J. Chem. SOC., Faraday Trans. I , 1982, 78, 273-28 I Calorimetric Study of the Adsorption of Carbon Monoxide, at 296 K, on Supported Nickel and Nickel-Copper Catalysts BY JEREMIA J. PRINSLOO~ AND PIERRE C. GRAVELLE* Institut de Recherches sur la Catalyse, CNRS, 69626 Villeurbanne, France Received 23rd February, 198 1 A calorimetric study of the adsorption of carbon monoxide, at 296 K, on a series of silica-supported nickel and nickel-copper catalysts was carried out. The experimental results were compared with previously published i.r. spectroscopic and magnetic data for the same systems. In agreement with the Chatt-Dewar model for CO chemisorption on metals, changes of the heat of formation of metal-CO bonds can always be associated with shifts of the frequency of the C-0 stretching vibration.The heat of the metalLC0 bond formation appears to be related to the position of the Fermi level at the metal surface. Formation of both linear and bridged adspecies, in definite proportions on each alloy, would obey a concerted mechanism, controlled by the electronic properties of the surface nickel atoms. Data on the heats of chemisorption of carbon monoxide on nickel surfaces are abundant: in their recent review,' Toyoshima and Somorjai cite nine determinations on polycrystalline samples and eleven on single-crystal surfaces. Therefore, the present study was not carried out with the aim to add one more set of values to the list. The purpose of the present study was to characterize calorimetrically the changes in the surface reactivity with respect to carbon monoxide which occur when nickel is alloyed with progressively larger amounts of copper and, by comparing these calorimetric results with those of previous volumetric,2 magnetic and infrared (i.r.) spectroscopic studies3 carried out with the same samples, to look for possible correlations between adsorptive and other properties of nickel surfaces.Correlations between the magnetic and electronic properties of nickel surfaces and their reactivity with respect to hydrogen have been recently pr~posed.~ EXPERIMENTAL The catalysts were prepared by adding the support (silica, 220 m2 g-') to aqueous solutions of ammonia containing nickel and copper nitrates in different proportion^.^ The solid was then filtered, washed and dried at 400 K for 24 h in order to decompose the metal complexes and evacuate ammonia.All the samples used in the present work come from the batches of catalysts which were prepared for previous studies on the nickekopper ~ y s t e m . ~ - ~ , Before each calorimetric experiment, a sample of catalyst (0.1-0.2 g) was reduced at 920 K in flowing hydrogen (4 dm3 h-l) for cu. 15 h, rapidly cooled to room temperature and transferred, in uucuo, into the calorimeter. This procedure, identical to that adopted in the preceding studies with these catalysts,2 was moreover carried out by means of the same vacuum line. In order to test the constancy of behaviour of successive samples, most calorimetric experiments were duplicated (column 5, table 1). t On sabbatical leave from the Chemistry Department of the Rand Afrikaans University, Johannesburg, South Africa.273TABLE I.-VOLUMETRIC AND CALORIMETRIC DATA FOR THE ADSORPTION OF CARBON MONOXIDE, AT 296 K, ON SILICA-SUPPORTED NICKEL AND NICKEL-COPPER CATALYSTS - amount of irreversibly heat of adsorption adsorbed from i.r. bond number average carbon average calculated spectroscopic metal diameter monoxide for per surface data3 composition of metal /(pmol CO) initial" e = 0.2 nickel atom /(atom surface Ni) catalyst [atom % (Cu)] particles/nm (m2 metal)-' /kJ mo1-l /kJ mol-l /kJ (atom Ni)-l (molec. adsorbed CO)-l Ni 0 6.7 4.8 _+ 0.2 142 (3) 141 86 1.65 Ni 98 Cu 2 1.52 5.8 6.1 kO.1 139 (2) 139 87 1.59 Ni 91 Cu 9 8.48 6.2 9.25 f. 0.15 133 (2) 133 92 1.45 Ni 88 Cu 12 11.8 6.1 6.95 +O. 15 125 (1) 124 89 1.40 Ni 80 Cu 20 20.2 6.0 6.25 & 0.15 123 (2) 120 95 1.30 Ni 73 Cu 27 26.7 5.6 4.95 & 0.15 120 (1) 112 97 1.24 Ni 54 Cu 46 46.2 13.3 - 111 (2) 94 100 1.11 " Figures in parentheses indicate the number of separate experiments from which the calorimetric data were obtained.J.J. PRINSLOO AND P. C. GRAVELLE 275 The composition of the samples and the average diameter of the metal particles determined from magnetic data6 are listed in table 1. Evidence has been presented elsewhere6 which demonstrates that the metal phase in all samples is completely reduced and that the surface composition of the alloy particles is very similar to that in the bulk. The heat-flow microcalorimeter, the ancillary volumetric line and the calibration procedures have already been described.' Throughout this work the calorimeter was maintained at 296 K.At this temperature carbon monoxide readily adsorbs on nickel surfaces but its dispropor- tionation to carbon dioxide and carbon does not occur.8 RESULTS The adsorption of carbon monoxide was carried out by introducing successive doses of adsorptive onto the catalyst sample, located in the calorimeter cell. The heat evolved, the quantity of gas adsorbed and the equilibrium pressure were measured at the end of the interaction of each dose, i.e. when the calorimeter had returned to thermal equilibrium. Introduction of doses was repeated until the equilibrium pressure reached ca. 200 Pa. Adsorption isotherms are, for all samples, similar to that reported in fig. 1 of ref. (2), in the case of the pure nickel sample.After the adsorption of the first doses the residual pressure is very low and the adsorbate is irreversibly held on the surface at 296 K. After the adsorption of further doses a measurable equilibrium pressure builds up, evacuation of which brings the desorption of the reversible fraction of the adsorbate. The amounts of irreversibly-held carbon monoxide are given in column 4 of table 1. Mass-spectrometric analyses of the gas phase in equilibrium with the sample have not revealed the presence of impurities at significant level. 2 4 6 Ni91 Cu9 8 k 10 adsorbed CO/pmol (m2 metal)-' FIG. 1.-Differential heats of adsorption of carbon monoxide as a function of the coverage of the silica-supported nickekopper catalyst (Ni 9 1 Cu 9). The calorimetric results are illustrated on fig.1, which reports the variation of the heats of adsorption as a function of the coverage of the metal surface by successive doses of carbon monoxide, in the case of two separate experiments with the Ni 91 Cu 9 sample. Heats recorded during the adsorption of the first doses are constant, within experimental uncertainty ( f 4 kJ mol-I). However, they progressively decrease when the coverage exceeds 3.5-4 pmol (m2 metal)-l, in the case of the Ni 9 1 Cu 9 sample (fig. 1). Constant heats of adsorption for low coverages have been also observed for276 ADSORPTION OF CO ON Ni AND Ni-Cu the other samples of nickekopper catalysts, at least when the copper content does not exceed 20-25 atom %. This observation is more quantitatively presented in columns 5 and 6 of table 1 where values of the initial (extrapolated) heat of adsorption and of the average heat of adsorption for a coverage of 0.2 are given.[In order to determine the quantity of adsorbed carbon monoxide which covers 20% of nickel atoms at the surface of each alloy, the following assumptions were made: (i) bulk and surface compositions are identical,6 (ii) the area covered by each adsorbed carbon monoxide molecule is defined by the bond-number deduced from either i.r. spectra (table 1 , column 8) or volumetric and magnetic data,2 (iii) carbon monoxide adsorbs irreversibly on copper atoms exposed at the alloy surface as it does on pure copper, i.e. 0.5 (pmol CO) (m2 metal)-']. Adsorption of carbon monoxide on copper being ten times smaller than on nickel (table 1 , column 4), the results presented in columns 5 and 6 of table 1 essentially characterize the adsorption of carbon monoxide on nickel atoms exposed at the surface of the alloy particles, at least when the copper concentration is small [< 20 atom % (Cu)].The difference which appears between initial and average heats of adsorption when the copper concentration exceeds 20% is probably explained by the adsorption of part of the adsorbate on exposed copper atoms : experiments with samples of silica-supported copper have indeed shown that the initial heat of adsorption of carbon monoxide on copper only amounts to 77 kJ mol-1 and that the differential heats steadily decrease with increasing coverage. DISCUSSION The observed stability of the differential heats of adsorption of carbon monoxide with increasing (low) coverage is a notable feature of this system.This result indicates that carbon monoxide chemisorption is not sensitive to surface heterogeneity which exists on supported metal particles and this confirms previous experimentalg and theoreticallo studies showing that, with this adsorption system, the role of surface orientation is only of minor importance [with, perhaps, the exception of the (1 11) plane]. Magnetic methods and i.r. spectroscopy have shown that upon adsorption at 296 K two distinct surface species, namely the linear and bridged forms of adsorbed carbon monoxide, are formed in the same proportion for all surface coverages3 The constancy of the adsorption heats (0 < 0.2) is in agreement with these previous results.The present calorimetric data also demonstrate that the adsorption of carbon monoxide, at least for OGO.2, does not modify the surface reactivity for the adsorption of further amounts of carbon monoxide. This result is in contrast with those obtained in the case of the adsorption of hydrogen on the same samples:4 preadsorbed hydrogen does create an induced heterogeneity and, as a result, the differential heats of adsorption of hydrogen regularly decrease with coverage. As previously di~cussed,~ the heat of adsorption of hydrogen always follows the changes in surface saturation magnetization and thus appears to be related, as is ferromagnetism for surface nickel atoms, to the density of electronic states near the Fermi level. Adsorption of carbon monoxide also decreases the saturation magnetization of the sample (fig.2), and it has been demonstrated2 that it is indeed this feature of carbon monoxide adsorption that explains the decrease in the nickel surface reactivity with respect to hydrogen after preadsorption of carbon monoxide. Therefore the constancy of the differential heats of adsorption for 0 < 0.2 shows that a relation between the adsorption energetics and the density of electronic states near the Fermi level is unlikely in the case of the adsorption of carbon monoxide.J. J. PRINSLOO AND P. C . GRAVELLE 277 adsorbed CO/cm3 (n.t.p.) (g metal)-’ FIG. 2.-Differential heats of adsorption, frequency of the stretching vibration of the linear adspecies and decrease of saturation magnetization3 as a function of the surface coverage of the Ni 98 Cu2 copper-nickel catalyst by adsorbed carbon monoxide.Chemisorption of carbon monoxide on nickel causes a nett electron transfer from the metal to the adspecies as shown by the observed increase in work function by ca. 1.5 eV at complete c~verage.~’ l2 Alloying nickel with copper, which produces the filling of holes in the d band of nickel by copper s electrons, as demonstrated by the resulting decrease in the nickel magnetic moment, should therefore increase the binding energy of carbon monoxide with the alloy surface. The shift in frequency of the C-0 stretching vibration in the adsorbed state towards lower values, observed in the i.r. spectra of carbon monoxide adsorbed on nickel alloyed with increasing amounts of copper (from e.g.2058 cm-1 for pure nickel to 2005 cm-l for Ni 28 Cu 72),3 is in agreement with this analysislO since it indicates that electron transfer from adsorbent to adsorbate becomes easier as the copper concentration increases. Therefore, the large decrease in the adsorption heats experimentally observed when carbon monoxide is adsorbed on samples of nickel alloyed with an increasing proportion of copper (table 1 , columns 5 and 6) is surprising and cannot be directly related to the frequency shifts in the i.r. spectra of carbon monoxide adsorbed on the same sample^.^ The most obvious feature of the i.r. spectra is not the frequency shifts but the evolution of the intensity of bands A and B, respectively corresponding to linear and bridged ad specie^.^ The bond numbers, listed in column 8 of table 1, quantitatively represent this evolution.Therefore, it is tempting to relate the variation in the adsorption heats to changes in the respective populations of linear and bridged species on different alloys. In fig. 3 the initial heats of adsorption of carbon monoxide are plotted as a function of the proportion of linear species at the surface of different nickekopper alloys. The linear correlation which is observed indicates that the heat of adsorption of carbon monoxide on any nickel-copper alloy may be satisfactorily explained by the respective proportions of linear and bridged species on its surface, the heat of formation of each adspecies being constant, within experimental uncertainty.From the linear correlation in fig. 3, values of 108 and 160 kJ mol-1 have been278 ADSORPTION OF C o ON Ni A N D Ni-Cu 1 n I N i54Cu46 Ni73Cu27 80 N i80 Cu20 m a 401 0 XgNru2 I 1 1 1 201 1;o 130 150 heat of adsorption/kJ mol-' FIG. 3.-Initial heats of adsorption of carbon monoxide on different nickel and nickekopper catalysts as a function of the proportion of linear species in the adsorbate, calculated from the i.r. ~pectra.~ computed for the heat of adsorption of carbon monoxide as, respectively, linear and bridged species. The difference between the two heats of formation (52 kJ mol-l) is larger than the value previously reported for the separation between the two states (21 kJ m ~ l - l ) . ~ When the surface coverage by carbon monoxide exceeds 0.2, the differential heats of adsorption decrease (fig.1 and 2). As discussed earlier, carbon monoxide adsorption on nickel is not very sensitive to surface structure: the decrease is not related to a pre-existing surface heterogeneity. Furthermore, it does not result from a modification in the respective proportions of linear and bridged species since the saturation magnetization of the sample decreases linearly for any increasing coverage (6' < 0.5) by adsorbed carbon monoxide (fig. 2). Similarly, throughout the adsorption experiment the ratio of intensities of the two i.r. adsorption bands A and B remains c o n ~ t a n t . ~ Therefore, the decrease in the adsorption heats for coverages > 0.2 is related to some induced heterogeneity. The changes in the adsorbate properties with increasing surface coverage, and particularly the positive shift in the frequency of the stretching vibration, have received several interpretations, invoking such phenomena as dipole-dipole interaction~l~ or collective vibrational modes.14 It seems, however, that the decrease in adsorption heats which has been observed for all samples (fig.1 and 2) is too large to be accounted for by direct or indirect dipolar-coupling effects. The changes in the metal-adsorbate bond strength as a function of coverage are probably due to a modification of the electronic structure of the metal atoms and consequently some delocalization ofthe adsorbate-adsorbent bond must be considered. The fact that the decrease in the adsorption heats appears, for all alloys with < 20% (Cu), at a constant coverage of exposed nickel atoms (0 = 0.2) may give a clue to the extent of delocalization. Induced effects develop when the number of bonds between carbon monoxide molecules and nickel exceeds one bond per five surface nickel atoms.Thus it appears that the electronic perturbation caused by adsorption extends on average to the four nearest neighbours of the nickel atom directly involved in the chemisorption process. A linear CO species would thus sit on an ensemble of five nickel atoms (and correspondingly a bridged CO species would require an ensemble of ten surface atoms). When the surface coverage by carbon monoxide exceeds 0.2, some adspecies are brought in close proximity so that the surface ensembles they need for adsorbing overlap and induced heterogeneity develops.This description is valid for all samples, at least when the copper concentration does not exceed 20-25%. Thus induced heterogeneity is characteristic of the properties of the chemisorptive bond andJ. J. PRINSLOO A N D P. C. GRAVELLE 279 it is not related, but may be indirectly, to the electronic properties of free surface nickel atoms since introduction of copper modifies the electron population and distribution at the nickel sites. As discussed earlier, the calorimetric results for samples containing 27% (Cu) and more integrate various contributions of the less-energetic adsorption of carbon monoxide on exposed copper atoms. It is therefore not possible to extend the preceding discussion to the case of copper-rich surfaces.As the surface coverage by adsorbed carbon monoxide increases, there is a shift of the stretching vibrations characteristic of both linear and bridged adspecies towards higher values. For instance, the shift attains 75 cm-1 in the case of band B, corresponding to bridged species, on the Ni 92 Cu 8 ample.^ In the case of the Ni 98 Cu 2 sample (fig. 2), the total shift for band A (linear species) is smaller. Although the i.r. spectroscopic data in fig. 2 are rather sketchy, they suffice to show that there exists a parallelism between the evolution with coverage of the differential heats and of the frequency of the stretching vibration. This in turn suggests that the Chatt-Dewar modeP5 for the chemisorption of carbon monoxide does apply. If, in agreement with thismodel, the adsorptionenergy is related to the back-donation of electrons from the metal to the 27c* orbital, then there should be a relation between the experimental heats of adsorption on the different copper-nickel alloys and the filling of the dstates of nickel by copper electrons. Since no values of the work-function for the homogeneous alloys have been reported to our knowledge, it is once more necessary to consider the previously published i.r.-spectroscopic data for the same ~amples.~ As discussed earlier, the calorimetric data collected in columns 5 and 6 of table 1 are in opposition to the observed negative shift of the i.r.bands maxima. However, these calorimetric data are expressed per mole of adsorbed carbon monoxide, and therefore from sample to sample they concern a variable number of surface nickel atoms, as shown by the bond numbers listed in column 8 of table 1.If a correlation is to be sought between the electron density at the metal site and the adsorption energy, then all calorimetric data should be referred to a single metal site. The initial heats of adsorption, recalculated per surface nickel atom involved in the chemisorptive bond, are given in column 7 of table 1 . As expected from the i.r.-spectroscopic data, the adsorption energy of carbon monoxide per nickel atom increases when copper is alloyed to nickel. Furthermore, fig. 4 shows that there exists, within experimental uncertainty, a correlation between calorimetric and i.r.- spectroscopic data. Note that the straight line in fig.4 indicates that a frequency close to 2000 cm-l should correspond to 108 kJ (atom Ni)-l, value previously calculated from the correlation presented in fig. 3 for the formation of the linear adspecies, and that the same frequency is also obtained by extrapolating the plot of v-CO frequencies for the A band against copper content towards copper-rich samples on which carbon monoxide adsorbs almost exclusively as a linear specie^.^ Fig. 4 indicates that, in agreement with the Chatt-Dewar model for the chemisorption of carbon monoxide on transition metals, there is a parallelism between the adsorption energy of carbon monoxide, expressed per nickel atom, and the intensity of the back-donation from the metal d states into the molecule 27c antibonding orbital, measured by the frequency of the C-0 stretching vibration.Therefore, the heat of formation of the Ni-CO bond appears to be related to the energetic position of the highest filled chemisorption level with respect to the highest occupied d state in the metal, i.e. with respect to the position of the Fermi level at the nickel-metal surface. However, the changes in adsorption energy resulting from the modifications of the electron density at the metal sites caused by alloying cannot simply be accommodated by the chemisorptive bond, since the correlation, presented in fig. 3, has indicated that either linear or bridged adspecies are formed on all nickekopper alloys with the280 ADSORPTION OF CO ON Ni AND Ni-Cu --. .? d E 2ol / Ni 2060 I 1 1 - 80 90 100 110 heat of adsorption/kJ (atom Ni)-' FIG.4.-Initial heats of adsorption of carbon monoxide, expressed per nickel atom involved in the chemisorptive bond, as a function of the frequency of the stretching vibration of the monodendate species at the surface of different nickel and nickel-copper catalysts. release of constant energy. Inspection of columns 5, 7 and 8 in table 1 reveals that it is only through a modification of the respective proportions of linear and bridged species (i.e. of the bond number) that chemisorption of carbon monoxide complies with the changes brought to the electronic structure of surface nickel atoms by the addition of copper. The formation of linear and bridged species, in definite proportion on any alloy, appears to be concerted and directly related to the electronic properties of the surface nickel atoms.The picture which emerges is that of a surface coordination complex including both linear and bridged carbonyl ligands in defined proportions for any Ni-Cu alloy and retaining its stoichiometry for all surface coverages. This description of the chemisorption of carbon monoxide is in agreement with (i) the linear decrease of saturation magnetization with increasing CO coverage and (ii) the existence of two adsorbed states at all surface coverages, demonstrated by i.r. spectroscopy in the case of the present samples and also by EELS (electron energy loss spectroscopy) on nickel single-crystal surfaces.ll CONCLUSIONS The comparison of the calorimetric data of the present study with data previously obtained with the same samples by magnetic methods and i.r.spectroscopy3 allows us to propose the following main conclusions, which complement the description of the adsorption of carbon monoxide on the surface of nickel or nickel-copper alloys at 296 K presented in the first article of this series.2 (i) The heats of formation of linear and bridged adspecies differ (108 and 160 kJ mol-', respectively) but they remain constant on all nickel-copper alloys. (ii) Linear and bridged adspecies are not separate states, independently adsorbed on distinct sites or patches at the nickel surface. The proportion of the two adspecies varies with the copper content of the alloy, so that the average energy of theJ. J. PRINSLOO AND P. C. GRAVELLE 28 1 metal-carbon-monoxide bond formation follows the same trend as the back-donation of electrons from the metal d states into the 2n antibonding orbital of the carbon monoxide molecule.The energy of adsorption of carbon monoxide, expressed per metal atom, thus appears to be related to the position of the Fermi level at the metal surface. (iii) The electronic perturbation caused by the chemisorptive bond extends on average to five surface nickel atoms. When the surface coverage by adsorbed carbon monoxide exceeds 0.2, i.e. when some ensembles of five nickel atoms start to overlap, induced heterogeneity develops and the heat of adsorption of further quantities of carbon monoxide decreases. (iv) Chemisorption of carbon monoxide is also associated with other electronic perturbations, such as the transformation of ferromagnetic nickel atoms (so.6 dS.*) into non-magnetic ones (so 8 O ) , demonstrated by magnetic measurement^.^ Carbon monoxide adsorption is not influenced by modifications of the electron density of states near the Fermi level as hydrogen adsorption Therefore, preadsorption of carbon monoxide (0 < 0.2) creates an induced heterogeneity for the subsequent adsorption of hydrogen, although the free surface still appears energetically homo- geneous to adsorbing carbon monoxide molecules. The authors thank J.A. Dalmon for the gift of the nickel and nickel-copper samples, P. Moral for the pretreatment of these samples, and G. A. Martin and M. Primet for many helpful discussions. I. Toyoshima and G. A. Somorjai, Cataf. Rev. Sci. Eng., 1979, 19, 105. J. J. Prinsloo and P. C. Gravelle, J. Chem. SOC., Faraday Trans. I , 1980, 76, 512. J. A. Dalmon, M. Primet, G. A. Martin and B. Imelik, Surf. Sci., 1975, 50, 95. J. J. Prinsloo and P. C. Gravelle, J. Chem. SOC., Faraday Trans. I , 1980, 76, 2221. G. A. Martin, B. Imelik and M. Prettre, J. Chim. Phys., Phys. Chim. Biol., 1969, 66, 1682. J. A. Dalmon, J. Catal., 1979, 60, 325. ’ P. C. Gravelle, Ado. Cataf., 1974, 22, 191. * G. A. Martin, M. Primet and J. A. Dalmon, J. Catal., 1978, 53, 321. T. N. Taylor and R. J. Estrup, J. Vac. Sci. Technof., 1973, 10, 26. lo G. Doyen and G. Ertl, Surf. Sci., 1974, 43, 197. l 1 J. C. Bertolini and B. Tardy, Surf. Sci., 1981, 102, 131. l 3 M. Moskovits and J. E. Hulse, Surf. Sci., 1978, 78, 397; M. Scheffler, Surf. Sci., 1979, 81, 562. l4 G. D. Mahan and A. A. Lucas, J. Chem. Phys., 1978,68, 1344. l5 J. Chatt, J. Chem. Soc., 1953, 1, 2939. H. H. Madden, J. Kuppers and G. Ertl, J. Chem. Phys., 1973, 57, 3401. (PAPER 1 /3 12) 10