GENERAL DISCUSSION Papers 5, 6 Dr. C. M. Quinn (Birmingham) said: Has Roberts any comment to make about the substantial invariance of the O(h) binding energy for adsorbed oxygen? The major influence on core levels for large separations of adatoms is likely to be the residual screened Coulomb potential due to neighbouring atoms, with little change in the local charge density. It is likely that the lack of change in the O(1s) level reflects the rather large distances between chemisorption sites and the probably equally large, but unknown distance of the adatoms from the substrate surface? Dr. C. R. Brundle (Bradford) said: More than one O(1s) peak is often observed during the chemisorption of oxygen on metals (even at 8 < 1). One of these always seems to be near 530 eV (tentatively+ 0.5 eV, confidence in an accuracy being limited by the varied calibration techniques used).The other peak, or peaks, if present, come from 0 to 4 eV higher binding energy. Many bulk metal oxides (transition and non- transition, including alkali metals) also have O( 1s) peaks close to 530 eV.2 These data suggest a rather simple interpretation. The oxygen atom is strongly electro- negative and all metals have a component of free or nearly free electrons. Con- sequently in bulk oxides involving only free metal electrons (e.g., CuO, NiO, but not Cu20 or Ni203) the oxygen atom is able to capture easily the electron density it requires to satisfy its electronegativity and become close to a formal " 02- ". Hence the O(1s) binding energies in the different metal oxides have similar values.This in turn suggests that the 530 eV O(1s) peak in chemisorption also represents an oxygen atom that has come close to " 02- ", i.e., the oxygen atom " thinks " it is in an oxide- like environment. This is perhaps evidence that one has in fact at least a reconstructed surface (deeper penetration may be occurring) and that in the cases where higher binding energy O(1s) peaks are also present these represent adsorbed oxygen in a non-reconstructed (and non " 02- " -like) environment. The important point is that this seems to be a rather general phenomenon (even for W and Mo) occurring even during the early stages of adsorption. Dr. R. W. Joyner (Brudford) (in part communicated) : Brundle suggests that re- construction is the cause of the almost identical oxygen Is binding energies observed in oxygen chemisorption on copper, nickel, silver, molybdenum, tungsten and platin~m.~ In drawing this conclusion from XPS data alone he ignores evidence from other techniques which suggests that the initial stage of oxygen interaction with metals is often chemisorption on top, without rearrangement of the metal atoms.For example, the work function of a nickel surface increases on oxygen exposure even at 440 K and then falls slowly, suggesting that chemisorption precedes incorporation, the latter being a slower processS4 LEED intensity against voltage analysis also supports the view that the initial structures formed by oxygen on Ni (100) and W(110) are non- R. W. Joyner and M. W. Roberts, Chem. Phys.Letters, 1974, 28,246. C. R. Brundle, to be published. R. W. Joyner and M. W. Roberts, Cheni. Phys. Letters, 1974, 28, 246. M, W. Roberts and B. R. Wells, Trans. Faraday. Sac., 1966,62, 1608. 90GENERAL DISCUSSION 91 reconstructed. , 2 For silver and platinum the totality of evidence weighs heavily against surface reconstruction, and lattice penetration is certainly not occurring. 3-5 Thus, as we have said previ~usly,~ the demand of the oxygen atom for an electronic environment associated with a binding energy of 530eV appears to be a unifying feature in oxygen chemisorption. It cannot, however, be accepted as demonstrating that reconstruction is the initial stage of oxygen chemisorption. The above discussion has assumed that it is apposite to compare binding energies measured with respect to the Fernii level, B.E.(,+ In photoelectron spectroscopy it has been more common to use : where 4 is the sample work function and B.E.(,, was assumed to be the absolute binding energy.For B.E.,,, to have thermodynamic validity it should be defined as the energy necessary to remove an electron to a state of rest in vacuum at infinity, the same definition as is used in gas phase photoelectron spectroscopy. It has recently been shown that for a conducting solid in electrical contact with earth (the usual situation in electron spectrometers) equation (1) is not valid and that the absolute binding energy is related to B.E.(F) by a fixed although unknown difference.’ Thus in the comparison discussed above it is appropriate to use the binding energy referred to the Fermi level, and not l3.EacF)+4, where q5 varies from system to system.Dr. D. Menzel (Mzmich) said: I agree with Joyner that the question of reference levels in photoelectron spectroscopy of adsorbates is a difficult one, and that it is most probably not appropriate just to add the average work function in order to relate binding energies measured on a surface (relative to EF) to binding energies measured in the gas phase. I do not agree with his conclusion, however, that the ‘‘ absolute ” binding energy of an electron in an atom on a surface is independent of the work function, although this is probably correct for a particle embedded in the solid. In order to relate these energies, comparison should be made of the energies required to take an electron to a point in space in which the chemically induced potential changes (including image force contributions for the metal) have decayed.The long- range dipole or multipole forces, which develop between different faces of the same solid, should not be of importance, on the other hand. It appears to me that the correction to be added to B.E.(F, should then be equal to the difference between EF and the potential in the vacuum above the surface atom. This will be equal to the macroscopic work function only for layers which can be considered as uniform dipole sheets. This may be a good model for a layer of alkali metal atoms or of a close-packed layer of physisorbed or chemisorbed identical particles. It will be a bad one for all layers with “ graininess ”, whether this may be caused by the adsorbates not covering all metal atoms even at maximum coverage (e.g., hydrogen on metals), or by the side-by-side existence of species with different dipole moments (e.g., p-CO and a-CO on W).It would be concluded, then, that B.E.(F)-ValUeS for the same particles on different faces of a metal should differ. This seems to be so for a-CO on W.8 In other cases, J. E. Demuth, D. W. Jepsen and P. M. Marcus, Phys. Rev. Letters, 1973, 31, 540. J. C. Bucholz, G. C. Wand and M. B. Lagally, Surface Sci., in press. H. A. Engelhardt, A. M. Bradshaw and D. Menzel, Ber. Bunsenges. Phys. Chem., 1972,76,500. H. A. Engelhardt, A. M. Bradshaw and D. Menzel, Surface Sci., 1972,40,410. A. E. Morgan and G. A. Somorjai, Surfuce Sci., 1968,12,405 ; J.Chem. Phys., 1969,51,3309. R. W. Joyner and M. W. Roberts, Chem. Phys. Letiers, 1974, 28, 246. ’ A. F. Carley, R. W. Joyner and M. W. Roberts, Chem. Phys. Letters, 1974, 27, 580. * A. M. Bradshaw, this Discussion p. 132.92 GENERAL DISCUSSION a compensation between the values of 4 (clean) and A$ may take place and make the B.E.,,,-values virtually identical. The fact that there is often little or no change of B.E.(F, for changing coverage and therefore work function can be understood by the notion that the local potential above the first adsorbed atom is (almost) the same as above an adatom in the full layer, if the adatom dipole moment is independent of coverage.' It should be realized, however, that core level binding energies often do show shifts with coverage (e.g., O(1s) of oxygen on W(l It is concluded that the proper value which would have to be added to B.E.(,) to compare to gas phase binding energy or even to compare values for different surfaces, is quite unclear at present.Dr. R. W. Joyner (Bradford) (communicated) : I agree with Menzel's concluding paragraph. It is still not clear what should be added to B.E.(,) for comparison with gas-phase binding energies, although studies of condensed materials rather that chemisorbed layers suggests that the correct value might be about 6 eV.3 Menzel concludes that the reference level should be the potential energy of an electron at rest in vacuum above the surface atom. This choice appears to me to violate the essential criterion that the reference level should be constant and indepen- dent of the system under study. Only if this condition is satisfied can valid com- parisons of binding energy be made and thermodynamic integrity be maintained.I therefore reiterate that the only appropriate absolute reference level is the potential energy of an electron at rest in vacuum at infinity with all that this implies concerning the binding energy referred to the Fermi Also, if we accept, with Menzel, that " it is probably not appropriate just to add the average work function ", the Fermi level becomes the only accessible and therefore the only practicable reference level. Dr. S. Evans and Dr. M. J. Tricker (Aberystwyth) (in part communicated) : We consider that it is most unwise to associate a constancy of binding energy with con- stancy of charge, formal or otherwise.It is already well established for bulk systems that to account adequately for an observed chemical shift it is necessary to conisder both the atomic charge and the so-called " Madelung '' potential arising from all the other charges surrounding the atom in questi~n.~ The contribution of this second term to the observed chemical shift may rival or even exceed that of the charge (valence) term : this is well demonstrated in Pb304, for example, where the two oxi- dation states of lead are not distinguished by XPS because the roughly doubled partial charge between Pb" and Pb'" is entirely offset by the difference in '' Madelung " terms between the Pb" snd PbrV sites.6 The fact that PbO, has a lower Pb 4f binding energy than PbO can similarly be explained by taking into account this " Madelung " potential.The approximate constancy (k 1 eV) of 0 (1s) binding energy in many metal-oxygen systems may well have an analogous explanation, especially for bulk oxides and where reorganisation of the surface occurs following chemisorption. This constancy of anion binding energies is also found in other systems; for in- stance, the three crystallographically inequivalent F atoms of LaF3 cannot be resolved G. E. Becker and H. D. Hagstrum, J. Vac. Sci. Tech., 1973, 10. 31. A. M. Bradshaw and D. Menzel, Vakuum-Technik, in press. R. W. Joyner and M. W. Roberts, unpublished results. this Discussion, and A. F. Carley, R. W. Joyner and M. W. Roberts, Chem. Phys. Letters, 1974, K. Siegbahn et al., ESCA (Almqvist and Wiksell, Uppsala, 1967).J. M. Thomas and M. J. Tricker, J.C.S. Faradar 11, 1975, 71, 317; D. E. Parry, J.C.S. Faraday IZ, 1975, 71, 344. 27, 580.GENERAL DISCUSSION 93 by XPS.l As we outlined above, the experiment tells us nothing in a direct way about the charge on 0 or F because of the " Madelung " term. What it does indicate, at this stage, is that the total potentials (i.e., Madelung+valence) at the anion sites are similar, for a given anion, in many environments. The approximate constancy of anion binding energy may thus originate in a tendency for the total potential at the different sites to become the same,2 rather than in a tendency of (for example) oxygen atoms in oxide systems to come " close to a formal 02- ". Furthermore, adsorbate binding energy data referenced to the Fermi level of different substrates are not, strictly, directly comparable, either with each other or with gas-phase data.They must first be corrected so that the zeros of the various energy scales coincide, since the energy required to excite an electron from the Fermi level into the lowest unbound state in the vacuum (the reference level of gas-phase UPS data) differs in each case. The best procedure available for this correction is probably to add the work function after adsorption to the observed Fermi level- referenced binding energy, although even this is only really valid for a complete (infinite) uniform layer of a monatomic adsorbate on a uniform, single crystal, sub- strate. The observed constancy of 0 (1s) binding energy thus also depends on the fortuitous constancy of this correction term, or, perhaps, on variations in it offsetting differences in partial charge and/or Madelung potential.Dr. R. W. Joyner (Bradford) (communicated): There is as yet no evidence for the importance of Madelung potential terms in considering binding energies of surface species. Our observations refer to oxygen 1s binding energies of the chemisorbed species, and that for chemisorbed carbon contamination on metal films, the carbon 1s level is usually found close to B.E.,,, = 285.0 eV. It is therefore plausible that atomic charge and " Madelung " terms can be equal on different metals, there being no obvious reason to invoke a compensation of these two terms. Evans and Tricker neglect the requirement of an absolute energy zero.The addition of a work function as they suggest, leads to unacceptable inconsistencies ; for example, an electron excited inside a crystal would therefore have a different binding energy depending on whether it leaves through a (100) or (1 10) layer, which have different work functions. Dr. C. M. Quinn (Birmingham) said: Brundle appears to be suggesting that an 02- surface species is a better defined " surface molecule " than a less charged adatom. Such suggestions tend to prolong the divisions between localized bonding concepts which encompass such adsorbate fragments as 02-, and band models in which adatoms are viewed more as extensions of the bulk solid. The surface molecule concept, currently being developed by several groups, 4-7 seems to provide a framework for re- conciling these two extremes.It appears * that the definition of a clearly localized surface molecule, e,g., a fragment like 02- or a discrete 0"-, depends largely on a difference term involving the first and second energy moments of the surface density of states at the adsorption site divided by the second energy moment of the bulk density states. These terms are determined by the substrate geometry and current R. G. Hayes and N. Edelstein, in Electron Spectroscopy, ed. D. A. Shirley (North Holland, Amsterdam, 1972), p. 771. M. J. Tricker, Inorg. Chem., 1974, 13, 742. R. W. Joyner and M. W. Roberts, Chem. Phys. Letters, 1974, 28, 246. T. B. Grimley, J. Vac. Sci. Tech., 1971, 7, 31. R. Haydock, V. Heine and M.J. Kelly, J. Phys. C, 1972, 5, 2845. J. R. Schrieffer, J. Vac. Sci. Tech., 1972, 9, 561. J. W. Gadzuk, Surface Sci., 1974, 43, 44. * M. J. Kelly, J. Pliys. C, 1974, 7, LT57.94 GENERAL DISCUSSION results, although somewhat preliminary, do not support such high localization as would be required for the formation of an 02- species at any surface. Dr. C. R. Brundle (Bradford) (communicated) : It seems that the use of ' formal "02-" ' may be misleading to those not well acquainted with inorganic chemistry conventions. Just as no inorganic chemist would assume that the use of V5+ to signify V205 implies five positive charges on the vanadium atom, " 02- " does not imply two complete negative charges on the oxygen. One could write the species as 0-", but this would be more uncommon.What is implied in my comment is that oxides having similar O(1s) binding energies have a similar (and optimum) net potential at the oxygen atom. For a comparison of chemisorbed 0 and the oxide of the same metal only, the conclusion that if the (Is) binding energies are similar in the two cases the chemisorbed oxygen is likely to be in an oxide-like environment is quite indepen- dent of the value for the absolute charge on " 02- " that might be assigned. Dr. C. M. Quinn (Birmingham) said: The binding energy N 530 eV observed by Roberts et al. does not appear to depend to any great extent on the nature of the substrate or the coverage. This behaviour contrasts with results for gas-phase species containing oxygen and also with the changes in binding energy observed for O(1s) and C(1s) in adsorbed carbon monoxide species.Core levels are influenced by the residual screened coulombic potential due to neighbours, here, both substrate and adsorbate, together with any appropriate changes in charge density. Richardson and I have recently initiated a programme of MSXa calculations on model clusters chosen to be representative of various conditions of adsorbed species. We find in preliminary non-SCF calculations"' that the superimposed coulombic potential at an oxygen adatom does not vary to any great low or high coverage on a lithium used the models 0 Li Li / \ (LiSO) Li extent for oxygen in an oxygen molecule, nor at (100) surface. For the adsorbed states we have 0 0 0 \ / 0 ' I \o with an 0-Li distance of about 3.0A.The non-SCF O(ls) energies are -515.32 eV in gas phase 02, -514.86 eV in Li,O and -514.69 eV in Li05 (centraI 0) and - 514.4 1 eV (peripheral 0). The lack of change in the O( 1s) values after superim- position of the potentials in the cluster reflects the rapid decay of the atomic potentials with radial distance. It seems that similar changes in charge density may be occurring in all the cases of oxygen adsorption where oxygen adatoms are the only surface species. A similar situation can be expected for the energy levels resulting from super- imposition of potentials due to adsorbed carbon monoxide and possible fragments. However, charge density changes for different configurations, viz., 0 C C co 0 may now substantially alter the O(ls) and C(1s) binding energies.* C. M. Quinn and N V. Richardson, to be published.GENERAL DISCUSSION 95 Prof. M. W. Roberts (Bradford) said: It may well be incorrect, at least with oxygen, to think in terms of discrete chemisorbed adatoms with a localized electron density, but instead to take an approach which is more akin to the collective electron outlook involving considerable delocalization. In this respect Quinn's results are interesting. On the other hand with CO (when in the molecularly adsorbed state), the electronic distribution as reflected by the O( 1s) binding energy parallels the heat of chemisorption' (fig. l), and we are dealing with a discrete surface molecule. If we accept that in both cases the O(1s) binding energy reflects the electron density (or electrostatic potential) in the neighbourhood of the oxygen then in oxygen chemisorption it is substantially invariant of the metal and not particularly dependent on the oxygen coverage.With CO it is sensitive to the metal-CO bond strength (fig. 1). We suggest also that there is a critical heat (- 300 kJ mol-l) above which CO dissociates at room temperature when the experimentally observed O(ls) value is about 530 eV. The criterion we have used for dissociation is the presence of atomic-like " orbitals " and the absence of " orbitals " akin to those in molecular CO. 537'0t 535.0 \ \ I I 1 I 100 2 00 300 - A Hab/kJ mol-' FIG. l.-Q, Platinum; a, nickel; 0, al state on tungsten; A, at state on tungsten; 0, copper; V, y state(s) on molybdenum. Prof. J. M. Thomas and Dr.D. E. Parry (Aberystwyth) said: Fig. 7 of the paper by Page and Williams contains some information which is of considerable general interest in regard to the reliability of photoelectron spectroscopy in monitoring valence bands of clean solids, particularly of systems such as graphite. Whilst we do not doubt that curve (b) in the figure is attributable to the graphitic solid formed by dehydrogenation, we would like to know more about the precise shape, and especially the total width, as determined by He I1 radiation, that Williams finds for his graphite valence band. In the first reported 3*4 XPS studies of the valence band of graphite, the total (What happens at lower energies, e.g., at - 30 eV in fig. 7 ?). R. W. Joyner. and M. W. Roberts, Cheni. Phys. Letters, 1974, 29, 447. A.F. Carley, R. W. Joyner and M. W. Roberts, Chetn. P ~ J L S . Letters, 1974, 27, 580. K. Hamrin, S. Johansson, U. Gelius, C. Nordling and K. Siegbahn, Physica Scriptlr, 1972, 2, 277. J. M. Thomas, E. L. Evans, M. Barber and P. Swift, Trcirrs. Furdaj, SOC., 1971, 67, 1875.96 GENERAL DISCUSSION width appeared to be 31 f2eV, disturbingly in excess of the figure of 19.5 eV obtained by the best theoretical computation of Painter and Ellis,l whose ab initio LCAO procedure was judged by Coulson to be within an electron volt or so of the correct value for graphite. have carefully reinvestigated the valence band structure of graphite, using high resolution XPS, and they conclude that the experi- mental width is not less than 24eV. (Examination of their fig.7, in which they '' eliminate artificial tailing " leads to the view that their XPS determined width could well be larger by another 4 eV). Kieser has recently quoted a band width of ca. 21 eV, but an examination of his data also shows uncertainty as regards the precise width. The implication, clearly, is that photoelectron spectroscopy can, under certain circumstances which are not yet fully understood, yield unreliable information at values of the energy far below the Fermi level. With UPS in place of XPS the difficulties could be more severe for another reason : final-state effects are not likely to be eliminated when such relatively low energy (40.8 or 48.4 eV) of stimulating radiation is employed. McFeely et al. Dr. P . M . Williams (Imperial CoZZege) said: The question raised by Thomas concerning the band-widths determined for solids by UPS and XPS is an impor- tant one.Both scattering and effects due to the coupling between photoelectron and hole states can broaden the experimentally observed structure. 5-7 With reference to fig. 7 of our paper, we reproduce there only the uppermost valence bands of graphite, as indicated in the text, to show the agreement between the ethylene-cracking product and pyrolytic graphite. The full He I1 (40.8 eV photon) spectrum for the latter shows a further prominent state, FWHM 3 eV centred at 18 eV below Ef (together with a weaker feature at 13.5 eV), so that the total bandwidth is close to 21 eV, in agreement with the data of Kieser * and the calculations of Painter and E1lis.l The XPS value for the bandwidth of 23 eV determined by McFeely et aL3 may, as these authors point out, be a slight overestimate due to the inaccuracy in fixing Ef and to scattering at the high binding energy side of the band ; similar considerations should apply to the XPS value of 31 eV measured by Thomas. The most important differ- ence between the high energy UPS (i.e. hw 2 30 eV) and XPS data appears to lie not in the appearance of final state effects in UPS (although these are undoubtedly of importance, particularly at He I energies) but in the different cross sections for excitation from the " s " and " p '' like parts of the band. In other respects, we believe our data (to be published elsewhere) to be consistent with the findings of McFeely et al. G. S. Painter and D. E. Ellis, Phys. Rev. B, 1970, 1, 4747. C. A. Coulson, 1971, personal communication to J. M. Thomas. F. R. McFeely, S. P. Kowalzyk, L. Ley, R. G. Cavell, R. A. Pollack and D. A. Shirley, Phys. Rev. B., 1974, 9, 5268. J. Kieser, Physica Fennica, 1974, 9, suppl. Sly 173. B. 1. Lundquist, Phys. Kondens. Mater., 1969, 9, 236. S. Doniach and M. Sunjic, J. Phys. C., 1970, 285. J. Kieser, Phys. Fennica, 1974, 9, suppl. S1, 173. ' See for example, G. K. Wertheim and S. Hufner, Phys. Rev. Letters, 1972, 28, 1028.