J. Chem. Soc., Faraday Trans. I , 1989, 85(8), 1907-1919 Tin Oxide Surfaces Part 19.-Electron Microscopy, X-Ray Diffraction, Auger Electron and Electrical Conductance Studies of Tin(rv) Oxide Gel Philip G. Harrison* Department of Chemistry, University of Nottingham, University Park, Nottingham NG7 2RD Martin J. Willett British Coal, HQ Technical Department, Ashby Road, Stafordshire DE15 OQD Physical and electrical properties of both unsintered and sintered tin(rv) oxide gel of high specific surface area and porosity are reported. Particle sizes range from 3 to ca. 5000 nm, the larger bodies apparently consisting of pressed agglomerations of smaller particles. The 3 : 1 0: Sn atomic ratio determined at the surface of the discs was consistent with a fully hydrox- ylated oxide surface.A slight oxygen deficiency from the ideal SnO, com- position is observed at all depths. Sintering for extended periods at 1273 K increased the minimum particle size from 3 to ca. 40 nm without causing any appreciable change in morphology on the micron scale. The electrical conductance of the unsintered oxide in air as a function of temperature was found to be extremely complex and exhibits hysteresis dependent upon the history of the sample. Unsintered discs initially displayed a sigmoid relationship between the conductance G and tem- perature T, with G typically reaching a maximum at 420 & 20 K, followed by a minimum at 520 +_ 50 K, thereafter increasing monotonically. Subsequent G-T cycles produced a wide range of behaviour. Although in general, the G-T characteristic exhibited a more linear form, the rate of change with temperature varied greatly.The initial sigmoid curve was largely unaffected by strict control of the water content of the test atmosphere, and hysteresis effects generally persisted, despite the adoption of more rigorous experi- mental methods. The air conductance behaviour of presintered samples was typically less complex, with an increasing tendency towards a monotonically rising G-T characteristic with extended duration of the sintering treatment. Significant differences were observed between the electrical behaviour of the unsintered tin(rv) oxide gel samples and that generally exhibited by ' commercial ' semiconductor gas sensors. However, these differences could be minimised by presintering of the oxide gel.The development of cheap solid-state electronic devices for the detection of toxic or flammable gases is a subject of prime importance in many areas. The electrical properties of tin(rv) oxide render it a n excellent material for the detection of very low levels of carbon monoxide, and several devices based on sintered pellets of SnO, are commercially available. '* * To date, investigations of operation of such sensors have generally been restricted t o studies of the electrical behaviour of the o ~ i d e . ~ - ~ Little attention has been paid t o the subtle relationship which must exist between the surface adsorption phenomena and the electrical changes thereby induced, although this intrinsic link is absolutely critical t o the successful operation of the sensor.Previous studies have demonstrated that a model in which the sensor conductance is effectively controlled by the population of negatively charged oxygen adsorbates is consistent with the observed 19071908 Tin Oxide Sutfaces beha~iour.~ The chemical reactions involved in the conductance mechanism, although unsubstantiated, have been assumed to involve CO adsorption, desorption of CO, (which produces an increase in conductance), and replenishment of the resulting surface oxygen vacancy by adsorption of molecular oxygen.’ As part of a programme designed to investigate more fully the mechanism of operation of tin(rv) oxide-based gas sensors, and particularly to elucidate the relationship between the nature of the adsorbate species and bulk oxide electrical conductance, we have already6 carried out infrared studies which have characterised the species formed on the surface of tin(rv) oxide from very low levels of carbon monoxide in air.However, the oxide employed in that study, tin(rv) oxide gel of high surface area, is substantially different in physical nature from that usually employed solely for electrical studies3-’ or used in commercial gas-sensing devices.’q2 We therefore have found it necessary to characterise the physical and electrical properties of the type of oxide used in our studies, and the results are described in this paper. Experimental Preparation of Tin(rv) Oxide Gel Tin(1v) oxide gel of high specific surface area was produced by the hydrolysis of redistilled tin(1v) chloride as described previ~usly.~ Great care was taken to ensure the purity of all materials and apparatus used in the preparation, and the quality of the final product was checked by X-ray fluorescence analysis using a Philips XRS PWI400 spectrometer.No significant impurities were detected. Small quantities (ca. 0.5 g) were ground for 6 min in a ‘Grindex’ agate ball mill grinder (Research and Industrial Instrument Co.) immediately prior to pressing into self-supporting discs (60 mg, 25 mm diameter) using a pressure of 34.5 MN m-,. Presintering of discs was performed under ambient atmospheres at a temperature of I273 & 20 K in a silica boat. After sintering, the samples were virtually indistinguishable in colour from untreated discs, there being a creamy hue in some cases, but were quite brittle. The surface of the oxide appeared shiny in comparison with the relatively matt as-pressed samples.There was also evidence of some shrinkage (ca. 10 YO) in the physical dimensions of the samples as a result of the treatment. Physical Characterisation Electron-microscopic studies were carried out using a Pye Stereoscan 600 scanning electron microscope, which has a quoted resolution of 10 nm. However, the insulating nature of the oxide prevented meaningful observations below ca. 50 nm. Auger electron spectroscopy was performed by Loughborough Consultants Ltd. Depth profiling was performed by using argon-ion bombardment. Owing to the particulate nature of the discs, the absolute depth values are only approximate, although intercomparison between the samples is valid.The quoted uncertainties in the 0:Sn ratios represent & 5 %, which was the ‘worst-case’ estimate of the error. Electrical Resistance Measurements Two methods of sample resistance measurement were employed depending upon whether the measurements were obtained under a flowing ambient atmosphere or under a rigorously controlled atmosphere. For both methods, the pressed disc oxide samples (ca. 2-3 mm across) were mounted on self-heated Platfilms (Rosemount Ltd.), which were spot-welded on to nickel-plated TO5 headers (Wesley Coe Ltd). Temperature calibrations of the Platfilm as a function of heater current were accomplished by attaching a Pt/Pt-lO% Rh thermocouple fabricated from 0.05 mm diameter wire toP. G . Harrison and M .J. Willet 1909 Table 1. Atomic 0 : S n ratios from AES analysis calcination pretreatment/h at 1273 K _________________~___ depth/nm 0 72 168 256 0 2.86k0.14 2.86f0.14 3.03k0.15 3.06k0.15 4 1.95+0.10 2.02k0.10 2.02+0.I0 2.05+0.10 13 1.92k0.10 1.96k0.10 1.98k0.10 2.01 kO.10 25 1.88k0.09 1.98k0.10 1.97+_0.10 2.00+_0.10 100 1.88 rfI0.09 1.89 k0.09 1.95 rfIO.10 1.98 & 0.10 200 1.86+0.09 1.84kO.09 1.92k0.10 1.97kO.10 400 1.87k0.09 1.98k0.09 1.90k0.10 1.95k0.10 the centre of the upper tile surface using gold paste ('Hanovia' A1644 liquid gold, Engelhard Ltd). Gold paste was also employed in the fabrication of all electrical contacts to the pressed discs. Contact resistances were at least two orders of magnitude below the minimum resistance values expected for the oxides, and were therefore neglected.Electrical resistance measurements under ambient atmospheres (flow rate ca. 50 cm3 min-') were made, employing a simple series load resistor circuit. The voltage drop across the load was measured using a Keighley Model 616 electrometer. In order to minimise problems due to polarisation of the samples, particularly those exhibiting high resistances, the supply voltage was connected to the oxide sample under study for only a few seconds to allow the measurements to be completed, and the polarity of the supply voltage was reversed at random. No significant errors due to polarisation of the samples were noted. This method also minimised any self-heating effects due to the measurement current. Measurements under rigorously controlled atmospheres were achieved mounting samples in a vessel connected to a conventional vacuum frame and using dry bottled 0,(21 %)-N,(79 O/O) mixtures (purified by P20, and KOH filters prior to entry to the test apparatus).The sample temperature in these tests was controlled by an automatic heater circuit which produced a current cycle with a triangular profile and a period of 48 h, providing a mean heating/cooling rate of ca. 27 K h-'. The oxide sample resistance was obtained using a circuit based on a logarithmic amplifier (Intesil 8048). By comparing the internal current flowing through the unknown resistance with an externally generated reference current, the logarithmic amplifier produced a signal directly proportional to the resistance of the oxide.The system provided an output of 1 V per decade of resistance, and was capable of handling over six decades of current input. A 1 kR resistor in series with the oxide sample limited the current flowing through it to < 0.5 mA, ensuring minimal self-heating. Calibration curves were obtained by using a range of known resistors in place of the oxide. Errors caused by variations in the reference current and the stabilised voltage supply were insignificant compared with the measured range of the device, providing automatic monitoring of properties over a wide range. Using this arrangement, three samples could be examined concurrently under reproducible atmospheric and thermal conditions. Results Physical Characterisation Electron Microscope Studies The unsintered ground powder contained a very wide range of particle sizes, between < 1 pm and cu.200 pm. The large particles appeared to be pressed agglomerates of 64 F A R I1910 Tin Oxide Surfaces smaller particles, and were covered in networks of fine cracks. Comparison with similar photographs illustrating material prepared 7 years earlier indicated no significant differences in the appearance of the powder on the micron scale due to aging. The surface of an unsintered 60 mg pressed disc consisted of relatively smooth areas of irregular shape ca. 1-2pm across, separated by more broken regions where the particle size appeared to be @ 1 pm. The overall impression was of a relatively flat but discontinuous surface above a porous structure. Viewed in profile, however, the surface is quite uniform, although the porous body of the disc comprised a range of particle sizes.The larger (ca. 30 pm) particles appeared to have fractured by cleaving, indicating a brittle nature, but there was no evidence of these protruding up through the surface, since the thickness of any given sample was found to be quite uniform (typically 43+ 1 pm for an unsintered 60 mg disc). There was, however, a large variation (ca & 5 pm) between the thicknesses of nominally identical discs fabricated from the same powder, demonstrating the irreproducibilities inherent in the pressing process. Examination of discs presintered for various periods indicated that there was little change in the appearance of the oxide or the cross-section due to the extended period at high temperatures.There was not an unambiguous correlation between the oxide disc thickness and the duration of the presintering treatment, although, in general, sintered discs tended to be thinner than unsintered ones. Discs produced from 120 mg rather than 60 mg of oxide powder did not have double the thickness after an identical pressing treatment. The increase was typically only ca. 50 YO, presumably indicating an increased packing density or reduced porosity within the disc body. In all other respects, the appearance of the two types was found to be quite similar. X- Ray Difr ac t ion Studies' Minimum crystallite sizes were estimated from measurements of the full width at half- maximum height parameter of the pressed discs using the [I 101 peak at 28 = 26.7'.The minimum crystallite size in the unsintered material is ca. 3 nm, but increases to ca. 36 nm following sintering for 72 h at 1273 K in air. Sintering for 168 h at 1273 K resulted in a smaller increase to ca. 45 nm. These values are similar to the data of Kittaka,g whilst Fuller" quotes crystallite sizes of 4-5 nm for tin(1v) oxide gel prepared by an analogous route. Auger Electron Spectroscopy The AES data confirmed the purity of the as-pressed oxide, whilst the sintering process resulted in some surface contamination (1-2 atom YO of silicon and potassium was typically detected perhaps by contact with the silica boat used to support the samples in the furnace). However, there was no evidence that these impurities had penetrated into the porous body of the disc.Atomic 0 : Sn ratios as a function of depth and the length of presintering are presented in table 1. The depth profiles of the 0: Sn ratio indicated that the value at any given depth was only weakly affected by increased length of sintering. There was a slight tendency for the 0: Sn ratio to rise with the period of presintering at 1273 K, although values covering pretreatments of between 0 and 256 h all fall within the same range of experimental uncertainty. In each case, the results at a depth of 400 nm indicated a slight oxygen deficiency as compared with the idealised SnO, composition, but there was no evidence for high concentrations of other oxides, e.g. SnO. In contrast to these relatively minor effects, the change in the 0: Sn ratio with depth in a given sample was highly significant. The ratio typically fell from ca.3 at the surface to ca. 2 at a depth of ca. 4 nm, with only very small changes occurring on further etching to 400 nm. For the sintered samples, some of the relative increase in the oxygen signalP. G. Harrison and M. J . Willet 191 1 0.0 h 2 u, -5.( on - - 10.1 . - . . * 9 0 .0 I I oxide temperature/K Fig. 1. oxide temperature /K Fig. 2. Fig. 1. Ambient air conductance-temperature characteristics of (a) a pressed disc of tin(1v) oxide and (b) a SMRE-type" sensor (temperature varied randomly). Fig. 2. Effect of heating rate upon ambient air conductance-temperature characteristics of tin(1v) oxide pressed discs (steadily rising temperature). ( a ) 24 h per temperature; (b) 1 h per temperature.at the surface could be attributed to the presence of oxygen-containing impurity compounds (related to the silicon and potassium signals already mentioned), but the comparable rise in the 0: Sn ratio of the unsintered (and apparently uncontaminated) sample indicated a second effect. Given the relatively constant value of the 0:Sn ratio at all other depths, oxygen-tin segregation at the unetched surface was not thought to be likely. The effect has therefore been attributed to the presence of adsorbed H,O- or 0,-derived groups within the analysis region, which are almost completely desorbed due to thermal effects during the first etching period. It is interesting to note that a 3 : 1 0: Sn atomic ratio is the value which would be expected from a fully hydroxylated tin(rv) oxide surface.* Electrical Characterisation Results of Ambient Atmosphere Studies Preliminary studies of the electrical conductance of SnO, pressed discs were performed by leaving the sample for approximately 1 week at each temperature and making electrical measurements at regular intervals during this period. On the assumption that 64-21912 Tin Oxide Surfaces I . n . . ... ;. ... . ' * a .. . * .,: - : . .. 2 u - -5.0- M 0 -10.0 I I 250 650 65 0 1050 I I oxide temperature/K oxide temperature /K Fig. 3. Comparison of ambient air conductance-temperature characteristics of various Sn0,- based sensors (steadily rising temperature) (all axes are identical). (a) SMRE-type; (b) Taguchi gas sensor type 71 1 ; (c) Taguchi gas sensor type 812; ( d ) Taguchi gas sensor type 813.the disc properties would have stabilised within this time, the sample temperature was varied at random within the range 290-890 K to avoid the introduction of effects due to the extended exposure to progressively greater temperatures. As the examples of mean sample conductance in fig. 1 show, two main conclusions can be drawn : no clear pattern in the conductance-temperature distribution emerges for the pressed-disc samples, and the behaviour exhibited by an SnO, SMRE-type sensor'' tested under the same conditions was significantly different from that of the pressed disc type. Fig. 2 illustrates the mean conductances of two sintered pressed-disc samples heated from room temperature to ca. 950 K at different rates: (i) taking 10 days to traverse the range, i.e.ca. 24 h at each temperature setting, and (ii) taking 3 days to traverse th'e range, i.e. taking ca. 1 h at each setting. In both cases, the sample was left overnight at elevated temperatures (ca. 16 h). As the trace for protocol (ii) shows, this resulted in significant drifting of the measured conductance value, particularly at intermediate temperatures. However, both methods illustrated a clear pattern in the conductance-temperature distribution which was not observed under random thermal cycling. A conductance maximum at ca. 420 K was followed by a minimum (slightly below the room temperature value) around 550 K. Thereafter, the conductance was found to rise monotonically up to the highest temperature studied, typically reaching a value between these extremes.The overall shape of the distribution could therefore be described as sigmoid, and although significant variations in the absolute conductance levels were exhibited by nominally identical samples, the same general curve was observed without exception. The two different thermal protocols were subsequently employed in various types of experiment : (ii) was particularly suitable for studying the effect of other parametersP. G. Harrison and M. J. Willet O.O h r2 9 - 5 . 0 - eo - 1913 (b) .. . . ..* . = . * 0 . * 0 0 0 0 0 0 0 0 0 0 . 0 0 O -7.0 - I o - * o o , , -10.0 250 650 1050 oxide temperature/# Fig. 4. -2.01 -12.01 I I8 oxide temperature/K oxide temperature/K Fig. 5. Fig. 4. Extreme examples of ambient air conductance hysteresis of (nominally identical) unsintered 60 mg pressed disc samples., rising temperature ; 0, falling temperature. Fig. 5. Typical examples of ambient air conductance hysteresis of SnO, pressed discs. (-) Rising temperature ; ( - . - a -) falling temperature) (all axes are identical). Presintering at 1273 K for (a) 0, (b) 4, ( c ) 14, ( d ) 24, (e) 48 and (f) 72 h. upon the conductance-temperature relationship, whilst (i) was used in examining the changes caused by exposure to gas mixtures. Attempts were also made to traverse the range much more rapidly, completing a full 290-+950-+290 K loop in a period of ca. 8 h and then leaving the sample at room temperature overnight. This means a thermal cycling rate of ca. 200 K h-l produced the same general behaviour as illustrated in fig.2, but the period spent at each temperature could not be reduced below 1 h without causing unacceptable inaccuracies in the measured conductance values, due to relatively slow equilibration of the samples. A brief comparison of the conductance-temperature distributions of various sensor types (including commercial Taguchi Gas Sensors) under protocol (ii) confirmed that widely different behaviour was exhibited by the various samples. In some cases, there was also significant drifting during the overnight (ca. 16 h) period. A detailed analysis of these variations (see fig. 3) was not attempted, since investigation of the pressed-disc samples was of greater significance to the present study, but reference will be made to the results later,Tin Oxide Surfaces -2.0' h Ez 2 -7.0 M - ' i * i ' : ..I I I ! : . ! I , : * . . 0.0 h Ez $ -5.0 M - - 1o.c -10.0 0 40 80 250 650 1050 oxide temperature/K - 12.0 1273 K presintering period/h Fig. 6. Fig. 7. Fig. 6. Effect of sintering period upon ambient air conductance of 60 mg pressed discs measured at (a) 293 and (6) 935 K. Fig. 7. Effect of repeated cycling under dry 0,-N, upon conductance hysteresis curves for a 60 mg pressed disc : unsintered sample (-) Rising temperature ; (- .-. -) falling temperature. (a) 1st cycle; (b) 4th cycle. When the same protocol was used for longer experiments in which the sample temperature was raised from room temperature to ca. 950 K and then reduced, significant hysteresis in the conductance distributions was observed, as the examples shown in fig.4 demonstrate. During tests on more than 30 samples, the behaviour in the falling-temperature arm of the loop was found to be very unpredictable, as the two extremes in this figure show. No change in the test conditions which would account for these fluctuations could be identified. The similarities in the rising-temperature arms of the loop for nominally identical samples implied that the hysteresis was attributable to the effect of high temperatures. Fig. 5 shows typical examples of the hysteresis loops obtained in experiments on samples presintered in air at 1273 K between 0 and 72 h. For clarity in this and subsequent figures, the loops are represented by line plots linking the points (ca. 30) in each arm. Broken horizontal portions of the curves at low temperatures indicate regions where the sample conductance was below the working range of the electrical measurement system.As already emphasised, there was great variation in the behaviour within a batch of nominally identical sensors, particularly those subjected to little or noP. G. Harrison and M . J. Willet 0.0 1915 - - 1 o.o/ h rz - M -10.0 :-5.0-{-N -.-.--./ I - - - I I - 10.01 I J 0.0 sintering. The examples shown have been selected from the results of more than 100 tests, in order to convey a general impression of the trends observed. With increased sintering time, there was a tendency for the absolute conductance at room temperature to fall, whilst the pronounced sigmoid shape of the rising-temperature curve was gradually flattened.However, the conductance at higher temperature was affected only very slightly by the presintering, The degree of hysteresis was greatly reduced in samples which had been presintered for more than ca. 24 h, and for the 72 h examples the uncertainty in the measurements on the increasing temperature arm was of the same order as the differences between the two halves of the loop (i.e. no significant hysteresis was observed). Fig. 6 illustrates the effect of increased sintering periods upon sample conductance at 293 K and ca. 935 K (data for increasing temperatures). Even allowing for the wide scatter in the data (note the logarithmic ordinate scale in this figure), there is a clear difference between the two sets of data, confirming the previous comments.In an attempt to reduce the scatter observed in the results (from unsintered samples in particular), tests were performed on discs prepared using 120 mg of tin(rv) oxide gel, i.e. of double weight. It was thought that if microcracking of the discs were responsible1916 0.0 h 2 $ -5.0 M - - - --- 0.0 Tin Oxide Surfaces (4 r -10.01 I I -10.01 I I 250 650 1050 oxide temperature/K Fig. 10. Effect of repeated cycling under dry 0,-N, upon conductance hysteresis curves for a 60 mg pressed disc: 72 h presintering at 1273 K. (-) Rising temperature; (.-.-) falling temperature; (--) limit of measurements. (a) 1st cycle, (b) 4th cycle. for the unpredictable behaviour, increasing the sample thickness might improve the reproducibility. However, no noticeable improvement was found, and the effects of sintering upon the behaviour of 120 mg discs were similar to those already illustrated in fig. 5 and 6 for 60 mg samples.Results of Controlled Atmosphere Studies Fig. 7-1 0 inclusive show the first and fourth complete rising/falling temperature traverses of sensors constructed using disc material subjected to 0, 24, 48 and 72 h of presintering at 1273 K in air. Again, data have been selected from tests on many sensors to convey a general impression of the trends observed. The aim of cycling through the temperature range more than once was to provide data on the degree of hysteresis in the conductance behaviour during repeated testing. Unless stated, all of the cycles were performed under the same sealed atmosphere. The data for unsintered samples were again found to exhibit significant scatter, although the general impression was of slightly increased reproducibility over the ambient atmosphere tests.There were fewer cases in which the falling temperature arm of the first cycle resembled the conductance-temperature behaviour usually associated with a presintered sample (i.e. a very large degree of hysteresis). This observation may be partly attributed to the reduced time spent at temperatures towards the top of the range under automatic, as opposed to manual, heater control. The results showed thatP. G. Harrison and M . J. Willet 0.0 h rz 0, -5.0 w - 1917 - - -10.0 I -10.0 250 650 1050 oxide temperature/K Fig. 11. Effect of evacuation-re-dosing on conductance-temperature characteristic of a 60 mg pressed disc [dry 0,(21 %)-N,(79 %) atmosphere switched between first and second traverses].(a) 1st traverse, low -+ high temperature; (b) (-) 2nd traverse, low + high temperature; (-.-.) 3rd traverse, high -+ low temperature. hysteresis in the conductance loops persisted throughout multiple thermal cycling. The falling temperature arms exhibited little or no evidence of a peak at around 420 K and a trough near 650 K. This behaviour contrasted with the strongly sigmoid curve observed for the unsintered material as the temperature was raised, as illustrated in For presintered samples tested under controlled atmospheres, the room-temperature conductance fell sharply with increased sintering time, whilst towards the upper end of the measuring temperature range there was relatively little variation.There was no significant effect on the degree of conductance hysteresis due to repeated thermal cycling. Increasing the degree of presintering generalIy reduced the extent of the hysteresis, as was found for tests under ambient atmospheres. To further investigate the sigmoid air conductance curve, a test was performed in which an unsintered sample was run up through the temperature range to ca. 900 K and then subjected to a brief evacuation (ca. 0.2 Pa) whilst cooling to room temperature. The sample cell was then re-dosed with dry 0,(21 %)-N,(79 O h ) and another full thermal cycle performed. The results of this experiment are shown in fig. 1 1, and indicate that, whilst there are differences between all three traverses of the temperature range, no significant effect may be attributed to the change in gas atmosphere.Again, the large degree of hysteresis in the electrical properties prevents unambiguous interpretation of the data. fig. 7.1918 Tin Oxide Surfaces Discussion Physical Nature of the SnO, Discs The as-pressed discs consisted of pure tin@) oxide gel with a high specific surface area and a porosity of ca. 60% (assuming a thickness of 4.5 x m and a diameter of 25 mm for a 60 mg sample). The chemical composition appeared homogeneous, with a slight oxygen deficiency at all depths. On heating, there was a rapid and significant loss of both mass and surface area, the former being assumed to be due to the loss of adsorbed water and surface hydroxyl groups in a variety of configurations.* Sintering in ambient air for extended periods at a temperature representing a significant fraction of the Tamman temperature increased the minimum particle size by at least one order of magnitude without causing any appreciable change in morphology of the disc on the pm scale.There was also evidence for a slight reduction in the oxygen deficiency within the lattice as a result of sintering, in agreement with the findings of Peterson,” who studied the effect of treating SnO, at 1323 K under oxygen atmospheres. Thermodynamic calculations performed by Peterson indicated that the equilibrium SnO(s) + +O,(g) -+ SnO,(s) (or the oxygen deficiency of the SnO,-, lattice) under various conditions could be used to rationalise the decrease in room temperature conductance of sintered as opposed to unsintered samples.Electrical Behaviour The electrical conductance of the as-pressed oxide in air as a function of measured temperature was found to be extremely complex and prone to significant hysteresis, dependent upon the history of the sample. On the first traverse of the range, unsintered discs displayed a sigmoid relationship between the conductance G and temperature T, with G typically reaching a maximum at 420+20 K, followed by a minimum at 520+ 50 K, increasing monotonically thereafter. Subsequent decrease of the temperature produced a wide range of behaviour. Although in general, the G-T characteristics exhibited a more linear form, the rate of change with temperature varied greatly. The initial sigmoid curve was largely unaffected by strict control of the water content of the test atmosphere, and hysteresis effects generally persisted, despite the adoption of more rigorous experimental methods.The air conductance behaviour of presintered samples was typically less complex, with an increasing tendency towards a monotonically rising G-T characteristic with extended duration of the sintering treatment. It is clear that there were significant differences between the electrical behaviour of the unsintered samples examined here, and that generally exhibited by ‘commercial ’ semiconductor gas sensors. However, these differences could be minimised by presintering of the oxide gel, which is perhaps not surprising since a sintering stage at temperatures well above the normal operating point is invariably a feature of the manufacturing of such devices.Gregg13 has discussed in detail the role of sintering in activated solids in some detail, noting that at relatively small fractions (ca. 30%) of the Tamman temperature ( qam), a surface diffusion is likely to reduce the number of defects in the near-surface crystal structure of the oxide without significantly altering the total pore volume. Goodman and Gregg14 have noted that for tin(rv) oxide, although the specific surface area is reduced by ca. one order of magnitude by sintering at 773 K due to surface diffusion (or, as suggested previously by us,’ by surface hydroxyl condensation), the total pore volume is virtually unaffected until qam is reached, when it collapses suddenly.These observations are consistent with the present findings ; treatment well below qam caused extensive losses in specific surface area without a largeP. G. Harrison and M . J. Willet 1919 fall in the porosity of the discs (as estimated from thickness measurements). No difference in the macroscopic morphology of the samples as a function of sintering time was detected despite the increase in minimum particle size. This is presumed to be because the plastic or viscous flow of particles during sintering occurs on a dimensional scale at least one order of magnitude below that resolved with the S.E.M. employed here. Therefore, the sintered discs suffered a large reduction in microscopic surface area, whilst retaining an open, porous macroscopic structure. These electrical data are discussed more fully in the following paper. M. J. W. is grateful for the support provided by British Coal during the course of this project, and for permission to publish this paper. The views expressed are those of the authors and not necessarily those of the British Coal Corporation. We thank Mr R. Hayward of British Coal for the electron microscopy studies and Mr B. Bellamy of the Materials and Surface Science Group, AERE, Harwell, for the XRD measurements. We thank S.E.R.C. for the award of a Research Grant. References 1 J. Watson and R. A. Yates, Electronic Engineering, 1985, 47. 2 S . Karpel, Tin and its Uses, 1986, 149, 1. 3 J. F. McAleer, P. T. Moseley, J. 0. W. Norris and D. E. Williams, J. Chem. SOC., Faraday Trans. 1, 4 J. F. McAleer, P. T. Moseley, B. C. Tofield and D. E.Williams, Proc. Br. Ceram. Soc., 1985, 36, 89. 5 H. Windischmann and P. Mark, J . Electrochem. Soc., 1979, 126, 627. 6 P. G. Harrison and A. Guest, J . Chem. SOC., Faraday Trans. I , 1989, 85, 1897. 7 E. W. Thornton and P. G . Harrison, J . Chem. SOC., Faraday Trans. 1, 1975, 71, 461. 8 See also P. G. Harrison and A. Guest, J . Chem. SOC., Faraday Trans. I , 1987, 83, 3383. 9 S. Kittaka, K. Morishige and T. Fujimoto, J. Colloid Interface Sci., 1979, 72, 191. 10 M. J. Miller, M. E. Warwick and A. Walton, J. Appl. Chem. Biotechnol., 1978, 28, 396. 11 U.K. Patent 1374375, 1974. 12 A. F. Peterson, Ph.D. Thesis (Rutgers State University, 1968). 13 S . J. Gregg, The Surface Chemistry of Solids (Chapman and Hall, London, 1961). 14 J. F. Goodman and S. J. Gregg, J. Chem. Soc., 1960, 1162. 1987, 83, 1323. Paper 8/01 165F; Received 21st March, 1988