Ultra-violet Photoelectron Spectroscopy Studies of Gas Adsorption on Transition Metals BY P. J. PAGE AND P. M. WILLIAMS* Dept. of Chemical Engineering and Chemical Technology, Imperial College Received 3rd June, 1974 The adsorption of CO and CzH4 on oriented single crystals of nickel has been studied by means of ultra-violet photoelectron spectroscopy (UPS). At He1 energies (21.2 eV), distinctly different d band densities of states are observed for the clean (1 1 l), (1 lo), and (100) faces which are interpreted in terms of bulk rather than surface band calculations. Following CO adsorption, states derived from the 772p antibonding level of CO may be distinguished in the structure close to E f in addition to the deeper 02p and r2p bonding states. The cracking of ethylene at high temperature over Ni(l10) is shown to produce graphitic overlayers. Preliminary studies of the oxidation of chromium are also reported.The study of adsorption phenomena on surfaces by means of ultra-violet photo- electron spectroscopy (UPS) has been the subject of considerable interest in recent years following the early work of Eastman.'. With ultra-high vacuum hemi- spherical analyser systems fitted with windowless discharge lamps, photons at energies of 21.2, 40.8, and 48.4 eV from helium excitation and 16.8, 26.9 eV from neon are now available. These energies permit the examination of adsorbate-induced valence levels over an energy window of 10-1 5 eV from the uppermost layers of a surface, so that direct changes in the bonding states of both adsorbate and host may be observed.Thus the adsorption of various gases on polycrystalline 3 9 and single crystal n i ~ k e l , ~ copper,6 molybdenum,' and tungsten have all been studied using UPS. In many of these investigations, an interpretation of the energy levels induced by the adsorbate has been attempted on the basis of a direct comparison with the molecular orbitals of the free molecule measured in the gas phase.9 This approach has been particularly successful, for example, in explaining the behaviour of CO on adsorption on a range of metals with relatively little adjustment in energy to the free molecular orbitals. The situation becomes more complex, however, in situations involving dissociative ad~orption,~. large hydrocarbon molecule^,^ or the possibility of more than one adsorbed * Under such circumstances, large substrate- adsorbate interactions, which may involve considerable charge transfer, may invali- date simple extrapolation from the free molecular case.A direct knowledge of the behaviour of the d electrons of the metal itself is neces- sary for a complete understanding of the nature of these interactions. In the present paper we therefore report changes within the d electron density of states of nickel between different crystal faces and following the adsorption of CO and C2H4. A preliminary extension of the work to the oxidation of chromium is also described. EXPERIMENTAL Single crystal discs of nickel ((lll), (110), and (100) orientations) and chromium ((111) orientation) were cut to within 1-2" and mechanically polished to a mirror finish on a vibratory polisher with 5000A alumina and Syton.These were welded directly to tungsten rods for 80P . J . PAGE AND P . M . WILLIAMS 81 support in the electron spectrometer (Vacuum Generators Ltd., ESCA3). Following bake- out, pressures -2x 10-lo Torr were routinely obtained, and crystals were cleaned in the specimen chamber of the instrument by repeated cycles of argon-ion bdbardrnent, oxidation, annealing and finally flashing to -1000°C. Surface impurities could be monitored by AlKa or MgKa excited XPS, although the precise form of the UPS spectrum was found to be the most sensitive test for surface contaminants. Spectroscopically pure gases (CO and C2H4) were admitted only to the specimen preparation chamber of the instrument via an all metal leak valve while the analyser remained at 2x Torr.Pressures in the analyser chamber increased to 7-8 x 10-lo Torr during running of the windowless He discharge lamp. Spectra were recorded digitally using a PDP/8f based data system,14 and all data are presented in " raw ", unsmoothed form. RESULTS CLEAN NICKEL SINGLE CRYSTALS Previous UPS work on polycrystalline nickel evaporated in ultra-high vacuum 2-4* l5 revealed considerable changes in the density of states within the d band on adsorption, in addition to the appearance of deeper lying levels. Interpretation of such be- haviour in terms of surface states has been suggested 3 9 l 5 but relatively few data are available on the density of states of pure nickel with which to compare these pre- dictions.With tungsten, the comparison of UPS spectra from different crystal faces has facilitated the identification of surface states lo* in addition to characterising features in the bulk band structure. Angular effects are important, however, for single crystal substrates when only electrons in specific directions of emission relative to the crystal surface are admitted to the analyser.12* 16. l7 In the present investiga- tion, therefore, single crystals of (lll), (110), and (100) orientation nickel were so oriented as to allow electrons emitted as closely as possible to the normal to the surface to enter the analyser. These directions were defined by a slit system rather than by a circular aperture, but the spectra do indicate a sufficiently high degree of angular selectivity despite this non-ideality.Fig. 1 compares He1 spectra recorded from the three crystal faces of Ni following cleaning, with the corresponding spectrum from a polycrystalline evaporated film for comparison. The (111) spectrum is dominated by a single state, fwhm 0.5 eV immediately below E,, with a weaker level 2.5 eV below Ef. A doublet feature with a splitting of just over 0.5 eV is seen in the (100) spectrum, whilst the (1 10) face shows three less distinct levels. The bandwidth and structure in the polycrystalline sample are consistent with an appropriately weighted sum over the single crystal densities of states. ADSORPTION OF co ON NICKEL Fig. 2-6 show the changes produced in the densities of states of the clean nickel on adsorption of CO to saturation coverage at room temperature. Thus, fig.2,3 and 4 show the He1 spectra over the d band region for (1 1 l), (1 lo), and (100) faces before and after exposure to CO. The He1 spectra from the CO saturated faces are com- pared in fig. 5. In all cases, particularly the (1 11) face (fig. 2) the effect of adsorption is a broadening in the d band together with the appearance of a lower energy broad state, 1.2 eV below Ef for (1 10) and (loo), and 1.5 eV for (1 11). He11 spectra from all three faces are similar (fig. 6) and show the now familiar 0- and n-bonded CO states at 8.0 and 11.0 eV below Ef. HIGH TEMPERATURE ADSORPTION OF C2H4 ON Ni (110) The interaction of hydrocarbons with Ni at elevated temperatures may be expected to produce complete dehydrogenation, and fig.7 shows the effect of exposing a clean82 GAS ADSORPTION ON TRANSITION METALS Ni(ll0) surface to C2H, at 600°C for 13 min at Torr. This spectrum is compared with that from vacuum-cleaved well-oriented pyrolytic graphite,' * also recorded using Hen photons. 1 I I I 1 L-f - 1 -2 -3 -4 binding energy /eV FIG. 1.-He1 UPS (21.2 eV) from (a) Ni(lll), (b) Ni(100), (c) Ni(ll0) and (d) polycrystalline Ni. Single crystals oriented so that near-normally emitted electrons enter analyser. Inset is calculated density of states from ref. (20).P. J . PAGE AND P. M. WILLIAMS 83 I 1 1 I i € f - I - 2 - 3 binding energy/eV FIG. 2.-He1 UPS from (a) clean Ni(l1 l), (6) Nit 11 1) saturated with CO (= 100 langmuirs) at 295 K. Shoulder mowed in (b) corresponds to peak in clean density of states in (a)..r-., . . . * *-.. -... . . - ... ..I. - ...--. __“C. 1 I I E f - I -2 -3 binding energy/eV FIG. 3.-He1 U P S from (a) clean Ni(1 lo), (b) Ni(ll0) saturated CO.I / 1 1 1 € f - 1 - 2 - 3 binding energy/eV FIG. 4.-He1 UPS from (a) clean Ni(lOO), (b) Ni(100) saturated CO. . .*.- . ...... ........ - 4::. .- .. - ..p*. .. .-~.%..-...*.:.I .. .- -***. (= 1 - .... . -- ,......'"-* "* (I 00) : __..I ..-,a . . . . *.n . . (110) :: ........... ..-.-.."--* . ............ -. I 8 t ET - I -2 -3 binding energyleV FIG. 5.-Comparison of CO-saturated surfaces: He1 UPS from (a) Ni(lll), (b) Ni(l10), and (c) Ni(100). The extra state near 1.2 eV below fi is seen for all faces ; feature near 0.6 eV only for the (1 11) face.P .3 . PAGE AND P . M. WILLIAMS k I I I €f -- 5 -10 -15 binding energy/eV 85 FIG. 6.-He11 UPS (40.8 eV) from CO-saturated surfaces : (a) Ni(l11), (6) Ni(lOO), and (c) Ni(l10). The 0 and n CO levels at 8.0 and 11 .O eV below E f are clearly visible.86 GAS ADSORPTION ON TRANSITION METALS OXIDATION OF CHROMIUM (111) Eastman has shown l9 that in moving from nickel (f.c.c., 3d84s2) to chromium b.c.c., 3d54s1) there is a considerable change in the density of d states as determined by UPS. It is therefore of interest to compare the effects of adsorption on these two metals. Little or no interaction could be detected with CO, but the effects of oxygen on the clean Cr (1 1 I ) surface are seen in fig. 8, where spectra excited with He1 and I I I ..... ..... .. . . a . . . ........... (b) ... (c) ......... ............... ......... ........... I I I Ef - I - 2 -3 binding energy/eV FIG. 8 . 4 ~ ) He1 UPS from Cr(ll1) following oxidation (30langmuir.s at room temperature); (b) HeII UPS from Cr following oxidation ; (c) HeII UPS from clean Cr(ll1) ; (d) He1 UPS from clean Cr. Note that a remnant of the sharp level immediately below E f persists in the He1 spectrum after oxidation. He11 photons for both clean and oxidised Cr (1 11) are shown. For the clean surface, a sharp state (fwhm -0.3 eV) is observed immediately below E,, particularly at HeI. This state is removed on exposure to O2 and a new level develops 2.0 eV below Ef. DISCUSSION CLEAN SURFACES A detailed discussion of the features in the densities of states for the clean Ni surfaces is beyond the scope of the present paper.Nevertheless, it is necessary to consider briefly the nature of the predominant d states in the spectra in the light of previous adsorption measurements on polycrystalline films and the present results on single crystals. Polycrystalline spectra (e.g., fig. Id) show a sharp peak immedi- ately below Ef followed by a broader d band and the attenuation of this sharp peak onP . J . PAGE AND P . M. WILLIAMS 87 chemisorption led to its possible interpretation in terms of a surface state.3* l 5 Spectra for the single crystal faces of Ni show considerable variation in structure with orientation, and the photo-emission from the (1 11) face in particular is dominated by a similar sharp peak at Ef. This high intensity for the (1 11) face (implying a lack of bulk contribution to the spectrum) does not in itself preclude an interpretation in terms of surface states as a similarly dominant peak on ( 1 0 ) tungsten has been so interpreted.1° However, whereas in the tungsten band structure, there is a favour- able hybridization gap immediately below Ef in which an evanescent surface mode may propagate, the only likely gap for such a state in Ni between the A; and A$ bands 2o occurs some 4-5 eV below Ef and so cannot provide a simple explanation of the sharp state.Furthermore, while the W( 100) surface state vanishes on exposure to less than a quarter of a monolayer of hydrogen, the sharp (1 11) state on Ni (fig. 2) in fact persists at saturation coverage of CO (arrowed).Its apparent attenuation relative to the second peak in the polycrystalline observations should be thought of for the CO chemisorption in terms of a relative increase in density of states structure 1-2 eV below Ef due to the appearance of the antibonding n2p level, rather than the quenching of an evanescent surface mode. The redistribution of d electrons due to population of, and admixture with the antibonding n2p level does not in itself neces- sarily imply the existence of a surface state. This interpretation removes the apparent anomaly in the behaviour of the sharp d state in the comparison of C 0 2 adsorption where, for a comparable heat of adsorption to CO, little attenuation is observed; in this case, however, there is no counterpart to the antibonding level which disturbs the d electron density of states in the CO adsorption.With H2S and 02, the drastic attenuation of the sharp d state is indicative of strong charge transfer in an ionic sense to form Ni-S or Ni-0 bonds ; the effective sampling depth of only 2-3 atomic layers ensures that the underlying metallic Ni contributes insignificantly to the observed intensity. We therefore discount the evanescent surface mode as a possible explanation of the sharp d state and consider the effect of the surface only in terms of narrowing of the bulk energy bands 21* 22 and possible s to d charge transfer.23 Such effects may be regarded as perturbations of the bulk band structure so that we may attempt an initial interpretation of the densities of states in fig. 1 in terms of the bulk density of states. Comparison with Zornberg’s 2o calculations suggests that for spectra in which only normally emitted electrons contribute, the peak in the (111) spectrum corresponds to the sharp L32 density of states maximum, and the doublet in (100) to the X2-X5 maxima ; the relative lack of sharply-defined structure on the (1 10) face is expected in view of the band dispersion along K.Inset in fig. 1 for comparison is Zornberg’s paramagnetic density of states, showing clearly the L32 maximum. This represents only a preliminary attempt at a band structure interpretation, and takes no account of exchange interaction splitting in the ferromagnetic case (which may be as large as 0.5-0.6 eV). A fuller treatment of the single crystal measurements is to be submitted elsewhere.13 ADSORPTION OF co ON NICKEL The immediately characteristic feature in the spectrum from all three crystal faces following adsorption is the appearance of levels 8.0 and 11.0 eV below Ef.These have previously been interpreted as deriving from the a2p and n2p bonding states of the CO molec~le,~’~ and their precise energies have been discussed by Eastman in terms of the relative strengths of a-d and n-d interactions between the adsorbed molecule and the surface d band. We make no further mention of them88 GAS ADSORPTION ON TRANSITION METALS here, except to note that there appears to be a slight shift of 0.2 eV in the 7c state to higher energy on the (1 10) face relative to the other two. Closer to the Fermi level, further adsorbate-induced effects are observed on all three faces.On both (100) and (1 lo), an extra level 1.2 eV below Ef masks the structure in the clean spectra; the corresponding state is observed 1.5 eV below Ef on the (111) face. This apparent shift for the (1 11) spectrum relative to the other two is not thought to indicate a genuine difference as the new level appears on a very low background density of states on (1 11) but overlaps underlying Ni structure, shifting the mean peak position to lower binding energy, on the other two faces. However, in the (1 11) case, between the shoulder in the CO-saturated spectrum, which corresponds to the peak in the clean density of states, and the level at 1.5 eV, a new band 0.6 eV below Ef appears. These observations cannot be interpreted, as was suggested in the polycrystalline case, simply in terms of attenuation of the sharp d state immediately below E,, although some decrease in intensity of the leading d band did occur following adsorption and the spectra are approximately normalized in all cases.Additional density-of-states structure does appear following adsorption as is conclusively shown on the (111) face. It is tempting, therefore, to assign the state 1.2 eV below Ef to the antibonding n2p state of the CO, previously unobserved in polycrystalline data because of over- lapping nickel d band emission. This state is populated by back donation from the Ni d bands following formation of the primary o-d bond on adsorption. Doyen and Ertl 24 have recently calculated the energy of the antibonding state as 0.63 eV below E,, and find an occupation number of 0.023 per molecule.Penn has pointed to the possibility of asymmetric shifts in energy of adsorbate levels due to degeneracy with metallic states, so that the agreement in energy for the antibonding level between theory and experiment is quite good. The intensity in the measured density of states peak, however, would seem to indicate a greater occupation number than calculated. Considerable admixing with d states should occur as a result of the above degeneracy, and the structure close to 0.6 eV following adsorption on the (1 11) face may merely reflect broadening of the d states due to such mixing. DEHYDROGENATION OF C2H4 Previous results on polycrystalline Ni films indicated that at high coverages of C2H4 at room temperature, a new set of adsorbate induced UPS levels develops, suggesting some decomposition of the adsorbate, possibly leaving a graphitic over- layer.3 Eastman has also shown using (1 11) orientation single crystals of nickel that, whereas at 100 K ethylene is non-dissociatively adsorbed, at 230 K dehydrogena- tion occurs producing an adsorbate spectrum identical to that from acetylene.At elevated temperatures (600°C in the present case), complete dehydrogenation would therefore be expected to occur, and the comparison between the ethylene-induced adsorbate spectrum in fig. 7 and that from well-oriented pyrolytic graphite supports this view. The predominant peak common to both spectra corresponds to the uppermost Q band at r whilst the other peak in the graphite spectrum corresponds to the flat n bands along the KM direction in the Brillouin zone.18 This latter peak would be partly obscured by the Ni d band and is not resolved as a separate band on the ethylene treated Ni surface.OXIDATION OF CHROMIUM (1 11) The clean spectra for Cr (1 11) again show a sharp state (fwhm 0.3 eV) immediately below E,, followed by a broader d band at both He1 and He11 energies (fig. 8). Once more, the sharpness of the structure is initially suggestive of a surface, rather than aP. J. PAGE A N D P . M. WILLIAMS 89 bulk state, and it is possible that the antiferromagnetic interactions in Cr at these temperatures may open up a gap close to the Fermi favourable for surface state propagation. No change could be detected in the sharp level following adsorp- tion of CO at room temperature, however, and further work will be required to identify its origins.Adsorption of O2 at room temperature proceeded even more readily than for Ni to produce a surface oxide (of, as yet, undetermined stoichio- metry). There is a clear withdrawal of electrons from the uppermost d band leaving a very low density of states at E,, and possibly opening up a band gap of 0.5 eV in the d manifold. A single broad d band (fwhm 1.5 eV at 2.5 below E,) remains following oxidation, which may correspond to chromium in a 3+(d3) charge state. Further work on this system is currently in progress. CONCLUSIONS UPS spectra for ( l l l ) , (110), and (100) faces of clean single crystal nickel show distinct differences which are interpreted in terms of a bulk, rather than a surface density of states.Following the adsorption of CO, a new level 1.2 eV below Ef is observed in addition to the 02p and n2p states of CO. This shallow level is tenta- tively assigned to the antibonding 712p state of CO, populated by back donation from the Ni, although effects of degeneracy and admixture with the Ni d bands may contribute. The cracking of C2H4 over Ni (110) at 600°C produces a graphitic overlayer by comparison with UPS data for pyrolytic graphite. NOTE (added after preparation of manuscript). Recent observations on clean polycrystalline foils of platinum show a sharp d band immediately below E,, and following adsorption of CO, the c and n levels of CO appear 9.2 and 11.6 eV below Ef.This represents not only a shift of both states to higher binding energy compared to CO on Ni, but also a decrease in the a-n separation from 3.0 eV for Ni to 2.4 eV for Pt ; further results on this system will be reported. D. E. Eastman in Electron Spectroscopy, ed. D. A. Shirley (North Holland, 1972), p. 487. D. E. Eastman and J. K. Cashion, Phys. Rev. Letters, 1971, 27, 1520. P. J. Page, D. L. Trimm and P. M. Williams, J.C.S. Faraday I, 1974, 70, 1769. R. Joyner and M. W. Roberts, J.C.S. Faraday I, 1974, 70, 1819. D. E. Eastman and J. E. Demuth, Phys. Rev. Letters, 1974, 32, 1123. E. L. Evans, J. M. Thomas, M. Barber and R. J. M. Griffiths, Surface Sci., 1973, 38, 245. ' S . J. Atkinson, C. R. Brundle and M. W. Roberts, Chem. Phys. Lett., 1974, 24, 175. A. M. Bradshaw, D. Menzel and M. Steinkilberg, Jup. J. Appl. Phys., 1974, to be published. D. W. Turner et al., Molecular Photoelectron Spectroscopy (Wiley, New York, 1970). B. J. Waclawski and E. W. Plummer, Phys. Rev. Letters, 1972, 29, 783. l o B. Feuerbacher and B. Fitton, Phys. Rev. Letters, 1972, 29, 786. l 2 B. Feuerbacher and B. Fitton, 1974, to be published. l3 P. J. Page and P. M. Williams, 1974, to be published. l4 P. J. Page and B. H. Blott, 1974, to be published. l 5 J. W. Linnett, Clzem. SOC. Autumn Meeting (University of East Anglia, 1973). l6 N. V. Smith, Phys. Rev. Letters, 1974, to be published. N. V. Smith, Phys. Rev. Letfers, 1974, to be published. l8 F. R. Shepherd and P. M. Williams, 1974, to be published. l 9 D. E. Eastman, J. Appl. Phys., 1969, 40, 1387. 2o E. I. Zornberg, Phys. Rev. B, 1970, 1, 244. 21 R. Haydock and M. J. Kelly, Surface Sci., 1973, 38, 139. 22 R. Haydock, V. Heine, M. J. Kelly and J. B. Pendry, Phys. Rev. Letters, 1972, 29, 868, 23 P. Fulde, A. Luther and R. E. Watson, Phys. Rev. B., 1973, 8,440. 24 G. Doyen and G. Ertl, Surface Sci., 1974, 43, 197. 2 5 D. R. Penn, Phys. Rev. Letters, 1972, 28, 1041. 26 S, Asano and J, Yamashita, J. Phys. SOC. Japan, 1967, 23, 714,