5 Spectroscopy of the Metal-Gas Interface By J. PRITCHARD Department of Chemistry Queen Mary College London E. 1. DESPITE the important role of solid surfaces in heterogeneous reactions and in catalysis it is only recently that direct investigations of their structure and composition have become possible. This review is concerned with recent applica-tions of spectroscopic methods to the metal-gas interface. Metal surfaces have attracted detailed study by electronic methods such as field emission field ionization low energy electron diffraction and work function measurements. Some perspective of the relatively minor role of spectroscopy to date is provided by a recent conference report.’ However this role will undoubtedly grow rapidly as appropriate techniques are developed as shown by the striking advances in Auger electron spectroscopy.1.r. spectroscopy provides the majority of other applications. 1 Auger Electron Spectroscopy The emission of characteristic Auger electrons from surface atoms ionised by a primary electron beam was observed by Lander2 and more recently by Scheibner and T h a r ~ . ~ Although the usefulness of this effect for chemical analysis of the surface layers of solids was recognised its exploitation awaited the development of electronic methods of differentiating the small Auger currents from the much larger background of secondary electrons. It is now a sensitive analytical t0ol,47~ and examples of its application and improvements in technique have multiplied during the year. Riviere6 has discussed the basis and potentialities of Auger electron spectro-scopy in relation to other methods of surface analysis.The electron detection efficiencies are for the electrostatic deflection analyser and for the retarding field analyser used by Harris’ and by Weber and Peria’ respectively in the first applications of the technique. To a first approximation the optimum ‘The Structure and Chemistry of Solid Surfaces,’ ed. G. A. Somorjai Wiley New York, 1969. E. J. Scheibner and L. N. Tharp J. Appl. Phys. 1967,38 3320. N . J. Taylor J. Vacuum Science Technology 1969 6 241. Chem. Eng. News 1969 47 No. 50 72. ’ J. C. Riviere Physics Bulletin 1969 20 85. ’ H. E. Bishop and J. C. Riviere J. Appl. Phys. 1969,40 1740. L. A. Harris J. Appl. Phys. 1968 39 1428. R. E. Weber and W.T. Peria J. Appl. Phys. 1967,38 4355. ’ J. J. Lander Phys. Rev. 1953,91 1382 66 J. Pritchard primary electron energy for ionizing surface atoms is 3 to 3.5 times the critical ionisation energy of the inner atomic level. It is estimated that the Auger current collected by the spherical screen used in a typical retarding field analyser would amount to 2 x 10- A for the case of KLL transitions generated in a monolayer of oxygen on a copper surface by a primary current of The pressing need for surface analysis particularly of single crystal metals in low energy electron diffraction (LEED) and other studies has encouraged the widespread development and application of the retarding field method with LEED optics. Auger electron spectroscopy equipment is now commercially available, either as an accessory to existing LEED systems or separately.Weber and John-son1* improved the sensitivity of secondary electron energy analysis in a con-ventional 3-grid LEED system to permit the detection of as little as 0.02 mono-layer of caesium on a silicon surface. An alternative detection system with 3-grid LEED optics has been described" which has the merit of allowing rapid alternation of LEED and Auger observations. Even 2-grid optics have been used successfully12 in a study of copper epitaxy on tungsten. Instead of the third grid screen a balanced bridge technique was employed to eliminate capacitative pick-up of the retarding potential modulation on the electron collector. A similar bridge technique has been used with 3-grid opticsI3 so that field penetra-tion can be reduced at high potentials by applying the modulated retarding poten-tial to two adjacent grids.Alternatively a fourth grid may be ~ s e d . ' ~ Improved resolution and accuracy is achieved for the detection of elements giving Auger peaks in the range 500 to 2000eV. A further improvement in sensitivity par-ticularly at high energies is obtained with a primary electron beam at a high angle of incidence.I4 This beam can be provided by an auxiliary gun yielding much higher currents than are possible with the LEED electron gun. Penetration of the bulk is reduced at high angles of incidence so that Auger emission from the surface layers increases in relative strength. Harris' has used the electrostatic deflection analyser to study the dependence of Auger emission on both the angle of incidence of the primary electron beam and the angle of ejection of the secon-dary electron.With a molybdenum sample containing sulphur and carbon im-purities little difference in behaviour of the Auger lines of molybdenum (190 eV), sulphur (1 50 eV) and carbon (272 eV) was found with varying angle of incidence, but with a specimen which had been heated and in which segregation of sulphur on to the surface had occurred there was a striking difference in the angular behaviour of the Auger lines from surface sulphur and the underlying molyb-denum. The main features of these results were discussed in terms of a simple model in which Auger electrons are assumed to be emitted isotropically both primary and Auger currents are attenuated exponentially and isotropically in passage through the crystal and scattered Auger electrons are not detected.A. l o R. E. Weber and A. L. Johnson J . Appl. Phys. 1969,40 314. l 1 T. E. Gallon I. G. Higginbotham and M. Prutton J . Phys. (E) 1969,2 894. l 2 A. R. L. Moss and B. H. Blott Surface Sci. 1969 17 240. l 3 H. E. Bishop and J. C. Rivibe Surface Sci. 1969 17 462. l4 P. W. Palmberg Appl. Phys. Letters 1968 13 183; Ref. 1 p. 29-1. l 5 L. A. Harris Surface Sci. 1969 15 77 Spectroscopy of the Metal-Gas Interface 67 The short mean free path of emitted electrons limits the depth of origin of detect-able Auger electrons. Palmberg and Rhodin16 deposited very thin layers of silver on gold and found the escape depth of 72 eV and 362 eV electrons to be about 4 A and 8 A respectively i.e.less than four atomic layers. The dependence of the intensity of Auger electrons originating from a surface layer on the thick-ness of the layer has been discussed on the basis of a simple Darwin scattering m0de1.I~ Because the ionisation cross-section of primary electrons shows a maximum at about 3 to 3-5 times the critical ionisation energy of the atomic level scattered primary electrons i.e. secondary electrons may sometimes be more effective than the original primary electrons in producing Auger electrons in situations where a wide range of Auger peaks is being scanned. Houston and Park' claim that because both Auger and secondary electron yields increase similarly with increasing primary energy the secondary electrons play an im-portant part in producing Auger signals and that the variation of scattering effects with composition of the solid will therefore complicate the use of Auger electron spectroscopy as a quantitative analytical tool.In practice Auger emission lines are superimposed on a stronger background of secondary electrons and the Auger spectrum is revealed by electronic dif-ferentiation of the secondary electron energy distribution. The derivative is obtained by applying a modulating potential to the energy analyser. An analysis7 shows that in the retarding field case the derivative current increases as the square of the amplitude of modulation up to an amplitude of half the r.m.s. width of the Auger peak and then levels off to a constant value. In the electro-static deflection (127" sector) analyser the derivative increases linearly with modulation amplitude up to an amplitude of one third of the r.m.s.width, passes through a maximum and then decreases at amplitudes greater than six times the peak width. Taylorlg has given a detailed analysis of the factors affecting resolution and sensitivity of measurement with both the 4-grid retarding field and 127" electrostatic sector analysers concluding that for low excitation voltages and wide Auger peaks the retarding field method is preferable but that the sector analyser is better for high energies and narrow peaks. However field penetration and scattering from the grids can cause considerable error in the apparent energy distribution of low energy secondary electrons.20 Whereas the spherical grid retarding field system has a high collection efficiency compared with the 127" sector analyser it suffers from the shot noise in the large total secondary electron current which is collected.Very recently a major improvement in the detection of Auger spectra has been achieved2' by using a cylindrical electrostatic analyser of high collection efficiency. The energy range 0 to lo00 eV can be scanned in 50 ms and the spectrum displayed on an oscillo-scope. Alternatively more conventional scanning speeds may be employed with l 6 P. W. Palmberg and T. N. Rhodin J . Appl. Phys. 1968 39 2425. '' T. E. Gallon Surface Sci. 1969 17,486. J. E. Houston and R. L. Park Appl. Phys. Letters 1969 14 358. l 9 N. J. Taylor Rev. Sci. Znstr. 1969 40 792.2 o P. S. P. Wei A. Y. Cho and C. W. Caldwell Rev. Sci. Znstr. 1969 40 1075. 2 1 P. W. Palmberg G . K. Bohn and J. C. Tracy Appl. Phys. Letters 1969,15 254 68 J. Pritchard very low primary beam currents (- A). In either case electron impact desorption effects would be greatly reduced. The precise location of an Auger peak in the energy distribution of secondary electrons is frequently complicated by the sloping background. The derivative curve is therefore unsymmetrical. Several authors have chosen for convenience to take the most clearly defined point in the derivative curve namely the minimum of the high energy wing. Bishop and Rivike13 have proposed that this should be a generally adopted convention. Analytical Applications.-Weber and Johnson have used Auger spectroscopy quantitatively in conjunction with LEED and work function measurements to characterise the surface structures of potassium on germanium (1 11).Calibra-tion of the potassium Auger signal at 252 eV was achieved by depositing measured amounts of K+ ions. The peak to peak amplitude was found to be linearly related to coverage. It was shown that a linear relationship should be found if the area under the Auger peak in the energy distribution is proportional to amount of potassium and the shape of the Auger peak is independent of coverage. The calibrated Auger signal was then used to measure the coverages correspond-ing to the LEED patterns which developed from a fully covered surface after partial desorption at higher temperatures. Work function changes could not be used for coverage measurements because the work function differed depending on whether the same coverage had been reached by adsorption at room tempera-ture or by desorption at higher temperatures.A similar study of caesium adsorption on the (100) and (1 10) planes of tungsten has been made,22 but without a completely independent coverage calibration. In this case the Auger measurements served to confirm conclusions drawn from LEED and work function measurements that the initial adsorption leads to a 42 x 2) ionic layer on top of which a (2 x 2) atomic layer first forms. Subse-quently the atomic layer develops into a close packed structure similar to bulk caesium but limited to a single layer and at the same time the secondary electron energy spectrum shows a 1.5 eV energy loss peak attributed to a two-dimensional plasma in caesium.The well-known minimum in the work function with increasing coverage corresponded to completion of the second layer (2 x 2) structure. In the epitaxial growth of copper films on tungsten(ll0) surfaces Auger spectra have shown no evidence for diffusion of copper into the substrate.12 Together with LEED and work function data the results indicate the formation of a strongly bound copper monolayer similar to a (111) plane of copper but strained in the tungsten[001] direction to match the substrate. Epitaxial growth of (1 11) oriented films takes place on top of this monolayer. The primary electron beam in either LEED or Auger electron spectroscopy may interfere with the system being studied.Alkali halides often used as sub-2 2 A. U. MacRae K. Muller J. J. Lander and J. Morrison Surface Sci. 1969 15 483. 23 A. U. MacRae K. Miiller J. J. Lander J. Morrison and J. C. Phillips Phys. Rev. Letters 1969 22 1048 Spectroscopy of the Metal-Gas Interface 69 strates for epitaxial metal films are particularly susceptible. Surface dissociation and the desorption of neutral chlorine atoms and molecules from potassium chloride leave a potassium-rich surface.24 During the growth of epitaxial silver films on such a substrate potassium migrates over the silver islands. Little change in the magnitude of the potassium Auger signal is seen until the formation of a con-tinuous silver film shields the substrate from the electron beam.25 The potassium Auger signal has been monitored also in a study of the cleavage of mica (musco-vite).26 The results indicate that the potassium ions separate equally onto the two cleavage faces.A valuable aspect of Auger electron spectroscopy is its ability to identify contaminants particularly those originating from the bulk of the solid.8 The segregation of impurities at the surface may profoundly affect the adsorptive properties of the solid and may cause structural changes in the surface layers. Examples of reconstruction are well known in LEED studies even with nominally clean surfaces. In such cases Auger spectroscopy may confirm that the surface is clean or it may reveal unsuspected contaminants. However Auger spectra must be interpreted cautiously as is shown by a series of papers concerned with silicon surfaces.The Si(l11) - (7 x 7) and Si(ll1) - (J19 x J19) diffraction patterns have been obtained reproducibly in many laboratories. Although it now seems probable that the (419 x J19) structure is associated with nickel imp~rity,~’ the (7 x 7) structure has for long been considered to be due to reconstruction of the clean surface. Bauer28 has suggested that both patterns arise from double scattering between thin films (iron or nickel silicides) and the unreconstructed substrate the supporting evidence being the presence in the Auger electron spectrum of peaks at 45 eV and 57 eV which were attributed to iron and nickel respectively. The heat treatment which produced the (7 x 7) structure caused the 45 eV peak to grow and the 57 eV peak to diminish.The reverse changes occurred when the (419 x 419) structure was developed. Furthermore added iron caused the 45 eV peak to grow and facilitated formation of the (7 x 7) structure. That the (7 x 7) structure is to be explained in this way has been disputed, however on the grounds that the rate of diffusion of iron in silicon is too low for sufficient accumulation at the surface in the time taken for this structure to ap-pear.29 Nor is the Auger evidence convincing as Grant and Haas3’ have found the same transitions in the Auger spectra of both a freshly cleaved silicon surface and the annealed surface giving a (7 x 7) LEED pattern. They conclude that all the peaks are due to silicon and that the (7 x 7) structure is not stabilised by impurities.Taylor3’ has extended the Auger measurements to include not only 2 4 P. W. Palmberg and T. N. Rhodin J . Phys. and Chem. Solids 1968,29 1917. 2 5 T. E. Gallon I. G. Higginbotham M. Prutton and H. Tokutaka Thin Solid Films, 2 6 J. P. Deville and S. Goldsztaub Compt. rend. 1969 268 (B) 629. ” A. J. van Bommel and F. Meyer Surface Sci. 1968,12 391. E. Bauer Physics Letters 1968 26A 5 3 0 ; Ref. 1 p. 23-1. 2 q J. W. T. Ridgway and D. Haneman Appl. Phys. Letters 1969,14 265. 3 0 J. T. Grant and T. W. Haas Appl. Phys. Letters 1969,15 140. 3 ’ N. J. Taylor Surface Sci. 1969 15 169; 1969 17 466. 1968 2 369 70 J. Pr it chard the low energy peaks but also the characteristic triads due to LMM transitions at 570 to 710 eV for iron and 700 to 850 eV for nickel. A comparison of the inten-sities of the low and high energy peaks from the pure metals with those observed on silicon at 44 and 56 eV led to the conclusion that the latter were due to silicon and not to metallic impurities.Peaks were observed at 106 90 73 56 44 and 35 eV and a shoulder appeared on the low energy side of the 90 eV peak. Apart from the small unidentified peak at 106 eV tentative assignments were made as follows 90 eV-L2,3VV Auger; 73 eV-first order bulk plasma loss peak of 90 eV Auger electrons ; 56 eV-second order bulk plasma loss peak ; 44 eV-L1L2,3V Auger; 35 eV-possibly third order plasma loss feature; and the shoulder on the 90 eV peak possibly related to the transition density of valence band electrons. These assignments have been criticised by Bishop and Riviere13 who suggest that plasma loss peaks are unlikely to have such high relative inten-sities and that alternative Auger transitions could account for the shoulder and the peaks at 73 and 35 eV.They also find that the 56 and 106 eV peaks are always observed together but not always in constant proportion to the 90 eV silicon peak and prefer to assign them to nickel impurity for which the calculated M2,3VV and MIVV Auger transitions would appear at 61 and 105 eV. The previous paragraph illustrates the present difficulties in utilising Auger spectroscopy for the unambiguous identification of small amounts of impurity. Little quantitative information has appeared yet on the detailed spectra of pure substances or the influence of chemical environment but preliminary results indicate that minor features of the spectrum may be strongly influenced by the latter.31 However in some cases identification is less controversial.Charig and Skinner32 have identified the carbon contamination of Si(ll1) by ethylene and their results corroborate the importance of such contamination for generating three-dimensional growth centres during the epitaxial growth of silicon by silane pyrolysis.33 PalmbergI4 has shown carbon to be responsible for the ring LEED pattern reported for the Pt( 100) surface. Not only is the 270 eV carbon Auger peak strongly evident but the ring corresponds to the lowest order lattice spacing in the basal plane of graphite. The carbon is apparently in the form of randomly oriented graphite layers.34 After removal of carbon by oxidation the Pt( 100) surface gives a complex LEED pattern similar to that given by Au( 100) and generally known as the (1 x 5) pattern.Auger spectra do not indicate any significant contamination on either surface and the general similarity of the spectra apart from a small energy displacement suggests that they are charac-teristic of the pure metals. Silicon segregation on Pt(100) at 1500°C has been identified by an Auger peak at 91.5eV. It could be eliminated by heating in oxygen.14 Carbon contamination may also appear as a result of electron beam cracking of residual gases during Auger or LEED observations. l 4 2 Electron Spectroscopy and Valence Electron Distributions. Electron spectra may be used to gain information on the energy distribution of electrons within adsorbed species or within the valence band at the solid surface.32 J. M. Charig and D. K. Skinner Surface Sci. 1969 15 277. 3 3 B. A. Joyce J. H. Neave and B. E. Watts Surface Sci. 1969 15 1. 3 4 J. W. May Surface Sci. 1969 17 267 Spectroscopy of the Metal-Gas Interface 71 Following the successful application of photo electron spectroscopy by Bordass and Linnett3' to methanol adsorbed on tungsten using 21.2 eV helium radiation at grazing incidence further exploitation of this technique may be expected. In principle similar information should be obtainable from the shapes of peaks in the Auger spectrum corresponding to transitions involving the valence band. Electron excited Auger spectra are complicated by the broad background of scattered electrons but in the favourable case of the 270eV KVV transition of graphite the transition density function was computed by Amelio and S ~ h e i b n e r .~ ~ This function combines the density of states and transition probabilities and was found to be in reasonable accord with the density of states in the valence band of graphite derived from band structure calculations. The more general application of this approach will be facilitated by progress in the interpretation of the true secondary electron energy distribution^.^^ An alternative but experimentally difficult approach which avoids the problem of secondary electron background is ion-neutralisation spectroscopy developed by H a g ~ t r u m . ~ ~ ~ ~ Auger emission is excited when ions such as He' are neutra-lised by electron transfer from surface orbitals.The theoretical basis of Auger ejection by ion neutralisation has been discussed by Wenaas and Ho~smon.~' Hagstrum and Becker41 have shown how the transition density function of a clean Ni(100) is changed when ordered adsorbed layers of oxygen sulphur and selenium are formed. On the clean surface it reflects the density of states in the d-band showing a peak about 1 eV below the Fermi level but with adatoms present the Auger transition probability is increased at energies corresponding to the bonding electrons. The adsorbates quoted give p(2 x 2) and 4 2 x 2) surface structures and lead to peaks in the transition density function well below the Fermi level and to a reduction in the nickel d-band peak.These peaks are taken to be the molecular orbital energy spectrum of surface species composed of adsorbate atoms and first layer nickel atoms and are related to the p-orbital energies of the adsorbate. Specific structures are proposed compatible with the orbital symmetry revealed by the splitting or non-splitting of the p-orbital peak, with the ionicity revealed by energy shifts relative to the free atom and with the observed work function changes. This work is perhaps the most significant recent development in surface spectroscopy. 3 Infrared Spectroscopy Although the use of silica-supported samples continues to be the basis of most i.r. studies of metal surfaces other approaches have been further developed. Bradley and French42 describe a novel technique for depositing metal aerosols on to potassium bromide plates.Aerosols of platinum palladium nickel, 3 5 W. T. Bordass and J. W. Linnett Nature 1969 222 660. 3 6 G. E. Amelio and E. J. Scheibner Surface Sci. 1968 11 242. 3 7 M. P. Seah Surface Sci. 1969 17 132; G. F. Amelio and E. J. Scheibner Ref. 1, 3 8 H. D. Hagstrum J . Appl. Phys. 1969,40 1398. 39 H. D. Hagstrum and G. E. Becker Ref. 1 p. 9-1. 40 E. P. Wenaas and A. J. Howsmon Ref. 1 p. 13-1. 4 1 H. D. Hagstrum and G. E. Becker Phys. Rev. Letters 1969 22 1054. 4 2 J. N. Bradley and A. S. French Proc. Roy. SOC. 1969 A 313 169. p. 11-1 72 J. Pritchard copper iron iridium silver and molybdenum were prepared by explosion of metal wires in argon at atmospheric pressure and allowed to settle on to five plates to provide a multilayer sample for i.r.measurements. With a mean particle size of about 45 nm a transmission of some 10 % at 2000 cm- ' resulted. Nickel, platinum and palladium exploded in argon containing carbon monoxide gave spectra comparable with supported metal samples but the spectra were much weaker if the carbon monoxide was added after the aerosols had been prepared, a,pd none were obtained from the other metals. This illustrates an inherent diffi-culty of the technique namely contamination and even if very pure argon is used the explosion may release contaminants from the walls. On the other hand the metal particles produced in this way are more stable and less prone to aggre-gate than those in evaporated metal films. In the Ni-CO system a band was consistently observed at 1880 but surprisingly no band appeared in the 2000 cm- ' region.Palladium similarly gave a single band at 1900 and platinum gave two bands at 1850 and 2070 cm- the latter being quite strong. Oxidation of adsorbed carbon monoxide followed by the decrease of absorbance was consistent with rate cc [CO]$,,[O,] as on silica-supported platinum. Single layer vacuum deposited metal films have been exploited for transmission i.r. spectroscopy of chemisorbed carbon monoxide.43 With films deposited at low temperatures and sufficiently thin to give about 70% transmission at 2000 cm-' spectra of carbon monoxide could be obtained from nickel cobalt iron, and iridium. With tungsten chromium and manganese it proved necessary to evaporate the metals in the presence of carbon monoxide presumably because the vacuum conditions were inadequate to prevent contamination.The spectra were comparable with those given by silica-supported metals but much less intense. Thicker films or films deposited at higher temperatures failed to yield any spectra with adsorbed carbon monoxide. A more detailed investigation of nickel has been made using films deposited under ultrahigh vacuum condition^.^^ The spectra depend on the temperature of deposition of the film an effect believed to be due to particle size variations. Low temperatures and small particles favour the appearance of a broad absorption band extending from 1900 to 1700cm-' as well as the higher frequency band at about 2070cm-'. Transmission peaks in the spectra are attributed to the effects of anomalous dispersion on the re-flection and absorption losses which in thin films are of comparable magnitude.The spectra obtained with nickel films deposited in the presence of carbon mon-oxide show additional features probably related to carbonyl formation and decomposition such as a band at 1610 cm- '. This band may be due to adsorbed carbon monoxide molecules in which the oxygen atom interacts with neighbour-ing nickel atoms. Similar frequencies are observed in the compounds45 4 3 F. S. Baker A. M. Bradshaw J. Pritchard and K. W. Sykes Surface Sci. 1968 12, 44 A. M. Bradshaw and J. Pritchard Surface Sci. 1969 17 372. 4 5 N. J. Nelson N. E. Kime and D. F. Shriver J. Amer. Chem. SOC. 1969,91 5173. 426 Spectroscopy of the Metal-Gas Interface 73 which have bands at 1682 and 1527 cm- ’ respectively due to the bridging CO groups to which aluminium triethyl groups are co-ordinated.It has been sug-g e ~ t e d ~ ~ that the spectra of carbon monoxide on nickel films and on supported nickel are due to the gas being adsorbed at relatively high coverage and that most of the strongly bound species do not contribute. In this connection recent LEED are interesting. On the (111) face the first carbon monoxide to be adsorbed appears to disproportionate to carbon and adsorbed carbon dioxide. Subsequently CO is adsorbed in a second layer. Greenler4’ has pursued his theoretical approach48 to reflection methods as a means of detecting i.r. absorption spectra of adsorbed species on metal surfaces. The absorption per reflection is a maximum at an incident angle of typically 88” and it decreases rapidly at higher or lower angles.At this angle several reflections are desirable if a conveniently measurable band is to be observed the optimum number being set by consideration of the signal to noise ratio. Although the absorption per reflection is less at lower angles more reflections may ade-quately compensate. Practical considerations limit the useful number of reflections between parallel mirrors but the number may be greatly enhanced by using curved reflecting surfaces. The sensitivity of the resulting design is illustrated by the experimental absorption spectrum of a 54 % film of cellulose acetate on silver mirrors. Using parallel mirrors and multiple reflections at high angles of incidence the absorption spectrum of a monolayer of chemisorbed carbon monoxide on copper has been determined.49 Copper mirrors were prepared by evaporation of the metal on to hinged glass plates under ultrahigh vacuum conditions.A single relatively strong band was found at 2105 cm- ’ in good agreement with transmission spectra from supported copper. An average extinction coefficient of about 2 x 10- l 8 molecule- cm2 was estimated and it is considered that reflection spectroscopy with single crystal surfaces should be practicable. Carbon monoxide adsorption on supported metals continues to receive atten-tion. Guerra” has compared the CO stretching frequencies in the high frequency ‘linear’ band for a range of metals. The bond order decreases in proportion to the number of missing d-electrons.This correlation is discussed in relation to the molecular orbital description of chemisorbed carbon monoxide proposed by Blyholder” and it is pointed out that changes in a-bonding cannot be ignored. Frequency shifts due to co-adsorption of oxygen ammonia and hydrogen sulphide are attributed to altered d-electron densities. Blyholder and Allens2 have extended the range of experimental data to cover the transition metals from vanadium to copper. Two main bands can be distinguished on each metal. The trend of frequencies in the high frequency band is consistent with Blyholder’s 46 T. Edmonds and R. C. Pitkethly Surface Sci. 1969 15 137. 4 7 R. G. Greenler J . Chem. Phys. 1969 50 1963. 4 8 R. G. Greenler J . Chem. Phys.1966,44 310. 4 9 A. M. Bradshaw J. Pritchard and M. L. Sims Chem. Comm. 1968 1519. 5 o C. R. Guerra J . Colloid Interface Sci. 1969 29 229. 5 1 G. Blyholder J . Phys. Chem. 1964 68 2772. 5 2 G. Blyholder and M. C. Allen J . Amer. Chem. Soc. 1969,91 3158 74 J. Pritchard original model where this band is associated with linear species (i.e. the terminal groups of molecular carbonyls) adsorbed in the middle of plane surfaces. On the other hand the low frequency band considered to be due to edge and corner sites, does not show comparable behaviour but remains at almost constant frequency. An attempt to reconcile this difference uses a more extensive molecular orbital model involving a square array of nine metal atoms with a two atom molecule adsorbed either on the middle atom or on a corner atom and utilising p-orbitals of the molecule and d-orbitals of the metal atoms.The estimated CO bond orders for nickel and chromium are in satisfactory accord with experiment. Caution is needed when seeking meaningful correlations of the kind described above because of the uncertain nature of the metal surfaces. As mentioned earlier it is likely that carbon monoxide adsorption on nickel is accompanied by the formation of carbon and carbon dioxide. Guerra has emphasized the lack of crystallinity in the supported samples of the metals Ir Os Re Rh and Ru,” and B l y h ~ l d e r ~ ~ used metal particles prepared by evaporation into a film of hydro-carbon oil. Variations of particle size in supported nickel are shown by van Hardeveld and HartogS3 to be accompanied by appreciable changes in the spectra.The nickel samples on which i.r. active nitrogen adsorption takes place, and which contain very small crystallites also give strong CO bands at low frequencies. Nitrogen adsorption is associated with B5 sites (i.e. sites providing five-fold co-ordination to nickel atoms) which occur on (110) or (113) planes, while the CO bands are attributed to linear species attached to nickel atoms having six or seven nearest neighbours. Carbon monoxide on supported gold gives a band at 2120 at low coverage, changing to 21 10 cm- ’ at high coverage.54 Higher frequencies previously reported for gold are due to incomplete reduction. Coverages estimated from parallel volumetric adsorption experiments enabled an extinction coefficient of about 3 x molecule-’ crn’ to be derived comparable with earlier values for carbon monoxide on platinum.Oxygen reacts readily with carbon monoxide on gold and neither gas behaves as a poison if added first. The supported gold also possesses unexpected activity in hydrogen adsorption promoting deuterium exchange with the OH groups of the silica support at room temperature. No band due to hydrogen on gold was observed but using alumina-supported iridium two bands due to chemisorbed hydrogen have been reported5’ at 2120 and 2050 cm-’. Deuterium gave bands at 1520 and 1490 cm-’ and no band due to HD was seen. These results are similar to previous ones for platinum and indicate dissociative adsorption. It has been shown recentlys6 that the hydrogen bands on platinum are associated with residual oxygen and the same may apply in the case of iridium.Other oxygen-containing surfaces appear active towards hydro-gen aqd yield clear spectra. Thus the pyrolysis of SiOCH3 groups on silica 5 3 R. van Hardeveld and F. Hartog Fourth Intern. Congress Catalysis Moscow preprint 5 4 D. J. C. Yates J . Colloid Interface Sci. 1969 29 194. ’’ F. Bozon-Verduraz J-P. Contour and G. Pannetier Compt. rend. 1969 269 (0, s 6 D. D. Eley D. M. Moran and C. H. Rochester Trans. Faraday Soc. 1968 64 2168. 70. 1436 Spectroscopy of the Metal-Gas Interface 75 generates SiH2 groups from which hydrogen can be desorbed at high temperatures and readsorbed at lower temperature^.'^ Similar behavior has been found with boria-silica' and on reduced germania.59 The effect of sulphur poisoning6' on the i.r. spectra of CO adsorbed on alumina-supported platinum is to drastically reduce the intensity of the band at 2080, characteristic of the clean surface and to produce a weak band at 2170cm-'. The latter is removed by evacuation and is attributed to CO weakly adsorbed on sulphur-poisoned sites. Bands occurred at 2080,2050,1995,1950 and 1920 cm-after adsorption of COS. Apart from the band at 2080 cm- which arises from CO as a decomposition product all the bands are due to COS. The lowest frequency band is considered to be due to adsorption on clean platinum sites. 1.r. spectra of adsorbed nitrogen have been discussed by several authors. Van Hardeveld and van Montfoort6' showed that the i.r. active nitrogen on nickel is associated with very small crystallites and they considered it to be strongly physically adsorbed on B5 sites rather than chemisorbed as originally proposed by Eischens.Mertens and Eisched2 have reinvestigated this system using combined i.r. magnetic susceptibility and gravimetric adsorption measurements. The integrated intensity of the band at 2202 cm- ' was found to be 18( & 7) x 10- '' cm molecule- about the same as for carbon monoxide on metals. The ratio of nitrogen molecules to nickel atoms was only 0.03 at saturation compared with 0-25 for carbon monoxide. No clear conclusions about the mode of adsorption could be deduced from the magnetic susceptibility changes. Ravi King and S h e ~ p a r d ~ ~ obtained a band at 2185 cm- ' on iridium and they consider that this, together with the band on nickel should be assigned to a chemisorbed species of the type M=k=N.It is interesting to note that although the stretching frequencies observed in most nitrogen complexes are considerably lower com-parable values of 2220 and 2225 cm- have now been found in complexes of This tends to support the view that the adsorbed species are chemi-sorbed. Van Hardeveld and van M ~ n t f o o r t ~ ~ have carried out an extensive investigation of the adsorption of isotopic forms of nitrogen. For 28N2 the i.r. band frequency is independent of coverage and has the value 2195 f 1 cm-', but the half-width increases with increasing coverage. From a consideration of molecular interactions and in view of the shifts observed when mixtures of isotopes are used it is concluded that the expected frequency increase must be offset by another unidentified factor.The extinction coefficient is approximately 2 x 10- cm2 molecule- and the isosteric heat of adsorption is 13.5 kcal mol- '. A consideration of intermolecular potential energies leads to the conclusion that only alternate B5 sites could be occupied at saturation and the number of sites '' C. Morterra and M. J. D. Low J . Phys. Chem. 1969 73 321. 5 9 M. J. D. Low N. Madison and P. Ramamurthy Surface Sci. 1969 13 238. 6 o E. S. Argano S. S. Randhava and A. Rehmat Trans. Faraday Soc. 1969 65 552. 6 1 R. van Hardeveld and A. van Montfoort Surface Sci. 1966 4 396. 6 2 F. P. Mertens and R. P. Eischens Ref. 1 p. 53-1. 6 3 A. Ravi D.A. King and N. Sheppard Trans. Faraday Soc. 1968 64 3358. 64 J. T. Moelwyn Hughes and A. W. B. Garner Chem. Comm. 1969 1309. 6 5 R. van Hardeveld and A. van Montfoort Surface Sci. 1969 17 90. C. Morterra and M. J. D. Low Chem. Comm. 1969 862 76 J. Pr it chard so calculated is in good agreement with the statistical calculations of van Harde-veld and Hartog.66 Spectroscopic studies of hydrogen adsorption on silica-supported nickel and platinum have yielded interesting results. Sheppard and Ward67 describe the experimental and interpretational methods and their application to the adsorp-tion of acetylene at room temperature. Bands due to =CH- CH2 and CH3 groups are characterised by their peak frequencies by the ratio of the optical densities of the bands near 2955 and 2925 cm- associated with the antisymmetic stretching modes of CH3 and CH2 groups and by the ratio of the total integrated intensities before and after hydrogenation.Interpretation is further aided by comparison with the spectra of model compounds. On nickel surfaces self-hydrogenation is revealed by the initial appearance of CH2 and CH3 groups, which are associated with surface alkyls. Bands assigned to =CH- groups are considered to indicate MCH=CHM species. After hydrogenation the spectrum of the n-butyl group appears. Platinum surfaces behave differently : the initial spectrum is weak and indicative of dissociative adsorption to a surface carbide. Hydrogenation greatly intensifies the spectrum and the spectra of alkyl groups CH3(CH2) (n 2 4) shows that rather more polymerization occurs on platinum than on nickel.Ethylene adsorption has been studied on the same metals over a range of temperatures.68 With platinum the main features of the spectrum indicate associatively adsorbed ethylene between - 78 and 150 "C. At - 145 "C a single strong band appeared which was ascribed to M2CH-CHM2 species. Hydro-genation generates ethane but it also leads to the appearance of weak bands assigned to n-butyl groups adsorbed on the surface. That dissociative adsorption also takes place is shown by the large increase of total intensity on hydrogenation. More complex behaviour is found with nickel in that dimerization and self-hydrogenation produce n-butyl groups before deliberate hydrogenation. At 150"c there is evidence for dissociation to a carbide methane is produced by hydrogenation.Results with but-1-ene are very similar69 with evidence for associative adsorption on platinum together with dissociative adsorption leaving a C4 carbide. Hydrogenation leads to butane but also generates some adsorbed butyl groups. On nickel at low temperatures both associative and dissociative adsorption take place and hydrogenation produces an almost complete con-version to a n-butane. At higher temperatures the spectra suggest multiple attachment and hydrogenation produces more adsorbed n-butyl groups than n-butane. It is surprising that gaseous n-butane and surface n-butyl groups do not seem to be in equilibrium. The spectroscopic results provide no evidence for 7c-bonded species resulting from ethylene or butene.Similarly an investigation7' of benzene adsorption on 6 6 R. van Hardeveld and F. Hartog Surface Sci. 1969 15 189. 6 7 N. Sheppard and J. W. Ward J . Catalysis 1969 15 50. 6 8 B. A. Morrow and N. Sheppard Proc. Roy. SOC. 1969 A 311 391. 69 B. A. Morrow and N. Sheppard Proc. Roy. SOC. 1969 A 311,415. 7 0 J. Erkelens and S. H. Eggink-du Burck J . Catalysis 1969 15 62 Spectroscopy of the Metal-Gas Interface 77 silica-supported nickel and copper shows a broad band at 2760 to 3060 cm- ’ characteristic of CH stretching vibrations at saturated carbon atoms. It is concluded that benzene loses its aromatic character on adsorption and gives a variety of adsorbed species. With palladium platinum and iron no i.r. bands were observed from initially adsorbed species and it seems probable that exten-sive CH bond dissociation takes place.Methanol and ethanol adsorptions on cobalt films in oil7’ give chemisorbed carbon monoxide (band at 1860 to 1890 cm- ’) as well as alkoxide and acyl groups. Propanol and butanol gave evidence only of alkoxide formation. Treatment with carbon monoxide removed these bands and gave the usual spectrum of chemisorbed carbon monoxide. The pattern of results was consistent with earlier work with iron and nickel. Bands at 3345 and 3280 cm- are reported for ammonia adsorbed on plati-n ~ m . ~ ~ They were observed in the course of an investigation of ammonia oxida-tion over silica-supported platinum and correspond to the asymmetric and symmetric stretching modes. The silica support also interacted strongly with the reactants.4 Other Spectroscopic Studies Other spectroscopic methods have not been widely applied to metal surfaces. The following examples indicate some topics which have been investigated and which may lead to significant developments. Kishi and Ikeda73 have explored the U.V. spectra of typical chelating agents on thin vacuum-deposited metal films. With acetylacetone and trifluoroacetylace-tone on the transition metals from vanadium to nickel peaks are found at around 300 nm while ethyl acetoacetate absorbs at 280 nm. These bands are assigned to n-n* transitions of chemisorbed P-diketonate anions. An additional band attributed to charge transfer (d-n*) is also observed on some of the metals. Significant differences between these spectra and those for the complexed metal ions are discussed in terms of the involvement of metal &orbitals in n-bonding.Subsequently these studies have been extended to the adsorption of 8-quinolin01~~ and of pyridine and 2,2’-bip~ridy1.’~ An entirely different kind of U.V. study is that of Moesta and B r e ~ e r . ~ ~ They have observed that the negative surface potential of carbon monoxide on eva-porated nickel films can be increased by up to 2.3 V by intense irradiation. A charge transfer process appears to be involved. The spectrum of the effect shows maxima at about 140 and 200nm. After illumination the changed surface potential relaxes quite slowly with a time constant of the order of 100 seconds. 7 1 G. Blyholder and L. D. Neff J. Phys. Chem. 1969 73 3494.72 D. W. L. Griffiths H. E. Hallam and W. J. Thomas Trans. Faraday Soc. 1968 64, ” K. Kishi and S. Ikeda J. Phys. Chem. 1969,73 15. 7 4 K. Kishi and S . Ikeda J. Phys. Chem. 1969,73 729. ’’ K. Kishi and S. Ikeda J. Phys. Chem. 1969,73 2559. 7 6 H. Moesta and H. D. Breuer Surface Sci. 1969 17 439. 3361 78 J. Pritchard M~Carroll’~ reports a visible chemiluminescence associated with the adsorp-tion of oxygen nitric oxide and carbon monoxide on clean polycrystalline tungsten. The effect is very feeble only one photon being emitted for every lo9 molecules of CO or NO and for every lo7 molecules of oxygen. The lumine-scence persists for several seconds. Mossbauer spectroscopy has considerable scope for application in metal sur-face chemistry. It has been used to monitor the growth of oxide films during the corrosion of and to relate the catalytic efficiency of supported iron catalyst for butene hydrogenation to the states of the iron particles.79 In the latter study samples of alumina-supported iron containing more than 10% by weight of iron have crystallites large enough to give Zeeman splitting. It is possible to follow reduction of the oxide to metallic iron through the intermediate valence state. Samples containing very small amounts of iron cannot be reduced to give metallic iron and are catalytically inactive, ” B. McCarroll J . Chem. Phys. 1969 50 4758. 7 9 M . C. Hobson jun. and H. M. Gager Fourth Intern. Congress Catalysis Moscow, A. M. Pritchard and C. M. Dobson Nature 1969 224 1295. preprint 48