J. Chem. SOC., Faraday Trans. I, 1989, 85(8), 1897-1906 Tin Oxide Surfaces Part 18.-Infrared Study of the Adsorption of very low Levels (2&50 ppm) of Carbon Monoxide in Air on to Tin(rv) Oxide Gel Philip G. Harrison* and Alan Guest Department of Chemistry, University of Nottingham, University Park, Nottingham NG7 2RD Three major types of surface species, unidentate carbonate, bidentate carbonate and carboxylate, are formed when tin(1v) oxide is exposed to dry air containing very low levels of CO (20-50 ppm). All three types appear to be formed immediately and simultaneously upon exposure to the CO-air mixtures, and the abundance of each continues to increase during the period of exposure. Little adsorption occurs at low oxide calcination temperatures when the surface retains a relatively high degree of hydroxylation, but abundances of the three surface species increase markedly a t temperatures > 570 K, and appear to reach a maximum at calcination temperatures of ca.590-610 K , declining thereafter. Although quite stable at lower tempera- tures, the surface bidentate carbonate and carboxylate slowly transform into surface unidentate carbonate at temperatures 2 ca. 400 K under an atmosphere of dry air. The chemisorption and electrical properties of tin(1v) oxide render it an excellent material for the detection of very low levels of carbon monoxide, and several solid-state sensors based on sintered pellets of SnO, are commercially To date, investigations of the mode of operation of such sensors have generally been restricted to studies of the electrical behaviour of the ~ x i d e .~ Little attention has been paid to the subtle relationship which must exist between the surface adsorption phenomena from atmospheres containing CO and the electrical changes thereby induced, afthough this relationship is absolutely critical to the successful operation of the sensor. It has previously4 been assumed that the chemical reactions occurring at the oxide surface 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. Our own earlier infrared studies of the chemisorption from CO-0, mixtures onto tin(rv) oxide gel, whilst showing that surface carbonate species are indeed formed, employed gas mixtures far too rich in CO to be applicable to the concentration regime expected in the operation of a sensor (< 100ppm).5 In this paper, therefore, we report infrared studies of the chemisorption from very low levels (20-50 ppm) of CO in air.Experimental Tin(rv) oxide gel was prepared as described previously.6 CO-air mixtures containing nominally 20, 30 and 50 ppm (accurate to within 1 ppm) CO were obtained from Rank-Hilger and purified by passing through a ' U '-tube packed successively with KOH pellets and a glass wool-P,O, mixture. In the text, the term 'dry air' is used to refer to a dried synthetic 0,(20 %)-N2(80 YO) gas mixture. The infrared cell used for these measurements comprised two gas cells of nominally identical pathlength, one positioned in the reference beam and the other in the sample 18971898 Tin Oxide Surfaces beam of the spectrometer, connected by glass tubing and equipped with a matched set of four sodium chloride windows, thus allowing absorptions due to gas-phase species to be compensated for by the spectrometer.To the sample cell was attached a vertical silica tube surrounded by a cylindrical furnace controlled by a ' Digi ' temperature controller (AEI). A winch arrangement permitted the tin(1v) oxide disc sample under examination, which was supported in a gold wire 'W'-cradle, to be raised and lowered between the furnace section and infrared cell as required. In turn, the furnace section was connected to a conventional vacuum system operating at < T0rr.l. In order to eliminate interference from water vapour, the spectrometer was purged for 35 min prior to use and throughout experimental runs by high-purity nitrogen, which was dried by passing through a silica gel column.Thermal pretreatment and the subsequent recording of spectra were usually carried out to a strict protocol so as to permit as great a comparability as possible between runs on different tin(rv) oxide disc samples. The general procedure was as follows: (i) the pressed disc of tin(1v) oxide gel (60 mg) was heated in the furnace section for 16-20 h under a dynamic vacuum of < lo-* Torr at the desired temperature, followed if necessary by a further 2 h at the same temperature in an atmosphere of 2 ppm of pure oxygen; (ii) the disc was lowered into the sample compartment of the infrared cell which was evacuated, and a background spectrum of an average of nine or more scans recorded; (iii) the appropriate gas or gas mixture was then admitted to the system (time zero), and a spectrum of the region 11 50-1750 cm-' was recorded at the ambient beam temperature (ca.329 K) (nine scans taking 13 min); (iv) further spectra (9 scans) were recorded at 20 min intervals for ca. 180 min. Times quoted in the Results section refer to the mid-point of the 13 min scan period. An applied gas pressure of 1 atml was employed for all the experiments with CO-air mixtures; pressure of carbon dioxide are as indicated in the Results section. Infrared spectra were recorded using a Perkin-Elmer 683 spectrophotometer equipped with a 3500 data station.Results Absorption onto Tin(rv) Oxide from Atmospheres containing CO and CO, The same general spectral features were observed for the surface adsorbates formed by adsorption from atmospheres containing either CO or CO, onto tin(rv) oxide under a wide range of conditions. The principal features of the spectra (after subtraction of the oxide background following calcination and, where appropriate, oxygen treatment) comprise broad bands at ca. 1600 and 1440 cm-' together with shoulders on the low- and high-wavenumber sides, respectively, a sharp band at 1223 cm-', and weak broad features at ca. 1380 and 1300 cm-', but their relative intensities varied with the heat pretreatment temperature of the oxide, the application or otherwise of oxygen treatment following heat treatment in vacuo, the composition of the gas phase, and the time of exposure of the oxide to the gas phase. As we have noted previously, exposure of tin(1v) oxide to an atmosphere of pure CO initially gives a similar spectrum rapidly followed by a loss of transmission due to bulk reduction of the oxide.6 The Effect of Increasing Temperature on the Surface-absorbed Species A tin(1v) oxide disc was heated for 17 h at 553 K under a dynamic .vacuum of ca.Torr. The deep-rust-coloured disc was then oxidised by heating in oxygen (20 Torr), again at 553 K, for 2.5 h, after which the colour was canary yellow. The disc was then exposed to an atmosphere of 30 ppm CO in air for 1 h, and the resultant spectrum t 1 Torr z 133.322Pa. 1 1 atm = 101 325 Pa.P.G. Harrison and A . Guest 1899 ' O f 0 'Ai 0 0 \ temperature/K Fig. 1. Plots of absorbance peak heights uerms temperature for the (a) 0 , 1450 cm-', (b) V, 1370 cm-', (c) 0, 1585 cm-', (d) A, 1223 cm-' and (e) A 1300 cm-' bands after adsorption of 30 ppm CO in dry air. Point A on the abscissa scale gives the peak heights of the bands in the original spectrum. The remaining points show the variation of peak height with calcination temperature under an atmosphere of dry air. comprises bands at ca. 1600 cm-', with weaker shoulders on the low-wavenumber side, ca. 1450, 1380, 1300 and 1223 cm-l. Replacement of the CO-containing gas phase by rapid evacuation to a pressure of Torr and admitting 760 Torr of dry air had little effect on the spectrum. However, heating the disc under an atmosphere of dry air resulted in a decrease in the intensities of the 1600 cm-' band and shoulders and the bands at 1300 and 1223 cm-l.Concomitantly, the bands at 1450 and 1380 cm-' increased in intensity. The variation with temperature of the peak height absorbance values for the 1585, 1450, 1370, 1300 and 1223 cm-' bands is illustrated in fig. 1. Absorption from CO-Air Gas Mixtures containing 20, 30 and 50 ppm CO Time-resolved difference spectra over a period of 180 min of the surface species formed on tin(1v) oxide heated at temperatures of 533-613 K from CO-air gas mixtures containing 20, 30 and 50 ppm CO were all qualitatively very similar, and exhibited the same general bands in the 1700-1 150 cm-l region as observed in the previous experiments.More detailed information concerning the growth and relative abundances of the respective surface species was obtained by measuring the absorbance peak heights for the 1585, 1540, 1450, 1300 and 1223 cm-' maxima. Spectra increased in intensity with increasing pretreatment temperature, and increasing time of exposure to the gas mixture, although the greatest changes occurred in the first 60 min of exposure. For all bands the peak absorbance heights were larger for mixtures containing 50 ppm CO. Rates of growth decrease markedly under both gas mixtures after ca. 1 h, and some reduction in1900 Tin Oxide Surfaces intensity of the 1300 and 1540 cm-' bands is observed after relatively long times. Substantially more surface-adsorbed species are formed on tin(1v) oxide which has been heated at temperatures 2 593 K.Pretreatment at lower temperatures gives rise to bands of significantly lower intensity, and in particular the 1223 cm-' is apparent on oxide discs heated and oxygen-treated at 533 K only after quite long exposure times. Discussion The Nature of the Absorbate Species Formed on Tin(rv) Oxide from Atmospheres containing CO and CO, The similarity of the spectra obtained indicate that the same surface species are formed from CO/O, mixtures and CO,, although their relative abundances may vary with the particular conditions of each experiment. The positions of the bands indicate that the surface species are carbonato in nature.' Possible sites for adsorption of CO and CO, on tin(1v) oxide are surface hydroxyl groups, single and bridging surface oxide atoms, and bare Lewis-acidic surface tin atoms? In addition, surface 0; species may be pre~ent.~ Coordination of CO to bare surface tin atoms may be excluded since no bands are observed in the region 2250-1 850 cm-', where matrix-isolated tin carbonyl molecules absorb." Carbon monoxide does adsorb at Sn2+ sites supported on S O , at 128 K giving a band at 2170 cm"," but no significant adsorption occurred at 233 K.Adsorption of CO at bare Zn2+ sites on the surface of zinc oxide has also been observed, which single- crystal studies at 80 K has indicated as through the carbonyl atom.', Interaction with surface hydroxyl groups to form a surface bicarbonato species may also be ruled out since neither the v(0H) nor 6(OH) vibrations are perturbed, and the 1223 cm-' assignment as the 6(COH) of a surface bicarbonate is precluded since it remains unshifted on deuteration.l3 Candidate surface species in the present case are unidentate carbonate, bidentate chelating or bridging carbonate, and unidentate or chelating carboxylate anions. The free C0:- ion D,, symmetry exhibits one Raman-active band at 1063 cm-l [v,(A,), symmetric CO stretching], one infrared-active vibration at 879 cm-' [v,(AL), out-of-plane deformation], and two of E symmetry at 141 5 cm-' (v3, antisymmetric CO stretching) and at 680cm-' (v4), in plane deformation, respectively, which are both Raman- and infrared-active.'* Coordination of the free ion lowers the point symmetry and causes the v3 mode to be split into two bands.The magnitude of the separation of these bands, Au, [not strictly v, but the separation of the higher and lower energy v(C=O) vibrations] may be used to distinguish different modes of coordination of carbonato groups. Values of ca. 100 cm-' are characteristic of unidentate carbonate, values of ca. 300 cm-l to bidentate (chelating) carbonate, whereas values b 400 cm-' are assigned to carbonate bridging two metal centres. The splitting Av3 has also been related to the polarising power of the cation in bidentate c~mplexes.'~ Band growth and desorption characteristics in the present study indicate that the pair of bands ca. 1600 and 1223 cm-' belong to the same surface species, as do the pairs of bands at ca. 1540 and 1300 cm-', and the bands at 1450 and 1380 cm-', i.e.only three general types of surface species are formed. Indeed, an F.t.i.r. spectrum recorded after only 30 s exposure to a 50 ppm CO-air mixture exhibits only six relatively sharp bands (fig. 2). Hence the broad nature of the bands, except for the 1223 cm-' band, and the presence of fine structure in many cases would suggest that several slightly different, but essentially similar, adsorption sites are available. This would be in accord with the rather poorly defined (in crystallographic terms) nature of the oxide surface in the present study . The bands at ca. 1600 and 1223 cm-', which we have also observed previously8,'3 from the adsorption of CO, onto tin@) oxide, we assign to the v,(A ,)[v(C-O,,) + v(C-O,)] and v,(B,)[v(C-O,) + d(0, - CO,,)] modes, respectively, of a surface bidentate carbonateP.G. Harrison and A . Guesl 1901 0.051 carboxylate bidentate carbonate 1600 1400 12oc 0 w avenumtxr/cm-' Fig. 2. F.t.i.r. spectrum recorded 30 s after exposure of tin(1v) oxide to 50 ppm CO in air. (I). The separation of these bands (377 cm-l) is intermediate between the values expected for chelating and bridging carbonato species and substantially higher than examples of purely chelating carbonate {e.g. the complexes [Co(NH,),CO,]X (X = C1, Br, I, NO,, ClO, and SO,), for which the separation is in the range 308-330 cm-' and [Co(en),CO,]Br, for which the separation is 297 cm-1}.16-18 The formation of a sur- face peroxycarbonate (11) similar to the [O,CO] ligand in the rhodium complex Rh(4-MeC6H,)[OOC(0)](PhP(CH,CH,CH,PPh,),], which could be formed by reaction of CO with one surface 0; and a surface-adsorbed oxide, is excluded since C-0 stretching bands characteristic of this species would be expected at ca.1660 and 1255 cm-l.19 The band envelopes at 1500-1 422 and 140 1-1 320 cm-l (separation ca. 100 cm-l) are readily assigned to the antisymmetric and symmetric CO stretching modes of a surface unidentate carbonate (111). Bands due to unidentate carbonate have been observed in several other systems. For example, on La,O, exposed to CO, where the corresponding bands occur at the bands at 1500 and 1390 cm-1,20 whilst the unidentate carbonate ligand in the pentammino cobalt complexes, [Co(NH,),CO,]X (X = C1, Br, I, NO,, and CIO,), which exhibit bands in the ranges 1449-1488 and 1351-1370cm-' with separations of 79-131 cm-1.16-1s1902 Tin Oxide Surfaces The 1560-1 520 cm-' group of bands, which appear as a series of shoulders on the low- wavenumber side of the more intense 158&1600 cm-l bands are associated with the envelope of bands at ca. 1300 cm-l.Bands observed at 1560 and 1330 cm-' on alumina2' and at 1575 and 1342 ern-',,, and 1570 and 1380 ~ m - ' , , ~ on zinc oxide have been assigned to surface carboxylate species. Two types of molecular alkali-metal carboxylates, [M+CO;-], with significantly different spectra, have been discerned in cocondensates of alkali metal with CO, in inert matrices. For example, the lithium complexes exhibit bands at 1755.7, 1750.9, 1221.4 and 1208.7 cm-' for the unidentate (C,) isomer and at 1568.6, 1574.9, 1329.9 and 1304.9 cm-' for bidentate (C2\,) isomer.24 In the present case, therefore, the species giving rise to the observed bands at 1540 and 1300 cm-' is assigned as a surface bidentate carboxylate (IV).0 II The band at 1182 cm-' is always present in the spectra, regardless of whether CO or CO, is the adsorbing gas and of the pretreatment temperature of the oxide. The band was always of low intensity and therefore more prominent for oxide discs pretreated at relatively low temperatures, but was not removed by D,O exchange. A weak band has been observed at 1180 cm-' in the spectra of matrix-isolated molecular alkali-metal carbonates but was not assigned.24 However, bands at 1850 and 1 180 cm-' in spectra of CO, adsorbed on alumina have been assigned to the CO stretching modes of a bridging surface carbonate.Assignment to a similar structure (V) would not be unreasonable in the present case, with the corresponding high-wavenumber mode being the weak band which occurs at ca. 1700 cm-'. Formation of Surface Carbonato Species Any satisfactory model of the adsorption has to account for the following observations : (i) The formation of surface unidentate carbonate, bidentate carbonate, and carboxylate occurs simultaneously, the relative abundances depending upon, particularly, the thermal pretreatment history of the oxide. (ii) All three surface species are formed at oxide pretreatment temperatures of 2 593 K, but very little formation occurs at pretreatment temperatures < 573 K. (iii) The most abundant species formed from CO-0, is surface bidentate carbonate, whilst the most abundant species formed from C0,-0, is unidentate carbonate.? (iv) Surface bidentate carbonate and carboxylate convert into surface unidentate carbonate under an atmosphere of dry air at temperatures > ca.400 K. In order to understand the chemisorption processes, it is necessary to examine the nature of the tin(rv) oxide surface present under the experimental conditions employed. We have previously described reconstructions of the probable available surfaces derived from the [loo], [I 101 and [loll planes of rutile-structure tin(xv) oxide,' and it is likely that these planes also occur in the microcrystalline particles of the tin(xv) oxide gel sample in t Assuming that the extinction coefficients for the major bands for both surface unidentate and bidentate carbonate species are similar.P.G . Harrison and A . Guest 1903 Fig. 3. (a) Model reconstruction of the [loo] surface after total dehydroxylation but no surface oxygen loss. Possible origins of 0, desorption are also indicated (A, loss from geminal bridging-non- bridging ; B, loss from geminal bridging-bridging ; C, loss from adjacent bridging-non- bridging). (6) Model reconstruction of the [ 1001 surface after partial surface oxygen loss. this study. Physisorbed water is lost from the oxide gel by 423 K, and further heating in uamo removes the surface hydroxyl groups in stages, with few hydroxyl groups remaining at temperatures in excess of 673 K8 When heated at moderately high temperatures (> ca.573 K) in uacuo, the oxide also loses oxygen, the oxygen deficiency being at least partially restored by treatment with molecular oxygen. The final result of the dehydroxylation process is shown in along with a reconstruction of the surfaces following subsequent thermal oxygen desorption. Surface deoxygenation must necess- arily occur via associative desorption of adjacent pairs of surface oxygen atoms, which can either be attached to the same surface tin atom or on adjacent tin atoms in the surface (fig. 3-5). Examples of all possible types are shown although not all will occur with equal probability, since desorption energies for different pairs will be different. The resulting oxygen-depleted surfaces will exhibit a greater number of bare surface tin sites.At the ambient temperatures employed in the CO-adsorption studies, adsorption of molecular oxygen at these sites will also produce O;,,, and O& as reactive surface oxygen specie^.^ Thus, the overall picture of the surface of the oxide under the conditions employed in this study is quite complex. However, this picture of the surface assists greatly in rationalising the present observations. The nature of the CO-chemisorption products and the lack of any surface carbonyl species [cf. transition-metal-exchanged tin(1v) oxide gelsz5] indicate the surface oxygen species are the sites for CO adsorption. At low pretreatment temperatures (< 573 K) the surface is still highly hydroxylated, and hence relatively few surface oxide sites are available resulting in little CO-adsorption.At higher temperatures, the abundance of1904 Tin Oxide Surfaces Fig. 4. (a) Model reconstruction of the [loll surface after total dehydroxylation but no surface oxygen loss. Possible origins of 0, desorption are also indicated (A, loss from geminal bridging- non-bridging ; B, loss from geminal bridging-bridging ; C, loss from adjacent bridging-non- bridging). (b) Model reconstruction of the [loll surface after partial surface oxygen loss. Fig. 5. (a) Model reconstruction of the [110] surface after total dehydroxylation but no surface oxygen loss. Possible origins of 0, desorption are also indicated (A, loss from adjacent bridging-non- bridging ; B, loss from geminal bridging-bridging). (b) Model reconstruction of the [110] surface after partial surface oxygen loss.adjacent pairs lating-bridging P.G. Harrison and A . Guest 1905 of surface oxygen atoms increases as does the formation of che- bidentate carbonate : At high evacuation temperatures, the occurrence of isolated surface oxygen species also increases, resulting in the formation of surface carboxylate : That both surface bidentate carbonate and carboxylate transform into unidentate carbonate at moderately high temperatures under an atmosphere of dry air would indicate that unidentate carbonate is the more thermodynamically stable surface species under these conditions. The conversion of surface carboxylate is via an oxidation process at carbon most probably involving a reactive adsorbed O;;a,, species : whilst transformation of bidentate carbonate into unidentate carbonate must just involve a change in the mode of coordination to the metal, perhaps in order to relieve strain in the bridging chelation : A similar process is probably responsible for the production of unidentate carbonate as a primary product from CO adsorption by interaction of a CO molecule with more strained pairs of surface oxygen atoms as, for example, found on the [loll plane.*1906 Tin Oxide Surfaces In contrast, unidentate carbonate is readily formed as a primary product from CO, via adsorption at surface oxygen.o=c=o 0 0 C \.+/ We thank the S.E.R.C. for the award of a research grant. References 1 2 3 4 5 6 7 8 9 10 1 1 12 13 14 15 16 17 18 19 20 21 22 23 24 25 J. Watson and R. A. Yates, Electronic Engineering, 1985, 47.S . Karpel, Tin and its Uses, 1986, 149, 1. J. F. McAleer, P. T. Moseley, J. 0. W. Norris and D. E. Williams, J. Chem. SOC., Faraday Trans. 1, 1987, 83, 1323. H. Windischmann and P. Mark, J. Electrochem. Soc., 1979, 126, 627. P. G. Harrison and E. W. Thornton, J. Chem. SOC., Faraday Trans. 1 , 1978, 74, 2597. P. G. Harrison and E. W. Thornton, J. Chem. SOC., Faraday Trans. 1, 1975, 75, 461. M. Aresta, C. F. Nobile, V. G. Albano, E. Forni and M. Manassero, J. Chem. SOC., Chem. Commun., 1975, 636. P. G. Harrison and A. Guest, J. Chem. SOC., Faraday Trans. I , 1987, 83, 3383. J. P. Joly, L. Gonzalez-Cruz and Y. Arnaud, Bull. SOC. Chim. Fr., 1986, 11. A. Bos, J. Chem. SOC., Chem. Commun., 1971, 26. B. Rebenstorf and R. Larsson, Acta Chem. Scand., Sect. A , 1980, 34, 239. K. L. D’Amico, M. Trenary, N. D. Shinn, E. I. Solomon and F. R. McFeely, J. Am. Chem. Soc., 1982, 104, 5102. P. G. Harrison and B. Maunders, J . Chem. SOC., Faraday Trans. I , 1984, 80, 1357. G. Hertzberg, Infrared and Raman Spectra of Polyatomic Molecules (Van Nostrand, New York, 1945), p. 178. J. P. Jolivet, Y. Thomas and B. Taravel, J. Mol. Struct., 1982, 79, 403. K. Nakamoto, J. Fujita, S. Tanaka and M. Kobayashi, J. Am. Chem. SOC., 1957, 79, 4904. J. Fujita, E. Martel and K. Nakamoto, J. Chem. Phys., 1962, 36, 339. J. A. Goldsmith and S. D. Ross, Spectrochim. Acta, Part A, 1968, 24, 993. L. Dahlenburg and C. Prengel, Organometallics. 1984, 3, 934. M. P. Rosynek and D. T. Magnuson, J. Catal., 1977, 48, 417. L. H. Little and C. H. Amberg, Can. J. Chem., 1962, 40, 1997. J. H. Taylor and C. H. Amberg, Can. J. Chem., 1961, 39, 535. S. Matsuchita and T. Nakata, J . Chem. Phys., 1962, 36, 665. Z. H. Kafafi, R. H. Hauge, W. E. Billups and J. L. Margrave, J. Am. Chem. SOC., 1983, 105, 3886. P. G. Harrison and E. W. Thornton, J. Chem. SOC., Faraday Trans. I , 1979, 75, 1487. Paper 8/01 11 1G; Received 18th March, 1988