Electrical Properties of Nanocrystalline WO for Gas Sensing

Comprehensive Summaries of Uppsala Dissertationsfrom the Faculty of Science and Technology 948

Electrical Properties of Nanocrystalline WO3

for Gas Sensing Applications

BY

ANDERS HOEL

ACTA UNIVERSITATIS UPSALIENSISUPPSALA 2004

i

Till Karin, Katarina, Fredrik, Mamma och Pappa

ii

Papers included in the thesis

I. Current and voltage noise in WO3 nanoparticle films A. Hoel, L.K.J. Vandamme, L.B. Kish, E. Olsson J. App. Phys. 91 (2002) 5221

II. Conduction invasion noise in nanoparticle WO3/Au thin–film devices for gas sensing application A. Hoel, J. Ederth, J. Kopniczky, P. Heszler, L.B. Kish, E. Olsson, C.G. Granqvist Smart Mater. Struct 11 (2002) 640

III. Gas sensing properties of nanocrystalline WO3 films made by advanced reactive gas deposition J.L. Solis, A. Hoel, L.B. Kish, S. Sauko, V. Lantto, C.G.Granqvist J. Am. Ceram. Soc. 84 (2001) 1504

IV. Gas sensing properties of films consisting of nanocrystalline WO3and Pd made by advanced reactive gas deposition A. Hoel, L.F. Reyes, S. Saukko, P. Heszler, V. Lantto, C.G.GranqvistSubmitted to Sens. Actuators

V. Gas sensor response studies of pure and activated WO3 nanoparticle films made by advanced reactive gas deposition L.F. Reyes, A. Hoel, S. Saukko, P. Heszler, V. Lantto, C.G.GranqvistSubmitted to J. Appl.Phys.

VI. Low level detection of ethanol and H2S with a WO3 nanoparticle gas sensor driven by square voltages R. Ionesco, A. Hoel, C.G. Granqvist, E. Llobet, P. HeszlerSubmitted to Sens. Actuators

VII. Small Polaron Formation in Porous WO3-x Nanoparticle Films J. Ederth, A. Hoel, G.A. Niklasson, C.G. Granqvist Submitted to J. Appl. Phys.

iii

Comments on my participation

I All experiments, part of analysis and most of the writing II All experiments except AFM and most of the writing III Part of sample production, SEM analysis and part in writing IV All experiments except XPS, all analysis and most of the

writingV Part of sample production, SEM analysis, part of the writing VI Sample production, part in experimental, part in writing VII Sample production, SEM analysis, X-ray analysis, part in

analysis and writing

Papers not included in the thesis

i. Ag-Mn nanoparticles: Three dimensional finite size effect of the spin glass state J. Ederth, A. Hoel, C.I. Johansson, L.B. Kiss, E. Olsson,C.G. Granqvist, P. Nordblad J. Appl. Phys. 86 (1999) 6571

ii. The microstructure of nanocrystalline tungsten oxide films made by reactive gas evaporation A. Hoel, R. Vajtai, L.B. Kiss, E. Olsson Proc. 51st Ann Meeting Scandinavian Soc. Electron Microscopy, Bergen, Norway (1999) 51

iii. Electrical properties of nanocrystalline tungsten trioxide A. Hoel, L.B. Kish, R. Vajtai, G.A. Niklasson, C.G. Granqvist,E. Olsson Proc. MRS 581 (2000) 15

iv. Infrared spectroscopy of electrochromic nanocrystalline tungsten oxide films made by reactive advanced gas deposition J.L. Solis, A. Hoel, V. Lantto, C.G. Granqvist J. Appl. Phys. 89 (2001) 2727

v. 1/f noise in WO3 nanoparticle films as a diagnostic tool A. Hoel, L.K.J. Vandamme, L.B. Kish, E. Olsson, Gy. Trefan Proc. 16th Int. Conf. Noise and Fluctuations (2001) 755

vi. Invasion noise in nanoparticle WO3/Au devices A. Hoel, J. Ederth, P. Heszler, L.B. Kish, E. Olsson, C.G. Granqvist Proc. SPIE 4590 (2002) 229

iv

vii. Nanoparticle films made by gas phase synthesis: Comparison of various techniques, and sensor applications P. Heszler, L.F. Reyes, A. Hoel, L. Landström , V. Lantto, C.G. Granqvist Proc. SPIE Vol. 5055, (2003) 106

viii. Detection of ethanol and H2S gases in air and in presence of both reducing and oxidizing species with a nanoparticle WO3 gas sensor R. Ionesco, A. Hoel, C.G. Granqvist, E. Llobet, P. Heszler Submitted to Sens. Actuators

ix. Improved gas response at room temperature of activated nanocrystalline WO3 filmsL.F. Reyes, S. Saukko, A. Hoel, V. Lantto, C.G. Granqvist, J. Lappalainen Proc. 20 th Nordic Semiconductor Meeting, Tampere, Finland

x. Nanomaterials for environmental applications: Novel WO3-basedgas sensors made by gas depositionA. Hoel, L.F. Reyes, P. Heszler, V. Lantto, C.G. Granqvist Submitted to Current Applied Physics

xi. Optical characterization and modeling of black pigments used in thickness sensitive solar selective absorbing paints T. Tesfamichael, A. Hoel, G.A. Niklasson, E. Wäckelgård, M.K. Gunde, Z.C. Orel Solar Energy, 69 (2000) 35

xii. Optical characterization of black pigment for solar selective absorbing paints T. Tesfamichael, A. Hoel, G.A. Niklasson, E. Wäckelgård, M.K. Gunde, Z.C. Orel Appl. Opt. 40 (2001) 1672

xiii. Optical electrical and microstructural properties of tin doped indium oxide films made from sintered nanoparticles A. Hultåker, A. Hoel, C.G. Granqvist, A. v. Doorn, M.J. Jongerius, D. Burgard Proc. MRS 703, (2002) 185

v

xiv. Photoelectrochemical and Optical Properties of Nitrogen Doped Titanium Dioxide Films Prepared by Reactive DC Magnetron SputteringT. Lindgren, J.M. Mwabora, E. Avendaño, J. Jonsson, A. Hoel,C.G. Granqvist, S.E. LindquistJ. Phys. Chem. B 107, (2003) 5709

xv. Electrical and optical properties of thin films consisting of tin-doped indium oxide nanoparticles J. Ederth, P. Johnsson, G. A. Niklasson, A. Hoel, A. Hultåker,P. Heszler, C.G. Granqvist, A. R. van Doorn, M. J. Jongerius,D. Burgard Phys. Rev. B 68, (2003) 155410

xvi. Surface morphologies of spectrally selective and polarization- dependent angular optical reflectors of SnOx:F coated anodized aluminiumM. Mwamburi, A. Hoel and E. Wäckelgård Sol. Energy. Mat. Sol. Cells (accepted)

xvii. Photoelectrochemical study of sputtered nitrogen-doped titanium dioxide thin films in aqueous electrolyte T. Lindgren, G.R. Torres, J. Lu, A. Hoel, C.G. Granqvist,S.E. LindquistSubmitted to Sol. Energy. Mat. Sol. Cells

vi

List of Symbols

RomanA Cross section area of a conductor a' Lattice constant Bd Broadening of diffraction line at full width, half maximum C=C’-iC’’ Complex capacitance C1, C2, C3 ContactsCp,dc Pre-exponential factor for dc conduction C Pre-exponential factor for optical absorption c Speed of light d, dc Thickness or distance in material dd Depth of depletion dhkl Interplanar distanceE, E0 Field strengthEA Activation energy EB Binding energy Ecb Conduction band level (for the bulk)Ecs Conduction band level (for the surface) EF Fermi energy Eg Band gap energy EH Polaron binding energy Ekin Kinetic energy Eop Energy of longitudinal optical phonon Ep Polaron binding energy Ephonon Phonon energy Et Trap energy Evb Valence band level (for the bulk) Evs Valence band level (for the surface) e Unit charge eVs Energy of band bending hkl Miller indices F Faraday’s constant f Frequency fLN Lognormal distribution function g Mean grain size gRDF Radial distribution function G Conductance

vii

Ggas Conductance after exposure to a test gas Gair Conductance in dry synthetic air Ha Adsorbed hydrogen Hi Hydrogen at interfacial sites of the metal-insulator interface h Planck’s constant

Planck’s constant/2I, I0 Intensity of electromagnetic wave I Currenti Current corresponding to one unit charge K Dimensionless apparatus constant k' Extinction coefficient kads, kdes, ki Rate constants kB Boltzmann’s constant kwv Wave vector l, lAl Length of a conductor ltp Distance from transfer pipe to substrate m Oxygen sensitivity constant N Number of carriers Ne Number of electrons Nh Number of holes NH Number of hydrogen atoms per unit area n’ Refractive index nuc Number of electrons per unit cell n Concentration of electrons nr Order of reflection n1/f Free carrier concentration P Polarization due to electrical field P’,P’’,P[O2] Oxygen partial pressure p Concentration of holes pdipole Equivalent dipole moment R ReflectanceR ResistanceRg Gas constant r Radiusrpolaron Polaron radius SG, SI, SR, SV Power density spectrum SI,th Power density of current noise under thermal equilibrium SI,sn Power density of current shot noise SN Power density of generation-recombination noise Sp Surface areaSs, St Sensitivity for sensor SV,th Power density of voltage noise under thermal equilibrium SV,1/f Power density of voltage noise with 1/f characteristics S Power density spectrum of

T TransmittanceT Absolute temperaturet Time for an electromagnetic wave through a specific materialtf Transfer time of an electron tr Response timeV VoltageVp Volume of a particle Y AdmittanceZ Impedancez Number of nearest neighbors 2a Diameter of aluminum channel

GreekRgas, Rair Variance of the sensor resistance

2N Variance of the number of charge carriers square 2 Variance of the quantity 2

Absorption coefficient 1/f Empirical 1/f noise parameter

Susceptibility0 Permittivity of free space

Exponent depending on optical transition p Polaron density in polarons/particle

Geometric mean diameterComplex dielectric function

0 Static dielectric constant High frequency dielectric constant Wavelength

µ Mobilityµn Mobility of electrons µp Mobility of holes

Angle of incidence Concentration of test gas in ppm

, Al Resistivity, 0 Conductivity* Conductivity constant sd Geometric standard deviation

o Operating temperatureA Sintering temperatureg-r Time constant for the generation-recombination process

Angular frequencyReducing agent

viii

ix

List of Acronyms and Abbreviations

AGD Advanced gas deposition CCD Charged coupled device DFA Discriminant factor analysis DWT Discrete wavelet transform EDS X-ray energy dispersive spectroscopy Emf Electromotive force ERATO Exploratory Research for Advanced Technology program ERDA Elastic recoil detection analysis FFT Fast Fourier transform ITO Indium tin oxide LGP Liquefied petroleum gas NFL The Studsvik Neutron Research Laboratory MISCAPS Metal-insulator-semiconductor capacitors MISFETS Metal-insulator-semiconductor field effect transistors MOSFET Metal-Oxide-Semiconductor field effect transistors PCA Principal components analysis SAD Selected area diffraction SEM Scanning electron microscope SLAD Studsvik Liquids and Amorphous materials Diffractometer TEM Transmission electron microscope XPS X-ray photoelectron spectrometry XRD X- ray diffraction

x

xi

Contents

1 INTRODUCTION .................................................................................11.1 Sensors in general.........................................................................11.2 Survey over solid state electronic gas sensors..............................21.3 Functional principles of semiconductor sensors...........................61.4 The crystal structures of tungsten trioxide (WO3)......................101.5 Prior work on gas sensing with WO3..........................................111.6 Scope of this work......................................................................11

2 SAMPLE PRODUCTION AND CHARACTERIZATION................132.1 Advanced reactive gas deposition ..............................................132.2 Electron microscopy...................................................................152.3 X-ray diffraction.........................................................................222.4 X-ray photoelectron spectrometry..............................................222.5 Elastic recoil detection analysis..................................................232.6 Neutron scattering ......................................................................232.7 Noise as a diagnostic tool ...........................................................242.8 Design and characterization of sensor devices ...........................282.9 Electrical and optical characterization........................................33

3 RESULTS AND DISCUSSION..........................................................413.1 Microstructure and composition analysis ...................................413.2 Conduction noise as a tool for gas sensing.................................513.3 Gas sensing properties at constant operating temperatures ........533.4 Temperature modulated gas sensors...........................................573.5 Results on optical and electrical characterization.......................58

4 CONCLUSIONS .................................................................................63

5 Elektriska egenskaper hos nanokristallin WO3 för gassensortillämpningar .................................................................................65

6 APPENDIX .........................................................................................68

ACKNOWLEDGEMENTS..........................................................................79

REFERENCES .............................................................................................81

xii

1 INTRODUCTION

During the last decades there have been huge investments and developmentswithin the area of nanoparticle based devices. There is interest within thin and thick film technologies for these nanoparticle films, with gas sensors andcatalysts being examples of applications.

There are several reasons why research and development on nanoparticleshave become an interesting area. One example is the large surface to volumeratio of a nanoparticle, where the ratio between surface Sp and volume Vprapidly increases with decreasing particle radius r as

r1

VS

p

p (1.1)

A football can be used to illustrate the effect of surface area fraction. A ball with a radius of about 0.15 m has a surface to volume ratio of 20 m-1, while for a nanoparticle with a radius 5 nm the ratio is 600 000 000 m-1.Nanoparticle films can be porous, with the degree of porosity depending on the fabrication parameters. Due to the porosity of the film, there will be a large surface exposed to the ambient gases. If there are chemical reactionscatalyzed by the surface, the large surface will enhance the amount ofreactions per unit time by several orders of magnitude, so that sensor applications become feasible.

Other important and interesting areas for nanoparticle films are for instancewithin optical, mechanical and data storage applications.

1.1 Sensors in general What do we actually mean by the term “sensor”? According to Webster's New World Dictionary of Computer Terms 1 the definition is:

“Device to detect and measure physical phenomena such as temperature,stress, heartbeat, wind direction and fire. Sensors translate physical stimuliinto electronic signals that may for example be input into computer”.

1

2

The canary bird has a very high sensitivity for carbon monoxide. The drawback is that the answer or output is the death of the bird. The sensor thus has a very poor repeatability. A few more criteria should therefore be added for an ideal dedicated sensor, which should be chemically selective, reversible, fast, highly sensitive, durable, highly resistant to contamination, simple to operate, small, easy to fabricate and be inexpensive to produce. Improvements of chemical sensors with respect to sensitivity and selectivity are of large interest for the industry as well as in our everyday life. In the industry, where hazardous gases may be present, the employees would be warned in an early stage so that the risk of exposure could be diminished. Determinations of exact concentrations of process gases or reaction gases are of great importance for many applications.

There is a large variety of examples where inexpensive and sensitive sensors could improve the quality and safety of our daily life. A well-known example is the fire alarm. A sensor capable of sensing the freshness of products in a refrigerator is another example.

There are two different kinds of sensor technologies. Considering the first one, gases can be investigated with flame emission spectroscopy, mass spectrometry, gas chromatography and paper detectors. This method requires either expensive equipment and/or well-experienced operators, and it is difficult or expensive to detect the concentrations at different sites. Even if the answer or the outcome of the investigation is very accurate, this type of sensor or equipment does not fulfill many of the criteria for an ideal sensor. The other method relies on detecting a chemical or an electrical response by use of a small sensor. Such sensors are easy to place at different sites, and the response can be monitored in a convenient way by using, for example, voltmeters and a computer. The latter type of sensors can be further divided into two groups, i.e., solid-state electronic sensors and electrochemical sensors. Since this work concerns the gas sensing properties of WO3nanoparticle films a short survey of solid state electronic gas sensors follow.

1.2 Survey over solid state electronic gas sensors Solid-state electronic sensors are widely used, and there are several well-known sensors belonging to this group. A calorimetric gas sensor can be used for the detection of combustible gases in air. The concentration of the specific chemical species is monitored and determined by the temperature change caused by the species to be detected. An example of such a device is the Pellistor gas sensor, which employs a thin platinum wire embedded in porous alumina-containing metal catalysts 2. The wire serves two purposes:

to electrically heat the sensor and the surrounding refractory alumina bead tothe wanted running temperature, and to measure the change in temperatureby detecting resistivity change. The construction of a Pellistor gas sensor is shown in Figure 1.1 3. A limitation of this sensor lies in the loss insensitivity in environments containing catalyst poisons and inhibitors.

Figure 1.1 Principle of the Pellistor gas sensor 3.

Another type of solid state sensor is frequently used in cars for detection ofthe ratio of air to fuel in order to adjust the injection for optimizing the combustion. The schematic of this sensor is shown in Figure 1.2 4 and belongs to the group of electrochemical sensors. (Sensors employed to detect the oxygen to fuel ratio at the stoichiometric point (the so called -point with air:fuel=15:1) is often called a -sensor.) A sample chamber and a referencechamber are separated by an oxygen ion conducting solid electrolyte. The surfaces next to the gas chambers are covered with porous metal electrodes,often of Pt, to enable the measurement of a potential difference between thechambers.

The electron reactions can be described as

eelectrolytO2Pte4gO 2-2 (anode) R (1.1)

Pte4gOeelectrolytO2 -2

2 (cathode) R (1.2)

These reactions represent an oxygen pumping principle, and this results in anelectromotive force according to

PPln

F4TR

Emf g (1.1)

3

where Rg is the gas constant, T the absolute temperature, F Faraday’sconstant, and P’ and P’’ represent the partial oxygen gas pressures in the reference chamber and sample chamber, respectively. The output of the sensor is the potential difference between the electrodes, which is related to the ratio of the partial pressure of oxygen in a reference gas and the analyzedgas 5.

Figure 1.2 Schematic view of an electrochemical oxygen sensor 4. The pressures of the two sides of the electrolyte are denoted P’ and P’’.

A third type of solid state sensor is the field-effect gas sensor of which there are two different types, viz. metal-insulator-semiconductor capacitors (MISCAPS) and metal-insulator-semiconductor field effect transistors (MISFETS). The semiconductor device normally consists of silicon (semiconductor) and silicon dioxide (insulator) and is normally called a MOS-device (Metal-Oxide-Semiconductor). Pd-MOSFET devices withhydrogen sensitivity were reported in 1975 6. They consist of an ordinarycapacitor or transistor with a 100-200 nm thick Pd gate electrode layer.Catalytic dissociation of hydrogen on the Pd causes the device to respond.The device exhibits sensitivity to hydrogen and hydrogen-containingmolecules. The adsorbed hydrogen (Ha) on the Pd surface diffuse to the interfacial sites of the metal-insulator interface (Hi). The diffusion through the metal layer is rapid (of the order of µs) 7-10, and the metal surface and the metal-insulator interface can therefore be assumed to stay in equilibrium.Hence the reaction of the hydrogen atoms can be represented simply as

Ha Hi R (1.3)k1

k2

4

where k1 and k2 are the rates for the different reactions. The device responds with a voltage drop V as the sensor is exposed to the gas under analysis, andthe voltage drop is related to the concentration through

0

dipolepNV H (1.2)

where pdipole is the equivalent dipole moment induced by adsorbed thehydrogen atom at the interface, NH is the number of hydrogen atoms per unit area at the interface, and 0 is the permittivity of free space.

A last example of a solid state sensor is the semiconductor gas sensor. Changes of low temperature conductance of a semiconductor due to changes in the composition of the surrounding gases were reported in 1953 11. The principle was used in 1962 to produce the first chemical sensors. They were based on polycrystalline ZnO between two metal electrodes and were used to detect reducing gases in air. The devices are often referred to as “oxidesensors”, “metal oxide sensors” or “ceramic gas sensors”.

A very popular semiconductor gas sensor is based on work by Taguchi and employed SnO2 12. The idea of the sensor is as follows: A bead of metaloxide is heated in an oxygen environment. The oxygen is adsorbed on thesurface of the metal oxide, and the oxygen adsorption/desorption reachesequilibrium. If a two-point resistance measurement is carried out across the sensor, a characteristic value for the resistance is detected. As the sensor is exposed to a sample gas, different chemical species will also be absorbed onthe metal oxide surface. The different adsorbed molecules react with preadsorbed oxygen and reaction products are created. Electrons are thenreleased and remain in the metal oxide so that they contribute to the conduction thereby decreasing the resistance. The resistance change is hence a measure of the concentration of the specific chemical species reacting with the oxygen. Figure 1.3 shows a schematic view of a Taguchi gas sensor 13.The Taguchi sensors are used for instance as alcohol sensors and as gas alarms for domestic applications.

Figure 1.3 Schematic view of a Taguchi gas sensor 13.

5

1.3 Functional principles of semiconductor sensors During the last decades, semiconductor sensors have been subject to extensive research. This thesis concerns the properties of the semiconductoroxide WO3 for gas sensing applications, and therefore a short introduction to semiconductor gas sensors will follow.

Semiconductor gas sensors are based on metal oxides with wide band gaps. A variety of different sensing materials have been studied over the years.SnO2 has been the most investigated material and was reviewed by Lantto in 1992 14. The first devices were based on thick films of this material. Thereare now commercial devices available that use resistance variations of the oxide for detection of low concentrations of flammable or toxic gases. Someexamples of gases are CH4, liquefied petroleum gas (LPG), H2, CO, H2S and NOx 15.

Semiconductor oxides are used for two different types of gas sensingapplications. One is where the partial pressure of oxygen is to be determined.The material that has been used commercially for this application (in the -sensor) is TiO2. Depending on material, the sensor may be operating at different temperatures. The sensor may operate at a high temperature intervalof 700 C and up; the mechanism responsible for the detection is then bulk conductance effects. Some examples of used materials are TiO2, Cr2O3 and Ga2O3. The relationship between oxygen partial pressure and electricalconductivity is given by

m12

B

A* ]P[OTk

Eexp (1.3)

where is electrical conductivity, * is a constant, EA is activation energyfor conduction, kB is Boltzmann’s constant, P[O2] is oxygen partial pressure and m is an oxygen sensitivity constant depending on the dominant type of bulk defect involved in the reaction between oxygen and the sensor. Othersensor materials, such as ZnO, SnO2 and Mn2O3, work better at low temperatures, such as below 500 C, and then the surface conductance is responsible for the detection of oxygen. The second application involves situations where the oxygen partial pressure remains constant and theconcentration of a minor constituent is to be determined. Some examples of such gases are H2, CO, CH4 and H2S 16.

Possible mechanisms behind the gas sensing response of an n-type metaloxide semiconductor are oxidation or reduction of the active sensor material,

6

ion exchange, direct gas adsorption and surface reactions with adsorbed gases 17. All of these mechanisms may be present, but for semiconductormaterials the observed sensor effects are dominated by direct gas adsorption and surface reactions with preadsorbed molecules. Depending on whether the sensor material is of n- or p-type and whether the species are reducing or oxidizing agents, the sensor characteristic will be different. Taking forexample an n-type sensor material and a reducing sample gas, there will be preadsorbed oxygen ions on the surface of the sensor. The gas will then react with oxygen-ions to form neutral molecules, leading to electron transfer tothe sensor material and a resulting decrease of the resistance 18. A moredetailed description of the phenomena is found below.

Electrons from the semiconductor are transferred to the surface and thenionize the oxygen adsorbates to form O2

- and O-. Thereby a negative chargeis generated on the surface. The surface layer of the semiconductor willtherefore be depleted of electrons, and a so-called depletion layer will be created 19.

Figure 1.4 Schematic diagram showing the surface layer and the correspondingelectron band structure 19. Adsorbed oxygen on the surface gives rise to a depletionlayer of electrons. The symbols are defined in the text.

This is shown in Figure 1.4, where Ecb and Ecs are the conduction band energy levels for the bulk and surface, respectively, EF is the Fermi energy,Evb and Evs are the valence band energy levels for the bulk and surface respectively, Et are trap energy levels of electrons at surface states due to theadsorbed oxygen and eVs is the height in energy of the band bending at thesurface, d denotes the distance from the surface and dd corresponds to the depth of depletion. The band gap energy Eg is given by the difference between Ecb and Evb. For WO3, the valence band is based on O p states while the conduction band is dominated by W d states 20. If the surface belongs to an ideal bulk material (large d in Figure 1.4) without grain boundaries, the

7

influence of the depletion layer is of small or no importance for theconduction along the surface. If, however, the active part of the sensor device is very thin, i.e. the thickness is comparable to dd, the depletion layerwill affect the whole conduction to a large extent.

Therefore, in order to achieve a surface controlled conduction mechanism inthe sensor, the sample thickness should not be much larger than the depletion layer. When considering a polycrystalline sample, each interface between grains will give rise to a band bending. The electron band structure along theline drawn in Figure 1.5a is shown in Figure 1.5b.

The depletion layer gives rise to a Schottky potential barrier. At anintergranular contact, the conduction is restricted by this Schottky barrier since the electrons have to overcome the energy barrier, eVs. The height of the barrier is decreased as the sensor is exposed to a reducing gas, which results in a decrease of the resistance. The Schottky barrier height is usuallyof the order of 0.1 eV for metal oxides 21.

The conduction may be described by the relation 13,22

TkVexp

B

s0

e (1.4)

where 0 is a factor depends on the mobility and contact area. The effect ofthe depletion layer increases as the size of the grains becomes smaller and ofthe same order of magnitude as the depletion layer width. Therefore, in a polycrystalline sensor, the resistance and the sensitivity depend on the sizes of the crystals or particles and the contact area between the particles 21,23.

Figure 1.6 illustrates three cases with different influence on the depletionlayer, inspired by Williams 22. Figure 1.6a shows a situation where the area of the depletion zone at the contact is less than the contact area. Thedepletion layer extends into the grain to a depth marked by the dashed line. The resistivity is about that of the undepleted region in the center of the neck. A closed neck is illustrated in Figure 1.6b. Here the depletion layersfrom the surfaces overlap, thereby resulting in a higher resistance in thecenter of the constriction. The situations described by Figure 1.6 a and b imply that the porous structure would respond to a gas in the same way as a thin film sensor.

8

Figure 1.5 Schematic view of a three-grain structure (a), together with the corresponding electron band structure (b). The band structure shown in (b) is along the line drawn in (a).

Figure 1.6 Schematic illustrations of three types of intergrain constrictionconditions. a) Open neck, b) Closed neck and c) Conduction limited by the point of contact.

Figure 1.6c describes a case that is applicable to a porous material but not to a thin film situation. Here the conduction is limited by the point of contact, and can be represented by Eq. (1.4). Sensors are often annealed to obtain the right conditions for surface controlled conduction and to optimize the influence of the specific gas under investigation. The transition from open neck, neck-controlled, to point-of-contact limited conduction of metal-oxidegas sensors was investigated by Wang 24. Calculations showed that the sensitivity to adsorbed gas increased rapidly as the grain size became smallerthan 40 nm, at which point the conduction mechanisms for neck-controlledand point-of-contact limited conduction merged into each other. The

9

increase of sensitivity with decreased grain size is in agreement with experimental results 23.

Sensors are not normally under equilibrium conditions but in a steady statesituation. Oxygen is adsorbed at a certain rate and is removed at the samerate by surface reactions with a reducing agent . The oxygen adsorption can be described by the rate equations13

adsOgO 223k R (1.4)

10

adsOadsOe 24k

2 R (1.5)

adsO2adsOe 5k2 R (1.6)

and the reaction of the gas to be detected can be represented by the reactions

egROadsO 226k

R (1.7)

egROadsO 7kR (1.8)

where the k’s are rate constants for the reactions 13.

The gas sensing material investigated in this work is WO3, which is amaterial with a large number of stoichiometric and substoichiometricstructures. In the next chapter, the crystallographic structures of WO3 are presented.

1.4 The crystal structures of tungsten trioxide (WO3)The structural configuration of the crystals of WO3 is perovskite – like withcorner sharing oxygen octahedra enclosing the tungsten atoms, as shown in Figure 1.7 25. The distortions from the ideal cubic perovskite – likestructures results in a number of different WO3 phases. The amount of distortion is temperature dependent and a WO3 crystal changes the phase, as cooling down, from tetragonal (900 C) – orthorhombic – monoclinic – triclinic and finally at -189 C to a low temperature monoclinic phase. As the displacements are small, there are sometimes difficulties in distinguishingbetween the different structures. WO3 easily forms sub – stoichiometricstructures, such as WOx with 2.5<x<3, also known as Magnéli phases.

Figure 1.7 Schematic illustration of the arrangement of the octahedra in a slightly substoichiometric WO3 crystal.

1.5 Prior work on gas sensing with WO3

Tungsten trioxide is a very interesting material, which attracts much current interest owing to its various possible applications. It is widely used in different thin film technologies, for example as the electrochromic film in smart windows 25,26 and in thermal control devices 27.

The first report on WO3 for gas sensing came already in 1967 in the work ofShaver 28. His studies concerned Pt-activated WO3 films and showed that the conductivity of the film increased by one order of magnitude on exposure to H2. During subsequent years, several reports on WO3-based sensors have been published. It has been found that WO3 can be used for detection of a variety of different gases such as H2, CH4, NH3, CO, NO, CHOH, O2, H2S,NO2, C2H5OH, O3, CH3SH, (CH3)3N, SO2, Cl2. Appendix 1 provides a shortsurvey of published results, organized according to the test gases.

1.6 Scope of this work The present work concerns mainly gas sensing properties of WO3. Advanced gas deposition was employed to make thick (~2-20µm) nanocrystalline filmsof WO3. This technique is capable of yielding layers comprised ofnanoparticles with sizes of ~5 nm. Microstructural determinations of these films were carried out using scanning and transmission electron microscopyas well as X-ray diffraction and neutron scattering techniques. Stoichiometryand atomic states were analyzed by use of elastic recoil detection analysisand electron spectroscopy for chemical analysis methods. The gas sensing

11

12

properties were enhanced as a consequence of the small particle size. Conduction noise was used as an analytical tool for quality assessment of thin films. A new method is also presented for gas sensing by use of conduction noise upon exposure of ethanol vapor. Temperature modulated WO3 sensor properties were investigated and the recorded resistance data was further analyzed using mathematical transformations for qualitative analysis. The transport properties, including optical in the near infrared region and thermal dependence of DC conduction, of thick tungsten oxide films were also investigated.

2 SAMPLE PRODUCTION AND CHARACTERIZATION

2.1 Advanced reactive gas deposition Already in 1908 Svedberg described a method for producing colloidalsolutions 29. This work provided convincing evidence on the validity of Einstein’s and von Smoluchowski’s theory on Brownian motion 30. In the middle of the 1900s the research on small particles accelerated. The developments within gas evaporation were promoted around 1981 with the Exploratory Research for Advanced Technology program (ERATO) initiatedin Japan. One outcome of these projects was the development of a deposition technique referred to as “gas evaporation” 32 or gas deposition. The original method of producing nanoparticles using gas evaporation methods is heating a metal in an inert atmosphere. As the material starts to evaporate, anoversaturated vapor zone is built up above the molten metal surface and the vapor is cooled by the inert gas, resulting in nucleation and growth of ultrafine particles. The particle size can be controlled by adjusting the metalvapor and the total pressure, respectively and thereby the growth conditions of the particles. Gas evaporation has become a leading method for production of high quality nanoparticles 33,34.

The classical technique of gas evaporation was introduced in 1976 byGranqvist and Buhrman. They found that isolated nanoparticles prepared byinert-gas evaporation followed a lognormal size distribution 31, which couldbe expressed by

2ln2lnlnexp

ln21

sdsdLN

xxf2

(2.1)

where sd is the geometric standard deviation and is the geometric meandiameter.

13

14

Lognormal distributions are usually explained as the outcome of Brownian coagulation models. These models are based on the theory of Smoluchowsky and treats coagulation in a closed system. In an advanced gas deposition unit, which has been employed in this work, the case is different though, and the gas evaporation takes place in an open system with particles constantly being removed and new material/vapor being added. The time spent in the growth zone is of importance. The primary mechanism for growth is vapor condensation and the time that a particle stays in the growth zone is depending on particle diffusion and drift through this growth zone. It has been shown that the particle size is a power function of growth time and this latter exhibit lognormal distribution for a mean passage time determined by the gas flow conditions. Consequently, the particle size distribution is also lognormal 35.

Figure 2.1 shows an illustration of the advanced gas deposition unit. Prior to evaporation the whole unit was evacuated to ~3*10-2 mbar. To obtain WO3nanoparticle films, we modified the original gas deposition arrangement and used a reactive ambient gas, i.e. synthetic air with a flow of ~10 l/min. The surface of the tungsten pellet (99.95% W, CERAC inc., USA) was oxidized and subsequently the oxide sublimed into vapor phase. The particles were formed in the oversaturated vapor zone. The particles grew in a growth zone in the lower chamber and, due to a pressure difference (P1 20 mbar, P2 0.15mbar), convection and a gas flow, they were transported into the deposition chamber, which was connected to a vacuum pump. The particles were collected by a narrow transfer pipe of 3 mm in diameter, for transport into the deposition chamber, or removed by an exhaust pipe to a vacuum pump. All collected particles were thereby obtained from a small region of the vapor zone, and therefore they were formed under approximately the same conditions, which result in a narrow size distribution.

The nanocrystalline films were fabricated at two different sublimation temperatures and different distances from transfer pipe to substrate; the WO3films were produced at ~1360 K (will further be denoted A) and at ~1470 K (denoted B). Samples are further, classified by their fabrication set up, i.e. their distance from transfer pipe to substrate (ltp), according to A1, A2 and A3 for ~3 cm, ~15 cm and ~25 cm, respectively. The width of the deposited line was between 7 –25 mm depending on the distance from transfer pipe to substrate.

The conditions for production of WO3 nanoparticles do not allow large variations of the pressure in the lower chamber. The particle sizes therefore cannot be varied to a large extent.

Figure 2.1 A schematic view of the equipment.

The substrates were normally attached to a x,y,z -table which alloweddifferent patterns to be made. The film thickness was determined with a WYKO NT-2000 interferometer or a with a Tencor Alpha-Step 200 mechanical stylus instrument. For production of samples for optical characterization a rotating plate was attached to the x,y,z-table. Therebylarger areas could be deposited and large thickness variations of the filmscould be avoided.

2.2 Electron microscopyA microscope operating with visible light is only capable of resolving detailsdown to ~500 nm. If smaller details are to be investigated, as is oftenrequired in materials science, one can use a beam of electrons instead of abeam of light. The electrons are accelerated over a voltage from a few 100eV to 30 keV, as is common for a Scanning Electron Microscope (SEM), or over a voltage from 100 keV to 1 MV, as is common for a TransmissionElectron Microscope (TEM).

15

Figure 2.2 Different types of signals an incoming electron beam can give rise to in a thin foil. These signals can be used for analysis in an electron microscope 36.

The wavelength is shortened dramatically and is no longer the limiting factor for the resolution of the microscope. Instead error sources inherent in the lenses, instabilities of acceleration voltage and lens currents, and mechanicalinstabilities become limiting factors. Typical optimum resolutions of the used microscopes are around 1-2 nm for SEM and around 0.15-0.3 nm for TEM. As the electrons travel through the material, their interaction willcause a variety of different signals that can be detected with various detectors. Figure 2.2 shows a schematic view of the different signals 36.

2.2.1 Scanning Electron Microscopy One of the most frequently used methods in surface analysis is scanningelectron microscopy. The incoming electrons, focused to a small spot, are scanned over the surface and interact with the material near the surface. The electrons are collected above the surface in a photomultiplier and are presented onto a computer screen; see Figure 2.3 37. Thus the surface can be imaged, and the elemental composition in the sample can be determinedusing X-ray energy dispersive spectroscopy (EDS).

The electron beam broadens due to the interaction with the material andgives rise to a typical pear formed interaction volume. This is illustrated in Figure 2.4a 38, depicting the different types of electrons and the regions where they are created. In this work a LEO 1550 Gemini instrument hasbeen used for imaging. An inlens detector was used for detection of secondary electrons and an example of a WO3 surface of a gas sensor is seen in Figure 2.4b. The collected secondary electrons originate from a volumeclose to the surface down to about 5 nm.

16

Figure 2.3 A schematic image of an SEM 37.

Figure 2.4 a) Interaction volumes in a typical SEM sample, and different features that can be used for characterization of the material 38. b) SEM micrographdepicting the surface of a WO3 sensor.

2.2.2 Transmission Electron Microscopy In transmission electron microscopy, the sample is exposed to a beam ofhighly accelerated electrons. When the electrons are passing through the sample, they interact with the material; see Figure 2.2. Since the sample is thinner than 100 nm, the beam broadening is small in comparison to that of abulk specimen. If the TEM is equipped with an EDS unit, the characteristicX-rays can be analyzed and the composition can be determined. Afterinteraction with the sample, the beam travels through the column and produces an image on a fluorescent screen, on a photographic plate or on a

17

post-column charged coupled device (CCD). A schematic view of a TEM is shown in Figure 2.5 39.

Figure 2.5 Schematic view of a TEM 39.

The elastically scattered electrons with small diffraction angle are coherentlyscattered and can therefore be used to determine the structure of the sampleby use of selected area diffraction (SAD). When an incoming wave front isinteracting with a crystal, every atom in the lattice will be a source of a coherent scattered wave. The scattered waves interfere with the waves coming from the neighboring atoms. The outcome is Bragg’s law ofdiffraction, which reads

rhkl nsind2 (2.2)

where dhkl is the interplanar distance, is the angle of incidence, nr the orderof reflection and the wavelength of the incoming wave. dhkl is for a cubic material described by the relation

222hkllkh

ad (2.3)

18

where a’ is the lattice constant and h, k and l are the Miller indices. Figure 2.6 shows an illustration of Bragg diffraction.

Figure 2.6 The Bragg description of diffraction. dhkl is the interplanar distance.

The difference between a microscope operated in image mode and diffraction mode is shown in Figure 2.7 36. The diffraction pattern is formedin the back focal plane of the objective lens. By using a selected area diffraction aperture in the first image plane, the sample area used for building up the diffraction pattern can be selected. Figure 2.8a depicts animage of a TEM sample and Figure 2.8b depicts a selected area diffractionpattern indicating a tetragonal structure of nanocrystalline WO3.

This work used a JEOL 2000 FX operating at 200 kV with an attached EDSunit and a FEI Tecnai F30 operating at 300 kV with attached Gatan EnergyFilter.

The most important factor for obtaining good TEM results is to have a verythin sample. Depending on the analysis, the optimal sample thickness canvary from 5 nm up to 100 nm. A convenient way to produce good TEMsamples is to use holey carbon copper grids. A copper grid is covered with a 10-20 nm thick carbon film, which supports the sample. The particles can be deposited directly onto the grid using the AGD-unit. A WO3 film can be deposited onto sodium chloride (NaCl) crystals. The NaCl can be dissolved in water and the film is then adsorbed to the grid.

19

Figure 2.7 Schematic ray diagram of a TEM in a) selected area diffraction modeand b) image mode 36.

Figure 2.8 (a) TEM micrograph showing WO3 nanoparticles deposited onto a carbon coated copper grid. (b) Selected area diffraction pattern from a tetragonal WO3 nanoparticle film.

Alternatively, the film can be scraped off from a substrate and then ground to a fine powder. The powder is then added to a liquid, and a small drop ofthe liquid together with the powder is placed on the carbon copper grid. Theliquid is evaporated and the particles remain on the carbon film. This methodwas used in this work for size distribution measurements.

If the sample has a multilayer structure and the different layers are to be imaged, a cross section sample is needed. In this case the sample preparationinvolves a large number of steps, and Figure 2.9 shows a schematic

20

illustration of them. A substrate is cut and the two pieces are glued together.After drying, the substrates are polished until they fit into a brass tube of 3 mm diameter. The tube is cut into ~1 mm slices and then polished down to~100µm. At this stage the slice is placed in a dimple grinder and a crater is produced.

Figure 2.9 Schematic illustration of the first steps involved in production of a TEM cross section sample. The steps are explained in the text.

When the thinnest parts of the slice are 10 – 20 µm, the dimpling is completed. Finally, the sample is placed in an ion mill, specifically a Gatan PIPSTM. The sample is milled from both sides with argon ions accelerated over a voltage of ~4 kV and with an inclination angle of 4-6 degrees. In order to obtain good conductivity and avoid charging of the sample, a thin layer of carbon is sputtered onto the upper side of the sample. Figure 2.10shows a TEM micrograph of a WO3 cross section. The image depicts particles being approximately 4 nm in size. Lattice fringes can be observedwithin the particles.

Figure 2.10 Cross sectional TEM micrograph of a WO3 sample, showingnanoparticles as well as lattice fringes. Above the nanoparticles a layer of glue is seen.

21

2.3 X-ray diffractionA widely used method to characterize crystalline materials is X- raydiffraction (XRD) 40. An illustration of the diffraction phenomena and Bragg’s law was shown in Figure 2.6 and represented by Eq. 2.1. CuKradiation is often used; its wavelength is 1.5406 Å. The technique can give information about crystallinity, orientation, composition, internal lattice strain and particle size.

XRD was used for determination of grain size and crystal structure of thickWO3 samples, for which the amount of particles can provide a high enough signal. The mean grain size g was determined by use of Scherrer’s formula40

cosBKg

d (2.4)

where K 0.9 is a dimensionless constant and Bd is the broadening of thediffraction line measured at full width, half maximum in radians. The XRDmeasurements were carried out using a Siemens D5000 diffractometer.

2.4 X-ray photoelectron spectrometry X-ray photoelectron spectrometry (XPS) is a frequently used method forinvestigation of solid materials, surfaces and thin films 41. By using XPS the chemical composition of the sample can be determined as well as the atomicstate of the different constituents. The method is very surface sensitive, and the collected signal originates from a layer not farther than ~2 nm from thesurface. The sample is exposed to monochromatic Al K X-rays ofwavelength 8.3401 Å or of energy ( ) 1.4866 keV. Due to the photoelectrical effect, electrons are emitted with a kinetic energy

Bkin EE (2.5)

where EB is the binding energy for the actual electron energy level. The kinetic energy Ekin is measured, and the binding energy can be calculated. Each atom is characterized by an individual set of electron binding energies. This XPS method is used in order to obtain information of the stoichiometryof the WO3, the oxidation state of W in pure WO3 films and the content and

22

23

oxidation state of Pd in WO3-Pd films. In this work, a PHI 5500 ESCA instrument and a PHI Quantum 2000 instrument have been used.

2.5 Elastic recoil detection analysis There are a variety of different types of ERDA instruments but in this work specifically a Time-of Flight Energy Elastic Recoil Detection (ToF-E ERD) system was employed 42,43. Heavy ions, specifically iodine ions, of high energy (~21 MeV) were recoiled at an angle of 45 after interaction with the sample. ToF-E ERD measurements are based on two carbon foil time detectors and one energy detector, whereby the mass of the scatterer could be deduced. An advantage of ERDA compared to many other methods is that light and heavy elements can be detected with a fair sensitivity. With the method the composition and depth profile of samples can be determined. The method can also give an estimation of the density in atoms/m2 for the sample under investigation. The characterization of WO3 nanoparticle samples was carried out at the Tandem accelerator laboratories at Uppsala University.

2.6 Neutron scattering In addition to X-ray diffraction, neutron scattering experiments were carried out on WO3 material produced with the AGD unit. The instrumentation used for the investigation is located at the research reactor facilities at Studsvik AB, where the produced neutrons are used for further experiments and research on neutron scattering within NFL (The Studsvik Neutron Research Laboratory). The advantages of using neutrons instead of X-rays are the shorter wavelength of 1.1158Å and a smaller energy of 65.7 meV of the neutrons. The former is comparable to typical atomic spacings and the latter to the vibrational energies. Thereby a higher resolution for interatomic distances is obtained and the materials under investigation are exposed to smaller radiation dose. A schematic picture of the SLAD - the Studsvik Liquids and Amorphous materials Diffractometer - instrument set-up is shown in Figure 2.11.

Figure 2.11 Schematic view of the neutron scattering instrument.

2.7 Noise as a diagnostic tool Electronic noise means random fluctuations of a physical quantity (e.g.,conductance, voltage etc.) in an electrical system. Noise exists in allmaterials and appears to have different characteristics depending on what mechanism is dominant in each case. Electrical noise is a stochastic signal, often described in terms of a variance 2 and represented as a power density spectrum S [ 2/Hz], where can be a voltage, current, resistance or optical power.

The following relation holds in general for homogenous crystals with ohmicbehavior and small relative fluctuations:

2R

2G

2I

2V

RS

GS

IS

VS

(2.6)

where SV, SI, SG, SR are the power density spectra for the quantities voltage V,current I, conductance G and resistance R, respectively. Some common typesof noise are thermal noise, shot noise, burst noise and 1/f noise. Next follows a short description of different types of noise, taken from work by Bell 44

and Ambrozy 45.

24

2.7.1 Thermal noiseIn homogenous conductors or semiconductors, noise is generated even without current flow. Conduction noise occurs as a result of thermal motionof charge carriers in the material. Current and voltage fluctuations, or noise in thermal equilibrium, are given by the relations

''fC2Tk4YReTk4fS BBth,I (2.7)

and

Y1ReTk4ZReTk4fS BBth,V

2B C

C1Cf2

1Tk4 (2.8)

where Y is admittance, Z is impedance, f is frequency, and C’ and C’’ denote real and imaginary parts of the complex capacitance, respectively. Both C’and C’’ may be dependent on frequency.

2.7.2 Shot noiseThe concept of shot noise originates from Schottky, known for his work onthe vacuum diode 46. Having a plate capacitor in mind, an electron with charge e is emitted from the cathode and travels without hindrance towards the anode in a time tf. This electron transfers a current pulse of magnitudei=-e/tf. The current for Ne electrons leaving the cathode and arriving at the anode is then I=-Nee/tf. The energies of the electrons, which are stochastically emitted, are randomly distributed. They result in a fluctuation of charges transmitted per time unit and therefore also in a currentfluctuation. If the current is amplified and connected to a loudspeaker, the sound is similar to that of the dropping of shots. This motivates the term“shot noise”. The power density spectrum is

2

f

fsn,I 2t

2tsineI2S (2.9)

25

which is the shot noise equation and is the angular frequency.

2.7.3 Generation-recombination noiseConduction fluctuations in semiconductors, due to fluctuations in the number of free charge carriers, is called generation-recombination noise. Theresistance of a homogenous semiconductor of length l and cross section area A is described by

ppenpn NµNµel

pneAlR

2(2.10)

where µn and µp are the mobilities and n and p are the concentrations of electrons and holes, respectively. The number of carriers Ne (electrons) and Np (holes) are not constant but fluctuate due to generation-recombination.Hence the resistance will fluctuate stochastically, and the generation-recombination noise is described by

rg

rg2N f21

4NfS (2.11)

where 2N is the variance of the number of charge carriers square and g-r

is a time constant for the generation-recombination process.

2.7.4 Burst noiseBurst noise is a pulse-type noise, also called “popcorn” noise. The noise appears as a square wave with random changeovers due to hot spots and their fluctuations in the material. The other noise components are superimposed onto the burst noise. The burst noise can be taken as anindicator of a poor device.

2.7.5 1/f noise 1/f voltage noise, SV,1/f, is caused by conductance fluctuations in a resistive sample with homogenous current density exposed to a constant current. It can be described by the empirical relation 47,48

26

Sf

ffV Vfn

VS

1

1/1,

2 (2.12)

where V is the average voltage drop across the sample, n1/f is the free carrier concentration, VS is the volume of the sample with homogenous current density and 1/f is a parameter in the range 10-6 < 1/f < 10-3 for good quality,homogeneous samples. Work by Vandamme has shown that this type of noise can be valuable for analysis and for quality assessment of electronicdevices 47.

2.7.6 Experimental set-upNoise in thin disc-shaped WO3 samples was measured using a probe station.The samples were produced using the advanced gas deposition unit, bydepositing WO3 nanoparticle films on ITO covered glass substrates. Aluminum contacts were evaporated onto the WO3 film to be able to measure the conduction noise. Figure 2.12a shows a schematic view of the samples, where d denotes the sample thickness, dc is the contact thickness, C1 is a contact on ITO and finally C2 and C3 are contacts on the WO3 -film.

The measurement set-up, is shown in Figure 2.12b, used a noise free voltagesource with a resistor in series that was at least 20 times the sampleresistance followed by a cross correlation of two low-noise voltage amplifiers (Brookdeal 5004) with their inputs in parallel and a double-channel Advantest R921E Digital Spectrum Analyzer. The Brookdeal amplifiers had each an input resistance of 100 M . The voltage noisespectra were investigated in the range from 1 Hz up to 100 kHz. The resistances and capacitances of the samples were measured using a Hewlett Packard 4274A Multi-Frequency LCR Meter (100 Hz – 100kHz). The values were used for comparison of the intensity of the calculated thermal noise andmeasured thermal noise.

27

Figure 2.12 a) Schematic view of the contact arrangement of the samples. d denotesthe sample thickness, dc is the contact thickness, C1 is a contact on ITO and finallyC2 and C3 are contacts on the WO3 –film. b) Set-up for voltage noise measurements.

2.8 Design and characterization of sensor devices

2.8.1 Gas sensor characterization at fixed temperature Gas sensing characterization of thick samples (~2-20µm) was carried outusing an experimental set-up shown in Figure 2.13. The temperature of thesample and the partial pressure of the gases were controlled during themeasurement. It was possible to attach three bottles of gas to the system.One gas was synthetic air (20% O2 and 80% N2) used for reference measurements, so the gas response of two different sample gases could bemeasured in one experimental run. The gases were mixed in a blender before insertion into the test chamber. The gas composition and flow werecontrolled by a computer. The computer also controlled the power amplifier,and thereby the temperature of the sensors, and the reading of the multimeter. The gases used for experiments were H2, CO, NO, NO2, SO2 and H2S at various concentration in dry synthetic air.

28

Gasoutlet

Figure 2.13 Experimental set-up for gas sensor measurements.

The sensor devices were constructed according to Figure 2.14 14. Contacts of Au were deposited onto the front side of Al2O3 substrates, and a meanderpattern of Pt, used for both heating and temperature measurement, was printed on the backside. Thick films of WO3 nanoparticles were deposited over the Au contacts.

Figure 2.14 Schematic view of the sensor device showing a) the Pt- resistor on thebackside of the substrate, b) the preprinted gold contacts and c) the covered goldcontacts after deposition of WO3

14.

The films were heated in air at sintering temperatures s in the range 470-870K for 1h. The electrical conductance between the Au contacts was measuredas a function of time by a two-point method upon gas exposure. Duringmeasurement, a voltage of 1 V was applied over the samples and the variation in current was recorded. In Figure 2.15 a sensor response of the conductance versus time can be seen, for a sensor exposed to 10 ppm of H2Sat room temperature.

29

10-10

10-9

10-8

10-7

0 20 40 60 80

Con

duct

ance

[S]

Time [min]

gas outgas in

10 ppm of H2S in synthetic

air at room temperature

Figure 2.15 Conductance response of a sample exposed to 10 ppm of H2S at room temperature.

The gas sensitivity Ss is defined here as the conductance ratio

air

gass G

GS (2.13)

where Ggas and Gair denote the conductance after exposure to a test gas andin pure synthetic air, respectively.

2.8.2 Temperature modulated gas sensorIt is often the case that more sensors are needed, one for each specific gas that could be present in the volume, or at sites, under investigation. Analternative to the fixed-temperature set-up mentioned above could be to have one sensor operating at different temperatures. A set-up for gas sensing was built consisting of mass flow controllers for the gases, a gas chamber and a PC for data collection via a multimeter HP 34401A. The sampling rate was set to 1.33 samples per second. The sensor was designed according to Figure 2.14. The Pt–heater on the back side of the sensor was connected to a circuitthat delivered a square voltage signal between 0.8 and 6 V with a frequencyof 36 mHz. This resulted in a continuous heating/cooling cycle with continuous adsorption and desorption events. The resistance of the sensor was measured and – due to the temperature-dependant variation of reaction rates and activation energies – each gas will give an unique signature to the sensor response. The resistance versus time under exposures to synthetic air, C2H5OH and H2S in different concentrations were measured.

30

31

For temperature modulated sensors the most commonly used method for extraction of features of the signal has been fast Fourier transformation (FFT)49,50. An alternative way to break down the sensor response into representative parameters is a discrete wavelet transform (DWT).

In FFT analysis, the sensor signal is represented by characteristic sines and cosines and the analysis requires a large number of sensor response periods to ensure a good definition of the harmonic peaks in the signal. In this work the data obtained through FFT stem from 540 samplings or ~20 sensor periods in the presence of a test gas. The amplitude of the dc component and the first 4 harmonics were extracted. Higher order harmonics were discarded because of low amplitudes so that they tented to be easily affected by noise.

In the DWT analysis, however the signal is represented by more complicated basis functions called wavelets. The DWT parameters can however be calculated over one single sensor response period, which makes the method to a fast extractor of important features from sensor dynamics. The data obtained through DWT stem from 28 samplings or one sensor period soon after insertion of a test gas. The first 4 wavelets coefficients were discarded since they correspond to very low frequencies and are affected by sensor drift 51. Higher wavelets coefficients, above the 16th, were also discarded since those coefficients correspond to high frequencies and may be affected by noise 52.

The outcome of FFT and DWT can further be analyzed using principal component analysis (PCA) or discriminant factor analysis (DFA) 53. PCA is a linear and unsupervised pattern recognition method. The purpose of PCA is to express the variables of the FFT/DWT sensor response in a new orthogonal coordinate system where the maximum of the variance is collected in a lower number of new variables, i.e. principal components. DFA, on the other hand, is a linear and supervised pattern recognition method. DFA computes the factors to minimize the variance within each class but also to maximize the variance between different classes. The obtained components can be used for qualitative analysis.

In Figure 2.16 a temperature modulated sensor response is shown. The sensor was operating at temperatures in the interval 210-310 C in synthetic air and after 180 s, 20 ppm of ethanol was introduced.

Figure 2.16 Sensor response to 20 ppm of ethanol in synthetic air when the sensor temperature varied between 210 and 310 C. The ethanol was inserted after 180 s.

The gas sensitivity St is defined here as the ratio

gas

airt R

RS (2.14)

where Rgas and Rair denote the variance of the sensor resistance upon exposure of a test gas and in pure synthetic air, respectively.

2.8.3 Conduction noise set-upConduction noise measurements were performed to test the gas sensing properties upon C2H5OH exposures of thin nanoparticle WO3/Au devices. A thin layer of Au (10-15 nm) was evaporated onto glass substrates. The Aulayer was subsequently covered with a thin layer of WO3 nanoparticles, using the advanced gas deposition unit. Figure 2.17 shows a TEMmicrograph showing a cross sectional view of a device. The thickness of the Au layer was about 12 nm and the thickness of the WO3 layer was one particle diameter, i.e. 3 nm.

The substrates were fabricated to obtain a narrow conductor with a width of approximately 0.5 mm, and is illustrated in Figure 2.18a. A four-point measurement set-up with a constant DC current was used. The measuredvoltage fluctuations represent the conductance (resistance) noise of thedevice. Typically, the resistance of the device was 70 and the applied current was 5 mA. The electronic circuit of the set-up is shown in Figure2.18b. The sample and the current generator were placed in a shielded box.The voltage fluctuations were amplified by a FET-input differential

32

preamplifier (80dB) and were sampled by a National Instrument 16-bit A/D converter with an anti-aliasing filter connected to a computer. The noise spectrum was then obtained by Fast Fourier Transformation of the recorded data in the frequency range 0.5- 30kHz.

Figure 2.17 TEM micrograph showing a cross section of a device. The dark area isthe thin gold film (~12 nm). WO3 nanoparticles with a particle size of ~3 nm are seen above this film.

Figure 2.18 a) Schematic view of a WO3/Au gas sensor device. b) Noisemeasurement set-up for gas sensing. The sample and current generator were placedin a shielded box.

2.9 Electrical and optical characterization In order to understand the conduction mechanisms in nanocrystalline WO3films, the optical properties as well as the temperature dependence of the

33

resistance were investigated. When a semiconductor material is exposed toan electromagnetic field, it interacts with the incoming waves. Depending on the material, the effect of the electromagnetic wave will be different andtherefore the response will also be diverse. Beginning at low frequencies, dipolar relaxation, ionic vibrations and finally electronic vibrations will appear if applicable to the material.

The polarization P obtained by an electrical field E is

34

10EP (2.15)

where is the complex dielectric function

phononsfcvedipole1 (2.16)

with contributions of susceptibilities from dipoles, valence electrons, free carriers and phonons.

2.9.1 Optical characterization

For a plane wave of electromagnetic radiation through an absorbing medium,the electric field component E is given by

tcdnic

dkee0EE (2.17)

where is the angular frequency, n’ is the refractive index of the medium , k’ is the extinction coefficient, d is the distance in the material, c is the speed of light and t is the propagation time for the radiation through the material.As the wave propagates through the material, the intensity I0 will decrease according to Beer-Lambert law as

dck2d ee 00 III (2.18)

where is the absorption coefficient.

The material of interest is often deposited as a thin film onto a substrate. Incoming radiation I0 interacting with the material results in reflected (R)and transmitted (T) components. This is visualized in Figure 2.19.

Figure 2.19 The incoming radiation I0 results in reflectance R and transmittance T.

In optical characterization, two different methods can be used. Either theoptical constants, i.e. n’ and k’, of the material are known and R and T canbe calculated. Alternatively, R and T are measured and the optical constantscan be determined.

In a thin film, the optical absorption coefficient is to a good approximationgiven by

TR1ln

d1

(2.19)

according to Hong’s method 54. The optical properties of the nanocrystallineWO3 were investigated by using a Perkin-Elmer Lambda 9 instrumentworking in the wavelength region 0.3< <2.5 µm or a Perkin-Elmer 983 instrument for wavelengths 2-50 µm. WO3 nanoparticle films of thickness 2-7 µm were deposited onto glass substrates and Si-substrates using a rotating sample holder. Thereby a more homogenous thickness was obtained over alarger area, which enabled optical characterization.

2.9.2 Mechanisms of optical absorption The energy gap between the conduction band and the valence band, the so-called band gap Eg, acts as a threshold for interband excitations. Theseexcitations are illustrated in Figure 2.20, where kwv is the wave vector. The direct transition involves a vertical transition whilst for the indirect

35

transitions contributions from phonons, i.e. Ephonon, are required. The bandgap can be derived by use of the equation

36

gE (2.20)

where the exponent depends on the actual type of optical transition. For crystalline semiconductors the exponent is 1/2, 3/2, 2 and 3 for directallowed, direct forbidden, indirect allowed and indirect forbidden transitions respectively 25.

Figure 2.20 Schematic illustration of semiconductor interband transitions: a)Allowed direct transition. b) Allowed indirect transition.

The optical absorption that appears below the bandgap energy is often related to transitions between localized states. The electrons may interact with the lattice vibrations of the material, phonons, and this interaction isreferred to as polarons. The idea of polarons was introduced by Landau in 1933 55. An electron placed in a conduction band of an ionic crystal, willaffect a neighboring electron at a distance r with a force e2/ 0r2. If, however, the ions did not move the force would be e2/ r2. So therefore an electron is affected by a potential barrier

re 0

112 (2.21)

where is the high frequency dielectric constant and 0 is the low frequency dielectric constant. The electron is said to be “self-trapped” or “trapped by digging its own hole”. A displacement created by the electron is so large that the electron get localized or trapped by the potential well caused by the displacement of the surrounding ions 56,57.

The polaron radius can be estimated from the expression 56

00ppolaron

114Eer

2(2.22)

There are different types of polarons, characterized by the lenghtscale of the polaron radius. If the ratio between lattice constant a’ and polaron radiusrpolaron is small, the polaron is defined as a large polaron. If the ratio however is equal to one or more, the polaron is classified as small 58.

There are different theories treating the problem of polarons 59-61 and several reports on polarons in WO3 have been published through the years62-66, to mention a few.

Optical absorption and electrical conductivity are related to each other by the equation

'cnRe

0 (2.23)

where ( ), for small polarons, is given by Austin and Mott 67 according to

TkE8E2

expTkE8

1JaeznBp

p

Bp

22uc

221

2(2.24)

where z is the number of nearest neighbors, nuc is the number of electrons per unit cell, J is the rigid band width, is the Planck’s constant divided by 2and Ep is the polaron binding energy. The thermal energy kBT is at roomtemperature equal to ~25meV. However, the energy of longitudinal opticalphonon Eop is often higher than the thermal energy. By substituting kBT withEop we obtain the small polaron absorption equation

opp

p

EE8E21C exp

2

(2.25)

37

where C is a constant. This function is a peak with maximum at( )max=2Ep.

2.9.3 Temperature dependence of the resistanceElectrical measurements can be used as a tool for characterization andunderstanding of material properties. For temperature dependent measurements, the WO3 nanoparticles were deposited onto patterned ITO substrates; a slit of 2.5 µm separated two ITO contacts and the WO3 filmbridged the contacts. Samples were cooled down to 77 K and the resistance was measured during the slow warming up sequence by use of a HP 3457A multimeter instrument.

The type of temperature dependence of the resistance that the sampleexhibits can give valuable information or support for the chosen polaron theory. The electronic conduction properties of large and small polarons are different to each other; for large polarons the mobilities are often rather high(µ>1 cm2/Vs) and falling with increased temperature, however, smallpolarons move with very small moblilities (µ<< 1 cm2/Vs) but increasing with temperature 68. A typical thermal behavior of the DC resistance withhopping conduction that could be predicted for small polarons is given by

TkETCR

B

Hdc,p exp (2.26)

where Cp,dc is a constant and EH is the polaron hopping energy 69.

The different energies EH and ( )max and their relations are illustrated inFigure 2.21 below. For hopping conduction the activation energy EH is approximately half of the polaron binding energy EP. The transfer from one state to the nearest neighboring state may also appear under influence of a photon but here the energy =4EH 67.

38

Figure 2.21 An illustration of the polaron potentials in energy versus distance. Theactivation energy EH for hopping conduction and for photon assisted hopping = 4EH are shown.

39

40

3 RESULTS AND DISCUSSION

3.1 Microstructure and composition analysis The microstructure and the composition of the WO3 nanoparticle films have been investigated by use of electron microscopy, XPS, ERDA, XRD,neutrons scattering as well as by IR-spectroscopy and 1/f noise measurements.

3.1.1 Size distributionFor gas sensing applications the particle size is of large interest. Figure 3.1shows size distributions for samples produced under similar conditions 70.The particles are deposited onto holey carbon copper grids or onto ITOcovered glass substrates and analyzed using SEM and TEM. The size distribution determined from the primary particles on the copper grid have amean size of about 3.1 nm (denoted “primary”), the size distribution indicated “on surface” originates from SEM images (see Figure 3.2b), and the size distribution indicated “embedded” originates from a cross section sample studied by TEM (see Figure 3.2a). The distributions are different,even though the particles were produced under the same conditions. This islikely due to a difference in agglomeration process for particles on a surfaceor in the cross section of a particle film. The straight lines in the log-probability plot correspond to lognormal size distribution so it follows thatnot only the isolated particles follow a lognormal size distribution but also the “embedded” and the particles on the surface.

41

1 1.01.115

10305070909599

99.999.99

0

on surface

embedded

primary

Particle size [nm]

Prob

abili

ty [%

]

Figure 3.1 Size distributions for samples produced under similar conditions and analyzed in different ways, (see details in the text).

Figure 3.2 a) TEM micrograph depicting a cross section. The dark layer correspond to the ITO film onto which the WO3 nanoparticles were deposited and to the right a thin layer of glue is seen. b) SEM micrograph showing the surface of a sampleproduced under the same conditions like for the film depicted in a).

3.1.2 XPS analysisIn Figure 3.3a, a characteristic XPS spectrum for WO3 is shown. The atomicstates of W and O are marked 71. The stoichiometry was estimated by the ratio between the O1s peak and W4f peak. For an as deposited sample a stoichiometry of WO2.6 was found. After annealing at 300 C the ratio was 2.75. In Figure 3.3b the W4f peaks for an as deposited sample as well as for asample annealed at 600 C are shown. The as deposited sample exhibits a broader peak, shifted to lower energies, corresponding to a mixture of lower

42

oxidation states, which is also indicating a sub-stoichiometric composition.After annealing the atomic states correspond to stoichiometric WO3.

1000 800 600 400 200 0

Inte

nsity

[arb

.uni

ts]

Binding energy [eV]

W 4f7/2

W 4f5/2

W 4d5/2

W 4d3/2

W 4p3/2

W 4p1/2

W 4s

O 1s

O: Auger

a)

40 38 36 34 32 30

As deposited

A=600 °C

Inte

nsity

[arb

.uni

ts]

Binding energy [eV]

b)

Figure 3.3 a) XPS spectrum for WO3 showing a wide energy range. The atomicstates are shown in the figure. b) XPS spectrum depicting the atomic state of W. Theannealed sample exhibits W4f

6+-states. The as deposited sample on the other hand, isshifted towards lower energies due to a mixture of lower oxidation states.

3.1.3 ERDA analysisAs deposited WO3 (A3, as defined in chapter 2.1) was analyzed using ERDAin order to estimate the stoichiometry as well as the density of the films. The WO3 nanoparticles were deposited onto a Si substrate. The raw data of the experiment is shown in Figure 3.4a, where the intensity of the recoiledIodine atoms is depicted in a two- dimensional time versus energy plot forthe WO3 sample. In the plot, the W and O are seen, but also C fromcontamination, Si from the substrate and the scattered I-ions.

Figure 3.4b shows the counts as a function of mass. An estimation of the stoichiometry is obtained by fitting and integration of the peaks originating from W and O, resulting in WO2.6. The density of the film was estimated to ~1.8 g/cm3 which is about 25 % of bulk data. This low density indicates an extreme porosity, adequate for gas sensing applications.

43

102

103

104

105

0 50 100 150 200 250 300

Cou

nts

Mass [atomic weight]

W

I

SiO

C

b)

Figure 3.4 a) Raw data from the ERDA measurements. For each element, thebrighter the core the higher the intensity. b) Counts as a function of mass for a WO3sample.

3.1.4 X-ray diffractionWO3 powder was produced with the AGD unit for neutron scattering experiments. The powder was produced according to the descriptions insection 2.1 (A3). The films were scraped off and the powder/film wascollected. The results were compared with a commercial WO3 powder fromAldrich 72. In Figure 3.5 XRD patterns for the two powders are shown. The mean particle size was, according to Scherrer’s formula, around 5 nm and 22nm for the nanocrystalline sample and for the reference powder,respectively. The structures of the two different powders were different; thenanocrystalline sample was of tetragonal phase and the reference powderwas of orthorhombic phase.

XRD was also employed for investigation of the sensor films. XRD- spectra for as deposited samples (a) and 873 K annealed samples (b) are shown inFigure 3.6. The as deposited sensors, produced according to A, consist of mainly tetragonal phase but the sensors produced according to B also contains a monoclinic structure (A and B defined in chapter 2.1). The particle sizes according to Scherrer’s formula were 6 nm (A) and 10 nm (B),respectively. At 300 K a transition towards monoclinic structure takes place, see further XRD–data in Paper III and V, and after annealing at 873 K onlythe monoclinic structure remains. The particle sizes were after annealing found to be around 25 nm (A) and 40 nm (B), respectively.

44

20 25 30 35 40 45 50 55 60

AGD powderRef. powder

Inte

nsity

[arb

.uni

ts]

Diffraction angle 2 deg

Figure 3.5 X-ray diffraction data for a nanocrystalline WO3 powder produced withthe AGD unit(----) and for a commercial WO3 powder ( ).

24 28 32 36 40

Inte

nsity

[arb

. uni

ts]

Diffraction angle2 [deg.]

*

* Al2O

3

m Monoclinic WO3

t Tetragonal WO3

*

*

t

m

tt

t t

a)

A

B

24 28 32 36 40

Inte

nsity

[arb

. uni

ts]

Diffraction angle2 [deg.]

*

*

*

b)m * Al

2O

3

m Monoclinic WO3

mm

m m

mm

A

B

Figure 3.6 X-ray diffraction patterns for a) as deposited sample and b) annealing at 873 K for 1h. (…) and ( ) corresponds to A and B type of samples, respectively.

3.1.5 Neutron scattering on powder The powder produced with AGD and the reference powder (Aldrich 72) were investigated by use of neutron scattering experiments. The corrected neutron scattering factors for the two powders are shown in Figure 3.7a. The data for the reference powder is shifted upwards for better visualization. The scattering factors are very similar despite the different phases and methodsof fabrication, see previous chapter. The AGD powder shows slightlybroader peaks, which is expected due to the smaller average grain size. The smaller particle size is also indicated at low wave number transfer values (Q-

45

values) in Figure 3.7a. A peak in the scattering factor has an onset for theAGD powder, which is caused by the grains.

0.0

0.50

1.0

1.5

0 2 4 6 8 10

AGD WO3

Ref WO3 1 % H

S [Q

]

Q [1/Å]

a)

-0.4

-0.3

-0.2

-0.1

0

0.1

0.2

0.3

0.4

0 2 4 6 8 10

AGD WO3 1 % H

Ref WO3 1 % H

g RDF

[r]

r [Å]

b)

Figure 3.7 a) Scattering factor and b) radial distribution function obtained fromneutron scattering for the nanocrystalline WO3 powder produced with the AGD unitand the commercial WO3 powder.

The radial distribution functions can be determined, from the neutron scattering data, by Monte Carlo simulations 73. The results are depicted inFigure 3.7b and the coordination numbers can be obtained from the curve. The coordination numbers of W-O and O-O nearest neighbors are presented in Table 3.1 below. There is almost no difference in the coordinationnumbers of the different powders. However, for stoichiometric WO3 one could expect the coordination numbers to be 6 for W-O and 8 for O-O. Therefore it is very likely that the reference WO3 powder was not stoichiometric either. Sub-stoichiometry has on the other hand also been reported earlier in work by Nanba et al. 74. Nanba found as deposited filmsproduced with electron beam vacuum evaporation to be substoichiometricwith a W-O coordination number of 4.5. After annealing at 300 C the WO3was found to be stoichiometric. Additionally, a sub-stoichiometry has withinthis work also been observed using other techniques like XPS, see chapter 3.1.2, and ERDA, see chapter 3.1.3, on samples produced with the AGD technique.

46

47

Table 3.1: Coordination numbers for W-O and O-O nearest neighbors according to neutron scattering experiments. In the table the expected values for stoichiometric WO3 are also presented.

W-O O-O

Expected values 6 8

AGD WO3 5.0 7.1Ref WO3 5.2 7.0

Neutron scattering is a method with high sensitivity to hydrogen. The hydrogen scattering cross sections are for coherent scattering 1.7568 barn and for incoherent scattering 80.26 barn. The incoherent scattering gives rise to a background scattering level; this contribution decreases for higher Q-values, and may be difficult to correct for. This may contribute to uncertainties in the radial distribution function and therefore also in the coordination numbers. Even though the neutron scattering spectrum was acquired in vacuum it is reasonable that an amount of water still was present in the powders as a consequence of an extreme surface area. In the next chapter an estimation of the content of water is presented.

3.1.6 IR-spectroscopy IR-spectroscopy was employed in order to estimate the water content in the nanocrystalline WO3 films (A3, as defined in chapter 2.1). In Figure 3.8 infrared absorption data obtained from thick films (7µm) of nanocrystalline WO3 films produced with the AGD unit is shown. The HOH bending mode as well as OH stretching modes are seen on top of the WO3 absorption. In Figure 3.8b the backgrounds have been subtracted so that the graph depicts the absorptions of only the water component. The measurements were carried out using s – and p - polarized light.

The averaged value of the integrated absorption from the different polarizations was used for the estimation of the content of water. Three different absorption modes have often been considered. The strongest or most intense absorption is located at 3200 - 3250 cm-1 and is characteristic for the H-OH stretching mode of surface water. Smaller absorption peaks appears at ~3400 cm-1 (hydroxylation) and at ~3000 cm-1 (hydration) 75-77.The amount of water was obtained by use of the calibrated absorption at 3200 cm-1 78,79. Given a 25 % relative density (compared to bulk WO3) of the film, see chapter 3.1.3, a molar ratio of H2O to WO3 of 22 % was obtained, which is in the same order of magnitude as observed in other oxide materials using the same method 80.

0

50

100

150

200

1.0 1.5 2.0 2.5 3.0 3.5 4.0

s-pol p-pol

[1/m

]

Wavenumber/1000 [1/cm]

OH-stretching

HOH-bending

a)

0

10

20

30

40

50

2.0 2.5 3.0 3.5 4.0

s-pol p-pol

[1/m

]Wavenumber/1000 [1/cm]

b)

Figure 3.8 a) Infrared absorption of a WO3 film. b) The absorption due to the OH-stretching bond for s- and p-polarized light.

3.1.7 Film characterization by 1/f noise Conduction noise in thin WO3 films was measured. A schematic view of themeasurement setup is shown in Figure 2.12. Figure 3.9 shows voltage noisefrom samples with different thickness. Unbiased samples showed thermalnoise, which were compared with calculated data obtained by using Eq. (2.7) and values of R and C’ (Table 3.2) derived from impedance measurements.Note the plateau-like shape below 10 kHz for unbiased samples. The plateau can be explained by shunting resistors consisting of needle-like channels of aluminum penetrating the porous structure of the WO3. As the aluminumcontact is made by evaporation, the widest pores are immediately filled withaluminum and channels are formed along the cross section of the thin WO3layer. An illustration of the aluminum channels is given to the right in Figure 3.10. This shunting results in a frequency independent resistance with a value roughly given by

2AlAl

alR (3.1)

where Al 10-5 cm is the resistivity of Al in the narrow channels (slightlyhigher than for bulk material due to additional surface scattering), lAl is the length of the aluminum channel penetrating the porous dielectric and 2a is the diameter of the channel.

48

10-18

10-17

10-16

10-15

10-14

10-13

10-12

10-11

10-10

100 101 102 103 104 105

d=125 nm, no biasd=130 nm, no biasd=140 nm, no biasd=140 nm, 4.5 mV bias

S v [V2 /H

z]

Frequency [Hz]

Figure 3.9 Spectral noise for three samples with different thickness. Also shown is the noise of the thickest sample when biased with 4.5 mV ( ). The solid lines were calculated from Eq. (2.7) with values of R and C’ (Table 3.2) obtained fromimpedance measurements.

Figure 3.10 Schematic view of the aluminum channels penetrating the porous structure of the WO3 nanoparticle films.

Table 3.2: Values of R and C’ for samples with thicknesses being 125 nm, 130 nmand 140 nm, used for calculation of voltage noise in Figure 3.9. f denotes frequency.

F [kHz] R125 [k ] C’125 [nF] R130 [k ] C’130 [nF] R140 [k ] C’140 [nF]0.10 0.60 40 2.6 20 6.2 200.12 0.60 37 2.6 19 6.2 170.20 0.58 28 2.6 14 6.0 130.40 0.58 19 2.6 8.5 5.7 7.91.0 0.57 10 2.5 4.9 5.3 4.42.0 0.57 6.3 2.4 3.3 5.1 3.24.0 0.56 4.3 2.3 2.5 4.8 2.410 0.55 2.9 2.2 1.9 4.3 1.820 0.54 2.3 2.0 1.6 3.9 1.540 0.52 1.9 1.9 1.4 3.4 1.3

100 0.47 1.6 1.6 1.2 2.7 2.1

49

Choosing the smallest possible value for lAl, which is the thickness of the dielectric layer (sample thickness) i.e., 130 nm, and choosing 2a 2 nm, we calculate R 4.2 k . This value of 2a is consistent with the expected cavitydiameters between the WO3 nanoparticles. Similar experimental results havebeen observed in ref 70,81. This calculated resistance is in agreement with the observed impedance and corresponds to a voltage noise level of 6 * 10-17

V2/Hz, which is of the same order of magnitude as the value observed fromthe noise measurement in Figure 3.9.

The criterion for observing 1/f noise above the thermal noise in a homogeneous conductor is given by SV,1/f > SV,th. Using Eqs. (2.7) and (2.11), and for 1/N = qµR/d2 where N is the total number of charge carriers, thisinequality becomes

TRk4fN1V Bf1

2 (3.2)

or

Tk4f1e Bf1

2E (3.3)

where µ is the mobility of the free charge carriers. Charge transfer byhopping is characteristic for materials with very low µ (below 1 cm2/Vs). If we assume f 1kHz, T = 300K and 1/f = 10-4, the necessary field strength E to observe 1/f noise often becomes stronger than the breakdown fieldstrength of the material. This explains why the 1/f noise is not commonlyobserved above the thermal noise in dielectrics.

Noise can reveal a sample defect caused by aluminum electrodes penetratinginto and through the thin dielectric. Such defects are not uncommon, andaluminum channels penetrating silicon integrated circuits were observed as early as in 1973 82. In the present case, applying a DC–voltage of 4.5 mVyields 1/f noise due to resistance fluctuations as observed in Figure 3.9. This effect is not due to fluctuations in the dielectric but to fluctuations in the resistance of the aluminum channels penetrating through the dielectric. Afterbiasing, the noise was measured again and no change in the noise spectrumwas observed, which means that the aluminum channels did not degrade dueto the bias conditions.

50

51

For thicker samples a different characteristic for the noise spectra was observed, which is due to a constriction of the Al channel, see further in Paper I.

3.1.8 Conclusions from structure and composition analysis The produced as deposited films consist of WO3 nanoparticles of lognormal size distributions with grain sizes from 3 nm (determined from isolated particles) to 10 nm (XRD) depending on analyzing techniques, agglomeration process and fabrication parameters. XPS and ERDA analysis, as well as neutron scattering experiments indicate a substoichiometric composition for the as deposited films, i.e. ~WO2.6. The films exhibit a very porous structure with a relative density of ~25% compared to bulk WO3.Film characterization by 1/f noise indicates a porous structure as well. After annealing the WO3 nanoparticles undergo additional oxidation towards stoichiometry. The as deposited films consisted of mostly tetragonal phase, but after annealing above 573 K, a phase transition towards monoclinic structure took place. The huge surface area combined with a high degree of porosity result in a significant amount of adsorbed water present in the films.

3.2 Conduction noise as a tool for gas sensing The conduction noise of thin WO3/Au devices were analyzed upon C2H5OHexposures. The sample and measuring setup are shown in Figure 2.18. Figure 3.11 shows a typical set of spectra before insertion (x) and after 18 min of exposure ( ). The noise level increased by about one order of magnitude. A time resolved response is shown in Figure 3.12. A spectrum was taken before insertion in the test chamber and each three minutes after exposure to alcohol vapor. The relative noise change here is a factor of 2.4.As the ethanol is removed, an increase is observed though less pronounced. No relative noise increase was observed for films without WO3, and therefore the WO3 or the WO3/Au interface must be responsible for the observed noise increase.

10-11

10-10

10-9

10-8

10-7

10-6

100 101

Vol

tage

noi

se[V

2 /Hz]

Frequeny [Hz]

After 18 min ofalcohol exposure

Before exposure

Figure 3.11 Typical set of spectra taken before alcohol insertion (x) and after 18 min of exposure ( ).

The conductance fluctuations can emanate from bulk or surface effects. The bulk conductance cannot change as a result of the exposure to alcohol, sotherefore the surface must account for the noise increase. Under exposure the total resistance only changes by a few percent, but the noise may change up to several orders of magnitude. The observed effects were reproducible, but the noise response decayed upon exposure due to ageing of the device.

There are different possible explanations for the observed effect.

1. The adsorbed ethanol molecules may modify the conductance of the WO3/Au surface.

2. Tungsten oxide may not be precisely stoichiometric, in the nanoparticles, even though the X-ray diffraction peaks correspondto WO383, see also chapter 3.1.2-3.1.4. Therefore WOx may easilyreduce adsorbed ethanol to form WO3 with a concomitant change of the conductance of the nanoparticles.

3. As the ethanol is adsorbed on the surface of the gold film, non-adiabatic energy dissipation, due to exothermic adsorption, maycreate hot, ballistic electrons in the metal. These electrons mayeasily penetrate the Schottky barrier that the metal/WO3semiconductor interface exhibits 84, thereby resulting in fluctuations in the number of charge carriers in the metal (andalso in the nanoparticles).

52

We call the observed effect “Invasion Noise” due to its appearance in bothinsertion and removal of the gas during the time when there is a netadsorption or net desorption.

0.5

0.75

1

1.25

1.5

1.75

2

2.25

2.5

0 30 60 90 120 150

Rel

ativ

e ch

ange

Time [min]

Ethanol insertion

Ethanol removal

Figure 3.12 Relative noise intensity change vs time for the device exposed to ethanolas indicated. The curve was drawn as a guide for the eye.

Conduction noise as a tool for gas sensing is a promising new method. The observed relative noise intensity change upon C2H5OH was up to several orders of magnitude, for further details see Paper II. In addition, an ageing effect was also observed which resulted in a decrease of sensitivity of the films.

3.3 Gas sensing properties at constant operating temperatures

Thick films of WO3 nanoparticle films were deposited onto substrates for gas sensing applications, see chapter 2.8.1 and illustration in Figure 2.14 Inorder to operate properly, an annealing procedure for the sensor materials is required. This effect is visualized in Figure 3.13 for a sensor produced according to B. The sensor exhibits the highest sensitivity after annealing at1h at 573 K.

The nanocrystalline WO3 sensors exhibit extreme room temperaturesensitivities towards H2S gas. In Figure 3.14, the room temperaturesensitivities for four different sensors are shown as a function of H2Sconcentration. The three pure WO3 sensors exhibit different levels ofsensitivity. The attached table shows the fabrication parameters, i.e.

53

temperature of the W pellet (T) and distance from transfer pipe to substrate (ltp) as well as the grain size (XRD – analysis) for the as deposited samples.

0

200

400

600

800

1000

1200

400 500 600 700 800 900

Sens

itivi

ty

Annealing temperature [K]

Figure 3.13 The sensitivity as a function of annealing temperature for a sample produced according to B. After annealing at 573 K a maximum in the sensitivity wasobtained.

100

101

102

103

104

1 10

WO3

A2

WO3 0.5% Pd A2

WO3

A1

WO3

B

Sens

itivi

ty

H2S concentration [ppm]

T [K] Grain size [nm] ltp

[cm]

A1A2B

136013601470

~5-6~5-6~10

3153

Figure 3.14 Room temperature sensitivities upon H2S exposure for four differentsensors reported in Papers III, IV and V as a function of H2S concentration. The table presents the different fabrication parameters.(see details in the text).

The sensors denoted A1 and A2 have the same structures and particle sizes according to XRD analysis. However, the sensors were deposited withdifferent distances from the transfer pipe and we speculate that the larger the distance the more porous the structure is and a higher sensitivity follows as itwas observed. For the sensors (B) a lower sensitivity was observed, whichprobably is due to the higher production temperature and thereby a larger

54

particle size, as observed from XRD-analysis. In addition, the sensor dopedwith 0.5 % Pd/W ratio exhibited enhanced low concentration sensitivity, and clearly 0.5 ppm of H2S gas was detected with a sensitivity of about 10.

By elevating the operating temperature of the sensor, the response and recovery times can be improved, as seen, for a pure WO3 sensor (A2) upon 10 ppm H2S exposure, in Figure 3.15a. The high room-temperaturesensitivity can also be combined with the high temperature recovery time byapplying a heating pulse of 473-530 K, as can be seen in Figure 3.15b.

100

101

102

103

104

0 5 10 15 20 25 30

Sens

itivi

ty

Time [min]

298 K

323 K

373 K

423 K

473 K

H2S in H

2S out

10 ppm H2S op

a)

100

101

102

103

0

2

4

6

8

10

0 10 20 30 40 50 60 70 80

Sens

itivi

ty

H2 S concentration [ppm

]

Time [min]

pulse of530 K

b)

Figure 3.15 a) Sensitivity dynamics for a pure WO3 sensor (A2) sintered at 473 Kand exposed to 10 ppm H2S at the shown operating temperatures op b) Sensitivityvs. time of a WO3 senor (A1) sintered at 750 K, exposed to different concentrations of H2S in synthetic air at room temperature. Each exposure was followed by a temperature pulse up to 530 K. The dashed lines in b) represent H2S concentration.

Figure 3.16 shows sensitivity versus H2S concentration for two doped sensors (B), with Al and Au, together with a pure WO3 sensor (B) operating at 600 K. Clearly, the sensitivity was influenced by the dopants and the bestperformance at these high temperatures was obtained for the sensor containing Au.

The selectivity for different species or gases is of huge interest for the sensor research community. Each gas has specific adsorption and desorptionproperties, which depend on the operating temperature. In order to determinethe optimal operating temperature for different gases, the sensitivity of anAl-activated sensor film (B) was tested at different operation temperatures in H2S, CO, and NO2 environment. Results are shown in Figure 3.17a for gas concentrations being 10 ppm H2S, 100 ppm CO, and 5 ppm NO2. The sensor exhibited different optimum operating temperatures, specifically being 400, 525, and 700 K for H2S, NO2, and CO, respectively. Importantly, the sensor

55

did not show any overlap in its maximum gas-specific sensitivities, whichimplies that chemical selectivity can be obtained when the sensor is operated at 400, 525, and 700 K. At optimal operating temperatures, the sensor is ~20 and ~2000 times more sensitive to H2S than to NO2 and CO, respectively,normalized to the concentration.

0

11

21

32

43

54

64

75

0 5 10 15 20 25 30 35 40

WO3 + Al B

WO3

B

WO3 + Au B

Sens

itivi

ty

H2S Concentration (ppm)

Top

= 600 K

Figure 3.16 Sensitivity as a function of H2S concentration for pure and doped nanocrystalline WO3 sensors at 600 K operation temperature.

0

20

40

60

80

100

120

140

0 100

1 103

2 103

3 103

4 103

300 400 500 600 700 800

NO2

CO

H2S

Sens

itivi

ty Sensitivity

op (K)

Figure 3.17 Sensitivity as a function of operating temperature for as-deposited Al-activated WO3 films exposed to 10 ppm H2S (right-hand scale), 5 ppm NO2 (left-hand scale), and 100 ppm CO (left-hand scale). The lines serve as guide for the eye.

It has been shown that the sensors exhibit an extreme room temperaturesensitivity towards H2S, for 10 ppm a sensitivity of more than three orders of magnitude, depending on the fabrication parameters, was observed. Byaddition of small quantities of Pd the low concentrations sensitivity was

56

improved and clearly 0.5 ppm of H2S was detected by a sensitivity of 10. The high room temperature sensitivity could be combined with the high temperature recovery times by applying a heating pulse. By use of differentoperating temperatures, one and the same sensor could be employed for detection of different gases and thereby obtaining selectivity.

3.4 Temperature modulated gas sensors The gas sensing properties of a temperature modulated WO3 sensor (A2) was measured in the temperature range 150 –250 C and the responses were similar as shown in Figure 2.16. The resistance changes due to the temperature modulation and due to complex serial and parallel adsorptionand desorption events as well as chemical reactions at the sensor surface. In Figure 3.18 the sensitivity for a sensor exposed to different concentrations of C2H5OH (2 –50 ppm) (a) and H2S (0.02 –1 ppm) (b) are shown. The sensor clearly manages to detect 2 ppm of C2H5OH and 20 ppb of H2S.

0

5

10

15

20

25

30

35

0 10 20 30 40 50

Sens

itivi

ty

Concentration [ppm]

S=0.54* +1.7C

2H

5OH

0

1

2

3

4

0 0.2 0.4 0.6 0.8 1

Sens

itivi

ty

Concentration [ppm]

S=-1.4* 2+3.9* +1

H2S

Figure 3.18 Sensor sensitivity S for different concentrations of (a) C2H5OH and (b) H2S . The fitting of the data is represented with the solid lines and are described by the equations for S.

Qualitative analysis was performed in order to distinguish between the ethanol and the H2S test gases using the same sensor. Coefficients from the sensor responses were extracted using either FFT or DWT methods followed by linear pattern recognition methods, such as PCA and DFA. Figure 3.19depict the outcome of the analysis of the sensor responses using (a) DWTtogether with DFA as well as (b) FFT together with PCA. DWT together with DFA is superior to the FFT analysis and gave a clear classification ofthe two gases.

57

Figure 3.19 Results of feature extraction by use of a) DWT together with DFA analysis and b) FFT together with PCA analysis. ( ) indicate ethanol and (+) H2S.

Temperature modulated sensor are clearly capable to detect very lowconcentrations of both ethanol, i.e. 2 ppm and H2S, i.e. 20 ppb. The methodwith temperature modulated sensors, together with mathematicaltransformations and pattern recognition, was found to be able to clearlydistinguish between the different gases C2H5OH and H2S, which means the sensor exhibits very good selectivity.

3.5 Results on optical and electrical characterizationThe WO3 nanocrystalline films were produced with a large distance (A3)from transfer pipe to substrate in order to obtain a large surface with homogenous thickness. Figure 3.20 shows experimental data for the optical absorption (full lines) together with the best fits of the calculated absorption (dotted lines) using Eq. (2.23). Fairly good agreement between theory and experiment is observed at low energies, whereas the agreement is less good on the high-energy side of the peak. Table 3.3 contains fitting parameters for the samples.

58

0.0

0.5

1.0

1.5

2.0

2.5

0.50 1.0 1.5 2.0

ExperimentalTheory

/105 (m

-1)

Energy (eV)

300

200

150

as dep

A=50 °C

100

Figure 3.20 The full line presents experimental absorption coefficient versus photon energy for nanoparticle tungsten oxide films annealed at the given temperatures A.The dotted line represents theoretical fit according to Eq. (2.23).

The resistance as a function of temperature for samples, as deposited and annealed at 323 K < A < 573 K, were measured in the temperature range 77 - 300 K and the results are depicted in Figure 3.21a. The resistance increases rapidly as the temperature decreases. Ln(R/T) as a function of the inverse temperature 1/T is shown in Figure 3.21b. The experimental data coincide with straight lines according to the theory for hopping conduction, i.e. Eq.(2.24). For each annealing temperature the activation energy for the hoppingconduction EH is calculated and presented in Table 3.3

s – and p- polarized light with an angle of incidence of 60 were used in order to find the energies of longitudinal (LO) and transversal (TO) opticalphonons. The p-polarized light excites both LO and TO phonons while the s-polarized light only excites the TO phonons. The largest LO phonon energy

op was found to be 0.112 eV.

59

0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

100 150 200 250 300

R [G

Ohm

]

Temperature [K]

As-depA=50 °C

100

150

200 300

a)

11

12

13

14

15

16

17

18

0.003 0.004 0.005 0.006 0.007 0.008ln

(R/T

)[K-1]

As-dep A=50 °C100

150200300

b)

Figure 3.21 a) The resistance as a function of temperature. b) Ln(R/T) versus theinverse temperature1/T for samples annealed at the given temperature A.

The polaron radius rpolaron, estimated by use of Eq. (2.20), is in the range of the lattice parameter, which indicates a small polaron situation. Thetemperature dependence of the resistance and the fact that the hoppingenergy is about half the polaron binding energy points strongly towardspolaron absorption and nearest neighbor hopping conduction in these nanocrystalline WO3 films. The fitting parameters for small polaron absorption theories according to Eq (2.23), i.e. C and Ep, are presented in Table 3.3 together with the experimentally obtained LO phonon energy opand activation energy for hopping conduction EH as well as the calculatedpolaron radius rpolaron and estimated polaron density p.

Table 3.3: Data for nanoparticle tungsten oxide films annealed at temperature A.The pre-expentional factor C , and the polaron binding energy EP are obtained fromfitting Eq. (2.23) to the optical absorption. Data are also given for the LO phonon energy op, the polaron hopping energy EH, the polaron radius rpolaron as well as an estimated polaron density p.

A[K]

C[105 m-1]

op[eV]

Ep[eV]

EH[eV]

rpolaron[Å]

p[polaron/particle]

As-dep 1.5 0.112 0.450 0.22 5.3 1.7273 1.67 0.112 0.460 0.19 5.2 1.9323 2.15 0.112 0.480 0.23 5.0 2.3423 0.91 0.112 0.425 0.25 5.6 1.1473 0.67 0.112 0.420 0.40 5.7 0.81573 0.33 0.112 0.4 0.51 6.0 0.44

60

61

Annealing at A 473 K causes a deviation from the relation Ep=2*EH as seen from Table 3.3. The reason for this deviation is not obvious. However, it is seen in Table 3.3 that these samples have a polaron density of less than one polaron per particle. One may speculate that surface oxidation produces a layer with stoichiometric, or almost stoichiometric, tungsten oxide around the individual nanoparticles or clusters of nanoparticles. In this layer, the charge carrier concentration may be very small, and the process leads to a decreasing polaron concentration. For the samples annealed at A 473 K, Epand EH are of the same order of magnitude. Due to effects of surface barriers and the low polaron concentration, the electrons have to be excited from the polaron states to the conduction band in order to participate in the electrical conduction process. Hence the activation energy should be of the order of the polaron binding energy Ep, in qualitative agreement with the data in Table 3.3.

There is an interest in determining the bandstructures and bandgap energies for WO3. Bandgap calculations are reported for a number of different crystallographic structures (chapter 1.4), and some examples are cubic 20,85,86, tetragonal 85,86 and monoclinic 85,86. In the work by Cora et al. a Hartree- Fock (HF) method was applied for the calculations. Hjelm et al. used the local density approximations (LDA) and De Wijs et al. employed LDA and generalized gradient approximations (GGA) for their calculations.

According to calculations, the cubic WO3 has an indirect bandgap, but tetragonal and monoclinic WO3 exhibit direct bandgaps. The bandgap for the tetragonal structure has been reported to be ~0.4 eV whilst for the monoclinic structure it is ~1.6 eV 86. Bandgap results obtained from calculations by use of LDA and GGA are always lower than what can be observed experimentally, 20 and by use of HF a too high value of the bandgap is obtained. However the bandgap of the monoclinic structure seems to be larger than what is obtained for the tetragonal structure.

0

0.5

1

1.5

2

0.5 1 1.5 2 2.5 3 3.5 4

As dep

A= 300 °C

A= 400 °C

h/(2

2/1

013 [m

-2 e

V2 ]

Energy [eV]

a)

0

500

1000

1500

2000

2500

0.5 1 1.5 2 2.5 3 3.5 4

As dep= 300 °C

= 400 °C

h/(2

1/2

[m-1

/2 e

V1/

2 ]Energy [eV]

b)

Figure 3.22 Adsorption data plotted to obtain a) the direct and b) the indirectbandgap. The dashed lines are for deriving the bandgap energies.

Figure 3.22 depicts adsorption data for different annealing temperatures andfor different values of according to Eq. (2.19). The transition fromtetragonal to monoclinic WO3 can be observed with a shift towards higherenergies. a) shows a plot applicable to direct transitions, i.e. =1/2, and b) to indirect transitions, i.e. =2. According to calculations, both tetragonal and monoclinic WO3 should have a direct bandgap. In Figure 3.22, a better fit isobtained for the graphs in a) compared to those in b). This would imply that the tetragonal and the monoclinic structures have a direct bandgap of ~3.3 eV and ~3.5 eV, respectively. In b) the fit can be drawn more arbitrary,ending up with Eg of 2.8-3.0 eV for tetragonal WO3 and 2.9 - 3.2 eV formonoclinic WO3. Earlier investigations often assumed indirect bandgap, asfor the cubic structure, and sometimes obtained Eg<3.0 eV, which is doubtfulin the light of new bandstructure result 85,86.

The main conclusion from the last chapter is that the optical adsorption andpossibly the conduction mechanism in the WO3 nanoparticle films are related to polarons. Both optical and electrical data are consistent and provide strong support to the formation of small polarons. Upon annealing A

473 K, the activation energy for hopping conduction increases and thepolaron concentration decreases. This latter is assumed to be related to additional oxidation at the surface of the nanoparticles. This implies an increased Schottky barrier at the grain boundaries and we think that this maybe the reason why the sensors require an annealing process to work properly(see Figure 3.13). Furthermore the bandgaps were found to be direct for both the tetragonal and monoclinic structures.

62

63

4 CONCLUSIONS

Advanced reactive gas deposition was applied to produce nanoparticle films of WO3 for sensor applications. The as deposited films were sub-stoichiometric with a composition of WOx, with 2.6<x<2.75. The small average grain size of about ~5 nm, the narrow size distribution, and the porosity of the films provided good conditions for sensor applications.

The deposited films had a relative density of ~25 % compared to bulk WO3,slightly depending on the fabrication parameters, resulting in very porous structures. The sensors exhibited extreme sensitivities upon exposure to H2S-gas; at room temperature 0.5 ppm could be monitored. By addition of small quantities of metals, the sensor properties were improved. Due to temperature dependent sensitivity for different gases, one and the same sensor can be used for detection of a variety of gases by use of different operating temperatures and hence it shows selectivity. With a temperature modulated gas sensor, as small quantities as 20 ppb H2S were clearly detectable. Further analysis of the temperature modulated sensor response using FFT and DWT, together with pattern recognition methods, resulted in selectivity and different gases (H2S and C2H5OH) could be distinguished by using the same sensor.

Optical measurements manifested that tetragonal and monoclinic structures have direct bandgaps, which is in accordance with electron band structure calculations.

Investigations on both the optical properties and the temperature dependence of the resistance indicate that the optical adsorption and conduction mechanism in these nanocrystalline WO3 films is related to small polarons. It has been shown that the concentration of polarons decreases upon annealing above 200 ºC, which is likely due to additional oxidation towards stoichiometric WO3 starting from the surface. As a consequence, the electron concentration, already depleted due to O2 adsorption, is further depleted in the vicinity of the surface and gives rise to further increased Schottky barriers. Intuitively, we think that this effect contributes to the observed extreme sensitivity of annealed nanocrystalline WO3 sensors.

64

65

5 Elektriska egenskaper hos nanokristallin WO3 för gassensortillämpningar

Det pågår en intensiv forskning på gassensorer av olika slag för diverse olika ändamål runt om i världen. Om man läser i ett uppslagsverk så finner man en beskrivning av sensorer: “ En sensor är en enhet för att detektera och mäta fysikaliska fenomen så som temperatur, hjärtslag, vindriktning och eld. Sensorer översätter fysikalisk stimulans till en elektrisk signal som kan insamlas med hjälp av dator”.

Det finns många olika typer av utrustningar för att detektera gaser. En typ av sensorer eller detektorer är gaskromatografer, masspektrometri eller pappersdetektorer. Dessa är antingen väldigt dyra eller kräver erfaren personal för att kunna användas. En annan typ av detektorer är en liten sensor där förändringen i gaskoncentration registreras med hjälp av förändringar i det kemiska eller elektriska gensvaret. Små sensorer kan placeras ut i stora antal och vid ohälsosamma miljöer utan att medföra stora kostnader eller riskera personalens hälsa. Speciellt intresse har givits åt lättantändliga eller toxiska gaser, varav några exempel är CH4, H2, CO, H2Soch NOX.

En ideal sensor skall ha följande egenskaper: kemiskt selektiv, ge snabbt gensvar, hög känslighet, tålig, högt motstånd för kontaminering, enkel att tillverka och använda samt billig. Mycket av forskningen på sensorer sker på metalloxider. Det har visat sig att nyckeln till framgång för dessa sensorerna är att utnyttja extremt små kristaller eller partiklar i nanometer skala, så kallade nanokristaller eller nanopartiklar.

Sensorerna fungerar enligt principen att syre adsorberas på nanopartiklarnas yta och ger upphov till ett på elektroner utarmat område. Detta innebär i sin tur att en barriär i gränsen mellan korn uppstår och man erhåller en hög resistans. Då sensorn exponeras för en gas kommer gasens molekyler (g) att reagera med det adsorberade syret (ads) och donera en eller flera elektroner i det utarmade området liknande vad som visas i följande reaktioner för gasen H2S

O2(g) + e– O2–(ads) R(5.1)

och

2H2S(g) + 3O2–(ads) 2H2O + 2SO2 + 3e– . R(5.2)

Barriären minskar då som följd av överföringen av elektroner varpåresistansen ändras. Figur 5.1 visar a) en tre-partikel struktur och b) denuppkomna barriären mellan partiklarna.

Figur 5.1 a) Schematisk bild av en tre-partikel struktur. b) Schematisk bild som visar den uppkomna barriären mellan partiklarna.

I detta arbete har en avancerad gasförångare används för att producera filmerav WO3. I enkelhet kan man likna tillverkningsprocessen i förångaren medlågan från ett stearinljus. Den varma förångningszonen kyls ner och då ångan kondenseras bildas partiklar. Partiklarna samlas upp och deponeras på ett substrat eller en sensorenhet. De bildade partiklarna är i storleksordningen 5 nm. Figur 5.2 visar en bild tagen med ett svepelektronmikroskop. Bilden visar ytan på en värmebehandlad sensor.Man ser tydligt att de tillverkade filmerna har en hög porositet. Den ringa partikelstorleken samt den porösa mikrostrukturen ger en stor yta som gör att dessa filmer lämpar sig väl för gassensortillämpningar.

66

Figur 5.2 Svepelektronmikroskopbild som avbildar ytan av en WO3 sensor med 0.5%Pd värmebehandlad vid 600 C.

Känligheten för en viss gas erhålls genom att mäta resistansen före och under exponering och beräkna kvoten. Sensorerna i detta arbete har i huvudsak undersökts med avseende på deras känslighet för H2S-gas. I Figur 5.3 ser vi hur känsligheten hos en sensor, som befinner sig vid rumstemperatur, varierar med tiden under exponering av olika koncentrationer av H2S gas. Vi ser tydligt att sensorn detekterar 0.5 ppm(miljondelarar). Sensorn kan nollställas, med hjälp av en värmepuls, varvid all adsorberad gas desorberas och sensorn är redo för en ny mätning.

100

101

102

103

104

0 5 10 15 20 25 30

0.5 ppm1 ppm2 ppm4 ppm6 ppm8 ppm10 ppm

Kän

slig

het

Tid [min]

H2S in H

2S ut

Värme -puls 10 min

Figur 5.3 Känligheten för en sensor som funktion av tid för olika koncentrationen av H2S – gas.

För temperaturmodulerade sensorer kan känsligheten förbättras ytterligareoch så små mängder som 20 ppb (miljarddelar) H2S kan detekteras. Omsensorsignalen analyseras med hjälp av matematiska metoder, så som FastFourier Transformation (FFT), kan man särskilja olika gaser med en och samma sensor.

67

68

6 APPENDIX

There is and has been an intense research on WO3 gas sensors. A variety of methods for detection have been applied. The most used method to detect the presence of a gas has been to monitor the resistance change upon exposure. A review of this kind of method is presented in Table 6.1 below. In addition to the table there are other reports, which have not been included because of their method of detection. Some examples are explained briefly below. A very powerful method for detection is temperature modulated sensors and, by use of mathematical transformations like FFT and DWT, one and the same sensor could be used for detection of mixtures of gases 52,87. An alternative method instead of varying the temperature is to expose the sensor to periodic concentrations and using PCA and pattern recognition techniques for classification. This method has successfully been used to differentiate between different perfumes, beverages and food 88. Another type of sensor is constructed like a diode with an interface like Pt-Pd/WO3/p-Si/Al and by biasing the sensor, a good sensitivity to NO gas was obtained 89.

A and o denote sintering temperature and operating temperature, respectively. In addition, some extra explanations to the table are as follows.

* nano-size coprecipitate of (W,Ti)O3 (denoted NTW) † TiO2 (4wt%)+WO3 (denoted TW) ¤ precursors size ** mixed oxide capacitor with change in capacitance

27 ppm NO + 1 ppm NO2 heterojunction of two pellets of WO3 and 3 % Nd2O3 doped

SnO2MEMS Micro electromachining structures # Response measured by transmittance using FTIR

electrochromic device, response under exposure of light

Table 6.1 A review of the reported research concerning WO3 gas sensors

69

Ref

eren

ce

90 91 92 93 91 94 93 95 96 97 98 99

SG

gas/G

air

1.5

8.4 62 1.9 15 1.

16

1.05

1.

14

1.03

7 4.

5

1.42

70 4 26 15 4 2.5

1.4

Con

c.

[ppm

]

1000

100

100

30 100

6750 25 50 500

50 100

10

o

[C

]

110

220 - 300

220

180

200

450

200

350

300

200

A

[C

]

- 650

800

(3h)

150(

0.25

h) +

60

0(0.

25h)

650

400

(6h)

+20

0 (1

h)

150(

0.25

h),

600

(0.2

5h)

700

(4h)

700

(5h)

- 300

500

Cry

stal

Size

[nm

]- - - - - - - - ~56 - -

Sam

ple

thic

knes

s[µ

m]

- 0.5 15 50 0.5

0.3 50 - - 0.5

“thi

n”

Add

itive

s/

Act

ivat

or~1

nm

Au

1-2

nm P

t 1-

2 nm

Au-

Pt

1-2

nm A

u

10/9

0Ti

O2/W

O3

-

1-2

nm P

t 1-

2 nm

Au-

Pt

1-2

nm A

u 70

nm

Bi

-

0.8

wt%

Au

-2

% C

u 2

% V

~1

6Å c

o-sp

utte

red

Au

-

Prod

uctio

n M

etho

d

LPC

VD

Sput

terin

g (r

.f)

Scre

en p

rintin

g

Dro

p co

atin

g

Sput

terin

g (r

.f)

Sput

terin

g (r

.f)

Dro

p co

atin

g

Past

e on

Al 2O

3-tub

e

Pow

der p

rod

from

tu

ngst

ic a

cid

Sput

terin

g (r

.f)

Sol-g

el d

rop

coat

ing

Sput

terin

g (r

.f)

Test

gas

H2

H2

CH

x

CH

4

CH

4

CH

4

NH

3

NH

3

NH

3

NH

3

NH

3

NH

3

70

Ref

eren

ce

93 100

101

91 102

93 103

104

105

106

107

108

109

SG

gas/G

air

-1.2

-(#)

1.1

1.2

3.3

3.5

1.2

7.1

-8.9

2

-3.4

-1

.8

0 -32

-200

-2

3

2.32

(**)

1.

61 (*

*)

3.18

(**)

- 4

7

-38

-65

Con

c.

[ppm

]

25

10*1

03

100

100

1300 25 194

163

200

100

200

40 200

o

[C

]

200

150

300

220 - 200

350

150

300

550

700

480

500

250

400

A

[C

]

600

(0.2

5h)

-

250(

4h)+

60

0(12

h)

650 -

150(

0.25

h)+

600

(0.2

5h)

400

(100

h)

400

(100

h)

600

(4h)

500

800

(2h)

600

(2h)

650

(8h)

Cry

stal

Size

[nm

]- 16 30 - - - -

50-2

>100 - - - - -

Sam

ple

thic

knes

s[µ

m]

50 - 0.3

0.5

1 m

m

50 15 -

0.6

mm

2000 -

0.75

Add

itive

s/

Act

ivat

or- -

10/9

0 Ti

/W

20/8

0 Ti

/W

1-2

nm P

t 1-

2 nm

Au-

Pt

1-2

nm A

u

- -

3 %

Bi 2O

3

9.5

% B

i 2O3

50 %

Bi 2O

3

-2.

5 w

t% B

i 2O3

-

33/ 6

6 Sr

SnO

3/WO

333

/ 66

BaZ

rO3/W

O3

33/ 6

6 B

aSnO

3/WO

3

BaC

O3+

WO

3

Ba 2

WO

5

1 %

Ag

-

Prod

uctio

n M

etho

d

Dro

p co

atin

g

Lase

r eva

pora

tion

of p

elle

t Sp

utte

ring

Sput

terin

g (r

.f)

Sint

ered

pow

der

from

cal

cina

tion

Dro

p co

atin

g

Mix

ed p

owde

r sc

reen

prin

ting

Mix

ed p

owde

r sc

reen

prin

ting

Past

e on

Al 2O

3-tub

e

Pow

der m

ixtu

re a

nd

sint

erin

g / p

ress

ing

Pow

der m

ixtu

re a

nd

sint

erin

g / p

ress

ing

Pow

der s

inte

red

to

pelle

tPu

lsed

lase

r de

posi

tion

Test

gas

CO

CO

CO

CO

CH

OH

NO

NO

NO

NO

NO

NO

NO

NO

71

Ref

eren

ce

110

92 111

112

94 113

114

115

116

SG

gas/G

air

-7.2

5

-25

-100

-180

-150

-1

40

-110

-5

0 -1

200

-30

-120

0

-40

-70

-100

-8

0

Con

c.

[ppm

]

280 5 150 5 800

10 1000 10 440

o

[C

]

300

300

25 200

180

180

250

190

250

150

200

200

A

[C

]

680

(10

min

)

800

(3h)

500

(1h)

600

(4h)

400

(6h)

+20

0 (1

h) a

ctiv

atio

n

-

400

(6h)

-

200

(1h)

Cry

stal

Size

[nm

]- - - - - 6.5

25-3

0

9 -

Sam

ple

thic

knes

s[µ

m]

15 15 300

0.3

0.15

0.05

-0.5

100-

200

0.30

Add

itive

s/

Act

ivat

orB

i 2O3

10/9

0Ti

O2/W

O3

- -

70 n

m S

b 70

nm

Au

70 n

m P

d 70

nm

Bi

- - - -Pt

60

nm

Pd 6

0 nm

A

u 60

nm

Prod

uctio

n M

etho

d

Scre

en p

rintin

g

Scre

en p

rintin

g

Sput

terin

g D

C

Sput

terin

g D

C

Sput

terin

g (r

.f)

Sput

terin

g (r

.f)

Rea

ctiv

eSp

utte

ring

(r.f)

R

eact

ive

Sput

terin

g (r

.f)

Rea

ctiv

eSp

utte

ring

(r.f)

R

eact

ive

Test

gas

NO

NO

NO

NO

NO

NO

NO

NO

NO

72

Ref

eren

ce

117

117

99 117

100

118

119

93 120

121

122

123

SG

gas/G

air

-25 -4

-1.7

5

-15

(#)

7.5

4.4

7.5 34 9.9

9.7 11 18 300

Con

c.

[ppm

]

10

50*1

03

10 10 1000 25 100 1 50 1

o

[C

]

200

200

400

200

150

420

420

420

200

220

170

300

100

20

A

[C

]

400-

700

(1h)

250

(4h)

+600

(3

0h)

500

600

(24h

)

- 500

500

450

(1h)

150(

0.25

h),6

00

(0.2

5h)

100-

400

700

(4h)

670

500

Cry

stal

Size

[nm

] - - - - 16

100-

2000

- - 10 - - ~40

Sam

ple

thic

knes

s[µ

m]

“thi

n”

“thi

n”

“thi

n”

“thi

n” -

“thi

n” - 50 40 20

“thi

ck”

7.5

Add

itive

s/

Act

ivat

or-

10/9

0 Ti

/W

- - - -5

at%

MoO

3

- -

7.7

wt%

Pt

7.2

wt%

Pd

0.5

wt%

Au-

parti

cles

SnxW

O3+

x

x=1.

25-2

.5

-

Prod

uctio

n M

etho

d

Sol-g

el o

f W

(OC

2H5) 6

in n

-bu

tano

l

Sput

terin

g (r

.f)

Sput

terin

g (r

.f)

Vac

. the

rmal

eva

p.

of W

O3 p

owde

r

Lase

r eva

pora

tion

of p

elle

t

Sol-g

el

Sol-g

el sp

inn

coat

ing

Dro

p co

atin

g

Gas

-eva

pora

tion

Past

e/sc

reen

prin

ting

Scre

en p

rintin

g of

po

wde

r mix

ture

Scre

en p

rint.

of a

rc

evap

. WO

3 pow

der

Test

gas

NO

x

NO

x

NO

x

NO

x

O2

O2

O2

H2S

H2S

H2S

H2S

H2S

73

Ref

eren

ce

124

125

126

127

91 128

94 93 129

130

131

132

105

SG

gas/G

air

50 14 20 3 2 24 6.5

2.3

2100 16 -4.5

3

-1.7

-50 -6

-23

-100

Con

c.

[ppm

]

50 5

5*10

-3

1 1 1 10 25 0.4 10 5 1 80

o

[C

]

100

200

330

215

220

200

180

300

226

320

350

150

300

A

[C

]

550

400

700

700

(2h)

600

650

250(

0.5h

)+

400(

1h)

400

(6h)

+

200

(1h)

150(

0.25

), 60

0 (0

.25h

) 65

0

600

500

300

(24h

)

600

(4h)

Cry

stal

Size

[nm

]- 30 70 - - - - - - ~16

12 - ~20 -

Sam

ple

thic

knes

s[µ

m]

“thi

ck”

-

21.4

“thi

n”

0.5

0.05

0.3 50 - 2 0.

15

0.15

-

Add

itive

s/

Act

ivat

orSn

xWO

3+x

x=0.

7-1.

72

-

5 %

SnO

2

-

1-2

nm P

t 1-

2 nm

Au-

Pt

1-2

nm A

u A

u

70 n

m B

i

- - - - - -

Prod

uctio

n M

etho

d

Scre

en p

rintin

g of

po

wde

r mix

ture

Scre

en p

rintin

g

Scre

en p

rintin

g of

W

O3 p

owde

r Sp

utte

ring

Rea

ctiv

e

Sput

terin

g (r

.f)

Sput

terin

g (r

.f)

Sput

terin

g (r

.f)

Dro

p co

atin

g

Dro

p co

atin

g of

pa

rticl

es p

rod.

with

la

ser a

blat

ion

HV

ther

mal

ev

apor

atio

n H

V th

erm

al

evap

orat

ion

HV

ther

mal

ev

apor

atio

n Pa

ste

on A

l 2O3-t

ube

Test

gas

H2S

H2S

H2S

H2S

H2S

H2S

H2S

NO

2

NO

2

NO

2

NO

2

NO

2

NO

2

74

Ref

eren

ce

133

134

135

109

107

136

137

138

125

139

92 140

141

SG

gas/G

air

-2.5

()

-9

-4

-150

-1

8 -1

70

- 29

-4

-2

-90

-400

-1

650

~-30

0

-3

-7

-80

-130

-8

0

-32

-190

-2

50

-700

-2

50

Con

c.

[ppm

]

1.5 2 4 200

80 20 30 100 1 10 6 1 50

o

[C

]

300

300

200

400

500

350

180

100

200

350

300

300

350

A

[C

]

1000

(5h)

1000

(5h)

500

650

(8h)

800

(2h)

500

800

(2h)

700

400

700

800

800

(3h)

400

(2h)

900

Cry

stal

Size

[nm

]1.

5 µm

1.5µ

m

- - - 30 70 23 60 30 70 3-9

3-9 - ~15

2 µm

Sam

ple

thic

knes

s[µ

m]

- 10m

m

-

0.75

2000 - - 20 30 -

“thi

ck”

15 6 -

Add

itive

s/

Act

ivat

or- - -

~3 %

Al 2O

3

~3 %

TiO

2

-

BaC

O3+

WO

3

Ba 2

WO

5

- -TW

†

NTW

*

- -

TW†

NTW

*

10/9

0Ti

O2/W

O3

- -0.

1 %

NiO

1.

0 %

NiO

10

% N

iO

Prod

uctio

n M

etho

d

Pelle

t fro

m p

owde

r

Pow

der s

inte

red

to

pelle

tPu

lsed

lase

r dep

. of

WO

3, A

l 2O3,

TiO

2

pow

der p

elle

t Pu

lsed

lase

r de

posi

tion

Pow

der m

ixtu

re a

nd

sint

erin

g/p

ress

ing

Scre

en p

rintin

g of

po

wde

r pro

d. b

y So

l- pre

cipi

taiti

onSc

reen

prin

ting

Scre

en p

rintin

g

Scre

en p

rintin

g

Scre

en p

rintin

g

Scre

en p

rintin

g

Sol-g

el sc

reen

pr

intin

gSc

reen

prin

ting

of N

iO

and

WO

3 pow

der

Test

gas

NO

2

NO

2

NO

2

NO

2

NO

2

NO

2

NO

2

NO

2

NO

2

NO

2

NO

2

NO

2

NO

2

75

Ref

eren

ce

142

119

143

144

117

101

112

145

117

146

147

94 148

114

SG

gas/G

air

-82

-178

2

-2

-2.2

-48

-180

-11.

6

-3

-2

-180

-22

-2.3

-20

-11

-650

-13

-130

Con

c.

[ppm

]

10 1 10 3 1 0.5

0.5 5 2 1 1 1 100 1 100

o

[C

]

200

300

250

100

200

300

300

200

300

200

200

300

180

300

250

A

[C

]

400

(1h)

450

(1h)

850

(1h)

250

(5h)

400-

700

(1h)

250

(4h)

+600

(12h

)

600

(4h)

600

(4h)

250

(4h)

+600

(3

0h)

600

600

(12

h)

400

(6h)

+20

0 (1

h) a

ctiv

atio

n

600

400

(6h)

Cry

stal

Size

[nm

] 34 18 - - ~4 - 30 - - - - 50

0

~60 - 60

25-3

0

Sam

ple

thic

knes

s[µ

m]

- - - - 0.8

“thi

n” - 300

0.15

“thi

n”

0.36

0

“thi

n”

0.3

0.3

0.05

-0.5

0

Add

itive

s/

Act

ivat

or-

~20

at.%

Ti

-M

oO3

- - -

10/9

0 Ti

/W

20/8

0Ti/W

- -

10/9

0 Ti

/W

10/9

0 Ti

/W

20/8

0 Ti

/W

70 n

m S

b

20/8

0 Ti

/W

-

Prod

uctio

n M

etho

d

Sol-g

el d

ip c

oatin

g

Sol-g

el sp

inn

coat

ing

Sol-g

el d

ip c

oatin

g

Sol-g

el d

ip c

oatin

g

Sol-g

el o

f W

(OC

2H5) 6

in n

-bu

tano

lSp

utte

ring

sput

terin

g D

C

Sput

terin

g D

C

MEM

S

Sput

terin

g (r

.f)

Sput

terin

g ( r

.f.)

Rea

ctiv

e

Sput

terin

g (r

.f.)

Sput

terin

g (r

.f)

Sput

terin

g (r

.f)

Sput

terin

g (r

.f)

Rea

ctiv

e

Test

gas

NO

2

NO

2

NO

2

NO

2

NO

2

NO

2

NO

2

NO

2

NO

2

NO

2

NO

2

NO

2

NO

2

NO

2

76

Ref

eren

ce

116

149

150

117

147

94 148

114

116

149

150

117

SG

gas/G

air

-1.8

-3

-6

.5

-4.7

-6

70

-30

-38

-8.1

-11

-650

-13

-130

-1.8

-3

-6

.5

-4.7

-6

70

-30

-38

-8.1

Con

c.

[ppm

]

10 10 1 1 1 1 100 1 100

10 10 1 1 1

o

[C

]

250

150

200

200

300

150

150

200

300

180

300

250

250

150

200

200

300

150

150

200

A

[C

]

200

600

(0.2

5h) +

30

0 (2

65h)

30

0(24

h)

500(

24h)

60

0 (2

4h)

600

(12

h)

400

(6h)

+20

0 (1

h) a

ctiv

atio

n 60

0

400

(6h)

200

600

(0.2

5h) +

30

0 (2

65h)

30

0(24

h)

500(

24h)

60

0 (2

4h)

Cry

stal

Size

[nm

]- - - - ~60 - 60

25-3

0

- - - -

Sam

ple

thic

knes

s[µ

m]

0.30

1.3

0.08

“thi

n”

“thi

n”

0.3

0.3

0.05

-0.5

0

0.30

1.3

0.08

“thi

n”

Add

itive

s/

Act

ivat

or-

Pt 6

0 nm

Pd

60

nm

Au

60 n

m

10/9

0 /i/

W

- -

20/8

0 Ti

/W

70 n

m S

b

20/8

0 Ti

/W

- -Pt

60

nm

Pd 6

0 nm

A

u 60

nm

10

/90

/i/W

- -

Prod

uctio

n M

etho

d

Sput

terin

g (r

.f.)

Rea

ctiv

e

Sput

terin

g (r

.f.)

Rea

ctiv

eTh

erm

al

evap

orat

ion

Vac

u. th

erm

al e

vap.

of

WO

3 pow

der

Sput

terin

g (r

.f.)

Sput

terin

g (r

.f)

Sput

terin

g (r

.f)

Sput

terin

g (r

.f)

Rea

ctiv

eSp

utte

ring

(r.f.

) R

eact

ive

Sput

terin

g (r

.f.)

Rea

ctiv

eTh

erm

al

evap

orat

ion

Vac

. the

rmal

eva

p.

of W

O3 p

owde

r

Test

gas

NO

2

NO

2

NO

2

NO

2

NO

2

NO

2

NO

2

NO

2

NO

2

NO

2

NO

2

NO

2

77

Ref

eren

ce

151

152

119

121

93 153

94 154

SG

gas/G

air

10 20 2.5 10 10 1.3 18 2 18

Con

c.

[ppm

]

100

100

100 1 25 800

10 1

o

[C

]

480

200

300

300

200

450

180

300

A

[C

]

700

(2h)

500

450

(1h)

700

150(

0.25

h), 6

00

(0.2

5h)

950

400

(6h)

+20

0 (1

h) a

ctiv

atio

n

400(

12h)

Cry

stal

Size

[nm

]- 21 - - - - -

Sam

ple

thic

knes

s[µ

m]

-

600-

800

- 20 50 1000 0.3

0.05

0

Add

itive

s/

Act

ivat

or-

4 w

t% S

iO2

-M

oO3

0.5

wt%

Au-

parti

cles

-

-1 w

t% A

g

70 n

m B

i

-

Prod

uctio

n M

etho

d

PVD

Pow

der

Sol-g

el sp

in c

oatin

g

Past

e/sc

reen

prin

ting

Dro

p co

atin

g

Past

e on

Al 2O

3-tub

e

Sput

terin

g (r

.f)

Sput

terin

g

Test

gas

C2H

5OH

C2H

5OH

C2H

5OH

CH

3SH

SO2

SO2

SO2

Cl 2

78

Ref

eren

ce

117

117

117

119

155

156

SG

gas/G

air

-5

-35

-18 -2

-100

0

3

Con

c.

[ppm

]

0.08

0.08

0.08

150

54 50

o

[C

]

400

400

400

150

200

70

A

[C

]

250

(4h)

+600

(3

0h

400-

700

(1h)

250

(4h)

+600

(3

0h)

450

(1h)

400

500

(10h

)

Cry

stal

Size

[nm

]- - - - 30 -

Sam

ple

thic

knes

s[µ

m]

“thi

n”

“thi

n

“thi

n” -

0.05

-

Add

itive

s/

Act

ivat

or10

/90

Ti/W

- -

MoO

3

- -

Prod

uctio

n M

etho

d

Sput

terin

g (r

.f)

Sol-g

el o

f W

(OC

2H5) 6

in n

-bu

tano

lV

acuu

m th

erm

al

evap

orat

ion

of W

O3

pow

der

Sol-g

el sp

in c

oatin

g

Sput

terin

g (r

.f)

Rea

ctiv

e

Sol-g

el d

ip c

oatin

g

Test

gas

O3

O3

O3

O3

O3

TMA

(CH

3) 3N

79

ACKNOWLEDGEMENTS

First of all I would like to thank the Advanced MicroEngineering (AME) graduate school for accepting me as a Ph.D student and for their financial support.

I would like to acknowledge and express my gratitude to my supervisors Peter Heszler and Claes-Göran Granqvist for their guidance and encouragement. Gunnar Niklasson is acknowledged for his enthusiasm and concern for my work. It has been interesting, instructive and enjoyable working at the division.

In addition I would like to thank my former supervisors Eva Olsson and Laszlo Kish, formerly at Uppsala University, for their support. Nils-Olof Erson and Michael Ottosson are acknowledged for their help whenever it was needed with the X-ray diffraction instruments. Thanks go to Torvald Andersson and Ola Wilhelmsson for their time spent with the XPS measurements. They really helped me a lot. Göran Possnert and Alenka Razpet are acknowledged for helping me with the ERDA experiment and analysis. Per Zetterström is acknowledged for his help in our project concerning the neutron scattering experiments. Jan Lindgren is thanked for fruitful discussions.

Thanks are also due to Lode Vandamme for the work we did together during my stay at Eindhoven Technical University. I also visited Oulu University for collaboration with Vilho Lantto, Sami Saukko and Tapio Lumiaho.Thank you all for your support; I had nice experiences being abroad, even if the visits were short.

I would like to thank Luis Reyes, my flat mate and colleague in Oulu, for teaching me a great deal about sensors and about Peru. I really hope to see the Machu Picchu in Peru some day. I also wish you good luck with your thesis.

All my colleagues and friends at the Ångström Laboratory receive special thanks for making the every day life so memorable. In particular, I would like to thank Maria, AnnaKarin, Anna-Lena, Kristina, Mattias, Jesper,

Ernesto, Staffan, Fredrik and Jörgen for all good times at work at Ångströmand during the conferences and outside work. There are a lot of occasions I will never forget.

Jag hade dock aldrig slutfört detta arbete utan stöd från mina syskon,Mamma och Pappa. Tack för allt.

Till sist vill jag tacka min bästa vän och fru Karin för all uppmuntran, kärlekoch för allt spännande som framtiden innehåller.

Uppsala, Februari 2004

Anders Hoel

80

81

REFERENCES

1 D. D. Spencer, Webster's New World Dictionary of Computer Terms (Prentice Hall, New York, 1992). 2 E. Jones, in Solid State Gas Sensors, edited by P. Moseley and B.Tofield (Adam Hilger, Bristol, 1987). 3 W. Göpel, J. Hesse, J. N. Zemel (Eds.), Sensors: A comprehensive

Survey, Vol. 2 (Chemical and Biochemical Sensors, Part I), (WCH, Weinheim, 1991).

4 R. M. A. Kocache, in Solid State Gas Sensors, edited by P. Moseley and B. Tofield (Adam Hilger, Bristol, 1987). 5 J. H. Visser, S. Shivaraman, C. Svensson, and L. Lundkvist, Appl. Phys. Lett. 26, 55 (1975). 6 I. Lundström, S. Shivaraman, C. Svensson, and L. Lundkvist, Appl. Phys. Lett. 26, 55 (1975). 7 I. Lundstrom, Sens. Actuators A 56, 75 (1996). 8 I. Lundstrom and L.-G. Petersson, J. Vac. Sci. Tech. A (Vacuum, Surfaces, and Films) 14, 1539 (1996). 9 I. Lundström, in Chemical Sensor Technology, Ed T. Seiyama (Kodansha LTD, Elsevier Tokyo, Vol. 2, 1989). 10 A. Spetz, F. Winquist, H. Sundgren, and I. Lundström, in G.Sberveglieri, Gas Sensors, (Kluwer, Dordrecht, The Netherlands, 1992). 11 W. H. Brattain and J. Barden, Bell. System Tech. J. 32, 1 (1953). 12 N. Taguchi, (US Patent No. 3 644 795, 1972). 13 M. J. Madou and S. R. Morrison, Chemical Sensing with Solid State Devices (Academic Press, San Diego, 1989). 14 V. Lantto, in Sberveglieri, G., Gas Sensors, (Kluwer, Dordrecht, The Netherlands, 1992). 15 G. Sberveglieri, Sens. Actuators B 23, 103 (1995). 16 P. Moseley, Sens. Actuators B 6, 149 (1992). 17 P. Romppainen, Acta Univ. Oulo. C47, Oulo, Finland (1988). 18 A. M. Stoneham, in Solid State Gas Sensors, edited by P. Moseley and B. Tofield (Adam Hilger, Bristol, 1987). 19 P. Romppainen and V. Lantto, J. Appl. Phys. 63, 5159 (1988).

82

20 A. Hjelm and C. G. Granqvist, Phys. Rev. B 54, 2436 (1996). 21 J. Tamaki, Z. Zhang, K. Fujimori, M. Akiyama, T. Harada, N. Miura, and N. Yamazoe, J. Electrochem. Soc. 141, 2207 (1994). 22 D. E. Williams, in Solid State Gas Sensors, edited by P. Moseley and B.Tofield (Adam Hilger, Bristol, 1987). 23 N. Yamazoe, Sens. Actuators B 5, 7 (1991). 24 X. Wang, S. S. Yee, and W. P. Carey, Sens. Actuators B 24-25, 454 (1995). 25 C. G. Granqvist, Handbook of Inorganic Electrochromic Materials, (Elsevier, Amsterdam, The Netherlands, 1995). 26 C. G. Granqvist, Sol. Energy Mat. Sol. Cells 60, 201 (2000). 27 A.-L. Larsson, L. Stenmark, and G. A. Niklasson, Proc. of 3rd Round table on micro/nano-technologies for space 15-17 May 2000, ESTEC Conference Centre, Noordwijk, The Netherlands, 249 (2000). 28 P. J. Shaver, Appl. Phys. Lett. 11, 255 (1967). 29 T. Svedberg, Studien zur Lehre von den kolloiden Lösungen /von The Svedberg (Uppsala, 1908). 30 The Nobel Foundation, http://www.nobel.se/chemistry/laureates/ 1926/svedberg-bio.html 31 C. G. Granqvist and R. A. Buhrman, J. Appl. Phys. 47, 2200 (1976). 32 C. Hayashi, R. Uyeda, and A. Tasaki, Ultra-fine particles: Exploratory

Science and Technologie (Noyes, NJ, USA, 1997). 33 H. Gleiter, Nanostruct. Mater. 1, 1 (1992). 34 R. W. Siegel, Ann.Rev.Mater. Sci 21, 559 (1991). 35 J. Söderlund, L. B. Kiss, G. A. Niklasson, and C.-G. Granqvist, Phys.

Rev. Lett. 80, 2386 (1998). 36 D. B. Williams and C. B. Carter, Transmission electron microscopy:a

text book for materials science (Plenum Press, New York, 1996). 37 S. Hogmark, S. Jacobsson, and Å. Rudolphi-Kassman, Uppsala

Univerrsity (1998). 38 Jeol, Instructions JEM-2000FX, Jeol Ltd., Tokyo. 39 L. Reimer, Transmission electron microscopy, 4 ed. (Springer-Verlag,

Berlin, 1997). 40 B. D. Cullity, Elements of X-ray Diffraction (Addison-Wesley,

Reading, Massachusetts, USA, 1956). 41 H. Siegbahn and L. Karlsson, Photoelectron Spectroscopy, Vol. XXXI

(Springer, Berlin, 1982). 42 H. J. Whitlow, G. Possnert, and C. S. Petersson, Nucl. Instr. and Meth.

B 27, 448 (1987). 43 Y. Zhang, H. J. Whitlow, T. Winzell, I. F. Bubb, T. Sajavaara, K.

Arstila, and J. Kainonen, Nucl. Instr. and Meth. B 149, 477 (1999). 44 D. A. Bell, Noise and the solid state (Pentech Press, Plymouth, 1985). 45 A. Ambrozy, Electronic Noise (McGraw-Hill Internaional Book

Company, New York, 1982).

83

46 W. Schottky, Ann. Phys. 57, 541 (1918). 47 L. K. J. Vandamme, IEEE-Trans. Electron Devices. 41, 2176 (1994). 48 F. N. Hooge, T. G. M. Kleinpenning, and L. K. J. Vandamme, Rep.

Prog. Phys. 44, 479 (1981). 49 A. Heilig, N. Barsan, U. Weimar, M. Schweizer-Berberich, J. W.

Gardner, and W. Gopel, Sens. Actuators B 43, 45 (1997). 50 S. Wlodek, K. Colbow, and F. Consadori, Sens. Actuators B 3, 63

(1991).51 E. Llobet, J. Brezmes, R. Ionescu, X. Vilanova, S. Al-Khalifa, J. W.

Gardnar, N. Barsan, and X. Correig, Sens. Actuators B 83, 238 (2002). 52 R. Ionescu, E. Liobet, X. Vilanova, J. Brezmes, J. E. Sueiras, J.

Calderer, and X. Correig, The Analyst 127, 1237 (2002). 53 R. Ionescu, Ph. D. Thesis, (Universitat Politècnica de Catalunya, Spain

2003).54 W. Q. Hong, J. Appl. Phys. D 22, 1384 (1989). 55 L. D. Landau, Sov. Phys. JETP 3, 664 (1933). 56 A. S. Alexandrov and N. F. Mott, Polarons and bipolarons (World

Scientific, Singapore, 1995). 57 D. Emin, Physics today, 34 (1982). 58 V. N. Bogomolov, E. K. Kudinov, and Y. A. Firsov, Soviet Physics-

Solid State 9, 2502 (1968). 59 T. Holstein, Ann. Phys. (N.Y.) 8, 343 (1959). 60 T. Holstein, Ann. Phys. (N.Y.) 8, 325 (1959). 61 H. Fröhlich, Adv. Phys. 3, 325 (1954). 62 L. Berggren, A. Azens, and G. A. Niklasson, J. Appl. Phys. 90, 1860

(2001).63 G. Niklasson, J. Klasson, and E. Olsson, Electrochemica Acta 46, 1967

(2001).64 E. Salje, Optics Comm. 24, 231 (1978). 65 G. A. Niklasson, L. Berggren, and A.-L. Larsson, (2003). 66 A.-L. Larsson, B. E. Sernelius, and G. A. Niklasson, Solid State Ionics

165, 35 (2003). 67 I. G. Austin and N. F. Mott, Adv. Phys. 18, 41 (1969). 68 D. Emin, Phys. Rev. B 48, 13691 (1993). 69 M. Ziese and C. Srinitiwarawong, Phys. Rev. B 58, 11519 (1998). 70 A. Hoel, L. B. Kish, R. Vajtai, G. A. Niklasson, C. G. Granqvist, and

E.Olsson, Proc. MRS 581, 15 (2000). 71 C. D. Wagner, D. Riggs, Davis, J. F. Moulder, and G. E. Muilenberg,

Handbook of X-ray photoelectron spectroscopy : a reference book of standard data for use in X-ray photoelectron spectroscopy (Perkin-Elmer Corp., Eden Prairie, Minnesota, USA, 1979).

72 http://www.sigmaaldrich.com, ProdNo=232785. 73 L. Puszai and R. L. McGreevy, Physica B 234-236, 357 (1997). 74 T. Nanba and I. Yasui, J. Sol. State Chem. 83, 304 (1989).

84

75 L. H. M. Krings and W. Talen, Sol. Energy Mat. Sol Cells 54,27(1998).

76 I. Porqueras and E. Bertran, Thin Solid Films 377-378, 8 (2000). 77 R. A. Nyquist, C. L. Putzig, and M. A. Leugers, The handbook of

infrared and Raman spectra of inorganic compounds and organic salts,Vol. 3 (Academic Press, San Diego, 1997).

78 D. N. Glew and N. S. Rath, Can. J. Chem. 49, 837 (1971). 79 J. Lindgren and Å. Persson, “Infrared absorption intensities of OD

stretching vibrations of HDO and D2O,” Report No. UUIC_B19_320 (1990).

80 J. Rodriguez, Ph.D. Thesis, (Uppsala University, 1999). 81 A. Hoel, R. Vajtai, L. B. Kish, and E. Olsson, Proc. 51 st Ann. Meeting

of the Scandinavian Soc. Electron Microscopy, Bergen, Norway, 51 (1999).

82 T. E. Price and L. A. Berthoud, Sol. State Electr.16, 1303 (1973). 83 E. W. McFarland, Private communications (2001). 84 H. Nienhaus, H. S. Bergh, B. Gergen, A. Majumdar, W. H. Weinberg,

and E. W. McFarland, Surf. Sci. 445, 335 (2000). 85 F. Cora, A. Patel, N. M. Harrison, R. Dovesi, and C. R. A. Catlow, J.

Am. Chem. Soc. 118, 12174 (1996). 86 G. A. De Wijs, P. K. De Boer, R. A. Groot, and G. Kresse, Phys. Rev.

B 59, 2684 (1999). 87 R. Ionescu and E. Llobet, Sens. Actuators B 81, 289 (2002). 88 M. Penza, G. Cassano, and F. Z. Tortorella, G., Sens. Actuators B 73,

76 (2001). 89 W. Zhang, H. Uchida, T. Katsube, T. Nakatsubo, and Y. Nishioka, Sens

Actuators B 49, 58 (1998). 90 D. Davazoglou and T. Dritsas, Sens. Actuators B 3887, 1 (2001). 91 W.-H. Tao and C.-H. Tsai, Sens. Actuators B 81, 237 (2002). 92 M. Gerlich, S. Kornely, M. Fleischer, H. Meixner, and R. Kassing,

Sens.Actuators B 93, 503 (2003). 93 A. A. Tomchenko, G. P. Harmer, B. T. Marquis, and J. W. Allen,

Sens.Actuators B 93, 126 (2003). 94 M. Penza, G. Cassano, and F. Tortorella, Sens. Actuators B 81, 115

(2001).95 T. Maekawa, J. Tamaki, N. Miura, and N. Yamazoe, Chem. Lett., 639

(1992).96 I. Jimenez, M. A. Centeno, R. Scotti, F. Morazzoni, A. Cornet, and J.

R. Morante, J. Electrochem. Soc. 150, H72 (2003). 97 B. T. Marquis and J. F. Vetelino, Sens Actuators B 77, 100 (2001). 98 E. Llobet, G. Molas, P. Molinas, J. Calderer, X. Vilanova, J. Brezmes,

J. E. Sueiras, and X. Correig, J. Electrochem. Soc. 147, 776 (2000). 99 G. Sberveglieri, L. Depero, S. Groppelli, and P. Nelli, Sens. Actuators

B 26-27, 89 (1995).

85

100 M.-I. Baraton, L. Merhari, H. Ferkel, and J.-F. Castagnet, Mat. Sci. & Engineering C 19, 315 (2002).

101 V. Guidi, D. Boscarino, E. Comini, G. Faglia, M. Ferroni, C. Malagu, G. Martinelli, V. Rigato, and G. Sberveglieri, Sens. Actuators B 65, 264 (2000).

102 L. Daza, S. Dassy, and B. Delmon, Sens. Actuators B 10, 99 (1993). 103 A. A. Tomchenko, Sens.Actuators B 68, 48 (2000). 104 A. A. Tomchenko, V. V. Khatko, and I. L. Emelianov, Sens. Actuators

B 46, 8 (1998). 105 M. Akiyama, J. Tamaki, N. Miura, and N. Yamazoe, Chem. Lett., 1611

(1991).106 T. Ishihara, H. Fujita, H. Nishiguchi, and Y. Takita, Sens. Actuators B

65, 319 (2000). 107 M. S. Kwak, J. S. Hwang, C. O. Park, N. Miura, and N. Yamazoe,

Sens. Actuators B 56, 59 (1999). 108 L. Chen and S. C. Tsang, Sens. Actuators B 89, 68 (2003). 109 H. Kawasaki, J. Namba, K. Iwatsuji, Y. Suda, K. Wada, K. Ebihara,

and T. Ohshima, Appl. Surf. Sci. 197-198, 547 (2002). 110 A. A. Tomchenko, I. L. Emelianov, and V. V. Khatko, Sens. Actuators

B 57, 166 (1999). 111 J.-J. Ho, Solid-State Electronics 47, 827 (2003). 112 T. S. Kim, Y. B. Kim, K. S. Yoo, G. S. Sung, and H. J. Jung, Sens.

Actuators B 62, 102 (2000). 113 D. Manno, A. Serra, M. Di Giulio, G. Micocci, and A. Tepore, Thin

Solid Films. 324, 44 (1998). 114 M. Penza, M. A. Tagliente, L. Mirenghi, C. Gerardi, C. Martucci, and

G. Cassano, Sens. Actuators B 50, 9 (1998). 115 M. Di Giulio, D. Manno, G. Micocci, A. Serra, and A. Tepore, J. Phys.

D: Appl. Phys. 30, 3211 (1997). 116 M. Penza, C. Martucci, and G. Cassano, Sens. Actuators B 50, 52

(1998).117 C. Cantalini, W. Wlodarski, Y. Li, M. Passacantando, S. Santucci, E.

Comini, G. Faglia, and G. Sberveglieri, Sens. Actuators B 64, 182 (2000).

118 K. Galatsis, Y. X. Li, W. Wlodarski, and K. Kalantar-zadeh, Sens. Actuators B 77, 478-483 (2001).

119 K. Galatsis, Y. X. Li, W. Wlodarski, E. Comini, G. Sberveglieri, C. Cantalini, S. Santucci, and M. Passacantando, Sens. Actuators B 83,276 (2002).

120 H.-M. Lin, C.-H. Hsu, H.-Y. Yang, P.-Y. Lee, and C.-C. Yang, Sens. Actuators B 22, 63 (1994).

121 M. Ando, S. Suto, T. Suzuki, T. Tsuchida, and C. Nakayama, Chem. Soc. of Japan, Chemistry Lett., 335 (1994).

86

122 J. L. Solis, V. Lantto, L. Häggström, B. Kalska, J. Frantti, and S.Saukko, Sens. Actuators B 68, 286 (2000).

123 J. L. Solis, S. Saukko, L. B. Kish, C. G. Granqvist, and V. Lantto, Sens. Actuators B 77, 316 (2001).

124 J. L. Solis and V. Lantto, Sens. Actuators B 48, 322 (1998). 125 I. Jimenez, J. Arbiol, G. Dezanneau, A. Cornet, and J. R. Morante,

Sens.Actuators B 93, 475 (2003). 126 S.-D. Han, I. Singh, H.-S. Kim, S.-T. Kim, Y.-H. Jung, and K. B.-K.,

Indian J. Chem. 41A, 1832 (2002). 127 I. Ruokamo, T. Kärkkäinen, J. Huusko, T. Ruokanen, M. Blomberg,

and H. Torvela, Sens. Actuators B 18-19, 486 (1994). 128 D. J. Smith, J. F. Vetelino, R. S. Falconer, and E. L. Wittman, Sens.

Actuators B 13-14, 264 (1993). 129 T. K. H. Starke, G. S. V. Coles, and H. Ferkel, Sens. Actuators B 85,

239 (2002). 130 D.-S. Lee, D.-D. Lee, J.-S. Huh, and Y.-H. Hong, J. Korean Phys. Soc.

35, S1092 (1999). 131 C. Cantalini, H. T. Sun, M. Faccio, M. Pelino, S. Santucci, L. Lozzi,

and M. Passacantando, Sens. Actuators B 31, 81 (1996). 132 S. Santucci, L. Lozzi, E. Maccallini, M. Passacantando, L. Ottaviano,

and C. Cantalini, J. Vac. Sci. Tech. A (Vacuum, Surfaces, and Films) 19, 1467 (2001).

133 Z. Ling, C. Leach, and R. Freer, Sens. Actuators B 87, 215 (2002). 134 Z. Ling, C. Leach, and R. Freer, J Europ. Ceram. Soc. 23, 1881 (2003). 135 Y. Zhao, Z.-C. Feng, and Y. Liang, Sens. Actuators B 66, 171 (2000). 136 H.-W. Ryu, K.-H. Park, I.-C. Kim, K.-J. Hong, W.-S. Lee, and J.-S.

Park, J. Kor. Phys. Soc. 42, L727 (2003). 137 D.-S. Lee, S.-D. Han, S.-M. Lee, J.-S. Huh, and D.-D. Lee, Sens.

Actuators B 65, 331 (2000). 138 Y.-K. Chung, M.-H. Kim, W.-S. Um, H.-S. Lee, J.-K. Song, S.-C. Choi,

K.-M. Yi, M.-J. Lee, and K.-W. Chung, Sens. Actuators B 60, 49 (1999).

139 D.-S. Lee, S.-D. Han, J.-S. Huh, and D.-D. Lee, Sens. Actuators B 60,57 (1999).

140 Y.-G. Choi, G. Sakai, K. Shimanoe, Y. Teraoka, N. Miura, and N.Yamazoe, Sens.Actuators B 93, 486 (2003).

141 W. Noh, Y. Shin, J. Kim, W. Lee, K. Hong, S. A. Akbar, and J. Park, Solid State Ionics 152, 827 (2002).

142 J. Shieh, H. M. Feng, M. H. Hon, and H. Y. Juang, Sens. Actuators B 86, 75 (2002).

143 W.-Y. Chung, Journal of the Korean Physical Society 41, L181 (2002). 144 L. G. Teoh, I. M. Hung, J. Shieh, W. H. Lai, and M. H. Hon,

Electrochemical and Solid-State Lett. 6, G108 (2003). 145 X. He, J. Li, X. Gao, and L. Wang, Sens. Actuators B 93, 463 (2003).

87

146 P. Nelli, L. E. Depero, M. Ferroni, S. Groppelli, V. Guidi, F. Ronconi, L. Sangallti, and G. Sberveglieri, Sens. Actuators B 31, 89 (1996).

147 M. Ferroni, D. Boscarino, E. Comini, D. Gnani, V. Guidi, G. Martinelli, P. Nelli, V. Rigato, and G. Sberveglieri, Sens. Actuators B 58, 289 (1999).

148 V. Guidi, E. Comini, M. Ferroni, G. Martinelli, and G. Sberveglieri, J. Vac. Sci. Tech. A 18, 509 (2000).

149 L. Sangaletti, E. Bontempi, L. E. Depero, R. Salari, M. Zocchi, P. Nelli, G. Sberveglieri, P. Galinetto, M. Ferroni, V. Guidi, and G. Martinelli, J. Mater. Res. 13, 1568 (1998).

150 L. Lozzi, L. Ottaviano, M. Passacantando, S. Santucci, and C.Cantalini, Thin Solid Films 391, 224 (2001).

151 T. Takada, T. Meakawa, and N. Dougami, Sens. Act. B 77, 307 (2001). 152 Y. D. Wang, Z. Chen, Y. Li, Z. Zhou, and X. Wu, So. State Electr. 45,

639 (2001). 153 Y. Shimizu, N. Matsunage, T. Hyodo, and M. Egashira, Electrochem.

69, 109 (2001). 154 F. Bender, C. Kim, T. Mlsna, and J. F. Vetelino, Sens. Actuators B 77,

281 (2001). 155 K. Aguir, C. Lemire, and D. B. B. Lollman, Sens. Actuators B 84, 1

(2002).156 M. Tong, G. Dai, and D. Gao, Mat. Chem. Phys. 69, 176 (2001).

Acta Universitatis UpsaliensisComprehensive Summaries of Uppsala Dissertations

from the Faculty of Science and TechnologyEditor: The Dean of the Faculty of Science and Technology

Distribution:Uppsala University Library

Box 510, SE-751 20 Uppsala, Swedenwww.uu.se, [email protected]

ISSN 1104-232XISBN 91-554-5894-7

A doctoral dissertation from the Faculty of Science and Technology, UppsalaUniversity, is usually a summary of a number of papers. A few copies of thecomplete dissertation are kept at major Swedish research libraries, while thesummary alone is distributed internationally through the series ComprehensiveSummaries of Uppsala Dissertations from the Faculty of Science and Technology.(Prior to October, 1993, the series was published under the title “ComprehensiveSummaries of Uppsala Dissertations from the Faculty of Science”.)

Electrical Properties of Nanocrystalline WO for Gas Sensing

Documents