Chemisorption and epitaxial growth: structural and kinetic studies of halogen interactions with vanadium and chromium surfaces

Vacuum/volume 33/numbers Printed in Great Britain

1 O-l Z/pages 707 to 714/l

0042-207X/83$3.00+ .OO Pergamon Press Ltd

983

Chemisorption and epitaxial growth: structural and kinetic studies of halogen interactions with vanadium and chromium surfaces A P C Reed and R M Lambert, Cambridge, CB2 IEP, UK

Department

of Physical

Chemistry,

University

of Cambridge,

Lensfield

Road,

and J S Foord,

Department

of Chemistry,

University

of Southampton,

Southampton,

SO9 5NH, UK

A unified surface chemistry has emerged from studies of halogen adsorption and halide corrosion on the (100) faces of Cr and V. Dissociated halogen overlayers form with high sticking probability at low coverages and pronounced precursor state effects are exhibited. As 8 increases above 0.5, attractive interactions originally present within the adlayer become repulsive and unr’axial compression of the adsorbed phase occurs. Thermal desorption is complex since thermal conversion yields a range of products: by consideration of the relevant thermodynamic and kinetic parameters governing the desorption process, a general model accounting for these observations has been formulated. Corrosion of the metal substrate occurs at high gas exposures resulting in the formation of dihalides by an island growth mechanism. These halide films adopt a common layer structure and orientation. Their stoichiometry and growth rate depend sensitively on the speed of interdiffusion through the film and the residence lifetime of the weakly bound X, precursor via which corrosion proceeds. Modifications to the V(lOO)/CI, chemistry upon coadsorbr’ng potassium were investigated. KCI forms in preference to VCI,, although microcrystallites of both halides result at high exposures. Diffusion of Cl between the two corrosion phases is facile at 300 K, and esd takes place only from the KCI phase.

1. Introduction

2. Experimental

Widespread attention has deservedly been directed towards the interaction of halogens and alkali metals with transition metal surfaces ‘--I * on two main counts. Firstly, the undoubted practical involvement ofsuch species as catalyst promoters and moderators provides an obvious impetus for their study. Secondly, they are good candidates with which trends in chemisorption and epitaxial growth may be probed, the latter providing an area of particular focus in this work. Recent studies on the chemisorption of Cl, on V and Br, on Cr have completed our study of halogen adsorption and corrosion on the (100) faces of these bee metals and a unified surface chemistry has emerged which permits the identification of key features of much wider applicability. These are described in this paper which presents structural and kinetic data concerning the formation of halogen overlayers and the subsequent nucleation and growth of halide corrosion phases for the four systems investigated (Cr/Cl,/Br, and V/CI,/Br,). Trends in behaviour and their correlation with changes in adsorbate and/or substrate will be described, and contrast drawn with other analogous adsorption systems involving Fe and W. Following this, variations to the V( lOO)/Cl, chemistry upon coadsorbing the highly electropositive K species will be dealt with.

All experiments were carried out in stainless steel vacuum chambers that have been described previously7*8*‘2. The Cr(lOO) and V( 100) specimens were prepared from 99.99% purity ingots. Temperature measurement was by means of a Pt/Pt-10% Rh thermocouple spot welded onto the rear face of the sample. Halogen gases were generated in situ by the solid state electrolysis of the appropriate silver halide”. Potassium ion dosing was carried out using a zeolite based ion sourcer4. The doses quoted (m-‘) refer to the flux of X, molecules and K+ ions respectively incident at the centre of the specimen. 3. Halogen adsorption formation. Depending on the total amount of halogen present on the surface, all the adsorption systems studied displayed two characteristic and markedly different types of behaviour. For low halogen exposures, the surface chemistry only 3.1. Overlayer

involves formation considered first.

of overlayers,

and

this

regime

will

be

Halogens were found to adsorb rapidly on Cr and V(100) at 300 K as monitored by AES, XPS and A0 measurements. Typical results (from Cr/Br,) are illustrated in Figure 1. The coverage was 707

A P C Reed,

Y 0

R M Lambert

Chemisorption

1

2

and eoitaxial growth

I

I

I

0,5

and J S Foord:

3

Br, exposure (mv2) X lo-l9 1. Bromine uptake on Cr(100) at 300 K followed using work IunctIon measuremenls (A) and Br (MVV) 55 eV:Cr (LMM) 483 eV Auger peak height ratio (0).

found to increase almost linearly with dose for exposures of up to - 5 x 10’” n-‘, after which the sticking probability appears to fall off rapidly, though adsorption does continue, and no saturation limit is found. In line with the conclusions of others who have studied halogen adsorption, we take the view that in the low exposure regime (up to -8 x IO’” m-‘),adsorption results in the formation of a single, dissociated halogen adlayer. This is consistent with XPS and UPS data which show that only one distinct halogen binding state is present, and the work function results where the large increase (A@= 1.1-1.5 eV for all four systems) fits in most naturally with this assertion. Furthermore, the LEED data described below confirm that dissociation takes place during adsorption. LEED studies revealed that the clean (100) surfaces do not exhibit the previously reported reconstructions’5*‘b but possess a (1 x 1)structureat 300 K”-19. A series ofordered LEED patterns was observed during the adsorption of either Cl, or Br, onto Cr( 100) as shown in Figure 2. As the chlorine coverage increased, a very poorly ordered c(2 x2) structure was observed which sharpened up to give ~hc pattern in Figure 2(b). This then developed continuously through a ~(2 x 4) intermediate to saturate at a ~(2 x 5) phase (Figures 2(c), (d), (e)). By comparison, bromine adsorption results in the formation of a sharp ~(2 x 2) pattern which transformed similarly but saturated at the c(2 x 4) stage. These limits represent completion of the first adlayer, since comparison with AES shows a sharp break in sticking probability occurs immediately after this coverage is reached, and new features characteristic of compound formation appear in the thermal desorption (TD) spectra (see Section 3.2). These patterns may be interpreted as shown in Figure 3 as arising from the compression of the c(2 x 2) structure first formed along a single direction ([OlO] or [OOI], depending on domain orientation). Comparison of relative surface coverages deduced by LEED and AES show good agreement on the basis of this model. Additionally, combination of this information with the known incident flux reveals that S, =0.9+0.2 in both cases, provided the overlayer scattering unit is a halogen atom. If the layer comprised X, species, this would imply that S, = 1.8 and hence this possibility is eliminated. It has been pointed out previouslyzO.” that an alternative explanation to the compression model outlined above may be applicable to the c(2 x 2)-+c(2 x 4) conversion. If the adspecies are 708

Figure 2. exposures 4.8 x 10” 7.5 x IO’”

LEED patterns produced on Cr(100) a~ 300 K by chlorine of (a) 0, 140 V, (1 x 1) substrate; (b) 3.5 x IO’” m-‘, 135 V; (c) m-l 124 v, ~(2 x 4); (d) 5.9 x 10’” n-*. 108 v; (c) m-2: 102 v, p(2 x 5).

C

(2x23

c (2x4)

Pl2x51

J

Figure 3. (a) Adlayer rcciprocai lattice. (b) Adlayer direct lattice; small circles refer to the overlayer, large circles to the substrate. (c) Overlayer s1ructures; shaded circles denote halogen atoms, open circles denote underlying metal atoms.

A P C Reed, R M Lambert and J S Food:

Chemisorption and epitaxial growth

considered to occupy the same high-symmetry sites throughout, a progressive reduction in antiphase domain boundary spacing effects the requisite coverage increase. However, this model leads to an unrealistically small interatomic spacing of 2.89 A. Furthermore, the absence of a sharp c(2 x 2) phase in the Cr/CI, study suggests that no strong preference for such high symmetry sites exists. The ‘high symmetry’ model is therefore not favoured in this case. The general processes described so far are frequently observed in halogen adsorption studies. Less commonly, it was noted that a sharp c(2x 2) LEED pattern is formed by Br on Cr(100) at coverages as low as 0=0.21; which implies that island formation occurs at such adatom concentrations. Behaviour indicative of island formation was also shown by V/Cl,, V/Br,, Cr/CI, and has been noted for Fe(lOO)/Cl,/Br, l-3 and W( 100)/Br,/I,ti. Such an occurrence might result from the strong precursor effects exhibited by these systems, since precursor species which initially sample occupied sites adsorb with highest probability into adjacent unoccupied sites. Alternatively, attractive interactions could exist within the overlayer at these coverages. To distinguish between these possibilities, the crucial observation was made that following desorption of the structurally uniform adlayer, when the thermally equilibrated arrangement will be adopted, island formation is again exhibited at 0<0.5. Thus it appears that attractive interactions do indeed exist within the adlayer at very low coverages, which at least partially accounts for the observed island growth behaviour. A point of particular interest in these studies concerns the manner in which the overlayer desorbs, since the range of products observed does not correspond to the initial state of the adlayer. We have endeavoured to examine the underlying factors responsible for such behaviour and to apply these to a consideration of other related adsorption systems. Typical TDS profiles are shown in Figure 4, and the products observed following desorption of the saturated adlayers (Q-0.8) are summarized below.

Coverage Associated

TDS peak

Adsorption system Cr/CI, /Br, VICI, /Br2

0.8>U>0.5

0<0.5

P

Y

CrCl,

CrCl

CrBr, VCI, VBr,

CrBr Cl

Br

Thus any model must seek to explain why the overlayer desorbs as the metal dihalide for 0 > 0.5, yet gives rise to a monohalogen product (X or MX) at lower coverages. Concentrating on the Cr/Br, adsorption system, thermodynamic energy cycles using tabulated bond energy data****’ readily reveal that MX,(,, is the energetically favoured desorption product in comparison with Br,,,, Br,,,, and CrBr,,,, the other species which could arise from the bromine overlayer. In order to explain the observed behaviour we postulate that (a) above f&,=0.5 when the stable c(2 x 2) structure is disturbed, the heat of adsorption of bromine shows a steady fall off and (b) a substantial activation barrier exists for the conversion of the bromine overlayer to CrBr, moieties, and this must be surmounted before such species can be evolved into the gas phase. An energy diagram, such as is shown in Figure 5 can then be constructed and the following points noted.

b

a

P

J 00

600

900

1200

1400

,oo

600

900

1200

Temperature (K)

Figure 4. Thermal desorption spectra from bromine dosed Cr(100) monitored at (a) 131 amu (CrBr’) and (b) 212 amu (CrBr;).

Figure 5. Energy diagram for the Cr(100)/Br, system.

(i) Although CrBr(,, is the exclusive desorption product when 0~0.5, it is energetically least favoured. This therefore suggests that the pre-exponential term for CrBr evolution is considerably greater than that for Br or Br,. (ii) For 0~0.5, desorption occurs as CrBr, rather than CrBr, because I&.> ECrRr.where the meaning of these terms is as in Figure 5. (iii) It is apparent from the figure that E 7-a - ECrBr= E,, -constant. Thus as compression sets in, E,,, falls, and transformation into the c( state outweighs desorption of CrBr. The model therefore readily accounts for the experimental observation that desorption occurs as a monohalogen species at low 0, but as CrBr, for 0>0.5; furthermore it predicts that the activation energy for CrBr, desorption from the overlayer should vary down to a limit equal to the Q state desorption energy (discussed in Section 3.2) as observed experimentally. Insertion of the appropriate figures for Cr/CI, into this model (where, in general, bond strength values are greater than for Cr/Br,) reveals that the relatioe disposition of the energy levels is 709

A P C Reed, R M Lambert and ./ S Foordc Chemisorption

and epitaxial

unaltered and hence it is unsurprising that an identical product distribution is found. However, a significant difference is noted where V is the substrate since the sublimation energy of the bulk metal is much higher than for Cr (512 cf396 kJ mol-I). This is responsible for increasing the value of E,,, to such an extent that despite the difference in pre-exponential factors, desorption by this route becomes energetically too expensive as 0 falls below 0.5 and therefore halogen atoms, rather than VX, desorb. 3.2. Halide growth. Although the overlayers discussed above saturate at coverages of U-0.8, further adsorption, as witnessed by AES, XPS and TDS takes place at higher gas exposures, and some important aspects of the interaction in this regime should be noted. Firstly, as can be seen in the TDS traces of Figure 4 (which serves as an exemplar of all four systems), the halogen starts to desorb from a more weakly bound (c() state when the exposure characteristics are exceeds 4.5 x 10” m-’ and no saturation displayed by the t( state which ultimately dominates the desorption profile. Secondly, LEED patterns (Figure 6) are observed, at

.

‘0

0 .D

-I Figure

6.

LEED

(-IO”

n-‘)in

patterns thesystems:

produced

following

(a)V(lOO)/CI,,49

gas exposures

V; (b)V(lOO)/Br,, 105 V: reciprocal Mice of four to the substrate lattice.

(c)G-( loO)/CI,. 104 V; (d) Cr( lOO)/Br,, 10.5 V; (e) domains of CrBr,, (v q 0 0). Directions r&r 710

high

growth

high concentrations in the a: state, in which the electrons apparently do not sample the M( 100) periodicity. Thirdly, distinct chemical shifts in the metal (2~) XP and metal (MVV) Auger spectra take place during population of the u state, which are indicative of the formation of a surface compound in which the metal oxidation state differs significantly from zero. Taken together these results firmly indicate that corrosion of the underlying metal substrate is occurring to produce a bulk halide phase which attains a considerable thickness, even at the low halogen pressures (< low5 Pa) employed in this work. The most obvious and important questions to be addressed relate to the stoichiometry, structure, rate and mechanism of growth of these new corrosion phases. The stoichiometry is of particular interest since a range of halides is known to exist22*23. However, both the di- and trihalides are known to sublime yielding vapours primarily consisting of the parent molecular ion. Therefore, the observation that in al[ the cases studied, MX,(,, is the predominant desorbing species, suggests this formulation for the developing corrosion phase. (Signals displayed in the TD spectra at lower masses result from fragmentation of MX,(,, in the mass spectrometer ion source.) Photoemission results confirm the view that it is the dihalide which develops during high gas exposures, typical data recorded in this case from the V/Br, system, being presented in Figure 7. Following an exposure of 10” mm2 at 300 K, the XPS data reveals a shift of 1.9 eV to higher binding energy in the V(2p) peaks. The magnitude of this shift is very similar to that observed on corrosion ol Cr and reference to standard compounds shows it to be characteristic of an M’-+M” conversion process. The UPS spectra arc also consistent with this view. Apart from the changes in the bromine 31, region of the spectrum (4.5-7.0 eV) which occur during corrosion of the substrate, another prominent change is the decay of emission intensity near E, and the concomitant growth of a feature at - 2.0 cV from surfaces which had been annealed at 550 K. In common with the interpretation of similar spectra from other ionic compounds it appears most likely that this feature arises from the -‘T,,+-“A Ig photoemission transition from the rl band of octahedrdily coordinated V2+ species in the VBr, lattice. Changes in the electron structure at the surface during bromine adsorption as rcvealcd by valence band Auger spectra (which arc also presented in Figure 7) support the interpretation ofthe results: there would appear to be no sensible way to explain the large energy shift of the V(MVV) transition during population of the u state unless it is accepted that the entire surface is converted to an epitaxial layer of VBr, of substantial thickness. It is worth remarking on the observation that the cntirc electron emission spectrum (XPS, UPS, AES) from the /la/i& layer shifts by 0.9 eV to lower kinetic energies when films formed at 300 K are annealed to 550 K; Figure 7(a) shows that substantial reductions in 41 accompany the process. The energy shifts that occur are unlikely to be true ‘chemical shifts’ in view of the uniform shift that occurs in the entire spectrum. Instead, it seems likely that they arise from changes in reference level caused by the development of an electric field across the halide layer. Films formed at 300 K appear to have an excess concentration of bromine in the outermost layer, which gives rise to a high work function and a negative potential drop across the halide film. Subsequent annealing then results in the diffusion of vanadium or bromine atoms through the layers causing a shift of reference level of the atoms in the surface region and a reduction in 4. The main oxidation state in the halide is lower than that which normally results from the direct interaction of the elements.

A P C h?mi

R M L8mb8rt and J S Foord: Chemisorption and epitaxial growth

I

Br.exposurelmolcculer maI

t--

J’512ELECTRONBINDING520ENERGY(evl526

Temprriture

d.

I

A

300K 300K 300K

20

30

ELECTRON ENERGY(cV) cp 50

I

8

ol=Efl 3 ELECTRONBlt&lNG :NERGY (cV)

Figure f.Changes induced by bromine adsorption on V( 100) and subsequent specimen heating: (a) work function; (b) AES of the V(MVV) and Br(MNN) region; (c) XPS of the V(2p) levels; (d) He 1 excited UPS.

Thermodynamic considerations suggest why this should be so. Although at partial halogen pressures such as exist in the beams employed (lo-’ Pa) MX, is unstable with respect to MX,, the reaction M,,+2MX,,,-,3MX,,,, is highly exoergic and so a combination of low incident flux and intimate contact with the underlying metal substrate favours the dihalide. The rate of halide growth was estimated by two methods. Firstly, it was noted that the a and fl binding states yield the same desorption products, and the amount of material giving rise to the /? state is readily derivable from the LEED data (Section 3.1). Measurement of the a desorption yields then also permits the quantity of bulk halide giving rise to the a peak in the TD profiles to be calculated. Secondly the ratio of the MO and Mu (2~) XPS signals were recorded at varying halide coverages and the data analysed to yield halide film thicknesses. Both methods indicated that halide growth occurs with a sticking probability of 0.02-0.1. This is ofcourse a much slower process than overlayer formation, a feature readily apparent in the AES uptake curve in Figure 1, which shows a sharp drop in slope when the overlayer saturates.

Insight can be obtained into the mechanism of the corrosion process by correlating LEED and TDS results, and examining the temperature dependence of the growth process. LEED observations on all systems consistently revealed that when amounts of corrosion product equivalent to several physical monolayers were present at the interface, overlayer domains still remained and much higher exposures were needed before these disappeared. This clearly indicates that corrosion follows the Volmer-Weber mechanism whereby 3-D halide microcrystallites form leaving large areas of the surface unperturbed, although eventually these islands coalesce. Interestingly, although the corrosion rate remained relatively constant with substrate temperatures of 300+50 K, it exhibited a rapid fall between 450-630 K as witnessed by the decrease in TDS yield suggesting that reaction of the impinging halogen does not occur within one scattering event, but rather that a precursor state is involved. Following Kisliuk” an incoming species trapped in such a mobile state on the halogenated surface may either (a) react to form a metal halide moiety at a particular active site, (b) desorb 711

A P C Reed, R M Lambert and J S Foord: Chemisorption and epitaxial growth

or (c) hop to another active site from where it may react, desorb, or undergo a further hopping motion. Under such circumstances the corrosion rate should then be given by

R

&I

T=l+um

where R ,=corrosion rate at the low temperature limit; R, =corrosion rate at temperature T; I.4 = v,,r, exp( - E,,,/RT)=rate constant of reaction (desorption) at a particular active site. This model predicts a low temperature limit for R, and then a reduction in rate as fd+f, with increasing T (which occurs provided E,> E, (see inset in Figure 8) as observed here). Qualitative agreement with this scheme was seen in all four cases. Equation (1) was tested quantitatively in the Cr/Br, study and the TD results are shown in Figure 8 along with a plot of ln[(R,/R,)-l] v l/T where the gradient yields Ed-E,-75 kJ mol-‘. A surprisingly high value of E,, 2 75f 20 W mol - ’ results implying that halogen corrosion crucially involves a weakly ckmisorbed mobile intermediate rather than the physisorbed species commonly assumed. The structure of the halide layers and relationship to the underlying metal substrate may be deduced from the LEED patterns, which are shown in Figure 6. The most obvious deduction is that epitaxial growth is encountered in all these systems and the obvious similarities between the four LEED ‘ring’ patterns presage the adoption of a common orientation and structure by the halide phases. The vanadium patterns (Figures 6(a), (b)) consist of 12 symmetrically spaced spots arranged in a ring about the 0,O beam, each spot being elongated along the ring circumference. In Figure 6(b) a second ring‘is visible with radius 1.74 times that of the first. Such patterns arise from the presence of two domains of hexagonal symmetry oriented at right angles to each other with some degree of rotational disorder about the surface normal. It is proposed that these patterns are generated by hexagonally close packed planes of halogen ions oriented parallel to the (100) surface with metal cations located in octahedral interstices between alternate planes. Such layer structures, known

a

a

b

,000

750 Temperature

(K)

Fipre8.

(a)212 amu thermal desorption spectra from Cr(1OO)dosed with 1.25x IO’* Br, m-’ at diNering substrate temperatures. (b) Plot of In[(R,/R,)l] vs T-’ to test equation (1) in text. Inset shows relationship of Ed and E,

712

as the CdCl, and CdI, forms (differing only in the precise ordering of the close packed layers), are adopted by the great majority of first row transition metal dihalides (MCI,, MBr, and MI,, M=Ti, V, Mn, Fe, Ni, CO)‘~*~~. The surface unit cell lattice parameters, measured from LEED patterns, are in remarkably good agreement with the values measured in the bulk compounds at 3.77 A for V/Br, and 3.6 A for V/Cl,. We also suggest that such an arrangement is adopted by the Cr compounds since this is the only sensible way of interpreting the LEED patterns. Although bulk Cr dihalides standardly adopt a rhombohedral form, the suggested layer structures only differ in the manner in which the CrX, octahedra, which constitute the two forms, are packed together. Good support for the layer compound interpretation comes from a report2 g ofa thin film hexagonal form for CrCl,. While the vanadium halide films possess an undistorted hexagonal structure, analysis, such as is shown in Figure 6(e) for the CrBr, system, reveals that the chromium halide films exhibit large deviations from hexagonal symmetry. Thus the basal plane is described by lattice parameters of 3.8 x 3.3 A2with an included angle of 110” for the CrBr, film, while the corresponding figures for the chloride structureare 3.5 x 3.9 A2with an included angle of 117”. In concluding this section, we tentatively suggest that this may be a manifestation of a Jahn-Teller type distortion, a phenomenon well established in the structure of bulk Cr” compounds. 3.3. Summary. The formation of dissociated halogen overlayers duringgasexposures of w 10” mm2 is observed with high sticking probability, metal-adsorbate charge transfer and strong precursor effects being exhibited in each of the four adsorption systems studied. In the case of Cr, an extensive series of cotipression structures is observed for coverages in the range O.S
A P C Reed, R M Lambert and J S Foord: Chemisorption

and epitaxial

Although no reliable TDS studies were carried out on Fe, it was postulated that halogen atom desorption would predominate from the overlayer. However, consideration of the model derived here leads to the expectation that desorption of FeX, should be found. Very recent work in our laboratory has in fact-shown this to be so”. Halogen atom desorption is observed in the case of W, and this is totally consistent with the present model since it simply arises from the high sublimation energy of W. The interesting phenomenon of low pressure halide growth has not previously been reported for Fe and W, for which no firm reason can be offered. However, because of the island growth mechanism corrosion is virtually undetectable at the onset except by TDS and

growth

this factor seems to have inhibited its observation since current work in this laboratory using TDS presents clear evidence for dihalide formation on Fe. 4. K +CI, coadsorption on V(100) Modifications to the metal/halogen surface chemistry upon coadsorbing the highly electropositive K species were examined. Attention was focused on the competition for halide growth between K and V, the structure and manner of growth of resulting compounds and the stability of the inhomogeneous interface under electron irradiation.

a

/I

b(i)

I

300

1

300

1200

BOO

600

Temperature

1

I_

0

I

800

L

1200

(I<)

T = SOOK

T.300K

-

I

000

I 1

1

r

20

LO

60

Charge

collected

ImC

1 -J

Figure 9. (a) Thermal desorption spectra from V(100) exposed to 2.5 x 1Ol8 K m-* and 1.3 x 10“’ Cl, m-* showing KCI desorption peak at -650 K. (b(i)) 42 x 2) LEED pattern, KC1 on V(100). 64 V. (b(ii)) Proposed real space model for c(2 x 2) KC1 structure on V(100). (c) esd ofchlorine from V(100) dosed with 6 x lOI K me2 and 2.5 x lOI Cl, m-’ showing effect of raising substrate temperature by 200 K.

713

A P C Reed,

R M Lambert

and J S Foord:

Chemisorption

and epitaxial

TD data proved revealing with regard to the nature of the chemical interaction between adsorbate species. Traces, from experiments involving the coadsorption of K and Cl, onto V(100) are shown in Figure 9(a). Such evidence provides a strong indication that KC1 forms on the surface since it is always a desorption product when the species are coadsorbed in any proportion. This is not unexpected in view of the attractive interspecies forces which exist and further support is afforded by LEED observations. A c(2x 2) pattern (Figure 9(b)(i)) was exhibited over a wide range of coverages (0.2<0,+,,
714

growth

which increases by two orders of magnitude when the sample is heated from 300 K to 500 K.

Acknowledgements

We are grateful to Johnson Matthey Limited for the loan of precious metals. APCR thanks BP Ltd and Newnham College, Cambridge for financial support.

References

’ P A Dowben and R G Jones, Surj”Sci, 84, 449 (1979). z R G Jones and D L Perry, SurfSci, 88, 331 (1979). ’ P A Dowben and R G Jones, SurfSci,88, 348 (1979). 4 H M Kramer and E Bauer. Sur~Sci. 107, 1 (1981). ’ K J Rawlings, G G Price and B J Hopkins, Sci,95, 245 (1980). 6 K J Rawlings, G G Price and B J Hopkins, Sur~Sci. 100, 289 (1980). ’ J S Foord and R M Lambert, SursSci, 115, 141 (1982). ’ P W Davies and R M Lambert, Sur[Sci, 95, 571 (1980). 9 M P Cox and R M Lambert, Surf Sci, 107, 547 (1981) and references therein. ” M W Roberts and C S McKee, Chemistry of the ~Meral-Gas Interfke. Clarendon, OUP (1978). ” G Brodcn and H P Bonzel, Surf Sci, 84, 106 (1979) and references therein. ” J S Foord, P J Goddard and R M Lambert, Sur/Sci, 94, 339 (1980). r3 P J Goddard and R M Lambert, Surf Sci, 67, 180 (1977). I4 R E Weber and L F Cordes, Rev Sci Iastrum, 37, 617 (1966). ” P W Davies and R M Lambert. SurfSci, 107, 391 (1981). ” G Gewinner, J C Peruchetti, A Jaegle and R Riedinger, Phys Reo Lett, 43, 935 (1979). r’ J S Foord, A P C Reed and R M Lambert, Surf Sci, 129, 79 (1983). ‘s V Jensen, J N Anderson, H B Nielsen and D L Adams, Surf Sci, 116,66 (1982). ” G Gewinner. J C Peruchetti and A Jaegle, Sur~Sci, 122, 383 (1982). ” M Huber and J Oudar, Surf Sci, 47, 605 (1975). ” J P Bibcrian and M Huber. SurfSci, 55, 259 (1976). I2 R Colton and J H Canterf0rd:Hnlide.s of tire First Row Trartsitioft Mefals. Wiley, London (1969). s3 Grnelins Handbuch der Anorganische Chemie, 52, (1962) and 48 (1967). 24 L Ley and M Cardona (Eds), Photoemission in Solids, vol II, Springer, Berlin (1979). 2s P J Kisliuk, J Phys Chern Solids, 3, 95 (1957) and 5, 78 (1958). s6 D McKie and C McKie, Crystalline Solids. Nelson, London (1974). *’ S Hino and R M Lambert, unpublished work. ‘s M L Knotek and P J Fcibelmann, Plrys Rev Left, 40, 964 (1978). 29 M L Knotek and P J Feibelmann, SurfSci, 90, 78 (1979).

Surf