Faraday Discuss. Chem. SOC., 1990, 89, 259-273 Core-level Shift Spectroscopy on Tungsten Surfaces Overlayer and Underlayer Adsorption G . P. Derby and D. A. King Department of Chemistry, University of Cumbridge, Lensjield Road, Cumbridge CB2 1 E W Core-level binding energy shifts in the 4f,,, level of W previously reported from a range of clean single-crystal surfaces are consistent with a simple linear relationship to an effective coordination number. Here we report data for the adsorption of H2, O2 and CO on W{110}, where conventional overlayers are formed, and for N2 on W{ 1 lo}, where an underlayer is formed. The observed adsorbate-induced shifts are consistent with a model which accounts for changes in the coordination of surface W atoms due to adsorp- tion, invariably leading to an increase in binding energy, and also changes due to charge transfer between adsorbate and substrate.It is pointed out that these two effects are additive for electron-withdrawing adsorbates, but tend to cancel each other with electron-donating adsorbates. Recent observa- tions for alkali-metal atom adsorption are consistent with this model, and do not imply non-ionic adsorption. At clean and adsorbate-covered metal surfaces the electronic core-level bonding energies of surface atoms are shifted with respect to those of bulk atoms.'-3 At a microscopic level, these surface core-level shifts (SCLS) originate from both initial-state effects, including purely environmental contributions and differences in electronic configurations between bulk and surface atoms, and final-state effects arising from modifications of the screening of the core hole following photoemission.For a wide range of clean W crystal planes the dependence of the magnitude of the core-level shift was found to be almost monotonically dependent on the effective coordination of atoms in the surface region, provided that the third coordination shell is i n ~ l u d e d . ~ With adsorbates, large shifts from clean surface binding energies are observed for electronegative species, consistently to higher binding energy, while for electropositive species such as the alkali metals relatively small shifts, to higher or lower binding energies, have been observed.5-8 The latter observation has recently led to the conclusion that the alkali metals are covalently bonded at all coverages on W{llO}, in paradoxical contrast with 0 on the same surface.' In the present paper we examine in some detail the core-level shift behaviour on W{ 110) with electronegative adsorbates, including H, 0, P-CO and virgin CO, in conventional overlayer sites and one case (P-N) where an underlayer is formed.The results are examined in terms of available microscopic models. Experimental The experiments were performed on the SRS at Daresbury in a VG ADES 400 ultra-high vacuum chamber. The core levels studied are the 4f,,,7, which have proved to be very surface sensitive at kinetic energies ranging from 20 to 70eV. A torroidal grating monochromator provided radiation at photon energies between 60 and 95 eV, well suited to the present work.The base pressure was better than 1 x lO-'"Torr,l- and a typical working pressure was less than 2 x lO-"'Torr. An angle-resolving analyser (4" half-width) was used to collect the electrons emitted normal to the surface. The light was incident i 1 Tom= 101 325/760 Pa. 259260 Core-level Shqt Spectroscopy on Tungsten Surfaces r 1 0.01 ' 1 32.4 32.0 31 .O 30.4 binding energy/eV Fig. 1. W 4f,,, core level spectrum taken from clean W(110) at 80 K, at normal emission. The photon energy is 70 eV, and the incidence angle is 70 O to the normal. The crosses represent raw data points. The data have been fitted, and the full-line peaks are the Doniach-Sunjic lines without any instrument or phonon broadening included. The full line through the data points is the sum of the individual components inclusive of all broadening factors.( a ) Bulk, ( b ) clean surface. at 70" to the surface normal, to maximize the relative surface sensitivity.6 The total resolution of the system, monochromator and analyser, was estimated to be 0.16 eV from the Fermi edge: this value is in good agreement with the calculated resolution for the system. The crystal was cut to within 0.5 O of the required plane, and cleaned by repeated heating to 1800 K in Torr oxygen and flashing in vacuum to 2400 K. The surface was considered to be clean when no further increase in the amplitude of the surface peaks, relative to the bulk, could be seen in the core-level spectra. The spectra from the clean surface were taken as multiple scans (170 s per scan), the crystal being flashed to 2500 K before the start of each scan to reduce surface contamination to a minimum.Temperature measurement was effected by use of a W-Re (3%)/W-Re (25%) thermo- couple. Results and Discussion The Clean Surface: Reference Spectrum On the closely packed { 110) surface of tungsten, only two, well separated, core levels can be seen. Data are shown in fig. 1. To extract the core-level binding energies from these data a non-linear least-squares fitting technique has been used, in which each displaced core level is represented by a single peak. The lineshape of the peak is influenced by several processes: the underlying shape is Lorentzian, arising from decay of the core hole, but excitations of electron-hole pairs at the Fermi level contribute to an asymmetry, resulting in a Doniach-Sunjic (DS) lineshape.Bulk and surface phonon excitations caused by the photoemitted electron additionally produce a Gaussian broad- ening, and inhomogeneities on the surface may also contribute to Gaussian broadening.50( 400 - I v) v) - G 300 : S eD .- w. 200 100 G. P. Derby and D. A. King * : , : : . : : : : : - : : ; :, 26 1 I . , . i . I . 33.0. 32.6 32.2 31.8 31.4 31.0 30.6 binding energy/eV Fig. 2. W 4f7,2 core-level spectrum taken under the same conditions as for fig. 1 from a fully oxidised W{ 1 lo} surface, showing an isolated bulk peak. ( a ) Oxide, ( b ) bulk. Finally, the finite resolution of the experimental system used to measure the spectrum must be included and this is also taken to be Gaussian. From the earlier work of Wertheim et al.,9 it was concluded that the bulk and surface peaks on W{ 110) have the same Lorentzian widths and asymmetry parameters (Lorent- zian FWHM, 50 meV; asymmetry parameter, 0.06).These values were assumed in the analysis depicted in fig. 1, and the binding energy of the bulk peak was fixed to the value determined from the peak position observed in the heavily oxidised surface (fig. 2), where the bulk peak is completely separzited from the surface peaks, with all selvedge peaks shifted to higher binding energy by +0.97 eV relative to the clean surface peak. On the clean surface, the surface-to-bulk peak separation is 300 meV, in agreement with the data of Duc et a1.l’ and Purcell et a1.,l1 and rather smaller than that reported recently by Riffe et al.(321 meV).’’ As reported below, we do observe that the gap between the surface and bulk peaks is narrowed with hydrogen adsorption, but great care was taken to avoid hydrogen contamination on the clean surface spectrum. The bulk peak in fig. l ( b ) is well fitted with a DS Lorentzian FWHM of 50 meV, asymmetry parameter 0.05, and Gaussian broadening (phonon + instrument) of 190 meV. In their recent paper, Riffe et al.” report a Lorentzian FWHM of 60 meV and asymmetry parameter of 0.035. Significantly, they also report difleerent values for the surface peak: 84meV and 0.063. While these differences are important to the physics of the emission process, they do not significantly affect the analyses from singular and stepped surfaces previously published, or those presented here.The Lorentzian broadening is swamped by the Gaussian broadening at the instrumental resolution deployed. Overlayer Adsorption We compare core-level shift data obtained on the W{ 110) surface after varying exposures to Hz, 02, and CO. The results are summarised in fig. 3.262 I I I Core-level Shift Spectroscopy on Tungsten Surfaces I I I I I I I P D I I I I I P s I I B I I I I I , I B I VI. ( f ) N, Ni B ] N: S binding energy/eV Fig. 3. Summary of W 4f,,, core-level shift data, showing the surface (S) and bulk ( B ) components of the clean-surface spectrum, and the positions of all lines fitted to the spectra with H, 0, p-CO (dissociated), virgin CO (molecular) and P-N adsorbed on W{ 1 lo}, respectively. Not all these peaks are observed at any one coverage (see text).( a ) Clean W( 1 lo), (6) H/ W( 1 lo), ( c ) O/ W( 1 lo), ( d ) P-CO/ W( 1 lo), ( e ) V-CO/ W( 1 lo), cf) P-N/ W( 110). Hydrogen on W{ 110) As hydrogen is adsorbed on W(110) to half-monolayer coverage at 300 K the surface core-level peak at a binding energy of 31.08 eV is rapidly attenuated and replaced by a component estimated to be ca. 110meV below the binding energy of the bulk peak, which is at 3 1.38 eV. An increase in the bulk peak intensity indicates a second component at the same binding energy as bulk W atoms. At low coverages a further peak is estimated at ca. 200 meV below the bulk peak binding energy. The results may be interpreted in terms of the structural model of Holmes and King," shown in fig. 4. It is suggested that the low-coverage peak shifted by -200meV from the bulk peak is due to a surface W atom bonding to a single H adatom; while the two peaks observed at higher coverage at -1 10 and 0 meV from the bulk peak are assigned to surface W atoms bonded to two and to three H adatoms, respectively.The shift ofG. P. Derby and D. A. King 263 Fig. 4. Hard-sphere representation of the Holmes and KingI3 model for the (2 x 1) overlayer structure formed by H on W{110}, showing two distinct adsorption sites, threefold hollow and short bridge. (Open circles represent top-layer W atoms.) the surface layer W 4f binding energy from that of the clean surface is therefore roughly proportional to the coordination with H adatoms, -100 meV per H atom. that the work function decreases almost linearly with hydrogen coverage on W{110}, the maximum observed change at saturation being ca.-500 meV. There is general Oxygen on W( 110) The influence of adsorbed oxygen is demonstrated in fig. 5 . The clean surface peak in the W 4f,,* spectrum is rapidly attenuated with exposure to 02, being eliminated after an exposure of one Langmuir. At very low exposures a peak appears to build in between the clean-surface and bulk peaks, shifted by ca. -80 meV with respect to the bulk peak. However, when the well known half-monolayer p(2 x 1) structure is formed, this peak is replaced by higher binding energy peaks, with a dominant peak at +370meV with respect to the bulk peak. This is attributed to surface W atoms bound to two oxygen adatoms. The deconvolution is consistent with two further peaks, shifted by +200 and +610meV, respectively, from the bulk peak, which are attributed to surface W atoms singly and triply bound to 0 adatoms.Once again, the shift appears to be roughly proportional to the coordination with adatoms, -200 meV per 0 adatom. When oxygen is adsorbed on W{110}, there is a small work function increase, of +0.20 eV at 8 = 0.5. At low coverages the work function actually decreases, which may be associated with adsorption at defects, reconstruction, or a pre-island phase. Carbon Monoxide on W( 1 10) The adsorption of CO on W{110} produces a diverse range of states, depending on substrate temperature and exposure.”-”’ At 100 K adsorption produces only a single molecular adsorption state, the virgin state, bound through the C end of the molecule.Heating the surface causes some desorption and some conversion into a dissociated state, the p state. A molecular state is coadsorbed with the p state at temperatures264 Core-level Shiji Spectroscopy on Tungsten Surfaces h c) v1 .- c 2 0.01 . J 32.2 32.0 31.8 31.6 31.4 31.2 31.0 30.8 30.6 0.oL . 32.2 32.0 31.8 31.6 31.4 31.2 31.0 30.8 30.6 binding energy/eV Fig. 5. Fitted core-level spectra from oxygen on W(l10) at room temperature for exposures of ( a ) 0.05 L and ( b ) 0.75 L. The narrow spectrum below each curve is the difference spectrum between the data points and the best fit line. below 500 K. The molecular virgin and dissociated p states are readily distinguished using XPS; for example, the C 1 s peak exhibits a binding energy at 285.5 eV for virgin CO which is shifted to 283.0eV in the dissociated state.” Core-level shift spectra for W4f,,, obtained on exposure to CO at a surface tem- perature of 100 K are shown in fig.6. The initial doses were not started until the crystal temperature was below 120 K after having been flashed more than once to greater than 2300 K to remove any contaminating residue from the previous dose. After each ‘cold’ surface spectrum was collected, representing the virgin CO overlayer alone, the crystal was heated to 600 K and re-scanned at this temperature to avoid any complications dueG. P. Derby and D. A. King 265 -__ - .-- 1.41 1.3j 1.21 1 la'! .o 0.9.- 0.8 .- 0.7 -- 0.6 -- : o.olpa+--+--L . *-- * --.--+- --*- % 33.0 33.2 33.4 33.6 33.8 34.0 34.2 34.4 34.6 34.8 o.o-------+--- ~ 33.0 33.2 33.4 33.6 ,333 34.0 34.2 34.4 34.6 34.8 kinetic energy/eV Fig.6. Fitted core-level spectra obtained after ( a ) adsorbing CO at 100 K, where the adlayer is in a molecular state, and ( b ) after heating to 600 K, where the adlayer is dissociated. ( a ) 100 monolayers, 100 K; ( b ) 600 K.266 Core-level Shift Spectroscopy on Tungsten Surfaces h Y m .- E 3 E E .- - a E 0.0 0.5 1.0 0.0 0.5 surface coverage/monolayer 1 .o Fig. 7. Variation of intensities of each of the adsorbate-induced components in the core-level spectra for CO on W(110) with surface coverage: ( a ) the /3 components from the 600 K spectra; (i) P s , (ii) Po, (iii) P I ; ( b ) the virgin components from the 100 K spectra; ( i ) V,, (ii) V,, (iii) PT.to a-CO readsorption. The latter spectra therefore correspond to the pure p-CO state. Fig. 6 includes examples of the fitted spectra for both the initial adsorbed layer and for the associated heat-treated layer. The effects of phonon broadening were included in the fitting procedure for the 600 K data by inclusion of extra Gaussian broadening" within the instrument broadening function parameters. From a comparison between fig. 6 and 5, there are strong similarities between the p(2 x 1) structure formed by the p-CO state and the p(2x 1) overlayer structure formed as the result of pure oxygen adsorption. We can therefore be confident about drawing comparisons between the respective core-level spectra. In effect, it would appear that C and 0 adatoms from p-CO produce a similar influence to an adlayer of 0 adatoms alone.Three peaks are clearly derived from the p-CO state, labelled P , , ps and pD. The variation of each component with surface coverage is illustrated in fig. 7 ( a ) . The PI peak is assigned to surface W atoms bound to a single adatom; the ps and pD peaks are assigned to a (2 x 1) structure with C and 0 atoms occupying three-fold hollow sites. This analysis follows the analysis of the O/W{110} system very closely. The pT. peak, lying on the high-binding-energy edge of the spectra, has low intensity but appears to be a real feature; by comparison with the O/ W{ 1 lo} analysis, this is assigned to surface W atoms triply coordinated to adatoms. From data based on a range of different techniques,'5-20 the adlayer formed at 100 K is composed of virgin CO alone.The two main peaks associated with this layer, V, and V, (fig. 6), develop in intensity with coverage as shown in fig. 7 ( b ) . The V, peak develops rapidly at very low coverages, and then decays as VB grows to become the dominant feature. Additionally, the binding energy attributed to V, increases with coverage, from +lo8 meV shifted from the bulk peak at 8 =0.1 monolayers to +280 meV at 0.7 monolayers. As it does so, it also increases in Lorentzian half-width from 210 to 328 meV. At all coverages this component is significantly broader than that observed for other components, e.g. 50 meV for the bulk peak. At saturation coverage in the virgin CO state on W{ 110) the work function is increased by 0.6 eV, showing net charge transfer to the adlayer. We tentatively assign VL and V, to two different molecularly adsorbed CO species: a linear species V,, with a binding- energy shift of -200 meV with respect to the bulk peak; and a bridged species, VB, with a shift of +lo0 to +280 meV, depending on coverage.This is consistent, for example, with the overall increase in work function of 0.6 eV observed at saturation coverage in the virgin state, when the VB peak is dominant: the bridged species is dominated byG. P. Derby and D. A. King 267 d.rr* backbonding, and hence by charge transfer to the adsorbate. On the other hand the linear species is dominated by 0- donation to the metal. The shift in the binding energy for VB can be attributed to crowding in the adlayer, with W atoms being coordinated to a single bridged species at lower coverages; to one bridge and one linear or two bridge species at higher coverages; and even to three bridged species at full coverage. The enhanced width of the peak attributed to the bridge-bonded species is attributed to a combination of two sources.First, the existence of a range of coordination possibilities for surface W atoms to linear and bridged species will lead to inhomogeneous broadening, and the increased half-width with increasing coverage appears to be con- sistent with this. Secondly, in the bridge-bonded position the CO 27r* level is partially occupied, straddling the Fermi level. Its occupancy will vary with the vibrations of the surface molecule, leading to a variable charge on the adsorbate, and hence an increased spread of core-level binding energies for the substrate atoms involved in the bonding.Underlayer Adsorption Based on ion-scattering spectra, Auger depth-profiling and LEED data Somerton and King concluded that dissociatively adsorbed nitrogen, P-N, on W( 1 lo} forms an under- layer structure, sandwiched between the first and second W layers at the surface.‘2 The adsorption rate of nitrogen on tungsten exhibits a large crystallographic ani~otropy,~’ with an initial sticking probability that varies from 0.73 on the (310) surface to 5 x lo-’ on W(110) at 300 K. Formation of the p-state is believed to occur via dissociation of the molecule at adjacent pairs of four-fold (100) sites, which on the close-packed { 1 lo} surface are only to be found at defects, such as steps.” The exothermicity of the chemisorption process produces ‘hot’ adatoms which are then free to migrate and populate the entire surface.At 300 K this process is inefficient but may be enhanced by repeated heating to 600 K and readsorption. The heating assists migration away from the region adjacent to the defect sites, allowing further dissociative adsorption to occur on cooling. This ‘cycling’ is carried out in an ambient pressure of 2 x lop6 mbar of N2 until a saturation coverage of 8=0.25 monolayer, as indicated by a (2x2) LEED pattern,23 is reached. We found the best quality LEED pattern to be achieved after a total exposure time of 3-5 min, involving 6-7 heating-cooling cycles. Core-level spectra taken from the (2 x 2)-N structure show a rather unnatural flat- topped distribution of photoelectron binding energies, as shown in fig.8. In this figure a direct comparison is made between the spectrum from the clean surface and that from W{ 1 10) (2 x 2)-N, obtained under identical conditions, and a further unusual feature is immediately obvious. In addition to almost total attenuation of the clean-surface surface peak S, the bulk peak B is attenuated to roughly half the clean-surface intensity. This contrasts with all of the overlayer cases discussed above, where virtually no attenuation of the bulk peak was observed. Qualitatively, the underlayer model provides a relatively simple means of understanding the spectrum. From our previous study of N on W{ an overlayer system, it was concluded that N induces shifts from the bulk position in surface W atoms of -100meV (coordinated to one N adatom), +200meV (two N adatoms) and +400 meV (three N adatoms).In the present system, each tungsten atom can be coordinated to a single surface W atom only, since the maximum coverage is only 0.25 monolayer, and yet there is clearly a component with a shift of ca. +500 meV from the bulk peak. This is readily understood in terms of the rumpled underlayer model for W{ 110) (2 x 2)-N, depicted in fig. 9. The coordination of W atom C’ to the rest of the surface is significantly reduced, and charge transfer from this atom to the N atom beneath it could be enhanced due to a reduction of screening by valence electrons of the substrate. Furthermore this N adatom is also coordinated to the underlayer atoms D, E and K (fig.9) which may be expected to produce a small core-level shift to higher268 Core-level Shift Spectroscopy on Tungsten Surfaces 33.0 34.0 35.0 kinetic energy/eV Fig. 8. A comparison of raw core-level spectra from ( a ) the clean W{ 1 lo} surface, and ( b ) W{ 1 lo} with a ( 2 x 2 ) nitrogen underlayer structure. The spectra were collected and recorded under identical conditions. ( i ) Bulk, (ii) surface. binding energies. These atoms would, in the clean surface, contribute to the intensity of the bulk peak, and we therefore also have a ready explanation for the observed attenuation of the bulk peak (fig. 8). Attempts to quantify this picture are frustrated by our inability to resolve the component peaks in the flat-topped spectrum.The spectrum can only be consistent with a large number of component peaks, each from W atoms differently coordinated within the selvedge. The simplest fitting procedure was applied to the data, based on the assumption that the lineshapes of all component peaks were identical to that of the bulk, and then using the minimum number of components to fit the data. The result is shown in fig. 10. The bulk peak B, the clean surface peak S and the peak Ns are readily identified, the latter being attributed to the rolled surface W atom C’ (fig. 9). This peak is shifted by 650 meV to higher binding energy compared with the clean-surface surface peak. Within the (2 x 2) structure, a proportion of the surface atoms are unaffected by the presence of the underlayer N adatoms, which is consistent with a residue from the clean-surface peak S.The peaks labelled N$ and Ng could be assigned to, respectively, underlayer W atoms at J (fig. 9) interacting with the underlayer N adatom, and shifted by 170 meV from the bulk position, and atoms of type A and B in the surface, shifted by 130 meV from the clean-surface peak S. Two minor features remain to be assigned, one at each end of the spectrum. The feature with a binding energy of 30.9 eV may be due to the generation of low-coordination defect atoms at the surface, produced in the process of generating the underlayer structure. However, it is also possible that these two features are artefacts of our analysis, and would be removed by allowing peaks Nk and Ng to broaden.G. P.Derby and D. A. King 269 Fig. 9. Hard-sphere representation of the surface rumpling generated at the W{ 110) surface by an underlayer N atom (filled circle) in the (2x2) structure. ( a ) Clean surface, ( b ) nitrogen underlayer. When a surface oxide is formed on W{ 1 lo}, the core-level spectrum, described above [fig. 2( b ) ] , shows a clear distinction between bulk and oxide W atoms. The centre-of- gravity of the oxide peak is shifted by 150 meV, to higher binding energy, from the bulk peak. This compares, for example, with a shift of 610 meV for a surface W atom bonded to three chemisorbed 0 atoms, as described above. The larger shift is a measure of the ionicity of the W-0 bond in the oxide layer. Note that the shift of the surface W atom peak associated with nearest-neighbour bonding to a single underlayer N atom is 620 meV with respect to the binding energy of the clean surface peak, considerably more than singly coordinated W atoms in the W{100}-N overlayer, of 300 meV.The implication is that the charge on the relatively unscreened surface W atom C’ in the presence of N is considerably larger than that on an embedded W atom bonded to a single N atom in the overlayer system N/ W{ 100). Discussion Adsorption of hydrogen, oxygen, nitrogen and carbon monoxide on metals invariably produces a shift in the core-level binding energy of surface metal atoms towards higher binding energies, as also demonstrated in the present work for these adsorbates on W(110). These results have led almost universally to an explanation based on electron transfer of metal d electrons to the adsorbate.However, Apai et aL5 studied core-level shifts in Pt single-crystal surfaces produced by the electron-donating adsorbates NH3 and K, and still reported shifts towards higher binding energies (as with electron- withdrawing adsorbates), apparently contradicting this model. They concluded that K causes a rehybridisation of surface Pt atoms, leading to a decreased d population.270 Core-level Shgt Spectroscopy on Tungsten Surfaces 0.8 0.6 0.4 .+ ' /i i . . 0 . o L . * . . '1 32.6: 32.2 31.8 31.4 ' 31.0 30.6 0.08 it binding energy/eV Fig. 10. Fitted core-level spectrum obtained from the W{ 1 lo} (2 x 2)-N structure; the components have been fitted on the assumption that they all have the same half-width as the bulk peak. Core-level shifts are more sensitive to d than to s and p occupancy, because the latter are spatially more diffuse.Referring to self-consistent calculations of Wimmer et ~ 1 . ~ they also argue that the valence electrons of the alkali-metal atom are polarized, but not significantly ionized: Cs is believed to be covalently bonded to the surface. For K on Pt and Au, Duckers and Bonze125 reported small shifts towards higher binding energy. Soukiassian et al.' reported shifts to lower binding energies, for the first time, for Cs on W{ 100) and Ta(100). The shifts were small: 50 meV for W and 100 meV for Ta, but nevertheless the authors concluded that the observations were consistent with donation into the surface transition-metal d band.Now Riffe et ul.' have reported results for Cs, K and Na on W{110} which show small shifts to lower binding energy, in the range 12-28 meV at monolayer coverage. Once again quoting the band structure calculations of Wimmer et ~ l . , ~ ' they conclude that there is little if any charge transfer to the substrate. This non-ionic bonding picture extends to low coverages. On the other hand, the results of Riffe et al.' for 0 on W{ 1 lo}, producing a shift to higher binding energy in agreement with the present results, are attributed by them to a substantial degree of charge transfer. The authors are led to conclude that 0 adatoms, which produce a small work-function increase of 0.2 eV on W{ 1 have a high degree of ionic character, while Cs, which produces a much larger work-function decrease of 3.9eV, is non-ionic.We shall see that this conclusion is incorrect. Adsorption of an atomic layer on W which is neither electron-withdrawing nor electron-donating must itself lead to an increase in the core level binding energy of top-layer substrate atoms. Not surprisingly, we have found this with Au adsorption,' and, trivially, we note that a W layer on its own lattice clearly shifts the underlying core-level energies towards the bulk position. Thus, Purcell et aL4 report a simpleG. P. Derby and D. A. King 27 1 empirical relationship, based on data from six different crystal planes of tungsten, between the core-level shift A and the surface coordination: Asn(eV) = 0.953 - 0.044 C N,/ rJ J = 1 where E is the core-level binding energy, and the effective coordination is considered important out to the third shell j at distance rJ and with an atomic occupancy Nj.When considering the effect of an adsorbate on the core-level binding energies of the substrate, two effects therefore need to be accounted for: increased coordination of substrate surface atoms, which will increase the core-level binding energy, and charge redistribu- tion. Eqn (1) does, of course, contain contributions from both effects even for the case where adatom and substrate are the same element. But we point out that a layer of tungsten adsorbed on W(110) causes a shift in the core-level binding energy of the original top layer to higher binding energy, of 300 meV. This does not imply that W is an electron-withdrawing adatom on its own lattice.When a shift is induced by an adsorbate, it is necessary to define the quantity where E A is the core-level binding energy of the top-layer substrate atom after adsorption. It is useful to consider separately the contribution to the surface atom core-level binding energy arising from changes in the environment external to the atom,29 Aenv, and changes arising from alterations in electronic configuration. The effective one-electron potential necessarily increases at a surface, decreasing the core-level binding energy of surface layer atoms with respect to their bulk counterparts, even if no charge redistribution is accounted for. Adsorption will cause an increase in Aenv, towards the value Ebulk. Charge redistribution in the initial state takes two forms: a configurational change within the atom involving interconversion between s and d states, producing a shift Asd, and a redistribution involving net charge donation between atoms, producing a shift A,, .In moving an atom from the bulk to the surface, a conversion occurs from d to s states, producing an increase in the core-level binding energy. Adsorption of an electron- donating species can be expected to increase the d state occupation. Finally, we must account for a shift due to final state relaxations, Arelax. Thus the total shift is given by: Benesh and Haydock’” show that for surface layer atoms on W{lOO} the terms Aen, and Asd are both large (0.6-0.8 eV) and of opposite sign. From their analysis and more recent results for W{lOO}’* the conclusion is inescapable that there is a large final-state relaxation shift, ca.0.4eV, contributing to the data from clean surfaces. This may be a result of a high density of states at the Fermi level, associated with surface states and resonances. Now we note that for the shift induced in top layer substrate atoms by electronegative adsorbates A,, should be positive, while for electropositive adsorbates it should be negative. Since the overall coordination shift is necessarily positive (towards higher binding energy), charge flow between adsorbate and substrate tends to cancel the shift in the case of electropositive adsorbates, and reinforce it in the case of electron- withdrawing adsorbates. This is in full agreement with the general observation that the observed shifts for alkali metals are usually very small, and can be towards higher or lower binding energies, never exceeding the bulk binding energy.The shifts for elec- tronegative adsorbates are large, often exceeding the bulk value. These small shifts for alkali metals on W{ 1 lo} are due to a cancellation of two effects, and not to the non-ionic nature of the bond: on the contrary, if the bond were non-ionic a large shift should be observed towards the core-level binding energy of bulk atoms. AAS = Aen, + Awi i- Ael+ Arelax272 Core-level Shift Spectroscopy on Tungsten Surfaces The data obtained for H, 0 and CO overlayer structures on W{ 1 lo}, summarised in fig. 3, can be readily rationalised in terms of this model.H adatoms, with a W,-H bond energy of ca. 280 kJ mol-' (2.9 eV), produce shifts of ca. +lo0 meV per bond from the clean-surface peak S. The charge redistribution at this surface actually produces a small work-function decrease, but the results are consistent with a coordination shift combined with a charge-transfer shift involving some transfer to the H adatom. For 0 and C adatoms, formed from 0, or from CO, where the bond energy to the substrate is ca. 800 kJ mol-' (8.3 eV), i.e. comparable to the bond energy of W to its own lattice, the observed shifts from the clean-surface peak are considerably larger, consistent with a large coordination shift and a significant donation of substrate d electrons to the adatoms. The molecular virgin CO state appears to produce two core level peaks, one at a binding energy between clean surface and bulk peaks and a second, broad peak at higher binding energy.We have assigned these, respectively, to linear and bridge-bonded species bonded to surface W atoms. Although the bridge-bonded species is less stable, since it develops at higher exposures, the W2-C bond strength is higher than that of W-C in the linear species: increased d v " backdonation strengthens this bond, while weakening the C-0 bond. This may lead to a larger coordination shift for W atoms bonded to the bridged species. The underlayer case, (2x2)-N, is significantly different. Here, at a low coverage (0.25 monolayer) a large shift is observed (peak N,) indicating that top layer W atoms are poorly screened: this will influence both the initial-state charge transfer and final-state relaxation, combining to give the observed large shift.In additiori, the multicomponent nature of the spectrum demonstrates bonding to second-layer W atoms, causing also a decrease in the bulk peak intensity. We are grateful to the S.E.R.C. for the provision of beamtime at the Daresbury Syn- chrotron Radiation Facility, and for a Studentship to G. P. Derby. We also acknowledge useful discussions with Dr Greg Benesh. References 1 P. H. Citrin and G . K. Wertheim, Phys. Rev. B, 1983, 27, 3167. 2 D. Spanjaard, C. Guillot, M. C. Desjonqukres, G . Triglia and J. Lecante, S u r - Sci. Rep., 1985, 5, 1. 3 K. G. Purcell, J. Jupille and D. A. King, in Solvuy Conference on Sucfuce Science, ed. F. W. d e Wette (Springer, Berlin, 1988), p.477. 4 K. G. Purcell, J. Jupille and D. A, King, Surf: Sci., 1989, 208, 245. 5 G . Apai, R. C. Baetgold, P. J. Jupiter, A. J. Viescas and I . Lindau, Sue- Sci., 1983, 134, 122. 6 P. Soukiassian, R. Riwan, J. Cousty, J. Lecante and C. Guillot, Surf: Sci., 1985, 152/3, 290. 7 D. M. Riffe, G. K. Wertheim and P. H. Citrin, Phys. Rev. Lerr., 1990, 61, 571. 8 A. Awobode, G . P. Derby and D. A, King, t o be published. 9 G. K. Wertheirn, P. H. Citrin and J. F. van der Veen, Phys. Rev. B, 1984, 30, 4344; 1988, 38, 7820. 10 T.-M. Duc, C. Guillot, Y. Lassailly, J. Lecante, Y. Jugnot and J. C. Vedrine, P h ~ x Rev. Lett., 1979, 43, 789. 11 K. G. Purcell, G. P. Derby and D. A. King, J. PhJs. Condensed Matter, 1989, I , 1373. 12 D. M. Riffe, G. K. Wertheirn and P. H. Citrin, Phys. Reu. Lett., 1990, 63, 1976. 13 M . W. Holrnes and D. A. King, Sue$ Sci., 1981, 110, 120. 14 €3. D. Barford and R. R. Rye, J. Chem. Phys., 1974, 60, 1046; E. W. Plummer and A. E. Bell, J. Vuc. 15 C. Kohrt and R. Gomer, SurJ Sci., 1971, 24, 77. 16 C. Leung, M. Vass and R. Gomer, Sur$ Sci., 1977, 67, 67. 17 R. Opila and R. Gomer, Surf Sci., 1983, 129, 563. 18 Ch. Steinbruchel and R. Gomer, Surj: Sci., 1977, 67, 21. 19 A. M. Bradshaw, D. Menzel and M. Steinkilberg, C'hem. Ph~t.5. Lett., 1974, 28, 516. 20 E. Umbach, J. C. Fuggle and D. Menzel, J. Electron Spectrosc. Relat. Phenorn., 1977, 10, 15. 21 E. Urnbach and D. Menzel, Surf Sci., 1983, 135, 565. 22 C . Somerton and D. A. King, Surf: Sci., 1979, 89, 391. 23 S. P. Singh-Boparai, M. Bowker and D. A. King, Surf: Sci., 1975, 53, 55. 24 J. Jupille, K. G . Purcell and D. A. King, Solid Stute Cornmun., 1986, 58, 529. Sci. Techno/., 1972, 9, 583.G. P. Derby and D. A. King 273 25 K. Duckers and H. P. Bonzel, Surf: Sci., 1989, 213, 25. 26 E. Wimmer, A. J. Freeman, M. Weinert, H. Krakauer, J. R. Hiskes and A. M. Karo, Phys. Rev. Lett., 27 E. Wimmer, A. J. Freeman, J. R. Hiskes and A. M. Karo, Phys. Rev. B, 1983, 28, 3074. 28 C . R. Brundle and J. Q. Broughton, in The Chemical Physics of Solid Surfaces and Heterogeneous Catalysis, ed. D. A. King and D. P. Woodruff (Elsevier, Amsterdam, 1990), Vol. 3A. 29 G. A. Benesh and R. Haydock, J. Phys. C, 1984, 17, L83. 30 J. Jupille, K. G. Purcell and D. A. King, Phys. Rev. B, 1989, 39, 6871. 1982,48, 1128. Paper 0/016526; Received 12th April, 1990