IGR Report on Grant GR R

IGR Report on Grant GR/R42511/01 Prof. D. E. Williams and Prof. I. P. Parkin, Department of Chemistry, University College London 1 Atmospheric Pressure Chemical Vapour Deposition as a Silicon Microfabrication-Compatible Route to Deposition of Gas Sensor Materials 1.1 Background/ Context This project demonstrated that a microfabrication compatible high throughput route to the gas sensor materials Cr2-xTixO3 and WO3-x can be achieved by atmospheric pressure chemical vapour deposition. We obtained films of controlled microstructure and porosity that are stable to thermal cycling. We discovered a new and potentially industrially important film growth phenomenon – the effect of an applied electric field on aerosol assisted chemical vapour deposition. In particular we demonstrated site directed deposition as well as control of morphology and particle orientation by using an alternating electric field on a sensor substrate. This has enabled us to produce gas sensors with what has been hitherto only an idealised theoretical microstructure. The new microstructures further confirm the theory of microstructure effects on sensor response. They show an increase in sensitivity to traces of nitrogen dioxide by more than 10 fold over commercial sensors. The work is in the process of being patented by UCL. 1.2 Aim The primary aim of the project was to prepare metal oxide gas sensor coatings by CVD. This involved direct deposition from available dualsource precursors in combination with more speculative single-source AA CVD routes. 1.3 Introduction 1.3.1 Present state of the art: The literature and commerce of low-cost gas sensors based on heated semiconducting oxides has been dominated for many years by tin dioxide. This material suffers from well-known problems of the strong dependence of response and baseline on the ambient humidity, and on response and baseline drift effects. In recent years, these problems have been overcome by the introduction of new sensor materials, as a result of work by DEW on chromium titanium oxide and tungsten trioxide. Titanium-substituted chromium oxide (CTO: Cr2-xTixO3, x < 0.4; typically x = 0.95) has proved to be a substance that offers very stable response and baseline over time, and very small effects of ambient humidity changes on response and baseline. It is useful for measurement of carbon monoxide, volatile organic compounds, H2S, SO2 and NH3 in the ppm range, and has been successfully commercialised for this purpose. Carbon monoxide sensors utilising CTO are the only sensors (as distinct from detectors) to have passed the new and testing UL2034 standard (Underwriters’ Laboratory, USA) as a component. The material is made by high-temperature, solidstate reaction of chromia and titania, and devices are made by screen-printing the powder (typical particle size 1 mm) onto alumina substrates having a printed gold electrode pattern and printed platinum heater track. Power consumption is typically 500 mW at the operating temperature (typically 400oC) Tungsten oxide, WO3, long known as a sensor material for NO2 and H2S in the 10 ppm range, has been shown to be excellent for measurement of ozone in the ppb range. It has been successfully commercialised for this purpose. Again, the sensors are prepared by screen-printing powder onto alumina substrates. Because ozone measurement requires a high air-flow rate over the sensor (because ozone is decomposed in the thermal boundary layer of the device and also on the walls of the sensor housing which are warmed by the device) the power consumption is typically 1 W. 1.3.2 Gas sensing mechanism and advantages of porous, film devices: The gas sensing mechanism of heated semiconducting oxides is a modulation of surface charge resulting in a modulation of conductivity. The sensors can either be made as a thin, dense layer, with thickness on the order of the space charge layer thickness, or as a highly porous, sintered mass of small crystallites. In practice, devices made by screen-printing thick (typically 100 mm) porous layers have proved to be more repeatable and less susceptible to surface poisoning (e.g. by siloxanes deriving from lubricating oils or polishes) than the dense, thin-film forms. Further, the thick-film structures allow a certain discrimination of gases by probing the concentration gradient that may be set up through the thick layer as a consequence of the reaction of the gas on the sensor surface. Different electrode gaps probe different depths of the sensor material. 1.3.3 Power reduction and capability improvement by integration with silicon microelectronics: A significant current world-wide research thrust is to integrate semiconducting oxide-based gas sensors with silicon microelectronics. There are two reasons. The first is that silicon microfabrication offers a route to very low power consumption structures, which have the further advantage of allowing fast temperature modulation, which then opens up other detection, zero-drift compensation, or discrimination schemes. The second is that these structures can be integrated into sensor arrays which can give improved accuracy and reliability by averaging over sensors or by ‘voting’ to remove signals due to failed devices (for example: devices which have suffered a fracture in the heater connections), or which can be used to improve discrimination of gases through the variations of signal caused by variations of the sensor temperature, or by variation of composition of the sensor material. The only work to date has, however been with SnO2-based devices, made for example by cracking SnCl4 vapour onto the devices using the device heater itself to ensure selective, local deposition, which has shown that to some degree, such integration can be achieved and the promised advantages delivered. The resultant thin layers are fairly dense. 2 The challenge faced by this project was to combine the performance advantages of porous sensor layers with the power advantages of silicon microfabrication and thin film devices. The aim was to demonstrate fabrication of porous sensing layers by atmospheric pressure chemical vapour deposition (APCVD). The need was to demonstrate that deposit morphology could be controlled and that mixed metal oxide sensors could indeed be made this way. We focused on CTO and WO3 based materials. A wide range of substrates were used for the investigation including glass, silica, silicon, silicon-microfabricated devices and alumina supported gas sensor tiles. The new sensors made were compared to conventional screen printed devices and correlations between microstructure and sensitivity sought. 2 • Project Objectives Preparation of Cr2-xTixO3 and WO3 films for gas sensor applications by APCVD and aerosol-assisted APCVD (AACVD). o Demonstration of substrate adherent layers of micrometer-scale thickness with controlled microstructure and (high) porosity, stable to thermal cycling Determination of growth parameters controlling the porosity and microstructure of the films. Determination of the gas sensing response of the films, correlation of film porosity and elemental composition with response. • • 3 Overview of project achievements • • • • • First demonstration of APCVD of solid-solution, mixed-metal oxide gas sensing layers (CTO : section 4.1). First demonstration of controlled morphology sensor films made by AACVD (WO3 : section 5.1). Discovery of DC and AC-electric-field induced changes in morphology and rate of film growth by AACVD (WO3 : section 5.1) Demonstration of theoretically ideal microstructures prepared by APCVD, and consequent fabrication of devices of very high gas sensitivity (WO3 : section 5.1.2). AACVD and APCVD deposition of sensor films on silicon microfabricated substrates 4.1 Chromium Titania Oxide Thin Films by APCVD APCVD reaction of CrO2Cl2, TiCl4 and ethylacetate produced a dense film that was adherent to the sensor substrate. The film was predominantly CTO but contained 5 atom % Cl as an impurity. This had a deleterious effect on sensor response. However, post annealing the films at 600 °C for 1 h in air (under a nominal flow of 5-10 % H2 in N2) led to removal of impurities (C, Cl) to below the detection limit of the spectroscopic techniques employed (£ 0.5 atm %), as described by equation 1. Cr2-xTixO3ClyCz (s) + H2(g) Æ Cr2-xTixO3(s) + zCH4(g) + yHCl(g) (1) SEM images (Figure 1) of the on-sensor film showed dense, spheroidal platelets of 1-4 microns in diameter sintered into a single film. A greater degree of deposition was observed between the gold electrodes, where film thickness was assessed. Upon reducing the film thickness to ca. 500 nm (deposition time of 30 s), significant changes in morphology were observed, with an increase in the range of both particle size (1-10 mm) and particle morphology. Annealing of the films caused no significant change in morphology. We needed to confirm that we had prepared the required solid solution rather than a simple physical mixture of the two constituent oxides. Cross-sectional SEM and EDXA showed homogeneous film thickness and stoichiometry as well as rapid film growth. For each sample, elemental ratios were consistent to within 0.5 atm % over many surface spots (1 mm focal width), with analysis by back-scattered electrons suggesting a single phase material. Surface composition of the films (determined by XPS analysis) showed chemical environments for CrIII, TiIV and O2- that match well with literature values for CTO [binding energies: Cr 2p3/2 576.7 eV (577.4 eV); Ti 2p3/2 459.3 eV (458.8 eV); O 1s 529.9 eV (531.0 eV)]. The CrIII 2p3/2 peak showed a smaller shoulder at ca. 579 eV (corresponding to surface CrVI species) than has been observed for CTO films prepared by other means. In addition, comparisons of XPS (surface, ca. 10 atomic layers) and EDXA (bulk, ca. 1-2 mm depth) analyses show surface segregation of titanium for the APCVD-prepared material which saturates at a lower bulk titanium concentration than that observed for material prepared by high temperature reactions of precursors. A comparison of titanium segregation of pre- and post-annealed films (600 °C; 72 h) showed, as expected if Ti was present in solid solution in Cr2O3 and that heat treatment increased the observed titanium surface segregation. Raman spectra were obtained for APCVD-Cr2-xTixO3 prepared on sensor substrates. The spectra for the CTO solid solution did not show any presence of TiO2 (both the anatase and rutile phases are strong Raman scatterers). Instead it bore close resemblance to that of Cr2O3 with a pronounced shift in peak positions to lower wavenumbers (ca. 55-80 cm-1). This is indicative of low quantities of substitutional doping (by titanium) into the chromia (corundum) structure. The combined analyses of Raman, back-scattered electrons and EDXA suggest the films to be a single phase of desired stoichiometry as opposed to an intimate mix of titania and chromia. In addition, there appears to be no support for the presence of CrTiO3 as a minor phase. Whilst compositional analysis techniques may show there to be no film contamination, dopant levels of contaminants are likely to have a profound effect on any gas sensing applications. It is also critical to ensure that the films are indeed of single phase rather than an intimate mix of chromia and titania beyond the detection limit of the SEM-EDXA employed. Consequently, determination of the activation energy of conductance of the films was carried out: the value for Eac obtained was 0.71 eV, which is in excellent agreement with literature for ceramic preparations of CTO (0.70 3 eV). This suggests that the film is a true solid solution rather than an intimate mix of particles of chromia and titania, since Eac (Cr2O3) = 0.40 eV. 4.1.1 Gas sensing properties of APCVD Cr1.95Ti0.05O3 films. Ethanol was used as the target reducing gas for these experiments. Figure 2 shows a change in film resistance to increasing reducing gas concentrations in both dry and humid air at 350 °C that is typical for 1.5 mm CVD prepared CTO films. General features of the curves are indicative of CTO, such as the slow return to baseline resistance, small effects of humidity and a relatively stable baseline. However, the gas response (R/R0) observed here is lower than would be expected when compared to conventional ceramic screen-printed films (ca. 1.2; cf 3 4). It was found that the response to ethanol increased with decreasing film thickness, suggesting that the microstructure is more characteristic of a continuous film, rather than a porous polycrystalline body formed by screen-printing. Three film thicknesses were deposited by APCVD (1500, 1000 and 500 nm). Each film was exposed to 80 ppm EtOH in dry air at 10 different temperatures, sequentially increasing then decreasing from 200 - 600 °C at intervals of 75 °C. There was minimal variation in the gas response of each film with temperature. 9.00E+04 80 ppm 8.00E+04 60 ppm 80 ppm 40 ppm Resistance (ohms) 7.00E+04 60 ppm 40 ppm 6.00E+04 20 ppm 20 ppm 5.00E+04 4.00E+04 0 % humidity 3.00E+04 0 5000 10000 15000 50 % humidity 20000 25000 30000 Time (s) Figure 1 (left) SEM of Cr1.98 Ti0.02 O3 deposited on sensor substrate. Film thickness is ca. 3 m m ( ≡ 180 s deposition time). The more particulate morphology (left of picture) was deposited on alumina. The more aggregated morphology (right of picture) was deposited onto the gold electrode. Figure 2 (right) Gas response to ethanol in air of APCVD- CTO sensor film (500nm) The 500 nm film showed comparable gas response to conventional screen-printed CTO at 400 °C Derivation of the relative contributions of bulk, surface and grain boundary sensitivity from analysis of gas response is accessible through application of a simple equivalent circuit model to a range of target gas concentrations, where a defined gas sensitive surface layer is in parallel with a gas insensitive region, each in turn lying in series with a gas sensitive 'grain boundary' approximation (Figure 3). Figure 4 shows fit of the model to the observed variation of resistance with gas concentration. The relative contributions of each element changed with layer thickness: the thicker layers behaved as if they were more dense. 6.00E+07 5.00E+07 Curve Fit Resistance (W ) 4.00E+07 Rs, Rb, 3.00E+07 2.00E+07 Rgb 1.00E+07 0.00E+00 1 10 100 1000 10000 100000 ln (ethanol concentration) Figure 3 (left) Equivalent circuit model for sensor film. Rb is considered independent of gas concentration. Rs and Rgb vary linearly with gas concentration. Figure 4 (right) Fit of equivalent circuit model to gas response of a 500nm APCVD-CTO film. 5.1 Tungsten Oxide studies 5.1.1 Variables influencing growth morphology in AACVD Figures 5 and 6 illustrate the different morphologies obtained by variation of the factors detailed below. Films of tungsten oxide were laid down on sensor substrates by atmospheric pressure vapour deposition. It was found that the microstructure could be manipulated very effectively by using an aerosol as the delivery vehicle for the single source tungsten precursor. A wide variety of parameters were found to affect the shape, size, coverage, orientation and morphology of the tungsten oxide sensor materials. Temperature – higher substrate temperature encouraged faster film growth – an onset temperature was found of 300°C for tungsten oxide deposition in most systems. 4 Precursor – A wide variety of tungsten alkoxide precursors were tried, subtle modifications of the microstructure were observed with change in class of precursor. Most experiments were done on W(OPh)6 as a single source WO3 precursor although acetates and halides were also investigated. Solvent – The dielectric constant and boiling point of the host aerosol solvent were key determinants in film morphology – films were grown from a wide variety of solvents. Microstructures from fibres to rods, to spherical particles to stalks to shell like to fully dense films were achievable simply from change the solvent. Toluene was found to be the best solvent for the formation of fiborous microstructures. Gas Flow- the gas flow velocity did have an affect on the amount of deposition. The orientation of the sensor substrate relative to precursor flow had a dramatic affect on microstructure. Introduction of turbulent flow either but orienting the sensor perpendicular to the gas stream or by using a step baffle prior to the sensor substrate gave faster film growth. Platinum seeding- the use of platinum seeding helped to localise deposition to specific sites- clustered around the platinum seed particle. Generally denser films were obtained using a Pt seed. The platinum seed also affected the growth dynamics and anenomi like structures were seen emanating from the seed. The use of a Pt seed often allowed deposition at slightly lower substrate temperature. Substrate type- the nature of the substrate- gold, silica, alumina or silicon did affect to some degree the temperature onset of deposition (ca 25-50°C). We were able to obtain the same wide range of morphologies on all substrates and we did not notice any substrate dependent morphology. Method of heating- the sensor substrates were heated in two ways by using the on chip platinum resistive heater track or by using the sensor on a heated glass substrate. Some minor differences in growth morphology were noted with deposition method. Application of an electric field during deposition - this somewhat fortuitous discovery enabled us to have the greatest control in morphology and enabled idealised microstructures to be laid down. We discovered this phenomena because we wanted to see if we could measure the change in resistance of a growing film as a consequence of measuring the deposition rate. What we found was that application of either ac or dc fields increased the rate of film growth on the sensor substrate. More importantly we found that the use of an ac field only promoted formation of fibrous growth that was aligned between the electrode gap. We studied a range of conditions and field strengths and managed to optimise the microstructure such that the tungsten oxide sensors made showed enhanced sensitivity to nitrogen dioxide and ozone compared to commercial materials. The sensors also maintained the microstructure and sensitivity to multiple thermal and gas cycling loops. Figure 5 Sensor morphologies obtained using different deposition conditions in APCVD of WO3 5 Figure 6 5.1.2 Effects of Application of an electric field during AACVD deposition Figure 6 illustrates the different growth morphologies that could be obtained by changing the field and frequency. We were able to obtain a structure comprising a thick and highly porous layer of fine interconnected needles spanning the electrode gap, that we had anticipated as minimising the ‘bulk’ and maximising the ‘surface’ contributions to the equivalent circuit shown in figure 3, and hence optimal for gas sensing. The resulting tungsten oxide sensors showed sensitivity to nitrogen dioxide and ozone enhanced by factors of more than 10 times compared to commercial preparations: figure 7 shows a representative result. The sensors also maintained the microstructure and sensitivity after many thermal and gas cycling loops. The behaviour was explicable first because the aerosol droplets were charged and hence the force upon them, near to the surface, was determined by the electric field; second because there was a frequency-dependent dipole induced in the droplets and in the solid material formed as the droplets evaporated and the precursor decomposed in the boundary layer near the substrate; and third because the local field distribution above the growing deposit was dependent upon the frequency-dependent conductivity of the deposit. The growth in a DC field was consistent with the aerosol droplets carrying a positive charge and hence being preferentially deposited upon the negative electrode. In an AC field, the droplet trajectory and hence the position of deposition would depend on the time of flight of an aerosol droplet near the surface relative to the time period of the signal: at higher frequency, growth was confined towards the electrode gaps. As WO3-x formed, the dipole induced in particles near the surface, and the local field near the surface, would depend upon the electrical conductivity, which is frequency dependent in a fashion dependent on the size scale of the conducting object. The result should be stabilisation of fibre growth and orientation of the resultant fibres along the electric field lines. It was at low ac field that we observed growth away from the electrode surface in alignment with the electric field, since at higher field the effect was obscured due to the amount of material deposited. We deduced that, as the fibres grew, a dipole could be induced. A dipole could not be induced in the fibres under d.c. because any electric field in the fibre would be dissipated due to the high conductivity of WO3-x; however, at sufficiently high frequency, the skin effect would cause a sufficiently high resistance of a fibre to allow a local electric dipole to be sustained. Thus, in a 100 DC field, the fibres were randomly oriented whereas at sufficiently high frequency, directed fibre growth was obtained. Gas Response (R/Ro) 10 Screen Printed 5 V d.c. 10 V, 50 Hz Figure 7. Gas response for a screen-printed WO3 sensor and one prepared in an applied electric field. 800 1 0 100 200 300 400 500 600 700 Concentration of NO 2 (ppb) 6 6. Project Plan Review The project plan did involve some minor changes. Prof Williams left UCL during the grant period to move to Unipath. Prof. Parkin carried on the supervision of the project. Prof. Williams regularly attended project meetings on a monthly basis, Prof Parkin dealt with weakly progress meetings. 7. Summary All of the objectives set out in the proposal have been achieved. We have demonstrated that APCVD is an ideal method for laying down mixed-metal oxide sensor materials. We have shown in particular that AACVD in conjunction with control of an electric field applied during the deposition gives excellent control over the microstructure of the resultant porous layers, and have demonstrated high-sensitivity gas sensors made in this way. We have shown that the route is indeed microfabrication-compatible. 8. Research Impact and Benefits to Society To date the research has yielded five publications in high quality international journals (J. Mater. Chem., Sens and Acctuators) and two refereed conference papers that were published in book [1-7]. A further four publications are planned. To some extent work was delayed form being published – especially on the electric field studies until a patient had been filed. The electric field work on WO3 will appear in a special feature addition of J. Mater .Chem. - issue 1 2005. Work has been presented as a talk or poster presentation at eight international conferences including Euro CVD, MC-5, ACS and ICMCTF and featured in a number of national and international Universities (15), including Indiana University, Ohio state, IUPUI, Purdue, Liverpool, Manchester, Herriot Watt, Cambridge, Oxford, Imperial College. The tungsten oxide sensor work has been filed for a patient through UCL. The new sensors have extraordinary sensitivity and incorporation of these materials is part of a start-up business planned through UCL based on gas sensing work – Novosense. The coated glass has also been supplied to various ceramic artists for incorporation into new pieces of art. The work has also been incorporated into the lecture demonstration that Prof. Parkin gives to school-children (6 presentations during the course of the project). The new films have also been grown on MOSFET (supplied to Prof. W. Wosik, Texas A+M), solid-state gas sensor devices and SIMOS substrates (Prof. Gardner). These collaborators have given conference presentations on their results obtained using the films that we grew on their devices. 9. Explanation of Expenditure The grant expenditure did not deviate from that stated in the original proposal within the virement allowed. City Technology provided more analytical support for the project than initially envisaged and also directly funded a CASE studentship in the sensor laboratory. We were also awarded some beam time from CCLRC Daresbury for the X-ray characterisation of the thin films. 10. Further Research and Dissemination Activities Dr G. Shaw has taken up employment in the NHS as an analytical scientific officer. Dr C. Blackman worked on this project for 3 months to work out some specific AACVD issues. He is now employed on a further PDRA contract at UCL with Prof. Parkin. Links to the final grant report and papers can be found from Prof. Parkin web home page – http://www.che.ucl.ac.uk/people/ipparkin. 11. References [1] D. E. Williams, G. Chabanis and I. P. Parkin, J. Meas. Sci. Tech., 2003, 14, 76 [2] G. Shaw, K. F. Pratt, I. P. Parkin and D. E. Williams, J. Mater. Chem., 2005, 14, (Jan 1 special issue) [3] G. Shaw, I. P. Parkin and D. E. Williams, J. Mater. Chem., 2003, 13, 2957 [4] G. Shaw, I. P. Parkin and D. E. Williams, CVD XVI 2004 p 777. [5] G. Shaw, I. P. Parkin and D. E. Williams, Advanced Mater (CVD) (in press). [6] C. Blackman and I. P. Parkin, Chem. Mater, submitted. [7] G. Shaw, K. F. Pratt, I. P. Parkin and D. E. Williams, Sensors and Actuators, 2004, (in press)