on new sensing mate-rials. SOI CMOS technology permits the on-chip deposition of new,exciting materials using CVD at 600e750 �C. For example, it is nowpossible to grow zinc oxide nanowires and CNTs directly onto the gas sen-sors.18,19 Similarly, it is also possible to grow a single layer of graphenedirectly on the chip at 650 �C with a nickel rather than copper catalyst.Although, the response of CNTs and graphene to gases has been shownrecently to suffer from poor sensitivity, selectivity, and response times, itmay be that synthesis via chemical route and functionalization will over-come these problems and open up a clear route to successfulcommercialization.

Such low-power, smart, and selective chemical sensors will be appro-priate in many applicationsdsafety, security, environmental monitoring,and health care to name a few. In the case of biomedical detection, a nonin-vasive breath test approach looks very promising, as it is easy to repeat thedetection process and it does not cause the discomfort associated with bloodtests. The breath test method is inexpensive and also point-of-care detectioncan be undertaken. It has been found that specific gases/vapors correspondto specific diseasesdfor example, acetone (for diabetes), nitric oxide (forasthma), carbon dioxide (for monitoring respiratory patients to evaluate theirlung and pulmonary function), etc. In such health care applications, the costof the sensor will be a major factor because one requires simple plug-and-play low-cost sensors. We envisage smart sensor modules that cost lessthan V5 each and are 1 mm by 1 mm and so could be packaged in amicro-SD card and inserted into tablets and even mobile phones. Thismeans that they can be used in home automation (IoT); for example, userswould be able to monitor many types of environmental gases, such ascombustible CH4 and toxic CO levels from domestic boilers. More chal-lenging will be addressing health care for an aging populationdagain, inhomes. With more advanced circuitry and signal processing techniques, it


should be possible to measure the breath gases mentioned above with resultswired straight to doctors’ surgeries or hospitals.

In last one decade, people have also been trying to develop sensors onlow-cost substrate, e.g., sensor on cloth, paper, glass, and plastic. This willreduce the sensor cost drastically provided batch fabrication is possible.However, one needs to use hybrid approach, i.e., separate interface elec-tronics chip, in this case.

Temperature of operation is an inherent property associated with aparticular material. In this respect, layered materials have clear advantagecompared with metal oxide counterpart. Recently, there have been reportsof self-power sensors. Here sensor device is interfaced with some energy har-vesting unit (energy harvesting is a process by which ambient energy [e.g.,solar, mechanical, and thermal] is captured and converted to electricity todrive a sensor device). It can be inconvenient to replace batteries in devicesthat need to work over long periods of time or in a remote place.

We believe that the next generation of computer tablets, smart phones,and wearable electronic systems will contain integrated gas sensors and, thus,CMOS gas sensors will find their way into everyday use in the next fewyears. Also, with the rapid expansion of cloud computing and the IoT, mo-bile sensors are recognized as an essential component in future ubiquitoussensor networks. Thus, the sheer numbers (trillions) of future sensor devices,i.e., “things” of IoT, will negate the possibility of individual sensor mainte-nance through human intervention and frequent battery replacement. Therequirement to deploy huge number of nodes of smart sensor systems will beuseful at smart city or smart building for environmental monitoring. Suchsensor systems need to communicate with each other and also with centralsensor hub. Each sensor will have unique identity and will be connected toexisting internet infrastructure. This IoT concept will be effectively realizedif development of low-power low-cost sensor system is possible, and in thisrespect sensor integration with CMOS platform might play a vital role.Thus, huge IoT market of trillions (w1012) of connected devices will driveand speed up to realize smart sensors; this is an exciting prospect for sensormanufacturers, investors, and end users.

Useful web addresses

http://www.alpha-mos.com/http://www.sonoplot.com


http://www.alpha-mos.com/

http://www.sonoplot.com

http://www.customsensorsolutions.comhttp://www.e2v.comhttp://www.figarosensor.com/https://www.marketsandmarkets.com/PressReleases/gas-sensor.asphttp://www.tesscorn.com/nanotechnology.htm

References[1] Gardner JW, Varadan VK, Awadelkarim OO. Microsensors MEMS and smart devices.

West Sussex, England: Wiley; 2001.[2] Cavicchi RE, Suehle JS, Kreider KG, Shomaker BL, Small JA, Chaparala P, Gaitan M.

Growth of SnO2 films on micromachined hotplates. Appl Phys Lett 1995;66:812e4.[3] Houlet LF, Tajima K, Shin W, Itoh T, Izu N, Matsubara I. Platinum micro-hotplates

on thermal insulated structure for micro-thermoelectric gas sensor. IEEJ Trans SensorsMicromachines 2006;126:568e72.

[4] Udrea F, Gardner JW, Setiadi D, Covington JA, Dogaru T, Lu CC, Milne WI. Designand simulations of SOI CMOS micro-hotplate gas sensors. Sensor Actuator B 2001;78:180e90.

[5] Guha PK, Ali SZ, Lee CCC, Udrea F, Milne WI, Iwaki T, Covington JA,Gardner JW. Novel design and characterisation of SOI CMOS micro-hotplates forhigh temperature gas sensors. Sensor Actuator B Chem 2007;127:260e6.

[6] Graf M, Barrettino D, Zimmermann M, Hierlemann A, Baltes H, Hahn S, Barsan N,Weimar U. CMOS monolithic metal-oxide sensor system comprising a micro-hotplate and associated circuitry. IEEE Sens J 2004;4:9e16.

[7] Iwaki T, Covington JA, Gardner JW, Udrea F, Blackman CS, Parkin IP. SOI-CMOSbased single crystal silicon micro-heaters for gas sensors. In: Proceedings of IEEE sensorsconference. Korea; 2006.

[8] Ali SZ, Udrea F, Milne WI, Gardner JW. Tungsten-based SOI microhotplates forsmart gas sensors. J Microelectromech Syst 2008;17:1408e17.

[9] Barrettino D, Graf M, SongWH, Kirstein K, Hierlemann A, Baltes H. Hotplate-basedmonolithic CMOS microsystems for gas detection and material characterization foroperating temperatures up to 500 �C. J Solid-State Circuits 2004b;39:1202e7.

[10] Briand D, Schoot BVD, Rooij NFD, Sundgren H, Lundstrom I. A low-power micro-machined MOSFET gas sensor. J Microelectromech Syst 2000;9:303e7.

[11] Afiridi MY, Suehle JS, Zaghloul ME, Berning DW, Hefner AR, Cavicchi RE,Semancik S, Montgomery CB, Taylor CJ. A monolithic CMOS microhotplate-based gas sensor system. IEEE Sens J 2002;2(6):644e55.

[12] Ghosh R, Midya A, Santra S, Ray SK, Guha PK. Chemically reduced graphene oxidefor ammonia detection at room temperature. ACS Appl Mater Interfaces 2013;5(15):7599e603.

[13] Burman D, Ghosh R, Santra S, Ray SK, Guha PK. Role of vacancy sites and UV-ozone treatment on few layered MoS2 nanoflakes for toxic gas detection. Nanotechnol2017;28(43):435502.

[14] Jha RK, Wan M, Jacob C, Guha PK. Ammonia vapour sensing properties of in situpolymerized conducting PANI-nanofiber/WS2 nanosheet composites. New J Chem2018;42(1):735e45.

[15] Ghosh R, Nayak A, Santra S, Pradhan D, Guha PK. Enhanced ammonia sensing atroom temperature with reduced graphene oxide/tin oxide hybrid film. RSC Adv2015;5:50165e73.


http://www.customsensorsolutions.com

http://www.e2v.com

http://www.figarosensor.com/

https://www.marketsandmarkets.com/PressReleases/gas-sensor.asp

http://www.tesscorn.com/nanotechnology.htm

[16] Ghosh R, Santra S, Ray SK, Guha PK. Pt-functionalized reduced graphene oxide forexcellent hydrogen sensing at room temperature. Applied Physics Letters 2015;107.153102, 5pp.

[17] Nayak AK, Ghosh R, Santra S, Guha PK, Pradhan D. Hierarchical nanostructuredWO3eSnO2 for selective sensing of volatile organic compounds. Nanoscale 2015;7:12460e73.

[18] Santra S, Guha PK, Ali SZ, Hiralal P, Unalan HE, Covington JA, Amaratunga GAJ,Milne WI, Gardner JW, Udrea F. ZnO nanowires grown on SOI CMOS substratefor ethanol sensing. Sensor Actuator B 2010b;146:559e65.

[19] Santra S, Ali SZ, Guha PK, Covington JA, Zhong G, Robertson J, Milne WI,Gardner JW, Udrea F. Post-CMOS wafer level growth of carbon nanotubes forlow-cost microsensorsda proof of concept. Nanotechnol 2010a;21:485301. 7pp.

[20] Bota SA, Dieguez A, Merino JL, Casanova R, Samitier J, Cane C. A monolithic inter-face circuit for gas sensor arrays: control and measurement. Analog Integr Circuits SignalProcess 2004;40:175e84.

[21] Barrettino D, Graf M, Hafizovic S, Taschini S, Hagleitner C, Hierlemann A, Baltes H.A single-chip CMOS micro-hotplate array for hazardous-gas detection and materialcharacterization. In: International solid-state circuit conference. San Francisco, CA; 2004.

[22] De Marcellis A, Depari A, Ferri G, Flammini A, Marioli D, Stornelli V, Taroni A. Un-calibrated integrable wide-range single-supply portable interface for resistance andparasitic capacitance determination. Sensor Actuator B Chem 2008;132. 477e184.

[23] Grassi M, Malcovati P, Baschirotto A. Wide-range integrated gas sensor interface basedon a resistance-to-number converter technique with the oscillator decoupled from theinput device. In: Proceedings of IEEE symposium on circuits and systems. Kos Island, Greece;2006.

[24] Grassi M, Malcovati P, Baschirotto A. A 141-dB dynamic range CMOS gas-sensorinterface circuit without calibration with 16-bit digital output word. IEEE J Solid StateCircuits 2007;42:1543e54.

[25] Barrettino D, Graf M, Zimmermann M, Hierlemann A, Baltes H, Hahn S, Barsan N,Weimar U. A smart single-chip micro-hotplate-based chemical sensor system inCMOS-technology. Int Symp Circuits Syst 2002;2:157e60.

[26] Koickal TJ, Hamilton A, Tan SL, Covington J, Gardner JW, Pearce T. Smart interfacecircuit to ameliorate loss of measurement range in chemical microsensor arrays. In: Pro-ceedings of the IEEE instrumentation and measurement technology conference. Ottawa, Ont.,Canada; 2005.

[27] Koickal TJ, Hamilton A, Tan SL, Covington JA, Gardner JW, Pearce TC. AnalogVLSI circuit implementation of an adaptive neuromorphic olfaction chip. IEEE TransCircuits Syst I: Regular Paper 2007;54(1):60e73.

[28] Gardner JW, Bartlett PN. Electronic noses: principles and application. Oxford: Oxford Uni-versity Press; 1999.

[29] Pearce T, Schiffman SS, Nagle HT, Gardner JW.Handbook of machine olfaction: electronicnose technology. Weinheim: Wiley VCH; 2003.

[30] Iwaki T, Covington JA, Gardner JW. Identification of different vapors using a singletemperature modulated polymer sensor with a novel signal processing technique.IEEE Sens J 2009;9:314e28.



IndexNote: ‘Page numbers followed by “f ” indicate figures, “t” indicates tables’.

A“Ablation plume”, 164Acceptor-doped materials, 366e367Adsorptionmolecular mechanism of, 261e265phenomena, 250e261processes, optical probing of, 245e246

Air constituents, competitive adsorptionof, 253e259

Ammonia, 310e311Analog-to-digital converter (ADC),

479e480Arc discharge, 192

BBase air resistance, 18e21Binary switch behavior, 323Breathing mode, 387BTEX, 223compounds, 223e225, 224t

CCalixarenes, 229e230Cambridge CMOS Sensor (CCS), 454Carbon nanomaterials, 226Carbon nanotubecharacterization ofRaman spectroscopy, 200e201, 200fscanning electron microscope (SEM),

203e204, 203fe204ftransmission electron microscopy

(TEM), 204e205, 205fX-ray diffraction, 201e202, 202f

preparation ofdip-coating, 198drop-coating, 197electron beam (E-beam) evaporation,

198e199screen-printing, 197e198spin-coating, 194e196

sensing mechanism of,205e206

synthesis ofarc discharge, 192chemical vapor deposition(CVD), 193

laser ablation, 192e19318650 cells, 190e191Charge carrier transport, 104e119Chemical vapor deposition (CVD),

164e165, 193, 474Chemoresistive gas sensors, 39e40Clean semiconductor surface, 117e118CNT/polymer nanocomposite sensing

materials, 209e210on fabric substrate, 215e216, 215fon textile substrate, 212e215

CO2, 329e331, 331fCommercial finite element method

program, 362Commercial sensor systems, 310e311Competitive adsorption of air constituents,

253e259Complementary metal oxide

semiconductor (CMOS)analog-to-digital converter (ADC),

479e480chemical vapor deposition (CVD), 474future trends, 483e485integrated multisensor and sensor array

systems, 480e483, 481f, 483finterface circuitry, 475e480driving circuits for microheaters, 476hybrid approach, 475monolithic approach, 475sensing material interface circuit, 478temperature control unit, 476e477

low-cost semiconductor gas microsensors,466e467

microcalorimetric complementary metaloxide semiconductor gas sensor,469e472

microheaters, 469

489 j

Complementary metal oxide semiconductor(CMOS) (Continued )

microresistive complementary metaloxide semiconductor gas sensors,467e469

new-generation microresistivesensors, 468

sensing materials, 472e475silicon on insulator (SOI), 468Taguchi gas sensors (TGS),

467e468, 467ftechniques, 474e475volatile organic compounds (VOCs), 472

Conduction electrons, 74Conduction mechanism in SMOXs gas

sensorsn-type SnO2, 62e67p-type CuO, 58e62sample preparation and experimental

conditions, 57e58Conductive polymers, 81, 81fConductometric sensor, 169e170, 348Contact potential differences (CPDs), 57,

58f, 62f, 64f, 97Conventional pellistors, 469e472CPDs. See Contact potential differences

(CPDs)Crocheting technique, 206e207CVD. See Chemical vapor deposition

(CVD)Cyclodextrins (CD), 226e229, 228t

DDangling bonds, 99e101Deep cavitands, 230e232Deep reactiveeion-etched (DRIE)

membrane, 436e437Depletion approximation, 105e106Depletion layer model, 56Deposition techniques, 162one-dimensional nanostructures,

166e169, 167ftwo-dimensional nanostructures,

163e166Derivatives, 229e230Device integration, 170Device structure, 9e10, 9f

Dielectric material-gate field-effecttransistor, 31

Dielectric thin films, 439e441Dilute limit, 118Diode-type sensors, 33e34Dip-coating, 198Dipole formation, interfacial layer,

119e120Direct and indirect thermoelectric gas

sensors, 350e352, 351fDonor-doped materials, 366e367Drift velocity, 76Drop-coating technique, 197, 211, 211fDynamic sensor operation, 332e334breathing mode, 387electrical multiparameter readout

(EMR), 386gas-sensitive field-effect transistors

(GasFETs), 388, 398e404combined methods, 403e404current compensation mode, 403gate biasecycled operation,401e403, 402f

SiC-FET sensors, 399e400temperature-cycled operation,399e400

metal oxide semiconductor (MOS) gassensors, 385, 388e398

field effect, 396e397, 398foptical excitation, 398temperature-cycled operation,390e396, 392f, 394fe395f

stability problems, 386systematic data evaluation, 386e387temperature-cycled operation

(TCO), 388transient response, 387

EElectrical gas sensors, 272Electrical multiparameter readout

(EMR), 386Electric field and capacitance, 104e109Electrode configuration, 85e90Electrode geometry, 90e95Electrode materials, influence of,

77e85

490 Index

Electrode-oxide interfaces, gas sensoroperation

charge carrier transport in, 104e119electrode configuration, 85e90electrode geometry, 90e95electrode-oxide semiconductor interfacescontacts with surface states and

interfacial layer, 99e103ideal contact of metal and oxide

semiconductor, 95e99image force effects on the barrier

height, 103e104gas/solid interactions indipole formation in interfacial layer,

119e120hydrogen adsorption in Schottky

barrier junction, 120e122Schottky barrier junction,

122e124influence of electrode materials, 77e85metals and conduction, 74e76semiconductor gas sensor, 72e73Taguchi sensor, 72

Electrode-oxide semiconductor interfacescontacts with surface states and interfacial

layer, 99e103ideal contact of metal and oxide

semiconductor, 95e99image force effects on the barrier height,

103e104Electrodes and CNT/polymer

nanocomposites for textile-basedsensors, 206e210

Electron affinity, 46e47Electron beam evaporation (EBE),

163, 198e199Electron energy band diagram, 102fEmbroidery technique, 207English Electric Valve (EEV) Ltd,

469e472E-textiles, 189, 191Eu3+:ZrO2

characterization, 289e290, 290foxygen sensing, 290e291, 292fpreparation, 289e290, 290fsensing mechanism, 291e294,

293fe294f

FFabrication, 10Fabric-based embroidered gas sensors,

212e213Fabric-based screen-printed gas sensors,

213e215, 214fField-effect transistor-type gas sensors,

27e31Field emission (FE), 113e114Flame spray pyrolysis (FSP), 431, 432fFocused ion beam scanning electron

microscope (FIB-SEM), 445e446

GGas response, disturbances to, 13Gas sensing characteristics, 11e13Gas-sensitive devices, 414e416Gas-sensitive field-effect transistors

(GasFETs), 388, 398e404combined methods, 403e404current compensation mode, 403gate biasecycled operation,

401e403, 402fSiC-FET sensors, 399e400temperature-cycled operation, 399e400

Gas sensor operation, electrode-oxideinterfaces in

charge carrier transport in, 104e119electrode configuration, 85e90electrode geometry, 90e95electrode-oxide semiconductor interfacescontacts with surface states andinterfacial layer, 99e103

ideal contact of metal and oxidesemiconductor, 95e99

image force effects on the barrierheight, 103e104

gas/solid interactions indipole formation in interfacial layer,119e120

hydrogen adsorption in Schottkybarrier junction, 120e122

Schottky barrier junction, 122e124influence of electrode materials, 77e85metals and conduction, 74e76semiconductor gas sensor, 72e73Taguchi sensor, 72

Index 491

Gas sensors, 32e34, 186e187Gas/solid interactions indipole formation in interfacial layer,

119e120hydrogen adsorption in Schottky barrier

junction, 120e122Schottky barrier junction, 122e124

Gold, 83

HH2O adsorbates, 259e261Homogeneous semiconductors, 365“Hot wire” type, 9e10H2O vapor, response to, 248e249“Hybrid ” chemical gas densor, 187Hydrogen adsorption in Schottky barrier

junction, 120e122Hydrogen-containing gases,

320e324, 322fHydrogen detection, 323e324

IIdeal contact of metal, 95e99Ideal gas sensors, 134III-nitrides, 243Image force effects on barrier height,

103e104Immersion-coating technique, 210, 211fImpurity stabilized, 118Inflammable gases, response to, 22e23InGaN/GaN nanowire heterostructure

arrays, 239e242Inhomogeneous semiconductors, 365Integrated multisensor and sensor array

systems, 480e483, 481f, 483fInterdiffusion, 118e119Interface circuitry, 475e480driving circuits for microheaters, 476hybrid approach, 475monolithic approach, 475sensing material interface circuit, 478temperature control unit, 476e477

Interfacial layer, 99e103structure of, 115e119

Internal sensitivities, 137Internet of Things (IoT), 134, 455Intrinsic emission, 279

Ionic direct thermoelectric gas sensors,378e379

Isothermal PoissoneBoltzmann equation,362

LLight-addressable potentiometric sensor

(LAPS), 144e147Limit of detection (LOD), 225Liquid phase growth methods, 168e169Llaser ablation, 192e193Low-cost semiconductor gas microsensors,

466e467Low temperature cofired ceramic

(LTCC), 312

MMacrocyclic compounds, 225e226Magnesium oxide (MgO), 327e328Metal-insulatoresemiconductor capacitor

type sensors, 32e33, 33fMetal insulator semiconductor (MIS) gas

sensor, 310e311Metal oxides (MOXs), 161e162,

200e206, 239e240, 272Metal oxide semiconductor capacitor,

142e144Metal oxide semiconductor field-effect

transistor (MOSFETs), 148e150Metal oxide semiconductor (MOS) gas

sensors, 151, 385, 388e398band diagram modulation, 153e154field effect, 396e397, 398foptical excitation, 398SnO2 bands, 151e152temperature-cycled operation, 390e396,

392f, 394fe395fMetal oxide sensing films, 194e199Metals and conduction, 74e76“Metalesemiconductor contact diode”,

33e34Metal-semiconductor interface, 104e109Metal-semiconductor interfacial zones,

116fMicrocalorimetric complementary metal

oxide semiconductor gas sensor,469e472

492 Index

Microelectromechanical system (MEMS)sensor, 9e10, 35

Microheaters, 469Micromachined semiconductor gas sensorsapplications, 454e458CMOS-compatible process, 439complementary metal oxide

semiconductor (CMOS), 414,437e441

concept and technologies, 416e425gas-sensitive devices, 414e416polysilicon heaters, 415temperature modulation, 415

manufacturing, 454e458microelectromechanical systems (MEMS)

devices, 413e414microhotplate performance, 418e421,

419fe420fmicrohotplate realization, 418e421,

419fe420fmicrohotplate reliability, 421e425micromachined field-effect gas sensors,

442e444, 443fe444fmicromachined metal oxide gas sensors,

425e437flame spray pyrolysis (FSP), 431, 432fpackaging, 435e437, 436ftemperature modulation,

432e435, 434fthick gas-sensitive films,

428e431, 429fthin gas-sensitive films, 425e428, 427f

MOSFET switch, 439nanostructured gas sensing layers,

445e450, 445f, 448fe449fpolymeric foil, 450e454printing semiconductor gas sensors,

452e454semiconductor gas sensors,

450e451, 452fproducts, 454e458robust high-temperature tungsten-based

SOI microhotplates, 441thermal design, 416e418, 417f

Microresistive complementary metaloxide semiconductor gas sensors,467e469

“Mixed potential” type sensors, 31e32

Molecular mechanism of adsorption,261e265

Monolayer formation, 118Monolayers, 118e119Mott barrier, 97e98MOX gas sensors, 189MOX materials, 187Multivariable sensing, 286e287

NNanocrystalline form, 274Nanostructured gas sensing layers,

445e450, 445f, 448fe449fNanowires, 448e449New-generation microresistive sensors,

468Nitrogen oxides, 310e311Nonhydrogen-containing gases,

324e327, 325fNonresistive sensors, 27NOx, 329e331, 331fn-type, 47e56n-type SnO2, 62e67

OOne-dimensional approach, 52One-dimensional nanostructures,

166e169, 167fOne-electrode configuration, 87fOperating temperature, 12e13Optical probing of adsorption processes,

245e246Ordinary least square (OLS), 287Organic precursor, 278Output characteristics, MOSFET, 149Oxide semiconductor, 95e99“Oxide semiconductor gas sensors”, 6Oxide semiconductor-gate field-effect

transistor, 29e31Oxide-semiconductor interface, tunneling

effects in, 112e115Oxidizing gases, response to, 23e25,

246e248Oxygen, 277, 328e329concentration cell type sensors, 31e32nitrogen mixtures, 374e375, 374fresponse to, 18e21

Index 493

PPalladiumegold, 85Palladiumesilver, 84Petrochemical industry, 223e224Photoluminescence-based gas sensorscollisional energy transfer process, 273conductometric sensing, 272fconventional chemiresistive gas

sensing, 272fluorescence, 272nanocrystalline form, 274oxygen sensing mechanism, 277progress, 275e277rare eartheactivated inorganic sensor

materials, 273e275SterneVolmer relationship, 273

Photoluminescent InGaN/GaN nanowirearrays, 243e245

Platinum, 84Platinumegold, 84e85Platinumesilver, 84PL response, 246e250dependence of, 250e253

Poisson’s equation, 52Polyimide (PI), 450e451Polymeric foil, 450e454printing semiconductor gas sensors,

452e454semiconductor gas sensors, 450e451,

452fPolysilicon heaters, 415“Precursors”, 164e165Pr3+:(K0.5Na0.5)NbO3oxygen sensing, 298e299potassiumesodium niobate, 298synthesis, 298

p-type, 47e56CuO, 58e62

QQuartz crystal microbalance (QMB), 327

RRaman scattering analyses, 289e290Raman spectroscopy, 200e201, 200fRatiometric sensor material, 277Receptor function, 14e18, 40e41“Reduced resistance”, 17e18Reducing gases, response to, 249e250

Resistor-type sensorsextensions, 25e27field-effect transistor-type gas sensorsdielectric material-gate field-effecttransistor, 31

oxide semiconductor-gate field-effecttransistor, 29e31

principle, 27e28solid electrolyte-gate field-effecttransistor, 28e29

gas sensing characteristics, 11e13gas sensors, 32e34nonresistive sensors, 27oxygen concentration cell type sensors,

31e32receptor function and transducer

function, 14e18response to inflammable gases, 22e23response to oxidizing gases, 23e25response to oxygen, 18e21semiconductor oxygen sensors, 13e14sensing materials and devices, 6e10

Response behavior, 246Response transients, 11e12, 11f

SScanning electron microscope (SEM),

203e204, 203fe204ffabric-based embroidered gas sensors,

212e213fabric-based screen-printed gas sensors,

213e215, 214fSchottky barrier junction, 97e98, 110f,

122e124hydrogen adsorption in, 120e122

Screen-printing technology, 41e42,197e198, 208e209

SEM. See Scanning electron microscope(SEM)

Semiconducting direct thermoelectric gassensors

acceptor-doped materials, 366e367commercial finite element method

program, 362conductometric gas sensors, 348definition, 353e379direct and indirect thermoelectric gas

sensors, 350e352, 351fdonor-doped materials, 366e367

494 Index

future trends, 380e381homogeneous semiconductors, 365inhomogeneous semiconductors, 365ionic direct thermoelectric gas sensors,

378e379isothermal PoissoneBoltzmann

equation, 362measurements, 367e378measurement techniques, 353e357,

355fe356fmodeling and simulation, 357e367,

358f, 360fmotivation for research, 347e349oxygen/nitrogen mixtures, 374e375,

374fpropane concentration profile, 371research activities, 353thermoelectric power, 349e350Wolkenstein isotherm, 365e366yttria-stabilized zirconia (YSZ), 348

Semiconducting metal oxide (SMOXs)gas sensors, 41e47, 41f

conduction mechanism inn-type SnO2, 62e67p-type CuO, 58e62sample preparation and experimental

conditions, 57e58n-type, 47e56p-type, 47e56

Semiconducting metal oxides (SMOXs),39e40

Semiconductor gas sensors, 4, 72e73,239e242

classification of, 5e6, 5fideal gas sensors, 134industrial processes and pollution control,

133light-addressable potentiometric sensor,

144e147metal oxide semiconductor capacitor,

142e144metal oxide semiconductor field-effect

transistor (MOSFETs), 148e150metal oxide semiconductors

(MOS), 151band diagram modulation, 153e154SnO2 bands, 151e152

resistor-type sensorsextensions, 25e27field-effect transistor-type gas sensors,27e31

gas sensing characteristics, 11e13gas sensors, 32e34nonresistive sensors, 27oxygen concentration cell type sensors,31e32

receptor function and transducerfunction, 14e18

response to inflammable gases, 22e23response to oxidizing gases, 23e25response to oxygen, 18e21semiconductor oxygen sensors, 13e14sensing materials and devices, 6e10

sensor blocks, 135evaluation of resolution, 139e142resolution, 138e139response curve, 136sensitivity, 137e138

solid-state sensors, 134Semiconductor oxygen sensors, 13e14Semiconductor technology, 243, 273Sensing materials, 472e475devices, 6e10

Sensing materials mechanismcarbon nanotube, 205e206

Sensirion Gas sensor Platform (SGP),455e457, 456f

Sensitizers, 8Sensor assembly, 199e200for textile-based gas sensorsdrop-coating technique, 211, 211fimmersion-coating technique,210, 211f

Sensor blocks, 135evaluation of resolution, 139e142resolution, 138e139response curve, 136sensitivity, 137e138

Sensor packaging, 335e336, 335f“Short-term effect of water vapor”, 13Silicon carbide field effect gas sensors

advanced data evaluation, 332e334ammonia, 310e311applications

Index 495

Silicon carbide field effect gas sensors(Continued )field tests, 336e338, 337fsensor packaging, 335e336, 335f

commercial sensor systems, 310e311dynamic sensor operation, 332e334field effect devices, 311low temperature cofired ceramic

(LTCC), 312metal insulator semiconductor (MIS) gas

sensor, 310e311nitrogen oxides, 310e311sensing mechanismsbinary switch behavior, 323CO2, 329e331, 331fhydrogen-containing gases,320e324, 322f

hydrogen detection, 323e324magnesium oxide (MgO), 327e328new material combinations, 327e328nonhydrogen-containing gases,324e327, 325f

NOx, 329e331, 331foxygen, 328e329quartz crystal microbalance(QMB), 327

tailor-made sensing layers,328e329, 328f

ultrahigh vacuum (UHV), 323voltage drops, 318e319

transduction, 312e327transducer platform, 313e316, 314ftransduction mechanisms,316e318, 318f

Silicon on insulator (SOI), 468Silver, 83Single-crystalline, 295SMOX-based sensors, 40e41Sm3+:TiO2

characterization, 278e280mathematical model, 282e285, 283fmultivariable sensing, 286e287, 286fPL-based oxygen sensing, 280e282, 281fpreparation, 278e280sensing mechanism, 282e285

Solid electrolyte-gate field-effecttransistor, 28e29

Solid-state sensors, 134Spin-coating, 194e196SterneVolmer law, 280Surface states, 99e103Synthesis of carbon nanotubearc discharge, 192chemical vapor deposition (CVD), 193laser ablation, 192e193

Systematic data evaluation, 386e387

TTaguchi gas sensors (TGS), 467e468, 467fTaguchi sensor, 72Tb3+:CePO4

characterization, 295, 296fgas sensing, 295e297, 296forthophosphates, 294preparation, 295, 296fX-ray photoelectron spectroscopy

(XPS), 297TEM. See Transmission electron

microscopy (TEM)Temperature-cycled operation

(TCO), 388Temperature modulation, 415,

432e435, 434fTemplate-assisted methods, 169Textile-based electrode, preparation ofcrocheting technique, 206e207embroidery technique, 207screen printing technique, 208e209

Textile-based gas sensors, sensorassembly for

drop-coating technique, 211, 211fimmersion-coating technique, 210, 211f

Thermal design, 416e418, 417fThermal evaporation (TE) technique, 163Thermionic field emission (TFE),

113e114Thermoelectric power, 349e350Thick gas-sensitive films, 428e431, 429fThin gas-sensitive films, 425e428, 427fTransducer function, 14e18Transduction principles and related novel

devices, 170e174Transient response, 387Transistor outline (TO) headers, 435

496 Index

Transmission electron microscopy (TEM),204e205, 205f

Transport mechanism across the junctionbarrier, 109e112

Tunneling effects, in oxide-semiconductor interface, 112e115

Two-dimensional nanostructures,163e166

UUltrahigh vacuum (UHV), 323“Utility factor”, 12e13

VVapor phase growth methods,

167e168, 446Volatile organic compounds (VOCs),

223, 472Voltage drops, 318e319

W“Wet” process, 10Wolkenstein isotherm, 365e366

XX-ray diffraction (XRD), 201e202,

202f, 289e290X-ray photoelectron spectroscopy

(XPS), 297

YYttria-stabilized zirconia (YSZ), 348

ZZirconium dioxide (ZrO2), 288e294ZnO nanowires, 451

Index 497


Date post:	02-Feb-2023
Category:	Documents
Upload:	khangminh22
View:	0 times
Download:	0 times

Semiconductor Gas Sensors

Documents