J. MATER. CHEM., 1994, 4( l),23-27 Metal Stannates and their Role as Potential Gas-sensing Elements Gary S. V. Coles,* Stephanie E. Bond and Geraint Williams Department of Electrical and Electronic Engineering, University of Wales, Swansea, UK SA2 8PP A selective gas sensor, sensitive to the presence of carbon monoxide in preference to the lower hydrocarbons, can be fabricated from a mixture of bismuth oxide and tin dioxide when sintered at 800°C. At temperatures above ca. 650°C a solid-state reaction takes place in which bismuth stannate (Bi,Sn,O,) is formed and in the above sensor all of the Bi203 is converted to the stannate. This material is one of a group of mixed oxide stannates which possess a pyrochlore structure and have the general formulae M2Sn207. Several of these materials can be produced by heating an intimate mixture of lanthanum metal oxides (M203where M =La, Nd, Sm, Gd, Yb, Dy, Tm and Ho) and tin dioxide at temperatures of 1500 "C.Sensors were produced containing these materials in an attempt to reproduce the behaviour of the original device and further understand the chemical, physical and topographical features responsible for conferring selectivity.However, none of the new sensors produce results consistent with those observed for the tin-bismuth system. It has subsequently been shown that when SnO, is subjected to heat treatment at 1500 "C, it can exhibit both resistance increases and decreases upon exposure to the same gas, depending on the operating conditions. In order to reduce these interfering effects the sintering temperature was lowered to 1350"C and sensors fabricated from pure tin dioxide fired at this temperature respond in a conventional manner to all reducing gases tested.A series of sensors produced from some of the SnOJM,O, materials listed above exhibit a general trend of increasing carbon monoxide and hydrogen sensitivity with decreasing M3+ ionic radius. Tin dioxide sensors, operated at elevated temperatures, respond via conductance modulations to a wide range of reducing and oxidising gases. Considerable effort has been directed towards improving the specificity of these devices. A common practice is to incorporate various additives with the SnO, powder prior to sensor fabrication.' For example, the addition of small quantities (typically ca.1 wt.% of the sensing material) of palladium or platinum can increase sensitivity towards reducing gases such as hydrogen and the lower hydrocarbons signifi~antly.,.~ Previous work by this group4 has established that a sensor element fabricated from a tin dioxide-bismuth oxide mixture, when heated to a temperature of SOO'C, responds selectively to carbon monoxide in the presence of methane. High-temperature treatment of the oxide mixture results in a solid-state reaction between the tin and bismuth oxides resulting in the formation of bismuth stannate: 2Sn0, +Bi203 +Bi,Sn,O, (1) X-Ray powder diffraction studies carried out by Roth' estab- lished that this compound possesses a pyrochlore structure.Furthermore, it appears that in addition to Bi, many other metal oxides with the general formula M203 react, when heated together in an intimate mixture with tin dioxide, to form M,Sn20, stannates possessing a pyrochlore structure.6 Metal oxides from the lanthanide series of the periodic table constitute the majority of these and can be substituted in place of bismuth oxide, thus providing a means of examining whether formation of pyrochlore-type stannates induce improved CO selectivity over CH, via an identical mechanistic process. Lanthanide metal stannates with the general formula M2Sn,0, belong to a family of compounds possessing a cubic pyrochlore structure. This has space group Fd3m (Oh7),com-prising eight M2Sn2060' formulae per unit cell., The structure consists of corner-sharing Sn06 octahedra surrounding hexag- onal vacancies, these containing M,O' tetrahedra.Fig. 1 (a) shows the 'pyrochlore unit' consisting of four octahedra sharing corners around a vacancy,8 while Fig. l(b) gives a representation of the octahedral arrangement within the pyr- ochlore frame~ork.~ All Sn-0 bond lengths are identical and are generally of the order of 0.20 nm. Two types of oxygen atom are attached to the M atom: six oxygen atoms associated with the octahedra (M-0 bond lengths are ca. 0.26 nm) plus octahedra Sn06 tetrahedra M&' 00'4BO OM Fig. 1 Diagrammatic representation of the octahedra and tetrahedra constituting the pyrochlore structure. The pyrochlore unit consisting of four corner-sharing SnO, octahedra is shown in (a), while (b)indi-cates the association of these units along with M,O' tetrahedr,i in the pyrochlore framework.two additional oxygen atoms (M-0' bond lengths are cu. 0.23 nrn)., Room-temperature X-ray diffraction was used for structural examination and to determine the approximate temperatures at which the respective lanthanide stannates form with a view towards possibly incorporating lower sintering temperatures in the sensor fabrication process. High-temperature X-ray diffraction was employed to further study the solid-state reaction between the oxides of tin and bismuth and to gain some insight into the mechanism by which the selective response of a sensor element composed of a sintered mixture of the two oxides arises.Experimental A Guinier-De Wolff X-ray diffraction camera manufactured by Enraf Nonius Delft employing Cu-Ka radiation, was used to generate diffraction patterns from powdered crystalline samples at ambient temperature. High-temperature X-ray diffraction apparatus comprising a Siemens D500 diffractometer instrument with 6-26 geometry was used to take X-ray diffraction patterns at temperatures up to 1000°C. The basic principles of the corresponding room-temperature technique were employed. In brief, the sample was contained in a high-temperature environmental cell subsequently mounted on an X-ray diffractometer. The diffractometer was controlled by a PDP-11 computer employing Siemens DIFFRAC 500 software.Temperature- resolved X-ray patterns were produced upon transferring data files to a PC spreadsheet. Two types of sensor were tested during the course of these studies. The first type was fabricated from an aqueous paste containing the Sn02-M203 mixture which had previously been subjected to high-temperature treatment. This was applied across the contact array of an alumina substrate supplied by Rosemount Engineering Ltd., allowed to dry and then sintered for a further 2 h at 1000°C. Alternative sensors employed tin dioxide-based pressed pellets as the active material and were prepared in the following manner. First, a disc of 13 mm diameter and 1 mm thickness was produced by pressing 0.4 g of the desired powder in a stainless-steel die under a pressure of 10 tons for 15 min.The discs were then sintered at the appropriate temperature (usually 1500 "C, ensuring complete solid-state reaction of the SnO, and M203 additive) for 2 h in air. A square section of the disc (ca. 2x2mm) was cut out and attached across the parallel contact pads of an alumina substrate by means of a conducting Pt paste. The complete pressed pellet-substrate assembly was then annealed for 1h at 750°C in order to attain the necessary mechanical stability of the Pt adhesive. Full details of the procedures adopted for blending gas mixtures and determining sensor response are given el~ewhere.~.~ Results and Discussion Room-temperature X-Ray Diffraction Analysis Table 1 gives a list of stannates prepared by heating an intimate stoichiometric mixture of tin dioxide and M203 powders.Initial attempts to form the stannates involved the use of a sintering temperature of 1000°C. However, with the J. MATER. CHEM., 1994, VOL. 4 exception of Bi,Sn,O, none of the systems studied showed any evidence of solid-state reaction when analysed by X-ray powder diffraction. At higher temperatures some stannates are clearly formed more readily than others, for example La,Sn,O, and Sc,Sn,O, (although this latter stannate deviates from a cubic pyrochlore structure). All the materials prepared differ considerably from bismuth stannate in that they require a significantly higher temperature, usually in excess of 1300 "C, to mediate the solid-state reaction involved.Reducing-gas Response of M"'-Doped Tin Dioxide Sensors Initial studies were performed on mixed oxides which had been presintered at 1500°C for 1 h. However, owing to the anomalous sensing behaviour exhibited by undoped tin diox- ide subjected to heat treatment at this temperature," no comparisons between the results given by these materials and the Sn0,-Bi,O, system could be made. The pretreatment temperature was therefore decreased to 1350°C and the duration of the exposure increased to 2 h. A cross-section of M203 additives was chosen from the series givtn in Table 1, covering a wide range of ionic radii from 1.016 A for La3+ to 0.85 A for Lu3+. It has been suggested6 that the size of the M3+ ion present in the M,Sn,O, lattice influences strongly the extent of distortion inherent in the cubic pyrochlore structure, where the greatest distortion of the SnO, octahedra is caused by the smallest M3+ ion.It may be, therefore, that the stannates of Lu, Yb and Y, which possess the smallest M3+ radii of the lanthanide metals studied are most closely related structurally to Bi,Sn,O,. Mixtures of tin dioxide and 15 wt.% M,O, (where M =Lu, Yb, Y, Gd, Sm, Nd or La) were wet ground and subjected to thermal treatment. X-Ray diffractograms of the powders formed confirmed that for each sample the majority of the M203 additive had been converted to the stannate. The gas- sensing properties of these materials were then evaluated after sensor fabrication. A summary of the results obtained for this series of sensors tested at two different operating temperatures in 1 vol.% concentrations of CO, CH4 and H, in dry air is given in Table2.Sensor response is represented by the ratio Ro:Rgaswhere Ro is the resistance in clean dry air and Rgasis the resistance in a reducing gas-air mixture. Therefore, the greater the RO:Rgasratio, the higher the sensitivity of the device to the specified gas. The absence of any results for the Sn0,-Lu203 material is explained by its highly conductive nature (sensor resistance at room temperature = 1.5 R) and therefore its unsuitability as a gas-sensor element. Interestingly, the onset of high conductivity matched the Table 1 Stannates prepared by heating an intimate stoichiometric mixture of tin dioxide and M,03 powders.Sintering times/h are given in parentheses compounds constituting reaction mixture Bi,03-2Sn0, La,O3-2SnO, Sc2O,-2SnO, Yb,O3-2SnO, Tm,0,-2Sn02 Sm,O3-2SnO, Nd,03-2Sn0, Gd,03-2Sn0, Er2O,-2SnO, Lu20,-2Sn0, Y20,-2Sn0, Dy,03-2Sn0, Ho,O,-2Sn02 Eu,03-2Sn0, temperature of stannate emergence /"c 650 (1) 1100 (24) 1200 (24) 1300 (24) 1200 (24) 1350 (2) 1350 (2) 1350 (2) <1420 (2) 1350 (2) 1350 (2) < 1500 (2) > 1500 (2) < 1500 (2) temperature at which the stannate is fully formed/"C 800 (1) 1300 (24) 1200 (24) 1420 (24) 1500 (24) 1420 (2) 1420 (2) 1420 (2) 1420 (2) 1500 (2) 1500 (2) 1500 (2) > 1500 (2) 1500 (2) literature values of stannate formation temperatures/"C 1250 ( 1)5 1550 (1)5 1400 (24)6 1400 (24)6 1400 (24)6 1500 (24)6 1550 (1)' 1400 ( 12)6 1400 (24)6 1400 ( 24)6 1400 (24)6 1500 ( 18)6 1500 ( 18)6 1400 (24)6 J.MATER. CHEM., 1994, VOL. 4 Table 2 Response of Sn02-M203 (15 wt.%) thick films, presintered at 1350°C for 2 h in air prior to sensor fabrication, to 1 vol.% reducing gas concentrations in dry air. operating temperature =300°C dry air dopant resistance RO/G Ro/Rco Ro/R,,, R0/RH2 none 1.8 x lo6 4.2 3.0 29 Yb203 9.5 x lo8 4.5 2.9 33 y2°3 3.4 x lo8 2.8 1.4 20 Gd203 4.2 x lo8 2.6 2.0 9.1 Sm203 8.4 x 10' 1.5 1.3 4.2 Nd203 1.9 x 109 1.5 1.1 6.2 7.8 x 10' 2.4 1.6 8.3 none" 6.0 x 104 31 12.2 134 Bi203" 1.5 x 107 9.8 1.1 19.4 operating temperature =430°C M3+ ion@ dopant radius/A ROIRCO ROIRCH.4 R0/RH2 none -2.4 4.7 29 Yb203 0.858 4.4 4.0 7.7 y2°3 0.893 5.9 3.2 14.3 Gd203 0.938 1.5 1.7 2.7 Sm203 0.964 1.7 1.7 2.1 Nd203 0.995 1.2 1.3 2.1 1.016 1.8 2.3 2.6 none" -1.o 13.3 2.4 Bi203" 0.960 2.3 1.o 5.0 "Sensors were pre-sintered at a temperature of 800 "C.emergence of the stannate since the conductance of Sn0,-Lu203 mixtures sintered at temperatures of 1300 "C or less were similar to pure tin dioxide. However, a striking feature of the remaining sensors are their high resistances, often several orders of magnitude greater than the undoped SnO, subjected to pretreatment at the same temperature. However, this is not wholly unexpected given the analogy with the highly resistive Sn02-Bi,Sn,07 sensor formed upon high-temperature sintering of the corresponding tin-bismuth oxide mixture.' The results listed in Table 2 indicate a general trend of decreasing CO and H, sensitivity and a greater degree of selectivity at the expense of methane response, especially at the lower temperature employed as the M3+ dopant ionic radius is increased.However, there is no real evidence of selectivity over the whole temperature range as is the case with the bismuth oxide system, since CH4 response increases for the majority of devices tested as the operating temperature is raised to 430°C. The lack of a positive result from these studies may be due to three possible causes.(i) It may be that the pyrochlore structure of Bi,Sn207 is not the important factor in determining selectivity to CO and H, in preference to methane. The action of the bismuth stannate may be to change the nature of the oxygen species, such as 0-or O,, present on the tin dioxide surface, thus controlling the types of reaction occurring at the semiconductor/gas interface. (ii) It is also possible that the family of stannates studied are not sufficiently similar to Bi,Sn207 and are therefore unable to moderate reactions in the same manner when incorporated in tin dioxide sensors. The stannates studied here possess cubic pyrochlore structures, though these become more distorted as the M3+ ionic radius of the M,Sn,07 material decreases., However, there appears to be some conflict in the literature concerning the validity of this previous statement.The findings of Vandenborre and HUSSO~,~ who used IR and Raman techniques to analyse some of the structural characteristics of M2Sn207 compounds, suggest that only lanthanum stannate from the series M=La, Sm, Gd, Yb and Lu exhibits some distortion of the SnO, octahedral network. (iii) The re-grinding step required after thermal pretreatment in order to produce an aqueous paste of the material may destroy possible effects caused by any solid-state diffusion of stannate to the surface of tin dioxide grains. This process is probably mfluen- tial if the action of Bi2Sn,07 is that of a molecular sieve on the surface of SnO, grains when incorporated into a tin dioxide-based sensor.In order to overcome possible problems caused by re-grinding the sensing material after the high-temperature sinter- ing step, active sensor elements were prepared from pressed pellets of the SnO,-M,O, mixtures as described in the Experimental. A summary of the results obtained for these devices is shown in Table 3. Sensor materials prepared by this method behave differently in several ways compared to the polycrystalline Sn0,-M,O, thick-film sensors described above. First, in order to attain maximum gas sensitivity, the sensor operating temperatures for pressed-pellet sensors are usually higher than previously observed. This is presumably owing to a lack of contact between the disc section and substrate surface.Consequently, heat transfer must occur across a narrow gap caused by protrusion of the Pt adhesive contact areas. Subsequent thermocouple measurements on the pressed-pellet-type sensors showed that the temperature of the disc section surface was 100-130°C lower than the sub- strate surface temperature. Secondly, it appears that sensitivity to CO and CH4 is greatly reduced while response to hydrogen remains very substantial. There seems also to be some degree of selectivity conferred upon addition of the lanthanide metal oxides, though the effect is not as startling as that observed for Bi,03-doped tin dioxide. Additives such as Yb, Sm, Gd and La appear to yield sensors exhibiting CO selectivity in the presence of methane, though the magnitude of the observed response is not sufficiently large to make any definite conclusions.In comparison to undoped SnO,, the hydrogen respmse of the lanthanide metal oxide-doped pressed pellets is greatly enhanced. Changes in conductance of up to 3 orders of magnitude are observed upon exposure of these types of sensor to 1 vol.% concentrations of H, in air. Fig. 2 shows a comparison of the performance of the two different types of sensor studied here, where the active material in both cases is composed of a Sn0,-Sm,O, (15 wt.%) mixture sinttx-ed at high temperature. The difference in the profiles of conductance uersus [H,] plots for the pressed pellet and thick-film sensors may arise from a change in the following physical properties.(i) Porosity: this is likely to be the most influential controlling factor of diffusion of reducing gas through the sensing material surface. The porosity of pressed pellets is expected to be greatly decreased when compared with a film of srntered Table 3 Response of various pressed-pellet-type sensors fabricated from Sn02-based mixtures to 1 vol.% inclusions of CO, CH4 or H2 in dry air. All pressed pellets tested were subjected to heat treatment at 1500°C for 2 h unless otherwise indicated substrate temperature dopant 1°C RoIRco none 515 1.26 1.10 8.9 Yb203 y2°3 Gd203 460 485 445 1.14 1.72 1.24 1.o 1.46 1.02 36 1500 98 Sm203 Nd203 La203 none" 495 475 435 480 1.59 1.31 1.52 1.37 1.03 1.22 1.06 1.61 260 120 68 10.8 Bi203" 400 2.09 1.o 20 "Heat treatment at 800°C for 2 h.J. MATER. CHEM., 1994, VOL. 4 0.01r IA 0.001 I ,1 UJ 100 1000 10000 53 1000r I 100 -(b) i10 -/x 1°1 /-* / u0.1100 1000 10000 [gas1(PPm) Fig. 2 Response characteristic obtained for sensor elements composed of presintered Sn0,-Sm,O, ( 15 wt.%) in (a)thick-film and (b)pressed-pellet form. Sensor operating temperatures are 310 and 495"C, respectively. x =CO; 0=methane; * =hydrogen polycrystalline SnO,, thus favouring the detection of the lightest gas, namely hydrogen. (ii) Geometry: it has been shown by several researchers" that a change in the thickness of the tin dioxide film can modify response to a given gas significantly and therefore affect sensor specificity.In the present case it should be noted that the thickness of a section of pressed disc is usually significantly greater than the depth of a sintered Sn02 film. High-temperature X-Ray Diffraction Analysis From the results presented above, the Sn02-Bi203 system appears unique in conferring CO selectivity in the presence of methane over the whole operating-temperature range of the sensor. Experiments have shown that a sensor composed entirely of Bi,Sn,O, does not respond to reducing gases when operated at elevated temperatures. Conversely, undoped tin dioxide is an excellent sensing material (see Table 2) which responds to all reducing gases tested. Further study of the system involved X-ray diffraction analysis of the sensor material when sintered at high temperature, which represents an integral part of the fabrication procedure. Fig.3 shows a temperature profile of the reaction initiated when a 2:l mixture of SnO, and Bi203 is heated. The temperature was raised from 200 to 800°C at a rate of 8°C min-l, patterns being obtained at 100"C intervals. The reac-720 700 680 660 640 620 600 dULtemp.room Fig. 3 X-Ray diffraction pattern of a 2Sn0,-Bi,03 mixture with increasing temperature/"C and after cooling to room temperature tion mixture was subsequently cooled at 25 "Cmin-' to room temperature. The complete temperature profile gives details of product formation at various temperatures, thus producing constructive information concerning mechanisms inherent within the reaction. The relevant peaks of an X-ray diffraction profile of the Bi,03 +2Sn0, system at room temperature constituted a reference guide for Fig.3. All profiles were taken with the machine set in the range 28= 18-38". Bi203 peaks existed at 28~19.8,21.8, 24.7, 28, 30.2, 32.5, 33.9, 35.1, 35.4, 37.0 and 37.6", whilst characteristic tin dioxide peaks appeared at approximate 28 values of 26.5 and 37.9'. Solid-state reaction commences at a temperature of ca. 650°C as shown by the appearance of the product Bi,Sn,O,. Also evident at 750°C (28=31.5") is a transitional peak corresponding to the 6 form of Bi203, thereafter absent at 800°C due to its reaction with SnO, for stannate generation.Its pattern indicated a cub& pyrochlore structure with lattice parameters a =b =c =10.7 A. Conclusion of the stannate for-mation process was indicated by the complete absence of reactants after a dwelling period of 25-35 min at 800 "C. Comparison of the room-temperature stannate peak and that present at 800°C showed a peak shift to the right upon cooling. Low intensity reflections present in the low-temperature pattern also indicated distortion of the structure upon cooling. A temperature-programmed X-ray diffraction profile of Bi203,shown in Fig. 4, comprises a temperature range set at 600-800 "C, with patterns recorded using 20 "C increments. The room-temperature pattern taken before heating is shown at the foot of the diagram. The set of patterns shows the transition of monoclinic a-Bi203 to the cubic S-Bi203form occurring within ca.740-760 "C, indicating characteristic peaks for both forms. These results concur with the findings of Levin and Roth12 who determined that the transformation takes place at a temperature of 730 f5 "C. The characteristic a-Bi2O3 peak was revealed at 28=ca. 32.5", the distinguishing 6-Bi203 peak appearing at 28~31.4".A shift in the Bi203 peak from 28~~27.3"(at room temperature) to 28~27.0"(at J. MATER. CHEM., 1994, VOL. 4 iBi203 I room temp. 850 800 750 700 650 600 500 400 300 200 Bi2Sn207 I/i,/,II1,I,I,I,IIlllll,l,l,~ll,//l~ll11­18 20 22 24 26 28 30 32 34 36 38 2Bldegrees Fig.4 Observed changes in the X-ray diffraction pattern of Bi,O, upon heating in the 600-800 "C region 800 "C)was detected. a-Bi203 and d-Bi203 peaks also showed corresponding deviation to the left upon increasing tempera- tures, indicating structural distortion. Conclusions Of the tin dioxide-metal stannate mixtures studied, the Sn02-Bi,Sn207 system appears to be unique in several ways. First, stannate formation via the solid-state reaction of the oxides of tin and bismuth proceeds at relatively low tempera- tures compared with systems incorporating lanthanum group metals. In addition, a sensor composed of a sintered mixture of Sn02 and Bi203 (15-18 wt.%), responds selectively to CO containing atmospheres in the presence of CH4.Gas sensors produced from mixtures of tin dioxide with other pyrochlore- forming M203compounds did not display the same character- istics. A general trend of increased reducing gas sensitivity, most notably to CO and H2, with decreasing M3+ ionic radius was observed for thick-film sensors prepared from Sn0,-M203 powders which had been subjected to heat treatment at 1350°C. However, none of the devices tested exhibited similar selectivity traits to the SnOz-Bi203, i.e. they did not exhibit negligible methane sensitivity over the whole operating temperature range. Sensors fabricated from pressed pellets of the presintered materials displayed greatly reduced CO and CH4 sensitivities. However, upon exposure to hydro- gen, large changes in sensor resistance of up to 3 orders of magnitude were observed.Temperature-programmed X-ray diffraction profiles of heated Bi203 and a Bi,03-2Sn02 mixture have revealed the transitional d-Bi203 phase present above 760 "C, whilst observing Bi203 heated within the temperature range 600-800°C. In addition, this appears upon heating the Bi203-2SnOz mixture to 800 "C, disappearing above 750 "C owing to its reaction with Sn02. Formation of bismuth stannate, Bi,Sn,O,, initially occurs at ca. 650 "C, the product peaks indicating a cubic pyrochlore structureo with correspond- ing lattice parameters of a=b=c= 10.7 A. Further infor- mation supplied by profiles of both Bi203 and Bi203-2Sn02, shows increased distortion of the structure (denoted by the appearance of low-intensity reflections and peak shifts) upon cooling. References 1 S. R. Morrison, Sens. Actuators, 1987, 12,425. 2 K. Ihokura, New Mater. New Processes, 1981,1,43. 3 N. Yamazoe, Y. Kurokawa and T. Seiyama,Sens. Actuators, 1983, 4,283. 4G. S. V. Coles, K. J. Gallagher and J. Watson, Sens. Actuators, 1985,7,89. 5 R. S. Roth, J. Res. Natl. Bur. Stand. (U.S.), 1956,56, 17. 6 F. Brisse and 0.Knop, Can. J. Chem., 1968,46,859. 7 M. T. Vandenborre and E. Husson, J. Solid State Chem., 1983, 50, 362. 8 H. Nyman, S. Andersson, B. G. Hyde and M. O'Keeffe, I. Solid State Chem., 1978,26, 123. 9 G. S. V. Coles, G. Williams and B. Smith, Sens. Actuators, R, 1991, 3, 7. 10 G.S. V. Coles and G. Williams, J. Mater. Chem., 1992,2,23. 11 D. E. Williams, in Solid State Gas Sensors, ed. P. T. Moseley and B. C. Tofield, Adam Hilger, Bristol, 1987, p. 71. 12 E. M. Levin and R. S. Roth, J. Res. Natl. Bur. Stand. (U.S. /,1964, 68(A), 189. Paper 3/04264B; Received 20th Julj,,1993