J. Chem. SOC., Faraday Trans. I , 1989, 85(2), 237-249 Structure and Reactivity of Zinc-Chromium Mixed Oxides Part 3.-The Surface Interaction with Carbon Monoxide Elio Giamello,* Bice Fubini and Massimo Bertoldi Dipartimento di Chimica Inorganica, Chimica Fisica e Chimica dei Materiali, Universita di Torino, Via P. Giuria 9, 10125 Torino, Italy Guido Busca Istituto di Chimica, Facolta di Ingegneria, Universita di Genova, Fiera del Mare, 16129 Genova, Italy Angelo Vaccari Dipartimento di Chimica Industriale e dei Materiali, Universita di Bologna, Viale del Risorgimento 4, 40136 Bologna, Italy The reactivity towards carbon monoxide of a series of Zn-Cr non- stoichiometric spinels (NSS) with an excess of zinc, active in the synthesis of methanol and higher alcohols, has been investigated by infrared spectro- scopy and adsorption microcalorimetry.A fraction of the total CO taken up is irreversibly oxidized to surface carbonates by the chromate groups pres- ent. A second fraction is strongly but reversibly coordinated in the form of a 'carbon-down' CO complex, the heat of adsorption being between 80 and 50 kJ mol-' and the C-0 stretching frequency in the range 2192-2205 cm-'. The strongest active sites for CO coordination are surface Cr3+ ions acting as a-acceptors, but also capable of d-n backdonation. The number of active sites for CO coordination is very low in the stoichiometric spinel ZnCr,O, (inactive in methanol synthesis), whereas it grows in the zinc-excess non- stoichiometric spinels owing to modifications of the ' collective ' properties of the solid introduced by the presence of the excess zinc into the spinel lattice.Zn-Cr mixed-oxide systems have been widely employed in the past in the catalytic synthesis of methanol from carbon monoxide and hydrogen,' and have recently found a new interest owing to the possibility of syngas conversion to mixtures of methanol and higher alcohols.2 In spite of the extensive utilization of this catalyst, information concerning the nature of its surface and the mechanisms of CO and H, activation and of the catalytic reaction is still lacking, owing to a lack of detailed surface characterization. We have recently carried out a systematic characterization of a family of well defined zinc-chromium mixed oxides having different Zn/Cr ratios. All the members of this series of compounds are monophasic in the composition range 33 : 67 < Zn : Cr < 50 : 50, and can be regarded as zinc-excess non-stoichiometric spinels (NSS) with an excess of Zn2+ ions located at B-type sites of the AB,O, spinel lattice (with octahedral Above the 50: 50 ratio the systems are biphasic, with a separate ZnO phase in addition to the ' spinel-like' one.Differences in surface properties and methanol decomposition between stoichiometric and non-stoichiometric spinels have been reported in previous paper^.^,^ The present one is devoted to a detailed investigation of the surface activity towards carbon monoxide by adsorption microcalorimetry and Fourier- transform infrared spectroscopy. As the CO molecule is directly involved in syngas catalytic reactions, the investigation of its interaction with the catalyst surface can afford useful information on the nature and number of active sites : 9 237 FAR 1238 Reactivity of Zn-Cr Mixed Oxides Table 1.Bulk and surface composition of some C-Zn-Cr samples Zn : Cr (bulk) Zn : Cr (surface) C r * / [ C r I + C r I 'I] 33 : 67 38:62 44: 56 50:50 22: 78 24: 76 41 : 59 47: 53 0.33 0.33 0.28 0.32 this approach has been adopted previously in the characterization of other systems active in methanol ~ynthesis.~-~ Experimental The Zn-Cr catalysts were prepared by coprecipitation from solutions containing Zn and Cr nitrates followed by successive calcination. The preparation procedure, sample composition and characterization have previously been described in detail.,, The samples investigated, hereafter denoted by their Zn/Cr ratio, are Zn : Cr = 33 : 67 (stoichiometric spinel, ZnCr,O,), Zn : Cr = 38 : 62, Zn : Cr = 44: 56 and Zn :Cr = 50 : 50 (all monophasic), and Zn : Cr = 75 : 25 (biphasic : ZnO + NSS phase). A sample of pure chromia was also employed.The catalysts were studied in their calcined and reduced states. The calcined catalysts (labelled C) were outgassed at 573 K for 1 h in order to eliminate surface poisoning due to atmospheric agents. The reduced catalysts (denoted R) were treated with 100 Torr CO (1 Torr z 133.3 Pa) for 20 h at 373 K and then outgassed at 573 K. Reduction with CO eliminated the surface chromates always present at the surface after calcination. X.P.S. experiments to measure the surface composition were performed in a Perkin Elmer PHI 5400 ESCA system at a pressure < 1 x Pa.All samples were measured with the same instrumental settings, and the sensitivity factors were previously determined with the same instrument on samples of single elements. The values employed were Zn(2p3/,) = 4.80 and Cr(2p) = 2.30 [relative to the F( 1 s) = 1 .OO]. In the case of Cr the area under both signals, 2p,/, and 2p,,, was considered taking into account the non-complete separation between the two signals. The infrared spectra were recorded at room temperature and 173 K using a Nicolet MXI Fourier-transform spectrometer. The samples were pressed in self-supporting discs and placed in a cell with KBr windows connected to a greaseless vacuum line. The heats of adsorption of CO were measured by means of a Tian-Calvet microcalorimeter connected to a volumetric apparatus, adsorbing small successive doses of the gas.Each adsorption run was followed by stepwise desorption.1° For each dose the amount adsorbed (or desorbed) and the related heat were measured, except in the case of the last dose of desorption for which, being obtained by direct evacuation, only the heat of adsorption is measurable. Results Surface Composition In table 1 the compositions in atomic percentage of both the bulk and the surface are reported for some Zn-Cr samples, as obtained by photoelectron spectroscopy. In no case was zinc enrichment of the surface observed : it is therefore possible to exclude the presence at the surface of amorphous or highly divided ZnO phases not detectable by X-ray diffraction.This hypothesis was advanced (inter alia) in a previous paper5 to explain the differences in surface behaviour between the stoichiometric and the zinc- excess, non-stoichiometric spinels.E. Giamello et al. 239 0 20 40 60 90 0.40 0.60 O. *O n, l-1 m-2 0 Fig. 1. CO adsorption onto a C-Zn : Cr = 75 : 25 sample. (a) Calorimetric isotherms (Qint us. p ) for three successive runs. Open symbols: adsorption; full symbols: desorption; 0 @, run I ; a#, run 11; H W , run 111. (b) Partial molar heat of adsorption us. amount adsorbed for the three runs reported in (a). We have also performed a rough determination of the Crv'/[Crvl + CrIII] surface ratio (on the basis of the areas of the 576.6 and 579.6 eV signals) by means of a peak-fitting program.For all samples we found values near 0.3 (table 1,3rd column) similar to those obtained by chemical analysis and previously reported in ref. (3). The Interaction of CO with C Samples In fig. 1 are reported the calorimetric isotherms (evolved heat, Qint, us. pressure,p) and the corresponding partial molar heats of adsorption (evolved heat per amount adsorbed for each dose, AQint/An, vs. amount adsorbed, n,) for three adsorption- desorption runs for CO onto the sample with C-Zn:Cr = 75:25. Run I1 was performed immediately after 9-2240 Reactivity of Zn-Cr Mixed Oxides 1800 1600 1400 1200 1000 800 vim-' Fig. 2. Infrared spectra of CO interaction onto C-Zn: Cr = 50: 50. Full line: background; dashed line: 100 Torr CO 2 min after immission.Dotted line: 100 Torr CO 30 min after immission. Dotkdash line: effect of heating 30 min at 473 K under 100 Torr CO. run I, while run I11 was obtained after a period of 20 days, during which the sample was kept under a pressure of 200 Torr CO at room temperature. The samples with different Zn : Cr ratios behave in a similar way, except for the values of the total amount adsorbed, which are different in each case. The three isotherms in fig. 1 (a) are not coincident, and a noticeable decrease in the heat evolved is observed on passing from run I to run 111; moreover, in the case of runs I and I1 a consistent fraction of the heat evolved is not recovered upon direct evacuation, indicating that a fraction of the adsorbate is irreversibly held at the surface.In the case of run 111, where adsorption and desorption are coincident, no irreversible fraction is observed. The partial molar heat curves [fig. 1 (b)] show that the heat of adsorption decreases with coverage in all cases, and the initial values in the three runs decreases from 230 kJ mol-' in run I to 70 kJ mol-' in run 111. The shape of the calorimetric peaks indicates the presence of slow adsorptive phenomena in the case of runs I and 11, whereas in the case of run I11 the adsorption is practically instantaneous. The experiment described above clearly indicates that the centres able to cause irreversible adsorption are progressively eliminated by prolonged contact with CO : after saturation in CO for 20 days the residual adsorption (fast, reversible, with a differential heat < 80 kJ mol-l) is exclusively related to the mere coordination of CO onto surface- unsaturated cations.The high value of the heat of interaction, as well as the slow kinetics in the irreversible process, indicate that a surface reaction occurs which is probably the oxidation of CO to surface carbonates. This is confirmed by infrared spectroscopy: fig. 2 reports the i.r. spectra recorded on a C-Zn:Cr = 50:50 sample in the region between 800 and 1800 cm-' recorded, respectively, after outgassing at 573 K (full line,E. Giamello et al. 0.004 24 1 I I I 0 0 .016 .012 .008 background), under 100 Torr CO 2 min after immission (dashed line), 30 min after immission (dotted line), after heating at 473 K in CO for 30 min (dotteddashed line). The narrow band at 1OOOcm-' and the wider one at 950cm-', already assigned to chromate~,~ are slightly eroded by CO at room temperature, with a parallel increase of absorption in the region between 1200 and 1600 cm-', indicating the formation of various types of surface carbonates.When the temperature of the system is raised to 473 K the decrease in the chromate bands and the increase in the carbonate bands become dramatic. The admission of CO also causes the appearance of a sharp band in the region typical of vco of metal carbonyl whose features will be discussed in the following section. The Interaction of CO with R Samples In fig. 3 are reported, as calorimetric isotherms, three adsorption-desorption cycles of CO successively performed on the surface of an R-Zn:Cr = 75:25 sample.The differences between the three isotherms [note that the ordinate scale in fig. 3 is expanded by comparison with fig. 1 (a)] are very small by comparison with those on the C sample [fig. 1 (a)], and a minor fraction of irreversibly adsorbed CO is observed in the first and second cycle only, due to residual CO oxidation. The third adsorption isotherm in fig. 3 (111) is completely reversible and almost coincides with the third one in fig. l(a), where prolonged contact with CO ensures the complete exhaustion of the oxidation process: the two treatments are therefore equivalent, and the reduction by CO at 373 K can thus be adopted as the standard method for eliminating the superimposition of surface oxidation on reversible adsorption of CO.This latter process on the whole series of R samples is, in all cases, fast and non-activated. The partial molar heat, corresponding to isotherm I11 in fig. 3, is reported in fig. 4, The heat of adsorption decreases with coverage from an initial value at ca. 80 kJ mol-' to ca. 45 kJ mol-'. The heat of desorption, in the more limited range of coverage in which it can be measured,242 80 - i 2- 40- 2 I 1 I I - c. I C J C . I I I 1 I 0 0.1 0.2 0.3 0 . 4 0.5 0.6 'E 3 =t 2 0.4 0.2 0 20 40 pmorr 60 Fig. 5. Volumetric isotherms (n, us. p ) for CO adsorption onto the whole series of stoichiometric and non-stoichiometric R-samples. ., Cr,O,; 0, ZnCr,O,; 0, Zn:Cr = 38:62; A, Zn:Cr = 44:56; 0, Zn:Cr = 50:50; V, Zn:Cr = 75:25.E. Giamello et al. 243 I I 1 I I 0 0.2 0.4 0.6 0.8 n, /pmol m-' 0 Fig.6. Partial molar heat of adsorption for CO onto the whole series of samples. Symbols as in fig. 5 . fits the trend shown by the heat of adsorption, indicating that the evacuation of surface sites exactly follows the inverse order of filling. The adsorption-desorption cycles described above have been performed for all the R samples of the series, and for a sample of pure chromia. Significant differences in adsorption capacity are found for the various samples. Fig. 5 reports the volumetric isotherms for the whole series of samples considered in the present paper. The corresponding heats of adsorption are reported in fig. 6 as a function of coverage, and exhibit a similar trend to that shown in fig. 4. The variations in adsorptive capacity (fig.5 and 6) do not involve different kinds of adsorption sites., but rather a different site energy distribution : in particular, inspection of fig. 6 indicates that the main differences among the various samples is determined by the different numbers of sites having heats of adsorption > 60 kJ mol-'. The six samples reported in fig. 6 can be easily divided into three groups: the stoichiometric oxides (Cr,O, and ZnCr,O,) which adsorb ca. 0.1 pmol CO m-,, the monophasic non-stoichiometric spinels on which the total adsorption is much higher, ranging between 0.6 and 0.8 pmol rn-, and the biphasic one, with an intermediate value. The non-stoichiometry of the solid is therefore related to a considerable improvement in the chemisorptive capacity. The amount of CO adsorbed by the various non-stoichiometric samples decreases with an increase in the zinc excess present in the solid : maximum adsorption is observed for the sample at the beginning of its departure from stoichiometry (Zn: Cr = 38 : 62).5 The total chemisorptive capacities of the various samples examined (taken at 60 Torr equilibrium pressure), as a function of composition, are reported in fig.7 (upper curve). The value reported for ZnO is from ref. (7). The lower curve in fig. 7 reports the amounts of CO adsorbed with heats of adsorption > 60 kJ mol-', as deduced from fig. 6. As the lower curve follows a trend nearly parallel to the upper one it can be confirmed that the pronounced differences in the CO adsorption capacity between the various samples must be ascribed to differences in high-energy sites, which are completely absent in the pure Zn0.7 When a prereduced sample (R) is put in contact with CO the only modification observed in the i.r.spectrum is the formation of an absorption band in the region of the244 Reactivity of Zn-Cr Mixed Oxides 1 Cr (atom %) 100 80 60 40 20 0 Zn (atom %) Fig. 7. CO chemisorptive capacity as a function of sample composition. Open symbols: total adsorption at 60 Torr. Full symbols: adsorption with qdiff 2 60 kJ mol-'. Table 2. Infrared frequencies for CO adsorption onto the whole series of R samples sample-Zn : Cr v/cm-' (298 K) vlcm-l(l70 K) 33:67 - 2170 38:62 2202 (ii) 2195 (i)-2205 (ii) 44: 66 2195 (it2202 (sh) 2195 (i)-2205 (sh) 50: 50 2192 2192 75 : 25 2192 2192 metal carbonyl, also present on C samples. Such bands have been observed at room temperature on all samples except the stoichiometric spinel (Zn:Cr = 33:67).If the spectra are carried out at 170 K, even in the case of ZnCr,O,, a small i.r. band appears, and on some of the non-stoichiometric samples a better resolution of the band in two component is visible. The i.r. frequencies recorded for the adsorption of CO at room temperature and at 170 K on the various samples are reported in table 2, and the corresponding i.r. spectra at 170 K are shown in fig. 8. The CO pressure was kept at 100 Torr in all cases. Three main facts can be outlined by inspection of the i.r. results : (i) the intensities of the spectra roughly follow the same order observed for the amounts adsorbed and reported in fig.6; (ii) the stoichiometric ZnCr,O, spinel exhibits at low temperatures a CO stretching band at 2170 cm-', whereas all the bands related to the non-stoichiometric spinels are at frequencies > 20 cm-I higher; (iii) in the case of the two samples exhibiting the highest adsorption capacities (Zn : Cr = 38 : 62 and Zn : Cr = 44 : 56) the low-temperature spectra reveal the presence of two distinct bands in the carbonyl region.E. Giamello et al. 245 1 - 100 2200 2100 ! 3 I 5 300 2200 2100 2300 2200 2100 2300 2200 2100 2300 2200 2100 vim-' Fig. 8. Infrared spectra in the carbonylic region of adsorbed CO. T = 170 K. (1) Zn : Cr = 33 : 67, (2) Zn:Cr = 38:62, (3) Zn:Cr = 44:56, (4) Zn:Cr = 50:50, ( 5 ) Zn:Cr = 75:25. Discussion Nature of the Interaction of CO with the Zn/Cr Samples The investigation of CO adsorption on calcined samples (C) has identified the presence of two types of interactions with the surface, i.e.redox reactions, with the consequent formation of carbonates, and CO coordination. The former phenomenon is due to the presence of chromates at the surface of the samples, easily observed by i.r. and X.p. spectroscopy. The reduction of the chromates by CO (very slow at room temperature) is highly exothermic, but involves a relatively low number of CO molecules. In the case of C-Zn : Cr = 75 : 25 for instance, the number of reactive chromate centres present at the surface is CQ. 25% of the total number of sites reacting with CO. The reversible adsorption of CO on NSS is characterized by heterogeneity in the adsorption sites.Inspection of the differential heat plot in fig. 6 suggests the presence of two distinct families of sites: the first, with a high differential heat (8MO kJ mol-'), is more abundant than the second, with a heat of adsorption of ca. 50 kJ mol-l. This interpretation is supported by i.r. data which show, at least in the case of R-Zn : Cr = 38 : 62 and R-Zn : Cr = 44: 56, the presence of two distinct absorptions (table 2 and fig. 8). The adsorption of CO at the surface of the non-stoichiometric spinels is thus characterized by a family of active sites capable of a strong but still reversible interaction with the CO molecule, having an energy of interaction in the range 8MO kJ mol-'. A frequency value higher than that of the free molecule indicates that the chemical bonding is mainly based on the interaction between the 50 orbital of CO and an acceptor site.This entails a net increase of the bond order between C and 0. Nature and Number of Active Sites Stoic h iome t r ic So lids The Cr3+ ions are thought to be present in octahedral coordination in both stoichiometric and Zn-excess they are also present in the same coordination in Cr,O,, which, however, has the corundum structure. The pure chromia has a chemisorptive capacity similar to that of ZnCr,O, (at least with the pretreatment adopted in the present work) in the adsorption of CO at room temperature (fig. 5). Infrared work carried out on a- chromia microcrystals" has shown that the frequency of adsorbed CO is 2185 cm-'246 Reactivity of Zn-Cr Mixed Oxides Fig.9. Schematic representation of an ideal (1 00) face of the ZnCr,O, spinel. The lattice parameter is ( a ) Taking as a reference the plane of external ions (Cr3+, 02-: open symbols) the other ions lie below this plane at distances of a/8 (Zn2+, full symbols) and a / 4 (02-, dashed symbols), respectively . (0 = 0), which shifts at high coverage to 2170 cm-l. This latter value is exactly the same as that measured by us on the zinc chromite at high coverage ( p = 100 Torr, table 1). The carbon monoxide adsorption on chromia has been assigned to the coordination onto coordinatively insaturated Cr3+ octahedral ions at the (0001) faces.ll The fact that ZnCr204 and Cr203 show similar trends in the differential heat of CO adsorption and exactly the same infrared frequency at high coverage, suggest that, on these two solids, the active sites for CO adsorption are the same, i.e.the coordinatively unsaturated octahedral Cr3+ ions. The absence of adsorption of CO onto zinc ions in the stoichiometric spinel can be explained by taking into account that, in AB204, the (100) and (1 11) are the most exposed crystal faces.12 In the former case the surface-exposed ions are exclusively Cr3+ and 02-, whereas the Zn2+ ions lie under the first surface layer with a complete tetrahedral coordination sphere (fig. 9). With regard to the ( 1 11) faces, it has been clarified12 that, although they could in principle assume two distinct configurations, only the one exposing octahedral sites is present at the surface.The previous description agrees with the very low value of Zn : Cr found by X.P.S. surface analysis in the case of the zinc chromite (table 1) and also with the lack of detection, on the same solid, of zinc hydrides observed by some of us on the non-stoichiometric solids. l 3 Non-stoichiometric Spinels In these solids the chromium ions are expected to maintain the same octahedral coordination typical of ZnCr,O,, but the presence of an excess of Zn2+ ions also in octahedral coordination4* implies that each Cr3' is surrounded by a 'mixed ' cationic sphere (at 2/2a/4) composed of both chromium and zinc ions. At the NSS surface, therefore, two kinds of coordinatively unsaturated ions of octahedral origin are exposed : Zn2+ ions and Cr3' ions 'modified ' by the presence of the former. The unambiguous identification of the nature of the site adsorbing CO at the surface of NSS, on the basis of the i.r.frequency only, is practically impossible because the observed frequency values are in agreement with those reported in the literature forE. Giamello et al. 247 adsorption either on Zn2+ or Cr3+ centres."' 14, l5 Some hypothesis can thus be made only considering the coupling between spectroscopic and energetic data. If CO were adsorbed onto Zn2+ the simple model of o-donation would apply for the chemical bond. In this model the CO stretching frequency and the heat of adsorption must both increase with an increase in the electrostatic field of the cation, which is related to the degree of coordinative unsaturation of the cation itself.The existence of a parabolic correlation law between the heat of adsorption and frequency has recently been proposed on the basis of the results obtained for various solids containing 'non-d' cations.l6 From an inspection of the correlation curve one can deduce that in the case of the Zn2+ cations of unreduced ZnO a frequency of ca. 2188 cm-' is associated with a heat of adsorption of 45 kJ mol-', while frequencies in the range 2192-2205 cm-l, like those monitored in the case of non-stoichiometric spinels, correspond to an energy in the range 50-55 kJ mol-'. If the zinc ions coordinate CO on non-stoichiometric spinels, they should have a higher coordinative insaturation than in ZnO. A typical example of highly exposed Zn2+ is that of the Zn2+ partially exchanged Y zeolites : in this system Angel1 and Shaffer found a stretching frequency for adsorbed CO at 2214 cm-', which was unambiguously assigned to CO adsorbed on highly uncoordinated zinc ions typical of the zeolite framework." As no energy data were available for this system, an ad hoc experiment was performed to measure the heat of adsorption of CO and the related frequency on a partially exchanged Zn/Y zeolite.The results will be reported elsewhere in detail;'* they confirm the frequency value from ref. (1 7) (2214 cm-l) and indicate a heat of interaction, due to CO coordination onto zinc ions, of 58 kJ mol-l, i.e. lower than that measured on NSS. These two values fall very close to the correlation curve in ref. (16). The hypothesis whereby zinc cations are the strongest active sites in CO adsorption at the NSS surface must therefore be disregarded.To explain the high heat of adsorption observed in correspondence with frequencies around 2200 cm-', it is necessary to consider the role of chromium ions as adsorption centres. It is generally accepted that the interaction of CO with a transition-metal ion bearing a high electric charge, like Cr3+, is based on simple o donation typical of those ions, without the possibility of d-n back-bonding.lg However, in a recent paper by Zecchina and co-workers this problem was reinvestigated on the basis of the results obtained for CO adsorption on pure a-chromia." The authors observed relevant frequency shifts of the CO stretching band of both the static and dynamic types, indicating, respectively, the existence of inductive effects and vibrational couplings in the CO/Cr203 system.This behaviour can be explained by a bond model involving in addition to a strong interaction between the 50 orbital and a suitable d orbital, a non- negligible contribution by d-n* delocalization. l1 The frequency value at zero coverage found on Cr20, (2185 cm-') is therefore determined by the balance of two opposite tendencies: a strong c interaction, which tends to increase the C-0 frequency, and back-bonding, which contributes to its reduction. The correlation between heat and frequency, typical of simple o coordination, does not apply in this model, and a lowering of the zero-coverage frequency by the n contribution corresponds to a higher heat of adsorption, as the latter value monitors the strength of the whole interaction.Therefore the heat and frequency values cannot fall on the correlation curve reported in ref. (16), which is typical of pure o coordination. In the case of NSS we thus assign the extended CO adsorption with a heat of interaction between 80 and 60 kJ mol-1 to Cr3+ ions present at the surface of the spinel-like system, thus modifying our previous a~signment.~ The heat of adsorption has the same initial value as that observed for stoichiometric oxides (fig. 6), indicating that the nature of the interaction is basically the same in the two cases and involves both a-bonding and n-backbonding. The assignment of the main fraction of CO adsorption to the surface chromium ions present on NSS is also in agreement with the data for chemisorptive capacity (fig.7), indicating that the adsorption of CO decreases on increasing the zinc content of the samples. For the248 Reactivity of Zn-Cr Mixed Oxides second, less abundant, family of sites exhibiting a heat of interaction < 60 kJ mol-1 and observed at the surface of non-stoichiometric spinels only, two hypotheses can be proposed. In the first the active sites are still Cr3+ ions in a different environment from those constituting the main fraction of adsorption sites : this is the case, for instance, of the ions located on a different crystal plane. In the second hypothesis the sites are constituted by surface Zn2+ ions: the frequency values corresponding to a heat of 50-55 kJ mol-' are expected to fall around 2200 cm-l,18 on the basis of the v us.AH relationship. l6 Finally, the remarkable increases observed on passing from ZnCr,O, to NSS, both in the amount and in the stretching frequency of the adsorbed CO, must be discussed. First of all, as already mentioned, the adsorbing Cr3+ ions in NSS are different from those of zinc chromite because of the presence of neighbouring octahedral zinc ions which can induce, by electrostatic effects, variations of the actual cationic charge at the adsorption sites, thus influencing the adsorption behaviour. We basically think, however, that the origin of this behaviour mainly stems from the different 'collective' properties of the systems investigated. It is well known, on the one hand, that Cr,03 and ZnCr,O, show similar conductivity behaviour.'" On the other hand we have recently reported5 the following.(a) All NSS are antiferromagnetic and show a resonant line in their e.s.r. spectra at the same g value observed for ZnCr,O,. The linewidth of the e.s.r. spectrum of NSS, however, is always larger than that of zinc chromite and regularly varies with composition, indicating a progressive variation of the magnetic coupling between Cr3+ ions. (b) Non-vibrational absorption bands are observed in the i.r. region in the case of NSS, whose maxima vary slightly with composition. This behaviour indicates the presence of semiconductivity phenomena, which are likely more pronounced than in the case of ZnCr,O,. It is therefore not surprising that CO adsorption, which involves electron donation towards the bands of the could occur at different extents on NSS and ZnCr,O,.Similar considerations explain the variation of ca. 30 cm-l of the frequency values at 100 Torr between stoichiometric and non-stoichiometric oxides : the delicate balance between electron donation and back-donation, which determines the value of v, basically depends on the capacity for transmission through the solid of the adsorbate-adsorbate interaction. The latter, in its turn, must be related to the electron distribution in the bands of the solid. Conductivity measurements are needed to confirm relationships between electronic properties and surface reactivity of the solids. Conclusions The investigation of the interaction of CO with the surface of Zn-Cr non-stoichiometric spinels has provided evidence of both the oxidation of the CO molecule to surface carbonates and the reversible coordination of the molecule mainly onto surface Cr3+ ions.The latter ions are able to form stable, but reversible, complexes with CO based on the mechanism of a-donation and n-back-bonding. The zinc-excess solids are able to chemisorb CO in larger amounts than the stoichiometric zinc chromite. The total amount adsorbed is related to the chromium content and therefore decreases with increasing zinc excess. The increase of the chemisorptive capacities towards CO is basically due to the unusual collective properties of the non-stoichiometric spinels, which contain Zn2+ ions in the B positions of the lattice. The Zn2+ ions (responsible for the hydrogen dissociation on non-stoichiometric spinels13 but not for the main fraction of CO adsorption) therefore play a fundamental role in determining the general properties of these solids.Like other systems active in the synthesis of methanol from carbon monoxide and hydrogen, the activation of CO consists of the 'carbon-down' coordination of theE. Giamello et al. 249 Helpful discussions with Prof. F. Trifiro’ are gratefully acknowledged, as well as financial support by the Italian Ministry of Education (Progetti di rilevante interesse Nazionale : Gruppo Struttura e reattivita’ delle superfici). References 1 G. Natta, in Catalysis, ed. P. H. Hemmet (Reinhold, New York, 1953), vol. 111, chap. 8. 2 M. Di Conca, A. Riva, F. Trifiro’, A. Vaccari, G. Del Piero, V. Fattore and F. Pincolini, Proc. 8th Znt. 3 G. Del Piero, M. Di Conca, F. Trifiro’ and A. Vaccari, in Reactivity of Solids, ed. P. Barret and L. C. 4 G. DelPiero, F. Trifiro’ and A. Vaccari, J . Chem. Soc., Chem. Comrnun., 1984, 656. 5 M. Bertoldi, G. Busca, B. Fubini, E. Giamello, F. Trifiro’ and A. Vaccari, J. Chem. SOC., Faraday 6 F. Trifiro’, L. Mintchev, G. Busca, A. Vaccari and A. Riva, J. Chem. SOC., Faraday Trans. 1, 1988, 84, 7 E. Giamello and B. Fubini, J . Chem. Soc., Faraday Trans. 1 , 1983, 79, 1995. 8 E. Giamello, B. Fubini and P. Lauro, Appl. Catal., 1986, 21, 133. 9 E. Giamello, B. Fubini and V. Bolis, Appl. Card., 1988, 36, 287. 10 B. Fubini, Rev. Gen. Thermodyn., 1979, 297. I 1 D. Scarano, A. Zecchina and A. Reller, Surf. Sci., 1988, 198, 11. 12 J. L. Hutchison and N. A. Briscoe, Ultramicroscopy, 1985, 18, 435. 13 G. Busca and A. Vaccari, J. Catal., 1988, 108, 481. 14 A. Zecchina, S. Coluccia, E. Guglielminotti and G. Ghiotti, J. Phys. Chem., 1971, 75, 2774. 15 D. A. Seanor and C. H. Amberg, J. Chem. Phys., 1965, 70, 2967. 16 C. Morterra, E. Garrone, V. Bolis and B. Fubini, Specfrochim. Acfa, Part A , 1987, 43, 1577. 17 C. L. Angel1 and P. Shaffer, J. Phys. Chem., 1966, 70, 1413. 18 V. Bolis, B. Fubini and E. Giamello, to be published. 19 Yu. A. Lokhov and A. A. Davidov, Kinet. Katal., 1980, 21, 1523. 20 J. F. Garcia de la Banda, J. Catal., 1962, 1, 136. 21 G. L. Griffin and J. T. Yates Jr, J. Chem. Phys., 1982, 77, 3351. 22 K. Klier, Adv. Catal., 1982, 31, 243. Congr. Catal. (Dechema, Frankfurt am Main, 1984), vol. 11, p. 173. Doufur (Elsevier, Amsterdam, 1985), p. 1029. Trans. I , 1988, 84, 1405. 1423. Paper 8/00566D; Received 15th February, 1988