A NOx sensor for high-temperature applications based on SiC

(1)

Linköping University

Institute of Technology

IFM

Master of Science Thesis LITH-IFM-A-EX–10/2286–SE

A NO

_x

sensor based on SiC for

high-temperature applications

by

Johan Midbjer

Supervisor: Ruth Pearce, IFM

Examiner: Anita Lloyd Spetz, IFM

(2)

thouroghly characterised. The sensor layers are a mixed oxide of CoO, MgO and MgO2 deposited by thermal evaporation with a porous platinum gate on top,

deposited by thermal evaporation or sputtering. The sensitivity and selectivity of the sensor is promising and is shown to depend upon the ratio between Co and Mg in the ﬁlm and a number of competing mechanisms are shown to take place on the sensor surface. Response and recovery of the device is still slow and there are some drift, which are suggested to be due to a restructuring sensor surface during operation that was found by SEM-studies. Finally, the oxide surface has been characterized by XPS and a novel process for deposition of the sensor layers by lift-oﬀ technique has been developed.

(3)

Acknowledgements

First of all, a huge thanks to Anita, who gave me the possibility to do this thesis and also has given me some valuable input and encouragement along the way. I’m also very grateful to Ruth, for being so patient with my messy thoughts and quickly changing interests and focuses; you really helped me ﬁnd a focus in the fog and in addition made sure I had a good time!

A special thanks goes to Ingemar for all the help with ﬁxing malfunctioning equipment, correcting my erronous operation of equipment and helping me with identiﬁcation of, and solutions to, measurement problems. Mike and Kristina helped me sort out all types of questions when Ruth couldn’t assist me and also gave me some extra ears to complain to, I owe you for that and miss those ears when I now move on to other tasks. Also, Linda who helped me conduct the XPS-measurements and Mats who gave me the possibility to do them have my gratitude.

The project group from the autumn project also has a major role in the creation of this thesis, without them my ﬁrst meeting with sensor development wouldn’t have been that exciting! All others, whos company I’ve had the pleasure of en-joying at work, you’ve made my days fun and something to look forward to. You all know who you are.

And last, but certainly not least, many thanks to Caroline for living with my constantly changing mood everyday. It’s said that it’s the smallest things that are those that mean the most, and you show me that every day.

(6)

(7)

Chapter 1

Introduction

1.1 Goals of the project

The goals of this project has been to investigate the possibility to construct a NOx sensor based on silicon carbide for operation at high temperatures. Previous

studies in a project course at Linköping university [11] had indicated that a struc-ture of magnesium and cobalt oxides with a porous platinum gate as a promising structure.

Therefore, this structure has been fabricated and thouroughly tested in gas ex-posure tests. Gas response performance with diﬀerent amounts of Co and Pt has further been evaluated and diﬀerent device structures has been utilized. Finally, possible sensing mechanisms are discussed.

1.2 Applications

Transport, household heating and energy production are keystones in the modern society. At the same time the environment and public health is suﬀering from pollution caused by the emissions of particles and gases from these processes. One of the most harmful components in exhausts from vehicles and power plants are nitrous oxides (NOx). It is therefore desirable to ﬁnd methods to decrease the

discharge of NOxinto the atmosphere to as close to zero as possible. Furthermore,

future public legislations in the EU (Euro 6) and US will demand an increased speed of reducing the NOx content in exhausts from heavy trucks.

The sensor is required to function in a vehicle exhaust environment, where temper-ature often are 500-1000℃ . Further requirements on the sensor is that it should be cheap to manufacture and operate. The inclusion of such a sensor in a vehicle exhaust system is one possible alternative route to meet the new legislations. By

(8)

detecting the amount of NOx in the exhaust, the function of other devices

de-creasing NOx-emissions, i.e. catalysts, can be monitored; emissions can also be

reduced even further through regulation of the combustion engine with the aid of feedback control. Perhaps even more importantly, by detecting the NOx levels

one can inject ammonia into the exhaust to reduce the NOx to water and

nitro-gen, a so called SCR process [6]. This technique is eﬃcient, but the proportions of NH3 and NOx need to be exact - otherwise some NOx or ammonia-slip will be

left in the exhaust.

There are also several other similar applications for a NOx-sensor besides heavy

trucks: domestic burners, gas turbines and other environments where combustion of hydrocarbons take place and NOxemissions are required to be reduced further.

(9)

Chapter 2

The sensor

This chapter describes the basic theory regarding sensors which is required to understand and evaluate the results within the project. Some basic information on MOS-technology and the materials used is given. But ﬁrst, a short background on NOx is required.

NOx is a common name for a group of highly reactive gases consistent of

oxygen and nitrogen, of which NO and NO2 are two of the more dominant ones;

NO and NO2 are toxic gases that are formed in a reaction between N2 and O2

which takes place at temperatures higher than 1700 K [3], i.e. in a gasoline or diesel engine. The amount of NO compared to NO2 are at 104.4℃ somewhere

between 9/1 and 5/1[3].

2.1 Basic principle of gas sensors

A gas sensor is a device that in the presence of the target gas react with a measur-able signal, i.e. a CO sensor might change its resistance in the presence of CO or emit light. The typical gas sensor that is useful in industrial applications however change its electrical properties, since this is the easiest to measure directly. The gas sensor is therefore placed in an electric circuit. In an ideal system, if no target gas is present, the electric properties of the system are constant. However, when the target gas adsorb, a bond formation of some form between the gas and the sensor surface is required; which change the work function of the materials in the sensor. The change in work function is detected by alterations in voltage, resistance or capacitance over the sensor. If the interaction is too strong, a more permanent bond will be formed, in which case the sensor would adsorb target gas until no sites for adsorbtion are present, and then the change in properties would stop. From this knowledge of the overall principle a few desirable features of a

(10)

sucessful NOx sensor can be noted:

1. The sensor should allow NO and NO2 to adsorb easily on the sensor surface

2. The sensor should also let NO and NO2desorb easily from the sensor surface

3. The sensor should change its electrical properties signiﬁcantly upon adsorb-tion of NOx on the sensor surface

a) The change in electrical functions should be the same for equal con-centrations of target gas, no matter how much of the target gas that has already adsorbed on the sensor surface

b) The change in electrical properties should increase with increasing con-centration of NOx

c) The change in electrical properties should be signiﬁcantly stronger for NOx than for other gas compounds, so that the dominant contribution

to the sensor response comes from the target gas

4. Other gas compounds should have signiﬁcantly weaker adsorbtion probabil-ity compared to NOx, both to avoid the surface being poisoned with other

gases, and to avoid a too large contribution to the response of the sensor

2.1.1 MOS-sensors

The sensor that has been investigated is a metal-oxide semiconductor (MOS) structure. The principle behind the MOS structure is as follows: An (ideally) non-conducting layer is sandwiched in between the gate and the substrate, which are connected individually to a circuit. If the potential in the gate and the sub-strate is diﬀerent, a capacitance is formed over the non-conductive layer, how large this capacitance becomes depends upon the potential diﬀerence across the insulator layer and the amount of charge carriers in the material[10] (Eq. 2.1).

dC ≡ dQ

dV (2.1)

When exposed to the target gas the non-conducting layer changes it electrical properties which can be detected as a change in either resistance, capacitance, voltage or current.

2.1.2 Materials

Substrate

The device in this project is based on hexagonal silicon carbide (4H-SiC). The rea-son for using SiC as the semiconductor is due to the large bandgap (4H-SiC: 3.23

(11)

2.2. SENSOR CHARACTERISTICS 7

eV[9]), which makes operation even at such elevated temperatures as 1 000℃ pos-sible[16]. The high operation temperature is possible due to that the thermal energy not being large enough to excite all electrons from the valence band to the conduction band even at these high temperatures. The substrate was delivered with a 5µm n-doped epilayer to decrease the naturally high amount of majority carriers (electrons) in a SiC region which is close to the insulating oxide.

The ohmic contacts at the backside of the capacitor surface and in the case of a FET-structure: drain, source and gate all have the same structure; consisting of 50nm Ni, annealed in Ar 10min at 950℃ , 50nm and 400nm of Pt deposited on the substrate. The Ni forms an ohmic contact with the substrate and the TaSix

is added on top to increase the adhesion of the overlying Pt layer. The top layer of Pt is deposited to protect the contact from oxygen in the high temperatures needed for measurements[16]. This setup has been shown to work during many years of research at Linköping university[1, 11, 18]

Gate

The gate consist of a thin semi-porous Pt ﬁlm. Apart from being conductive this metal has also been shown to facilitate the decompostion of NOx in i.e.

cata-lysts[3]. Such a decompostion is likely to increase the probability of adsorbtion of NOx on the sensor surface, due to that both NOx and compounds from

disso-ciated NOx are available for adsorption on the sensor surface; which is made up

by part platina islands and part oxide mixture.

Oxides

The oxide layers are made up of Mg and Co oxide. Mg has two roles: to reduce the sensitivity to hydrogen by not binding strongly to hydrogen molecules, ex-perimentally shown by Andersson et al[1]. Furthermore, Mg has previously been shown to form rather strong bonds with nitrates and nitrides often formed by decomposing NOx[20]. Co has been shown to favour NOx adsorbtion and

de-composition when in a spinel phase [20]. Furthermore, Yu et al also showed that Mg stabilizes the reduction eﬀect that Co normally shows in a high temperature environment, so that hydrogen access to the Co ions in the oxide is obstructed, thus increasing the longevity of the sensor. By tuning the ratio of Co-oxide to Mg-oxide it should therefore be possible to engineer a surface with the desired properties.

2.2 Sensor characteristics

When dealing with sensors there are a number of common benchmarks of the sen-sor performance. For full understanding these benchmark concepts are presented

(12)

in short form below. For a more thourough treatment, the article [4] by D’Amico et. al. is recommended.

2.2.1 Transfer function

The transfer function, denoted f(x), describes the electrical output from the sensor as a function of the concentration of the target gas. In an ideal case the transfer function is linear, but is often more complicated.

2.2.2 Response

The response is the change in the electric signal from the sensor. This change is of course aﬀected by changes in concentration of target gas, but the response is not related to the amount of change in concentration of the target gas. There are several alternative deﬁnitions of response, but in this report (Eq. 2.2) is used. f(x0) corresponds here to the response at a standard state of output during

exposure to syntethic air (20% O2, 80% N2) which is often referred to as the

baseline.

Response, ∆S≡ f(x1)− f(x0) (2.2)

2.2.3 Sensitivity

Sensitivity also describes the change of the output signal, but here it is coupled to the change in input signal (concentration of target gas) which caused the change of output signal. Sensitivity is more easily described as the slope or derivative of the curve f(x)[4] (Eq. 2.3). Note especially that sensitivity is not a constant but will, in general, be diﬀerent for diﬀerent temperatures or other physical changes in the measurement environment. It is also to be noted that the overall sensitivity is, in general, depending on a row of internal sensitivities, i.e. to all the gases present, see D’Amico et. al. for further details.

Sensitivity≡ d

dxf (x) (2.3)

2.2.4 Selectivity

Selectivity describes the strength of response towards the target gas compared to the magnitude of response towards another gas upon individual exposure to each of the gases (Eq. 2.4). Selectivity is thus always determined individually for each reference gas that is tested, i.e. one NOx-sensor can be very selective

towards hydrogen but not selective at all towards CO. In a more theoretical view, the selectivity is the same as the quotient of the internal sensitivity to the target gas and the internal sensitivity to the other gas[4] (Eq.2.5).

(13)

Selectivity≡ ∆f (xtarget gas)

∆f (xref. gas) (2.4) Selectivity≡ i_S target gas i_S ref. gas

, whereiSx is the internal sensitivity to compound x

(2.5)

2.2.5 Cross-sensitivity

Cross-sensitivity is closely related to selectivity, but with the important diﬀerence that cross-sensitivity describes the change in response toward the target gas in

the presence of another speciﬁc gas.

2.2.6 Stability and drift

In order to have good stability, the sensor output shouldn’t change unless the conditions (gas concentration, temperature etc.) are changed. If there is a change in sensor output this is referred to as drift ; it is desired for all sensors to exhibit as small drift as possible, since it will aﬀect the accuracy of the sensor.

2.2.7 Response and recovery time

To describe how quickly the output signal is changed when target gas concentra-tion is introduced this report uses the T90 standard which gives the time needed

to go from baseline to 90% of the maximum response for each test gas concentra-tion as response time (Fig.2.1). The recovery time is by the same standard the time taken for the sensor output to change from maximum response to 10% of the max response.

2.2.8 Repeatability and Reproducibility

Two other important quality measures used are repeatability and reproducibility. The former describes whether or not one sensor gives the same response for several exposures to the same conditions when tests are conducted at diﬀerent times. Reproducibility on the other hand states whether or not two diﬀerent sensors, manufactured in the same way, give the same response when exposed at the same conditions to exactly the same gas environments.

2.2.9 Detection limit

The lower detection limit, xmin is a quantity that gives the smallest concentration

of target gas that can be detected by the sensor. One common deﬁnition of this for commercial purposes is Eq. 2.6, where σ0 corresponds to the standard deviation

(14)

Figure 2.1: Deﬁnition of response and recovery time

of the signal at baseline. In reality, the resolution of the sensor depends on the operation point. Therefore, the detection limit should be given together with information on at what conditions and what operation point this detection limit has been determined.

(15)

Chapter 3

Manufacturing,

characterization and testing

This chapter will describe the techniques used for production of the sensors and characterization of the sensors after production. An in-depth discussion of the test equipment is included.

3.1 Deposition techniques

The sensor devices were processed mostly by the use of thermal evaporation to deposit oxides, gate metal and contact pad materials onto the silicon carbide substrate. Also, some devices had the metal gates sputtered and the transistor devices also utilized the lift-oﬀ technique to pattern evaporated material on the substrate.

3.1.1 Thermal evaporation

For thermal evaporation, the substrate is mounted on a holder and placed inside a vacuum chamber; the materials to be deposited are placed in crucibles inside the same vacuum chamber (Fig. 3.1a). After pumping down to high vacuum the crucible with the deposition material is heated by i.e. appliyng an electric current through the crucible until the material melts and evaporates at a controlled rate; in the gas phase the evaporated species transfer up to the substrate, where it condense as the solid phase again. Compared to other deposition methods it’s simple and easy-to-use but has some limitations; materials with a high melting point cannot be evaporated since the crucible will fail if the temperature gets too close to its melting point. Also, ﬁlm adhesion is often poor compared to

(16)

energy methods like sputtering or ion arc deposition and control of growth modes and thicknesses aren’t as accurate as for epitaxial methods or sputtering.

3.1.2 Sputtering

When sputtering the plasma plays a central role. Plasma is a gas which is electri-cally conductive since it contains a non-negligible number of charge carriers. This plasma is created from an inert gas (i.e. Ar), which is leaked into the vacuum vessel. The plasma is formed by applying a strong electric field between the depo-sition material (the target) and the substrate, which ionizes the Ar. The plasma contains a large number of Ar ions, which are accelerated by the electric field towards the target; when the ions collide at high speed with the target, target atoms will be ejected. These ejected atoms adsorb on the substrate (Fig.3.1b). The sputtering technique allows for a better process control and a more diverse range of materials than can be deposited compared to thermal evaporation. Films deposited can be porous or dense, pure metal or metal and oxide or even mixed oxides. Whether porous or dense films are grown is controlled by adjusting the pressure. At normal conditions (Ultra high vacuum) sputtered films are denser than when evaporated, but sputtering at elevated pressures result in porous films. In addition, co-sputtering, where metal and oxide are deposited simultaneously to obtain a mixed, more thermally stable surface mimicing a porous metal on top of an oxide can be perfomed. This is rather common in sensors with porous metal gates, since all porous metal films are restructured by thermal and catalytic etching[18]. For this project, however, co-sputtering hasn’t been utilized due to limitations in time.

3.1.3 Lift-oﬀ

Lift-off is often used for deposition on more complex substrates, e.g. FET-transistors, where smaller or more complex structures are required. A photoresist film is first deposited over the whole substrate. The photoresist is then patterned by exposure to UV-light through a pattern mask; The UV-light degrades the pho-toresist, removing it from the surface in the desired pattern. The patterning of our samples was reformed by ACREO AB.

The desired sensor layers are thereafter deposited over the whole of the sub-strate. The desired patterning is achieved by leaving the specimen to soake in acetone or other strong solvent until the soluble film dissolves and the film de-posited on top of that is lifted away from the substrate, leaving dede-posited films only in the patterned areas. This technique does not allow annealing before, since this will cure the lift-off layer, resulting in difficulties in dissolving it. Therefore, devices produced by lift-off have been annealed at the same conditions (500℃, 16hrs, air) after all depositions has been made.

(17)

3.1. DEPOSITION TECHNIQUES 13

(a) A schematic drawing of an evaporation system

(b) A schematic drawing of a sputtering system

Figure 3.1: Layouts of processing systems

3.1.4 Preparatory cleaning

To ensure a good growth and decrease the amount of impurities in the deposited ﬁlms, the SiC substrate was cleaned before it was put in the vacuum chamber for deposition. The cleaning was done by exposure to UV-radiation in an air environment for 10 minutes. The UV-radiation creates ozone, O3, from the

nat-ural oxygen in the air. Ozone is highly reactive and reacts with adsorbates on the substrate and thus removes these impurity adsorbates and leaves a clean SiC surface for depostion.

3.1.5 Annealing

Annealing is a process that is performed for two reasons: first of all, to make the films deposited adopt a thermally stable configuration; secondly, it’s used for adaption of the films to a certain gas environment, normally oxygen or air.

After film deposition, the substrates are put in the oven at the target temperature (500 - 1 500 ℃) for times varying between 1 and more than 24 hours, film side upwards. During this time some gas typically air, N2 or O2, is flowed through the

oven, over the devices. The elevated temperature give a supply of extra thermal energy, which allows for reconstruction of the film to a more thermally stable con-figuration. The flowing gas at the same time reacts with the film, which results in the film being saturated with species of the flowed gas.

(18)

Figure 3.2: A sensor with heater and Pt-100 mounted on the 16-pin header

3.2 Sensor packaging

Once the sensor chip ﬁlms have been deposited, the sensors are in principle ready and it should be possible to start measurements. However, a few additional details are required before plug-in. These are often referred to as packaging. After packaging the sensor is ready to be put in a measurement system. Packaging normally involves:

• A standardised contact platform. A standard 16-pin header was used • A heater. Since the sensor performance depends very much upon

tempera-ture it’s important to be able to change the temperatempera-ture of the sensor chip fast and accurate. The temperature sensor should be placed close to the sensor to increase the accuracy further

• A temperature sensor. To be able to control the temperature of the sen-sor chip it’s absolutely necessary to have a temperature sensen-sor that gives continous information about the temperature at or near the sensor chip

To make this packaging, the heaters were mounted onto the 16-pin header by means of spot-welding. Thereafter, the sensor chips and the temperature sensor (Pt-100, Platinum sensor) were glued onto the heater using a ceramic glue (Ceramabond 571-P, Aremco Ltd). The temperature sensor was spot welded onto the 16-pin header. The sensor chips were instead connected to the 16-pin header by connecting the sensor chip and the contact pins on the header by means of

(19)

3.3. THE DEVICE 15

gold-ball bonding and a thin gold wire. For some of the structures (Device 1-3), the ﬁlm adhesion of the pad contact wasn’t good enough to use gold-ball bonding to make the gold wire stick. In these cases a conductive silver glue (Nanotach®,NBET ECH) was used for connecting the gold wires to the contact

pad of the sensors.

3.3 The device

An outline of the device can be seen in ﬁgure 3.3. The various thicknesses of devices is listed in table 3.1.

Figure 3.3: Overall device structure

Device 1 Device 2 Device 3 Device 4 Device 5 Device 6

MgO [Å] 150 150 150 150 150 150

Co [Å] 12 50 100 50 75 50

Pt [Å] 200 200 150 150 150 150

Sputtered Pt? No No No Yes Yes Yes

Type Capacitor Capacitor Capacitor Capacitor Capacitor FET Table 3.1: A summary of devices produced and tested

(20)

3.4 Characterisation methods

To examine the overall structure of the ﬁlm surface and especially the degree of porousity of the top Pt-layer the samples were examined by SEM both before and after operation. This also gave the possibility to investigate the reasons behind sensor failures more closely. To better understand what compounds that contribute to the sensing mechanism the top-most 100Å of the oxide layer surface was investigated by means of XPS.

3.4.1 Scanning Electron Microscopy, SEM

SEM is a form of microscopy where electrons, instead of photons, are used to create an image of the surface. The higher energy of electron beams (0.1-50 keV)[12] allows for a higher resolution of the images. In modern SEMs it’s possible to reach resolutions less than 1nm, and 20nm features are resolved by most SEMs in operation today[12]. The image in SEM is based upon interaction from the sample with the injected electron beam. Unlike for optical microscopy, a SEM picture is built up by information on both topography, electrical properties and elemental compostion of the sample.

Instrument layout

A typical layout of an SEM instrument can be found in Fig.3.4. The electron beam is generated at the source and focused into a narrow beam by the aperture and the condenser lenses, the scan coil is used to control the beam towards the spot intended. The objective lens is used for focusing the image to obtain a clear image. Upon irradiation, a lot of phenomena arise in the specimen: secondary, auger and backscattered electrons are ejected and characteristic x-rays are emitted. All of these can be used to obtain information about the specimen, but morphology and topography pictures mostly utilises the secondary electrons (SE) and the backscattered electrons (BSE) to create the image. These electrons are detected by the electron detector, which can be of several diﬀerent types. The energy of the SE are less than 50 eV[12], thus, detection of SE can only be used to characterize the top-most of the specimen, whereas the BSE with their higher energies (up to irradiation energy) can travel deeper into the material and give information also about the bulk.

The image

The image is created by means of scanning the electron beam over the sample collecting the electrons emitted from the specimen at each spot scanned. The SE can be collected by appliyng a positive bias to the detector, which is then placed at one side of the sample. The larger energies of the BSE makes them impossible

(21)

3.4. CHARACTERISATION METHODS 17

Figure 3.4: A typical layout of an SEM-instrument

to collect by biasing; the detectors are therefore placed so that they cover a large solid angle collection[12]. All of the images from SEM are given in grayscale, where light colors correspond to one extreme of a material property and dark colors correspond to the other extreme of the same property. Which properties that are expressed in the pictures diﬀers depending on the imaging mode used and here the two most common imaging modes are inspected in more detail.

SE-mode

The most common operation mode is SE-mode, where only SE are collected and analyzed. In SE-mode, topography of the sample can be examined at high mag-nification (1-10 nm[12]). The low energy of the SE also implies that they are effected by positively and negatively biased areas in the specimen, resulting in positively biased areas appearing more dark and vice versa for negatively biased areas. In addition, the trajectories of the SE can be affected by magnetic fields caused by ferromagnetive domains in the specimen. The image created by the SE

(22)

is a combination of all these types of information, which makes SE an ideal choice for examination of the morphology of the specimen.

BSE-mode

One of the disadvantages of the low-energy SE, are that the trajectories aren’t straight in the electric field applied by the detector. The energetic BSE however, move in straight trajectories also in the relatively strong electric filed from the SE-detectors. This facilitates a better shadow effect in the topographic analysis, at least for lower magnifications. When increasing the magnification too much, the large information volume from the BSE simply result in a worse resolution compared to SE-mode. The resolution can be improved by energy level filter-ing[12]. The most important advantage of the BSE-imaging lies in the possibility to also find phases in the specimen with different atomic number, ¯Z. This is

pos-sible since the kinetic energies of the BSE are highly dependant on the masses of the particles against which the BSE has been scattered. It should be noted that this information is also present in the SE spectra, since 50-80% of the SE-spectra originates from SEs that have been exicted by backscattered electrons on their return path through the specimen. But in conclusion, ¯Z is more dominant in the

BSE-spectra. It is also possible to retrieve some information on crystal structure and ferromagnetism from the BSE, but that is outside the scope of this thesis.

3.4.2 X-ray photoelectron spectroscopy, XPS

XPS is a non-destructive technique which relies upon stimulating emission of secondary electrons from the atom shells by irradiating the sample with X-rays and analyzing the energy spectra of these secondary electrons. XPS can be used for analysing the elemental composition of the top 5-10 nm of a the sample, to determine the chemical state of materials in the sample, or, to determine empirical formulas of pure materials. In addition, XPS can also be used to detect adsorbants on the material surface and can thus be used to i.e. analyze the surface coverage of adsorbed gas species on the sample surface. By depth proﬁling it’s possible to investigate deeper into the material, but then the non-destructivity is lost. Due to the precise energy resolution needed and to avoid oxidation of the surfaces, experiments must be performed in UHV[2].

Detection principle

The X-rays used in XPS are typically of energies between 1-1.5 keV[5]. When the photons enter into the material, they will be adsorbed by individual atoms; this will release energy that can break the bond between one electron and the atom core and add kinetic energy to the electron, making it escape the atom. This phenomenon is known as the photoelectric eﬀect (Fig.3.5) - irradiation with

(23)

3.4. CHARACTERISATION METHODS 19

a photon of energy larger than the bond energy of one electron causes emission of that electron.

Figure 3.5: Illustration of the photoelectric eﬀect phenomenon

The x-rays can penetrate deep into the material and cause a photoelectric eﬀect. However, the energies of the free electrons are quickly attenuated, and only the electrons from the uppermost 5-10 nm of the material will escape with a energy large enough to be detected. This attenuation is what limits how deep into the specimen XPS can be used for analysis. The escaping electrons are collected by an electron collection lens into a detector with an electron energy analyzer and an electron detector, which gives the energy distribution of the electrons and counts the number of electron hits. This data gives the XPS spectra.

Analyzing the spectra

The spectra is usually given as hits, or number of electrons detected, versus the energy of the electrons detected. Elements give rise to specific sets of peaks cor-responding to the electron distribution of each element (eg. 1s, 2s, 2p etc.). The relative intensity distribution of these ”fingerprints” directly gives the percentage of different elements in the materials.

The energy of the electrons detected can be converted into the corresponding binding energy, by use of Eq.3.1. Here ϕ is the work function of the spectroscopic instrument used. Since this and the energies of detection (Ekinetic) and of the

x-ray radiation are known, the whole spectra can be converted to show the binding energy directly, which is normally the case in XPS instruments.

(24)

Ebinding= EXray− (Ekinetic+ ϕ) (3.1)

3.5 Measurement method

The measurement method used depend on which structure the underlying sub-strate has. Most of the structures used are simple capacitive structures as in ﬁg.3.6a. This is a structure that is well suited for researching new sensor layers; it’s easy and relatively cheap to manufacture and simple to mount. However, the signal is often rather noisy and responses can be small. Therefore, in a more applied context, a FET or other transistor structure is often used; responses are then larger and the signal more stable. The long term stability is also improved, since the layout of the FET-device increases the passivisation of the gate[13] and also eliminates the need for a constant current through the whole device.

(a) Capacitive structure (b) MISiC-FET structure

Figure 3.6: The diﬀerent sensor structures used

3.5.1 Capacitive measurements

The principle behind measuring on capacitive structures can be summarised as follows. The sensor is placed in an electric circuit that measures voltage and/or resistance, while keeping the capacitance in the device at a constant level. The sensor signal is the change in voltage and/or resistance.

Measurements were performed using the system described in Fig. 3.7, con-sisting of a gasmixer system, an electronics box, a temperature controller, a gas measurement cell, a capacitance meter and a computer with software to control

(25)

3.5. MEASUREMENT METHOD 21

Capacitance and voltage/resistance signal Feedback control signal Raw data [V/Ω] Heater voltage Sensor signal Gas mix flow Gas flow specifica$on Gas mixer

system Measurement cell Electronics box Capacitance meter Gas Mix seq.

Temperature controller Ref. temp

Gas waste Ref. capacitance

Figure 3.7: The capacitive measurement system

the gasmixer system and record data from the electronics box and capacitance meter. The gasmixer system was custom-built at the department of Applied Physics and gas mixtures are created by ﬂow regulators controlled with the Gas-Mixer software. A picture of the reality can be seen in ﬁg.3.8. The temperature controller is handled manually by setting a reference temperature which must be changed manually.

Figure 3.8: A set-up with two measurement cells

The electronics box and the capacitance meter (Boonton 7200 Series Capaci-tance meter) produce the raw data from the measurements together. The

(26)

capaci-(a) Accumulation (b) Depletion (c) Inversion

Figure 3.9: Band diagrams for an ideal p-type MIS capacitor

tance meter analyzes the signal in terms of voltage/resistance and capacitance and send a feedback control signal to the electronics box until the reference capaci-tance is reached. The voltage/resiscapaci-tance signal is then forwarded to the computer, where it is recorded. The capacitance meter utilizes a measurement frequency of 1 MHz and can measure in the capacitance range from 0-2000 pF with a resolution of 0.1-1 pF, depending on which measurement regime that is used[19].

3.5.2 CV-curve regions

The capacitive device has three regions of operation: accumulation, depletion and inversion (Fig. 3.9). In the MIS-junction, band bending and concentration of carriers occur due to the diﬀerent fermi levels in the material. Depending on the type of doping and the voltage applied, the carrier distribution in the junction is diﬀerent. For a p-type doped substrate majority carriers (holes) will be attracted towards the metal if a negative bias is applied to it. However, they cannot pass through the oxide. This is the accumulation mode, named after the accumulation of carriers close to the insulator. The whole semiconductor is conductive, since the carriers can easily move in the semiconductor. The total capacitance in this mode is constant and close to the capacitance of the oxide.

As the bias is increased, the concentration of carriers close to the junction is decreased. At zero bias, the bands are completely ﬂat. But as soon as the bias becomes positive, the bands in the semiconductor are bent downwards and this will subsequently decrease the number of majority carriers at the insulator border even more. This results in the formation of a depletion layer close to the insulator, where the carrier concentration is lower than in the semiconductor bulk. The depletion layer behaves as an extra insulator layer and the total capacitance is thus decreased, which is characteristic of the depletion mode. The reason why the depletion region isn’t always found at positive voltages is simply that the model used to describe these modes rely on an ideal model where the insulator

(27)

resistance is inﬁnite, all charges exist only in the semiconductor and there is no diﬀerence in Fermi level between metal and semiconductor. In reality, this isn’t entirely true, and these non-ideal properties cause the CV-curve to shift.

If the bias is increased further so that the intrinsic Fermi level at the insulator surface crosses over the Fermi level in the semiconductor, a state called inversion is reached. In inversion mode, minority carriers are accumulated close to the insulator surface. If a measurement frequency larger than or equal to 1 MHz is used, the minority carriers cannot move fast enough to follow the rapidly changing electric ﬁeld. As a consequence, the total capacitance is again constant and much less than the insulator capacitance. For some structures, it might be possible to use frequencies down to 100 kHz and with the normal sensor measurement frequency of 1 MHz all set-ups show this behaviour.

Measurement procedure

Before the measurement can begin, a suitable reference capacitance must be se-lected. This is done by setting the temperature controller to the desired value. Thereafter, there are two ways to proceed; the first is to record the capacitance for voltages between -5 and +5 V, while flowing first 100ml/minute syntethic air and then record a second CV-curve for 100ml/minute of 500ppm of NO in syn-tethic air over the sensor. The two obtained C-V curves are then plotted and compared to find the capacitance for which the difference between the two curves is maximal: this is the reference capacitance for the sensor at that temperature. This method is very suitable if the contact with the sample is poor. If the contact is good, the large difference will always be in the depletion region where the slope of the C(V)-curve is steepest (Fig.3.10). Therefore, in most applications, this can be used to directly choose the working capacitance. The first method always pro-vides an extra certainty in the result but demand more time for measurement set up, which is often unnecessary. The chosen reference capacitance for the sensor is then set in the computer software, and the same procedure is repeated for all sensors in each measurement.

After programming the desired gas-mixing sequence in the GasMixer software, the measurement can be started in the computer software and will run until the gas-mixture sequence is ﬁnished while saving raw data continously. This setup does not allow for readout of sensor response during experiment, which is another drawback as malfunctions cannot be detected until after the measurement is completed.

(28)

Figure 3.10: A ideal CV-curve

3.5.3 Transistor measurements

As mentioned, transistor-based structures are somewhat more complex to process, but in return provide a more stable output with larger responses, thus improving the sensitivity and detection limit of the sensor. The gas sensor FET has, as can be seen in fig.3.6b, four contacts: source, drain, gate and substrate. The key to the function is that by adjusting the electric field in the channel, which is done by adjusting the voltage on the gate contact, the current between source and drain can be controlled. In the device used, gate and drain contact are interconnected and biased, while source and substrate are both grounded. In this setup, the device is normally off, that means that no current can flow through the device. To enable a current flow, the drain voltage must be increased.

IDS =

k

2(VGS− VT)

2_{, where: k =} Z

LµnCox (3.2)

When operating as a sensor, the most common operation mode is punch-through, when the drain voltage is large enough to cause depletion all the way into the source area, and thus carriers are injected directly from drain to source. The drain current is then given by Eq.3.2. The current thus depend upon the carrier mobility in the channel and the capacitance of the gate oxide as well as the gate width (Z) and lenght (L). In this operation mode the FET-device has a

(29)

characteristic, unsaturated I-V characteristic[18].

Figure 3.11: The transistor measurement system

Measurement method

The key to the operation as a sensor, is that the I-V characteristics are altered when the work function of the gate is changed by adsorbing gas molecules. To detect this change, the device is operated in either a constant current or constant voltage mode and the change in current or voltage is detected. However, since the adsorbed species in fact contribute with further charges and dipoles, the constant current is the more reliable mode of operation[18]. The setup utilized for this project uses a constant current setup and electronics and LabView® software from National Instruments to control the measurement and record the measure-ment data (Fig.3.11).

(30)

(31)

Chapter 4

Results

In this chapter the results in the scope of the thesis will be presented. They will be presented for each device tested and summarized in the next chapter. All devices are based on SiC and has a structure as shown by Fig. 3.3, with the MgO layer being approximately 150Å thick with variyng thicknesses of the Co- and Pt-layers. A full list of the devices used can be found in table 3.1. Due to high noise levels all sensor signals have been ﬁltered by nearest neighbour averaging prior to evaluation and presentation.

4.1 Device number 1 (50Å Co, 200Å Pt)

Earlier results [11] suggested a NOx-sensitive structure based on CoO and MgO.

Therefore, the same structure was remade, to verify that the structure is indeed sensitive to NOx. Gas exposure tests were performed at 400℃ with NO and

NO2 as test gases (Fig. 4.2). It is seen that the sensor is sensitive to NO and

NO2 with a downward response of 50-250mV. The recovery time is in the order

of minutes to infinite, but baseline is stable. Thereafter the devices failed, and no useful information could be extracted from the data of later measurements. The devices were also inspected by SEM after and prior to usage; images can be seen in Fig. 4.1. A significant restructuring of the surface of the Pt and the underlying oxide was noted. Also, the film exhibited small pores already before operation.

4.2 Device number 2 (12Å Co, 200Å Pt)

Since the amount of Co has been shown to play a major role in the catalysis function of similar structures[20], the eﬀect of Co concentration in the ﬁlm was examined. Gas exposure tests were performed at 400℃ with NO and NO2 as

(32)

(a) Device 1 (50Å Co, 200Å Pt) before op-eration, showing a smooth ﬁlm with small pinholes through both oxide and Pt-layers

(b) Device 1 (50Å Co, 200Å Pt) after oper-ation. Both the Pt and the oxide has clus-tered together.

Figure 4.1: SEM-images of device surface both before and after operation

test gases (Fig. 4.4). Below is shown that the sensor is clearly sensitive to NO and NO2 with a downwards response and unstable behaviour. Recovery is slow

(5 minutes to inﬁnite), but baseline is stable. During continuoed exposure the devices failed, and no further information could be extracted. The devices were also inspected by SEM after and prior to usage; images can be seen in Fig. 4.3. For this device, the oxide was restructured into large grains even before operation, and the SiC substrate also exhibit cracks (Fig. 4.3b, 4.3a). The size of the holes in the oxide ﬁlm and the Pt- and oxide agglomerates have increased during operation, indicating a continouing restructuration of the surface.

4.3 Device number 3 (100Å Co, 150Å Pt)

To test if more Co alter the sensor properties, a device with a twice as thick Co-layer as Device 1 was produced. Also, a 50Å thinner Co-layer of Pt was deposited to investigate how this affected the structure of the Pt-film and the perfomance of the sensor. SEM-images of the surface before and after operation are seen in Fig. 4.5. Some restructuring of the films could be noted, but less prominent than for Device 1 and 2, despite being a thinner film. Unfortunately, due to packaging issues, only one gas exposure measurement could be made on this structure before the contact failed. This measurement was performed at 300℃ with NO (Fig. 4.6) and showed a similar behaviour as for Device 1, once stabilized, with an improved baseline stability. It is interesting to note that the sensor needs to be exposed to higher concentration of NO before it becomes sensitive to NO at all.

(33)

4.4. DEVICE NUMBER 4 (50Å CO, 150Å PT) 29

4.4 Device number 4 (50Å Co, 150Å Pt)

For reference and further measurements, a sensor with the same amount of Co as in device 1 but with a 50Å thinner layer of Pt was fabricated. The first batch of sensors suffered from identical problems as Device 3, and no measurements could be performed. But a second batch was produced, where the platinum film was also sputtered to ensure porousity. SEM-images of the surface before and after operation was taken (Fig. 4.7), revealing a porous Pt film and pronounced step edges from the SiC being visible in the oxide films, possibly indicating a strong adhesion between oxide and SiC. The Pt-film contains some smaller clusters after operation, but much less prominent than for the evaporated films.

Gas exposure tests were conducted with NO (Fig. 4.8) as well as NO2 (Fig.

4.9) along with some selectivity and cross-sensitivity tests (Fig. 4.8,4.10). Only a selected amount of data is given, but this data represent the behaviour of the device, showing response to NO between 30 and 150mV. There is also a response to NO2, but response and recovery is slower and drift more prominent. Also, response

to NO and NO2 are in diﬀerent directions. For lower concentrations of NO, the

response is also in the opposite direction compared to at higher concentration during some measurements (Fig. 4.13, 4.8. Selectivity of the sensor is promising: only CO, C3H6and H2caused any response, and the responses are smaller than for

both NO and NO2. The cross-sensitivity to oxygen is excellent, since response to

NO and NO2 do not change magnitude with diﬀerence in oxygen concentration. A

drawback is the systematic drift of the baseline. For all conducted measurements, this has been ongoing during the whole measurement time, up to 22 hours. Since no ﬂow of gas takes part between measurement, nothing can be said about if this drift is just a stabilization process or an ongoing process. The sensor showed good stability, lasting for approx. 100 hours of measurements before failing. The given measurement data comes from two diﬀerent senor dots produced in the same way.

4.5 Device number 5 (75Å Co, 150Å Pt)

For Device 5, a 75Å CO layer was deposited to try and combine properties of Devices 3 and 4. As for Device 4, the Pt-film was deposited by sputtering. The surface was investigated by SEM before and after operation (Fig. 4.14). Before a porous Pt film and pronounced step edges from the SiC being visible in the oxide films, which also contained some holes, indicating a strong adhesion between oxide and SiC. After operation the Pt has clustered into slightly larger islands, but less prominent than for the evaporated films. This was most clear around scratches or other larger holes through the oxide layer. The oxide is stable.

Gas exposure tests were conducted with NO (Fig. ??,4.15) as well as NO2

(Fig. 4.19) along with some selectivity and cross-sensitivity tests (Fig. 4.17,4.16). Only a selected amount of data is given, but this data represents the behaviour of

(34)

the device. Directions of response are diﬀerent for NO and NO2, as also seen for

Device 4 (Fig. 4.9, 4.8). However, selectivity (Fig. 4.17, 4.18) and baseline drift is poorer than for Device 4. Also for this sensor, doping with NO/NO2 is needed to

obtain a good response. Stability of these sensors were overall better, they lasted and gave a rather consistent output for more than 150 hrs of measurements.

(35)

(a) Output from device 1 (50Å Co, 200Å Pt) when exposed to 10 min pulses of NO (25-500ppm), 400℃. The pre-annealing with NO results in the direction of response being constant. A clear response and a long recovery time can be noted.

(b) Output from device 1 (50Å Co, 200Å Pt) when exposed to 10 min pulses of NO2(25-250ppm), 400℃. Response time is longer compared to for NO and

direction of response is the same. Recovery time is inﬁnite, and after the exposure to 250ppm NO2 the sensor is broken. All the responses show a

double peak feature at maximum response.

(36)

(a) SEM-images of device 2 (12Å Co, 200Å Pt) oxide surface before operation, showing an already restrucutred oxide surface

(b) SEM-images of device 2 (12Å Co, 200Å Pt) Pt surface after operation. Both Pt and oxide are already clustered.

(c) SEM-images of device 2 (12Å Co, 200Å Pt) oxide surface after operation, showing more large pinholes in the oxide ﬁlm com-pared to before operation

(d) SEM-images of device 2 (12Å Co, 200Å Pt) Pt surface after operation, showing even larger Pt and oxide grains.

Figure 4.3: SEM-images of device 2 (12Å Co, 200Å Pt) surface before and after operation

(37)

(a) Output from device 2 (12Å Co, 200Å Pt) when exposed to 10 min pulses of NO (25-500ppm), 400℃. The pre-annealing with NO result in the same direction of response for all NO-pulses. Responses are between 70 and 200mV, and recovery time slow and overall behaviour noisy compared to Device 1. All the responses show a double peak feature at maximum response, in agreement with Device 1.

(b) Output from device 2 (12Å Co, 200Å Pt) when exposed to 10 min pulses of NO2(25-500ppm), 400℃. Responses are unclear, and the overall behaviour

very noisy compared to Device 1. As for Device 1, the sensor failes after some time of exposure.

(38)

(a) Surface of Device 3 (100Å Co, 150Å Pt) before operation, showing a Pt ﬁlm with larger grains compared to devices 1 and 2.

(b) Surface of Device 3 (100Å Co, 150Å Pt) after operation, showing slight agglomera-tion of the Pt-ﬁlm to larger islands, but less clear than for Devices 1 and 2. Also, some holes in the oxide ﬁlm have formed.

Figure 4.5: SEM-images of device 3 (100Å Co, 150Å Pt) Pt surface before and after operation

Figure 4.6: Output from sensor device 3 (100Å Co and 150Å Pt) when exposed to pulses of NO, 300℃. The response to initial exposures does not recover, but upon exposure to 500 ppm of NO, a change in direction of response occurs. Responses after this are of the order 30-150mV and recovery improved compared to devices 1 and 2.

(39)

(a) Surface of Device 4 (50Å Co, 150Å Pt) before operation. Note the clearly visible steps i the sensing layers and the small holes through the ﬁlm

(b) Surface of Device 4 (50Å Co, 150Å Pt) after operation. The steps in the sensor layers are still clearly visible and some small clustered Pt grains are visible.

Figure 4.7: SEM-images of device 4 (50Å Co, 150Å Pt) Pt surface before and after operation

Figure 4.8: Response of sensor dot number 1 of type 4 at exposure to 10 min NO (25-500ppm) pulses with 20, 10 and 5% O2 background, 400℃. The direction of

response changes for exposures under and over 50ppm of NO, and some minor drift can be seen. The cross-sensitivity to O2 is low.

(40)

Figure 4.9: Response of sensor dot number 1 of type 4 at exposure to 10 min NO2 (25-250ppm) pulses, 400℃. Initially, the responses are weak. As for device 3

with NO, there is a need of exposure to high concentration of NO2 before a good

sensistivity to NO2 can be obtained. The direction of the response for the device

is diﬀerent between NO and NO2.

Figure 4.10: Output from sensor dot number 1 of type 4 at exposure to 10 min pulses of H2, CO and C3H6, 400℃. Some minor responses could be seen. These

are in the same direction as for NO, and never larger than 30mV. The CO gas ﬂow did not work during this measurement, and response to CO can instead be seen in Fig.4.12.

(41)

4.5. DEVICE NUMBER 5 (75Å CO, 150Å PT) 37 250 25 50 100 200 250 250 200 100 50 25 ppm NO 2 50 Å Time (approx. 4 hrs) B -0.2 -0.1 0.0 R e sp o n se [ V ]

Figure 4.11: Response of sensor dot number 1 of type 4 when ﬁrst exposed to 250 ppm NO2 and thereafter equal pulses as in Fig. 4.9, 400℃. The doping with NO2

increases the sensitivity of the sensor, but response and recovery is slower for NO2

compared to NO, and still in the opposite direction. After this measurement this sensor failed.

25 50 100 200 500 500 200 100 50 25 ppm CO

Figure 4.12: Response of sensor dot number 1 of type 4 at exposure to 10 min CO (25-500ppm) pulses, 400℃. The direction of response is the same as for NO, and smaller than 20mV. Note the sharp transient peaks in the responses.

(42)

Figure 4.13: Output from sensor dot number 5 of type 4 when ﬁrst exposed to 500 ppm NO and thereafter equal pulses as in Fig. 4.9, 400℃. Also here, the doping increased the sensitivity of the device and caused all responses to go in the same direction. The doping eﬀect is temporary since the response to 50mV NO is again in the opposite direction. Response and recovery is faster than for NO2, but drift is more prominent.

(43)

(a) Surface of a Device 5 (75Å Co, 150Å Pt) Pt-ﬁlm before operation. Note the small holes and islands in the Pt-ﬁlm and the sharp steps

(b) Surface of a Device 5 (75Å Co, 150Å Pt) oxide before operation. Note the small holes and the sharp steps

(c) Surface of a Device 5 (75Å Co, 150Å Pt) Pt-ﬁlm after operation. Note the slightly larger islands and still sharp steps

(d) Surface of a Device 5 (75Å Co, 150Å Pt) oxide after operation, still smooth with sharp steps

Figure 4.14: SEM-images of device 5 (75Å Co, 150Å Pt) pt and oxide surface before and after operation

4.6 Device number 6 (50Å Co, 150Å Pt)

To investigate the sensor performance in a FET-structure the structure with the amount of cobalt that had been tested the most as a capacitor was put in a FET-structure instead. The contact of the samples were not excellent, but allowed for some gas exposure tests to be conducted. For the FET, response and recovery is much slower, but the sensor showed response to both NO and NO2. Response

to NO2 is signiﬁcantly stronger than to NO (Fig. 4.23, 4.21). Most surprising is

the signiﬁcant drift that was shown for all measurements, but in particular upon introduction of diﬀerent concentrations of H2 in the background (Fig. 4.22).

(44)

Figure 4.15: Output from sensor dot number 3 of type 5 (75Å Co and 150Å Pt) at exposure to 10 min NO (25-500ppm) pulses, 400℃. Direction of response is changing upon exposure with 500ppm NO as for Device 3, and the response and recovery is also much faster after exposure to 500ppm NO. The sensor signal exhibit some drift, which is decreasing towards the end of the measurement.

4.7 XPS-measurements

The surface of the oxides in device 4, without Pt deposited, were examined after annealing by XPS to determine the exact composition of the uppermost 8nm of the sensor surface. The resulting spectra are shown in Fig. 4.24, 4.26, 4.25, 4.27. The surface of the sensor is a mixture of CoO and Mg-oxides with two diﬀerent oxidation states. Furthermore, the high amount of C and Si on the surface reveals that the holes in the oxide ﬁlm exhibited by SEM-studies of devices 4 and 5 (Fig. 4.14b, 4.7a) go all the way through the oxide to the SiC substrate. Note also that the surface is almost free from other pollutants, apart from a small portion of Na, probably released from the glass in the oven used for annealing. This means that there are mainly three agents on the surface that can interact strongly with the gas phase: Pt, CoO and the Mg-oxide.

(45)

4.7. XPS-MEASUREMENTS 41

Figure 4.16: Output from sensor dot number 3 of type 5 (75Å Co and 150Å Pt) at exposure to 10 min NO (25-500ppm) pulses with 20, 10 and 5% O2 background,

400℃. The responses are very similar to Fig.4.15 and undependent of the back-gropund oxygen, showing that the cross-sensitivity to oxygen is low. As for Device 4, the direction of response is changed upon exposure to 100ppm NO.

Figure 4.17: Output from sensor dot number 3 of type 5 at exposure to 10 min pulses of H2, CO and C3H6, 400℃. The CO gas ﬂow did not work, and data on

CO exposure can be found in Fig.4.18. The direction of response is the same as for NO and opposite to NO2. Selectivity is poorer compared to Fig.4.10. Also, a

(46)

Figure 4.18: Output from sensor dot number 3 of type 5 at exposure to 10 min pulses of CO, 400℃. Response is slower and weaker than in Fig.4.12 and in the same direction as for NO.

250 25 50 100 200 250 250 200 100 50 25 ppm NO 2 75 Å Time (approx. 4 hrs) -0.2 -0.1 0.0 0.1 0.2 R e sp o n se [ V ]

Figure 4.19: Output from sensor dot number 3 of type 5 when ﬁrst exposed to 250 ppm NO2 and thereafter 10 min pulses of NO2 (25-250ppm), 400℃. The

initial response to 250ppm is slow, but for remaining pulses response times are shorter. Both recovery and response are slower than for NO-exposure and also in the opposite direction. There is a systematic drift upon NO2 exposure, possibly

(47)

Figure 4.20: Output from sensor dot number 3 of type 5 when ﬁrst exposed to 500 ppm NO and thereafter equal pulses as in Fig. 4.15, 400℃. Response and recovery is similar to Fig. 4.15. Unlike Device 4, the eﬀects of doping with 500ppm NO causing the responses also at concentrations below 100pm NO to go in the same direction as for concentrations above 100ppm NO are retained after air exposure for 1 hr. 2.62 2.64 2.66 2.68 2.70 2.72 2.74 25 50 100 200 500 500 200 100 50 25 ppm NO R e sp o n se [ V ] Time(approx.15 hrs) 25 50 100 200 500 500 200 100 50 25 ppm NO

Figure 4.21: Output from the FET SiC transistor at exposure to 13 min pulses of NO (25-500ppm), 400℃. The response and recovery is slower compared to the capacitive devices 1-5 and drift very prominent . The magnitude of the responses are also smaller compared to devices 1-5.

(48)

Figure 4.22: Output from the FET SiC transistor at exposure to 13 min pulses of NO (25-500ppm) with varying H2 background, 400℃. Drift is still prominent

with slow response and recovery. There is a big shift in baseline between 10% and 5% H2 background, and responses and recovery to NO is quicker with 5% H2

background than with 10% H2 background.

Figure 4.23: Output from the FET SiC transistor at exposure to pulses of 13 min pulses of NO2(25-250 ppm), 400℃. Response is faster compared to NO and in the

same direction as NO. Sharp peaks up and down are observed at the beginning of NO2 exposure and also at the end on NO2 exposure during the second half of

(49)

Figure 4.24: XPS-spectra of the oxide surface of Device 4 after annealing. Peaks from CoO and Mg-oxide with two exictation states can be seen. There are also peaks from Si and C, indicating that there are holes through the oxide to the SiC. Some minor Na pollutant is present, probably from the glass in the annealing oven.

Figure 4.25: Zoomed in XPS-spectra of the O2(1s)-peaks, showing a split in two

caused by both Mg-oxides and CoO being present at the surface and a broadening of the top of the Mg-oxides, indicating that there are several oxidation states of this present.

(50)

Figure 4.26: Zoomed in XPS-spectra of the Co(2p) tops, showing that the type of Co-oxide present is CoO

Figure 4.27: Zoomed in XPS-spectra of the Mg(2p) tops, showing that there are several oxidation states of Mg present at the surface, indicating several types of Mg-oxides in the ﬁlm.

(51)

Chapter 5

Discussion

5.1 Material characterization

5.1.1 Pt ﬁlm stability

As seen from the SEM-pictures (Fig. 4.1a, 4.3b, 4.5a) the samples processed by thermal evaporation does not exhibit a porous Pt-film. Furthermore, after operation for at least 8 hours at elevated temperatures the film is significantly re-structured (Fig. 4.3d, 4.3c, 4.1b, 4.5b), leaving the Pt dispersed in crystal islands over the whole surface. For Device 2, oxide was restructured aldready during annealing. The restructuring is less dominant for device 3 (100 Å Co). That evaporated Pt-films are rather dense on these oxides is expected from earlier re-sults[11], but the restructuring of the film is a new feature compared to these results, although common in other cases [15]. The question of whether it is an ongoing or time-limited process is still unknown, but most likely it is due to a high mobility of Pt-atoms on the mixed-oxide surface and to interaction with the gas phase, as decribed for other films and gases [15, 7] and thus an ongoing process until clustered into large islands and thereafter continouing, but at a slower pace.

This restructuring is somewhat problematic, since it changes the sensor char-acteristics [15] and eventually even causes failure of the sensor. On the other hand, for the evaporated ﬁlms, the island formation intially increases the surface area compared to the more dense Pt-ﬁlm originally deposited. Hence, a lot of three phase boundaries where gas, metal and mixed oxide can meet are created. These points are the most reactive in terms of the sensor mechanism, and the reactivity of the surface is increased [3]. However, if the Pt is clustered in isolated islands, the conductivity of the gate that is necessary for operation is lowered and eventually lost.

(52)

5.1.2 Oxide surface

The XPS analysis of the oxides in Device 4 (Fig. 4.24) shows that the uppermost 80Å of the oxide ﬁlm is made up of several types of Mg-oxides, most likely MgO and MgO2, and CoO. It also reveals high content of Si and C, indicating vertical

holes through the oxide, most probably caused by a restructuring of the oxides during annealing. These holes were also visible in the SEM-studies of the same oxide. This shows that a mixture of Mg and Co-oxides at the surface can indeed be obtained through the processing scheme used, at least for Co thicknesses up to 50Å. Furthermore, it is clear that there are three main agents in the sensor response: Mg-oxides, Co-oxides and porous Pt. Furthermore, the holes are most likely unevenly distributed over the surface, which would aﬀect CV-characteristics and therefore also the sensor responses.

A feature of the sputtered platina ﬁlms seen by the SEM is that the step edges are clearly visible in the oxide and Pt-ﬁlm. It is reasonable that this is a result of the oxides and the thereupon deposited porous Pt growing with a strong adhesion and almost epitaxially on the SiC substrate and thus taking the structure of the underlying SiC substrate, which has natural step edges.

5.1.3 Evaporation and sputtering of Pt

Under normal conditions, Pt films of thicknesses less than 200-250Å are normally porous, regardless of whether they are produced by evaporation or sputtering [15]. But for these sensor structures, the Pt films are of more dense than porous character (Fig. 4.1, 4.5, 4.3b) whereas the sputtered films are relatively porous (Fig. 4.7, 4.14). The reason why the evaporated films are not porous as expected is most likely due to a rather strong interaction between the mixed oxide and the Pt, which results in a rather dense and a stressed platinum film growing on top. This is also a plausible explanation for the bad adhesion of the contact pads on all devices produced by thermal evaporation, since the stressed Pt-film in known to have a weaker adhesion to SiO2 [17]. SEM-studies further revealed the sputtering

ﬁlms agglomerating less than evaporated ﬁlms of the same thickness, which is favourable for sensor response stability.

5.2 Sensor characteristics

All the devices produced showed a response in the order several to 130 mV to NO and NO2. Devices 1 and 2 showed response in the same direction for both

NO and NO2 (Fig. 4.2b, 4.2a, 4.4b, 4.4a) in contrast to Devices 4 and 5 (Fig.

4.16, 4.19, 4.8, 4.9). One plausible explanation to this is a diﬀerence in the oxide mixture, since these devices were produced on diﬀerent wafers, that might have more or less prominents steps. The doublet peak feature of the responses

(53)

of Devices 1 and 2 together with the diﬀerent direction of response for Devices 4 and 5, indicate that several competing mechanisms take place on the sensor surface. This is expected, since there are multiple pathways of reaction for NO and NO2 on Co-oxide [20]. Mg [20] and Pt [3] are also known to interact with

NOx. The responses seen from the sensor is a combination of responses from

all these reaction pathways. For devices 1 and 2, the NO2 measurements ended

in the sensor failing, indicating that the dominant mechanism of response for these devices are also causing agglomeration of the Pt due to interaction with the strongly oxidizing NO2 in a similar way as reported by Spetz et al. for H2 and

Pt [15] and by Eriksson et al. for Pd [7]. This is supported by the agglomerated Pt films showed by the SEM-studies. For Devices 4 and 5, agglomeration was not as predominant, indicating that interaction with the oxides were more prominent in the responses from these devices. That there is a difference in direction of response for NO between device 3 (Fig. 4.6) and Devices 4 and 5 (Fig. 4.16, 4.8) indicate that a large amount of Co in the surface of the oxide film causes a downwards response toward NO. Possible reasons to this is presented in section 5.2.1.

Cross-sensitivity to oxygen was tested for Devices 4 and 5 (Fig. 4.8, 4.16) and showed undetectable changes in the response of both devices for varying oxygen backgrounds, which can be assigned to the high degree of oxygen content at the surface, providing oxygen also at low atmospheric oxygen content. Selectivity for NOx towards CO, H2 and C3H6 was also tested for devices 4 and 5 (Fig. 4.10,

4.12, 4.17, 4.18), showing a very large diﬀerence in selectivity. Device 4 exhibits excellent selectivity to NOx, only responding slightly to CO and C3H6 and high

concentrations of H2 and never responding more than 30 mV, comparable to a

response to 25 ppm of NO. Device 5, however, exhibits a stronger response to H2,

CO and C3H6 and is thus not as selective. The diﬀerence in selectivity indicates

that the ratio of Co and Mg oxides at the surface are critical for tuning the properties of the sensor. The very sharp peak in the beginning of the response toward CO for Device 4 (Fig. 4.12) is an interesting property. One reason for this behaviour could be that CO at ﬁrst removes a lot of O2 from the surface,

forming CO2, but as the surface coverage of O2 decreases, there is room also for

CO to adsorb directly on the surface, resulting in a equlibrium of charge transfer between the oxide and CO being oxidized and CO being adsorbed, something that is supported by the thereafter very stable output during CO exposure and the fast recovery after exposure.

5.2.1 Possible reaction mechanisms

As mentioned before, the opposite direction of response for NO and NO2 in

De-vices 4 and 5 (Fig. 4.16, 4.19, 4.8, 4.9) as well as the doublet peak feature exhibited for Device 1 and 2 (Fig. 4.2b, 4.2a, 4.4b, 4.4a) indicate that there is

A NOx sensor for high-temperature applications based on SiC

Linköping University

Institute of Technology

IFM

A NO

x

sensor based on SiC for

high-temperature applications

by

Johan Midbjer

Supervisor: Ruth Pearce, IFM

Examiner: Anita Lloyd Spetz, IFM

Contents

Acknowledgements

Chapter 1

Introduction

1.1

Goals of the project

1.2

Applications

Chapter 2

The sensor

2.1

Basic principle of gas sensors

2.2

Sensor characteristics

Chapter 3

Manufacturing,

characterization and testing

3.1

Deposition techniques

3.2

Sensor packaging

3.3

The device

3.4

Characterisation methods

3.5

Measurement method

Chapter 4

Results

4.1

Device number 1 (50Å Co, 200Å Pt)

4.2

Device number 2 (12Å Co, 200Å Pt)

4.3

Device number 3 (100Å Co, 150Å Pt)

4.4

Device number 4 (50Å Co, 150Å Pt)

4.5

Device number 5 (75Å Co, 150Å Pt)

4.6

Device number 6 (50Å Co, 150Å Pt)

4.7

XPS-measurements

Chapter 5

Discussion

5.1

Material characterization

5.2

Sensor characteristics

_x