Photoluminescent Spectra of Surface States in Alkaline Earth Oxides BY SALVATORE COLUCCIA,? A. MICHAEL DEANE AND ANTHONY J. TENCH* Chemistry Division, AERE Harwell, Oxfordshire Received 15th May, 1978 Photoluminescence has been observed from high surface area alkaline earth oxides with exciting light of a much lower frequency than that expected from the band gaps of the bulk oxides. This Iuminescence can be quenched by oxygen and hydrogen and the quenching is reversible under some conditions. It seems probable that both the excitation and the luminescence spectra observed arise from excitons in the surface region. The different behaviour towards oxygen and hydrogen has been used to develop a model in which the excitation site, where absorption of light occurs, is thought to be associated with anions on the surface in states of unusually low coordination such as might be found at step, edge or corner sites on the surface. The luminescence is thought to be associated with a cation on the surface in a similar state of coordination, Pure MgO in single crystal form shows no absorption in the near ultraviolet but high surface area powders have been shown to give a luminescent spectrum when excited by U.V.light with an energy much less than the band gap of the bulk solid. The absorption and emission were attributed to the presence of local surface states, intrinsic to the oxide, which correspond to ions in sites of unusually low coordination or to the presence of high index planes on the surface. Molecular species such as hydroxyl groups on the surface can also give rise to absorption and luminescent spectra which are significantly different from those associated with the intrinsic ions.Similar surface states should be present on other oxides and it is likely that ions in unusually low states of coordination play a major role in determining the reactivity of the solid surface. With this in mind, we have studied the photoluminescent behaviour of the alkaline earth oxides, both in vacuu and in the presence of oxygen and hydrogen, to try to establish the chemical properties of the sites involved in the absorption and luminescent processes. The results from the excitation data are compared with reflectance meas~rements.~ EXPERIMENTAL The apparatus used was developed during this work from that described previously.' A 250 W xenon lamp provided the excitation source and excitation wavelengths were selected using two Spex f4 monochromators coupled to form a double monochromator.The studies were limited to 2 > 230nm (5.40eV) because of the very low intensity of the exciting light at shorter wavelengths. Emission spectra were analysed using a similar f4 monochromator. The excitation and emission band pass wits 5 nm for all the results reported here. When the emission wavelengths permitted, a sharp cut filter (Corning 3-74) was used before the emission monochromator to eliminate higher orders of the exciting wavelength. A small fraction of the monochromatic exciting light was detected with it photo- multiplier, type 95263 (silica window, bialkali cathode) and emission spectra were detected with a photomultiplier, type 9798B (glass window, S20 cathode). An Ortec photon counting t On leave from the Institute of Physical Chemistry, University of Turin, Italy.291 32914 SURFACE STATES I N OXIDES system was used to compare the photon pulse rates from the two photomultipliers and to present spectra corrected for variations of excitation intensity with wavelength or time. No corrections were made for the variation of photomultiplier sensitivity with wavelength but both multipliers were used in wavelength ranges where their photon sensitivity with wavelength was fairly uniform. Variable temperature measurements below ambient were carried out with the specimen cell mounted in a flow of cooled nitrogen gas, while for those above ambient the specimen was mounted in a small furnace equipped with two windows at right angles. Preliminary measurements of luminescent lifetimes were made on an Applied Photo- physics nanosecond spectrometer, model SP2.Lifetimes were measured using the mono- chromator to select an excitation wavelength and the integral emission over all luminescent wavelengths was collected on the photomultiplier. Magnesium oxide was prepared by controlled thermal decomposition of high purity hydroxide or carbonate by standard techniques '9 and calcium and strontium oxides from the Specpure carbonates to give surface areas of 100-200 m2 g-l for MgO, 50-70 m2 g-' for CaO and 10-20 m2 g-l for SrO. Most preparations were carried out on a vacuum line with a base pressure about Torr), but one preparation of magnesium oxide was carried out on a grease free system with a final pressure about N m-2 (lo-' Torr).A low area specimen of barium oxide was prepared by outgassing a specimen of Hopkin and Williams " fine chemical " barium oxide to 1200 K in vacuum. A further specimen was prepared by decomposing barium nitrate on magnesium hydroxide in the manner described by Zecchina and Stone.4 Each preparation was carried out in a silica ampoule attached to a rectangular Spectrasil cell 2mm thick and the powder shaken into the cell after preparation. The effect of hydrogen and oxygen on the luminescence was studied using Specpure gases with the specimen attached to the vacuum line while mounted in the spectrometer ; gas pressures quoted were measured after equilibration with the surface.N m--2 RESULTS LUMINESCENCE OF CLEAN OXIDE SURFACES All the alkaline earth oxides could be excited with ultraviolet light to show a broad luminescence band which in general moved to lower energy as the cation size increased (table 1, fig. 1 to 4). The excitation spectra show a doublet for all the oxides, however the maxima of the peak at higher energies could not be observed clearly for MgO and CaO because of the high noise levels below 230nm. The position and shape of the luminescence band varied significantly with the excitation TABLE 1 .-ABSORPTION,3 ' EXCITATION AND EMISSION SPECTRA OF ALKALINE-EARTH OXIDE POWDERS absorption excitation emission oxide lev lev lev MgO 5.70; 4.58 >5.40; 4.52 3.18 CaO 5.52; 4.40 >5.40; 4.40 3.06 SrO 4.64; 3.96 4.43 ; 3.94 2.64 BaO 3.60; 3.22 3.70 2.67 wavelength indicating that more than one component was present.The intensity of the luminescence and the ratio of the excitation peaks varied to some extent from one preparation to another and considerable care was needed to obtain reproducible results. The luminescent species were present after < 1 s irradiation in the excitation waveband and this was taken as evidence that the luminescence was not the result of a photochemical reaction. At the illumination intensities used, ultraviolet induced changes were limited to intensity variation of a few per cent ; spectra were normally recorded with the specimen in equilibrium with the excitation source.S. COLUCCIA, A. M . DEANE AND A .J . TENCH 2915 Some samples of MgO were prepared with Fe, Cr or Mn added (20 to 100 p.p.m. of impurity) to examine whether the luminescence described above was associated with impurity transition metal ions. All these samples had the same broad lumines- cence but in addition for the CrlMgO sample both the surface and bulk luminescence of Cr3+ ions was observed around 710 nm (1.75 eV). Surface contamination by carbon compounds does not seem to cause the luminescence since a specimen of magnesium oxide prepared under completely grease-free conditions showed a similar spectrum to that of other preparations. Deliberate attempts to contaminate the surface with carbon compounds resulted in a loss of the luminescence spectrum. eV 60 5.0 4.0 3.0 2 .o ___ -- , ’ IT---- 7 I 200 300 400 500 600 wavelength /nm cence, (6) emission excited at 230 nm, (c) emission excited at 274 nm.eV FIG. 1 .--Photoluminescence spectra of MgO at 300 K. (a) Excitation spectrum of 390 nmllumines- 6.0 5.0 4 .O 3 .O 2 .o r----’i” ’ 1 I 200 300 400 500 600 wavelength/nm FIG. 2.-Photoluminescence spectra of CaO at 300 K. (a) Excitation spectrum of 405 nm lumines- cence. Emission excited at (b) 282, (c) 310 and (d) 330nm.2916 SURFACE STATES IN OXIDES Several specimens of MgO and SrO were prepared with the surface area reduced by sintering in oxygen or water vapour at various temperatures and then evacuating at 1200 K. Some scatter in intensity was observed between the different preparations, but sintering the same sample progressively in air or O2 leads to a reduction in the intensity of the emission and after long sintering periods the reduction was very marked.The luminescence of the oxides showed strong temperature dependence ; that of magnesium oxide showed an order of magnitude decrease with temperature from 150 to 400 K. Since the emission is not a single peak the data do not justify further analysis at present. eV 60 5.0 4 0 30 2 .o wavelength/nm FIG. 3.-Photoluminescence spectra of SrO at 300 K. (a) Excitation spectrum of 470 nm lumines- cence. Emission excited at (6) 280, (c) 315, ( d ) 330 and (e) 350 nm. eV 6.0 5.0 4 0 30 2.0 i 200 I_-- i I 300 LOO 500 I 6 00 wavelength /nm FIG. 4.-Photoluminescence~spectra at 300 K of BaO supported on MgO. (a) Excitation spectrum of 465 nm luminescence. Emission excited at (b) 270, (c) 330 and (d) 360 nm.S .COLUCCIA, A . M. DEANE AND A . J . TENCH 2917 Luminescent lifetimes were calculated for MgO and SrO on the assumption that the observed decay curves were the sum of several first order decay curves (table 2). The apparatus was not suitable for the determination of lifetimes $= 100 p, and a specimen of polycrystalline barium hydroxide, known to have a 2 s lifetime showed only a background of uniform luminescence over the 1 0 0 ~ s time span of the apparatus. The excitation spectrum of BaO coated MgO [fig. 4(43 showed a band at 270 nm (4.59 eV) very similar to that of the supporting material together with a second band at 335 nm (3.70 eV) characteristic of BaO. The band at 270 nm (4.59 eV) coincides with the excitation band of MgO though comparison with the BaO powder indicates that the BaO could be responsible for some contribution.However, the emission is very different from that of MgO (fig. 1) and is attributed to the BaO on the surface. TABLE 2.-LUMINESCENCE LIFETIMES OF SURFACE STATES ON MAGNESIUM AND STRONTIUM OXIDE excitation lifetime oxide energy/eV IPS MgO 4.52 1, 7, 25 SrO 4.40 40, 100 SrO 3.94 22, 50 EFFECT OF ADSORBED GASES ON LUMINESCENCE All adsorbate gases tested (02, H2, C02, H20 and organic vapours) quenched the luminescence in the oxides in the pressure range 1 to 100 N m-2 to 1 Torr), but quenching by oxygen and hydrogen was observed at much lower pressures without significant modification of the surface. In contrast, the other gases react chemically with the surface and in addition to quenching the intrinsic surface luminescence, several adsorbates formed new luminescent species which will be considered separately.6 The reversible quenching of luminescence from MgO by oxygen has already been described 1’ but further measurements have shown that there is a small irreversible quenching at room temperature; the intensity was restored to only about 80 % of its original value after evacuation.The initial quenching by oxygen produced similar reductions of intensity in each of the excitation bands, but after exposure to high pressures of oxygen (400 N m-2, 3 Torr) at 295 K under U.V. irradiation the higher energy band was irreversibly reduced to half its original value. At 295 K quenching to half the original intensity required 0.7Nm-’ (5x Torr) of oxygen but at 550 K the same proportional reduction required 700 N m-2 (5 Ton) of oxygen while at 575 K oxygen quenching was not detected under 700 N m-2 of oxygen.The luminescent spectra from CaO and SrO were quenched by oxygen, the amount of irreversible quenching increasing along the series MgO, CaO and SrO. The effect of H2 on the luminescence of clean MgO is dependent on whether the sample is left in contact with the gas in the dark or under U.V. irradiation. In the former condition, contact with 10 Torr H2 for 10 min produces only minor changes in the original spectra. However, if H2 is admitted on MgO under U.V. irradiation the luminescence is quenched reversibly only if the gas is pumped off immediately, but irreversibly if left to stand for a few hours.During the quenching experiments small variations (about 5 nm) were noticed in the position of the excitation maximum. These changes give some indication that the excitation peak contains at least two unresolved components which are present in varying proportions. The spectrum of strontium oxide showed the variations unambiguously; a more detailed study of that oxide has been made.291 8 SURFACE STATES I N OXIDES The excitation band of strontium oxide shows considerable changes in shape when hydrogen is adsorbed (fig. 5). A more quantitative picture of this was obtained by resolving the envelope into five component gaussian bands. These bands remained unchanged in position and width during the quenching experiments ; however, their relative contribution to the spectrum changed markedly after addition of hydrogen.Those bands on the low frequency side of the envelope spectrum are preferentially eroded by hydrogen and this is irreversible at 295 K. There was no change of the luminescence band shape except a slight narrowing (105 to 90 nm) on the first hydrogen addition but a progressive shift of the band to shorter wavelengths was found as quenching proceeded. The addition of oxygen after the excess hydrogen had been pumped off caused a further reduction of intensity, and an increase of luminescence band width to its original value, but no frequency shift. After reactivation at 1200 K the luminescence band was restored close to its original value and the differences between the original clean surface and the reactivated surface were typical of the differences observed between samples prepared separately.eV 50 L O I 3 wavelength/nm FIG. 5.-Excitation spectra of SrO at 300 K. (a) SrO in vacuum, (b) SrO in contact with 1 N m-2 (8 x Torr) of Hz. In contrast to the effect of hydrogen, quenching of the luminescence from strontium oxide by oxygen at low pressure caused no changes of shape in the excitation band as the intensity decreased. The luminescent band position and shape were also unaltered, The luminescence spectrum was asymmetric along the linear wavelength abscissa and replotted on a linear energy abscissa, the luminescence excited at 3 15 nm (3.94 eV) was a single symmetrical gaussian with a small long wavelength tail which never exceeded 5 % of the height of the main band. The behaviour of this part of the spectrum when quenched showed that it was caused by a separate weak lumines- cence centred about 625 nm (1.98 eV).Because of its low intensity this second luminescence band has not been studied further. DISCUSSION All the alkaline earth oxides studied show a characteristic luminescent spectrum when excited by light in the near U.V. This is not found for the pure single crystal material but is characteristic of the relatively high surface area (20-150 m2 g-l) materials studied. Similarly, apart from BaO the first excitation levels of the bulkS . COLUCCIA, A . M. DEANE A N D A . J . TENCH 2919 crystals (table 3) are at much higher energy than the observed excitation bands of the luminescent spectra.These factors lead us to suppose that the observed luminescence and corresponding excitation spectra are associated with energy levels differing from those present in the bulk material and of major significance in the small crystallites because of the large ratio of surface to bulk ions. It has been previously argued for MgO that these new levels do not result from impurities but are associated with the reduced coordination of ions at the surface; the similarity in behaviour of the other alkaline earth oxides supports this picture. This general model is confirmed by the sensitivity of the luminescence to a wide range of gases. The quenching induced by these gases is reversible in some cases but in others a reaction takes place with the surface and the luminescence is destroyed irreversibly at that temperature, indicating the enhanced reactivity of those ions with a reduced coordination.The luminescence process consists of two parts, the initial energy absorption and the subsequent emission of light ; both of these need to be considered in some detail before going on to develop a model. The excitation spectra observed for the crystallites in vacuo are essentially identical with the surface absorption bands reported by Zecchina and Stone 3 9 (table 1) although the latter measurements were made under conditions where the luminescence had been quenched by an oxygen overpressure of 133 N m-2 (1 Torr). The absence of any significant difference between the absorption and excitation bands indicates that the adsorption of sufficient oxygen to quench the luminescence does not modify the levels responsible for the absorption of light.It is clear from the reflectance data that at higher pressures, particularly in the case of BaO, modification of the surface energy levels does occur but this appears to be associated with higher surface coverages where strong adsorp- tion occurs. The evidence indicates that absorption and emission of light may take place at different sites on the surface and it is particularly interesting to look at the effect of hydrogen on the luminescence. Quenching of the emission was found for several of the oxides; the detailed study of SrO shows clearly that the excitation band was changed in shape (fig. 5) for a sample which was partially quenched.Hydrogen reacts at different rates with species absorbing in different parts of the excitation band causing the observed change of shape. No corresponding change of shape was found in the luminescence band as it decreased in intensity. Quenching by oxygen caused no change of shape in either the excitation or the emission bands, only a decrease in intensity. Since we might expect oxygen and hydrogen to adsorb on different sites it seems likely that the absorption and emission processes occur at different, but probably closely related, sites on the surface and that energy can be transferred along the surface. NATURE OF EXCITATION A N D EMISSION SITES It is convenient to use a chemical approach to try to define the sites on the surface. In general we associate oxygen adsorption with a part of the surface which is oxygen deficient or has a local excess of cations.Although adsorption studies do not give a picture of the surface on the atomic scale, studies using other techniques, such as electron spin resonance,' show that adsorbed forms of oxygen such as 0; are associated with a nearby cation on the surface and a hyperfine interaction is frequently observed if the cation has a non-zero nuclear spin. The position is less clear for hydrogen, in some situations it is known to react with the cationY8* e.g., ZnO to form a hydride but in general we may expect hydrogen to react with that part of the surface which is cation deficient or has a local excess of oxygen ions. For SrO most of the loss of intensity of the emission is not easily reversible and the hydrogen must2920 SURFACE STATES I N OXIDES be strongly held on the surface possibly as a hydroxyl.Based on the arguments above we suggest that the sites on the surface responsible for emission are those that are oxygen deficient or have a local excess of cations, whereas the sites responsible for the absorption of light are cation deficient or have a local excess of oxygen ions. Such sites will represent situations where the coordination of the ion is markedly lower than found in the bulk crystal. These ideas enable us to examine a range of possibilities in more detail. From a comparison of the excitation spectra of the oxides it is clear that the excitation energy decreases with increasing cation radius from Mg2+ to Ba2+ ; these spectra have been shown to arise from energy absorption at the surface (where the surface is defined as those atomic layers able to react rapidly with an adsorbate). An intrinsic bulk absorption in the U.V.can arise at energies just less than the band gap from a charge transfer process to give excitons (electron-hole pairs) which can be considered as M+O-. In the surface region, it seems likely that less energy would be required for the charge transfer process because of the reduced Madelung energy at the surface. It is possible to make a qualitative calculation of the intrinsic surface state energies using the approach of Levine and Mark lo where the surface ions are considered to be equivalent to bulk ions except for their reduced Madelung constant. The band gap ratio 8 of surface (&) to bulk (&) given by where y = c,/cb is the ratio of surface to bulk Madelung constant, and where I is the ionization potential (in eV) of the cations, A the electron affinity (in eV) of the anion, c b the bulk Madelung constant, r the lattice parameter and 2 the valence of the ions.Values for y and p are available lo and since the alkaline earth oxides will not be 100 % ionic the calculation has been carried out for both 2 = 1 and 2. Calculated values for the surface band gaps on the surface (100) plane and on higher index planes are shown in table 3. = (Y -PM1 - P h p = 0.0347 r(1-A) CbZ TABLE 3.-ALKALINE EARTH OXIDE SURFACE BAND GAPS (ev) CALCULATED FROM THE MADELUNG POTENTIAL lo CaO SrO BaO surface Eb = 8.7 11, 12 Eb = 7.712, 13 Eb = 6.7 12 E b = 4.4 1 3 MgO planes coordination Z = 1 Z = 2 z-1 z=2 z = l 2-2 z = l z=2 5 8.25 8.00 7.32 7.08 6.37 6.15 4.19 4.03 4 7.11 6.26 6.37 5.54 5.57 4.79 3.66 3.11 4 6.11 4.70 5.51 4.15 4.84 3.56 3.19 2.29 3 4.18 1.74 3.90 1.54 3.47 1.23 2.30 0.73 (100) (110) (210) (211) Eb denotes the bulk band gap This theoretical approach can be considered only as an approximation ; however, it does reproduce the main features of the results and it predicts a decrease in excitation energy through the series MgO to BaO for the various index planes.The energies for the (100) surfaces are only slightly shifted from the bulk while those for the higher index planes are much closer to the observed data. However, the energies of the surface excitons would be expected to be slightly lower than the calculated band gaps.On an atomic scale this means that ions whose coordination number is t 5 must be involved in the photoluminescence process, i.e., edge and corner sites on the surface. This appears to be an appropriate model since there is no evidence to suggest anyS. COLUCCIA, A. M. DEANE AND A. J . TENCH 292 1 non-stoichiometry of the oxide and samples prepared in vacuo will adsorb both hydrogen and oxygen. In the charge transfer model it seems most plausible that an electron is transferred from an anion in a position of low coordination to a cation. Independent evidence for the existence of anions on the surface which can easily lose an electron comes from the studies on the interaction of electron acceptors with the surface where it has been shown that nitrobenzene or related molecules 14* l5 can abstract electrons from -0.1 % of the surface ions.It appears probable that some oxygen anions in sites of very low coordination such as corner or edge positions are responsible for both phenomena in oxides that have been evacuated at high temperatures. This is in agreement with the ideas proposed by Cordischi et all6 suggesting that on MgO outgassed at 1200 K the predominant surface centres are coordinatively unsaturated 02- ions. Other sites appear to make only a minor contribution to the luminescence spectrum. The excitation spectra are characterised by two main bands; this is taken as strong evidence that there are two sites on the surface which absorb at different energies and which correspond to sites of different coordination number.The lower energies will correspond to the lower coordination numbers (table 3). In addition to the main doublet structure there is evidence that both bands are complex. For example the more detailed analysis of the excitation spectrum for SrO shows several components. These components are most likely to result from minor differences in the environment of the ions such as changes in the next nearest neighbour ions. It has already been argued that luminescence occurs at a different site, which is reactive towards oxygen, and hence at an exposed cation. This luminescence arises from a process such as, M+O- + M2+02-+hv at an exposed metal ion site, and the luminescence energy decreases in the expected way through the oxide series (table 1).We can understand the processes occurring at the surface in terms of the formation of an exciton by adsorption of light at a cation-anion pair involving an oxygen ion of low coordination. This exciton, after moving over the surface in a random fashion, can decay at a cation-anion pair which involves a cation site of low coordination, with the emission of light. The excitation and emission sites may be close together but the lifetime is sufficient for the exciton to have moved some distance over the surface. This picture is consistent with the way in which hydrogen and oxygen affect the two sites independently. We only observe the radiative pathways but many of the excitons will decay via non-radiative processes. The measured lifetimes of the luminescent states to s, table 2) are relatively long and probably indicate triplet state formation but the data so far recorded do not justify detailed analysis.The curves have been interpreted as several different first order lifetimes arising from different luminescent sites on each oxide, consistent with the other experimental evidence. However it is possible that the lifetime is much longer than s and that the decay curve reflects the non- radiative decay through a higher order process. The large temperature dependence shows that non-radiative processes are dominant at room temperature. The sites at which absorption and luminescence occur cannot be defined un- ambiguously, although the earlier arguments lead to the conclusion that they must be associated with anions and cations respectively, in positions of low coordination on the surface.For the absorption process such sites could be associated with oxygen ions on edge or corner sites of the crystal, cation vacancies or possibly where cations are missing from an edge or corner site and oxygen ions are exposed. E.s.r. data2 922 SURFACE STATES IN OXIDES indicate that holes trapped at cation vacancies can be produced by U.V. irradiation and these holes are destroyed by hydrogen. However, the trapped holes absorb at much longer wavelengths.l* The sites linked to the emission could be surface anion vacancies, similar point defects or cations at step and corner sites where the coordina- tion is less than 5. The bulk anion vacancy with an electron (Ff centre) in MgO absorbs at 250 nm (5.0 eV) and emits at 396 nm (3.13 eV) and for CaO it is 340 nm (3.65 eV) and 370 nm (3.35 eV) respectively.lg* 2o Even after allowing for modified values at the surface, these do not agree well with the observed values for the powders and there is no indication of any significant formation of trapped electron centres from the e.s.r.data. There is, however, an interesting correspondence between the broad luminescence observed at slip planes in the bulk and that from the surface. Luminescence from deformed crystals 21 9 22 of magnesium, calcium and strontium oxides gave excitation peaks of about 0.2 eV above those reported here (table 1) and emission about 0.4 eV lower than the values for the surface luminescence. Impurity ions did not contribute and the luminescence was thought to be associated with vacancy clusters; such a cluster can be regarded as a small hole in the bulk crystal where surface luminescence occurs.We would like to thank Miss K. Mahmood and Miss J. Abbott, who carried out some of the measurements described ; S. Coluccia and A. J. Tench acknowledge financial support from NATO. A. J. Tench and G. T. Pott, Chern. Phys. Letters, 1974,26,590. S. Coluccia, A. M. Deane and A. J. Tench, Proc. 6th Int. Congress on Catalysis, London, 1976, 1, 171. A. Zecchina, M. G. Lofthouse and F. S. Stone, J.C.S. Furaday I, 1975,71, 1476. A. Zecchina and F. S. Stone, J.C.S. Faraduy I, 1976,72,2364. P. B. Merkel and W. H. Hamill, J. Chem. Phys., 1971,55,2174. S. Coluccia, A. M. Deane and A. J. Tench, to be published. J. H. Lunsford, Catalysis Rev., 1973, 8, 135. A. L. Dent and R. J. Kokes, J. Phys. Chern., 1969,73,3772. lo J. D. Levine and P. Mark, Phys. Rev., 1966,144,751. l 1 G. H. Reiling and E. B. Hensley, Phys. Rev., 1958,112, 1106. l 2 V. I. Neeley, Ph.D. Tliesis (University of Oregon, Eugene, 1963). l3 H. H. Glascock and E. B. Hensley, Phys. Rev., 1963, 131, 649. l4 A. J. Tench and R. L. Nelson, Trans. Faraday Soc., 1967, 63,2254. l 5 M. Che, C. Naccache and B. Imelik, J. Catalysis, 1972, 24, 328. l6 D. Cordischi, V. Indovina and M. Occhiuzzi, J.C.S. Furaday I, 1978, 74,456. l 7 A. J. Tench and J. F. J. Kibblewhite, J.C.S. Chem. Comm., 1973, 955. l9 G. P. Pells and A. E. Hughes, AERE R8686,1977. 2o A. E. Hughes and B. Henderson, Point Defects in Solids, ed. J. H. Crawford and L. M. Slifkin 21 T. J. Turner, N. N. Isenhower and P. K. Tse, Solid State Comm., 1969, 7, 1661. 22 Y. Chen, M. M. Abraham, T. J. Turner and C. M. Nelson, Phil. Mag., 1975, 32,99. * R. P. Eischens, W. A. Pliskin and M. J. D. Low, J. Catalysis, 1962, 1, 180. Y. Chen and W. A. Sibley, Phys. Rev., 1967, 154,842. (Plenum Press, N.Y., 1972), vol. 1, chap. 7. (PAPER 8/896)