I. A Nuclear Magnetic Resonance Study of Metal Carbonlys in the Solid State, and II. Studies of the Surface Chemistry of Rhodium Supported on Alumina Citation Gleeson, James William (1982) I. A Nuclear Magnetic Resonance Study of Metal Carbonlys in the Solid State, and II. Studies of the Surface Chemistry of Rhodium Supported on Alumina. Dissertation (Ph.D.), California Institute of Technology. doi:10.7907/vmgw-qf86. https://resolver.caltech.edu/CaltechTHESIS:05152018-092853858 Abstract Part I The principal components of the 13 C nuclear magnetic resonance chemical shift tensors of metal carbonyls containing between one and six metal atoms were determined from the powder patterns of the solid compounds. The tensors of terminally-bound CO groups are highly anisotropic (380 ± 60 ppm) and nearly axially symmetric. The tensors of bridging CO groups are much less anisotropic, due to significant asymmetry in the electron orbitals about the CO internuclear axis. The tensors vary only slightly for different transition metals. There is no intramolecular rearrangement of the metal carbonyls in the solid state at frequencies ≳ 10kHz, except in Fe 3 (CO) 12 . Part II The surface chemistry of rhodium supported on alumina was studied using infrared spectroscopy and quantitative measurements of the gases adsorbed and evolved during various procedures. First, the behavior of alumina-supported Rh upon heating in the presence of CO, CO 2 , O 2 and H 2 was studied. The loss in the capacity to adsorb CO after heating to 525 K increases in the order O 2 , H 2 , vacuum < CO 2 < CO. Upon heating in CO, some CO is oxidized to CO 2 with oxygen from the surface, while the dicarbonyl-forming Rh I is reduced to Rh°. The Rh° agglomerates, accounting for the substantial loss in capacity to adsorb CO. Upon heating in CO 2 , the dicarbonyl-forming Rh I is also deactivated. There is little loss in the capacity to adsorb CO upon heating in H 2 , O 2 or vacuum. Second, the adsorption of H 2 S and its interaction with CO on Rh supported on alumina was studied. The dissociation of H 2 S on the Rh at 300 K produces H 2 and is inhibited by preadsorbed CO. Rh also facilitates the reaction of H 2 S with surface oxygen below or at 373 K, in which water is produced. After exposure of the Rh to H 2 S, CO adsorbs in the linear, but not in the dicarbonyl or bridging modes. Exposure of a CO-precovered surface to H 2 S displaces much of the bridging CO, but only slowly removes the dicarbonyl and linear CO. Exposure to H 2 S strongly inhibits the removal of adsorbed CO by O 2 , but exchange of adsorbed and gas phase CO occurs readily. Item Type: Thesis (Dissertation (Ph.D.)) Subject Keywords: (Chemistry) Degree Grantor: California Institute of Technology Division: Chemistry and Chemical Engineering Major Option: Chemistry Thesis Availability: Public (worldwide access) Research Advisor(s): Beauchamp, Jesse L. Thesis Committee: Beauchamp, Jesse L. (chair) Weinberg, William Henry Chan, Sunney I. Grubbs, Robert H. Defense Date: 17 November 1981 Funders: Funding Agency Grant Number Caltech UNSPECIFIED Office of Naval Research (ONR) N00014-77-F-0008 Office of Naval Research (ONR) N00014-79-F-0014 Record Number: CaltechTHESIS:05152018-092853858 Persistent URL: https://resolver.caltech.edu/CaltechTHESIS:05152018-092853858 DOI: 10.7907/vmgw-qf86 Default Usage Policy: No commercial reproduction, distribution, display or performance rights in this work are provided. ID Code: 10887 Collection: CaltechTHESIS Deposited By: Mel Ray Deposited On: 15 May 2018 18:54 Last Modified: 28 Apr 2025 22:08 Thesis Files Preview PDF - Final Version See Usage Policy. 42MB Repository Staff Only: item control page

I. A NUCLEAR MAGNETIC RESONANCE STUDY OF METAL CARBONYLS IN THE SOIJD STATE. AND TI. STUDIES OF THE SURFACE CHEMISTRY OF RHODIUM SUPPORTED ON ALIDA:INA Thesis by James W. Gleeson In Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy California Institute of Technology Pasadena, California i982 (Submitted November 17, 1981) i i. ACKNOWLEDGEMENTS I thank Professor W. Henry Weinberg for valuable advice during my final two years at Caltech. The efforts of Professor Sunney I. Chan in maintaining the financial support for the research ~re much appreciated. I am indebted to the late Professor Robert W. Vaughan for his inspiring example as a scientist and as a person. I thank the present and former members of the Vaughan and Weinberg groups, especially Mike Duncan, Mike Templeton, Jeff Reimer, and Albert Highe, for helpful advice and discussions and for being pleasant to work with. The assistance of the staff of the Chemistry and Chemical Engineering Division, especially John Yehle, George Griffith, Ray Reed, Seichi Nakawatase, and Eric Siegel, in constructing and maintaining the equipment used in these studies is greatly appreciated. Professor George R. Rossman was very generous in allowing the use of his infrared spectrometer. Kathy Lewis was most patient and accommodating in typing this thesis. The financial support of Caltech and the Office of Naval Research is gratefully appreciated. Finally, and most importantly, I would like to thank two people for assistance of many sorts during the past five years, as well as the previous twenty one, my parents, Mr. James A. and Mrs. Dora R. Gleeson. iii. ABSTRACTS Part I The principal components of the 13 C nuclear magnetic resonance chemi- cal shift tensors of metal carbonyls containing between one and six metal atoms were determined from the powder patterns of the solid compounds. The tensors of terminally-bound CO groups are highly anisotropic (380 ± 60 ppm) and nearly axially symmetric. The tensors of bridging CO groups are much less anisotropic, due to significant asymmetry in the electron orbitals about the CO internuclear axis. The tensors vary only slightly for different transition metals. There is no intramolecular rearrangement of the metal carbonyls in the solid state at frequencies~ 10kHz, except in Fe 3 (C0) 12 . Part II The surface chemistry of rhodium supported on alumina was studied using infrared spectroscopy and quantitative measurements of the gases ~ adsorbed and evolved during various procedures . First , the behavior of alumina-supported Rh upon heating in the presence of CO, C0 2 , 02 and H2 was studied. The loss in the capacity to adsorb CO after heating to 525 K increases in the order 0 2 , H 2 , vacuum < C0 2 < CO. Upon heating in CO, some CO is oxidized to C0 2 with oxygen from the surface, while the dicarbonyl-forming Rh1 is reduced to Rh 0 . The Rh 0 agglomerates, accounting for the substantial loss in capacity to adsorb CO. Upon heating in C0 2 , the dicarbonyl-forming Rh1 is also deactivated . There is little loss in the capacity i v. to adsorb CO upon heating in H2 , 0 2 or vacuum. Second, the adsorption of H2 S and its interaction with CO on Rh supported on alumina was studied. The dissociation of H2 S on the Rh at 300 K produces H 2 and is inhibited by preadsorbed CO. Rh also facilitates the reaction of H 2 S with surface oxygen below or at 373 K, in which water is produced. After exposure of the Rh to H 2 S, CO adsorbs in the linear, but not in the dicarbonyl or bridging modes . Exposure of a CO-precovered surface to H 2 S displaces much of the bridging CO, but only slowly removes the dicarbonyl and linear CO . Exposure to H 2 S strongly inhibits the removal of adsorbed CO by 0 2 , but exchange of adsorbed and gas phase CO occurs readily. v. TABLE OF CONTENTS Acknowledgments Abstract. . . . Part I: . . iii 13 A c Nuclear Magnetic Resonance Study of Metal Carbonyls in the Solid State . . . . . . . 1 Chapter 1: 2 Chapter 2: Introduction . . . . 13 c Nuclear Magnetic Resonance Chemical Shift Tensors of Metal Carbonyls . . . Part I I: Surface Chemistry of Rhodium Supported on Alumina 12 46 Chapter 3: Introduction 47 Chapter 4: Deactivation of CO Adsorption Sites upon Heating Rh/ A1um in a in CO, C:OuOl.and H . • 2 52 H S Adsorption and Interaction with CO on 2 Rh/Alumina . . . ........ . 88 Chapter 5: APPENDIX: c2 H2 Adsorption and Interaction with CO on Rh/Alumina Conclusions . . . . 120 133 -1- Part I: A 13C Nuclear Magnetic Resonance Study of Metal Carbonyls in the Solid State -2- Chapter 1: Introduction - 3 Many chemical reactions require breaking bonds involving carbon , hydrogen and/ or oxygen atoms that typically have bond energies of ~ 80 kcal/mol. This can often be facilitated by the process of catalysis, in which weaker, temporary bonds are made between one or more of the reactants and an additional material, the catalyst, that is recovered unchanged after the reaction. Bonds to transition metals are well suited for this and transition metal catalysts are consequently used extensively in industry. An example of a reaction catalyzed by a transition metal is the oxidation of CO, shown in Fig. 1. Although the thermodynamics of the reaction are not affected by the presence of the catalyst since there is no net change in the catalyst, the kinetics are affected strongly. The activation energy for dissociating the oxygen molecule is nearly eliminated in the presence of the metal. In contrast , it is more difficult for the dissociated oxygen to react with the CO in the presence of the metal, and additional energy is needed to break the bonds to the metal in desorbing the product. The overall effect of the catalyst is , however, highly beneficial toward reaction, such that the reaction proceeds readily at room temperature in the presence of a catalyst (1), whereas a temperature of ,.... 900°C is needed merely to sustain this reaction after ignition in the absence of a catalyst (4). To understand the reason for the high catalytic activity of transition metals, we .consider the energies, number and symmetries of the valence orbitals that bind to the reactants (5-7). The example above illustrated the point that one or more of the reactants must be activated through binding to the metal, but the binding must not be so strong as to prevent further reaction and desorption of the product. In fact, typical binding energies of various species to transition metals in metal complexes and on metal surfaces lie in the range 25-70 kcal/rnol (8-10), somewhat below the energies of the bonds to be broken in the reactants and reformed in the products. In each -4transition metal atom, there are empty s, p and d valence orbitals of various symmetries so that, in principle, different reactants may be bound in proximity to one another, increasing the chances of reaction. While the remaining sites of metal atoms of a metal surface are involved in bonds to other metal atoms, the remaining sites in metal complexes are occupied by ligands that can be chosen to optimize the catalytic properties of the complex. Transition metals bind species such as hydrogen atoms and alkyl groups with covalent bonds qualitatively similar to those in organic compounds . Bonds to the most electronegative elements probably _possess significant ionic character. In addition to these types of bonding found in the main group elements, transition metals also participate in coordinate or donor-acceptor bonding-, illustrated in Fig . 2 for the bonding of CO to a transition metal. The CO "donates" a pair of electrons from one of its filled orbitals to an empty orbital of the metal and accepts electrons from the metal into an unoccupied orbital. The bonding of phosphines, molecular nitrogen, nitriles, isonitriles and other molecules to transition metals is of this type . The ability of transition metals to bind reactants in more than one manner increases the number of ways it can promote the reaction . It is common for a reactant bound in a coordinate manner to change to bonding in a covalent manner after reaction with another species. A striking example of this is found in a molybdenum complex in which both hydrogen and ethylene -ligands are bound. The hydrogen ligand exchanges with the hydrogens of the coordinatively bound ethylene at a rate of "' 500 Hz at 7°C, almost certainly via a covalently bound ethyl group ( 11). This process models an important step in metalcatalyzed olefin hydrogenation. Despite the much greater importance of transition metals dispersed on rela- -5tively inert supports as catalysts in industry, most of the fundamental understanding of the bonding of molecules to transition metals has been obtained from studies of molecular transition metal complexes. In addition to the information concerning the syntheses and reactions of the metal complexes, a variety of physical methods are available for their characterization, among them: X-ray crystallography, infrared, optical, nuclear and electron magnetic resonance and photoelectron spectroscopies, methods for measuring magnetic susceptibilities, and mass spectrometry (12). The use of many of these techniques to study the binding of reactants to dispersed transition metals is hindered by particular features of the dispersed metals. For example, their amorphous nature prevents the use of X-ray crystallography, while their light scattering properties inhibit the use of optical spectroscopy. The reduced concentration of metal-reactant bonds relative to that in solid or even dissolved metal complexes causes substantial problems in the application of nuclear magnetic resonance and sometimes infrared spectroscopy. Recently, however , there has been a burgeoning of sensitive electron spectroscopies appropriate for studying transition metal crystals and molecules bound to them for gas phase pressures less than about 1 o- 4 Torr . The use of nuclear magnetic resonance spectroscopy to study metal complexes has focused primarily on determining the geometrical structures of the metal complexes and the intramolecular rearrangements that they may undergo (13-16). These studies are carried out with the metal complex in a solution, in which the resonances of the various nuclei of a given element depend on the precise chemical environment of the nucleus . They d iffer from each other by amounts of the order of ppm and are often easily resolved in the spectrum. The temperature dependence of the resonance frequencies can provide information about intramolecular rearrangements of the complex (15,16). -6For compounds in the solid state in general, it is much more difficult to resolve the resonances of the nuclei of a given element. This is because there are several effects on the resonance frequency of a nucleus that are dependent on the orientation of the molecule with respect to the externally applied magnetic field, such as the dipole-dipole interactions between nuclei. the chemical shift interaction, and the interaction of a quadrupolar nucleus with electric field gradients at the nucleus. These are averaged by the rapid tumbling of the molecule in a liquid but are not averaged in the solid state. An advantage of obtaining the spectrum of a solid is that the unaveraged interactions can provide more information about the mole~ule than the averaged interactions. Recently, there have been advances which allow one to selectively reduce the effects of the various interactions on the observed spectrum by applying the exciting radiation to the sample in a particular manner (17-19). In Part I of this thesis, the bonding of CO in transition metal carbonyls will be examined with 13 C nuclear magnetic resonance. It is desirable to understand the bonding of CO to metals because of the importance of metal-catalyzed reactions involving CO: reduction by hydrogen to methane and higher hydrocarbons (20,21), oxidation to C0 2 (1,2) and reaction with alcohols to make carboxylic acids, and with olefins and hydrogen to make aldehydes (7). Supported transition metals are generally used in the first two reactions, while soluble metal complexes are used in the latter two. There is much effort toward comparing the bonding of CO in metal complexes to that on metal surfaces (9), and it seems clear that the general features of the bonding are similar in the two cases. The metal carbonyls will be studied by measuring the nuclear magnetic resonance spectra of the solids. The major interaction that will determine the shape of the spectrum will be the unaveraged chemical shift interaction. The spectra -7will be simple enough. to be interpreted in terms of certain aspects of the electronic structure of these compounds. Comparisons among the metal carbonyls and between metal and organic carbonyls will be made. -8References 1. T. Engel and G. Ertl, Advan. Catal. 28, 1 (1979) . 2. G. Ertl. Proc. 7th Intern. Congr. Catal., Preprints, P2, Tokyo, 1980. 3. JANAF Thermochemical Tables, 2nd ed., Office of Standard Ref. Data, Wash- ington, 197 1. 4. R. V. Green, in Encyclopedia of Chemical Technology, 2nd ed., Vol. 4, p. 430, Wiley-Interscience, New York, 1964. 5. F . A. Cotton and G. Wilkinson, Advanced Inorganic OLemistry, 3rd ed., p. 528, Wiley-Interscience, New York, 1972. 6 . The Nature of the Surface OLem.ical Bond , T. Rhodin and G. Ertl, eds., North-Holland, Amsterdam, 1979. 7. G. W. Parshall, Homogeneous Catalysis, p. 5, Wiley-Interscience, New York, 1980. 8. J . A. Connor , Top . Curr . Chern. 71, 71 (1977) . 9 . E. L. Muetterties, T. N. Rhodin, E. Band, C. F . Brucker and W. R . Pretzer, Chern. Rev. 79, 91 ( 1979) . 10. I. Toyoshima and G. A. Somorjai, Catal. Rev.-Sci. Eng. 19, 105 (1979) . 11. J. W. Byrne, H. U. Blaser and J. A. Osborn, J. Amer. Chern. Soc. 97, 387 1 ( 1975) . 12. R. S. Drago, Physical Methods in Inorganic Chemistry, Reinhold , New York, 1965. 13 . B. E . Mann , Advan . Organometal. Chern. 12, 135 (1974) . 14. 0. A. Gansow and W. D. Vernon, in Topics in Carbon-13 JV1JR Spectroscopy, Vol. 2, G. C. Levy, ed ., Wiley-Interscience, New York, 1976 . -915. E. Band and E. L. Muetterties, Chern . Rev. 78, 639 (1978). 16. J. Evans, Advan. Organometal. Chern. 16, 319 (1977). 17. M. Mehring, High Resolution Nk!R Spect'Tioscopy in Solids, Springer-Verlag, New York, 1976 . 18. U . Haeberlin, High Resolution in Solids - Selective Averaging , Academic Press, New York, 1976 . 19. R . W. Vaughan, Ann . Rev. Phys. Chern. 29, 397 (1978) . 20. V. Ponec , Catal. Rev.-Sci. Eng. 18, 151, (1978). 2 1. M.A. Vannice, Catal. Rev.-Sci. Eng . 14, 153 ( 1976) . -10- c 0 (g) + 0 (g) Ea ""'54 ~G = 54 = -113 c 0 (g) + ~ 0 2 (g) c 0 2 (g) E a ~ 5 C 0 (ad)+ 0 (ad) ___,.;----t) C 0 2 (ad ) E a Figure 1. = 25 Catalysis of CO Oxidation by a Group VIII Transition Metal ( M). The activation energy, E , and ~G are a in kcal/mol for the reaction at 400 K {1-3). are for low coverages of co(ad) and o(ad). Values Figure 2. c 0 Donor-Acceptor Bonding of CO to a Transition Metal OM - ~ I 1--' 1--' I - 12- Chapter 2: 13 C NJlR Chemical Shift Tensors· of Metal Carbonyls - 13- Abstract The principal components of the 13 C NMR chemical shift tensors· ot metal carbonyls containing between one and six metal atoms were determined from their powder patterns. The tensors of terminally-bound CO groups are highly anisotropic (380 ± 60 ppm) and nearly axially symmetric. The tensors of bridging CO groups are much less anisotropic, due to significant asymmetry in the elec- tron orbitals about the C-0 internuclear axis. The tensors vary only slightly for different transition metals. There is no intramolecular rearrangement of the metal carbonyls in the solid state at frequencies greater than 10 kHz, except in Fes(C0) 12 . - 14- 1. Introduction Despite the plethora of studies of the interaction of CO with transition metals, many fundamental details of the bonding are not completely understood. From semiempirical and even sophisticated a.b initio theoretical calculations of metal carbonyls, there has yet to emerge a consistent description of such aspects as the relative importance of a and rr bonding, and variations in electronic structure with the type of transition metal and bonding geometry (1). Relative and absolute electronic energy levels have been determined experimentally using electronic absorption and photoelectron spectroscopies, respectively (1,2). Vibrational and NMR spectroscopies are sensitive to specific a~pects of the electronic structure, such as the degree of rr-backbonding from the metal to the CO and to specific electronic excitation energies (2-4). The isotropic chemical shifts measured in 13 C NMR studies of liquids have been correlated to other parameters such as the CO stretching frequency (3-4) and atomic charge populations from theoretical studies (5). Unfortunately, a number of changes in the bonding can cause changes in the isotropic chemical shift that are comparable to the differences observed. NMR studies of metal carbonyls have been more successful in characterizing intramolecular rearrangements in solution that indicate the lability of the carbonyl bonds and serve as models for motion of CO adsorbed on metal surfaces (6-8). There is inherently a greater amount of information in the NMR of solids than liquids since in liquids thE. angular dependence of the chemical shift interaction, as well as that of other interactions, is averaged by the rapid tumbling of molecules. The chemical shift tensor can be defined by three principal components: a 11 , a 22 and a 33 , and three principal directions in the molecule or crystal along which the magnetic field at the nucleus is shifted from the externally applied field by the amounts equal to the three principal components (9-11). - 15- For a chemical shift tensor with axial symmetry, which will at least be approximately the case throughout this study, two of the components and directions coincide: ( 1) The chemical shift along the unique direction is defined as a 11 . We further note that the average, or isotropic, chemical shift and the chemical shift anisotropy are defined, respectively, as: (2) and (3) NMR spectra of solids can also be afiected by intramolecular rearrangements (12). Since the chemical shift tensor depends on the angle of the applied magnetic field with respect to the molecule, not simply the average chemical environment of the observed nucleus, a greater number of rearrangements may change the spectra of solids than of liquids, and in ways more characteristic of the rearrangement. In general, the principal components, but not the principal directions of the chemical shift tensor, are determined from polycrystalline samples, or powders. However, we will see that the tensors of the CO groups of metal carbonyls are axially symmetric, or nearly so, the unique direction being along the C-0 internuclear axis. Thus we obtain, at least approximately, the complete tensors from the powder patterns of metal carbonyls. Initial studies of the 13 C NMR chemical shift tensors of metal carbonyls have been carried out. Mahnke et al. (13) by relaxation time measurements in solution, evaluated the chemical shift tensors of Ni(C0) 4 and Fe(C0) 5 , which were - 16- later confirmed by NMR measurements of the solids (14). By analyzing the diamagnetic and paramagnetic contributions of each chemical shift component, they argued that the differences among the spectra of CO, Ni(C0) 4 and Fe(C0) 5 could be explained in terms of rr-backbonding. This will be discussed later with respect to the present results. Spiess et al. (14) showed that the temperature- dependent changes in the spectrum of solid Fe(C0) 5 were the result of exchange between axial and equatorial CO groups. Such exchange also occurs in the liquid state (15,16). In the present study, we examine the sensitivity of the chemical shift interaction in metal carbonyis to the type of transition metal, the presence of metalmetal bonding and the mode of CO bonding, i.e. terminal or bridged. We will see that the mode of bonding has a particularly large effect. The motional properties of metal carbonyls in the solid state will also be discussed. 2. Experimental Procedures The Cr(C0) 6 and W(C0) 6 were obtained from Alfa-Ventron and the other metal carbonyls were obtained from Strem Chemicals. The 13 CO (90% enriched) was obtained from Merck, Sharp and Dome. The metal carbonyls were enriched in 15 C by d_ i rect exchange of 15 CO with the metal carbonyls. Enrichment of Cr(CO)a. Mo(C0) 6 and W( C0) 6 was carried out in the gas phase at 360 K. while exchange with the other carbonyls was performed in solution at. 300 K. The typical extent of enrichment was 5%, except for (1] 5 -C 5 H5 ) 2 Fe 2 (C0) 4 , which was 80% of the carbonyl groups. Samples were ,.,.. 0.5 g each. NMR spectra of the polycrystalline samples were taken on a largely homebuilt spectrometer, similar to that described previously (17). An external lock system held the 2.35 Tesla magnet to drifts of less than 2 ppm over eight hours. A single coil (5 mm J.D.) double resonance probe (18), tuned for 1H resonance at - 1799.8 MHz and 13 C resonance at 25.1 MHz, was used to obtain the spectrum of (17~-C:5H:5)2Fe2(C0) 4 . A decoupling field of 12 Gauss was applied continuously at the 1 H resonance frequency while the 13 C resonance was observed. The spectra of the other carbonyls ~obtained using a single resonance probe with a B mm J.D. coil and tuned to 25.1 MHz All chemical shift values are reported on the a scale ( -o scale) relative to tetramethylsilane (TMS). Spin-lattice relaxation times of the 13 C nuclei were "' 20 min. and depended on the orientation of the molecules with respect to the external magnetic field. With a pulse repetition time ·of 50 min, spectra were obtained that were free of distortions from relaxation effects. To remove experimental artifacts, pulses with 0° and 180° phase shifts were alternately applied to the sample and the decays alternately added and subtracted. Approximately 128 free induction decays were accumulated for each sample, and then Fourier transformed. The spectrum of a sample at a low temperature was obtained by first allowing the sample to magnetically equilibrate in the magnet at 300 K. The sample was then brought to 77 K by immersing in liquid N2 between the magnet pole faces but outside the NMR coil. Finally, the sample was quickly placed in the coil, pulsed and a decay recorded. The temperature of the sample at this point is estimated to be 100 ± 20 K. With this procedure, possible problems resulting from long spin-lattice relaxation times at low temperatures were avoided. Also, the purpose of these experiments was to determine if there were gross changes in the spectra due to the presence of temperature-dependent motions on the NMR timescale. The.r efore, it was not necessary to know the temperature precisely. - 183. Results Principal components of the chemical shift tensor for the 13 C nuclei of the CO groups of nine metal carbonyls were determined from their 13 C NMR powder patterns. The principal components of these carbonyls, along with those of related carbonyls measured by other authors are listed in Table 1. Illustrative spectra are presented in Figs. 1-4 and are described below. A. Mo(C0) 6 , Cr{CO}a and W(C0) 6 Shown in Fig. 1 is the 13 C NMR spectrum of Mo( C0) 6 at 300 K. The solid line is a nonlinear least-squares fit to the experimental points of a theoretical chemical shift powder pattern convoluted with a Lorentzian broadening function. The chemical shift tensor of the carbon nucleus is axially symmetric due to the four-fold rotational symmetry about the Mo-C-0 internucleus axis in the molecule (19) (slight deviations from the idealized symmetry in the crystal are ignored here). The principal components of the chemical shift tensor, a 11 and a l· obtained from the fit to the Mo(C0) 6 spectrum may be used to calculate the isotropic chemical shift, a, using Eq. (2). The value found, -203 ppm, should be equal both to the center of mass of the spectrum, -208 ppm, as well as to · the isotropic chemical shift obtained from the spectrum of dissolved Mo(C0) 6 , -202 ppm (20) in the absence of intermolecular effects. The agreement is well within the exper- imental error of 15 ppm. Similar agreement was found for other metal carbonyls (13.14.21-24), and is typical of 13 C chemical shifts in general (9). Only the a l components, which were observed easily, were determined directly from the powder patterns of Cr(C0) 6 and W(C0) 6 because of the low signal-tonoise. The values of a 11 were calculated from Eq. (2) using the values of a for - 19- these carbonyls in solution (20). Theoretical powder patterns calculated with these values matched the experimental spectra well within experimental error . We note that it was not necessary to include effects of motional averaging of the chemical shift interaction in fitting the experimental spectra. That the chemical shift interaction is not being averaged for the metal carbonyls here is also supported by the similarity of the principal components determined to the unaveraged components of CO (25.26), Ni(C0) 4 , Fe(C0) 5 (13.14) and Rh2(C0) 4 Cl2 (24), as seen in Table 1. Thus, for the hexacarbonyls, . the frequency of intramolecular rearrangements that would average the chemical shift interaction must be less than 10 kHz, the linewidth due to chemical shift anisotropy. To investigate possible changes in the 13 C chemical shift interaction from bonding CO to a metal cluster rather than to a single metal atom in ·a complex, several cluster carbonyls were examined. In Ru3 (C0) 12 and Os 3 (C0) 12 , four CO groups are bound terminally to each metal atom. There are two equatorial CO groups in the plane of the metal triangle and one axial CO on each side of this plane and perpendicular to it, resulting in \) .lh symmetry for these molecules (27.28). In Ir 4 (C0) 12 , three CO groups are bound terminally to each atom of the Ir tetrahedron resulting in T d symmetry for the molecule (29). The spectrum of Ru 3(CO}t 2 at 300 K is shown in Fig. 2, along with a fit of a single theoretical chemical shift powder pattern. We would expect the spectrum to be a superposition of two equally intense chemical shift powder patterns from the axial and equatorial CO groups. Also, the two tensors are not required to be axially symmetric by the molecular symmetry. However, we see that a single powder pattern of a nearly axially symmetric tensor fits the data quite well. As with the carbonyls containing a single metal atom, the isotropic chemical shift, - zo-201 ppm. and. the center of mass of the spectrum, -196 ppm are within experimental error of the isotropic chemical shift in solution, -199.7 ppm (30). Here, also, there is no evidence for averaging of the chemical shift interaction by intramolecular rearrangements. The signal-to-noise of the spectra of . Os 3 (C0} 12 and Ir4 (C0) 12 was low. The chemical shift tensors of both. carbonyls were assumed to be axially symmetric. Also, a in solid Os 3 (C0) 12 was set equal to that of dissolved Os 3 (C0) 12 (30) and used to calculate u 11 as before. All spectra described so far have been of CO bound terminally to a single metal atom of a complex or cluster. To investigate the bonding of CO to more than one metal atom in a multicentered. or bridging configuration, the powder patterns of Rh6 (C0) 16 and (1] 5 -C 5 H 5 hFe 2 (C0) 4 were measured. Rh 6 (C0) 16 contains triply-bridging CO groups and (1] 15 -C5 H5 ) 2 Fe 2 (C0} 4 doubly-bridging CO groups. The spectra are shown in Fig. 3. In Rh 6 (C0) 16 , two CO groups are bound terminally to each Rh atom, while the remaining four triply bridge alternate faces of the Rh octahedron (31). The spectrum will be a superposition of the powder patterns of the terminal and bridging CO groups, with a relative intensity of 3:1. The chemical shift t;.ensor of the bridging CO groups is axially symmetric due to the three-fold rotational symmetry about the C-0 axis. With the above two restrictions, in addition to setting the isotropic chemical shifts in the solid equal to those in solution {32), the experimental spectrum is fit uniquely and well. The chemical shift tensor of the terminally-bound CO groups in Rh6 (CO)u3 is typical of terminally-bound CO groups in general, both in anisotropy and near - 21 axial symmetry. The anisotropy of the tensor of the bridging CO, however, is only 194 ppm, or about half that of terminally-bound CO. The change results almost exclusively from a shift of a 11 . Evidence that the difference is not due to motional averaging is (a) the typical powder pattern of the terminal CO groups, (b) the lack of exchange on the NMR timescale of CO groups among the Rh atoms of Rb.a(C0) 16 in solution at 343 K (28), and (c) the similarity of the spectra of the solid at 300 and 100 K. The only change in the spectrum that occurs at 100 K is that the intensity from 0 to +80 ppm increases to match the theoretical fit to the 300 K spectrum more closely. This could be due to freezing out some low frequency vibrational motions that average the chemical shift interaction to a small extent. To determine the chemical shift tensor of a doubly-bridging CO group, the 1 H-decoupled powder pattern of (775 -C0 H:s) 2 Fe 2 (C0) 4 was ~easured. There is one CO is bound terminally to each of the Fe atoms, while the other two symmetrically bridge the Fe-Fe bond (33-35). There are cis and trans forms, depending on the relative positions of the cyclopentadienyl and terminal CO groups. The infrared spectrum of the sample used here indicates a mixture of the two forms. However, the isotropic chemical shifts in solution differ by only 0.2 ppm (36,37), an insignificant amount in this study. The experimental spectrum shown in Fig. 3b is due to the carbon nuclei of the CO groups since they were selectively enriched to 80% by exchange with 13 CO. The small contribution of the carbons of the cyclopentadienyl groups was removed by measuring the spectrum of an unenriched sample and subtracting an appropriate intensity from the spectrum of the enriched sample. The number of fitting parameters was reduced as for Rh6 (C0) 16 to obtain a unique fit. - 22- The chemical shift tensor of the terminal CO groups was nearly axially symmetric and had an anisotropy of 444 ppm, which are typical of terminal CO groups in other metal carbonyls. The anisotropy of the tensor of the bridging CO group is even less than that of the triply-bridging CO group in Rh 6 (C0) 16 , 138 ppm vs. 194 ppm, again because the a 11 component has shifted. The deviation from axial symmetry of the tensor of the doubly-bridging CO is somewhat greater than that of terminally-bound CO, with a 33 - a 22 = 20 ppm. Although the cis and trans forms of (1] 5 -C 5 H 5 ) 2 Fe 2 (C0)4 rearrange on the NMR timescale in solution at 238 K (36-38), motional averaging is unlikely in the solid based on the typical powder pattern of the terminal CO groups and the similarity of the powder patternsof the sample at 300 and 100 K. Thus, the tensors of both doubly- and triply-bridging CO groups are much less anisotropic than those of terminally-bound CO groups. In solid Fe 3 (CO)I 2 , two CO groups doubly bridge a single edge of the Fes triangle, while the ten terminally-bound CO groups are of four types based on slightly different chemical environments (39). powder patterns from terminal We would expect a superposition of and bridging CO groups, as in ('ry 5 -C 5 H 5 ) 2 Fe 2 (C0) 4 , but with a relative intensity of 5 terminal:l bridging. In the spectrum of Fe3 (C0) 12 at 300 Kin Fig. 4, however, we see that this is not the case. There is no indication of a feature that could be attributed to the a 11 component of the powder pattern of the bridging CO groups, in contrast to the spectrum of ('7] 5 -C 5 H 5 hFe 2 (C0) 4 . More definitively, there is intensity over much, but not all of the range expected for terminal CO groups. The spectrum is well fit with a powder pattern of a single theoretical chemical shift tensor with an anisotropy of 327 ppm, compared to 425 ppm for Fe(C0) 5 (14). However, the - 23broadening function that must be convoluted with the chemical shift powder pattern to fit the experi.Inental spectrum is more than five ti.Ines as large as that used for the other carbonyls, and is physically unreasonable. The spectrum of the sample at 100 K is shown in Fig. 4b. Although the spike has been reduced in .intensity somewhat the spectrum has not changed greatly from that of the sample at 300 K. Thus, any intramolecular rearrangement that averages the chemical shift interaction must have a small activation energy. 4. Discussion A. Chemical Shift Tensors of Terminal CO Groups The 13 C chemical shift tensors of all CO groups bound terminally in metal carbonyls are quite similar; Anisotropies lie in the range 320-440 ppm, and are large in comparison with the anisotropies of tensors for carbon nuclei in other compounds (9), which are usually less than 200 ppm. Even when not required to be so by the local symmetry about the carbon nucleus, the tensors of terminally-bound CO groups are nearly axially symmetric, with asymmetry In order to interpret these results in terms of the bonding and electronic structure about the carbon nucleus, we first review the analysis of Mahnke. Sheline and Spiess (MSS) (13) for CO, Ni(C0) 4 and Fe(C0) 5 • They use the conventional theoretical development (40) in which the chemical shift components are divided into diamagnetic and paramagnetic contributions : (4) which, for axial symmetry, are given by: (5 ) -24- (6) (7) (B) where e, m and c are the electron charge, electron mass and speed of light, respectively; r 1 is the length of the vector (xi, y1, zi) between the electron and gauge origin, La and Ll are the electron orbital angular momenta about and orthogonal to the unique axis, z, and !O> and !K> are the ground and an excited state of the molecule with energies E 0 and EK, respectively. The indices K and i extend over all the excited· states and all the electrons of the system, respectively. rhe relative values of O'(d) and O'(p) depend on the choice of gauge, but their sum, u, a physical observable, does not. However, by choosing the carbon nucleus as the gauge origin, uCd) is dominated by the C19 electrons, while the valence electrons, which dominate the bonding, largely determine u (p). Calculations by MSS using self consistent field and semiempirical wavefunctions and Flygare's approximation of Eqs.(5) and (6) above (41) show that this is the case for CO, Ni(C0) 4 and Fe(C0) 5 . We expect this to hold for the other metal carbonyls. MSS point out that for CO, as for all linear molecules, < OILziK> is zero (the wavefunctions are eigenfunctions of ~ and are orthogonal to each other). Therefore u fv> is zero, according to Eq. (7). Since the values of u 11 and u f1> for CO, Ni(C0) 4 , and Fe(C0) 5 are approximately equal, uiv> must also be approximat ely zero for Ni(C0) 4 and Fe( C0) 5 . The electron orbitals in CO are symmetric -25about the C-0 internuclear axis. This symmetry is not significantly reduced upon bonding CO to Ni and Fe in Ni(C0) 4 and Fe(C0) 5 . In physical terms, there is free circulation of the electrons about the C-0 internuclear axis. The data presented here indicate that this is a general result for CO bound terminally in metal carbonyls containing between one and six transition metal atoms. B. Chemical Shift Tensors of Bridging CO Groups and Comparison with Organic Carbonyls For the doubly-bridging CO groups in (175 -C5 H 5 ) 2 Fe 2 (C0) 4 and the triplybridging CO groups in Rh 6 (C0) 16 , the a 11 values are more negative, i.e. are shifted in the paramagnetic direction, by 182 and 266 ppm, respectively, compared to the a 11 values of free and terminally-bound CO. Since we expect any variation in a f1> to be much less than this (13), there is a substantial paramagnetic contribution to a 11 for bridging CO. Thus, the electron orbitals of bridging CO are not symmetric about the C-0 internuclear axis, in contrast to those of terminallybound CO. In physical terms, the circulation of electrons about the C-O internuclear axis is not free for bridging CO. We note that a similar phenomenon occurs in the bonding of cyano and acetylene groups. The difference beween the a 11 values of HCN and CN bound terminally in a Pt compound is only 13 ppm (42). Upon replacing the hydrogens of HCN or C2 H 2 with methyl groups, however, the a 11 values shift 40-50 ppm in the paramagnetic direction (43-46). We may also compare the chemical shift tensors of metal-bridging CO to those of organic carbonyls. Although experimental chemical shift values are not available for H2 CO, reliable theoretical calculations have been made (47), and are consistent with the experimental values for CH3 CHO and (CH 3 ) 2 CO (46). In addition to the principal components of the chemical shift tensor, the -26calculations also provide their associated principal directions which are needed to fully define the chemical shift tensor. The association of principal values and directions for terminal and triply-bridging CO groups is unambiguous due to the axial sym.metry about the c-o internuclear axis. The chemical shift values of CO, H 2 CO and the triply-bridging CO groups of Rb.a(C0) 16 are presented in Fig. 5 as a correlation diagram. Principal com- ponents along a given direction with respect to the C-0 internuclear axis are connected by dashed lines. Because of the reduced symmetry in doubly-bridging CO, its principal components cannot be connected unambiguously in this diagram. It seems likely, though, that the least negative component corresponds to the chemical shift parallel to the C-0 internuclear axis as interminal and triply-bridging CO. In any case, the chemical shift components of, and therefore the bonding in, H 2 CO are much different from those in free and metal-bound CO. Thus, despite the geometric similarity of doubly-bridging CO and H 2 CO, and the similar CO bond distances and stretching frequencies of triply-bridging CO (48,49) and H 2 CO (50,51), the bonding in H 2 CO is not a good model for the bonding of CO to metals. C. Variations in Chemical Shift with Transition Metal Variations in the 13 C chemical shift tensors among the carbonyls of different transition metals are most clearly demonstrated through the a 1 values, as shown in Fig. 6. Although the error in any particular measurement is relatively large, two trends are evident: decreasingly paramagnetic values of a l both in going across and down in the transition series. We also note that there are values both above and below the value of a l for free CO. To explain these trends, we again review the analysis due to MSS. - 27- As with the u 11 component, variations in a 1 are due primarily to variations in the paramagnetic term. In free CO, there is only one low-lying excited state of the proper symmetry to contribute to ulP). In a molecular orbital formalism, this excited state is formed from the ground state by promoting an electron from the 5a to the 2rr· orbital. MSS argue that to the extent that the 2rr orbital is stabilized by backhanding from the d orbitals of the metal, the 5u ~ 2rr excitation energy will be lowered and the paramagnetic contribution to u 1 will be increased. Based on the increasingly paramagnetic shifting of CTl observed in the series CO - Ni(C0) 4 < Fe(C0) 5 , they conclude that rr-backbonding is significant in Fe(C0) 5 but much less so in Ni(C0) 4 . In contradiction to the above analysis, however, the 2rr orbital would be destabilized, not stabilized, through interaction with filled d orbitals. It would be better to think of the metal d orbitals, which are intermediate in energy between the 5a and 2rr orbitals of CO, as stabilizing the excited state formed from the 5a ~ 2rr transition. The excitation energy would thus be lowered and the paramagnetic contribution to u _lP> would be increased by the presence of the d orbitals, as is observed for Fe( C0) 5 . This stabilization of the excited state would decrease as the d energy level decreases. Photoelectron spectra of carbonyls containing one (1,2.52) and more than one (53,54) metal atom indicate that the d energy levels generally decrease with increasing atomic number of the metal within a row or column. Therefore, the above arguments may be used to explain the decreasely paramagnetic value of a l that is observed for increasing atomic number of the metal for the carbonyls in general. Since lower d levels would also lead to increased a bonding and decreased rr bonding, the a l values, and thus the a values for terminal CO, may be related to these fundame~tal properties, as concluded by MSS. The chemical shift may also be correlated with other properties that are determined -28- by the extent of u and rr bonding such as metal-CO bond energies and CO and metal-C stretching force constants. Although the variation in a l among the transition metals was explained by differences in the stabilization of the excited state formed from a 5u ~ 2rr transition, this is highly simplified. In metal carbonyls, the metal and CO orbitals are mixed and form a new set of orbitals based on the particular molecular symmetry. There are then many excited states that contribute to a l· In addition, it is not obvious to what extent the matrix elements < OIL~K> and < KIL1/r 3 IO> change upon bonding CO to a metal or among the various metal carbonyls. The simplified approach of effectively considering the bonding to the metal as a perturbation of the bonding in CO is supported by the small magnitude of the trend to be explained. In addition, the trend holds in the cases where there is no change in molecular symmetry: for M(C0) 6 , where M = Cr, Mo and W, and for Ru3 (C0) 12 and Os 3 (C0) 12 . Also, calculations show that the largely metal-derived orbitals contribute ~ 5 ppm to the chemical shift values of Cr(CO)e and W(CO)e (55). However, by considering only the stabilization of the excited state from a 5a ~ 2rr transition by the metal d orbitals, a l of metal-bound CO is always predicted to be more paramagnetic than a l of free CO. However, the a l values of Os 3 (C0) 12 , Rh 6 (C0) 16 and Ir4 (C0) 12 are less paramagnetic than that of free CO. Thus, a more sophisticated analysis is necessary to relate the a 1 values of metal carbonyls to that of free CO to determine the extent of rr backhanding in the metal carbonyls. D. Intramolecular Rearrangements of Metal Carbonyls in the Solid State A final aspect that we would like to consider is that of intramolecular rearrangements in the metal carbonyls. Of the carbonyls studied here, only in Fea(COh 2 is the chemical shift interaction motionally averaged. In previous -29- studies, Fe(CO)~. but neither Ni(C0) 4 (14) nor Rh 2(C0) 4 Cl 2 (24), was shown to be undergoing intramolecular rearrangement. Most rearrangements would change the angle of the external field with respect to the C-0 internuclear axis, averaging the chemical shift interaction and narrowing the spectrum. Thus, there is little intramolecular rearrangement of the metal carbonyls in the solid state, in contrast to those of dissolved carbonyls containing one (8) and more than one (6. 7) metal atoms. This is probably the result of intermolecular packing forces in the crystal lattice. An example of a rearrangement that cannot be taking place in the carbonyls in the present study, ignoring Fe 3 (C0) 12 for the moment, at rates ~ 10 kHz, is twisting of three or four CO groups with respect to the rest of the molecule, e .g. : A~ t £ ' c \ / I I k'B Other examples would include most exchanges of CO groups among metal atoms in which the CO groups change from terminal to bridging positions or vice versa in the intermediate or transition state, e.g.: A mechanism that would not affect the NMR spectrum would be cycling of the axial CO groups on one side of the Ru 3 triangle in Ru 3 (C0)12: ~B c B~ -30However, in only one metal carbonyl, [(1]0 -C5 H5 )RhCO] 3 , has a similar mechanism been clearly demonstrated (56). The final class of rearrangements that will be discussed involves motion of the metal cluster within the "shell" formed by the CO groups (57). In the liquid state, it is difficult to distinguish this mechanism from the above types in which some CO groups move with respect to the rest of the molecule since both redistribute CO groups among the various sites of the molecule. Motion of the metal cluster within the CO shell, though, would leave the spectrum of the solid relatively broad, whereas the other mechanisms would lead to substantial narrowing. Even the partial narrowing expected in the former case would be enough to discount it from occurring in most of the metal carbonyls, with the exception of Fe 3 (C0) 12 . In Fe 3 (C0) 12 , rotation of the Fe 3 cluster would exchange CO groups among terminal and b~idging positions. The weak temperature dependence of the powder pattern indicates that the rate of rearrangement is at least comparable to the linewidth. In the fast exchange limit, this mechanism would not change a l substantially, but the value of a 11 observed would become"' [5( +70) - 179] /6 = +28 ppm, based on the chemical shift values of Fe( C0) 5 and the bridging CO groups of (1]5 -C5 H5 ) 2 Fe 2 (C0) 4 . Small librational motions and deviations from idealized symmetry (C2 v for the molecule, icosahedral for the carbonyl shell) would result in further narrowing of the spectrum and smoothing of the shoulders, consistent with the observed spectrum. This mechanism is further supported by the fact that the Fe-Fe distances may either expand or remain the same during the reorientation of the Fe 3 cluster with little or no change in the positions of the CO groups (57). Intermolecular interactions which would be expected to hinder CO motion in the solid state would have little effect on this mechanism. A crucial test of this mechanism would be motional narrowing of the 57 Fe NMR - 31 - spectrum. 5. Conclusions From measurements of the principal components of the 13C NMR chemical shift tensors of a series of metal carbonyls, we find that: 1. The chemical shift tensors of CO groups bound terminally in metal carbonyls are highly anisotropic (380 ± 60 ppm) and axially symmetric. They vary only slightly with the transition metal. The electron orbitals are symmetric about the c-o internuclear axis. 2. The tensors of CO groups that bridge two or more metal atoms are much less anisotropic due to significant asymmetry in the electron orbitals about the C-0 internuclear axis. This suggests that there is a greater perturbation of the electronic structure of CO in bridged than in terminal bonding to a metal. However, the bonding in metal-bridging CO is distinct from that in organic carbonyls, which is not apparent from a consideration of isotropic chemical shifts or vibrational frequencies. 3. There is little or no intramolecular rearrangement of the metal carbonyls in the solid state at frequencies greater than 10kHz, except in Fes(CO)t2· We have seen that it is possible to draw conclusions about the bonding and electronic structure of metal carbonyls from chemical shift measurements. To put these on a quantitative basis requires wavefunctions and energies of the ground and excited states that are more accurate than those presently available for metal carbonyls. The present results also serve as important references for interpreting 13 C NMR spectra of CO that is adsorbed or that has reacted on metal surfaces (24). In particular, terminal and bridging CO should be easily distinguishable from each other and from CO that has reacted with hydrogen or - 32- some organic group , based on their chemical shift powder patterns and/or isotropic chemical shifts. Acknowledgments This study was supported by the Office of Naval Research under contracts N00014-77-F-0008 and N00014-79-F-0014. Helpful comments were offered by Drs. W. H. Weinberg, T. M. Duncan, W. A. Goddard and S. I. Chan. . -33- References 1. E.g., see A. H. Crowley. Prog. Inorg. Chern. 26, 45 (1979). 2. P. S. Braterman, Metal Carbonyl Spectra (Academic, New York, 1975). 3. 0. A. Gansow and W. D. Vernon, in Topics in Carbon-13 NMR Spectroscopy, edited by G. C. Levy, (Wiley-Interscience, New York, 1976), Vol. 2, Ch. 5. 4. B. E. Mann, Advan. Organometal. Chern. 12, 135 (1974). 5. A. F. Schreiner and T. L. Brown, J . Am. Chern. Soc. 90, 3366 (1968) . 6. E. Band and E. L. Muetterties, Chern. Rev. 78, 639 (1978) . 7. J. Evans, Advan. Organometal. Chern. 16, 319 (1977). 8. R. D. Adams and F. A. Cotton, in Dynamic NMR .Spectroscopy, edited by L. M. Jackman and F. A. Cotton, (Academic Press, New York , 1975), Ch. 12. 9. M. Mehring, High Resolution Nl.JR Spectroscopy in Solids, Vol. 11 of NMR: - Basic Principles and Progress (Springer, New York, 1976) . 10. U. Haeberlen, High Resolution NMR in Solids: Selective Averaging, Advan. Magn. Res on. Suppl. 1 (Academic, New York, 1976). 11. R. W. Vaughan, Ann. Rev. Phys. Chern. 29, 397 ( 1978). 12. H. W. Spiess, Chern. Phys. 6, 217 (1974). 13. H . Mahnke , R. K. Sheline and H. W. Spiess, J. Chern. Phys . 61, 55 (1974). 14. F. A. Cotton, A. Danti, J. S. Waugh and R. W. Fessenden, J. Chern. Phys. 29, 1427 ( 1958). 15. R. Bramley, B. N. Figgis and R. S. Nyholm, Trans . Faraday Soc. 58, 1893 (1962) . 16. R. W. Vaughan, D. D. Elleman. L. M. Stacy, W. K. Rhim and J. W. Lee, Rev. Sci. - 34- Instrum. 43, 1356 (1972) . 17. M. E. Stoll, A. J. Vega and R. W. Vaughan, Rev. Sci.Instrum. 48, 800 (1977). 18. L. 0. Brockway, R. V. G. Ewens, and M. W. Lister, Trans. Farad . Soc. 34, 1350 (1938). 19. 0. A. Gansow, B. Y. Kimura, G. R. Dobson and R. A. Brown, J . Am . Chern. Soc. 93, 5922 (1971). 20. B. E . Mann, J. Chern. Soc., Chern. Commun. 1173 (1971) . 21. P. C. Lauterbur and R. B. King, J. Amer. Chern. Soc. 87, 3266 (1965). 22. P . Chini, S. Martinengo, D. J. A. McCaffrey and B. T. Heaton, J. Chern. Soc., Chern. Commun. 310 (1974). 23 . T. M. Duncan, J. T. Yates, Jr. and R. W. Vaughan, J. Chern. Phys. 73, 975 ( 1980). 24. A. A. V. Gibson, T. A. Scott and E. Fukushima, J. Magn. Res on. 27, 29 ( 1977). 25. K.. Jackowski and W. T. Raynes, Mol. Phys. 34, 465 (1977) . 26. R. Mason and A. I. M. Rae, J. Chern. Soc (A), 778 (1968) . 27. E. R. Corey and L. F. Dahl, Inorg. Chern. 1. 521 (1962) . 28. S . H . H. Chasten and F . G. A. Stone, J. Chern. Soc. (A) , 500 (1969). 29. A. Forster, B. F. G. Johnson, J. Lewis, T. W. Matheson, B. H. Robinson and W. G. Jackson, J. Chern. Soc., Chern. Comrnun., 1042 (1974). 30. E. R. Corey, L. F. Dahl and W. Beck, J. Am. Chern. Soc. 85, 1202 ( 1963). 31. B. T. Heaton, A. D. C. Towl, P. Chini, A. Furnagilli, D. J. A. McCaffrey and S. Martinengo, J . Chern. Soc., Chern. Comrnun., 523 (1975). 32. 0 . S. Mills, Acta Crystallogr. 11, 620 ( 1958). -3533. R. F. Bryan, P. T. Greene, D. S. Field and M. J. Newlands, J. Chern. Soc., Chern. Commun., 1477 (1969). 34. R. F. Bryan and P. T. Greene, J. Chern. Soc. (A), 3064 (1970). 35. 0. A. Gansow, A. R. Burke and W. D. Vernon, J. Am. Chern. Soc. 94, 2550 (1972). 36. D. C. Harris, E. Rosenberg and J. D. Roberts, J. Chern. Soc., Dalton Trans., 2399 ( 1974). 37. R. D. Adams and F. A. Cotton, J. Am. Chern. Soc. 95, 6589 (1973). 38. F. A. Cotton and J. M. Troup, J. Am. Chern. Soc. 96, 4159 (1974). 39. N. F. Ramsey, Phys. Rev. 86, 243 (1952); 78, 699 (1950). 40. W. H. Flygare and J. Goodisman, J. Chern. Phys. 49, 2314 (1968). · 41. M. E. Stoll. R. W. Vaughan, R. B. Saillant and T. Cole, J. Chern. Phys. 61, 2896 (1974). 42. F. Millett and B. P. Dailey, J. Chern. Phys. 54, 5434 (1971). 43. S. Mohanty, Chern. Phys. Lett. 18, 581 (1973). 44. S. Kaplan, A. Pines, R. G. Griffin and J. S. Waugh, Chern. Phys. Lett. 25, 78 (1974). 45. A. Pines, M.G. Gibby and J. S. Waugh, Chern. Phys. Lett. 15, 373 (1972). 46. R. Ditchfield, quoted in B. R. Appleman and B. P. Dailey. Ad van . .Magn. Reson. 7. 231 (1974). 47. P. Chini. G. Longoni and V. G. Albano, Advan. Organometal. Chern. 14, 285 ( 1976), and references therein. 48. F. A. Cotton and G. Wilkinson, Advanced Inorganic Interscience, New York, 1972), 3rd ed., p. 693. Chem:isry, (Wiley- - 3649. R. B. Lawrence and M. W. P. Strandberg, Phys. Rev. 83, 363 (1951) . 50. E. S. Ebers and H . H. Nielsen, J. Chern. Phys. 6, 311 (1938) . 51. C. Furlani and C. Cauletti,. in Structure and Bonding 35: New Theoretical Aspects, edited by J . D. Dunitz et al. (Springer-Verlag, New York, 1978), p. 119. 52. E. W. Plummer, W. R. Salaneck and J. S . Miller, Phys. Rev. B 18, 1673 (1978). 53. J. C. Green, E. A. Seddon and D. M. P. Minges, J. Chern. Soc., Chern . Commun. 94 (1979). 54. D. Cozak and S. Butler, Can. J. Chern. 55, 4056 (1977) . 55. R. J. Lawson and J. R. Shapley, J. Am. Chern. Soc. 98, 7433 ( 1976). 56. B. F . G. Johnson , J . Chern. Soc., Chern. Cornmun., 703 ( 1976). Table 1 Principal Components, Isotropic Values and Anisotropies of the lJC NMR Chemical Shift Tensors of Metal Carbonyls(a,bJ 0 Cr(CO) 6 +70 -353 -353 6 +75 -343 -343 6 +71 -324 -324 Ru (co) 12 3 Os (co) 12 3 lr (C0) 4 12 Rh (co) 6 16 +63 -319 -348 +55 -292 ~292 +53 -277 -277 Mo(C0) W(C0) {a) solid -203 -201 0 Ref. soln (a so 1n ) -212 423 20 -202 417 20 -192 395 20 -200 397 30 -176 347 30 -167 322 terminal .+80 -305 -315 ~181 bridging -102 -296 -296 -231 390 194 terminal +85 -354 -354 -211 444 bridging -179 -309 -328 -272 138 36,37 Fe (CO) 12 3 +6 -316 -325 -212 327 30,36 co - +90 (d)- 316 (d)- 316 (d) -181 406 26 . 5 (n -c 5H ) 2 Fe 2 (co) 5 4 32 4 Ni(C0) {~) +69 -326 -326 -194 -191 395 21 Fe(CO) +70 -355 ~355 -213 -212 425 22 (e) 5 Rh (CO) C1 ( fJ 23 -180 402 +99 -299 -306 -186 2 4 2 {a)Values are in ppm, relative to TMS. {bJUncertainties are ±15 ppm for the components and ±30 ppm for the anisotropies. {a)Where values of 5 are not listed; they are assumed to be equal to 5soln (see text). (dJCalculated using the estimate for ucr in the absence of motionai averaging (25). (e) Ref. (14). {f/R.e f. (24) . -38Figure Captions Figure 1: 15 C NMR spectrum of polycrystalline Mo(C0) 6 at 300 K. The solid line is a .fit of a theoretical chemical shift powder pattern to the experimental points. The principal components of the chemical shift tensor are indicated, along with the isotropic chemical shift. The separation between points is 4 ppm. Figure 2: 15 C NMR spectrum of polycrystalline Ru 3 (C0) 12 at 300 K. The theoret- ical fit (solid line) is for a single chemical shift powder pattern. Figure 3: 15 C NMR spectra of polycrystalline: (1'J~-C~H 5 ) 2 Fe 2 (C0) 4 at 300 K. (a) Rh6 (C0) 16 and {b) The spectra were deconvoluted as described in the text into the contributions from the terminal (---) and bridging (···)CO groups in each compound. Figure 4: 13 C NMR spectra of Fe 3 (C0) 12 at: (a) 300 K and {b) 100 K. The princi- pal components for Fe( C0) 5 are indicated. While there is considerable spectral intensity near a l· there is little intensity near a 11 at both temperatures. Figure 5: Principal components of the chemical shift tensor for various types of CO bonding. Components associated with the same direction of the external magnetic field, H0 , with respect to the C-0 internuclear axis are connected by dashed lines. The components for the triplybridging CO groups in Rh 6 (CO)t 6 are connected in the diagram. The components of the doubly-bridging CO groups in (1] 5 -C 5 H 5 )2Fe2(C0)4 ( ···) cannot be associated with directions unambiguously from the powder patterns alone and are not connected. Figure 6: Values of the a l components of terminally-bound CO in metal carbonyls taken from Table 1. In cases of nonaxial symmetry, the value - 39- of (u 22 + u 33 )/2 was used. The value for Fe is that of Fe(C0) 5 . The trend to lower magnitude of ,. l with increasing atomic number across and down the transition metals is evident. -40- E 0.. 0.. r(') ~ r(') I O(J) II b-1 ~-~~~~~~~ E 0.. 0.. r(') ~ @~ 11-- 0 row o> C\J11-<! 0 N I _j II w lb 0::: <.0 --... 0 u II 0 ~ Figure 1 b N (D li c:: OQ ...... 1-::j + 200 c 0 3 Gjl Ru • 0 c 0 co co -200 ! o- =-201 ppm = -348 ppm -400 1rcr33 o-22 =-319 ppm o-,(ppm), RELATIVE TO TMS c 0 (C0) 12 I ....... I ~ -42- a - .. · ... .... .... b / / + 200 0 -200 / .. : -400 a-, (ppm), RELATIVE TO TM S . - Figure 3 -43- A ·'• • -355 ppm ~ I • I I •'.,I a 300K *I * lOOK b +200 0 -200 -400 r;-, (ppm),RELATIVE TO TMS Figure 4 -44- 0 u C\.J I 0 u 01 c 01 -o ~ ....0 0 _u 0u~ w·E Q) ~ ~ ~ Q) ~ 0 0 0 0 0 C\J (\j I + S lAJ 1 01 3 1\ I_L\1l3 d ' ( w d d) ' fl Figure 5 .0 0 ~ I -45- -350 Cr (f) ~ ~ 0 ~ w -325 > _____________ jree ~ co <( _j w 0:: ... E -3oo 0... 0... -275 6 8 10 METAL VALENCE ELECTRONS Figure 6 -46- Part n: Surface Chemistry of Rhodium Supported on Alumina -47- Chapter 3: Introduction -48In Part I of this thesis, an example of the use of NMR spectroscopy in elucidating some details of the bonding in a metal complex was presented. An important application of such knowledge is in understanding the functioning of metal catalysts, since the catalytic properties of the metal are determined largely by how they bind the reactants at their surfaces. Of course, it would be more relevant to study the surface-adsorbed species itself. This is a difficult task. The proposed form of the adsorbed species, which is a reaction intermediate, must be consistent with the kinetics of the overall reaction, though the kinetics rarely, if ever, imply a specific intermediate. More definitive are experiments in which the reactants are isotopically labelled and the distribution of the label in the product determined (1). For example, in a recent study of methanol synthesis from CO and H over rhodium supported on 2 13 16 12 18 titania, c o and c o were used as reactants and the major products were 13C H 16 OH an d 12CH3180H (2). 3 The absence of scrambling of the labels implies that the carbon and oxygen atoms bound in the methanol must have been bound to each other in all the intermediate steps of the reaction. However, the exact nature of the reaction intermediates is still not known. It is desirable to observe the reaction intermediate directly, such as by spectroscopic means. Unfortunately, as remarked in Chapter 1, the application in studies of surface species of most of the physical methods used in characterizing molecular metal complexes is hindered by insufficient sensitivity or by particular characteristics of the catalyst material. By far the most widely used ·spectroscopic method in studies of species adsorbed on supported metals has been infrared -49spectroscopy (3-S). A given functional group often has characteristic vibrational frequencies that are resolvable from those of other functional groups and thus can be used to identify the species. more specific information can be obtained. Often, For example, if a surface hydroxyl group is involved in hydrogen bonding can be determined by the frequency of its OH stretching mode (3,4). A drawback of infrared spectroscopy is that the intebsities of the peaks in a spectrum often cannot be related to the concentration of the species responsible for that peak {3). Thus, it is highly beneficial to have an independent means of quantifying the effects seen in the infrared spectrum, such as through mass balances or by another spectroscopic technique, such as NMR (6). In Part II of this thesis, the surface chemistry of rhodium supported on alumina is studied using infrared spectroscopy and quantitative measurements of the gases taken up by and evolved from the samples during the various procedures. Supported Rh is an especially active catalyst for olefin hydrogenation (7,8) and is also active for hydrogenation of CO to methane (9) and oxygenated hydrocarbons (10), reduction of NO, and oxidation of CO (11). Supported Rh is unusual in binding two CO (or NO (12,13)) groups to a single Rh atom (9), in addition to binding .CO in linear and bridging modes on clustered Rh, similarly to other Group VIII transition metals (14). These sites are distinguishable using infrared spectroscopy (9). In the first study, the changes in these three CO adsorption sites upon heating in CO, co 2 , and H2 will be .examined. In another study, the · · lnteractlon o f H S , a common catalyst poison, with Rh/alumina is exam2 ined. Supported Rh was shown to be especially resistant to poisoning -50by SOz (11), and it is of interest to determine if this is related to its unusual surface chemistry. The effect of H S on the reactivity of 2 adsorbed co toward co(g) and 02 is also examined. -51- References 1. A. Ozaki, Isotopic Studies of Heterogeneous Catalysts, Academic Press, New York, 1977. Chern.~' 937 (1981). 2. A. Takeuchi and J. R. Katzer, J. Phys. 3. L. H. Little, Infrared Spectra of Adsorbed Species, Academic Press, New York, 1966. 4. M. L. Hair, Infrared Spectroscopy in Surface Chemistry, Dekker, New York, 196 7. 5. G. L. Haller, in Spectroscopy in Heterogeneous Catalysis, p. 19, Academic Press, New York, 1979. 6. T. M. Duncan, J. T. Yates'· Jr. and R. w. Vaughan, J. Chern. Phys. J.l., 975 (1980). c. A. Schmit and L. L. Van Reijen, Advan. Catal. .!Q_, 61 (1945). 7. G. 8. 0. Beeck, Rev. Mod. Phys. ll_, 61 (1945). 9. A. 10. P. C. Ellgen, W. J. Bartley, M. M. Bhasin and T. P. Wilson, in c. Yang and c. w. Garland, J. Phys. Chern. ~' 1504 (1957). Hydrocarbon Synthesis from CO and Hydrogen, E.LKugler, ed., p. 147, Academic Press, New York, 1979. 11. J. C. Summers and K. J. Baron, J. Catal. ·s7, 380 (1979). 12. H. Arai and H. Tominaga, J. Catal. ~' 131 (1976). 13. H. C. Yao, s. Japar and M. Shelef, J. Catal. SO, 407 (1977) · 14. N. Sheppard and T. T. Nguyen, Advan. Infrored Raman Spectr. ~' 67 (1978). -52- Chapter 4: ''Deactivation of . CO Adsorption Sites upon Heating Rh/ Alumina in CO. C02 • H2 and 0 2 " -53Abstract The behavior of alumina-supported Rh upon heating in the presence of CO, C02 , 0 2 and H 2 was studied using infrared spectroscopy and quantitative measurements of the gases taken up and evolved during various procedures. The loss in the capacity to adsorb CO after heating to 525 K increases in the order 02, H2 vacuum < C0 2 < CO. Upon heating in CO, some CO is oxidized to C02 with oxygen from the surface, while the dicarbonyl-forming Rh1 is reduced to Rh 0 . The Rh0 agglomerates, accounting for the substantial loss in capacity to adsorb CO. Upon heating in C0 2 , the dicarbonyl-forming Rh1 is also deactivated. There is little loss in capacity to adsorb CO after heating in H2 , 0 2 or vacuum. -541. Introduction There has been much interest in the fundamental interactions of small molecules on supported Rh. In a classic infrared study of CO on alumina- supported Rh, Yang and Garland (1) identified three modes of adsorption: 0 c· \ c 0 0 0 c I c I Rh ' / -Rh-Rh- -Rh- which we will refer to , respectively, as dicarbonyl, linear and bridged. The linear and bridged modes occur on two- or three-dimensional Rh clusters, and are common to most other supported and unsupported. Group VIII transition metals (2). Only on supported Rh, Ru (3) and Os (4-6) has CO been shown to adsorb in the dicarbonyl mode. The Rh dicarbonyl appears to be isolated from the clustered Rh (7.:..10), though this is still controversial (11). The. oxidation state of the dicarbonyl Rh also appears to difier from that of the clustered Rh . Whereas the clustered Rh is almost surely Rh0 , several studies suggest that the dicarbonyl Rh is Rh1 (Rh I+). The intensitit:s of the infrared bands of the dicarbonyl are not changed by treatment of alumina-supported Rh with 0 2 at 300 K either before or after exposure to CO or even by treatment with 0 2 at 473 K before exposure t o CO. This suggests that the dicarbonyl Rh may be oxidized to some extent. However, the bands are lower in frequency than those of CO in Rhm complexes (1.10.13). Alumina-supported Rha(COha that had been treated under vacuum or with H2 above 425 K. a procedure that produces a surface that is similar to that obtained by reducing an alumina-supported Rh salt (14), has been studied using X-ray photoelectron spectroscopy (15). The results are consistent with the existence of Rho and/or Rh1 species, but not Rhn or Rhm species . The absence of strong ESR signals (9.16,18) and the linewidth of 13 C NMR spectra (9) from -55Rh/ alumina with adsorbed CO indicate that there are few unpaired electrons. Thus, there can be few magnetically isolated dicarbonyls containing Rhrr or Rh0 . We note that the Os species that forms supported di- and tri-carbonyls is also oxidized ( Osii) ( 4.18 ). The dicarbonyl of Ru, the only other metal shown to form a supported multicarbonyl species, is probably oxidized also, since its infrared bands are very similar in frequency to those of the osii dicarbonyl (3). However, the results of temperature programmed reduction of highly dispersd Rh/alumina by H2 have been interpreted as indicating reduction to Rh 0 (19). The Rh dicarbonyl species also ditrers in reactivity from CO adsorbed on the clustered Rh. The CO of the dicarbonyl exchanges with gas phase CO more readily (20) and is more easily desorbed upon evacuation (1,7-11,21), but is less easily removed by 0 2 (1.10.13) and NO (22) than i~ the CO on the clustered Rh. The dependence of various Rh-catalyzed reactions on the dispersion and oxidation state of the Rh (23-27) may often be based on such reactivity di.trerences. It has been noticed that the dicarbonyl species is gradually removed both during adsorption-desorption-readsorption cycles (1,7,11) and in in situ infrared studies of CO hydrogenation over supported Rh catalysts (21). Fujimoto et al. (21) have demonstrated that the first step in the decomposition of dicarbonyl species upon heating is loss of a CO. Whether t he remaining site is especially reactive or inactive, and whether this change is reversible or not, is not clear. In the present study, changes in the sites for CO adsorption on Rh/ alumina upon heating are investigated. We used infrared spectroscopy and quantitative measurements of the CO taken up and the gases evolved while heating between 300 and 525 K. Heating to 525 K was found to cause a substantial loss in capa- city to adsorb CO. Employing estimates of CO and Rh among the various adsorp- -56tion modes based on the NMR results of Duncan et al. (9), a model to account for this deactivation is proposed. By heating Rh/alumina in the presence of C0 2 , 02 or H2 instead of CO, a strong dependence of deactivation on the gas phase is observed. 2. Experimental Procedures The 2.0% by weight Rh dispersed on alumina was prepared as described previously (7). Briefly, RhC1 3 ·3H 2 0 is dissolved in a water:acetone ( 1:1 0) suspension of alumina and sprayed onto a CaF 2 disc preheated to 365 K. The disc, with a coverage of .... 11 mg of sample per cm2 is mounted in a stainless steel infrared cell with CaF 2 windows. The cell is then attached to a stainless steel vacuum system and heated under vacuum to 450 K. The sample is reduced with three charges of H2 (Matheson, 99.9995%) at 425 K and 200 Torr, evacuated at 450 K for 12 hr and cooled to 300 K. The procedure for heating the samples in various gases was to till the cell with - 50 Torr of the gas, close ofi the cell, heat at the desired temperature for 12 hr and then cool to 300 K. The alumina samples were prepared exactly as were the Rh/ alumina samples, with the exception that RhCl:r 3H 2 0 was not added to the water:acetone suspension of the alumina. Quantitative measurements of the gases taken up (removed from the gas phase) by and evolved from the samples were made using either the infrared samples, or, for greater precision, 0.2 g samples packed loosely in the bottom of glass tubes and pretreated in situ as above. The measurements were made in the constant volume system at pressures of a few Torr to ± 0.01 Torr with a capacitance manometer. The amounts of CO, C0 2 and H2 in the gas phase were conveniently monitored manometrically by condensing the C0 2 in a glass tube at 77 K and condensing CO and C0 2 on dehydrated alumina at 77 K. The methane formed on heating to 525 -57K was determined by mass spectrometry or infrared spectroscopy. The concentration of all other gases was less than 5 x 1 o-e mol per gram of sample, as shown using mass spectrometry. Transmission infrared spectra were obtained with a Perkin-Elmer 180 grating spectrometer with a linear absorbance scale. The double beam mode was used without any compensating sample in the reference beam. The spectral resolution at 2000 cm- 1 was set at 2.6 cm- 1 . All spectra were recorded with the sample at-310K. The 19 CO (Merck Isotopes, 90.% enriched) and the unenriched CO (Matheson, Research) were obtained in glass breakseal bulbs and used without further purification. The 0 2 (Matheson, Research) was passed through a trap at 77 K before use. The C0 2 (Matheson, Bone Dry) was purified by three freeze-pumpthaw cycles and sublimed at 195 K immediately before use. The a lumina was "Alon-C" (Cabot), a fumed alumina with a particle size of ..... 15 nm. Two lots with BET surface areas of 90 and 115 m 2 I g were used and are designated as alumina-90 and alumina-115. 3. Results A. Infrared Spectra Rh/ Alumina Heated in CO Qualitative changes in the types and amounts of CO adsorption sites present after heating Rh/ alumina in CO above room temperature were followed using infrared spectroscopy. Shown in Fig. 1 is the spectrum of Rh/ alumina-90 exposed to 50 Torr of CO at 300 K. The peaks, accordin~ to the original assignment (1), for which there is now general agreement (7-14,21,22), are due to t;b.e symmetric and asyp:1metric CO stretches of the dicarbonyl species at 2103 em - l and 2034 cm- 1 , respectively, and the stretches of CO bound to one and more than one Rh atom of a Rh cluster at 2070 and 1875 em - l , respectively. The -58- rotational structure about the missing Q branch at 2143 cm- 1 of gas phase CO is also seen. The ratio of CO molecules adsorbed per Rh atom in this sample is 1.0. The peak assignments for the infrared spectra in Figs. 1-4 are summarized in Table 1. After isolating the sample, heating at 425 K for 12 hr, and cooling to 300 K, the intensity ·of the dicarbonyl doublet has decreased to ,.... 35% of that before heating. There is little change in the peaks due to linear and bridging CO. When 19 CO adsorbed on Rh/alumina is heated to 425 K, a broad peak at . . . 2280 cm-1 from adsorbed 13 C0 2 appears. The corresponding peak of adsorbed 12 C0 2 is not easily seen after heating in 12 CO because of spectral interference by atmospheric C0 2 . After heating this sample at 525 K for 12 hr and cooling to 300 K, the dicarbonyl doublet is absent, while again the bands from linear and bridging CO have not changed greatly. There is a further increase in the 13 C0 2 band at 2280 em - 1 when 19 CO is used. This band is removed upon evacuation at 300 K, indicating that the C0 2 must be weakly adsorbed. After evacuation, treatment with H 2 and reevacuation, all at 450 K, which was the procedure used to reduce the RhCls dispersed on the alumina, there are no bands from adsorbed CO. After reexposing the sample to 50 Torr of CO at 300 K, the bands are similar in intensity, but \.. shifted somewhat higher in frequency compared to those after heating at 525 K. The higher frequency peak of the dicarbonyl appears at 2110 cm- 1 , but is much weaker than before heating. Similar results are obtained when the sample is heated in 100 Torr at 0 2 at 450 K before the H 2 treatment in the regeneration procedure. 1 At frequencies outside the CO stretching region (2200-1700 cm- ) there is no change upon adsorption of. CO at 300 K. Only small increases were observed -59- near 1475 and 1280 cm- 1 after heating at 425 K. Upon heating at 525 K, however. there are several changes. The broad peak at ,..,. 3700 - 3000 em - l at 300 and 425 K is due to the OH stretch of hydroxyl groups of the alumina surface. After heating to 525 K, the reduced intensity on the high frequency side and increased intensity on the low frequency side of this peak, along with the appearance of a peak at 1625 em - 1 • indicates that there has been some combination of the hydroxyl groups to form molecular water. The peaks at 3019 and 1307 cm- 1 • as well as the rotational structure associated with the former peak. are assigned to gas phase CH 4 . Since no 13 CH4 is formed when 13 CO adsorbed on Rh/ alumina is heated, it is suggested that some of the acetone used in the sample preparation is retained by the sample and decomposes upon heating. No CH stretching bands were observed before the sample was heated to 525 K. In contrast to the H2 0 and CH 4 bands, the other peaks below 1700 cm- 1 in Fig .. 1 remain after the sample is evacuated and treated with 0 2 or H2 at 450 K. Their identification is facilitated by reference to the infrared spectra of alumina heated in the presence of CO or C0 2 at 425 and 525 K. which are shown in Fig. 2. Upon exposure of alumina to either CO or C02 at 300 K, only a small amount of gas is taken up (~ 2 J.Lmol/ g) and there is little change in the infrared spectra. Heating alumina in CO and C0 2 does cause changes in the infrared spectra. The most prominent bands present after heating alumina at 525 Kin CO are at 1594, 1395 and 1383 cm- 1, and are assigned to a surface formate species (28,29). Peaks at 1470 and 1455 cm- 1 are due to carbonate species, referred to variously as free. basic or ionic (30). The peaks at 1649 and 1231 em - l are due to carbonate species involved in more covalent bonding, probably the bicarbonate ion in this case (31). Heating alumina in C0 2 produces the same three species, but more of the carbonate and bicarbonate and less of the formate than are formed by heating in CO. -60- Referring back to the spectrum in Fig. 1 of Rh/ alumina heated to 525 Kin CO, we see that of these three species, only the ionic carbonate is present. The band at- 1415 cm- 1 lies at the lower end of the range of asymmetric OC0 2 stretching frequencies of ionic carbonates, and may indicate that there is yet another type. The bands at "' 1365 and 1280 em -l, unaccompanied by corresponding bands above 1450 cm- 1 are probably not due to a carbonate species, but may be from the COH bending and/or CHx deformation modes of an acetone residue . B. Infrared Spectra of Rh/ Alumina Heated in C02 There is no change in the infrared spectrum in Fig. 3 upon exposure of Rh/ alumina-90 to 55 Torr of C0 2 at 300 K. There has been much controversy about the conditions under which C0 2 will dissociate on Rh (32.33), and relevant variables include temperature, exposure, dispersion of the Rh, and support material. There is also little change in the spectrum after heating the sample in C0 2 at 425 K. After heating for 12 hr at 525 K and cooling to 300 K, though, substantial intensity appears in the CO stretching region. The presence of only two peaks, with maxima at 2044 and 1855 em -l, indicates that CO is adsorbed in the linear and bridging modes but not in the dicarbonyl mode ~ After the sample is evacuated at 300 K, and then exposed to 50 Torr of CO, the peaks of linear and bridg- . ing CO are more intense and shifted to higher frequencies compared to those before exposure to CO. Again, no dicarbonyl species are present. The spectral lineshape is quite _comparable to but "' 1.5 times more intense than that obtained after heating Rh/ alumina-90 in CO at 525 K. The sample that had been heated in C0 2 was further heated in CO at 525 K. This caused little change in the peak intensities or frequencies. As with heating Rh/ alumina in CO, the low frequency part of the spectrum of Rh/ alumina heated in C0 2 indicates the - 61presence of carbonate, water and methane, but no formate or bicarbonate species. C. Infrared Spectra of CO Adsorbed on Rb./ Alumina Previously Heated at 525K. The effect of heating Rh/ alumina in vacuo on the sites for adsorption of CO was also studied using infrared spectroscopy. However, Rh/alumina-115 was used instead of Rh/ alumina-90 as in the above studies. The dispersion of the Rh was greater for Rh/alumina-115 because of the greater surface area of the alumina. For reference, the infrared spectrum of CO adsorbed at 300 K on unheated Rh/alumina-115 is shown in Fig. 4. The dicarbonyl doublet is more intense, per unit weight and surface area of sample and more intense relative to the bands of linear· and bridging CO, than for CO adsorbed on Rh/ alumina-90 The band of the linear CO is present but is not resolved from the bands of the dicarbonyl. This was demonstrated with a different Rh/alumina-115 sample. It was prepared similarly and had a similar infrared spectrum after CO was adsorbed at 300 K. Upon exposure to 10 Torr of 0 2 at 300 K, there was a loss of intensity at 2070-2060 cm- 1 and at,... 1850 cm- 1 due to selective removal of the linear and bridging CO, as observed by other workers using similar treatments (10,12,34) . A cell containing Rh/alumina-115 which had been reduced in H2 and evacuated is closed, heated at 525 K for 12 hr, and cooled to 300 K. The sample is then exposed to CO at 300 K, and an infrared spectrum recorded. Some redistribution of the CO adsorption sites is evident from the specctrum in Fig. 4 . The dicarbonyl doublet is somewhat less intense and the peaks of linear and bridging CO are more intense compared to those in the spectrum of CO adsorbed on unheated Rh/alumina-115. The change is not nearly so great as upon heating in -62the presence of CO or C0 2 , though. The spectrum also indicates, as after heating Rh/ alumina in the presence of CO or C0 2 , that ionic carbonate, water and methane (evacuated before adsorption of CO), but not formate or bicarbonate, are formed. D. Gas Uptake and Evolution Measurements In order to obtain quantitative information about ·the redistribution and deactivation of the sites for adsorption of CO on Rh/ alumina that were indicated by the infrared spectra, quantitative measurements were made of the gases taken up by and evolved from the samples upon heating. The results are summarized in Table 2. Experiments were also carried out with 13 C-enriched CO and the gases analyzed mass spectrometrically. These were used to verify the accuracy of the manometric measurements and the lack of interference from any residue of acetone that may have been retained during the sample preparation. Quantitative measurements for Rh/ alumina heated in CO were made for both Rh/alumina-115 and Rh/alumina-90 samples. Upon exposure of Rh/alumina115 and -90 to a few Torr of CO at 300 K, 286 and 194 x 1 a-a mol of CO per gram of sample, respectively, are taken up, that is, removed from the gas phase. These are equivalent to ratios of CO molecules adsorbed per Rh atom in the sample of 1.47 and 1.00, respectively. After heating Rh/alumina-115 at 425 Kin CO, during which the only major change in the infrared spectrum was a decrease in intensity of the dicarbonyl doublet, CO and C0 2 are evolved in a ratio of 3.4 to 1. Upon heating to 525 K. after which there are no infrared bands of the dicarbonyl, CO and C0 2 are again evolved, in a ratio of 3.1 to 1. The amount of CO adsorbed at this point can be estimated by the amount of CO that readsorbs after evacuation, H2 treatment, and reevacuation at 450 K and cooling to 300 K, since the infrared spectra of the sample before and after this procedure are - 63very similar (see Fig. 1). The amount of CO adsorbed on Rh/alumina-90 is greater than that on the Rh/alumina-115, both in absolute value, 38 vs. 14 .u.mol/g, and relative to that adsorbed on an unheated sample, 5 vs. 20%. In both cases, this is a dramatic loss in the capacity for CO adsorption. The total amount of CO and co.a evolved upon heating plus the CO that r~mains adsorbed account for 275 of the 286 ,umol/ g or 96.% of the CO taken up at 300 K by Rh/alumina-115. The amount of H2 in the gas phase after heating Rh/alum.ina115 in CO at either 425 or 525 K is less than 11 ,umol/g. Also, there is less than 1 ,umol/ g of C0 2 and H 2 after heating alumina in CO without Rh at 425 and 525 K. From. the data in Table 2, we also observe a strong dependence of the decrease in capacity for CO adsorption on Rh/alum.ina upon the gas present during heating to 525 K. After heating Rh/alumina-115 in CO and C02, the respective losses are 95 and 83% of the amounts of CO adsorbed on unheated samples. The losses after heating Rh/alumina-115 in the presence of 02, H2 and after heating an evacuated sample are much less: 3, 19 and 32%, respectively. After heating in 0 2 , H2 or in vacuo, there are one or more CO molecules adsorbed per Rh atom, indicating that dicarbonyl species are present. This had been determined independently for CO adsorbed on Rh/alumina-115 previously heated in vacuo from the infrared spectrum in Fig. 4. We also note that Rh/alumina-115 which had been heated in H 2 is deactivated by subsequently heating in CO at 525 K, during which CO(g) and C0 2 are evolved. There is little additional deactivation of Rh/alumina-115 which had been .heated by C02 by subsequently heating in CO, in agreement with the infrared spectra in Fig. 3. 4. Discussion A. Deactivation of Rh/ Alumina Upon Heating in CO -64- We first consider the possibility that sites for adsorption of CO on Rh/ alumina that had been previously heated in CO could be blocked by species strongly adsorbed on the Rh. Surface carbon could be deposited by disproportionation of CO: (1) which occurs on several Group VIII transition metals (35-37). However, all but 11 ,umol/g (4%) of the carbon atoms of the CO adsorbed on unheated Rh/alumina-1 15 is accounted for as CO(g)• CO(ad) or C0 2 after heating to 525 K. Therefo.r e , this is an upper limit for the concentration of surface carbon after heating. It is not reasonable that 11 J,.Lmol/ g of C(ad) could block the adsorption of 272 ,umol/g, or all but 4%, of the CO that adsorbs on unheated Rh/ alumina115. By the same argument, the concentration of other carbon-containing species formed from CO, such as adsorbed carbonates, bicarbonates or carboxylates, is also limited to 11 ,umol/g. Therefore, their formation cannot explain the extent of deactivation observed. We also consider that adsorbed oxygen, which could be formed by dissociation of CO, could block the sites for adsorption of CO. However, the number of oxygen atoms in the CO(g) and C0 2 evolved and CO{ad) after heating Rh/ alumina115 in CO is greater than the number of oxygen atoms in the CO adsorbed on the sample before heating. Thus, it is unlikely that oxygen is being deposited on the Rh. It indicates, rather, that there is a net transfer of oxygen from the surface to the CO in forming C0 2 . Another possible cause of the deactivation of sites for adsorption of CO is that the Rh agglomerates upon heating in CO, so that a large fraction of the Rh is no longer able to adsorb CO. The deactivation is facilitated by the CO since heating Rh/alumina in vacuo does not cause much deactivation. We also recall -65- that while heating Rh/alumina in CO, some of the CO reacts with oxygen from the surface to form C0 2 . Little C0 2 was formed in the absence of Rh. The amount of C0 2 that is produced increases with increasing initial dispersion of the Rh (as determined by the initial uptake of CO) as shown in Table 3. Since there is greater deactivation of the more highly dispersed Rh/ alumina. this further supports the involvement of C0 2 formation in the deactivation process. In addition, we recall that the dicarbonyl species, which contain Rh1, are eliminated by heating Rh/ alumina in CO. The Rh of the dicarbonyl species would have a greater tendency to agglomerate if it were reduced from Rh1 to Rh 0 because of a loss in ionic bonding to the alumina support. These observations suggest that CO is oxidized to C0 2 and that Rh1 is reduced to Rh 0 upon heating Rh/ alumina in CO as in the reaction (2) As a check of this mechanism, we predict the amount of C0 2 that would be formed by the reaction in Eq. (2) and compare it to the observed value. The amount of C02 observed is within 20% of that actually formed while heating, based on the amount of CO adsorbed before heating that is not accounted for as CO(ad)· CO(g) or C0 2 after heating. We can predict the amount of C02 that would be formed according to Eq. (2) by estimating the amount of dicarbonyl-forming Rh1 that is present before heating. Using CO uptake and primarily 13 C NMR measurements , Duncan et al. (9) determined the distribution of CO among the dicarbonyl, linear and bridging sites. For samples essentially identical to the Rh/ alumina-90 samples used in this study, they found that the ratio of linear to bridging CO was 2 to 1, that the rati~ of CO adsorbed on the clustered Rh to the surface Rh atoms of the Rh clusters was 3 to 4, and that 12% of the Rh of a freshly prepared sample was not involved in bonding to CO in any of the three modes of CO adsorption. - 66- We assume that these results are true for the samples in the present study. We can then use the ratio of total CO adsorbed to total Rh atoms in the sample to calculate the distribution of Rh in the dicarbonyl, clustered and inactive forms (as well as the distribution of CO in the dicarbonyl, linear and bridging modes). The sensitivity of the calculated distributions to the assumptions made is such that an error of 5% in the amount of inactive Rh or an error of 0.05 in the ratio of CO to surface Rh atoms for the Rh clusters or an error of 0.1 in the overall CO to Rh ratio would result in an error of 15-20% in the calculated values. The Rh and CO distributions of three samples of different dispers ions before heating in CO at 525 K were calculated. The amount of C02 produced upon heating these samples in CO, along with the calculated concentrations of c!icarbonylforming Rh1 are shown in Table 3. The ratio of Rh1 to C0 2 formed for all three samples agrees with the stoichiometry of Eq. (2) well within the uncertainty of ± 0.4 determined by an uncertainty in the calculated Rh1 concentration of 20%. Thus, Eq. (2) can account for most, if not all, of the C0 2 produced upon heating. B. Mechanism of Reduction of Rh1 by CO It is possible that the Rh1 could facilitate the reaction of CO directly with oxygen of the surface. However, in studies of supported Rh formed from Rha(C0) 18 (13.14) or Rh4 (C0) 12 (38), H 2 0 was found to promote the conversion of dicarbonyl species to Rh that adsorbs CO in the linear and bridging modes. The surfaces as prepared in the present study are highly hydroxylated (2900 o2 .urnol(OH)/g , or 15 per 100 A , as determined by 1H NMR) . Also, the infrared spectra indicate that molecular water _is formed upon heating the samples to 525 K. presumably by condensation of the surface hydroxyl groups: - 67- (3) Therefore, we consider the implications of the results of the present study for the possible participation of H2 0 in the deactivation process. Water is known to react with CO groups bound in transition metal complexes to produce C0 2, during which either the H2 0 or the metal may be reduced. The former would be part of the water gas shift reaction: (4) which is catalyzed by metal carbonyls (39-41), Ptii and Pt 0 complexes (42.43), and Rh complexes such as RhHL3 and Rh(OH)(CO)~. L = P(i- C3 H7 )s (44). Water or Ho- reacts with metal-bound CO to form, presumably, a carboxylate or formate intermediate, which decomposes to C0 2 and a metal hydride complex. Eventually, the hydrogen bound to the metal forms H2. An example in which the metal of a complex is reduced during the reaction of H 20 with metal-bound CO is the reduction of the Rhm complex, RhC1 3 to the Rh1 complex, Rh 2 (C0) 4 C~ (45). There is most likely some bonding between Rh1 atoms in crystalline Rh2 (C0) 4 Cl2 (46). It is possible to further reduce the Rh 1 to the Rh 0 of Rh 6 (C0) 16 in the presence of H 2 0 according to the following stoichiometry (47): (5) The reduction of the Rh in the di-Rh1 complex may be compared to the reduction of H2 0 with the complex Rh(OH)(CO)(P(i-C 3 H7 )) 3 (44), which contains only one Rh1 atom. Two electrons are transferred per molecule of CO that is oxidized to C0 2 • Apparently, when the metal is able to accept these two electrons, the metal, rather than the H20 . is reduced. The present study is a case in which the metal rather than H20 is reduced. - 68- The ratio of Rh1 present before· heating to the C0 2 that is produced while heating agrees well with the ratio predicted assuming the Rh1 is reduced to Rh0 [Eq. (2)] . Also, agglomeration of the Rh0 that r~sults provides an explanation of the deactivation of the sites for adsorption of CO. In contrast, both the reduction of H2 0 as in the water gas shift reaction [Eq. (4)] and the reduction of H2 0 and Rh1 simultaneously are not consistent with the results of the present study. Unless the water gas shift reaction was limited kinetically, which seems unlikely, the amount of C0 2 formed would not be related to the amount of dicarbonyl-forming Rh1 that is present, in contrast to the results in Table 3 . If the Rh1 and H 2 0 were reduced simultaneously, as in: (6) the ratio predicted for the Rh1 present before heating to C0 2 formed while heating would be 1.0, compared to the observed ratio of 1.9 ± 0.4 (Table 3). H 2 that is present in the gas phase or adsorbed on the Rh or alumina or that has reacted to f orm CH 4 after heating Rh/alum.ina-115 in CO is less than 5 J,Lmol/g, as determined by mass spectrometric and gas uptake and evolution measurements. This is equivalent to a ratio of H2 to C0 2 formed of less than 0 .08, compared to ratios of 1 and 0.5 that would be predicted using Eqs . ( 4) and (6), respectively. If H 2 0 does facilitate the reduction of Rh1 by CO in the present study, the Rh 1 must be able to accept the two electrons transferred for each CO that is oxidized. Production of hydrogen that could reduce other Rh1 species cannot be involved since heating Rh/alumina in H 2 did not cause much deactivation (see below). We recall that H2 0 was reduced . by CO in the presence of 1 Rh(OH)(CO)(P(i-C 3 H 7 )) 2 (44), which contains one Rh1 atom, but that the Rh was reduced by CO in the di-Rh1 complex, Rh 2 (C0) 4 Cl2 (47). We then speculate that - 69- each Rh1 is near at least one other Rh1 before it is reduced to Rh0 . Structures of supported Rh derived from Rh 6 (C0) 16 and Rh4(C0) 12 in which the Rh atoms are near each other, perhaps bonded to each other have been proposed (13,14,38). Dicarbonyl species are formed by addition of oxygen to the supported complexes. The complexes can be reformed by addition of CO and H20, and the analogy to the synthesis of Rh 4 (C0) 12 and Rh 6 (C0) 16 from Rh2(C0)4Cl2 (47) has been noted (13.38). Since there are Rh-Rh bonds in the complexes from which the supported Rh was formed, it was not apparent that dicarbonyl-forming Rh1 species of supported Rh that was formed by reducing RhCls in H 2 would also be near each other. The results of the present study suggest that this is the case. Though the dicarbonyl-form.ing Rh1 species may be near each other at some time before they are reduced by CO, this does not imply that the dicarbonyl species are near each other. During thermal programmed desorption of COcovered Rh/ alumina, Fujimoto et al. observed that only CO desorbs at temperatures less than 373 K, during which the dicarbonyl doublet in the infrared spectrum is removed (21). At higher temperatures, C0 2 desorbs . Apparently one CO is removed from the dicarbonyl species before the Rh1 is reduced. It is thus possible that the formation of the d.icarbonyl. which is known to be an activated f. process (12,20), could involve migration of Rh1 species away from each other in order to bind two CO molecules per Rh1. C." Rh/ Alumina Heated in CO vs. H 2 The reduction of Rh1 to Rh0 by heating Rh/ alumina in CO at 525 K is nearly quantitative, as shown by the agreement of the Rh1:C0 2 ratios with those predicted by Eq. (2) and by the lack of the dicarbonyl doublet in the infrared spectrum after heating in CO (Fig. 1). In contrast, only ..... 4 2 of the 125 ,umol/g, - 70or 34% of the Rh1 is reduced by heating at 525 K in H 2 , according to calculations of the distribution of CO among the various adsorption modes that was described above. We may speculate that the binding of two CO groups to a single Rh1 facilitates the reduction in CO, as it does the exchange with gas phase CO (20). However, the thermal programmed desorption results discussed above suggest that one CO is removed from the dicarbonyl before the Rh1 is reduced. It is also unlikely that a metastable state is formed after the removal of CO from the dicarbonyl since its reformation is activated, as was its original formation (48). The difference in the ability of CO and H 2 to reduce the Rh1 may be due to the stronger bonding of CO to the Rh1 sites, increasing its availability for reaction with oxygen of the surface. More likely, the reaction of CO directly with oxygen of the surface or with H 2 0 is a lower energy path by which the Rh1 may be reduced, compared to the paths available for reduction by H 2 • Thus, the specific reactions that a gas may undergo, rather than simply the reducing or oxidizing nature of the gas strongly influence the effect of the gas on the supported Rh. D. Rh/ Alumina Heated in C0 2 The infrared spectra and uptake measurements suggest that the effects of heating Rh/ alumina in C0 2 at 525~ K are similar to those of heating in CO. Rh/ alumina heated in C0 2 (or CO) adsorbs CO in the linear and bridging but not in the dicarbonyl mode, as shown in the infrared spectra. The capacity for adsorption of CO after heating Rh/ alumina-115 in C0 2 is 4 8 J..i.rnol/ g, or only 17% of the amount of CO adsorbed on an unheated sample. This is similar to the amount of CO adsorbed in the linear and bridging modes on unheated Rh/ alumina (34 J..i.mol/ g) , according to the calculations described above. - 71 We might suppose that adsorbed CO, formed by decompostion of C0 2 on the clustered Rh, could migrate to and reduce the Rh1 to Rh 0 as proposed for heating in the presence of CO. The newly formed Rh 0 would most likely migrate to a Rh cluster. The adsorbed 0 may be absorbed into the Rh cluster, a process that occurs for 0 adsorbed on Rh(lll) at 400 K (49). Considering 1/2 0 for each of the 125 p.mol/g of Rh1 of Rh/alumina-115 and one 0 for each of the 11 p.mol/ g of adsorbed CO formed from C0 2 dissociation, one calculates an O:Rh ratio of 0.38. However, there is little, if any, color change upon heating Rh/alum.ina in C02, in contrast to the darkening upon heating· CO. The lack of color change is not due to a difference in the color of a Rh0 cluster and a partially oxidized Rh cluster since a sample which had turned dark upon heating in CO did not become lighter when heated at 525 Kin 0 2 for 12 hr. It thus appears that the mechanism proposed for deactivation by heating Rh/ alumina in CO is not responsible for the deactivation by heating in C02. Another possibility is that oxygen, which could be formed by the dissociation of C0 2, could bind to the dicarbonyl-forming sites and block subsequent binding of CO at these sites . This is unlikely, based on thermochemical estimates. We assume that a species with the stoichiometry Rh20 is stable with respect to decomposition to 0 2 and Rh metal and that the stability of a dicarbonyl-forming Rhi with an additional oxygen bound to it is similar to that of Rh20s. The r-eaction (7) is endoergi c by more than about 75 kcal/mol (50). The reaction is more favorable if the CO that is produced is adsorbed on the Rh . However , only ..... 34 ,u.mol/ g of CO is adsorbed on linear and bridging sites of unheated Rh/ alumina- - 72115, or """29% of the amount of Rh1 that forms dicarbonyl species . Thus, there are not enough sites to adsorb the CO that would be formed by deactivation of all of the dicarbonyl species according to the stoichiometry of Eq. (7). Another argument against deactivation of Rh/ alumina by the reaction in Eq. (7) is that we might have expected that heating in 0 2 would deactivate Rh/ alumina by a similar process . In contrast to this, heating in 0 2 caused almost no loss in the capacity of Rh/ alumina to adsorb CO, as shown in Table 2. In other studies (10,12,34), after treatment of Rh/ alumina with 0 2 at 300 or 473 K. the amount of dicarbonyl species formed upon exposure to CO had either increased or remained unchanged compared to that without the pretreatment in 0 2 . A new peak in the infrared spectrum appeared at 2128-2120 cm- 1 after the pretreatment in 0 2 and subsequent exposure to CO (10,12). No peak is observed at these frequencies in the infrared spectra after heating Rh/ alumina in C0 2 and after subsequent exposure to CO in the present study. Yet another possibility is that the C0 2 could react with the dicarbonylforming Rh1 to form a species that would block the adsorption of CO on these sites. There are Rh complexes in which carbonate (51,52), bicarbonate (51.53) and carboxylate (51,53,54) groups are bonded in unidentate and bidentate configurations. These Rh complexes are often synthesized by reaction of C0 2 with other Rh complexes (51-54}. However, upon heating Rh/alumina in C0 2 , there is little change in the infrared spectrum at the frequencies expected for such species (51-54). Uncertainty about the extinction coefficients · of the infrared peaks that would be due to Rh carbonate and carboxylate species prevents us from concluding that these species are not present , though. Thus, additional experiments are necessary to determine the cause of deactivation of the dicarbonyl-forming sites of Rh/alumina which results from heating in C02 . Perhaps 13 C NMR or X-ray photoelectron spectroscopy will be useful in resolving this problem . -73(THIS PAGE IS BLANK, DUE TO ERROR IN PAGINATION) -74- E. Surface Species Bound to the Alumina The final point we would like to discuss is the difference in surface species bound to the alumina after heating in the presence and absence of Rh. Heating alumina in · either , CO or C0 2 produces surface formate, carbonate and bicarbonate species , whereas in the presence of Rh. only carbonates are formed. We might expect formate and bicarbonate species to be formed by reaction with surface hydroxyl. groups as follows: (8) and (9) These reactions may take pl.ace with hydroxyl groups, that are blocked by the Rh or are replaced by the chloride ions introduced as RhC13 . Peri (55) observed that infrared bands in the region 1900 - 1350 cm- 1 that appeared upon exposure of a lumina to CO or C02 were not present if the al.umina was previously exposed to chloride or fluoride ions. The concentration of chloride ion remaining after reduction of Rh/ alumina in H 2 is 305 ,umol/ g (9) and the Rh concentration is 1 95 ,umol/ g. These are much less than the total hydroxyl group concentration of 2900 ,umol/ g. This may indicate that the formate and bicarbonate species are formed from particular hydroxyl groups that are also preferentially blocked by the Rh or are replaced by chloride ions. 5. Conclusions The behavior of Rh/ alumina upon heating in the presence of CO, c.c.Jl.., H~.,and 02 was studied using infrared spectroscopy and quantitative measurements of gases takenup and evolved during various proced.ures . By combining the results with estimates of the distribution of CO and Rh among the dicarbonyl, linear - 75- and bridging modes of CO adsorption we find that: A. The loss in the capacity of Rh/ alumina to adsorb CO upon heating at 525 K in the presence of various gases increases in the order: 0 2 , H 2 , vacuum < C02 < CO. B. The Rh1 that forms dicarbonyl species is reduced with CO by 525 K, yielding C0 2 and Rh0 • The Rh 0 agglomerates, accounting for the substantial loss in capacity for adsorption of CO . There is little or no disproportionation of CO . C. Reduction of Rh1 by heating in CO is much more extensive than by heating in H 2 most likely because of a greater reactivity of CO toward oxygen of the surface or the stronger bonding of CO to the Rh1, increasing its availability for reaction. D. The dicarbonyl-forming Rh1 species are also deactivated by heating in C0 2 . The resulting loss in capacity to adsorb CO is nearly as great as that from heating in CO . Acknowledgments This study was supported by the Office of Naval Research under contract N00014-79-F-00 14. We thank Dr. G. R. Rossman for the use of his infrared spec- trometer and M. Templeton for obtaining the mass spectra . Helpful comments were offered by Drs . W. H. Weinberg. T. M. Duncan and S. I. Chan. -76References 1. Yang, A. C. and Garland, C. W., J. Phys. Chern. 61. 1504 (1957). 2. Sheppard, N. and Nguyen, T. T., Advan. Infrared Raman Spectr. 5, 67 (1978). 3. Kellner, C.S. and Bell, A. T., J. Catal. 71, 296 (1981). 4. Smith. A. K., Besson, B., Basset, J.M., Psaro, R., Fusi, A. and Ugo, R., J. Organometal. Chern. 192, C31 (1980). 5. Deeba, M. and Gates, B. C., J. Catal. 67, 303 (1981). 6 . Knozinger, H. and Zhao, Y., J. Catal. 71, 337 (1981). 7. Yates, J. T., Jr, Duncan, T. M., Worley, S.D. and Vaughan , R. W., J. Chern. Phys. 70, 1219 (1979). 8 . Yao, H. C. and Rothschild, W. G., J. Chern. Phys. 68, 4774 (1978). 9. Duncan, T. M., Yates, J . T., Jr . and Vaughan, R. W., J . Chern. Phys. 73, 975 (1980) . 10. Rice, C. A., Worley, S. D. , Curtis, C. W., Guin, J . A. and Tarrer, A. R., J. Chern. Phys. 74,6487 (1981). 11. Yates, D. J. C., Murrell, L. L. and Prestridge, E. B., J. Catal. 57,41 (1979). 12. Primet, M., J. Chern. Soc. Faraday Trans. I 74, 2570 (1978). 13. Smith, A. K., Hugues, F ., Theolier, A., Basset, J. M., Ugo, R., Zanderighi, G. M., Bilhou, J. L., Bilhou-Bougnol, V. and Graydon , W. F., lnorg. Chern. 18, 3104 (1979). 14. Watters , K. L.. Howe. R. F., Chojnacki, T. P ., Fu. C.-M .. Schneider, R. L. and Wong, N. B., J . Catal. 66, 424 ( 1980). 15. Andersson. S. L. T., Watters. K. L. and Howe, R. F., J . Catal. 69, 212 (1 981). -7716. Gleeson, J. and Vaughan, R. W., unpublished observations. 17. Yao, H. C. and Shelef, M.. Proc. 7th Intern. Congr. Catal., paper A21. Preprints, Tokyo ( 1980). 18. Knozinger, H., Zhao, Y., Tesche, B., Barth. R., Epstein, R .. Gates, B. C. and Scott, J. P., Faraday Discus. Chern. Soc. 72, in press. 19. Yao, H. C., Japar, S. and Shelef, M., J. Catal. 50, 407 (1977). 20. Yates, J. T., Jr .. Duncan, T. M. and Vaughan, R. W., J. Chern. Phys. 71, 3908 (1979). 21. Fujimoto, K., Kameyama, M. and Kunugi, T., J. Catal. 61, 7 (1980). 22. Arai, H. and Torninaga, H., J. Catal. 43, 131 ( 1976). 23. Yates, D. J. C. and Sinfelt, J. H., J. Catal. 8 , 348 (1967). 24. Yao, H. C., Yao, Y.-F. Yu and Otto. K., J. Catal. 56, 21 (1979). 25. Okanoto, et al., J. Catal. 58, 82 (1979). 26. Iizuka, T. and Lunsford, J. H., J. Mol. Catal. 8, 391 (1980). 27. Fuentes, S. and Fuentes, F., J. Catal. 61, 443 ( 1980). 28. Greenler, R. G., J. Chern. Phys. 37, 2094 ( 1962). 29. Solymosi, F., Erdohelyi, A. and Kocsis, M., J. Catal. 65, 428 (1980). 30. Little, L. H., Infrared Spectra of Adsorbed Species, p. 74, Academic, New York (1966). 31. Parkyns, N.D., J. Phys. Chern. 75, 526 (1971). 32. Solymosi, F. and Erdohelyi, A., J. Catal. 70 , 451 (1981). 33. Ii-£uka, T. and Tanaka, Y., J. Catal. 70, 449 (1981), and references therein. - 7834. Cavanaugh, R. R. and Yates, J. T., Jr., J . Chern. Phys . 74, 4150 (1981 ). 35. Ponec , V., Catal. Rev. Sci. Eng. 18, 151 (1978). 36. Rabo, J. A., Risch, A. P . and Poutsma, M. L., J. Catal. 53, 295 (1978) . 37 . Doering, D. L.. Poppa, H. and Dickinson, J . T .. J. Vac. Sci. Tech. 18, 460 (1981). 38. Theolier, A., Smithg, A. K., Leconte, M., Basset, J. M., Zanderighi, G. M., Psaro, R. and Ugo , R., J. Organometal. Chern. 191, 415 (1980). 39. King , A. D. et al., J. Amer. Chern. Soc. 103, 299 (1981). 40 . Ungermann, C., Landis, V., Moya, S. A., Cohen, H., Walker, M., Pearson, R. G., Rinker, R . G. and Ford, P. C., J. Amer. Chern. Soc. 101, 5922 ( 1979). 41. Darensbourg, D. J. and Rokicki, A., in Catalytic Activation of Carbon Monoxide (P. C. Ford, Ed.), p . 1 07, American Chemical Society, Washington, D.C., 1981. 42. Chen, C.-H. and Eisenberg, R., J. Amer. Chen1. Soc. 100, 5968 (1978) . 43. Yoshida, T.. Ueda , Y. and Otsuka, S., J. Amer. Chern. Soc. 100, 3941 (1978). 44. Yoshida, T., Okano, T. and Otsuka, S., in Catalytic Activation of Carbon Monoxide (P. C. Ford, Ed.), p. 79, American Chemical Society, Washington, D.C., 1981. 45 . James, B. R., Rempel, G. L. and Ng, F. T. T., J. Chern. Soc .- A, 2454 (1969). 46 . Dahl, L. F., Martell, C. and Wampler, D. L., J. Amer. Chern. Soc . 83, 1761 (1961) . 47 . Chini, P. and Martinengo, S., Inorg. Chim. Acta 3, 315 (1969) . 48. Duncan, T. M. , PhD Thesis, California Institute of Technology (1 980) . 49. Thiel, P . A., Yates, J. T., Jr. and Weinberg, W. H., Surface Sci. 84, 54 (1979) . 50. JANAF Thermochemical Tables, 2nd ed. (D. R. Stull and H. Prophet, Project Directors) . U.S. National Bureau of Standards, Washington, D.C., 1971. - 7951. Eisenberg, R. and Hendriksen, D. E., Advan. Catal. 28, 79 (1979). 52. Krogsrud, S., Komiya, S., Ito, T., Ibers, J. A. and Yamamoto, A., Inorg. Chern. 15, 2789 ( 1976). 53. Yoshida, T., Thorn, D. L., Okano, T., Ibers, J. A. and Otsuka, S., J. Amer. Chern. Soc. 101, 4212 (1979). 54. Volpin, M. E. and Kolmnikov, I. S., Organornetal. Reactions 5, 313 (1975). 55. Peri, J. B. , J. Phys. Chern. 70,1482 (1966); 72,2917 (1968) . - 80- Figure Captions Figure 1: Infrared spectra of Rh/alumina heated in 12 CO. (a) Clean Rh/alumina exposed to 50 Torr of 12 CO at 300 K; (b) The sample in (a) heated at 425 K for 12 hr; (c) The sample in (b), heated at 525 K for 12 hr; (d) The sample in (c) evacuated, exposed to 200 Torr of H2. and reevacuated, all at 450 K (after which there were no infrared peaks due to CO stretching), then exposed to 50 Torr of 12 CO at 300 K. Figure 2: 12 Infrared spectra of alu,mina heated in 50 Torr of: A. 12 C0 . 2 CO and B . (a) 50 Torr ofc.oor C0 2 at 300 K; (b) The sample in (a), heated at 425 K for 12 hr; (c) The sample in (b). heated at 525 K for 12 hr. Figure 3: Infrared spectra of Rh/alum.ina-90 heated in 12 C0 . 2 (a) Clean Rh/alurnina exposed to 55 Torr of 12 C0 2 at 300 K; (b) The sample in (a), heated at 425 K for 12 hr; (c) The sample in (b). heated at 525 K for 12 hr. ; (d) The sample in (c), evacuated at 300 K ar,J exposed to 50 Torr of 12 CO at 300 K. Figure 4: Infrared spectra of CO adsorbed at 300 K on: (a) unheated Rh/alumina-115; (b) Rh/alumina-115 previously heated at 525 K for 12 hr, then cooled~ to · 300 K before exposure to CO. Note that the absorbance scale has been compressed by a factor of two, relative to the previous figures. 1-' (!) ti ~ OQ 1-'• 1-Tj <! m (f) 0 0::: m <! z w 0.4 u 0 (Jl 0 - . ~ 0 0.8 2062 2072 ,. I 21'00 I 2103l I 2000 2034 1900 1875 19001 I 1700 1600 WAVENUMBER ( cm-1) 1800 1500 1625 ~/ I .~ ~·' 1400 1300 '·'-..e:.,.j I r .\ ' ·, ~ .. . . . /-·-·'~.,. 130711280 1475/ I / :·/ iI i 14,15 1365 -U- Resolu.tion = 2.6 cm- 1 co on Rh/alumina (d) H2 at 450 K, then CO at 300 K (c) 525 K (a) 300 K . (b) 425 K 12 I I 1-' 00 ...... ~ N (l) cli (JQ 0 _ - / yfr (c) 525K 15~(b) 425K -- 1 _, 1470 - / I JV , 't 1\ I I \ B. 12 co 2 on alumina I -- __/<::' ___ 141o I 1455 <l: al (f) 0 0:: (I) <( 1700 - I ~ 1600 - ( a) 300K 1500 '/ \ ~ \\ ~~-. \ ..-...1400 1300 -~ 1 1381 '-../'-'' 1 11396 WAVENUMBER (cm- 1) -......: '~,-=../) ij ,~j/ I I \ r-1451 / I\ I z 1646 \ \ 1395 r-1383 I I 0.2 (b) 425K (a) 300K co on alumina 00 N ... 0 / I / \ ...____.,., ..," / / I / A. 12 w u 0 Ql H a .......... 0.2 /I I l / i ; ( c ) 525 K 1\ I 1594 VJ ro ~ ~ (JQ I-'• 1-rj <{ m ,U) 0::: 0 w u z <{ m 0 ()) 0 1---1 0 1---1 ........._ 0.4 2100 I / \ 2000 ••••• .,, / " ··.\'•\ ,, I I I I \ : f ; ; ~ f \.1 i .J 2143 l I ....::r .-.·.·,.·,,. .::::· 2044 : £(c) 1900 1700 1600 12 WAVENUMBER (cm- 1) 1800 I .-- (a) 300 K r--(b) 425 K f\ ------------ (d) +CO at 300 K ~ i ~ 525K l\ 2070 I 1500 1400 I 1430 I 1472 C0 2 on Rh /alumino 1300 1307 I I VJ 00 ~ ~ (I) c (JQ ~ ..... <{ m (f) 0 m n: z<{ 0 ~ -0 w CJl 0 t--4 0 --- I I 0 I 0.8~ 1.6 I ! \ I A \ 2034 2100 !J \ ]vr 2000 \i '(\' 11~nc.c. ! \ j \ 2103 I 1860 1700 1600 I 1630 I - 1460 1400 at 525K 1500 r-{b) previously heated WAVENUMBER (cm- 1) 1800 I heated ,--(a) unheated previously 12co on Rh/alumina 1300 l I I ~ 00 -85- Table 1 Frequencies and Assignments of Peaks in the Infrared Spectra between 2400 and 1200 cm-1 Frequency -1 (em ) 'V2284 2143 2103, 2034 Other Modes of Species Vibrational mode physicallyadsorbed 13co 2 gas phase 12 co dicarbonyl asymmetric OCO stretch CO stretch symmetric CO stretch asymmetric CO stretch 2072-2044 linear CO CO stretch 1900-1850 bridging CO CO stretch 1625 molecular water HOH bend 1649-1646, 1231 bicarbonate or · bidentate carbonate C=O stretch asymmetric OCO stretch 1594-1592, 1396, 1381 formate asymmetric OCO stretch CH in-plane bend symmetric OCO stretch 1475-1430 ionic ·carbonate asymmetric oco 2 stretch 1 415 '1 365 1280 carbonate and/or alcohol impurity asymmetric oco 2 stretch, COH bend, and/or CH deformation x methane CH 1307 the Species 4 deformation ~3400 (OH stretch) 2915 (CH stretch) 3019 (CH stretch) - 66- Table 2 Amounts of CO and C0 Adsorbed on and Evolved from Rh/Alumina-115 upon 2 Heating in Various Gases (a3b3c) Gas Heating T (K) co 300 C0 Additional Procedures Evolved Evolved 425 78 525 120 525 H2 co 286 ( 195) 525 2 co(ad) co (g) (evac, H , eva c) 2 at 450 K, + co 14 (38) evac at 300 K, + co 37 (39) (evac, H2' evac) at 450 K, + co 48 evac 450 K, + co 233 heat in co at 525 K 525 02 525 (a ) (evac, H2, evac) at 450 K, + CO (evac, H2, eva c) at 450 K,+ co + co 187 11 277 194 Values are in ~mol/g ± 5 1-!mo 1/g. (b ) Values in parentheses are for Rh/alumina-90. ( c ) Rh (d) · concentration = 195 f-lmo 1/ g. . Amount of co 2 in the gas phase or weakly adsorbed. 2~1 2 (d) Total 286 ( 195) (28±10) 39 43 l 233 241 - 87- Table 3 The Amount of C0 2 Produced upon Heating Rh/Alumina Samples of Various Dispersions in CO at 525 K for 12 hr. ( a) co Formed Rh Rhl 1 ca 1c (b) co _ Formed 2 Sam~le co(ad) Total Rh Rh/alumina -115 1. 47 63 125 2.0 Rh/alumina-115 previously heated in H at 525 K 2 1. 20 43 83 1.9 Rh/alumina-90 1. 00 28 ± 10 53 1. 9 2 (]lmo 1/g) calc (llmo 1/g) = 195 11mol/g. 1 Rh d;co fanned ca 1cu 1ate 2 (a)Total Rh concentration (b) The theoret i ca 1 for Eq. (2) is 2. 0 -88- Chapter 5: H2 S Adsorption and Interaction with CO on Rh/ Alumina -89- Abstract The adsorption of H 2 S and its interaction with CO on Rh supported on alumina was studied using infrared spectroscopy and quantitative measurements of the gases taken up by and evolved from the samples. The dissociation of H 2 S on the Rh at 300 K produces H 2 and is inhibited by preadsorbed CO. Rh also facilitates the reaction of H 2 S with surface oxygen at ~ 373 K, in which water is produced. After preexposure of the Rh to H 2 S, CO adsorbs in the linear, but not in the dicarbonyl or bridging modes. Adsorption .of H 2 S onto a COprecovered surface displaces much of the bridging CO, but only slowly removes the dicarbonyl and linear CO. Adsorption of H 2 S strongly inhibits the removal of adsorbed CO by 0 2 , but exchange with CO(g) occurs readily. - 901. Introduction The modification of transition metal catalysts by adsorption and decomposition of sulfur-containing compounds is of great importance industrially (1-3). The activity of the catalyst is generally lowered by adsorbed or absorbed sulfur. Occasionally, intentional exposure to sulfur-containing compounds is used to change the selectivity of a catalyst (1.2). The interaction of sulfur-containing compounds with Rh catalysts is especially interesting since among several of the Group V1II transition metals, Rh was shown to be particularly resistant to S0 2 poisoning of the catalytic reduction of NO (4). The resistance of Rh and other metals to sulfur-poisoning was greatly enhanced by adding small amounts of 0 2 to the reactants (4.5). After NO was reduced in the presence of metal foils, oxygen was found in both the surface and subsurface region of the metal (5). It is possible that this oxygen inhibited the adsorption of sulfur or bulk sulfide formation and/ or reacted with the sulfur deposited. One may speculate that the resistance of supported Rh to sulfurpoisoning may be related to a strong interaction of Rh with the oxygen of the support. The interaction of Rh with the support is reflected in its binding of CO. Using vibrational spectroscopies (6-15), it has been shown that CO adsorbs on supported Rh in three distinct modes: 0 0 c c \ I Rh 0 0 c c I -Rh- /" - Rh-Rh- which we will refer to, respectively, as dicarbonyl, linear and bridged. The linear and bridged modes, which are typical of CO adsorbed on supported and unsupported Group VIII transition metals (16), occur on two- or three-dimensional - 91 - clusters of Rh. Only on supported Rh, Ru {17) and Os (18-20) is there convincing evidence of dicarbonyl formation. The Rh of this species is most likely in an oxidation state of + 1 and may be isolated from the clustered Rh, as discussed elsewhere (8-10.12.14.21,22). Thus, it is particularly difficult to reduce all of the supported Rh to Rh 0 , presumably due to a strong interaction of Rh with the support. There have been two studies of the interaction of H 2 S with CO on supported Rh. Guerra {23), using infrared spectroscopy, found that exposing silica- supported Rh to H 2 S inhibited subsequent adsorption of CO in the bridging, but not in the dicarbonyl or linear modes. Similarly, other workers observed that CO adsorbed in the linear but not in the bridging mode on supported Ni that had been exposed to H 2 S or CS 2 (24-28). Kroeker et al. (29), us ing inelastic electron tunneling spectroscopy, found that adsorption of CO on alumina-supported Rh in all three modes was inhibited by preexposure to H 2 S. Upon exposing a COprecovered surface to H 2 S. the CO that was adsorbed in the dicarbonyl and bridging, but not in the linear mode was displaced. They also obtained evidence that one CO molecule could be bound to a dicarbonyl-forming Rh that was perturbed by H 2 S . In the present study we investigate the adsorption of H 2 S and its interaction with CO on alumina-supported Rh. We us~ infrared spectroscopy and also measure the amounts of H 2 S and CO taken up by Rh/ alumina and gases evolved during their adsorption to quantify the effects seen in the infrared spectra. By combining these results with estimates of the CO and Rh among t he various adsorption modes, we determine the effect of H 2 S on the modes of CO adsorption and the effect of CO on the modes which H 2 S is found to adsorb in. We also examine the effect of H2 S on the exchange of adsorbed and gas phase CO and the removal of adsorbed CO by 02. - 92- 2. Experimental Procedures The 2.0% by weight Rh on alumina was prepared as described previously (10). Briefly, RhCl3 ·3H 2 0 was dissolved in a water:acetone ( 1:1 0) suspension of alumina and sprayed onto a CaF 2 disc preheated to 365 K. The disc, with a coverage of - 11 mg of sample per cm2 • was mounted in a stainless steel infrared cell with CaF 2 windows. The cell was then attached to a stainless steel vacuum system and heated under vacuum to 450 K. The Rh was reduced with three charges of H2 (Matheson, 99.9995%) at 425 K and 200 Torr . The sample was then evacuated at 450 K for 12 hr and cooled to 300 K. We will refer to the sample at this point as clean Rh/ alumina. The alumina samples were prepared exactly as were the Rh/ alumina samples, except that no RhC1 3 ·3H 2 0 was added to the water:acetone suspension of alumina. The amounts of H2 S and CO taken up by the samples and gases evolved during their adsorption were measured manometrically in the constant volume system at pressures of a few Torr with a Baratron capacitance manometer to ± 0.01 Torr. Due to the low mass of the infrared samples, precision uptake measurements were made on o_~ samples packed loosely in the bottom of glass tubes and pretreated in situ as above. After the exposures to CO and H2S. the only gas phase species present at concentrations greater than 5 ,umol/ g were CO, H 2 S, H 2 0 and H2 as shown by mass spectrometry. During the exposures, these components were conveniently monitored manometrically by condensing the H2 0 in a glass tube at 195 K, condensing the H2 0 and H 2 S in a glass tube at 77 K, and condensing the CO on dehydrated alumina at 77 K. Transmission infrared spectra were obtained with a Perkin-Elmer 180 grating spectrometer with the absorbance scale. The double beam mode was used without any compensating sample in the reference beam. The spectral resolu- - 93- tion at 2000 cm- 1 was set at 2.6 cm- 1 . All spectra were recorded with the sample at approximately 310 K. The 15 CO (Merck Isotopes, 90% enriched) and the unenriched CO (Matheson, Research) were obtained in glass breakseal bulbs and used without further purification. The 0 2 (Matheson, Research) was also used without further purification. The H 2 S (Matheson, C.P .) was purified by three freeze-pump-thaw cycles and sublimed at 195 K immediately before use . The alumina was ''Alon-C" (Cabot) , a fumed alumina with a particle size of~ 15 nm and a BET surface area 3. Results A. Infrared Spectra (i). Adsorption of H 2 S on Rh/Alumina The infrared spectra of alumina and clean and co- precovered Rh/alumina after exposure to H 2 S are shown in Fig. 1. Before exposure to H 2 S, there is a broad band in the spectrum at - 3700-3000 em - 1 due to the OH stretching mode of the hydroxyl groups of the alumina surface. Exposure at 300 K of clean Rh/alumina to 4.5 Torr of H 2 S causes four changes in the infrar-ed spectrum. A band appears at 1630 cm-,..1 and there is increased inten. . sity on the lower frequency side of the band at 3700-3000 em - 1 . These are assigned to the HOH bending and OH stretching modes, respectively, of molecular water. The very weak band that appears at ~ 2480 cm- 1 is assigned to an SH stretching mode of an HS or H2 S species involved in hydrogen bonding. For comparison, the asymmetric stretching mode for gaseous H 2 S at 2627 em - 1 (30) is shifted down in frequency through hydrogen bonding to 2544 ern - 1 in solid H 2 S (31) and to 2482 em - 1 for H 2 S dissolved in pyridine (32). It will be argued later that the peak at 2067 em - l is due to adsorbed CO impurity that is - 94- perturbed by adsorbed sulfur. After alumina and CO-p recovered Rh/ alumina are exposed to 4.5 Torr of H 2S at 300 K, the bands due to molecular water are present, indicating that some H20 is formed. However, these bands are less intense than for those after exposure of Rh/alumina to H 2 S. The band at 2480 is present after exposure of COprecovered Rh/alumina, but not alumina, to H 2 S. {ii}. Sequential Adsorption of CO and H 2 S A Rh/ alumina sample was exposed to CO and then to H 2 S and another sample was exposed to H 2 S and then to CO. Their infrared spectra are shown in Fig. 2. After CO is adsorbed on clean Rh/ alumina, the intense doublet of the dicarbonyl species at 2103 and 2034 em -I and the broad band of the bridged-bonded CO near 1850 cm- 1 are seen (Fig. 2A). A different sample, prepared similarly and having a similar infrared spectrum after adsorption of CO, was then exposed to 10 Torr of 0 2 at 300 K. There was a loss of intensity at 2065-2070 cm- 1 and at 1850 cm- 1 due to selective removal of the linear and bridging CO, as observed by other workers using similar treatments (9,14,33). Thus, there is intensity due to the linear CO, but it is not resolved from the intense doublet of the dicarbonyl. For CO adsorbed on H 2S-precovered Rh/ alumina, the maximum of the dominant peak is at 2079 cm- 1 c:-s .shown in Fig. 2B, compared to 2060-2070 crn- 1 for linearly-adsorbed CO on clean Rh/alumina. The broad band of the bridging species is missing completely. The lower frequency peak of the dicarbonyl doublet at ~ 2036 cm- 1 is now only a shoulder on the peak of the linear CO, while the higher frequency peak is not resolved from that of the linear CO. The integrated intensity of the spectrum decreases by 10% upon evacuating the sample for 2 hr at 300 K. - 95Upon exposure of CO-precovered Rh/ alumina to H2 S, the peak of the bridging CO decreases substantially in intensity and shifts down in frequency from 1850 to 1820 cm- 1 (Fig . 2A) . The spectrum between 2150 and 1950 cm- 1 increases in intensity by ..... 25% and appears to be similar to a superposition of a dicarbonyl doublet and the spectrum of CO adsorbed on H 2 S-precovered Rh/ alumina. The latter was subtracted from the spectrum of a CO-precovered surface which had been exposed to H 2 S to better resolve the peaks of the dicarbonyl from that of the linear CO. As shown in Fig. 2C, the intensity of the lower frequency peak of the dicarbonyl is reduced by ..... 15-20% upon exposure of a CO-precovered surface to H 2 S. Also , the intensity of the peak of the linear CO increases to a value somewhat greater than that after adsorption of CO on an H 2 S-precovered surface. Extended exposure of a CO-p recovered surface to H 2 S ( 120 hr) causes the intensity of the spectrum to decrease by"" 25%, compared to that after exposure for 1.5 hr . {iii). Exchange of Adsorbed CO with 13 CO (g) We wished to examine the effect of exposure to H2 S on the ability of the CO adsorbed on Rh/ alumina to exchange with CO in the gas phase and also to verify that the infrared bands in the region 2200-1700 cm- 1 are due to CO stretches. To Cio this, the samples were exposed to excess 13 C-enriched CO (enriched CO/adsorbed CO = 200) at 300 K and a pressure of 12 Torr. Nearly all CO adsorbed on H2 S-precovered Rh/ alumina exchanges with 13 CO(g) in 20 hr, as shown in Fig. 3B. The shoulder at ,.... 2070 cm- 1 is assigned to adsorbed 12 CO that either does not exchange with 13 CO(g) or that is from the 10% 12 CO in the 13 C- enriched CO(g) used for the exchange. The exchange that takes p!ace is ,.... 90% complete after 0.5 hr. This is somewhat slower than the exchange in the absence of H 2 S, which is complete in~ 0.2 hr (10). - 96- Much of the CO adsorbed on Rh/ alumina, which had then been subsequently exposed to H 2 S also exchanges with 13 CO(g) in 20 hr. as shown in Fig . 3A. However, a significant fraction of the adsorbed species with infrared intensity at 2070-2080 em - l does not exchange. This is indicated by the peak at .-.J 2078 em - 1 and the loss of intensity between the peaks of the dicarbonyl, both of which are now resolved at 2051 and 1988 em -I. The high frequency peak of the mixed dicarbonyl, Rh(1 2 C0)( 13 CO) would be expected at . . . 2086 cm- 1 (10). It may contribute to the 2078 cm- 1 peak and shift it somewhat higher in frequency. The bridged-bonded CO that remains after exposure to H2 S and has its CO stretch at 1820 cm- 1 does not exchange with 13 CO(g). Again, the exchange that does occur is ...... 90% complete after 0.5 hr. {iv). Reactivity of Adsorbed CO Toward 0 2 The effect of exposure to H 2 S on the reactivity of the CO adsorbed on Rh/ alumina toward 0 2 was also investigated. The samples were exposed to excess 0 2 (0 2 / adsorbed CO = 200) at 300 K and a pressure of 12 Torr. Even after 20 hr, the intensity o·f the peaks from CO adsorbed on an H2 S-precovered sur- face is reduced by only . . . 35%, as shown in Fig. 3B. In the absence of H 2 S, the linear and bridging CO species are removed by 0 2 within 0.2 hr as found for similar samples here and by other workers (14,33). 0 2 preferentially removes the, linearly-adsorbed CO, allowing the higher frequency peak of the Rh(1 3 C0) 2 doublet to become visible as a shoulder at . . . 2051 crn- 1 and supporting the assignment of the 1988 em -I peak to the lower frequency peak of this doublet. At least some of the adsorbed species that did not exchange with 13 CO(g) are also not removed by 0 2 , since the shoulder at . . . 2070 em -I is still present. CO adsorbed on Rh/ alumina which is then exposed to H 2 S is also only slowly removed by 0 2 . Much of the peak at . . . 2078 cm- 1 due to adsorbed species that - 97- did not exchange with 13 CO(g) remains after the sample is exposed to 0 2 . An additional check that the peak at 2078 cm- 1 is not from adsorbed hydrogen rather than unreactive adsorbed CO was performed. After exposure to 0 2 the sample was evacuated and exposed to 12 Torr of D2 at 300 K. A very broad peak with a maximum at 2700 em - 1 due to OD stretching is observed, indicating that D2 is able to dissociated on this surface. The infrared spectrum between 2200 and 1700 em -l, however, is unchanged. Adsorbed hydrogen would have been expected to exchange with dissociated D2 and infrared peaks due to adsorbed hydrogen would have been shifted down in frequency by a factor of v'2 for adsorbed deuterium. B. Gas Uptake and Evolution Measurements Shown in Table 1 are the results of manometric measurements of the H 2 S and CO taken up, that is, removed from the gas phase, by various surfaces. On clean Rh/ alumina, 220 x 10-6 mol of H2 S per gram of sample are taken up, or nearly 1.1 molecules of H 2 S for each Rh atom in the sample. Of the hydrogen in the H 2 S taken up, 28% appears in the gas phase as H 2 , most of it within 0.25 hr. To test for reversible adsorption of the H 2 S and for the presence of H 2 0, which was suggested by the infrared spectra in Fig. 1. the sample was heated at 373 K for 2 hr while continuously· trapping the desorbing H2 S and H2 0 at 7i K: · We see in Table 2 that of the H 2 S taken up, 38 ,umol/ g (H 2 S/Rh = 0.19), or 17%, desorbs as H 2 S, in addition to 5 ,umol/g of H 2(s) (H 2 /Rh = 0.03). Also, a considerable amount of H 2 0 desorbs, . . ., 70 ,umol/g (H 2 0/Rh = 0.36) or 32% of the amount of H2S taken up. Due to adsorption of H2 0 on the alumina and on the walls of the vacuum system, the uncertainty in the measurement of H 2 0 is large, and is estimated to be ± 25 ,umol/ g. We cannot be sure what fraction of the H20 which des orbs upon heating is formed upon adsorption of H2 S at 300 K and what fraction is formed - 98- while heating. The amount of CO that adsorbs on this sample after this heating procedure is the same , within experimental error, as that which adsorbs on unheated, H 2 S-precovered Rh/ alumina. The adsorption of H2 S on alumina and the desorption of species upon heating at 373 K: were also carried out. Of the 36 ,umol/ g of H 2 S taken up, 28 ,umol/ g, or 78% des orbs as H 2 S , in addition to - 6 ,umol/ g of H2 0. Thus, 74% of the H2S that de sorbs from Rh/ alumina may be attributed to desorption from the alumina support, while most of the water is formed only in the presence of Rh. After exposure of CO-precovered Rh/ alumina to 4.5 Torr of H 2 S for 0.25 hr. 104 ,umol/ g of H2 S are taken up and 8 ,umol/ g of H2 <g) are evolved. The amount of H 2 S taken up that does not produce H2 (g) is equal, within experimental error, to that taken up in the same time in the absence of preadsorbed CO . Also , after 0.25 hr, only 14 ,umol/g of CO, or 5% of that adsorbed previously, was displaced into the gas phase. Between 0.25 and 10 hr of exposure to H 2 S , the amount of H 2 S taken up is similar to the amounts of H 2 (g) evolved and CO displaced into the gas phase during this period. The effect of exposing Rh/ alumina to H2 S on the amount of CO that it takes up was also examined. On clean Rh/ alumina 224 ,umol/g of CO (CO/Rh = 1.15) is taken up within 0.25 hr. and 268 ,umol/g (CO/Rh = 1.38) is· taken up in 20 hr. On an H2 S-precovered Rh/alumina, the CO/Rh ratios are only 0 .10 and 0.21 at 0.25 and 20 hr. respectively, or 8% and 15%, respectively, of those in the absence of preadsorbed H2 S. Less than 1 ,umol/ g of CO adsorbs on the alumina support. Less than 1 ,umol/ g of H2 S could be displaced into the gas phase by exposure of H 2 S-precovered Rh/alumina to CO. To determine if there was any hydrogen adsorbed on the Rh/alumina after it was exposed to H2 S , we attempted to displace adsorbed hydrogen by CO. To prevent adsorbed hydrogen from desorbing - 99upon evacuation, the H 2 S which was not taken up by the Rh/ alumina was removed by condensing at 77 K and 99.% of the H2 (g) above the sample was removed without evacuating the sample itself. After then exposing the sample to CO, there was less than 1 ,u.mol/ g of H2 (g)' 4. Discussion A. Adsorption of H 2 S on Rh/ Alumina When H2S adsorbs on clean Rh/ alumina, H2 (g) is evolved. H 2 S has been shown to adsorb dissociatively on Group VIII transition metal crystals (34-36) and films {37-39), producing H2 <s> and adsorbed sulfur: (1) This reaction occurs at temperatures as low as 193 K (37-39) . The sulfur adsorbed on crystal surfaces of Rh (36) and other metals (40) was shown to occupy high coordination sites by using Low Energy Electron Diffraction. The surfaces of Rh (40) and Pt (34,35) crystals that had been heated in excess H2S at 525-575 K had 0 .25 and 0.5-0.75 adsorbed sulfur atoms per surface metal atom, respectively. There is evidence that there is also adsorbed sulfur in high coordination sites on the clustered Rh of the present study. Preexposing Rh/ alumina to H2S prevents CO from absorbing in the bridging mode, as shown in the infrared spectra of Fig . 2B. In addition, bridging CO is displaced by subsequent exposure to H 2 S. Preadsorbing CO does, however, slow the dissociation of H 2 S. as reflected in the amount of H2 (g) evolved (Table 1). We also wish to compare the coverage of sulfur adsorbed on the metal crystals with that on the Rh/ alumina. The amount of adsorbed sulfur atoms is equal to the amount of H2 (g) evolved at or below 373 K, 66 ,umol/ g, if all of the - 100H 2 <s) is produced by reaction ( 1). We can calculate the amount of Rh that is involved in dicarbonyl formation, that is on the surface of the Rh clusters, and that is not involved in bonding to CO by using the results of Duncan et al. (21), who studied Rh/ alumina samples that were prepared similarly to those used here. By using CO uptake and primarily 13 C NMR measurements, they found that the ratio of linear to bridging CO was 2.0, that the ratio of CO molecules adsorbed on the clustered Rh to the surface Rh atoms of the clusters was 0.75, and that 12% of the Rh of a freshly prepared sample was not involved in bonding to CO in any of the three adsorption modes. Assuming that these are true for the samples in the present study, we can use the ratio of the total CO molecules adsorbed to the total Rh atoms in the sample to calculate the distribution of Rh in the dicabonyl, clustered and inactive forms (as well as the distribution of CO in the dicarbonyl, linear and bridging modes). The results of this calculation are given in Table 3. The sensitivity of the calculated distributions to the assumptions made is such that an error of 5% in the amount of inactive Rh or an error of 0.05 in the ratio of CO to surface Rh atoms for the Rh clusters or an error of 0.1 in the overall CO to Rh ratio would result in an error of 15-20% in the calculated values. T.P.e calculated ratio of adsorbed sulfur atoms to the surface Rh atoms of the Rh clusters is then 66/59 = 1.1. This is larger than any of the values found for Rh (36) or Pt (34,35) crystal surfaces (0.25-0.75). The larger values for Rh/ alumina indicate either that there is more extensive sulfur deposition on the clustered Rh than on the crystal surfaces or that some H2(g) is produced by adsorption of H2 S on the dicarbonyl-forming Rh. We are unable to differentiate between these possibilities at this time. - 101In the present study, H 2 0 is also formed upon exposure of the samples to H 2 S . The infrared spectra in Fig. 1 suggest that more H 2 0 is produced when H 2 S is adsorbed at 300 K on clean Rh/ alumina than on alumina. This is in accord with the amounts of H 2 0 that desorb upon heating these surfaces at 373 K. Only ,... 9% of H20 formed by adsorption of H 2 S on clean Rh/alumina can be accounted for by adsorption of H 2 S on the alumina. The H20 may be formed by the reaction of H 2 S with surface oxygen: (2) It is likely that this reaction is facilitated by the dicarbonyl-forming Rh1, rather than the clustered Rh. The Rh1 facilitates the reaction of CO with surface oxygen at 425-525 K (22). The amount of C0 2 produced upon heating Rh/ alumina in CO, 63 ,umol/ g, is similar to the amount of H 2 0 produced from the adsorption of H 2 S on Rh/alumina, 64 Mmol/ g, and is approximately half of the amount of Rh1 that is present before these reactions. Reactions similar to (2) may occur in the treatment of metal oxides with H 2 S in preparing metal sulfide catalysts (41). Of the hydrogen in the H 2 S that is taken up by clean Rh/alumina at 300 K, all but ,... 20% (46 Mmol/ g) is accounted for by the H 2 evolved at 300 and 373 K and by the H2 S and H 2 0 des orbed at 373 K. Only 2 ,umol/ g of H 2 are taken up by clean Rh/alumina at 373 K. Thus, the hydrogen that is not accounted for is not bound directly to the Rh. It is presumably either still bound to the sulfur or is bound to the surface oxygen but is not desorbed as H 2 S or H 2 0 below 373 K. This hyQ.rogen may contribute to the bands in the infrared spectra (Fig. 1) that were assigned to SH and OH groups. Preadsorbed CO does not inhibit greatly the uptake of H 2 S by Rh I alurn_ina. The amount of H 2 that is evolved upon exposure of CO-precovered Rh/ alumina to H 2 S for 0.25 and for 10 hr and after heating to 373 K is 14, 41 and 83%, - 102respectively, of the H 2 that evolves under the same conditions from Rh/ alumina that had been exposed only to H 2 S. This suggests that the amount of H 2 S taken up that eventually dissociates to yield H 2 is similar for clean and CO-precovered Rh/ alumina, but that the dissociation itself is inhibited by the CO. In contrast, the H 2 S taken up by clean Rh/ alumina that produces H 2 0 appears to be inhibited by preadsorbed CO, as shown by the infrared spectra (Fig. 1) and by the amount of H 2 0 that desorbs upon heating (Table 2). The difference in the amount of H 2 0 that des orbs from clean and CO-p recovered Rh/ alumina is similar to the diference in the amount of H 2 S taken up by the two surfaces. As with H 2 S adsorbed on clean Rh/ alumina, a substantial amount ( 62 ,umol I g) of the hydrogen in the H 2 S adsorbed at 300 K on CO-p recovered Rh/ alumina is not accounted for by the H 2 , H 2 S and H 2 0 desorbed at or below 373 K. Again, the hydrogen may be bound as SH and/ or OH groups based on the observation of infrared bands that were assigned to these species. B. Effect of H 2 S on the Adsorption of CO Preadsorption of H 2 S on Rh/ alumina greatly inhibits formation of the dicarbonyl species and adsorption of CO in the bridging but not in the linear mode, as indicated by the infrared spectra in Fig. 2B. The amount of CO taken up by H2Sprecove~d Rh/alumina, 41 ,umol/ g, may be compared to the calculated amount of CO adsorbed in the dicarbonyl and linear modes on clean Rh/alumina, 224 and 30 ,umol/ g, respectively. Thus, a maximum of 37% of the dicarbonyl- forming Rh can adsorb one CO. We may expect some CO to adsorb in the linear mode on the clustered Rh, based on the results of others using supported Ni (24-28). However, there is less CO adsorbed in the linear mode on a clean surface than CO on an H 2 S-precovered surface. Therefore, it seems likely that, on an H 2 S-precovered surface, CO is adsorbed linearly on both the clustered and - 103dicarbonyl-forming Rh. The effect of H 2 S on the adsorption of CO on Rh/alumina is less if the CO is adsorbed before than after exposure to H2 S. Only 25 of the 268 ,umol/ g, or 9% of the CO adsorbed on a clean surface is displaced into the gas phase by exposure to 4 .5 Torr of H2 S for 1.5 hr. The intensity of the infrared peak of the bridging CO is reduced by a factor of "'4 by exposure to H 2 S, probably due to displacement of the CO from the high coordination sites preferred by adsorbed sulfur, as discussed above. The infrared peaks of the dicarbonyl also decrease somewhat in intensity upon exposure of CO-precovered Rh/ alumina to H2 S for 1.5 hr. It is likely that a significant fraction of the 25 ,umol/g of displaced CO was bound in the dicarbonyl and bridging modes. Therefore, at least some of the 30 ,umol/ g CO that is bound linearly to the clustered Rh appears to be present after exposure of Rh/ alumina to H 2 S for 1.5 hr. However, there must be a loss of 10-25 ,umol/ g, or 4-10% in the combined amount of CO adsorbed in the linear and dicarbonyl modes, depending on whether the bridging CO is displaced to other adsorption sites or into the gas phase. Surprisingly, there is an increase of ,.... 25% in the infrared intensity between 2150 and 1950 em - 1 under these conditions. It would seem possible that there could be infrared intensity in the region 2150-1950 cm- 1 due to adsorbed hydrogen formed by dissociation of the H2S. Metal-hydrogen stretching modes in metal hydride complexes occur at these frequencies (38). Evidence that there is little adsorbed hydrogen present ( < 5 ,umol/g) after adsorption of H2 S on clean Rh/alumina was presented above . When clean Rh/ alumina samples prepared similarly to those used here were exposed to 500 Torr of H2 , no changes in the infrared spectrum resulted {43). Thus, even if adsorbed hydrogen was present, it is not likely that it would be - 104observed in the infrared spectra. In addition, Kroeker et al. (29) did not observe any vibrational modes in the tunneling spectrum due to deuterium after Rh/ alumina was exposed to D2 S. Thus, the increase of infrared intensity in the region 2150-1950 cm- 1 is not due to adsorbed hydrogen. Instead, the increase is attributed to an average increase in the infrared extinction coefficient of CO adsorbed in the linear and/or dicarbonyl modes. We can estimate the extinction coefficient of CO adsorbed linearly on clean Rh/alumina by exposing CO-precovered Rh/alumina to excess 0 2 at 300 K and at a pressure of 10 Torr and then measuring the loss in intensity in the region 2150-2000 cm- 1 . This treatment does not change the intensity of peaks of the dicarbonyl but removes the peaks of the linear and bridging CO (9,14,33). The amount of linearly-adsorbed CO is estimated as above. The extinction coefficient for CO adsorbed on H2S-precovered Rh/ alumina is obtained directly from the CO uptake and the intensity of the infrared spectrum [Fig . 2B(b)], s ince CO adsorbs almost exclusively in the linear mode . We find that CO adsorbed linearly on H2S-precovered Rh/ alumina has an integrated extinction coefficient which is - 5 times that for CO adsorbed on clean Rh/alumina. We may suppose that much of the increase of the infrared intensity in the region 2150-1950 em - 1 upon exposure of CO-p recovered Rh/ alumina to H2S could a lso be due to an increase in the extinction coefficient of t he linearlyadsorbed CO by the H 2 S. We can remove much of the contribution of the linear CO from the spectrum by subtracting from the spectrum, that of CO adsorbed onto H 2S-precovered Rh/ alumina, which was done in Fig . 2C. It can then be seen that the peaks of the dicarbonyl have decreased in intensity compared to those before exposure to H 2 S. A displacement into the gas phase of both CO groups of 11.% of the dicarbonyls would account for all of the CO displaced into the gas phase by exposure to H2 S. The CO displaced can also be accounted for if all of - 105the bridging CO and one CO of 19% of the dicarbonyls are displaced. ln studies by other workers using supported Ni (25.26.28), it was sometimes noticed that the intensity of the bands due to linearly-bound CO could be increased by either exposure of the surface to H 2 S before adsorbing CO or by exposing a CO-precovered surface to H2 S. This was interpreted as indicating that the number of linear CO species was increased by exposure to H 2 S (28). The results of the present study would suggest that the increase in intensity of the band due to linear CO could alternatively be caused by an increase in its infrared extinction coefficient. From the arguments presented above, the infrared peak at 2067 em -I which is present after clean Rh/ alumina is exposed to H2 S is not due to adsorbed hydrogen. One might suppose that lhe C0 2 impurity (0 .1%) in the H 2 S could produce adsorbed CO by decomposition on the Rh. However, only 1 ,umol/ g of C0 2 is adsorbed on a clean surface at 300 K. Also there is no change in the infrared spectrum between 2200 and 1700 em - 1 after exposure to 50 Torr of C02 at 300 K. (22). There is, however, a small amount of adsorbed CO present on the "clean" surface , as shown by the very weak band at '""'2040 cm- 1 in the background spectra of Fig . 2B. Subsequent adsorption of H2 S could increase the intens~ty of the band of linear CO by increasing its ex_tinction coefficient, as ~ -. described above, or by displacing bridging CO to linear adsorption sites. The peak could be shifted up in frequency from '""' 2040 em -I before exposure to H2S by an electronic effect of adsorbed sulfur and/ or by interactions with other adsorbed CO ( 44). The results in the present study are somewhat different from those of other investigators . After exposure of H2 S-precovered Rh/silica to CO, Guerra (23) observed the band of linear CO but not that of bridging CO in the infrared spec- - 106- trum, as found here and by other workers using supported Ni (24-28). However, Kroeker et al. (29) did not observe bands in the inelastic electron tunneling spec trum from either linear or bridging CO on H2 S-precovered Rh/alumina. Kroeker et al. also did not observe a band from the dicarbonyl, in agreement with the present results (the bands of the dicarbonyl were extremely weak), whereas Guerra observed bands which most likely should be assigned to dicarbonyl species. Exposure of a CO-precovered surface to H 2 S decreased the intensity of the bands of the dicarbonyl and bridging CO more in the study of Kroeker et al. than in the present study. The difference in results most likely indicates the sensitivity of the nature of the supported Rh to the support used and the method of preparation. C. Effect of H 2 S Adsorption on the Reactivity of Adsorbed CO The dicarbonyl species and most of the CO adsorbed linearly on Rh/ alumina that had been exposed to H 2 S before or after exposure to CO were found to exchange with 13 CO(g) within 0.5 hr. Since the desorption of the CO is much slower than this, as shown by its stability toward evacuation, the exchange must take place by an associative mechanism, as found for Rh/ alumina without H2 S (10). The bridging species that remain after H2 S adsorption on a CO-precovered surface does not exchange with 13 CO(g) or with adsorbed CO that exchanges with 13 CO(g), even after 20 hr. Since the exchange of linear and bridging CO would be expected to take place much faster than this (45), these bridging CO groups must be isolated from other adsorbed CO, presumably by adsorbed S, HS or HzS species. In contrast to the exchange of adsorbed CO with 13 CO(g), the removal of adsorbed CO with 0 2 is inhibited greatly by previous exposure of Rh/ alumina to H 2 S. The presence of adsorbed dioxygen has been proposed to explain isotopic - 107exchange experiments of 0 2 over Rh/ alumina that had been derived from Rh6 (C0} 16 dispersed on alumina (13). We might have expected for the samples used here that dioxygen could displace adsorbed CO as in the exchange reaction with 13 CO(g} . Arguing against this, however, is that H 2 S inhibits the removal of adsorbed CO by 13 CO(g) much less than by 0 2 . The rate of oxidation of adsorbed CO by 0 2 over a Pt( 100} surface that had been partially covered with adsorbed sulfur by decomposition of H 2 S was shown to be limited by the rate of dissociative adsorption of 0 2 (34). Inhibition of 0 2 dissociation by H 2 S could also prevent CO removal by 0 2 in the present study. 5. Conclusions We have studied the adsorption of H 2 S and its int~raction with CO on Rh/ alumina by using infrared spectroscopy and gas uptake and evolution me asurements . We find that: 1. On clean Rh/ alumina, H 2 S adsorbs dissociatively at 300 K, producing H 2 , and also reacts with surface oxygen at~ 373 K, producing water. 2. Little or no CO adsorbs in the dicarbonyl and bridging modes on an H 2 Sprecovered surface, while the amount of adsorption in the linear mode is not changed greatly. H 2 S adsorbing onto a CO-precovered surface displaces much of the bridging CO, but only slowly removes the dicarbonyl and linear co. 3 . Exposure of Rh/ alumina to H 2 S strongly inhibits the removal of adsorbed CO by 0 2 , but exchange of adsorbed CO with CO(g) takes place readily. From the present results , we can conclude that the particular resistance of Rh catalysts to poisoning by S0 2 (4) cannot be due to the presence of dicarbonylforming Rh1 species that do not bind sulfur. It is possible that the Rh1 and - 108clustered Rh may react differently toward H 2 S and S02 or that the ease of removal of sulfur from these sites may be different. Acknowledgments This study was supported by the Office of Naval Research under contract N00014-79-F-0014. We thank Dr. G. R. Rossman for the use of his infrared spectrometer and M. Templeton for obtaining the mass spectra. Helpful comments were offered by Drs. W. H. Weinberg and T. M. Duncan. - 109- References 1. Oudar, J .• Catal. Rev. Sci. Eng. 22, 171 (1980). 2. Madan, R. J. and Shaw, H., Catal. Rev. Sci. Eng. 15, 69 (1977). 3. Shelef, M., Otto, K. and Otto, N.C., Advan. Catal. 27, 311 (1978). 4. Summers, J. C. and Baron, K., J. Catal. 57, 380 (1979). 5. Tsai, J., Agrawal, P. K., Sullivan, D. R., Katzer, J. R. and Manogue, W. H., J. Catal. 61,204 (1980). 6. Yang, A. C. and Garland, C. W., J. Phys. Chern . 61, 1504 (1957). 7. Arai, H. and Torninaga, H., J. Catal. 43, 131 ( 1976). 8. Yao, H. C. and Rothschild, W. G., J. Chern. Phys. 68, 4774 ( 1978). 9. Primet, M., J. Chern. Soc. Farad. Trans. I 74, 2570 (1978). 10. Yates, j, T., Jr., Duncan, T. M., Worley, S. D. and Vaughan, R. W., J. Chern. Phys. 70, 1219 (1979). 11. Yates, D. J. C., Marrell, L. L. and Prestridge, E. B., J. Catal. 57, 41 (1979). 12. Smith, A. K., Hugues, F., Theolier, A., Basset, J. M., Ugo, R., Zanderighi. G. M., Bilhou, J. L., Bilhou-Bougnol, V. and Graydon, W. F., Inorg. Chern. 18, 3104 (1979). 13. Watters, K. L., Howe, R. F., Chojnacki, T. P., Fu, C-M., Schneider, R. L. and Wong, N.-B., J. Catal. 66, 424 (1980). 14. Rice, C. A., Worley, S. D., Curtis, C. W., Guin, J. A. and Tarrer, A. R., J. Chern. Phys. 74, 6487 (1981). 15. Kroeker, R. M., Kaska, W. C. and Hansma, P. K., J. Catal. 57, 72 (1979). 16. Sheppard, N. and Nguyen, T. T., Advan. Infrared Raman Spectr. 5, 67 (1978). - 11017. Kellner, C. S. and Bell, A. T., J. Catal. 71, 296 (1981). 18. Smith, A. K., Besson, B., Basset, J. M., Psaro, R., Fusi, A. and Ugo, R., J. Organometal. Chern. 192, C31 (1980). 19. Deeba, M. and Gates, B. C., J. Catal. 67, 303 (1981). 20. Knozinger, H. and Zhao, Y., J. Catal. 71, 337 (1981). 21. Duncan, T. M., Yates, J. T., Jr. and Vaughan, R. W., J. Chern. Phys. 73, 975 (1980). 22. Gleeson, J. W. and Vaughan, R. W., submitted. 23. Guerra, C. R., J. Col. Interface Sci. 29, 229 ( 1969). 24. Garland, C. W., J. Phys. Chern. 63, 1423 ( 1959). 25. Crell, W., Hobert, H. and Knappe, B., Z. Chern. 6, 396 (1968). 26. Marx, von G., Hobert, H., Hopfe, V., Knappe, B., Vogelsberger, W., Mackrodt, P. and Meyer, K., Z. Chern. 12, 444 (1972). 27. Rochester, C. H. and Terrell, R. J., J. Chern. Soc. Farad. Trans. I, 609 (1977). 28. Rewick, R. T. and Wise, H., J. Phys. Chern. 82, 751 ( 1978). 29. Kroeker, R. M., Kaska, W. C. and Hansma, P. K., J. Catal. 63, 487 (1980). 30. Allen, H. C. and Plyler, E. K., J. Chern. Phys. 25, 1132 (1956) . 31. Reding, F. P. and Hornig, D. F., J. Chern. Phys. 27, 1024 (1957). 32. Josien, M. L. and Saurnagne, P., Bull. Soc. Chirn . France, 937 (1956). 33. Cavanagh, R. R. and Yates, J. T., Jr ., J. Chern. Phys. 74, 4150 (1981). 34. Benzel, H. P . and Ku, R., J. Chern. Phys . 58, 4617 (1973); 59, 1641 ( 197.3). 35 . Fischer, T. E. and Kelemen, S. R .. J. Catal. 53, 24 (1978). (1980). - 111 36. Hengrasmee, S., Watson, P. R., Frost, D. C. and Mitchell, K. A. R., Surface Sci. 87, L249 ( 1979); 92, 71 (1980). 37. Saleh, J. M., Kemball, C. and Roberts, M. W., Trans. Faraday Soc. 57, 1771 (1961) . 38. Roberts, M. W. and Ross , J. R. H., Trans. Faraday Soc. 62, 2301 (1966). 39. Saleh, J . M., Trans. Faraday Soc. 67, 1830 (1971); 66, 1830 (1970). 40. Somorjal, G. A. and Van Hove, M.A., Structure and Bonding 38, 123 ( 1979). - 41. Weisser, 0. and Landa, S., Sulfide Catalysts, The?:r Properties and Applica- tions, p. 23, Pergamon Press, New York, 1973. 42. Kaesz, H. D. and Saillant, R. B., Chern. Rev. 72, 231 (1972) . 43. Yates, J . T., Jr., Worley, S. D., Duncan, T. M. and Vaughan, R. W., J. Chem. Phys. 70, 1225 (1979). 44. Crossley, A. and King, D. A., Surface, Sci. 68, 528 (1977) . 45. Engel, T. and Ertl, G., Advan. Catal. 28, 1 (1979). 46. Yates, J. T., Jr., Duncan, T. M. and Vaughan, R. W., J . Chem. Phys. 71, 3908 (1979} . - 112Figure Captions Figure 1. Infrared spectra of H 2 S adsorbed at 300 K on (a.) Rh/ alumina for 3.5 hr, (b) CO-precovered Rh/ alumina for 1.5 hr, and (c) alumina for 1.5 and 3. 5 hr. The background is the spectrum of Rh/ alumina before exposure to H2 S. The samples were evacuated for 0.2 hr following exposure to H2 S before the spectra were recorded. The peaks in the region 2200-1700 cm- 1 are due to adsorbed CO. Figure 2. Infrared spectra for H2 S and CO adsorbed sequentially on Rh/alumina at 300 K. A. (a.) A clean surface exposed to 2.5 Torr of 12 CO for 20 hr; {b) the sample in (a.}, evacuated for 0.2 hr, then exposed to 5 Torr of H2 S for 1.5 hr; (c) for 120 hr. B. (a.} A clean surface exposed to 5 Torr of H2 S for 3.5 hr and evacuated for 0.2 hr; (b) the sample in (a.), exposed to 2 Torr of 12 CO for 20 hr. The background is the spectrum before exposure to H2 S. C. The difierence spectrum between a CO-precovered surface exposed to H 2 S (A {b)) and CO adsorbed on an H 2 S-precovered surface (B {b)). The spectrum of CO adsorbed on a clean surface (A {a)) is shown for comparison. Figure 3. Infrared spectra of the exchange of 13CO(g) and the reaction of 02 with Rh/alumina that had been exposed sequentially to H2 S and 12 CO. A. A CO-precovered surface exposed to H2 S for 120 hr [same as Fig. 2A(c)] {a) evacuated for 1 hr. then exposed to 10 Torr of 13 CO .for 20 hr; (b) the sample in (a), evacuated for 0.2 hr, then exposed to 10 Torr of 0 2 for 20 hr. -B. (a.} An H2 S-precovered surface exposed to 2 Torr of 12 CO for 20 hr [same as Fig. 2B{b)]; {b) the sample in (a) evacuated for 2 hr. then exposed to 10 Torr of 13 CO for 20 hr; (c) the - 113sample in {b) evacuated for 0.2 hr, then exposed to 10 Torr of 0 2 for 20 hr. t-rj ~ (I) ~ (JQ ...... 0.8 0.0 3500 3000 ', 2000 ,_ ._..., ' 1800 ,.,-, rr2067 1 WAVENUMBER (cm 1) 2500 I \ ~ 0.2 I \ I \ (f) 0::: 0 I I II I \ background (b) CO-precovered adsorbed CO Rh/alumina r A------------, (c) alumina (a) Rh /alumina m .. w u z 04 <t . go.6 0 ....-0 '-..... ....- 1600 I 1630 H2S Adsorbed on Various Surfaces at 300K 1400 I I ~ ~ ~ - 115- A. H 2 S on 12 CO- precovered Rh /alumino - lf- Resolution = 2.6 cm- 1 (c) + H2 S, 120 hr (b) + H2 S, I. 5 hr ~--~r--(a) 12 CO 0 r-2079 01 0 /' j \ w u z <{ CD a:: B. I 2067~ \ 12 CO on H 2 S - precovered Rh /a lumina r-2036 ~~(b) +'2co / 'J \ (a) H2S ...,./ 0 (f) CD <{ 2103 I 2032 I C. Difference Spectrum r\ / .:: 2000 1900 WAVENUMBER (cm Figure 2 1800 1 ) -116- 1.0 0.8 13 co exchange and 0 2 exposure A. H2S after 2088 I 2co- precoverage 1 0.6 a ~ ~ 0 0.4 Ol 0 . 0.2 w u z 1820 <( CD 0:: I 0 (f) 0 CD <( ~-----+------~------+-----~------~ 20791 0.4 0.2 0 2100 2000 1900 WAVENUMBER (cm- 1) Figure 3 1800 - 117- Table 1 The Amounts of H S and CO Taken up by Various Surfaces and of Gases 2 Evolved during their Adsorption at 300 K. (a-c) Adsorbate H S 2 Time of Exposure (h r) Uptake Rh/alumina 0.25 10 160 220 CO-p recovered 0.25 1•5 10 104 130 165 A1umi na 0.25 10 27 36 Rh/alumina 0.25 20 22.4 268 H S-precovered 2 Rh/alumina 0.25 19 <1 20 41 <1 0.25 20 <1 <1 Surface Rh/alumina co Alumina (a) Values are in 1-lmol/g ± 5 1Jmol/g. (b) Rh concentration = 195 1-lmol/g. (c)PH S = 4.5 Torr; Pco = 2."5 Torr. 2 Preadsorbate Displaced into Gas Phase H2 (g) Evo,lved 51 61 14 25 55 8 14 25 <1 <1 <1 -118- Table 2 Hydrogen Balance for Adsorption of H S at 300 K and Desorption of Species 2 at 373 K from Various Surfaces(a~b) Iotal H X( e) Produced 2 Desorbed SEecies fraction of H S Uptake H S H o(dJ H2 2 f-mo 1/g 2 2 H0 2 Uptake H Evolved 2 at 300 K Rh/alumina 220 61 38 70 5 174 .79 CO-p recovered Rh/alumina(c) 165 25 32 15 30 102 .62 alumina 36 <1 28 6 <1 34 .94 Surface (a) Values are in flmo 1/g ±. 5 fAmO 1I g . (b) Rh concentration = 195 lJmol/g. (c)CO uptake 268 lJmol/g; 55 1Jmol/g of CO displaced at 300 K; 138 and 11 1Jmol/g of CO and co , respeGtively desorbed at 373 K. 2 (d) ±25 11mol/g. (eJx = s, o, - . -119- Table 3 Calculated Distribution of Rh and CO for CO-covered Rh/alumina{aJ Rh Distribution 11mol/g co distribution )lmol/g dicarbonyl 1"12 dicarbonyl 224 clustered 59 1 i near 30 inactive 23 bridging 15 total 194 total 269 (a) Total CO/Total Rh = ] . 38. - 120- APPENDIX: C2 H2 Adsorption and Interaction with CO on Rh/ Alumina - 121- 1. Introduction Heterogeneous and homogeneous catalysts containing Rh are among the most active for hydrogenation and dimerization of unsaturated hydrocarbons (1-3). It has also been found that the activity for hydrocarbon reactions may depend on the oxidation state of the Rh. For zeolite-supported Rh, Rh 0 was much more active than Rh1 for hydrogenation of C2 H 2 , while Rh1 was more active than Rh0 for the dimerization of C2H4 (4). On alumina, Rh also exists as Rh0 and Rh1 (5). Rh0 adsorbs CO in the linear and bridging modes whereas Rh1 forms a dicarbonyl species (5). They can be distinguished by their vibrational spectra (6-8) . We report here a preliminary study of the adsorption of C2 H2 and its interaction with CO on alumina-supported Rh using infrared spectroscopy and measurements of the amounts of C2H 2 and CO taken up. 2. Experimental Procedures The preparation of the 2.0% Rh/ alumina and the procedures for recording the infrared spectra were described in Chapters 4 and 5 of this thesis. The alumina used in this study had a BET surface area of 115 m 2 /g . The C2 H2 (Matheson, C.P.) was purified by three freeze-pump-thaw cycles and sublimed at 178 K. All adsorption measurements and treatments of the samples after the initial preparation of the Rh) alumina were made at 300 K. 3. Results A. Uptake of C2 H 2 After exposure of Rh/ alumina to 2 Torr of C2 H2 at 300 K for 38 hr. the sample and the gases above it were exposed to a trap at 77 K for 5 min. The condensed material was analyzed by gas chromatography and 1.5 and 0.1 ,umol/ g of C2H4 and C2 H 6 , respectively, were found. We assume that components other than - 122- C2 H 2 are not present in the gas phase and calculate the amount of C2 H 2 taken up by the sample, i.e. removed from the gas phase, .f rom the decrease in pressure during exposure to C2 H 2 • Data for the uptake measurements are listed in Table 1. The uptake of C2 H 2 is approximately linear in time 11'2 from....., 15 hr to at least 38 hr. as shown in Fig. 1. After 38 hr. the uptake is 211 ,umol/ g, which is equivalent to a ratio of molecules of C2H 2 taken up to Rh atoms in the sample of 1.09. A ratio of 1.37 was reached for another sample that exhibited the same type of behavior. The sample was then evacuated for 3 ·h r at 300 K and reexposed to 2 Torr of C2 H 2 . After ,. . . 20 hr. the uptake again became linear in time 11'2 Less detailed studies of C2 H2 adsorption on alumina and CO-precovered Rh/ alumina were carried out and the results are illustrated in Fig. 1. After exposure of alumina to 1.0 Torr of C2H2 for 30 hr. only 15 ,u mol/g of C2H2 are taken up. After exposure of CO-p recovered Rh/ alumina to 1.5 Torr of C2H2 for ..... 15 hr. 134 ,umol/g of C2 H 2 are taken up, which is 72.% of that taken up in 15 hr by clean Rh/alumina. B. Infrared Spectra and Uptake Measurements of C2H2 and CO Adsorbed Sequentially A Rh/ alumina sample was exposed to 5 Torr of 12 CO for 15 hr, at which point the amount of CO taken up was 284 ,umol/ g, or 1.46 molecules per Rh atom in the sample. The changes that occur in the infrared spectrum in the region recorded, 4000-1200 em -l, compared to before exposure to CO, are due to the CO stretching modes of CO adsorbed · on the Rh, as shown in Fig. 2A( a). The infrared spectrum of CO adsorbed on Rh/ alumina has been explained in previous studies (6-8). The intense peaks at 2103 and 2034 cm- 1 are assigned to the - 123dicarbonyl species and the broad peak at "-~ 1855 cm-1 is assigned to CO adsorbed in a bridging configuration on clustered Rh. The presence of a peak from CO adsorbed linearly on the clustered Rh may be inferred from the loss of intensity at 2070-2060 cm- 1 upon exposure of CO-precovered Rh/ alumina to 10 Torr of 0 2 at 300 K. This treatment removes the linear and bridging CO but not the dicarbonyl species (8-10). The sample which gave the infrared spectrum in Fig. 2A(a) after exposure to CO, was evacuated for 0.2 hr and then exposed to 1.5 Torr of C2 H 2 at 300 K. The uptake of C2 H2 on this sample after 15 hr was 134 ,umol/ g. or 0.69 molecules per Rh. During this time, the amount of CO displaced into the gas phase was 32 ,u.mol/g, or 11.% of the CO previously adsorbed on the Rh/alumina and equivalent to a displaced CO to Rh ratio of 0.16. The infrared spectrum after exposure to C2H2 for 15 hr is shown in Fig. 2A(b). The peaks of the dicarbonyl have changed little by exposure to C2H 2 . The band of the bridging CO has decreased in intensity and shifted lower in frequency while that of the linear CO has decreased in intensity and/ or shifted lower in frequency. A Rh/ alumina sample was exposed to 2 Torr of C2 H2 for 10 hr, at which point the ratio of C2 H2 molecules taken up to the number of Rh atoms in the sample was 1.37. Without evacuating the sample, the infrared spectrum from 4000 to 1200 cm- 1 was recorded. No changes were observed compared to before C2H2 was adsorbed on the sample [(a) in Fig.~B]. The sample was then evacuated for 3 hr. exposed to 2.5 Torr of 12 CO. After 15 hr. 82 ,umol/g of CO had been taken up, or 29.% of that by a clean surface in the same time. At this point the infrared spectrum was recorded. All the changes occurred between 2200 and 1700 em - l and are shown in Fig. 2B(b ). The peaks of the dicarbonyl species at 2099 and 2030 cm- 1 , and the band of bridging CO at - 124- ..... 1830 cm- 1 are evident. They are, however, substantially reduced in intensity, those of the dicarbonyl by ..... 70%, compared to the intensities of the peaks from the dicarbonyl and bridging CO adsorbed on clean Rh/ alumina shown in Fig. 2A(a). Changes in the infrared intensity from the linearly-adsorbed CO are difficult to follow in the spectra presented due to the much greater intensity of the dicarbonyl doublet. 4. Discussion Preadsorbing C2H2 on Rh/alumina was found to inhibit the adsorption of CO on the dicarbonyl-forming Rh1 and on the clustered Rh. Thus, to inhibit adsorption of CO on both types of Rh, C2 H2 must also interact with both types of Rh. Only 1% of the C2 H2 that is taken up is hydrogenated to C2 H4 or C2 H6 • The lack of hydrogenation is in contrast to the results for Rh/ silica exposed to C2 H2 (11}. The C2H2 is presumably hydrogenated by adsorbed hydrogen from acetylene that had dissociated. In the present study, then. either little of the C2H2 dissociated, or the C2 H4 or C2 H6 that would be expected to form reacted further. We obtained values of molecules of C2 H2 taken up per Rh atom in the sample as large as 1.37, with no indication that the uptake had reached a maximum. I The amount of CO taken up by the sample was 1.46 molecules per Rh atom. The ratio of CO molecules adsorbed on the Rh clusters to clustered Rh atoms is only 0.75 ( S }. The _greater overall ratio observed is due to formation of dicarbonyl species in addition to the adsorption of CO on clustered Rh. Formation of a species with more that one acetylene bound to a single Rh atom could explain the large values of C2 H 2 taken up by a clean Rh/ alumina surface. Examples of organometallic complexes with more than one C2 H4 bound to a single Rh atom - 125- Considering the adsorption of C2 H2 on CO-p recovered Rh/ alumina, however, this explanation would appear unlikely. A sample which had adsorbed 1.46 molecules of CO per Rh was subsequently exposed to C2 H2 for 15 hr . The ratio of CO displaced into the gas phase per Rh was 0 .16. However, there were 0.69 molecules of C2 H2 taken up per Rh, or 72% of the C2 H2 taken up in the same time by clean Rh/alumina. The infrared spectra in Fig. 2A supports the idea that the adsorbed CO has not been perturbed greatly by exposure to C2 H 2 . Thus an average of 1.99 molecules of CO + C2 H2 would be bound per Rh , c·a using very strong steric interactions. An explanation for the large values of C2 H 2 taken up per Rh which seems more likely is that the C2 H 2 is oligomerizing or polymerizing. During hydrogenatio~ of C2 H2 over Rh/ alumina, C4 hydrocarbons were observed (12). Also, after exposure of Rh/silica to C2 H2 , C4 hydrocarbons and benzene could be desorbed by evacuating at 291 K (11). Preadsorbing CO on the Rh/silica did not greatly affect the amount of benzene formed (12). Also, preadsorbed CO on Rh/ alumina did not strongly affect the subsequent uptake of C2 H2 in this study. Oligomerization or polymerization of C2 H 2 would not greatly relieve the steric crowding on the Rh if all of the product was bound to the Rh. For example, the saturation coverage of benzene chemisorbed on the Ni( 100) and ( 1 1 1) surfaces at 300 K is one molecule per 8 and 12 Ni surface atoms, respectively (13). However, the benzene may migrate onto the alumina support or desorb into the gas phase in the present study. The linearity of the amount of C2 H2 taken up versus (time of exposure) 1AZ would be consistent with the uptake being limited by difiusion of benzene from a constant number of Rh sites onto the support. This is not the only explanation of the shape of the uptake versus time, and the dependence of the shape on temperature and C2 H2 pressure would certainly be required to fully characterize the kinetics of the uptake process . - 126- It is interesting that Reid et al. (16) also proposed that surface species migrated from the Rh onto the silica or alumina support. They proposed, however, that the species that migrated was C2 H2 , rather than a product of oligomerization. It is not clear why the migration of C2 H2 would be facilitated by the Rh, as they observed. None of their data contradicts the migration of a product of oligomerization, rather than C2 H2 . From the infrared spectra in Fig. 2A, we saw that CO adsorbed on the clustered Rh was perturbed by subsequent exposure to C2 H2 while the dicarbonyl species was apparently nearly unaffected. Preadsorbing CO on Rh/ alumina only reduced the uptake of C2 H2 by"' 28% . Also , preadsorbing CO on Rh/ silica did not afiect the formation of benzene greatly (11). Thus , we suspect that there is much oligomerization and/ or polyrnerization of C2 H 2 on clustered Rh. This is somewhat surprising since the clustered Rh is most likely in the zerovalent state (5), and Rh1, but not Rh0 was found to be active for C2 H 4 dimerization (4). The final point we comment on is the lack of change in the infrared spectrum after as much as 1.37 molecules of C2 H2 were taken up by Rh/alumina. We do not wish to discuss the reason we did not observe changes, since the factors that determine the extinction coefficients of bands in infrared spectra in general (15) and in infrared . spectra of surface species in particular (16.17) are not well understood. We wish only to point out that considerable caution must be used in interpreting the intensities and absence of bands in the infrared spectra of adsorbed hydrocarbons. Conclusions We have studied C2 H2 adsorption and interaction with CO on aluminasupported Rh with infrared spectroscopy and measurements of the amounts of - 127- C2 H2 and CO taken up. We find that: 1. C2 H 2 adsorbs on both the dicarbonyl-f"orming and clustered Rh. 2. There is oligomerization and/ or polymerization of the C2 H2 upon adsorption on the Rh, at least some of which takes place on the clustered Rh. Acknowledgments This work was supported by the Office of Naval Research under contract N00014-79-F-0014. - 128References 1. R. S. Mann and T. R. Lien, J. Catal. 15, 1 (1969). 2. R. E. Harmon, S. K. Gupta and D. J. Brown, Chern. Rev. 73, 21 (1973). 3. R. C:ramer, J. Am.er. Chem. Soc. 87, 4717 (1965). 4. Y. Okamoto, N. Ishida, T. Imanaka and S. Teranishi, J. Catal. 58, 82 (1979). 5. Chapter 4 of this thesis, and references therein. 6. A. C. Yang and C. W. Garland, J. Phys. Chern. 61, 1504 (1957). 7. J. T. Yates, Jr., T. M. Duncan, S. D. Worley and R. W. Vaughan, J. Chern. Phys. 70, 1219 (1979). B. M. Primet, J. Chern. Soc. Faraday Trans. I 74, 2570· (1978). 9. R. R. Cavanagh and J. T. Yates, Jr., J. Chern. Phys. 74, 4153 (1981). 10. C. A. Rice, S. D. Worley, C. W. Curtis, J. A. Guin and A. R. Tarrer, J. Chern. Phys. 74, 6467 (1981). 11. J . U. Reid, S. J. Thomson and G. Webb, J. Catal. 30, 378 (1973). 12. G. C. Bond and P. B. Wells, J. Catal. 5, 419 (1966). 13. J. C. Bertolini, G. Dalmai-Imelik and J. Rousseau, Surface Sci. 67, 478 (1977). 14. J. U. Reid. S. J. Thompson and G. Webb, J. Catal. 30, 372 (1973). 15. D. Steele, Advan. Infrared Raman S~. 1, 232 (1975). 16. L. H. Little, Infrared Spectra of Adsorbed Species, p. 382, Academic Press, NewYork, 1966. 17. H . A. Pearce and N. Sheppard, Surface Sci. 59, 205 ( 1976). - 129- Table 1 . The Amounts of c H and CO Taken up by Var1ous Surfaces at 300 K. (a ~ b) 2 2 Time of Exposure Adsorbate Surface (h r) Pressure (Torr) Uptake (]Jmo 1/g) Rh/alumina 15 38 2.0 2.0 186 211 15 1. 5 134 alumina 15 1•0 15 Rh/a1umina 15 5 284 c2 H2-precoverd 15 2.5 82 Preadsorbed CO Displaced into Gas Phase (pmo 1/g) CO-p recovered 32 Rh/alumina co Rh/alumina (a )Rh concentration = 195 ~mol/g. (b)The measurements were made manometrically assuming that only c H , 2 2 c2 H4 , c2 H6 and CO were present in the gas phase. - 130- Figure CaptionsFigure 1. C2H 2 taken up by various surfaces at 300 K. The uptake was measured manometrically, assuming that only C2 H 2 , C2 H 4 , CO and C2H 6 were present in the gas phase. The gas pressures during uptake on the Rh/alumina, CO-precovered Rh/alumina and alumina were 2.2-1.8, 1.5-1.4 and 1.0-0.9 Torr, respectively. The Rh concentration was 195 ,umol/ g. Figure 2. Infrared spectra for C2 H 2 and CO adsorbed sequentially on Rh/ alumina at 300 K. A. (a) Clean Rh/alumina exposed to 5 Torr of 12 CO for 15 hr; (b) the sample in (a) evacuated for 0.2 hr, then exposed to 2 Torr of C2 H 2 for 15 hr. B. (a) Clean Rh/alumina exposed to 2 Torr of C2 H 2 for 10 hr; (b) the sample in (a) evacuated for 0.2 hr, then exposed to 2.5 Torr of CO for 15 hr. During all of these treatments, there were no changes in the infrared spectra between 4000 and 1200 em - l at frequencies outside the region shown. - 131- time of adsorption I hr 10 20 80 40 , , """ 160 / 200 1.0 160 0.8 ......... l. Rh/alumina 00 l r-i 0 s::(.. ........... QJ ~ CJ N 120 N evacuation co 0.6 ~ ........... :::0 ::r 0.. :::1 N ;:I:: ~ N c.:> ::I: 80 0.4 CO-precovered Rh/alumina --- 40 0.2 alumina ~ 0 0 20 40 60 ;,; 2 (time of adsorption) Figure 1 80 ;,; I min 2 0 100 -132- 1.2 2103 I 2034 A. C 2 H 2 on 12 CO-precovered Rh/alumina I 0.8 -' 0 0 O'l 0 .. 0.4 w u z <( en 0.0 0:: 0 (/) (]] B. <( 0.4 2099 I (\ / \ I 2030 1 '- ./ CO on C2H2 - precovered Rh/alumina I /\ 12 _...-/(b) +CO \~ ~(a) C H 2 1830 ---- 0.0 2100 2 I I 2000 1900 1800 1 WAVENUMBER (cm ) Figure 2 - 133- CONCLUSIONS - 134- In this thesis, . the binding of small molecules to transition metals was studied with the goal of better understanding the fundamental interactions between reactants and catalysts. In Part I, the bonding of the CO in metal carbonyls was examined by determining the 15 C nuclear magnetic resonance chemical shift tensors of several metal carbonyls. It was found that for CO bound terminally, the electron orbitals are symmetric about the C-0 inter- nuclear axis. In contrast, for CO bound in a bridging configuration between two or three metal atoms, there is significant asymmetry in the electron orbitals.Both types of bonding are unlike that in organic carbonyls. In Part II of this thesis, the surface chemistry of rhodium supported on alumina was studied using infrared spectroscopy and quantitative measurements of gases taken up and evolved during various procedures. In the first study, the deactivation of the CO adsorption sites upon heating the alumina-supported Rh was found to depend strongly on the gas phase present while heating. The deactivation was much greater if the sample was heated in CO or C0 2 than in 0 2 , H2 or in vacua. The specific reactions that the gases undergo, rather than simply the reducing or oxidizing nature of the gas determines the extent of deactivation. In the second study, another deactivation mechanism was examined, the poisoning of CO adsorption sites of alumina-supported Rh by exposure to H 2S. Though the adsorption of CO was not completely inhibited by H2S, nor was the ability of adsorbed CO to exchange with gas phase CO, the reactivity of adsorbed CO toward oxygen was greatly reduced. The particular resistance of Rh catalysts to S0 2 poisoning cannot be due to an inability of the dicarbonyl-forming Rh to bind sulfur, as might have been speculated. Some. of the present results may be relevant to the use of Rh- - 135- containing catalysts. For example, in reactions in which the presence of dicarbonyl-form.ing Rh is advantageous, it would be important to limit the temperature during the reaction or the concentrations of CO, C0 2 and perhaps H2 0 and to limit the concentration of sulfur-containing com- pounds. It is hoped that the value of identifying various surface species , as well as estimating their concentrations, has been demonstrated . The observation of a surface species does not, however, imply that it is an active int ermediate in the catalyzed reaction. It is quite possible that it is involved in an unproductive side reaction. Changes in its concentration during a reaction must be consistent with the kinetics of the overall reaction if it is an active intermediate. Also, isotopic labeling of the reactants and analysis of the distribution of the label in the products can assist in demonstrating the involvement of a surface species in a reaction. In the future, we should expect a trend toward studies in which surface species are observed while kinetic and/or isotopic measurements are made in order to determine mechanisms of catalytic reactions.