Langmuir 1987, 3, 395-400
395
eV) films are associated with Fe d-d transitions. The absorption of Fe3+ decreases sharply upon hydration.
in the passive layer.
Conclusions (1)Reversible formation of two monolayers of Fez+film and the quasi-reversible hydration of Fe3+film have been demonstrated. (2) Fe2+-Fe3+conversion between -0.3 and -0.2 V/NHE is associated with constant film thickness region. Presence of oxygen reduces the conversion giving rise to a less hydrated thin film. (3) Hydration of the passive layer following Fe2+-Fe3+ conversion attains 80% of water content. (4) Absorption spectra for Fez+ (2.05 eV) and Fe3+(2.3
Acknowledgment. We are grateful for support from the Office of Naval Research and for the planning discussion with Dr. Frank Herr. We also express our appreciation to Prof. M. Hall for discussions of molecular orbital interpretations. We thank E. Halverson from the Surface Science Center a t Texas A&M University for carrying out some of XPS measurements and his help in data analysis. Registry No. Fe, 7439-89-6; Na2B407,1330-43-4; H3B03, 10043-35-3; 0 2 , 7782-44-7.
X-ray Photoelectron Spectroscopic Characterization of CO Bonding Sites on Supported Small Rhodium Clusters G . Apai* Corporate Research Laboratories, Eastman Kodak Company, Rochester, New York 14650
B. G. Frederick Department of Chemistry and Cornell Materials Science Center, Cornell University, Ithaca, New York 14850 Received July 3, 1986 Vapor deposition of Rh4(CO),,and Rh(CO),G clusters onto UHV cleaned carbon and planar amorphous alumina supports at low temperatures produced two bonding states of CO which have been identified from XPS core levels as terminal and bridging CO species. The C 1s and 0 1s binding energies for the carbonyl ligands agree closely with the assignments for on-top and bridging CO on rhodium single-crystalsurfaces. Differences in the binding energy splitting between terminal and bridging for the two clusters may be ascribed to the two forms of bridging species, edge-bridging and face-bridging CO on Rh4(C0)12and Rh6(C0)16, respectively. Preferential population of bridging CO ligands over terminal ligands occurs during progressive decomposition of the cluster under UHV conditions. High Rh 3d binding energies for the fully decarbonylated clusters indicate that the clusters remain small during CO loss.
Introduction Comparisons between adsorbate behavior on small metal particles and adsorbate behavior on surfaces of bulk metals are fundamentally important for a systematic understanding of heterogeneous catalytic reactions. The vast majority of UHV surface spectroscopic experiments have used single-crystal surfaces to characterize bonding sites of adsorbed atoms and molecules. Bonding site studies, with systems consisting of only a few metal atoms with well-defined geometry, could provide the information necessary to make this comparison and also be of practical interest, because the supported clusters represent model heterogeneous catalysts. Metal carbonyl clusters provide a straightforward way of investigating both size and adsorbate effects. X-ray photoelectron spectroscopy (XPS) is particularly useful for comparisons of the type proposed because identification of bonding sites, especially for CO species, and quantification of peak intensities is straightforward, provided the photoemission lines are not complicated by energy-loss features. For single-crystal studies, XPS has been used to identify differences in core-level binding energies for molecules bonded in different adsorption sites.1.2 Recently, DeLouise et al.3p4 (1) Norton, P. R.; Goodale, J. W.; Selkirk, E. B. Surf. Sci. 1979,83,189. (2) Kiskinova, M.; Pirug, G.; Bonzel, H. P. Surf. Sci. 1983, 133, 321.
0743-7463/81/2403-0395$01.50/0
characterized CO bonding sites on Rh single crystals using XPS. Few XPS studies of transition-metal carbonyl complexes exist which identify species in different bonding sites. Several studiesk8 of gas-phase carbonyl complexes have reported core levels for both terminal and 2-fold bridge-bonded CO. However, we find no XPS references concerning bonding site information on supported clusters or face-bridging CO species that could provide information about hollow-site bonding on single-crystal surfaces. Therefore, in this study we investigated the discrete bonding sites for CO ligands associated with supported carbonyl clusters and characterized what effects the variations in electronic structure due to cluster decomposition might have had on CO bonding modes.
Experimental Section Experiments were performed in a VG ESCALAB Mark I1 spectrometer equipped with a Mg K a X-ray source and multi(3) DeLouise, L. A.; White, E. J.; Winograd, N. Surf. Sci. 1984, 147,
ncn
,504.
(4) DeLouise, L. A.; Winograd, N. Surf. Sci. 1984, 138, 147. (5) Plummer, E. W.; Salaneck, W. R.; Miller, J . S. Phys. Reo. B 1978, 18, 1673. (6) Avanzino, S. C.; Jolly, W. L. J . Am. Chem. SOC.1976, 98, 6505. (7) Chen, H. W.; Jolly, W. L.; Kopf, J.; Lee, T . H.J. Am. Chem. SOC. 1979.101. 2607. (8) Avkzino, S. C.; Jolly, W. L.; Malmquist, P.-A.;Siegbahn, K. Znorg. Chem. 1978, 17,489.
0 1987 American Chemical Society
396 Langmuir, Vol. 3, No. 3, 1987 I
'
I
'
Apai and Frederick I
a) Rh&O),,/C
'
I
'
I
a ) Rh6(CO)i6/C T S 190K
Rh 3ds
I
I
I
01s
T
I
-
fn cC
317
313
309
305
Binding energy (eV) Figure 1. Rh 3d core-level spectra for Rh4(CO)=and Rh&!O)16 chemisorbed on amorphous carbon surfaces at temperatures below 200 K. The Rh 3d spectrum for Rh4(C0)12in part b shows broadening on the low-bindingenergy side of the core levels attributable to slight decomposition during transfer from the preparation chamber to the analyzer chambers when the sample could not be continuously cooled. channel detection. The analyzer resolution for data acquisition was -0.2 eV. Operating pressures were typically in the low mbar range. Samples were prepared by using vapor deposition of rhodium carbonyls onto cold supports. Rh4(C0)12and R b (CO),, cluster complexes were obtained from Strem Chemicals. This method proved important because we have found that other methods, such as powder on tape, powder films formed from a slurry, or solution impregnation onto powders, produced excess carbon contamination and decomposition of the compound in UHV because of introduction to vacuum at ambient temperature. These methods have been used by several groups, including Conrad et al.? Andersson et al.,'O and Wehner et al." However, the studies have neither reported nor discussed C 1s or 0 1s core-level spectra for carbonyl species. Without a more comprehensive analysis of the integrity of the supported clusters, the transition-metal binding energies previously reported may not be indicative of the intact carbonyl clusters. Rh&O), has a vapor pressure which is high enough to evaporate in UHV without heating above room temperature. Rb(CO)16 clusters were evaporated at 373 K, which is a typical evaporation temperature for a number of carbonyl complexes. Similar evaporation methods have been reported by Legare et al.12 for various transition-metal carbonyl complexes which contained only terminal CO ligands. However, their studies focused primarily on electron-beam decomposition at low temperatures to investigate aggregation of the metal on various surfaces. In our work, the substrates were maintained at temperatures lower than 200 K during and after the vapor depositionsto elucidate ligand-metal bonding sites and changes in ligand geometry during decomposition. Two different supports, carbon and alumina, were chosen to investigate both the carbonyl C 1s and 0 1s peaks. Both were prepared in a planar format to eliminate inhomogeneous charging problems. A clean, amorphous carbon substrate was formed from a sheet of graphite by argon sputtering. A clean, unhydroxylated amorphousalumina surface was prepared by sputtering 5 N aluminum foil to eliminate
-
(9)Conrad, H.; Ertl, G.; Knozinger, H.; Kuppem, J.; Latta, E. E. Chem. Phys. Lett. 1976, 42, 115. (10) Andersson, S.L.T.;Watters, K. L.; Howe, R. F. J. Catal. 1981, 69,212. (11) Wehner, P.S.;Mercer, P. N.; Apai, G. J. Catal. 1983, 84, 244. (12) Legare, P.; Sakisaka, Y.; Brucker, C. F.; Rhodin, T. N. Surf. Sci. 1984, 139,316.
I
I
I
I
542
538
534
530
I
Binding energy (eV) Figure 2. 0 1s photoemission spectra of Fth&O)16 and Rh&O)12 chemisorbed on clean, amorphous carbon surfaces at low temperatures. Peaks T and B represent terminal and bridging CO species, respectively. S represents an 0 1s satellite peak. the 0 1s core level and then exposing the sample, heated at 373 K, to loo0 L of ultrahigh purity OF Referencing of the XPS data was made for amorphous carbon by assigning the C 1s binding energy at 284.4 eV and for thin planar aluminas by assigning the Alo 2p level at 72.7 eV. These values yielded consistent binding energies for the metal atom and carbonyl peaks. The A13+ 2p core level was not a good reference, because its binding energy varies with surface preparation (Le., variations in the stoichiometry). A thorough descriptioh of the apparatus and the technique for sample preparation is discussed ~eparate1y.l~
Results For both sputtered graphite and planar unhydroxylated alumina surfaces, the Rh 3d5,, core-level binding energy for the Rh4 and R b carbonyl clusters deposited a t low temperatures is 309.4 f 0.2 eV. This value is significantly higher than binding energies previously reportedlOJ1from solvent impregnation techniques. The rhodium atom surface coverage ranged between 1 and 3 monolayers. Typical Rh 3d spectra are shown in Figure 1. The fullwidth half-maximum (fwhm) of the Rh 3d6/2peaks is 1.2 eV, which is considerably narrower than values previously reported a t -2.2-4.0 eV.l0 At the temperatures we used (ca. 200 K), binding energies remained stable on carbon for an hour or more, whereas changes on alumina occurred within 15 min. The interactions responsible for these changes involve cluster-support interactions and their effects on chemisorbed species, which will be discussed in detail in a subsequent paper.13 The 0 is spectrum for Rh&O)16, vapor deposited onto an amorphous surface of graphite at low temperature (193 K), shows two peaks a t 534.1 and 532.1 eV. Figure 2a represents the spectrum of the 0 1s region immediately after evaporation. The two sharp peaks are assigned to terminal bonded CO a t 534.1 eV and to face-bridge bonded CO a t 532.1 eV. In addition, a broad satellite peak at 540.6 eV accompanies the main photoemission peaks. The 0 1s carbonyl spectrum for Rh(CO),, prepared under the same conditions also shows two features (see Figure N
(13) Frederick,B. G.; Apai, G.; Rhodin, T. N., submitted for publication in J.Am. Chem. SOC.
Langmuir, Vol. 3, NO.3, 1987 397
XPS Characterization of CO Bonding Sites Table 1. Binding Energy of Carbonyl Species on top/terminal bridging Rh system 0 1s c 1s 0 1s c 1s 530.7 285.3 532.1 286.1 Ru(ll1)" 531.2 BE (edge bridging) > BE (face bridging), reflects the larger degree of back-bonding for the bridging ligands. This trend is also consistent with the a-acidic nature of various CO species. Muetterties et al.15 reported that bridging carbonyl ligands, particularly face bridging, are better a-acids than terminally bonded carbonyl ligands. These trends have recently been substantiated by theoretical calculations of the electronic structure of rhodium and iridium carbonyl clusters.24 Several interesting correlations can be made from comparisons of the binding energies in Table I. Apart from the absolute binding-energy increase of C 1s and 0 1s core levels of the supported clusters compared to the values for chemisorbed CO on metallic rhodium, binding-energy differences for bonding configurations of CO can be compared. We observe that, for both CO/Rh(lll) and Rh4(C0)12/C,the difference in binding energy between terminal and bridging carbonyl 0 1s is identically 1.4-1.5 eV, while the difference in C 1s (23) Jolly, W. L.; Avanzino, S. C.; Rietz, R. R. Inorg. Chem. 1977, 16, 964. (24) Miessner, V. H. 2. Anorg. Allg. Chem. 1983, 505, 187.
Langmuir, Vol. 3, No. 3, 1987 399
XPS Characterization of CO Bonding Sites binding energies is 0.8 and 0.7 eV, respectively. The consistency of the data suggests a similarity in the bonding environment of the two bridging CO species. The bridging carbonyl in the Rh4(CO),, cluster is known to be edge bridging; the C-0 stretching frequency a t 1880 cm-' is consistent with this assignment.% High-resolution electron energy loss (EELS) measurements for CO chemisorbed on Rh(ll1) showed C-0 stretching frequencies a t 2070 cm-' assignable to atop species and a t 1870 cm-' which was assigned to a molecularly bridge-bonded species.26 The observed binding-energy difference between terminal and face-bridging co in Rh6(C0)16is 2.0 e v , indicating that hollow-site chemisorption of CO on more open Rh single-crystal surfaces should be distinguishable from both 2-fold bridge sites and terminal sites. The small binding-energy difference between terminal and bridge-bonded sites for CO/Rh(331) appears to indicate either a change in electronic interaction (i.e., back-bonding) or an interaction of the carbonyl species with atoms associated with step sites. Data concerning carbonyl clusters point to two factors that influence distribution of the carbonyl ligands between terminal and bridging modes. These are the dissipation of electronic charge and the metal atom coordination number. The former is the dominant influence. Because bridging carbonyl ligands, particularly face bridging, are better ?r-acids than terminally bonded carbonyl ligands, their population is favored with enhanced electron density from the cluster core. Thus, if we consider intact supported Rh&O),, clusters which showed the correct stoichiometry as discussed earlier, a cluster-support interaction may be responsible for the low ratio of terminal/ bridging CO species (i.e., observed ratio of 2.1:l rather than 3:1, as predicted). If the support surface were to donate charge to the cluster, then a lower ratio might be expected. In addition, as the original supported carbonyl clusters lose CO ligands, the electron density on the metal framework favors population of bridging positions to retain the highest possible level of coordination. Thus, the ratio of terminal/ bridging CO continuously decreases upon cluster decomposition. Our observation that the bridging CO C 1s of Rh4(C0),, falls closer in energy to the terminal CO peak than that for Rh6(CO)16would be consistent with the Tacidity argument discussed above. The bridging ligands for the Rh&O)16 cluster are face bridging, whereas those on the Rh4(C0)12cluster are edge bridging. Metal atom coordination number is also a factor. It has been pointed out that, for a cluster complex, the number of ligands and their mode of bonding will be influenced by the relative size of the solid angle described by the metal atom and its nearest metal atom neighbor^.'^ When the solid angle is small, the metal-ligand coordinationnumber can be expected to be higher than when the-angle is large, as in a large metallic cluster. In addition, it is known that small angles permit, but do not require, a higher relative percentage of terminally bound carbonyl ligands than observed for larger clusters. If we assume that the metal cluster size is constant during decarbonylation, it is entirely reasonable, based upon the above-described cluster pattern, that the ratio of bridging ligands to terminal ligands increases as the total number of ligands decreases, since the metal atoms will tend to maintain the largest degree of coordination possible. This is quite different from metallic Rh(ll1) and Rh(331) surfaces where bridging CO species are not pop(25) Creighton, J. A.;Heaton, B. T. J . Chem. SOC., Dalton T i a m . 1981, 1498. (26) Dubois, L.H.; Somorjai, G. A. Surf. Sci. 1980, 91, 514.
ulated until high levels of surface CO c0verage.4~ The facile transition of CO ligands from a high terminal to bridging ratio into a more evenly distributed ratio up0.n ligand desorption is not totally unexpected. In transition-metal carbonyls, evidence exists for fluxional behavior (i.e., scrambling of linear and bridge CO species).*' 13C NMR line shapes for CO chemisorbed on supported Rh clusters are also interpreted as being due to this kind of fluxional behavior.28 Therefore, the change in the ratio of terminal to bridging CO ligands may only be interpreted as a change in the equilibrium population of the two sites, and not necessarily as a difference in the desorption mechanism for the two sites. The relatively constant splitting between terminal- and bridge-bonded 0 1s peaks over a wide range of decomposition indicates that the ligand environment (that is, the cluster framework) is not dramatically altered. Only in the sample warmed to room temperature does the splitting decrease from 1.9 to 1.5 eV. This may, in some way, be attributable to possible cluster rearrangement or aggregation. The Rh 3d5,, binding energy of 308.0 eV observed in the sample warmed to room temperature may be compared with the 308.2-eV value found for bare rhodium clusters of ca. six atoms. The continued presence of carbonyl ligands in the partially decomposed cluster is responsible for some part of the high binding energy. A further decrease in the Rh 3d6 binding energy to 307.6 eV under vacuum heating redects the value for a bare rhodium cluster which has undergone some agglomeration. CO satellite intensity gives some idea about cluster integrity. The intensity of the dominant satellite is larger for the C 1s level than for the 0 1s core level. A correlation of the intensity has been made with the number of metal atoms in the ~arbonyl.~ Plummer et al. point out that there is a general decrease in the satellite intensity as the carbonyl clusters get bigger. No observation of satellite peaks has been made for CO chemisorbed on Rh(ll1). They also show that the shake-up energy increases with the number of metal atoms and saturates at -7 eV for six metal atoms in the cluster. For the clusters studied, no changes in satellite intensity were observed during cluster decomposition. In addition, no changes in satellite energy relative to the primary core peaks were observed. Thus, we have corroborating evidence, in addition to high Rh 3d binding energies, substantiating that the clusters are not aggregating upon initial deposition. N
Conclusions Study of rhodium carbonyl clusters, formed by vapor deposition methods, has provided bonding site information which is normally only available from chemisorption studies done on clean, single-crystal surfaces. The cluster results are in good agreement with single-crystal results and offer further insight into the assignment of molecular chemisorption bonding sites. XPS spectra resolve peaks associated with bridge-bonded CO and terminally bonded CO. The 0 1s core-level splitting shows a larger effect than the C 1s core-level splitting, which can be accounted for in terms of initial-state differences due to bonding effects. The different 0 1s binding energies for terminal-, edge-, and face-bridging species can be ascribed to the magnitude of back-bonding for each. Binding energy and core-level intensity changes, resulting from CO desorption during cluster decomposition, indicate a preference for bridging (27) Dorn, H.; Hanson, B. E.; Motell, E. Inorg. Chin. Acta 1981,54, L71. (28) Duncan, T.M.; Yates, J. T.;Vaughan,R.W. J.Chem. Phys. 1980, 73,975.
Langmuir 1987, 3,400-404
400
CO ligands vs. terminal CO ligands. We have explained this using two general concepts from organometallic chemistry (dissipation of electronic charge and metal atom coordination number). The interaction of these carbonyl clusters with various supports, and the potential for observing interactions of the adsorbate ligands with the substrate oxide, appears to offer potential. Studies of the interaction of carbonyl
clusters with oxide supports are in progress.
Acknowledgment. Support from Eastman Kodak Company, Cornel1 Materials Science Center, and NSF DMR-83-03742is D a t e f a y acknowledged, 85 well as is the Participation of Prof. N. RhodinRegistry No. CO, 630-08-0; Rh, 7440-16-6; Rh4(C0)12, 19584-30-6;Rh&O),,, 28407-51-4;C, 7440-44-0;A1203,1344-28-1.
Oscillating Behavior of the Iron Anodes: Effect of Electrode Wetting 0. Teschke,* F. Galembeck, and M. A. Tenan Instituto de Fisica e Instituto de Quimica, Universidade Estadual de Campinas, 13100 Campinas, SP,Brasil Received July 25, 1986. I n Final Form: November 25, 1986 The interdependence of electrode wetting and electrochemical oscillation is studied for iron anodes in aqueous sulfuric acid solutions. The effects of solution concentration and of surface-active agents on the oscillations were measured. The results show that the oscillation parameters (oscillation frequency and current amplitude) are dependent on the electrode wettability. A mechanism for passivating film stripping is advanced, based on the Marangoni effect.
Introduction Chemi~all-~ and enzyme-catalyzed reactions4g5may take place in a periodic manner. There are many electrochemical systems which also display oscillatory behavior.6s A large proportion of the cases of periodicity in electrochemical systems described in the literature are concerned with oscillations encountered during anodic polarization of various metals, in particular, the iron electrode in sulfuric acid so1utions.+l2 These oscillations are directly connected with the characteristic instability of passivating films. The iron electrode reactions which occur at voltages close to the current oscillation region in 1 M H2S04solutions are the following: (a) Below +250 mV vs. SCE the electrode is active. It is anodically dissolved with the formation of Fe2+ions. Above -140 mV vs. SCE the dissolution rate increases, reaching values of 0.1-1.0 A/cm2. At about these dissolution rates the solubility product of ferrous sulfate is reached near the surface, and crystals of this compound (1) Hedges, E. S.;Myers, J. E. The Problem of Physico-Chemical Periodicity; Longmans Green: New York, 1962. 1921, 43, 1262. (2) Bray, J. J . Am. Chem. SOC. (3) Pearls, M. G.; Cullis, C. F. Trans. Faraday SOC.1951, 47, 616. Fields, R. J.; Kovos, E.; Noyes, R. M. J . Am. Chem. SOC.1972,94,8469. (4) Higgins, J. Proc. Natl. Acad. Sci. U.S.A. 1964, 51,'989. (5) Higgins, J. J. Ind. Eng. Chem. 1967, 59, 18. (6) Bonhoeffer, K. F.; Langhammer, G. Z. Elektrochem. 1948,52,60, 67. Bonhoeffer, K.F.; Gerisher, H. Z. Elektrochem. 1948,52, 149. (7) Liu, S. W.; Keizer, J.; Rock, P. A.; Stenschke, H. R o c . Natl. Acad. Sci. U.S.A. 1974, 71, 4477. (8) Wojtowicz, J.; Marincic, N.; Conway, B. E. J. Chem. Phys. 1968, 48, 4333. (9) Gilroy, D.;Conway, B. E. J.Phys. Chem. 1965, 69,1269. (10) Thalinger, M.; Volmer, M. Z.Phys. Chem. 1930, 150, 401. (11) Russel, P. P.;Newman, J. J . Electrochem. SOC. 1983, 130, 547. (12) Epelboin, I.; Gabrielli, C.; Keddam, M.; Lestrode, J. C.; Takenouti, H. J . Electrochem. SOC. 1972, 119,1632.
0743-7463/87/2403-0400$01.50/0
nucleate and spread over the surface. Despite the saturation in the solution near the surface, the sulfate layer remains porous and allows the reaction to proceed. (b) Polarization to about +300 mV vs. SCE results in the sudden replacement (within milliseconds) of the sulfate layer with a porous free thin oxide layer. The growth of oxide is accompanied by a decrease of reaction rate by a factor of lo6, and the metal becomes passive. A model for the oscillation phenomena occurring on anodically passivated electrodes was proposed by Frank et al.13 The fundamental features of the model are the following: When the electrode is in the active state, a high current density flows, making the concentration of H+ ions around the electrode vary rapidly. When the electrode becomes more positive than a threshold value (transition potential), passivation occurs. However, the concentration of H+ ions is soon rebuilt due to diffusion; the electrode is then activated and electrical current increases starting a new oscillation cycle. Current oscillations are thus dependent on the coupling between the rate of the reactions taking place a t the electrode and the H+ transport processes; changes in the local pH induce the switching between the two states. More recently, we have shown that iron anodes in aqueous sulfuric acid also display capillary ~ s c i l l a t i o n s . ~ ~ This means that the liquid level moves up and down around a half-immersed iron wire electrode, in phase with the current-time oscillations. The correlation between capillary oscillations and current oscillations suggests that electrochemical oscillations are determined by surface phenomena, as well as by the diffusion-limited electrochemical reactions. (13) Franck, U.F.; Fitz Hugh, R. 2.Elektrochem. 1961, 65, 156. (14) Teschke, 0.; Galembeck, F.; Tenan, M. A. J. Electrochem. SOC. 1985, 132, 1284.
0 1987 American Chemical Society
