J . Chem. Soc., Faraday Trans. I , 1984,80, 1479-1489 Paramagnetic Metal and Oxygen Species Observed with [Co,(CO),,l and [Rh,(CO),,I Carbonyl Clusters Pyrolysed on ?-Alumina and Zirconia Supports BY TIZIANA BERINGHELLI, ANTONELLA GERVASINI, FRANCA MORAZZONI* AND DONATELLA STRUMOLO Dipartimento di Chimica Inorganica e Metallorganica, Universita di Milano, Via Venezian 21, 20133 Milano, Italia AND SECONDO MARTINENGO Centro di Studio per la Sintesi e la Struttura dei Composti dei Metalli di Transizione nei bassi stati di ossidazione, Via Venezian 21, 20133 Milano, Italia AND LUCIANO ZANDERIGHI Dipartimento di Chimica Fisica ed Elettrochimica, Universita di Milano, Via Venezian 21, 20133 Milano, Italia Received 26th July, 1983 The pyrolysis of y-Al,O,- and Zr0,-supported [Co,(CO),,] and [Rh,(CO),,] gives highly dispersed metal systems whose electronic properties have been studied by e.s.r.spectroscopy. Paramagnetic formal Rh" species (g,, = 2.21, gl = 2.11) were observed on Rh/y-Al,O, samples. The pattern of the hyperfine structure suggests that metal-metal interactions are present and that paramagnetic transition-metal carbonyl clusters can be assumed as model compounds. No paramagnetic species involving the metal were observed for either y-Al,O,- or Zr0,-supported cobalt samples. This behaviour is consistent with a higher acidic character of y-Al,O, with respect to ZrO,. On contact with 0, and CO the following paramagnetic species were observed: O;-A13+ (g,, = 2.034, g , = 2.00) over Co/y-Al,O, samples, O;-Rh"' (8, = 2.09, g, = 2.03, g , = 2.00) over Rh/y-Al,O, samples, O;-RhlI1 and O;-Zr4+ (8, = 2.03, g , = 2.00) over Rh/ZrO, samples and CO-RhII (g = 2.034) over Rh/y-Al,O, samples.The 0, and CO bond strengths are higher for the y-Al,O,-supported samples than for the Zr0,-supported samples and are tentatively related to the catalytic activity of these metal systems. In spite of the technological importance of supported transition-metal systems, there have been few extensive physico-chemical investigations on the nature of the active metal species. Specifically, whenever it was observed that the activity and selectivity depended on the properties of the support, the presence of a strong metal-support interaction was postu1ated.l Because of the dearth of experimental data it is still difficult to assess the relevance of such an effect.It is well known that supported mononuclear transition-metal complexes can give highly dispersed metal catalysts2 and that molecular spectroscopy and magnetic techniques can be helpful in characterizing the electronic structure of the surface metal centres. However, the properties of these systems are different from those of conventional metal catalysts. In recent years another technique, the decomposition of supported metal carbonyl cluster^,^ has been tested for the preparation of highly dispersed metal catalysts. It is reasonable to expect that the activity of such systems is more analogous to that of 14791480 PARAMAGNETIC METAL AND OXYGEN SPECIES conventional metal catalysts and that the electronic configuration of the small metal aggregates is more amenable to investigation by molecular physical methods. These reasons, together with the increasing interest being shown in catalytic processes assisted by transition-metal clusters4 and the lack of extensive physico-chemical characterization of such catalysts, led us to consider the possibility that: (i) the paramagnetic metal species which are possibly present on the surface of pyrolysed supported-metal carbonyl clusters could be investigated by e.s.r.spectroscopy and (ii) new paramagnetic species could be generated by reaction with oxygen and/or carbon monoxide, which are used to probe the electronic properties of both paramagnetic and diamagnetic surface species. The e.s.r. work reported in the present paper was carried out on systems differing in either the supported transition-metal carbonyl cluster or the support metal oxide in order to investigate the different roles of the transition metal and of the metal-support interaction in the generation of the paramagnetic species.The interpretation of the e.s.r. spectra was based on previous work on paramagnetic transition-metal car- bony1 cluster^.^? The reaction with oxygen was considered because the paramagnetic 0, molecule can easily interact, depending on the electronic properties of the surface transition-metal centres, with either paramagnetic or diamagnetic species and may often produce new paramagnetic oxygenated products. Moreover, the use of 0,, which usually acts as a z-acceptor system, could be diagnostic of the ability of the metal to transfer electrons from its dn orbitals to the z* molecular orbitals of the acceptor molecule.Several catalytic processes probably require this mechanism and the catalytic activity could thus depend on the amount of electron back-donation. The interpretation of the e.s.r. spectra of the oxygenated product was based on extensive investigations of the oxygen paramagnetic species generated by supported transition- metal ions' and by undiluted or supported transition-metal complexe~.~~~ Finally, the reaction of the surface paramagnetic metal species with CO was tested, in view of the possible use of these systems as catalysts for the CO+H, reaction.1° EXPERIMENTAL PREPARATION OF [Co,(CO),,]/y-Al,O,, ZrO, AND [Rh,(CO),,]/y-Al,O,, ZrO, Solutions of [Co,(CO),,] and [Rh,(CO),,] in anhydrous deaerated n-pentane were added dropwise, under an inert atmosphere, to deaerated y-Al,O, or ZrO, suspended in the same solvent.The carbonyl solutions lost their colour on chemisorption of the cluster on the support. When the addition was complete the mixture was allowed to equilibrate for 2 h while being stirred; the solid was then filtered off. Samples were dried in vacuo at room temperature for 2 h and stored under a N, atmosphere. Contact with air was avoided as it could affect the formation of the paramagnetic species after decarbonylation. For example, if [Rh,(CO),,]/y- A1,0, was allowed to oxidize in air until the Rh(CO), fragment species was formedll and subsequently decarbonylated (see later), no paramagnetic species were observed.The percentages of the supported compound, expressed as g of carbonyl metal cluster per 100 g of the support were: [Co,(CO),,]/y-Al,O,, 2.9 %; [Co,(CO),,]/Zr02, 2.5 %; [Rh,(CO),,]/y- Al,O,, 3 % ; [Rh,(CO),,]/ZrO,, 2.5 % . The i.r. spectrum of [Rh,(CO),,]/y-Al,O, shows two absorptions attributable to CO vibrations of [Rh,(CO),,], as already observed by Ichikawa.', [Rh,(CO),,] and [Co,(CO),,] were prepared as in ref. (13). y-Al,O, was from Akzo Kemie, CK 300, 50-150 mesh, thermally pretreated for 1 h at 200 "C, 1 h at 300 "C, 1 h at 400 "C and 6 h at 550 "C under an inert atmosphere (surface area 200 m2 ggl). ZrO, was from Strem, containing 98% ZrO, and 2% y-Al,O,, 50-150 mesh, thermally pretreated at 500 "C for 7.5 h under an inert atmosphere (surface area 70 m2 g-l).VACUUM TREATMENT OF THE SAMPLES [Co4(C0),,]/y-A1,0,, ZrO, and [Rh,(CO),,]/y-A120,, ZrO, were treated in vacuo ( lo-, Pa) for 2 h at 180 "C (cobalt cluster) and 250 "C (rhodium cluster). The decarbonylation progressT. BERINGHELLI et al. 1481 was monitored by i.r. spectroscopy. Treatments in uucuo were performed in a small flask connected to an e.s.r. tube (internal diameter 3 mm). Contacts with gas at controlled pressures were carried out on a standard gas-vacuum line. The e.s.r. spectra were recorded on a Varian E-109 spectrometer equipped with an automatic Varian temperature control. COMPUTATIONAL METHOD FOR THE SPIN CONCENTRATION The spin concentration was determined by double integration of the area under the resonance lines, taking as reference area that of the Varian weak pitch, which contains 1013 spin per cm of length.The sensitive region of the e.s.r. cavity was 1 cm. The apparent density of samples was 1 g cm-, for y-Al,O,-supported samples and 3.5 g cmP3 for Zr0,-supported samples. The broad shape of some resonances and the overlap between others limited the accuracy of the quantitative evaluation and thus only the orders of magnitude of the spin concentration can be safely assumed. RESULTS DETECTION AND CHARACTERIZATION OF PARAMAGNETIC SPECIES SIJPPORT SYSTEMS y-Al,O, was treated in vacuo ( Pa) at 180 and 250 OC, then contacted with argon (26.6 kPa). The only observed resonances were those due to the high-spin tetragonally distorted Fe3+ centres ( g = 4.25). The three-line low intensity signals at g x 2 could be due to very small quantities of low-spin rhombic Fe3+.14 Control experiments on thermally treated y-Al,O, contacted with either 0, or CO, under the same conditions as the samples, showed no changes in the support resonance lines, thus ruling out interaction of these gases with the paramagnetic centres of the support.ZrO, was treated in vacuo ( lo-, Pa) at 180 and 250 "C, and contacted with argon and then with 0, and CO under the same conditions as the samples. The support, after thermal treatment, showed the resonance lines of a radical species and were not affected by contact with the gases. As the observed ZrO, signal could interfere with the signals of the metal-containing samples, their spectra were corrected accordingly. DECARBONYLATED SYSTEMS Co/y-Al,O,, under an argon atmosphere (26.6 kPa), gave symmetrical resonance lines, whose linewidth (12 G) and position ( g = 2.00) suggest the presence of a radical species.The resonance lines were not observed before decarbonylation of [Co,(CO),,]/y-Al,O,. It is difficult to draw any conclusion as to the nature of this paramagnetic species, although the cobalt valence electrons are not responsible for this signal. On the other hand, because of the low intensity of these lines it was decided not to investigate the nature of this radical species further. Rh/y-Al,O, under an argon atmosphere (26.6 kPa) gave strong asymmetrical lines, whose position (gll = 2.21, gl = 2.1 1) and width (fig. 1) suggest the location of an unpaired electron on Rh centres.The lines are strongly temperature dependent; they decrease in intensity and broaden at room temperature, as expected for 4d transition- metal-ion resonances, for which the spin-lattice relaxation time controls the linewidth. The presence of hyperfine structure in the perpendicular region of the resonances indicates interaction with more than one Rh(l = 1/2) nucleus. Although the spectra are easily reproducible, the number of lines is difficult to define. Measurements of the hyperfine coupling constant are not warranted as the spectral pattern depends on the pyrolysis temperature (see fig. 1). The multi-line nature of the spectra seems to arise from a contribution of several polynuclear paramagnetic species with very similar g values rather than from the hyperfine coupling of a single species.The ground state of the paramagnetic polynuclear species can be assigned by assuming a conventional paramagnetic centre6 and it requires a molecular orbital comprised mainly of the same 49 FAR 11482 PARAMAGNETIC METAL AND OXYGEN SPECIES 4 \r 200 G Fig. 1. X-band e x . spectra recorded under an argon atmosphere (26.6 kPa) at - 150 "C: (a) y-Al,O, ; (6) Rh/y-Al,O, from [Rh,(CO),,]/y-Al,O, pyrolysis at I50 "C ; (c) Rh/y-Al,O, from [Rh,(CO),,]/y-Al,O, pyrolysis at 250 "C. atomic orbitals suggested for the conventional centre by the crystal-field analysis of the g tensor. The values of the g-tensor components of a mononuclear Rh paramagnetic species with gI1 > gl 2 2 could be diagnostic of two different electronic configurations, d7 RhT1 with the unpaired electron in the dz2--.2/2 orbital or d9 Rho with the unpaired electron in the same orbital, corresponding to compressed or elongated tetragonal symmetry, respectively.Contrasting reports on the assignment of the Rh oxidation state based on e.s.r. data have been p~b1ished.l~ In the present case the clustered Rh atoms could be represented magnetically by a d9 or a d7 centre; both are in agreement with an oxidative interaction between Rh and y-Al,O, as the number of interacting Rh atoms is not known. A definite conclusion on the electronic configuration of the Rh species will be drawn from an analysis of the interaction with oxygen.T. BERINGHELLI et al. 1483 Fig. 2. X-band e.s.r. spectra, recorded at - 150 "C: (a) ZrO,; (b) Rh/ZrO, recorded under an 0, (10 Pa) atmosphere; (c) Rh/y-Al,O, recorded under an 0, (10 Pa) atmosphere; ( d ) Rh/ZrO, recorded under an 0, (10 Pa) and CO (10 Pa) atmosphere; ( e ) Co/y-Al,O, recorded under an 0, (10 Pa) atmosphere.Co/ZrO, and Rh/ZrO, were treated analogously to the y-Al,O,-supported samples and did not give resonance lines other than those due to the support. REACTION WITH OXYGEN Co/y-Al,O, was contacted with 0, (26.6 kPa) at room temperature for 5 min. The e.s.r. spectrum shows strong new resonance lines having a different shape and higher intensity than the radical species under an argon atmosphere. The intensity and resolution of the new signals increase at low pressure (lo-, Pa) and the magnetic anisotropy becomes evident (fig. 2).The spectrum is obtained either at low (- 150 "C) or at room temperature. This behaviour, together with the values of the g-tensor 49-21484 PARAMAGNETIC METAL AND OXYGEN SPECIES components (gll = 2.034, gl = 2.00) suggests that the resonances can be attributed to 0; fixed on A13+ c e n t r e ~ . ~ * l ~ The bonding interaction between 0; and A13+ cannot be removed by vacuum or vacuum and thermal treatment (1 80 "C, 1 0-3 Pa), in agreement with the usual behaviour of the A13+-O; species.17 Rh/y-Al,O, was contacted with 0, (26.6 kPa) at room temperature for 5 min and did not reveal any resonance lines; the lines seen under the argon atmosphere are no longer noticeable. The sample, evacuated to 10 Pa, showed new resonance lines (fig. 2) whose intensity and resolution are strongly dependent on the temperature.At - 150 "C g-tensor anisotropy is evident (gl = 2.09, g , = 2.03, g , = 2.00). If we consider the rhombic anisotropy of the g tensor, the value of its principal components and the relative intensity of the e.s.r. lines, satisfactory agreement is found with the g-tensor properties of the monomeric superoxo compound [Rh(en),C1(O,)]+ls and of dimeric superoxo-bridged compounds such as [(Rh(en),Cl),(p and O,)I3+ l8 [(Rh(bpy),C1),(p-0,)]3+.1B The values of the g-tensor components suggest that the unpaired electron occupies a 0, n* orbital, with a small contribution from the Rh d . orbitals. However, the amount of the actual charge on 0, cannot be defined from the e.s.r. data and the location of the unpaired electron on the 0, n* orbitals does not necessarily mean that one negative charge is transferred to 0,.For simplicity, whenever we refer to the RhIL1-0; formulation, we mean that the negative charge is localized mainly on 0,. The anisotropic character of the g tensor implies that the 0-0 bond is bent with respect to the Rh-0 bond direction, but we cannot decide between monomeric and dimeric superoxo species. The suggestion of a superoxo- bridged species does not disagree with the hypothesis that Rh-Rh interactions are not fully removed by the supporting processes. The formation of 0x0-bridged rhodium clusters was also tentatively suggested for Rh/y-Al,O, samples obtained from pyrolysis of [Rh6(CO)l,]/y-Al,0,.20 The interaction between 0, and Rh/y-Al,O, is not removed by vacuum treatment ( Pa); vacuum and thermal treatment (250 "C) remove 0, from the coordination, but new e.s.r.lines, different from those of the unoxygenated precursor, appear. These lines are probably related to the irreversible oxidation of Rh and therefore the usefulness of 0, as a probe for the electronic state of the surface metal centres or as model of catalytically important electron-acceptor systems is doubtful. Irreversibly modified systems were not studied. RhIII-0; formulation requires that the spectrum observed after pyrolysis is that of a RhI1 species and that the oxygen adduct is formed by a one-electron-transfer process, as invoked for d7 centres.,' Co/ZrO,, after contact with O,, did not give any resonance lines in addition to those observed for 21-0,.Rh/ZrO, was contacted with 0, (26.6 kPa) for 5 min at room temperature. Comparison with the spectra of ZrO,, analogously treated, did not reveal any new paramagnetic species. On evacuation at 10 Pa two different new paramagnetic species appeared at - 150 "C (fig. 2): one was identical to that observed on interaction of 0, with Rh/y-Al,O, and the other is almost temperature independent and could be easily separated from the first by recording the spectrum at room temperature. The values of the g-tensor components (g,, = 2.03, gl = 2.00) suggest that the latter signal could be attributed to 0; fixed on Zr4+ centres;,, as the assignment of the perpendicular component is doubtful when strong ZrO, signals are present, other evidences on the nature of this resonance will be given in the following section.Both 0; signals disappear on vacuum treatment Pa) at room temperature for 24 h, thus ruling out the possibility that the RhI"-O; species found on the Rh/ZrO, sample is due to the low content of y-Al,O, in ZrO,.T. BERINGHELLI et ai. L 1485 Fig. 3. X-band e.s.r. spectrum of Rh/y-Al,O, recorded under a CO (10 Pa) atmosphere. REACTION WITH co Co/y-Al,O,, ZrO, and Rh/ZrO, were contacted with CO (26.6 kPa) at room temperature. No new resonance lines appeared. Rh/y-Al,O, was treated with CO (26.6-10 kPa) and revealed strong modification of the e.s.r. spectrum with respect to that observed under argon (fig. 3). Signals attributed to formal Rhl* polynuclear species disappear. The new resonance lines, whose intensity does not depend on the CO pressure, are because of their width still indicative of the location of the unpaired electron on the Rh centre.However, their small paramagnetic anisotropy and the displacement of the g-tensor value from that of the free electron suggest that the extent of electron location on Rh decreases on interaction with the CO molecule. In the absence of hyperfine structure, no direct proof of the nuclearity of the CO adduct is available. Thermal treatment (100 "C) in uucuo (lo-, Pa) removes the Rh-CO interaction. REACTION WITH co AND 0, On successive contact with CO and O,(26.6 kPa), and further evacuation to 10 Pa, Co/y-Al,O, reveals the lines of 0; fixed on A13+, whose stability properties have been already described, and Co/ZrO, does not show any other resonances beside those of the support.Rh/y-Al,03 reveals the lines of 0; fixed on A13+, in addition those of the Rh-CO interaction product. Neither Rh-CO or A13+-O; interactions could be removed by vacuum treatment. Thermal treatment (100 "C) in vacuu Pa) removes only the Rh-CO interaction. Rh/ZrO, reveals only the species attributed to 0; fixed on Zr4+ (fig. 2). Following vacuum treatment ( Pa) for 4 h the intensity of these lines decreases and the resonances due to 0; fixed on Rh become evident. After 24 h under vacuum at room temperature all the reduced forms of oxygen are removed from the ZrO, surface. The formation of 0, reduced species is a reversible process. The behaviour observed following contact with CO and then with 0, can be interpreted as follows : on y-Al,O,-supported samples the CO molecule successfully competes with oxygen for the interaction with Rh, so that the only available coordination sites for 0; are provided by the support and the adduct can be formulated as A13+-O;. If we invoke the same competitive process in the case of Zr0,-supported samples we can conclude that the 0, coordinates on the Zr4+ centres, thus confirming that these e.s.r.signals are attributable to Zr4+-O; species. Vacuum1486 PARAMAGNETIC METAL AND OXYGEN SPECIES treatment switches 0; from Zr4+ to Rh, probably as a consequence of the partial removal of CO from Rh. DISCUSSION The values of the magnetic tensors of the species are collected in table 1, together with the number of paramagnetic centres derived from the e.s.r.analysis. The assignment of the g-tensor components was based on the relative intensity of the resonances. We have made the following observations. (i) The number of paramagnetic centres on Rh/y-Al,O, is much lower than the number of supported Rh, units (ca. 0.02%). This can be explained by supposing that besides the Rh, units interacting with y-Al,O,, higher aggregates also exist. It is also likely that not all the A13+ centres have the same acidic strength, as found for thermally pretreated y-Al,O,, and their electron-attracting power is not sufficient to produce electron-deficient paramagnetic centres. (ii) 0, interaction with Rh/y-Al,O, involves an amount of Rh centres slightly less than or comparable to the number of paramagnetic Rh centres observed before contact with oxygen. This behaviour seems to be in keeping with that observed when the species interacting with 0, are ionic and monon~clear.~~ (iii) The presence of paramagnetic metal centres before oxygenation is not a necessary condition for electron transfer to 0,.Indeed the reduction of 0, over Rh/ZrO, samples proceeds in the absence of a detectable number of paramagnetic metal centres, producing the same number of oxygenated centres as found over Rh/y-Al,O, samples; moreover, 0, reduction over Co/y-Al,O, produces more paramagnetic centres than in the starting material, confirming that no relationship exists between the initial and final paramagnetic species. On the other hand, only paramagnetic centres can generate paramagnetic CO adduc t s. The above results suggest that the formation of paramagnetic species by decar- bonylation of supported clusters depends on either the support and/or the metal properties.Indeed the presence of detectable paramagnetic centres for only y- Al,O,-supported samples is consistent with the greater acidic character of y-Al,O, compared with ZrO,. The larger number of paramagnetic centres found for Rh samples with respect to Co samples (whatever the nature of the cobalt paramagnetic species) is due to better overlap between the A13+ p orbitals and the Rh d, orbitals, the overlap being caused by a larger expansion of the Rh orbital lobes. The resonance lines found for Rh/y-Al,O, samples suggest that the Rh-Rh interaction in the supported Rh cluster is not totally removed by the metal-support interaction; therefore the metal-metal interaction is more likely to occur for the e.s.r.-inactive Zr0,-supported samples, where the metal-support interaction is less than on y-Al,O,. Note that if the support does not remove the metal-metal interaction, the vacancy in the valence electrons generated by the interaction of 7-Al,O, with rhodium should be distributed among several metal centres, as usually observed in stable paramagnetic metal carbonyl clusters where hyperfine structure is e ~ i d e n t .~ * , ~ The oxidizing effect of y-Al,O, on a metal centre could be transferred to another atom, which may not be in contact with the oxide surface. The possibility that a perturbation generated on a metal centre, for example after a change in the ligand molecule, is transferred to other metal centres in a cluster is well known in the chemistry of metal carbonyl interaction with the support, in our metal systems, could result in such an effect.The results of the e.s.r. investigation concerning the chemisorption of 0, and COTable 1. E.s.r. dataa for paramagnetic centres for supported metalsb no. of 0, co paramagnetic pressure pressure centres paramagnetic /spin g-l speciesC sample /Pa /Pa g,, g x x or yy g y y or xx notes c04(c0)1 2/YmA12'3 0 10 0 10 10 Rh4(C0)12/y-A1203 0 0 10 Rh4(C0)12/Zr02 0 10 0 10 0 0 10 10 0 0 10 10 0 0 10 10 2.00 2.034 2.034 2.2 1 2.09 - 2.034 2.034 { 2.034 - - 2.03 2.00 2.00 2.00 2.1 1 2.03 - 2.034 2.00 2.034 2.03 2.00 2.00 - - 2.00 2.00 2.00 2.1 1 2.00 - 2.034 2.00 2.034 - 2.00 2*oo 1 - 2.00 0.28 x 1014 radical 0.56 x 1015 ~ 1 3 + - 0 ; 1.12 x 1015 ~ 1 3 + - 0 ; 0.46 x 10'' RhlI1- 0, - - 0.56 x 10l6 RhII 0.28 x 10l6 RhII-CO 1 0.42 x 1015 ~ 1 3 + - 0 ; 1.38 x 1015 RhII-CO - stable at 180 "C and lop3 Pa stable at 180 "C and lop3 Pa multinuclear species thermally removed at 250 "C, - c3 % lop3 Pa with irreversible z sample modification p Pa z z % # thermally removed at 100 "C, stable at 100 "C, Pa removed by vacuum treatment at Pa, reversible process on vacuum treatment ( Pa) Rh"I-0, is formed as well as - - zr4+-0; a From spectra at on y-Al,03 = 0.32 x - 150 "C.1020 g-l. Rh4 units supported on y-Al,03 = 0.24 x 1020 g-l; Rh, units supported on ZrO, = 0.17 x 1020 g-l; CO, units supported The meaning of the oxidation state indicated for the paramagnetic species is detailed in the text.1488 PARAMAGNETIC METAL AND OXYGEN SPECIES molecules can be interpreted as follows.Oxygen is reduced by the supported metal clusters with the reaction pathway and products depending on the electronic structure of the metal and/or the metal-support interaction. y-Al,O,-supported Co is not rich enough in d, electrons to stabilize bonding with oxygen. It is well known that bonding between oxygen and transition-metal ions is partially a d, - n* bond interaction; in the presence of A13+ centres it is likely that electron transfer from the cobalt d, orbitals to the A13+p orbitals successfully competes with that to the 0, n* orbitals. The y-Al,O, surface acts as an electron-transfer medium for the electrons which reduce the 0, fixed on AP+ to 0;.By a different reaction pathway 0, could interact with cobalt and after reduction could migrate over A13+; however, the formation of O;, in the presence of CO, is difficult to explain. Reduction of 0, does not take place over Co/ZrO, because of the lower ability of this support to accept electrons from the cobalt. In the case of Rh/y-Al,O, and Rh/ZrO, the metal has more expanded d, orbitals than cobalt, thus allowing better overlap with the oxygen n* orbitals. Bonding with oxygen is partially covalent, as suggested by the anisotropy and the temperature dependence of the e.s.r. resonances; moreover, it is very likely that the internuclear axis of 0; is bent with respect to the Rh-0 axis as the interaction with 0, should depend strongly on the expansion of the d, orbital lobes.On the other hand, the reduction of 0, by Rh/y-Al,O, after contact with CO can be explained only by electron transfer from rhodium ' via y-Al,O, ' as the Rh centres are presumably bound to CO. With ZrO, supports both Rh and Zr centres are involved in the fixation of O;, but it is improbable that 0, can accept electrons from rhodium 'via support' because of the low electron-accepting properties of ZrO,. Most likely the superoxide anion is produced on Rh centres, then partially migrates towards Zr centres. Note that the strength of Rh-0, bonding is lower on Zr0,- than on y-Al,O,-supported samples, so that by simple vacuum treatment 0; jumps between Rh and Zr and is then removed from the coordination. The same conclusion can be reached for the CO interaction; indeed the vacuum treatment removes CO from the Rh/ZrO, samples, while only thermal treatment destabilizes the Rh-CO bonding in y-Al,O,-supported samples.The higher strength of RhIII-0; and RhII-CO bonds in y-Al,O,-supported samples is due to the higher acidic character of Rh centres for y-Al,O,-supported samples. This situation is favourable to a CJ interaction between Rh and the chemisorbed molecules. CONCLUSIONS One of the most important problems related to the physico-chemical investigation of supported-metal systems is concerned with knowledge of the oxidation state of the metal. There seems to be evidence that the selectivity of such systems depends on metal centres having a positive oxidation state.l'~,~ By dispersing metal centres over supports of different acidity it is possible to obtain the metal in an oxidized form following interaction with support.Conventional catalysts, obtained by reduction of supported metal ions, show, in general, oxidized centres whose amount is limited to those interacting with the support. The formation of large metal particles hinders the interaction and lowers the reactivity of the metal centres towards the molecules involved in the catalytic reactions. On the basis of our results it seems reasonable that, when the metal precursor is a pyrolysed cluster, the metal on the surface is not a mononuclear system and the valence positive charge generated by support interaction can reasonably be distributed among all the interacting metal centres. Thus by means of metal clusters we can increase the availability of active metal centres and have more chance than withT.BERINGHELLI et al. 1489 mononuclear metal complexes to affect the reactivity through the choice of the support. Finally, although the number of paramagnetic centres found on the surface is small compared with the total number of metal atoms, they can still act as catalytic sites in some catalytic reactions, as has been shown in the case of Ag-supported systems.17 Samples of Rh/y-Al,O, and Rh/ZrO,, obtained from pyrolysis of [Rh,(CO),,], have been tested as catalysts for the CO + H, reaction27 and show that: (i) y-Al,O,-supported samples produce methane and higher hydrocarbons and (ii) Zr0,-supported samples selectively form oxygenated products.Our results on the e.s.r.-active species show that the interaction of n* acceptor molecules such as CO and 0, with Rh is stronger on y-Al,O, than on ZrO,. As a consequence, on y-Al,O, the stronger metal-CO interaction weakens the C-0 bond and leads to hydrocarbon formation; on ZrO, the weaker interaction cannot lead to dissociation of the C-0 bond and oxygenated products are obtained. We thank M.P.I. and the Italian C.N.R. (Progetto Finalizzato Chimica Fine e Secondaria) for financial support. S. J. Tauster, S. C. Fung and R. L. Garten, J. Am. Chem. SOC., 1978, 100, 170. R. H. Grubbs, Chemtech., 1977, 7, 512. B. C. Gates and J. Lieto, Chem. Technol., 1980, 195; 1980, 248. B. C. Gates, in Chemistry and Chemical Engineering of Catalytic Processes, ed. R. Prins and G.C. A. Schuit (Sijthoff & Noordhoff, Amsterdam, 1980), p. 427. G. Longoni and F. Morazzoni, J. Chem. SOC., Dalton Trans., 1981, 1735. T. Beringhelli, F. Morazzoni and D. Strumolo, J. Organomet. Chem., 1982, 236, 109. J. H. Lunsford, Catal. Rev., 1973, 8, 135. (a) F. Campadelli, F. Cariati, P. Carniti, F. Morazzoni and V. Ragaini, J. Catal., 1976, 43, 167; (b) G. Mercati, F. Morazzoni, M. Barzaghi, P. Carniti, V. Ragaini and F. Campadelli, J. Chem. SOC., Faraday Trans. I , 1979, 75, 1857. M. Barzaghi, T. Beringhelli and F. Morazzoni, J. Mol. Catal., 1982, 14, 357. A. Ceriotti, S. Martinengo, L. Zanderighi, G. Tonelli and A. Iannibello, Atti del ZZZ Congress0 Nazionale di Catalisi, Torino, Italy, 1982, p. 26; A. Ceriotti, S. Martinengo, L. Zanderighi, G. Tonelli, A. Iannibello and A. Girelli, J. Chem. SOC., Faraday Trans. I, 1984, 80, 1605. l 1 G. C. Smith, T. P. Chojnacki, S. R. Dagupta, K. Ivatake and K. L. Watter, Znorg. Chem., 1975, 14, 1429; A. K. Smith, F. Hughues, A. Theolier, J. M. Basset, R. Ugo, G. M. Zanderighi, J. L. Bilhou, V. Bilhou-Bougnol and W. F. Graydon, Znorg. Chem., 1979, 18, 3104. l 2 M. Ichikawa and K. Shikakura, in New Horizons in Catalysis, ed. T. Seiyama and K. Tanabe (Elsevier, Amsterdam, 1980), p. 925. I3 (a) S. Martinengo, G. Giordano and P. Chini, Znorg. Synth., 1980,20,209; (b) R. Ercoli, P. Chini and M. Massi-Mauri, Chim. Ind. (Milan), 1959, 41, 132. l4 J. Caspary, D. A. Lanzo and C. Niziak, Biochemistry, 1981, 20, 3868. l5 M. D. Sastry, K. Savitri and B. D. Joshi, J. Chem. Phys., 1980,73, 5568. l6 A. A. Gezalov, G. M. Zhabrova, V. V. Nikisha, G. B. Pariiskii and K. N. Spiridonov, Kinet. Catal., l 7 R. B. Clarkson and A. C. Cirillo Jr, J . Catal., 1974, 33, 392. l8 J. B. Raynor, R. D. Gillard and J. D. Pedrosa de Jesus, J. Chem. SOC., Dalton Trans., 1982, 1 165. Is H. Caldararu, K. De Armond and K. Hanck, Inorg. Chem., 1978, 17, 2030. 2o K. L. Watters, R. F. Howe, T. P. Chojnacki, Chia-Min Fu, R. L. Schneider and Ning-Bew Wong, J. 21 F. Basolo, B. M. Hoffman and J. A. Ibers, Acc. Chem. Res., 1975, 8, 384; B. B. Wayland and 22 M. Setaka, S. Fukuzawa, Y. Kirino and T. Kwan, Chem. Pharm. Bull., 1968, 16, 1240. 23 J. R. Katzer, G. C. A. Schuit and J. H. C. Van Hooff, J. Catal., 1979, 59, 278. 24 C . E. Strouse and L. F. Dahl, Discuss. Faraday Soc., 1969, 47, 93. 25 K. J. Karel and J. R. Norton, J. Am. Chem. SOC., 1974, 96, 6812. 26 S. Sivasankar, A. V. Ramaswamy and P. Ratnasamy, J. Catal., 1977, 46, 420. 27 M. Ichikawa, J. Chem. SOC., Chem. Commun., 1978, 566. 1968, 9, 462. Catal., 1980, 66, 424. A. R. Newman, J. Am. Chem. SOC., 1979, 101, 6472. (PAPER 3/1295)